PDF Complet de La Technologie en Suède

2005:30
DOCTORA L T H E S I S
Seasonal Snow Storage

for Space and Process Cooling
Kjell Skogsberg
Luleå University of Technology

Department of Civil and Environmental Engineering
Division of Architecture and Infrastructure
2005:30|: -1544|: - -- 05⁄30 -- 

Seasonal Snow Storage for Space and Process Cooling
Doctoral Thesis 2005:30
Kjell Skogsberg

Luleå University of Technology, Sweden
Akademisk avhandling för avläggande av teknologie doktorsexamen, som med vederbörligt

tillstånd av tekniska fakultetsnämnden vid Luleå tekniska universitet kommer försvaras
offentligen den 11 oktober 2005 kl 1000, i sal F1031 vid Luleå tekniska universitet.
Opponent: Edward L. Morofsky, Public Works and Government Services Canada
Ordförande: Professor Bo Nordell, Avdelningen för arkitektur och infrastruktur, Institutionen
för Samhällsbyggnad, Luleå tekniska universitet
Academic thesis for the degree Doctor of Philosophy, which with the permission of the Board
of Faculty Engineering, will be publicly defended on October 11th 2005 at 10 am, in room
F1031 at Luleå University of Technology.
Opponent: Edward L. Morofsky, Public Works and Government Services Canada
Chairman: Professor Bo Nordell, Division of Architecture and Infrastructure, Department of
Civil and Environmental Engineering, Luleå University of Technology, Sweden
Doctoral Thesis 2005:30
Kjell Skogsberg

Skogsberg K (2005). Seasonal Snow Storage for Space and Process Cooling.
Foreword
This thesis is the partial fulfilment of the requirements for the degree of Doctor of
Philosophy (PhD). The work was carried out at the Division of Architecture and
Infrastructure. Financial funding was provided by Luleå University of Technology and
The Swedish Energy Agency.
It is my sincere hope that the knowledge gained during this project will be a small
contribution to a sustainable energy usage in the future. I am absolutely convinced that
we must and can do this change to enable a good life for future generations. One can be
certain that technology will improve, but despite that this work concerns technology
development I am certain that a sustainable society is mainly a matter of choice!
During this project I have met many inspiring persons and also received a lot of help
and encouragement. I would like to thank all these people, and first is of course my
supervisor Professor Bo Nordell. Without him the snow cooling plant at the Sundsvall
Regional Hospital would not exist, and without that project there would not have been
any PhD project for me. I am also grateful for all help, encouragement and interesting
discussions, both on the cooling topic and others.
Next in line is the staff at The County Council of Västernorrland. They were also
necessary for the snow cooling project in Sundsvall. Without their enthusiasm my work
would have been much more boring and difficult. I am grateful to all of you that have
been involved in this project and then especially Per-Erik Larsson, Esko Niiranen, Rolf
Nilsson and Kent Hedin. Hans E Ericsson at Sweco in Sundsvall and Bengt Österlund at
TAC in Sundsvall have also been of important help.
My friends and colleagues in Japan have also meant a lot both as professionals and as
friends. I especially want to mention Professor Kiyoshi Ochifuji and Professor
Katsunori Nagano at Hokkaido University, Sapporo, Professor Masayoshi Kobiyama at
Muroran Institute of Technology, and Seiji Kamimura at The Nagaoka University of
Technology. There are also a number of other professors, engineers, PhD students and
students that made my visit in Japan a wonderful memory for life.
Colleagues and friends in Annex 14 of the Energy Conservation Through Energy
Storage Implementing Agreement, International Energy Association, has also meant a
lot, both because of the warm atmosphere in the group and of course also from a
knowledge point of view. I am truly sorry that lack of time prevented me from
participating more than I did.
Thank you also to Professor Folke Björk and research engineer Christer Hägglund at the
Division of Building Technology at the Royal Institute of Technology (KTH), Sweden,
forgiving me the opportunity to conduct experiments on thermal conductivity of wet
cutter shavings, for practical help, and for fruitful and interesting discussions.
I also thank other colleagues at LTU; Professor Angela Lundberg, Professor Anders
Sellgren, Professor Maria Viklander, Professor Sven Knutsson, Jurate Kumpiene, Anna-
Maria Gustafsson, Jenny Lindblom, Göran Tuomas, Maria Engström, Rolf Engström,
Roger Lindfors and many more, both for help, support and joy.
I
Thanks also to my friends Mattias, Lilian, Crister, Sofi, Fredrik, Martin, Anders,
Rønnaug, Karina, Elin, Louis and some more, mostly for just being there; to my mother
and father for encouragement and taking care of my son Simon when writing this packet
of paper; to the rest of my family for just being there and not making a huge fuzz of all
this PhD:ing; and of course to my son Simon, who with joy participated in different
experiments and at other times forced me to think of something else whenever you was
around. I love you!
Hopefully I have not forgotten to mention anyone that has helped me during the years as
a PhD student. It is however likely that some persons that ought to be named here are
not, and it is my sincere wish that anybody who misses her/his name can accept my
apology and still be proud of her/his contribution.
Thank you everybody!
September 8, 2005 Kjell Skogsberg
II
Summary
The world’s cooling demand has increased considerably during the last decades due to
increased population, industrialisation, comfort demands, electronic equipment usage
and new building technologies. Conventional cooling is often produced by electrically
driven devices. One less prime energy-consuming alternative is to use stored winter
cold in snow and ice for cooling during the summer. This ancient technique is feasible
in large parts of the world.
The snow/ice can be stored indoors, on the ground, in the ground and underground. This
study focuses on in ground storage, in an open pond, where the cold energy is extracted
by water that is cooled by direct contact with the snow.
Open pond snow storage must be thermally insulated; hence, different insulation
alternatives were discussed. Cutter shavings were studied in laboratory experiments and
numerical modelling. The surface melt rate of snow covered with cutter shavings
increased with increased solar radiation, air velocity, air temperature, and decreased
insulation layer thickness. The surface melt rate was similar with initially wet and
initially dry wood chips. The evaporation rate contributed significantly to the energy
balance. It was concluded that evaporative cooling is an important part of the thermal
insulation qualities of wood chips. It was also found that heat transfer from the rain and
ground is usually relatively small.
The Sundsvall Regional Hospital snow cooling plant in Sweden has successfully
operated since 1999. Natural and artificial snow is stored in a slightly sloping, shallow
pond of watertight asphalt. During these years, the plant has delivered the main part
(77-93%) of the cooling, totalling 655-1,345 MWh. The snow was thermally insulated
by a 0.1-0.2 m layer of wood chips. The total coefficient of performance, including
construction energy, was 2.0-6.6 times greater than that of a conventional chiller
system.
The environmental impact of a snow cooling plant and a chiller system was compared,
for both existing and “environmentally optimised systems”. Of the existing systems, the
chillers had the largest impact concerning climate change, acidification and nitrification,
while the snow cooling system meant more photochemical ozone emissions. The
dominating impact sources of the snow cooling system were fuel and electricity. In the
construction phase, ground insulation had the greatest impact.
In future open pond storage, a more compact design (deeper storage) is suggested to
reduce maintenance and melt loss. Total cooling costs were estimated to be 0.29-0.47
SEK kWh-1 for new open pond storage, i.e. lower than that of district cooling.
The study also comprised the remarkable mass loss that occurs from freezing water.
Each freezing means a water loss that corresponds to 2.5-10% of formed ice. This
phenomenon was studied since repeated freezing and thawing will evaporate large
amounts of water. The freeze loss was estimated to have little effect on the function of
snow storage in Swedish conditions, though the loss might be considerable at other
locations.
III
IV
Sammanfattning
Kylanvändningen i världen har ökat markant de senaste decennierna. Detta beror bl.a.
på ökande befolkningsmängd, industrialisering och komfortkrav samt fler elektriska
apparater. De senaste decenniernas byggnadsutformning har dessutom i många fall lett
till ökade kylbehov.
Konventionella kylmaskiner drivs vanligen av el. En alternativ kylteknik är att använda
lagrad vinterkyla, t.ex. genom att utnyttja kyla som lagrats i snö- och is. Detta är möjligt
i stora delar av världen. Snö/is kan lagras både inomhus, på mark, i mark och under
mark. Den här studien fokuserar på lagring i mark, i öppna dammar, där kylan utvinns
genom att vatten cirkuleras genom snön.
I öppna lager är det nödvändigt att isolera snön för att minska smältförlusterna. Olika
isoleringsalternativ diskuteras och laboratorieexperiment och numerisk modellering har
genomförts på kutterspån. Smältförlusten ökade med ökande solstrålning, vindhastighet
och lufttemperatur, samt då lagertjockleken minskade. Smälthastigheten var ungefär
lika stor för torra och blöta spån. Avdunstningen från spånlagret bidrar markant till
träspånets goda värmeisolerande egenskaper på snö. Smältningen p.g.a. värmeflöde från
mark och regn befanns vara liten för de flesta förhållanden.
Snökylanläggningen vid Regionsjukhuset i Sundsvall har varit i framgångsrik drift sen
1999/2000. Naturlig och konstgjord snö lagras i en grund, svagt sluttande bassäng av
vattentät asfalt. Anläggningen har sedan dess levererat 77-93% av det totala kylbehovet
under somrarna (655-1345 MWh). Snön har isolerats med 0.1-0.2 m tjockt lager av
träflis. Anläggningens årliga totala kylfaktor (COP), som inkluderade energin för att
bygga anläggningen, var 2.0-6.6 gånger bättre än den för ett konventionellt
kylmaskinsystem.
Miljöpåverkan av en snökylanläggning och ett kylmaskinsystem jämfördes, både för det
existerande systemet och för ett ”miljöoptimerat” system. Av de existerande systemen
hade kylmaskinsystemet störst påverkan på kategorierna klimat, försurning och
övergödning, medan snökylan påverkade kategorin ozonemissioner mest. Den största
föroreningskällan från snökylanläggningen var motorbränsle och driftelektricitet, medan
markisoleringen gav störst negativ påverkan från själva anläggningen.
I nya snökylanläggningar bör man sträva efter att göra lagret mer kompakt, d.v.s.
djupare, för att minska smältningen och underlätta driften. Kylkostnaden för en ny
anläggning av samma typ som i Sundsvall uppskattas till 0.29-0.47 SEK kWh-1, vilket
är lägre än motsvarande (erhållna) priser för fjärrkyla.
I arbetet ingick också en studie av de anmärkningsvärda massförluster som uppkommer
då vatten fryser till is. Vid varje frystillfälle försvinner vatten motsvarande 2.5-10% av
ismängden. Fenomenet studerades eftersom stora mängder vatten försvinner på detta
sätt vid upprepad tining/frysning. Frysförlusterna bedömdes ha liten effekt på snölagers
funktion under svenska förhållanden, men på andra platser kan det medföra stora
förluster.
V
VI
Kjell Skogsberg

This thesis treats seasonal snow storage for cooling applications, with focus on open
pond storage. It consists of an extensive review, four papers and three reports.
Paper I Skogsberg K and Nordell B (2001). The Sundsvall hospital snow storage.
Cold Regions Science and Technology, 32, 63-70. Elsevier Science B.V.
Paper II Skogsberg K (2001). The Sundsvall Regional Hospital snow cooling plant-
Results from the first year of operation. Cold Regions Science and
Technology, 34, 135-142. Elsevier Science B.V.
Paper III Skogsberg K and Lundberg A (2005). Wood chips as thermal insulation of
snow. Accepted for publication in Cold Regions and Science Technology,
Elsevier Science B.V., April 25, 2005.
Paper IV Skogsberg K (2005). Numerical modelling of heat and moisture transfer
through a cutter shaving layer on snow. To be submitted.
Report I Skogsberg K (2005). Capillary suction in cutter shavings and sawdust.

Experiments and a model approach. Technical report. Division of
Architecture and Infrastructure, Department of Civil and Environmental
Engineering, Luleå University of Technology.
Report II Skogsberg K (2005). Convective heat and moisture transfer in a wood chips
layer on snow. A first modelling approach. Technical report. Division of
Architecture and Infrastructure, Department of Civil and Environmental
Engineering, Luleå University of Technology.
Report III Skogsberg K (2005). Mass losses of freezing water. Technical report.
Division of Architecture and Infrastructure, Department of Civil and
Environmental Engineering, Luleå University of Technology.
VII
VIII
List of content
Foreword ............................................................................ I
Summary ............................................................................ III
Sammanfattning.................................................................. V
Nomenclature...................................................................... 2
1 Cooling ...................................................................... 3
2 Objective and scope ................................................... 4
3 Seasonal snow and ice storage.................................. 5
3.1 Techniques and examples.............................................. 6
3.2 The Sundsvall Hospital snow cooling plant.................... 11
3.3 Boundary conditions for snow cooling............................ 22
3.4 Snow and ice production ................................................ 24
3.5 Thermal insulation........................................................... 27
4 Natural melting ........................................................... 29
4.1 Moisture and heat transfer.............................................. 29
4.2 Surface melt .................................................................... 37
4.3 Rain melt ......................................................................... 39
4.4 Ground melt .................................................................... 40
4.5 Design model for a snow cooling plant........................... 42
5 Mass loss of freezing water ........................................ 45
6 Construction of open pond snow storage systems ..... 47
6.1 Pond construction ........................................................... 47
6.2 Thermal insulation and cold carrier system.................... 48
6.3 Snow and ice production ................................................ 48
7 Environmental aspects ............................................... 49
7.1 Environmental impact comparison ................................. 49
7.2 Snow pollution................................................................. 51
8 Economy .................................................................... 53
8.1 Pond construction ........................................................... 54
8.2 Thermal insulation........................................................... 55
8.3 Operation ........................................................................ 55
8.4 Cooling cost model ......................................................... 55
9 Conclusions ................................................................ 59
9.1 Further research ............................................................. 62
Acknowledgement............................................................... 62
References ......................................................................... 63
-1-
NOMENCLATURE
D vapour diffusivity in air [m2 s-1]
g gravitational acceleration [m s-2=N kg-1]
k permeability of porous material [m2]
M molar weight [kg kmol-1]
m mass (weight) [kg]
conv
m convective moisture transfer rate [kg m-2 s-1]
p pressure [Pa=N m-2=kg m-1 s-2]
P energy flux, power [W]
R the universal gas constant [J kmol-1 K-1]
u velocity [m s-1]
v vapour content [kg m-3]
vsat,surf surface saturation vapour content [kg m-3]
wactive(z) maximum active capillary moisture content at height z [kg m-3]
wpassive(z) maximum passive capillary moisture content at height z [kg m-3]
weq hygroscopic equilibrium moisture content [kg m-3]
whygr maximum hygroscopic moisture content [kg m-3]
zcap capillary front height [m]
zcap,active maximum active capillary height [m]
zcap,passive maximum passive capillary height [m]
Z capillary layer thickness [m]
D convective heat transfer coefficient [W m-2 K-1]
E convective moisture transfer coefficient [m s-1]
Gvap vapour permeability coefficient [m2 s-1]
O thermal conductivity [W m-1 K-1]
U density [kg m-3]
P dynamic viscosity [Ns m-2 =Pa s]
Ppor vapour permeability resistance coefficient of porous material [-]
X kinematic viscosity [m2 s-1]
' difference
gradient [X m-1]
Subscripts
abs absorbed
cold cooling (energy)
cond condensation
conv convection
el electricity
fl fluid
int interface
moist moisture
sat saturation
surf surface
unsat unsaturated
vap vapour
wat water
wb wet bulb
1 € = 9.1 SEK = 1.5 CD$
-2-
1 COOLING
Cooling, i.e. removing heat, is used for space and process cooling. The cooling temperature
for different applications varies from far below freezing to above 0oC. The worlds’ cooling
demand has increased considerably during the last decades as a result of an increasing
population and greater industrialisation combined with increased comfort demands,
electronic equipment usage and new building technologies (Paksoy, 2003:1). Turnover of
the refrigeration and air conditioning sectors in the main part of the European Union (EU)
was about 20 G€ in 2003 (AREA, 2004). Space cooling in Sweden has doubled since 1990,
and despite the cold climate, Sweden has among the largest comfort cooling areas per
capita within the EU (SR, 2005).
Compressor driven chillers, absorption chillers, free cooling, evaporative cooling, desiccant
cooling, radiative cooling and stored cold can all achieve cooling. The cold can be
produced close to the consumer or distributed in pipes, i.e. district cooling. There are at
present (Spring 2005) 24 district cooling systems in Sweden.
Chillers are heat pumps for cooling and heating, or most commonly for cooling only. The
cooling media used in chillers and heat pumps are often flammable, poisonous,
environmentally hazardous, or a combination thereof. Chillers are also noisy, require space,
and are aesthetically disadvantageous. Another type of chiller called an absorption chiller is
driven by heat.
Free cooling is low temperature energy extracted from ground, water, and air requiring only
a fan or a pump to utilise the cold. Ground and groundwater temperatures reflect the annual
mean air temperature. In areas with low winter temperatures deep surface water tends to
maintain a low temperature because of its greater density at low temperatures. Free cooling
from the air is used more directly during cold nights and the winter season.
Evaporative cooling, i.e. passing air over a wet surface, decreases the air temperature
because of the energy used by evaporation. The evaporation rates increase by creating a
vacuum in the evaporation chamber. In desiccant cooling, air is first dehumidified by
contact with a desiccant material and thus heated. The warm air is cooled by the return air
from the building and then humidified and thereby further cooled before entering the
building.
In radiative cooling, long wave radiation is emitted from a surface. The surface can reach
temperatures below the ambient by high emissivity in the infrared “atmospheric window”
(8-13 Pm) and low absorptivity for other wavelengths. The emitted power increases with
increased ambient temperature, the decreased temperature difference between ambient and
radiation surface, and clear skies. At an ambient temperature of 20oC it is possible to
achieve a surface temperature difference of about -35oC when the emitted power is low, or
emitted powers of 100 W m-2 when the temperature difference is small (Eriksson, 1985).
In cold storage, energy is stored as sensible heat, latent heat, or heat of chemical reactions.
In large-scale storage systems that utilise sensible heat, thermal energy is usually stored in
water or in the ground. In phase change materials (PCM) like snow, ice, paraffin, fatty
-3-
acids, and mixtures of inorganic salt hydrates, the majority of energy is stored in the phase
change, i.e. as latent heat. In thermo chemical energy storage (TCES) the energy is stored
in the heat of reaction (Setterwall and Alexanderson, 1996).
The length of the storage cycles vary from diurnal to seasonal. Diurnal storage is used when
the temperature difference between day and night is great, or when the cooling (energy)
cost varies enough throughout the day (Dorgan and Elleson, 1993). Seasonal cold storages
utilise low winter temperatures. Stored cold can be used for both direct cooling and as a
heat sink for cooling machines, to reduce driving energy consumption.
Seasonal ice storage for cooling is an ancient technology that was common until the
beginning of the 20th century when chillers were introduced. The oldest snow/ice storage
reference found is an ancient Chinese book of poetry mentioning ice as being stored
underground (Morofsky, 1985). During recent decades different cooling systems with
seasonal snow and ice storage have been developed. Both natural and artificial snow and
ice are used. No energy or power limitations for snow/ice cooling systems exist.
Snow and ice are renewable and not poisonous. Depositing urban snow in central snow
storages is economically and environmentally feasible. Shorter transports are beneficial in
both aspects and pollution control is possible by keeping the contaminated snow in a pond.
2 OBJECTIVE AND SCOPE

The main aim of this project was to develop seasonal snow storage systems and improve
models for dimensioning. A second purpose was to clarify the well-known good thermal
insulation qualities of wood chips on snow.
The study was limited to storage of natural and artificial snow and ice produced with
natural winter cold.
The focus was on pond storages whose upper surfaces were exposed to air. With open snow
storages thermal insulation is necessary. Mainly wood chips and especially cutter shavings
were treated; other alternatives were discussed.
-4-
3 SEASONAL STORAGE OF SNOW AND ICE

In seasonal snow/ice storage, frozen water is stored from winter to summer when the cold is
utilised. Natural and artificial snow/ice can be stored indoors, on the ground, in open
ponds/pits and underground, Figure 1. Snow/ice stored indoors is done in a more or less
thermally insulated building. In an underground cavern, no insulation is needed. When the
snow/ice is stored on the ground or in ponds thermal insulation is necessary. This work
focuses on the seasonal storage of snow/ice in open ponds.
Figure 1 Principle snow storage methods; indoor, on ground, in open ponds and
under ground.
The basic idea of this system is to store snow/ice in a generally watertight pond where a
cold carrier is cooled by the snow. The cold carrier can either be circulated between the
load and the snow or be rejected after it has been used for cooling, Figure 2. The main issue
is that the cold carrier utilises the latent heat of fusion, i.e. melting. Systems where the cold
carrier is in direct contact with the snow as well as closed systems where the cold carrier is
separated from the snow are possible. Air, melt water, sea water, groundwater or some
other fluid might be used as cold carrier. The cooling power is controlled by varying the
flow rate of the cold carrier.
LOAD LOAD
Figure 2 Principles to extract cold from the snow; circulation and one-way systems.
Snow is precipitation in the form of ice crystals, whose density varies from 100-700 kg m-3
(Gray and Male, 1981; Nagano et al, 2000; Viklander, 1997; Sundin, 1998). Ordinary solid
ice (ice 1h) with a hexagonal structure has a density of about 920 kg m-3. Artificial ice is
made from water in a pond and by spraying water. Artificial snow is made with different
types of snow guns. The distinction between ice of sprayed water and snow produced with
snow guns is not defined. The issues in snow production for cooling applications are how
much frozen water is produced in a certain time and at what cost.
Thermal insulation on snow can be loose fill, sheets and self supporting superstructures.
Loose fill insulation includes various types of wood chips, rice shell, debris (mineral
particles), etc. Organic refuse such as sawdust (wood chips) and crop residues is the
traditional thermal insulation material of ice and snow (Taylor, 1985). Sheets can be both
-5-
plastic and filled tarpaulins, e.g. with straw. Superstructures are generally rigid
constructions that are partially or totally removed or opened during the winter.
Distinguishing between natural and forced melting in snow pond cooling applications is
appropriate. Snow melting in warmer surroundings is here denoted as natural melting.
Natural melting is divided in ground melt, rain melt and surface melt, Figure 3. Surface
melt occurs by heat transfer from the sun, air and sky. Heat transfer through sides and the
bottom of the storage causes ground melt.
Figure 3 Natural snow melt in open pond snow storage.
Forced melting is a result of cold extraction, i.e. by heat transfer from the cold carrier.
Much of the cooling energy of snow is in the phase change, i.e. melting. This latent heat of
fusion corresponds to an 80oC temperature increase of liquid water. For a typical air
condition application, the melting thus corresponds to about 90% of the cooling energy. To
utilise the latent energy, circulating the cold carrier through or next to the snow is
necessary.
3.1 Techniques and examples

Numerous snow and ice storages techniques for cooling applications have been suggested
and implemented. In Japan, about 100 projects have been realized during the last 30 years,
while about 50-100 snow and ice storage systems are found in China (Kobiyama, 2000).
The USA, Canada and Sweden have also made efforts in the field.
3.1.1 Himuros and Yukimoros
A Himuro is a house, or a room, where vegetables are stored with ice to be preserved. In a
Yukimuro, snow is used instead of ice. The storage room can be situated in or on the
ground. Both are traditional ways to use snow/ice for cold storage in Japan. The food is
stored on shelves and the snow/ice is stored in cages or in one section of the room. The
food/snow ratio varies depending on the climate and thermal insulation of the building. The
cold is distributed by natural convection and can be roughly controlled by shutters, curtains,
-6-
etc. (Okajima et al, 1997). The temperature in a Himuro/Yukimuro is a few degrees above
0oC with a relative humidity of about 90-95%. One drawback is that the humidity and
temperature cannot be controlled with any precision (Kobiyama, 1997).
In a Japanese-Chinese project a partly buried Himuro was constructed in 1994 (Suzuki et
al, 1997). Water was added to crushed lake ice and a compact ice block was formed. After
an exceptionally warm summer and autumn one-third of the original ice block, 1053 m
(LWH), remained. Watermelons were stored for two months, longer than ever before in
this area, and vegetables were stored for three months. Hence, it was concluded that the
method was successful. The temperature in the food storage room varied from 0 to 10oC
during the entire year.
In a Canadian study, 380 tons of swedes (turnips) were stored in a partly empty storage
with 64.5 m3 of snow (41 tons) to one side of the building. A thin layer of sawdust
insulated the snow, and a fan and channel circulated the air. The swedes were stored from
the end of December to the end of June. When the temperature became too low a door was
opened; when it became too high, above 3oC, some sawdust was removed. At the end of
June the storage temperature was 8oC. The remaining vegetables were of good quality. It
was estimated that with 75-80 m3 of snow and an increased exposed snow surface, the
temperature could be kept below 6oC (Vigneault, 2000).
3.1.2 Indoor snow storages
In the Japanese All-Air-System, air is used as the cold carrier. Warm air is blown through
holes in the snow. The snow is covered with a thin water layer that absorbs particles and
gases, i.e. the air is cleaned during cooling. The hole spacing is about one meter and the
snow lies on a steel grid (Iijima et al, 1999).
The air temperature and relative humidity (after the snow cooling) vary with airflow rate,
inlet temperature and length of the snow holes, but depend only slightly on the diameter of
the hole. In the All-Air-System, temperature and humidity to the user part can be
controlled.
Three systems for temperature and humidity control have been developed. In the D control
system only the outdoor air can be controlled, Figure 4, considered suitable when only a
rough control of the temperature and humidity is needed and for buildings where air re-
circulation is not wanted. Air re-circulation (No. 1 in Figure 4) can, however, be included.
In the E control system outdoor air mixed with re-circulated air flows through the snow
holes. The other part of the re-circulated air is mixed with cooled air before entering the air-
conditioning area. This system is most common and allows the humidity to be well
controlled. In the DE system the temperature and humidity are precisely controlled. Here,
both outdoor and circulated air are controlled. However, this system is more complicated to
construct (Kobiyama et al, 1997).
-7-
Figure 4 The All-Air-System with D control (from Kobiyama et al, 1997).
Japanese indoor storage systems using water as a cold carrier also exist (Kaneko et al,
2000). A system with open water circulation intended for premises where water might be a
nice part of the design though not for cooling purpose was also under development at the
Muroran Institute of Technology in 2000 (Kobiyama, 2000). It is not known if the system
was ever implemented.
3.1.3 Icebox and Fabrikaglace
The Icebox and Fabrikaglace are Canadian inventions, both consisting of an un-insulated
box inside an insulated shelter. Thinly sprayed water layers only a few mm thick are frozen
in the inner box by cold air blown over the water. Without any area size constraints, it is
possible at some locations to form ice blocks more than 20 m thick during one winter. To
extract cold the melt water is pumped to a heat exchanger and then re-circulated over the
ice. Thermal expansion and ice creeping is handled by flexible walls. The tested plants
were from a few MW up to 250 MWh with cooling powers 8-1600 kW, but the high
construction cost required large plants or high cooling powers to become economically
feasible. The initial system was called Icebox, while the fully automated enclosed design
was known as Fabrikaglace (Buies, 1985; Morofsky, 1984).
The Icebox’s coefficient of performance (COP) was about 90-100 (Abdelnour et al, 1994;
Morofsky, 1982). It was concluded several years ago that the technique had not yet been
widely used because of the high construction cost and lack of knowledge by architects,
engineers and others (Canada, 2000).
Vigneault (2000) reports a Fabrikaglace of about 88 m2 that made 325 m3 of ice. Melt water
was used to pre-cool the vegetables after harvest, and was then re-circulated to the ice. The
water temperature after the ice was about 0.5oC. The cooling rate of the vegetables was
almost doubled relative the conventional system. Electricity usage decreased 83.6%, but the
investment cost was too high. It was concluded that ice blocks of, e.g., 0.28 m (in
Vancouver), 4.4 m (Toronto), 7.7 m (Ottawa) and 14.0 m (Winnipeg) thicknesses could be
made during a normal winter.
-8-
3.1.4 Under-ground storages

Snow storage in underground caverns has not yet been implemented. Johansson (1999)
investigated the prerequisites of snow cooling during the summer, focusing on under-
ground storage. Storages of 25,000-150,000 m3 were studied, assuming a snow density of
650 kg m-3. Simulations yielded a natural melting in Stockholm of about 3-6% during the
first years and 1-3% the tenth year. Based on typical district cooling costs, the studied
applications had payback times of less than one year.
In 1995 potatoes and rice were stored in an abandoned road tunnel in Japan, 2404.74 m
(LWH). During the winter, snow was taken into the tunnel and covered with aluminium-
coated tarpaulins, and the tunnel was divided into sections with plastic sheets. The produce
was stored in bags beside the snow piles and under the tarpaulins, with rice from the same
harvest being stored in a grain magazine and at a research centre for comparison. The rice
in the tunnel had best quality after two months. The potatoes under the tarpaulins retained
their quality the longest. It was concluded that it is possible to store food produce year
around if the tunnel is cooled with air during the winter (Suzuki et al, 1997).
3.1.5 Open pond storages
A number of different open snow and ice storage techniques have been studied and
implemented, where snow/ice is stored in ponds or pits with low or non-permeable sides
and bottom. The idea of cooling sprayed water with winter cold was patented in the USA
1836 (Taylor, 1985).
In Ottawa, Canada, a storage for 90,000 m3 of snow in an abandoned rock quarry was
studied, 120809.5 m (LWD). The mean cooling load was 7,000 kW. Simulations
demonstrated the necessity of insulation, and a light coloured PE tarpaulin on plastic
insulation was suggested, but no satisfactory insulation materials were found. A melt water
re-circulation system for cold extraction was suggested, with an estimated payback time of
10 years (Morofsky, 1981).
In the late 1970s, an ice pond measuring 20205 m (LWD) for comfort cooling was built
at Princeton University, USA. The pond contained 1,000 tons of ice and was made
watertight with a polyethylene sheet. The ice was produced by spraying water. It was
initially insulated with an aluminized mylar sheet covered with an inflated plastic cover.
Due to a storm, the cover was ripped and a 0.3 m layer of straw between the two tarpaulins
was used for the remainder of the season. Melt water was re-circulated by a device on the
snow surface (Kirkpatrick et al, 1985).
The following year a circular pond, 17 m diameter and 3 m deep, was constructed. Ice
totalling 600 m3 was stored under a rigid dome with a plastic cover that was lifted off
during ice making. At the end of the winter a polyurethane foam blanket with a heat
transfer coefficient of 0.75 W m-2 K-1 was placed onto the ice. The success of the research
plants led to a full scale plant with 7,000 tons of ice, 40506 m (LWD), for comfort
cooling of a 12,000 m2 office building. Snow was made with three snow guns and insulated
with a self supporting 493710 m (LWH) dome, consisting of two polyester reinforced
-9-
PVC sheets with 0.23 m of fibreglass and 0.3-0.6 m of air in between. The dome was lifted
5.5 m during ice making. The plant began operation in 1985. Inflatable tents, rigid arch
structures and different insulation sheets were also investigated. Inflatable tents were
rejected because of their continuous electricity consumption as well as warm air being
blown into the dome. Rigid structures were too expensive and insulation sheets involved
handling problems. Hence, no really satisfactory insulation was found (Kirkpatrick et al,
1987).
In another project for industrial cooling, four ponds measuring 35-600 m3 were built for
year-round cooling. The biggest pond was for summer cooling and the medium pond for
winter cooling. In the smallest ponds water was cooled at night and used for air
dehumidification in a storage room. Ice was produced with water sprayers (Taylor, 1985).
Vigneault (2000) reports on a 270 m3 ice pond, 9152 m (LWD), made in soil of low
hydraulic permeability for 200 tons of ice. The pond was filled with water during the
autumn and kept free from snow to increase ice production. At the end of the winter the
pond was insulated with 0.065 m Styrofoam and a light colour tarpaulin. Cold bottom water
was pumped to a cooling battery via a filter and then back into the pond. The cold was used
for potato storage. The average cooling load was 27.4 kW and the estimated COP was 2-4
times better than for a conventional system.
The New York State Energy Research and Development Authority in collaboration with
Snomax Technologies built a full-scale ice pond for research and to demonstrate where the
pond was sealed with a rubber sheet (CADDET, 1990). Operational energy savings of 89%
over a conventional mechanical cooling system and high reliability were reported. Ice
making was improved by using a bacterial protein for ice nucleation.
Näslund (2000) investigated a district cooling system in Sundsvall, Sweden, with sea water
and stored snow. The cooling load was 7,900 kW and 7,450-8,560 MWh. The idea was to
use natural snow from streets and squares and snow made from snow guns or water
spraying. The estimated snow proportion of the cooling load was 43.6-66.8%, requiring
122,500 m3 of snow. Two layers of 0.01 m plastic sheets fastened with weights, with a
thermal conductivity 0.05 W m-1 K-1, were recommended as insulation.
A study at the Polarbröd bakery in Älvsbyn, northern Sweden, investigated if snow could
replace part of the driving energy for the cooling machines (max power 1 MWel, mean
power 243 kWel). It was found that the required low cooling temperatures of -18oC to -38oC
restricted the advantages of snow cooling (Antonsen et al, 1998).
Näslund (2001) investigated a combined system for an industrial application with river
water as the base load and snow cooling for peak loads. The continuous cooling demands
were 1,500 kW at both 5oC and 15oC. This indicated a necessary snow volume of 121,230
m3 (78,800 ton) of snow to be stored in a 1201003 m (LWD) pond with a watertight
asphalt bottom. The estimated investment cost was 8,633,000 SEK (~948,700 €).
André et al (2001) investigated a snow cooling plant in Luleå, northern Sweden, for year-
round operation with a winter base load of 2 MW and a summer peak load of 6 MW. The
- 10 -
idea was to combine the local snow deposit with artificial snow. The suggested system
consisted of two main ponds and one smaller pond to facilitate maintenance.
The possibility to use snow for atomic power plant cooling, and thereby increasing
electricity production, was investigated in a Japanese study. With 38,400,000 tons of snow,
the seawater temperature should be lowered 9oC and an extra 1,800 MWh of electricity
could be produced annually. The low extra electricity output was due to the short diurnal
operation time. A layer of 0.15-0.3 m rice husk was suggested as thermal insulation
(Kamimura and Toita, 2004). In a connected experiment, 960 m3 of snow were stored under
a 0.15 and 0.30 m layer of rice husk, covered with straw mats and anti-scattering net to
avoid the rice husk being blown away. With 0.3 m of insulation the height decrease rate
was about 0.01 m d-1 during mid-summer, and the rice husk was concluded as an excellent
insulating material (Nakamura et al, 2004).
In a similar project involving the snow cooling of a Swedish waste gas power plant,
3,200,000 m3 (~2,080,000 ton) of snow would increase the production of electricity by 30.7
GWh annually. The snow should be stored in a 40040020 m (LWD) pond with
watertight sides, but an open bottom, since the groundwater table was situated above the
pond bottom level. Here, 0.2 m of wood chips was suggested as insulation (Falk et al,
2001).
3.1.6 Water purification
Another application is to combine ice production and water purification, since impurities
are pressed out as water freezes. It is possible to freeze sea water or polluted groundwater
and use the cleaned melt water for both cooling and drinking. In Greenport, New York, the
salinity of sea water was reduced from 3.0% to 0.00005% by freezing (Taylor, 1985).
3.2 The Sundsvall Hospital snow cooling plant

The snow cooling plant at the Sundsvall Regional Hospital in mid Sweden is an open pond
solution designed for 60,000 m3 of snow. It was first filled with snow during the winter of
1999-2000 and taken into cooling operation June 2000. To our knowledge it is the first
open storage with melt water re-circulation and wood chips as thermal insulation. Monthly
mean air temperatures during the cooling season (May to August) vary from 8 to 15oC,
while daytime temperatures often reach 25-30oC. The annual mean air temperature of
Sundsvall from 1961-1990 was 3.2oC. The surface area of the hospital is 190,000 m2 and
the plant is owned and managed by the County Council of Västernorrland (CCV).
The project started with theoretical studies and a small field experiment storage in 1998. A
total of 200 m3 of snow was stored in a 126 m (LW) enclosure with one-meter wooden
walls and 0.05 m thick polystyrene sheets as side insulation. Snow above the walls was
covered with 0.2 m of sawdust and a tarpaulin. With about 75% of the snow remaining at
the end of the summer, the melt rate was slightly lower than expected (Jansson, 1998).
In the theoretical pilot study by Nordell and Sundin (1998), a cooling need of 1,000 MWh
was assumed. The natural melting of 15,000 m3 and 30,000 m3 snow piles without thermal
- 11 -
insulation, as well as 0.1 and 0.2 m of sawdust as insulation was simulated. The simulated
snow pile had the shape of a frustum cone with an initial height of 4.0 m, diameter 105.6 m
and a side slope of 26.6o. The ratio h/d was kept constant throughout the melting. The snow
density was assumed to be 650 kg m-3.
The Surface Melt Rate (SMR) was calculated as the effective heat conduction through the
sawdust. Air temperature and a thermal conductivity of 0.35 W m-1K-1, the mean values of
dry sawdust (0.10 W m-1 K-1) and water (0.60 W m-1K-1), were used in the calculations.
The ground melt was calculated as one-dimensional heat conduction from the 6oC
groundwater, located at a depth of about one meter below the pond’s surface. The
groundwater temperature was assumed constant due to its high flow. Rain melt was
calculated with a temperature of 10oC from evenly distributed rain throughout the entire
summer, 240 mm in total.
Simulations demonstrated that an un-insulated 30,000 m3 snow pile would be melted away
by mid-June; consequently, the storage had to be thermally insulated. With 0.1 m of
sawdust as thermal insulation the remaining snow volumes was 12,169 m3, and 19,040 m3
with 0.2 m of sawdust, Figure 5. With 0.2 m of sawdust, the surface melt was 9,149 m3, the
ground melt was 1,421 m3 and rain melt was 390 m3 of snow. It was suggested that the
project should be implemented, preferably with a deeper storage and smaller upper surface.
35000
No insulation snötipp 100 mm sawdust 200 mm sawdust
30000
25000
volume (m3)
20000
15000
10000
5000
0
Feb 25 Mar 16 Apr 5 Apr 25 May 15 Jun 4 Jun 24 Jul 14 Aug 3 Aug 23
Figure 5 Natural melting of 30,000 m3 of snow with 0.1 and 0.2 m of sawdust and
without thermal insulation (from Nordell and Sundin, 1998).
3.2.1 Construction
The snow storage is a shallow pond (130642 m, LWD) with a slightly sloping (~1%)
asphalt surface, Figure 6. The storage is built for 60,000 m3 of snow, about 40,000 tons.
The watertight 0.1 m asphalt layer overlays 0.5 m gravel, 0.1 m insulation and 0.8 m of
sand.
- 12 -
Figure 6 The Sundsvall snow pond. The white construction is the pump house. The
right picture shows the storage with snow and wood chip insulation, in the
beginning of the summer 2001.
Cold is extracted by pumping melt water through heat exchangers connected to the cooling
system of the hospital. The heated melt water is re-circulated to the snow, where it is cooled
as new snow melts. The 36 original re-circulation inlets are controlled by valves located at
the sides of the storage. The water outlet is two openings in the pump house, in the lower
corner of the pond. The water is cleaned by a coarse filter, oil and gravel filter, and
automatically rinsed fine filters. Two pumps (0.035 and 0.050 m3 s-1) pump water to the
heat exchangers (1000+2000 kW). The snow cooling plant is outlined in Figure 7. The
comfort cooling system of the hospital also includes one 800 kW chiller. The system
primarily runs on snow cooling, with the support of the chiller when necessary.
Hospital
Hospital
Wood Chips
Filters Heat Exchangers
Pumps
Pumps
Snow
+2oC +5oC
+7
+8oC +12oC
+10
Return Pipes
Figure 7 Outline of the Sundsvall snow cooling plant.
- 13 -
3.2.2 Years of operation

Much experience has been gained since the first cooling season in the summer of 2000.
Snow
Natural snow from streets and squares and artificial snow from snow guns were used. The
artificial snow accounted for 37-70%. The stored snow volumes, measured with a geodetic
total station, were 18,800-40,700 m3, Table 1. The snow density, measured at five depths at
five locations in 2002 and at five depths at four locations in 2003, varied from 578-735 kg
m-3, with mean densities of 664 and 637 kg m-3. Snow and wood chips were spread and
compacted (the snow) during the first year with an excavator and a tractor. A snow groomer
was then used, making spreading much more efficient. According to CCV, cavities in the
snow will appear without compaction.
Different snow and ice making systems have been tested. Fan type snow guns and a
LowEnergyTower (section 3.4) were used during the first years. The fan type machines
required much maintenance and surveillance, while with the LowEnergyTower, large
amounts of snow landed outside the storage, possibly caused by improper location. During
the last years, two fan type snow guns mounted on five-meter high towers have been used,
and are now only operated in the wind direction. This works well according to CCV. In
2001, a pipe on the pump house roof was also used to spray autumn rain stored in the pond
onto the ice surface. The pipe was removed because the water melted the snow instead of
producing ice.
Cooling
The cooling demands varied largely over the summers and between the years. Of the total
cooling load (655-1345 MWh), 77-93% was delivered by the snow system. With increased
operational experience, the snow cooling has met a larger part of the total cooling load
during a larger part of the cooling season. The maximum cooling load increased during the
first years partly due to the expansion of the hospital’s cooling system and partly because of
the warmer weather, Table 1. The cooling system will continue to grow to about 3,000 kW
and 3,000 MWh by about 2010 (Larsson, 2005).
The snow cooling system has met most cooling demands until the end of June or beginning
of July, except for peak periods when the chiller was also used. The main part of the snow
system was designed for higher cooling loads than the extracted, but the re-circulation
system was found to be too small. Another problem was algae growth in the heat
exchangers; see the section “Melt water re-circulation”. The snow cooling seasons ended
between the middle of August and the beginning of September, Figure 8 and Figure 9.
- 14 -
2000
1800 2002 Snow
Chiller
cooling energy [kWh h ]
1600
-1
1400
1200
1000
800
600
400
200
0
13
18
15
24
07
20
04
01
29
27
10
7-
4-
5-
5-
6-
6-
6-
7-
8-
8-
9-
-0
-0
-0
-0
-0
-0
-0
-0
-0
-0
-0
02
02
02
02
02
02
02
02
02
02
02
Figure 8 Hourly cooling energy from snow cooling system and chiller in 2002.
2000
2004 Snow
cooling energy [kWh h ]
1800
-1
1600 Chiller
1400
1200
1000
800
600
400
200
0
18
27
10
24
07
20
04
01
15
29
13
7-
8-
9-
8-
4-
5-
5-
6-
6-
6-
7-
-0
-0
-0
-0
-0
-0
-0
-0
-0
-0
-0
04
04
04
04
04
04
04
04
04
04
04
Figure 9 Hourly cooling energy from snow cooling system and chiller in 2004.
The commonly used Coefficient of Performance (COP) only describes the performance of
operation. To invoke the total performance, a new coefficient COPtotal was developed.
COPtotal describes the ratio between delivered cooling energy and total energy needed to
produce the cold, including the total driving energy and annual material depreciation,
defined as the total energy required manufacturing and constructing the system divided by
estimated technical lifetime (Hagerman, 2000; Wichmann, 2003). COPtotal is always lower
than COP, but is a more relevant coefficient when comparing different systems.
To make a fair comparison, the COPtotal of the chiller was based on assuming the same
cooling amount from the chiller as from the snow cooling plant. COPtotal was 4.3-17.2 of
the snow cooling plant, 2.0-6.6 times better than the chiller system, while ordinary COP
was 5.2-26.6. This variation was due to thee different amounts of snow cooling energy and
used electricity. The amount of artificial snow was calculated as the ratio of water for snow
making over the estimated snow amount, i.e. losses at snow making were not included.
- 15 -
Table 1 Operation data 2000-2004 at the snow cooling plant at the Sundsvall
Regional Hospital.
1 2 2 2
Year 2000 2001 2002 2003 2004 2005
3
Snow volume [m ] 18,800 27,400 40,700 36,800 35,400 39,900
Share of artificial snow [%] 49% 59% 57% 37% 52% 70%
Snow cooling period 6/6-29/8 26/4-22/8 25/4-29/8 6/5-17/8 28/4-3/9
Total cooling energy [MWh] 655.5 1159.1 1345.3 1068.4 870.5
Snow cooling energy [MWh] 607.9 897.2 1125.9 894.5 799.6
Snow cooling proportion [%] 92.7% 77.4% 83.7% 83.7% 91.9%
Maximum total cooling power [kW] 1366 1648 2004 2034 1919
Maximum snow cooling power [kW] 1366 1148 1873 1508 1594
COPsnow [-] 5.2 15.7 23.8 7.4 6.9
COPsnow,total [-] 4.3 11.2 16.0 6.2 5.7
COPsnow,total/COPchiller,total [-] 2.0 3.3 6.6 2.6 2.4
1 th
Snow cooling started June 6
2
Municipal water was complemented with water from own well
Thermal insulation
A 0.2 m wood chip layer with 20-150 mm large pieces thermally insulates the snow. In
2000, about 75% of the wood chips were saved for the following year. The rest was burnt at
a nearby co-generation plant. The second year most of the wood chips were saved and a
small quantity mixed with a large proportion of gravel and sand was deposited. After the
third cooling season, the decay of the wood chips and amount of debris had increased too
much; thus, it was used as filling at a nearby ski slope. The cogeneration plant would not
use the wood chips because of too much debris. In 2003 new wood chips were bought and
all wood chips from 2003 and 2004 were saved. In 2005, 420 m3 of fresh wood chips were
added. Small cavities due to uneven melting were self-filled with insulation from the
closest surrounding.
In 2003, CCV made a minor experiment with 25 m insulation sheets, U-value 13.6 W m-2
K-1, covering an area of about 1010 m. This insulation initially worked well, but as the
snow started to melt unevenly, the sheets slid apart. The staff had to adjust the sheets
frequently, and it was concluded that this solution was not as good as wood chips (Larsson,
2005).
Melt water re-circulation
The melt water re-circulation system extracts the cold and delivers it to the hospital. Stable,
low temperatures (<~3oC) were achieved from about mid-June to mid-July. These
temperatures were supposed to be caused by longer periods of operation and proper re-
circulation strategy of water, though this could not be verified because of a lack of data.
Water flow rates during these periods corresponded to a detain time of 1.5-3 days,
assuming that 2/3 of the volume below water surface was occupied by snow.
- 16 -
After the end of July/mid August, water temperatures before entering the heat exchangers
started to increase because of decreasing snow and shortages in the re-circulation system.
After an intense rain in 2000, the outlet water temperature was 15 oC.
In the autumn of 2001, the storage was reconstructed to render a higher water level in the
pond, with the aim to receive a longer detain time and thereby ensure a lower temperature
of re-circulated water. Direct re-circulation, from pump house to pond, was also enabled
when the outlet water was too warm.
Wood chips fell off because snow around the inlet valves melted quickly. As snow then
became directly exposed to sun and wind, the snow melt rate increased and more wood
chips fell off. This resulted in an increased natural melting as well as re-circulated water
finding short cuts along the sides of the pond. In a minor experiment water was sprayed
evenly over a 200 m2 area for four weeks. Here, snow melted without any formation of
cavities.
To decrease side-melting and outlet water temperatures, new re-circulation hoses were
installed. The hoses were placed on top of the snow the first year, resulting in holes and
exposed snow. In 2001 the hoses were placed at the far side, and in 2002-2004 the hoses
were placed underneath the snow. In 2002 the snow groomer damaged the hoses, but in
2003 and 2004 outlet water temperature decreased. Water temperatures from 2002 and
2004 are seen in Figure 10 and Figure 11.
The average snow cooling power, water flow rate through heat exchangers, water
temperature before the heat exchangers and water temperature increase in the heat
exchangers, in June and July 2001-2004, are seen in Table 2.
25
melt water temperature [ C]
2002 before hx
o
20 after hx
15
10
0
20
04
18
01
15
29
13
27
10
24
07
4-
5-
5-
6-
6-
6-
7-
7-
8-
8-
9-
-0
-0
-0
-0
-0
-0
-0
-0
-0
-0
-0
02
02
02
02
02
02
02
02
02
02
02
Figure 10 Melt water temperatures before and after the heat exchangers (hx),
in 2002.
- 17 -
25
melt water tempearture [ C]
o 2004 before hx
20 after hx
15
10
0
0
7
-2
-0
-1
-0
-1
-2
-1
-2
-1
-2
-0
4
9
-0
-0
-0
-0
-0
-0
-0
-0
-0
-0
-0
04
04
04
04
04
04
04
04
04
04
04
Figure 11 Melt water temperatures before and after the heat exchangers (hx), in
2004.
Table 2 Melt water outlet data for 2001-2004.
June+July 2001 2002 2003 2004

Average snow cooling load [kW] 457.5 453.4 476.6 276.9
Average water flow through heat exchangers [m3 d-1] 1842.6 2567.7 2272.7 1595.0
Average water temperature before heat exchangers [oC] 4.1 3.3 3.0 2.2
Average water temperature increase in heat exchangers [oC] 4.9 3.6 3.8 3.7
The temperature increase in the heat exchangers was usually 1-8oC. The decreased
temperature difference in July 2002 and 2004 was due to warmer outlet water and algae
growth in heat exchangers; see below. The system tries to compensate for the warmer outlet
water with a larger flow, but as the algae content increased, the flow resistance and
maximum flow rate decreased. The heat exchangers were cleaned with lye at the end of
July 2002 and beginning of August 2004, and the temperature difference thus increased
again. The decreased water temperature difference in the heat exchangers also depended on
decreased temperatures on the secondary side of the heat exchangers, from 13.0oC in 2001
to 8.1 oC in 2004, before heat exchangers.
Avoiding short cuts and increasing outlet water temperature is possible by applying re-
circulation melt water on top of the snow, though CCV has so far avoided this because they
do not want pollution to be blown by wind to the hospital. Despite the polluted snow,
spraying re-circulated water would be possible if it can be continuously filtered. During the
autumn of 2005, the height of the south pond side will be increased to enable a horizontal
snow surface. The idea is to obtain a more favourable incidence angle of solar radiation,
and thereby reduce melt. There are also plans to drill wells for snow making, install a water
- 18 -
spraying system, increase the capacity of the re-circulation system, and build two new
water outlet points in the centre of the pond bottom (Larsson, 2005).
The algae growth discovered in August 2001 escalated in 2002. This decreased the heat
transfer in the heat exchangers and the water flow rate. The heat exchangers were cleaned
with lye one to three times per summer from 2002-2004. CCV believed that an increased
decay rate of the old wood chips and high temperatures in the heat exchangers caused the
growth. Algae growth decreased radically during 2003 when new wood chips were used.
The algae problem can probably be avoided by using new wood chips every year or by
removing wood chips from the pond during the summer. These solutions not only increase
the possibilities to burn the wood chips after the cooling season, but also the cost. An
alternative would be to clean the heat exchangers when the occurring pressure drop exceeds
a certain value.
Evaporative cooling
The amount of melt water let out to a nearby ditch is measured both during the cooling
season and when emptying the storage after the season. During the cooling season melt
water is continuously let out after the heat exchangers to keep a constant water level in the
pond. Comparing the estimated snow amount plus rain versus the water amount let out
from the system is thus possible. Problems due to an opened valve before 2002 and
measuring equipment in 2003 only enabled comparisons in 2002 and 2004.
In 2002 and 2004 the total mass of snow and rain was 28,637 and 25,213 tons. The
measured water discharge was 25,082 and 24,430 tons, giving differences of 3,555 and 783
tons. Evaporated, this water corresponds to 296.9 (2002) and 65.4 (2004) kWh m-2,
assuming an area of 13064 m. There are, however, some unaccounted water losses, e.g.
from the rinsing of fine filters and emptying of the pond after the cooling season. The solar
radiation from May to August in 2002 and 2004 was 659.0 and 593.2 kWh m-2, according
to the Swedish Meteorological and Hydrological Institute (SMHI). Assuming a solar
radiation reflectivity of 20%, the evaporative cooling as energy corresponded to 56.3%
(2002) and 13.8% (2004) of the solar radiation. This agrees well with results of experiments
and simulation, see section 4.2.
Weather
Air temperature, relative humidity, precipitation, solar radiation, air velocity and wind
direction were measured at a weather station close to the pump house. The summers of
2000-2004 were warmer and wetter than the mean summer of 1961-1990. In 2002, the
summer mean temperature was 4.1oC warmer than the average year. In 2001 and 2002,
precipitation was 52.2% and 71.2% greater than the average year, Table 3.
The mean values of 1961-1990 are from the Sundsvall Airport, located about 30 km from
the snow cooling plant. Good correlation was found between the measured air temperature
and precipitation and values from the airport climate station during 60 days the summer of
2000. The airport climate station is run by SMHI.
- 19 -
Table 3 Monthly mean temperatures and total monthly precipitation during the
cooling seasons of 2000-2004 at the Sundsvall Regional Hospital Snow
Storage. Mean values between 1961-1990 are from the Sundsvall airport 30
km away.
o
Mean temperature [ C] Precipitation [mm]
May June July August May June July August Total
2000 9.5 12.6 16.5 15.0 45.4 72.3 146.4 37.3 301.4
2001 9.3 15.1 17.8 15.4 33.0 15.4 68.8 221.8 339.0
2002 11.4 17.6 18.4 19.4 84.0 95.0 62.0 21.2 262.2
2003 8.7 14.1 18.9 15.6 41.0 79.6 13.0 134.2 267.8
2004 9.3 14.0 16.5 16.7 26.8 52.8 97.0 88.2 264.8
1961-1990 7.8 13.4 15.3 14.0 35 41 58 64 198
Figure 12 shows measured snow volumes in 2001 and simulated melt by the model used in
the pilot study, run with measured climate and cooling data of 2001. At the beginning the
curves fit but then the actual melting was faster. This was probably because of the increased
un-insulated snow surface area as the cooling season continued and wood chips fell off, and
because the area that absorbs solar radiation decreased with time in the model. In the real
snow storage solar radiation is absorbed over almost the whole area the whole summer,
either by wood chips, snow or water in the pond.
30 000
Calculated
25 000
snow volume [m ]
3
Measured
20 000
15 000
10 000
5 000
0
26-apr 16-maj 5-jun 25-jun 15-jul 4-aug 24-aug
Figure 12 Calculated and measured snow volume of 2001.
The ground temperature under the storage pond was measured both above and below the
ground insulation at 0.6 m depth. Above the insulation the ground temperature followed the
air temperature with a slight trailing when no water or snow covered the storage bottom,
otherwise it was 1-5oC above zero. The ground temperature below the insulation was about
4-5oC from spring to the end of snow cooling season, when after it increased a few degrees.
Environment
A number of melt water quality analyses of heavy metals, hydrocarbons, oxygen demands,
and nutrients were made. Measurements were made in the snow storage, in the stream
where melt water is discharged and in the recipient, totally at 7 locations. Reference
- 20 -
measurements were made at a nearby location not affected by the cooling plant. The results
were compared with Swedish environmental quality criteria (SEPA, 1990; SEPA, 1999).
The concentration of non-biodegradable compounds (measured as COD-Mn) was rather
high, especially in the second half of the cooling season. These were supposed to origin
from the wood chips (Ericsson, 2003). The content of phosphorus and some heavy metals
in the snow were high, but in the outlet water most substances were considerably reduced.
This was related to particle adsorption, as particles settled in the snow basin, which agrees
with performed studies (Viklander and Malmqvist, 1993). The effect on the recipient was
limited or non detectable for all parameters, relative the reference location (Ericsson, 2003).
The phosphorus content in the snow was probably caused by droppings from dogs and
birds (Malmqvist, 1983).
An unknown amount of the substances described above were lost in the oil and gravel
separator and by flushing of the fine filters. An attempt to make a full mass balance failed.
3.2.3 Economy
The total investment cost of the Sundsvall snow cooling plant was about 14.5 MSEK, or
1.59 M€. Since this full scale operation plant was made also for experiments, research and
demonstration it was difficult to distribute the cost in an appropriate way, e.g. the cost of
reconstructions. An estimated cost split up is shown in Table 4 (Larsson, 2005).
Table 4 Investment costs for Sundsvall Regional Hospital snow cooling plant.
[kSEK] [k€]
Pond construction 4800 527.5
Ground thermal insulatio 1000 109.9
Fence + vehicle approach 700 76.9
Pump house 1000 109.9
Pumps, pipes etc. 4000 439.6
Electrical installations 1000 109.9
Control system 600 65.9
Planning 1400 153.8
TOTAL 14,500 1593.4
The costs of cooling operations during 2002/2003 and 2003/2004 are seen in Table 5. The
cost of the first three years was not available, but CCV estimates that the total operation
cost has decreased each year.
The “Snow handling incl. production” was expensive in 2002/2003, when a contractor was
responsible. The depositing of urban snow became an income from the winter 2002/2003,
since CCV received a fee of 150 SEK per lorry-load of snow. “Cold production” means
operation cost during the summer, excluding electricity and water. “Post season work”
includes removing and storing wood chips, removing sediments and cleaning the pond.
- 21 -
Table 5 Operation cooling cost from the Sundsvall Regional Hospital snow cooling
plant (Larsson, 2005).
2002/2003 2003/2004
Artificial snow share 37% 52%
3
Artificial snow amount [m ] 13,984 18,408
Snow handling incl. production [SEK] 1,170,030 733,624
Natural snow [SEK] -141,000 -98,100
Cold production [SEK] 199,781 167,932
Post season work [SEK] 147,652 134,291
Electricity [SEK] 35,551 45,893
Water [SEK] 34,587 33,485
Operation cooling cost [SEK kWh-1] 1.62 1.27
“Electricity” depends on the amount of artificial snow. The cost of water has decreased due
to their private well being used since 2002. This also increased electricity usage. The
municipal water cost is 5 SEK m-3 (Larsson, 2005).
Larsson (2005) predicts that the operation cooling cost in 2004/2005 will be 0.85 SEK
kWh-1, and about 0.50 SEK kWh-1 in 2010. Thanks to the commitment of CCV, occurring
problems have been solved, necessary reconstructions have been made and the plant has
worked satisfactorily.
3.3 Boundary conditions for snow cooling

The boundary conditions for snow storages on the ground are social, political, economical,
environmental, natural and system related. Here, the social, political, economical and
environmental boundary conditions are treated briefly, while the natural and system
boundary conditions are treated more extensively.
The usage of snow cooling systems presupposes that the technique is known. However, the
awareness of modern snow and ice cooling systems is limited to a fairly small number of
countries. Also, in countries that have snow cooling systems, knowledge about the
technique is generally low among engineers and authorities at all levels. There is also a lack
of education in system construction and design. Global warming and the increasing cost of
fossil fuel, however, force all countries to increase the usage of renewable energy (Paksoy,
2003:2). Further, the Kyoto Protocol should be an incentive to more snow cooling plants.
The economical boundary condition varies largely around the world and generally depends
on the cost of water, electricity, fuel, labour and land. Natural snow means a cost if it is
collected only for cooling, but might be an income if snow depositors pay a dumping fee.
Snow cooling can be an alternative to other technologies as well as enable cooling systems
at locations where conventional systems are not applicable. Estimations of the actual
cooling cost are further treated in chapter 8.
- 22 -
Environmental boundary conditions concern both chillers and snow cooling. Some
examples are the reduction of green house gases, elimination of ozone depleting gases,
urban noise reduction, aesthetic concerns and peak shaving. Snow contaminations, noise
from snowmaking, transports and land/ground usage are also relevant (Paksoy, 2003:2).
Snow pollution might be both a drawback and a benefit for snow cooling. The negative
aspect is that the pollution is concentrated to one location, the positive that it can be
controlled and treated.
3.3.1 Natural boundary conditions
Natural boundary conditions involve available land area, air temperature, humidity,
precipitation, solar radiation, water availability, groundwater flow and level, soil properties
and groundwater usage.
Snow precipitation and the deposited amount of natural snow vary greatly. For example, in
Stockholm 0-2,000,000 m3 was deposited annually from 1961-1992 (Johansson, 1999),
whereas in Sundsvall 12,000-267,000 m3 of snow was deposited annually from 1995-1999
(Näslund, 2000). Even if the annual snow precipitation over two years is the same, the
deposited amount varies greatly. One explanation is that an intense snow fall means that the
snow has to be removed. If the same amount of snow falls during a longer period less snow
is removed and deposited.
Artificial snow is preferably produced by cold water at a low air temperature, to some limit
where the equipment does not work. Low humidity is also beneficial, since evaporation
from water droplets, i.e. heat emission, and thus the freezing rate then increase. If
producing artificial snow/ice is at all possible, producing enough of it is possible, since
short and/or warm winters can be compensated by more snow guns. The potential snow
amount varies with the type of snow gun, see section 3.4.
During the cooling season, the climate affects both natural melting and cooling demand.
Hence, the difference in the amounts of snow needed between a normal year and a warm
year is doubly influenced. Natural variations in the climate must be carefully considered.
Since seasonal snow storage is a long-term investment with an estimated technical life of
40 years, global warming will also influence predictions.
Water availability is important if using artificial snow. Saving pond water during a period
and using it for snow making when the temperature is low enough is possible, as well as
reusing water if pollution problems can be handled. Small particles (<1 mm) are beneficial
for ice nucleation, but dissolved pollution is not, since it decreases the freezing point
(Ågren, 2005). It might, however, be possible to treat pond water after the cooling season.
Soil properties influence pond construction and the groundwater flow. The groundwater
flow and groundwater level influence both natural melting and suitable storage
constructions, see sections 4.4 and 6.1. Large groundwater flows increases the ground melt.
In watertight ponds, snow pollution is concentrated and treatable. However, if the pond is
permeable, studying the migration of pollutants is necessary. This is important if nearby
groundwater is used for drinking or irrigation, see section 7.2.
- 23 -
3.3.2 System boundary conditions

The required water temperatures and load characteristics determine if snow cooling is
suitable for a certain application. One general advantage is that snow cooling enables
cooling temperatures close to 0oC without any risk of freezing.
The melt temperature of snow is about 0oC, whereas seawater melts at about -2oC.
Decreasing the melting temperature, and thereby the water temperature, by adding salt to
the re-circulation water would be possible. This might have negative consequences, though
the subject was not studied further. It is also possible to reduce the driving energy of
freezers by using melt water instead of air as a heat sink for the cooling media. This is
beneficial, since the temperature difference between the cold and the warm side decreases
as well as water having greater heat transfer capacity than air. However, further
development of chillers to enable condenser temperatures of about 5oC is necessary (Hill,
2004).
The required cooling energy and cooling power influence the plant size and design. In
general, a snow cooling system has no power limit, since the cold carrier is circulated
through the snow at the same rate as it is pumped to the object that is being cooled. In an
open water circulation system the delay between water inflow and outflow requires a buffer
to meet load variations, i.e. it is necessary with a certain amount of water in the pond.
3.4 Snow and ice production

Both natural and artificial (produced) snow might be used for cooling. Unless the storage is
used as a snow deposit, snow or ice must be collected or produced. Collecting snow or ice
exclusively for cooling is expensive (Johansson, 1999). Therefore, it might be economically
feasible to produce snow or ice at the plant.
The easiest way to produce ice is to freeze a water-filled pond during the winter. To
enhance ice production it is possible to remove the snow, pump water from the pond
bottom to the ice surface or apply water intermittently. Energy usage is less than 0.1 kWh
ton-1. Spraying water in the air increases the ice production because the direct contact
between water and air increases. Energy usage is about 0.5 kWh ton-1. Water is taken from
rivers, lakes, sea, ground or community net. The ice/snow production is increased with
lower water temperature. Pre-cooling is an option if the water is warm.
Production increases further with snow guns and the energy usage is 1-3 kWh ton-1. Two
examples of snow guns are given here.
One snow gun type is called LowEnergyTower, Figure 13. Here, pressurized water and air
are sprayed separately and mixed in the air. The water flow rate is adapted to air
temperature by valves. One example is TopGun 11 from JL Toppteknik (1999), which
works with air pressures of about 7 bars and water pressures of 15-50 bars. It uses about 18
kWel and produces 5-60 tons of snow per hour.
With a fan type snow gun, pressurized water (8-35 bar) and air are sprayed into an open,
barrel like construction with a fan at one end, Figure 14. Water and air are blown out from
- 24 -
the device and snow is formed. This type of snow gun has its own air compressor. Areco
Supersnow fan type snow gun (Areco, 1999) uses about 20-25 kWel and produces 10-45
tons of snow per hour. Fan type snow guns should not be operated below -30oC.
Figure 13 LowEnergyTower snow gun. Photo; Pär K Olsson, JL Toppteknik.
Figure 14 Fan type snow gun. Photo; Kjell Skogsberg.
The Snow Production Potential (SPP) [ton h-1] is related to the wet bulb temperature (Twb),
i.e. the temperature that occurs when a wet bulb is ventilated by air. Energy from the air is
used to evaporate water, thus decreasing the air temperature. A lower relative humidity
- 25 -
gives a bigger difference between the ordinary (dry) air temperature and wet bulb
temperature. According to manufacturers, SPP varies with wet bulb temperature, Figure 15.
Snow production is thus a balance between waiting for lower temperatures to make cheaper
snow and the risk of a snow shortage.
70
60 Supersnow
TopGun 11
50
SPP [ton h-1]
40
30
20
10
0
-35 -30 -25 -20 -15 -10 -5 0
T,wb [oC]
Figure 15 Snow Production Potential SPP [ton snow h-1] of two snow guns, as
function of wet bulb temperature (T,wb). Areco Supersnow is a fan type
snow gun and JL Toppteknik TopGun 11 is a LowEnergyTower type
snow gun. Snow production starts at -2oC.
SPP was investigated for five Swedish and four European cities. The hourly 1995 wet bulb
temperatures from Meteonorm (2000) were adjusted with the difference between the local
mean temperature from 1995 and the local mean temperatures from 1961-1990. SPP was
calculated for five different years; the adjusted curve from 1995 (base year) and four cases
where temperatures were adjusted one and two degrees above and below the base year.
This corresponds to years that were very warm, warm, normal, cold and very cold. It was
also assumed that only 75% of the maximum potential could be achieved, since snow guns
are normally not run during shorter periods below freezing. The results are seen in Table 6
and Table 7.
It was concluded that it is possible to produce snow in large parts of Europe. Taylor (1985)
estimated that 500 hours below freezing were necessary for ice ponds, which are found in
all of Canada, large parts of the USA and Europe, northern Asia, and at altitudes over 3,000
m in most remaining parts of the world.
- 26 -
Table 6 Snow Production Potential SPP [ton year-1 snow gun-1] for eight cities, with
JL Toppteknik TopGun 11 snow gun.
SPP [ton year-1 snow gun-1]
Tmean+2 Tmean+1 Tmean Tmean-1 Tmean-2
Luleå 84 769 95 670 107 087 119 207 131 345
Borlänge 37 976 47 128 57 594 70 011 83 576
Stockholm 19 802 24 770 30 862 38 180 46 828
Gothenburg 6 508 8 588 11 370 15 709 22 007
Lund 4 762 6 438 9 317 13 446 18 720
Hamburg 3 883 5 272 7 322 10 875 15 504
Bergen 2 442 4 147 6 810 10 735 16 924
Warsaw 18 847 23 974 29 905 36 873 44 911
Vienna 6 334 9 076 13 264 18 282 24 506
Table 7 Snow Production Potential SPP [ton year-1 snow gun-1] for eight cities, with
Areco Supersnow snow gun.
SPP [ton/(year*snow gun)]
Tmean+2 Tmean+1 Tmean Tmean-1 Tmean-2
Luleå 105 692 116 366 127 352 138 774 150 466
Borlänge 63 847 77 160 90 041 103 882 117 299
Stockholm 32 264 41 171 51 498 61 864 73 909
Gothenburg 10 597 13 745 18 496 27 033 39 899
Lund 8 674 11 249 16 582 24 656 34 767
Hamburg 7 178 9 386 12 900 20 119 29 573
Bergen 5 741 9 575 15 461 23 821 36 915
Warsaw 32 159 39 995 48 845 58 873 69 625
Vienna 11 069 16 204 24 723 34 125 45 157
3.5 Thermal insulation

Numerous possible thermal insulations for snow storage exist. Insulation material affects
plant operation strategies and labour costs. Different insulations have different lives and
environmental impacts. So far, no satisfactory thermal insulation has been found.
Superstructures are too expensive and disturb snow production. Insulation sheets might
slide apart due to uneven snowmelt and involve other handling problems, e.g. algae growth.
There are also a number of loose fill alternatives, with different advantages and drawbacks.
One general disadvantage is that applying and removing loose fills is work intensive.
Wood chips are a traditional thermal insulation of snow. Different kinds of wood chips
exist, e.g. wood powder, sawdust, cutter shavings and bark. Larger pieces of wood chips are
used in the Sundsvall snow cooling plant. According to Kosonen (2005) wood powder is
the best wood chips insulation on snow. No scientific work has however been found on this
subject. Thermal insulation qualities of cutter shavings are treated in section 4.2. One
general drawback with wood chips is its decay and deteriorating thermal insulating
- 27 -
properties. New material must be added each year; after some years, all the material must
be replaced (Skogsberg, 2003). Bark is another alternative, but no studies were found. Bark
is darker with a small or non-existent capillary transfer and whose decay is likely to be
slower.
One material not investigated in a snow cooling context is debris (mineral particles). Many
glaciers are covered by a layer of debris. The measured glacier melt rate increases for very
thin layers (<0.03 m), but decreases for thicker layers, relative to bare ice. A 0.10 m of
debris reduced the melt rate 35%-66% (Mattson and Gardner, 1991; Mattson et al, 1993;
Kayastha et al, 2000; Pelto, 2000; Takeuchi et al, 2000). With 0.40 m of debris the melt rate
was reduced 59%-85% (Mattson and Gardner, 1991; Kayastha et al, 2000). The increased
ablation rate with thin debris layers was because of increased solar radiation absorptivity.
The decreased ablation rate for thicker layers was because of thermal resistance, heat
storage and evaporation (Mattson and Gardner, 1991; Takeuchi et al, 2000). With urban
snow, the debris insulation implies a one-time cost, since the amount of debris increases
each year. Another advantage is that there is no decay. It is unknown how organic
compounds influence the thermal qualities of debris.
The choice of thermal insulation is intimately connected with a cold carrier system, see
section 6.2. This subject obviously needs further studies.
- 28 -
4 NATURAL MELTING
In an open snow storage, heat and moisture transfer are closely connected to natural
melting due to rain, capillary transfer in loose fill insulations, evaporation and
condensation. Natural melting is divided in surface melt, rain melt and ground melt.
4.1 Moisture and heat transfer

4.1.1 Moisture in air and materials
Here, moisture denotes both liquid and vapour water and water denotes only the liquid
phase. Air vapour content (v) [kg m-3] depends on air temperature (Tair) [oC] and relative
humidity (RH) [%]. One common relation for maximum vapour content in air (vsat) as a
function of temperature between 0 and 30oC is from the DIN 4108 (1981).
n 8.02
§ T · § T ·
M water a ¨ b air ¸ 18.02 288.68 ¨1.098 air ¸
© 100 ¹ © 100 ¹
v sat ,air (1)
R Tair ,K 8314.3 273.15 Tair
Mwater is the molar weight of water [kg kmol-1], R is the universal gas constant [J kmol-1 K-
1
] and Tair,K is air temperature in [K]. The coefficients a and b are empirical.
A porous material, or specimen, can be both permeable and impermeable. Here, porous
means a material that is porous and permeable. The moisture in a porous material can be
part of the structure or bonded in different chemical and physical ways. Moisture that is
able to vaporise is henceforth called moisture (Ahlgren, 1972).
The moisture content of a material (w) [kg m-3] can be measured up to 15 different ways.
One common method is the gravimetric, where a specimen is dried in an oven at 105oC
until constant weight is attained. The moisture content will then be low and in equilibrium
with the relative humidity in the oven. The residual moisture ratio of the specimen will
typically be about 0.5% of moisture per kg dry material (Skaar, 1988). In a porous material
w depends on RH, Tair, temperature of the material, if the material is in contact with water
and the history of the material. Some notions are introduced to help comprehend better. The
porous material is divided in a solid part and a pore part, Figure 16. The moisture content is
m moist
w (2)
V
where mmoist is moisture weight [kg] and Vwc is sample volume [m3]. Another measure is
moisture ratio (u) [kg kg-1][%],
m moist
u (3)
m dry
where mdry is weight of the dried wood chips [kg].
- 29 -
solid total pore volume
hygroscopic volume
whygr wcr wcap wsat

Figure16 Schematic picture of a porous material with the pore volume divided into
moisture relevant sections. The relative relations are not accurate.
For compactness comparisons the dry density (Udry) [kg m-3] is needed. Assuming the same
volume of wet and dry sample the dry density is
m dry m wet
ȡ dry . (4)
V V 1 u
The relation between moisture content and moisture ratio is
m moist ȡ dry
w u ȡ dry . (5)
m dry
A porous material in humid air with constant relative humidity and constant temperature
will eventually reach its hygroscopic equilibrium moisture content. Increasing hygroscopic
moisture content is called absorption, whereas decreasing is desorption. During absorption
moisture is first adsorbed in monomolecular layers, then in multiple layers and finally by
capillary condensation. The hygroscopic moisture content mainly depends on RH, called
sorption curve or isotherm, Figure 17A. The sorption curve follows different paths
depending on material, and if the material is in absorption or desorption, called hysteresis.
The air temperature dependence is small; w only increases about 1% per 10oC decrease for
a given RH (Stamm, 1964). It is custom to relate w only to RH at 20oC (Hagentoft, 2001).
The slope of the sorption isotherm, dw/dRH, is denoted [; Figure 17B. It is used to define
the moisture diffusivity coefficient of the hygroscopic moisture content shown below.
The sorption isotherms of Figure 17A are based on spruce measurements with Udry=430 kg
m-3 and converted to cutter shavings with Udry=90 kg m-3 by a factor 90/430=|0.2093, after
Hedenblad (1996). The macro porosity of a wood chip layer with Udry=90 kg m-3 is then
about 80%. Density and moisture characteristics vary with wood type. The dry density of
spruce (Picea abies) is 390-480 kg m-3, and 480-530 kg m-3 for pine (Pinus silvestris)
(Saarman, 1992). The moisture ratio sorption isotherms of spruce (Udry=420 kg m-3) and
pine (Udry=510 kg m-3) are almost identical (Ahlgren, 1972); therefore, the moisture content
sorption isotherm of pine is about 20% higher than that of spruce.
- 30 -
A B
35 100
30 w,des
w,abs 80
25
wwc [kg m ]
-3
dw/dRH
20 60
15 40
10
20
5
0 0
0% 20% 40% 60% 80% 100% 0 10 20 30
RH [%] w [kg m-3]
Figure 17 A) Sorption isotherms for cutter shavings with dry density 90 kg m-3,
based on curves of Swedish spruce with dry density 430 kg m-3 (after
Hedenblad, 1996). B) Slope of the mean sorption isotherm of A. (w,des
and w,abs are hygroscopic equilibrium moisture contents of absorption
and desorption)
At the fibre-saturation point, no water is found in the cell cavities, though the cell walls are
fully saturated with water. This corresponds to an equilibrium moisture content at 100%
RH, denoted whygr. This point is not well defined, and sorption curves therefore normally
stop at around 98% RH. Below the fibre-saturation point a moisture content change causes
swelling or shrinking. The maximum shrinkage is 3-10%, depending on direction and wood
quality (Skaar, 1988). RH and pressure in the capillary condensed water satisfy Kelvin’s
law at equilibrium,
R Tair,K ȡ water R Tair,K ȡ water
p pore p sat Tair ln RH | ln RH . (6)
M water M water
Assuming Tair=20oC the pore water pressures are -691 bar at RH=60% and -14 bar at
RH=99%. The large negative pore water pressures cause a suction that is balanced by the
surface tension of water (see next section). Larger pressure differences require smaller
pores.
The sorption isotherms of a specimen are measured in climatic chambers that control
temperature and humidity, or in a sealed chamber with some saturated salt solution at
constant temperature (Nevander and Elmarsson, 1994).
Reaching the critical moisture content level wcr, Figure 16, by hygroscopic absorption is not
possible. To get water in the cell cavities, the material has to be in contact with water. The
maximum water content that a material can reach by capillary suction, by contact with
water, is called the capillary saturated water content, wcap. To achieve a totally saturated
material, wsat, all trapped air must be removed. This can be done be repeated boiling or
vacuum treatment.
- 31 -
4.1.2 Moisture transfer

In porous materials, moisture is transported as vapour and water. The transfer rate is
calculated as the product of a transfer coefficient and some potential gradient (v, w, p,
T). Vapour transfer takes place by diffusion ( m diff ) and mass convection ( m
conv ). Liquid
transfer might be divided in flow due to pressures above atmospheric pressure mainly
caused by gravity and denoted Darcy flow, and due to sub atmospheric pressure, denoted
suction (s). Suction is the difference between air and water pressure, i.e. a negative
pressure. In Darcy flows, water is pressed into the pores of the material, with the transfer
mainly occurring in the bigger pores. In sub-atmospheric pressure the transfer occurs by
adhesive forces and mainly in the smaller pores.
Moisture convection is both a surface and a bulk phenomenon. When air passes through a
conv ) [kg m-2 s-1] into the material is
porous material the convective moisture transfer rate ( m
conv
m v v
V (7)
air in out
is the air volume flow rate [m3 m-2 s-1=m s-1]. The condensation rate is
where Vair
cond
m ȕ v air v sat,surf (8)
where E is the moisture transfer coefficient [m s-1] and vsat,surf is the saturation vapour
pressure of the surface [kg m-3]. E depends on the air velocity and surface properties, often
related to the convective heat transfer coefficient Dconv [W m-2 K-1] by the Lewis
approximation,
D conv
E conv (9)
U air c p ,air
where cp,air is the air heat capacity at constant pressure [J kg-1 K-1]. In a porous material
evaporation, or desorption, occurs when vsat,surf>vair and w>weq,des (Hagentoft, 2001).
Vapour diffusion in a porous material is described by Fick’s diffusion law, which in one
dimension (1D) is
D dv
diff
m G vap v (10)
P por dx
where m diff is the diffusive vapour flow rate [kg m-2 s-1], Gvap is the vapour permeability
coefficient [m2 s-1], v is the vapour content gradient [g m-3 m-1], D is the vapour
diffusivity in air [m2 s-1] , Ppor is a resistance coefficient [-] (Ppor1), dv is the vapour
content difference [g m-3] and dx is the distance over which the vapour content difference
occurs [m]. The vapour permeability coefficient of sawdust is 1510-6 m2 s-1 (Nevander and
Elmarsson, 1994). The actual moisture transfer at hygroscopic moisture content might be
purely diffusion or a mix of diffusion and suction or surface migration (Hagentoft, 2001).
The water flow through a porous layer can be estimated by Darcy’s law,
- 32 -
k ǻp

m ȡ wat (11)
ȝ ǻx
where k is the material permeability [m2], P is the dynamic viscosity [N s m-2=Pa s=kg m-1
s-1], and 'p is the pressure difference [Pa] over the distance 'x [m]. Either suction or over-
pressures causes the transfer. Here, over-pressures are mostly relevant for groundwater
flow.
The permeability is dependent on the type of liquid transfer. Darcy’s law, Equation 11, is
mainly used for saturated flow with over-pressures, e.g. groundwater flow, since suction
curves are normally not available. It is common to use empirically determined
permeabilities. The air permeability of cutter shavings varies from 210-10 m2 for
compressed material (Udry=240 kg m-3) to 1.1510-8 m2 for loosely filled (Udry=100 kg m-3)
(Nevander and Elmarsson, 1994). Typical permeabilities in soils vary between 1.510-16 m2
(clay till) to 1.510-10 m2 (coarse sand) (Domenico and Schwartz, 1990).
Suction occurs because of adhesive attraction forces between water molecules and the
material, as well as the cohesive forces between water molecules that cause surface tension
or surface energy (V) [N m-1=J m-2]. The relation between the cohesive and adhesive forces
is described by the contact angle T. An increased T implies relatively stronger cohesive
forces than adhesive forces. T of water and ordinary building materials is approximately
zero, i.e. a water droplet will be flattened to a thin layer, or absorbed by a porous material.
If a small diameter tube is inserted through a liquid surface, the liquid might either rise or
sink inside the tube, depending on the relation between cohesion and adhesion. In a glass
tube the mercury surface will sink, while a water surface will rise. The surface tension of
mercury is 0.486 N m-1 and 0.073 N m-1 for water. The curved surface that appears is called
meniscus, and is concave with water. The pressure is below atmospheric under the concave
meniscus. The pressure difference is
2 ı cosș 2ı
p air p water s ș|0 | . (12)
rmenisc rtube
where r is mean radius [m]. The potential capillary height (zcap) of water with ordinary
building materials is calculated from a pore radius, Equation 13. The capillary height of
materials with various pore sizes also depends on whether the moisture content is
increasing or decreasing.
2V
z cap . (13)
rpore U wat g
The capillary height of sand is 0.04-3.5 m and more than 10 m in clay (Nevander and
Elmarsson, 1994). The potential capillary height of wood is hundreds of meters (Skaar,
1988). Water is transported from low to high suction, i.e. from larger to smaller pores. The
moisture transfer in porous materials might also be calculated by
- 33 -
liq
m D w w (14)
where the moisture diffusivity is defined differently depending on moisture content. For
hygroscopic moisture content Dw is derived from Equation 10 for the temperature of
interest,
dv dv dRH dw 1 dw
diff
m į vap į vap į vap v sat,air (T) (15)
dx dRH dw dx dw dx
dRH
į vap v sat,air (T)

w D w w
ȟ
For capillary moisture contents, i.e. above the critical level, Dw is derived from Darcy’s
law, Equation 11. Here, ksuc is the permeability of the unsaturated porous material and
depends on the moisture content.
k suc ȡ wat dp k suc ȡ wat dp dw k suc ȡ wat

liq
m w D w w . (16)
ȝ dx ȝ dw dx dw
ȝ
dp
The relation between moisture content and suction is called the water retention curve. A
large negative pressure exists when only small pores are filled with water. As the moisture
content increases the negative pressure decreases. The slope of the absorption retention
curve is negative and becomes more negative with an increased w.
The moisture diffusivity Dw thus varies with moisture content, according to Figure 18, with
Dw of brick and light-weight concrete. The general shape of the curve shows lowest
moisture diffusivity at the critical moisture content wcr (Hagentoft, 2001).
- 34 -
1E-05
brick
1E-06 light-weight concrete
Dw [m2 s -1]
1E-07
1E-08
1E-09
1E-10
0 50 100 150 200 250
w [kg m-3]
Figure 18 The moisture diffusivity for light-weight concrete and brick (after
Nevander and Elmarsson (1994).
In capillary transfer, moisture is transported upwards from a liquid surface by suction.

When the capillary transfer stops, an active moisture content gradient occurs where w
decreases with height. Above a certain height the moisture content decreases drastically to
hygroscopic moisture contents. A very wet porous material being drained will attain a new
moisture gradient through the material, called the passive capillary moisture content, where
both total moisture content and capillary height are greater than before the soaking
(Nevander and Elmarsson, 1994). The transfer rate decreases with height, i.e. also with
time. The time dependent capillary moisture flow into a porous material in contact with
cap [kg m-2 s-1] is given by
liquid water, m
A
cap
m (17)
2 t
where A is the water sorption coefficient [kg m-2 s-½] and t is the time the material has
been in contact with the water [s]. Water sorption coefficient values for wood are 0.004
(perpendicular to the fibre direction) and 0.02 kg m2 s-½ (parallel to the fibre direction). The
coefficient A, however, decreases towards zero with time (Hagentoft, 2001). The capillary
front height (zcap) [m] as a function of time is
z cap B t (18)
where B is the water penetration coefficient [m s-½]. B also decreases with time. The
capillary transfer rate through a porous layer of thickness z is
AB
cap
m . (19)
2z
The use of two independent state variables is necessary for correct non-isothermal moisture
flow calculations. Simplifications are used because such data are usually not available
(Hagentoft, 2001).
- 35 -
4.1.3 Heat transfer
Conduction
Heat conduction means a heat transfer from more to less energetic particles due to
interactions between the particles. Pure heat conductivity is described by Fourier’s law,
which for the one dimensional (1D) case is
dT
Pcond O (20)
dx
where Pcond is conductive heat flux [W m-2], O is thermal heat conductivity [W m-1 K-1] and
dT/dx is the temperature gradient [K m-1]. The thermal conductivity of the wood chips
depends on wood quality, chips structure, compaction, temperature and moisture content.
The thermal conductivity of dry sawdust and cutter shavings is 0.08-0.14 W m-1 K-1,
depending on compaction (Mörtstedt and Hellsten, 1999). The main part of the heat
conduction occurs in the wood chips, since Oair|0.025 W m-1 K-1. The thermal conductivity
of cutter shavings Ocs increases almost linearly with moisture content, to about 0.3 W m-1 K-
1
for w=300 kg m-3 (Skogsberg, 2004). The suggested equation was
Ȝ cs 0.1 0.0007 w . (21)
For the transient case in 1D with varying thermal conductivity the temperature distribution
is given by
wT w § wT ·
ȡ cp ¨Ȝ ¸ (22)
wt wz © wz ¹
where cp is specific heat at constant pressure of the material [ J kg-3 K-1]
Thermal convection
Thermal convection is a heat transfer due to the random molecular motion (diffusion) and
bulk motion of a fluid. The heat transfer rate is proportional to the temperature difference
between the surface and the fluid. Both surface and internal convective heat transfer exist.
The latter is when a fluid passes through a permeable layer. The surface convective heat
transfer (Pconv) [W m-2] from air to the surface is
Pconv D conv Tair Tsurf (23)
Dconv is the convective heat transfer coefficient [W m-2 K-1]. It varies from 1 to 106 and
might be calculated from empirical expressions or formulas with dimensionless numbers,
where
Nu O fl
D conv (24)
L
Ofl is the thermal conductivity of the fluid [W m-1 K-1] and L is a characteristic length,
dependent on type of problem, [m]. The dimensionless Nusselt number (Nu) is a function
of other dimensionless numbers. When forced convection is present the Reynolds number
- 36 -
(Re) is important. Nu is proportional to Re to the power of a positive number that depends

on the problem. Re is proportional to uair and conversely proportional to the kinematic
viscosity of air (Xair) [m2 s-1]. A common empirical expression to determine Dconv for air
velocities (uair) parallel to an external surface is (Hagentoft, 2001),
°6 4 u air u air d 5 m s 1
D conv ® (25)
°̄7.41 u air 0.78 u air t 5 m s 1
The velocity of a fluid at the surface of an impermeable material is zero. The velocity field
above the surface is calculated by mass and energy conservation and Newton’s second law
of motion, which states that the sum of all forces acting on the control volume must equal
the net rate at which momentum leaves the volume (Incropera and DeWitt, 1996).
The fluid velocity (uint) will exceed zero at the interface between the fluid and the
permeable material, where the relative velocity (uint/uair) depends on permeability, surface
roughness, layer thicknesses and fluid velocity outside the boundary layer (Prinos et al,
2003). For a material with high permeability the ratio will be closer to one, i.e. uint will be
closer to uair. At a certain distance into the permeable layer, the velocity profile changes and
Darcy flow occurs.
Radiation
Thermal radiation is energy transported by electromagnetic waves, emitted by all materials
with a temperature above 0 K. The radiation exchange between a wood chips surface and
air includes both short wave and long wave (lw) radiation. Solar radiation reflectivity, or
albedo (rsolar) at a small spot of cutter shavings, was 0.15-0.32 (Skogsberg and Lundberg,
2005). The absorbed solar radiation (Psolar,abs) [W m-2] is
Psolar,abs Psolar 1 rsw (26)
The long wave radiation exchange between a wood chips surface and the sky, assuming
grey surfaces and the sky area much bigger than the wood chips surface, is
4 4
P12 V H wc Tsky , K Tsurf , K . (27)
Here, V is the Stefan-Boltzmann constant 5.6710-8 W m-2 K-4, Tsky,K and Tsurf,K are surface
temperatures [K], and H is wood chip surface emissivity [-]. The emissivity of wood is 0.82-
0.92 (Kreith and Kreider, 1978). The sky temperature can be estimated in different ways.
One method is to set the sky temperature equal to the air temperature when the sky is
cloudy and use Equation 28 when the sky is clear (Hagentoft, 2001).
Tsky, K 273.15 1.2 Tair 14 (28)
4.2 Surface melt

Much of the natural melting in open pond snow storage occurs by surface melt. Surface
melt here means melting caused by heat transfer from air and solar radiation to snow. Rain
- 37 -
is treated separately. With permeable insulations, heat and moisture transfers are closely
connected. The energy balance [W m-2] of the thermal insulation is
Psw Plw Pconv Psto Pcond (29)
where Psw is the net short wave (solar) radiation, Plw is the net long wave radiation, Pconv is
the convective net heat transfer to the insulation, including moisture transfer, i.e.
condensation, evaporation (negative) and sensible energy to/from melt and condense water.
Psto is the energy increase of thermal insulation and Pcond is the thermal conduction from the
insulation to the snow.
Heat transfer through a thermal insulation without moisture transfer is rather
straightforward if the material properties and climate parameters are known. Heat and
moisture transfer through wood chips on snow, i.e. the traditional thermal insulation on
snow, was largely unknown before this study. An emphasis was put on this subject.
Heat and moisture transfer through cutter shavings on snow was studied in one laboratory
experiment (PAPER III) and in numerical modelling (PAPER IV). In the experiment cutter
shaving layers of 0.025-0.075 m were investigated. In the modelling the layers were 0.1-0.3
m. Air was blown through a thermally well insulated box with snow and cutter shavings.
Solar radiation was simulated with spotlight bulbs, Figure 19.
Figure 19 Pictures of the experiment box. The box did lean to achieve melt water
outflow. The right picture is a view from the air outlet.
The melt and evaporation rates were studied for cutter shavings layer thickness (0.025-
0.075 m), air velocity (corresponding to 2.3-6.8 m s-1 at standard height 10 m), light
intensity (0-190 W m-2), air temperature (11.8-20.1oC), absolute air humidity (0.9-4.9 g m-
3
), cutter shaving moisture content (10-12% and 336-362%) and hindered evaporation.
The surface melt rate (SMR) was 0.75-2.38 kg m-2 h-1 with light and 0.50-1.28 kg m-2 h-1
without light. No interactions were significant. The multiple linear regression model was
significant at the 99% confidence level, Equation 31.
- 38 -
SMR 0.97 0.097 Z 0.164 u air 0.00175 i light 0.102 Tair 0.192 AH (31)
SMR is surface melt rate [kg m-2 h-1], Z is cutter shaving layer thickness [cm], uair is air
velocity [m s-1], ilight is light intensity [W m-2], Tair is air temperature [oC] and AH is
absolute air humidity [g m-3]. The ilight coefficient of Equation 31 cannot be directly used for
solar radiation.
ER was 17.0-118.2 g m-2 h-1 with light and 21.0-60.8 g m-2 h-1, without light. Without light
the influence of air temperature, air humidity and cutter shaving layer thickness were
significant. Without light no factors were significant. The ratio between evaporation and
surface melt rates was small as mass, 2.2-9.5%, but considerable as energy, 16.8-71.2%.
With increased ratio ER/SMR the surface melt rate decreased.
With initially wet cutter shavings, the SMR was about the same as with initially dry cutter
shavings, while ER increased from 17.0-78.8 g m-2 h-1 to 212.8-303.0 g m-2 h-1. This
indicated a larger thermal conductivity with wet cutter shavings was counteracted by
increased evaporation, both night and day.
In the numerical modelling SMR dependence on air temperature (10-30oC), relative air
humidity (60-80%), solar radiation (0-700 W m-2), air velocity (2-10 m s-1), initial cutter
shaving moisture content and cutter shaving layer thickness (0.1-0.3 m) was investigated.
Capillary transfer was calculated with a redefined (capillary) moisture diffusivity
(REPORT I) and a height dependent linear dampening function. The maximum active
capillary height was set to 0.13 m, while the upward capillary transfer above this height
was only allowed during certain conditions.
A thermal resistance between snow and cutter shavings of 0.01 m2 K W-1 was used in the
simulations. It was assumed that the wood chips absorbed 10% of the melt water, unless the
moisture content in the lowest part was too large. The influences of chosen maximum
active capillary height, absorbed melt share ratio, dry density, solar reflectivity, long wave
emissivity, convection function, and thermal resistance between cutter shavings and snow
were investigated.
SMR was 0.051-1.507 kg m-2 h-1, and increased with increased solar radiation, air
temperature, air velocity, and decreased layer thickness, Equation 32. SMR with initially
wet cutter shavings was the same as or less than with initially dry cutter shavings, due to
greater evaporation or evaporation instead of condensation.
SMR -0.09 0.00014 Psolar 0.0575 Tair 0.0012 Tair u air 0.18 Tair Z (32)
Psolar is incident solar radiation, Tair is air temperature, uair is air velocity and Z is cutter
shaving layer thickness.
ER was from -0.074 kg m-2 h-1 (evaporation) to 0.023 kg m-2 h-1 (condensation). The ratio
ER/SMR was from -75.5% to 31.7% as energy, i.e. evaporation was a significant part of the
energy balance. The relative importance of evaporation increased with a thinner layer of
wood chips. The sensitivity analysis of the simulation model showed that the most
- 39 -
important parameters for the SMR were assumed turbulence penetration depth, thermal
resistance between shavings and snow, absorbed melt share and dry density.
The effective thermal conductivity was calculated with SMR, cutter shaving layer
thickness, air temperature and 0oC snow temperature. It was determined in the laboratory
experiment to 0.32 and 0.21 W m-1 K-1, with and without light, and 0.42 W m-1 K-1 in the
numerical modelling.
4.3 Rain melt

Rain melt depends on the temperature and amount of rain that comes into the storage. With
a superstructure (roof) rain does not affect the storage. With large sheets as insulation rain
water might pour off as long as the snow pile is above ground level. With a permeable
insulation some rain remains in the insulation and gradually evaporates. The evaporated
rain proportion depends on rain characteristics, i.e. intensity and duration. Very intense rain
might influence the insulation layer and thereby increase the surface melt.
Rain melt simulations might be based on measurements or by distributing the total rain
precipitation over the whole summer. Rain temperature roughly equals air temperature. 500
mm rain with mean temperature of 15oC means a heat contribution of 8.7 kWh m-2, i.e.
melting of 0.09 tons of ice.
4.4 Ground melt

Ground melt is the heat transfer to the pond caused by heat conduction from the
surrounding ground and leakage of cold pond water.
4.4.1 Ground melt by heat conduction
Heat transfer through the snow pond bottom was simulated in 2D for different ground
conditions. The simulations were made on the same cross section of snow pond, measuring
642 m (WH) with side slope 1:4 (14.0o). The total heat transfer was made for a 130 m
long pond, similar to the Sundsvall plant. The pond volume was 13,693 m3 and the
groundwater depth was 0.8 m below pond bottom. The storage construction consisted of
0.10 m asphalt, 0.50 m of gravel, 0.10 m of insulation and 0.10 m of sand. The simulation
cross section extended 15 m outside the snow pond and 10.2 m below the sand. The flow of
groundwater was simulated in the 10.2 m layer with 35% porosity, assuming the flow was
along the cross section, i.e. from one long side to the other. The groundwater gradient was
1%. Material properties used are shown in Table 8
- 40 -
Table 8 Materials and properties used in the ground melt simulations.

Density cp l Dyn. visc. Kin. visc.
-3
[kg m ] [J kg-1 K-1] [W m-1 K-1] [kg m-1 s-1] [m2 s-1]
Asphalt 2100 900 0.8 - -
Gravel 1800 775 0.8 - -
Insulation 30 1210 0.03 - -
Sand 1800 775 0.8 - -
Ground 2000 800 1.5 - -
Water (5 oC) 1000 4206 0.568 1.519×10-3 1.519×10-6
The heat transfer simulations were made with monthly mean ground surface temperatures.
To obtain a realistic initial ground temperature profile, a stationary solution was calculated
for October 31st, after which the temperature change with time was calculated from
November to August. The total heat transfer was calculated with the assumption that the
heat transfer at every section along the bottom was the same, except at the short side slopes
where an average of the “windward” and “lee” slope transfers was used.
The simulations were made with Fluent (2003). All materials except the groundwater zone
were considered as isotropic solids. The boundary conditions were surface temperatures
and groundwater inlet velocity and temperature. Total heat transfer and the resulting
amount of melted snow are shown in Table 9.
Calculations were made with reference plant settings varying one parameter at the time,
Table 9. One “worst case” simulation with high groundwater velocity, high groundwater
temperature, and no ground insulation was also made.
Table 9 Parameter settings and heat transfer through the pond bottom and
corresponding snow melt, assuming all heat melted snow.
Ground type k vgw Tgw Insulation Heat transfer Snow melt
2 -1 o
- [m ] [m s ] [ C] - [kWh] [ton]
Reference plant Sandy till 1.55×10-14 2.86×10-9 4.5 Yes 36,366 391.9
vgw+ Coarse sand 1.55×10-10 2.86×10-5 4.5 Yes 47,689 513.9
-16
vgw- Clay till 1.55×10 2.86×10-11 4.5 Yes 36,317 391.4
-14 -9
Tgw+ Sandy till 1.55×10 2.86×10 10.0 Yes 36,976 398.5
No insulation Sandy till 1.55×10-14 2.86×10-9 4.5 No 43,311 466.7
Worst Case Coarse sand 1.55×10-10 2.86*10-5 10.0 No 150,056 1617.0
The simulated snow melt was lower compared to the calculated melt in the pilot study (924
tons), except for the “Worst Case” scenario with all parameters at the worst setting. The
biggest effect on heat transfer occurred by increasing permeability from 1.5510-14 m2 to
1.5510-10 m2 and removing the 0.1 m ground insulation layer. With 20,000 tons of snow
- 41 -
the ground melt was about 2.0-2.6%, and 8.3% in the worst case scenario. The ground melt
in the pilot study was 4.7%. For normal conditions ground melt is thus considered small.
4.4.2 Ground melt by water leakage
If the snow pond is permeable, a water exchange between ground and melt water will occur
where the flow direction depends on the water levels (pressure) inside and outside the pond.
The out leakage of cold pond water is negative from an energy point of view, and might
also be negative for environmental reasons; see section 7.2. The leakage of groundwater is
positive if the groundwater temperature is lower than the re-circulated water, unless the in
leakage is too large.
Table 10 shows the diurnal groundwater flow and connected heat transfer for three
permeabilities and two groundwater table gradients. Permeabilities were taken from
Domenico and Schwartz (1990). A temperature difference of 5oC between the groundwater
and melt water was assumed in the calculations.
The heat transfer with clay and sandy till was small. With a 100 m2 cross section and coarse
sand, the losses equal a heat transfer of 4 to 40 MWh during 120 days, corresponding to
melting of 43 to 430 tons of snow. It should, however, be noted that the permeability within
different soil groups varies widely.
Table 10 Groundwater flow and connected heat transfer for three permeability’s and
two groundwater table gradients, with 5oC temperature difference between
groundwater and melt water.
Vflow Eflow
Soil k [m2] dh/dx [-] dp/dx [Pa m-1] [m3 m-2 d-1] [kWh m-2 d-1]
Clay till 1E-16 0,01 100 0,0 0,0
Sandy till 1E-14 0,01 100 0,0 0,0
Coarse sand 1E-10 0,01 100 0,6 3,3
Clay till 1E-16 0,001 10 0,0 0,0
Sandy till 1E-14 0,001 10 0,0 0,0
Coarse sand 1E-10 0,001 10 0,1 0,3
4.5 Design model for a snow cooling plant

Two snow cooling models have been developed, one for mainly economic calculations
(section 8.4) and one for more detailed design, as described here.
The storage ponds in both models took the shapes of an upside down frustum of a cone,
Figure 20. The models allowed a limited amount of snow above the edge, assuming that the
“top snow” formed the segment of a sphere. It was also assumed that each ton of snow
corresponded to 100 kWh of cold, i.e. melting and heating of 0oC ice to 6oC water.
- 42 -
Top width of bank
Storage top diameter
Height above bank Top snow
Ground level
Bank height
Pit depth
Bottom diameter
Cone height
Figure 20 Outline of snow storage pond in the calculation models.
The design variables were initial snow volume, upper pit diameter, side slope, pit depth,
thermal insulation material and thickness, thermal conductivity of the ground, and cooling
demand. The climatic variables used were annual mean air temperature, and hourly values
of solar radiation, air temperature, air velocity, precipitation, and distance to undisturbed
ground temperature. This distance was not easily estimated, but simulations of the
Sundsvall storage showed that the ground melt was rather small and thus not a crucial part
of the calculations, section 4.4. The model will be publicly available and sinus functions of
air velocity, precipitation and cooling demand are given as defaults.
Hourly values of forced melt, rain melt, ground melt, and surface melt are calculated. The
top snow was assumed to melt first. Rain melt was calculated with positive monthly
temperatures and upper pond diameter, since rain falling into the pit area is added to the
melt water. In the ground melt calculations, the heat flow goes through the pond area in
contact with the snow. The surface melt rate was calculated by Equation 32, and the forced
melt is defined by cold extraction.
- 43 -
- 44 -
5 MASS LOSS OF FREEZING WATER

Heat is emitted as water freezes, with some heat being emitted as evaporated water. This
phenomenon, freezing losses, was investigated by Nilsson and Sjödin (1986) and Milanov
and Nilsson (1988). They found that if a mass of water is frozen, the relative mass loss
(evaporated mass/initial water mass) is 2.5-3%. It was believed that the losses could add up
to certain amounts in a wood chip covered snow storage during the spring because of clear,
cold nights and large surfaces.
Freezing losses were studied in two laboratory experiments, i.e. water losses from initially
wet wood chips and losses from water filled cans. In the wet wood chips, the influence of
moisture content and air temperature was studied. In the can study, the influence of surface
area, initial water weight, initial water temperature, and a layer of initially dry wood chips
on top of the water/ice were studied. The sides and bottoms of the cans were thermally
insulated.
The magnitude of the freezing losses depended mainly on the frozen water mass. When all
water was frozen the losses were about 2.5-3% of the initial water weight, corresponding to
20-30% of the total energy transfer. This ratio agreed with reported studies. With a small
portion of frozen water the relative freezing losses increased, e.g. if 100 g of water is frozen
the freezing losses might be from 2.5 g to about 10 g, depending on the total amount of
water.
The freezing rate was important for the freezing loss rate, but unimportant for the relative
freezing losses when a certain proportion of the water was frozen. Water in a can froze
faster when the surface area was larger and when the water surface was closer to the can
edge.
The moisture content with initially wet wood chips was 11.6-236.1%. The relative freezing
losses decreased from almost 20% when a small portion of the wood chips were frozen,
down to 2.5-3% when the frozen portion increased. However, it was concluded that such
mass losses will not contribute to large melt water losses from the snow storage during
Swedish circumstances, since the melt rate and thus the capillary moisture transfer to the
wood chip layer is small during the spring. These losses might be important during other
circumstances.
- 45 -
- 46 -
6 CONSTRUCTION OF OPEN POND SNOW STORAGE SYSTEMS

Different seasonal snow and ice storage systems exist. The best system for a certain
application depends on the cooling load, cooling need, climate, soil/rock type, land
availability, groundwater level, groundwater flow, aesthetic concerns, and costs of snow,
land and thermal insulation.
Pond construction, thermal insulation and cold carrier system, snow and ice production, and
collection strategies are discussed here. Other issues, e.g. snow unloading and distribution,
insulation handling, melt water filtering, handling of sediments, and cleaning of heat
exchangers and the pond, are not treated.
6.1 Pond construction

The pond represents a large part of the construction cost, and its design is important to
obtain good functionality. Few studies have focused on this and no detailed study was done
within this project. Here, general conclusions from several realised and studied projects are
presented.
Since the main part of natural melting is the surface melt, it is reasonable to build more
compact storages. The pumps and pump house are preferably placed outside the pond to
facilitate construction and maintenance, as well as enable gravitational flow through oil and
gravel filters, if necessary.
The out leakage of cold water should be small and the storage must hold a certain amount
of water to buffer load variations. An acceptable leakage is also a function of snow cost. If
the storage goes below groundwater level and the groundwater flow is small, having a pond
with relatively high permeability is possible. Heat transfer by groundwater flow through a
permeable pond is generally small, section 4.4.2. Polluted snow in permeable ponds needs
further studies.
If the groundwater table is below the storage bottom or if the groundwater flow is high, a
“watertight” pond is needed, accomplished by lining the pond with asphalt or different
plastic (e.g. polyethylene or PVC) or rubber sheets.
From a geotechnical point of view the best ground for snow pond construction is coarse-
grained material or solid rock. If the ground consists of finer material there might be
problems with frost heave. Studying how frost heave might affect the plant is then
necessary. Frost heave under the pond bottom is avoided by filling the pond with water or
snow. Frost heave susceptible soil at the storage sides might require ground insulation or
other measures to avoid frost heave. For environmental reasons it is desirable to avoid
ground insulation (Knutsson, 2005).
- 47 -
6.2 Thermal insulation and cold carrier system

The cold carrier system and the thermal insulation affect each other. Different thermal
insulation materials are treated in section 3.5. The snow storage has no limit in cooling
power with proper design, except from capacity of pipes, pumps, and heat exchangers.
Seawater, groundwater, melt water, air or some other fluid might be used as cold carrier.
From an energy point of view it is beneficial to use the cold carrier with the lowest
temperature. Environmental, economical or other considerations might however restrain
continuous inflow and lead to re-circulation solutions.
A closed system, with pipes underneath the snow, was investigated in a Canadian study. It
was rejected because of economical reasons and potentially more severe consequences
from failure (Morofsky, 1981). A closed pipe system was also discussed in Sundsvall, but
this idea was rejected for the same reasons. No further closed systems were reported.
In open cold carrier systems, the cold carrier is in direct contact with the snow, thereby
improving the heat transfer capacity. The inlets can be placed at the sides, bottom and top
of the storage. In general, it is important to avoid water short cuts and uneven melting that
might negatively influence the insulation.
It is also important to have low cold carrier temperatures from the storage. The detain time
must thus be long enough, and the outlet water should come directly from contact with
snow. Therefore, it is favourable to distribute the re-circulation water over the storage area.
With a superstructure as thermal insulation, the circulation system is only a matter of
properly mixing snow and water, and inlets can be placed anywhere under the construction.
With sheet insulation it is necessary to place the water inlets under the insulation.
It is important that the snow melts evenly with loose fill insulation. Experiments in
Sundsvall showed that water re-circulation must be spread, otherwise local melt will cause
wood chips to fall off. If the inlets are placed under the snow pile the insulation should stay
intact unless the surface collapses by internal cavities. However, such re-circulation
systems have not been studied thoroughly.
Spraying water on the snow during warm days will increase the water temperature, and
decrease the air temperature. The total effect on the surface melt rate was not analysed.
Another aspect of spraying is that smaller particles and solutes might drift with the wind.
6.3 Snow and ice production

Unless enough natural snow is available, the necessary amount of snow or ice can be
produced next to the plant in many parts of the world. Depending on the type of snow or ice
producing system being used, the production capacity depends on equipment, water
temperature, water pressure, air pressure, air temperature, relative humidity, air velocity
and wind direction, see section 3.1.3 and 3.4. Noise might decrease production if the
equipment cannot be run at any time. Automatic systems are also available. The subject of
snow/ice production for seasonal storage will benefit from further studies.
- 48 -
7 ENVIRONMENTAL ASPECTS
Most products affect the environment during construction, operation and deconstruction,
i.e. when it is regarded as being consumed. Deconstruction is re-usage, recycling,
combustion, composting and depositing. Material extraction and transports shall also be
included when the environmental impact of a product is studied. Different products affect
differently at various stages. By comparing the total environmental impact for different
products the least harmful can be chosen, or the worst elements replaced (Rydh et al, 2002).
7.1 Environmental impact comparison

Wichmann (2003) compared the environmental impact of a snow cooling plant based on
data from the Sundsvall snow cooling plant and a chiller system. The chiller system
included two cooling machines of the same type used as backup at the Sundsvall hospital
plant. Operation data were mean values from 2000-2002. The comparison was made based
on the assumption that both systems had covered the total cooling demand. Their estimated
lifetimes were 40 years for the snow cooling system and 20 years for the chillers system.
The electricity production (nuclear, hydro, biofuel, etc.) was assumed as have the average
mix for Sweden, Denmark and Finland.
The environmental impact was investigated for the whole lifetime of the systems, i.e.
construction, operation and deconstruction. The results were quantified and compared in
four categories of potential environmental impact, viz climate change, acidification,
nitrification and formation of photochemical ozone. The units were [ton CO2 equivalents a-
1
] for climate change and [kg SO2/NO3/O3 equivalents a-1] for acidification, nitrification
and formation of photochemical ozone.
It was found that the chillers system had the largest environmental impact concerning
climate change, acidification and nitrification, while the snow cooling system had the
largest environmental impact in the category formation of photochemical ozone, Figure 21.
The main part of the environmental impact for both systems originates from the operation
phase, but the construction phase was also significant. The deconstruction phase accounted
for less than 1% in all categories for both systems.
The operation impact shares of the snow cooling system for each category were 81.0%,
82.7%, 91.4% and 82.3%. The dominating impact sources of the snow cooling system were
fuel and electricity. In the construction phase the thermal ground insulation of polystyrene
dominated the impact in all categories of the snow cooling system.
The chillers’ impact during operation was completely dominated by electricity. Impact of
the cooling media was significant in the climate change category, 8.3%. The cooling media
was R134a (CH2FCF3), and estimated annual leakage was 2% (SWEDAC, 2003).
- 49 -
250
[kg SO2/NO3/O3-equivalents a ]
-1
snow cooling
[ton CO2-equivalents a ]
-1
200 chiller
150
100
50
0
climate change acidification nitrification ground ozone
Figure 21 Total environmental impact from snow cooling and chillers systems
concerning climate change, acidification, nitrification and formation of
photochemical ozone. Units are [ton CO2 equivalents a-1] for climate
change and [kg SO2/NO3/O3 equivalents a-1] for the others.
One comparison with “environmentally optimised systems” was also made, where
hydropower electricity, renewable fuels and a storage construction without ground thermal
insulation and asphalt were assumed. The absolute values decreased drastically, Figure 22
(observe scales!) and the snow cooling system had less impact in all studied categories.
20
[kg SO2/NO3/O3-equivalents a ]
-1
snow cooling
[ton CO2-equivalents a ]
-1
15
chiller
10
0
climate change acidification nitrification ground ozone
Figure 22 Environmental impact from “environmentally optimised” snow cooling
and chillers systems concerning climate change, acidification, nitrification
and formation of photochemical ozone. Units are [ton CO2 equivalents a-1]
for climate change and [kg SO2/NO3/O3 equivalents a-1] for the others.
- 50 -
7.2 Snow Pollution

Urban snow is polluted. The sources are anthropogenic and natural, local and far distant.
Pollution mainly comes from air, traffic, animal faeces, litter and slipperiness control
(salts). Pollution is largely influenced by traffic intensity and how long the snow has been
on ground (Viklander, 1997).
During melting, most dissolved substances and smaller particles leave with melt water,
while larger and adsorbed particles tend to remain on the ground (Westerström, 1995;
Viklander, 1997). The pollution concentration in sediments can be very high (Droste and
Johnston, 1993). If polluted snow is to be used in a permeable pond, investigating the
environmental impact is critical.
In watertight snow storage, the pollution can be handled since the melt water outlet to the
environment is controlled. Larger particles stay in the pond or in the filters. If debris is used
as insulation it will mix with mineral particles from urban snow, and the total amount of
particles will therefore increase. It is thus necessary to occasionally remove a proportion.
Recycled sand and gravel might need treatment. It was found in a Swedish study that this is
possible and economically feasible (Andersson, 2005). If melt water is intended for reuse in
snowmaking, it is important to investigate if the treatment is necessary.
- 51 -
- 52 -
8 ECONOMY
The economic incentive for snow storage depends on the alternative cost and if a new
cooling system is needed or not. Brief summaries of different studies are presented below.
Johansson (1999) investigated district cooling prices for seven Swedish systems and found
the consumer cost to be roughly equal to the alternative cooling cost. This implies that
district cooling suppliers had set a price that was close to the conventional cost for this
relatively new technology. The mean cooling prices were divided into three parts, i.e. a one
time connection cost of about 1,750 SEK kW-1 (190 € kW-1), an annual cooling load cost of
about 175 SEK kW-1 (19 € kW-1) and an energy cost of about 175 SEK MWh-1 (19 € MWh-
1
). This indicated a payback time 0.5-6 years for a 100,000 m3 snow storage (rock cavern),
depending on the connected peak powers.
A current follow-up of district cooling prices revealed a connection cost of 1,760 SEK kW-
1
, a load cost of 260 SEK kW-1 a-1 and an energy cost of 250 SEK MWh-1. With 20 years
depreciation time and a 6.0% interest rate for the connection cost, the cooling price of 3,850
MWh with different peak loads became 0.46-1.32 SEK kWh-1, Table 11.
Table 11 Calculated annual cost for 3500 MWh of district cooling for different peak
powers.
Peak power Connection Power cost Energy cost Total cost Total cost
[kW] [SEK] [SEK] [SEK] [SEK/kWh] [€/MWh]
2,000 306,890 520,000 962,500 0.46 51.1
4,000 613,779 1,040,000 962,500 0.68 74.7
6,000 920,669 1,560,000 962,500 0.89 98.3
10,000 1,534,448 2,600,000 962,500 1.32 145.5
Kobiyama (1997) made an economic comparison between the All-Air air-conditioning

system and a conventional system, for Japanese conditions. If the land cost was added to
the snow clearing account, the snow system was far cheaper than the conventional system
for air-conditioning areas larger than 200 m2, because of a reduced electricity need.
Otherwise, it was not possible to give a general answer due to large land cost variations.
Taylor (1985) estimated that seasonal ice ponds were profitable when air conditioning of
buildings larger than 10-20,000 m2, because of lower operation costs and size advantages.
The estimated total ice cost was 10-25 US$ ton-1, for an ice storage of 10,000-200,000 tons.
The cost of thermal enclosure insulations accounted for more than two-thirds of the total
cost. Land costs up to about 20 US$ m-2 was acceptable for this technique.
Kirkpatrick et al (1985) made a general comparison between a high cost and a low cost
scenario, since the alternative cooling cost varied by a factor of 5. A storage with initially
26,000 m3 (14,300 tons) of snow/ice was insulated with aluminized insulation sheet or an
inflatable tent. Three to six snow guns were used. The estimated lifetime was 20 years. The
- 53 -
estimated payback times were 0.8-14 years with plastic sheets and 5-86 years for the tent
alternative.
Abdelnour et al (1994) studied three methods to store winter coldness, i.e. icebox, ice pond
and snow pit, and three air condition applications; in a separate flat building (8 kW) and
two office buildings (45 and 1,600 kW) in Montreal, Halifax and Toronto. Toronto and
Montreal have continental climates, while Halifax has a more temperate maritime climate.
The most extreme climate years over a 20-year period were studied. In this study the
estimated lifetime was 20 years, interest rate 6%, annual maintenance of 1% of the
investment cost and an electricity cost of 0.07 CD$ kWh-1. Snow transportation was not
included. They found that the snow pit with deposited snow was cheaper than the
alternative cost for all applications and locations. The ice box was competitive for smaller
applications in Halifax because of a short cooling season with high cooling loads. The ice
ponds were too expensive, possibly because the investigated system only enabled a 1.6 m
ice layer.
In a Canadian study, swedes (turnips) were stored with snow thermally insulated by
sawdust; see section 3.1.1. The construction cost was negligible and the annual operation
cost was estimated to 700 CD$ (465 €). Construction and operation costs of a conventional
system were estimated to 22,000 CD$ (14,700 €) and 400 CD$ (265 €) Vigneault (2000).
Vigneault (1985) found that a Fabrikaglace to pre-cool about 1,000 tons of vegetables
would reduce the cost of electricity by 72%, and that capital costs were comparable to those
of a conventional system. The system could be further improved, though it was concluded
that the technique was viable and an economic option.
Today (2005), the main obstacles for snow cooling in Canada are the low energy cost and
the preference of companies for weather independent automatic systems (Vigneault, 2005).
The estimated payback time of a Swedish snow cooling plant for 6,000 MWh of cold
(~120,000 m3 of snow) is approximately 3 years, based on experiences from Sundsvall;
however, a greater cooling power would reduce this time (Skogsberg, 2003).
8.1 Pond construction

Pond construction includes excavation and pond construction, perhaps a watertight lining,
and a cold carrier system. Some cost examples are given.
The excavation cost of a 505010 m (LWD) pit in Sweden is about 1.3 MSEK (0.14 M€),
according to one company. Lining this pit with almost watertight steel tongues, permitting
vertical sides, costs about 4.3-4.8 MSEK (0.47-0.52 M€) (Stoltz, 2005). For large-scale
plants a welded plastic sheet with a protective geo membrane on each side costs 70-100
SEK m-2. A sand and gravel layer is needed if vehicles are to be driven in the pond
(Johansson, 2005). The cost of sand and gravel is about 65-300 SEK m-3, depending on
quality.
- 54 -
Assuming a cooling power of 2,000 kW and a heat exchanger temperature difference of

5oC, the necessary water flow rate is almost 0.1 m3 s-1. With an area 5050 m, four water
sprayers (BEVA, 2005) are needed, at a total cost of 30,000 SEK (3300 €). The cost of a
suitable pump with speed regulation is about 75,000 SEK (8240 €) (ITT, 2005). Pipes and
control system were not included. It might be advisable to initially operate the system
manually the first year to learn how to operate it.
8.2 Thermal insulation

The cost of thermal insulation is a matter of investment, handling and lifetime. According
to experiences from Sundsvall, wood chips have to be complemented almost every year and
exchanged after two to four years, because of decay and contamination. CCV pays about
115 SEK m-3 (13 € m-3) for larger pieces of wood chips (Larsson, 2005).
Debris is a single investment cost or at least a limited occasional refill cost, since urban
snow contains large amounts of mineral particles. Rejecting loose fill insulations implies a
cost because of transport, and there might also be a cost to deposit the polluted material.
No really good insulation sheets have been found yet. They are too small, too expensive,
too fragile or too difficult to handle. Superstructures are to expensive and seem to hinder
snow production. Perhaps a superstructure is economically feasible if only using natural
snow. This subject obviously needs further research.
8.3 Operation
The operation cost depends on storage construction and is largely connected with snow
production, and handling of snow and insulation. It is, thus, beneficial to construct the
storage for reduced snow handling and reduced natural melt, to lower the necessary amount
of snow. A deeper storage with a smaller upper area will reduce natural melt. Snow
handling is reduced since less snow will have to be moved from where it is dumped. With
moveable water sprayers or snow guns, making the snow roughly where desired is possible.
This question needs further attention.
Snow
Urban snow might be a cost, an income or free, depending on local regulations. If natural
snow is collected for only cooling it is a cost, whereas if the plant is used as a snow deposit,
the owner might be able to charge the depositors or at least get the snow for free.
The cost of produced snow depends on climate, type of snow guns and the cost of
electricity and water. Johansson (1999) estimated the total snow production cost for
100,000 m3 of snow (40,000-45,000 tons), with a production time 100-500 hours and 8.5%
interest rate, to be 173,000-323,000 SEK (19.000-35,500 €) with a system of nine
LowEnergyTower snow guns, and 178,000-262,500 SEK (19,500-29,000 €) with six fan
type snow guns.
- 55 -
8.4 Cooling cost model

To estimate the cooling cost, a simplified melt model was made where a large number of
parameters were varied, e.g. monthly mean cooling demand, pond top diameter, bank
height, snow height above bank, side slope, type and thickness of thermal insulation, plastic
liner at sides and in bottom, monthly mean temperatures, annual precipitation, proportion
artificial snow, number of snow guns, interest rate, and a number of other economical
parameters. Values were mostly estimated from the Sundsvall plant. The storage was
assumed to have the shape of an upside down frustum of a cone, with a limited amount of
snow above the surrounding bank, see Figure 20.
The model estimates the snow volume needed to meet the total heat gain from natural and
forced melting. Pit design is chosen with guidance from the model. The main outcome is
the cooling cost including investment depreciation. The estimated cooling cost can be
easily compared for different designs. It is appropriate to compare the physical results of
the economic model with the more detailed design model.
Rain melt is calculated with top diameter and positive monthly temperatures, since all rain
falling into the pond mixes with snow. Ground melt is calculated with the upper snow
surface area. Forced melting is calculated from the given cooling demand. Finally, surface
melt is calculated based on the mean snow area. The surface melt rate depends on the
chosen thermal insulation.
Cooling cost comparisons
The cost of 3,850 MWh of cooling was compared for a number of alternatives, Table 12.
The excavation cost was 55 SEK m-3, bank shaping 20 SEK m-3, lining with polypropylene
sheets 85 SEK m-2, 0.4 m sand and gravel layer, sand and gravel cost 100 SEK m-3, and
”other” construction costs (pumps, pipes, electrical installations, control system, planning
etc.) was 6.2 MSEK. The snow production system assumed snow guns of the
LowEnergyTower type. The pump and heat exchanger cost was estimated to 300+100 SEK
kW-1. The assumed pipe cost was 500,000 SEK.
The calculated cooling cost was 0.29-0.47 SEK kWh-1 (31.9-51.6 € MWh-1). The lowest
cost was for a solution with 25% artificial snow instead of 50%, and for a solution without
a watertight lining on the sides and bottom. The most expensive cooling cost was for the
solution with a 10-year depreciation time instead of 30 years.
The corresponding district cooling prices, from Table 11, were 0.68 SEK kWh-1 with
cooling load 4,000 kW and 1.11 SEK kWh-1 with cooling load 8,000 kW. The district
cooling cost was calculated with a 20-year depreciation time and 6% interest rate.
- 56 -
Table 12 Cooling costs for different alternatives.
Depreciation time for snow storage [y]
Total cooling cost [SEK/kWh]

Electricity price [SEK/kWh]
Artificial snow share [%]
Storage top diameter [m]
Storage (pit) depth [m]

Cooling need [MWh]
Thermal insulation
Insulation layer [m]
Snow volume [m3]
Bank height [m]

Peak load [kW]
Interest [%]
Lining [Y/N]
Wood chips 3,850 3,696 0.2 68,623 50% 94 16.6 5.4 Y 0.70 30 6 0.31
Plastic insulation 3,850 3,696 0.03 69,537 50% 94 18.6 5.5 Y 0.70 30 6 0.33
Debris 3,850 3,696 0.4 78,309 50% 98 20.3 5,9 Y 0.70 30 6 0.33
Wood chips 3,850 8000 0.2 68,623 50% 95 14.8 5.5 Y 0.70 30 6 0.35
Wood chips 3,850 3,696 0.4 64,572 50% 92 16.1 5,4 Y 0.70 30 6 0.32
Wood chips 3,850 3,696 0.2 69,061 25% 95 14.8 5.5 Y 0.70 30 6 0.29
Wood chips 3,850 3,696 0.2 75,051 50% 110 7.5 5.4 Y 0.70 30 6 0.33
Wood chips 3,850 3,696 0.2 68,623 50% 94 14.8 5.4 N 0.70 30 6 0.29
Wood chips 3,850 3,696 0.2 68,623 50% 94 14.8 5.4 Y 1.00 30 6 0.33
Wood chips 3,850 3,696 0.2 68,623 50% 94 14.8 5.4 Y 0.70 10 6 0.47
Wood chips 3,850 3,696 0.2 68,623 50% 94 14.8 5.4 Y 0.70 30 3 0.25
- 57 -
- 58 -
9 CONCLUSIONS
There are a number of studied and suggested seasonal snow and ice storage systems using
natural and artificial snow and ice for cooling purposes. Sufficient precipitation or periods
of low temperatures are found in many parts of the world. Snow/ice is stored indoors, on
the ground and in the ground, and in watertight and low permeable ponds. Most cooling
energy is stored in the latent heat of fusion. To extract this energy the cold carrier must melt
the snow, either in direct contact or through pipes in or under the snow. The current project
focuses on open pond storages with open circulation system.
In open pond snow storages, thermal insulation of the snow is necessary. Different
alternatives with advantages and disadvantages exist. Self-supporting structures work well,
but are expensive. Insulation sheets tested in the different reported projects were either too
expensive, too fragile, too labour intensive, and/or connected with handling problems.
The loose fill insulation tested or investigated are wood chips, rice shell and debris (mineral
particles). Two general disadvantages with loose fill insulation are the extensive handling
required and their falling off if the snow sides are too steep. The falling off problem might
be handled by even melting of the snow. One general advantage with loose fill insulation is
its ability to self-adjust, i.e. fall into smaller pits in the snow surface.
This study focuses on wood chips, the traditional thermal insulation of ice, and especially
cutter shavings. In a wood chip layer on snow the heat and moisture transfer are intimately
connected. Dry cutter shavings have low thermal conductivity, though during the
summertime on snow, the shavings are mostly wet as they absorb water from the snow.
Heat and moisture transfer through cutter shavings was studied in a laboratory experiment
and numerical modelling. It was found that the surface melt rate of snow covered with
cutter shavings increased with an increased solar radiation, air velocity and air temperature,
and a decreased cutter shaving layer thickness. Air humidity was significant for the melt
rate in the experiments but not in the modelling, possibly because of thinner cutter shaving
layers in the experiments. The surface melt rates were 0.75-2.38 kg m-2 h-1 in the
experiments and 0.051-1.507 kg m-2 h-1 in the numerical modelling. The average effective
thermal conductivity was 0.32 and 0.21 W m-2 K-1 in the experiments, with and without
light, and 0.42 W m-2 K-1 in the modelling.
The evaporation rate contributes significantly to the energy balance of the cutter shaving
layer. Laboratory experiments revealed the ratio between evaporation and surface melt rates
(as energy) to be 16.8-71.2%. The numerical model yielded corresponding ratios from -
31.7% (condensation) to 75.5% (evaporation). Numerical modelling showed the ratio to
decrease with increased layer thickness, though this was not seen in the experiments. It was
also seen that the surface melt rate was roughly the same with initially wet and initially dry
wood chips. It was concluded that the evaporative cooling effect is important in the thermal
insulation qualities of wood chips.
- 59 -
The principle results of this study are applicable for “similar” materials like sawdust, sheep
wool, straw, etc., but not for materials without capillary water transport like mineral wool.
In a Japanese project without cold extraction, rice husk was used as thermal insulation. It
was concluded that a 0.3 m layer would sufficiently reduce the melting. It should also be
possible to use debris as insulation, since the ablation rate of glaciers decreases
significantly (59-85%) with a 0.4 m debris layer.
There is also heat transfer from rain and the ground. The rain melt is small, e.g. 500 mm of
10oC rain on 1 m2 melts about 30 kg of snow. Ground melt is caused by heat conduction
and, in permeable ponds, out leakage of cold water. The groundwater flow through a cross
section corresponded to at most 3.3 kWh m-2 d-1, or 40 MWh for a 100 m2 section over 120
days. In Sundsvall with snow corresponding to about 2,600 MWh, the ground melt is thus
0.2-1.5% of the initial snow.
Numerically modelling the Sundsvall plant revealed a heat transfer from the ground of
about 36.4 MWh. Without thermal ground insulation and with coarse sand and a annual
mean air temperature of 10oC, the heat transfer increased to 150.0 MWh, which
corresponds to 5.8% of 40,000 m3 of snow (snow density 650 kg m-3).
Heat is emitted as water freezes, with some heat being emitted as evaporated water. This
phenomenon, called freezing loss, was investigated in a laboratory experiment. It was
found that the freezing loss from water filled cans and wet wood chips primarily depended
on the frozen water mass. The loss was about 2.5-3% of the initial water mass when all
water was frozen. This corresponded to about 20-30% of the total energy transfer. With a
small portion of frozen water, the relative freezing loss increased to almost 10%. It was
concluded that the loss might contribute to a considerable melt water loss under certain
circumstances, though unlikely for typical Swedish conditions.
The Sundsvall Regional Hospital snow cooling plant began cooling operation in June 2000.
It was built for 60,000 m3 of snow, corresponding to about 4000 MWh. During the winters
of 2000-2004, 18,800-40,700 m3 of natural and artificial snow were stored in the slightly
sloping, shallow pond, 13064 m, with a watertight asphalt bottom. The artificial snow
portion was 37-70%. The hospital annually used 655-1,345 MWh of comfort cooling, with
the snow cooling plant delivering the main part, 77-93%. The snow was thermally insulated
by a layer of larger wood chips pieces. The plant has mostly worked satisfactory.
The snow cooling seasons started at the end of April or beginning of May, and ended
sometime between the middle of August and beginning of September. The conventional
COP was 5.2-23.8, and the total COP, including construction energy, was 4.3-17.2. The
total COP of the snow cooling plant was 2.0-6.6 times better than a chiller system, based on
data from the back-up system in Sundsvall.
Melt water re-circulation caused wood chips to fall off, due to the snow becoming exposed
to sun and wind. This increased the surface melt and led to increased outlet temperatures,
since short cuts formed along the edge of the pond. Minor tests and experiments showed
that re-circulation pipes under the snow and the spraying of return water could handle these
- 60 -
problems. Another problem was the decay of wood chips and mixing with sand and gravel
from the snow. New wood chips were bought every year, and were eventually replaced
after three years.
The water temperature after the heat exchangers decreased from 4.1oC in 2001 to 2.2oC in
2004, while the temperature difference in the heat exchangers also decreased, from 4.9oC to
3.7oC. This occurred due to lower temperatures at the secondary side and algae growth in
the heat exchangers. It was believed that substances from the wood chips and conditions in
the heat exchangers caused the growth. The algae problem was handled by lye cleaning,
one to three times per summer.
Concentrations of phosphor and some heavy metals were rather high in the snow, but the
concentration was considerably reduced in the outlet water. This reduction was related to
particle adsorption. In the outlet water, a considerable concentration of non-biodegradable
compounds was detected, probably related to the wood chips. The investment cost was 14.5
MSEK (1.59 M€), and the total cooling cost varied between 1,140 and 2,176 SEK MWh-1
(125.3 and 239.1 € MWh-1).
The geometry should be more compact in future open pond storage, i.e. deeper storage, to
reduce natural melt and facilitate handling. If the soil properties and groundwater require a
watertight pond, this is preferably done with rubber or plastic sheets, due to their flexibility.
The out leakage of polluted water from a permeable pond requires further studies. In a
watertight pond melt water can be treated.
The environmental impact of a snow cooling plant and a chillers system was compared,
both for existing systems and for “environmentally optimised systems”. Values were based
on experiences from Sundsvall The impact was investigated for the estimated lifetime of
the systems, and the results were compared in four categories of potential environmental
impact; climate change, acidification, nitrification and formation of photochemical ozone.
The chillers system had the largest environmental impact concerning climate change,
acidification and nitrification, while the snow cooling system had a larger formation of
photochemical ozone. The operation phase caused the main part of the impact for both
systems, though the construction phase was also significant. The dominating impact
sources of the snow cooling system were fuel and electricity. In the construction phase the
polystyrene ground insulation dominated the impact. In the environmentally optimised
systems the absolute values decreased drastically, and the snow cooling system had less
impact in all studied categories.
Different studies show optimistic predictions for seasonal snow and ice storage systems,
but they are not yet commercially viable, except for Japan and perhaps China. The success
of snow cooling depends on the awareness and development of the techniques,
environmental considerations, climate, soil and groundwater properties, cooling demands,
and the costs of water, primary energy, labour and land. According to Vigneault (2005) the
main obstacles seem to be the large investment costs, labour intensive systems, problems
with the thermal insulation and low electricity prices.
- 61 -
Recent estimations gave 0.5-6 years payback time of cavern snow storage. Snow and ice
are natural and non-poisonous, and if depositing urban snow in central snow storages is
possible, both money and the environment are saved. The resulting shorter transports are
beneficial in both aspects, and controlling pollution is possible by keeping the contaminated
snow in a watertight enclosure. Estimations reveal lower cooling costs for new open pond
storage (0.29-0.47 SEK kWh-1) than for comparable district cooling prices.
9.1 Further research

Further research is divided into topics concerning open pond snow cooling plants and the
function of wood chips on snow.
The main issue concerning open pond snow cooling plants is cost, which is intimately
connected to function. A deeper storage will decrease natural melt, but finding appropriate
solutions of thermal insulation and water re-circulation that works together is necessary.
One interesting thermal insulation alternative is debris, whose handling and function must
be studied further. Reducing the costs of snow and insulation handling, other equipments in
the system and electrical installations are also important. Snow production has been
developed for other purposes than snow storage, and is therefore a subject where
considerable improvements can be made.
Concerning snow cooling in general, it is important to identify the best alternative for
different applications. Perhaps the most promising system is underground caverns, since
these can be located in densely populated areas and thereby reduce snow transports and
deliver cold without long connection pipes. This solution requires both environmental and
technical investigations.
The heat and moisture transfer in wood chips is not fully elucidated. Questions concerning
the proportion of melt water that is transported upwards to the wood chip layer and heat and
moisture convection are highly relevant. To further understand the processes in the wood
chips layer, it is necessary to combine experiments with mathematical modelling.
ACKNOWLEDGEMENT
I acknowledge Prof Bo Nordell for all help with this text and The Swedish Energy Agency
(Statens Energimyndighet) and Luleå University of Technology for financing the study.
- 62 -
REFERENCES
Abdelnour R, Labrecque B and Underdown A (1994). Technoeconomic analysis of three
seasonal cooling technologies; Ice box, frozen pond and waste snow pit.
Proceedings of Calorstock ´94 Conference on Thermal Energy Storage, Helsinki,
Finland.
Ahlgren L (1972). Moisture fixation in porous building materials. Report 36, Division of
Building Technology, The Lund Institute of Technology. (In Swedish. Pictures and
results in English).
Andersson J O (2005). Ragn-sells AB, Umestans företagspark 366, S-903 47 Umeå. Private
communication.
André E, Hogdin S and Derstroff T (2001). District cooling at Luleå University of
Technology. Student report, Division of Water Resources Engineering, Department
of Environmental Engineering, Luleå University of Technology, Sweden.
Antonsen G, Lindberg J and Siri N M (1998). Snökyllager för Polarbageriet, Älvsbyn
(Snow cooling for the Polar bakery, Älvsbyn). Student report, Division of Water
Resources Engineering, Department of Environmental Engineering, Luleå
University of Technology, Sweden.
AREA (2004). Homepage of Air conditioning and Refrigeration European Association,
http://www.area-eur.be/ March 28th 2004.
Areco (1999). Product file from Areco Snow system, Nifsåsvägen 11, 831 52 Östersund,
Sweden. http://www.snowmaking.com/.
BEVA (2005). BEVA Bevattningsteknik, Box 625, Radiovägen, 27B, S-135 26 Tyresö.
Private communication.
Buies S (1985). Engineering of a life-size Fabrikaglace. Centre de Recherche du Québec
(CRIQ), Canada. Technical report No. FAB-85-051.
CADDET energy efficiency (1990). An ice pond system for industrial process cooling.
IEA/OECD US 90.211/2A.H02.
Canada (2000). Cooling in all climates with Thermal Energy Storage. State-of-the-Art for
Canada in Annex 14, a project within ECES (Energy Conservation through energy
Storage Implementing Agreement), IEA (International Energy Agency).
DIN 4108 (1981). Thermal insulation in buildings: DIN 4108, part 1-5. Deutches Institut
für Normung, Berlin, 1981.
Domenico P A and Schwartz F W (1990). Physical and Chemical Hydrogeology. Second
Edition. Johan Wiley and Sons, Inc., ISBN 0-471-59762-7.
- 63 -
Dorgan C E and Elleson J S (1993). Design Guide for Cool Thermal Storage. ASHRAE
American Society of Heating and Refrigerating and Air-Conditioning Engineers,
1791 Tullie Circle, Atlanta, Georgia 30329. ISBN 1-883413-07-9.
Droste R L and Johnston J C (1993). Urban snow dump quality and pollutant reduction in
snowmelt by sedimentation, Canadian Journal of Civil Engineering, 20:9-21.
Ericsson H E (2003). Snökylanläggning, Sundsvalls sjukhus. Uppföljning av kontroll-
program 2001-2002 (Snow cooling plant, The Sundsvall Regional Hospital. Follow-
up of the control program 2001-2002). Commision 1658119 101, Sweco VIAK.
Eriksson T S (1985). Surface Coatings for Radiative Cooling to Low Temperatures.
Unpublished. Department of Physics, Chalmers University of Technology,
Gothenburg, Sweden.
Falk A, Hansson S, Johansson J and Sandström C (2001). Snölager- En kall rapport. Snö
för kylning av kraftvärmeverk (Snow storage- A cold report. Snow for cooling of a
heat and power plant). Student report, Division of Water Resources Engineering,
Department of Environmental Engineering, Luleå University of Technology,
Sweden.
Fluent (2003). Fluent version 6.1.18, Fluent Inc., 10 Cavendish Court, Centerra Resource
Park, Lebanon, NH 03766, USA.
Gray D M and Male D H, editors (1981). Handbook of Snow: principles, processes,
management and use. Pergamon Press, Toronto. ISBN 0-08-025375-X.
Hagentoft C-E (2001). Introduction to building physics. Studentlitteratur, Lund, Sweden,
ISBN 91-44-01896-7.
Hagerman A (2000). Snökylanläggning kontra kylmaskiner (Snow Cooling Plant vs.
Chillers). Undergraduate thesis. Division of Resource Management, Mid Sweden
University.
Hedenblad G (1996). Materialdata för fukttransportberäkningar (Material data for moisture
calculations). T19:1996, Byggforskningsrådet, Stockholm, Sweden.
Hill P (2004). Projektförslag beträffande utveckling av snökyla i Sundsvall (Suggestions
for projects concerning snow cooling in Sundsvall). Division of applied thermo-
dynamics and refrigeration, Department of Energy Technology, KTH, Sweden.
Iijima K, Kobiyama M, Hanaoka Y, Kawamura M, Toda H (1999). Absorbability of
contaminants from air by snow cooling system. Paper for the 8th Indoor Air 99
Conference in Edinburgh, Scotland.
Incropera F P and DeWitt D P (1996). Fundamentals of Heat and Mass Transfer. John
Wiley and Sons Inc., USA. ISBN 0-471-30460-3.
ITT (2005). ITT Flygt AB, Gesällvägen 33, S-174 87 Sundbyberg. Private communication.
- 64 -
Jansson L (1998). Kylning Sundsvalls Sjukhus. Test snölager (Cooling of the Sundsvall
Hospital. Experiment with snowpile). Appendix to final report on project 803 939 D
“Prestudy of ice storage for comfort cooling at the Sundsvall Regional Hospital”.
County Council of Västernorrland, Sweden.
JL Toppteknik (1999). Product information booklet, JL Toppteknik, Box 126, 686 23
Sunne, Sweden. http://www.toppteknik.se/.
Johansson P (1999). Säsonglagring av kyla i bergrum (Seasonal cold storage in rock
caverns). Master thesis 1999:184 CIV, Luleå university of technology, Department
of Environmental Engineering, Division of Water Resources Engineering, Sweden.
Johansson P (2005). Gävle vägtrummor AB, 801 04 Gävle, Sweden. Private
communication.
Kamimura S and Toita T (2004). Concept of Electric Power Output Control System for
Atomic Power Generation Plant Utilizing Cool energy of Stored Snow. Presented at
the 5th International Conference on Snow Engineering, July 5-7th 2004, Davos,
Switzerland.
Kaneko Y, Kobiyama M, Nagaoke H, Sato T (2000). Air-conditioning System for
Apartment Residences by Using Water made from Snow. Paper for Iscord 2000.
Conference in Hobart, Australia.
Kayastha R B, Takeuchi Y, Nakawo M, and Ageta Y (2000). Practical prediction of ice
melting beneath various thickness of debris cover on Khumbu Glacier, Nepal, using
a positive degree-day factor. Proceedings of the workshop Debris-Covered Glaciers
held in Seattle, Washington, USA, September 2000. The International Association
of Hydrological Sciences, Publication 264:71–81.
Kirkpatrick D L, Masoero M, Rabl A, Roedder C E, Socolow R H, and Taylor T B (1985).
The Ice Pond Production and Seasonal Storage of Ice for Cooling. Solar Energy
Vol. 35, No. 5, pp 435-445.
Kirkpatrick, D.L., McGarity, A.E. and Norford, L.K. (1987). Design and operation of an ice
pond for cooling a large commercial office building. ASHRAE Transactions nr 93
(1987) s 1749-1759
Knutsson S (2005). Division of Soil Mechanics, Department of Civil and Environmental
Engineering, Luleå University of Technology. Private communication.
Kobiyama M (1997). Economic Estimation of All-Air Type Snow Air-Conditioning
System. Proceedings of Megastock, the 7th International Conference on Thermal
Energy Storage. Conference in Sapporo, Hokkaido, Japan (page 611-616).
Kobiyama M (2000). Department of Mechanical Engineering, Muroran Technical
University, Japan. Private communication.
- 65 -
Kobiyama M, Wang A, Takahashi T, Yoshinaga H, Kawamoto S and Iijima K (1997). Air-

conditioning system by stored snow. Proceedings of Snow Engineering Conference,
Balkem, Rotterdam, Netherlands. ISBN 90-5410-865-7.
Kosonen I (2005). Country Visions Oy, Käpytie 5, 40520 Jyväskylä, Finland. Private
communication.
Kreith F and Kreider J F (1978). Principles of Solar Energy. Hemisphere Publishing, New
York. ISBN 0-89116-678-5
Larsson P-E (2005). Landstingfastigheter Västernorrland (The County Council of
Västernorrland), Sweden. Private communication.
Malmqvist P-A (1983). Urban storm water pollutant resources. An analysis of inflows and
outflows of nitrogen, phosphorus, lead, zinc, and copper in urban areas. Dissertation
series, Department of Sanitary Engineering, Chalmers University of Technology,
Gothenburg, Sweden. ISBN 91-7032-106-X.
Mattson L E and Gardner J S (1991). Energy exchanges and ablation rates on the debris-
covered Rakhiot Glacier, Pakistan. Zeitschrift für Gletscherkunde und
Glazialgeologie. 1989 (published 1991), 25(1):17-32.
Mattson L E, Gardner J S, and Young G J (1993). Ablation on debris covered glaciers: an
example from the Rakhiot glacier, Punjab, Himalaya. Proceedings of the
Kathmandu Symposium, November 1992. The International Association of
Hydrological Sciences, Publication 218, 1993.
Meteonorm (2000). Meteonorm Version 4.00, Meteotest, Fabrikstrasse 14, CH-3012 Bern,
Switzerland. Homepage; http://www.meteotest.ch/en/firma.
Milanov T and Nilsson J (1988). Frysförluster av vatten (Freezing losses from water).
SMHI Hydrologi, No 20. (in Swedish)
Morofsky E (1981). Project Snowbowl. Public Works of Canada (PWC) Contract EN 280-
0-3650
Morofsky E (1982) Long-term latent energy storage- the Canadian perspective. US China
conference on energy, resources and environment.
Morofsky E (1984). Developing and introducing an innovative building cooling technique:
Strategy formulation based on market and technology considerations. Issues in
strategic management, Adm. 6395.
Morofsky E (1985). Cooling with ice – Canadian progress. Proceedings of the Third
Conference on Energy Storage for Building Heating and Cooling. September 22-26,
1985, Toronto, Canada, p. 337-340.
Mörtstedt S-E and Hellsten G (1999). Data och diagram. Energi- och kemitekniska tabeller
(Data and diagram. Tables of energy and chemistry technology). Liber AB. ISBN
91-47-00805-9. In Swedish.
- 66 -
Nagano K, Mochida T, Miyazaki T and Shimakura K (2000). Field experiments on

consolidation of stored snow in the ground and a computational simulation for the
underground snow storage. Proceedings of 8th International Conference on Thermal
Energy Storage. Stuttgart, Germany. pp 441-446.
Nakamura Y, Sugiyama H, and Kamimura S (2004). Experimental Study on Open Air
Storage of Snow with Simple Insulation Using Agricultural Wastes. Presented at
Switzerland.
Nevander L E and Elmarsson B (1994). Fukthandbok – praktik och teori. AB Svensk
Byggtjänst, Andra reviderade utgåvan. ISBN 91-7332-716-6. (in Swedish)
Nilsson J and Sjödin N (1986). Frysförluster av vatten (Freezing losses from water). SMHI
Hoh PM No 67, Norrköping. (in Swedish)
Nordell B. and Sundin E. (1998). Snöupplag för säsongslagring av kyla (Snow Deposit for
Seasonal Storage of Cold). Division of Water Resources Engineering, Luleå
University of Technology, Sweden. April 1998.
Näslund M (2000). Fjärrkyla i Sundsvall baserad på sjövatten och lagrad snö (District
cooling in Sundsvall based on sea water and stroed snow). Master thesis 2000:132
CIV, Luleå university of technology, Department of Environmental Engineering,
Division of Water Resources Engineering, Sweden.
Näslund M (2001). Kylvattenförsörjning till SSAB Hardtech. Förprojektering (Supply of
cooling water to SSAB Hardtech. Prestudy). Consultant report, Energidalen i
Sollefteå AB, 881 52 Sollefteå, Sweden.
Okajima K, Nakagawa H, Matsuda S and Yamasita T (1997). A cold storage for food using
only natural energy. Proceedings of Snow Engineering Conference, pp 569-572,
Balkem, Rotterdam, Netherlands. ISBN 90-5410-865-7.
Paksoy H Ö (editor) (2003:1). General State-of-The-Art Report, Subtask 1. Annex 14,
Cooling in all climates with thermal energy storage. July 2003. International Energy
Agency (IEA), Energy Conservation through Energy Storage (ECES).
Paksoy H Ö (editor) (2003:2). Feasible boundary conditions and system configurations for
cooling with TES, Subtask 2. Annex 14, Cooling in all climates with thermal
energy storage. August 2003. International Energy Agency (IEA), Energy
Conservation through Energy Storage (ECES).
Pelto M S (2000). Mass balance of adjacent debris-covered and clean glacier ice in the
North Cascades, Washington. Debris-Covered Glaciers (Proceedings of a workshop
held at Seattle, Washington, USA, September 2000). The International Association
of Hydrological Sciences, Publication 264:35–42.
Prinos P, Sofialidis D and Keramaris E (2003). Turbulent flow over and within a porous
bed. Journal of Hydraulic Engineering, ASCE, Vol. 129, No. 9, Sept. 1, 2003.
- 67 -
Rydh C J, Lindahl M and Tingström J (2002). Livscykelanalys – en metod för

miljöbedömning av produkter och tjänster (Life cycle assssment – A method for
estimating environmental impact of products and services). Studentlitteratur, Lund,
Sweden. ISBN 91-44-02447-9.
Saarman E (1992). Träkunskap (Wood science). Sveriges Skogsindustriförbund, Markaryd.
ISBN 91-7322-726-9.
SEPA Swedish Environmental Protection Agency (1990). Bedömningsgrunder för sjöar
och vattendrag (Quality criteria for lakes and water courses). General guidelines
90:4. ISBN 91-620-0042-X.
SEPA Swedish Environmental Protection Agency (1999). Bedömningsgrunder för
miljökvalitet- sjöar och vattendrag (Environmental quality criteria- lakes and water
courses). Report 4913, ISBN 91-620-4913-5.
Setterwall F and Alexanderson K (1996). Phase change materials and chemical reactions
for thermal energy storage. State of the Art 1996. Department of Chemical
Engineering and Technology, Transport phenomena, Royal Institute of Technology,
Sweden. ISSN 1104-3466.
Skaar C (1988). Wood-Water Relations. Springer-Verlag Series in Wood Sciences. New
York, USA. ISBN 0-387-19258-1.
Skogsberg K (2003). The Sundsvall Regional Hospital snow cooling plant – results from
three years of operation. Proceedings of FutureStock, the 9th International
Conference on Thermal Energy Storage, September 1-4, 2003, Warsaw, Poland.
Skogsberg K (2004). Thermal conductivity of wet wood chips- A laboratory study.
Technical report. Division of Renewable Energy, Luleå University of Technology.
Skogsberg K and Lundberg A (2005). Wood chips as thermal insulation of snow. Accepted
for publication in Cold Regions and Science Technology, Elsevier Science B. V.,
April 25, 2005.
SR (2005). Interview from the program “Vetenskap” (Science) in Sveriges Radio (The
Swedish National Radio), August 8th, 2005.
Stamm A J (1964). Wood and cellulose science. Ronald Press, New York, USA.
Stoltz A (2005). Skanska Sverige AB, Region Väg och anläggning i norr, 971 76 Luleå,
Sweden. Private communication.
Sundin E (1998). Snow deposit melt and atmospheric icing: analysis and modelling by
weather station data. Doctoral thesis 1998:37. Division of Water Resources
Engineering, Luleå University of Technology. ISSN 1402-1544.
Suzuki T, Kobayashi S, Tsushima K, Shao S, Teng Y and Liu G (1997) A case study on the
utilization of snow and ice as natural cold energy source for low-temperature
storage materials. Snow Engineering, p 553-558. ISBN 90 5410 865 7
- 68 -
SWEDAC, The Swedish Board for Accreditation and Conformity Assessment (2003).
Homepage; www.swedac.se.
Takeuchi Y, Kayastha R B, and Nakawo M (2000). Characteristics of ablation and heat
balance in debris-free and debris-covered areas on Khumbu Glacier, Nepal
Himalayas, in the pre-monsoon season. Debris-Covered Glaciers (Proceedings of a
workshop held at Seattle, Washington, USA, September 2000). IAHS Publ.
264:53–61.
Taylor T B (1985). Ice ponds. American Institute of Physics (AIP) Conference
Proceedings, November 25, 1985. Volume 135, Issue 1:562-575.
Vigneault C (1985). Precooling vegetables using a block of ice produced by winter
coldness. Proceedings of the Third Conference on Energy Storage for Building
Heating and Cooling. September 22-26, 1985, Toronto, Canada, p. 345-349.
Vigneault C (2000). Winter coldness storage. 2nd Expert meeting of Annex 14,
Implementing Agreement of Energy Conservation trough Energy Storage (ECES),
IEA, Nova Scotia, Canada, April 6-7, 2000.
Vigneault C (2005). Agriculture and Agri-Food Canada, HRDC, 430 Gouin BLVD, Saint-
Jean-sur-Richelieu, Quebec J3B 3E6, Canada. Private communication.
Viklander M (1997). Snow Quality in Urban Areas. Doctoral Thesis 1997:21. Division of
Sanitary Engineering, Department of Environmental Engineering, Luleå University
of Technology, Sweden. ISSN 1402-1544.
Viklander M and Malmqvist P-A (1993). Melt water from snow deposits. Proceedings of
the Sixth Conference on Urban Storm Drainage. September 12-17, 1993, Niagara
Falls, Ontario. p 429-434.
Westerström G (1995). Chemistry of snow melt from an urban lysimeter. Water Quality
Research Journal Canada, 30 (2): 231-242.
Wichmann P (2003). Miljökonsekvensanalys av snökylanläggning och kylmaskin
(Environmental impact study of a snow cooling plant and a chiller). Master Thesis
2003:322 CIV, Division of RenewableEnergy, Department of Environmental
Engineering, Luleå University of Technology, Sweden. ISSN 1402-1617.
Ågren L (2005). JL Toppteknik, Box 126, 686 23 Sunne, Sweden. www.toppteknik.se/.
- 69 -
- 70 -
Paper I
Skogsberg K and Nordell B (2001). The Sundsvall hospital snow

storage. Cold Regions Science and Technology, 32, 63-70. Elsevier
Science B.V.
Cold Regions Science and Technology 32 Ž2001. 63–70
www.elsevier.comrlocatercoldregions
The Sundsvall hospital snow storage

Kjell Skogsberg ) , Bo Nordell 1
DiÕision of Water Resources Engineering, Lulea˚ UniÕersity of Technology, S-971 87 Lulea,
˚ Sweden
Received 17 July 2000; accepted 24 November 2000
Abstract
During the summer, the regional hospital in Sundsvall in central Sweden requires 1000 MW h of cooling with a
maximum cooling power 1500 kW. From the summer of 2000, seasonally stored snow will be utilised to meet the cooling
demand. A storage area of 140 = 60 m with a capacity for 60,000 m3 Ž40,000 tons. of snow was constructed in 1999.
Initially, about half of this volume will be stored. The storage consists of a shallow pit made of watertight asphalt. A layer of
wood chips covering the snow reduces the natural melting to 20–30% of the total volume. Meltwater from the snow storage
is pumped to the hospital. After cooling the hospital, the heated meltwater is re-circulated to the snow storage. When all the
snow has melted, the wood chips will be burnt in a local heating plant. Lulea˚ University of Technology is responsible for the
scientific evaluation of the project. This paper describes the construction and the simulated operation of the snow storage
system. q 2001 Elsevier Science B.V. All rights reserved.
Keywords: Snow; Cooling; Energy storage; Renewable energy
1. Introduction last decades, however, new techniques using snow

and ice for both comfort cooling and food storage
Ice storage for cooling was used already in an- have been developed.
cient Greece. Ice was harvested from lakes and rivers Fabrikaglace is an un-insulated box inside an
and stored in barns that were thermally insulated by insulated shelter. Thin layers of water Ža few mil-
sawdust ŽTaylor, 1985.. This technique was common limeters. are frozen by cold air blown at the water
in Europe and North America until the beginning of surface, at suitable locations more than 20 m of ice
the 20th century, when cooling machines were intro- can be formed. To extract the cold, meltwater is
duced ŽMacCracken and Silvetti, 1987.. During the pumped to a heat exchanger and then re-circulated to
the ice. Flexible walls have been used to manage the
problem with ice-creeping and thermal expansion of
the ice. The size of existing Fabrikaglaces varies
) between a few up to 250 MW h with cooling powers
Corresponding author. Fax: q46-920-91697.
E-mail addresses: kjell.skogsberg@sb.luth.se ŽK. Skogsberg.,
from 8 to 1600 kW ŽBuies, 1985.. The coefficient of
bo.nordell@sb.luth.se ŽB. Nordell.. performance ŽCOP. is about 95 ŽAbdelnour et al.,
1
Fax: q46-920-91697. 1994.. The high construction cost requires large-scale
0165-232Xr01r$ - see front matter q 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 5 - 2 3 2 X Ž 0 0 . 0 0 0 2 1 - 5
64 K. Skogsberg, B. Nordellr Cold Regions Science and Technology 32 (2001) 63–70
plants or high cooling power ŽMorofsky, 1984.. The transportation cost ŽJohansson, 1999.. To reduce the
technique has not been widely used so far because snow volume, it is helpful to add water before
of, among other reasons, higher initial cost and lack compacting the snow ŽVigneault, 2000.. One major
of awareness and knowledge by architects, engineers problem with urban snow deposits is that the melt-
and others ŽCanada, 2000.. water is highly polluted ŽViklander, 1994.; by using
Ice ponds are based on the concept to use ponds, snow storage, this contamination can be controlled
or pits, to store snow, ice or slurry. The ponds have a and treated.
rubber or plastic fabric bottom and some top insula- Short Term ŽDiurnal. Ice Storage with ice gener-
tion. The icersnow is produced with nozzles or ated by refrigeration systems is another technique
snow blowers. Cold is extracted by re-circulation of that has become increasingly popular in North Amer-
meltwater. Ice ponds can combine cold storage and ica, Europe and Japan. This is economically feasible
water cleaning since impurities are separated from because of the varying electricity cost during the
the ice as it freezes. Cold storage capacities vary day. In Japan, for instance, the peak price around
from 50 to 650 MW h with cooling powers from 8 to noon is five times higher than night price ŽJapan,
1600 kW ŽTaylor, 1985.. The COP varies between 2000..
50 and 2000 ŽAbdelnour et al., 1994.. Snow and ice have excellent qualities for the
Himuros and Yukimuros are traditional Japanese storage of winter cold for cooling because of its 08C
buildings made of stone or wood with one part for melting temperature and its high heat capacity and
ice or snow and another part as a storage for vegeta- heat of fusion. The heat amount required Ž E . to raise
bles, fruits, etc. ŽOkajima et al., 1997.. Natural con- the temperature of ice from T1 to 08C, melt it, and
vection and occasionally shutters control the air cir- raise the water temperature to T2 is given by
culation, resulting in a low and stable air temperature
with high relative humidity. Both artificial and natu- E s Ž 0 y T1 . Cice q L q T2 C water kJ kgy1 . Ž 1.
ral icersnow are used. There is no size constraint for
The heat capacity of ice at y58C, Cice s 2.1 kJ kgy1
HimurosrYukimuros ŽKobiyama, 1987..
Ky1 ŽDorsey, 1940.; the heat of fusion of water
The All Air System is another Japanese technique
L s 333.6 kJ kgy1 ; and the heat capacity of water at
where air is cooled by direct contact with snow. In
58C, C water s 4.2 kJ kgy1 Ky1 ŽHobbs, 1974..
this system, both temperature and humidity can be
controlled ŽKobiyama et al., 1997. and pollutants
removed from the air ŽIijima et al., 1999.. This
system has mainly been used for food storage, but 2. The Sundsvall snow storage
there are also examples of comfort cooling ŽKaneko
et al., 2000.. There are more than 50 snow and ice The Sudsvall Regional Hospital, with a floor area
storage cooling plants in Japan today. Some of them of 190,000 m2 , is located in central Sweden. Na-
are test plants while the rest are commercial plants tional regulations and environmental targets made
ŽKobiyama, 1997.. ¨
the County Council of Vasternorrland to look for
Seasonal Snow Storage for cooling has been stud- new solutions in comfort cooling. The hospital’s
ied in Ottawa, Canada ŽMorofsky, 1982.. The idea cooling demand is 1000 MW h ay1 with a maximum
was to use snow that has to be removed from streets cooling power of 1500 kW. The chosen alternative
and squares and transport it to a snow deposit; a was to use the nearby snow deposit as a cold source
deposit of 90,000 m3 of snow covered with a reflec- for the cooling system.
tive and insulating cover was investigated. It was
found that the best technique for cold extraction was
re-circulation of the meltwater. The mean cooling 2.1. Principle
load of 7 MW resulted in an estimated payback of 10
years. It would also be possible to store snow in In seasonal snow storage, a major part of the cold
underground rock caverns where such plants would is stored in the phase change from water to ice.
be constructed in urban areas to reduce the snow Therefore, the heat transfer occurs by circulating the
K. Skogsberg, B. Nordellr Cold Regions Science and Technology 32 (2001) 63–70 65
Fig. 1. Snow storage principle. The meltwater is re-circulated through the snow.
cooling media through the storage, melting the snow, sulting snowmelt calculations were roughly verified
Fig. 1. The snow is covered with some type of by a performed small-scale test at the Sudsvall hospi-
insulation to prevent natural melting. In the Sundsvall tal during the summer of 1998.
case, a layer of wood chips is used to reduce the heat
transfer from air to snow and act as a cooler. This 2.2. Construction
cooling effect is a result of the wood chips’ absorp-
tion and evaporation of water. The construction work of the Sundsvall Snow
The function of the snow storage was simulated in Storage was completed in November 1999 ŽFig. 2..
a feasibility study ŽNordell and Sundin, 1998.. Re- Snow was stored during the winter and at the begin-
Fig. 2. Picture of the Sundsvall hospital snow storage, November 1999.

ning of April 2000 the storage was thermally insu- ŽFig. 3.. The cooling power is controlled by the
lated by a layer of wood chips. The system was put meltwater flow rate. This type of snow storage has
into operation in June 2000. no limit in extraction power, but its limitations are in
Snow is stored onto a watertight, slightly bowl- the capacities of the pumps, heat exchangers, etc. As
shaped asphalt surface with the area 140 = 60 m. A long as there is snow in the storage, the temperature
possible maximum snow depth of 9 m means a of the meltwater will be close to 08C. To avoid an
storage volume of 60,000 m3. A covering layer of increasing water level in the pit, part of the meltwa-
wood chips Ž0.2 m. thermally insulates the snow. ter is occasionally diverted from the system.
During the first winter of operation, 30,000 m3 of Both natural and artificial snow will be used. The
snow was stored corresponding to approximately Sundsvall storage will be filled from December to
20,000 tons of snow Žsnow density ; 650 kg my3 . February by urban snow transported from the clear-
and approximately 2000 MW h of cold. ing of streets and squares of the surroundings. Dur-
The asphalt surface is slightly sloped to flow the ing snow deficient winters, artificial snow will be
meltwater towards the water outlet at the pump produced by snow blowers.
house. Here, the water is cleaned by a coarse meshed The snow required to meet the cooling demand of
filter and an oil and gravel filter. It is pumped by two the hospital is approximately 10% of all snow re-
frequency controlled pumps Ž0.035 and 0.050 m3 moved from the city. The overall cost of the project
sy1 . to the heat exchangers Ž2 = 1000 kW.. Between is 1.6 M EURO; however, this sum also includes
the pumps and the heat exchangers are automatically costs that are not directly part of the snow cooling
rinsed fine filters. system. In a recent study ŽNaslund,
¨ 2000., based on
The meltwater leaves the storage at a temperature experience from the Sundsvall snow storage, it was
close to 08C. In the heat exchangers, where the warm concluded that the construction cost of a similar
side is cooled from 128C to 78C, it is heated to 108C. storage of 120,000 m3 snow Ž; 6,000 MW h. would
At the present maximum cooling power demand of be about 0.8 M EURO. The estimated technical
1500 kW, a water flow rate of 0.045 m3 sy1 is lifetime of the snow storage plant is 40 years. The
required. The meltwater is then re-circulated to the estimated pay back time is approximately 3 years for
snow storage to be cooled down and form more a snow storage with the same basic conditions as
meltwater. This re-circulated water is distributed over Sundsvall hospital system, which are typical for
the storage area by 36 individually adjustable valves space cooling in Sweden. A greater cooling power
Fig. 3. Outline of the Sundsvall snow storage system.

Fig. 4. The temperature series used in snow melt simulations.
would reduce the pay back time. For a large-scale height Ž h s 4.0 m., diameter Ž d s 105.6., and side
system Ž5000 MW h, 5 MW. the pay back time is slope angle Ž b s 26.68.. The ratio hrd was kept
about 2 years. constant throughout the melting. The air temperature
series used in the simulation is shown in Fig. 4. The
snow density was assumed to be constant at 650 kg
3. Snowmelt simulations my3 ŽViklander, 1994..
Two different models were used to estimate the
Snowmelt simulations for a 30,000 m3 snow stor- snowmelt from the air. In the un-insulated case a
age, without insulation and with 0.10 and 0.20 m of degree–days ŽDD. model was used. In the insulated
sawdust, were performed to estimate the necessary case, heat conduction was assumed to control the
sawdust thickness. Melting was divided into forced heat transfer through the sawdust.
melting caused by cold extraction and natural melt- The DD model is the most common method for
ing caused by heat transfer from the environment. snowmelt modelling and gives satisfactory results
The forced melting was based on the estimated ŽSundin, 1998.. It only requires mean diurnal tem-
cooling demand of the hospital. perature, snow density, and initial values of pile
geometry Žlength, width, and height..
3.1. Natural melting The DD model can be expressed as
Natural melting occurs by heat exchange between V̇ s kTair ,diurnal A Ž 2.

the snow pile and its environment. Consequently,
natural melting depends on climate, geometry of the where V˙ is the meltwater flow rate wm d , k is
3 y1 x
storage, and thermal insulation. the degree day constant wm Ky1 dy1 x, Tair,diurnal is the
The natural melting was divided into three parts: diurnal mean air temperature above freezing point
heat transfer from the air, from ground, and from w8Cx and A is the snow pile area facing air, A s f Ž V .
rain. Heat transfer from the air and the ground is a wm2 x.
function of the snow pile area, while heat transfer The meltwater flow rate can also be written as
from rain only depends on the size of the storage rs
area. In the simulations, the storage geometry was V̇ s DV0 Ž 3.
simplified by the shape of a cut cone with an initial rw
where rs is snow density wkg my3 x, r w is water wm2 x, z is the thickness of sawdust layer wmx, DTtop
density wkg my3 x and DV0 is diurnal snow volume is the temperature differences Tair, diurnal y Tsnow surface
decrease wm3 dy1 x.Eqs. Ž2. and Ž3. gives wKx, D z is the thickness of sawdust wmx, F is the
rw latent heat of snow wJ kgy1 x and 86400 is the number
DV0 s kT A. Ž 4. of seconds per 24 h ws dy1 x.
rs air ,diurnal
The heat transfer from the ground was simplified
Before the meteorological spring Ždefined as when as a one-dimensional heat flow from the ground
Tair,diurnal is roughly stable above "08C, based on water under the snow pile, by heat conduction from
meteorological records ŽSMHI, 2000.., k s 0.003 m the 68C ground water to the 08C bottom of the snow
Ky1 dy1 was used and from April 4th, k s 0.011 m pit. The ground water level at this location is about 1
Ky1 dy1 . These k-values were empirically decided m below the pit surface. Since the ground water flow
ŽSundin, 1998.. is high, it was assumed that its temperature was
In the insulated cases, the heat conduction equa- constant. This resulted in a heat flow of 2.4 W my2 ,
tion was used to determine the heat transfer through which is at the upper limit of the ground heat flow.
the area facing the upper surroundings Žfrom the air.. The total snowmelt depends on the bottom area of
Thermal conductivity of dry sawdust Ž12% water the snow, which decreases with time.
content. varies from 0.08 to 0.14 W my1 Ky1 , Approximately one-third Ž240 mm. of the annual
depending on the density. Thermal conductivity of precipitation in Sundsvall falls between April and
water is 0.60 W my1 Ky1 . In these calculations, a August ŽSwedish National Atlas, 1995.. Assuming
thermal conductivity of 0.35 W my1 Ky1 , the mean that the rainfall is evenly distributed during the
value of 0.10 and 0.60 was assumed for wet sawdust. summer and that the rain temperature is 108C, rain
The natural snowmelt because of the heat flow from results in a snowmelt of 2.1 m3 dy1 or 390 m3
the air is then given by totally.
DTtop
lins A top
Dz
DV1 s 86,400 Ž 5. 4. Results
Frs
where DV1 is the natural snow melt volume wm3 The calculated natural melting of the snow stor-
dy1 x, lins is thermal conductivity of sawdust wW age is shown in Fig. 5. The necessity of thermal
my1 Ky1 x, A top is the top area of remaining snow insulation is demonstrated by the fact that an un-in-
Fig. 5. Natural melting of 30,000 m3 of snow with 0.1 and 0.2 m of sawdust and without thermal insulation ŽDD-model..
Fig. 6. Total melting for initial snow volume 30,000 m3 with 0.1 and 0.2 m of sawdust as thermal insulation.
sulated pile would be gone by June 17th, i.e. long remains with 0.1 m sawdust and 6300 of snow
before the end of the cooling season. With thermal remains with 0.2 m of sawdust. The major part of the
insulation, the remaining snow volumes were 12,169 natural snowmelt, about 83% in the thermally insu-
m3 for 0.1 m of sawdust and 19,040 m3 for 0.2 m of lated case Ž0.2 m of sawdust., is caused by heat
sawdust. For 0.2 m of sawdust the maximum natural transfer from the air. Heat transfer from groundwater
meltwater rate is 77.4 m3 dy1 and occurs at the end contributes with 13% and rain with 4%. Since the
of June. Heat transfer from the air contributes the natural melting is highly dependent on the area
major part of the natural melt; 16,128 m3 with 0.1 m facing the air, a more compact shape would reduce
of sawdust and 9149 m3 with 0.2 m of sawdust. Heat the melting significantly.
transfer from the ground melts 1313 m3 of snow The storage technique is an example of utilising
with 0.1 m of sawdust and 1422 m3 with 0.2 m of renewable energy. Since the temperature of the melt-
sawdust. Heat transfer from rain melts 390 m3 of water is constantly 08C, snow storage has no power
snow. limit of cold extraction. However, detain time of the
When including the cold extraction of 1000 MW re-circulated water in the pit and the chosen capacity
h Ž15,000 m3 of snow. the total snowmelt becomes of pipes, pumps, and heat exchangers means a limit
28,181 m3 for the pile with 0.1 m of sawdust and of the cooling power.
23,676 m3 for the pile with 0.2 m ŽFig. 6.. There is a great potential of snow storage for
industrial, agricultural, and comfort cooling applica-
tions in large parts of the world. It is a natural
5. Discussion and conclusions technology where local resources are utilised to save
prime energy resources. In Sweden, there is a grow-
A thermally insulated 30,000 m3 snow storage ing interest for large-scale snow cooling. The snow
covers the cooling demand Ž1000 MW h, 1500 kW. collected from cities is polluted, but with this snow
at the Sundsvall hospital over the summer ŽMay– storage technique, the meltwater can be analysed and
August.. treated.
The natural snowmelt was calculated for the un- Based on the summarised pre-study ŽNordell and
insulated case and for the cases of 0.1 and 0.2 m of Sundin 1998., the Sundsvall Snow Storage was con-
sawdust as thermal insulation. By natural melting, structed during 1999. Snow was stored during the
the un-insulated snow would be gone by mid-June. winter 2000 and the storage was thermally insulated
With 0.2 and 0.1 m of sawdust approximately 60% by a 0.2 m layer of the wood chips at the beginning
and 40% of the snow remained at the end of August. of April. The system was put into operation in June.
When cold extraction is included, 1800 m3 of snow The maximum cooling power of the system is 2000
kW, which means a meltwater flow rate of 0.060 m3 Iijima, K., Kobiyama, M., Hanaoka, Y., Kawamura, M., Toda, H.,
sy1 at an assumed temperature difference of 88C. 1999. Absorbability of contaminants from air by snow cooling
system. Indoor Air 99 Conference, Proceedings Edinburgh,
The operation of the Sundsvall plant will be Scotland, pp. 37–42.
monitored, evaluated and reported. The total con- Japan, 2000. Cooling in all climates with Thermal Energy Storage.
struction cost of the plant was about 1.6 M EURO. State-of-the-art for Japan. Annex 14, a project within Interna-
In a more recent study, it was concluded that the tional Energy Agency, Energy Conservation through Energy
construction cost could be considerably reduced. The Storage Implementing Agreement.
¨
Johansson, P., 1999. Sasongslagring av kyla i bergrum ŽSeasonal
estimated lifetime for the snow storage plant is about storage of snow in a rock cavern.. Master thesis ISSN 1402-
40 years. The estimated pay back time is about 3 1617, LTU-EX-99r184-SE.
years for a new project with the same conditions as Kaneko, Y., Kobiyama, M., Nagaoke, H., Sato, T., 2000. Air-con-
the Sundsvall hospital system and decreases with ditioning system for apartment residences by using cold water
increased cooling power and size. made from snow. Iscord 2000 Conference, Hobart, Australia.
Kobiyama, M., 1987. Himuro type storage shed using the ice
The excellent thermal insulation qualities of wood stored in winter season. International Symposium on Cold
chips for snow storage is not only because of the Regions Heat Transfer Proceedings, pp. 167–170.
relatively low thermal conductivity, but also that Kobiyama, M., 1997. Economic estimation of all-air type snow
absorbed water evaporates, thereby cooling the snow air-conditioning system. Megastock ’97, Conference on Ther-
storage. This particular quality of wood chips Žand mal Energy Storage, Sapporo, Japan.
Kobiyama, M., Wang, A., Takahashi, T., Yoshinaga, H.,
sawdust. will be studied further at Lulea˚ University Kawamoto, S., Iijima, K., 1997. Air-Conditioning System by
of Technology. Snow pollution and possible treat- Stored Snow. Snow Engineering Balkema, Rotterdam, ISBN
ment methods of the meltwater will also be studied. 90 5410 8657.
MacCracken, C.D., Silvetti, B.M., 1987. Charging and discharg-
ing long-term ice storage. ASHRAE Trans. 93, 1766–1772.
Acknowledgements Morofsky, E.L., 1982. Long-term latent energy storage—the
Canadian perspective. US China Conference on Energy, Re-
sources and Environment.
This study is part of Kjell Skogsberg’s PhD stud- Morofsky, E.L., 1984. Developing and introducing an innovative
ies and was funded by the Swedish Energy Office building cooling technology: Strategy formulation based on
ŽEnergimyndigheten.. The project is part of the IEA market and technology considerations. Issues in strategic man-
Energy Storage Program, Annex 14 ACooling in all agement, 6395.
climates with thermal energy storageB, see http:rr ¨
Nordell, B., Sundin, E., 1998. Snoupplag ¨ sasongslagring
for ¨ av
kyla ŽSnow Deposit for Seasonal Storage of Cold., Water Res
cerve.cu.edu.trrannex14r. Eng, Lulea˚ University of Technology, April 1998 Žin Swedish..
¨
Naslund, ¨
M., 2000. Fjarrkyla i Sunsvall baserad pa˚ sjovatten
¨ och
lagrad sno¨ ŽDistrict Cooling at Sundsvall Based on Seawater
References and Seasonally Stored Snow., Water Res Eng, Lulea˚ Univer-
sity of Technology, CIV 2000:132. May 2000 Žin Swedish..
Abdelnour, R., Labrecque, B., Underdown, A., 1994. Technoeco- Okajima, K., Nakagawa, H., Matsuda, S., Yamasita, T., 1997. A
nomic analysis of three seasonal cooling technologies; ice cold storage for food using only natural energy. Snow Eng.
block, frozen pond and waste snow pit. Calorstock ’94, Con- 569–572, ISBN 90-5410-865-7.
ference on Thermal Energy Storage, Helsinki, Finland. SMHI homepage: www.smhi.se 2000-04-20.
Buies, S., 1985. Engineering of a life-size fabrikaglace. Centre de Swedish National Atlas, 1995. Climate, Lakes and Watersheds. Žin
´
Recherche du Quebec ŽCRIQ., Technical report no. FAB-85- Swedish. ISBN 91-87760-31-2.
051. Sundin, E., 1998. Snow Deposit Melt and Atmospheric Icing.
Canada, 2000. Cooling in all climates with thermal energy stor- Lulea˚ University of Technology, Division of Water Resources
age. State-of-the-art for Canada. Annex 14, a project within Egineering, ISSN 1402-1544.
International Energy Agency, Energy Conservation through Taylor, T.B., 1985. Ice ponds. AIP Conference Proceedings, pp.
Energy Storage Implementing Agreement. 562–575.
Dorsey, N.E., 1940. Properties of Ordinary Water-Substance in all Vigneault, C., 2000. Winter coldness storage. Paper for IEA
its Phases: Water Vapour, Water, and all the Ices. American Annex 14 meeting in Canada April 6–7, 2000.
Chemical Society, New York, Reinhold, n 99-0184307-3. Viklander, M., 1994. Melting of Urban Snow Deposits—A Water
Hobbs, P.V., 1974. Ice Physics. Oxford Univ. Press, ISBN 0 19 Quality Study. Licentiate Thesis 1994:19L. Division of Sani-
851936 2. tary Engineering. Lulea˚ University of Technology.
Paper II
Skogsberg K (2001). The Sundsvall Regional Hospital snow cooling

plant-Results from the first year of operation. Cold Regions Science
and Technology, 34, 135-142. Elsevier Science B.V.
Cold Regions Science and Technology 34 (2002) 135 – 142
www.elsevier.com/locate/coldregions
The Sundsvall Regional Hospital snow cooling plant—results

from the first year of operation
Kjell Skogsberg *
Division of Water Resources Engineering, Luleå University of Technology, S-971 87 Luleå, Sweden
Received 2 September 2001; received in revised form 14 November 2001; accepted 14 November 2001
Abstract
During the summer of 2000, the Sundsvall Regional Hospital (central Sweden) required 655 MW h of comfort cooling with
a maximum cooling power of 1366 kW. Of the cooling demand, approximately 93% was attained by the use of 19,000 m3 of
snow stored from the winter. The majority of the snow/ice was natural while the rest was artificially made with snowguns and a
hose. The snow was stored in a 140 60-m shallow pond of watertight asphalt. A 0.2-m layer of wood chips thermally
insulated the snow. The operation of the plant experienced only minor problems. Luleå University of Technology is responsible
for the scientific evaluation of the project. This paper compiles the results and experiences of the snow cooling plant during the
first year of operation in 2000. D 2002 Elsevier Science B.V. All rights reserved.
Keywords: Snow; Cooling; Energy storage; Renewable energy; Pilot plant
1. Introduction and artificial snow and ice for comfort cooling and
food storage have been developed. Seasonal snow/ice
One way towards a sustainable society is to reduce storage has a great technical potential in industrial,
both the use of non-renewable energy and the amount agricultural, and comfort cooling applications in large
of energy presently used for different tasks. One parts of the world. One way of extracting cold from a
energy efficient method for cooling is by using ice snow/ice storage is based on the re-circulation of the
and snow stored from the winter. Ice storage for cooling media between the cooling load and the snow/
cooling is an old technology that was used in ancient ice (Fig. 1).
Greece when ice was harvested from lakes and rivers Skogsberg and Nordell (2001) briefly reviewed the
and stored in barns. The ice was thermally insulated literature on seasonal snow and ice storage, which
by sawdust (Taylor, 1985). This technique was com- also included a description of the first Swedish large-
mon in Europe and North America until the beginning scale snow cooling plant at the Sundsvall Regional
of the 20th century, when chillers were introduced Hospital in central Sweden. Luleå University of
(MacCracken and Silvetti, 1987). During the past few Technology is responsible for the scientific evaluation
decades, however, new techniques using both natural of the project. In this study, different types of thermal
insulation on snow/ice are evaluated, with a focus on
*
Fax: +46-920-91697. evaporative materials such as wood chips and saw-
E-mail address: kjell.skogsberg@sb.luth.se (K. Skogsberg). dust.
0165-232X/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 5 - 2 3 2 X ( 0 1 ) 0 0 0 6 7 - 2
136 K. Skogsberg / Cold Regions Science and Technology 34 (2002) 135 – 142
Fig. 1. Snow storage principle; the meltwater is re-circulated between the load and the snow.
The objective of this article is to present experience a maximum cooling power of 1500 kW (1999). The
and measured results from the first year of operation monthly mean air temperatures during the cooling
at the Sundsvall snow cooling plant. season (May to August) vary from about 8 to 15 °C.
The annual mean air temperature of Sundsvall is 3.2
°C. Snow is stored in a shallow pond with a watertight
2. The Sundsvall Regional Hospital snow cooling asphalt surface (140 60 m) and a possible snow
plant storage volume of 60,000 m3 (Fig. 2).
Cold is extracted by pumping meltwater through
The Sundsvall Regional Hospital snow cooling heat exchangers connected to the existing cooling
plant was constructed during 1999/2000 and began system of the hospital. The heated meltwater is then
operating in June 2000. The hospital, with a floor area re-circulated to the snow where it is cooled and new
of 190,000 m2, needs about 1000 MW h of cooling with meltwater is formed. The meltwater is cleaned by a
Fig. 2.
K. Skogsberg / Cold Regions Science and Technology 34 (2002) 135 – 142 137
coarse meshed filter, oil and gravel filter, and automati- The pile was covered with a wood chip layer of
cally rinsed fine filters. It is then pumped by two approximately 0.2 m in April. It took about 10 days for
frequency-controlled pumps (0.035 and 0.050 m3 s 1) a tractor and an excavator to distribute the wood chips
to the counter flow heat exchangers (2 1000 kW). evenly over the snow surface. After the cooling season
Preliminary simulations by Nordell and Sundin the wood chips were removed from the pond. About
(1998) indicated that an un-insulated snow pile would 75% were saved for the following year and the rest was
be gone by mid-June, i.e. long before the end of the returned to the producer due to the lack of storage
cooling season. Thus, the pile had to be thermally space. The handling of the wood chips required 270
insulated. The traditional thermal insulation of snow/ man-hours. The returned wood chips were cleaned of
ice is different kinds of wood chips, sawdust, etc. sand and gravel before being burnt at a nearby co-
Although no scientifically oriented work was found generation plant.
on the function of wood chips on snow/ice, it was The snow cooling plant was used for cooling from
decided to use wood chips because of its easy access June 6th to August 28th. At the beginning of the
and low price. The outline of the snow cooling plant is cooling season, the meltwater temperature was bet-
shown in Fig. 3. ween 0 and + 1 °C. Very soon the snow on the sides of
During the autumn 1999, the pond was filled with the storage melted and shortcuts to the pump house
water to test its watertightness. When this water began were formed. Therefore, the meltwater temperature
to freeze, the bottom water was pumped on top of the was increased to about + 3 °C. The snow at the sides
ice to accelerate the ice production. The storage was melted first since the recirculation valves are located at
filled with snow and ice until the beginning of April. the sides of the storage. To avoid even higher temper-
The maximum snow volume (at the beginning of May) atures hoses were installed to return the melt water to
was 18,800 m3, which was less than the planned 30,000 the centre of the snow pile. In July, the intense rain
m3. Since the start of the snow cooling season was diluted and heated the melt water to between + 5 and
delayed by more than one month, the smaller snow + 10 °C as well as causing the outlet water to be
volume covered almost the entire season. Approxi- brownish due to washed-out substances from the
mately 70 –75% of the volume was natural snow/ice wood chips. The melted sides also meant that the
from streets and squares. Snowguns and a hose pro- snow was directly exposed to the air (Fig. 4), which
duced the rest of the snow. The snowguns ran for a total resulted in an increased snow melt. This problem,
of about 240 h. One snowgun produces 50– 100 m3 though, was not attended to during this summer of
(density about 400 kg m 3) of snow per hour. The operation.
Snowguns required considerable maintenance and sur- The running cost of the first year was about 15.3
veillance; at most they required de-icing every 4 h. EUR MW h 1 (a total of about 0.01 MEUR). The total
About 760 man-hours were spent on the snow handling investment cost of the snow cooling plant was approx-
at the snow plant. imately 1.3 MEUR.
Fig. 3. Outline of the snow cooling plant at the Sundsvall Regional Hospital, Sweden.
Fig. 4.
The owner, the County Council of Västernorrland the weather station at the snow pond was installed, air
(CCV), summarised the first year as one filled with new temperature and precipitation measurements by the
challenges, but still less problems than an ordinary year Swedish Meteorological and Hydrological Institute at
with chillers. The snow cooling plant delivered the the Midlanda Airport (20 km from Sundsvall) were
desired cooling, even though the melt water temper- used.
ature part of the season was higher than expected. The snow volume was measured five times with a
After the first year of operation, the storage was geodetic total station. The snow density was assumed
reconstructed to make it possible to attach re-circula- to be constant, 730 kg m 3, based on a few measure-
tion hoses directly to the valves and to have a higher ments in May.
water level in the pond. In this way, CCV hoped to The cooling energy from the snow cooling system
avoid shortcuts and snow exposed to the air, thereby was calculated as
ensuring a lower water temperature and a slower Es ¼ m_ m cp ðTm2 Tm1 Þt ð1Þ
natural snow melt. A pump was also installed to
where Es is the cooling energy from the snow storage
simplify the emptying of the pit and to enable the
[kW h]; ṁm is the melt water flow [kg/s]; cp is the heat
pumping of pond water to the snow blowers and water
capacity of melt water [4180 J/kg K]; Tm1 is the melt
hoses. The maximum melt water outlet flow to the
water temperature before heat exchanger [°C]; Tm2 is
recipient stream was reduced from 0.045 to 0.005 m3
the melt water temperature after heat exchanger [°C]; t
s 1 in order to avoid sediment transport. This measure
is the time step [h].
only affects the emptying time after the cooling season.
The cooling energy from the chiller was calculated
as
3. Measurements and evaluations Ec ¼ m_ c cp ðTc2 Tc1 Þt ð2Þ
where Ec is the cooling energy from the chiller [kW
The air temperature, precipitation, and global radi- h]; ṁc is the cooling circuit water flow [kg/s]; cp is the
ation were measured hourly. Before June 16th, when heat capacity of cooling circuit water [4180 J/kg K];
Tc1 is the cooling circuit water temperature before heat 11th was 363 mm and global radiation between June
exchanger [°C]; Tc2 is the cooling circuit water tem- 16th and September 11th was 282 kW h m 2. The
perature after heat exchanger [°C]. climate corresponded well with an average year except
The running coefficient of performance (COPrun), for the intense precipitation in July that was almost
defined as the ratio between the delivered cooling twice the normal. The climate and the cold extraction
energy and input energy, was calculated. COPrun inclu- resulted in the reduction of the snow storage volume as
ded electricity used in the two systems and all energy shown in Fig. 6.
for the handling of the snow and wood chips. The co- The extracted cooling energy per day from the two
oling power was calculated as the mean power per hour. systems, Eqs. (1) and (2), is shown in Fig. 7. The total
The ground temperature under the storage pond was cooling energy from the snow cooling plant was 607.9
measured above and below the ground insulation, at MW h and the maximum cooling power was 1366 kW
three locations. (June 30th, 2:00 p.m.). The total cooling energy from
Meltwater of urban snow deposits is highly conta- the chiller was 47.6 MW h and the maximum cooling
minated (Viklander, 1997). The Environment and Pub- power was 762 kW (August 29th, 2:00 p.m.). The
lic Health Office of Sundsvall required analysis of chiller also ran between May 9th and June 6th, before
relevant pollution parameters of the outlet meltwater, the storage was in operation; however, for this period
e.g. heavy metals, hydrocarbons, oxygen demands, and no values of cooling energy exist.
nutrition content, since the snow cooling plant at the The COPrun of the snow plant was 10.5, and 2.2 for
Sundsvall Hospital is the first of its kind. These param- the chiller. The COPrun of the snow cooling plant would
eters were also measured in 1998, i.e. the year before have been approximately 5% higher if it had been
the snow storage was put into operation. Measurements possible to separate the electricity that was not used
were made at five locations: in the snow storage, in the directly for the cooling system (lamps, computers, etc.).
outlet stream, in two small ponds downstream, and at To compare the two systems within a larger context,
the inlet and outlet of a small lake downstream. a COPtotal was also calculated. Except for the operating
energy, the denominator of this coefficient also
included the yearly material depreciation, i.e. the total
4. Results and discussion energy required to manufacture and construct a system
divided by the estimated lifetime, based on values by
Climate data from the year 2000 are shown in Fig. 5. Hagerman (2000). The COPtotal of the snow cooling
The total precipitation between May 1st and September plant was 8.6 and the chiller 2.2. The COPtotal of the
Fig. 5. Climate data for The Sundsvall Hospital from May 1st to September 11th 2000.
Fig. 6. Measured snow volume at five occasions from May 1st to August 31st, during the cooling season 2000. The last value was estimated.
chiller was based on the assumption that the chiller had ature followed the air temperature with a slight trailing
produced the same amount of cooling energy as the when water or snow was not covering the storage
snow cooling system. This was necessary to make a fair bottom. When snow or water insulated the pond bot-
comparison between the two systems since the relative tom the temperature rose to slightly above zero.
contribution of the material depreciation increases with According to the authorities, the melt water quality
a decreased level of delivered cooling energy. in the recipient was such that further treatment had to
During the whole season, the ground temperature be considered, in particular concerning non-biodegrad-
below the insulation was rather stable, ranging from able compounds, phosphorus, and lead. However,
+ 4 to + 7 °C, and varied with time and location. In the other heavy metal concentrations were generally low
pre-study, a ground water temperature of + 6 °C was (The Environment and Public Health Office of Sunds-
used, which corresponded well to the measured tem- vall, 2000). The fact that urban snow is contaminated
peratures. Above the insulation, the ground temper- is in a way advantageous for this technique, since the
Fig. 7. Diurnal cooling energy need, extracted from the snow cooling plant and the chiller.
melt water can easily be controlled and treated. Con- input, was included, the total COP became 8.6 for the
tamination of the snow and substances from the wood snow cooling system and 2.2 for the chiller.
chips might cause problems in the future for the Despite some minor problems the plant worked well
equipment being used. and met the required cooling energy and power. The
The pre-study showed that heat transfer from the air melting process was not ideal because the sides of the
caused the main part of the natural melting. Because of snow pile melted first, thereby directly exposing the
this and since about 85% of the running cost of the first snow to air. This resulted in warmer re-circulation
year was connected with the handling of wood chips, it water because of shortcuts and a faster snowmelt.
is clear that the top insulation problem needs to be The first problem was handled by using hoses to return
examined further. This concerns both the handling of a part of the meltwater to the central parts of pile. The
wood chips and minimising the heat losses through the second problem was not attended to.
top insulation. Different types of insulation needs to be The ground temperature above the insulation was
evaluated, e.g. sawdust, plane chips, plastic mat- close to the air temperature when the pond bottom was
tresses, and rigid buildings, and also alternative ways not covered by water or snow and slightly above zero
of applying the insulation. One specific quality with when it was covered. The temperatures below the
materials such as wood chips is that they absorb ground insulation varied from + 4 °C to + 7 °C during
meltwater that in turn evaporates, especially during the entire year. This corresponds well to the assumed
sunny and windy days. This evaporation cools the pile. ground temperature in the pre-study.
The construction cost will also be reduced in future The meltwater quality in the recipient was such
snow storage systems. Based on the experience from that further treatment had to be considered, in partic-
Sundsvall, it has been concluded that the construction ular concerning non-biodegradable compounds, phos-
cost of a similar storage for 120,000 m3 of snow phorus, and lead. The snow storage makes it possible
( f 6000 MW h) would be about 0.8 M EURO. The to treat the meltwater.
estimated pay back time is approximately 3 years; The investment cost of the plant was 1.3 MEUR,
however, a greater cooling power would reduce this and the first year running cost was about 15.3 EUR
time. The estimated technical life of a snow storage MW h 1 (total 0.01 MEUR).
plant is 40 years. The thermal insulation qualities of wood chips for
snow storage, which will be studied further at Luleå
University of Technology, are not only due to the low
5. Conclusions thermal conductivity, but also due to evaporative cool-
ing. Snow pollution and possible treatment methods of
The first cold extraction period of the Sundsvall the meltwater will also be studied.
Hospital Snow Cooling Plant was between June 6th
and August 29th 2000. Its maximum snow volume was
18,800 m3, of which 75% was natural snow from Acknowledgements
nearby streets and squares. The rest of the ice/snow
was artificially made with snowguns and a water hose. This study is part of Kjell Skogsberg’s PhD. studies
The pile was thermally insulated with 0.2 m of wood and was funded by the Swedish Energy Office
chips in April. After the cooling season, 75% of the (Energimyndigheten). I acknowledge Prof. Bo Nordell
wood chips was saved for the following year and the and Prof. Anders Sellgren for their constructive work
rest was burnt in a co-generation plant. with this article, and the staff at The County Council of
Between June 16th and September 11th, 655 MW h the Regional Hospital of Sundsvall for all their help.
of cooling energy was delivered to the hospital, of
which 93% came from the snow cooling system. The
References
maximum cooling power was 1366 kW. The running
COP of the snow cooling system was 10.5 while the Hagerman, A., 2000. Snökylanläggning kontra kylmaskiner (Snow
corresponding value of the chiller was 2.2. When Cooling Plant vs. Chillers). Undergraduate thesis. Division of
material depreciation, based on lifetime and energy Resource Management, Mid Sweden University. (In Swedish.)
MacCracken, C.D., Silvetti, B.M., 1987. Charging and discharging Taylor, T.B., 1985. Ice ponds. AIP Conf. Proc., s562 – s575.
long-term ice storage. ASHRAE Trans. 93, 1766 – 1772. The Environment and Public Health Office, 2000. Inspektion av
Nordell, B., Sundin, E., 1998. Snöupplag för säsongslagring av kyla fjärrkyleanläggning vid Sundsvalls sjukhus (Inspection of the
(Snow Deposit for Seasonal Storage of Cold). Water Resources district cooling plant at the Sundsvall Hospital). Dnr 99-720.
Engineering, Luleå University of Technology, April 1998. (In Sundsvalls Kommun, Miljökontoret. (In Swedish.)
Swedish.) Viklander, M., 1997. Snow Quality in Urban Areas. Doctoral Thesis
Skogsberg, K., Nordell, B., 2001. The Sundsvall Hospital snow 1997: 21. Division of Sanitary Engineering, Luleå University of
storage. Cold Reg. Sci. Technol., vol. 33/1. Elsevier, pp. 63 – 71. Technology. ISSN 1402-1544.
Paper III
Skogsberg K and Lundberg A (2005). Wood chips as thermal

insulation of snow. Accepted for publication in Cold Regions and
Science Technology, Elsevier Science B.V., April 25, 2005.
Skogsberg K and Lundberg A (2005). Wood chips as thermal insulation of snow.
Wood Chips as Thermal Insulation of Snow

Skogsberg Kjell1, Lundberg Angela2
1
Division of Renewable energy, Luleå University of Technology, 971 87 Luleå, Sweden
2
Division of Applied Geology, Luleå University of Technology, 971 87 Luleå, Sweden
Abstract
Snow (and ice) cooling is in some areas an alternative cooling technique to chillers.
However, to utilise snow cooling during the summer, storage is needed. Snow storages
were traditionally insulated by sawdust. A large demonstration plant is in operation
since 2000 for cooling of the regional hospital in Sundsvall, Sweden. This storage is
thermally insulated by larger wood chips. This paper summarizes laboratory testing of
some key factors that influence the functioning of cutter shavings (another kind of wood
chips) as thermal insulation. The melt rate significantly increased with increased wind
velocity, light intensity (from spotlight bulbs), air temperature and air humidity. A one
cm thicker layer of cutter shavings decreased the melt rate about 0.097 kg m-2 h-1,
within the studied layer limits 2.5 to 7.5 cm. The melt rate with initially wet cutter
shavings was about the same as for initially dry cutter shavings. This was explained by
the increased evaporation which balanced increased thermal conductivity. Evaporation
was found to be an important part of the energy balance and reduced the melt rate. Solar
light reflectivity of new cutter shavings was measured in a separate outdoor experiment,
found to be about 20%.
Keywords
Thermal, insulation, renewable, energy, snow, cooling, cutter shavings
1 Introduction
The worlds cooling demand has increased considerably during the last decades as a
result of increased population and industrialisation in combination with increased
comfort demand, electronic equipment usage, and new building technologies (Paksoy,
2003). The turnover of the refrigeration and air conditioning sectors in the main part of
EU was about 20 G€ in 2003 (AREA, 2004). Conventional cooling techniques often
require cooling media that are flammable, poisonous, and/or environmentally
hazardous. In addition, cooling machines cause noise and an aesthetic disadvantage of
buildings.
One alternative cooling technique utilises natural and/or artificial snow which is stored
from winter to summer. The cold is extracted by re-circulation of melt water, air, or
some other energy carrier in the snow storage (Skogsberg and Nordell, 2001). Snow can
be stored on ground, inside a building or under ground. When the snow is stored on
ground, thermal insulation of the snow is necessary (Nordell and Sundin, 1998). At the
first large scale snow cooling plant in Sweden, larger wood chips (2-15 cm) were used
-1-
as thermal insulation. Saw dust is the traditional thermal insulation of snow but there are
alternatives, e.g. different plastic materials, encapsulated hay (Taylor, 1985), rice shell
(Nakamura, 2004) and other types of wood chips. The term wood chips here include
wood powder, saw dust, cutter shavings and larger wood chips.
1.1 Objectives
The objectives of this paper were to establish the dominating factors influencing the
melt and evaporation rates of snow, thermally insulated with cutter shavings, and the
magnitudes of their effects.
The hypothesis was that the high thermal resistance of different kinds of wood chips on
snow is explained by evaporative cooling and relative low thermal conductivity. The
influence of cutter shaving layer thickness, wind velocity, light intensity, air
temperature, air humidity, cutter shavings moisture content and hindered convection
were treated in this study. The solar light absorptivity of cutter shavings was measured
in a separate outdoor experiment. Heat conduction, surface roughness, water transport
potential and compaction were also assumed to affect the surface melt rate and were
partly discussed but not examined in the study.
2 Heat and mass transfer

In a snow cooling plant the melting can be divided into forced and natural melting.
Forced melting is defined as caused by re-circulation of the energy carrier. The natural
melting can be divided into surface melt, rain melt and ground melt. Rain and ground
melt contributes to less than 20% of the natural melt when a 30,000 m3 snow pile in
Sundsvall is covered with a 0.2 m layer of larger pieces of wood chips (Skogsberg and
Nordell, 2001). The present study comprises natural surface melt of snow covered with
a layer of cutter shavings.
The heat and mass transfer through a wood chip layer on snow are closely related,
Figure 1. Most melt water percolates downward through the snow but a fraction is
transported upwards through the chip layer due to capillary forces and evaporation.
There may also be condensation in the wood chip layer.
Radiation
Air
Heat convection Water transport

Wood chips
Heat conduction
Snow Water transport
Figure 1 Principle of heat and mass transfer through a layer of wood chips on snow.
Evaporation requires energy and cools the wood chip layer, and thus the melt rate
decreases. As the latent heat of vaporisation is about 7.5 times greater than the latent
-2-
heat of fusion, also minor evaporation rates decreases the melt rate significantly
(Skogsberg, 2001). We found no studies of evaporation from wood chips covered snow,
but the open field mean evaporation from snow is about 4-13 g m-2 h-1 (Harding, 1986;
Kojima et al., 1985; Bengtsson 1980). Since the snow storage remains throughout the
summer, evaporation can also be compared with summer values. The mean potential
evapotranspiration April-September in Sweden is 68-137 g m-2 h-1 (Eriksson, 1981).
Typical evaporation rates for grass and heather plains in the UK, during wet conditions,
are 50-150 g m-2 h-1 (Calder, 1986).
The radiation exchange at the wood chips surface contains short (solar) and long wave
(heat) radiation. During the summer the magnitude of heat radiation is much smaller
than solar radiation, but matters especially at night. If the snow is covered the radiation
properties change. Reported solar radiation reflectivity (also called albedo) varies from
0.8 for fresh snow to 0.1 for old snow along roads (USACE, 1956; Bengtsson and
Westerström, 1992; Lundberg and Beyrel, 2001; Semádeni-Davies, 1999). No albedo of
wood was found, but albedo of humid soil is 0.1 (Hagentoft, 2001).
The convective heat and mass transfer through the wood chip layer is influenced by
ambient air properties (temperature, humidity and wind velocity) and wood chips
properties (surface roughness, water transport capacity, and compaction).
The thermal conductivity of wood chips depends on wood quality, chips structure,
compaction and moisture content. The thermal conductivity of dry (12% moisture
content by mass) sawdust and cutter shavings varies from 0.08-0.14 W m-1 K-1,
depending on compaction (Mörtstedt and Hellsten, 1999). The thermal conductivity of
water is 0.60 W m-1 K-1. Skogsberg (2004) found that the thermal conductivity of cutter
shavings varied roughly linearly from about 0.1-0.30 W m-1 K-1 when moisture ratio
varied from 11-380%, with dry density 70-100 kg m-3. In the pre-study of the Sundsvall
snow cooling plant the surface melt was calculated by assuming an effective wood chips
thermal conductivity of 0.35 W m-1 K-1 and using the temperature difference between
air and snow (Skogsberg and Nordell, 2001).
The density of natural snow covers varies from 100-500 kg m-3 (Gray and Male, 1981).
After consolidation with 0.25 bar Nagano et al. (2000) found that the snow density was
400-500 kg m-3. For some large snow deposits in Luleå (Sweden) the snow density was
found to be about 650 kg m-3 (Viklander, 1997) and 700 kg m-3 (Sundin, 1998). The
outflow from a melting snow pack is affected by liquid water in the snow. Old snow
holds relatively little water due to coarse-grained structure with large pores. Lundberg
(1997) reports a maximum liquid water content of 5.5% by volume (13.8% by mass) for
old snow in nature, density about 400 kg m-3.
2.1 Debris covered glaciers

The debris layer covering many glaciers reminds of a wood chip layer on snow, since
heat and mass (water) is transferred through the layer. The glacier ablation (melt) rate,
relative bare snow and ice, increases for very thin layers (<0.03 m) but decreases for
thicker layers. With 0.10 m of debris the melt rate was reduced 35%-66% (Mattson and
Gardner, 1991; Mattson et al., 1993; Kayastha et al., 2000; Pelto, 2000; Takeuchi et al.,
2000). Corresponding reductions with 0.40 m of debris was 59%-85% (Mattson and
Gardner, 1991; Kayastha et al., 2000).
-3-
The increased ablation rate with thin debris layers is caused by increased absorptivity of
solar radiation. The decreased ablation rate for thicker layers was explained by the
insulating effect of, and heat storage in, the debris layer (Mattson and Gardner, 1991;
Takeuchi et al., 2000). Mattson and Gardner also found that one day with wet debris the
evaporation was 24% of the energy loss.
3 Laboratory experiment
In this study melt and evaporation rates of snow thermally insulated with cutter
shavings were investigated. Air was blown through a thermally well insulated box with
snow, covered with a layer of cutter shavings. Light bulbs in the ceiling simulated solar
radiation. The box measured 1.25u0.65u0.90 m (LuWuH), Figure 2 and Figure 3.
Tair Light
RH
Fan Cutter shavings T of diff. levels
Snow
Melt water box Outflow
Scale
Figure 2 Section outline of the test rig. An induced airflow passed over the snow. Air
temperature, relative humidity, total weight, temperatures and melt water
flow in the cutter shaving layer were recorded.
Figure 3 Pictures of the test rig. The rig did lean to achieve melt water outflow. The
right picture is a view from the air outlet.
The influence on melt and evaporation rates by five main and two complimentary
factors was studied. The main factors were cutter shaving layer thickness (zcs), wind
velocity (vw), light intensity (ilight), air temperature (Tair) and absolute air humidity (AH).
AH was calculated from relative air humidity (RH) and Tair. The interaction influences
of the main factors were also studied. The complimentary factors were cutter shaving
-4-
moisture content and hindered evaporation. Solar light reflectivity (rsw) of cutter
shavings was measured in a separate outdoor experiment. The studied responses were
evaporation rate (ER) and surface melt rate (SMR), defined as melt rate of snow directly
under the cutter shaving layer (explained later). The experimental factors studied are
seen in Table 1.
Table 1 Studied experimental factors.

Factor and abbreviation Unit Range Measured Controlled Comment
Wood chips layer thickness zcs [cm] 2.5-7.5 X X
Wood chips moisture content wcs [%] 10-362 X X
Wind velocity1 vw [m s-1] 0.6-1.8 X X
Convection in wood chips layer Y/N X Hindered by plastic sheet
Light intensity2 ilight [W m-2] 0-190 X X By spotlight bulbs
Air temperature Tair [oC] 11.8-20.1 X
Absolute air humidity AH [g m-3] 0.9-4.9 Calculated
Relative air humidity RH [%] 7.2-34.1 X
Total weight3 mtot [kg] 108-144 X
Outflow m out [g 15 min-1] 147.6-633.6 X
Wood chips temperature Tcs [oC] 0-24.5 X
Wood chips solar light reflectivity rsw [%] 0.15-0.32 X Measured on a small spot
Surface Melt Rate SMR [kg m-2 h-1] Calculated from Outflow
Evaporation rate ER [g m-2 h-1] Calculated from mtot
The air outlet channel was built as a contraction to reduce variations in wind velocities,
following recommendations of Morel (1975), and thus rendering possible representative
measurements of wind velocities. With the used contraction ratio of four and a certain
shape, the wind velocity variations at the outlet shall be reduced by 94%. The measured
mean vw after the contraction, with standard deviations, were 3.44 ± 0.17, 6.20 ± 0.22,
and 9.46 ± 0.29 m s-1. This corresponded to 0.6, 1.2, and 1.8 m s-1 in the box. These
wind velocities corresponded to 2.3, 4.5, and 6.8 m s-1 at standard height 10 m,
assuming a roughness parameter of 0.0005 m (same as for mowed grass) and a
logarithmic wind profile according to Monin (1970). Solar radiation was simulated with
eight 60 W spotlight bulbs, light efficiency 5% according to manufacturer. The melt
water was led into a covered plastic box after passing a tipping bucket, both on the
scale. The passage between the box and the tipping bucket was covered with a plastic
sheet to reduce evaporation losses there.
Measurements were made every third minute and mean values were recorded every 15
minutes. Melt water flow was summed every 15 minutes. In some runs the temperature
of the cutter shaving layer, Tcs, was measured at four levels and in some runs only at the
surface of the layer. The instruments used were a Toledo load cell (max weight 600 kg,
r0.1 kg), a tipping bucket (3.6 g flip-1, r2.0 %), two Rotronic Air-Probe YA-100-
Hygrometers (RH r 3.0%, Tair r 0.5qC), eight pT-100 (r1.0oC), a hot wire anemometer
Swema ATD 81 (r 5%) and a MacSolar SLM 018c solar radiation power measuring
unit (r 2%). Accuracies except tipping bucket were according to manufacturers.
Incoming and reflected short-wave radiation of the outdoor experiment was measured
with Kipp and Zonen CM 6 pyranometers. All instruments were calibrated before each
group of experiments.
-5-
Before each run the box was filled with fresh natural snow, outdoor air temperature -5
to +5oC. Two test runs preceded each new group of experiments, to check instruments
and reduce water absorption in the wood frame during real runs. Each block of runs (see
below) was finished before the next started. After each run old cutter shavings were
removed and new and old snow was mixed. About 20% of the cutter shavings closest to
the snow surface were replaced before each new run. This was made to reduce
contaminations from snow, hygroscopic moisture equilibrium time, and effects of aging.
The remaining cutter shavings were reused after minimum one week in room climate, to
restore the hygroscopic equilibrium level. Measurements showed that less than 10% of
the moisture change from the previous run remained after three to six days.
The experiment was divided in five parts; Set 1 was a first study to roughly investigate
effects of different factors and equipment function (not reported here), Set 2 was to
quantify the effects with light, Set 3 was to quantify the effects without light, and Set 4
was to estimate SMR without evaporation by putting a diffusion proof plastic sheet just
beneath the cutter shaving layer surface. In Set 1-4 the cutting shavings moisture
content was in equilibrium with air humidity. Set 5 was to indicate the effect on SMR
and ER with initially wet cutter shavings.
Melting snow contains liquid water. The measured outflow thus included melt from the
sides of the box and melted water stored in the recently melted snow. The reported
SMR values were thus calculated by adding the measured outflow and evaporation
rates, subtracting previously stored water in the snow, and finally subtracting bottom
and side melt rates. The liquid water content was measured with a dilution method
(Davis et al., 1985) in three experiments, taken from Set 3. The measured average snow
liquid water content was 16.4% (13.6-21.8%) by mass, and it was assumed to be 16.4%
during all experiments. The measured liquid water content corresponded well to
Lundberg’s (1997) results of 13.8% for old snow. The side and bottom melt was
calculated assuming one-dimensional heat conduction, corresponding to 8.5-31.7% of
the snow melt rates.
The reported (calculated) SMR were two hour mean values, measured when the melt
rate was judged most stable (about 4-11 hours after run start). Since all snow and melt
water was on the scale the weight change was assumed to be because of evaporation or
condensation. The reported ER were measured weight change over the whole run
length, i.e. an overall mean. Snow density measurements were performed in Set 3.
The solar light reflectivity of cutter shavings was measured outdoor during spring, on a
small spot (1.1 m2, layer thickness 0.1 m). Since the spot was small, the measured
reflectivity was influenced by the surrounding snow and the measured values were re-
calculated according to Olseth (2000).
rmeasured rsurrounding 1 ' R2

rtrue where ' (1)
' R h2
2
Here rtrue is the true reflectivity of the measured surface [-], rmeasured is measured
reflectivity of the surface [-], rsurrounding is reflectivity of the surroundings [-], R is radius
-6-
of the studied surface [m], and h is the measurement height [m]. Due to difficulties with
instrument no measurements of long wave radiation were performed.
4 Statistical methods
In Set 2 and 3 both main (single) and interaction effects were analysed. An interaction
effect is when the effect of one factor depends on the level of some other factor/-s (Box
et al., 1978). Assume an experimental design with two factors A and B and response y.
A and B were kept at two levels each, high and low. If the response y increased with 3
when Alow was changed to Ahigh, at both Blow and Bhigh, there was no indication of
interaction. On the other hand if y increased 3 at Blow and 6 at Bhigh, there was an
interaction effect. The effect of changing A thus depended on the level of B.
In Set 2 and 3 the main factors zcs, vw and ilight were kept at chosen levels, while Tair and
AH was measured but not controlled. In Set 3 ilight was zero. Set 2 and 3 were run in two
blocks of 10 and 5 runs each. One block was a share of experiments that might be
expected to be more homogenous than the whole group of experiments. Here the block
divider was order, i.e. time. The second blocks were replicates of the first blocks
concerning zcs, vw and ilight but not concerning Tair and AH. Within blocks the order was
randomized.
The statistical analysis was both qualitative and quantitative. Model factor coefficients
showed the measured quantitative effect, while P-values and confidence levels showed
the certainty by which the factor influences the response.
The confidence level, usually 95% or 99%, returns a confidence interval around the
experiment mean value of some factor, which with the chosen certainty includes the true
mean value. If the interval excludes zero, it is said that the factor is significant on this
level. A higher confidence level causes a relative bigger confidence interval. The
absolute interval depends on how spread the responses were. The achieved P-value of a
factor or model shows the statistical risk that there is no effect of this factor or model,
i.e. a lower P-value means larger probability that the factor has an effect on the
response. If a factor has a P-value of 0.003 it is 0.3% risk that a value as large as the
observed would occur if the factor actually has no effect, i.e. it is very likely that the
factor has an effect in this case (Box et al., 1978).
The R2 statistic indicates the percentage of variability for which the model accounts.
The standard error of estimate, i.e. the standard deviation of residuals, were also
reported and can be used to construct prediction limits for new observations.
Set 4 and 5 were made with middle settings of zcs, vw and ilight of Set 2 to facilitate
comparisons.
5 Results
Results are presented in this section and discussed in section 6. All setups and results
are seen in Table 2.
-7-
Table 2 Factors and results of Set 2-5.

Run Block Order zcs vw ilight Tair RH AH SMR Evaporation ER
[cm] [m s-1] [W m-2] o
[ C] [%] [g m-3] [kg m-2 h-1] [kg total] -2
[g m h ]
-1
2.1.1 1 9 7.5 1.8 190 15.1 18.0 2.3 0.86 1.1 63.7
2.1.2 1 7 7.5 0.6 50 16.7 16.9 2.4 0.49 0.8 41.0
2.1.3 1 6 2.5 0.6 190 16.9 21.2 3.1 1.07 0.8 51.8
2.1.4 1 5 2.5 1.8 190 16.1 23.2 3.2 1.61 1.3 71.9
2.1.5 1 4 7.5 0.6 190 16.1 15.8 2.2 0.86 0.7 60.5
2.1.6 1 3 5.0 1.2 120 16.4 16.4 2.3 0.96 0.4 78.8
2.1.7 1 1 2.5 0.6 50 16.3 29.9 4.2 1.28 0.6 118.2
2.1.8 1 2 7.5 1.8 50 15.4 25.6 3.4 0.75 0.8 64.6
2.1.9 1 10 5.0 1.2 120 14.8 20.5 2.6 0.93 1.0 78.1
2.1.10 1 8 2.5 1.8 50 17.5 23.9 3.6 1.32 0.3 56.8
2.2.1 2 19 7.5 1.8 190 17.9 31.9 4.9 2.09 0.3 24.2
2.2.2 2 16 7.5 0.6 50 15.8 34.1 4.6 0.81 0.3 19.2
2.2.3 2 17 2.5 0.6 190 17.4 31.9 4.7 2.12 0.4 22.6
2.2.4 2 13 2.5 1.8 190 18.0 23.7 3.6 2.38 1.1 55.3
2.2.5 2 11 7.5 0.6 190 16.7 17.4 2.5 0.81 0.7 45.9
2.2.6 2 15 5.0 1.2 120 16.3 31.7 4.4 1.16 0.3 20.8
2.2.7 2 12 2.5 0.6 50 16.4 19.8 2.8 1.35 0.7 41.5
2.2.8 2 14 7.5 1.8 50 16.6 25.3 3.6 0.97 0.7 35.5
2.2.9 2 18 5.0 1.2 120 17.5 29.1 4.4 1.64 0.3 17.0
2.2.10 2 20 2.5 1.8 50 17.1 29.7 4.3 1.82 1.0 53.5
3.1.1 1 2 5.0 1.2 0 15.4 20.7 2.7 0.94 0.5 21.0
3.1.2 1 4 7.5 1.8 0 17.2 13.3 2.0 0.63 0.7 43.6
3.1.3 1 5 2.5 1.8 0 16.7 13.3 1.9 1.28 1.1 60.8
3.1.4 1 3 7.5 0.6 0 16.6 14.1 2.0 0.54 0.7 37.9
3.1.5 1 1 2.5 0.6 0 14.3 13.5 1.7 0.97 0.7 45.3
3.2.1 2 7 5.0 1.2 0 15.3 7.2 0.9 0.54 0.9 51.5
3.2.2 2 9 7.5 1.8 0 11.8 28.2 3.0 0.50 0.4 28.1
3.2.3 2 10 2.5 1.8 0 14.8 15.7 2.0 0.69 0.6 42.8
3.2.4 2 8 7.5 0.6 0 15.0 20.3 2.6 0.69 0.5 31.6
3.2.5 2 6 2.5 0.6 0 16.0 10.4 1.4 0.83 1.0 58.6
4.1 - 1 5.0 1.2 120 16.5 12.8 1.8 0.72 0.5 29.7
4.2 - 2 5.0 1.2 120 17.5 8.7 1.3 0.75 -1.4 -84.1
4.3 - 3 5.0 1.2 120 15.1 10.5 1.4 0.59 0.3 17.6
5.1 - 1 5.0 1.2 120 20.1 23.0 4.0 1.51 5.6 303.0
5.2 - 2 5.0 1.2 0 17.9 12.3 1.9 1.19 5.1 212.8
5.1 Surface Melt Rate

SMR of Set 2, with light, was 0.75-2.38 kg m-2 h-1. All single factors (zcs, vw, ilight, Tair,
AH) were significant at the 95% confidence level, i.e. indicating that they significantly
influenced the response (P; 0.0025, 0.0291, 0.0054, 0.0246, 0.0051). No interactions
were significant.
SMR of Set 3, without light, was 0.50-1.28 kg m-2 h-1.Only zcs was significant at the
95% confidence level. At the 90% confidence level, zcs, Tair and AH were significant (P;
0.0095, 0.0923, 0.0715).
Since no large problems were found in the SMR analysis in Set 2 and 3 a common
multiple linear regression was made, Equation 2. The model was significant at 99%
confidence level and factors at 95% confidence level, with vw on the border (P-values;
-8-
0.0000, 0.0503, 0.0115, 0.0171, 0.0004). R2 was 82.6% and the standard error of
estimate was 0.232. There was no indication of serial auto-correlation in the residuals.
SMR 0.97 0.097 z wc 0.164 v w 0.00175 i light 0.102 Tair 0.192 AH (2)
where SMR is the surface melt rate [kg m-2 h-1], zcs is cutter shaving layer thickness
[cm], vw is wind velocity [m s-1], ilight is light intensity [W m-2], Tair is air temperature
[oC] and AH is absolute air humidity of inlet air [g m-3]. The observed and predicted
response of the model is in Figure 4.
2,4
2
observed
1,6
1,2
0,8
0,4
0
0 0,4 0,8 1,2 1,6 2 2,4
predicted
Figure 4 Observed vs. predicted Surface Melt Rate (SMR) for the model of Set 2 and 3.
SMR of Set 4, with a plastic sheet on top of the cutter shavings, was 0.59-0.75 kg m-2 h-
1
. In Set 5, with initially wet cutter shavings, SMR was 1.51 and 1.19 kg m-2 h-1. SMR
of comparable runs with initially dry cutter shavings (Set 2) was 0.93-1.64 kg m-2 h-1.
The initial moisture ratios (water weight over dry cutter shavings weight) of Run 5.1
and 5.2 were 362% and 336%. After Run 5.2 the moisture ratio was reduced to 66%,
101% and 237% in top, middle and bottom of the cutter shaving layer.
5.2 Evaporation rate

ER of Set 2, with light, was 17.0-118.2 g m-2 h-1. Only the block factor was significant
at 99% confidence level (P; 0.0004). The average ER of block 1 and 2 was 68.5 and
33.6 g m-2 h-1. When the block factor was divided in Tair and AH these factors were not
significant (P; 0.3257, 0.2058). The ratio ER/SMR was 1.0-9.2% (average 4.8%) as
mass and 7.4-66.6% (average 34.6%) as energy.
ER of Set 3, without light, was 21.0-60.8 g m-2 h-1. Here Tair, AH and the interaction
effects zcsTair and zcsAH were significant at the 90% confidence level (P; 0.0339,
0.0210, 0.0358, 0.0547). The ratio ER/SMR was 2.2-9.5% (average 5.9%) as mass and
16.8-71.2% (average 43.9%) as energy. Since there were significant interactions in Set 3
but not in Set 2, no common ER analysis was made.
In Set 4, with plastic sheet, ER was 29.7, -84.1, and 17.6 g m-2 h-1. It thus failed to
hinder mass transfer. No explanation was found of the measured condensation in run
-9-
4.2, see section 6.1 for further discussion. ER of Set 5 was 303.0 and 212.8 g m-2 h-1,
with and without light. ER of Set 2 runs with same settings were 17.0-78.8 g m-2 h-1
with light and 21.0-51.5 g m-2 h-1 without light.
5.3 The other measurements

For calculated SMR, cutter shaving layer thickness, measured Tair, and Tsnow=0oC the
effective thermal conductivity of cutter shavings in Set 2 was determined to 0.15-0.81
W m-1 K-1, average 0.32 W m-1 K-1. In Set 3, the effective thermal conductivity of the
cutter shaving layer was 0.11-0.32 W m-1 K-1, average 0.21 W m-1 K-1.
There were too few Tcs measurements to make any certain statements, but some results
are however presented. In Set 2, with light, Tcs at the surface was higher than Tair with
one exception, with zcs=2.5 cm and ilight=50 W m-2. The cutter shavings surface
temperature increased with light intensity, but it could not be confirmed that increased
vw decreased the temperature difference between Tcs and Tair. Without light, Set 3, Tcs at
the surface was about the same as Tair. Tcs decreased roughly linearly through the cutter
shaving layer, but it was difficult to tell for zcs=2.5 cm, since there were problems to
locate the probes exactly at the correct heights.
The total snow density, inclusive liquid water content, of Set 3 and 5 was 313-461 kg m-
3
, average 411 kg m-3. This is about two thirds of the snow deposit densities reported.
The measured cutter shavings solar light reflectivity sank from 0.50 initially to about
0.45 after a month, probably caused by cutter shavings getting wet. Recalculation to
reduce the effect of the surrounding snow gave that the true solar light reflectivity
decreased from about 0.32 to 0.15.
6 Discussion
6.1 Surface Melt Rate

With light (Set 2), SMR was influenced by zcs, vw, ilight, Tair and AH, as expected. Within
the studied limits, increased zcs decreased SMR. An increased level of any of the other
factors increased SMR, no matter the level of the other factors. The convective net heat
transfer must thus have been from air to snow, and not from the warmer cutter shavings
surface to the air, since increased wind velocity increased the melt rate. The increased
air humidity gave decreased evaporation rate which led to increased SMR. Without light
(Set 3), the effect of vw decreased or disappeared. There were no significant interaction
effects in Set 2 and 3.
The combined multiple linear model of Set 2 and 3, Equation 2, was significant on the
99% confidence level. The model showed that with a 1 cm thicker cutter shaving layer
SMR decreases approximately 0.10 kg m-2 h-1. This corresponded to a SMR change of
2.9-13.0%. According to the achieved model, the same effect is caused by decreasing
the level of vw, ilight, Tair and AH with 0.60 m s-1, 55.4 W m-2, 0.95oC and 0.51 g m-3,
respectively.
Both SMR and ER of Run 4.1 and 4.3 decreased, relative the comparable runs 2.1.6 and
2.1.9, even though ER was not completely hindered by the plastic sheet. By using
Equation 2, from the analysis above, and vw=0 (hindered internal convection), the
-10-
calculated SMR corresponded well to the experimental SMR. In run 4.2 both SMR and
experiment factors were comparable with run 4.1 and 4.3. Since the SMR and ER
energies of Run 4.2 were of the same magnitude (68.6 and -61.8 W m-2) the
condensation would be expected to drastically influence the melt rate. This was not the
case. Neither did AH increase during the run. We thus suspected some kind of
measurement error.
In Set 5, with initially wet cutter shavings, SMR was roughly as calculated from
Equation 2. This indicated that larger thermal conductivity with wet cutter shavings,
initially three times larger, was counteracted by increased evaporation. Set 5 was
comparable with conditions after a rainfall.
In this experiment ordinary spot light bulbs were used to simulate solar radiation, and
the short wave radiation was measured with an instrument calibrated for solar radiation.
With high ilight (190 W m-2) the Tcs increase was comparable with non reported summer
measurements at the Sundsvall snow cooling plant, with a much higher solar radiation
(~600-800 W m-2). This implied that measured ilight values in the experiment
corresponded to higher solar radiation values. One possible explanation is that thermal
radiation from the bulbs was important, but not measured by the instrument. The ilight
coefficient of Equation 2 can thus not be used for solar radiation. The main finding was
however that light affected the melt rate.
Unlike reported melt rates from glaciers covered with a 2.5 cm debris layer, the melt
rate with zcs=2.5 cm decreased drastically (about 70%) relative an uncovered snow
surface. Cutter shavings increased the solar light absorptivity relative bare snow, but
thermal resistance and evaporation counteracted this effect.
Since the thermal insulation quality of wood chips on snow and ice depends on heat
conductivity, water transport capacity, evaporation capacity, absorptivity and relative
surface area, cutter shavings are favourable because of large surface area and airy
structure. A saw dust layer is favourable because of higher compactness, decreasing the
convective heat transfer. In an early study Skogsberg (2002) found that two snow piles,
one with 0.1 m of cutter shavings and one with 0.2 m of saw dust, melted at about the
same rate, i.e. indicating that cutter shavings are more efficient as thermal insulation on
snow. There was however some questions concerning different conditions in this study.
Bark is darker, more compact and has poor water transporting qualities which is why it
is assumed to be a poorer thermal insulation material on snow and ice. But in the end
enough thermal insulation is only a matter of layer thickness, e.g. ablation rate of a
glacier covered with a 0.4 m debris layer is reduced with 59-85%.
One alternative to natural materials is some plastic thermal insulation. The evaporative
cooling effect is then absent, but the heat conductivity is much lower (typically 0.04-0.1
W m-1 K-1). It is then possible to select a reflective surface, but so far there have been
problems to find sheets large enough at acceptable cost. This problem needs further
attention.
The principal results of this study are probably applicable for “similar” materials like
sheep wool, straw, rice shell etc. Insulation materials like mineral wool, without
capillary water transport, will benefit from lower thermal conductivity but suffer from
omitted evaporative cooling.
-11-
It must be noted that reported SMR were two hour mean values, and it was difficult to
judge which time period to use. It took some time before the snow water holding
capacity was filled and the outflow rate roughly equalled the melt rate. By waiting too
long, there became a gap between the walls and the snow, which quickly increased by
turbulent wind effects, especially with high wind and thin cutter shaving layer
thickness.
6.2 Evaporation Rate

The measured ER in this experiment was 17.0-118.2 g m-2 h-1, mean 54.9 g m-2 h-1. This
was higher than the mean values of open snow field evaporation (4-13 g m-2 h-1,
Harding (1986); Kojima et al. (1985); Bengtsson (1980)), and ranges well into the
summer evapotranspiration rates from open fields (50-150 g m-2 h-1, Calder (1986)).
The result is reasonable since air and surface temperatures were summer like, but the
active process of transpiration was not present in the cutter shaving layer. There was no
significant influence of vw and ilight at all. It was expected that at least vw would be
found to affect ER, since evaporation often is described by aerodynamic methods that
directly includes wind velocity.
In Set 2 and 3 the ratio ER over SMR as mass was small, in average 4.8 and 5.9%, but
in the energy balance evaporation was considerable, in average 34.6 and 43.9%. ER
included heating to 20oC, since the vapour leaving the cutter shaving layer was heated,
either before or after evaporation.
By omitting evaporation, the cutter shavings temperature and thus the melt rate will
increase. The melt rate increase with omitted evaporation was not directly seen, since a
higher cutter shavings temperature also increased radiation heat losses. The net effect of
ER was investigated by plotting SMR against the ratio ER/SMR, Figure 5, since an
increase of vw and ilight increased both SMR and ER. Analysis showed a statistically
significant relationship at the 99% confidence level, i.e. a relative high ER decreased
SMR.
2,4
2
1,6
SMR
1,2
0,8
0,4
0
0 0,02 0,04 0,06 0,08 0,1
ER/SMR
Figure 5 The Surface Melt Rate (SMR) versus the ratio SMR over Evaporation rate
(ER).
-12-
ER of Set 4 is discussed together with SMR of Set 4 in the SMR section above. ER of
Set 5 was clearly larger than for initially dry cutter shavings, showing that evaporation
will increase after a rainfall, both night and day.
Since the box was a wood frame with painted wood bottom and plastic insulation glued
to the frame there was a leakage risk through edges and screw holes, in spite of silicone
sealing. To ensure that evaporation really took place, one experiment was conducted
with a plastic box inside. The measured ER was about the same magnitude as for
comparable runs. In future experiments it is suggested to put the snow in a plastic box
and increase the thermal insulation to decrease the disturbing side and bottom melts. It
is also suggested that side and bottom melt rates are investigated experimentally instead
of assumed from calculations.
6.3 Effective thermal conductivity and solar light reflectivity

The average effective thermal conductivities with light (Set 2), and without light (Set 3),
of 0.32 and 0.21 W m-1 K-1 were lower than the value used in the prestudy of the
Sundsvall Hospital snow cooling plant (0.35 W m-1 K-1). The thermal resistance was
thus larger, i.e. the cutter shaving layer was a better thermal insulation than expected.
The true (re-calculated) solar light absorptivity of new, dry, cutter shavings was about
80%. It will increase as cutter shavings get wet and darken with age. This, together with
possible water transport deterioration, makes it important to study time impact, since it
is economical to reuse cutter shavings for some years.
7 Conclusions
In this study the surface melt rate (SMR) and evaporation rate (ER) of snow, thermally
insulated with a cutter shaving layer, was studied. SMR was 0.49-2.38 kg m-2 h-1. The
experiment showed that a factor change within the studied limits (zcs=2.5-7.5 cm,
vw=2.3, 4.5, 6.8 m s-1 at 10 m height, ilight=0-190 W m-2, Tair=11.8-20.1oC, AH=0.9-4.9
g m-3) changed SMR as;
cutter shaving layer thickness (zcs) +2 cm SMR: -0.19 kg m-2 h-1
wind velocity (vw) +1 m s-1 SMR: +0.16 kg m-2 h-1
light intensity (ilight) +100 W m-2 SMR: +0.18 kg m-2 h-1
air temperature (Tair) +2oC SMR: +0.20 kg m-2 h-1
air humidity (AH) +1 g m-3 SMR: +0.19 kg m-2 h-1.

The multiple linear SMR model of the main Sets (2 and 3) was statistically significant at
the 99% confidence level. There were no interaction effects within the studied limits.
ER varied from 17.0-118.2g m-2 h-1. One run gave condensation, and we could not find
any explanation why. In Set 2 (with light), only the block effect was significant.
Without light (Set 3), the main factors Tair and AH and the interaction effects zcsTair and
zcsAH were significant.
-13-
The average ratios ER/SMR of the main Set were 5.2% as mass and 37.1% as energy.
With larger ratio SMR decreased. The evaporative cooling effect was thus an important
part of the cutter shaving layer energy balance and the melt rate would increase
drastically without evaporation. SMR with wet cutter shavings was about the same as
with dry. This indicated that increased ER roughly balanced the increased thermal
conductivity.
In the future it is important to study how aging changes the properties of wood chips. In
order to further understand the processes in the wood chip layer, it is necessary to
combine experiments with mathematical modelling.
Acknowledgements
This study is part of Kjell Skogsberg's PhD studies and was funded by the Swedish
Energy Agency and Luleå University of Technology. We also acknowledge Prof. Bo
Nordell and Prof. Anders Sellgren for their constructive work with the article, Prof.
Kerstin Vännman and Robert Lundqvist for assistance with statistics, and Rolf
Engström, Helena Beyrel and The Division of Fluid Mechanics at LTU for practical
assistance.
References
AREA (2004). Homepage of Air conditioning and Refrigeration European Association,
http://www.area-eur.be/ March 28th 2004.
Bengtsson L (1980). Evaporation from a snow cover. Review and discussion of
measurements. Nordic Hydrology 11:221-234.
Bengtsson L and Westerström G (1992). Urban snowmelt and runoff in northern
Sweden. Hydrological Sciences Journal, 37:263-275. ISSN 0262-6667.
Box G E P, Hunter W G, and Hunter J S (1978). Statistics for experimenters. John
Wiley and Sons Inc., Canada. ISBN 0-471-09315-7.
Calder I R (1986). The influence of land use on water yield in upland areas of the UK.
Journal of Hydrology, 88:201-212.
Davis R E, Dozier J, LaChapelle E R, and Perla R (1985). Field and laboratory
measurements of snow liquid water by dilution. Water Resources Research,
21(9):1415-1420.
Eriksson B (1981). The “potential” evapotranspiration in Sweden. SMHI Reports, RMK
28, RHO 27. Liber Allmänna Förlag, 40 pp. ISSN 0347-2116.
Gray D M and Male D H, editors (1981). Handbook of Snow: principles, processes,
management and use. Pergamon Press, Toronto. ISBN 0-08-025375-X.
Hagentoft C-E (2001). Introduction to building physics. Studentlitteratur, Lund, ISBN
91-44-01896-7.
-14-
Harding R J (1986). Exchange of energy and mass associated with a melting snowpack.
Proceedings from Budapest Symposium, The International Association of
Hydrological Sciences, Wallingford. Publication 155:3-15.
Kayastha R B, Takeuchi Y, Nakawo M, and Ageta Y (2000). Practical prediction of ice
melting beneath various thickness of debris cover on Khumbu Glacier, Nepal,
using a positive degree-day factor. Proceedings of the workshop Debris-Covered
Glaciers held in Seattle, Washington, USA,September 2000). The International
Association of Hydrological Sciences, Publication 264:71–81.
Kojima K, Ishikawa N, Motoyama H, and Yamada Y (1985). Evaporation rate of snow
at the surface of a snow cover - observations in Sapporo and Moshiri, Hokkaido,
Low Temperature Science Series A-44:49-62, (in Japanese with English
summary).
Lundberg A and Beyrel B (2001), Ash on snow - a tool to prevent flooding? Nordic
Hydrology 32(3): 195-214.
Lundberg A (1997). Laboratory calibration of TDR-probes for snow wetness
measurements, Cold Regions Science and Technology 25:197-205.
Mattson L E and Gardner J S (1991). Energy exchanges and ablation rates on the
debris-covered Rakhiot Glacier, Pakistan. Zeitschrift für Gletscherkunde und
Glazialgeologie. 1989 (published 1991), 25(1):17-32.
Mattson L E, Gardner J S, and Young G J (1993). Ablation on debris covered glaciers:
an example from the Rakhiot glacier, Punjab, Himalaya. Proceedings of the
Kathmandu Symposium, November 1992. The International Association of
Hydrological Sciences, Publication 218, 1993.
Monin A S (1970). The Atmospheric boundary layer. Annual Review of Fluid
Mechanics, Vol. 2. SBN 8243-0702-X, California, USA.
Morel T (1975). Comprehensive Design of Axisymmetric Wind Tunnel Contractions.
Journal of Fluids Engineering, The American Society of Mechanical Engineers,
Paper No. 75-FE-17.
Mörtstedt S-E and Hellsten G (1999). Data och diagram. Energi- och kemitekniska
tabeller (Data and diagram. Tables of energy and chemistry technology). Liber
AB. ISBN 91-47-00805-9. In Swedish.
Nagano K, Mochida T, Miyazaki T and Shimakura K (2000). Field experiments on
consolidation of stored snow in the ground and a computational simulation for the
underground snow storage. Proceedings of 8th International Conference on
Thermal Energy Storage. Stuttgart, Germany. pp 441-446.
Switzerland.
Nordell B and Sundin E (1998). Snow Deposit for Seasonal Storage of Cold. Internal
report, Div. of Water Resources Engineering, Luleå University of Technology,
April 1998. (In Swedish)
-15-
Olseth J A (2000). Geophysical Institute, University of Bergen, Bergen, Norway.

Paksoy H Ö (editor) (2003). Annex 14, Cooling in all climates with thermal energy
storage. General State-of-The-Art Report, Subtask 1. July 2003. International
Energy Agency (IEA), Energy Conservation through Energy Storage (ECES).
Pelto M S (2000). Mass balance of adjacent debris-covered and clean glacier ice in the
North Cascades, Washington. Debris-Covered Glaciers (Proceedings of a
workshop held at Seattle, Washington, USA, September 2000). The International
Association of Hydrological Sciences, Publication 264:35–42.
Semádeni-Davies A F (1999). Urban snowmelt processes: modelling and observation.
PhD 1026, 1999, Department of Water Resources Engineering, Lund University.
Skogsberg K (2001). Seasonal Snow Storage for Cooling Applications. Licentiate thesis
2001:51. Division of Water Resources Engineering, Luleå University of
Technology, Sweden. ISSN 1402-1757.
Skogsberg K (2002). Träspån som termisk isolering på snö. En studie vid Polarbageriet,
Älvsbyn, Sverige. (Wood chips as thermal insulation on snow. A study at the
Polar bakery, Älvsbyn, Sweden.) Internal experiment report, Division of Water
Resources Engineering, Luleå University of Technology.
http://www.sb.ltu.se/~kjesko/homepage/publications.html. In Swedish.
Skogsberg K (2004). Heat conduction through wet cutter shavings– An initial
laboratory experiment. Internal experiment report, Division of Renewable Energy,
Luleå University of Technology.
http://www.sb.ltu.se/~kjesko/homepage/publications.html.
Skogsberg K and Nordell B (2001). The Sundsvall hospital snow storage. Cold Regions
Science and Technology 32:63-70, Elsevier Science B.V.
Sundin E (1998). Snow deposit melt and atmospheric icing: analysis and modelling by
weather station data. Doctoral thesis 1998:37. Luleå University of Technology,
Division of Water Resources Engineering. ISSN 1402-1544.
Takeuchi Y, Kayastha R B, and Nakawo M (2000). Characteristics of ablation and heat
balance in debris-free and debris-covered areas on Khumbu Glacier, Nepal
Himalayas, in the pre-monsoon season. Debris-Covered Glaciers (Proceedings of
a workshop held at Seattle, Washington, USA, September 2000). IAHS Publ.
264:53–61.
USACE, U.S. Army Corps of Engineers (1956). Snow hydrology - summary report of
the snow investigations. North Pacific Division, Corps of Eng., Portland, Oregon,
437 pp.
Viklander M (1997). Snow Quality in Urban Areas. Doctoral Thesis 1997:21. Division
of Sanitary Engineering. Luleå University of Technology. ISSN 1402-1544.
-16-
Paper IV
Skogsberg K (2005). Numerical modelling of heat and moisture

transfer through a cutter shaving layer on snow. Division of
Architecture and Infrastructure, Luleå University of Technolgy.
To be submitted.
Skogsberg K (2005).
Numerical modelling of heat and moisture transfer through a cutter shaving layer on snow.
Numerical modelling of heat and moisture transfer through a

cutter shaving layer on snow
Kjell Skogsberg
Division of Architecture and Infrastructure, Luleå University of Technology, Sweden
e-mail: kjell.skogsberg@ltu.se, phone: +46 (0)920 491 496
Abstract
This study concerned numerical modelling of heat and moisture transfer in 0.1-0.3 m
cutter shaving layers on snow, aiming to investigate how the surface melt rate depends
on air temperature, relative air humidity, solar radiation, air velocity, initial wood chip
moisture content and wood chip layer thickness. The capillary transfer was described by
a new defined function, based on established moisture transfer functions for unsaturated
porous materials and experimental results. Convection was calculated by simplified
functions where heat convection was included as apparent thermal conductivity. It was
found that the melt rate (0.051-1.507 kg m-2 h-1, or 4.7-139.6 W m-2) increased with
increased solar radiation, air temperature, and air velocity, and with decreased layer
thickness. With initially wet cutter shavings the stationary melt rate was the same or less
as with initially dry cutter shavings, due to greater evaporation or evaporation instead of
condensation.
Background
Seasonal snow and ice storage for cooling applications is an ancient technique that has
received renewed attention during the last 30 years. The traditional thermal insulation of
snow and ice was a layer of organic refuse such as sawdust or crop residues (Taylor,
1985). No studies of the mechanisms behind the good thermal insulation qualities of
sawdust were found before this project started.
In Sundsvall, Sweden, approximately two-thirds of a 30,000 m3 snow pile with 0.2 m
sawdust as thermal insulation remained after the summer. About 84% of natural melting
is caused by heat transfer through the wood chip layer (Skogsberg and Nordell, 2001).
The surface melt is thus important in open snow storage with wood chips as thermal
insulation.
A wood chip layer is permeable both to air and moisture transfer as rain, condensation,
evaporation and capillary transfer of melt water. In a laboratory experiment with snow
thermally insulated by 0.025-0.075 m cutter shaving layers, evaporation was found to be
a minor part in the mass balance, but significant in the energy balance. Moisture transfer
thus greatly influences heat transfer (Skogsberg and Lundberg, 2005). There are
different kinds of wood chips, i.e. wood powder, sawdust, cutter shavings and wood
chippings. This study focuses on cutter shavings as thermal insulation of snow.
1
Skogsberg K (2005).
Moisture in air and materials

The vapour content in air depends on air temperature and relative humidity. The
moisture content in a porous material, defined as moisture that is able to evaporate
(Ahlgren, 1972), depends on relative humidity, air and material temperatures, the
moisture history of the material, and if the material is in contact with free water. The
moisture content w [kg m-3] of a porous material is
m moist
w (1)
V
where mmoist is moisture weight [kg] and V is sample volume [m3]. Another way to
describe the amount of moisture is the moisture ratio u [kg kg-1= -],
m moist
u (2)
m dry
where mdry is the weight of the dry material [kg]. Moisture content and moisture ratio
are related by dry density Udry [kg m-3].
m moist ȡ dry
w u ȡ dry (3)
m dry
A porous material in humid air will eventually reach its equilibrium moisture content.
This hygroscopic moisture content mainly depends on the relative air humidity. A
temperature dependency also exists, though the moisture content in the relevant
temperature area only increases by about 1% per temperature decrease of 10oC (Stamm,
1964). It is thus customary to relate moisture content to only relative humidity for
normal building temperatures, where the moisture content commonly is decided for
20oC. This dependency is called a sorption isotherm and follows different paths
depending on material and if the moisture content is increasing or decreasing, Figure 1.
Increasing hygroscopic moisture content is called absorption (abs) and decreasing
desorption (des) (Hagentoft, 2001).
35
30 w,des
25 w,abs
wwc [kg m ]
-3
20
15
10
5
0
0% 20% 40% 60% 80% 100%
RH [%]
Figure 1 Sorption isotherms for wood chips with a dry density of 90 kg m-3, based on curves of
Swedish spruce with a dry density of 430 kg m-3 (after Hedenblad, 1996). wdes and wabs
are moisture content during desorption and absorption.
Gravimetric is a common method to measure the moisture content, where a specimen is

dried at 105oC until constant weight is attained. The moisture content will then be low
2
Skogsberg K (2005).
and in equilibrium with the relative humidity in the oven. The residual moisture ratio of
the specimen will typically be about 0.5% and is considered as dry (Skaar, 1988).
Density and moisture characteristics vary with wood type. The moisture ratio sorption
isotherms of spruce and pine are almost identical, though the moisture content sorption
isotherm of pine is about 20% higher than that of spruce. The sorption isotherms of
Figure 1 are based on spruce measurements converted to wood chips, assuming that
they have the same moisture ratio sorption isotherm as solid wood (Saarman, 1992).
At the so-called fibre-saturation point there is no water in the cell cavities, though the
cell walls are saturated. This is the equilibrium moisture content of saturated air. Below
this point a moisture content change will cause swelling or shrinking. From dry wood to
the fibre-saturation point the swelling is 3-10%, depending on direction and wood
quality (Skaar, 1988).
Water exists in cell cavities above the fibre-saturation point. To achieve this moisture
content level the material has to be in contact with liquid water. At the critical moisture
content wcr, chains of water filled pores are found and capillary moisture transfer starts.
Water is then transported into the porous material due to pressure differences. The
maximum moisture content a material can achieve by capillary suction is called the
capillary saturated water content.
Numerical modelling
Combined heat and moisture transfer through a horizontal layer of cutter shavings on
snow was simulated by a 1-D finite-difference method with an explicit scheme
performed in Microsoft Excel. The layer was divided into 21 cells of the same size,
except for the top and bottom cells that were half the thickness. Simulations were made
according to Table 1. Each simulation was given a code, e.g. 20-70-6-350-0.2 to
indicate air temperature Tair=20oC, relative humidity RH=70%, air velocity uair=6 m s-1,
incident solar radiation Psolar=350 W m-2 and cutter shaving layer thickness Z=0.2 m.
Table 1 Simulation set-ups, where e.g. 10-60-2-700-0.1/0.2 means that Tair=10oC, RH=60%, uair=2
m s-1, and Psolar=700 W m-2, were made with both Z=0.1 m and Z=0.2 m. Basic setting was
initially dry cutter shavings and Z=0.2 m. One factor of initial conditions and layer
thickness was changed at each simulation.
Tair RH uair Psolar Initially Z [m]
[oC] [m s-1]
-2
Code [%] [W m ] dry wet 0.1 0.2 0.3
10-60-2-0-0.2 10 60% 2 0 X X X
10-60-2-700-0.1/0.2 10 60% 2 700 X X X
10-60-10-0-0.2 10 60% 10 0 X X
10-60-10-700-0.2/0.3 10 60% 10 700 X X X X
10-80-2-0-0.2/0.3 10 80% 2 0 X X X
10-80-2-700-0.2 10 80% 2 700 X X X
10-80-10-0-0.1/0.2 10 80% 10 0 X X X X
10-80-10-700-0.2 10 80% 10 700 X X
20-70-6-3500.1/0.2/0.3 20 70% 6 350 X X X X X
30-60-2-0-0.2/0.3 30 60% 2 0 X X X
30-60-2-700-0.2 30 60% 2 700 X X X
30-60-10-0-0.1/0.2 30 60% 10 0 X X X X
30-60-10-700-0.2 30 60% 10 700 X X
30-80-2-0-0.2 30 80% 2 0 X X X
30-80-2-700-0.1/0.2 30 80% 2 700 X X X
30-80-10-0-0.2 30 80% 10 0 X X
30-80-10-700-0.2/0.3 30 80% 10 700 X X X X
3
Skogsberg K (2005).
The time steps were chosen by trial and error and increased gradually to reduce the
number of iterations. The maximum time step varied with settings. The snow
temperature Tsnow was set to 0oC, dry density at 100 kg m-3 and reference dry density at
91.5 kg m-3. The initial layer temperature was linear from 0oC at the bottom to air
temperature at the cutter shaving layer surface. Initially, the cutter shavings were in
hygroscopic equilibrium with air humidity.
The simulations continued until the temperature and moisture content were constant or
fluctuated between two values. The final surface melt rates were used in the multiple
regression analysis described below.
Moisture transfer
In a wood chip layer on snow, liquid water transport occurs by drain and capillary
transfer, and vapour transfer by diffusion and convection. Evaporation and condensation
links them together.
In the capillary layer liquid water is transferred both upwards and downwards. Vapour
is transferred downwards because of the vapour content gradient caused by the
temperature gradient. This problem is preferably handled by two independent state
variables, but such variables were not available. Capillary transfer, however, dominates
the moisture transfer in the capillary layer, being about 50-150 times greater than the
downward vapour diffusivity.
Capillary transfer
Capillary transfer from a cell starts when the moisture content exceeds the critical
moisture content (70 kg m-3), plus a threshold value of 1 kg m-3 to avoid instability in
the calculations. Upward capillary transfer occurred until the moisture content of the
above cell reached the maximum active capillary moisture content.
Based on performed experiments the maximum active capillary height was set to 0.13
m. An upward capillary transfer above the maximum active capillary height was
allowed if the moisture content in the cell above was larger than 71 kg m-3. This is only
possible by condensation, drain or initially wet wood chips. Downward capillary
transfer was not limited. When the moisture content exceeded the maximum passive
moisture content the surplus drained. The active and passive maximum moisture
contents were described by one linear and one exponential function (Skogsberg, 2005).
cap [kg m-2 s-1] was calculated by a commonly used moisture
Capillary transfer m
transfer model with a redefined (capillary) moisture diffusivity Dw,cap [m2 s-1] and a
linear dampening function R(z) [-], where z is the height to include the height
dependence of transfer and raise rates.
ǻw z R z ǻw
cap
m D w,cap R(z) D w,cap 0 d z d z cap (4)
ǻz zR ǻz
The capillary moisture diffusivity depends on the moisture content. It was established
by capillary experiments, an ordinary moisture diffusivity equation, and an accepted
4
Skogsberg K (2005).
capillary transfer function (Skogsberg, 2005). The coefficient zR was set to 0.22 m, to
enable capillary moisture transfer at the upper capillary level.
Hygroscopic moisture content
The relative humidity was set to 100% from the highest point of the capillary layer and
the dew point level, and downwards. The dew point level is the highest level where
absolute humidity exceeded the saturation vapour content of the cutter shavings. In the
hygroscopic layer the relative humidity was the vapour content ratio between free air
and air saturation, calculated from the cutter shaving temperature vsat(T) [kg m-3] DIN
4108 (1981),
n 8.02
§ T · § T ·
M water a ¨ b air ¸ 18.02 288.68 ¨1.098 air ¸
© 100 ¹ © 100 ¹
v sat (T) (5)
R 273.15 Tair 8314.3 273.15 Tair
Equation 5 is commonly used for air temperatures between 0 and 30oC, where Mwater is
the molar weight of water [kg kmol-1], R is the universal gas constant [J kmol-1 K-1], and
a and b are empirical coefficients. The hygroscopic moisture content was calculated
from the sorption isotherm in Figure 1.
Convective moisture transfer – Evaporation and condensation

Vapour diffusion and convection in the cutter shaving layer were not included due to
limitations of the program; hence, air transfer was implicitly given. A simplified
evaporation/condensation (e/c) function was however included where rate and direction
depended on depth, air velocity and permeability (Skogsberg, 2005).
There was no e/c below the capillary front, the highest level with at least critical
moisture content. When the dew point level was equal to or above the capillary front,
condensation down to the capillary front was found. When the dew point level was
below the capillary front there was evaporation at the capillary front.
Evaporation occurred when the moisture content above the capillary front exceeded the
hygroscopic equilibrium moisture content plus the threshold value of 1 kg m-3.
Evaporation and condensation above the capillary front were limited to 10% of
Equation 6 to avoid instability. The e/c rate was
§ U ·
conv
m ȕ´ v air vsat T ¨ ȕconv dry Eslope Z z ¸ vair v sat T (6)
¨ U ¸
© dry , ref ¹
E´ is a modified convective moisture transfer coefficient [m s-1], vair is the air vapour
content [kg m-3], Econv is the ordinary convective moisture transfer coefficient by a
surface, Udry and Udry,ref are the current and reference dry densities of the cutter shaving
layer [kg m-3], and Eslope is an assumed coefficient that influences how fast the mass
convection rate decreases with a depth [m s-1], here Eslope=0.0523+0.016uair. E´ has the
same principal look as D´, see Figure 2 below. It is related to convective heat transfer by
the Lewis formula approximation
5
Skogsberg K (2005).
Į conv
ȕ conv (7)
ȡ air c p,air
where Dconv is the convective heat transfer coefficient [W m-2 K-1] described below, Uair
is the air density [kg m-3], and cp,air is the heat capacity of air at a constant pressure [J
kg-1 K-1].
Heat transfer
Convective heat transfer

The fluid velocity is zero at the surface of an impermeable material. The velocity at the
interface between a porous material and a fluid depends on permeability, surface
roughness and fluid velocity outside the boundary layer. The interface velocity for a
material with high permeability will be closer to fluid velocity outside the boundary
layer. The air flow comes from the pressure gradient and changes gradually with depth
into the permeable layer, from highly turbulent at the surface to laminar at some depth
(Prinos et al, 2003). In performed simulations the convective heat transfer Pconv [W m-2]
at the upper cutter shaving layer interface was calculated according to Skogsberg
(2005),
Tsurf is the cutter shavings temperature at interface [oC] and Dconv the convective heat
transfer coefficient [W m-2 K-1], determined from an empirical expression for air
velocity parallel to external surfaces (Hagentoft, 2001).
D conv ® (9)
°̄7.41 u air 0.78 u air t 5 m s 1
Internal heat convection was included as an apparent thermal conductivity. The actual
thermal conductivity was multiplied by a heat convection factor D´, which depended on
air velocity, depth and permeability, Equation 10. Permeability was implicitly included
by dry density. Heat convection inside the layer was therefore always positive. This is
reasonable since the air temperature will exceed cutter shaving temperature close to the
wood chip surface.
ª1 § U dry ·º
Pconv Pcond D' Pcond « ¨ D conv D slope Z z ¸» (10)
¨
«¬ K © U dry, ref ¸»
¹¼
where Dslope is an estimated factor that decides how the convective heat transfer
decreases with depth [W m-3 K-1], here Dslope=73+19u. The factor K was used to scale
down the convective heat transfer to a multiple of the thermal conduction, here K=8 m
s-1.
Laboratory snow melt measurements were made on snow thermally insulated by 0.025-
0.0.75 m cutter shaving layers (Skogsberg and Lundberg, 2005). These showed that the
convective net heat transfer was from air to snow, also when air temperature was lower
than the wood chip surface temperature. The convective heat transfer was 0.2-5.8 times
6
Skogsberg K (2005).
greater than the conductive heat transfer. It was assumed that the convective heat
transfer was from both vertical and horizontal air exchanges. Figure 2 shows D´ for
different air velocities.
6
Alfa´ (u=2)
5
Alfa´ (u=6)
4 Alfa´ (u=10)
Alfa`[-]
0
0,00 0,05 0,10 0,15 0,20
height [m]
Figure 2 Suggested heat convection factor D´ (Alfa´) for different air velocities u. The convective
heat transfer [W m-2] is achieved by multiplying the convection factor with conductive
heat transfer.
Thermal conduction
The conductive heat transfer Pcond [W m-2] was calculated by the Fourier’s law,
dT
Pcond Ȝ cs (11)
dz
where Ocs is the thermal conductivity of the cutter shavings [W m-1 K-1], and dT/dx is
the temperature gradient [K m-1]. The thermal conductivity of wood chips depends on
wood quality, chips structure, compaction, temperature and moisture content (Mörtstedt
and Hellsten, 1999). The thermal conductivity used was measured for cutter shavings
and linearly increased with moisture content (Skogsberg, 2004).
Ȝ cs 0.1 0.0007 w (12)
For the transient case in 1D the temperature distribution is found by

wT w § wT ·
c ¨Ȝ ¸ (13)
wt wz © wz ¹
The volumetric heat capacity of a moist material c [ J m-3 K-1] is
c U dry c p,dry w c p, wat (14)
where cp is heat capacity at constant pressure of dry material and water [J kg-1 K-1].
Radiation
The radiation exchange at the cutter shaving surface consists of both short wave (sw)
and long wave (lw) radiation. The absorbed solar radiation (sw) is
Psolar ,abs Psolar 1 rsw (15)
where Psolar,abs is the absorbed part of incident solar radiation [W m-2] and rsw is solar
radiation reflectivity (albedo) [-]. Here, 0.2 was used according to the author’s own
7
Skogsberg K (2005).
experiments (Skogsberg and Lundberg, 2005). Long wave radiation exchange between a
cutter shaving layer surface and the sky, assuming grey surfaces and Asky>>Awc, is
4 4
P12 V H wc Tsky , K Tsurf , K . (16)
V is the Stefan-Boltzmann constant 5.6710-8 W m-2 K-4, TK is the temperatures in [K]

and Hwc is the surface emissivity of wood chips [-]. The emissivity of wood varies
between 0.82-0.92 (Kreith and Kreider, 1978), a value of 0.87 was used in the
calculations. The sky temperature was estimated as the average between cloudy and
clear conditions, where the cloudy sky temperature was set equal to the air temperature
and clear sky temperature as in Equation 17 (Hagentoft, 2001).
Tsky, K 273.15 1.2 Tair 14 (17)
Thermal resistance and moisture transfer at cutter shavings-snow interface

Heat and moisture transfer between snow and cutter shavings are greatly unknown, but
an upwards moisture transfer that increases with an increased melt rate is reasonable.
The surface melt rate is thus highly dependent on the contact between the cutter
shavings and snow, i.e. thermal resistance. By assuming a thermal conductivity
somewhere between 0.2-0.4 W m-1 K-1 and a distance between the cutter shavings and
snow of 0.001-0.005 m, the thermal resistance is 0.0025-0.025 m2 K W-1. A thermal
resistance of 0.01 m2 K W-1 was used in the simulations.
Based on measured evaporation rates the moisture inflow to the bottom cell was
assumed as an absorbed melt share (AMS) of the surface melt rate, chosen as 10%.
Sensitivity of numerical model

To investigate model sensitivity a number of simulations with alternative variable
settings, one at the time, were made, Table 2. Another capillary moisture diffusivity
(Dw,cap) function was also tested with a critical moisture content of 50 kg m-3 instead of
70 kg m-3 and minimum Dw,cap of 110-9 m2 s-1 instead of 2.710-10 m2 s-1, increasing the
capillary moisture diffusivity for all moisture contents.
Table 2 The sensitivity tested factors were maximum active capillary height zcap, absorbed melt
share ratio AMS, dry density Udry, solar reflectivity (albedo) of cutter shavings rsolar, long
wave emissivity Hlw, alternative convection depth dconv and thermal resistance between
cutter shavings and snow Rsnow.
zcap AMS Udry rsolar Hlw dconv Rsnow
[m] [%] [kg m-3] [%] [-] [m] [m2 K W-1]
Basic 0.13 10% 100 20% 0.87 0.15 0.01
Alternative 0.15 5% 90 30% 0.92 0.04 0.001
Alternative 20% 120 0.1
Statistical Analysis of Surface Melt Rates

A multiple regression analysis was made with air temperature, relative humidity, air
velocity, solar radiation, cutter shavings layer thickness and all their two-way
interactions as independent variables, and the surface melt rate SMR as a dependent
variable, defined as the melt rate of snow directly under the cutter shaving layer. An
interaction effect is when the effect of one factor depends on the level of some other
8
Skogsberg K (2005).
factor(s) (Box et al, 1978). In each step the variable with the highest P-value was
rejected from the model until all remaining variables had P-values below 0.05, i.e.
significant at 95% confidence level.
Nine of seventeen simulations were conducted with both initially wet and initially dry
cutter shavings. The rainy day share from 2000-2004 during April 15th-August 31st was
46%, 27%, 49%, 47% and 54%, i.e. there were rainfalls about every second day.
Therefore, the probability of approaching stationary conditions from drier and wetter
cutter shavings is equal. In the regression analysis the mean SMR of initially dry and
wet runs were used.
Results and discussion

SMR was 0.051-1.507 kg m-2 h-1 (4.7-139.6 W m-2), averaging 0.501 kg m-2 h-1 (46.4 W
m-2). Simulation 30-60-10-0-0.1 had the largest SMR while simulation 10-80-2-0-0.3
had the smallest SMR (see Table 1). The effective thermal conductivity using air
temperature, snow temperature 0oC and SMR was 0.129-0.654 W m-1 K-1. The
influence of different factors on SMR is treated in the “Multiple regression” section
below.
ER as mass varied from -0.074 kg m-2 h-1 (evaporation) to 0.023 kg m-2 h-1
(condensation), or -51.4 to 15.7 W m-2 as energy. ER averaged -0.020 kg m-2 h-1 (-14.2
W m-2). The ratio of ER over SMR was -10.1% to 4.2% as mass and -75.5% to 31.7%
as energy, i.e. a significant part of the energy balance. The relative importance of
evaporation increased with a thinner layer of wood chips. Simulation results are seen in
Appendix 1.
With initially wet cutter shavings, the SMR decreased in six of nine simulations by
0.9%, 1.2%, 6.4%, 6.4%, 17.5% and 18.5%, and was unaffected in the others. The two
largest decreases were due to evaporation instead of condensation, where evaporation
occurred at a higher level than condensation. In three of the other four cases there was
greater evaporation with initially wet cutter shavings, in two cases at a higher level. The
evaporative cooling effect was thus relatively greater than the increased thermal
conductivity caused by greater moisture content. In one simulation with a 6.4% SMR
decrease, a lower condensation rate at a higher level with initially wet cutter shavings
was seen, due to a decreased vapour content difference. The modified convective
moisture transfer coefficient convection E´ here increased by 19% with initially wet
cutter shavings, while the vapour content difference decreased 90%.
SMR increased 38.6%-104.7% in the five simulations with a cutter shaving layer
thickness of 0.1 m instead of 0.2 m. With a 0.3 m cutter shaving layer SMR decreased
19.3%-68.2%.
The capillary layer thickness was the same or increased with initially wet cutter
shavings, and was 0.05-0.19 m with Z=0.2 m. With a 0.1 m cutter shaving layer the
capillary layer thickness was 0.05-0.09 m, and with a 0.3 m cutter shaving layer it was
0.12-0.15 m.
The SMR of Skogsberg and Lundberg’s (2005) laboratory experiment (0.490-2.380 kg
m-2 h-1) was greater than simulated SMR because of thinner cutter shaving layers
9
Skogsberg K (2005).
(0.025-0.075 m). The effective thermal conductivity was 0.136-0.612 W m-1 K-1, i.e.
quite similar to present simulations. ER was of the same range, 0.017-0.118 kg m-2 h-1.
Sensitivity analysis
The effect on SMR from investigated sensitivity variables is seen in Table 3. The
decrease of convection penetration from 0.15 to 0.04 m depth when SMR decreased
30.5% was the greatest influence, since less amounts of warm air were transferred into
the cutter shaving layer. As well, ER decreased to zero.
By increasing the thermal resistance between snow and cutter shavings from 0.01 to 0.1
m2 K W-1, SMR decreased 22.0%, the capillary layer decreased slightly and ER
decreased 1.5%. By decreasing the absorbed melt share ratio from 10% to 5%, the SMR
decreased 13.7% despite a 31.6% decrease in ER. This was due to decreased thermal
conduction caused by a thinner capillary layer and lower moisture contents. With a dry
density 120 kg m-3, SMR increased 5.8% due to an increased thermal conductivity. ER
also increased because the capillary transfer rate increased. Other SMR changes were
small.
The alternative capillary moisture diffusivity function decreased SMR 1.1% because of
a 19.3% increase in ER. This was caused by increased capillary transfer rate due to
increased moisture content at 0.13 m.
Table 3 Sensitivity of SMR due to changes of different variables in the model.

zcap AMS Udry rsolar Hlw dconv Rsnow Alternative Dw,cap
[m] [%] [kg m-3] [%] [-] [m] [m2 K W-1] [m2 K W-1]
Basic 0.13 10% 100 20% 0.87 0.15 0.01
Alternative 1 0.15 5% 90 30% 0.92 0.04 0.001
SMR-change 0.1% -13.7% -1.5% -1.1% 0.1% -30.5% 2.5% -1.1%
ER-change -0.2% -31.6% -31.7% -16.5% -1.3% -100.0% -5.5% 19.3%
Alternative 2 20% 120 0.1
SMR-change 0.1% 5.8% -22.0%
ER-change -0.2% 4.0% -1.5%
Multiple regression
The significant factors, with P-values in parenthesis, were solar radiation (0.0091), air
temperature (0.0000), Tairuair (0.0000) and TairZ (0.0000). The model P-value was
0.0000, Equation 18.
SMR=-0.09+0.00014Psolar+0.0575Tair+0.0012Tairuair-0.18TairZ (18)
This means that increased solar radiation, air temperature and air velocity increased
SMR, while an increased layer thickness decreased SMR. The effect of an increased air
temperature in the fifth term is less than the combined effect of the third and fourth
term, where air temperature is also included. The SMR of Equation 18 diverged -0.144
to 0.126 kg m-2 h-1 from the numerical modelling results, averaging 0.004 kg m-2 h-1,
Figure 3.
The results are mostly applicable for cutter shavings as thermal insulation since the
thermal conductivity and capillary transfer functions in the simulations were for cutter
10
Skogsberg K (2005).
shavings. It is thus uncertain how these results are adapted to other wood chip materials.
However, the principal results should be the same.
1,6
1,4 SMR,numerical modelling

SMR,calculated
1,2
-1]
SMR [kg m h
1,0
-2
0,8
0,6
0,4
0,2
0,0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27
Figure 3 Surface melt rates SMR from numerical modelling and Equation 18.
Conclusions
The surface melt rate (SMR) of snow thermally insulated with 0.1-0.3 m cutter shaving
layers was 0.051-1.507 kg m-2 h-1. SMR increased with increased solar radiation (Psolar),
air temperature (Tair) and air velocity (uair), and decreased with an increased layer
thickness (Z). The main effects at a 95% significance level were Tair and Psolar. The
significant two-way interactions were Tairuair and TairZ.
With initially wet cutter shavings, stationary SMR was unaffected or decreased (0%-
18.5%) relative to corresponding cases with initially dry cutter shavings due to
increased evaporation rates, decreased condensation rates, or evaporation instead of
condensation. The moisture convection rate and direction (evaporation/condensation) is
thus important for SMR.
The sensitivity analysis showed that turbulence penetration depth, thermal resistance
between shavings and snow, absorbed melt share and dry density affected SMR the
most.
Further work
The most important factors to include in further studies are air penetration in the cutter
shaving layer, more detailed moisture diffusivity measurements at different
temperatures, passive moisture content, and absorption of melt water. It would also be
useful to compare cutter shavings and sawdust more thoroughly.
Acknowledgements
I hereby thank Professor Bo Nordell for help and encourage, Professor Angela
Lundberg and Professor Anders Sellgren for fruitful discussions, Stenvalls trä AB in
Luleå for wood chip sponsoring, Luleå University of Technology, and The Swedish
Energy Agency for financing this project.
11
Skogsberg K (2005).
References
Ahlgren L (1972). Moisture fixation in porous building materials. Report 36, Division
of Building Technology, The Lund Institute of Technology. (In Swedish. Pictures
and results in English).
Box G E P, Hunter W G, and Hunter J S (1978). Statistics for experimenters. John
Wiley and Sons Inc., Canada. ISBN 0-471-09315-7.
DIN 4108 (1981). Thermal insulation in buildings: DIN 4108, part 1-5. Deutches
Institut für Normung, Berlin, 1981.
91-44-01896-7.
Hedenblad G (1996). Materialdata för fukttransportberäkningar. T19:1996,
Byggforsknings-rådet, Stockholm, Sweden.
Kreith F and Kreider J F (1978). Principles of Solar Energy. Hemisphere Publishing,
New York. ISBN 0-89116-678-5
Mörtstedt S-E and Hellsten G (1999). Data and diagram. Tables for Energy and
Technical Chemistry. ISBN 91-47-00805-9. Liber AB, Stockholm, Sweden.
Prinos P, Sofialidis D and Keramaris E (2003). Turbulent flow over and within a porous
bed. Journal of Hydraulic Engineering, ASCE, September 2003.
Skaar C (1988). Wood-Water Relations. Springer-Verlag Series in Wood Sciences.
Springer-Verlag, New York, USA. ISBN 0-387-19258-1.
Skogsberg K and Lundberg A (2005). Wood chips as thermal insulation on snow.
Accepted for publication, Cold Regions Science and Technology, Elsevier, 05-04-
25.
Skogsberg K and Nordell B (2001). Sundsvall hospital snow storage. Cold Regions
Science and Technology 32 (2001) 63-70, Elsevier.
Skogsberg K (2004). Thermal conductivity of wet wood chips- A laboratory study.
Technical report. Division of Renewable Energy, Luleå University of Technology.
Skogsberg K (2005). Seasonal Snow Storage for Space and Process Cooling. Doctoral
thesis 2005:30, Division of Architecture and Infrastructure, Department of Civil
and Environmental Engineering, Luleå University of Technology.
Stamm A J (1964). Wood and cellulose science. The Ronald Press Company, New
York.
12
Skogsberg K (2005).
Appendix 1
Surface Melt Rates from numerical modelling
Initially dry
T,air [oC] 10 10 10 10 10 10 10 10 20 30 30 30 30 30 30 30 30
RH [%] 0,6 0,6 0,6 0,6 0,8 0,8 0,8 0,8 0,7 0,6 0,6 0,6 0,6 0,8 0,8 0,8 0,8
u,air [m s-1] 2 2 10 10 2 2 10 10 6 2 2 10 10 2 2 10 10
P,rad,sw 0 700 0 700 0 700 0 700 350 0 700 0 700 0 700 0 700
Z [m] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2
T0 0,1 0,3 0,1 0,2 0,1 0,3 0,2 0,3 0,4 0,5 0,6 0,8 0,8 0,6 0,6 1,0 0,9
T1 0,4 1,3 0,7 1,2 0,6 1,4 0,9 1,4 1,8 1,7 2,4 3,0 3,5 2,1 2,6 3,4 4,0
T2 0,7 2,4 1,2 2,1 1,0 2,6 1,6 2,6 3,3 3,1 4,3 5,3 6,3 3,6 4,7 6,1 7,1
T3 1,0 3,4 1,8 3,1 1,5 3,8 2,4 3,8 4,8 4,5 6,2 7,7 9,1 5,3 6,8 8,9 10,3
T4 1,3 4,5 2,3 3,9 1,9 5,0 3,1 4,9 6,3 6,0 8,2 10,0 11,7 7,1 9,0 11,6 13,2
T5 1,7 5,6 2,7 4,8 2,4 6,2 3,6 5,7 7,7 7,6 10,2 11,9 13,9 9,0 11,2 13,8 15,6
T6 2,0 6,7 3,1 6,4 2,9 7,4 4,1 6,4 9,0 9,2 12,3 13,5 15,6 10,9 13,5 15,7 17,6
T7 2,3 7,8 3,6 8,2 3,4 8,6 4,5 7,0 10,2 10,8 14,4 14,9 17,1 12,8 15,9 17,3 19,4
T8 2,6 8,9 4,2 9,7 3,9 9,9 4,8 7,7 11,1 12,5 16,6 16,1 18,5 14,8 18,3 18,7 20,9
T9 3,0 10,3 4,8 11,0 4,3 11,1 5,1 8,8 12,0 14,0 18,7 17,2 19,7 16,6 20,6 19,9 22,3
T10 3,3 13,2 5,3 12,2 4,7 12,8 5,4 10,2 12,9 15,4 20,5 18,1 20,9 18,2 22,6 21,0 23,5
T11 3,6 16,3 5,9 13,3 5,1 15,5 5,7 11,5 13,6 16,7 22,7 19,0 22,1 19,7 24,5 22,1 24,7
T12 4,2 19,0 6,3 14,2 5,4 18,4 6,0 12,7 14,3 17,8 26,5 19,8 24,3 21,1 26,3 23,0 25,8
T13 4,8 21,6 6,8 15,1 5,7 21,0 6,2 13,8 15,3 18,9 30,6 20,6 26,5 22,4 28,7 23,9 27,0
T14 5,3 24,0 7,2 16,0 6,0 23,5 6,4 14,8 16,8 19,9 34,4 21,5 28,5 23,6 32,2 24,8 28,8
T15 5,8 26,2 7,5 16,8 6,3 25,8 6,6 15,8 18,4 21,0 38,0 22,8 30,5 24,7 36,1 25,6 30,6
T16 6,3 28,3 7,9 17,5 6,5 28,0 6,9 16,7 20,0 22,3 41,3 24,0 32,3 25,2 39,7 25,9 32,4
T17 6,8 30,3 8,2 18,2 6,8 30,0 7,3 17,5 21,5 23,6 44,5 25,1 34,0 25,7 43,2 26,5 34,1
T18 7,2 32,1 8,5 18,8 7,2 31,9 7,8 18,3 22,9 24,8 47,5 26,2 35,7 26,3 46,5 27,2 35,7
T19 7,6 33,9 8,8 19,4 7,6 33,7 8,3 19,1 24,2 26,0 50,3 27,2 37,2 27,0 49,6 27,9 37,2
T20 8,0 35,5 9,1 20,0 8,0 35,4 8,8 19,8 25,5 27,2 53,0 28,2 38,7 27,7 52,6 28,6 38,7
w0 152,3 206,9 150,1 197,3 307,1 192,6 307,1 203,4 307,1 450,0 307,1 450,0 307,1 450,0 307,2 450,0 307,1
w1 150,4 205,5 146,5 195,7 296,5 190,3 296,6 201,5 296,6 408,3 296,7 408,3 296,5 408,3 298,1 408,3 297,2
w2 148,3 204,0 142,2 193,9 285,9 187,8 286,3 199,2 285,9 362,9 286,3 362,9 286,5 362,9 287,1 362,9 286,0
w3 146,0 202,3 136,8 192,0 275,3 185,0 275,4 196,7 275,6 336,4 275,7 336,4 275,6 336,4 275,4 336,4 275,4
w4 143,2 200,4 129,4 189,8 264,7 181,6 265,0 193,7 264,8 317,6 264,8 317,5 264,9 317,5 264,8 317,5 265,1
w5 140,0 198,3 117,9 71,0 254,2 177,7 254,4 190,0 254,3 302,9 254,3 302,9 254,3 302,9 254,2 302,9 254,3
w6 136,0 196,0 99,2 15,7 243,6 173,0 243,7 185,4 243,7 291,0 243,6 291,0 243,6 291,0 243,9 291,0 243,6
w7 130,9 193,3 70,0 12,9 239,3 167,1 233,0 179,0 233,1 280,9 233,1 280,9 233,1 280,9 233,0 280,9 233,0
w8 124,1 190,1 19,9 11,9 239,3 159,2 222,4 77,5 222,4 272,2 222,4 272,2 222,4 272,2 222,5 272,2 222,4
w9 113,9 70,0 18,0 11,2 239,3 147,2 211,8 50,8 211,8 264,5 211,8 264,5 211,8 264,5 211,9 264,5 211,8
w10 101,6 10,1 16,7 10,6 239,3 70,0 201,2 16,4 201,2 257,6 201,2 257,6 201,2 257,6 201,3 257,6 201,2
w11 70,0 8,9 15,7 10,1 239,3 12,1 190,6 14,5 190,6 251,3 70,0 251,3 70,0 251,3 190,6 251,3 190,6
w12 19,8 7,7 14,9 9,7 239,3 10,9 180,0 13,3 180,0 245,7 14,0 245,6 17,3 245,6 180,0 245,6 180,0
w13 18,1 6,7 14,3 9,3 239,3 9,8 169,4 12,4 70,0 240,4 11,4 240,4 14,0 240,4 70,0 240,4 70,0
w14 16,8 5,8 13,9 8,9 235,5 8,8 158,8 11,7 18,2 235,6 9,9 25,9 12,4 235,6 13,7 235,5 18,9
w15 15,7 5,1 13,5 8,6 166,9 7,9 144,9 11,2 15,1 29,0 8,6 20,9 11,4 231,0 11,4 231,0 15,3
w16 14,9 4,5 13,1 8,3 166,9 7,1 110,0 10,9 13,4 22,6 7,4 17,7 10,6 226,8 10,0 226,8 13,4
w17 14,3 4,0 12,8 8,0 28,1 6,4 25,8 10,7 12,3 18,6 6,3 15,7 10,0 222,8 8,7 27,4 12,3
w18 13,8 3,5 12,6 7,7 25,6 5,9 23,5 10,5 11,5 16,1 5,5 14,3 9,5 27,7 7,6 24,2 11,5
w19 13,4 3,2 12,4 7,5 23,6 5,4 21,6 10,3 10,9 14,5 5,0 13,3 9,2 24,6 6,7 21,5 11,2
w20 13,0 2,8 12,2 7,2 22,0 4,9 20,0 10,1 10,4 13,4 4,8 12,6 8,9 21,9 6,2 19,3 10,9
SMR [kg m-2 h-1] 0,070 0,279 0,119 0,241 0,145 0,292 0,285 0,302 0,471 0,536 0,615 0,915 0,902 0,632 0,678 1,059 1,021
ER [kg m-2 h-1] -0,007 -0,028 -0,012 -0,024 0,000 -0,029 0,004 -0,030 -0,027 0,023 -0,025 0,023 -0,023 0,009 -0,028 0,002 -0,006
O-eff [W m-1 K-1] 0,129 0,516 0,220 0,447 0,269 0,542 0,528 0,560 0,437 0,331 0,380 0,565 0,557 0,391 0,419 0,654 0,630
13
Skogsberg K (2005).
Initially wet
T,air [oC] 10 10 10 10 20 30 30 30 30
RH [%] 0,6 0,6 0,8 0,8 0,7 0,6 0,6 0,8 0,8
u,air [m s-1] 2 10 2 10 6 2 10 2 10
P,rad,sw 0 700 700 0 350 700 0 0 700
Z [m] 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2 0.2
T0 0,1 0,2 0,3 0,2 0,4 0,6 0,7 0,5 0,9
T1 0,4 1,2 1,4 0,9 1,8 2,4 2,9 1,9 3,7
T2 0,7 2,1 2,6 1,6 3,2 4,2 5,2 3,4 6,7
T3 1,0 3,1 3,8 2,4 4,7 6,1 7,5 5,0 9,7
T4 1,3 3,9 5,0 3,1 6,2 8,1 9,7 6,7 12,4
T5 1,7 4,8 6,2 3,6 7,7 10,1 11,4 8,4 14,7
T6 2,0 6,4 7,4 4,1 9,0 12,1 12,9 10,2 16,6
T7 2,3 8,2 8,6 4,5 10,1 14,3 14,2 12,0 18,2
T8 2,7 9,7 9,9 4,8 11,0 16,4 15,3 13,8 19,7
T9 3,0 11,0 11,1 5,1 11,9 18,4 16,3 15,5 21,0
T10 3,3 12,2 12,9 5,4 12,7 20,3 17,2 17,1 22,1
T11 3,6 13,3 15,5 5,7 13,5 22,0 18,1 18,5 23,3
T12 4,2 14,2 18,4 6,0 14,2 24,4 18,9 19,8 24,3
T13 4,8 15,2 21,1 6,2 14,9 28,2 19,6 21,0 25,3
T14 5,3 16,0 23,5 6,4 16,3 32,3 20,4 22,3 26,4
T15 5,8 16,8 25,8 6,6 18,0 36,2 21,2 23,5 28,2
T16 6,3 17,5 28,0 6,9 19,6 39,9 22,4 24,1 30,4
T17 6,8 18,2 30,0 7,3 21,2 43,3 23,8 24,7 32,5
T18 7,2 18,8 31,9 7,8 22,7 46,6 25,2 25,3 34,5
T19 7,6 19,4 33,7 8,3 24,1 49,7 26,5 25,9 36,4
T20 8,0 20,0 35,4 8,8 25,4 52,7 27,8 26,8 38,2
w0 152,3 197,7 192,5 305,5 307,1 307,2 307,1 450,0 307,1
w1 150,5 196,1 190,2 296,9 296,6 298,1 296,5 408,3 297,0
w2 148,4 194,3 187,7 286,4 285,9 285,1 285,9 362,9 286,1
w3 146,0 192,4 184,8 275,6 275,4 275,9 275,5 336,4 275,6
w4 143,3 190,2 181,5 265,1 264,8 265,4 264,9 317,5 264,9
w5 140,0 71,0 177,6 254,4 254,4 254,6 254,2 302,9 254,3
w6 136,1 15,7 172,9 243,7 243,7 243,9 243,6 291,0 243,6
w7 131,0 13,3 166,9 233,0 232,9 233,1 233,0 280,9 233,0
w8 124,2 12,5 159,0 222,4 222,4 222,4 222,4 272,2 222,4
w9 114,0 11,8 146,8 211,8 211,9 211,8 211,8 264,5 211,8
w10 101,7 11,3 70,0 201,2 201,3 201,2 201,2 257,6 201,2
w11 70,0 10,9 12,1 190,6 190,6 190,6 190,6 251,3 190,6
w12 20,7 10,5 10,9 180,0 180,0 70,0 180,0 245,6 180,0
w13 19,0 10,1 9,8 169,4 169,4 13,3 169,4 240,4 163,4
w14 17,7 9,8 8,8 158,8 20,3 11,5 156,0 235,5 110,0
w15 16,7 9,5 7,9 144,9 16,6 10,1 110,0 231,0 20,9
w16 15,9 9,2 7,1 110,0 14,6 8,8 23,2 226,8 16,5
w17 15,3 8,9 6,4 25,8 13,4 7,6 19,0 222,8 14,3
w18 14,8 8,6 5,9 23,5 12,6 6,6 16,5 219,1 13,0
w19 14,4 8,4 5,4 21,6 12,0 5,8 14,9 215,6 12,2
w20 14,0 8,1 4,9 20,0 11,2 5,1 13,8 25,8 11,5
SMR [kg m-2 h-1] 0,070 0,242 0,292 0,235 0,467 0,607 0,745 0,592 0,956
ER [kg m-2 h-1] -0,007 -0,024 -0,029 -0,018 -0,033 -0,036 -0,028 0,002 -0,051
O-eff [W m-1 K-1] 0,129 0,449 0,542 0,435 0,433 0,375 0,460 0,366 0,590
14
Skogsberg K (2005).
Alternative heights
T,air [oC] 10 10 10 10 20 20 30 30 30 30
RH [%] 0,6 0,8 0,6 0,8 0,7 0,7 0,6 0,8 0,6 0,8
u,air [m s-1] 2 10 10 2 6 6 10 2 2 10
P,rad,sw 700 0 700 0 350 350 0 700 0 700
Z [m] 0.1 0.1 0.3 0.3 0.1 0.3 0.1 0.1 0.3 0.3
T0 0,5 0,4 0,2 0,0 0,9 0,2 1,4 1,2 0,2 0,5
T1 1,6 1,1 1,1 0,3 2,3 1,1 3,6 3,2 0,9 2,4
T2 2,6 1,9 2,0 0,5 3,7 2,0 5,9 5,2 1,7 4,5
T3 3,7 2,7 2,9 0,8 5,1 2,9 8,2 7,3 2,5 6,8
T4 4,7 3,4 3,9 1,0 6,6 3,9 10,3 9,4 3,4 9,1
T5 5,8 3,9 4,9 1,3 8,0 5,0 12,0 11,5 4,3 11,6
T6 6,8 4,4 6,0 1,6 9,2 6,0 13,3 13,6 5,2 14,1
T7 7,9 4,8 7,1 1,9 10,2 7,2 14,5 15,7 6,2 16,7
T8 8,9 5,1 8,3 2,2 11,1 8,4 15,5 17,9 7,2 19,4
T9 9,9 5,4 10,1 2,6 11,9 10,1 16,4 19,9 8,7 22,1
T10 11,0 5,7 12,3 3,2 12,7 12,4 17,3 21,7 10,7 24,7
T11 13,5 5,9 14,1 3,8 13,3 14,6 18,0 23,3 12,7 27,0
T12 16,3 6,1 15,4 4,4 13,9 16,9 18,7 24,8 14,6 29,5
T13 19,0 6,3 16,4 5,0 14,5 19,3 19,4 26,2 16,6 31,6
T14 21,4 6,5 17,2 5,6 15,0 21,4 20,0 28,0 18,5 33,3
T15 23,7 6,7 17,9 6,2 15,8 22,8 20,5 31,4 20,5 34,8
T16 25,8 6,9 18,6 6,7 17,2 23,9 21,1 35,2 22,5 36,0
T17 27,8 7,1 19,1 7,3 18,9 24,8 21,7 38,7 24,5 37,2
T18 29,7 7,3 19,6 7,8 20,4 25,5 23,1 42,1 26,1 38,2
T19 31,4 7,7 20,0 8,1 21,9 26,2 24,5 45,3 27,2 39,1
T20 33,1 8,2 20,4 8,3 23,3 26,8 25,8 48,3 28,1 39,9
w0 221,3 198,6 307,1 307,1 307,1 307,1 307,1 307,2 307,1 450,0
w1 220,4 197,4 291,2 291,6 302,1 291,2 302,2 302,4 291,2 381,7
w2 219,4 196,0 275,5 276,8 297,0 275,3 296,9 296,9 275,3 336,4
w3 218,3 194,6 259,5 259,4 291,5 259,5 291,6 291,3 259,5 309,8
w4 217,2 193,0 243,6 243,6 285,9 243,6 285,9 286,0 243,6 291,0
w5 216,1 191,3 227,7 227,7 280,4 227,7 280,7 281,0 227,7 276,4
w6 214,9 189,5 211,8 211,8 275,6 211,8 275,5 275,4 211,8 264,5
w7 213,6 187,4 195,9 196,0 270,2 196,0 270,1 270,1 195,9 254,4
w8 212,2 185,2 180,0 180,1 264,8 180,0 264,9 264,8 180,0 245,6
w9 210,7 182,7 18,7 29,5 259,4 29,5 259,6 259,5 29,5 237,9
w10 71,0 179,8 10,5 29,5 254,3 29,5 254,2 254,2 29,5 231,0
w11 11,0 176,6 9,8 29,5 249,0 29,5 248,9 248,9 29,5 24,2
w12 8,8 172,8 9,4 29,5 243,6 29,5 243,6 243,6 29,5 17,0
w13 7,7 168,4 9,1 29,5 238,3 14,1 238,3 238,4 29,5 14,4
w14 6,7 163,0 8,9 29,5 233,0 12,5 233,0 70,0 29,5 13,3
w15 5,9 156,1 8,7 29,5 70,0 12,0 227,7 14,3 29,5 12,6
w16 5,2 146,7 8,5 28,1 17,2 11,7 222,4 11,8 21,8 12,1
w17 4,8 130,9 8,4 24,9 14,5 11,4 70,0 10,3 16,7 11,8
w18 4,5 70,0 8,2 22,8 13,0 11,2 20,1 9,1 14,4 11,4
w19 4,2 23,1 8,1 21,4 12,0 11,0 16,8 8,0 13,3 11,1
w20 4,0 21,3 8,0 20,5 11,3 10,8 14,8 7,0 12,7 10,8
SMR [kg m-2 h-1] 0,570 0,395 0,195 0,051 0,946 0,197 1,507 1,344 0,170 0,532
ER [kg m-2 h-1] -0,057 -0,039 0,000 0,000 -0,074 0,000 -0,070 -0,065 0,000 0,015
O-eff [W m-1 K-1] 0,528 0,366 0,541 0,142 0,438 0,274 0,465 0,415 0,158 0,493
15
Report I
Skogsberg K (2005). Capillary suction in cutter shavings and sawdust.

Experiments and a model approach. Technical report. Division of
Architecture and Infrastructure, Department of Civil and
Capillary suction in cutter shavings and sawdust
Experiments and a model approach
Kjell Skogsberg
Luleå University of Technology
kjell.skogsberg@ltu.se, phone: +46(0)920/491 496, fax: +46(0)920/491 697
Skogsberg K. Capillary suction in cutter shavings and sawdust- Experiments and a model approach
Content
Abstract ......................................................................................................................................................- 1 -
Aim and limitations ....................................................................................................................................- 1 -
Moisture in air and porous materials ..........................................................................................................- 2 -
Hygroscopic moisture -3-
Liquid moisture -5-
Moisture transfer ........................................................................................................................................- 5 -
Diffusion -5-
Mass convection -6-
Darcy flow -6-
Suction -7-
Non-isothermal moisture transfer - 11 -
Experiments..............................................................................................................................................- 11 -
XP1; Capillary suction from a free water surface - 11 -
XP2; Maximum hygroscopic moisture ratio - 12 -
XP3; Moisture content change of initially wet cutter shavings during capillary conditions - 12 -
XP4; Moisture content with capillary conditions and simulated condensation - 12 -
Results ......................................................................................................................................................- 13 -
XP2-4 - 16 -
Discussion ................................................................................................................................................- 18 -
XP2-4 - 19 -
Suggested capillary moisture transfer model - 20 -
Conclusions ..............................................................................................................................................- 22 -
Acknowledgements ..................................................................................................................................- 23 -
References ................................................................................................................................................- 23 -
Abstract
Cooling represents a significant part of the energy used today, and this amount is
increasing. An alternative and less electricity intensive cooling technique is to store snow
or ice from the winter and use it for cooling during the remainder of the year. For snow
and ice storages on and in the ground, thermal insulation is necessary to reduce the
wasted melting. A wood chip layer is the traditional thermal insulation of snow. In this
layer heat and moisture transfer are intimately connected.
This study concerns the capillary transfer in cutter shaving and sawdust layers and
comprises an experiment where the capillary properties were estimated, and a model to
enable the invoking of capillary transfer in finite difference modelling.
The transfer equation, product of a new capillary moisture diffusivity and the moisture
content gradient, is combined with a height dependent damping function, a height and
dry density dependent maximum capillary moisture content and a maximum capillary
height. The capillary height of most experiments was less than the 0.2 m wood chip
layer, due to the minimum contact distance between the individual wood chips.
Aim and limitations

The aim of this study was to suggest a model for capillary moisture transfer in cutter
shavings and sawdust, and enable finite modelling of heat and moisture transfer in these
materials. This first study does not answer all questions about capillary transfer in wood
chips.
-1-
Moisture in air and porous materials

Moisture is here denoted as both liquid and vapour water, while water denotes the liquid
phase. The vapour content in air (vair) [kg m-3] depends on air temperature (Tair) [oC] and
relative humidity (RH) [%]. Although different empirical relations exist, there is no
theoretically derived function for maximum vapour content in air (vsat,air) as a function of
temperature. One common relation for air temperatures between 0 and 30oC is from the
DIN 4108 (1981),
n 8.02
§ T · § T ·
M water a ¨¨ b air ¸¸ 18.02 288.68 ¨¨ 1.098 air ¸¸
© 100 ¹ © 100 ¹
v sat,air (1)
R Tair,K 8314.3 273.15 Tair
Mwater is the molar weight of water [kg kmol-1], R is the universal gas constant [J kmol-1
K-1] and Tair,K is air temperature in [K]. The expression also includes the empirical
coefficients a and b.
A porous material, or specimen, can be both permeable and impermeable, though here it
is only permeable. The moisture content of a porous material can be divided in different
ways, both as a part of the structure and bonded in different chemical and physical ways.
From a practical aspect, making a division in moisture able and not able to evaporate is
suitable (Ahlgren, 1972). The moisture able to evaporate is henceforth named moisture.
The moisture content of a material is denoted w [kg m-3] and can be measured in 15
different ways. One common method is the gravimetric, where a specimen is oven dried
at 105oC until constant weight is attained. The moisture content will then be low and in
equilibrium with the relative humidity in the oven. The residual moisture ratio of the
specimen will typically be about 5 g of moisture per kg dry material (0.5%), considered
as dry (Skaar, 1988).
In a porous material, w depends on RH, Tair, the temperature of the material, if it is in
contact with water, and the moisture history of the material. To facilitate this explanation,
some notions are introduced. The porous material is divided into a solid part and a pore
solid total pore volume
hygroscopic volume
whygr wcr wcap wsat

part, Figure 1.
Figure 1 Schematic picture of porous material with the pore volume divided into moisture
relevant sections. The relative relations are not accurate.
The moisture content of a wood chip layer, wwc [kg m-3], is

m moist
w wc (2)
Vwc
where mmoist is moisture weight [kg] and Vwc is the initial wood chips sample volume
[m3]. Another way to measure the moisture in wood chips is the moisture ratio (uwc) [kg
kg-1],
-2-
m moist
u wc (3)
m wc,dry
where mwc,dry is the weight of the dried wood chips [kg]. The dry density of the wood
chips (Uwc,dry) [kg m-3] is needed to compare compactness. Assuming that the volumes of
the wet and dry wood chip layers is the same the dry density is
m wc,dry m wc, wet
ȡ wc,dry . (4)
Vwc Vwc 1 u
The relation between moisture content and moisture ratio is

m moist ȡ wc,dry
w wc u ȡ wc,dry (5)
m wc,dry
Hygroscopic moisture
A porous material in humid air with constant relative humidity and constant temperature
will eventually reach equilibrium moisture content. During absorption, moisture is first
adsorbed in monomolecular layers, then in multiple layers and finally by capillary
condensation, Figure 2.
A B C
Figure 2 Adsorption in a monomolecular layer (A), in multiple layers (B), and by capillary
condensation. RH is lowest in A and highest in C.
The hygroscopic moisture content mainly depends on the relative air humidity, called a
sorption curve or isotherm, Figure 3A. During capillary condensation the change in
moisture content increases for a given RH change. Increasing hygroscopic moisture
content is called absorption and decreasing desorption. The sorption curve follows
different paths depending on material and if the material is in absorption or desorption,
i.e. w depends on the moisture history of the material – called hysteresis. The equilibrium
moisture contents of absorption and desorption are denoted wabs and wdes. The Tair
dependence is small; w only increases about 1% per 10oC decrease for a given RH
(Stamm, 1964). It is custom to relate w only to RH at 20oC (Hagentoft, 2001).
The slope of the sorption isotherm dw/dRH, Figure 3B, is used to define the moisture
diffusivity coefficient of hygroscopic moisture content, Equation 9.
-3-
A B
35 100
30 w,des
w,abs 80
25
wwc [kg m ]
-3
dw/dRH
20 60
15 40
10
20
5
0 0
0% 20% 40% 60% 80% 100% 0 10 20 30
RH [%] w [kg m-3]
Figure 3 A) Sorption isotherms for cutter shavings with a dry density of 90 kg m-3, based on
curves of Swedish spruce with dry density 430 kg m-3 (as per Hedenblad, 1996). B)
Slope of the mean sorption isotherm of A.
It was assumed that cutter shavings had the same sorption isotherm of moisture ratio as
solid wood, scaled to the lower dry density. The sorption isotherms of Figure 3 are based
on spruce measurements with Udry=430 kg m-3 (Hedenblad, 1996) and converted to cutter
shavings with Udry=90 kg m-3 by a factor 90/430=|0.2093. The macro porosity of a wood
chip layer with Udry=90 kg m-3 is then about 80%. Density and moisture characteristics
vary with wood type. The dry density is 390-480 kg m-3 for spruce (Picea abies) and 480-
530 kg m-3 for pine (Pinus silvestris) (Saarman, 1992). The moisture ratio sorption
isotherms of spruce (Udry=420 kg m-3) and pine (Udry=510 kg m-3) are almost identical
(Ahlgren, 1972), which is why the moisture content sorption isotherm of pine is about
20% higher than that of spruce. At equilibrium RH and pore water pressure in the
capillary, condensed water satisfies Kelvin’s law,
R Tair,K ȡ water R Tair,K ȡ water
p pore p sat Tair ln RH | ln RH . (6)
M water M water
By assuming Tair=20oC and Uwater=1000 kg m-3, large negative pore water pressures are
obtained for different RH, Table 1. The negative pore water pressures cause a suction
that is balanced by the surface tension of water (see section Suction). A larger pressure
difference requires a more curved surface, i.e. smaller pores.
Table 1 Equilibrium pore water pressures (ppore) of some relative humidity’s (RH).
RH 50% 60% 70% 80% 90% 95% 99%
ppore [Pa] -93,753,235 -69,092,909 -48,242,900 -30,181,800 -14,250,782 -6,937,794 -1,359,382
ppore [bar] -938 -691 -482 -302 -143 -69 -14
At the fibre-saturation point, no water in the cell cavities though the cell walls are fully
saturated. This is the equilibrium w at 100% RH, denoted whygr. This point is not well
defined; hence, the sorption curves normally stop at around 98% RH. The hygroscopic
moisture ratio of spruce cutter shavings (Udry=90 kg m-3) at RH=98% is about 30.7%, as
per the above.
Sorption isotherms are measured by storing a specimen in either climatic chambers
where the temperature and humidity are controlled or in a sealed chamber with a
saturated salt solution at constant temperature. A certain salt and temperature give a
certain RH. In a sealed chamber, evaporation from a water surface will theoretically take
place until no difference in vapour content between surface and air exists, i.e.
-4-
vair=vsurf,sat(Tsurf). RH then theoretically equals 100%, though this is difficult to reach

(Nevander and Elmarsson, 1994).
A moisture content change below the fibre-saturation point causes swelling or shrinking.
The maximum shrinking is 3-10%, depending on direction and wood quality (Skaar,
1988).
Liquid moisture
Reaching the critical moisture content level wcr, Figure 1, by hygroscopic absorption is
not possible. To achieve water in the cell cavities the material has to be in contact with
water. The maximum water content a material can reach by capillary suction is called the
capillary saturated water content, wcap. To achieve a totally saturated material, wsat, it is
necessary to remove trapped air, done by repeated boiling or vacuum treatment.
Moisture transfer
In porous materials moisture might be transported as both vapour and water. The transfer
rate is calculated as the product of a transfer coefficient and some potential gradient (v,
w, p, T). Vapour transfer occurs as both diffusion ( m diff ) and mass convection
(m conv ). Liquid transfer might be divided in flow due to pressures above atmospheric
pressure that are mainly caused by gravity (denoted Darcy flow) and due to pressures
below atmospheric. Darcy flow means water is being pressed through the pores of the
material, occurring mainly in the bigger pores. For water pressures below atmospheric
the transfer occurs by suction (s), occurring mainly in the smaller pores. Suction is the
difference between air and water pressure, i.e. a negative pressure. Transfer by suction is
divided in liquid transfer and capillary suction where the latter is connected to absorption
from a water surface.
Diffusion
Fick’s diffusion law describes vapour diffusion in air, which in 1D is
dv
diff,air
m D v D (7)
dx
diff,air is diffusive vapour flow rate [kg m-2 s-1], D is vapour diffusivity in air [m2
where m
s-1], v is the vapour content gradient [g m-3 m-1], dv is the difference in vapour content
[g m-3] and dx is the distance over which the vapour content difference occurs [m]. The
vapour diffusivity D is replaced by a vapour permeability coefficient Gvap [m2 s-1], since
the diffusion in a porous material is smaller than in air,
dv D dv
diff
m į vap (8)
dx ȝ por dx
where Ppor is a resistance coefficient [-],Ppor1. The vapour permeability coefficient of

sawdust is 1510-6 m2 s-1 at 20oC (Nevander and Elmarsson, 1994).
Using the moisture content gradient to calculate isothermal moisture transfer during
hygroscopic moisture contents is also possible. The moisture diffusivity Dw [m2 s-1] is
calculated for the temperature of interest by,
-5-
dv dv dRH dw 1 dw
diff
m į vap į vap į vap v sat,air (T) (9)
dx dRH dw dx dw dx
dRH
į vap v sat,air (T)
w D w w
ȟ
Another way to calculate the moisture transfer at hygroscopic moisture contents is by the
Kirchhoff potential, which efficiently handles the nonlinearities of moisture transfer. The
Kirchhoff potential \ [kg m-1 s-1] is defined as
v x ,t
\ x, t ³ v ref
G vap v dv . (10)
The 1D moisture flow then becomes

wȥ wv
diff
m į vap . (11)
wx wx
The actual moisture transfer at hygroscopic moisture content might be purely diffusion or
a mix of diffusion and suction or surface migration (Hagentoft, 2001).
Mass convection
Mass convection is both a surface and bulk phenomenon. In bulk convection when air
conv ) [kg m-2 s-1]
passes through a porous material the convective moisture transfer rate ( m
to the porous material is
conv v
m Vair air,in v air,out (12)
where V is the air volume flow rate [m3 m-2 s-1=m s-1], and vair,in and vair,out is air
air
vapour contents before and after the passage. The convective moisture transfer rate to a
surface, i.e. condensation, is
conv,surf
m ȕ v air v sat,surf (13)
where E is the moisture transfer coefficient [m s-1] and vsat,surf is the saturation vapour
pressure of the surface [kg m-3]. The value of E depends on the air velocity and surface
properties, and is often related to the convective heat transfer coefficient Dconv [W m-2 K-
1
] by the Lewis formula approximation,
D conv
Econv (14)
Uair c p , air
where Uair is the air density [kg m-3] and cp,air is the air heat capacity at constant pressure
[J kg-1 K-1]. One way to calculate Dconv is by using dimensionless numbers, though there
are also empirical relations. In a porous material, evaporation or desorption occurs when
vsat,surf>vair and w>weq,des (Hagentoft, 2001).
Darcy flow
The water flow in a porous layer is commonly estimated by Darcy’s law,
k ǻp

m ȡ wat (15)
ȝ ǻx
-6-
Here, k is the material permeability [m2], P is the dynamic viscosity [N s m-2][Pa s] and
'p is the pressure difference [Pa] over the distance 'x [m]. Either suction or over-
pressures causes the transfer. The permeability k is dependent on the type of fluid
transfer, whereas the intrinsic permeability k0 [m2] is a material parameter independent of
transfer. Equation 15 is mainly used for saturated laminar flow, e.g. ground water flow,
since suction curves are normally not available for building materials. It is common to
use empirically decided permeability, but it can also be estimated theoretically (Yen,
1990).
The air permeability of cutter shavings varies from 210-10 m2 for compressed material
(Udry=240 kg m-3) to 1.1510-8 m2 for loosely filled (Udry=100 kg m-3) (Nevander and
Elmarsson, 1994). The air permeability for medium density fibreboard is 1.610-12 and
6.810-14 m2 for brick (Hagentoft, 2001).
Rainfall on a wood chip layer causes a downward water transport, initially because of
gravity and then gradually by pressure differences. It is reasonable to believe that the
wood chip layer acts as a soil matrix with root channels and worm holes, where the
“channels” correspond to the voids between the wood chips, here called macro pores.
Bevin and Germann (1982) showed that rainwater is transported downwards into the soil
matrix until the rain intensity exceeds the infiltration rate. Until then, the channels were
dry. The rainwater then flowing into the channels is at atmospheric pressure and is
transported purely due to gravity. In a Japanese study where snow was thermally
insulated with rice shell, the rainwater found or made channels through the insulation
layer (Nakamura et al, 2004).
Suction
Suction occurs because of adhesive attraction forces between water molecules and the
material, as well as cohesive forces between water molecules. Molecules at the surface of
solids and liquids have higher energy than interior molecules because they are bonded by
fewer molecules, called surface energy [J m-2] or surface tension (V) [N m-1]. A liquid
droplet in a gas strives to minimize its surface energy and forms a sphere. The pressure
inside the sphere exceeds atmospheric pressure, whose absolute value depends on V and
sphere radius. When the droplet comes in contact with a solid material the new shape of
the droplet depends on the relation between cohesion and adhesion. If the cohesive forces
are stronger than the adhesive forces, the droplet will more or less keep its shape. If the
adhesive forces are stronger the droplet will wet, and float out on the material. The
relationship between cohesive and adhesive forces is described by the contact angle T.
An increased T means relatively stronger cohesive forces than adhesive. T of water and
ordinary building materials is approximately zero, i.e. a water droplet will be flattened to
a thin layer or absorbed by a porous material.
If a small diameter tube is inserted through a liquid surface, the liquid might rise or sink
inside the tube depending on the relation between cohesion and adhesion. In a glass tube
a mercury surface will sink, while a water surface will rise. The surface tension of
mercury is 0.486 N m-1 and 0.073 N m-1 for water. The appearing curved liquid surface is
called meniscus, which is concave for rising liquids and convex for sinking fluids. Under
the concave meniscus the pressure caused by the adhesive and cohesive forces is below
atmospheric and depends on V, T and mean radius of the meniscus (rmenisc). The
difference between atmospheric and water pressure, i.e. suction, at the capillary interface
is
-7-
2 ı cosș 2ı
p air p water s ș|0 | . (16)
rmenisc rtube
The capillary height of water (zcap) in, e.g., building materials is calculated from the pore
radius(r):
2V
z cap . (17)
r U wat g
A smaller diameter capillary tube in the middle section and smooth shaped transition
between thinner and thicker parts are now assumed, Figure 4A-C. In case A the left
directed force at 1 is greater than the right directed force at 2. However, there is also a
right directed force from the curved walls and the total of the right directed forces is
larger; hence, the water will move right until equilibrium, case B. It is also possible to
examine the problem by suction, where suction at 1 is less than the suction at 2 in case A
due to the meniscus radius being greater at 2. When the tube is held vertically, case C,
the water will have its equilibrium moved slightly downwards because of weight. If the
water’s weight is too big, it will pour out.
C
B 1
A
1 1 2
2
transport direction
2
Figure 4 Principles of suction in a horizontal and vertical capillary tube.
The capillary height of sand is 0.04-3.5 m and more than 10 m in clay (Nevander and
Elmarsson, 1994). The potential capillary height in wood is hundreds of meters (Skaar,
1988). The stationary capillary height for ground materials of various pore sizes also
depends on moisture history.
Liquid transfer
When water pressure is lower than air pressure the liquid transfer might be calculated by
liq
m D w w (18)
where the moisture diffusivity for w>wcr is defined starting from Darcy’s law (Equation
15),
k suc ȡ wat dp k suc ȡ wat dp dw k suc ȡ wat
liq
m w D w w . (19)
ȝ dx ȝ dw dx dw
ȝ
dp
Here, ksuc is the permeability of the unsaturated porous material and depends on moisture
content. Dw at moisture contents above wcr is thus defined as
k suc U wat
Dw . (20)
dw
P
dp
-8-
The water retention curve in Figure 5A describes the relation between moisture content w
and suction s of a hypothetical material (Hagentoft, 2001). When only small pores are
water filled, a large negative pressure exists, see Table 1. As the moisture content
increases the negative pressure decreases. Figure 5B shows the slope of the absorption
retention curve, i.e. dw/ds, as a function of w. The slope of dw/dp increases with w.
A B
w [kg m-3]
60 0,0
moisture content w [kg m-3]
w,des 0 10 20 30 40 50
dw abs /ds [kg m Pa ]

50 -0,5
-1
w,abs
40
-3
-1,0
30
-1,5
20
10
-2,0
0 -2,5 -3
0 20 40 s [Pa] 60 80 100 w [kg m ]
Figure 5 Principle water absorption and desorption retention curve and slope of the absorption
retention curve as a function of w, where suction (s) is negative pressure.
The moisture diffusivity Dw is a function of moisture content w, showing in Figure 6 a

typical graph with an initial decrease followed by an increase. Wood is a fine porous
material with values similar to lightweight concrete. The lowest moisture diffusivities are
at the critical moisture content wcr (Hagentoft, 2001).
Initially, Dw decreases because the slope of the sorption isotherm (dw/dRH) increases,
Equation 9 and Figure 3B. The Dw increase above wcr is explained by an increased
permeability (ksuc). More pores included in liquid transfer causes an increased
permeability, e.g. Engelmark (1986). The negative retention curve slope increases with w
and thus has a decreasing effect on Dw, though the effect of increased permeability is
greater. Viscosity and density are independent of w.
It is common in building moisture transfer applications to use constant moisture
diffusivities in both the capillary and hygroscopic regions, due to the difficulties of
finding accurate data (Hagentoft, 2001).
1E-05
brick
1E-06 light-weight concrete
Dw [m2 s -1]
1E-07
1E-08
1E-09
1E-10
0 50 100 150 200 250
w [kg m-3]
Figure 6 The moisture diffusivity for lightweight concrete and brick (after Nevander and
Elmarsson, 1994).
-9-
Capillary transfer
The transport mechanisms for capillary transfer are the same as for liquid transfer, but the
notion is used for upward transport from a liquid surface. When the transfer stops there is
a moisture content gradient, called active, where w decreases with height. Above a
certain height the moisture content is below the critical level. If a very wet porous
material is drained it will attain a new moisture gradient through the material where both
total moisture content and capillary height are greater than before the soaking, i.e. the
passive capillary moisture content (Nevander and Elmarsson, 1994). The transfer rate
decreases with height. The capillary moisture flow into a porous material in contact with
liquid water, m cap [kg m-2 s-1], as function of time is
A
cap
m (21)
2 t
where A is the water sorption coefficient [kg m-2 s-½] and t is the time the material is in
contact with the water [s]. Tabulated A for wood is 0.004 kg m-2 s-½ perpendicular to
fibre direction and 0.02 kg m-2 s-½ parallel to fibre direction (Hagentoft, 2001). By
integrating Equation 21 the total amount of water absorbed by the material, mcap [kg], is
m cap A t . (22)
A is not an ordinary coefficient because it decreases with time (Nevander and Elmarsson,
1994), which explains why either a maximum amount of absorbed water or a time
dependent A must be used. A can be calculated directly from Equation 22 or from the
secant module. The capillary front height zcap [m] as a function of time is
z cap B t (23)
where B is the water penetration coefficient [m s-½]. B also decreases with time and
approaches zero as zcap,active is reached. No values of B for wood or wood chips were
found in the literature. The capillary water transport rate [kg m2 s-1] through a porous
layer of thickness z, in contact with a water surface, might be described by combining
Equations 21 and 23,
AB
cap
m . (24)
2z
Equations 21-24 are empirically derived and the active physical mechanisms are implicit
in the A and B coefficients. Height dependent dampening is invoked by the time square
root of Equations 22 and 23. Another way to describe the effect of gravity is by the
vertical capillary permeability gradient,
dk cap dk cap dw
(25)
dz dw dz
where kcap of soils can be approximately related to suction by expressions of the type
k cap k 0 e ms . (26)
Here, k0 is the intrinsic permeability and m is a constant (Gardner, 1959). The capillary
permeability thus increases with increased moisture content. During saturated conditions
kcap=k0.
- 10 -
Non-isothermal moisture transfer

For correct non-isothermal moisture flow calculations, it is necessary to use two
independent state variables like
m Dvw v Dwv w (27)
where Dvw Dvw v, w and D wv D wv v, w . Vapour transport varies with temperature

because the vapour saturation content depends on temperature. The liquid transport
depends on temperature mainly by viscosity, Equation 15. Other potential combinations
might be used as well. Such data are usually not available and thus simplifications are
used. Hagentoft (2001) suggests
° G v0 v w wcr
m moist ® 0 '0
(28)
°̄ G v v Dw w w t wcr
where G v0 is the coefficient of pure vapour permeability [m2 s-1] and Dw'0 is a constant
coefficient of pure liquid transport.
Experiments
Four experiments were conducted to investigate the capillary moisture transfer rate and
moisture contents in cutter shavings and sawdust, with a focus on cutter shavings. The
moisture ratio was calculated by the wet and dry weight of the samples, Equation 3. The
samples were dried in a 105oC oven until the weight decrease stopped. Both wood chips
types were a mix of spruce and pine. Assuming 50% of each, a corresponding dry density
of solid wood is 470 kg m-3.
XP1; Capillary suction from a free water surface

In XP1 a perforated bottomed plastic column (i=0.094 m, height 0.25 m) was filled
with a 0.2 m layer of cutter shavings (CS) or sawdust (S) and placed into a can with a
constant water level, Figure 7. To avoid evaporation from the water table and the column,
a lid and a plastic bag on top of the column were used.
Figure 7 Column for capillary coefficient measurements. The lid was to avoid evaporation from
the water table.
The column was initially filled with 0.005 m of cutter shavings and placed into the can to
exclude the drowned bottom cutter shavings from the measured weight. After two hours
it was assumed that the cutter shavings were saturated, and the column was weighed.
The column was then filled with cutter shavings to 0.2 m above the water level, weighed
again, and then placed into the can where the water level was adjusted and the measuring
- 11 -
started. The water table varied ±3 mm. All cutter shavings (except in CS8) were stored
for more than a month in the same room before experiments. The cutter shavings of CS8
were from another source. No cutter shavings were reused.
The capillary front and weight increase were measured numerous times during each test.
The column was lifted out of the water one minute before weighing, when most of the
free water was drained. A total of 9 tests with different dry densities were made, Table 2.
The water sorption coefficient A was calculated both as the secant module of the water
absorption curve (denoted A1), and from the total amount of absorbed water according to
Equation 22 (denoted A2). The water penetration coefficient B was calculated as a
function of the capillary front position, Equation 23. The mean capillary front height was
measured with a scale, Figure 7.
After the test, the moisture ratio was measured at different heights of the column and
recalculated to moisture content with dry density, Equation 5, both above and below zcap.
To avoid wrong cutter shavings in the capillary moisture measurements, the samples
were taken 0.01-0.04 m above the water level and 0.005-0.025 m below the capillary
front. The dry density was calculated from initial moisture ratio, Equation 4.
XP2; Maximum hygroscopic moisture ratio

The maximum hygroscopic moisture ratio uhygr, i.e. the maximum moisture a porous
material can absorb from humid air, was investigated in two experiments by putting an
open box with cutter shavings into a water filled closed box until the weight increase
stopped. Both boxes were placed into a closed insulated box to obtain a stable
temperature, Figure 8. The room temperature was 20-22oC.
Figure 8 Experiment for measuring the maximum hygroscopic moisture ratio, uhygr.
XP3; Moisture content change of initially wet cutter shavings during capillary conditions
An experiment with initially wet cutter shavings was conducted to get an indication of
the passive moisture ratio gradient. The column was filled with wet cutter shavings and
placed in the water filled can, like in XP1. Moisture ratio at different levels, total
moisture content, and dry density were measured after the experiment.
XP4; Moisture content with capillary conditions and simulated condensation

A test was made to investigate the effect of condensation where water was slowly
injected into the hygroscopic layer of the column. Before injection the test was treated as
in XP1, until the capillary rose and the weight increase stopped. Condensation was then
simulated by an infusion set device, where water was injected into the middle of the
hygroscopic layer for 1046 h (43.6 days). A reasonable condensation rate is about 200 g
- 12 -
m-2 h-1, corresponding to 1.4 g h-1 for the column area. The smallest and used stable flow
was about 7.1 g h-1.
Results

The capillary height and moisture weight quickly increased fast in the beginning, before
gradually slowing and eventually stopping, like in run CS3, Figure 9. In most runs the
capillary rise did not continue throughout the whole 0.20 m cutter shaving layer.
Measured and calculated results are seen in Table 2. The high initial moisture ratios of
CS7 could not be explained. The test lengths were 4.8-24.2 days. It would have been
favourable to continue most experiments to ensure that the increase of capillary height
and moisture weight had stopped, especially CS6 and CS8. It was, however, judged that
longer tests would have meant only small result changes.
0,20 250
acc. moisture incr. [g]

200
0,15
150
zcap [m]
0,10
100
z,cap
0,05 acc. moisture incr. 50
0,00 0
0 100 200 300 400 500 600
time [h]
Figure 9 Capillary height and moisture weight increase of run CS3, with 0.20 m cutter shaving
layer.
Table 2 Measured and calculated results from XP1, where u0 is initial moisture ratio, A1 and A2
are water sorption coefficients from absorption rate and total amount of absorbed water, B
is water penetration coefficient (B), uhygr is moisture ratio in hygroscopic layer and zcap is
capillary height. CS-cutter shavings, S-sawdust.
Test u0 Dry density Length A1 A2 B uhygr zend

[%] [kg m-3] [h] [kg m-2 s-½] [kg m-2 s-½] [m s-½] [%] [m]
CS2 - 55.0 144 0.006-0.807 0.016-0.029 1.0-3.0 *10-4 - 0.107
CS3 10% 91.3 580 -0.017-0.169 0.021-0.070 0.9-3.3 *10-4 - 0.134
CS4 10% 91.2 115 0.014-0.107 0.044-0.053 1.9-2.3 *10-4 17% 0.119
CS5 12% 100.2 116 -0.025-0.207 0.042-0.104 1.8-3.8 *10-4 21% 0.119
CS6 11% 60.1 139 0.006-0.066 0.017-0.033 1.1-2.2 *10-4 19% 0.077
CS7 26% 130.8 208 -0.003-0.294 0.057-0.147 1.4-3.9 *10-4 25% 0.123
CS8 11% 70.5 116 0.035-0.184 0.053-0.092 3.0-5.2 *10-4 - 0.197
S1 11% 148.3 239 -0.003-0.687 0.084-0.344 2.2-7.6 *10-4 - 0.200
S2 11% 141.4 168 -0.014-0.756 0.081-0.378 2.1-8.0 *10-4 - 0.167
- 13 -
The negative values of A1 only occurred after zcap was reached, probably caused by a
varying water table or different amounts of water being poured out before weighing or
both. Both A2 and B decreased slowly with time. All A1, A2 and B curves were briefly
similar to those in Figure 10.
0,07 7,0E-04
A1
0,06 6,0E-04
A2
0,05 B 5,0E-04
A [kg m-2 s -0.5]
B [m s -0.5]
0,04 4,0E-04
0,03 3,0E-04
0,02 2,0E-04
0,01 1,0E-04
0,00 0,0E+00
0 20 40 60 80 100 120 140
time [h]
Figure 10 Water sorption and water penetration coefficients of run CS2.
The capillary height in all tests was almost horizontal over the column, with less than one
cm variation. A trend towards greater maximum active capillary height (zcap,active) with
increased dry density was seen, but CS8 (the unfilled marker) diverged greatly, Figure
11. CS8 was made with a new set of cutter shavings, and both CS8 and CS6 (the
enlarged marker) were stopped before stationary conditions.
0,20
0,15
zcap,end [m]
0,10
0,05
0,00
0 25 50 75 100 125 150
dry density [kg m-3]
Figure 11 Capillary height of cutter shavings at test end, as a function of dry density. CS8
(unfilled marker) and CS6 (enlarged marker) were stopped before the capillary height
rise had stopped.
The moisture ratio in the capillary layer at test end was measured at three to five different
levels, except test CS2 where no measurements were made, Figure 12.
- 14 -
0,20
CS3 (91.3)
CS4 (91.2)
0,15
CS5 (100.2)
height [m]
CS6 (60.1)
0,10 CS7 (130.8)
CS8 (70.5)
0,05
0,00
0% 50% 100% 150% 200% 250% 300% 350%
moisture ratio [%]
Figure 12 Capillary moisture ratio of CS3-CS8 as function of height. Values within brackets are
dry densities.
The calculated capillary moisture contents in the cutter shavings capillary layer, as a
function of height, are seen in Figure 13. The R2 values of the linear regressions were
0.964 (CS3), 0.957 (CS4), 1.000 (CS5), 0.991 (CS6), 0.988 (CS7) and 0.966 (CS8).
0,20
CS3 (91.3)
CS4 (91.2)
0,15
CS5 (100.2)
CS6 (60.1)
height [m]
0,10 CS7 (130.8)

CS8 (70.5)
0,05
0,00
0 100 200 300 400
moisture content [kg m-3]
Figure 13 Capillary moisture content of CS3-CS8 as a function of height, with best-fit linear
regressions. The legend values in brackets are dry densities.
The capillary moisture contents of CS3 and CS4, with almost identical compactness
(Udry=91.3 and 91.2 kg m-3), are seen in Figure 14.
0,20
0,15
height [m]
0,10
0,05
0,00
0 50 100 150 200 250
-3
moisture content [kg m ]
Figure 14 Capillary moisture content for CS3 (filled markers) and CS4 (unfilled markers), dry
densities 91.3 and 91.2 kg m-3, as a function of height.
Moisture increase, capillary height increase and capillary moisture content of runs with
sawdust S1 and S2 are seen in Figures 15 and 16.
- 15 -
0,20
S1
0,15 S2
height [m]
0,10
0,05
0,00
0 100 200 300 -3 400 500
Figure 15 Total moisture and capillary height increase of S1 and S2.
The hygroscopic layers were 0-0.123 m. The hygroscopic moisture ratio, measured in the
middle part of the hygroscopic layer, changed from 10 to 17% (CS4), 12 to 21% (CS5),
11 to 19% (CS6), 26 to 25% (CS7) and 11 to 19% (S2). Hygroscopic moisture ratios of
CS2 and CS3 were not measured. No hygroscopic layer was detected in CS8 and S1.
600 0,30
500 0,25
moist incr. [g]
400 0,20
height [m]
300 0,15
S1 moist
200 S2 moist 0,10
S1 height
100 0,05
S2 height
0 0,00
0 50 100 150 200 250
time [h]
Figure 16 Capillary moisture content of S1 and S2.
XP2-4
XP2 consisted of two runs to estimate the maximum hygroscopic moisture ratio uhygr of
cutter shavings, i.e. the equilibrium moisture ratio with RH|100%. Moisture ratio
increase and final moisture ratios, i.e. 37.3% and 38.7%, are seen in Figure 17.
50%
40%
30%
u [%]
20%
ucr,1
10% ucr,2
0%
0 500 1000 1500 2000 2500
time [h]
Figure 17 Hygroscopic moisture ratio increase of cutter shavings with RH|98%.
XP3 was one test lasting 816 h (34 days), with initially wet cutter shavings to estimate
the maximum passive moisture content. Dry density was 91.8 kg m-3 with an initial
moisture content of 214.8 kg m-3. Upwards drying was hindered. The maximum passive
- 16 -
moisture contents together with active dittos of similar dry densities (CS3 and CS4) are
shown in Figure 18.
0,20
0,18 XP3
0,16
CS3/CS4
0,14
0,12
height [m]
0,10
0,08
0,06
0,04
0,02
0,00
0 100 200 300 400
Figure 18 Moisture ratios of XP3 and runs CS3/CS4.
In XP4, with a dry density of 94.0 kg m-3, the moisture content change by simulated
condensation in the hygroscopic cutter shaving layer was measured. Before the real
experiment could begin, the cutter shavings were at maximum active capillary moisture
content, Figure 19.
0,20 400
acc. moisture incr. [g]
0,15 300
zcap [m]
0,10 200
z,cap
0,05 100
acc. moisture incr.
0,00 0
0 200 400 600 800 1000
time [h]
Figure 19 Capillary height and weight increase of cutter shavings in XP4 during preparation.
Condensation was then simulated for 120 hours at about 0.175 m. After 2.5 days of
simulated condensation the whole layer had capillary moisture content. The inflow rate
was significantly lower during the last 2.5 days. When the run was stopped and the tube
was lifted out the high flow rate returned. The average inflow rate was 7.1 g h-1 (1.0 mm
h-1), with initially higher values (10.0 g h-1). The total inflow was 850 g (122.5 mm) of
which 411 g of water poured downwards to the free water. Some leakage was removed to
keep the water surface at roughly constant level. The moisture contents of XP4 and the
results of XP3 are seen in Figure 20.
- 17 -
0,20
0,15
height [m]
0,10
XP3
0,05
XP4
0,00
0 50 100 150 200 250 300 350 400
Figure 20 Moisture content after simulated condensation (XP4) for initially wet cutter shavings
(XP3).
Discussion

The dry densities of cutter shavings varied from 55.0 to 130.8 kg m-3. It was estimated
that the dry density of a real cutter shaving layer on snow would be between 90-110 kg
m-3. The capillary height of cutter shavings was 0.077-0.197 m, and 0.167-0.20 m for
sawdust. With thicker layers and longer tests, the capillary height of CS8 and S1 would
probably have exceeded the present cutter shaving layer thickness of 0.2 m.
The only plausible explanation found for the low capillary height was that it
corresponded to the minimum contact distance between the cutter shavings, 0.15-0.30
mm for a capillary height of 0.1-0.2 m. A correlation between dry density and capillary
height was indicated, Figure 11.
The capillary moisture ratio decrease was roughly linear with height; all R2 values were
above 0.96. A second-degree polynomial, however, increased R2 even further, i.e. there
was a slight curvature with moisture ratios in the top and bottom of the capillary layer
above the linear regressions. However, it was difficult to make any specific conclusions
with only three measurements. There was also a vague tendency that the moisture ratio at
the capillary front decreased with increased maximum capillary height.
The decreased capillary moisture content with height is congruent with soil
measurements, explained by smaller and smaller unfilled pores at higher levels. At some
level the adhesive forces were too small to lift the water further. The capillary moisture
content increased with increased dry density, but no such relation of the capillary
moisture ratio existed.
In CS4-CS6 the hygroscopic moisture content increased 3.5-5.2 g. This moisture
originated mainly from the water surface, since the initial air vapour content
corresponded to less than 1% of the increase. The time required for a 4.5 g increase by
diffusion is 90-181 h. This corresponds roughly to the experiment durations of 115-208 h
(CS4-CS8).
Theoretically the diffusion should have continued until the air was saturated and there
was maximum hygroscopic moisture ratio above the capillary level. The measured
moisture ratio in the hygroscopic layer was 17-25%, i.e. considerably lower than in XP2
(37.3-38.7%). In XP2, between 70 and 200 h were required to reach a moisture ratio of
30%, and 250 and 700 h to reach 35%. It is thus reasonable to assume that the
hygroscopic moisture ratio should have increased further with time.
- 18 -
Wood swells as the hygroscopic moisture content changes porosity, thus affecting air
transport. In spruce the macro porosity (volume between wood chips) of dry cutter
shavings with Udry=100 kg m-3 is about 76.7%. Wood swell caused by moisture is
between 3-10%. A 7% swell in all directions means a volume increase of about 22.5%.
The porosity then decreases by about 5% to 71.5%, i.e. a noticeable change. The effect
increases with an increased dry density. There might also be a porosity decrease caused
by water film on the cutter shavings in the capillary layer, though this question was not
treated here.
The water sorption coefficient A was calculated in two ways, from the secant module of
the moisture absorption curve (A1) and the total moisture uptake (A2, Equation 22). A1
fluctuated more than A2 because the rate was more sensitive to remaining subsurface
water and the varying water level. A2 slowly decreased towards zero, and values ranged
from the highest tabulated values of solid wood to one magnitude larger.
The water penetration coefficient B of cutter shavings was 0.9-3.910-4 m s-½ (CS2-CS7).
In CS8, where zcap,active was clearly higher, B was 3.0-5.210-4 m s-½. B of sawdust was
2.1-8.010-4 m s-½, and was thus about the same size as concrete (wcr=0.5) and cement
mortar (B=2.1-5.110-4m s-½). B of brick (Udry=1700 kg m-3) is 14.110-4 m s-½. B also
slowly decreased towards zero.
Initial moisture increase and capillary rise of S1 and S2, and thus A and B, were almost
identical despite a rather large difference between total moisture increase and final
capillary height. Sawdust of both runs was from the same box, with dry density and
initial moisture ratio being similar. No explanation was found for the difference.
XP2-4
In XP2 the measured maximum hygroscopic moisture ratios were 37.3% and 38.7%. No
RH measurements were made, but it was assumed that RH was close to 100%. Thus uhygr
of cutter shavings is at least 38%, i.e. 8% higher than Ahlgren’s (1972) measurements of
solid wood.
The maximum passive moisture content of cutter shavings in XP3 was significantly
higher than the active ditto, Figure 18, and there was capillary moisture content in the
whole layer. The initial moisture ratio was 234% and the total moisture content increased
from 298.0 to 486.1 g. The two uppermost moisture ratios decreased while the others
increased. The moisture content decreased with height. By linear, second degree
polynomial, power and exponential extrapolations the passive capillary heights were
0.19, 0.27, 0.30, and 0.42 m. The R2 values were 80.2%, 93.8%, 98.9% and 99.0%. XP3
reminded of conditions after an intense rain fall when all cutter shavings initially are wet.
The moisture contents of XP3 and XP4 were about the same at the bottom and top of the
layers, but the moisture content inside the layer was greater in the condensation run,
Figure 20. This is reasonable since water was continuously added at about 0.175 m
height. The condensate was transported to the surface of the cutter shaving layer,
showing that the maximum capillary height might increase to more than 0.2 m by
condensation. Evaporation was not measured, but was estimated to be small.
Since the preparation time of XP4 was 1046 hours to reach stable active moisture
contents, it was also seen that the weight and height decrease really stopped and not just
drastically slowed down. The capillary rise and capillary moisture weight increase rates
were comparable with similar runs in XP1.
- 19 -
Suggested capillary moisture transfer model

The main purpose of this study was to suggest a method to calculate capillary moisture
transfer and to receive magnitudes of moisture transfer. Such a method is developed
below, comprising;
x new function to calculate capillary transfer
x capillary moisture diffusivity curve for cutter shavings
x height dependent maximum active and passive capillary moisture contents
x height dependent moisture transfer diffusivity decrease.
Moisture transfer by suction depends on the same physical mechanisms as capillary
transfer, though in the literature a model without physical parameters except layer
thickness was used, Equation 18. To enable finite modelling a new capillary moisture
diffusivity (Dw,cap) was defined by setting Equation 18 and Equation 24 equal,
AB ǻw AB
cap
m D w,cap D w,cap (29)
2 z cap z cap 2 ǻw
By using the measured coefficients of water sorption (A2 of Equation 22) and water
penetration (B of Equation 23), and assuming constant moisture contents at bottom and
capillary front, Dw,cap was calculated at each time step. The moisture difference ('w)
used was (wbottom-wcr), where wbottom was the measured bottom moisture content after
each experiment. No moisture content measurements were made during the tests, but it
seemed reasonable that the lowest cutter shavings achieved maximum moisture content
rather fast. Almost all measurements were made more than 24 hours after the run began.
The uppermost measured capillary moisture ratios of cutter shavings after the
experiments were 155-202%, corresponding to 109.5-257.3 kg m-3. According to the
maximum hygroscopic moisture ratio measurements reported above (XP2), uhygr is about
40%. The true ucr should thus be between 40 and 150%. Dw,cap was calculated with two
moisture ratios, 60 and 120%, Table 3.
Table 3 Calculated capillary moisture diffusivities (Dw,cap) through the capillary layer, assuming
constant moisture content at bottom and capillary front, Equation 29 (e.g. the average
Dw,cap of CS2, with ucr=60%, is 2.4 *10-8 [m2 s-1]).
Dw,cap X10-8 [m2 s-1] CS2 CS3 CS4 CS5 CS6 CS7 CS8 S1 S2
Udry [kg m-3] 55.0 91.3 91.2 100.2 60.1 130.8 70.5 148.3 141.4
wbottom/wcr for ucr=60% 151/33 251/55 251/55 275/60 165/36 360/78 194/42 408/89 389/85
Dw,cap low-high [m2 s-1] (60%) 0.9-3.7 0.6-7.5 2.3-3.4 1.7-8.5 0.7-2.8 1.2-8.9 4.2-12.6 1.6-22.3 1.8-30.6
Dw,cap average [m2 s-1] (60%) 2.4 2.7 2.8 4.4 1.7 3.9 8.4 8.8 12.7
wbottom/wcr for ucr=120% 151/66 251/110 251/109 275/120 165/72 360/157 194/85 408/178 389/170
Dw,cap low-high [m2 s-1] (120%) 1.2-5.1 1.0-11.8 3.3-4.9 2.2-11.5 1.0-3.9 1.6-11.7 5.4-16.2 1.8-26.3 2.1-37.0
Dw,cap average [m2 s-1] (120%) 3.4 4.2 4.0 5.9 2.3 5.1 10.8 10.3 15.4
The calculated Dw,cap was 0.6-16.110-8 m2 s-1 for cutter shavings and 1.6-37.010-8 m2 s-1
for sawdust. These were the same magnitudes as the diffusivities (Dw) of brick and
lightweight concrete for corresponding moisture contents, Figure 6. Dw,cap of runs CS3-
CS5 (Udry=91.2-100.2 kg m-3) were 0.6-11.510-8 m2 s-1.
- 20 -
It was assumed that Dw,cap of cutter shavings and sawdust vary similarly with moisture
content as with the moisture diffusivity of other porous materials. In the hygroscopic
area, with moisture contents up to about 30 kg m-3, the moisture diffusivity was given by
Equation 9. The saturation vapour content of 15.0oC and cutter shavings vapour
permeability at 20oC, i.e. 1510-6 m2 s-1 (Nevander and Elmarsson, 1994), was used in the
calculations. The achieved moisture diffusivity curve of hygroscopic moisture contents
was smoothened slightly.
The main part of the moisture transfer in the experiments was, however, in the capillary
layer, where the measured moisture content of CS3-CS5 was 164-292 kg m-3. Since the
upper and lower measurements were made a short distance from the borders the true
capillary moisture content span was bigger. With results from Table 3 and assuming that
the main part of the moisture transfer occurs at wt125 kg m-3, the most relevant
magnitudes of Dw,cap were identified. For w between 125 and 350 kg m-3, the assumed
Dw,cap was 3.910-9-2.010-6 m2 s-1.
Dw,cap between maximum hygroscopic moisture content (30-40 kg m-3) and 125 kg m-3
was estimated from corresponding curves of brick and lightweight concrete. In CS3-CS5
the estimated critical moisture content was 55-120 kg m-3, the assumed critical moisture
content was 70 kg m-3. The critical moisture content of brick is 30 kg m-3 and 50 kg m-3
for lightweight concrete. The suggested Dw,cap curve is seen in Figure 21.
1E-05
moisture diffusivity [ms ]
2 -1
1E-06
1E-07
1E-08 Brick
1E-09 LWC
Dw,cap
1E-10
0 50 100 150 200 250 300 350
-3
Figure 21 Dw,cap curve of cutter shavings (Udry|90-100 kg m-3) and Dw curves of brick and light-
weight concrete (LWC).
Experiments showed that the capillary moisture transfer rate and capillary height increase
rate decreased significantly with height, explaining why a linear damping function R(z)
was suggested, Equation 30. The chosen coefficient zR must be bigger than Z, to also
enable capillary moisture transfer at the upper capillary level.
ǻw z R z ǻw
cap
m D w,cap R(z) D w,cap (30)
ǻz zR ǻz
The three runs with dry densities between 90-100 kg m-3 had an active capillary height of
0.12-0.13 m. It was suggested that the maximum active capillary height with initially dry
cutter shavings be set to 0.13 m. If the moisture content above this level exceeds critical
moisture content, by condensation or rain, there will also be capillary transfer above this
level.
One active and one passive height dependent maximum moisture content relation was
suggested. The moisture contents of CS3, CS4, CS5, CS7 and CS8 were used to make a
linear dry density dependent function for the active maximum moisture content function,
- 21 -
ȡ dry ȡ dry
w max,active z K1 K 2 z 281 969 z . (31)
ȡ dry,ref ȡ dry,ref
Measured and calculated active moisture contents are seen in Figure 22. The reference
dry density was 91.5 kg m-3
0,20
Calc (80)
Calc (100)
0,15 Calc (130)
CS8 (70.5)
height [m]
CS3/CS4 (91.3/91.2)
0,10 CS5 (100.2)
CS7 (130.8)
0,05
0,00
0 100 200 300 400 500
-3
Figure 22 Lines show calculated maximum active moisture content from Equation 31 and
markers measured maximum active moisture contents from experiments. Dry densities
are within brackets.
For initially wet wood chips, the exponential moisture content regression from XP3
combined with the ratio of actual and reference dry density was suggested, Equation 32.
Caution is needed for dry densities far from that of XP3 (91.8 kg m-3), since only one run
with initially wet wood chips was made.
§ z ·
ln¨ ¸
U dry © 5.12 ¹
w max,passive z (32)
U dry,ref 0.0167
Conclusions
Based on experiments and established moisture diffusivity relations of porous materials,
a method for capillary moisture transfer in cutter shavings was suggested,. The transfer
equation is the product of a new capillary moisture diffusivity and the moisture content
z R z ǻw
gradient, combined with a height dependent damping function, m D w,cap .
zR ǻz
The model also contains height and dry density dependent maximum capillary moisture
content and maximum capillary height. The critical moisture content, where capillary
transfer starts, of a cutter shaving layer with dry density about 100 kg m-3 was assumed
to be 70 kg m-3. To achieve more exact moisture diffusivities, performing measurements
where the moisture content variation during the test is known is necessary.
The wood chip layers in the experiments were 0.2 m with dry density 60.1-148.3 kg m-3.
The capillary height in cutter shaving layer was 0.077-0.197 m and 0.167-0.2 m in
sawdust. A vague correlation between dry density and maximum capillary height in the
experiments was noticed. The low capillary height was due to the minimum contact
distance between the individual wood chips, calculated to 0.15-0.30 mm.
- 22 -
There was a clear moisture content decrease with height and increase with dry density.
Linear regressions of maximum active capillary moisture content and height correlated
well, possibly depending on few measurements of each test. A vague trend of decreased
moisture gradient with increased height was seen.
One single experiment with initially wet cutter shavings gave a significant moisture
content curvature. This passive moisture content was prominently higher than the active
moisture content gradient at all levels. The exponential and power regressions improved
the degree of explanation from R2=0.80 of the linear model to 0.99. This test lasted 34
days. The moisture contents were perhaps not stationary because no measurements were
made during the test. Further research is valuable, since rain and condensation might be
common in a wood chip layer on snow.
The moisture transfer between individual wood chips above the maximum capillary level
occurs by diffusion and convection. If there is condensation in the hygroscopic layer,
upwards and downwards capillary transfer will eventually occur until the condensation
capillary layer meets the ordinary capillary layer. This was seen in one run with
simulated condensation where the capillary layer increased from 0.15 to 0.2 m, i.e. the
whole layer.
Acknowledgements
I acknowledge Prof. Johan Claesson, at the Department of Building Technology and
Building Physics, Chalmers University of Technology, for elucidating discussions and
Prof. Bo Nordell for all help with the study and the report. I also acknowledge Stenvalls
trä AB in Sikfors, Sweden, for sponsoring of cutter shavings and sawdust, and The
Swedish Energy Association and Luleå University of Technology for funding this
project.
References
Ahlgren L (1972). Moisture fixation in porous building materials. Report 36, Division of
Building Technology, The Lund Institute of Technology. (In Swedish. Pictures
and results in English).
Bevin K and Germann P (1982). Macropores and water in soils. Water Resources
Results, 18; 1311-1325.
Engelmark H (1986). Infiltration och avrinning snösmältningsperioden. Fältobservationer
och numeriska simuleringar (Infiltration and Run-Off During the Period of
Snow Melt. Field observations and numerical simulations). Licentiate Thesis
1986:08 J, Division of Water Resources Engineering, Luleå University of
Technology.
DIN 4108 (1981). Thermal insulation in buildings: DIN 4108, part 1-5. Deutches Institut
für Normung, Berlin, 1981.
Gardner W R (1959). Solutions of the Flow Equation for the Drying of Soils and Other
Porous Media. Proceedings of the Soil Science Society of America, 23;183-187.
91-44-01896-7.
- 23 -
Hedenblad G (1996). Materialdata för fukttransportberäkningar (Material data for

moisture calculations). T19:1996, Byggforskningsrådet, Stockholm, Sweden. (in
Swedish)
Switzerland.
Yen Y-C (1990). Friction loss through a uniform snow layer. Elsevier, Cold Regions
Science and Technology, 1990;18(1), p. 83-90.
Saarman E (1992). Träkunskap (Wood science). Sveriges Skogsindustriförbund,
Markaryd. ISBN 91-7322-726-9.
Skaar C (1988). Wood-Water Relations. Springer-Verlag Series in Wood Sciences. New
York, USA. ISBN 0-387-19258-1.
Stamm A J (1964). Wood and cellulose science. New York, Ronald Press.
- 24 -
Report II
Skogsberg K (2005). Convective heat and moisture transfer in a wood

chips layer on snow. A first modelling approach. Technical report.
Convective heat and moisture transfer in wood chips on snow
A first modelling approach
Kjell Skogsberg
Convective heat and moisture transfer in wood chips on snow- A first modelling approach
Content
Introduction and limitations 1
Heat and moisture transfer in a wood chip layer on snow 1
Convective heat transfer 2
Convective moisture transfer 2
Fluid flow over and within a porous layer 3
Own experiments and observations 4
Discussion 5
Model suggestions 6
Convective heat transfer inside a cutter shaving layer on snow 6
Convective moisture transfer inside a cutter shaving layer on snow 7
Conclusions 8
Acknowledgements 9
References 9
Introduction and limitations

For many centuries, snow and ice have been stored from winter to summer for
different cooling applications. The technique has received renewed attention during
the last 30 years. The traditional thermal insulation of snow and ice is a layer of wood
chips (Taylor, 1985).
In a pilot study of a 30,000 m3 snow storage in Sundsvall, Sweden, it was found that
all snow would melt away by the middle of June without thermal insulation. With 0.2
m of sawdust as thermal insulation, approximately two-thirds of the snow pile would
remain after the summer, thus available for cooling. About 84% of the melting was
caused by heat transfer through the wood chip layer (Skogsberg and Nordell, 2001).
A wood chip layer is permeable to both air and moisture transfer, e.g. rain,
evaporation and capillary transfer of melt water. In a laboratory experiment with snow
thermally insulated by 0.025-0.075 m of cutter shaving layers, evaporation was a
significant part of the energy balance (7.8-71.2%) (Skogsberg and Lundberg, 2005).
This study aims to suggest a method to include heat and moisture convection in
numerical modelling of a cutter shaving layer on snow. The principal ideas are also
applicable on layers of sawdust, wood powder and larger pieces of wood chips.
Heat and moisture transfer in a wood chip layer on snow

Heat is transferred to and from a wood chip layer on snow by radiation and
convection. Heat transferred to the snow causes melting, with some melt water being
absorbed by the wood chips, transported upwards by capillary and evaporated.
Condensation can also appear in the wood chip layer, causing moisture transfer
downwards. Moisture transfer affects heat transfer and vice versa. The moisture might
be transferred as both vapour and liquid water. The fluxes connected with the energy
balance to, from and through the wood chip layer on snow are shown in Figure 1.
1
air
m Prad Pconv vapour
m
Air
Wood chips Conduction Convection Moisture transfer
Snow
Pmelt water
m
Figure 1 Heat and mass transfers to and from a wood chip layer on snow, exclusive rain. P is
energy flux [W m-2] and m
is mass flow [kg m-2 s-1].
Convective heat transfer

Thermal convection means fluid flow heat transfer. Natural convection occurs as a
result of temperature dependent density differences, whereas forced convection is by
external forces from, e.g., a fan or the wind. Natural convection will not occur in a
horizontal wood chip layer since air density increases downwards. Thermal
convection Pconv [W m-2] between air and a general surface is
-2 -1
The convective heat transfer coefficient Dconv [W m K ] is derived from empirical
relations. One common relation uses the dimensionless Nusselt number Nu,
Nu O fl
D conv (2)
L
Ofl is the thermal conductivity of the fluid [W m-1 K-1] and L is a characteristic length
[m], depending on the specific problem. The dimensionless Nusselt number is the
ratio between actual heat transfer and conductive heat transfer and is measured or
calculated by other dimensionless empirical numbers. When forced convection is
present the Reynolds number Re is important,
u air L Uair u air L
Re (3)
Xair Kair
uair is air velocity [m s-1], Xair kinematic viscosity of air [m2 s-1], Uair air density [kg m-
3
] and Kair dynamic viscosity of air [Pa s]. The Nusselt number is proportional to the
Reynolds number to a power of a positive number depending on the specific problem.
Other empirical relationships of the convective heat transfer coefficient for many
different applications also exist. For forced convection with air velocity parallel to the
surface, these expressions are common in building physics (Hagentoft, 2001),
D conv ® (4)
°̄7.41 u air 0.78 u air ! 5 m s 1
Convective moisture transfer

The convective moisture transfer rate to/from a surface [kg m-2 s-1], i.e. condensation
and evaporation, is
2
conv
m ȕ v air vsat, surf (5)
where E is the convective moisture transfer coefficient [m s-1], vair is air vapour
pressure [kg m-3] and vsat,surf is the saturation vapour pressure at the surface [kg m-3]. E
depends on air velocity and surface properties, and is often related to the convective
heat transfer coefficient Dconv by the Lewis formula approximation,
D conv
E (6)
Uair c p , air
where cp,air is air heat capacity at constant pressure [J kg-1 K-1].
A porous material will eventually be in hygroscopic moisture equilibrium with the
relative humidity RH [%]. A weak temperature dependency that is mostly neglected
for building relevant temperatures also exists. Moisture is absorbed when the material
moisture content is less than the hygroscopic equilibrium moisture content, and the
vapour content of the air is less than that of the surface saturation. However,
condensation occurs when the vapour content of the air is greater than that of the
surface saturation.
Moisture is evaporated when the surface saturation vapour content exceeds air vapour
content, and the material moisture content exceeds the maximum hygroscopic
moisture content. Moisture is desorbed (emitted) when the surface saturation vapour
content exceeds air vapour content, and the moisture content exceeds the equilibrium
hygroscopic moisture content (Hagentoft, 2001).
Fluid flow over and within a porous layer

When modelling fluid flow the conservation of mass, energy and momentum must be
considered. The momentum equation states that the sum of all forces acting on the
control volume must equal the net rate at which momentum leaves the volume
(Incropera and DeWitt, 1996). At the surface of an impermeable material the fluid
(here air) velocity is zero. The air velocity at the interface between a porous and a free
layer (uint) is above zero. The relative velocity at the interface (uint/uair) depends on
permeability, porosity, layer thicknesses, air viscosity and air velocity in the free layer
(Prinos et al, 2003).
For fluid flows over a porous bed, numerous studies have investigated velocity
distribution and mass exchange, of the fluid and of components in the fluid. Laminar
and turbulent conditions have also been studied. The Reynolds averaged Navier-
Stokes equations (RANS) describe the flow in the free region. In the porous region
different approaches like the Darcy, Darcy-Brinkman and Darcy-Brinkman-
Forchheimer models have been used (Wahlgren, 2001). Prinos et al. (2003) is the first
study where RANS were used in both the free and the porous layers, in conjunction
with a low-Reynolds number k-H turbulence model, and simulated water flow in and
above a porous layer with porosity 0.44-0.83 and permeability 4.110-4-5.510-7 m2.
The free flow velocities were 0.16-0.38 m s-1, the Reynolds numbers were 5,600-
23,900 and the Darcy numbers were 2.610-2-7.710-5. The Darcy number is
§ k ·
¨ Da ¸ (7)
© h2 ¹
3
where h is usually only the porous layer thickness, though here it was the total
thickness, i.e. porous plus free layer thickness. Prinos et al. (2003) found that during
turbulent conditions free flow velocity and total discharge decreased with a porous
layer relative to an impermeable layer. Increased permeability decreased free flow
velocity and total discharge due to a strong momentum exchange at the interface and
penetration of turbulence into the porous layer.
The turbulence penetration depth increased with increased permeability. The interface
velocity was 24-34% of the free flow velocity. During laminar conditions free flow
velocity and total discharge increased, with the interface velocity being 3-6% of free
flow velocity. For both turbulent and laminar flow the interface velocity will be closer
to free flow velocity with increased permeability.
The permeability of cutter shavings is 1.1510-8 m2 for loose layers (U|100 kg m-3) to
2.010-10 m2 for a compact layer (U|240 kg m-3) (Nevander and Elmarsson, 1994).
Permeability will decrease downwards in the wood chip layer on snow because wet
wood chips swell and water will occupy a greater part of the pores.
Own experiments and observations

Skogsberg and Lundberg (2005) studied in a laboratory test the surface melt rate SMR
[kg m-2 h-1] of snow, thermally insulated with a 0.025-0.075 m layer of cutter
shavings. Both initially wet and initially dry cutter shavings were used. The
experiment box measured 1.250.650.90 m (LWH). SMR increased with increased
air velocity (2.3-6.8 m s-1) when air was both warmer and colder than the wood chips
surface, i.e. the convective net heat transfer was from air to snow. The effective
thermal conductivity of the cutter shaving layer was between 1.22-6.77 times the
estimated actual thermal conductivity, averaging about 2.6 times, i.e. a type of Nu
value.
Evaporation occurred from both initially wet and initially dry wood chips. The main
part of the evaporated moisture came from the snow. In this small scale laboratory test
the importance of vertical air exchange versus horizontal air flow was not clarified.
The expected evaporation rate increase in the experiment was insignificant due to air
velocity.
The capillary (wet) layer with initially dry cutter shavings was only a few millimetres
thick after one to two days. However, occasional observations at real snow storages
have shown a significant capillary layer, >0.1 m, during normal summer conditions. It
was not clarified if a capillary layer of this thickness is possible without rainfall or
condensation.
Skogsberg (2005) investigated capillary transfer from a water surface into a top sealed
column filled with 0.2 m layer of cutter shavings or sawdust. The maximum active
capillary height, with initially dry wood chips, was 0.11-0.2 m. With initially wet
wood chips similar to conditions after a rainfall, the (passive) capillary height was 0.2
m and capillary moisture content was significantly greater than with initially dry
wood chips. The capillary moisture content decreased with height and increased with
dry density. The hygroscopic moisture content, above the capillary layer, also
increased during the tests. Moisture transfer between the wood chips above the
capillary layer was due to diffusion. In a test with simulated condensation in the
4
hygroscopic layer, the condensation layer and the original capillary layer eventually
became one capillary layer.
Discussion
With still air and at stationary conditions, air and wood chip temperatures inside the
layer will be the same, while the heat transfer through the wood chip layer will be due
to conduction only. For horizontal airflows the temperature of wood chips and air will
equal after a certain distance, i.e. convection will not influence inside the wood chip
layer.
The porosity of cutter shavings with a dry density of 100 kg m-3 is 75-80%.
Calculated Re with uair=2-10 m s-1, dynamic viscosity K=18.110-6 Pa s and a
characteristic height of 10 m is about 1.3-6.6106. Da is about 10-8-10-12.
Prinos et al. (2003) studied more permeable materials than wood chips for much
lower Re numbers. It was therefore not obvious how to use their findings in the
current case, though a turbulent exchange of air in the porous layer will exist.
This vertical air exchange contributes to increased surface melt if the air penetrates to
a depth warmer than the wood chips. The common assumption of local temperature
equilibrium between fluid and solid material, the Oberbeck-Boussinesq
approximation, is thus not valid here. The convective heat transfer then occurs from
wood chips to air in the uppermost wood chip layer and from air to wood chips below
this level, denoted ziso. Since the down going air is heated and the wood chip
temperature decreases with depth, a superficial ziso is reasonable.
Viscosity decreases and air density increases downwards due to the varying
temperature of the wood chips from about 0oC by the snow to about air temperature at
the surface. Assuming normal summer conditions in central Sweden (Tair=15-25oC,
RH=60-80%), the dew point temperature (Tdp) will be reached at a certain depth (zdp)
in the wood chips. Condensation conditions are found beneath this level. If the dew
point level is beneath the capillary layer thickness, evaporation at stationary
conditions will occur, since v(zcap)>vair, see left part of Figure 2. If the dew point
temperature level is in the hygroscopic layer, there will at least be initial
condensation, see right part of Figure 2. In the capillary layer the air is saturated. In
the hygroscopic layer, above the capillary layer, the wood chips are at hygroscopic
equilibrium and RH is below 100% (Hagentoft, 2001).
When the moisture content at the condensation level has reached a certain value
moisture transfer starts both upwards and downwards. Evaporation starts once
moisture is transported above the condensation level, unless the temperature
conditions have changed and then condensation continues at this level. The
condensation layer eventually merges with the capillary layer. Once the two layers
have merged, there might be both evaporation and condensation, but the capillary
height will not increase above the active capillary height without condensation at a
higher level.
During stationary condensation water will drain to the snow when the moisture
content exceeds the maximum passive level, requiring some time because the
maximum passive capillary moisture contents are significantly greater than the active
contents.
5
z [m]/T [oC] z [m]/T [oC]

Air
Z/Tsurf Z/Tsurf
Hygroscopic part
Tdp
RH<100%
Wood chips zcap zcap RH=100%
Capillary part
Tdp
RH=100%
0 0
Snow
Evaporation Condensation
Figure 2 Principle sketch of wood chip layer on snow with lower capillary layer and upper
hygroscopic layer. The left part shows evaporation conditions with zdp<zcap and the
right part condensation conditions where zdp>zcap.
At stationary evaporation conditions the capillary suction and evaporation transfer

rates are equal. This might happen below, at or above the maximum capillary height.
Another possibility is alternating evaporation and condensation with no net moisture
transfer.
Evaporation requires energy from its surroundings, leading to a cooling of the
surroundings and decreasing melt rate, and thus acts as a buffer towards an increased
melt rate when the air velocity increases. If both SMR and ER increase with an
increased air velocity the convective heat transfer must increase faster than energy
transfer by evaporation.
During evaporation, most energy transfer to moisture is by phase change. Assuming
that the vapour leaves the wood chip layer at 30oC, the phase change corresponds to
about 90% of the total enthalpy change.
In the capillary layer, a downward diffusive vapour transfer also exists because the
vapour content decreases downwards. The downward transfer is about 0.5-2% of the
upward capillary moisture transfer.
Model suggestions
Convective heat transfer inside a cutter shaving layer on snow

Here, a function enabling convective heat transfer calculations inside a cutter shaving
layer on snow is suggested, where actual thermal conductivity is multiplied by a heat
convection factor D´ [-]. It is thus assumed that convective heat transfer inside the
wood chip layer is from air to wood chips. The heat convection factor depends on air
velocity, permeability (by dry density) and depth by an estimated factor Dslope [W m-3
K-1]. An estimated scaling factor K [W m-2 K-1] was necessary to transform
convective heat transfer into a multiple of thermal conductivity,
ª1 § Udry ·º
Pconv Pcond D' Pcond « ¨ D conv D slope Z z ¸» (8)
¨ Udry , ref ¸»
¬« K © ¹¼
6
Udry is the current dry density of wood chips [kg m-3], Udry,ref is a reference wood chips
dry density (here 91.5 kg m-3), z is height and Z is the wood chip layer thickness [m].
From experiments, the estimated reasonable values of the factors were K=8 W m-2 K-1
and Dslope=73+uair19 W m-3 K-1. Figure 3 shows how the heat convection factor D´
changes with air velocity,
6
u,air=2 m s-1
5
u,air=6 m s-1
4 u,air=10 m s-1
Alfa`[-]
3
2
1
0
0,00 0,05 0,10 0,15 0,20
height [m]
Figure 3 Suggested thermal convection factor D´ (Alfa´) for different air velocities uair.
Figure 4 shows how the convection factor changes with dry density. Here uair=6 m s-1.
5
Density=91.5 kg m-3
4 Density=150 kg m-3
3
Alfa´ [-]
0
0,00 0,05 0,10 0,15 0,20
height [m]
Figure 4 Suggested thermal convection factor D´ (Alfa´) for two different dry densities.
Convective moisture transfer inside a cutter shaving layer on snow

Here, a function to enable convective moisture transfer calculations in a wood chip
layer on snow is suggested with a mass convection coefficient E´ [m s-1], analogous to
the suggested heat convection coefficient above. The method assumes that
evaporation and condensation only occur down to the capillary front. Vapour transfer
through the wood chip layer is not included. The evaporation/ condensation rate
decreases with dry density and depth by a chosen factor Eslope,
§ ȡ ·
conv
m ȕ´ v air v sat,surf ¨ ȕ conv dry ȕ slope Z z ¸ v air v sat T (9)
¨ ȡ dry,ref ¸
© ¹
Figure 5 shows how E` varies with air velocity and actual and reference dry density,
for Eslope=0.0523+uair0.016 (from experiments and estimations). E` varies with dry
density just as D´.
7
0,04
u,air=2 m s-1
0,03 u,air=6 m s-1
Beta´ [m s ]
-1
u,air=10 m s-1
0,02
0,01
0,00
0,00 0,05 0,10 0,15 0,20
height [m]
Figure 5 Suggested moisture transfer coefficient E´ (Beta´) for different wind velocities.
Conclusions
Two analogous functions for convective transfer inside a cutter shaving layer on snow
were suggested, one for heat Pconv [W m-2] and one for moisture transfer m
conv [kg m-2
-1
s ]. The functions are based on ordinary surface convection rates where the ordinary
convective coefficients were replaced by a heat convection factor D´ [-] and a
moisture convection factor E´ [m s-1]. Both functions decrease with dry density and
depth at an estimated rate, described by Dslope [W m-3 K-1] and E slope [s-1]. The
convective heat transfer function also included an estimated scaling factor K. The
functions were
ª1 § ȡ dry ·º
Pconv Pcond Į' Pcond « ¨ Į conv Į slope Z z ¸»
K
«¬ ©¨ ȡ ¸»
dry,ref ¹¼
§ ȡ ·
conv
m ȕ´ v air v sat,surf ¨ ȕ conv dry ȕ slope Z z ¸ v air v sat T
¨ ȡ ¸
© dry,ref ¹
Here, Pcond is actual thermal conductivity [W m-2], Dconv the convective heat transfer
coefficient [W m-2 K-1], Udry and Udry,ref are current and reference dry density of wood
chips [kg m-3], z and Z are height and wood chip layer thickness [m], vair the air
vapour pressure [kg m-3], vsat,surf the saturation vapour pressure at the surface [kg m-3]
and Econv the convective moisture transfer coefficient [m s-1].
The estimated reference dry density of a cutter shaving layer was 91.5 kg m-3.
Estimated reasonable values of the factors were K=8 W m-2 K-1, Dslope=73+uair19 W
m-3 K-1 and Eslope=0.0523+uair0.016 s-1. Actual air transfer in the layer was not
directly included.
During stationary conditions, evaporation, condensation or no net moisture transfer
might exist. Stationary evaporation from the capillary front can occur at any level up
to the passive capillary height. At stationary evaporation the capillary transfer
matches the evaporation rate. During evaporation the relative humidity and thus the
moisture content and thermal conductivity in the hygroscopic layer will increase
slightly. This was not included in the model suggestion.
8
Stationary condensation might occur at any level, though when condensation occurs
below the active capillary height, the capillary layer strives to increase, and
condensation or evaporation will be at a higher level. If condensation occurs above
the capillary layer, condensation and capillary layers will eventually merge.
Evaporation or condensation might be in the following stationary conditions. In
initially wet wood chips the capillary front is lowered until the evaporation is equal to
the capillary transport, or reaches a level where condensation occurs. There might also
be no net convective moisture transfer at stationary conditions, with alternating
evaporation and condensation.
This report does not claim to fully answer the questions of convective heat and
moisture transfers in a cutter shaving layer on melting snow, though the suggested
functions include the most important factors, give reasonable magnitudes and enable
numerical modelling. The principle idea is also applicable on layers of sawdust, wood
power and larger pieces wood chippings.
Acknowledgements
I want to acknowledge Professor Bo Nordell and Professor Anders Sellgren at Luleå
University of Technology for discussions, help and patience.
References
Hagentoft C-E (2001). Introduction to building physics. Studentlitteratur, Lund,
Sweden. ISBN 91-44-01896-7.
Incropera F P and DeWitt D P (1996). Fundamentals of Heat and Mass Transfer. John
Wiley and Sons Inc., USA. ISBN 0-471-30460-3.
Prinos P, Sofialidis D and Keramaris E (2003). Turbulent flow over and within a
porous bed. Journal of Hydraulic Engineering, ASCE, Vol. 129, No. 9, Sept.
1, 2003.
Skogsberg K (2005). Capillary suction in cutter shavings and sawdust. Experiments
and a model approach. Technical report. Division of Architecture and
Infrastructure, Luleå University of Technology.
Skogsberg K and Lundberg A (2005). Wood chips as Thermal Insulation of Snow.
Accepted for publication in Cold Regions Science and Technology, Elsevier
Science B.V.
Skogsberg K and Nordell B (2001). The Sundsvall hospital snow storage. Cold
Regions Science and Technology, 32, 63-70. Elsevier Science B.V.
Wahlgren P (2001). Convection in Loose-fill Attic Insulation. Doctoral Thesis P-01:3,
Department of Building Physics, Chalmers University of Technology,
Sweden. ISBN 91-7291-079-8.
9
Report III
Skogsberg K (2005). Mass losses of freezing water. Technical report.

Environmental Engineering, Luleå University of Technology
Mass loss of freezing water
Kjell Skogsberg
Skogsberg K (2005). Mass loss of freezing water.
Mass loss of freezing water
Kjell Skogsberg
Division of Architecture and Infrastructure, Luleå University of Technology
Abstract
As water freezes, heat is released. A significant part of this heat is emitted by evaporation or
sublimation. With repeated freezing and thawing, the effect of this phenomenon, called mass
loss, accumulates. It was therefore feared as an important water loss (and cold gain) factor in
open-pond, seasonal snow storage containers which were thermally insulated by wood chips.
Laboratory tests of mass loss due to freezing were conducted on wet wood chips. Water and
wood chips were frozen in cans that were thermally insulated at the bottom and sides. The
influence of surface area, initial water mass, initial water temperature, prevented pressure
increase in the water and wood chips on the loss of mass due to freezing were investigated.
The loss of mass mainly depended on the amount of water which was frozen. The mass loss
which occurred was typically 2.5-3% of the ice mass, corresponding to 20-30% of the total
energy transfer. The loss of mass is considerably greater in the moments directly following
the onset of freezing.
1 Background
Moisture is evaporated from freezing water. Nilsson and Sjödin (1986) analysed the
phenomenon of mass loss as a result of freezing. They found that about 2.7 g of water
evaporates when 100 g of water is frozen to ice and that freezing should result in a loss of
mass whenever water freezes in nature. In springtime, when several melting and freezing
periods occur, this phenomenon could add up to significant losses in mass. Milanov and
Nilsson (1988) performed a more detailed study and established a model for the calculation
of mass loss due to freezing. They conducted three experiments with a water-filled can and
one with a wet scouring cloth, when all water was frozen. The air temperature was about –15
o
C, relative humidity 70-80%, and estimated wind velocity 0.5 ms-1. They found that the loss
of mass due to freezing was approximately 3% of the initial quantity of water.
The vapour pressure gradient between the water/ice surface and air governs the direction, and
partly the magnitude, of the vapour transport. Evaporation increases with decreased relative
humidity and increased surface temperature. Wind velocity and surface roughness also affect
the degree of evaporation.
When a water-filled can freezes from the top the pressure in the unfrozen water increases due
to thermal expansion, since the density of ice (~920 kgm-3) is lower than that of water
(~1,000 kgm-3). The pressure increases until the temperature pressure equilibrium is reached,
or until some water is removed. The balance pressure of –10 oC and constant volume is 1,178
bar (Nordell, 1990).
Sublimation means that a substance transforms from solid into gaseous phase without
melting. A 5% relative change in humidity affects the rate of sublimation of a small sphere as
much as a temperature change of 2 oC at an air temperature of -10 oC and a relative humidity
-1-
of 70%. The rate increases almost linearly with wind velocities above 1 ms-1 (Pomeroy and
Gray, 1995).
The main objective of this laboratory study was to analyse how different parameters affect
the loss of mass due to freezing, as well as to investigate whether the phenomenon is an
important factor in seasonal snow storage.
2 Experiments
The loss of mass of wet wood chips due to freezing and the influence of some presumed
relevant factors were studied. The test was performed in cans filled with water or wood chips.
The analysed factors were surface area, initial water mass, initial water temperature,
prevented water pressure increase and a layer of initially dry wood chips on top of the
water/ice.
Four PVC cans, one steel can and one bucket of poly ethylene were used in the experiments
(see Table 1). The sides and bottoms of these cans were thermally insulated by 0.05 m of
polystyrene (O=0.027 W m-1 K-1).
Table 1 Data of cans.
Can Diameter Height Area Weight Material

[m] [m] [cm2] [g]
1 0.103 0.100 83.3 246.1 PVC
2 0.150 0.100 176.7 497.1 PVC
3 0.188 0.100 277.6 795.7 PVC
4 0.295 0.100 683.5 1529.0 PVC
Steel top 0.135 0.069 143.1 117.7 Steel
bottom 0.093 67.9
Bucket 0.300 0.250 706.9 1246.7 PE
The freezer used in the experiment was a top opening household freezer measuring 0.95 x
0.50 x 0.68 m (L x W x H), with a lid made of 0.05 m polystyrene board (O=0.027 W m-1 K-
1
). The main scale was placed on a table above the freezer and the can was placed on a plate
inside the freezer. The plate hung in a string attached to the bottom socket of the scale
(Figure 1).
The main scale (Mettler, PT15) measured up to 5 kg with errors d 0.2 g. The second scale
(Mettler, PT11), which was used to measure initial and final weights, was calibrated against
the main scale with accuracy d 0.3 g, up to 5 kg. The temperature was measured with an
electrical resistance thermometer (Pentronic, pT-100) and humidity with an electrical
resistance humidity sensor (Rotronic, YA-100C). The thermometer was calibrated with
ice/water slurry and boiling water and the humidity sensor with humidity standard solutions.
The dry can was filled with water of desired temperature and placed on the plate in the
freezer. After completing the tests, two holes (=5 mm) were drilled through the ice, water
was poured out and the ice mass was measured.
-2-
scale
temperature and
humidity gauge
fan
insulation
water/ice
plate
Figure 1 Outline of the experimental set-up.
To distinguish between the loss of mass due to initial evaporation and the loss of mass due to
freezing, the rate of evaporation from an ice-free water surface was measured, until the
surface was covered with ice. The effect of prevented water pressure increase under the ice
was also studied. This was done by using a submerged oil-filled balloon connected to a
drainage system that evacuated the oil as ice formed, Figure 2. The rate of sublimation was
also measured.
Balloon where
oil was collected
pipe
air
ice
Balloon with oil
and lead balls
water
Figure 2 Device for preventing pressure increase in the water under the ice.
Wet wood chips were initially tested at –2 oC or –5 oC in twelve experimental runs during a
period of five hours. This length corresponded to the mean diurnal freezing hours during
April 2000 and 2001, in Sundsvall. The initial moisture content varied from 11.6-236.1%.
The loss of mass due to freezing was measured: the ice mass was estimated by measuring the
freeze depth and relating it to the initial moisture content.
3 Results and discussion

In the water-filled cans, ice was formed (almost exclusively) from the top. The upper ice
surface became uneven because of the trapped water pressure which pressed the ice and some
water on top of the ice surface. The mean temperatures ranged from –9 oC to –15 oC and the
-3-
relative humidity from 60-90%. Some typical runs are shown in this report. More data and
information are found in Skogsberg (2001).
Ice needles began to form within 5-35 minutes. The surface was covered with ice within 15-
60 minutes. The evaporation rate was relatively constant both from the water surface and a
thin ice cover, Figure 3. However, there were huge variations between the results of the
experiments.
6
5
mass loss [g]
4
3
2
1
0
0 10 20 30 40 50 60 70 80 90
time [min]
Figure 3 Initial mass loss (due to evaporation) with time;

can 2, m0=1400g, T0=+7.0 oC.
The influence of the area of the can on the loss of mass due to freezing is shown in Figures 4-
5. Larger cans (and thus larger surface areas) increased the rate of mass loss since the rate of
freezing was greater.
20
Can2 (No.1.1) Can2 (No.1.9)
15
mass loss [g]
Can3 (No.1.2) Can3 (No.1.10)
10
0
0 1 2 3 4 5 6
time [h]
Figure 4 Mass loss measurements for two different can areas;

m0=800g and T0= +1.5 oC. Area of can 2 area of can 3.
50
Can 2 (No.1.7) Can 2 (No.1.15)
40 Can 3 (No.1.8) Can 3 (No.1.16)
mass loss [g]
30 Can 4 (No.2.6)
20
10
0
0 1 2 3 4 5 6 7 8 9 10 11 12
time [h]
Figure 5 Mass loss measurements for three different can areas,

m0=1400 g and T0=+7.0 oC. Area of can 2 area of can 3 area of can 4.
-4-
The mass loss rate also increased with water volume since the water surface was closer to the
edge of the cans (Figures 6-7).
20
800g (No 1.5) 800g (No 1.13)
15
mass loss [g]
1400g (No 1.7) 1400g (No 1.15)
10
0
0 2 4 6 8 10 12
time [h]
Figure 6 Mass loss measurements for different initial mass of water;

can 2 and T0= +7.0 oC.
25
800g (No 1.2) 800g (No 1.10)
20
mass loss [g]
1400g (No 1.4) 1400g (No 1.12)

15
10
0
0 1 2 3 4 5 6 7
time [h]
Figure 7 Mass loss measurements for different initial water masses,

can 3 and T0= +1.5 oC.
There was a slight indication that water which was initially warmer increased the rate of mass
loss due to freezing (Figures 8-9). The evaporation from warm water was greater than that
from cold water. With an initial water temperature of +7 oC, the water particles will also mix
until all water is close to +4 oC, due to the maximum water density at this temperature.
15
+1.5oC (No 1.1) +1.5oC (No 1.9)
mass loss [g]
+7.0 (No 1.5) +7.0 (No 1.13)

10
0
0 1 2 3 4 5 6
time [h]
Figure 8 Mass loss measurements for two different initial water temperatures;
can 2 and m0=800g.
-5-
40
+1.5oC (No.1.4) +1.5oC (No 1.12)
30
mass loss [g]

+7.0oC (No 1.8) +7.0oC (No 1.16)
20
10
0
0 2 4 6 8 10 12 14 16
time [h]
Figure 9 Mass loss measurements for two different initial water temperatures;
can 3 and m0= 1400g
Without preventing the pressure increase, water was pressed up against the ice and the
indications of its influence on mass loss due to freezing were contradictory (Figures 10-11).
No conclusion was drawn on this issue.
40
No 1.3 No 1.11
30 No 3.1 (dp=0) No 3.2 (dp=0)
mass loss [g]
20
10
0
0 2 4 6 8 10 12 14 16
time [h]
Figure 10 Mass loss with and without water pressure increase;

can 2, m0= 1400g, and T0=+1.5 oC (dp=0 is no pressure increase).
40
No 1.4 No 1.12
30
mass loss [g]
No 3.3 (dp=0) No 3.4 (dp=0)
20
10
0
0 2 4 6 8 10 12 14 16
time [h]
Figure 11 Mass loss with and without an increase in water pressure;

can 3, m0= 1400g, and T0=+1.5oC (dp=0 means there is no increase in
increase).
Three longer freezing tests were conducted, where the useful time vs. actual freezing time
were 70 of 70 hours, 95 of 114 hours and 35 of 203 hours, due to scale problems. Formed ice
corresponded to 97.2%, 92.3% and 97.9% of initial water mass. No detectable loss of mass
occurred after the freezing had stopped, i.e., there was very little or no sublimation from the
solid ice (Figure 12).
-6-
60
Run 4.1
50 Run 4.2
mass loss [g]
40 Run 4.3
30
20
10
0
0 10 20 30 40 50 60 70 80 90 100
time [h]
Figure 12 Comparison of mass loss/sublimation for long runs. The water was never
completely frozen.
The evaporated water molecules originated either from the upper ice surface or from the
ice/water interface. It was not possible to establish where the evaporated water came from.
The loss of mass and amount of ice thus increased with time until all water was frozen. The
relative mass loss (mass loss/ice mass) decreased with increased ice thickness and ice mass
(Figures 13 –14).
10%
9%
8% can 1 can 2 can 3 can 4
FL/ice mass
7%
6%
5%
4%
3%
2%
1%
0%
0 10 20 30 40 50 60
ice thickness [mm]
Figure 13 The ratio of mass loss due to freezing (FL) and ice mass as a function of ice
thickness.
10%
9% can 1 can 2 can 3 can 4
8%
FL/ice mass
7%
6%
5%
4%
3%
2%
1%
0%
0 200 400 600 800 1000 1200
ice mass [g]
Figure 14 The ratio of mass loss due to freezing (FL) and ice mass as a function of ice
mass.
-7-
The increased mass loss with respect to ice thickness conflicted with the findings of Milanov
and Nilsson (1988), who obtained a 3% loss of mass by freezing a scouring cloth; this was
the same relative loss that was obtained with water-filled cans. Their results thus indicated
that ice thickness was not an important parameter. The increase of relative mass loss with
respect to ice mass was however expected.
In contrast to the study conducted by Milanov and Nilsson (1988), only a portion of the water
was frozen in this study. Thus a comparison could be drawn between relative mass loss and
the proportion of frozen water (Figure 15).
10%
can 1 can 2 can 3 can 4
8%
FL/ice mass
6%
4%
2%
0%
0% 20% 40% 60% 80% 100%
ice mass/initial water mass
Figure 15 The ratio of mass loss (FL) and ice mass as a function of the frozen
proportion.
It was thus concluded that the relative mass loss initially decreased until about 25% of the
water was frozen. Thereafter the relative mass loss was about 2.5-3%.
The ratio of evaporation energy and freezing energy as a function of different ice parameters
has the same principal appearance as the mass loss/ice weight figures, but the ratio is greater.
Evaporation energy/freezing energy as a function of ice mass/initial water mass is seen in
Figure 16.
evaporation energy/
80%
freezing energy
60%
40%
20%
0%
0% 20% 40% 60% 80% 100%
ice mass/initial water mass
Figure 16 The ratio of evaporation energy and freezing energy as a function of the
frozen proportion.
With a 5 cm layer of initially dry wood chips on top of the water, the final loss of mass (after
14 and 16 hours) decreased by approximately two-thirds, compared to corresponding
experiments without wood chips. This was because vapour condensed on the cold wood chips
after evaporating from the ice/water. This was observed when the can was taken out of the
freezer.
-8-
The loss of mass from initially wet wood chips decreased with frozen proportion, from
almost 20%, down to 2.5-3% (Figure 17). The results corresponded well to the findings
above. With a freezer temperature of -5 oC instead of –2 oC, the frozen quantity and the
relative loss increased. The frozen depth was 0-0.05 m.
There was no detected mass loss from the hygroscopic moisture (moisture content at
equilibrium with the air vapour) in the wood chips. No frozen (crisp) wood chips were found
in the experiments with hygroscopic moisture content.
20%
15%
FL/mice [%]
10%
5%
0%
0% 5% 10% 15% 20% 25%
mice/mmoist [%]
Figure 17 The mass loss and ice mass ratio as a function of the frozen proportion.
With initial hygroscopic moisture contents no mass loss due to freezing or
frozen wood chips were identified.
In the lab report (Skogsberg, 2001), it was suggested that the loss of mass due to freezing
could contribute to significant water loss from a snow storage area which was covered with
wood chips. The losses occur if melted water is transferred by capillary action to the wood
chips during daytime and lost due to freezing action during clear cold nights. Nevertheless,
the degree of surface melting during the spring is small, with a 0.2 m wood chip layer in the
Swedish climate. This means that the capillary transfer from the snow/wood chip interface is
small. In the spring, the moisture in the wood chips are more likely to originate from
condensed vapour. In other climates and circumstances the loss of melted water might be
considerable.
4 Conclusions
The loss of mass loss of freezing water contributes significantly to heat transfer. The
magnitude of the phenomenon named freezing loss depends mainly on the frozen water mass.
When all water was frozen, the losses were typically about 2.5-3% of the initial water weight,
corresponding to about 20-25% of the total energy transfer. This ratio corresponded that
found in reported studies. With a small portion of frozen water the relative loss of mass
increased, but too few experiments were conducted to enable the magnitude to be determined
in detail. The conclusion, however, is that if 100 g of water is frozen then the loss of mass
might range from 2.5 g to more than 10 g, depending on the total amount of water.
The freezing rate significantly influenced the rate of mass loss, but did not influence the
relative mass loss when a certain proportion of water was frozen. Water in a can froze faster
when the surface area was larger and when the water surface was closer to the edge of the
can. There was also a slight tendency towards larger losses in mass with water that was
initially warmer (7 oC instead of 1.5 oC). This was related to increased vapour content and
subsequent water particle mixture to approach the density maximum at +4 oC.
-9-
The relative mass loss from initially wet wood chips was of the same magnitude as that
obtained in the wet scouring cloth test, i.e., 3% when more than 15% of the wood chips were
frozen. It was concluded that the loss of mass might contribute to a significant loss of melted
water if the daytime melting rate is high and the wood chip layer freezes at night-time for
longer periods.
4.1 Sources of error

The main scale used was a large potential source of error. This problem was realised rather
late. During a 100-day test the scale failed on 3 occasions, when the recorded mass showed
discrepancies of up to twenty grams. There was also a slight, unavoidable pendulum
movement of the can plate that hung under the scale, but the result of this error was estimated
to be less than one gram.
Water had condensed on the can when it was removed from the freezer. The amount was less
than one gram. The ice mass was measured by drilling a hole in the ice and pouring out the
water, and approximately one gram of water was left in the can.
It was not possible to keep the temperature and humidity in the freezer constant.
Acknowledgements
I acknowledge our technician Mr Rolf Engström who never complained about repairing the
cans and helped in many other ways. I also acknowledge my supervisor Prof. Bo Nordell for
initiating this study and for his constructive comments on this report. I thank Prof. Anders
Sellgren for his constructive comments. This study was funded by Luleå University of
Technology and the Swedish Energy Agency (Energimyndigheten).
References
Nilsson J and Sjödin N (1986). Freezing losses from water. SMHI Hoh PM No 67,
Norrköping. (in Swedish)
Milanov T and Nilsson J (1988). Freezing losses from water. SMHI Hydrologi, No 20. (in
Swedish)
Pomeroy J W and Gray D M (1995). Snowcover. Accumulation, relocation and management.
National Hydrology Research Institute Science Report No. 7. ISBN 0843-9052.
Nordell B (1990). Measurement of P-T coexistence curve for ice-water mixture. Cold
Regions Science and Technology, 19 (1990) 83-88. Elsevier Science Publishers B.V.,
Amsterdam.
- 10 -

PDF Complet de La Technologie en Suède

Uploaded by

Copyright:

Available Formats

PDF Complet de La Technologie en Suède

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PDF Complet de La Technologie en Suède

Uploaded by

Copyright:

Available Formats

2005:30

Seasonal Snow Storage

Luleå University of Technology

2005:30|: -1544|: - -- 05⁄30 -- 

Doctoral Thesis 2005:30

Division of Architecture and Infrastructure

Akademisk avhandling för avläggande av teknologie doktorsexamen, som med vederbörligt

Doctoral Thesis 2005:30

Division of Architecture and Infrastructure

September 8, 2005 Kjell Skogsberg

Seasonal Snow Storage for Space and Process Cooling

Division of Architecture and Infrastructure

Report I Skogsberg K (2005). Capillary suction in cutter shavings and sawdust.

Seasonal Snow Storage for Space and Process Cooling

1 € = 9.1 SEK = 1.5 CD$

2 OBJECTIVE AND SCOPE

3 SEASONAL STORAGE OF SNOW AND ICE

Figure 3 Natural snow melt in open pond snow storage.

3.1 Techniques and examples

Figure 4 The All-Air-System with D control (from Kobiyama et al, 1997).

3.1.4 Under-ground storages

3.2 The Sundsvall Hospital snow cooling plant

3.2.2 Years of operation

Table 2 Melt water outlet data for 2001-2004.

June+July 2001 2002 2003 2004

Figure 12 Calculated and measured snow volume of 2001.

3.3 Boundary conditions for snow cooling

3.3.2 System boundary conditions

3.4 Snow and ice production

Figure 13 LowEnergyTower snow gun. Photo; Pär K Olsson, JL Toppteknik.

Figure 14 Fan type snow gun. Photo; Kjell Skogsberg.

3.5 Thermal insulation

4.1 Moisture and heat transfer

solid total pore volume

whygr wcr wcap wsat

4.1.2 Moisture transfer

į vap  v sat,air (T)

k suc  ȡ wat dp k suc  ȡ wat dp dw k suc  ȡ wat

In capillary transfer, moisture is transported upwards from a liquid surface by suction.

4.1.3 Heat transfer

(Re) is important. Nu is proportional to Re to the power of a positive number that depends

4.2 Surface melt

4.3 Rain melt

4.4 Ground melt

Table 8 Materials and properties used in the ground melt simulations.

4.5 Design model for a snow cooling plant

Top width of bank

Storage top diameter

Height above bank Top snow

Figure 20 Outline of snow storage pond in the calculation models.

5 MASS LOSS OF FREEZING WATER

6 CONSTRUCTION OF OPEN POND SNOW STORAGE SYSTEMS

6.1 Pond construction

6.2 Thermal insulation and cold carrier system

6.3 Snow and ice production

7.1 Environmental impact comparison

7.2 Snow Pollution

Kobiyama (1997) made an economic comparison between the All-Air air-conditioning

8.1 Pond construction

Assuming a cooling power of 2,000 kW and a heat exchanger temperature difference of

į vap v sat,air (T)

k suc ȡ wat dp k suc ȡ wat dp dw k suc ȡ wat

rmeasured rsurrounding 1 ' R2

V is the Stefan-Boltzmann constant 5.6710-8 W m-2 K-4, TK is the temperatures in [K]

where Ppor is a resistance coefficient [-],Ppor1. The vapour permeability coefficient of