Journal of Hydrology 282 (2003) 104–115

www.elsevier.com/locate/jhydrol

Temperature index melt modelling in mountain areas

Regine Hock*
Department of Physical Geography and Quaternary Geology, Stockholm University, SE-106 91 Stockholm, Sweden

Abstract
Temperature index or degree-day models rest upon a claimed relationship between snow or ice melt and air temperature
usually expressed in the form of positive temperatures. Since air temperature generally is the most readily available data, such
models have been the most widely used method of ice and snow melt computations for many purposes, such as hydrological
modelling, ice dynamic modelling or climate sensitivity studies. Despite their simplicity, temperature-index models have
proven to be powerful tools for melt modelling, often on a catchment scale outperforming energy balance models. However,
two shortcomings are evident: (1) although working well over long time periods their accuracy decreases with increasing
temporal resolution; (2) spatial variability cannot be modelled accurately as melt rates may vary substantially due to
topographic effects such as shading, slope and aspect angles. These effects are particularly crucial in mountain areas. This paper
provides an overview of temperature-index methods, including glacier environments, and discusses recent advances on
distributed approaches attempting to account for topographic effects in complex terrain, while retaining scarcity of data input.
In the light of an increasing demand for melt estimates with high spatial and temporal resolution, such approaches need further
refinement and development.
q 2003 Elsevier B.V. All rights reserved.
Keywords: Temperature index models; Degree-day factor; Melt modelling; Glacier mass balance

1. Introduction process. Melt models generally fall into two cat-

egories: energy balance models, attempting to quan-
Melt modelling is a crucial element in any attempt tify melt as residual in the heat balance equation, and
to predict runoff from snow-covered or glacierised temperature-index models assuming an empirical
areas, as well as to assess changes in the cryosphere relationship between air temperatures and melt rates.
associated with climate change. In mountainous Such a relation was first used for an Alpine glacier by
regions, snow and ice significantly affect catchment Finsterwalder and Schunk (1887) and has since then
hydrology by temporarily storing and releasing water been widely applied and further refined (e.g. Clyde,
on various time scales (Jansson et al., 2003). This 1931; Collins, 1934; Corps of Engineers, 1956;
causes distinct annual and diurnal discharge variations Hoinkes and Steinacker, 1975; Braithwaite, 1995).
significantly differing from those of conventional Temperature index models have been the most
landscapes (Röthlisberger and Lang, 1987). Hence, common approach for melt modelling due to four
success of runoff modelling in such areas largely reasons: (1) wide availability of air temperature data,
depends on accurate quantification of the melt (2) relatively easy interpolation and forecasting
possibilities of air temperature, (3) generally good
* Tel.: þ46-8-16-47-84; fax: þ 46-8-16-48-18.
E-mail address: regine.hock@natgeo.su.se (R. Hock). model performance despite their simplicity and

0022-1694/$ - see front matter q 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0022-1694(03)00257-9
 R. Hock / Journal of Hydrology 282 (2003) 104–115 105

(4) computational simplicity. Applications cover a reason for the close relationship between melt and air
wide range including prediction of melt for oper- temperature. In addition, temperature is in part
ational flood forecasting and hydrological modelling affected by global radiation (Kuhn, 1993; Ohmura,
(WMO, 1986). Most operational runoff models, e.g. 2001), the secondary source of heat for melt.
HBV-model (Bergström, 1976), SRM-model (Marti-
nec and Rango, 1986), UBC-model (Quick and Pipes,
1977), HYMET-model (Tangborn, 1984) and even 3. Degree-day factors
versions of the physically based SHE-model (Bøggild
et al., 1999) rely on temperature-index methods for 3.1. Definition and influencing factors
melt modelling. Temperature index models also
provide the mass balance forcing for most ice dynamic Temperature index models or degree-day models
models (e.g. Oerlemans et al., 1998) and they have are based on an assumed relationship between
been used to predict the response of glacier mass ablation and air temperature usually expressed in the
balance to climate change (e.g. Bøggild et al., 1994; form of positive temperature sums. The most basic
Braithwaite and Zhang, 1999). This paper provides a formulation relates the amount of ice or snow melt, M
synthesis of the use of temperature-index methods for (mm), during a period of n time intervals, Dt (d), to the
melt modelling including glacierised areas. It is sum of positive air temperatures of each time interval,
concluded that temperature-index models need to be T þ (8C), during the same period, the factor of
enhanced in order to bridge the gap between restricted proportionality being the degree-day factor, DDF,
data availability and increasing demand for high expressed in mm d21 8C21.
resolution estimates of melt rates in space and time.
X
n X
n
M ¼ DDF T þ Dt ð1Þ
i¼1 i¼1
2. Physical basis of temperature-index models
Commonly, a daily time interval is used for
Many studies have revealed a high correlation temperature integration, although any other time
between melt and air temperature. Braithwaite and interval, such as hourly or monthly can also be used
Olesen (1989) found a correlation coefficient of 0.96 for determining degree-day factors. Reported degree-
between annual ice ablation and positive air tempera- day factors from glaciers and snow-covered basins
ture sums. Although involving a simplification of including site characteristics are summarized in Table
complex processes that are more properly evaluated 1, elaborating on previous similar tables by e.g.
by the energy balance of the glacier surface, Braithwaite and Zhang (2000) or Singh et al. (2000a).
temperature-index models often match the perform- Results reveal a large variability from site to site.
ance of energy balance models on a catchment scale Values are derived from different integration periods
(WMO, 1986). The reason for the success of air ranging from a few days (e.g. 3 days; Singh and
temperature as the sole index of melt energy in spite Kumar, 1996) to several years (e.g. 512 days over a 6-
of the predominance of net radiation as a source of year period; Braithwaite, 1995), limiting direct
melt energy is attributed to the high correlation of comparison. Degree-day factors are computed either
temperature with several energy balance components from direct measurements, using snow lysimeter
(Ambach, 1988b; Sato et al., 1984; Braithwaite and outflow (e.g. Kustas and Rango, 1994) or ablation
Olesen, 1990; Lang and Braun, 1990). Ohmura (2001) stakes (e.g. Braithwaite et al., 1998), or from melt
analyzed the physical basis of temperature-index obtained by energy balance computations (e.g. Arendt
models and stressed the role of longwave atmospheric and Sharp, 1999). For the same site, values might be
radiation: Usually, the latter is by far the most sensitive to the way they are derived, for instance,
important heat source for melt, and together with the how mean daily temperature is computed (Singh et al.,
sensible heat flux, provide about 3/4 of the entire 2000a) or which temporal average is used (Arnold and
energy source for melt. Both heat fluxes are highly MacKay, 1964). Use of daily temperature mean can be
affected by air temperature, which provides the main misleading during times of temperature fluctuations
 106
Table 1
Average degree-day factors (DDF) on snow and bare ice in mm d21 8C21

Site DDF snow DDF ice Latitude Altitude (m, a.s.l.) Period Reference

Glaciers
Alps/New Zealand/America
Aletschgletscher (Switzerland) 5.3 468270 N 3366 3 Aug– 19 Aug 1973 Lang, 1986
11.7 2220 2 Aug– 27 Aug 1965 Lang, 1986
Morenoglacier (Argentina) 7.1 508280 S 330 12 Nov 1993–1 Mar 94 Takeuchi et al., 1996
John Evans Glacier (Canada) 5.5 798400 N 260 27 Jun–29 Jun 1996 Arendt and Sharp, 1999
4.1 820 19 Jun–14 Jul 1996 Arendt and Sharp, 1999
3.9 820 23 May –1 Jul 1998 Arendt and Sharp, 1999
3.9 1180 25 Jun–19 Jul 1996 Arendt and Sharp, 1999
2.7 1180 31 May –19 Jul 1998 Arendt and Sharp, 1999

R. Hock / Journal of Hydrology 282 (2003) 104–115

7.6 260 4 Jul –16 Jul 1996 Arendt and Sharp, 1999
8.1 820 15 Jul–19 Jul 1996 Arendt and Sharp, 1999
5.5 820 2 Jul –19 Jul 1998 Arendt and Sharp, 1999
Scandinavia/Spitzbergen/Iceland
Alfotbreen (Norway) 4.5a 6.0a 618450 N 850–1400 1961–1990 Laumann and Reeh, 1993
Hellstugubreen (Norway) 3.5a 5.5a 618340 N 1450–2200 1961–1990 Laumann and Reeh, 1993
Nigardsbreen (Norway) 4.0a 5.5a 618410 N 300–2000 1961–1990 Laumann and Reeh, 1993
4.4a 6.4a 1964–1990 Johannesson et al., 1995
Storglaciären (Sweden) 3.2 678550 N 1550 5 Jul –7 Sep 1993 Hock, 1999
6.0 1370 5 Aug– 12 Aug 1993 Hock, 1999
6.4 1370 19 Jul–27 Aug 1994 Hock, 1999
5.4 1250 9 Jul –4 Sep 1994 Hock, 1999
Vestfonna (Spitzbergen) 13.8b ,808N 310–410 26 Jun–5 Aug 1958 Schytt, 1964
Satujökull (Iceland) 5.6a 7.7a ,658N 800–1800 1987–1992 Johannesson et al., 1995

Himalaya
Dokriani Glacier 5.9 318450 N 4000 4 Jun –6 Jun 1995 Singh and Kumar, 1996
5.7 7.4 4000 4 days (1997 –98) Singh et al., 2000a,b
Glacier AX010 7.3 8.1 278450 N 4956 Jun– Aug 1978c Kayastha et al., 2000a
8.7 8.8 5072 Jun– Aug 1978d Kayastha et al., 2000a
11.6 5245 1 Jun –31 Aug 1978 Kayastha et al., 2000a
Khumbu Glacier 16.9 288000 N 5350 21 May –1 Jun 1999 Kayastha et al., 2000b
Rakhiot Glacier 6.6 358220 N 3350 18 Jul–6 Aug 1986 Kayastha et al., 2000b
Yala Glacier 9.3 288140 N 5120 1 Jun –31 Jul 1996e Kayastha, 2001
10.1 5270 1 Jun –31 Jul 1996e Kayastha, 2001
Greenland
Thule Ramp 12.0b 768250 N 570 Jul 1954 Schytt, 1955
7.0b 570 Aug 1954 Schytt, 1955
Camp IV-EGIGf 18.6 698400 N 1013 Melt season 1959 Ambach, 1988a
Table 1 (continued)
Site DDF snow DDF ice Latitude Altitude (m, a.s.l.) Period Reference

GIMEXg profile 8.7 678060 N 341 10 Jun–31 Jul 1991 Van de Wal, 1992

R. Hock / Journal of Hydrology 282 (2003) 104–115

9.2 678060 N 519 15 Jun–6 Aug 1991 Van de Wal, 1992
20.0 678040 N 1028 15 Jun–6 Aug 1991 Van de Wal, 1992
Qamanarssup sermia 2.8a 7.3a 648280 N 370–1410 1979–1987 Johannesson et al., 1995
8.2 790 512 days (1980–86) Braithwaite, 1995
Nordboglacier 7.5 618280 N 880 415 days (1979–83) Braithwaite, 1995
Kronprins Christian Land 9.8 798540 N 380 8 Jul –27 Jul 1999 Braithwaite et al., 1998
Hans Tausen Ice Cap 5.9 828490 N 540 2 Jul –5 Aug 1994 Braithwaite et al., 1998

Non-glaciated sites
Gooseberry Creek, Utah 2.5 ,388N 2650 23 Apr–9 May 1928 Clyde, 1931
Weissfluhjoch 4.5 468480 N 2540 Snowmelt season Zingg, 1951
3 basins in USA 2.7– 4.9 Several seasons Corps of Engineers, 1956, p. 243
Former European USSR 5.5 7.0 1800–3700 Kuzmin, 1961, p. 117
12 Sites in Finland 2.8– 4.9 ,60–688N 1959–1978 Kuusisto, 1980
a
Best-fit values comparing degree-day model to measured net balance means.
b
Surface type not given, but probably partially snow.
c
Averaged over 47 and 45 days for snow and ice, respectively.
d
Averaged over 11 and 81 days for snow and ice, respectively.
e
At 5120 and 5270 m, a.s.l. 5 and 10% of ablation was snow ablation, respectively.
f
Expedition glaciologique international au Grönland.
g
Greenland Ice Margin Experiment.

107
108 R. Hock / Journal of Hydrology 282 (2003) 104–115

around the freezing point. Mean temperature may be

negative indicating no melt, whereas melt conditions
may have prevailed during part of the day. Hence, the
degree-day factor will be overestimated. Conse-
quently, a high time resolution for temperature
integration is desirable.
Most of the variation in degree-day factors
(Table 1) can be attributed to differences in relative
importance of individual energy components provid-
ing energy for melt, since energy balance character-
istics vary considerably in space and time. High
shares of sensible heat flux in the heat balance are
generally associated with low degree-day factors and
vice versa (Ambach, 1988a,b). Ambach (1988a)
attributed smaller degree-day factors at lower
elevations of the Greenland ice sheet to larger ratios
of sensible heat to melt energy resulting from higher Fig. 1. Schematic plot illustrating qualitatively how degree-day
factors are affected as variables increase. Degree-day factors
values of air temperature and wind speed compared to increase as solar radiation and elevation increase and as portion
higher elevations. Due to relatively large turbulent of sensible heat flux and albedo decrease.
fluxes, including condensation, maritime environ-
ments are likely to have lower degree-day factors
than more continental climate regions. For different reached only about 40% of that for ice, while the
reasons, low degree-day factors can also be expected corresponding percentage was about 70% on four
for dry high-radiation areas where sublimation plays a Scandinavian glaciers. Differences were inferred to
major role in the heat budget. Due to high energy reflect a smaller albedo contrast on the Scandinavian
consumption involved, sublimation reduces consider- glaciers than on the Greenland glaciers (Braithwaite,
ably the energy available for melt, and thus corre- 1996). Braithwaite (1995) also found evidence for
sponding degree-day factors (Lang and Braun, 1990). degree-day factors to depend on air temperature.
Such conditions can, e.g. be found at high elevations Higher values tended to occur at lower temperatures,
in the tropics and subtropics (e.g. Wagnon et al., indicating a non-linearity in the melt-temperature
1999), but temporarily also in mountain areas relationship. A detailed discussion on the complex
subjected to föhn- or chinook-events (Barry, 1992). interplay between environmental factors and degree-
During such events, despite relatively high air day factors can be found in Lang and Braun (1990).
temperatures, melt is subdued on the lee-side of the
mountains due to low air humidity favouring 3.2. Spatial variations
sublimation, as, e.g. observed on individual days on
Aletschgletscher, Switzerland (Lang et al., 1977). Despite the dependency of degree-day factors on
Energy partitioning will vary with different the relative importance of the sensible heat flux in the
climate, seasons and surfaces, resulting in a variation heat budget, regional patterns of degree-day factors
in degree-day factors, since these implicitly account are not detectable in the reported data (Table 1). This
for all terms of the heat budget. Under otherwise may be due to the complexity of the processes,
similar conditions, degree-day factors are expected to including counteracting mechanisms (Fig. 1). In
increase with increasing elevation, with increasing addition, any local signal might be subdued by
direct solar radiation input and with decreasing albedo variation resulting from different periods and inte-
(Fig. 1). Degree-day factors for snow tend to be gration times on which reported degree-day factors
considerably lower than those for ice, due to generally are based. Degree-day factors are subject to signifi-
higher albedo of snow compared to ice. On two cant small-scale variability. Braithwaite et al. (1998)
Greenlandic glaciers, the degree-day factor for snow compared degree-day factors obtained at several
 R. Hock / Journal of Hydrology 282 (2003) 104–115 109

ablation stakes spaced only a couple of metres apart threshold temperature accounts for the fact that melt
and found a ^ 10% variation between stakes. does not necessarily occur at air temperatures . 0 8C
(Kuhn, 1987) as well as for potential uncertainties in
3.3. Temporal variations air temperatures related to measurement and extra-
polation errors. On the other hand, Braithwaite (1995)
The degree-day factor can be expected to vary observed a tendency for melt to occur even when air
seasonally due to variations in direct solar radiation temperatures were zero.
and, in case of snow, due to metamorphic evolution of Many temperature-index based runoff models
the snow cover causing a decrease in albedo. Degree- employ a seasonally variable melt factor. For instance,
day factors over snow at 12 sites across Finland the UBC-runoff model (Quick and Pipes, 1977) adopts
tended to increase sharply in the beginning of April a monthly variable melt factor, while the HBV-ETH
and roughly doubled during this month due to snow model determines the melt factor from sinusoidal
ripening and associated albedo decline (Kuusisto, interpolation between a minimum value on December
1980). Over ice, seasonal variations in surface albedo, 21 and a maximum value on June 21 (Braun et al.,
and hence in degree-day factors, tend to be less 1993). Schreider et al. (1997) applied the IACRES-
pronounced. Braithwaite and Olesen (1993) detected runoff model with the degree-day factor varying
no evidence of distinct seasonal variation in degree- according to an albedo-related factor computed for
day factors, analysing six years of summer data over each month as a function of mean monthly tempera-
ice on Qamanârssûp sermia in Greenland, although ture. A dependency of the degree-day factor on albedo
monthly averages ranged from 6.2 to was also suggested by Arendt and Sharp (1999) in
9.0 mm d21 8C21. Despite uncertainties in determin- order to model melt at three sites on an Arctic glacier,
ing hourly melt rates, field studies indicate strong resulting in improved model performance compared to
diurnal variations in degree-day factors. Singh and application of a constant degree-day factor. Modelling
Kumar, (1996) report variations ranging from 0 to glacier runoff of Zongoglacier in Bolivia, Rigaudière
15 mm d21 8C21 roughly following diurnal cycles in et al. (1995) used a significantly lower degree-day
shortwave incoming radiation. factor in the dry season than in the wet season to
account for the pronounced seasonality in melt rates
that characterizes the glacier despite of low seasonal
4. Models temperature variability typical of the tropics.
Simple degree-day formulations are also a com-
4.1. Simple formulations mon tool to assess the sensitivity of glacier mass
balances to climate change (e.g. Laumann and Reeh,
In practise, the degree-day approach often assumes 1993; Bøggild et al., 1994; Braithwaite and Zhang,
the form 1999; Braithwaite and Zhang, 2000). Models are
( generally calibrated according to measured mass
fm ðTd 2 T0 Þ; Td . T0 balance gradients and then re-run while perturbing
M¼ ð2Þ
0; Td # T0 climate variables, mainly air temperature. Results
must be interpreted with caution as the inherent
(e.g. Gottlieb, 1980; Lang, 1986; Braun et al., 1994), assumption that degree-day factors remain constant
where M is daily melt, Td is daily mean temperature, under a different climate may not be true. However,
T0 is a threshold temperature beyond which melt is such studies can yield valuable information about
assumed to occur, and fm is a melt factor. It is regional differences in sensitivity (e.g. Braithwaite
emphasized that, when Td – T0 ; fm is not the same as and Zhang, 1999).
the degree-day factor, DDF, in Eq. (1). Therefore, I
suggest to use the term melt factor instead of degree- 4.2. Extended formulations
day factor in any formulations deviating from Eq. (1)
in order to avoid confusion with the original definition The fact that melt factors are influenced by all
of a positive degree-day factor. The inclusion of a components of the energy balance has prompted many
110 R. Hock / Journal of Hydrology 282 (2003) 104–115

attempts to improve the method by incorporating melt rates. However, these effects are not accounted
more variables, such as wind speed, vapour pressure for in the basic degree-day approach, as the degree-
or radiation components. Lang (1968) concluded from day factor is generally assumed constant in space,
multiple regression techniques that for hourly means although in glacier applications the factor might
inclusion of global radiation and for daily means depend on whether the surface is ice or snow (e.g.
inclusion of vapour pressure improves melt water Braun et al., 1993; Hock, 1999). In temperature-index
runoff computations. Also based on statistical anal- based melt-runoff models, spatial variation of melt
ysis, Zuzel and Cox (1975) suggested that daily snow rates across the basin is usually considered in a crude
melt estimates could be improved by including net manner. Elevation bands are often the only criterion
radiation, vapour pressure and wind speed rather than for spatial discretisation. Hence, melt rates will only
air temperature alone. vary as a function of elevation resulting from an air
There is a gradual transition from simple degree- temperature lapse rate. The HBV-ETH-runoff model
day approaches to energy-balance-type expressions improves upon this simplification by subdividing
by increasing the number of input variables into further into three aspect classes (Braun et al., 1994). A
model formulations. A widely quoted method has multiple factor, constrained by optimization pro-
been proposed by Anderson (1973). His so-called cedures, is then applied for each class to account for
combination method applies a simple degree-day enhanced melt on south-facing slopes and reduced
approach during dry periods and a simplified melt on north-facing slopes.
empirical energy balance formulation during rainy Only very few studies have attempted to apply
periods. The UBC-runoff-model (Quick and Pipes, temperature-index methods in a fully distributed
1977) and the HYMET-runoff-model (Tangborn, manner, allowing for spatially variable melt estimates
1984) employ the daily temperature range in addition on a grid (Table 2). Common to most of them is that
to air temperature as climatic input for their melt melt rates are computed as a function of a radiation
routines, as a measure of cloud-cover and, hence solar index, which heavily is affected by topographic
radiation. Several studies have added a radiation term effects, thus addressing the large variability of melt
to Eq. (2) in the general form rates in mountain terrain. Dunn and Colohan (1999)
divide the basin into five slope and three aspect
M ¼ fm T þ aR ð3Þ
classes (N, S, W/E) and vary the melt factor as a
where a is a coefficient and R is the shortwave function of snow albedo, rainfall rate and monthly
radiation balance (Martinec, 1989; Kane and Gieck, values of the ratio of direct solar radiation received by
1997) or net radiation (Martinec and de Quervain, the sloping surface to that received by a horizontal
1975; Kustas and Rango, 1994). These extended surface. The monthly values are derived from
expressions have usually been tested at the site scale published tables. Cazorzi and Fontana (1996) and
yielding better results compared to solely temperature Hock (1999) propose fully distributed temperature-
based simulations. E.g. Kustas and Rango (1994) index melt models computing variable melt rates for
obtained an increase in r 2 by nearly 40% when each grid element of a digital elevation model
moving from the basic degree-day approach to the incorporating topographic effects. Cazorzi and Fon-
extended expression, comparing daily model results tana (1996) used monthly raster maps of clear-sky
with observations. Brubaker et al. (1996) added net global radiation, each map obtained as average of
radiation to the temperature-index based SRM-runoff accumulated values between 21 December and the
model, but assumed net radiation uniform over the middle of each month.
basin. Hock (1999) proposed a model varying the melt
factor for every hour and each grid cell according to
4.3. Distributed temperature-index models the temporal and spatial variation of clear-sky direct
solar radiation. The latter can be computed from
It is well known that in mountain regions melt is standard algorithms on solar geography without the
heavily affected by topographic effects, such as slope, need of meteorological data using a digital elevation
aspect and shading, yielding high spatial variability in model. By adopting an hourly resolution, not only
 R. Hock / Journal of Hydrology 282 (2003) 104–115 111

Table 2
Spatially distributed temperature-index melt models

Melt factor is function of Area characteristics model is applied Spatial discretization Time step Reference
to

Monthly maps of clear-sky Upper Cordevole catchment, Italy, Grid-based (20 £ 20 m) Hour Cazorzi and Fontana,
global radiation 7 km2, 1815–3150 ma.s.l. 1996
Monthly values of ratio of Catchment on River Dee, Scotland 5 slope and 3 aspect Day Dunn and Colohan,
direct solar radiation received by 293 km2, 334–1309 ma.s.l. classes 1999
surface to that received by
a horizontal surface
Parameterized snow albedo
Rain fall rate
Hourly potential direct solar radiation Storglaciären (Sweden), 3 km2, Grid-based (30 £ 30 m) Hour Hock, 1999
1120–1730 ma.s.l.
Accumulated temperature index Sacramento and San Joaquin basins, Grid-based (2 £ 2 km) Hour Daly et al., 2000
California 180–9500 km2,
500–4400 ma.s.l.

spatial variability in degree-day factors is considered, distribution of melt rates and diurnal melt and
but also their strong diurnal variability is taken into discharge cycles. Fig. 2 compares runoff simulations
account, since direct radiation is subject to pro- using this model (b) to those based on the classical
nounced diurnal cyclicity. Hence, the model con- degree-day method (a), and a distributed energy-
siderably improves simulation of both, spatial balance model (c, Hock, 1998). While seasonal

Fig. 2. Hourly data of wind speed, u (m s21), air temperature, T (8C), precipitation, P (mm h21), simulated and measured hourly discharge, Q
(m3s21) of Storglaciären, Sweden, from July 11 to September 6, 1994. Melt calculations are based on classical degree-day method (a), modified
temperature-index model including potential direct solar radiation (b; Hock, 1999) and energy balance model (c; Hock, 1998).
112 R. Hock / Journal of Hydrology 282 (2003) 104–115

discharge variations are reasonably well modelled, it the relative contributions of energy balance com-
is obvious that the pronounced diurnal discharge ponents in the heat budget, which in turn vary with
cycles, typically for melt-affected regimes, are not weather and surface type. Thus, the classical degree-
properly resolved by the simple degree-day method. day method is only adequate for ‘average conditions’,
This aspect is considerably improved when using spatially defined to the catchment scale and tempor-
the modified temperature-index method, while only arily restricted to periods exceeding a couple of days
little additional improvement in model performance (Lang and Braun, 1990). Nevertheless, in connection
is achieved when adopting an energy balance with runoff models, temperature-index melt models
approach. generally yield good results even on a daily basis, as
Although strictly speaking not a temperature- daily deviations are smoothed by the basin response
index model, but working on similar grounds as the (Rango and Martinec, 1995). Since there is no
models described above, Williams and Tarboton universal numerical value, as evident from Table 1,
(1999) present a model to spatially distribute degree-day factors need to be adjusted to each
measured melt for every time step. Three com- application, and hence treated as calibration (tuning)
ponents of melt are distinguished: a spatially parameters within reasonable limits in runoff- and
uniform component, an elevation dependent com- mass balance models. Alternatively, predetermining
ponent, and one that is proportional to solar degree-day factors based on measurements or basin
radiation. For every time step snow melt measure- characteristics (Martinec and Rango, 1986) appears
ments at selected index-locations are related to problematic. Due to large small-scale variability
elevation and solar radiation by regression tech- (Hock, 1999), especially in mountain terrain,
niques. The resulting relationship combined with the degree-day factors obtained from point measurements
spatial pattern of direct radiation and elevation is can generally not be assumed representative on the
then used to compute spatially distributed melt over catchment scale. Deriving degree-day factors from
the entire catchment. physiographic basin characteristics might yield a
Daly et al. (2000) distribute the snow melt factor range of reasonable values, but not a single values
spatially according to an antecedent temperature considering the complexity of processes and inter-
index rather than a radiation index. The melt factor actions between atmospheric and surface character-
increases as a function of an accumulated temperature istics affecting the numerical value of the degree-day
index up to a threshold value beyond which the melt factor. Consequently, in degree-day driven run-off
factor remains constant. models, degree-day factors are more properly eval-
uated by optimization procedures using historic data
to minimize discrepancies between model results and
5. Concluding remarks observations.
The majority of proposed modifications to the
The degree-day method has been in use in many original degree-day method primarily aim at more
variants for more than a century, although more accurately capturing seasonal variations in degree-day
physically based energy balance models more prop- factors, basically adopting one of two strategies. (1)
erly account for the processes determining melt. Degree-day factors are varied directly as a function of
However, these are often not practical due to large time of year, or physical surface properties, such as
data requirements and uncertainties about spatial snow density or albedo. Most operational runoff
variability of some of these data. Hence, due to models employ such a seasonally variable degree-day
generally good performance, low data requirements factor and additional model parameters are usually
and simplicity, temperature-index methods are most introduced to contain its variation. (2) Degree-day
common, and will also in the future retain their factors are varied indirectly by adding more input
foremost position in snow and glacier melt modelling. variables, often radiation components. These
However, awareness of their limitations is necess- approaches lead to a gradual transition to simplified
ary. Degree-day factors exhibit considerable spatial energy balance methods. Such attempts have mainly
and temporal variability, since they depend on been tested at the point scale, often improving upon
 R. Hock / Journal of Hydrology 282 (2003) 104–115 113

simpler formulations. However, they are not easily References

transferable from the point to the catchment scale. A
daily time step is usually adopted in temperature- Ambach, W., 1988a. Heat balance characteristics and ice ablation,
index runoff modelling. However, such a time step is western EGIG-profile, Greenland, Seventh Northern Research
Basins Symposium/Workshop: Applied Hydrology in the
insufficient to accurately capture melt-induced diurnal Development of Northern Basins, May 25 –June 1, Copenhagen
discharge cycles or flood flows. Hence, models need Danish Society for Arctic Technology, Ililissat, Greenland, pp.
to be developed that account for the diurnal variability 59 –70.
in degree-day factors. Ambach, W., 1988b. Interpretation of the positive-degree-days
factor by heat balance characteristics—West Greenland. Nord.
While temporal variability of degree-day factors
Hydrol. 19, 217 –224.
has received considerable attention, it is surprising Anderson, E.A., 1973. National weather service river forecast
how little research has focussed on the development system/snow accumulation and ablation model, NOAA Tech-
of spatially distributed temperature-index models nical Memorandum NWS HYDRO-17, US Department of
specifically accounting for the large spatial varia- Commerce, Silver Spring, MD, 217 pp.
Arendt, A., Sharp, M., 1999. Energy balance measurements on a
bility in degree-day factors in mountain regions, in
Canadian high arctic glacier and their implications for mass
particular in steeply sided terrain. A reason might be balance modelling. In: Tranter, M., et al. (Eds.), Interactions
that degree-day melt modelling has traditionally Between the Cryosphere, Climate and Greenhouse Gases,
been driven by the need for operational runoff Proceedings of the IUGG Symposium, Birmingham 1999:
modelling with primary interest in basin runoff IAHS Publ. no. 256, pp. 165–172.
Arnold, K.C., MacKay, D.K., 1964. Different methods of calculat-
response and not in snow line retreat or spatially ing mean daily temperatures, their effects on degree-day totals
distributed melt estimates. However, in recent years in the high Arctic and their significance to glaciology. Geogr.
an increasing need for spatially distributed estimates Bull. 21, 123 –129.
of melt rates has been identified (Kirnbauer et al., Barry, R.G., 1992. Mountain weather and climate, second ed.,
1994). To cater to this demand, while retaining Routledge Physical Environment Series, London, 402 pp.
Bergström, S., 1976. Development and application of a
simple data input requirements, distributed tempera- conceptual runoff model for Scandinavian catchments.
ture-index models need to be developed. Some (Department of Water Resources Engineering, Lund Institute
recent attempts to vary degree-day factors in a fully of Technology/University of Lund, Bulletin Series A, No. 52,
distributed manner include potential direct solar 134 pp.).
Bøggild, C.E., Reeh, N., Oerter, H., 1994. Modelling ablation and
radiation, thus exploiting the close correspondence
mass-balance sensitivity to climate change of Storstrømmen,
between its spatial pattern with that of melt. Such northeast Greenland. Glob. Planet. Change 9, 79 –90.
approaches are intriguing as meteorological data Bøggild, C.E., Knudby, C.J., Knudsen, M.B., Starzer, W., 1999.
input is not enhanced, while yielding more realistic Snowmelt and runoff modelling of an arctic hydrological basin
spatial variations in melt estimates. These models in east Greenland. Hydrol. Proc. 13, 1989– 2002.
Braithwaite, R.J., 1995. Positive degree-day factors for ablation on
need further testing and refinement. Nevertheless, it
the Greenland ice sheet studied by energy-balance modelling.
seems, that inclusion of potential direct solar J. Glaciol. 41 (137), 153–160.
radiation into temperature-index models can Braithwaite, R.J., 1996. Models of ice–atmosphere interactions for
improve traditional methods with respect to spatial the Greenland ice sheet. Ann. Glaciol. 23, 149–153.
distribution and also diurnal variation in melt rates, Braithwaite, R.J., Olesen, O.B., 1989. Calculation of glacier
ablation from air temperature, West Greenland. In: Oerlemans,
without need for more data. J., (Ed.), Glacier Fluctuations and Climatic Change, Glaciology
and Quaternary Geology, Dordrecht, pp. 219–233.
Braithwaite, R.J., Olesen, O.B., 1990. Response of the energy
balance on the margin of the Greenland ice sheet to temperature
Acknowledgements changes. J. Glaciol. 36 (123), 217–221.
Braithwaite, R.J., Olesen, O.B., 1993. Seasonal variation of ice
Comments on the manuscript by Roger ablation at the margin of the Greenland ice sheet and its
Braithwaite, Andrew Fountain, Peter Jansson and sensitivity to climate change, Qamanârssûp sermia, West
Greenland. J. Glaciol. 39 (132), 267–274.
Herbert Lang are gratefully acknowledged. Thanks to Braithwaite, R.J., Zhang, Y., 1999. Modelling changes in glacier
A. Arendt, S. Daly, R. Kayastha and R. Van de Wal mass balance that may occur as a result of climate changes.
for providing additional information for the tables. Geogr. Ann. 81A (4), 489– 496.
114 R. Hock / Journal of Hydrology 282 (2003) 104–115

Braithwaite, R.J., Zhang, Y., 2000. Sensitivity of mass balance of Kane, D.L., Gieck, R.E., 1997. Snowmelt modeling at small
five Swiss glaciers to temperature changes assessed by tuning a Alaskan arctic watershed. J. Hydrol. Engng. 2 (4), 204–210.
degree-day model. J. Glaciol. 46 (152), 7–14. Kayastha, R.B., 2001. Study of glacier ablation in the Nepalese
Braithwaite, R.J., Konzelmann, T., Marty, C., Olesen, O.B., 1998. Himalayas by the energy balance model and positive degree-day
Errors in daily ablation measurements in northern Greenland, method. PhD Thesis. Graduate School of Science, Nagoya
1993–94, and their implications for glacier climate studies. University, 95 pp
J. Glaciol. 44 (148), 583 –588. Kayastha, R.B., Ageta, Y., Nakawo, M., 2000a. Positive degree-day
Braun, L.N., Grabs, W., Rana, B., 1993. Application of a conceptual factors for ablation on glaciers in the Nepalese Himalayas: case
precipitation-runoff model in the Langtang Khola basin, Nepal study on glacier AX010 in Shoron Himal, Nepal. Bull. Glaciol.
Himalaya. In: Young, G.J., (Ed.), Snow and Glacier Hydrology, Res. 17, 1– 10.
Proceedings of the Kathmandu Symposium 1992: IAHS Publ. Kayastha, R.B., Takeuchi, Y., Nakawo, M., Ageta, Y., 2000b.
no. 218, pp. 221– 237. Practical prediction of ice melting beneath various thickness of
Braun, L.N., Aellen, M., Funk, M., Hock, R., Rohrer, M.B., debris cover on Khumbu Glacier, Nepal, using a positive degree-
Steinegger, U., Kappenberger, G., Müller-Lemans, H., 1994. day factor. In: Nakawo, M., Raymond, C.F., Fountain, A. (Eds.),
Measurement and simulation of high alpine water balance Debris-Covered Glaciers, IAHS Publ. no. 264, pp. 71– 81.
components in the Linth-Limmern head watershed (North- Kirnbauer, R., Blöschl, G., Gutknecht, D., 1994. Entering the era of
eastern Switzerland). Z. Gletscherkd. Glazialgeol. 30, 161–185. distributed snow models. Nord. Hydrol. 25 (1/2), 1 –24.
Brubaker, K., Rango, A., Kustas, W., 1996. Incorporating radiation Kuhn, M., 1987. Micro-meteorological conditions for snow melt.
inputs into the snowmelt runoff model. Hydrol. Proc. 10, J. Glaciol. 33 (113), 263–272.
1329–1343. Kuhn, M., 1993. Methods of assessing the effects of climatic
Cazorzi, F., Fontana, G.D., 1996. Snowmelt modelling by changes on snow and glacier hydrology. In: Young, G.J., (Ed.),
combining air temperature and a distributed radiation index. Snow and glacier hydrology. Proceedings of the Kathmandu
J. Hydrol. 181, 169–187.
Symposium 1992: IAHS Publ. no. 218, pp. 135– 144.
Clyde, G.D., 1931. Snow-melting characteristics. Utah Agricultural
Kustas, W.P., Rango, A., 1994. A simple energy budget algorithm
Experiment Station Bull. 231, 1–23.
for the snowmelt runoff model. Water Resour. Res. 30 (5),
Collins, E.H., 1934. Relationship of degree-days above freezing to
1515–1527.
runoff. Trans. Am. Geophys. Union, Reports and Papers,
Kuusisto, E., 1980. On the values and variability of degree-day
Hydrol., 624 –629.
melting factors in Finland. Nord. Hydrol. 11 (5), 235–242.
Corps of Engineers, 1956. Summary report of the snow investi-
Kuzmin, P.P., 1961. Melting of snow cover, Israel Program for
gations, snow hydrology. US Army Engineer Division (North
Scientific Translation, 290 pp.
Pacific, 210 Custom House, Portland, Oregon), 437 pp
Lang, H., 1968. Relations between glacier runoff and meteorolo-
Daly, S.F., Davis, R., Ochs, E., Pangburn, T., 2000. An approach to
gical factors observed on and outside the glacier, IUGG General
spatially distributed snow modelling of the Sacramento and San
Assembly, Berne, International Association of Scientific
Joaquin basins, California Hydrol. Proc. 14, 3257–3271.
Dunn, S.M., Colohan, R.J.E., 1999. Developing the snow component Hydrology: IAHS Publ. no. 79, pp. 429–439.
of a distributed hydrological model: a step-wise approach based Lang, H., 1986. Forecasting meltwater runoff from snow-covered
on multi-objective analysis. J. Hydrol. 223, 1–16. areas and from glacier basins. In: Kraijenhoff, D.A., Moll, J.R.
Finsterwalder, S., Schunk, H., 1887. Der Suldenferner. Zeitschrift (Eds.), River Flow Modelling and Forecasting, D. Reidel
des Deutschen und Oesterreichischen Alpenvereins 18, 72– 89. publishing company, pp. 99 –127, Chapter 5.
Gottlieb,L., 1980. Development and applications of a runoff modelfor Lang, H., Braun, L., 1990. On the information content of air
snowcovered and glacierized basins. Nord. Hydrol. 11, 255–284. temperature in the context of snow melt estimation. In: Molnar,
Hock, R., 1998. Modelling of glacier melt and discharge. Zürcher L., (Ed.), Hydrology of Mountainous Areas, Proceedings of the
Geographische Schriften 70, Geogr. Inst. ETH Zürich, ISBN 3- Strbske Pleso Symposium 1990: IAHS Publ. no. 190, pp.
906148-18-1, 140 pp 347 –354.
Hock, R., 1999. A distributed temperature-index ice- and snowmelt Lang, H., Schädler, B., Davidson, G., 1977. Hydrological
model including potential direct solar radiation. J. Glaciol. 45 investigations on the Ewigschneefeld—Gr. Aletschgletscher.
(149), 101–111. Z. Gletscherkd. Glazialgeol. 12 (2), 109– 124.
Hoinkes, H.C., Steinacker, H., 1975. Hydrometeorological impli- Laumann, T., Reeh, N., 1993. Sensitivity to climate change of the
cations of the mass balance of Hintereisferner, 1952–53 to mass balance of glaciers in southern Norway. J. Glaciol. 39
1968–69, Proceedings of the snow and ice symposium, Moscow (133), 656 –665.
1971. IAHS Publ. no. 104, pp. 144–149. Martinec, J., 1989. Hour-to-hour snowmelt rates and lysimeter
Jansson, P., Hock, R., Schneider, T., 2003. The concept of glacier outflow during an entire ablation period. In: Colbeck, S.C.,
storage—a review. J. Hydrol.. (Ed.), Glacier and Snow Cover Variations, Proceedings of the
Johannesson, T., Sigurdsson, O., Laumann, T., Kennett, M., 1995. Baltimore Symposium, Maryland 1989: IAHS Publ. no. 183, pp.
Degree-day glacier mass-balance modelling with applications to 19 –28.
glaciers in Iceland, Norway and Greenland. J. Glaciol. 41 (138), Martinec, J., de Quervain, M.R., 1975. The effect of snow
345–358. displacement by avalanches on snow melt and runoff, Snow
 R. Hock / Journal of Hydrology 282 (2003) 104–115 115

and Ice Symposium. Proceedings of Moscow Symposia, 1971: Singh, P., Kumar, N., 1996. Determination of snowmelt factor in the
IAHS Publ. no. 104, pp. 365 –377. Himalayan region. Hydrol. Sci. J. 41 (3), 301 –310.
Martinec, J., Rango, A., 1986. Parameter values for snowmelt runoff Singh, P., Kumar, N., Arora, M., 2000a. Degree-day factors for
modelling. J. Hydrol. 84, 197– 219. snow and ice for Dokriani Glacier, Garhwal Himalayas.
Oerlemans, J., Anderson, B., Hubbard, A., Huybrechts, P., J. Hydrol. 235, 1–11.
Johannesson, T., Knap, W.H., Schmeits, M., 1998. Modelling Singh, P., Kumar, N., Ramasastri, K.S., Singh, Y., 2000b. Influence
the response of glaciers to climate warming. Clim. Dyn. 14, of a fine debris layer on the melting of snow and ice on a
267–274. Himalayan glacier. In: Nakawo, M., Raymond, C.F., Fountain,
Ohmura, A., 2001. Physical basis for the temperature-based melt- A. (Eds.), Debris-covered Glaciers, Proceedings of the Work-
index method. J. Appl. Meteorol. 40, 753–761. shop on Debris-covered Glaciers, Seattle, 2000: IAHS Publ. no.
Quick, M.C., Pipes, A., 1977. UBC watershed model. Hydrol. Sci. 264, pp. 63 –69.
Bull. 221, 153–161. Takeuchi, Y., Naruse, R., Skvarca, P., 1996. Annual air-temperature
Rango, A., Martinec, J., 1995. Revisiting the degree-day method for measurement and ablation estimate at Moreno Glacier,
snowmelt computations. Water Resour. Bull. 31 (4), 657– 669.
Patagonia. Bull. Glacier Res. 14, 23–28.
Rigaudière, P., Ribstein, P., Francou, B., Pouyaud, B., Saravia, R.,
Tangborn, W.V., 1984. Prediction of glacier derived runoff for
1995. Un modèle hydrologique du glacier du Zongo, Rapport
hydro-electric development. Geogr. Ann. 66A (3), 257–265.
No. 44, ORSTOM, Bolivie. 90pp.
Van de Wal, R., 1992. Ice and climate. PhD Thesis, Utrecht
Röthlisberger, H., Lang, H., 1987. Glacial hydrology. In: Gurnell,
University, 144 pp
A.M., Clark, M.J. (Eds.), Glacio-fluvial Sediment Transfer, An
Wagnon, P., Ribstein, P., Kaser, G., Berton, P., 1999. Energy
Alpine Perspective, Wiley, New York, pp. 207–284, Chapter 10.
balance and runoff seasonality of a Bolivian glacier. Glob.
Sato, A., Takahashi, S., Naruse, R., Wakahama, G., 1984. Ablation
and heat balance of the Yukikabe snow patch in the Daisetsu Planet. Change 22, 49 –58.
mountains. Ann. Glaciol. 5, 122–126. Williams, K.S., Tarboton, D.G., 1999. The ABC’s of snowmelt: a
Schreider, S.Y., Whetton, P.H., Jakeman, A.J., Pittock, A.B., 1997. topographically factorized energy component snowmelt model.
Runoff modelling for snow-affected catchments in the Aus- Hydrol. Proc. 13 (12–13), 1905–1920.
tralian alpine region, eastern Victoria. J. Hydrol. 200, 1–23. WMO, 1986. Intercomparison of models for snowmelt runoff.
Schytt, V., 1955. Glaciological investigations in the Thule Ramp Operational Hydrology Report 23 (WMO No. 646).
area. SIPRE Report 28, Snow, ice and permafrost research Zingg, T., 1951. Beziehung zwischen Temperatur und Schmelz-
establishment, Corps of Engineers, US Army, Wilmette, wasser und ihre Bedeutung für Niederschlags-und Abflussfra-
Illinois. 88 pp gen. IAHS Publ. no. 32, 266 –269.
Schytt, V., 1964. Scientific results of the Swedish glaciological Zuzel, J.F., Cox, L.M., 1975. Relative importance of meteorolo-
expedition to Nordaustlandet, Spitzbergen, 1957 and 1958. gical variables in snowmelt. Water Resour. Res. 11 (1),
Geogr. Ann. 46 (3), 243–281. 174 –176.