This article has been accepted for inclusion in a future issue of this journal.

Content is final as presented, with the exception of pagination.

IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING 1

LiDAR-Based Approach for Urban Ventilation

Corridors Mapping
Marzena Wicht and Andreas Wicht

Abstract—Cities, recognized as one of the roughest surfaces, urbanization index nears 74% in Europe and is fairly stable [2].
significantly reduce wind velocity. This leads to a greater concen- Whereas aforementioned megacities attract a lot of debates, lots
tration of pollution and increased temperature in urban areas. A of other urbanized areas lack the publicity, and therefore, funds
proper urban design may promote the air flow and mitigate these
negative phenomena. This paper describes a novel approach to required to adopt settlements for climate change.
detect and analyze the potential urban ventilation corridors. The On the one hand, more and more people live in cities and
concept is based on the morphometric analyses, where roughness this share might swell, according to United Nations [1], to 66%
parameters (such as roughness length and displacement height), in 2050. On the other hand, Shepherd reports [3] that cities
as well as porosity are investigated as opposed to previous studies. occupy only 2% of land surface. As a result, we have approx-
Commonly available light detection and ranging (LiDAR) data en-
able more detailed terrain study, as the topography and obstacle imately 3.3 billion people [4] living in a relatively small area,
features (type, dimensions, and porosity) might be investigated. In which at the same time contribute approximately 80% of the
comparison to the previous studies and long adapted practices, this global greenhouse gas emissions [5]. Such density generates
approach takes advantage of all information incorporated in Li- numerous problems for our everyday living. The worsening the
DAR data to deliver even more accurate results. The comparison of sanitary conditions and elevated urban air temperature, relative
the detected corridors, with the officially designed corridors in our
study area in Warsaw, suggests inaccuracies in their extent (only to rural areas, had already been noticed at the beginning of 19th
40% of their areas might still be functioning) and calls for an eval- century by Howard [6] toward the end of the industrial revo-
uation. The results deliver areas that may be recognized as urban lution, which brought many factories and following workers to
ventilation corridors and serve as air regeneration and ventilation the cities. Since then many researchers have tried to find a solu-
system. They may also be used for developing new solutions for tion to mitigate the negative effects of the dense, urban living.
continuously evolving cities.
The idea of garden cities [7], green belts [8], or restoring nat-
Index Terms—Drag coefficient for vegetation, light detection and ural river path [9] have been developed over time, yet the need
ranging (LiDAR), morphometric methods, roughness parameters, to find better methods increases. More and more cities suffer
urban ventilation corridors.
from the urban heat island (UHI) both in tropical and temper-
ate zones [10]–[12] and the disturbed urban heat budget. These
I. INTRODUCTION problems exist due to the energy and water budget variations in
the built environment [13], as the urban fabrics, land cover, and
INCE 2014 more than half of human population lives in
S urbanized areas [1]. The urbanization index varies greatly
over the world, from 38% of sub-Saharan area to more than
metabolism are the most contributing factors. On average, an
urban citizen is faced up with higher air pollution and temper-
ature as well as decreased access to the wind speed and green
80% in North America. The pace of this process also varies in
areas relative to rural inhabitant. The latter have been proved to
different geographical regions. Southeast Asian urban regions
improve city’s climate and structure, as they are often a source
and their megacities, such as Tokyo, Delhi, or Shanghai, are one
of fresh and clean air and provide relaxation areas. On a city
of the fastest urbanizing areas in the world and, therefore, have
scale, if designed properly, they can also facilitate air flow and
caught more scientific attention. Recently, the research focus
penetrate the inner city with the fresh and cool air and discharge
has switched toward African agglomerations—as the next big
warm and polluted air to the outskirts. Since cities are one of
urban boom—yet the overall urban growth occurs elsewhere
the roughest aerodynamic boundaries, modeling and designing
as well. Europe, with stagnant or even declining population, is
such ventilation corridors pose many challenges and requires
getting older and at the same time more densely built-up. The
thorough investigation.
Surface roughness influences the wind velocity—the higher
Manuscript received September 28, 2017; revised November 14, 2017 and
December 25, 2017; accepted December 27, 2017. (Corresponding author: the roughness, the slower the wind. To describe and model this
Marzena Wicht.) component in urban context, several parameters are known in
M. Wicht is with the Department of Photogrammetry, Remote Sensing the literature. Roughness parameters illustrate how efficiently
and GIS, Warsaw University of Technology Warsaw, Warsaw 00-661, Poland
(e-mail: mwicht@gik.pw.edu.pl). wind travels above specified canopy layer [14], [15]. By calcu-
A. Wicht is with Geodata Science Team, GfK Geomarketing GmbH, Bruchsal lating and analyzing many urban aerodynamic indices, we can
76646, Germany (e-mail: andreas@wichts.net). detect areas, which—due to their roughness characteristics and
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. spatial allocation—promote air flow. Yet, their determination
Digital Object Identifier 10.1109/JSTARS.2018.2791410 poses many challenges, as there is no straightforward method to

1939-1404 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

2 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING

derive them. Altogether, we can divide urban scale aerodynamic also wins recognition [15], [29]–[31]. The method developed by
assessment methods into three categories [15], [16]: Millward-Hopkins has also been implemented in wind resource
1) reference (classification) based methods; calculation [32]–[34], where he also integrates light detection
2) micrometeorological (anemometric) methods; and ranging (LiDAR) and digital elevation model (DEM) data in
3) morphometric methods. order to increase accuracy and eliminate gross errors. Although
The reference method relates to the roughness estimation the method studies in detail wind velocity above the canopy
based on the earlier measurements. Davenport [14] and Wieringa layer, it is possible, that it might also be used in urban ven-
[17] created reference tables for different land-use classes—but tilation corridor mapping. While studying wind speeds and/or
for homogenous surfaces, whereas Grimmond and Oke [18] directions on a microscale, wind tunnel simulations and com-
focused on urban canopy. Unfortunately, the method requires putational fluid dynamics (CFD) models are used. Although
visual photography assessment, which may lead to subjective CFD studies consider wind velocities (which influences heat
results and diligence errors. In addition, they are severely limited transferred inside street canyons and courtyards) [35] and they
with wind direction inclusion. describe single buildings in detail, they are exclusively imple-
Multiple micrometeorological methods exist in the literature mented on very small areas due to high computational costs.
(detailed overview in [16]) and they use sensors located at dif- Various CFD models are used in engineering flow analysis,
ferent heights across the city, where the logarithmic wind profile building and structural design, pollution dispersion, and wind
is valid flow modeling [15]. They include physical models that try to
u ∗ z − zd represent the real urban setting, and therefore, model the wind
ūz = ln (1) flow along the streets and buildings. Even when they are em-
κ z0
ployed to map wind resources (with aid of LiDAR data), they
where ūz is the mean horizontal wind speed at height z, u∗ is the deliver more accurate results (in comparison to the analytical
friction velocity, z0 is the aerodynamical roughness length, zd is approach), however, at the expense of computational time [36].
the zero-displacement height, and κ = 0.40 is the von Karman Running these simulations requires power of supercomputers
constant [19]. Naturally, the logarithmic wind law applies during and remains still a challenge [37], [38].
neutral weather conditions and might be used approximately up In our previous studies, we adopted the approach proposed
to 200 m. Yet, this method requires high amount of observations in [15], [24], and [29], where we calculated the aerodynam-
[20] and is sensitive to errors. Moreover, the campaigns are very ical roughness length (z0 ) and zero-displacement height (zd )
time- and cost-intensive, and due to private property or local and conditioned their values (as suggested by [18] and [39])
laws, they might not be allowed in all desired locations. to retrieve areas of potential ventilation corridors. However, we
Numerous morphometric methods have been developed over noticed two essential features, which were being omitted: the
the years and each one has its own assumptions, preferred terrain topography and the vegetation. These concepts assumed
application, and limitations. Adolphe [21] developed a set of that the terrain is flat, and the vegetation does not obstruct air
indicators (density, rugosity, porosity, sinuosity, occlusivity, flow. For this reason, we decided to include these two features,
compacity, contiguity, solar admittance, and mineralization), by using building and vegetation database obtained from LiDAR
which he then used to derive urban climate characteristics. In data. We calculated the aforementioned parameters for both to-
comparison to prevailing urban climate studies, he does not pographies (flat terrain and digital surface model—DSM) and
base his calculations on (1), therefore, he does not derive dis- added an extra parameter—porosity, which has helped us to
placement height and roughness length, but refers to rugosity vary between bluff and porous bodies in order to model how
(absolute and relative). Although his assumptions take into con- they block the wind.
sideration the mean obstacle height and width depending on The main objective of the study was to compare the two ap-
the wind direction, the approach was tested just on a small proaches. One utilized in many studies, yet neglecting some fea-
neighborhood [21], [22] and not on city scale, and the results tures; second—a novel developed concept including topography
were not validated, which makes it harder to compare with and vegetation in possible urban ventilation corridor mapping.
other morphometric studies. Most of them investigate various The second approach also requires consideration on the condi-
implications of displacement height and roughness length (al- tioning values, since they were developed for the theoretically
together called as roughness parameters) on the urban climate flat terrain, which we also tested and discussed.
and they all derive those from logarithmic wind profile. Com-
monly known and referred to are methods developed by Rau-
pach [23], Bottema [24], Macdonald [25], Grimmond and Oke II. STUDY AREA
[18], Millward-Hopkins [26], and Kanda [27]. According to Similarly to our previous studies [30], [40], we located our in-
Kent [16], they can be split into two types based on the fact vestigation area in Warsaw for two reasons: it has historically de-
how do they treat obstacle height—as a mean for a given ref- signed ventilation corridors, so not only can we compare two ap-
erence idea, or as variable/maximum height. Kanda [27] and proaches, we can also contrast their results with the actual design
Millward-Hopkins [26] use this approach, as they argue Hm ax and use it for the evaluation. Second of all, Warsaw represents
is more suitable scaling parameter to map roughness element an average European metropolis—it lies in the temperate zone,
heterogeneity. Another possible method of division is the ref- which is one of the biggest cities in Europe, has a population of
erence area. Many studies utilize various type of mesh (regular 1.8 million inhabitants and almost twice as many city workers.
or adaptive) [20], [26], [28], whereas the “voronoi” approach It occupies more than 500 km2 and has a population density of
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

WICHT AND WICHT: LIDAR-BASED APPROACH FOR URBAN VENTILATION CORRIDORS MAPPING 3

3391 people/km² [41]. Also, the biggest river in Poland (Vistula

River) flows through the city for 28 km. in addition, Warsaw has
quite an interesting topography with clearly visible post-glacier
riverbed, which extends between 5 and 10 km within Warsaw
administrative boundaries. This riverbed makes the most iconic
topographic feature–Warsaw escarpment—present in everyday
lives of each inhabitant. It runs along the riverbed and the height
differences vary from 10 to 22 m and its inclination reaches 30°.
Such higher ground just above the vast river decided the city’s
location centuries ago and influenced greatly Warsaw’s spatial
development.
Although the annually average precipitation and air temper-
ature is considered typical for the temperate zone in Europe,
Warsaw also faces challenges due to the climate change. Sum-
mer flash flood events occur regularly and endanger the lives
of the inhabitants, the air pollution gradually worsens—lately,
the increase in smog caught a lot of attention, and the elevated
air temperature lowers the thermal comfort of living. Warsaw,
among many other cities, suffers from UHI that can get as high
as 9 °C [42]. An increasing number of these violent weather
events, as well as their randomness, urges local government to
develop and implement new strategies that might help adapting
to the climate change.
A. Warsaw Ventilation Corridors
Fig. 1. Designed ventilation corridors in Warsaw in 1992 and 2006.
These needs were noticed much earlier in Warsaw—at the
beginning of the 20th century—when population was getting
close to one million and the sanitary conditions were decreas- 6) Jerozolimski—has the best wind conditions and runs par-
ing. City planners were developing various ideas to increase tially through one of the biggest transportation valley—
standard of living. They proposed a radiocentric-linear growth Jerozolimskie Alley, not many supply areas.
model of Warsaw, which was supposed to slow down the urban 7) Western Railway—aids Jerozolimski corridor in air ex-
sprawl and provide enough relaxation areas. Several green belts change, occupies western railway infrastructure.
were supposed to prevent city-spreading and a system of urban 8) Bemowski—being connected to the Kampinowski
ventilation corridors (consisting of urban green areas) was to primeval forest and big share of vegetation and recreation
ventilate and regenerate the air in the city. The idea changed and areas supplies city with clean and cool air.
was updated through the turbulent socialist period in Poland and In 2006, the newer, updated Study of the Conditions and
was finally officially ratified in 1992. It determined nine ven- Directions of the Spatial Management of Warsaw (pol. SUiKZP,
tilation corridors, which were legally binding till 2006. These hereafter the Study) changed their extent. Even while preparing
were as follows. the Study, in 2001, the officials commissioned a comprehensive
1) Vistula river with adjacent Wilanowski corridor—28 km research evaluating the performance of the air ventilation system
long, mostly unoccupied water body with adjacent coastal (ventilation corridors), which clearly states how important it
regions (vegetation, meadows, etc.) and partially nonin- was to preserve and protect the corridors’ areas. Yet, the study
tensive housing mostly discharges the polluted and warm decided to diminish the Bródnowski corridor’s role by removing
air from the center. it from the official air ventilation and regeneration system and
2) Bródnowski—has rather small effect in the city-scale ven- substituting it with the Northern Railway corridor. Although it
tilation, due to the prevailing wind-directions (mostly is well supplied by the fresh and cool air coming from Jabłonna
western), but has a huge meaning locally, as it has one forests, it lacks continuity and goes through vast industrial areas
of the best supply areas (vast forests). (including CHP Żerań coal power plant). We can see the changes
3) Eastern Railway—runs partially through the eastern rail- of the air ventilation and regeneration system from 1992 to 2006
way infrastructure. in Fig. 1, and Table I presents the areawise modification of each
4) Podskarpowy—runs along Warsaw’s escarpment, its corridor in those two periods.
biggest role is discharging heated and polluted air from While the extent of the system was designed over 100 years
the city center. ago, urban planners did not have the access to the modern tech-
5) Mokotowski—starts with huge green or sparsely occupied nology and modeling possibilities. Even the evaluation analysis
areas of Warsaw’s Chopin Airport to reach one of the from 2001 was based on the land cover classification and evalu-
biggest green areas in Warsaw, over the years gradually ation as in [14]. That is why, we believe that this issue should be
built up. investigated using morphometric methods, as we have a unique
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

4 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING

TABLE I practice to use only first returns to construct the interpolating

VENTILATION CORRIDORS AREAS (SQUARE KILOMETER) IN 1992 AND 2006
surface for DTM/DSM generation. Usually, the first return is
the highest for an airborne survey, where the scanner is located
ID Name 1992 2006 Difference above the terrain. Using this first return means that some of
1+4 Vistula + Wilanowski 39.58 66.21 67% the detail is missing and not all data are used. This becomes
2 Bródnowski 16.05 0 −100% the problem with off-nadir scans (e.g., from unmanned aerial
3 East Railway 4.32 5.43 26% vehicle—UAV). Moreover, when there are clouds or high noise,
5 Podskarpowy 6.79 10.85 60%
6 Mokotowski 8.66 11.33 31% first returns are often discarded, and the remaining returns are
7 Jerozolimski 3.69 4.14 12% not renumbered, which leads to further mistakes. For this reason,
8 West Railway 6.01 3.65 −39% we implemented the algorithm proposed by Khosravipour [45],
9 Bemowski 9.07 7.19 −21%
10 North Railway 0.00 7.16 100% which triangulates all relevant LiDAR returns using constrained
Sum 105.34 115.96 Delaunay algorithm to derive a DSM.
In order to take the topography into account, we have ob-
tained the height attribute for each obstacle from the DSM.
possibility to assess the original design and see if there are However, while calculating it, we noticed a flaw in this simpli-
any other areas (previously omitted) that could aid the air flow. fied approach, where we calculated the roughness parameters
Particularly, since Warsaw’s air quality has been declining. with the true altitude values (in terms of elevation above sea
level). That is why we decided on the absolute altitude values
III. DATA AND METHODS for the two approaches—one with the absolute building height
(i.e., four-storey building ∼ 12 m) and one with true altitude in
A. Data terms with the local lowest point detected in Warsaw (i.e., two
The main objective of the study was to compare the commonly buildings with the same absolute building height, but laying on
used approach—in deriving potential urban corridors—with our top and below the escarpment would differ in height attribute
new concept, where we include the vegetation and topography, by the DEM value). The choice of these data ensured that we
as two essential variables in the wind flow calculation. In our account for topography and vegetation, as opposed to previous
previous approaches [30], [40], we utilized SPOT imagery to studies.
retrieve the building outlines and cadastral geodatabase for the
quality check and the height attribute. These two parameters— B. Methods
obstacles dimension and spatial alignment—are crucial in calcu- To describe the turbulent air motion during the neutral at-
lating roughness indices. Other studies have used various input mospheric stratification, the logarithmic wind law is used [46],
data: Gál and Unger [15] used a three-dimensional (3-D) build- which is based on the idealized wind speed profile relations
ing database available for Szeged. Bechtel [43] used SAR data (1) [28] and only two roughness parameters. Yet, as Kent and
and worked on a normalized DSM taking all possible obstacles Grimmond [28] report, exactly those morphometric methods
into account, yet he did not derive zo and zd parameters but resemble the most wind speeds estimated up to 200 m above
rather focused on topology, statistics, and texture of all rough- the canopy layer in central London, even with 25% mean wind
ness elements for a specific wind direction. Suder and Szy- speed uncertainty.
manowski [29] obtained building database from LiDAR aided In order to calculate these parameters, we adopted the ap-
with CityGML data, and Ketterer et al. [31] benchmarked three proach proposed by Gál and Unger [15] and created corre-
models based on 2.5-D cadastral database (one of which in- sponding lot-areas for each obstacle—including buildings and
cluded vegetation). Including topography and vegetation in the vegetation—for the whole city. Since a cadastral database could
roughness calculation is, therefore, not a standard yet. In order not be used and we believe that the grid approach (as in [11] and
to obtain this information, we need more precise data. [47]) is also flawed (due to the fact that cities are rarely based
For this reason, we acquired LiDAR data from the Cen- on a regular grid), we created lot-areas based on the Voronoi
tral Documentation Center of Geodesy and Cartography (pol. approach, as in Fig. 2(a) and (b). They were essential for the
CODGiK). Unfortunately, the latest available datasets were process, as we relate all parameters to these lot areas and their
from 2012, as the 2015 campaign is still in progress. The av- area accounts for AP —plan area (later used in plan area ratio).
erage pulse density for big agglomerations such as Warsaw is Therefore, to illustrate the texture (roughness) of an urban
12 p/m2 . The data supplied by the CODGiK were preclassi- surface, the aerodynamical roughness length (z0 ) and the zero-
fied (not classified; ground; low vegetation (0–0.4 m); medium plane displacement height (zd ) are used in this study. The ap-
vegetation (0.4–2 m); high vegetation; buildings; noise; water proach, modified after Bottema [24], [48], includes the calcula-
bodies) and for whole Warsaw resulted in more than 500 GB tion of frontal area index (used as an input for zd calculation),
of data. With such an amount of point-cloud data, the prepro- which is derived from
cessing software had to be chosen carefully. For this reason, n
we compressed the ∗.LAS to ∗.LAZ files, extracted building AF i
λF = i=1 (2)
and vegetation layers and calculated the DEM and DSM of AT
Warsaw with LAStools [44]. We also noticed the need to im- where AF represents frontal area (facade) of an obstacle, and
prove our terrain/surface model derivation, as it was a common AT —lot area.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

WICHT AND WICHT: LIDAR-BASED APPROACH FOR URBAN VENTILATION CORRIDORS MAPPING 5

Fig. 3. Lot, building, and vegetation statistics required for roughness param-
eters calculation.

these were focused only on building obstacles, which tend to be

wider than 5 m. It still is a compromise between accuracy and
computational time. We decided to decrease this interval to 3 m,
as we also wanted to detect the trees’ facades. The lines were
defined for eight (inter) cardinal wind directions, to take many
wind flow directions and different facades into account. λF was
calculated per each lot area corresponding to each obstacle for
eight directions.
Once the frontal area index was retrieved, we could calculate
roughness length and displacement height. The roughness length
describes the relationship between the wind and the drag force
and, as Bottema reports [24], [48], can be modeled by
 
κ
z0 = (h − zd ) exp − √ (3)
0.5CD λF
where κ—von Karman’s constant 0.4 [19], CD —drag coeffi-
cient, λF —frontal area index.
Zero-plane displacement height is the height at which wind
speed drops to zero because of obstacles in its way and is derived
from
zd = h(λP )0.6 (4)
Fig. 2. Development process of lot areas: (a) Calculating Voronoi polygons
for each polygon vertex; (b) computed lot areas; (c) lot areas overlaid with where h stands for volumetrically averaged building height and
parallel lines for λF computation.
λP is the plan-area ratio.
In general, roughness parameters demonstrate how efficiently
In order to derive λF , we had to determine facade length for wind travels above the specified canopy layer, which was cru-
different directions, as presented in Fig. 3. Depending on the cial for this study. However, in many studies, (3) is usually
wind direction, the frontal area of an obstacle varies. The same approximated and some constants are neglected and written in a
building or vegetation might block more wind from the east simpler form (as in [15]). We wanted to differentiate as much as
than from the southwest. To detect these facades for various possible between the porous (vegetation) and bluff (buildings)
wind directions, we superimposed our study area with a set of obstacles. Besides the porosity index, described below, the drag
parallel lines—Fig. 2(c). For each Voronoi polygon, a number created by vegetation varies in comparison to the bluff body
of “hits” on the facade is calculated and multiplied by the inter- with the similar structure. As Kent and Grimmond [28] argue,
val. Other studies [15], [30] used an interval of 5 m, however, this drag difference cannot be satisfied with simple frontal area
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

6 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING

Fig. 4. Spatial distribution of zero-displacement height (zd ), roughness length (z0 ) and porosity (P) for: (a) Flat terrain approach and (b) topography consideration.

porosity as the ratio of air volume to UCL volume to the same

reference area and derive it from
AT h − V
P = (5)
AT h
where h is the mean urban/vegetation height derived from the
entire study area, V is the volume of the building/vegetation. The
last parameter, volume, was also decreased, as the vegetation
volume is significantly lower in comparison to the building
volume. We approximated that a tree with similar dimensions
of a bluff body has only 40% of the original bluff body volume.
Fig. 5. Possible ventilation corridors in two approaches: Flat and with topog- Once all parameters (z0 , zd , and P) were calculated per lot
raphy consideration. area, we wanted to normalize them. Ratti et al. [50] suggest
averaging the results; however, we felt that the real wind shares
should be considered. Therefore, we averaged the results ac-
index decrease. It should rather settled separately for buildings cording to the prevailing winds, as seen in Fig. 1.
and vegetation. Although the drag coefficient itself is a complex 1) Potential Ventilation Corridor Mapping: Previous stud-
matter and is dependent on many variables (foliage type or a ies [39] suggest certain requirements of the potential ventilation
season), we decided to utilize an averaged value from other nu- corridors that a given area has to meet. Gál and Unger summa-
merical studies—0.2 [28] for the vegetation and 0.8 [15] for the rized it in [15] and for the purpose of this study, we considered
building layer. the most important three (for the roughness parameters):
Another parameter that we wanted to include in our modeling 1) roughness length lower than 0.5 m;
was the porosity value. As mentioned before, bluff and porous 2) negligible zero-plane displacement (lower than 3);
bodies influence wind and can penetrate investigated area dif- 3) at least 50 m width.
ferently. We believe that this index is of value, as the vegetation As for the porosity, the index takes values between 0 and
does not “block” the air flow completely, yet it slows it down 1, 0 meaning the flow is blocked and 1 the air flows without
significantly depending on the wind speed, vegetation density the obstruction. We tested several possibilities for the poros-
and type, and wind direction [49]. Gál and Unger [15] define ity threshold and after reviewing Kents [28] discussion on the
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

WICHT AND WICHT: LIDAR-BASED APPROACH FOR URBAN VENTILATION CORRIDORS MAPPING 7

TABLE II
POTENTIAL VENTILATION AREAS (KILOMETER) WITHIN CORRIDORS FROM
1992 AND 2006 IN TWO APPROACHES

Flat Topo

ID 1992 2006 1992 2006

1+4 16.57 41.7% 27.64 69.8% 14.86 37.5% 23.00 34.7%

2 5.55 0.0% 0.00 0.0% 5.47 34.1% 0.00 0.0%
3 0.99 28.9% 1.57 36.4% 0.34 7.9% 0.78 14.4%
5 1.59 25.2% 2.74 24.5% 1.10 9.9% 1.59 14.6%
6 5.07 61.4% 6.66 98.1% 3.52 51.9% 4.65 42.9%
7 1.44 16.0% 1.81 20.9% 0.84 9.8% 1.07 9.5%
8 1.82 34.2% 1.42 38.4% 1.01 27.3% 0.83 20.1%
9 2.41 60.7% 2.22 36.9% 1.73 28.9% 1.54 42.3%
10 0.00 16.5% 1.19 13.1% 0.00 0.0% 0.98 13.6%
Sum 35.45 45.24 28.88 34.45

This comparison shows clearly how sensitive the aerodynamical

roughness length is (zd ) to the height attribute, as the values in-
creased rapidly, once the absolute local altitude was calculated.
The zero-plane displacement height computation has shown sat-
isfactory results in variation between bluff and porous bodies—
large forests in the south and southwest part of Warsaw are
clearly visible, even in the topography approach.
The porosity values, on the other hand, show a slightly differ-
ent spatial pattern. The riverbed is still clearly visible, but some
areas show much better values in comparison to z0 and zd . It
gives us some additional information on the type of obstacle—
Fig. 6. Community gardens as a difference example between two approaches whereas roughness parameters focus mainly on the height. Still,
in detected ventilation corridors. the large woodland areas show decreased porosity, which was
expected.
various tree types and leaf on/off season, we averaged the value In potential ventilation corridors mapping, we applied the
to 0.6. conditions suggested by Matzarakis et al. in [39] and applied
In order to meet condition 3, we decided to transform the multiple times ever since, however, that study from 1992 was
Voronoi polygons into a 50 m × 50 m grid. Most studies simply also based on the relative building heights. It is visible, how the
average the Voronoi value depending on the share in grid, we conditions that were set are incredibly height-sensitive. Natu-
wanted however to move forward and distinguish between areas rally, to describe the geometry, or roughness of a surface, the
that are blocked or not. There is a certain flaw to the Voronoi ap- height of the obstacles is crucial. Still, so is the topography.
proach, which can be resolved in this step. Lot-areas are created When aforementioned (in Section III-B1) conditions were ap-
corresponding to each obstacle, which means that theoretically plied, we retrieved 102.9 km2 for the flat and 72.5 km2 for the
the whole city is covered with Voronoi polygons with computed topography approach, which sums up to 19.9% and 14% of
parameter values. It means that even some larger, open soil or the Warsaw area. All possible ventilation areas for the two ap-
lawns, meadows, etc., would, theoretically, be assigned an over- proaches are visible in Fig. 5. Altogether, a significant greater
estimated roughness value. Therefore, in the gridding step, we detection can be seen for the flat approach. When we investi-
excluded all of the grids that in fact do not intersect any of the gated some of the areas that increased in roughness and porosity
obstacles (buildings or vegetation) and appointed the value be- indices values in the flat approach, we concluded that the biggest
low the conditions specified in [39]. This allowed us to detect areawise differences were due to the community gardens, as seen
the potential ventilation corridors even more accurately. in Fig. 6.
Once we ran the calculation for both approaches (flat-terrain In comparison to our former studies, thanks to refining the grid
assumption and topographic), we were able to compare their outcome with obstacle-intersection information, we were able to
results. detect much smaller and finer corridors, which actually may help
with ventilation. Apart from the Jerozolimski corridor (whose
name originates from the Jerozolimskie Alley), other significant
IV. RESULTS
transportation routes have been detected in our current approach.
The visual evaluation of the spatial variability of the three In the northern part, we can see the S8 (E67) route contour, as
parameters is presented in Fig. 4. We can easily distinguish well as main railway axis—connecting corridor 3 and 8.
between the flat [see Fig. 4(a)] and the topography approach Generally, increased detection was also visible in the single,
[see Fig. 4 (b)], as the Vistula riverbed is easily recognizable. predefined corridors. As seen in Table II, the biggest differences
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

8 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING

between the approaches can be seen in corridor 6 (Mokotowski), height, and some larger parks and small huts are being detected
8 (West Railway), and 3 (East Railway). As we studied these as blocking area, or the porosity threshold was set too low. In ad-
changes, we also noticed that it was mostly high vegetation that dition, whereas for all corridors negative changes are noticeable
was excluded in the calculation. In Mokotowski corridor, the for the topographical approach (under-detection), it must not
difference can be as high as 22% and covers vast high vegeta- mean a flaw in the calculation. Perhaps, the previous approach
tion areas stretching from Chopin Airport up to the Wygledów
˛ was actually “over detecting” the potential ventilation areas, as it
housing estate. A very similar process is visible for both rail- did not account for the high vegetation effect. Regardless of the
ways corridors: in each case between 15% and 22% is lost to reason, we question the validity of the conditions proposed over
the high vegetation. 20 years ago, in an era, where we have relatively easy access
In general, the inclusion of vegetation in our calculation has to high accuracy DSM data and we can efficiently implement
changed the outcome a lot. In our previous approaches, such it. We believe that including the topographic information yields
as in [30] and [40], the minimum percentage of well-ventilated better results, they need, however, to be followed up by the in
areas in a corridor was not lower than 25% and for Vistula cor- situ measurements to assess the proper adjustment.
ridor it reached even 92%. Overall detection of well-ventilated Second of all, we plan to further investigate the porosity in-
areas within existing corridors reached more than 60%, whereas dex. In this study, we utilized the approach suggested by Gál
inclusion of topography, vegetation, and bluff/porous bodies de- and Unger [15] and varied the volume value (decreased by a
creased it to little over 40% (see Table II). On average, 60% of constant) for the vegetation and building layer. We are aware,
all corridors areas were covered with well-ventilated areas. This however, that porosity is highly complex matter and has many
has significantly decreased now, since even the Vistula corridor uncertainties—season (leaf-on versus leaf-off), vegetation type
efficiency lies between 34% and 42%, depending on the ap- (coniferous versus deciduous and deciduous species), density or
proach. Surprisingly, the best performing corridor—according depth of a patch in a given wind direction. Given the spatial vari-
to our analysis—is Mokotowski corridor. It has the highest share ability of large woodland areas and the fact that forest account
of well-ventilated areas. for 14% of whole Warsaw, we believe that more detailed inves-
One more negative process is visible from the data. In 2006, tigation of the high vegetation influence on the urban ventilation
the city council decided to withdraw from the Bródnowski (2) is highly required.
corridor at the gain of the Northern Railway corridor. Our anal- One very important matter must still be studied in greater de-
ysis has shown—independent from the approach chosen—that tail. Whereas our analysis delivered the overall share of the areas
whereas around 35% of Bródnowski corridor was still function- that might promote the air flow, we did not regard their dimen-
ing, it was only true for around 15% of the Northern Railway sions and continuity. The only aspect applied was the minimum
corridor. Even more important notice is that the latter does not width of 50 m. In our further studies, we plan to investigate the
have any access to the supply areas, which directly weakens its length of each patch (Matzarakis et al. [39] suggest the width has
performance. to be substantial, at least 1000 m in one direction). Second mea-
sure worth considering is connectivity with other potential cor-
ridors but as well with the supply fields. Some areas do show po-
V. DISCUSSION AND CONCLUSION tential (such as the southern part of the Northern Railway corri-
The results above show how we can use roughness and poros- dor), yet they lack connection to the supply fields (such as forests
ity parameters in defining potential urban ventilation corridors. or meadows) and, therefore, loose the most important feature
The proposed new approach includes vegetation—as an impor- of the ventilation corridor—bringing fresh and cool air into the
tant variable in the wind speed decrease—and the topography city. That is also directly connected with the investigation of
of the terrain. potential pollutants located in the direct neighborhood or in the
In our current study, we wanted to compare an existing ap- supply fields. Some areas in Warsaw are not connected to any of
proach (for over 20 years) in mapping potential ventilation cor- the CHPs and, therefore, large parts of the single-family houses
ridors with our new concept. In order to retrieve both vegetation there are using gas or coal stoves for heating. This phenomenon
and topography information, we utilized LiDAR data. We de- is even worse for the western municipalities outside Warsaw—
liberately used the local absolute altitude—treating the lowest such as Ożarów Mazowiecki or Pruszków. This poses potentially
detected point in Warsaw as local zero—as we wanted to in- an even bigger threat, since the prevailing wind directions for
clude topography in our calculation, yet did not want to work Warsaw are western. It could mean that ventilation corridors
in true altitude values. Such increase in height attribute value might transport even worse air to the city center. For this reason,
generated lower detection rate, in comparison to the flat ap- we want to expand our study area and locate the possible threats
proach. In many detailed investigated areas, it turned out, that and distinguish which areas are suitable for air regeneration.
the “blocked” areas (in flat detected as free) were community On the way to adapt to climate change, cities have devel-
gardens. Examples, Fig. 6(a)–(c) focuses on the three biggest oped many mitigation strategies—increasing surface albedo,
contrasting areas. The fact that these are community gardens decreasing impervious areas share or increasing urban green ar-
makes it worth considering—such areas are usually character- eas. Locating potential urban ventilation corridors bears many
ized by low roughness and one might think that they contribute advantages—wind flow increases our thermal comfort, sinks
overall to the air flow. This under-detection may result from the air temperature, disperses the air pollution and joins the
two issues: either the roughness parameters are too sensitive to aforementioned mitigation measures. In addition, such areas
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

WICHT AND WICHT: LIDAR-BASED APPROACH FOR URBAN VENTILATION CORRIDORS MAPPING 9

not only contribute to the city’s climate, they also increase our [19] U. Högström, “Review of some basic characteristics of the atmospheric
standard of living. We strongly believe that results from our surface layer,” in Boundary-Layer Meteorology 25th Anniversary Volume,
1970–1995: Invited Reviews and Selected Contributions to Recognise Ted
studies might be used by local planners, who may find any Munn’s Contribution as Editor Over the Past 25 Years, J. R. Garratt and
wastelands or unoccupied land and buildings along these paths P. A. Taylor, Eds. Dordrecht, The Netherlands: Springer-Verlag, 1996,
and secure them to promote the air flow in the city. Since urban pp. 215–246.
[20] C. S. B. Grimmond, “Aerodynamic roughness of urban areas derived from
climates are likely to worsen, and the number of urban inhabi- wind observations,” Boundary-Layer Meteorol., vol. 89, no. 1, pp. 1–24,
tants is rising, we should use any means necessary to change this 1998.
tendency. [21] L. Adolphe, “A simplified model of urban morphology: Application to an
analysis of the environmental performance of cities,” Environ. Plan. B,
Plan. Des., vol. 28, no. 2, pp. 183–200, 2001.
REFERENCES [22] Y. Gan and H. Chen, “Discussion on the applicability of urban morphol-
ogy index system for block natural ventilation research,” Procedia Eng.,
[1] United Nations, World Urbanization Prospects: The 2014 Revision. New vol. 169, pp. 240–247, 2016.
York, NY, USA: Department of Economics and Social Affairs, Population [23] M. R. Raupach, “Simplified expressions for vegetation roughness length
Division, United Nations, 2014. and zero-plane displacement as functions of canopy height and area index,”
[2] United Nations, “Urbanization: A majority in cities,” Population Fund, Boundary-Layer Meteorol., vol. 71, no. 1, pp. 211–216, 1994.
2007. [Online]. Available: https://www.unfpa.org/pds/urbanization.htm [24] M. Bottema and P. G. Mestayer, “Urban roughness mapping—Validation
[3] J. M. Shepherd, “A review of current investigations of urban-induced techniques and some first results,” J. Wind Eng. Ind. Aerodynamics, vol. 74,
rainfall and recommendations for the future,” Earth Interact., vol. 9, no. 12, pp. 163–173, 1998.
pp. 1–27, 2005. [25] R. W. Macdonald, R. F. Griffiths, and D. J. Hall, “An improved method for
[4] Worldometers, “Current world population,” 2014. [Online] Available: the estimation of surface roughness of obstacle arrays,” Atmos. Environ.,
http://www.worldometers.info/world-population/ vol. 32, no. 11, pp. 1857–1864, 1998.
[5] D. Hoornweg, “Cities and climate change: An urgent agenda,” Sustain. [26] J. T. Millward-Hopkins, A. S. Tomlin, L. Ma, D. Ingham, and
Low-Carbon City Develop. China, pp. 3–33, 2012. [Online]. Available: M. Pourkashanian, “Estimating aerodynamic parameters of urban-like
https://siteresources.worldbank.org/EXTNEWSCHINESE/Resources/ surfaces with heterogeneous building heights,” Boundary-Layer Meteo-
3196537-1202098669693/4635541-1335945747603/low_carbon_city_ rol., vol. 141, no. 3, pp. 443–465, 2011.
full_en.pdf [27] M. Kanda, A. Inagaki, T. Miyamoto, M. Gryschka, and S. Raasch, “A new
[6] L. Howard, The Climate of London, 3rd ed. London: W. Phillips, J. and aerodynamic parametrization for real urban surfaces,” Boundary-Layer
A. Arch, 1818. Meteorol., vol. 148, no. 2, pp. 357–377, 2013.
[7] H. Ebenezer, Garden Cities of Tomorrow. Cambridge, MA, USA: MIT [28] C. W. Kent, S. Grimmond, and D. Gatey, “Aerodynamic rough-
Press, 1984. ness parameters in cities: Inclusion of vegetation,” J. Wind Eng.
[8] M. Amati and M. Yokohari, “Temporal changes and local variations in Ind. Aerodynamics, vol. 169, pp. 168–176, 2017. [Online]. Available:
the functions of London’s Green Belt,” Landscape Urban Plan., vol. 75, http://www.sciencedirect.com/science/article/pii/S0167610516307346
no. 1, pp. 125–142, 2006. [Online]. Available: http://www.sciencedirect. [29] A. Suder and M. Szymanowski, “Determination of ventilation chan-
com/science/article/pii/S0169204605000101 nels in urban area: A case study of Wrocław (Poland),” Pure Appl.
[9] S.-G. Han, S.-H. Mun, and J.-H. Huh, “Changes of the micro-climate Geophys., vol. 171, no. 6, pp. 965–975, 2014. [Online]. Available:
and building cooling load due to the green effect of a restored stream in http://link.springer.com/article/10.1007/s00024-013-0659-9/fulltext.html
Seoul, Korea,” in Proc. Int. Building Perform. Simul. Assoc. Conf. Exhib., [30] M. Wicht, A. Wicht, and K. Osińska-Skotak, “Detection of ventilation
Beijing, China, 2007, pp. 1131–1138. corridors using a spatio-temporal approach aided by remote sensing data,”
[10] K. Klysik and K. Fortuniak, “Temporal and spatial characteristics of Eur. J. Remote Sens., vol. 50, no. 1, pp. 254–267, 2017. [Online]. Available:
the urban heat island of Lodz, Poland,” Atmos. Environ., vol. 33, http://www.tandfonline.com/doi/abs/10.1080/22797254.2017.1318672
pp. 3885–3895, 1999. [Online]. Available: http://www.sciencedirect. [31] C. Ketterer, M. Gangwisch, D. Fröhlich, and A. Matzarakis, “Compar-
com/science/article/pii/S1352231099001314 ison of selected approaches for urban roughness determination based
[11] M. S. Wong, J. E. Nichol, P. H. To, and J. Wang, “A simple on Voronoi cells,” Int. J. Biometeorology, vol. 61, no. 1, pp. 189–198,
method for designation of urban ventilation corridors and its applica- 2017.
tion to urban heat island analysis,” Building Environ., vol. 45, no. 8, [32] J. T. Millward-Hopkins, A. S. Tomlin, L. Ma, D. B. Ingham, and
pp. 1880–1889, 2010. [Online]. Available: http://www.sciencedirect. M. Pourkashanian, “Aerodynamic parameters of a UK city derived from
com/science/article/pii/S0360132310000776 morphological data,” Boundary-Layer Meteorol., vol. 61, pp. 447–468,
[12] A. Velazquez-Lozada, J. E. Gonzalez, and A. Winter, “Urban heat island 2013.
effect analysis for San Juan, Puerto Rico,” Atmos. Environ., vol. 40, no. 9, [33] J. T. Millward-Hopkins, A. S. Tomlin, L. Ma, D. B. Ingham, and
pp. 1731–1741, 2006. [Online]. Available: http://www.sciencedirect. M. Pourkashanian, “Assessing the potential of urban wind energy in a
com/science/article/pii/S135223100500909X major UK city using an analytical model,” Renewable Energy, vol. 60,
[13] P. Osmond and E. Sharifi, “Guide to urban cooling strategies,” Low pp. 701–710, 2013.
Carbon Living CRC, Univ. New South Wales, Tech. Rep. CRC LCL [34] J. T. Millward-Hopkins, A. S. Tomlin, L. Ma, D. B. Ingham, and
RP 2014, Jul. 2017. [Online]. Available: https://www.researchgate. M. Pourkashanian, “Mapping the wind resource over UK cities,” Re-
net/publication/318982885_Guide_to_Urban_Cooling_Strategies newable Energy, vol. 55, pp. 202–211, 2013.
[14] A. Davenport, C. Sue, B. Grimmond, O. Tim, and J. Wiering, “Esti- [35] Y. Toparlar et al., “CFD simulation and validation of urban microclimate:
mating the roughness of cities and sheltered country,” Proc. 12th Conf. A case study for Bergpolder Zuid, Rotterdam,” Building Environ., vol. 83,
Applied Climat., Asheville, NC, pp. 96–99, 2000. [Online]. Available: pp. 79–90, 2015.
https://goo.gl/yHCC93 [36] N. Lukač, G. Štumberger, and B. Žalik, “Wind resource assessment using
[15] T. Gál and J. Unger, “Detection of ventilation paths using high- airborne LiDAR data and smoothed particle hydrodynamics,” Environ.
resolution roughness parameter mapping in a large urban area,” Build- Model. Softw., vol. 95, pp. 1–12, 2017.
ing Environ., vol. 44, no. 1, pp. 198–206, 2009. [Online]. Available: [37] Y. Ashie, T. Tokairin, T. Kono, and K. Takahashi, “Numerical simulation
http://www.sciencedirect.com/science/article/pii/S0360132308000322 of urban heat island in a ten-kilometer square area of central Tokyo,”
[16] C. W. Kent et al., “Evaluation of urban local-scale aerodynamic param- Annual Report of the Earth Simulator Center, pp. 45–48, Apr. 2006.
eters: Implications for the vertical profile of wind speed and for source [Online]. Available: http://www.jamstec.go.jp/esc/publication/annual/
areas,” Boundary-Layer Meteorol., vol. 164, pp. 183–213, 2017. annual2006/pdf/3project/chapter1/045ashie.pdf
[17] J. Wieringa, A. Davenport, C. S. B. Grimmond, and T. R. Oke, “New [38] Y. Ashie and T. Kono, “Urban-scale CFD analysis in support of a climate-
revision of Davenport roughness classification,” in Proc. 3rd Eur. African sensitive design for the Tokyo Bay area,” Int. J. Climatol., vol. 31, no. 2,
Conf. Wind Eng., Eindhoven, The Netherlands, 2001, pp. 285–292. pp. 174–188, 2011.
[18] C. S. B. Grimmond and T. R. Oke, “Aerodynamic properties of urban [39] A. Matzarakis and H. Mayer, “Mapping of urban air paths for planning in
areas derived from analysis of surface form,” J. Appl. Meteorol., vol. 38, Munich,” Wiss. Ber. Inst. Meteorol. Klimaforsch. Univ. Karlsruhe, vol. 16,
no. 9, pp. 1262–1292, 1999. pp. 13–22, 1992.
 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination.

10 IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING

[40] M. Wicht and A. Wicht, Eds., “Urban ventilation corridors-performance Marzena Wicht received the M.Sc. degree in spa-
evaluation using remotely sensed data,” in Proc. 2017 Joint Urban Remote tial planning in 2012 from the Warsaw University
Sens. Event, IEEE, 2017, pp. 1–4. of Technology, Warsaw, Poland, where she has been
[41] Central Statistical Office (GUS), “Size and structure of the working toward the Ph.D. degree in research concen-
polish population.” [Online]. Available: http://demografia.stat.gov.pl/ trated on the spatial morphometric methodology to
bazademografia/Tables.aspx. Accessed on: Jul. 19, 2016. derive and evaluate urban ventilation corridors with
[42] K. Błażejczyk et al., “Urban heat island and bioclimatic comfort in War- the Department of Photogrammetry, Remote Sensing
saw,” in Counteracting Urban Heat Island Effects in a Global Climate and GIS, Faculty of Geodesy and Cartography, since
Change Scenario. Berlin, Germany: Springer-Verlag, 2016, pp. 305–321. 2012.
[43] B. Bechtel et al., “Towards an urban roughness parameterisation using She has also been involved with the Z_GIS De-
interferometric SAR data taking the metropolitan region of Hamburg partment of Geoinformatics, University of Salzburg,
as an example,” Meteorologische Zeitschrift, vol. 20, no. 1, pp. 29–37, Salzburg, Austria, and the Frankfurt University of Applied Sciences, Frankfurt,
2011. Germany. Her research interests include urban areas and processes that occur
[44] M. Isenburg, “LAStools-efficient tools for LiDAR processing,” 2012. [On- there, in particular, how violent weather events affect cities, how urban climate
line]. Available: http://www.cs.unc.edu/∼isenburg/lastools/. Accessed on: might be improved by appropriate spatial planning, and how remote sensing can
Oct. 9, 2012 be used in these cases.
[45] A. Khosravipour, A. K. Skidmore, and M. Isenburg, “Generating spike- Mrs. Wicht’s research work has gained an international approval expressed
free digital surface models using LiDAR raw point clouds: A new approach in one-year DAAD scholarship, and the Best Student Paper Awards at the Joint
for forestry applications,” Int. J. Appl. Earth Observ. Geoinformation, Urban Remote Sensing Event 2017 and the SPIE Remote Sensing 2017.
vol. 52, pp. 104–114, 2016.
[46] H. Tennekes, “The logarithmic wind profile,” J. Atmos. Sci., vol. 30, no. 2,
pp. 234–238, 1973.
[47] E. Ng, “Air ventilation assessment system for high density planning and
design,” in Proc. Passive Low Energy Archit. Int. Conf., Switzerland, 2006,
pp. 1–6.
[48] M. Bottema, “Urban roughness modelling in relation to pollutant dis-
persion,” Atmos. Environ., vol. 31, no. 18, pp. 3059–3075, 1997. Andreas Wicht received the M.Eng. degree in geoin-
[Online]. Available: http://www.sciencedirect.com/science/article/pii/ formation and public works from the Frankfurt Uni-
S1352231097001179 versity of Applied Sciences, Frankfurt, Germany, in
[49] T. G. Sachsen, “Die wirkung von vegetation in randstädtischen luftleitbah- 2012.
nen: Studien zur kaltluft in der stadt aachen,” Lehr-und Forschungsgebiet From 2011 to 2014, he held positions as a Project
Physische Geographie und Klimatologie, Achen, 2014. [Online] Avail- Engineer and Research Assistant with the Frankfurt
able: http://publications.rwth-aachen.de/record/229472 University of Applied Sciences. He is currently a
[50] C. Ratti, S. Di Sabatino, and R. Britter, “Urban texture analysis with part of the Geodata Science Team, GfK Geomarket-
image processing techniques: winds and dispersion,” Theor. Appl. Climat., ing GmbH, Bruchsal, Germany.
vol. 84, no. 1–3, pp. 77–90, 2006.