5.0 7.

Article

Land Surface Temperature

Differences within Local Climate
Zones, Based on Two Central
European Cities

Jan Geletič, Michal Lehnert and Petr Dobrovolný

Special Issue
The Application of Thermal Urban Remote Sensing to Understand and Monitor Urban Climates
Edited by
Dr. Benjamin Bechtel, Dr. Iphigenia Keramitsoglou, Dr. Simone Kotthaus, Dr. James A. Voogt and
Dr. Klemen Zakšek

https://doi.org/10.3390/rs8100788
 remote sensing
Article
Land Surface Temperature Differences within Local
Climate Zones, Based on Two Central European Cities
Jan Geletič 1,2, *, Michal Lehnert 3 and Petr Dobrovolný 1,2
1 Department of Geography, Faculty of Science, Masaryk University Brno, Kotlářská 2, 611 37 Brno,
Czech Republic; dobro@sci.muni.cz
2 Global Change Research Institute of the Czech Academy of Sciences, Bělidla 986/4a, 603 00 Brno,
Czech Republic
3 Department of Geography, Faculty of Science, Palacký University Olomouc, 17. listopadu 12,
771 46 Olomouc, Czech Republic; michal.lehnert@gmail.com
* Correspondence: geletic.jan@gmail.com; Tel.: +420-541-211-214

Academic Editors: Benjamin Bechtel, Iphigenia Keramitsoglou, Simone Kotthaus, James A. Voogt,
Klemen Zakšek, Richard Müller and Prasad S. Thenkabail
Received: 20 June 2016; Accepted: 19 September 2016; Published: 22 September 2016

Abstract: The main factors influencing the spatiotemporal variability of urban climate are quite
widely recognized, including, for example, the thermal properties of materials used for surfaces
and buildings, the mass, height and layout of the buildings themselves and patterns of land use.
However, the roles played by particular factors vary from city to city with respect to differences in
geographical location, overall size, number of inhabitants and more. In urban climatology, the concept
of “local climate zones” (LCZs) has emerged over the past decade to address this heterogeneity. In this
contribution, a new GIS-based method is used for LCZ delimitation in Prague and Brno, the two
largest cities in the Czech Republic, while land surface temperatures (LSTs) derived from LANDSAT
and ASTER satellite data are employed for exploring the extent to which LCZ classes discriminate
with respect to LSTs. It has been suggested that correctly-delineated LCZs should demonstrate the
features typical of LST variability, and thus, typical surface temperatures should differ significantly
among most LCZs. Zones representing heavy industry (LCZ 10), dense low-rise buildings (LCZ 3)
and compact mid-rise buildings (LCZ 2) were identified as the warmest in both cities, while bodies
of water (LCZ G) and densely-forested areas (LCZ A) made up the coolest zones. ANOVA and
subsequent multiple comparison tests demonstrated that significant temperature differences between
the various LCZs prevail. The results of testing were similar for both study areas (89.3% and 91.7%
significant LST differences for Brno and Prague, respectively). LSTs computed from LANDSAT
differentiated better between LCZs, compared with ASTER. LCZ 8 (large low-rise buildings), LCZ 10
(heavy industry) and LCZ D (low plants) are well-differentiated zones in terms of their surface
temperatures. In contrast, LCZ 2 (compact mid-rise), LCZ 4 (open high-rise) and LCZ 9 (sparsely
built-up) are less distinguishable in both areas analyzed. Factors such as seasonality and thermal
anisotropy remain a challenge for future research into LST differences.

Keywords: land surface temperature; local climate zones; ASTER; LANDSAT; analysis of variance;
Prague; Brno; Czech Republic

1. Introduction
The main factors influencing the spatiotemporal variability of urban climate are quite widely
recognized, including, for example, the thermal properties of materials used for surfaces and buildings,
the mass, height and layout of the buildings themselves and patterns of land use (e.g., [1,2]). However,
urban climate research has long been limited by unsatisfactory quantification of these factors and

Remote Sens. 2016, 8, 788; doi:10.3390/rs8100788 www.mdpi.com/journal/remotesensing

Remote Sens. 2016, 8, 788 2 of 18

non-standardized descriptions of them [3]. To address these problems in urban climatology, the concept
of local climate zones (LCZs) has emerged over the past decade [4].
Stewart and Oke [4] define local climate zones as “regions of uniform land cover, surface structure,
construction material and human activity that span hundreds of meters to several kilometers on a
horizontal scale”. The thermal climate of each LCZ is determined by its surface character: structure
(building and tree height/density), cover (permeability), fabric (albedo, thermal admittance) and
metabolism (waste heat from transportation and space heating/cooling). Unique combinations of
these properties provide a distinctive thermal regime for each LCZ, particularly the characteristic
temperature regime at screen height, best observed in clear weather conditions and areas of simple
relief [4,5].
The concept of the LCZ emerged out of the need to improve the documentation of atmospheric
heat island observations, but its relation to spatial and temporal temperature variability has recently
become a key research question [4,6,7]. Most recent studies investigating urban temperature fields
utilize the LCZ classification and largely support the correspondence of LCZs with air temperature
fields in cities and their surroundings (e.g., [6,8–10]). However, a number of authors working with
LCZs have pointed out that the influence of thermal, radiative, metabolic and surface-cover properties
on the formation of local climate may significantly vary with respect to the geographical location of
the zone, the size of the city, the position within the city (distance from the city center, neighboring
zones) and relief [11–13]. Simultaneously, Bechtel and Daneke [14] and Lelovics et al. [9] created the
first LCZ mapping methods and applied the LCZ concept to secondary uses. With such an extension
of LCZ beyond its primary function, certain new methodological problems appeared: the size of a
spatial mapping unit, the method used for generalization, the temporal variability of the physical
properties of the environment, the objectification and standardization of the classification procedure,
and others [8,9,14]. The extension of LCZ to regional mapping significantly widens the field of LCZ
applications. For instance, Alexander et al. [15] revealed distinctive heat fluxes for some LCZs and
concluded that LCZ classification provides a useful sampling framework for the derivation of the
land-cover fractions needed to run surface-energy balance models. Zuvela-Aloise et al. [16] and
Geletič et al. [17] used LCZs as input to MUKLIMO_3, a non-hydrostatic micro-scale air temperature
model. Further, Skarbit et al. [18,19] worked on the relations between LCZ and LST. The authors show
that the different LCZ classes have different surface temperature characteristics as for their typical
values and their frequency distribution. The latter will be addressed in this study in more detail
because the statistical significance of the differences between the typical LST of climate zones will
be analyzed.
From the nature of LCZ classification, standard or special-purpose near-surface air temperature
measurements may be best suited to analyzing temperature contrasts among LCZs [6]. However, such
measurements, even supplemented by measurements from a dense network of automatic stations,
often prove insufficient to describe and understand fully the spatial variability of air temperature
in urban areas [17,20]. These measurements are insufficient especially in terms of their ability
to characterize temperature spatial distribution in detail, as the result may be dependent on the
number of “point” measurements, the method of interpolation and more. Thus LSTs derived from
various airborne or satellite remote sensing systems may be a suitable alternative, as their spatial
coverage is complete. Moreover, spatial and temporal resolution is constantly improving [21–23].
In spite of the fact that surface temperature modulates the air temperature of the boundary layer
atmosphere [21], relationships between the two types of temperature are quite complicated in urban
environments [24–26]. Apparent urban LSTs are modified by numerous static (e.g., land cover type,
urban morphology) and dynamic (e.g., geometry of thermal imagery acquisition, solar elevation)
factors. These factors not only modify the energy balance of urban areas [1,4], but they also generate
effects, such as thermal anisotropy [27]. Nevertheless, the use of land surface temperature for
the analysis of temperature spatial variability and the estimation of urban heat island intensity
remains challenging.
Remote Sens. 2016, 8, 788 3 of 18

From all of this, it becomes clear that the LST characteristics typical of LCZs (if such definitive
factors indeed exist) should be characterized in more detail in order to understand the influence of local
climate zones on local climate formation better. Case studies using remote sensing data to investigate
land surface temperatures (LSTs) from various cities may therefore contribute significantly to the study
of LCZs. The particular aim of this study is to analyze relations between surface temperatures and
local climate zones, taking two cities in the Czech Republic as examples. A new GIS-based method is
used for LCZ delimitation. Compared to the widely-used method of Bechtel and Daneke [14], this
method does not use thermal imagery, which is an important aspect as the paper explores the extent to
which LCZ classes are delineated with respect to land surface temperatures.

2. Materials and Methods

2.1. Study Area

This study focuses on Prague and Brno, the two largest cities in the Czech Republic. The capital
Prague, with around 1.3 million inhabitants, as well as Brno, with fewer than 0.4 million inhabitants,
may be considered as medium-sized cities typical of the central European region (Figure 1, Table 1).

Figure 1. Locations of Brno and Prague within central Europe and demarcation of the study areas with
cadastral borders of compact urban development.

Table 1. Basic data, study areas.

Size of Cadastral Number of Mean Latitude Longitude

Location ′ Center)
Study Area Area Inhabitants Elevation (City (City′ Center)
′ ◦ ′ ′
Brno and surroundings 25 × 25 km 8266 ha 400,000 259 m 49 12 N 16◦ 37′ E
Prague and surroundings 35 × 25 km 49,600 ha 1,275,000 288 m 50◦ 05′ N 14◦ 25′ E

In terms of urban morphology and relief, the cities are quite similar. Brno and Prague are both
located in hilly landscapes. Altitudes vary from under 200 m to over 500 m in Brno and under 200 m
to 400 m in Prague (mean elevations are 259 m for Brno and 288 m for Prague). Both cities have grown
in concentric zones, reflecting the historical stages of their development. Building patterns show this
clearly, and neither city lost significant proportions of their architectural heritage to the Second World
War (clearly-defined historical center, residential buildings old and new, industrial areas, housing
estates, modern shopping centers, malls and allotments). Because of this, neighborhoods with irregular,
as well as regular (grid and radial grid) street plans are present in both cities. Differences between
Remote Sens. 2016, 8, 788 4 of 18

the cities are especially the presence of a large river (the Vltava) in Prague and a higher proportion
of open vegetated (low plant) spaces in Brno. The landscapes beyond the city’s boundaries are of a
predominantly agricultural character, with agricultural areas represented by low plants (cultivated
plants like maize, barley, wheat or rapeseed; growing period beginning in April and harvest in August;
fields not irrigated). Patches of orchards (LCZ B) and mixed forests (LCZ A) complete the landscape
mosaics of the hinterlands (Brno also has large forested areas north of the city).
Mean annual temperature stands at 9.1 ◦ C for Brno and 8.4 ◦ C for Prague (1961–2000 reference
period). The annual temperature regime is characterized by a single maximum (the warmest month
being July with mean temperatures of 19.3 ◦ C in Brno and 18.2 ◦ C in Prague), with a single minimum
(the coldest month is January with a mean temperature −1.9 ◦ C in Brno and −1.4 ◦ C in Prague). Mean
annual precipitation totals are quite similar in both (505 mm for Brno and 526 mm for Prague) [28,29].

2.2. Local Climate Zones

A GIS-based approach was taken for the delimitation of LCZs [30]. This method had already
been developed and tested in the area of Brno and its surroundings. It is based on the measurable
physical properties of the environment and a clearly-defined decision-making algorithm. The algorithm
derives from the basic physical parameters defined by Stewart and Oke [4]: building surface fraction
(BSF), pervious surface fraction (PSF), impervious surface fraction (ISF) and height of roughness
elements (HRE). These were supplemented by number of buildings (NoB). For the classification
process, the study area was divided into a regular grid with cells of 100 m × 100 m, each cell exhibiting
these physical parameters.
In the first step, cells were divided by BSF into two basic classes: building types and natural land
cover types. In the second step, the cells in the first step categorized as building types (BSF ≥ 10%)
were classified into particular LCZs on the basis of the smallest deviation from the optimum interval,
as defined by Geletič and Lehnert [30] (against the background of values suggested by Stewart and
Oke [4]; see Table 2) in terms of physical parameters (BSF, PSF, ISF and HRE). Industrial zones (LCZs 8
and 10) were refined using NoB. Cells that had been categorized as land-cover types in the first step
were automatically classified into particular LCZs in the second step by means of the vector land use
geodatabase [30].

Table 2. Selected surface cover properties of the local climate zones valid for the central European
region ([4], modified); BSF, building surface fraction; ISF, impervious surface fraction; PSF, pervious
surface fraction; HRE, height of roughness elements.

LCZ Type BSF (%) ISF (%) PSF (%) HRE (m)
1 Compact high-rise 40–60 40–60 <10 >25
2 Compact mid-rise 40–70 30–50 <20 10–25
3 Compact low-rise 40–70 20–50 <30 3–10
4 Open high-rise 20–40 30–40 (30–50) 30–40 >25
5 Open mid-rise 20–40 30–50 (30–60) 20–40 10–25
6 Open low-rise 20–40 20–50 30–60 3–10
7 Lightweight low-rise 60–90 <20 <30 2–4
8 Large low-rise 30–50 40–50 <20 3–10
9 Sparsely built 10–20 <20 60–80 3–10
10 Heavy industry (40–70) 20–30 (30–60) 20–40 (<10) 40–50 (10–20) 5–15
A Dense trees <10 <10 >90 3–30
B Scattered trees <10 <10 >90 3–15
C Bush, scrub <10 <10 >90 <2
D Low plants <10 <10 >90 <1
E Bare rock or paved <10 >90 <10 <0.25
F Bare soil or sand <10 <10 >90 <0.25
G Water <10 <10 >90 −
Note: Values without brackets are defined by Stewart and Oke [4]; values in brackets indicate modified values
defined by Geletič and Lehnert [30].
Remote Sens. 2016, 8, 788 5 of 18

When all cells had been assigned to an appropriate LCZ, the LCZ areas were delineated using
a majority filter to smooth the results of classification. Independent testing of the decision-making
algorithm for defining the percentage coverage for individual LCZs demonstrated close agreement
with areas defined on the basis of expert knowledge (e.g., there was 90% agreement for Brno, not shown
here). New LCZ maps for Brno and Prague were created for this particular study.

2.3. Land Surface Temperature

Land surface temperature (LST) was obtained from two satellites: LANDSAT-8 and Terra,
sensor ASTER. LANDSAT-8 carries two instruments: an operational land imager (OLI) sensor and a
thermal infrared sensor (TIRS). The OLI sensor has nine bands (Bands 1–7 and 9 at 30-m resolution,
panchromatic Band 8 at 15-m resolution), while TIRS has two bands (Bands 10 and 11, collected at
100-m resolution and re-sampled to 30 m). ASTER consists of three separate instrument subsystems:
visible near-infrared (VNIR; Bands 1–3, 15-m resolution), short-wave infrared (SWIR; Bands 4–9,
at 30-m resolution) and thermal infrared (TIR; Bands 10–14, at 90-m resolution).
Altogether, 16 thermal images, eight for each city, four from LANDSAT-8 and four from ASTER,
were used for the comparison with LCZs. Basic scene information is summarized in Table 3. For the
interpretation of further analyses carried out in this study, it is important to emphasize that all thermal
images were recorded in the morning daylight hours (except 19 July).

Table 3. Satellite imagery used for local climate zone (LCZ) evaluation.

Time Cloud Solar Solar

City Scene ID Satellite Date
(UTC) Cover 1 Elevation Azimuth
BRNO AST_20020402 ASTER 2 April 2002 09:57:53 0% 44.250 159.398
BRNO AST_20090928 ASTER 28 September 2009 09:56:29 0% 37.792 165.303
BRNO AST_20130806 ASTER 6 August 2013 09:56:37 0% 54.506 153.383
BRNO AST_20140719 ASTER 19 July 2014 20:47:55 0% −14.399 328.189
BRNO LT8_20130618 LANDSAT-8 18 June 2013 09:46:54 0% 61.294 145.506
BRNO LT8_20130906 LANDSAT-8 6 September 2013 09:46:58 0% 45.369 156.246
BRNO LT8_20140520 LANDSAT-8 20 May 2014 09:44:27 0% 58.345 149.068
BRNO LT8_20150320 LANDSAT-8 20 March 2015 09:44:31 0% 38.296 154.861
PRAGUE AST_20060817 ASTER 18 July 2006 10:08:14 0% 51.457 156.961
PRAGUE AST_20070921 ASTER 21 September 2007 10:08:42 0% 39.606 164.853
PRAGUE AST_20080331 ASTER 31 March 2008 10:08:17 0% 42.738 160.003
PRAGUE AST_20110628 ASTER 28 June 2011 10:08:10 0% 60.817 151.613
PRAGUE LT8_20130515 LANDSAT-8 15 May 2013 09:58:53 0% 56.473 152.642
PRAGUE LT8_20130727 LANDSAT-8 27 July 2013 09:52:42 0% 55.845 148.504
PRAGUE LT8_20130803 LANDSAT-8 3 August 2013 09:58:54 0% 54.258 149.681
PRAGUE LT8_20150606 LANDSAT-8 6 June 2015 09:56:06 0% 59.396 148.417
1 Cloud cover for study areas, not full-scene.

Two algorithms were used for the estimation of LST: split-window and multispectral. The
split-window technique uses two thermal infrared (TIR) bands, typically located in the atmospheric
window between 10 µm and 12 µm [31]. In this study, the algorithm applied was based on the
estimation of top-of-atmosphere spectral radiance and at-satellite brightness temperature. LANDSAT-8
provides two thermal bands, Band 10 and Band 11 [32]. However, LANDSAT-8 does not provide data
for surface emissivity calculation. Because of this, a land-surface emissivity (LSE) algorithm was used,
estimating emissivity from the normalized difference vegetation index (NDVI) [33]. The scheme of
LANDSAT-8 scene processing is summarized in Figure 2.
Remote Sens. 2016, 8, 788 6 of 18

Figure 2. LST calculation from LANDSAT-8 OLI/TIRS images.

μ μ
μ μ µm [34]. Two different
The thermal infrared bands on ASTER are located between 8 µm and 12
approaches to the calculation of LST can be taken. The first exploits Level-1B data, which are produced
by applying radiometric calibration and geometric correction coefficients. The multispectral algorithm
for ASTER multiplies thermal bands by a scale factor, then atmospheric corrections are applied. The
normalized emissivity (NEM) algorithm was used for LST calculation [34]. A schematic representation
of this method appears in Figure 3. The ASTER product AST_08 (L2 surface kinetic temperature) may
also be downloaded, but this option has only been available since 1 April 2016.

Figure 3. LST calculation from ASTER thermal images.

Remote Sens. 2016, 8, 788 7 of 18

2.4. Comparison of Land-Surface Temperatures in Local Climate Zones

The analysis herein is based on the assumption that individual LCZs should demonstrate certain
features typical of a given LST regime. LST fields were overlaid with LCZs, and typical LSTs were
calculated for each zone. Differences between mean LCZ temperatures were evaluated by one-way
analysis of variance (ANOVA). Before applying ANOVA, a careful assessment of assumptions for the
application of this method was made. This comprised, in particular, testing for normality using the
Kolmogorov–Smirnov tests, Q-Q plots and comparisons of LCZ temperature variability. When the
ANOVA F-test indicated statistically-significant differences in LST, the Tukey HSD test was employed
to control for the effect of multiple comparisons [35] and to reveal which LCZs differentiate significantly
as to their mean LST. The analysis was performed separately for all 16 temperature fields constructed
(two study areas with eight temperature fields each). Finally, the score counting the number of tests
indicating significant temperature differences was calculated for each LCZ. Interpretation of this score
is straightforward: LCZs with higher scores are better differentiated from other zones in terms of their
typical LSTs. The other way round, LCZs with lower scores are less distinguishable from other zones
in terms of their typical LSTs.

3. Results

3.1. Local Climate Zones

The LCZ distributions exhibit concentric patterns, especially for Prague. The historical centers of
Brno and Prague form compact areas of LCZ 2 (Figures 4 and 5), clearly surrounded by LCZ 5, the most
frequent type of development in both cities. Towards the outer boundaries of compact development,
there is an increase in areas classified as LCZ 6. Together with LCZ 9, LCZ 6 is also typical of the
villages around the cities. An increased occurrence of LCZ B is typical of the boundary between
compact development and its surrounding landscape. LCZ B could, however, be found even near the
centers of both cities (parks). LCZs 8, 10 and E appear especially on the outer part of the cities; these
zones are frequently present in projections of development (city) into the surrounding countryside.
Nevertheless, there are also the sites of LCZ 8, 10 and E in the core areas of both Brno (Figure 4) and
Prague (Figure 5). LCZ D dominates in Prague’s surroundings, as it does in the landscape south of
Brno, but to the north of Brno lie large areas of LCZ A. Also in the latter area is a large body of water,
the Brno town reservoir, which creates a 259-ha area of LCZ G. Similarly, a line of LCZ G marks the
course of the River Vltava in Prague.

Figure 4. (a) Spatial distribution of LCZs and examples of (b) LANDSAT-8, scene LT8_20150606, and
(c) ASTER, scene AST_20110628; surface temperature variability in the Prague area; see Table 3 for
the scene ID. LCZ key: 2, compact mid-rise; 3, compact low-rise; 4, open high-rise; 5, open mid-rise;
6, open low-rise; 8, large low-rise; 9, sparsely built-up; 10, heavy industry; A, dense trees; B, scattered
trees; C, bush, scrub; D, low plants; E, bare rock or paved; F, bare soil or sand; G, water.
Remote Sens. 2016, 8, 788 8 of 18

Figure 5. (a) Spatial distribution of LCZs and examples of (b) LANDSAT-8, scene LT8_20130618, and
(c) ASTER, scene AST_20130806; LST variability in the Brno area; see Table 3 for the scene ID. LCZ
key: 2, compact mid-rise; 3, compact low-rise; 4, open high-rise; 5, open mid-rise; 6, open low-rise;
8, large low-rise; 9, sparsely built-up; 10, heavy industry; A, dense trees; B, scattered trees; C, bush,
scrub; D, low plants; E, bare rock or paved; F, bare soil or sand; G, water.

3.2. Land Surface Temperature

Brno and Prague show similar spatial distribution of surface temperatures. Industrial zones
and the city centers with historical buildings were always warmer compared to their surroundings
(Figures 4 and 5). Away from the densely built-up areas, the warm parts of the cities also occurred near
large patches of relatively flat, impervious surfaces (car parks, railway marshalling yards, stations, etc.).
Hotspots were very often associated with large commercial and distribution areas and the international
trade-fair grounds (which are quite extensive) regardless of whether they lay in the core of the city or
in the countryside. On the other hand, bodies of water (the Brno reservoir and Prague’s River Vltava)
and forested areas formed the coldest localities (see Figures 4 and 5).

3.3. Local Climate Zones and Land Surface Temperatures

The spatial distributions of LCZs and LST presented above indicate that there is a certain degree
of correspondence. Further analyses were therefore carried out to investigate their relations.
Box-plots summarizing typical LSTs for individual LCZs show relatively consistent results for the
two cities (Figures 6 and 7). However, the LSTs for LCZs derived from individual scenes cannot
be further compared directly as they represent different days within a year, a period spanning
20 March–28 September (Table 3). Particular LSTs arose not only out of static factors (e.g., land cover),
but were influenced by dynamic elements, such as the synoptic situation on a given day. Therefore,
LSTs of LCZs were compared in terms of rank order within a set of eight analyzed scenes for each
city. The mean LSTs of LCZs were sorted in descending order for each scene, and mean rankings were
calculated from all scenes individually.
The majority highest mean LST emerged for LCZ 10 (heavy industry), which was the warmest for
all eight scenes in Prague and for seven scenes in Brno. LCZ 3 (dense mix of low-rise buildings) and
LCZ 2 (compact mid-rise buildings) were the second- and third-warmest zones in Brno and Prague.
Next, higher LSTs were associated with LCZ 8 (large low-rise) in both study areas. The lowest ranking
and, thus, the coolest LCZ was related to bodies of water (LCZ G) followed by LCZ A, areas with
dense tree cover. Lower mean LSTs also occurred in other zones where pervious surfaces prevailed:
LCZ B (scattered trees) and LCZ C (mixture of shrubs, woody trees and agricultural areas).
The box-plots in Figures 6 and 7 present not only typical (median) LSTs for LCZs, but also permit
the evaluation of their variability and the occurrence of outliers. Whereas LCZ temperature variations
are more or less comparable and boxes and whiskers symmetrical, indicating no significant departures
Remote Sens. 2016, 8, 788 9 of 18

from normality, outliers typically occur in certain zones; they are analyzed further in the Discussion
section below. The normality of the LST distribution in each LCZ was further tested more formally,
since it is an important assumption for subsequent analysis of variance. A Gaussian curve was fitted to
the LST frequency distribution for each zone and differences addressed by the Kolmogorov–Smirnov
test. Figure 8 presents an example for a single date. Significant departures from normality occurred for
several zones in Brno (LCZ 10 and G) and one in Prague (LCZ 10).

Figure 6. Box-plots with LSTs in LCZ classes in Brno for ASTER (a–d) and LANDSAT-8 (e–h) images;
see Table 3 for the scene ID; the line within the box indicates median. The bottom of the box is the
first quartile, and the top is the third quartile. Whiskers represent the lowest value still within 1.5 IQR
(IQR = third quartile −−first quartile) and the highest value still within 1.5 IQR. Black crosses indicate
outliers. The top row with blue numbers indicates the order of LCZ according to the average LST
(1, the warmest zone; 14, the coldest zone).
Remote Sens. 2016, 8, 788 10 of 18

Figure 7. Box-plots with land surface temperature (LST) in LCZ classes in Prague for ASTER (a–d) and
LANDSAT-8 (e–h) images; see Table 3 for the scene ID; for the explanation, see Figure 6.
Remote Sens. 2016, 8, 788 11 of 18

Figure 8. Frequency distribution of land-surface temperatures in LCZ classes completed with a fitted
normal distribution and with the results of the Kolmogorov–Smirnov test. D stands for the testing
criterion, and p is the corresponding p-value. A p-value less than 0.05 indicates departure from
normality. The example features Prague, LANDSAT-8, 6 June 2015.

The results of the normality tests and comparable variations in LCZ land-surface temperatures
indicated that ANOVA may be used for testing whether the differences in LCZ mean temperatures
outlined above are significant or not.
Indeed, the results of the one-way ANOVA F-test (p < 0.001) determined that there exist significant
differences among the means of LCZ land surface temperatures. Moreover, these results stand for the
two study areas and for all 16 dates analyzed (scenes). Subsequently, Tukey HSD tests revealed which
pairs of LCZs significantly differentiate in terms of their mean LSTs (Figures 9 and 10).
Interpretation of all tests for both cities and analyzed scenes is provided in a simple binary format
in Figures 9 and 10. If the result of the test indicates no significant difference in average LSTs between
two zones in question (p-value higher than 0.05), there is a blue point in the corresponding cell of
the table. Blue points represent “negative” result in our analysis (“misses”), that is no difference in
average LSTs for a given pair of LCZs. In contrast, if the result of the test shows significant difference
(p-value lower than 0.05), there is an empty space in the corresponding cell of the table. Significant
differences in average LSTs represent “positive” results in this analysis and may be denoted as “hits”.
Specifically, for both study areas, LCZ pairs with statistically-significant differences in mean LCZ
temperatures (“hits”) prevail, while pairs with no temperature differences (“misses”) are less frequent,
and such LCZ pairs are different in the individual scenes analyzed. This result strongly supports the
method used for LCZ delimitation in both cities. That is, LCZs show typical surface temperatures
that are largely quite different from the surface temperatures of other zones. The robustness of this
finding is further confirmed by the fact that the relative number of “hits” is comparable for both cities
(89.3% for Brno and 91.7% for Prague). Slightly larger differences may be found between the LSTs
calculated from the two satellite data sources. Whereas ASTER-based LSTs indicate 89.2% hits, the
Remote Sens. 2016, 8, 788 12 of 18

LANDSAT-8-based LSTs show 91.9% hits. This implies that LSTs computed from LANDSAT-8 data
better differentiate among various LCZs compared to those from ASTER.
A closer look at multiple comparison results facilitates the identification of zones that are well
differentiated in terms of mean surface temperatures (Figure 11). The highest number of “hits”
(significant temperature differences) was registered for LCZ 8 (large low-rise), LCZ 10 (heavy industry)
and also for LCZ D (low plants). Such results confirm what is generally known of the spatial LST
distribution in the areas analyzed. LCZs 8 and 10 consist of areas where impervious surfaces with
lower proportions of vegetation cover prevail and where less surface moisture is available, features that
are responsible for the higher surface temperatures of these zones. In both Czech cities, industrial areas
are well delimited from neighborhoods that consist largely of LCZs with lower surface temperatures.
In spite of the fact that LCZ D occurs in rural areas in which low plants prevail, typical land surface
temperatures in this zone are markedly higher compared to their surroundings, due to the radiation
response of soil (Figures 4 and 5); see the Discussion section for details. LCZs with the lowest number
of “hits” (those less distinguishable in terms of typical surface temperatures) may be identified in
the same way. They stand for LCZ 2 (compact mid-rise), LCZ 4 (open high-rise) and LCZ 9 (sparsely
built-up) in both areas analyzed. While LCZs 4 and 9 are characterized by somewhat lower LSTs,
LCZ 2 represents zones with the highest mean surface temperatures; see Figures 6 and 7 and the
Discussion section below.

Figure 9. Results of the Tukey-HSD test for all combinations of LCZ classes (red numbers and letters)
and for all analyzed scenes in Brno; see Table 3 for the scene ID; blue points indicate pairs of LCZs for
which LSTs are not significantly different (p > 0.05), while empty cells denote LCZs pairs for which
average LSTs are significantly different.
Remote Sens. 2016, 8, 788 13 of 18

Figure 10. Results of the Tukey-HSD test for all combinations of LCZ classes (red numbers and letters)
and for all analyzed scenes in Prague; see Table 3 for the scene ID; blue points indicate pairs of LCZs
for which land surface temperatures are not significantly different (p > 0.05), while empty cells denote
LCZs pairs for which their average LSTs are significantly different.

Figure 11. Relative number of multiple comparison tests indicating significant differences (p < 0.05)
in LST (“hits”) among LCZs in Brno and Prague; LCZ F does not occur in Brno; dotted lines indicate
mean values for the two cities.
Remote Sens. 2016, 8, 788 14 of 18

4. Discussion
This study is based on satellite (ASTER and LANDSAT-8) observations; therefore, bird’s eye-view
surface temperatures from nadir (more precisely, nadir and very small off-nadir angles) are analyzed.
The results here cannot therefore be equalized with respect to the general thermal responses of any
particular LCZ, as vertical surfaces are neglected [20]; the results of this study thus provide a partial,
but coherent, insight into the nature of surface temperatures in LCZs. Some noteworthy particularities
and queries are addressed previously.
It was mentioned above that the highest LSTs were observed in LCZ 10 (heavy industry). However,
when interpreting the thermal behavior of LCZ 10, it must be taken into account that LCZ 10 delineation
was based on the Geletič and Lehnert [30] method, which employed a higher percentage of ISF and
BSF for delineation of LCZ 10 than that originally suggested by Stewart and Oke [4]. Nevertheless, the
higher land surface temperatures of industrial areas in daytime have previously been described by
many authors, e.g., [28,36,37]. Unexpectedly, the LSTs of LCZ 8 (large low-rise) were in most cases
considerably lower than the LSTs of LCZ 10. However, LCZ 8 was characterized by high statistical
variability in LST. This may correspond to the different thermal properties implicit in the large roof
surfaces that occur quite often in LCZ 8 [38]. LCZ A (dense trees) was identified as a lower LST
zone, but a large number of positive outliers are evident (i.e., highest LST). This is a reaction to
the relatively high number of patches, mostly consisting of clear-cuts of areas that are quite often
below the spatial resolution of the data used for LCZ classification. On the other hand, the high LST
variability of LCZ G (water) should be considered as realistic, since temperature differences reflect the
different characters of individual bodies of water (depth, rate of flow, etc.). The LST of LCZ D also
exhibited a relatively wide interquartile range with frequent outliers, especially for the upper part of
the distribution. Considerable temperature differences between LCZ D and other zones are especially
clearly expressed at the beginning and the end of the growing season, periods characterized by lower
proportions of vegetation and the stronger radiation response of bare soil (lower albedo, reduced latent
heat flux). This corresponds with the findings of [20,39,40] who described different spatial patterns of
LST for spring and autumn compared to summer due to seasonal changes in land cover. Therefore,
when working with LCZs, seasonal changes in land cover must be taken into consideration.
There were statistically-significant differences in mean LSTs between most LCZs, but LCZs 2, 4 and
9 were recognized as zones less distinguishable from other LCZs. A feature common to these three
zones is the higher heterogeneity of land cover typical of central European cities. For instance, LCZ 2
might exhibit a range of building patterns (e.g., street canyons, courtyards) of various orientations
to incoming solar radiation. This may significantly influence radiation response regarding both real
surface temperature differences and the magnitude of effective anisotropy [14,27]. Moreover, the LSTs
of LCZs 2 and 3 are similar because the percentages of PSF, ISF and BSF are also similar and the same
types of materials for infrastructure, roofing, etc., are used (at least in central European conditions).
One particular factor that differentiates these two zones is the height of the buildings; however,
considering the nadir (or very close to nadir) view position of the sensors, the effect of vertical surfaces
on apparent surface temperature is neglected, and simultaneously, dense development significantly
reduces increase in shaded area (in the sense of direct solar radiation) with building height [41].
This analysis has indicated certain differences between the two sensor data sources (ASTER
and LANDSAT-8). There may be several reasons for this. Each span leads to different LST retrieval
algorithms to be used for LST calculation from ASTER and LANDSAT, arising out of the varying
viewing geometry (IFOV (instantaneous field of view)) of the two satellite systems. In addition
to this, land surface temperatures may be biased by several other factors arising from the nature
of the remote sensing method. A higher proportion of vertically-oriented surfaces in urban areas
compared to rural environments causes uneven solar heating of those surfaces and induces a thermal
anisotropy effect [41]. Different solar heating of vertical surfaces arising out of variations in viewing
geometry renders the interpretation of land surface temperatures problematic [27], and the effect of
thermal anisotropy may influence the results of this study. According to Krayenhoff and Voogt [27],
Remote Sens. 2016, 8, 788 15 of 18

thermal anisotropy in an urban environment depends strongly on solar elevation and irradiance, and
it increases in tall, closely-spaced buildings. Further, compact and high-rise zones generate greater
anisotropy than open low-rise zones, while the regularity of street orientation increases anisotropy.
As follows from the description of the study areas (see Section 2) and also from the results of LCZ
mapping (Section 3), typical values of several of the above parameters should suffer less from the
thermal anisotropy effect. Very high and closely-spaced buildings are quite rare in Prague and Brno.
The average building height in the LCZ 2 zone of Prague is 19.9 m (typically between five and seven
stories) and 20.9 m in Brno (typically 5–6 stories). Compact mid-rise zones (LCZ 2) constitute 1.32%
and 0.31% of the Prague and Brno areas, respectively. Finally, streets are not regularly oriented in the
two cities in the fashion typical of some larger cities elsewhere, e.g., in North America.
Moreover, one may assume that the negative effect of thermal anisotropy is partly suppressed in
this study because eight different thermal images acquired from May–September were used. Different
dates of data acquisition have slightly different viewing geometry and, thus, partly “average” the
negative effect of thermal anisotropy. In spite of this, not only the bias due to thermal anisotropy,
but also other aspects of LST (e.g., atmospheric corrections, LST~air temperature relationships) should
be further clarified. However, these problems are beyond the scope of the current article.
It is important to stress that the same sets of warmest (coldest) LCZs were identified by both data
sources. Last, but not least, an important precondition for LST comparison between different zones is
mutual independence of LCZ classification and the surface temperature field. The GIS-based method
used for LCZ delimitation in this study met this requirement. This is an important methodological
aspect of this study with respect to widely-used imagery-based methods [14,42,43], including thermal
bands for LCZ delimitation.

5. Conclusions
This paper compared the spatial distribution of land surface temperatures with local climate zones
for the cities of Brno and Prague in the Czech Republic, employing data provided by two satellites.
The cities studied are characterized by a similar spatial distribution of local climate zones previously
constructed on the basis of a GIS method. This method utilizes data completely independent of those
used for LST derivation, a feature that permitted the use of LSTs for unbiased investigation of the
extent to which LCZs correspond with surface temperatures.
The key findings of this study may be summarized as follows: (i) regardless of the city studied or
the satellite data employed, LSTs show typical surface temperatures that differ significantly between
zones (in 89.3% and 91.7% of all tests applied for Brno and Prague, respectively); (ii) the warmest
(LCZ 10, heavy industry; LCZ 3, dense low-rise buildings; LCZ 2, compact mid-rise buildings)
and the coldest (LCZ G, water bodies; LCZ A, areas with dense trees) zones were identified in the
two study areas; (iii) LCZ 8 (large low-rise), LCZ 10 (heavy industry) and LCZ D (low plants) were
well differentiated in terms of surface temperatures; LCZ 2 (compact mid-rise), LCZ 4 (open high-rise)
and LCZ 9 (sparsely built-up) constitute less clearly distinguishable zones in both cities.
It may be concluded that our findings generally support the concept of LCZs and the GIS-based
method used for their delimitation for two typical central European cities of different sizes. Although
the concept of LCZ was developed for the classification of air temperature measurements, the results
of this paper confirm that individual LCZ prove also characteristic features of surface temperatures.
Such conclusions were indicated in other studies [18,19,44]. In combination with the use of the
GIS-based method, this LCZ-LST comparison may be considered the most important finding of this
this contribution. Certain questions, such as seasonality in LST differences and thermal anisotropy
(complete surface temperature differences), remain open to future research.

Acknowledgments: This contribution was prepared within the “UrbanAdapt—Development of urban adaptation
strategies using ecosystem-based approaches to adaptation” project, supported by Grant EHP-CZ02-OV-1-036-2015
from Iceland, Liechtenstein and Norway and by the Ministry of Education, Youth and Sports of CZ within the
National Sustainability Program I (NPU I), Grant Number LO1415. J.G. was also supported by the project
Remote Sens. 2016, 8, 788 16 of 18

MUNI/A/1315/2015. All ASTER images were retrieved from https://lpdaac.usgs.gov, maintained by the NASA
EOSDIS Land Processes Distributed Active Archive Center (LP DAAC) at the USGS/Earth Resources Observation
and Science (EROS) Center, Sioux Falls, South Dakota. The data product for the images was provided by NASA.
The authors would like to thank the four anonymous reviewers who helped improve the manuscript. Tony Long
(Svinošice) helped work up the English.
Author Contributions: J.G. performed the data analysis, prepared graphics and scripts and formatted the final
manuscript. M.L. wrote the first draft and worked with J.G. on the paper. P.D. interpreted the analysis results and
contributed to final text preparation. All authors contributed to the final manuscript and approved it.
Conflicts of Interest: The authors declare that there is no conflict of interest.

Abbreviations
The following abbreviations are used in this manuscript:
BSF Building Surface Fraction
ISF Impervious Surface Fraction
LCZ Local Climate Zones
LST Land Surface Temperature
PSF Pervious Surface Fraction

References
1. Arnfield, A.J. Two decades of urban climate research: A review of turbulence, exchanges of energy and
water, and the urban heat island. Int. J. Climatol. 2003, 23, 1–26. [CrossRef]
2. Grimmond, C.B.S.; Ward, H.C.; Kotthaus, S. Effects of urbanization on local and regional climate.
In The Routledge Handbook of Urbanization and Global Environmental Change, 1st ed.; Seto, K.C., Solecki, W.D.,
Griffith, C.A., Eds.; Routledge: London, UK; New York, NY, USA, 2016; pp. 169–187.
3. Stewart, I.D. A systematic review and scientific critique of methodology in modern urban heat island
literature. Int. J. Climatol. 2011, 31, 200–217. [CrossRef]
4. Stewart, I.D.; Oke, T.R. Local climate zones for urban temperature studies. Bull. Am. Meteorol. Soc. 2012, 93,
1879–1900. [CrossRef]
5. Stewart, I.D.; Oke, T.R. Local climate zones and urban climatic mapping. In The Urban Climatic Map:
A Methodology for Sustainable Urban Planning, 1st ed.; Ren, C., Ng, E., Eds.; Routledge: New York, NY, USA,
2015; pp. 397–401.
6. Stewart, I.D.; Oke, T.R.; Krayenhoff, E.S. Evaluation of the ‘local climate zone’ scheme using temperature
observations and model simulations. Int. J. Climatol. 2014, 34, 1062–1080. [CrossRef]
7. Gál, T.; Bechtel, B.; Unger, J. Comparison of two different local climate zone mapping methods.
In Proceedings of the 9th International Conference on Urban Climate, Toulouse, France, 20–24 July 2015.
8. Lehnert, M.; Geletič, J.; Husák, J.; Vysoudil, M. Urban field classification by “local climate zones” in a
medium-sized Central European city: The case of Olomouc (Czech Republic). Theor. Appl. Climatol. 2015,
122, 531–541. [CrossRef]
9. Lelovics, E.; Unger, J.; Gál, T.; Gál, V. Design of an urban monitoring network based on local climate zone
mapping and temperature pattern modelling. Clim. Res. 2014, 60, 51–62. [CrossRef]
10. Przybylak, R.; Uscka-Kowalkowska, J.; Araźny, A.; Kejna, M.; Kunz, M.; Maszewski, R. Spatial distribution
of air temperature in Toruń (Central Poland) and its causes. Theor. Appl. Climatol. 2015. [CrossRef]
11. Bokwa, A.; Hajto, M.J.; Walawender, J.P.; Szymanowski, M. Influence of diversified relief on the urban heat
island in the city of Kraków, Poland. Theor. Appl. Climatol. 2015, 122, 365–382. [CrossRef]
12. Leconte, F.; Bouyer, J.; Claverie, R.; Pétrissans, M. Using local climate zone scheme for UHI assessment:
Evaluation of the method using mobile measurements. Build. Environ. 2015, 83, 39–49. [CrossRef]
13. Lindén, J.; Grimmond, C.S.B.; Esper, J. Urban warming in villages. Adv. Sci. Res. 2015, 12, 157–162. [CrossRef]
14. Bechtel, B.; Daneke, C. Classification of local climate zones based on multiple earth observation data.
IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2012, 5, 1191–1202. [CrossRef]
15. Alexander, P.J.; Mills, G.; Fealy, R. Using LCZ data to run an urban energy balance model. Urban Clim. 2015,
13, 14–37. [CrossRef]
Remote Sens. 2016, 8, 788 17 of 18

16. Zuvela-Aloise, M.; Bokwa, A.; Dobrovolný, P.; Gál, T.; Geletič, J.; Gulyas, Á.; Hajto, M.; Hollosi, B.; Kielar, R.;
Lehnert, M.; et al. Modelling urban climate under global climate change in Central European cities.
In Proceedings of the EGU General Assembly 2015, Vienna, Austria, 12–17 April 2015.
17. Geletič, J.; Lehnert, M.; Dobrovolný, P. Modelled spatio-temporal variability of air temperature in an urban
climate and its validation: A case study of Brno (Czech Republic). Hung. Geogr. Bull. 2016, 65, 169–180.
[CrossRef]
18. Skarbit, N.; Gal, T.; Unger, J. Airborne surface temperature differences of the different local climate zones
in the urban area of a medium sized city. In Proceedings of the 2015 Joint Urban Remote Sensing Event
(JURSE), Lausanne, Switzerland, 30 March–1 April 2015; pp. 1–4.
19. Gémes, O.; Tobak, Z.; van Leeuwen, B. Satellite based analysis of surface urban heat island intensity.
J. Environ. Geogr. 2016, 9, 23–30. [CrossRef]
20. Dobrovolný, P.; Krahula, L. The spatial variability of air temperature and nocturnal urban heat island
intensity in the city of Brno, Czech Republic. Morav. Geogr. Rep. 2015, 23, 8–16. [CrossRef]
21. Voogt, J.A.; Oke, T.R. Thermal remote sensing of urban climates. Remote Sens. Environ. 2003, 86, 370–384.
[CrossRef]
22. Grimmond, C.B.S. Progress in measuring and observing the urban atmosphere. Theor. Appl. Climatol. 2006,
84, 3–22. [CrossRef]
23. Weng, Q. Thermal infrared remote sensing for urban climate and environmental studies: Methods,
applications, and trends. ISPRS J. Photogramm. 2009, 64, 335–344. [CrossRef]
24. Gallo, K.P.; Owen, T.W. Satellite-based adjustments for the urban heat island temperature bias.
J. Appl. Meteorol. 1999, 38, 806–813. [CrossRef]
25. Unger, J.; Gál, T.; Rakonczai, J.; Mucsi, L.; Szatmári, J.; Tobak, Z.; van Leeuwen, B.; Fiala, K. Air temperature
versus surface temperature in urban environment. In Proceedings of the 7th International Conference on
Urban Climate, Yokohama, Japan, 29 June–3 July 2009.
26. Schwarz, N.; Schlink, U.; Franck, U.; Großmann, K. Relationship of land surface and air temperatures and its
implications for quantifying urban heat island indicators—An application for the city of Leipzig (Germany).
Ecol. Indic. 2012, 18, 693–704. [CrossRef]
27. Krayenhoff, E.S.; Voogt, J.A. Daytime thermal anisotropy of urban neighbourhoods: Morphological causation.
Remote Sens. 2016, 8, 108. [CrossRef]
28. Dobrovolný, P.; Řezníčková, L.; Brázdil, R.; Krahula, L.; Zahradníček, P.; Hradil, M.; Doleželová, M.; Šálek, M.;
Štěpánek, P.; Rožnovský, J.; et al. Klima Brna. Víceúrovňová Analýza Městského Klimatu, 1st ed.; Masarykova
Univerzita: Brno, Czech Republic, 2012.
29. Tolasz, R.; Brázdil, R.; Bulíř, O.; Dobrovolný, P.; Dubrovský, M.; Hájková, L.; Halásová, O.; Hostýnek, J.;
Janouch, M.; Kohut, M.; et al. Atlas Podnebí Česka: Climate Atlas of Czechia, 1st ed.; Praha, Olomouc: Český
Hydrometeorologický Ústav, Czech Republic, 2007.
30. Geletič, J.; Lehnert, M. A GIS-based delineation of local climate zones: The case of medium-sized Central
European cities. Morav. Geogr. Rep. 2016, 24, 25–35.
31. Sobrino, J.A.; Li, Z.L.; Stoll, M.P.; Becker, F. Multi-channel and multi-angle algorithms for estimating sea and
land surface temperature with ATSR data. Int. J. Remote Sens. 1996, 17, 2089–2114. [CrossRef]
32. U.S. Geological Survey, Department of the Interior. LANDSAT 8 (L8) Data Users Handbook (Version 2.0).
2016. U.S. Geological Survey LANDSAT Missions Web Site. Available online: https://landsat.usgs.gov/
documents/Landsat8DataUsersHandbook.pdf (accessed on 30 May 2016).
33. Rozenstein, O.; Qin, Z.; Derimian, Y.; Karnieli, A. Derivation of land surface temperature for Landsat-8 TIRS
using a split window algorithm. Sensors 2014, 14, 5768–5780. [CrossRef] [PubMed]
34. Gillespie, A.R.; Rokugawa, S.; Hook, S.J.; Matsunaga, T.; Kahle, A.B. Temperature/Emissivity Separation
Algorithm Theoretical Basis Document (Version 2.4). 1999. NASA’s Earth Observing System Project Science
Office Web Site. Available online: http://eospso.nasa.gov/sites/default/files/atbd/atbd-ast-05-08.pdf
(accessed on 30 May 2016).
35. Livezey, R.E. Field intercomparison. In Analysis of Climate Variability: Applications of Statistical Techniques,
2nd ed.; von Storch, H., Navarra, A., Eds.; Springer: New York, NY, USA; Heidelberg, Germany, 1995;
pp. 161–178.
36. Roth, M.; Oke, T.R.; Emery, W.J. Satellite-derived urban heat islands from three coastal cities and the
utilization of such data in urban climatology. Int. J. Remote Sens. 1989, 10, 1699–1720. [CrossRef]
Remote Sens. 2016, 8, 788 18 of 18

37. Stathopoulou, M.; Cartalis, C.; Chrysoulakis, N. Using midday surface temperature to estimate cooling
degree-days from NOAA-AVHRR thermal infrared data: An application for Athens, Greece. Sol. Energy
2006, 80, 414–422. [CrossRef]
38. Akbari, H.; Levinson, R. Evolution of cool-roof standards in the US. Build. Energy Res. 2008, 1, 1–32.
[CrossRef]
39. Dobrovolný, P. The surface urban heat island in the city of Brno (Czech Republic) derived from land surface
temperatures and selected reasons for its spatial variability. Theor. Appl. Climatol. 2013, 112, 89–98. [CrossRef]
40. Wang, J.; Huang, B. Future urban climatic map development based on spatiotemporal image fusion for
monitoring the seasonal response of urban heat islands to land use/cover. In The Urban Climatic Map:
A Methodology for Sustainable Urban Planning, 1st ed.; Ng, E., Ren, C., Eds.; Routledge: New York, NY, USA,
2015; pp. 408–417.
41. Voogt, J.A.; Oke, T.R. Effects of urban surface geometry on remotely-sensed surface temperature. Int. J.
Remote Sens. 1998, 19, 895–920. [CrossRef]
42. Bechtel, B.; Alexander, P.J.; Böhner, J.; Ching, J.; Conrad, O.; Feddema, J.; Mills, G.; See, L.; Stewart, I. Mapping
local climate zones for a worldwide database of the form and function of cities. ISPRS Int. J. Geo-Inf. 2015, 4,
199–219. [CrossRef]
43. Danylo, O.; See, L.; Bechtel, B.; Schepaschenko, D.; Fritz, S. Contributing to WUDAPT: A local climate zone
classification of two cities in Ukraine. IEEE J. Sel. Top. Appl. Earth Obs. Remote Sens. 2015, 9, 1841–1853.
44. Lin, Z.; Xu, H. A study of urban heat island intensity based on “local climate zones”: A case study in
Fuzhou, China. In Proceedings of the 4th International Workshop on Earth Observation and Remote Sensing
Applications (EORSA), Guangzhou, China, 4–6 July 2016.

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).