Methods to assess heat exposure: A comparison of fine-scale approaches within the German city of Karlsruhe

UCLIM-00260; No of Pages 13 Urban Climate xxx (2016) xxx–xxx

Contents lists available at ScienceDirect

Urban Climate journal homepage: http://www.elsevier.com/locate/uclim

Methods to assess heat exposure: A comparison of fine-scale approaches within the German city of Karlsruhe Mady Olonscheck ⁎, Carsten Walther Potsdam Institute for Climate Impact Research, P.O. Box 601203, 14412 Potsdam, Germany

a r t i c l e

i n f o

Article history: Received 14 March 2016 Received in revised form 13 November 2016 Accepted 4 December 2016 Available online xxxx Keywords: Heat wave day Heat exposure City quarter Multiple regression analysis Population density Cross validation

a b s t r a c t Knowledge on the most exposed areas of a city constitutes an important basis for suitable short and long-term planning. We present and compare three different methods that allow us to assess the potential heat exposure for the smallest administrative spatial units, the quarters, in the German city of Karlsruhe which was repeatedly affected by heat waves in the past. The three methods are based on (1) meteorological station data from the city and its hinterland, (2) a city climate index and (3) remote sensing data. The aim is to answer the question whether different approaches provide different levels of heat exposure. By comparing the three methods we could identify regions by cross validation where the level of heat exposure is highly confident. Regions where one model result deviates from that of another, give interesting insights in the interrelation of features of the method and circumstances in the study area. Regions were all the three models showed different results remained very rare. The results may be relevant for decision-makers who want to implement small-scale measures for heat mitigation but only have limited resources available. © 2016 Elsevier B.V. All rights reserved.

1. Introduction Extended periods of intense heat can have severe impacts on people and infrastructures. During the summer of 2003, large areas of Western Europe suffered from extraordinarily high temperatures. In Europe as a whole, 70,000 people (including 7000 in Germany) died due to the stresses and strains of this heat wave (Robine et al., 2007). Older people (Hondula et al., 2012; Medina-Ramon and Schwartz, 2006), infants (Loughnan et al., 2010; McGeehin and Mirabelli, 2001), those with a low socio-economic status (Harlan et ⁎ Corresponding author. E-mail addresses: [email protected] (M. Olonscheck), [email protected] (C. Walther).

http://dx.doi.org/10.1016/j.uclim.2016.12.001 2212-0955 © 2016 Elsevier B.V. All rights reserved.

Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

2

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

al., 2006; Huang et al., 2011) or with pre-existing diseases (Vandentorren et al., 2006) are particularly susceptible to heat. However, not only individuals, but also the health sector is affected. Dolney and Sheridan (2006) have shown a significant correlation between periods of extreme heat and an increase of emergency calls. Parts of a city that are specially heat exposed also have a higher risk of damage to infrastructure (Rosenzweig et al., 2011). Existing studies analyze the spatial distribution of heat exposure in urban regions using different input data and methods. For decision makers the question arises as to which method to use and which results to trust when it comes to implementing proper adaptation measures. This paper focuses on comparing the results of three methods. We are interested in answering the following question: Does using different methods to assess the relative heat exposure within a city necessarily lead to different results? There are many studies that have examined the relationship between temperature and different influencing factors. The temperature distribution within a city is influenced by natural and anthropogenic factors. The natural factors include, for example, altitude which is negatively correlated to temperature (Alcoforado and Andrade, 2005) and vegetation which can provide cooling and fresh air for cities (Bowler et al., 2010; Hart and Sailor, 2009; Heinl et al., 2015; Peng et al., 2012; Radhi et al., 2013; Senanayake et al., 2013). Population density, which is positively correlated with temperature (Bottyán et al., 2005; Klysik and Fortuniak, 1999; Radhi et al., 2013), is often used as a proxy for anthropogenic factors like the compactness and heat-storing capacity of buildings (Goh and Chang, 1999; Pandey et al., 2014), the percentage share and type of sealed surfaces (Akbari et al., 2001) and heat emissions caused by industrial or residential activities (Hart and Sailor, 2009; Nkemdirim and Truch, 1978). Following the IPCC definition, we define exposure as “the presence of people […] that could be adversely affected (IPCC, 2014)”. Thus, a high exposure means circumstances of extraordinary heat with potentially stressful effects on human health and well-being. Vice versa, a low exposure is characterized by low or negligible risk of heat impacts. Three kinds of data collection methods can be found concerning the assessment of spatial differences in exposure to temperature in cities. Firstly, there are studies that are based on measured temperature data from mobile sensors or meteorological stations (Alcoforado and Andrade, 2005; Bottyán et al., 2005; Hart and Sailor, 2009; Steeneveld et al., 2014). Secondly, there are analyses based on urban climate maps (Acero et al., 2013; Hebbert, 2014; Houet and Pigeon, 2011; Ren et al., 2010). Sismanidis et al. (2015) describe advances in a novel methodology of using geostationary satellite thermal infrared data for deriving land surface temperature data with a high spatial and temporal resolution. Thirdly, there are analyses that use remote sensing data (Effat and Abdel Kader Hassan, 2014; Heinl et al., 2015; Li et al., 2011; Senanayake et al., 2013; Zhou et al., 2013). Additionally, there are some studies that use a combination of measured near-surface data and remote sensing data (Kato and Yamaguchi, 2005; Radhi et al., 2013; Sodoudi et al., 2014). Studies that use two different methods only compare the temperature at different meteorological stations or at different times of the day, but do not compare the results from using data from different sources. For Karlsruhe, which was one of the most affected cities in Germany during the heat wave of 2003, we critically review three methods to assess the relative heat exposure on a small scale. We choose the spatial level of quarters (in German: Stadtviertel), which is the smallest administrative unit in the city of Karlsruhe. Although quarters in Germany are administrative divisions with no regulatory power or jurisdictional autonomy, it is helpful and often even necessary for a decision maker to have small-scale information on heat exposure in order to focus adaptation measures on the most exposed parts of the city. Different quarters together form a district (in German: Stadtteil), which in many large cities worldwide is the only unit of subdivision. Some districts in Karlsruhe have committees that have to be consulted regarding important issues, but decisions cannot be taken at this administrative level. As different parts of a city may not be evenly affected by extreme weather events, methods for the assessment of climate change impacts for quarters are required. Thus, we introduce three possibilities to determine exposure to heat waves, which are based on different data sources: meteorological stations, a city climate index, and remote sensing. These methods differ in terms of their data requirements, processing effort, necessary knowledge, and resources, and the accuracy of the results may also differ (see Discussion). Therefore, it is the aim of this study to compare the results of these three methods. We also describe the strengths and weaknesses of the three methods in terms of revealing the heat distribution in Karlsruhe. Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

3

2. Materials and methods 2.1. Data The applied meteorological data both for the past analysis of extended periods of high temperature and the exposure analysis are from the data portal WebWerdis (Web-based Weather Request and Distribution System) of the German Weather Service (DWD, 2015) and KIT (2015). Altitude data are from a digital elevation model of Europe (Jarvis et al., 2008) and population density data are from commune level population data and the CORINE data set CLC 2006 (Gallego, 2010), both with a spatial resolution of 100 m. We used the land surface temperature data MOD11A2 from the Moderate Resolution Imaging Spectroradiometer (MODIS) Aqua satellite platform (NASA, 2015), which provides data at a roughly 1000 m spatial resolution, approximately at 1:30 p.m. local time. Additionally, we applied processed data from the city climate model FITNAH (Gross, 1991) that characterize green areas and fresh air corridors (balancing zones) and areas with increased heat stress and air pollution (atmospheric load zones). The data were generated during a local climate analysis by the Neighborhood Association of Karlsruhe (Nachbarschaftsverband Karlsruhe, 2011).

2.2. Study region Karlsruhe (173.46 km2, 300,000 inhabitants) is a city situated in southwest Germany, which is bordered by the river Rhine in the west and the Black Forest in the South. Karlsruhe consists of 27 districts which are further subdivided into 70 quarters. The size of the districts is 6.4 km2 on average and ranges between 1.4 km2 and 23.0 km2. The average size of the quarters is 2.4 km2 and varies between 0.2 km2 for the smallest and 14.0 km2 for the largest. Analyzing the CORINE data for Karlsruhe shows that the city is covered by artificial surfaces (43%), agri-cultural areas (28%), forest and semi-natural areas (27%) and wetlands and water bodies (2%). According to Köppen's climate classification, the city is in the Cfb class, which is typical for large parts of central Europe, and is characterized by a temperate and humid climate with warm summers (Köppen, 1923). The annual average temperature of 10.2 °C is N2 °C higher than the German average. The city has been repeatedly subject to extended periods of intense heat, for example in 2003, 2005, 2006 and 2015 (DWD, 2015; KIT, 2015). During the 2003 heat wave, the meteorological station Karlsruhe-city had the highest mean daily maximum temperature of all of the meteorological stations in Germany with available data (5.4 °C above the average of all stations in August). Figure 1 shows the occurrence rates of heat wave days at the

Figure 1. Heat wave days per year based on observed climate data at the meteorological station Karlsruhe-city. The operator of the meteorological station changed in 2009 while the location remained the same. The black bar marks missing values in the year 1945. Note: Data for 2015 were only available until the end of September. Data source: 1876–2008 from DWD (2015) and 2009–2015 from KIT (2015).

Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

4

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

meteorological station Karlsruhe-city (Hertzstraße) in the period 1876 to 2015. A heat wave day (HWD) is a day of a series of at least three consecutive days with a maximum temperature of at least 30 °C (Hutter et al., 2007; Huynen et al., 2001; Kyselý, 2004; Porebska and Zdunek, 2013). 2.3. Methods The distribution of heat in Karlsruhe was assessed by three different methods. They are aimed to cover the various input variables and procedures that can be used to gain information on the heat exposed parts of a city. An overview of the data and the applied method is given in Table 1, while a detailed description follows below. 2.3.1. Method 1: meteorological station data input The first method for the assessment of the heat exposure uses daily maximum temperature from meteorological stations in the region of Karlsruhe. As the number of meteorological stations within a city is often relatively small, this method is applied to additionally use daily temperature data from meteorological stations in the vicinity of Karlsruhe. The core of the method is to uncover the functional relationship between non-meteorological influencing variables (e.g. topography and population density) and the regional distribution of heat. The area for selecting meteorological stations in the hinterland of Karlsruhe was restricted in size, in order to avoid that the influence of different climates is larger than the effect of the local topography and the population density (our influencing variables). The smallest buffer size around the city that provides an appropriate number of meteorological stations for the following analysis is 80 km around the city boundary of Karlsruhe (Walther and Olonscheck, 2016). For the 39 meteorological stations within this buffer, we calculated the mean number of HWDs per year, introduced above, for the period 2003 to 2011. With a multiple regression approach the obtained heat wave day occurrence rates were explained by the values of the two influencing variables in the area surrounding the meteorological stations: altitude (ALT) (Figure 2, left) and population density (PD) (Figure 2, right). Walther and Olonscheck (2016) were able to show that these variables are most suitable for explaining the local heat exposure in Karlsruhe. The following equation describes the functional relationship gained from the values of each of the meteorological stations (MS): HWDMS ¼ β0þ β1  ALTMS þ β2  PDMS þ σ MS

ð1Þ

This provides the model parameters (β0, β1, β2) applied later and the unexplained residuals of the regression at each meteorological station σMS. To take a certain influencing zone into account, the variable Table 1 Overview of data and methodology of the three compared methods. Method

Data input

Meteorolo-gical Daily maximum temperature from meteorological stations in station data the urban area and its input surrounding in combination with altitude (CGIAR) and population density data (CORINE) City climate Urban climate model output index input

Methodology

Data output – assessment of heat exposure

Data sources

Regression analysis with inverse distance interpolation

City quarter specific heat wave days

DWD (2015); Gallego (2010); Jarvis et al. (2008)

City climate index – a relation between the balancing and the load effect per city quarter Surface temperature per km2

Gross (1991); Nachbarschaftsverband Karlsruhe (2011)

Identification of areas with a balancing and an atmospheric load effect Remote sensing Satellite measurements of surface Maps of surface input temperature temperature distribution are interpreted as heat exposure in the city

NASA (2015)

Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

5

Figure 2. Influencing variables altitude (left) and population density (right) in the quarters of Karlsruhe. Note: A map shows the location of Karlsruhe in Germany. Data sources: DWD (2015), Gallego (2010), Jarvis et al. (2008).

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

6

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

Table 2 Definition of the balancing (blue) and the loading (red) effect of the city climate index.Data source: Nachbarschaftsverband Karlsruhe (2011).

Zones with a balancing effect – Zones with an atmospheric load Cold air supply capacity – Stress due to heat

Very high (4) High (3) Moderate (2) Low (1)

Part of a fresh air corridor (air exchange) – Increased stress due to air pollution Yes

No

Very high (5) High (4) Moderate (3) Low (2)

Very high (4) High (3) Moderate (2) Low (1)

population density was averaged in a buffer of 3000 m around the meteorological stations (see Walther and Olonscheck, 2016). The model parameter and the residuals from Eq. (1) are in a second step used to gain high resolution information about the heat distribution within the city. The altitude and the population density data are averaged for each of the city quarters of Karlsruhe (ALTQ, PDQ) which delivers in combination with the model parameters (β0, β1, β2) and interpolated residuals (σQ) the quarter specific number of heat wave days (HWDQ): HWDQ ¼ β0þ β1  ALTQ þ β2  PDQ þ σ Q

ð2Þ

The quarter specific heat wave days gained with this method are used as an estimation of the heat exposure distribution within the city of Karlsruhe.

2.3.2. Method 2: city climate index input The second method is compiling a city climate index that consists of balancing and atmospheric load zones. The well-known numerical simulation model Flow over Irregular Terrain with Natural and Anthropogenic Heat Sources FITNAH (Gross, 1991; Nachbarschaftsverband Karlsruhe, 2011) was applied in an Ecological Tolerance Study (ETS) carried out by the Neighborhood Association of Karlsruhe. Within this analysis, two effects are distinguished which do not consider administrative boundaries, but are solely based on climatic parameters: zones with a balancing effect and zones with a high atmospheric load. For both zones, criteria were developed in the ETS. First, zones with a balancing effect (such as green areas) feature the ability to deliver cold air; these are shown in Table 2 in blue. In the ETS, such areas are defined according to the official landuse plan of the city and are divided into four categories according to their cold air supply capacity, namely very low (1) to very high (4). In addition, we considered the contribution fresh air corridors make to air exchange and thus cooling (also shown in blue in Table 2). Fresh air corridors have been manually assigned in the ETS based on the depth contours of the valley incisions, obstacle-free areas of the upper Rhine valley and cold air outlet areas of undeveloped zones with a slope of N1°. The balancing values increase by one value if the zone is part of a fresh air corridor (“Yes” in Table 2) and stay the same if there is no air exchange (“No” in Table 2). Second, atmospheric load zones (shown in red in Table 2) can be characterized by heat stress and increased air pollution (Nachbarschaftsverband Karlsruhe, 2011). Air pollution is assumed to be an approximation of anthropogenic heat emissions (industry, traffic). The load zones that cause heat stress are ranked from very low (1) to very high (4), depending on the deviation of the Predicted Mean Vote (PMV), a dimensionless, subjective degree of uneasiness, from mean values in the city. Based on Nachbarschaftsverband Karlsruhe (2011) we assumed that the atmospheric load increases by one value (“Yes” in Table 2) if there is an increase in air pollution that exceeds the value of 80 μg/m3 NO2 and stays the same if this is not the case (“No” in Table 2). Similar categorizations have been used in urban climate map studies (see Introduction). Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

7

For each quarter, we determined both the spatial share of different balancing and atmospheric load values and combined them to a total city climate index (CCI): 5 X

CCI ¼

C¼1 5 X

AL; C  LC ð2Þ AB; C  BC

C¼1

AL,C thereby describes the percentage share of different atmospheric load areas which is multiplied by the value of the corresponding atmospheric load LC of the area concerned. The result is then divided by the product of the percentage share of the area with a balancing effect AB,C and its value BC. The variable C refers to the five different loading and balancing categories. Low CCI values show that the balancing effect predominates in a part of the city (green in Figure 4) while high values demonstrate high heat stress (red and orange in Figure 4). In contrast to the other two methods, this approach allows for the consideration of the structure of the city. The relationship between the balancing effect and the atmospheric load enhancing effect per city quarter is used as an estimation of the heat exposure distribution within the city of Karlsruhe. 2.3.3. Method 3: remote sensing input The third method to assess the exposure to heat in a city is based on satellite data. While in-situ measured air temperature data have a high temporal resolution and often a long recording period, remotely sensed data have a higher spatial resolution and almost global coverage (Tomlinson et al., 2011). Although the correlation between land surface temperature (from MODIS) and air temperature depends on many factors, e. g. advection, solar radiation, surface material and sky view factor (Kloog et al., 2012), a variety of studies have shown that these two variables are closely related (Prigent, 2003; Prihodko and Goward, 1997). The MODIS data have been assessed over a widely distributed set of locations and time periods using several ground-based measurements and validation efforts, the accuracy of which can be within 1 K in most cases (Wan, 2008). Because satellite observations are only available under clear-sky conditions, we used the 8-day averaged data to compensate for data gaps. Against this background, we applied the 1:30 p.m. data for the summer months from June to August (JJA) for the years 2003–2011. The land surface temperature data averaged per city quarter are used to assess the heat exposure distribution within the city of Karlsruhe. 2.3.4. Sensitivity analysis and creation of the final map As we aim to compare the three models regarding their ability to assess heat exposure in Karlsruhe, the unequal output data need to be harmonized. This is done by dividing the city quarter values for each method into different classes of heat exposure ranging from very high (red) to low (green). The decision on how many classes to include was made based on a sensitivity analysis which was conducted also in order to find out the effect of the number of heat exposure classes on the final result. We checked the sensitivity by summing the mean standardized variance between the three methods for each quarter for different numbers of heat exposure classes in the following way:

DG ¼

3 1 X Q Q jCi −Cj j=ðG−1Þ X 3 i; j¼1;i¿ j Q

N

ð3Þ

Q DG is the mean standardized variance between the three methods for G classes. CQ i and Cj are the heat exposure classes for the quarter Q using the methods i and j respectively. G is the number of classes and N is the number of quarters. We divided by the maximum possible difference in G classes (G − 1) to normalize the deviation and to enable a comparison. We then created a comparison map which gives an indication of the similarity of the results using the three methods for each quarter (Figure 4, upper right).

Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

8

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

Figure 3. Mean standardized variance between the results of the three methods for different numbers of heat exposure classes.

3. Results The results based on different methods are influenced by the way the data are classified. Regarding the dependency of the result on the number of heat exposure classes applied, overall the sensitivity (see Eq. (3)) decreases with an increasing number of classes (Figure 3). Other than for two, three and four classes, the choice of the number of classes only has a small influence on the similarity of the results obtained by the different methods. Regarding the number of classes for the heat exposure, we opted for five classes. This is a compromise between a low mean standardized variance and a manageable number of classes. We created the classes on the basis of quintiles (20th, 40th, 60th, and 80th). Figure 4 (left column) shows that on a qualitative level the three exposure maps are comparable. The quarters in the southeast are higher in altitude, for the most part have a low population density, and have more open or green space, which leads to a very low level of the potential heat exposure. Very high and high levels of heat exposure can be found in particular in the central parts of the city (city district CD2 and CD5) as well as in parts of the districts CD1, CD3, CD4, CD6, CD8, CD16, CD17 and CD27, which are all characterized by a high level of compaction. Differences and similarities can be seen in the comparison plot in Figure 4 (upper right). We summed the differences in the heat exposure classes for the three methods in each quarter. That let to five possible combinations listed in Table 3. While no quarter shows a maximal discrepancy between the results in the three methods (“pink” label), for four quarters do the three models provide strongly deviating results (“orange” label in the comparison map of Figure 4). For 20 quarters there is a weak consensus between the classes (“light blue” label). Differences between the methods occur above all in areas with special characteristics such as highly sealed areas with low population density or green areas with a high population. The same holds true for districts and quarters that are partly highly sealed and partly vegetated. Surprisingly however, for the majority of quarters (66%), all methods provide a comparable heat exposure. This includes the two classes: same result (“dark blue” label, 21 quarters) and two methods show the same exposure while the third reveals a deviation of only one heat exposure class (“blue” label, 25 quarters).

4. Discussion When comparing the map showing heat wave days (method 1) with the map of the city climate index (method 2), it becomes clear, that most quarters with a low number of heat wave days in the past also have a low atmospheric load, whereas those with a high number of heat wave days have a high atmospheric load. Nevertheless, there are also districts of the city (like CD3 and CD11) with partly very high atmospheric loads (city climate index of 5) despite a moderate or low frequency of heat wave days. These districts are only moderately populated (inducing low heat exposure values in method 2), but are more highly sealed than other quarters, meaning that there is a high atmospheric load in combination with a low ability of these areas to deliver cold air (which leads to a high city climate index). Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

9

Figure 4. Level of the potential heat exposure for each quarter of Karlsruhe obtained by the three methods (left column). In the upper right corner the result of the comparison of the three methods is shown. Note: In the text we refer to the city district (CD) numbers which are assigned to the district names in the lower right corner of this figure. (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)

Conversely, there are quarters with a high or very high number of heat wave days per year, but a low or moderate atmospheric load such as parts of the city districts CD9, CD12 and CD26. The former quarter, for example, has a high population density, but is influenced by the nearby green and forest areas, which have a heat balancing effect. For parts of the city districts CD9, CD11, CD12, CD13, and CD15, the method based on data from meteorological stations (method 1) reveals higher values than the method based on remote sensing data (method 3). Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

10

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

Table 3 Overview on the classification of the applied heat exposure classes. Distance sum of the three methods

Class name

Color

Description

0 2 4 6 8

Consensus Almost consensus Weak consensus Stronger discrepancy Maximal discrepancy

Dark blue Blue Light blue Orange Pink

Same heat exposure level in all three methods One method one class difference (e.g. [1,1,2])) One method two classes difference (e.g. [1,1,3]) Different combinations possible (e.g. [1,3,4],[1,4,4]) Different combinations possible (e.g. [1,3,5],[1,5,5])

Note: The numbers in the description column refer to the degree of heat exposure. 5 = Very high to 1 = Very low.

Method 1 reflects the higher population density in those quarters. The reasons why the averaged characteristics of the surface material in that region - which influence the output of method 3 via the surface temperature - do not confirm the results of methods 1 and 2 can be twofold. Firstly, it can reflect a weakness of method 3, as the resolution is relatively low in comparison to the size of the quarter (e.g. central parts of CD15 or northern parts of CD13). Second, the local average surface characteristics could be influenced by “colder” surfaces, e.g. highly vegetated parts, such as the grass strip along the rivers or urban green spaces. On the other hand, there are parts of city districts with lower values for the meteorological station method compared to the results achieved with remote sensing data, such as in CD1, CD4, CD7 and CD17. For these differences the following explanations might be used: It is possible that method 1 underestimates the heat burden in that region, maybe because the population density data taken from CORINE data set are too low. While for most quarters the CORINE population data are slightly higher than the statistical data of the city (about plus 9% on average), they are 12% lower for the western part of CD4. An argument against this is the fact that the second method also shows relatively low heat exposure values for that region. Another reason for underestimation of the heat exposure in method 1 is a low population density but a high level of sealing. The size of the city quarter is also relatively small in comparison to the buffer around the stations that can be seen as an influencing zone – this can lead to both overestimation and underestimation. Industrial areas are not reflected in the population density variable – a drawback which could be improved by the application of other influencing variables. For example in the northern part of CD17 there is a technology park, which is highly sealed but not inhabited by people. In this region, method 3 and partly also method 2 show warmer conditions than method 1. Differences in the results might be also caused by the fact that land surface temperature, albeit closely related to air temperature, is only an approximation for the latter. Thus, it could be that the local ground cover is enabling a good discharge of surface heat which therefore is not heating up the atmospheric layer above. The comparison of the heat exposure based on the city climate index (method 2)with the results of the surface temperature derived from remote sensing (method 3) reveals differences for some quarters. Parts of the districts CD3, CD7, CD9, CD11, CD15 and CD19 clearly show lower heat exposure values in the remote sensing method, while the western part of CD4 has a higher value. The latter region is determined by a high level of sealing in parts, but also by allotments and a local recreation area, which functions as a fresh air lane and causes a low city climate index. The eastern part of CD6 is characterized by the highly sealed old airfield which explains the very high heat exposure in the remote sensing method, but it is influenced by the surrounding green areas which leads to a moderate heat exposure based on the city climate index. The western part of CD8 is, for example, characterized by the Rhine harbor which has little vegetation. The large water surfaces produce moderate heat exposure values in the remote sensing method. However, as the harbor and surrounding transport facilities also cause a lot of air pollution, the city climate index is quite high. Figure 4 reveals that the accordance in heat exposure levels per city quarter of Karlsruhe is relatively high despite the diverse methods and varying applied input data. If three methods show the same result, this is a valid sign for the reliability of each of the methods – under the presented assumptions. For those city quarters with an agreement, decision makers can freely choose the method that best fits their resources. Differences regarding the heat exposure level for some quarters can be explained by specific features of the methods in combination with special circumstances on the ground. As mentioned, the underlying reasons can be shortcomings in the input data (e.g. underestimations in population density), weaknesses in the underlying assumptions (e.g. relation between land surface temperature and air temperature) or problems arising from Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

11

data resolution or necessary aggregation of the model output. Table 4 provides a comparison of the results regarding heat exposure in the three applied methods and gives possible explanations for the differences. The methods for exposure assessment differ in terms of their underlying data requirements, processing efforts and costs, necessary knowledge and resources, as well as accuracy. In poorer countries and in regions in the Southern Hemisphere, small-scale data on independent influencing variables – that we used for method 1 – may be lacking. However, given the availability of such data, this method can be quite easily applied for other cities or regions. For the second method introduced a model of the city climate is necessary which is usually not available. Thermal, high resolution remote sensing data are essential for the third method. Given the existence of a city climate model or adequate data from meteorological stations and regarding the influencing variables, both methods are fairly cheap. In contrast, new scan flights can be expensive. However, the creation of a city study, which can provide the data for a city climate index can be quite time consuming. The highest spatial accuracy is offered by an analysis based on a city climate index, as the data are normally based on local measurements and modeling. The low spatial resolution of free satellite data can be improved by new scan flights. The temporal resolution of the data used for the city climate index and the remote sensing method is generally rather limited. While the temporal resolution of the data from the meteorological stations is quite high, the exactness of the influencing variables might be low as this type of data is often modeled. 5. Conclusion Numerous studies analyze the effect of heat waves on human health. However, in order to efficiently direct adaptation measures for the reduction of possible heat-related impacts on the population, small scale analyses regarding the most exposed parts of a city are needed but are often lacking. We introduced and compared three different methods for the assessment of heat exposure in city quarters. The application to the German city of Karlsruhe, which was repeatedly affected by heat waves in the past, showed large areas in the city with a very high accordance in the results of the methods. Thus, our analysis underlines that using different methods for the same question does not necessarily mean getting different results. The results can be interpreted as a cross validation meaning a high level of confidence in the accuracy of the heat exposure classification for most of the quarters. This suggests that decision makers could freely decide which of the three methods are most appropriate for them given their available resources. But this has to be done in an informed way as we also identified combinations of method characteristics and urban circumstances that lead to divergent interpretations of the three methods. In those quarters, where the methods yield different results, those responsible should be cautious of drawing premature conclusions and should better use more than one method for these special cases. The presented maps are a good basis for the implementation of small scale heat mitigation and adaptation measures particularly for older more temperature sensitive people. The suggested methods can also be used for

Table 4 Comparison of the results regarding heat exposure in the three applied methods and possible explanations for the differences. Qualitative comparison regarding the heat exposure

Explanation for differences

Method 1 b Method 2 Method 1 N Method 2 Method 1 N Method 3

Moderately populated while highly sealed. Highly populated but influenced by nearby vegetation. Resolution in Method 3 is relatively low in comparison to the size of the quarter; local peaks in population density cannot be adequately represented; local average surface characteristics could be influenced by “colder” surfaces. Moderately populated while highly sealed; surface temperature only approximation of air temperature. High sealing but large influence by green areas. Cool water surfaces but little vegetation and high air pollution in the harbor.

Method 1 b Method 3 Method 3 N Method 2 Method 3 b Method 2

Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

12

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

the assessment of heat in other cities. If data are available, implementing one of our methods on a smallscale level can be a useful contribution to reaching decisions on appropriate and focused local actions.

Acknowledgments The research leading to these results has received funding from the European Community's Seventh Framework Programme under Grant Agreement No. 308497 (Project RAMSES) and from the Federal Ministry for Education and Research of Germany who provided support under the framework of the PROGRESS Initiative (Grant No. 03IS2191B). We are very thankful to Alison Schlums and Stefanie Lyn Becker for proof-reading this document.

References Acero, J.A., Arrizabalaga, J., Kupski, S., Katzschner, L., 2013. Deriving an urban climate map in coastal areas with complex terrain in the Basque Country (Spain). Urban Clim. 4:35–60. http://dx.doi.org/10.1016/j.uclim.2013.02.002. Akbari, H., Pomerantz, M., Taha, H., 2001. Cool surfaces and shade trees to reduce energy use and improve air quality in urban areas. Sol. Energy 70 (3):295–310. http://dx.doi.org/10.1016/S0038–092X(00)00089–X. Alcoforado, M.J., Andrade, H., 2005. Nocturnal urban heat island in Lisbon (Portugal): main features and modelling attempts. Theor. Appl. Climatol. 84 (1–3):151–159. http://dx.doi.org/10.1007/s00704–005–0152–1. Bottyán, Z., Kircsi, A., Szegedi, S., Unger, J., 2005. The relationship between built-up areas and the spatial development of the mean maximum urban heat island in Debrecen, Hungary. Int. J. Climatol. 25 (3):405–418. http://dx.doi.org/10.1002/joc.1138. Bowler, D., Buyung-Ali, L., Knight, T., Pullin, A., 2010. Urban greening to cool towns and cities: a systematic review of the empirical evidence. Landsc. Urban Plan. 97:147–155. http://dx.doi.org/10.1016/j.landurbplan.2010.05.006. Dolney, T.J., Sheridan, S.C., 2006. The relationship between extreme heat and ambulance response calls for the city of Toronto, Ontario, Canada. Environ. Res. 101 (H. 1), 94–103. DWD, 2015. Selected meteorological data in XML format and online products. (Retrieved April 7, 2015 from). http://www.dwd.de/ webwerdis. Effat, H.A., Abdel Kader Hassan, O., 2014. Change detection of urban heat islands and some related parameters using multi-temporal Landsat images; a case study for Cairo city, Egypt. Urban Clim. 10:171–188. http://dx.doi.org/10.1016/j.uclim.2014.10.011. Gallego, F.J., 2010. A population density grid of the European Union. Popul. Environ. 31 (6):460–473. http://dx.doi.org/10.1007/s11111– 010–0108–y. Goh, K.C., Chang, C.H., 1999. The relationship between height to width ratios and the heat island intensity at 22:00 h for Singapore. Int. J. Climatol. 19 (9):1011–1023. http://dx.doi.org/10.1002/(SICI)1097-0088(199907)19:9b1011::AID-JOC411N3.0.CO;2-U. Gross, G., 1991. Das dreidimensionale, nichthydrostatische mesoscale-Modell FITNAH (the three-dimensional non-hydrostatic mesoscale model FITNAH). Meteorol. Rundsch. 43, 97–112. Harlan, S.L., Brazel, A.J., Prashad, L., Stefanov, W.L., Larsen, L., 2006. Neighbor-hood microclimates and vulnerability to heat stress. Soc. Sci. Med. 63:2847–2863. http://dx.doi.org/10.1016/j.socscimed.2006.07.030. Hart, M.A., Sailor, D.J., 2009. Quantifying the influence of land-use and surface characteristics on spatial variability in the urban heat island. Theor. Appl. Climatol. 95 (3–4):397–406. http://dx.doi.org/10.1007/s00704–008–0017–5. Hebbert, M., 2014. Climatology for city planning in historical perspective. Urban Clim. 10:204–215. http://dx.doi.org/10.1016/j.uclim. 2014.07.001. Heinl, M., Hammerle, A., Tappeiner, U., Leitinger, G., 2015. Determinants of urban–rural land surface temperature differences – a landscape scale perspective. Landsc. Urban Plan. 134:33–42. http://dx.doi.org/10.1016/j.landurbplan.2014.10.003. Hondula, D.M., Davis, R.E., Leisten, M.J., Saha, M.V., Veazey, L.M., Wegner, C.R., 2012. Fine-scale spatial variability of heat-related mortality in Philadelphia County, USA, from 1983–2008: a case-series analysis. Environ. Health 11:11–16. http://dx.doi.org/10.1186/1476069X-11-16. Houet, T., Pigeon, G., 2011. Mapping urban climate zones and quantifying climate behaviors - an application on Toulouse urban area (France). Environ. Pollut. 159:2180–2192. http://dx.doi.org/10.1016/j.envpol.2010.12.027. Huang, G., Zhou, W., Cadenasso, M.L., 2011. Is everyone hot in the city? Spatial pattern of land surface temperatures, land cover and neighborhood socioeconomic characteristics in Baltimore, MD. J. Environ. Manag. 92 (7):1753–1759. http://dx.doi.org/10.1016/j. jenvman.2011.02. 006. Hutter, H.P., Moshammer, H., Wallner, P., Leitner, B., Kundi, M., 2007. Heatwaves in Vienna: effects on mortality. Wien. Klin. Wochenschr. 119 (7–8):223–227. http://dx.doi.org/10.1007/s00508-006-0742-7. Huynen, M., Martens, P., Schram, D., Weijenberg, M.P., 2001. The impact of heat waves and cold spells on mortality rates in the Dutch population. Environ. Health Perspect. 109 (5):463–469. http://dx.doi.org/10.2307/3454704. IPCC, 2014. Annex II: Glossary. In: Mach, K.J., Planton, S., von Stechow, C. (Eds.), Climate Change 2014: Synthesis ReportContribution of Working Groups I, II and III to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change[Core Writing Team, R.K. Pachauri and L.A. Meyer (eds.)]. IPCC, Geneva, Switzerland, pp. 117–130. Jarvis, A., Reuter, H.I., Nelson, A., Guevara, E., 2008. Hole-filled SRTM for the globe version 4. Available from the CGIARCSI SRTM 90 m database. (Retrieved January 20, 2014 from). http://srtm.csi.cgiar.org/. Kato, S., Yamaguchi, Y., 2005. Analysis of urban heat-island effect using ASTER and ETM+ data: separation of anthropogenic heat discharge and natural heat radiation from sensible heat flux. Remote Sens. Environ. 99:44–54. http://dx.doi.org/10.1016/j.rse.2005. 04.026. KIT, 2015. Personal provision of the data at 01.10.2015. Institute of Meteorology and Climate Research of Karlsruhe Institute of Technology.

Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001

M. Olonscheck, C. Walther / Urban Climate xxx (2016) xxx–xxx

13

Kloog, I., Chudnovsky, A., Koutrakis, P., Schwartz, J., 2012. Temporal and spatial assessments of minimum air temperature using satellite surface temperature measurements in Massachusetts, USA. Sci. Total Environ. 432:85–92. http://dx.doi.org/10.1016/j.scitotenv.2012. 05. 095. Klysik, K., Fortuniak, K., 1999. Temporal and spatial characteristics of the urban heat island of Lodz, Poland. Atmos. Environ. 33 (24–25): 3885–3895. http://dx.doi.org/10.1016/S1352-2310(99)00131-4. Köppen, W., 1923. Die Klimate der Erde (The Climates of the Earth). Walter de Gruyter, Berlin. Kyselý, J., 2004. Mortality and displaced mortality during heat waves in the Czech Republic. Int. J. Biometeorol. 49 (2):91–97. http://dx. doi.org/10.1007/s00484–004–0218–2. Li, J., Song, C., Cao, L., Zhu, F., Meng, X., Wu, J., 2011. Impacts of landscape structure on surface urban heat islands: a case study of Shanghai, China. Remote Sens. Environ. 115 (12), 3249–3263. Loughnan, M., Nicholls, N., Tapper, N., 2010. Hot Spots Project - A Spatial Vulnerability Analysis of Urban Populations to Extreme Heat Events. McGeehin, M., Mirabelli, M., 2001. The potential impacts of climate variability and change on temperature-related morbidity and mortality in the United States. Environ. Health Perspect. 109:185–189. http://dx.doi.org/10.2307/3435008. Medina-Ramon, M., Schwartz, J., 2006. Temperature, temperature extremes, and mortality: a study of acclimatization and effect modification in 50 United States cities. Occup. Environ. Med. 64 (12):827–833. http://dx.doi.org/10.1136/oem.2007.033175. Nachbarschaftsverband Karlsruhe, 2011. Ökologische Tragfähigkeitsstudie für den Raum Karlsruhe (Ecological viability study for the region Karlsruhe). NASA, 2015. EOSDIS – NASA's Earth Observing System Data and Information System. (Retrieved April 7, 2015 from). http://reverb.echo. nasa.gov/reverb/#utf8=%E2%9C%93&spatial_map=satellite&spatial_type=grid. Nkemdirim, L.C., Truch, P., 1978. Variability of temperature fields in Calgary, Alberta. Atmos. Environ. 12 (4), 809–822 (1967). Pandey, A.K., Singh, S., Berwal, S., Kumar, D., Pandey, P., Prakash, A., Lodhi, N., Maithani, S., Jain, V.K., Kumar, K., 2014. Spatio-temporal variations of urban heat island over Delhi. Urban Clim. 10:119–133. http://dx.doi.org/10.1016/j.uclim.2014.10.005. Peng, S., Piao, S.L., Ciais, P., Friedlingstein, P., Ottle, C., Breon, F.-M., Nan, H., Zhou, L., Myneni, R.B., 2012. Surface urban heat island across 419 global big cities. Environ. Sci. Technol. 46 (2):696–703. http://dx.doi.org/10.1021/es2030438. Porebska, M., Zdunek, M., 2013. Analysis of extreme temperature events in Central Europe related to high pressure blocking situations in 2001–2011. Meteorol. Z. 22 (5):533–540. http://dx.doi.org/10.1127/0941-2948/2013/0455. Prigent, C., 2003. Land surface skin temperatures from a combined analysis of microwave and infrared satellite observations for an allweather evaluation of the differences between air and skin temperatures. J. Geophys. Res. 108:1–14. http://dx.doi.org/10.1029/ 2002JD002301. Prihodko, L., Goward, S., 1997. Estimation of air temperature from remotely sensed surface observations. Remote Sens. Environ. 60 (3): 335–346. http://dx.doi.org/10.1016/S0034–4257(96)00216–7. Radhi, H., Fikry, F., Sharples, S., 2013. Impacts of urbanisation on the thermal behaviour of new built up environments: a scoping study of the urban heat island in Bahrain. Landsc. Urban Plan. 113:47–61. http://dx.doi.org/10.1016/j.landurbplan.2013.01.013. Ren, C., Ng, E.Y.-Y., Katzschner, L., 2010. Urban climatic map studies: a review. Int. J. Climatol. 31:2213–2233. http://dx.doi.org/10.1002/ joc.2237. Robine, J., Cheung, S.L., Roy, S.L., Oyen, H.V., Herrmann, F.R., 2007. Report on Excess Mortality in Europe During Summer 2003. pp. 1–15. Rosenzweig, C., Solecki, W.D., Hammer, S.A., Mehrotra, S., 2011. Climate Change and Cities: First Assessment Report of the Urban Climate Change Research Network. Cambridge University Press, Cambridge. Senanayake, I.P., Welivitiya, W.D.D.P., Nadeeka, P.M., 2013. Remote sensing based analysis of urban heat islands with vegetation cover in Colombo city, Sri Lanka using Landsat-7 ETM+ data. Urban Clim. 5:19–35. http://dx.doi.org/10.1016/j.uclim.2013.07.004. Sismanidis, P., Keramitsoglou, I., Kiranoudis, C.T., 2015. A satellite-based system for continuous monitoring of surface urban Heat Islands. Urban Clim. 14 (2), 141–153. Sodoudi, S., Shahmohamadi, P., Vollack, K., Cubasch, U., Che-Ani, A.I., 2014. Mitigating the urban heat island effect in Megacity Tehran. Adv. Meteorol. 1:1–19. http://dx.doi.org/10.1155/2014/547974. Steeneveld, G.J., Koopmans, S., Heusinkveld, B.G., Theeuwes, N.E., 2014. Refreshing the role of open water surfaces on mitigating the maximum urban heat island effect. Landsc. Urban Plan. 121:92–96. http://dx.doi.org/10.1016/j.landurbplan.2013.09.001. Tomlinson, C.J., Chapman, L., Thornes, J.E., Baker, C., 2011. Review Remote sensing land surface temperature for meteorology and climatology: a review. Meteorol. Appl. 18:296–306. http://dx.doi.org/10.1002/met.287. Vandentorren, S., Bretin, P., Zeghnoun, A., Mandereau-Bruno, L., Croisier, A., Cochet, C., Ribéron, J., Siberan, I., Declercq, B., Ledrans, M., 2006. August 2003 heat wave in France: risk factors for death of elderly people living at home. Eur. J. Pub. Health 16:583–591. http://dx.doi.org/10.1093/eurpub/ckl063. Walther, C., Olonscheck, M., 2016. Analyzing heat exposure in two German cities by using meteorological data from both within and outside a city. Meteorol. Appl. 23 (3), 541–553. Wan, Z., 2008. New refinements and validation of the MODIS land-surface temperature/emissivity products. Remote Sens. Environ. 112 (1):59–74. http://dx.doi.org/10.1016/j.rse.2006.06.026. Zhou, B., Rybski, D., Kropp, J.P., 2013. On the statistics of urban heat island intensity. Geophys. Res. Lett. 40:1–6. http://dx.doi.org/10. 1002/n2013GL057320.

Please cite this article as: Olonscheck, M., Walther, C., Methods to assess heat exposure: A comparison of fine-scale approaches within th..., Urban Climate (2016), http://dx.doi.org/10.1016/j.uclim.2016.12.001