Urban Climate 34 (2020) 100678

Contents lists available at ScienceDirect

Urban Climate

journal homepage: www.elsevier.com/locate/uclim

Land surface temperature and energy expenditures of households in the Netherlands: Winners and losers T ⁎ Bardia Mashhoodi

Landscape Architecture and Spatial Planning Group, Department of Environmental Sciences, Wageningen University & Research, the Netherlands

ARTICLE INFO ABSTRACT

Keywords: This study aims to examine the associations between land surface temperature (LST) and annual Land surface temperature household energy expenditure (HEE) in urbanized zones of the Netherlands. To do so, satellite Urban heat island images of 96 days at four diﬀerent overpassing local times (10:30 a.m., 1:30 p.m., 10:30 p.m., Household energy expenditure 1:30 a.m.) are retrieved, and annual average maximum (LST-Max) and average minimum (LST- Energy poverty Min) LST at each zone are calculated. Employing geographically weighted regression, controlling Household energy consumption for 11 control variables, the results indicate that the impact of LST on HEE could not be enhanced unless the interactions between LST and location-speciﬁc circumstances are taken into consideration. In this line, four types of socio-spatial characteristics are distinguished: (1) losers, where higher levels of both LST-max and LST-Min are associated with higher HEE, characterised by high population density; (2) peak losers and trough winners, where LST-Max is associated with higher HEE and LST-Min with lower HEE, characterised by large households and elderly citizens; (3) peak winners and trough losers, where LST-Max is associated with lower HEE and LST-Min with higher HEE, characterised by private-rental dwellings and large building surface to volume ratio; (4) winners, where both LST-Max and LST-Min are associated with lower HEE, characterised by old buildings.

1. Introduction

1.1. Land surface temperature and household energy expenditure: two phenomena of critical societal importance

Land surface temperature (LST) and household energy expenditure (HEE) are two phenomena of critical societal importance in the European Union's member states and the Netherlands. On the one hand, due to climate change and the steady trend of urbanization, the number of Europeans exposed to high levels of land surface temperature is substantial and is expected to grow steadily (Ward et al., 2016). In the Netherlands the impervious surfaces and building masses in the cities of Amsterdam and Rotterdam have created a patchwork of urban heat islands, occasionally more than 10 °C warmer than suburban areas (van der Hoeven and Wandl, 2015a; van der Hoeven and Wandl, 2015b). On the other hand, household energy expenditures – or, in its severe form, energy poverty- endangers health and well-being of a substantial portion of the population in the majority of European countries (Thomson et al., 2017). In the Netherlands, according to a report of the EU Energy Poverty Observatory (2019), almost 5% of households are facing diﬃculties with adequately warming their dwellings or paying their energy bills. The report emphasizes that meeting energy expenditures is particularly challenging for households living in social housing – estimated to be as much as 2.25 million dwellings (Government of the Netherlands, 2019).

⁎ Corresponding author at: P.O. box 47, 6700 AA Wageningen, the Netherlands. E-mail address: [email protected]. https://doi.org/10.1016/j.uclim.2020.100678 Received 10 January 2020; Received in revised form 26 April 2020; Accepted 23 July 2020 2212-0955/ © 2020 The Author. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/BY/4.0/). B. Mashhoodi Urban Climate 34 (2020) 100678

Fig. 1. The underlying mechanism of the impact of LST on energy consumption of households.

This study aims to establish links between these two phenomena of critical societal importance: land surface temperature and household energy expenditure. The manuscript consists of ﬁve main parts. First, a knowledge gap in the existing studies is illustrated, and the objective and approach of this study are set. In the second section, the method and data of this study are described. In the subsequent sections, the results of the study are presented, discussed, and conclusions are drawn.

1.2. The underlying mechanism of the impact of land surface temperature on household energy consumption

The previous studies illustrate a pathway for the impact of LST on household energy consumption, consisting of three mechanisms (Fig. 1). The ﬁrst mechanism is the impact of LST on the ambient air temperature of buildings. In his conceptualization of urban energy balance, Oke (1988) describes the sun and the anthropogenic heating as the primary sources of heat in urban areas (Eq. (1)):

∗ Q +=++∆+∆QQQFHE Q S QA ∗ where Q is the net radiation received in an area, that is the difference between incoming solar radiation and outgoing long-wave radiations, and QF is anthropogenic heat. These sources of heat transfer to four types of heat in an area: QH, the sensible heat; QE, the latent heat, typically the heat consumed for evaporation of water; ΔQS, the heat stored in the surfaces of the area; ΔQA, the net transfer of heats trough flow of fluids. The first mechanism in the pathway is the impact of the heat stored in surfaces of an area, among them LST, on air temperature. In a study on microclimate in Paris, Lemonsu and Masson (2002) showed that a substantial ∗ portion of net radiation (Q ) is stored in urban surfaces (ΔQS), and subsequently is transferred into sensible (QH) and latent heat (QE). ∗ Similarly, Offerle et al. (2006), in a study on heat fluxes in central European cities centers, observed that QH and QE closely follow Q and ΔQS. The second mechanism is the association between outdoor and indoor temperature. In a study of the greater Boston and Massachusetts, Nguyen et al. (2014) found that there is a nonlinear relationship between outdoor and indoor temperature. The study shows that the association is robust in warm seasons (Pearson correlation, r = 0.91), and relatively weaker, yet substantial, in cold seasons (r = 0.40). A similar observation is made through a three-year analysis of a group of dwellings in New York by Tamerius et al. (2013). Oreszczyn et al. (2006) who found that indoor temperature can be predicted based on outdoor temperature, when a set of socioeconomic variables are controlled for, among them building age and energy efficiency, household size and age of the head of household. The third mechanism refers to three types of final energy consumption within dwellings in response to indoor thermal comfort and, indirectly, ambient air temperature of the dwellings: space heating, water heating and space cooling. Analyzing 15,000 Dutch dwellings, Santin et al. (2009) concluded that an increase in ambient temperature during days, evenings and nights results in an increase in energy consumption for space and water heating. In their study of the energy demand of aged residents in the U.S., Liao and Chang (2002) found that an increase in both heating- and cooling degree days is associated with higher energy consumption for water heating. In case of space heating, the authors suggested, the association with heating degree days is positive, whereas that with cooling degree days is negative. Santamouris et al. (2001) suggested that energy consumption for space heating can be lowered as LST could tolerate the cold ambient temperature. In a study on air conditioning load in the city of Athens, Hassid et al. (2000) concluded that LST is associated with higher space cooling consumption. Santamouris et al. (2001) found that the space cooling load of buildings could increase for 100% in general, and up to 150% in the peak, in response to urban heat island of 10 °C magnitude. Where the previous studies well elaborate the pathway between LST and energy consumption of households, the impact of LST on energy consumption of households is not studied. In the next part, this knowledge gap is discussed.

1.3. A knowledge gap: does land surface temperature aﬀect energy expenditures of households?

Although a variety of previous studies have estimated the impact of land surface temperature on households' gross energy consumption, there is no previous study available which estimates its impact on the expenditure of the households. Santamouris et al., (2001), for instance, concluded that due to surface temperature, heating load in city centres could be 30% to 50% lower compared to rural areas. A study conducted by Hassid et al. (2000) on the city of Athens found that due to heat island effect in urban areas peak hourly demand for electricity may increase up to 100%. An estimation of buildings' electricity load for cooling purposes in London by Kolokotroni et al. (2007) suggested that due to urban heat island effect the load in the urban areas is up to 25% higher than that in the rural areas. A study on urban and rural areas of the Netherlands concluded that energy consumption of households in half of the areas is significantly affected by land surface temperature, accounting for 0.8% of total consumption (Mashhoodi et al.,

2 B. Mashhoodi Urban Climate 34 (2020) 100678

2020). Two studies have approached the notion of household energy expenditure but remained short of conducting a comprehensive empirical study. In what they referred to as a “preliminary study”, Azevedo et al. (2016) concluded that land surface temperature is correlated with higher levels of electricity consumption when corrected for households' level of income. They, however, remained short of offering a multivariate statistical model, including other influential factors. Tsilini et al. (2015) in a study titled as “Urban gardens as a solution to energy poverty and urban heat island” has simulated the ambient temperature of a building block and concluded that in the scenarios of lower surface temperature the indoor temperature is lower in summers. The authors, however, have offered no measurement or analysis of “energy poverty” in the manuscript. Although previous studies have established links between household energy expenditure and a variety of determinants, there is no study available which considers land surface temperature as a potential determinant. For instance, using detailed interviews in the UK, Middlemiss and Gillard identified six challenges to energy vulnerable households: “quality of dwelling fabric, energy costs and supply issues, stability of household income, tenancy relations, social relations within the household and outside, and ill health” (2015, pp.146). Bouzarovski and Petrova (2015) concluded that residents of privately rented dwellings are more likely to be energy poor. Robinson argued that energy expenditures and energy poverty are vastly affected by “exclusion from the economy, time- consuming and unpaid reproductive, caring or domestic roles, exposure to physiological and mental health impacts, a lack of social protection during a life course; and coping and helping others to cope” (Robinson, 2019, pp.222). Legendre and Ricci concluded that French retirees are more likely to be at the margin of energy poverty as “Once individuals find themselves alone, they live in homes that are too big and bear the related costs”(2015, pp.626). Analyzing a longitudinal survey on living standards in Spain, Phimister et al. (2015) observed that being unemployed increases the chance of being in energy poverty by 50%. A knowledge gap in the previous studies is apparent. There is no empirical study available which establishes a link between two highly societal-relevant issues of land surface temperature and household energy expenditure. It remains unclear if land surface temperature affects all household similarly or such impact is more significant in case of a group of household with a particular socioeconomic profile. If the impact varies due to particular environmental factors, level of urbanization, or housing tenure, is not yet studied.

1.4. Objective and approach of this study

This study aims to analyse the impact of land surface temperature on households' energy expenditures by posing two research questions. First, does land surface temperature have a significant impact on increasing, or decreasing, household energy expenditures and, if that is so, how sensitive is the level of expenditures to an increase in land surface temperature? Second, is the impact of land surface temperature uniform in case of all households, or such an impact varies given the socioeconomic profile of the households and the geographic circumstances of the neighbourhoods in question? The analysis consists of two steps. First, a statistical model based on the assumption that the impact of land surface temperature is spatially uniform is employed, i.e. ordinary least square regression model (OLS). The results of the OLS model help control whether or not the level of collinearity between land surface temperature and other control variables is small enough. Second, a statistical model based on the assumption that the impacts on household energy expenditures are spatially variant, geographically weighted regression model (GWR), is employed. Employment of the GWR model is in line with the theoretical framework recently developed in the field of energy poverty studies which underlines that energy expenditure of household need to be understood in regards with location- specific circumstances (see Robinson et al., 2018; Mashhoodi et al., 2019b). Ultimately the diagnosis of the OLS and GWR models are compared, and the model which provides a better estimation used as the basis for answering the research questions of this study. This study includes ten types of the determinants of household energy expenditures, in addition to land surface temperature. The determinants are used as control variables, and the interaction terms between them and land surface temperature are included in the statistical models. The control variables are previously identified as potential determinants of households' energy expenditures and consumption:

1. Household size, as the number of children and economies of scale tend to increase in larger households (Middlemiss and Gillard, 2015; Anderson et al., 2012); 2. Privately rented dwellings, as the level of investment in building retrofit tends to be less in these dwellings compared to the owner-occupied or publicly rented dwellings (Kholodilin et al., 2017; Bouzarovski and Petrova, 2015); 3. Residents older than 65 years old, as such residents tend to live in larger dwellings alone and spend a longer time at home (Legendre and Ricci, 2015; Harrison and Popke, 2011); 4. Population density as a proxy for the level of urbanization (York, 2007; Mashhoodi et al., 2019a); 5. Building age, an indicator of buildings energy efficiency (Brunner et al., 2012; Fahmy et al., 2011); 6. Building surface to volume ratio, as it affects the thermal transfer between indoor and outdoor spaces (Bernabé et al., 2015; Steemers and Yun, 2009); 7. Number of warm or cold days, as it affects the energy use for space heating and cooling as well as the frequency of bathing and showering (Wiedenhofer et al., 2013; Reinders et al., 2003); 8. Humidity, as it affects the felt temperature of the residents (Alfano et al., 2011; Chow et al., 2010); 9. Wind speed, as it affects the thermal loss of the buildings, the level of energy use for ventilation and the ambient temperature of the buildings (Sanaieian et al., 2014; van Moeseke et al., 2005); 10. Level of vegetation, as it affects the felt temperature (Taleghani, 2018; Letu et al., 2010).

3 B. Mashhoodi Urban Climate 34 (2020) 100678

2. Methods and data

2.1. Method

At the ﬁrst step of the analysis an ordinary least square linear regression model, OLS, which holds all variables as a spatially invariant determinant of household energy expenditures (HEE) is adopted. The OLS model is formulated as below (Eq. 2).

HEEi =+ β0 βLST LSTi +∑∑ βk xik + γk LSTiik x + ε i k k where HEEi, LSTi, and xik respectively indicate the level of household energy expenditure, land surface temperature and measurement of control variable k in residential zone i. βLST, βk, and γk respectively denote coeﬃcients of land surface temperature, control variable k, and the coeﬃcient of the interaction terms between land surface temperature and the control variable k. β0 and εi show intercept and random error at zone i. At the second step of analysis, a geographically weighted regression model, GWR, which holds all the determinants of household energy expenditures as spatially variant determinants is adapted as (Eq. 3).

HEEi =+ β0 (,) μνi i βUST (,) μνLST i ii +∑∑ βki (,) μνxiik + γμνLSTxki (,)iiiki + ε k k where (μi,νi) is longitude and latitude of the centroid of the residential zone i. β0(μi,νi), βLST(μi,νi), βk(μi,νi), and γk(μi,νi) are respectively intercept, coefficient of land surface temperature, coefficient of the control variable k, and the coefficient of the interaction terms between land surface temperature and the control variable k, estimated specific to the location i. The location-specific coefficients are calculated as formulated by Eq. 4:

β (,ϑ)μ= ( XWμTT (,ϑ)) X−1 XWμ (,ϑ)y where β (,ϑμ ) is the unbiased estimate of β, and W(μ,ϑ) is an adaptive gaussian spatial weight matrix:

2 2 ⎧⎛ dij ⎞ ⎪⎜⎟1,θ− if dij < Wij = θ ⎨⎝ ⎠ ⎪ ⎩ 0, otherwise where Wij is the weight assigned to zone j in the GWR estimation speciﬁc to zone i. dij is the distance between the centroid of the two zones. θ is the adaptive bandwidth, i.e. the number of closest zones of zone i which are included in the calculation of spatial weight matrix. The bandwidth size is set at the value that minimize the Corrected Akaike Information Criteria (AICc) of the overall GWR model, calculated using the GWR 4.0 tool developed by Nakaya et al. (2009). Ultimately the performance of OLS and GWR models are compared employing four tests and the model providing a better estimation of HEE is selected: adjusted R2, AICc, and the magnitude and the randomness of the spatial distribution of the residuals.

2.2. Dependent variable

The case study area comprises 1262 urbanized residential zones, the so-called wijk in Dutch, of the Netherlands (Fig. 2). According

Fig. 2. The percentage of households' disposable income spent on energy expenses in the urbanized zones of the Netherlands in 2014.

4 B. Mashhoodi Urban Climate 34 (2020) 100678

Table 1 The satellite images providing data on land surface temperature. Each image represents the average value of 8 days.

Satellite image Overpassing local time Time period

MODIS/Terra (MOD11A2) 10:30 a.m. T01: 02/02/2014–09/02/2014 T02: 26/02/2014–05/03/2014 10:30 p.m. T03: 22/03/2014–29/03/2014 T04: 15/04/2014–22/04/2014 MODIS/Aqua (MYD11A2) 1:30 a.m. T05: 09/05/2014–16/05/2014 T06: 02/06/2014–09/06/2014 1:30 p.m. T07: 26/06/2014–03/07/2014 T08: 20/07/2014–27/07/2014 T09: 29/08/2014–05/09/2014 T10: 30/09/2014–07/10/2014 T11: 01/11/2014–08/11/2014 T12: 03/12/2014–10/12/2014 to the criteria set by the Dutch central bureau of statistics (CBS) the areas with a minimum number of 500 dwellings per square kilometre are designated as an urbanized zone (Centraal Bureau voor de Statistiek, 2014). Household energy expenditures (HEE) is the percentage of disposable income per capita spent on energy expenditures. It is calculated based on three types of data provided by CBS: gas and electricity consumption per capita in 2014 and average disposable income per head (ibid). The average price of gas and electricity in the Netherlands is provided by Eurostat database on the energy prices in EU member states (Eurostat, 2015).

2.3. Independent variables

The independent variable of this study is the average annual land surface temperature in the urbanized areas (LST). To retrieve land surface temperature across the study areas, four sources of satellite data with four diﬀerent overpassing times are used (Earthdata, 2019): MODIS Terra day (10:30 a.m.), MODIS Terra night (10:30 p.m.), MODIS Aqua day (1:30 p.m.), and MODIS Aqua night (1:30 a.m.). In order to grasp the annual variation of LST, 12 satellite images from the four source are retrieved, each image providing data on the land surface temperature of 8 days. The satellite images represent 96 days of 2014 which are regularly spread across the year and provide reliable measurements of surface temperature – as indicated by the Quality Assurance band, and cover all study areas (Table 1). Fig. 3 summarises the variation of the average LST observed across the Netherlands, retrieved from the four satellite images in the 12 time periods. The spatial resolution of the satellite images is 1 km per 1 km which suﬃces for the purpose of this study, that is aggregating the values at the scale of residential zones. Fig. 3 represents the average annual maximum LST (LST-Max), and average annual minimum LST (LST-Min) across the Netherlands, the two indicators used in order to summarise the annual status of LST (Fig. 4).

2.4. Control variables

This study uses 11 control variables (Table 2). Four of the variables representing the socioeconomic status of the households are provided by CBS (Centraal Bureau voor de Statistiek, 2014). Household size is the average number of residents in a household in the residential zones. Private rent (%) shows the percentage of dwellings that are neither owned-occupied nor owned by housing associations or municipalities. Age group 65+ (%) is the percentage of residents aged 65 years or older. Population density represents the number of registered residents per square kilometre of the zone. Two of the control variables are retrieved from the GIS database of

Fig. 3. Average LST values in the 12 time periods.

5 B. Mashhoodi Urban Climate 34 (2020) 100678

Fig. 4. Average annual maximum LST (left) and average annual minimum LST (right) in the Netherlands, 2014.

Table 2 Descriptive statistics of control variables.

Variable Mean Minimum Maximum SD

Household size 2,23 1,30 3,50 0,30 Private rent (%) 41,12 2,00 100,00 15,04 Age group 65+ (%) 18,41 1 65 6,75 Population density 3164,00 23,00 21,656,00 2931,91 Building age 40,04 0 164 18,45 Surface to volume ratio 0,2495 0,1031 0,3367 0,0393 Humidity (%) 80,72 79 83 0,71 Wind speed (km/h) 40,26 29 68 6,65 NDVI 0,58 0 1 0,09 CDD 45,22 17 76 14,23 HDD 1554,74 1430,29 1657,22 52,70 the buildings of the Netherlands – the so-called 3D BAG database (Esri Netherlands, 2016): Surface to volume ratio, accounting for the ratio of the area of the outer surfaces of the buildings to their volume; Building age, representing the median age of the buildings in the zones. Four climate-related control variables are retrieved using the observations recorded at the 28 meteorological stations of the Royal Netherlands Meteorological Institute, KNMI (KNMI, 2018) and based on the guideline for spatial interpolation provided by the KNMI scientific team (Sluiter, 2012). Humidity is calculated based on ordinary kriging interpolation of the observations, with exponential variogram. Wind-speed at the height of 10 m is calculated based on the two-layer model of the planetary boundary layer interpolation (Stepek and Wijnant, 2011), using measurements of wind speed at the KNMI meteorological stations. Aerodynamic roughness length at the residential zones, as a precondition for the application of the two-layer model, is retrieved from European land cover database, CORINE (European Environment Agency, 2016), based on the classification proposed by Silva et al. (2007). The last control variable is annual average NDVI, Normalized Difference Vegetation Index, in 2014, representing the average monthly values observed by the satellite images. The images are retrieved from MODIS/Terra Vegetation Indices Monthly L3 Global 1 km SIN Grid V006 (Erathdata, 2019). Two measures quantify the annual status of air temperature in the zones: cooling degree days (CDD), the total number of degrees by which temperature exceeded a threshold temperature and thus energy consumption for space cooling is triggered; heating degree days (HDD), the total number of degrees by which temperature went below a threshold temperature and thus energy consumption for space heating is triggered. To do so, the maximum daily temperature (TX), minimum daily temperature (TN) and mean monthly temperature (TM) in the zones are retrieved - by universal kriging interpolation of the observations at the meteorological stations,

6 B. Mashhoodi Urban Climate 34 (2020) 100678

with an external drift of logarithmic distance to the shore. Following the study by Spinoni et al. (2015), The base temperature (Tb) for CDD is set at 22 °C, and that for HDD at 15 °C. Ultimately CDD and HDD are calculated as (Eq. 6 to 9):

⎧ 0, if TbX≥ T ⎪ TTXb− , if TMb≤< T T X ⎪ 4 CDDi = ⎨TTTTXb− bN− − , if TNbM≤< T T ⎪ 24 ⎪ ⎩ TTifTTMb−≤, bN

182 CDD= ∑ CDDi i=1

⎧ TTifTTbM−≥, bX ⎪TTbN− T Xb− T − , if TMb≤< T T X ⎪ 24 HDDi = ⎨ TTbN− , if TNbM≤< T T ⎪ 4 ⎪ ⎩ 0, if TbN≤ T

183 HDD= ∑ HDDi i=1 where CDDi is the daily cooling degrees from April 1st to September 30th, and HDDi is the daily heating degrees from October 1st to March 31st.

3. Results

3.1. Comparison between the performance of OLS and GWR models

Table 3 represents the diagnosis of the OLS and GWR models. The OLS model seeks the global impact of land surface temperature (LST-Max and LST-Min), the control variables, and the interaction terms between LST-Max and LST-Min with the control variables on HEE. The result of the model indicates that the level of collinearity between the predictors - measured by the variance inflation factor, VIF- falls within the acceptable range. The comparison between the performance of the OLS and GWR models show that the latter outperforms the former. Thus the spatial variation of the impacts needs to be taken into consideration. Adjusted R-square, the measurement of goodness-of-fit, of the GWR model is more than 62%, compared to 48% of the OLS model. The measurements of AICc show that the GWR model is significantly more informative than the OLS model. (Note that the lower levels of AICc indicate better performance.) The GWR model outperforms the OLS model in terms of residuals. First, the Moran's I measurement of residuals of the GWR model is significantly closer to zero, indicating a more random spatial distribution of the residuals. Regarding the magnitude of the residuals, the ANOVA test of residuals indicates a significant improvement consequent to the employment of the GWR model. The average standardized coefficient of LST-Max and LST-Min in the GWR model shows that the absolute magnitude of the impact is comparable to those of Private rent(%), Building age, CDD and HDD, four of the principal determinants of HEE (Table 3).

3.2. The interaction of land surface temperature with the socio-spatial context

The results of the GWR model indicate in more than 95% of the zones there is a significant (p-value≤0.05) interaction between land surface temperature with social and housing variables. In other words, in more than 95% of the zones there are households whose energy expenditures are significantly affected by the level of LST-Max or LST-Min. The type of association, however, varies across the zones. The results indicate that in more almost 48% of the zones, higher values of both LST-Max and LST-Min affect the level of HEE. These zones comprise of in-land urbanized zones – i.e. those distant from the North Sea. In more than 45% of the zones, only LST-Min has a significant interaction with social and household variables. The latter zones are mainly located in the west side of the so-called Randstad region and by the North Sea. The respective number for LST-Max, however, is negligible, 2.3%. In sum, households of different social and housing status are affected by the minimum level of LST in a vast majority of zones, 93%. The impact of the maximum level, however, is limited to a significantly lower number, 50%. (Fig. 5).

3.3. Winners and losers: four types of interactions between land surface temperature and the socio-spatial context

In order to summarise and compare the magnitude of estimated interaction terms across the Dutch zones, the average standardized coeﬃcient of interaction terms is calculated. Based on the average interaction terms, four types of socio-spatial characteristics could be diﬀerentiated: (1) losers, (2) peak losers and trough winners, (3) peak winners and trough losers, and (4) winners (Fig. 6).

7 B. Mashhoodi Urban Climate 34 (2020) 100678

Table 3 Diagnosis of the OLS and GWR model.

Variable OLS results GWR results

B Β VIF B mean B SD β mean β SD

(Constant) −29,318 −28,688 50,492 −0,232 0,462 Independent variable LST-Max 0,019 0,041 1857 0,010 0,037 0,021 0,079 LST-Min −0,052 −0,095 3459 0,008 0,040 0,015 0,073 Control variable Household size 0,154 0,086 4281 0,162 0,387 0,091 0,216 Private rent (%) 0,010 0,268 2700 0,008 0,008 0,215 0,208 Age group 65+ (%) 0,022 0,274 2116 0,027 0,007 0,336 0,084 Population density −4,15E−05 -0,224 4533 −3,50E−05 1,80E−05 −0,190 0,095 Building age 0,008 0,277 2880 0,009 0,004 0,298 0,153 Surface to volume ratio 0,353 0,026 1904 −0,217 1349 −0,016 0,098 Humidity (%) 0,321 0,419 3270 0,282 0,494 0,367 0,644 Wind speed 0,016 0,191 3482 0,021 0,009 0,259 0,113 NDVI −0,584 −0,101 2732 −0,242 0,484 −0,042 0,084 CDD 0,019 0,491 5079 0,011 0,026 0,291 0,690 HDD 0,002 0,217 3201 0,004 0,008 0,357 0,754 Interaction terms LST-Max*Household size 0,135 0,086 3779 0,083 0,156 0,053 0,100 LST-Max*Private rent −0,002 −0,071 2408 −0,001 0,002 −0,044 0,059 LST-Max*Age group 65+ 0,005 0,067 1590 0,003 0,004 0,040 0,047 LST-Max*Population density 6132E-06 0,037 2362 2,00E-06 1,10E-05 0,011 0,067 LST-Max*Building age -0,002 -0,057 2064 −1,32E−03 0,002 −0,048 0,068 LST-Max*Surface to volume ratio −0,594 −0,050 1445 −0,881 0,669 −0,073 0,056 LST-Min*Household size −0,116 −0,064 4160 −0,036 0,314 −0,020 0,173 LST-Min*Private rent 0,003 0,094 2569 0,005 0,004 0,140 0,108 LST-Min*Age group 65+ −0,007 −0,090 2120 −0,005 0,008 −0,071 0,106 LST-Min*Population density 2,36E−05 0,151 3153 0,000 0,000 0,123 0,079 LST-Min*Building age −0,002 −0,078 2903 -0,001 0,004 -0,054 0,154 LST-Min*Surface to volume ratio 0,981 0,073 1803 2015 1370 0,149 0,101 R-square 0,497 0,699 Adjusted R-square 0,486 0,623 AICc 2769,4843 2512,310 Residual Moran's I 0,149 0,014 Bandwidth NA 72 GWR ANOVA table SS DF MS F Global residuals 634,816 1236,000 GWR improvement 255,862 228,534 1120 GWR residuals 378,955 1007,466 0,376 2976

B: unstandardized regression coefficient β: standardized regression coefficient. OLS coefficients significant at p-value < 0,05 are underlined. VIF: variance inflation factor

The ﬁrst type of socio-spatial characteristics is called as the losers, i.e. when and where both higher levels of LST-max and LST- Min are likely to contribute to an increase in the level of household energy expenditures. The results show that at higher levels of population density, an increase at any level of LST imposes extra expenditures on households. LST-Min is found to have a more substantial impact in the highly populated areas than LST-Max. The second type of socio-spatial characteristics is called as peak losers and trough winners, i.e. when and where an increase in LST-Max imposes a higher level of energy expenditures on households, whereas an increase in LST-Min helps to reduce the expenditures. It is obtained that an increase in the maximum level of land surface temperature imposes higher energy expenditure on large-size households and elderly citizens, whereas these social groups beneﬁt from an increase in the minimum level of LST. The third type of socio-spatial characteristics is called as peak winners and trough losers, i.e. when and where an increase in LST- Max helps to reduce HEE, however, an increase in LST-Min imposes a higher level of expenditures on the households. It is found that the maximum level of land surface temperature contributes to mitigating energy expenditure of households living in private-rental dwellings and those who live in the dwellings with a relatively large surface to volume ratio. The residents of such housing tenures, however, face a higher level of energy expenditures when the minimum level of land surface temperature increases. The fourth type of socio-spatial characteristics is called as the winners, i.e. when and where both LST-Max and LST-Min contribute to reduce energy expenditures of households. The results show that in the zones where the median age of buildings, as an indicator for the quality of the buildings, is higher, any kind of increase in land surface temperature lowers HEE. The results show that the average impact of LST-Max and LST-Min are roughly equal.

8 B. Mashhoodi Urban Climate 34 (2020) 100678

Fig. 5. The signiﬁcant interactions of LST with social and housing variables.

Fig. 6. Average standardized interaction between the maximum and the minimum level of land surface temperature with social and housing variables.

4. Discussion

4.1. The mixed impacts of land surface temperature

The results of this study show that land surface temperature has a mixed impact on energy expenditure of diﬀerent social groups and housing types. In other words, when an increase in the level of LST could impose a higher level of expenditures on some households, it can contribute to mitigate those of others. The impact of land surface temperature also has a spatiotemporal variation. It has a spatial variation as the impact varies from one residential zone to another. The impact of LST has a temporal dimension, as the minimum level of LST, i.e. the temperature during the night, can exert a diﬀerent impact than the maximum level of LST, i.e. the temperature during the days. The mixed impact of land surface temperature is presumably due to its associations with two types of households' end-use of energy. First, through tolerating cold air temperature, land surface temperature can reduce the amount of energy used for space heating (Santamouris et al., 2001), which accounts for 64% of Dutch households' gross energy consumption (Eurostat, 2019). It needs to be

9 B. Mashhoodi Urban Climate 34 (2020) 100678 noted that gas, as a cheaper source of energy than electricity, account for almost 87% of space heating in the Netherlands. Therefore, the share of space heating from household energy expenditure is arguably less than its share from gross consumption (ibid). Second, at higher levels of land surface temperature the level of energy used for space cooling, bathing, and showering may increase (Kolokotroni et al., 2007; Hassid et al., 2000). A Eurostat survey shows that space cooling and hot water account for nearly 17% of Dutch households' energy consumption (Eurostat, 2019). As electricity –an energy source more expensive than gas- is the only energy source for space cooling (ibid), supposedly the share of cooling of total energy expenditure is signiﬁcantly larger than its share of gross consumption. Additionally, as buildings in the Netherlands are rarely equipped with air conditioning facilities (Menkveld and Beurskens, 2009) and therefore extra appliances, e.g. small fans, are used during the heatwaves, presumably a part of space cooling consumption is concealed in the category of “lighting and appliances”–which account for another 17.4% of consumption in the Netherlands (ibid). The mixed impact of land surface temperature and the spatiotemporal variation of such impact, as the results of this study suggest, are the combined impact of these opposite forces that occasionally outnumber or neutralize one another. When and where the increasing impact of land surface temperature outnumber, neutralize, or being outnumbered by its decreasing impact is determined by the underlying geography of the areas in question. In the next subsection, the interactions between land surface temperature and its embedding geography are discussed.

4.2. The winners and the losers

Both land surface temperature and energy consumption are embedded in a multidimensional geography comprising a variety of socioeconomic and housing characteristics. One cannot enhance the impact of land surface temperature without taking its interaction with the characteristics mentioned above into consideration. The results of this study point out that in more than 95% of the residential zone of the Netherlands there is at least one factor which is intensifying, or undermining, the impact of land surface temperature. The presence of such interactions, presumably, is the main reason behind the observed mixed impact of land surface temperature. In the following paragraphs, four types of socio-spatial characteristics, with regards to the observed interactions with urban surface temperature, are discussed: (1) losers, (2) peak losers and trough winners, (3) peak winners and trough losers, and (4) winners. The first type of socio-spatial characteristics is called as losers. This type of characteristics exemplifies the circumstances under which an increase in both the maximum and the minimum levels of land surface temperature results in a higher level of energy expenditure. This type of characteristic is prominent in the areas in which population density is relatively higher. Presumably, any increase in LST of such areas results in an increase in energy expenditure for space cooling and water heating, due to more frequent bathing, and such an increase outnumbers the reduction in space heating expenditures. This impact is presumably due to higher building density, which gives rise to heat transfer between dwellings (Wolf and McGregor, 2013) and co-presence of higher numbers of people at in- and outdoor spaces (Gabriel and Endlicher, 2011). The second type of socio-spatial characteristics is called as peak losers and trough winners. This type of characteristic typifies high presence of large households and elderly citizens. Presumably, the observed associations are due to the sensitivity of these social groups to extreme warm weathers, reflected by the maximum level of LST, which imposes a higher level of energy expenditures for space cooling and water heating. This result is in line with the previous studies indicating that children's relatively small body mass to surface ratio makes them exceptionally vulnerable to heatwaves and dehydration (Kovats and Hajat, 2008), and the studies un- derlining the difficulties that elderly citizens, particularly those in home care or institutions, face during heatwaves (Hajat et al., 2007). The other side of the coin of high sensitivity to temperature is the need of these social groups for a proper space heating during the nights, which benefit them from an increase in the minimum level of LST. Land surface temperature, as previously suggested, could help to warm the dwellings in the areas with cold air temperature (Zhu et al., 2010), as the Netherlands, and thus to mitigating energy expenditure of households for space heating. The third type of socio-spatial characteristics is called as peak winners and trough losers. This type of characteristic typifies high presence of private rental housing stock and the buildings with a large surface to volume ratio. The observed impact of LST on the private rental housing stock is presumably related to the overrepresentation of one-person, young households in the housing tenure. A study by Scanlon and Kochan (2011) shows that more than 50% of the households in the tenure in the Netherlands are single, compared to less than 30% couple without children and about 10% couples with children, and that around 60% of the renters are in the age groups between 30 and 65 years old. As a single citizen intends to spend more time outside his or her dwelling during the days, for the purposes as like work, shopping, sport, outside eating, his or her energy behaviour is more sensitive to thermal comfort during the nights (De Lauretis et al., 2017), represented by LST-Min. Additionally, given the age and family structure of the renters, presumably, the need for space heating in the housing tenure is less than other tenures (Paauw et al., 2009), and of the impact of LST- Min on increasing energy expenditures for space cooling outnumbers its impact on the reduction in space heating related expenditures. The other side of this coin of high presence of single, young households is a lower sensitivity to daily temperature than the residents of other housing tenures. Presence of privately rented dwellings, therefore, shows a negative interaction with LST-Max. In terms of building surface to volume ratio, the observed impact of LST could be explained by different functions of building geometry in the normal and extreme conditions. In a study on the energy impact of thermal comfort in different urban blocks in the Neth- erlands, Taleghani (2018) show that discomfort hours during summers is significantly higher in the building with a higher surface to volume ratio. In this respect, the overall warmth in a neighbourhood, represented by LST-Min, could impose extra energy expenditures on households. In contrary, in the event of extreme heat, represented by LST-Max, the buildings with a higher surface to volume ratio offer the opportunity for passive space cooling (Zinzi and Agnoli, 2012), and thus contribute to lowering energy

10 B. Mashhoodi Urban Climate 34 (2020) 100678 expenditure of households. The fourth type of socio-spatial characteristics is called as winners. This type of characteristic typifies high presence of old buildings. The characteristics of this type, i.e. benefiting from an increase in both LST-Max and LST-Min, suggests that in such geographic circumstances the reduction in energy expenditures for space heating outnumbers other possible impacts of LST. An estimation used in a policy report to the European Commission of the European Union stated that only 6% of Dutch dwellings are equipped with an air conditioning system (Menkveld and Beurskens, 2009). In this respect, the observed impact of LST could be explained by the fact that the most prominent difference between an old and a new building in the Netherlands is their efficiency in terms of space heating. As space heating account for 64% of Dutch households' gross energy consumption (Eurostat, 2019), presumably any increase in buildings' ambient temperature contributes to lowering energy expenditures of the residents of relatively older buildings. Therefore, such households benefit from higher levels of both LST-Max and LST-Min.

5. Conclusion

The result of this study urges for adding land surface temperature to the scope of studies on households' energy expenditure and energy poverty. The magnitude of the impact of land surface temperature is comparable to that of well-established determinants of energy poverty, such as the presence of privately rented dwellings, building age and number of heating- and cooling degree days. The results of this study indicate that the impact of land surface temperature on household energy expenditure has a spatiotemporal dimension. It has a spatial dimension as the effect varies from one location to another, and has a temporal dimension as the impact of LST during the days could differ from that of LST during the nights. The core conclusion of this study, in this respect,is that there is no global rule which captures the impact of LST, however, given the social and housing pattern of the neighbourhoods in question, different impacts can be expected. Land surface temperature could have different impacts in different socio-spatial contexts. In order to properly place land surface temperature in the policies regarding energy poverty and energy transition, case-specific policy instruments need to be introduced. As different social groups are likely to face higher than usual expenditures due to the maximum or the minimum diurnal land surface temperature, safety nets specifictodifferent households could be introduced. The policies could also monitor location-specific si- tuations and, if necessary, introduce measures to mitigate surface temperature. For instance, high levels of land surface temperature in highly-populated and unusually warm areas could be mitigated by urban cooling strategies and introduction of green and blue infrastructures. Over the course of the coming decades, the impact of land surface temperature could become more intense and less mixed. At this moment, land surface temperature has a mixed impact on energy expenditures of households, i.e. in some cases it provokes higher expenditures whereas in some cases it contributes to the mitigation of the costs. Given the trends in global warming and population ageing, however, in the coming decades land surface temperature could become a significant stimulus of higher energy expenditure for a substantial number of households. It is projected that the frequency of heatwaves in Europe will substantially increase. Examining different scenarios of climate change and heatwaves in central Europe, a paper by Lhotka et al. concluded that “severe heat waves are likely to become a regular phenomenon” (Lhotka et al., 2018, pp. 1043). The number of inhabitants older than 65 years old, which according to the results of this study suffer from higher energy expenditures due to the maximum levels of land surface temperature, is set to rise in the coming decades sharply. In 2014, statistical data used in this study, 2.92 million Dutch inhabitants were older than 65 years, accounting for 17% of the total population. In 2030, according to the projection of the Dutch Central Bureau for Statistics (Centraal Bureau voor de Statistiek, 2019), the numbers will be respectively 4.23 million and 23%. Currently, a marginal portion of the dwellings in the Netherlands are equipped with air conditioning system (Menkveld and Beurskens, 2009), and the share of space cooling account for less than one-fifth of end energy use of Dutch households. In the coming future, however, climate change will presumably affect both direct energy expenditures for space cooling, i.e. cost of energy consumed by the appliances, and indirect energy expenditures, i.e. cost of installation of new space cooling appliances (Hamlet et al., 2010; Sailor and Pavlova, 2003). Future studies need to consider both types of costs. As of today, land surface temperature has a mixed impact on energy expenditure of households. In the near future, however, it can potentially become a reason for severe energy poverty of a substantial portion of households - exerting an impact which is neither mixed nor desirable. The analysis carried out by this study is based on the annually aggregated energy expenditures of Dutch residential zones, due to lack of access to data at a finer temporal resolution. Further studies, using seasonal or monthly data on energy consumption, could elaborate further on the associations between LST and energy expenditure of households. Sharing of more detailed data by the Dutch central bureau for statistics (CBS) and energy companies could particularly open new opportunities for the researchers in this field to contribute to the mitigation of GHG emissions and tackling energy poverty in the Netherlands.

Declaration of Competing Interest

The authors declare that they have no conﬂict of interest.

References

Alfano, F.R.D.A., Palella, B.I., Riccio, G., 2011. Thermal environment assessment reliability using temperature—humidity indices. Ind. Health 49 (1), 95–106. Anderson, W., White, V., Finney, A., 2012. Coping with low incomes and cold homes. Energy Policy 49, 40–52. Azevedo, J.A., Chapman, L., Muller, C.L., 2016. Urban heat and residential electricity consumption: a preliminary study. Appl. Geogr. 70, 59–67.

11 B. Mashhoodi Urban Climate 34 (2020) 100678

Bernabé, A., Bernard, J., Musy, M., Andrieu, H., Bocher, E., Calmet, I., Kéravec, P., Rosant, J.M., 2015. Radiative and heat storage properties of the urban fabric derived from analysis of surface forms. Urban Clim. 12, 205–218. Bouzarovski, S., Petrova, S., 2015. A global perspective on domestic energy deprivation: overcoming the energy poverty–fuel poverty binary. Energy Res. Soc. Sci. 10, 31–40. Brunner, K.M., Spitzer, M., Christanell, A., 2012. Experiencing fuel poverty. Coping strategies of low-income households in Vienna/Austria. Energy Policy 49, 53–59. Centraal Bureau voor de Statistiek, 2014. Wijk- en buurtkaart 2014. https://www.cbs.nl/nl-nl/dossier/nederland-regionaal/geografische%20data/wijk-en- buurtkaart-2014 (Accessed 08/03/2018). Centraal Bureau voor de Statistiek, 2019. Forecast: 18.4 Million Inhabitants in 2060. https://www.cbs.nl/en-gb/news/2017/51/forecast-18-4-million-inhabitants-in- 2060#71becea3-4ca8-4b17-9bf7-8dae4c6a9f7e. Chow, T.T., Fong, K.F., Givoni, B., Lin, Z., Chan, A.L.S., 2010. Thermal sensation of Hong Kong people with increased air speed, temperature and humidity in air- conditioned environment. Build. Environ. 45 (10), 2177–2183. De Lauretis, S., Ghersi, F., Cayla, J.M., 2017. Energy consumption and activity patterns: an analysis extended to total time and energy use for French households. Appl. Energy 206, 634–648. Earthdata, 2019. https://earthdata.nasa.gov/ Accessed 22/07/2019. Esri Netherlands, 2016. 3D BAG. http://www.esri.nl/nl-NL/news/nieuws/sectoren/nieuw-in-arcgis-voor-leefomgeving accessed 09/03/2017. EU Energy Poverty Observatory, 2019. Member state reports, Netherlands. https://www.energypoverty.eu/observatory-documents/netherlands Accessed 04/12/ 2019. European Environment Agency, 2016. https://www.eea.europa.eu/data-and-maps/data/clc-2012-raster Accessed: 08/03/2018. Eurostat, 2015. Half-Yearly Electricity and Gas Prices, Second Half of Year 2012–14 (EUR per kWh). http://ec.europa.eu/eurostat/statistics-explained/ index.php?title=File:Half-yearly_electricity_and_gas_prices,_second_half_of_year,_2012%E2%80%9314_(EUR_per_kWh)_YB15.png&oldid=238282, Accessed 08/ 03/2018. Eurostat, 2019. Energy Consumption and Use by Households. https://ec.europa.eu/eurostat/web/products-eurostat-news/-/DDN-20190620-1. Fahmy, E., Gordon, D., Patsios, D., 2011. Predicting fuel poverty at a small-area level in England. Energy Policy 39 (7), 4370–4377. Gabriel, K.M., Endlicher, W.R., 2011. Urban and rural mortality rates during heat waves in Berlin and Brandenburg, Germany. Environ. Pollut. 159 (8–9), 2044–2050. Government of the Netherlands, 2019. Rented housing. https://www.government.nl/topics/housing/rented-housing Accessed 04/12/2019. Hajat, S., Kovats, R.S., Lachowycz, K., 2007. Heat-related and cold-related deaths in England and Wales: who is at risk? Occup. Environ. Med. 64 (2), 93–100. Hamlet, A.F., Lee, S.Y., Mickelson, K.E., Elsner, M.M., 2010. Effects of projected climate change on energy supply and demand in the Pacific northwest and Washington state. Clim. Chang. 102 (1–2), 103–128. Harrison, C., Popke, J., 2011. “Because you got to have heat”: the networked assemblage of energy poverty in eastern North Carolina. Ann. Assoc. Am. Geogr. 101 (4), 949–961. Hassid, S., Santamouris, M., Papanikolaou, N., Linardi, A., Klitsikas, N., Georgakis, C., Assimakopoulos, D.N., 2000. The effect of the Athens heat island on air conditioning load. Energy Build. 32 (2), 131–141. Kholodilin, K.A., Mense, A., Michelsen, C., 2017. The market value of energy efficiency in buildings and the mode of tenure. Urban Stud. 54 (14), 3218–3238. KNMI, 2018. http://www.sciamachy-validation.org/climatology/daily_data/selection.cgi. Kolokotroni, M., Zhang, Y., Watkins, R., 2007. The London heat island and building cooling design. Sol. Energy 81 (1), 102–110. Kovats, R.S., Hajat, S., 2008. Heat stress and public health: a critical review. Annu. Rev. Public Health 29, 41–55. Legendre, B., Ricci, O., 2015. Measuring fuel poverty in France: which households are the most fuel vulnerable? Energy Econ. 49, 620–628. Lemonsu, A., Masson, V., 2002. Simulation of a summer urban breeze over Paris. Bound.-Layer Meteorol. 104 (3), 463–490. Letu, H., Hara, M., Yagi, H., Naoki, K., Tana, G., Nishio, F., Shuhei, O., 2010. Estimating energy consumption from night-time DMPS/OLS imagery after correcting for saturation effects. Int. J. Remote Sens. 31 (16), 4443–4458. Lhotka, O., Kyselý, J., Farda, A., 2018. Climate change scenarios of heat waves in Central Europe and their uncertainties. Theor. Appl. Climatol. 131 (3–4), 1043–1054. Liao, H.C., Chang, T.F., 2002. Space-heating and water-heating energy demands of the aged in the US. Energy Econ. 24 (3), 267–284. Menkveld, M., Beurskens, L.W.M., 2009. Renewable Heating and Cooling in the Netherlands. D3 of WP2 from the RES-H Policy Project. A Report Prepared as Part of the IEE Project "Policy Development for Improving RES-H/C Penetration in European Member States (RES-H Policy)". https://publications.ecn.nl/bs/2009/ECN- B–09-013. Middlemiss, L., Gillard, R., 2015. Fuel poverty from the bottom-up: characterising household energy vulnerability through the lived experience of the fuel poor. Energy Res. Soc. Sci. 6, 146–154. Nakaya, T., Fotheringham, A.S., Charlton, M., Brunsdon, C., 2009. Semiparametric Geographically Weighted Generalised Linear Modelling in GWR 4.0. Nguyen, J.L., Schwartz, J., Dockery, D.W., 2014. The relationship between indoor and outdoor temperature, apparent temperature, relative humidity, and absolute humidity. Indoor Air 24 (1), 103–112. Offerle, B., Grimmond, C.S.B., Fortuniak, K., Kłysik, K., Oke, T.R., 2006. Temporal variations in heat fluxes over a central European city Centre. Theor. Appl. Climatol. 84 (1–3), 103–115. Oke, T.R., 1988. The urban energy balance. Prog. Phys. Geogr. 12 (4), 471–508. Oreszczyn, T., Hong, S.H., Ridley, I., Wilkinson, P., Warm Front Study Group, 2006. Determinants of winter indoor temperatures in low income households in England. Energy Build. 38 (3), 245–252. Paauw, J., Roossien, B., Aries, M.B.C., Santin, O.G., 2009. Energy Pattern Generator Understanding the Effect of User Behaviour on Energy Systems. ECN, Petten, pp. 9–10. Phimister, E., Vera-Toscano, E., Roberts, D., 2015. The dynamics of energy poverty: evidence from Spain. Econ. Energy Environ. Policy 4 (1), 153–166. Reinders, A.H., Vringer, K., Blok, K., 2003. The direct and indirect energy requirement of households in the European Union. Energy Policy 31 (2), 139–153. Robinson, C., 2019. Energy poverty and gender in England: a spatial perspective. Geoforum 104, 222–233. Robinson, C., Bouzarovski, S., Lindley, S., 2018. Underrepresenting neighbourhood vulnerabilities? The measurement of fuel poverty in England. Environ. Plan. A: Econ. Space 50 (5), 1109–1127. Sailor, D.J., Pavlova, A.A., 2003. Air conditioning market saturation and long-term response of residential cooling energy demand to climate change. Energy 28 (9), 941–951. Sanaieian, H., Tenpierik, M., van den Linden, K., Seraj, F.M., Shemrani, S.M.M., 2014. Review of the impact of urban block form on thermal performance, solar access and ventilation. Renew. Sust. Energ. Rev. 38, 551–560. Santamouris, M., Papanikolaou, N., Livada, I., Koronakis, I., Georgakis, C., Argiriou, A., Assimakopoulos, D.N., 2001. On the impact of urban climate on the energy consumption of buildings. Sol. Energy 70 (3), 201–216. Santin, O.G., Itard, L., Visscher, H., 2009. The effect of occupancy and building characteristics on energy use for space and water heating in Dutch residential stock. Energy Build. 41 (11), 1223–1232. Scanlon, K., Kochan, B., 2011. Towards a Sustainable Private Rented Sector: The Lessons from Other Countries. LSE London. The London School of Economics and Political Science. Silva, J., Ribeiro, C., Guedes, R., 2007, May. Roughness length classification of Corine land cover classes. In: Proceedings of the European Wind Energy Conference, Milan, Italy. vol. 710. pp. 110. Sluiter, R., 2012. Interpolation Methods for the Climate Atlas. KNMI Technical Rapport TR–335. Royal Netherlands Meteorological Institute, De Bilt, pp. 1–71. Spinoni, J., Vogt, J., Barbosa, P., 2015. European degree-day climatologies and trends for the period 1951–2011. Int. J. Climatol. 35 (1), 25–36. Steemers, K., Yun, G.Y., 2009. Household energy consumption: a study of the role of occupants. Build. Res. Inf. 37 (5–6), 625–637. Stepek, A., Wijnant, I.L., 2011. Interpolating wind speed normals from the sparse Dutch network to a high resolution grid using local roughness from land use maps. In: Koninklijk Nederlands Meteorologisch Instituut, Technical Report TR-321.

12 B. Mashhoodi Urban Climate 34 (2020) 100678

Taleghani, M., 2018. Outdoor thermal comfort by different heat mitigation strategies-a review. Renew. Sust. Energ. Rev. 81, 2011–2018. Tamerius, J.D., Perzanowski, M.S., Acosta, L.M., Jacobson, J.S., Goldstein, I.F., Quinn, J.W., Rundle, A.G., Shaman, J., 2013. Socioeconomic and outdoor meteorological determinants of indoor temperature and humidity in new York City dwellings. Weather, Clim. Soc. 5 (2), 168–179. Thomson, H., Snell, C., Bouzarovski, S., 2017. Health, well-being and energy poverty in Europe: a comparative study of 32 European countries. Int. J. Environ. Res. Public Health 14 (6), 584. Tsilini, V., Papantoniou, S., Kolokotsa, D.D., Maria, E.A., 2015. Urban gardens as a solution to energy poverty and urban heat island. Sustain. Cities Soc. 14, 323–333. Van der Hoeven, F., Wandl, A., 2015a. Amsterwarm: mapping the landuse, health and energy-efficiency implications of the Amsterdam urban heat island. Build. Serv. Eng. Res. Technol. 36 (1), 67–88. Van der Hoeven, F.D., Wandl, A., 2015b. Hotterdam: Hoe ruimte Rotterdam warmer maakt, hoe dat van invloed is op de gezondheid van inwoners, en wat er aan te doen is. TUDelft Boukunde. Van Moeseke, G., Gratia, E., Reiter, S., De Herde, A., 2005. Wind pressure distribution influence on natural ventilation for different incidences and environment densities. Energy Build. 37 (8), 878–889. Ward, K., Lauf, S., Kleinschmit, B., Endlicher, W., 2016. Heat waves and urban heat islands in Europe: a review of relevant drivers. Sci. Total Environ. 569, 527–539. Wiedenhofer, D., Lenzen, M., Steinberger, J.K., 2013. Energy requirements of consumption: urban form, climatic and socio-economic factors, rebounds and their policy implications. Energy Policy 63, 696–707. Wolf, T., McGregor, G., 2013. The development of a heat wave vulnerability index for London, United Kingdom. Weather Clim. Extrem. 1, 59–68. Mashhoodi, B., Stead, D., van Timmeren, A., 2020. Land surface temperature and households’ energy consumption: Who is affected and where? Applied Geography 114. Mashhoodi, B., Stead, D., van Timmeren, A., 2019a. Local and national determinants of household energy consumption in the Netherlands. GeoJournal 1–14. Mashhoodi, B., Stead, D., van Timmeren, A., 2019b. Spatial homogeneity and heterogeneity of energy poverty: a neglected dimension. Ann. GIS 25 (1), 19–31. York, R., 2007. Demographic trends and energy consumption in European Union nations, 1960–2025. Soc. Sci. Res. 36 (3), 855–872. Zhu, K., Blum, P., Ferguson, G., Balke, K.D., Bayer, P., 2010. The geothermal potential of urban heat islands. Environ. Res. Lett. 5 (4), 044002. Zinzi, M., Agnoli, S., 2012. Cool and green roofs. An energy and comfort comparison between passive cooling and mitigation urban heat island techniques for residential buildings in the Mediterranean region. Energy Build. 55, 66–76.