1
The temporal variation in the urban heat
island effect of Amsterdam from 2017-2021
Simone Hibma (Bsc student)
John van Boxel (examiner)
Emiel van Loon (coassessor)
30
thof May 2021
2
Abstract
The urban heat island (UHI) effect is present in cities all over the world. Heat can result in excess mortality and has an impact on human health, especially for vulnerable people. Heat also correlates with higher energy demand. The urban heat island intensity changes over time. This research aims to show to what degree meteorological extremes and seasonality affect the UHI effect in Amsterdam. The difference between the measured temperature from a rural and urban weather station results in the UHI effect. Certain days are classified as summery, ideal, and heat wave days based on KNMI
guidelines. Several tests are performed to assess whether or not the UHI effect is higher on these days than all data and the meteorological summer (JJA). The data is also divided into seasons and tested for significant differences. For the location AMSIJ and Osdorp, the UHI effect on summery, ideal, and heat wave days is significantly higher than for all data and for JJA. For Amstelveen, the UHI effect on summery, ideal, and heat wave days is not higher or not significantly higher compared to JJA.
Seasonality in the UHI is most important in Amstelveen. There are also significant differences between seasons in the UHI effect in AMSIJ and Osdorp. However, the trend is not as pronounced as it is for Amstelveen. These results are relevant, because summers are expected to get warmer. Moreover, meteorological conditions similar to summery days, ideal days and heat wave days will likely increase. Thus, the UHI intensity will likely increase in the future.
3
Introduction
Cities greatly influence climate change because urban-based activities and residents generate a high proportion of global greenhouse gas emissions. The reverse is also true, climate change also has a significant impact on cities. Some of these effects are increased flooding risk, drought, air pollution, and heat stress (Kleerekoper et al., 2012; Revi et al., 2014). In cities, the urban heat island (UHI) is an important phenomena. The UHI effect is defined as the difference in temperature within urban areas compared to rural areas (Gartland, 2008; Lauwaet et al., 2015).
In cities, several characteristics lead to a different surface energy budget, thus resulting in the UHI effect. There are five ways how the surface energy budget is influenced (Gartland, 2008;
Kleerekoper et al., 2012):
1. More absorption of short-wave radiation due to a larger share of low albedo surfaces, 2. Higher levels of air pollution, which lead to re-emitted longwave radiation,
3. Anthropogenic heat production through the use of energy,
4. A lower turbulent heat transport, because of a lower wind speed, due to the buildings, 5. The low degree of impervious surfaces, resulting in less water available for potential
evapotranspiration.
In the Netherlands, the UHI effect is a problem since the two large heatwaves in 2003 and 2006. In the Netherlands, between 1400 and 2200 fatalities in the summer of 2003 may have been heat-related since this was the excess mortality during the weeks of the heatwave (Garssen et al., 2005). Before that, the UHI effect was not of much concern in the Netherlands, because in a mild climate, temperature extremes are not as prevalent (Van der Hoeven & Wandl, 2014). However, extreme heat can be fatal. This is mainly a risk for people at of a higher and lower age group and people with a higher vulnerability (Revi et al., 2014). A higher vulnerability can be underlying
illnesses like depression, cardiovascular and cerebrovascular conditions and diabetes (Kovats & Hajat, 2008). Problems that may occur due to heat are exacerbating air pollution (Campbell-Lendrum & Corvalan, 2007; Rosenzweig et al., 2012) and increase the demand for energy in summers to cool buildings (Lemonsu et al., 2012). In other words, since heat waves in the 2000s, heat is increasingly forming a problem to cities in the Netherlands.
The UHI intensity is higher during certain times of the year. In a study about 1449 Chinese cities, the UHI intensity was consistently higher during summers (Li et al., 2019). A lower cloud cover and lower wind speed correlate with a higher UHI intensity in a study in Buenos Aires (Figuerola and Mazzeo, 1998). In a study about Amsterdam, the UHI intensity was higher during a heatwave (Van der Hoeven & Wandl, 2014). When downscaling the RCP 8.5, the yearly average UHI effect does not significantly intensify (Lauwaet et al., 2015). However, the number of days with a maximum
temperature >25 °C will double, and the days with a maximum temperature >30 °C will triple in the more extreme scenarios of the KNMI (Van der Hoeven & Wandl, 2014). Overall, the temperature will increase most in summer, leading to more summery days (Attema et al., 2014). Globally heat waves will become more frequent (Revi et al., 2014). Moreover, KNMI experts expect that heat waves will become more frequent in the Netherlands. So, climate change will increase the severity and impacts of the UHI effect due to more heat waves (Van der Hoeven & Wandl, 2014).
Concluding, heat waves will become more frequent, summery days will also become more frequent, and summers will become warmer as well in the future. It is interesting to research the UHI intensity during these circumstances because these are the moments that will impact human health and energy consumption. The objective of the study is to show to what degree meteorological extremes and seasonality affect the UHI effect in Amsterdam, which leads to the following research question.
4
What temporal changes are there in the urban heat island effect of Amsterdam from 2017-2021?
When considering:
• meteorological extremes (ideal weather, warm weather and a heatwave) • seasonal differences
5
Method
Comparison to other research
Weather stations are used in this research to provide the opportunity to examine variation in the UHI effect over time. It is impossible to analyse the UHI effect in an ideal experiment (i.e., measure without the city and then with the same factors measured with the city). That is the reason that several approaches have been created (Lowry, 1977). These approaches include mobile sensors, weather stations, remote sensing, vertical sensing, and energy balances (Gartland, 2008; Hawkins et al., 2004). The use of mobile sensors requires equipment that is not available. Calculating the energy balance is not possible due to the lack of available open-source data. Vertical sensing provides data on the effects of the UHI on the boundary layer (can reach a height of 1–1,4 km). Since the impact on people is of interest, the urban canopy layer is researched. Thus, vertical sensing does not provide the data that is of interest. Remote sensing results in land surface temperature, therefore it also does not give the urban canopy layer UHI effect. Remote sensing is also primarily used on days with a low cloud cover. Thus, it is hard to examine the difference of days with high cloud and low cloud cover. Remote sensing is useful when the spatial difference in the UHI effect is reviewed.
The research project is not the first study of the UHI effect in Amsterdam. Van der Hoeven and Wandl (2014) used remote sensing to access the UHI during a heatwave. Research has also been performed on how to mitigate the UHI effect in Den Haag and Utrecht (Kleerekoper et al., 2012). Steeneveld et al. (2011) examined the UHI effect in other Dutch cities by using amateur weather stations, but not in Amsterdam. So, some research is similar to this research. However, this research is still different from previously conducted research.
The data
The study area is the city of Amsterdam. The UHI effect in this region is sometimes more in accordance with that for a metropolitan urban area (the Randstad) of 7 million inhabitants (Van der Hoeven & Wandl, 2014). So, the UHI effect in this region is substantial. Furthermore, due to
urbanisation, it is expected that the UHI effect will increase (CPB & PBL, 2015; Gartland, 2008). The Randstad has not been studied because the distribution of available weather stations is even more limited if the entire Randstad is assessed (WOW-NL, 2021).
The data is acquired from WOW-NL (2021). WOW-NL (Weather Observations Website Nederland) is a big data project of the KNMI that provides open-source data from weather stations. The data gathered from this source is from amateur weather stations, which do not fully comply with the World Meteorology Organization standards. These weather stations offer accessible data for locations in urban areas as well as outside cities. The amateur weather stations are used because official weather stations (KNMI stations) are mainly located in rural areas (Steeneveld et al., 2011).
After inspecting the data for all available amateur weather stations in the region, three stations (AMS-IJ, Osdorp, & Amstelveen) have been selected that offer data from the urban area. So one of the weather stations is located in Amstelveen, which is outside the city of Amsterdam. This weather station is still being considered because it is so close to Amsterdam, the area is very urbanised, and this weather station is located near Schiphol. The weather station of Schiphol is a weather station that provides the temperature for the rural area. The closeness of these weather stations makes it less likely that meteorological differences could explain the temperature difference. Therefore, the location Amstelveen is likely to provide relevant data. The weather stations that provide the temperature for the rural area (Schiphol and Den Ilp) are situated outside of Amsterdam. Within the city’s boundaries, there are no weather stations that could provide data on the temperature, which could be considered rural. Additionally, the data is acquired from the weather station Schiphol, derived from KNMI (2021b). This is an official weather station, so it does adhere to the World Meteorology Organization
6 standards. Characteristics of all the weather stations are available in Table 1 and the locations of the weather stations that are used are visible in Figure 1. For all five weather stations that are used, there is data available from the 6th of February 2017 until now (the 28th of March 2021).
Table 1. Specification on the location of the weather stations that are used the information is from WOW-NL (2021).
After exporting the data from WOW-NL (2021), the data is imported into R Studio. The data from the WOW-NL has different intervals of available data per weather station. All have multiple data entries per hour. To make the data more uniform the average temperature has been determined for each day. After that, the UHI effect is calculated. The urban temperature (Turban) is the data from
AMS-IJ, Osdorp and Amstelveen. The rural temperature (Trural)is the data from Schiphol and Den Ilp. The
UHI effect is determined using the following formula. Calculations for all weather stations will result in six different UHI effects per day (e.g. AMSIJ-Schiphol, AMSIJ-DenIlp, Schiphol, Osdorp-DenIlp, Amstelveen-Schiphol, Amstelveen-DenIlp).
UHIeffect = Turban-Trural
Station name Station ID Elevation (m) Latitude Longitude
AMS-IJ 926656002 0 N 52.38 E 4.93
Osdorp 924276001 3 N 52.35 E 4.80
Amstelveen 933736001 -2 N 52.28 E 4.86
Ilp 960736001 0 N 52.46 E 4.91
Schiphol 916696001 -4 N 52.32 E 4.79
Figure 1. Locations of the weather stations. The weather stations in urban area
are AMS-IJ, Osdorp and Amstelveen. The weather stations in rural area Den Ilp and Schiphol (Google Earth, 2020).
7 After that, the data is also classified based on season, summery days, ideal days and heat wave days. To classify specific data, previously mentioned terms have to be operationalised. A heatwave day is defined as a day that is part of at least five consecutive days with maximum daily temperatures of over 25°C, including three days with maximum daily temperatures of over 30°C (KNMI, 2021c). A summery day is defined as a day with maximum daily temperatures of over 25°C (KNMI, 2021e). An ideal day is a day with a daily average wind speed below 3.3 m/s, which is still considered a weak wind (KNMI, 2021d) and a cloud cover that is less or equal to 1 eighth, which is considered virtually cloudless (KNMI, 2021a). The assignment of all these categories is determined based on data from Schiphol. The seasons are classified based on meteorological seasons: MAM (March, April & May), JJA (June, July & August), SON (September, October & November) and DJF (December, January & February).
Available data
Within this timeframe, there also a limited amount of data points based on classification (Table 2 & Appendix A for the dates). The global radiation, cloud cover, wind speed and wind direction are all from the weather station at Schiphol, so these are complete.
Table 2. Amount datapoints of UHI effect for each location. Dates from missing data in Appendix A.
AMSIJ Osdorp Amstelveen
UHI (all data) 1509 1459 1510
UHI (JJA) 368 368 347
UHI (summery) 113 101 113
UHI (ideal) 30 26 30
UHI (heat wave) 45 45 45
UHI (seasons) 1093 1043 1094
Subsequently, the statistical analysis can be done. First, the test for summery, ideal and heat wave day are discussed. The UHI effect summery, ideal and heat wave day is not always normally distributed. So, the tests that are performed are a Welch’s t-test (assumes normality) and a Wilcoxon rank-sum test (does not assume normality) (Ruxton, 2006). There is likely a higher UHI effect on summery, ideal and heat wave day. So, the alternative hypothesis for these tests is that the mean is greater on those days. The null hypothesis is that there is no difference between the UHI effects. The UHI effect on summery days, ideal days and heat wave days are tested against the UHI effect on all available data (6-2-2017 until 28-3-2021) excluding the days that are tested against. Additionally summery days, ideal days and heat wave days are tested against all data from JJA (excluding the days that it is tested against). To minimise the effect of autocorrelation every fifth day of all data and JJA are selected. A Welch’s t-test and a Wilcoxon rank-sum test are also performed on these subgroups. The mean of the resulting five p-values has been calculated to assess whether or not the null
hypothesis can be rejected or not. Secondly, the difference between seasons is assessed. The seasons of 2018, 2019 and 2020 are assessed because these are years in which all seasons are complete. To test a difference of the UHI effect between seasons, an ANOVA-test has been done with the Tukey-HSD as a post hoc test (Abdi & Williams, 2010).
After these tests, several statical analyses are done to explain the difference. A simple regression has been done to assess the maximum temperature (at Schiphol) as a predictor for the UHI effect, for all data and for each season. Furthermore, multiple regression tests are done to examine if global radiation and windspeed are good predictors for the UHI effect (Uyanık & Güler, 2013). These tests are both for all data and each season. Lastly, an ANOVA-test has been done with the Tukey-HSD as a post hoc test to test if there is a difference in the UHI effect per wind direction (Abdi & Williams, 2010).
8
Results
Summery days
Summery days are days with a maximum temperature of 25°C or higher measured at Schiphol (KNMI, 2021e). These days are tested against all data (6-2-2017 until 28-3-2021) and the
meteorological summer (JJA). In Figure 2a the distribution for the UHI effect is displayed for each group that is tested for all combinations of locations. Only relevant scores are given in the results section. The exact p-values for all tests are in Appendix B.
For the location of AMSIJ the mean UHIAMSIJ - Schiphol on summery days is 1.50 °C, and the
UHIAMSIJ- Den Ilp on summery days is 1.97 °C. The p-values for both Welch’s t-test and Wilcoxon test
are lower than 0.01. This also applies to when every fifth day is selected of the group all data & JJA. So, in cases for the location AMSIJ, when considering summery days is the null hypothesis rejected. Therefore the mean UHI on summery days also higher than on other days.
For the location Osdorp the mean UHIOsdorp - Schiphol on summery days is 1.14 °C, and the
UHIOsdorp- Den Ilp on summery days is 1.60 °C. The p-values for all these tests were also lower than 0.01.
So, also, at this location, the null hypothesis can be rejected.
Lastly, for Amstelveen the mean of the UHIAmstelveen - Schiphol on summery days is 1.07 °C and
the UHIAmstelveen- Den Ilp on summery days 1.54 °C. These means are higher in comparison to the mean
UHI for the rest of the days, which are 0.59 °C (Amstelveen - Schiphol) and 0.56 °C (Amstelveen - Den Ilp). When comparing summery days with JJA, the null hypothesis that there is no difference is accepted. For Amstelveen-Schiphol, the mean of the JJA is 1.40 °C, so that is even higher than on summery days. Only when using a Wilcoxon test and not selecting every fifth day, a p-value is found. This indicates a significant difference when comparing summery days and JJA. Compare to the other urban locations, the mean UHI effect on summery days is lowest in Amstelveen, and the mean UHI effect in JJA is highest in Amstelveen.
Several simple regression tests were done to examine the importance of the daily maximum temperature in relation to the UHI. For all locations, the daily maximum temperature is good predictor for the UHI effect if all data is used. For AMSIJ and Osdorp, the p-value is higher and sometimes not significant for winter and autumn. For Amstelveen, the intercept is below 0 for summer and autumn. Only for Amstelveen-DenIlp, the daily max temperature has an intercept above 0 and a significant p-value for summer. For all locations the intercept of daily max temperature in winters is below zero. The exact p-values and intercepts are in the table in Appendix C.
Ideal days
An ideal day is a day with a daily average wind speed below 3.3 m/s and a cloud cover that is less or equal to 1 eighth, measured at Schiphol. These days are tested against all data (6-2-2017 until 28-3-2021) and the meteorological summer (JJA). In Figure 2b, the distribution for the UHI effect is displayed for each group that is tested for all combinations of locations. Only relevant scores are given in the results section. The exact p-values for all tests are in Appendix B.
The mean UHIAMSIJ - Schiphol on ideal days is 2.17, and the UHIAMSIJ- Den Ilp on ideal days is 2.68
°C. The mean UHIOsdorp - Schiphol on ideal days is 1.46 °C and the UHIOsdorp- Den Ilp on ideal days 1.98 °C.
All tests for these two locations yielded p-values lower than <0.01. Therefore the null hypothesis, that there is no difference, can be rejected in all cases. For Amstelveen, the results of the statistical test are different. The mean UHIAmstelveen - Schiphol on ideal days is 0.55 °C. This is lower than the mean UHI for
all data, which is 0.63 °C, and the mean UHI for JJA, which is 1.37 °C. So, the UHIAmstelveen - Schiphol on
9 higher mean than on all days. The mean UHIAmstelveen- Den Ilp for JJA is 1.53 °C. So, the mean
UHIAmstelveen- Den Ilp is higher in JJA than on ideal days.
To test the importance of wind speed and global radiation further, several multiple regression tests were performed. For all location, when all data was tested, these yielded a p-value far below 0.01. So, both global radiation and daily mean wind speed are good predictors for the UHI effect. The multiple regression for each season also resulted in a p-value below 0.01, except for Amstelveen. Global radiation is in winter not a good predictor for Amstelveen-DenIlp and the intercept below 0. Global radiation is not a good predictor for Amstelveen-Schiphol in autumn , because the p-value is above 0.1. The exact p-values are in a table in Appendix C.
Heat wave days
A heatwave day is defined as a day that is part of at least five consecutive days with maximum temperatures of over 25°C, including three days with maximum temperatures of over 30°C (KNMI, 2021c). A summery day is defined as a day with maximum temperatures of over 25°C. These days are tested against all data (6-2-2017 until 28-3-2021) and the meteorological summer (JJA). In Figure 2c the distribution for the UHI effect is displayed for each group that is tested for all combinations of locations. Only relevant scores are given in the results section. The exact p-values for all tests are in Appendix B.
The mean UHIAMSIJ - Schiphol on heat wave days is 1.75, and the mean UHIAMSIJ- Den Ilp on heat
wave days is 2.23 °C. For Osdorp, the mean UHI effect on heat wave days is slightly lower, 1.36 °C (UHIOsdorp – Schiphol) and 1.83 °C (UHIOsdorp- Den Ilp). The means on heat wave days are significantly higher
in all tests for AMSIJ and Osdorp. For Amstelveen, the UHI effect on heat wave days is significantly higher compared to all data. The mean UHIOsdorp - Schiphol on heat wave days is 1.19 °C, and the
UHIOsdorp- Den Ilp on heat wave days is 1.66 °C. However, for UHIOsdorp - Schiphol, the mean UHI for JJA is
higher. For UHIOsdorp- Den Ilp, the p-value is lower than 0.1 when the Welch’s t-test and Wilcoxon test
are performed without subgrouping every fifth day. When every fifth day is selected, the p-value is slightly above 0.1.
10
Figure 2. Comparing the urban heat island effect of the six research locations on a) summery days, b) ideal days and c) heat wave days, to all data from 2017 – 2020 and to JJA of all years (excluding the days tested
against). Summery day = a day with maximum temperatures of over 25°C for Schiphol’s weather stations. Ideal day = a day with a daily average wind speed < 3.3 m/s & cover that is less or equal to 1 eighth for Schiphol’s weather stations. Heat wave day = a day that is part of at least five consecutive days with maximum temperatures of over 25°C, including three days with maximum temperatures of over 30°C for Schiphol’s weather stations.
11
Seasons
The time frame is 2018, 2019 and 2020. The seasons are tested for difference with an ANOVA test with the Tukey-HSD as a post hoc test. In Figure 3, the distribution for the UHI effect is displayed for the seasons per year. Only relevant scores and trends are given in this section. The exact p-values for the post-hoc (Tukey-HSD) are in Appendix D. When Den Ilp is the rural weather station, there seems to be a greater difference between seasons. There are more significant p-values (when α = 0.05). Thus, the significant p-values are more convincing evidence of a difference if these are also found when Schiphol is the rural weather station.
For UHIAMSIJ-Schiphol, the p-values are below 0.01 for spring-autumn, summer-autumn and
summer-winter but each for only one year. So, these significant differences are not consistent over the three years. With a timeframe of all three years there are no p-values below 0.05. For UHIAMSIJ-DenIlp,
p-values are below 0.05 for summer-winter and summer-autumn for all three years. When the data from all three years is tested there is a p-value below 0.05 for spring-autumn, spring-winter, summer-autumn and summer-winter found. For Osdorp, the p-values are below 0.05 for summer-winter, summer-autumn and spring-autumn for at least two years, with both Schiphol and Den Ilp as a rural weather station. With a timeframe of all three years there are p-value below 0.05 for spring-autumn, spring-winter, summer-autumn and summer-winter. Lastly, for Amstelveen, p-values lower are than 0.01 when comparing season to each other for all years except autumn-winter. Only for Amstelveen-DenIlp 2019 a p-value of 0.0016 is found for autumn-winter. When the data from all three years is tested there is a p-value below 0.01 for all seasons, except for UHIAmstelveen-Schiphol there is no significant
difference between autumn-winter.
To further study the seasonal difference, the wind direction and wind speed for each season are examined. The UHI in Amstelveen is highest when the wind comes from the North, which is most prevalent in summer and spring (Appendix E). The UHI in Amstelveen is lowest when the wind comes from the South, which is more common in autumn and winter. As mentioned before, the daily mean wind speed is a good predictor for UHI for each season and each location, indicated by p-values below 0.01. A frequency table with the distribution of wind direction per season and the boxplots per location are in Appendix E.
12
Figure 3. Difference between urban heat island effect for meteorological seasons for each year. Every boxplot (a-f) is a comparison between different locations. MAM= March, April & May. JJA= June, July &
13
Discussion
The question is whether or not there are temporal variations in the UHI effect of Amsterdam from 2017-2021. First, the findings for summery, ideal and heat wave days will be discussed. Then, the observed seasonal differences will be addressed. Next, the points of improvement for further research are discussed. Lastly, what the findings mean in the context of expected climatological changes is addressed.
For the locations AMSIJ and Osdorp, the UHI effect is significantly higher on summery days, ideal days and heat wave days, when compared to all data and to summers (Fig. 2 and Appendix B). The UHI effect in Amstelveen is not always higher on summery days, ideal days and heat wave days. The mean UHIAmstelveen-Schiphol in the summer is higher than that on summery days, ideal days and heat
wave days. So, the UHIAmstelveen-Schiphol is never significantly higher on summery days, ideal days and
heat wave days. The mean UHIAmstelveen-DenIlp of summers is only higher than ideal days, but a
significant difference was not observed when summery days and heat wave days are compared to summers. Autocorrelation probably played a role when the p-values were significant. Because when selecting every fifth day (for summers), the p-values were never below 0.01. Most noteworthy is that the UHIAmstelveen-Schiphol of ideal days is not even higher than the mean of all UHI Amstelveen-Schiphol data.
This is unexpected because wind speed and cloud cover are the meteorological parameters that influence the UHI effect the most (Figuerola and Mazzeo, 1998). When wind speed is low and cloud cover is low, the UHI effect also tends to be higher (Levermore et al., 2018; Mohammed et al., 2020). However, the UHI effect of Amstelveen has the most seasonality (Appendix F), and summer months have the highest UHI (Figure. 3). So, it is especially difficult for ideal days to result in a higher UHI, because these days are more distributed over the entire year. Summery days are mainly recorded at the end of spring, in summer and at the beginning of autumn. All heat wave days are in summer in this timeframe. For summery and heat wave days, it is also expected that these would result in significantly higher UHI intensity. Warmer weather, especially heat waves, results in a higher UHI effect (Van der Hoeven & Wandl, 2014). The results found for AMSIJ and Osdorp are mostly in line with other research, whereas the results for Amstelveen seem to contradict earlier research.
For the difference in seasons, the results also differ per location. The UHI intensity is greater in summer and spring (Schatz & Kucharik, 2014; Wu et al., 2019). For Amstelveen, there is a difference between each season except between winter and autumn (Appendix D). This is mostly in line with the expectation. The highest UHI effect is during summer. During summer and spring, there is also more wind from the North, where Amsterdam's city is located. The wind from the city is probably warmer than when it is from rural areas. The Amstelveen weather station is close to the southern border of Amstelveen. So wind from the south will cause a lower UHI. For the location Osdorp, the difference between season was mainly found when comparing summer to autumn and winter, and when comparing spring to autumn, which is somewhat in line with the expectation. These results were only not found for summer-winter 2019 and summer-autumn 2020 (for the UHI Osdorp-Schiphol), when the p <0.05 is used. For AMSIJ, there were few significant differences between seasons
when UHIAMSIJ-Schiphol is considered. For UHIAMSIJ-DenIlp the summer-winter and summer-autumn for all
year are significant, which is in line with expectation. The seasonal difference can also be observed in the trendline of the scatterplot in Appendix F.
The AMSIJ weather station is on the border of the IJ, north of the large open water of the IJ. Large open water can lead to cooling during the day (Cosgrove & Berkelhammer, 2018). The cooling effect during the day happens in spring and early summer. In the autumn, a lake can have a warming effect (Schatz & Kucharik, 2014). Since the difference of daily average temperature was used to measure the UHI effect and not hourly data, it is unclear if these effects play a role at this location. It could explain the lack of seasonal difference for AMSIJ. Further research in this area could determine if large open water's cooling and warming effect plays a role. When the wind direction is from the North, the UHI effect is larger. This is not what would be expected if the IJ would have a cooling
14 effect. Since AMSIJ is located to the South of the IJ. The wind direction from the North occurs most in summer & spring, so daily maximum temperature or the wind speed may explain the higher UHI better than wind direction. The Amstelveen weather station is also located on the cities' edges, so the observed UHI effect is likely lower (Van der Hoeven & Wandl, 2014). At Amstelveen, there does seem to be lower UHI intensity. The means of the UHI effect tend to be the lowest of the three locations. For AMSIJ, it does not seem to be the case. The means of the UHI effect tend to be the highest. This could indicate that the weather station on AMSIJ might be too closer to buildings. As noted before, these amateur weather stations do not comply with World Meteorology Organization standards.
In further research, more locations should be analysed so there is more of understanding of the spatial distribution of the UHI, because it is a very location specific phenomenon. The limited
timeframe is also a point of improvement. The number of summery, ideal and heat wave days was small, so UHI on those days was not always normally distributed. Especially, for heat wave days it would have been smart to also select every fifth day, because very likely these days show an autocorrelation. That there is likely greater UHI intensity on summery, ideal and heat wave days in these locations is important to know because these more extreme weather phenomena are likely to increase.
Extreme weather, like heat waves, will likely increase (Russo, 2014). The number of days with a maximum temperature of 20 °C will likely double, and days with a maximum temperature of 30 °C or higher will likely triple in the extreme scenarios of the KNMI (Van der Hoeven & Wandl, 2014). Summers are already the warmer months and have already a higher UHI intensity. The temperature for summers will increase, probably leading to even greater UHI intensity during those months (Attema et al., 2014). Lastly, wind speed in the region of Schiphol has decreased (Stepek et al., 2013). If this trend continues, it will increase the UHI effect. If the UHI intensity gets higher the negative impacts will likely increase too. These effects on health, and energy consumption will likely increase if there is no adaptation.
15
Conclusion
Temporal variations in the UHI effect of Amsterdam are found for the period 2017-2021. The results are not the same for each location because the UHI effect also can be highly location-specific. A higher UHI effect is observed for summery days, ideal days and heat wave days compared to all data from over three years. When comparing these days to summer, the UHI effect is only
significantly higher in Osdorp and AMSIJ. For Amstelveen, the UHI intensity is possibly lower due to the location near the city's edge. In Amstelveen, there is a large seasonal difference, with the highest UHI intensity in summer. For AMSIJ, the seasonal difference is the lowest. The IJ near the AMSIJ location probably has a tempering effect of the UHI effect. Thus there is little seasonal difference. For the location Osdorp, seasonal differences have been found, although not all differences were
statistically significant. A higher UHI intensity is expected for summery days, ideal days and heat wave days and in spring and summer. The observations are somewhat in line with this expectation or can be explained by location-specific characteristics.
These observation matter, because meteorological conditions similar to summery days, ideal days and heat wave days are likely to increase. Moreover, summers are also expected to get warmer. Thus, the UHI intensity is likely to increase in the future.
Data Availability
Datasets and R code related to this article can be found at https://doi.org/10.6084/m9.figshare.c.5444367
16
References
Abdi, H., & Williams, L. J. (2010). Tukey’s honestly significant difference (HSD) test. Encyclopedia
of research design, 3(1), 1-5.
Attema, J., Bakker, A., Beersma, J., Bessembinder, J., Boers, R., Brandsma, T., ... & van Zadelhoff, G. J. (2014). KNMI’14: Climate Change scenarios for the 21st Century–A Netherlands
perspective. KNMI: De Bilt, The Netherlands.
http://publicaties.minienm.nl/documenten/knmi-14-climate-change-scenarios-for-the-21st-century-a-netherla
Campbell-Lendrum, D., & Corvalán, C. (2007). Climate Change and Developing-Country Cities: Implications For Environmental Health and Equity. Journal of Urban Health, 84(S1), 109– 117. https://doi.org/10.1007/s11524-007-9170-x
Cosgrove, A., & Berkelhammer, M. (2018). Downwind footprint of an urban heat island on air and lake temperatures. Npj Climate and Atmospheric Science, 1(46).
https://doi.org/10.1038/s41612-018-0055-3
CPB & PBL. (2015). Nederland in 2030 en 2050: Twee referentiescenario’s. Retrieved from https://www.cpb.nl/sites/default/files/publicaties/download/cpb-pbl-boek-19-wlo-2015-nederland-2030-en-2050.pdf
Figuerola, P. I., & Mazzeo, N. A. (1998). Urban-rural temperature differences in Buenos Aires.
International Journal of Climatology, 18(15), 1709–1723.
https://doi.org/10.1002/(SICI)1097-0088(199812)18:15<1709::AID-JOC338>3.0.CO;2-I
Garssen, J., Harmsen, C., & De Beer, J. (2005). The effect of the summer 2003 heat wave on mortality in the Netherlands. Eurosurveillance, 10(7), 13–14. https://doi.org/10.2807/esm.10.07.00557-en
Gartland, L. (2008). Heat Islands: understanding and mitigating heat in urban areas. Earthscan. London, 215 pp.
Google Earth. (the 15th of September, 2020). Amsterdam, the Netherlands. 52° 22' 11.32" N, 4° 51' 36.09" E. Retrieved March 26, 2021.
Hawkins, T. W., Brazel, A. J., Stefanov, W. L., Bigler, W., & Saffell, E. M. (2004). The Role of Rural Variability in Urban Heat Island Determination for Phoenix, Arizona, Journal of Applied
Meteorology, 43(3), 476-486.
https://doi.org/10.1175/1520-0450(2004)043<0476:TRORVI>2.0.CO;2
Kleerekoper, L., van Esch, M., & Salcedo, T. B. (2012). How to make a city climate-proof, addressing the urban heat island effect. Resources, Conservation and Recycling, 64, 30–38.
https://doi.org/10.1016/j.resconrec.2011.06.004
KNMI. (2021a). Achtergrond daggegevens van het weer. Retrieved March 26, 2021, from https://www.knmi.nl/kennis-en-datacentrum/achtergrond/daggegevens-van-het-weer KNMI. (2021b). Daggegevens van het weer in Nederland. Retrieved March 26, 2021, from
https://www.knmi.nl/nederland-nu/klimatologie/daggegevens
KNMI. (2021c). Hittegolven. Retrieved March 26, 2021, from https://www.knmi.nl/nederland-nu/klimatologie/lijsten/hittegolven
KNMI. (2021d). Windschaal van Beaufort. Retrieved March 26, 2021, from
17 KNMI. (2021e). Warme dagen. Retrieved March 26, 2021, from
https://www.knmi.nl/kennis-en-datacentrum/uitleg/warme-dagen
Kovats, R. S., & Hajat, S. (2008). Heat Stress and Public Health: A Critical Review. Annual Review of
Public Health, 29(1), 41–55. https://doi.org/10.1146/annurev.publhealth.29.020907.090843
Lauwaet, D., Hooyberghs, H., Maiheu, B., Lefebvre, W., Driesen, G., Van Looy, S., & De Ridder, K. (2015). Detailed Urban Heat Island Projections for Cities Worldwide: Dynamical
Downscaling CMIP5 Global Climate Models. Climate, 3(2), 391–415. https://doi.org/10.3390/cli3020391
Lemonsu, A., Kounkou-Arnaud, R., Desplat, J., Salagnac, J.-L., & Masson, V. (2012). Evolution of the Parisian urban climate under a global changing climate. Climatic Change, 116(3–4), 679– 692. https://doi.org/10.1007/s10584-012-0521-6
Levermore, G., Parkinson, J., Lee, K., Laycock, P., & Lindley, S. (2018). The increasing trend of the urban heat island intensity. Urban Climate, 24, 360–368.
https://doi.org/10.1016/j.uclim.2017.02.004
Li, Y., Wang, L., Zhang, L., Liu, M., & Zhao, G. (2019). Monitoring Intra-annual Spatiotemporal Changes in Urban Heat Islands in 1449 Cities in China Based on Remote Sensing. Chinese
Geographical Science, 29(6), 905–916. https://doi.org/10.1007/s11769-019-1080-9
Lowry, W. P. (1977). Empirical estimation of urban effects on climate: a problem analysis. Journal of
Applied Meteorology and Climatology, 16(2), 129-135.
https://doi.org/10.1175/1520-0450(1977)016<0129:EEOUEO>2.0.CO;2
Mohammed, A., Pignatta, G., Topriska, E., & Santamouris, M. (2020). Canopy Urban Heat Island and Its Association with Climate Conditions in Dubai, UAE. Climate, 8(6), 81.
https://doi.org/10.3390/cli8060081
Revi, A., Satterthwaite, D.E., Aragón-Durand, F., Corfee-Morlot, J., Kiunsi, R.B.R., Pelling, M., Roberts, D.C., & Solecki W. (2014). Urban areas. In: Climate Change 2014: Impacts,
Adaptation, and Vulnerability. Part A: Global and Sectoral Aspects. Contribution of Working Group II to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Field, C.B., V.R. Barros, D.J. Dokken, K.J. Mach, M.D. Mastrandrea, T.E. Bilir, M. Chatterjee, K.L. Ebi, Y.O. Estrada, R.C. Genova, B. Girma, E.S. Kissel, A.N. Levy, S. MacCracken, P.R. Mastrandrea, and L.L.White (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 535-612.
RIVM. (2017). Stedelijk hitte-eiland effect (UHI). Retrieved from
https://www.atlasleefomgeving.nl/stedelijk-hitte-eiland-effect-uhi
Rosenzweig, C., Solecki, W. D., Hammer, S. A., & Mehrotra, S. (2012). Climate Change and Cities: First Assessment Report of the Urban Climate Change Research Network. Cambridge, UK: Cambridge University Press. https://doi.org/10.1017/CBO9780511783142
Russo, S., Dosio, A., Graversen, R. G., Sillmann, J., Carrao, H., Dunbar, M. B., . . . Vogt, J. V. (2014). Magnitude of extreme heat waves in present climate and their projection in a warming world. Journal of Geophysical Research: Atmospheres, 119(22), 12,500-12,512.
https://doi.org/10.1002/2014jd022098
Ruxton, G. D. (2006). The unequal variance t-test is an underused alternative to Student’s t-test and the Mann–Whitney U test. Behavioral Ecology, 17(4), 688–690.
18 Schatz, J., & Kucharik, C. J. (2014). Seasonality of the Urban Heat Island Effect in Madison,
Wisconsin. Journal of Applied Meteorology and Climatology, 53(10), 2371–2386. https://doi.org/10.1175/jamc-d-14-0107.1
Steeneveld, G. J., Koopmans, S., Heusinkveld, B. G., van Hove, L. W. A., & Holtslag, A. A. M. (2011). Quantifying urban heat island effects and human comfort for cities of variable size and urban morphology in the Netherlands. Journal of Geophysical Research, 116(D20).
https://doi.org/10.1029/2011jd015988
Stepek, A., Wichers Schreur, B., & Wijnant, I. (2013, July 22). Variatie van het windklimaat van de luchthaven Schiphol in relatie tot het baangebruik. Retrieved May 11, 2021, from
https://edepot.wur.nl/337283
Uyanık, G. K., & Güler, N. (2013). A Study on Multiple Linear Regression Analysis. Procedia -
Social and Behavioral Sciences, 106, 234–240. https://doi.org/10.1016/j.sbspro.2013.12.027
Van der Hoeven, F., & Wandl, A. (2014). Amsterwarm: Mapping the landuse, health and energy-efficiency implications of the Amsterdam urban heat island. Building Services Engineering
Research and Technology, 36(1), 67–88. https://doi.org/10.1177/0143624414541451
WOW-NL. (2021). WOW-NL: Jouw weer op de kaart! Retrieved March 26, 2021, from https://wow.knmi.nl/
Wu, X., Zhang, L., & Zang, S. (2019). Examining seasonal effect of urban heat island in a coastal city.
19
Appendix A
Table 3. Dates of missing data for each weather station.
Den Ilp Amstelveen AMSIJ Osdorp
Temperature 26/4/2020 & 27/4/2020 26/4/2020 & 27/4/2020 9/3/2020, 26/4/2020 & 27/4/2020
2/5/2018 - 21/6/2018, 26/4/2020 & 27/4/2020
UHI (for summery days) - - - 6/5/2018 - 8/5/2018,
14/5/2018, 23/5/2018, 26/5/2018 -30/5/2018, 6/6/2018 & 7/6/2018
UHI (for ideal days) - 26-4-2020 26-4-2020 3/5/2018, 4/5/2018,
7/5/2018, 8/5/2018 & 26/4/2020
UHI (for heat wave days) - - - -
UHI (for season) - 26/4/2020 & 27/4/2020
(MAM) 9/3/2020, 26/4/2020 & 27/4/2020 (MAM) 2/5/2018 - 31/5/2018 (MAM) 1/6/2018 - 21/6/2018 (JJA), 26/4/2020 & 27/4/2020 (MAM)
20
Appendix B
Table 4. Test results when comparing summery, ideal and heat wave days with all data and JJA for each location. Fractions larger than 1.00E-4 (0.01%) have been expressed as percentages. Percentages larger than 10% are rounded. P-values that are not significant at the 5% significance level are highlighted.
x y test p-value (AMSIJ-DenIlp) p-value (AMSIJ-Schiphol) p-value (Osdorp-Schiphol) p-value (Osdorp-DenIlp) p-value (Amstelveen-Schiphol) p-value (Amstelveen-DenIlp)
summerday all data Welch’s t-test 4.5E-12 1.5E-23 8.2E-18 9.3E-23 2.5E-08 1.0E-22
Wilcoxon test 6.4E-14 1.7E-27 3.0E-21 2.8E-30 2.2E-09 4.1E-26
Welch’s t-test (5th
day) 1.8E-10 3.1E-22 4.3E-15 3.1E-21 4.8E-07 5.1E-20
Wilcoxon test (5th day) 1.0E-11 3.1E-22 5.5E-16 3.4E-23 6.3E-08 1.1E-19
JJA Welch’s t-test 4.9E-11 4.7E-19 1.1E-16 1.7E-19 100% 24%
Wilcoxon test 2.4E-11 2.9E-19 2.9E-17 4.8E-22 100% 4.44%
Welch’s t-test (5th day) 3.8E-06 1.2E-09 3.6E-11 6.8E-14 100% 71%
Wilcoxon test (5th day) 3.5E-06 4.1E-09 1.8E-09 5.4E-11 100% 45%
idealday all data Welch’s t-test 7.6E-10 7.7E-17 3.6E-07 4.7E-10 74% 0.72%
Wilcoxon test 9.2E-13 1.5E-16 1.6E-10 1.1E-12 69% 0.35%
Welch’s t-test (5th day) 4.5E-10 7.8E-19 4.2E-07 2.0E-10 68% 0.43%
Wilcoxon test (5th day) 2.7E-12 1.0E-15 1.0E-09 6.7E-12 58% 0.23%
JJA Welch’s t-test 3.5E-09 6.4E-16 3.2E-06 2.6E-08 100% 100%
Wilcoxon test 2.9E-11 8.1E-14 9.5E-09 9.7E-10 100% 100%
Welch’s t-test (5th day) 5.9E-08 1.5E-15 6.9E-06 7.2E-09 100% 99%
Wilcoxon test (5th day) 1.3E-08 5.7E-11 3.5E-07 3.5E-07 100% 99%
heatwaveday all data Welch’s t-test 1.9E-10 4.9E-13 6.5E-13 2.2E-11 2.4E-05 7.1E-15
Wilcoxon test 1.1E-11 2.0E-15 6.4E-16 8.8E-16 1.5E-05 1.9E-13
Welch’s t-test (5th
day) 1.6E-10 1.1E-13 1.2E-12 1.9E-11 3.5E-05 1.3E-15
Wilcoxon test (5th day) 3.1E-11 1.0E-14 6.5E-14 3.1E-14 1.4E-05 1.8E-12
JJA Welch’s t-test 4.4E-10 4.9E-11 3.5E-12 6.4E-08 91% 8.36%
Wilcoxon test 1.0E-10 4.9E-12 2.4E-14 1.9E-09 98% 8.94%
Welch’s t-test (5th
day) 3.2E-07 4.5E-08 4.9E-10 6.1E-07 87% 15%
21
Appendix C
Table 5. Simple regression p-values and intercepts for test between UHI effect and max temperature and multiple regression p-values and intercepts for test between UHI effect and daily mean windspeed & global radiation. Fractions larger than 1.00E-4 (0.01%) for the p-values have been expressed as percentages. P-values percentages larger than 10% are rounded (highlighted values are not significant at P<5%).
AMSIJ-Schiphol AMSIJ-DenIlp Osdorp-Schiphol Osdorp-DenIlp Amstelveen-Schiphol Amstelveen-DenIlp
time predictor(s) intercept p-value intercept p-value intercept p-value intercept p-value intercept p-value intercept p-value
all year max.temperature 1.8E-02 7.6E-12 4.2E-02 5.1E-43 2.1E-02 1.2E-21 4.5E-02 1.9E-53 4.3E-02 1.3E-50 6.7E-02 1.0E-90
DJF max.temperature -4.6E-03 65% -7.9E-03 45% -1.5E-02 6.8% -1.9E-02 6.0% -7.3E-02 6.4E-23 -7.6E-02 3.6E-15
MAM max.temperature 2.7E-02 7.9E-06 5.2E-02 1.3E-13 3.7E-02 6.2E-11 6.4E-02 5.0E-18 2.1E-02 7.9E-04 4.6E-02 1.4E-13
JJA max.temperature 7.0E-02 5.1E-15 1.2E-01 1.5E-28 8.5E-02 5.8E-29 1.4E-01 1.7E-40 -2.0E-02 1.9% 2.9E-02 4.3E-04
SON max.temperature 1.0E-02 23% 2.2E-02 2.3% 1.6E-02 1.2% 2.8E-02 4.8E-04 -1.3E-02 5.1% -1.7E-03 83%
all year dailymeanwindspeed -1.5E-01 5.8E-84 -1.6E-01 2.2E-82 -1.2E-01 3.5E-82 -1.3E-01 7.4E-63 -6.9E-02 2.6E-17 -7.7E-02 4.7E-20
global radiation 2.0E-04 1.1E-24 4.5E-04 1.0E-94 2.3E-04 8.4E-46 4.9E-04 7.9E-112 4.9E-04 7.0E-99 7.4E-04 7.5E-185
DJF dailymeanwindspeed -1.0E-01 2.4E-20 -1.0E-01 9.5E-18 -9.7E-02 8.0E-26 -9.5E-02 5.8E-16 -6.4E-02 7.8E-09 -6.1E-02 1.8E-05
global radiation 1.4E-03 7.5E-34 1.5E-03 6.0E-34 1.1E-03 1.9E-32 1.2E-03 7.7E-24 -2.9E-04 6.8E-03 -2.0E-04 15%
MAM dailymeanwindspeed -1.2E-01 4.7E-18 -1.2E-01 2.7E-16 -1.1E-01 1.4E-22 -1.1E-01 1.1E-14 -1.1E-01 2.4E-11 -1.1E-01 1.8E-13
global radiation 3.0E-04 1.8E-13 5.4E-04 8.4E-33 2.8E-04 1.2E-15 5.1E-04 3.5E-28 1.3E-04 3.6E-03 3.7E-04 1.4E-18
JJA dailymeanwindspeed -2.0E-01 3.3E-27 -2.4E-01 7.2E-34 -1.4E-01 1.7E-17 -1.9E-01 4.5E-20 -5.3E-02 6.3E-03 -1.0E-01 1.3E-10
global radiation 3.3E-04 2.9E-12 6.1E-04 6.9E-30 3.5E-04 2.0E-15 6.4E-04 7.5E-29 1.6E-04 2.5E-03 4.5E-04 2.6E-23
SON dailymeanwindspeed -2.0E-01 3.5E-33 -2.2E-01 5.1E-31 -1.6E-01 9.4E-34 -1.7E-01 1.6E-26 -1.0E-01 1.1E-10 -1.2E-01 4.6E-10
22
Appendix D
Table 6. The p-value of the Tukey-HSD test between different seasons for AMSIJ (p-values that are significant
at p<0.05 are highlighted). AMSIJ-Schiphol AMSIJ-DenIlp time 2018 2019 2020 2018-2020 2018 2019 2020 2018-2020 MAM-JJA 0.615 0.854 0.277 0.997 0.032 0.831 0.122 0.839 MAM-SON 0.858 0.259 0.034 0.136 0.867 0.122 0.000 0.000 MAM-DJF 0.408 0.998 0.168 0.163 0.000 0.057 0.000 0.000 JJA-SON 0.974 0.043 0.785 0.087 0.003 0.013 0.000 0.000 JJA-DJF 0.030 0.771 0.993 0.105 0.000 0.005 0.000 0.000 SON-DJF 0.092 0.346 0.911 1.000 0.003 0.988 0.989 0.102
Table 7. The p-value of the Tukey-HSD test between different seasons for Osdorp (p-values that are significant at p<0.05 are highlighted). Osdorp-Schiphol Osdorp-DenIlp time 2018 2019 2020 2018-2020 2018 2019 2020 2018-2020 MAM-JJA 0.829 0.571 0.317 0.996 0.037 0.669 0.225 0.825 MAM-SON 0.046 0.046 0.007 0.000 0.004 0.046 0.000 0.000 MAM-DJF 0.367 0.819 0.000 0.000 0.000 0.010 0.000 0.000 JJA-SON 0.002 0.001 0.409 0.000 0.000 0.001 0.000 0.000 JJA-DJF 0.045 0.138 0.000 0.000 0.000 0.000 0.000 0.000 SON-DJF 0.696 0.310 0.034 1.000 0.839 0.956 0.018 0.072
Table 8. The p-value of the Tukey-HSD test between different seasons for Amstelveen (p-values that are significant at p<0.05 are highlighted).
Amstelveen-Schiphol Amstelveen-DenIlp time 2018 2019 2020 2018-2020 2018 2019 2020 2018-2020 MAM-JJA 0.000 0.000 0.000 0.000 0.000 0.000 0.003 0.000 MAM-SON 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 MAM-DJF 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 JJA-SON 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 JJA-DJF 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 SON-DJF 0.765 0.410 0.467 0.637 0.856 0.002 0.080 0.001
23
Appendix E
Table 9. Frequency table of the amount of days for each wind direction for each season.
North East South West
MAM 80 107 76 133
JJA 78 60 81 149
SON 43 66 157 98
DJF 26 65 162 131
Table 10. The average wind speed for each season, calculated from the daily wind measured at Schiphol.
Average wind speed (m/s)
MAM 4.94
JJA 4.36
SON 4.76
24
Figure 4. Comparison of the urban heat island effect for each wind direction, every boxplot (a-f) is a comparison between different locations. North = when the wind direction at Schiphol is > 315° and <= 45°. East =
25
Appendix F
Figure 5. The urban heat island effect over total timeframe, for every plot (a-f) is a comparison between different locations. Summery day = a day with maximum temperatures of over 25°C for Schiphol’s weather
stations. Ideal day = a day with a daily average wind speed < 3.3 m/s & cover that is less or equal to 1 eighth for Schiphol’s weather stations. Heat wave day = a day that is part of at least five consecutive days with maximum temperatures of over 25°C, including three days with maximum temperatures of over 30°C for Schiphol’s weather stations.
