SPATIAL HOTSPOT ANALYSIS OF BUCHAREST S URBAN HEAT ISLAND (UHI) USING MODIS DATA

Transcription

1 Annals of Valahia University of Targoviste. Geographical Series (2018), 18(1): DOI: /avutgs ISSN (Print): , ISSN (Online): Copyright by Department of Geography. Valahia University of Targoviste SPATIAL HOTSPOT ANALYSIS OF BUCHAREST S URBAN HEAT ISLAND (UHI) USING MODIS DATA Georgiana GRIGORAȘ 1, Bogdan URIȚESCU 2 1 National Institute for Aerospace Research "Elie Carafoli" - INCAS, Bvd. Iuliu Maniu no. 220, 6th District, , Bucharest, Romania; grigoras.georgiana@gmail.com 2 University of Bucharest, Faculty of Geography, Bvd. Nicolae Bălcescu no.1, 1st District, , Bucharest, Romania; bogdan_uri@yahoo.com Abstract The aim of the study is to find the relationship between the land surface temperature and air temperature and to determine the hot spots in the urban area of Bucharest, the capital of Romania. The analysis was based on images from both moderate-resolution imaging spectroradiometer (MODIS), located on both Terra and Aqua platforms, as well as on data recorded by the four automatic weather stations existing in the endowment of The National Air Quality Monitoring Network, from the summer of Correlation coefficients between land surface temperature and air temperature were higher at night ( ) and slightly lower during the day ( ). After the validation of satellite data with in-situ temperature measurements, the hot spots in the metropolitan area of Bucharest were identified using Getis-Ord spatial statistics analysis. It has been achieved that the "very hot" areas are grouped in the center of the city and along the main traffic streets and dense residential areas. During the day the "very hot spots represent 33.2% of the city's surface, and during the night 31.6%. The area where the mentioned spots persist, falls into the "very hot spot" category both day and night, it represents 27.1% of the city's surface and it is mainly represented by the city center. Keywords: Land Surface Temperature, Air temperature, MODIS, satellite data, hot spot, Getis-Ord statistics, spatial analysis 1. INTRODUCTION The difference between air temperatures and surface temperatures within the city as compared to suburban/rural areas of the city is known as the Urban Heat Island (UHI) (Deilami et al., 2018; Oke, 1982, 1976). Expansion of cities and the land use changes (LULC) have led to the intensification of the Urban Heat Island phenomenon in different cities around the world (Fu and Weng, 2016; Gaur et al., 2018; Nastran et al., 2018; Singh et al., 2017; Zhang et al., 2017). The Urban Heat Island phenomenon has been extensively studied in recent decades both in the meteorological approach, where the air temperature above the city is studied (UHI), as well as analyzing surface heating that covers the land (SUHI) (Arnfield, 2003; Mohajerani et al., 2017; Stewart, 2011), due to the desire of understanding the mechanisms that intensify it and needing to find and apply measures to mitigate it. Increasing the temperature in the urban environment creates problems for human health through the influence it has on: deterioration of air quality, contributing to increased tropospheric ozone concentrations, increasing the energy consumption of the population, contribution to the intensification of heat waves (Chapman et al., 2017; Debbage and Shepherd, 2015). In Bucharest, the UHI phenomenon was highlighted and studied both in terms of air temperature by direct measurements (Tumanov et al., 1999) and using numerical modeling for weather forecast (Iriza et al., 2017) as well as on surface temperature using its extension (Cheval and Dumitrescu, 2015, 2009) as well as the connection with certain categories of land us (Cheval et 14

2 al., 2009; Iojă et al., 2014), using satellite data (Zoran, 2011). Continuing on this topic, the present paper aims to determine the hot spots of the Bucharest's UHI using a spatial statistical analysis applied to MODIS satellite data, representing Land Surface Temperature (LST) values from the summer of DATA AND METHODS The studied area is represented by Municipality of Bucharest and its surrounding areas, a total of 25 x 25 km. Version 6 of MOD11_L2 and MYD11_L2 products were used, which provides land surface temperature (LST) and emissivity with spatial resolution of 1 km. These products contain data in HDF-EOS format (Hierarchical Data Format) recorded by MODIS instruments (moderate-resolution imaging spectroradiometer) which operates on both the Terra and Aqua spacecraft (and have been downloaded from (NASA, 2018). Data with good quality and no cloud cover were selected for the analyzed area, from June, July and August of 2017, the second most warm year in the history of worldwide weather records. Satellite data were compared with the temperatures recorded at the air quality monitoring stations in Bucharest which belongs to the National Air Quality Monitoring Network (ANPM, 2018). The locations of the stations from which the hourly temperature data was used are shown in Figure 1. Station B1 is located on the shore near Lake "Morii", station B4 is located in the "Bagdasar-Arseni" hospital garden, station B5 is located in "Drumul Taberei" park, and B7 is located in the courtyard of the research platform in Magurele, where a number of research institutes are located. Figure 1: Study area with Land Use/Land cover classification from Urban Atlas (Copernicus Land Monitoring Service); location of stations where air temperature measurements were available The relationship between surface temperatures derived from satellite data and air temperature measured in situ was determined by calculating the Pearson correlation coefficient (r): 15

3 Where: şi - average of variables X and Y; si - variables values with i=1,,n. Pearson's r can range from -1 to indicates a perfect negative linear relationship between variables,1 indicates a perfect positive linear relationship between variables and 0 indicates no linear relationship between variables. Hot spot analysis has been achieved by calculating the Getis-Ord statistic for surface temperature in context with neighboring cell temperatures. The value is an z-score and shows where characteristics with high or low values are clustered. To be a statistically significant hot spot, a feature will have a high value and at the same time it will be surrounded by other features with high values. The Getis-Ord statistics are calculated according to the formula (ESRI, 2018): where i is the resultant statistics (z-scores and p- values) for pixel i, xj is the LST value for pixel j, w i,j is the spatial weight between pixel i and neighboring pixel j, n is equal to the total number of pixels, and and S are mean and variance: and The output of the Gi* statistic (the z-score) represents the statistical significance of clustering for a specified distance (ESRI, 2018). The z score was compared with the range of the values for seven confidence levels (table 1): (values less than -2.58), (values with range from to -1.96), -0.1 (values with range from to -1.65), 0 (values with range from to 1.65), 0.1 (values with range from 1.65 to 1.96), 0.05 (values with range from 1.96 to 2.58) and 0.01 (values greater than 2.98). The seven levels correspond to seven classes that LST values were assigned to; most important for analyzing are the very cold spot and very hot spot, classes that defines the areas with extreme values. Table 1. Classification based on p value and z score Significance Level (p Value) Critical Value (z Score) Class No Class name < Very cold spot Cold spot Cool Spot Not significant Warm spot Hot spot 0.01 > Very hot spot 16

4 3. RESULTS AND DISCUSSION In order to analyze the hotspots of the Bucharest's Urban Heath Island, viewed as a whole, both in terms of air temperatures and surface temperatures, the following steps were taken: (i) in the first step the measured data from the ground stations were analyzed; (ii) in the second step the satellite data was validated by determining the relationship between the air temperatures, available from the air quality stations, and land surface temperatures determined from satellite data; (iii) finally, hot spots are identified using Getis Ord Gi * statistics available in ArcGIS. Data from 4 stations were used, 3 located on the territory of Bucharest (B1, B4, B5) and one in the suburban area (B7). The statistics for these data are presented in Table 2. Table 2. Temperature statistics for the 4 stations Statistic B1 B4 B5 B7 parameters min max avg stdev It is noted that on average the values of the temperatures are approximately the same at all stations. The largest deviation from the average is the station located in the suburban area, because, on the one hand, here, during the night, the recorded temperatures values are lower than those inside the city, on the other hand this station is located very close to the road and buildings, so the local influences increase the thermal extremes during the day. The smallest deviation from average is obtained at station B1, located near "Morii" Lake, where the air temperature is influenced by the water temperature, and is warmer at night and colder during the day. Also, an inter-stations comparative analysis of the measured hourly values was performed Strong correlations with Person's correlation coefficient values above 0.95 were obtained for all the analyzed situations. These results are shown in Table 3 and indicate that the measured temperatures have the same trend of variation, not being particularly affected by local factors. The lowest value of the correlation coefficient (0.958) was obtained between data sets from stations B1 and B7, stations where the thermal extremes are higher, as discussed above. Table 3. Correlation coefficients between temperature values recorded in the 4 stations Name of stations B1 B4 B5 B7 B B B B Using MODIS satellite products and ArcGIS software, maps of average surface temperature were made for the summer of These are shown in Figure 2 and highlight the Bucharest's Urban Heat Island. Analyzing Figure 2, it is noticed that the temperatures have higher values inside the city, in the suburban areas or in the settlements with compact areas of residence and the phenomenon of the UHI is highlighted. Temperatures within the boundaries of Bucharest Municipal area are higher than in suburban areas and in those in the mentioned localities. Also, it is noticed that the temperature difference between the city and the suburban areas is lower during the day compared to the night. From satellite data LST values were extracted at the corresponding points to the locations of the stations where the air temperature is measured. For each location, sets of pairs data (air temperature - LST from MODIS) were compared and analyzed. 17

5 Linear regression analysis is used to find equations that fits data from ground stations to satellite ones. The relationship between the two sets of variables was determined for both daytime and nighttime. To measure the strength of the linear relationship between the two variables, the correlation coefficient Person, for each individual case, was determined. The resulting regression equations and correlation coefficient values are shown in Table 4. Figure 2. Mean Land Surface Temperature determined from MODIS day (left), and for the night (right), for the summer of 2017 Table 4. Correlation coefficient values and regression equations obtained by comparing air temperatures with MODIS-derived land surface temperatures at locations corresponding to air quality monitoring stations station part of the day regression equation r B1 B4 B5 B7 day y= x 0.72 night y= x 0.84 day y= x 0.72 night y= x 0.80 day y= x 0.71 night y= x 0.87 day y=8+0.79x 0.77 night y= x 0.86 Between the two data sets there were obtained very good correlations (over 0.7), since correlated values represent, on the one hand, the air temperatures measured at one point and on the other hand, the temperature of the surface mediated by a pixel of 1 km 2. Lower values of the correlations for the daytime compared to those for the nighttime were obtained because during the day the surfaces exposed to solar radiation are heated differently, and have higher temperatures than those of the air. After sunset the surfaces that have absorbed solar radiation during the day and have stored heat, they cool off re-emitting the heat, so during the night, in urban areas the land surface 18

6 temperatures and air temperatures have close values (compared to those in daytime). Thus, for the night time, correlation coefficients of over 0.8 are obtained between data sets for all stations. Figure 3. Scatter plots Air Temperature versus Land Surface Temperature for day-time (left) and night-time (right), for the location of air quality monitoring stations B1, B4, B5, B7 The results of hot spot analysis performed using Getis Ord Gi * spatial statistics are shown in Figure 4. The analysis was carried out differential for both periods during the day and at night, determining their position and extension. From the analyzed area, the surfaces classified as: worm spot represent 11.7% of the surface during the day and 5.4% - night, hot spot represent 8%-day and 4.5% night, while very 19

7 hot spot represent 5.6% - day and 5.8% - night. Also, during nighttime these areas are compact and clearly delineate the urban area. During the day, the hot spots area has interruptions due to the influences exerted by the water-covered areas and the green areas, and follows the shape of districts and industrial areas, built along major traffic routes. The three classes representing cold spots have the following percentages values of total surface area: cool spot 28.6% - day and 18.2% - night, cold spot 21% - day and night, very cold spot 4.2 % - day and 24.8% - night. The class not significant in which higher or lower temperatures are not surrounded by other similar temperatures (there are not clustered), represent 20.9%, from the analyzed area during daytime and 10.7% for the nighttime. Figure 4. Hot spot analysis in Bucharest area during daytime (left), nighttime (right) Using spatial analysis, the persistent "very hot spot" area, which occurs both during the day and the night, was determined. This area is delineated with a red contour in Figure 5, it has 58.2 km2 and represents 4.18% of the analyzed area and 24.1% of the Bucharest's area (delimited with black contour in Figure 5) and it is located in the city center, the zone with the highest construction density (shown in purple in Figure 1 and called "continuous urban fabric"). Figure 5. Very hot spot area persistent both day and night during summer

8 4. CONCLUSION Determining the hot spots inside the UHI is one of the steps needed to establish and implement measures to mitigate the Urban Heat Island phenomenon. This paper presents the identification of hotspots using ArcGIS s Getis-Ord Gi* statistic. To determine at what degree the surface's hot spots characterizes the air temperature's hot spots, the temperature values from in situ measurements were analyzed and correlated with those determined from MODIS satellite images, during daytime and nighttime. Linear relationships were obtained characterized by correlation coefficients over 0.7 during daytime and over 0.8 during nighttime. The analysis of the available data from the summer of 2017, showed that within the city the average air temperature varies by less than 0.24 C, while the land surface temperature may have large variations between pixels (up to 20 C), depending on the land cover type. Analyzing the Land Surface Temperature mean values, determined from MODIS, for the summer of 2017, higher temperatures have been obtained inside the city compared to the surrounding areas, both during daytime and nighttime, but with a higher gradient at night, thus highlighting the Bucharest's Urban Heat Island. The area classified as very hot spot both daytime and nighttime is represented by the Bucharest s city center and represents 20.1% from the city's surface. This is the critical area where measures to mitigate the Urban Heat Island phenomenon are needed. REFERENCES ANPM, (2018). Calitate Aer Rețeaua Națională de Monitorizare a Calității Aerului (RNMCA) [WWW Document]. URL locale=ro Arnfield, a. J., (2003). Two decades of urban climate research: a review of turbulence, exchanges of energy and water, and the urban heat island. Int. J. Climatol. 23, Chapman, S., Watson, J.E.M., Salazar, A., Thatcher, M., McAlpine, C.A., (2017). The impact of urbanization and climate change on urban temperatures: a systematic review. Landsc. Ecol. 32, Cheval, S., Dumitrescu, A., (2015). The summer surface urban heat island of Bucharest (Romania) retrieved from MODIS images. Theor. Appl. Climatol. 121, Cheval, S., Dumitrescu, A., (2009). The July urban heat island of Bucharest as derived from modis images. Theor. Appl. Climatol. 96, Cheval, S., Dumitrescu, A., Bell, A., (2009). The urban heat island of Bucharest during the extreme high temperatures of July Theor. Appl. Climatol. 97, Debbage, N., Shepherd, J.M., (2015). The urban heat island effect and city contiguity. Comput. Environ. Urban Syst. 54, Deilami, K., Kamruzzaman, M., Liu, Y., (2018). Urban heat island effect: A systematic review of spatiotemporal factors, data, methods, and mitigation measures. Int. J. Appl. Earth Obs. Geoinf. 67, ESRI, (2018). How Hot Spot Analysis: Getis-Ord Gi* (Spatial Statistics) works [WWW Document]. URL (accessed ). Fu, P., Weng, Q., (2016). A time series analysis of urbanization induced land use and land cover change and its impact on land surface temperature with Landsat imagery. Remote Sens. Environ. 175, Gaur, A., Eichenbaum, M.K., Simonovic, S.P., (2018). Analysis and modelling of surface Urban Heat Island in 20 Canadian cities under climate and land-cover change. J. Environ. Manage. 206, Iojă, C.I., Grădinaru, S.R., Onose, D.A., Vânău, G.O., Tudor, A.C., (2014). The potential of school green areas to improve urban green connectivity and multifunctionality. Urban For. Urban Green. 13, Iriza, A., Dumitrache, R.C., Ştefan, S., (2017). Numerical modelling of the Bucharest urban heat island with 21

9 the WRF-urban system. Rom. J. Phys. 62, Article no Mohajerani, A., Bakaric, J., Jeffrey-Bailey, T., (2017). The urban heat island effect, its causes, and mitigation, with reference to the thermal properties of asphalt concrete. J. Environ. Manage. 197, NASA, (2018). EARTHDATA [WWW Document]. URL (accessed ). Nastran, M., Kobal, M., Eler, K., (2018). Urban heat islands in relation to green land use in European cities. Urban For. Urban Green. In Press. Oke, T.R., (1982). The energetic basis of the urban heat island. Q. J. R. Meteorol. Soc Oke, T.R., (1976). The distinction between canopy and boundary layer urban heat islands. Atmosphere (Basel). 14, Singh, P., Kikon, N., Verma, P., (2017). Impact of land use change and urbanization on urban heat island in Lucknow city, Central India. A remote sensing based estimate. Sustain. Cities Soc. 32, Stewart, I.D., (2011). A systematic review and scientific critique of methodology in modern urban heat island literature. Int. J. Climatol. 31, Tumanov, S., Stan-Sion, A., Lupu, A., Soci, C., Oprea, C., (1999). Influences of the city of Bucharest on weather and climate parameters. Atmos. Environ. 33, Zhang, X., Estoque, R.C., Murayama, Y., (2017). An urban heat island study in Nanchang City, China based on land surface temperature and social-ecological variables. Sustain. Cities Soc. 32, Zoran, M., (2011). Satellite observation of Urban Heat Island effect, in: Proceedings of the Global Conference on Global Warming. Lisabon, pp