Acta Sil va e et Ligni 125 (2021), 13–24 13 Izvirni znanstveni članek / Original scientific article THE INFLUENCE OF LAND USE ON THE SPATIAL DISTRIBUTION AND INTENSITY OF HEAT ISLANDS IN SLOVENIA VPLIV RABE T AL NA PROSTORSKO RAZPOREDITEV IN INTENZIVNOST TOPLOTNIH OTOKOV V SLOVENIJI Anica SIMČIČ 1 , Petra PEČAN 2 , Mojca NASTRAN 3 , Milan KOBAL 4 (1) Slovenian Forestry Institute, anica.simcic@gozdis.si (2) University of Ljubljana, Biotechnical Faculty, Department of Forestry and Renewable Forest Resources, petra.pecan@bf.uni-lj.si (3) University of Ljubljana, Biotechnical Faculty, Department of Forestry and Renewable Forest Resources, mojca.nastran@bf.uni-lj.si (4) University of Ljubljana, Biotechnical Faculty, Department of Forestry and Renewable Forest Resources, milan.kobal@bf.uni-lj.si ABSTRACT Heat islands (HI) are a common anthropogenic phenomenon and are defined as artificial surfaces (urban areas) that have a higher average temperature than their surroundings (rural areas). The aim of this work was to determine the influence of land use on the spatial distribution and intensity (HI i ) of HI in Slovenia. The MODIS Land Surface Temperature (LST) and Corine Land Cover (CLC) databases were used to perform the analysis. Within the identified HI, two HI levels were determined based on temperature difference. The results revealed a statistically significant negative correlation between HI i and both forest co- ver and forest fragmentation (forest edge density and ratio of mean forest patch size to HI size). Artificial surface was positively correlated with HI i . The results contribute to the understanding of the spatial distribution of HI and HI i and provide informati- on for spatial planning and policy-making to reduce the negative impact of HI. Ključne besede: heat island (HI), forest cover, artificial surface, forest fragmentation IZVLEČEK Toplotni otoki (HI) so pogost antropogeni pojav, ki ponazarjajo temperaturno razliko med umetnimi površinami (urbana območja) in okolico (ruralna območja). Namen prispevka je ugotoviti vpliv rabe tal na prostorsko razporeditev in intenziteto HI (HI i ) v Sloveniji. Uporabili smo karto temperature površja MODIS LST in podatke rabe tal Corine Land Cover (CLC). Na pod- lagi temperaturnih razlik smo določili dve ravni HI. Rezultati so pokazali statistično značilno negativno korelacijo med HI i in gozdnatostjo in/ali fragmentiranostjo gozdov (gostota gozdnega roba in razmerje povprečna velikost gozdne zaplate - velikost HI). Umetne površine vplivajo na povečanje HI i . Rezultati prispevajo k razumevanju prostorske razporeditve HI in HI i ter dajejo informacije za prostorsko načrtovanje in oblikovanje politik za zmanjšanje negativnega vpliva HI. Key words: toplotni otok (HI), gozdnatost, umetne površine, fragmentiranost gozda GDK 91+907.4+922.2(497.4)(045)=111 Prispelo / Received: 16. 2. 2021 DOI 10.20315/ASetL.125.2 Sprejeto / Accepted: 17. 5. 2021 1 INTRODUCTION 1 UVOD It is expected that 66 % of the world’s population will live in cities by 2050 (Gunawardena et al., 2017). With the rapid expansion of cities, urban sprawl is encroaching into green spaces, leading to their fra- gmentation into heterogeneous forms with irregular boundaries or even total loss (Kong, 2014a). Urbani- sation increases the frequency and severity of extre- me weather events, such as heat islands (hereafter HI) (Komac et al., 2017; Zalar et al., 2017). HI are defined as urban areas with significantly higher temperatures compared to their rural surroun- dings (Schwarz et al., 2011). The main cause of urban HI is the absorption and radiation of solar energy by urban structures and other anthropogenic heat sour- ces (Rizwan et al., 2008). Due to heat stress, HI affect urban ecosystems, energy demand, air quality and consequently public health and well-being (Luo et al., 2007; Rizwan et al., 2008). In addition to urbanisation, activities such as ag- riculture, grazing, forestry and irrigation can directly alter heat flow at the land surface and indirectly alter atmospheric circulation. Consequently, local tempera- ture increases occur (Jusuf et al., 2007), leading to the formation of HI. The potential to reduce HI is associ- ated with urban geometry and urban greening. Street trees and urban park design are some of the most im- portant elements that determine the incidence of sun- light and wind speed in urban greening design (Jamei 14 Simč ič A., P ečan P ., Nastr an M., K obal M.: The infl uenc e o f land use on the spa t ial d istr ibut ion and int ensi t y o f hea t ... et al., 2016). Effective techniques to reduce HI include changes in land use and land cover, as well as urban land use patterns (i.e. urban geometry), the use of low albedo objects and building materials, and the use of different types of (building) greening. The relationship between land use and climate has increasingly drawn the attention of urban planners and policy makers to the identification, formulation and implementation of urban development strategies to improve the well- being of the population (Meerow and Newell, 2017). Most studies have identified urban green spaces as one of the most efficient and simple nature-based solutions to reduce urban HI (e.g. Norton et al., 2015), especially due to their cooling and shading effects and absorption of short-wave radiation (Kong et al., 2014b). Studies examining the effects of forests and other green spaces on average temperature variabil- ity have shown that temperatures are lower in forests than in other areas (Dugord et al., 2014; Kong et al., 2014b; Nastran et al., 2019). Due to their thermal ben- efits, urban forests play an important role in reducing HI intensity (hereafter HI i ) and improving the quality of life in cities. On the other hand, the amount of artifi- cial land is a crucial factor affecting the size of urban HI and HI i (Lin et al., 2016). Moreover, some studies (e.g. Dugord at al., 2014; Kong et al., 2014b; Nastran et al., 2019) show that urban HI depend not only on the pro- portion of a particular land use, but also on its spatial configuration. Urban planners should prevent deforestation of green areas for construction purposes and ensure the even distribution of green areas in cities. Trees and vegetation reduce surface and air temperature through shading and evapotranspiration. Vegetation in urban environments can reduce energy consumption, improve air quality, reduce greenhouse gasses, manage stormwater, improve water quality and reduce noise. It also enhances the aesthetic value of an area and is a popular recreational area during heat waves, when temperatures outside forests threaten human health. The dynamic cooling effects of green and forest areas are therefore becoming increasingly important (Olivei- ra et al., 2011; Aram et al., 2019). The main objective of this study was to analyse how land use influences the spatial distribution and inten- sity of HI in Slovenia. The correlation between the selected land use cover and HI i is shown with forest fragmentation indices (forest edge density and ratio of mean forest patch size to HI size). The three hypoth- eses underlying this study are as follows: • The effect of forest cover on HI i is statistically sig- nificant and negative. • The effect of artificial surface on HI i is statistically significant and positive. • The effect of forest fragmentation on HI i is statisti- cally significant and negative. 2 DAT A AND METHODS 2 PODATKI IN METODE 2.1 Study area 2.1 Območje raziskave The study area covered the entire territory of Slove- nia, a Central European country with traditionally scat- tered and relatively small urban areas. Slovenia has a continental climate with hot summers and cold win- ters in the plateaus and valleys to the east and north, and a Mediterranean climate near the coast. The Slo- venian littoral refers to the short coastline and south- east strip of predominantly karstic topography. The towns are predominately located on flat terrain, while smaller villages are located on steep terrain. According to Corine Land Cover (hereafter CLC) data, there were 34 land cover categories in Slovenia in 2012, out of the 44 defined by the CLC methodol- ogy. The proportion of different types of land cover in Slovenia is as follows: forests and other semi-natural land (58 %), agricultural land (35 %), artificial sur- faces (3 %), wetlands (2 %) and water bodies (2 %) (EEA, 2016). 2.2 Source of data 2.2 Viri podatkov 2.2.1 Land use map (Corine Land Cover - CLC) 2.2.1 Karta rabe tal (Corine Land Cover - CLC) We used the 2012 CLC map for the whole Sloveni- an territory (EEA, 2016). The original methodology for the Slovenian CLC used in this analysis is based on vi- sual photo interpretation of Landsat satellite imagery at a scale of 1:100,000. The classification is hierarchi- cal at three levels. The first level comprises five CLC classes: 1) artificial surfaces, 2) agricultural land, 3) forest and semi-natural land, 4) wetlands and 5) water bodies. The second level comprises 15 land use catego- ries and the third includes 44 land use categories. In this analysis, we used the following categories of land uses: • artificial surfaces (CLC code: 112 - discontinuous, urban fabric, 121 - industrial or commercial units, 122 - road and rail networks and associated land, 124 - airports, 131 - mineral extraction sites and 133 - construction sites) • green urban areas (CLC code: 141 - green urban areas and 142 - sport and leisure facilities) • agricultural areas (CLC code: 211 - non-irrigated ara- Acta Sil va e et Ligni 125 (2021), 13–24 15 ble land, 221 - vineyards, 222 - fruit trees and berry plantations, 231 - pastures, 242 - complex cultivation patterns and 243 - land principally occupied by agri- culture, with significant areas of natural vegetation) • forests (CLC code: 311 - broad-leaved forest, 312 - coniferous forest, 313 - mixed forest, 321 - natural grasslands and 324 - transitional woodland-shrub) • wetlands (CLC code: 411 - inland marshes, 511 - water courses and 512 - water bodies) Other land uses were not present in the HI in Slo- venia. 2.2.2 Map of the land surface temperature of Europe (LST) 2.2.2 Karta temperature površja Evrope (LST) The MODIS Land Surface Temperature (LST) map of Europe with a spatial resolution of 250 × 250 m was used for the determination of HI and measurement of HI i . The dataset per each pixel included four temporal records per day. The LST map is based on the analy- sis of records reconstructed from MODIS satellite data from 2001 to 2011 (Metz et al., 2014) and includes night-time and daily temperatures. Ten-year average temperature data eliminates abnormalities in the oc- currence of HI. The LST dataset is available online for download and direct usage in GIS programs on the GFOSS Blog webpage (Metz et al., 2014). Since the correlation of HI i and HI with urban green infrastructure is highest in summer (Chang et al., 2007; Hamada and Ohta, 2010; Imhoff et al., 2010; Li et al., 2011), we used the average temperature of the warmest quarter of the year, i.e. June, July and August. HI smaller than 50 ha were eliminated from the study, so every HI polygon could be represented with at least 8 image cells (temperature values). Fig. 1: Example of elimination of higher level HI (adapted by: Obu and Podobnikar, 2013; Kralj, 2017) Slika 1: Primer omejevanja kotanj višjega reda (prirejeno po: Obu and Podobnikar, 2013; Kralj, 2017) 16 Simč ič A., P ečan P ., Nastr an M., K obal M.: The infl uenc e o f land use on the spa t ial d istr ibut ion and int ensi t y o f hea t ... 2.3 Determination of HI 2.3 Določanje HI The determination of HI was based on a GIS algo- rithm used for watershed delineation in hydrological analyses, where delineation was based on the simula- tion of waterflow over the surface raster map. We re- placed the surface raster map with the temperature raster map and multiplied by the constant value of -1 to create a map of inverse temperature where the local minimum value (sink) represented the highest value of temperature and vice versa (Fig. 1). The further proce- dure was the same as described by Doctor and Young (2013), Obu and Podobnikar (2013) and Kobal et al. (2015) and is based on water flow simulations on a surface raster map (inverse temperature) and incorpo- rated three steps: (i) watershed delineation (increas- ing temperature along the gradient from rural to urban area), (ii) confining sinkholes (HI) and (iii) confining higher rank sinkholes (higher level of HI). The spatial distribution of HI was defined in ArcMap 10.5.1. (ESRI, 2017) using the tools from the “Hydrol- ogy” toolbox implemented in Model Builder. The flow direction for each raster cell was identified, and each raster cell received a value representing the total num- ber of raster cells that drained into it. In the second step, effluent levels confined the sinkholes (HI), and the raster cells below the effluent level were designated as part of the sinkhole (HI). The input data in this step included a layer containing delineated watersheds and a layer containing inverse temperature information. For each delineated watershed, the raster cell with the lowest inverse temperature value among the water- shed boundary raster cells was defined as the effluent level. The output consisted of a list of watersheds with elevation information for the effluent raster cells, and the raster cells below the effluent level were assigned as forming an HI. In the third step, sinkholes (HI) were ranked according to their locations and sizes with re- spect to surrounding sinkholes (HI). The 1 st level of sinkholes (HI) are the smallest and are located within larger sinkholes (HI) of a higher rank. Figure 1 contains a graphical presentation of sinkholes (HI) of different ranks. When a smaller sinkhole (HI) was located within a larger sinkhole (HI), the effluent level was first deter- mined for the smaller sinkhole (HI). HI intensity (HI i ) was defined here as the range of temperature inside the above described methodology for the delineation of HI. Based on the LST map, the size and spatial distri- bution of HI were determined by extracting polygons with an area greater than 50 ha, where the tempera- ture difference was more than 0.1 °C compared to the neighbouring area. 2.4 Variables considered and statistical analyses 2.4 Obravnavane spremenljivke in statistična analiza To determine the two HI levels, we evaluated the effects of various variables on HI i . For this study we Fig. 2: Spatial distribution of HI and HI i in Slovenia - 1 st level Slika 2: Prostorska porazdelitev HI in HI i v Sloveniji - raven 1 Acta Sil va e et Ligni 125 (2021), 13–24 17 chose forest area because it is the most common land cover in Slovenia, and artificial surface area because it has been proven that HI are most intense in built up ar- eas. We considered two variables as a forest fragmen- tation indicator: i) forest edge density and ii) ratio of mean forest patch size to HI size. Simple linear regression with log transformation of variables was used to relate HI i to corresponding at- tributes at the single HI level. All statistical analyses were done in Python3 using the nump y and st atsmod - e ls packages. 3 RESUL TS 3 REZUL T ATI 3.1 Spatial distribution and intensity of HI in Slovenia 3.1 Prostorska razporeditev in intenziteta HI v Sloveniji We identified 392 and 60 HI at the 1 st and 2 nd level of HI, respectively (Fig. 2 and Fig. 3). HI size ranges from 0.5 to 551.8 km 2 . The mean size of HI is 5.40 km 2 (1 st level) and 60.00 km 2 (2 nd level). HI of the 1 st level occur in large cities and their surroundings (e.g. Lju- bljana, Kranj, Celje, Velenje, Murska Sobota, Lendava, Koper and Nova Gorica), while HI of the 2 nd level occur in larger cities and form “HI agglomerations” (e.g. Lju- bljana - Kranj, Celje - Velenje, Maribor - Ptuj, Murska Sobota - Lendava, Krško - Zagreb, Novo mesto, Carin- thia region - Austria). HI i ranges from 0.1 to 1.8 °C, with a mean value of 0.19 °C (1 st level) and 0.58 °C (2 nd level). Forest cover within HI ranges from 0 to 100.0 % (Table 1). Mean forest cover within HI with the lowest HI i is 53.2 % and 9.8 % with HI i = 1.3 °C (1 st level). Mean forest cover within HI with the lowest HI i is 60.7 % and 18.8 % with HI i = 1.8 °C (2 nd level). The artificial surface within HI ranges from 0 to 73.0 % (Table 1). The lowest value of HI i is observed in the area without artificial surfaces (both levels of HI). The highest value of HI i is observed in HI with 47 % of artificial surface (1 st level) and 30 % of artificial surface (2 nd level). At the 2 nd level of HI, with the high- est share of artificial surfaces (41.3 %), HI i increases by 1.1 °C compared to the surroundings. Forest edge density ranges from 0.0 to 59.7 m/ha and is on average lower at the 2 nd level of HI (17.00 m/ ha). The HI with the highest value of HI i has 5.31 m/ha of forest edge (1 st level). The forest edge density of the lowest value of HI i at the 2 nd level of HI level range from 15.9 m/ha to 36.97 m/ha, while the HI with the high- est value of HI i vary from 9.29 m/ha to 15.05 m/ha of forest edge (Table 2). The ratio between the mean forest patch size and size of HI ranges from 0.01 (smaller forest patches, for- ests are more fragmented) to 0.99 (larger forest patch- es, forests are less fragmented) (Table 1). HI at the 1 st level have a higher value of HI i with a higher ratio of the mean forest patch size to HI size, while HI at the 2 nd level have a lower value of HI i up to 1. The largest forest patch size is 7.12 km 2 with an intensity of 0.3 °C. 3.2 The influence of land use on the intensity of HI in Slovenia 3.2 Vpliv rabe tal na intenziteto HI v Sloveniji 3.2.1 The influence of forest cover on HI i 3.2.1 Vpliv gozdnatosti na HI i The results of the statistical analysis confirm a sta- tistically significant negative effect (p < 0.05) of forest cover on HI i (Fig. 4). The effect of forest at the first level Table 1: Characteristics of HI for both levels Preglednica 1: Značilnosti HI na obeh ravneh HI HI 1 st level Raven 1 2 nd level Raven 2 Number / Število 392 60 Size range / Razpon velikosti 0.5–151.1 km 2 0.6–551.8 km 2 Mean size / Povprečna velikost 5.40 ± 1.05 km 2 60.00 ± 24.76 km 2 Total area / Skupna površina 2115.56 km 2 3600.06 km 2 HI i HI i Range / Razpon 0.1–1.3 °C 0.1–1.8 °C Mean / Povprečje 0.19 ± 0.02 °C 0.58 ± 0.11 °C Forest cover Gozdnatost Range / Razpon 0.0–100.0 % 6.1–100.0 % Mean / Povprečje 46.38 ± 3.15 % 38.23 ± 6.66 % Artificial surface Delež umetnih površin Range / Razpon 0.0–73.0 % 0.0–41.3 % Mean / Povprečje 5.76 ± 1.19 % 7.81 ± 2.50 % Forest edge density Gostota gozdnega roba Range / Razpon 0.0–59.7 m/ha 0.0–38.3 m/ha Mean / Povprečje 18.2 ± 1.29 m/ha 17.00 ± 2.15 m/ha Mean forest patch size / HI size Povprečna velikost zaplate gozda / velikost HI Range / Razpon 0.01–0.99 0.01–0.99 Mean / Povprečje 0.36 ± 0.03 0.19 ± 0.07 18 Simč ič A., P ečan P ., Nastr an M., K obal M.: The infl uenc e o f land use on the spa t ial d istr ibut ion and int ensi t y o f hea t ... is smaller (for a 10 percentage point increase in forest cover, HI i decreases by 4.7 %) than that at the second level (for a 10 percentage point increase in forest co- ver, HI i decreases by 18.4 %). 3.2.2 The influence of artificial surface on HI i 3.2.2 Vpliv deleža umetnih površin na HI i The results of the statistical analysis confirm a statically significant positive (p < 0.05) effect of arti- ficial surfaces on HI i at both levels (Fig. 5). The effect of artificial surfaces at the 1 st level is lower (for a 10 percentage point increase in artificial surface cover, HI i increases by 21.6 %) than for the 2 nd level (for a 10 per- centage point increase in artificial surfaces cover, HI i increases by 48.5 %). Fig. 3: Spatial distribution of HI and HI i in Slovenia - 2 nd level Slika 3: Prostorska porazdelitev HI in HI i v Sloveniji - raven 2 Fig. 4: The influence of forest cover on HI i for both levels of HI in Slovenia Slika 4: Vpliv gozdnatosti na HI i za obe ravni HI v Sloveniji Acta Sil va e et Ligni 125 (2021), 13–24 19 3.3 The influence of forest fragmentation on the intensity of HI in Slovenia 3.3 Vpliv fragmentiranosti gozda na intenziteto HI v Sloveniji 3.3.1 The effect of forest edge density on HI i 3.3.1 Vpliv gostote gozdnega roba na HI i The results also confirm a significant negative (p < 0.05) relationship between HI i and forest edge density for the 1 st level of HI and the 2 nd level of HI (Fig. 6). The influence of forests on the HI of the 1 st level is lower (for a 10 m/ha increase in forest edge density, HI i decreases by 11.0 %) than that at the 2 nd level (for a 10 m/ha inc- rease in forest edge density, HI i decreases by 39.0 %). 3.3.2 The influence of the ratio of mean forest patch size to HI size on HI i 3.3.2 Vpliv razmerja povprečne velikosti gozdne zaplate - velikost HI na HI i The correlation between the ratio of mean forest patch size to HI size and HI i is statistically significantly negative (p < 0.05) (Fig. 7). The influence of the ratio of mean forest patch size to HI size on HI i at the 1 st level is significantly lower (for a 10 % increase in the ratio of mean forest patch size to HI size, HI i decreases by 1.1 %) than that at the 2 nd level (for a 10 % increase in the ratio of mean forest patch size to HI size, HI i decreases by 3.1 %). HI Equation p R 2 1 st level 1. raven log(HI i ) = -1.629 - 0.005 × COV f < 0.001 0.07 2 nd level 2. raven log(HI i ) = -0.081 - 0.017 × COV f < 0.001 0.39 Table 2: Linear model equations, level of statistical sig- nificance (p), and coefficient of determination (R 2 ) for the dependence of HI intensity (HI i ) on forest cover (COV f ) for both levels of HI Preglednica 2: Enačbe linearnega modela, statistična značilnost (p) in koeficient determinacije (R 2 ) za odvisnost intenzivnosti HI (HI i ) od deleža gozdov (COV f ) znotraj HI za obe ravni HI Equation p R 2 1 st level 1. raven log(HI i ) = -1.973 + 0.021 × COV a < 0.001 0.20 2 nd level 2. raven log(HI i ) = -1.161 + 0.047 × COV a < 0.001 0.36 Table 3: Linear model equations, level of statistical sig- nificance (p), and coefficient of determination (R 2 ) for the dependence of HI intensity (HI i ) on artificial surfaces (COV a ) for both levels of HI Preglednica 3: Enačbe linearnega modela, statistična značilnost (p) in koeficient determinacije (R 2 ) za odvisnost intenzivnosti HI (HI i ) od deleža umetnih površin (COV a ) znotraj HI obeh ravni Fig. 5: The influence of artificial surface on HIi for both lev- els of HI in Slovenia Slika 5: Vpliv deleža umetnih površin na HIi za obe ravni HI 20 Simč ič A., P ečan P ., Nastr an M., K obal M.: The infl uenc e o f land use on the spa t ial d istr ibut ion and int ensi t y o f hea t ... 4 DISCUSSION AND CONCLUSION 4 RAZPRAVA IN ZAKLJUČEK 4.1 The impact of artificial surfaces on the intensity of HI 4.1 Vpliv deleža umetnih površin na intenziteto HI Numerous studies have shown that artificial surfa- ces have a warming effect in urbanized areas, while ve- getation has a cooling effect (Oliveira et al., 2011; Aram et al., 2019). Our study shows that the share of artificial areas has a statistically significant influence on HI i (Fig. 5). Our results showed that the highest HI temperature increase (1.3 °C) is in areas with 47 % artificial are- as, while the lowest temperature increase (0.1 °C) is in areas without artificial areas for the 1 st level of HI. The- se findings are consistent with those of other studies. For example, Kong et al. (2014b) showed that the ave- rage surface temperature in Nanjing, China is typically higher on artificial surfaces and lower in green spaces. Yang et al. (2013) analysed the occurrence of urban HI in Beijing (China), where air temperatures increased linearly with increasing area of artificial surfaces. 4.2 The impact of forest cover on the intensity of HI 4.2 Vpliv gozdnatosti na intenziteto HI Our study revealed that forest cover has a statisti- cally significant negative effect on HI i . The highest va- lue of HI i (1.8 °C) was observed for HI from the second level, in areas with 20.7 % artificial surfaces, and the smallest increase (0.1 °C) was observed in areas wi- thout artificial surfaces. The highest temperature inc- rease (1.3 °C) for HI from the 1 st level was observed in areas with 9.8 % forest cover, while the lowest increase (0.1 °C) was observed at the 2 nd level in areas with 53.2 % forest cover (Fig. 4). Similarly, in Hanoi (Vietnam) researchers found that the warmest districts are loca- ted along the main roads and industrial zones, while colder areas are closer to green spaces and water bo- dies (Tran et al., 2017). Dugord et al. (2014) claim that the function of forests and other green spaces is redu- ced when HI i is moderated. The cooling effect of forests was studied in Nanjing, China, where it was found that the daily mean temperature in the forest was 1.9 °C lo- wer compared to the temperature of the surrounding Fig. 6: The length of the forest edge calculated for the sur- face of forests within HI Slika 6: Dolžina gozdnega roba glede na površino HI Table 4: Linear model equations, level of statistical signifi- cance (p), and coefficient of determination (R 2 ) for the de- pendence of HI intensity (HI i ) on forest edge density (FED) for both levels of HI Preglednica 4: Enačbe linearnega modela, statistična značilnost (p) in koeficient determinacije (R 2 ) za odvisnost intenzivnosti HI (HI i ) od gostote gozdnega roba (FED) znotraj HI za obe ravni HI Equation p R 2 1 st level 1. raven log(HI i ) = -1.649 - 0.011 × FED < 0.001 0.06 2 nd level 2. raven log(HI i ) = -0.116 - 0.040 × FED < 0.001 0.19 Acta Sil va e et Ligni 125 (2021), 13–24 21 concrete surface, while the maximum daily tempera- ture was up to 3.2 °C lower than that of the surroun- ding concrete surface (Kong et al., 2016). Furthermore, Kong et al. (2014b) in Nanjing, China found that a 10 % increase in forest cover reduced air temperature by 0.83 °C. These studies concluded that a sufficient amo- unt of green space with small, fragmented forest areas would effectively ensure the cooling of the area. 4.3 The impact of forest edge density on the intensity of HI 4.3 Vpliv gostote gozdnega roba na intenziteto HI Various fragmentation variables observed in other studies prevent a direct comparison of the results; ne- vertheless, some common patterns can be identified. We found that HI with higher intensity have less forest edge, and HI with lower intensity have more forest edge (Fig. 6). Li et al. (2016) pointed out that the spa- tial distribution and fragmentation of green space has a significant influence on surface temperature, someti- mes even greater than the land use itself. In our study, the fragmentation of the forest edge had a statistically significant effect. Our findings are in agreement with the results of other research (Chang et al., 2007; Jusuf et al., 2007) showing that the fragmentation, type, size, and shape of green spaces have a positive effect on temperature reduction. The proximity, location and size of green spaces were statistically significant factors for the re- duction in air temperature (Chang et al., 2007; Jusuf et al., 2007), although how the fragmentation of gre- en spaces contributes to the reduction of surface tem- perature was not considered. By understanding how vegetation cools the environment in an urban area, it may be possible to maximize the cooling potential of green spaces and to incorporate these features in ur- ban planning for the express purpose of reducing the HI effect. 4.4 The impact of the ratio of the mean forest patch size to HI size on the intensity of HI 4.4 Vpliv povprečne velikosti gozdne zaplate na intenziteto HI The ratio of the mean forest patch size to HI size in our study ranges from 0 (smaller forest areas, more HI Equation p R 2 1 st level 1. raven log(HI i ) = -2.071 - 0.117 × MFPS < 0.001 0.12 2 nd level 2. raven log(HI i ) = -1.701 - 0.306 × MFPS < 0.001 0.50 Table 5: Linear model equations, level of statistical signifi- cance (p), and coefficient of determination (R 2 ) for the de- pendence of HI intensity (HI i ) on the mean forest patch size (MFPS) for both levels of HI Preglednica 5: Enačbe linearnega modela, statistična značilnost (p) in koeficient determinacije (R 2 ) za odvisnost intenzivnosti HI (HI i ) od povprečne velikosti zaplate gozda (MFPS) znotraj HI za obe ravni Fig. 7: The correlation between the mean forest patch size and the size of HI Slika 7: Razmerje med povprečno velikostjo gozdne površine in površino HI 22 Simč ič A., P ečan P ., Nastr an M., K obal M.: The infl uenc e o f land use on the spa t ial d istr ibut ion and int ensi t y o f hea t ... fragmented) to 1 (larger forest areas, less fragmented). We confirmed that temperatures are lower in HI with larger, less fragmented forest patches and are higher in those with smaller, more fragmented forest patches (Fig. 7). We cannot compare our results with many other works because Nastran et al. (2019) found that the effects of forest configuration on urban HI i in Euro- pe are not the same for all climatic regions. In principle, larger and more fragmented forest areas are associated with lower urban HI i , except in Central and Mediter- ranean Europe. Dugord et al. (2014) reported similar results in Berlin, Germany. They found that (although scattered vegetation is important to regulate urban temperatures) large and/or spatially aggregated green spaces are more important for temperature reduction. On the other hand, dense and interconnected traffic areas tend to increase temperatures. However, the po- tential cooling effect of green spaces depends strongly on the degree of tree cover (Dugord et al., 2014). Slovenia is near Central and Mediterranean Europe, so our results coincide with the aforementioned works. Other studies have shown opposite results. Park et al. (2017) found a positive effect of small green spaces on urban HI reduction in Seoul, Korea, where the tempe- rature dropped by 1.66 °C due to the presence of small green spaces. Kim et al. (2016) showed the influence of the shape of green spaces on surface temperatures in Austin, Texas. The number of surfaces was presented as one influencing factor. Many surfaces have a positi- ve influence on LST reduction. Li et al. (2016) pointed out that the shape of the surface has a greater influence on temperature than the land use itself; it is better to distribute them evenly over the face rather than to con- centrate them in one place. While most research only provides a qualitative description of cooling effects, quantifiable effects and statistically significant correlations are missing. Kong et al. (2014a), Durgod et al. (2014) and Li et al. (2011, 2016) provide statistical analyses of the size, shape and fragmentation of green spaces. Kong et al. (2014a) emphasise that fragmented green spaces are also ef- fective for cooling purposes when the forest cover is solid. Furthermore, a correlation analysis between mean patch size, patch density and an aggregation in- dex of forest with temperature showed that for a fixed share of forest vegetation, fragmented green space also has an effective cooling effect (Kong et al., 2014a). The results provide urban planners, architects and natural resource managers with important theoretical and practical information on urban green space plan- ning and management that they can use to adapt to and mitigate the effects of HI that will evolve with climate change. These lessons should be taken into account in the formulation of planning strategies at higher levels and incorporated into lower-level planning guidelines. More variables need to be introduced and calculated to better understand which variable corelates most strongly with HI intensity, which would be helpful in- formation for urban decision makers. 5 SUMMARY 5 POVZETEK Pričujoča študija prikazuje kvantitativno raziskavo vpliva rabe tal na prostorsko razporeditev in intenziv- nost pojavljanja toplotnih otokov v Sloveniji. Toplotni otoki (HI) so rezultat antropogenega delovanja v ur- banih območjih in ponazarjajo temperaturno razliko med urbano in ruralno krajino. Toplotni otoki so bili v raziskavi določeni za tista območja, znotraj katerih je temperatura površja najtoplejše četrtine leta (ju- nij, julij, avgust) presegala več kot 0,1 °C v primerjavi s temperaturo okolice. V Sloveniji se pojavljajo tako v urbanih kot ruralnih predelih države. Raziskava preučuje vpliv gozdnih in umetnih povr- šin na razporeditev HI, vpliv gostote gozdnega roba ter razmerje povprečna površina gozdne zaplate - velikost HI na pojavljanje HI. Podatki o rabi tal so bili povze- ti po evropski karti rabe tal Corine Land Cover (EEA, 2016), podatki o povprečni temperaturi najtoplejše če- trtine leta pa so bili pridobljeni na spletni strani GFOSS Blog (Metz et al., 2014). Območje raziskave je zajemalo celotno območje Slovenije ter tista območja sosednjih držav, do koder segajo vplivi HI. Analiza je podrobneje raziskala vpliv različnih rab tal (umetne površine, zele- ne mestne površine, kmetijske površine, gozdne povr- šine in močvirnate površine). Vpliv rabe tal na prostorsko razporeditev HI in in- tenziteto HI (HI i ) je bil obravnavan na dveh ravneh. Znotraj posameznih HI smo izračunali delež posame- zne rabe tal, gostoto gozdnega roba ter povprečno velikost gozdne zaplate glede na velikost toplotnega otoka. Pri tem so bile posamezne kategorije rabe tal združene v 5 kategorij: umetne površine, kmetijske po- vršine, gozdne in deloma ohranjene naravne površine, močvirnate površine in vodne površine. Analiza je bila opravljena v programu ArcMap 10.5.1 (ESRI, 2017). Za prostorsko določitev in razmejitev HI so bila upora- bljena orodja iz paketa ArcGIS ''Hydrology''. V prvem koraku smo z orodjem ArcGIS ''Sink'' določili najnižje vrednosti rastrskih celic. V drugem koraku smo z orod- jem ''Flow Direction'' določili smer padcev tempera- ture (obrnjene vrednosti) in z uporabo ''Watershed'' določili rastrske celice z najnižjimi vrednostmi (obr- njene vrednosti). Nato smo določili najvišje vrednosti Acta Sil va e et Ligni 125 (2021), 13–24 23 za vsako rastrsko celico in jih združili v posamezne HI. Območje HI smo določili z dvigom povprečne tempera- ture za 0,1 °C v primerjavi z okolico. Zaradi prostorske ločljivosti temperaturne karte so bili iz raziskave izlo- čeni vsi otoki, manjši od 50 ha, tako da je bilo vsako območje HI predstavljeno z vsaj 8 slikovnimi celicami. Korelacija med deležem gozda in HI i je je statistično značilno negativna na obeh ravneh HI. Analiza vpliva gozda (slika 4) na pojavljanje HI (1. raven) kaže, da je najvišja HI i (1,3 °C) pri 9,81 % gozdnatosti. Najnižjo HI i (0,1 °C) pa so imeli tisti HI, ki so imeli v povprečju 53,15 % gozdnih površin. Statistično značilno negati- ven vpliv na HI i ima tudi gostota gozdnega roba (slika 6). Najvišjo HI i (1,3 °C) smo zabeležili pri 5,30 m/ha dolžine gozdnega roba. Najnižjo HI i (0,1 °C) pa je imel HI z 59,7 m/ha gozdnega roba. Delež umetnih površin ima statistično značilen pozitiven vpliv na HI i na obeh ravneh (slika 5). Navedeni izsledki lahko prostorskim planerjem pomagajo pri izbiri najustreznejših ukrepov za blaže- nje učinkov HI, ki so lahko škodljivi tako za okolje kot zdravje ljudi. 6 ACKNOWLEDGEMENTS 6 ZAHVALA This study is based on the results of the first author’s master’s thesis at the Department of Forestry and Renewable Forest Resources, Biotechnical Faculty, University of Ljubljana. The first author would like to express gratitude to the Pahernik Foundation for su- pporting her master’s study. 7 REFERENCES 7 VIRI Aram F., Higueras Garciá E., Solgi E., Mansournia S. 2019. Ur- ban green space cooling effect in cities. Heliyon, 5, 4, e01339. DOI:10.1016/j.heliyon.2019.e01339 Chang C.R., Li M.H., Chang S.D. 2007. A preliminary study on the local cool-island intensity of Taipei city parks. Landscape & Urban Planning, 80, 4: 386–395. DOI 10.1016/j.landur- bplan.2006.09.005 Doctor D.H., Young J.A. 2013. An evaluation of automated GIS tools for delineating karst sinkholes and closed sinks from 1-me- ter lidar-derived digital elevation data. V: 13th sinkhole con- ference, NCKRI symposium 2, Carlsbad, New Mexico, 6-10 May 2013. Land L., Doctor D.H., Stephenson B. (ur.). Carlsbad, National Cave and Karst Research Institute: 449–458. DOI 10.5038/9780979542275.1156 Dugord P .A., Lauf S., Schuster C., Kleinschmit B. 2014. Land use patterns, temperature distribution and potential heat stress risk - the case study Berlin, Germany. Computers, Environment and Urban Systems, 48: 86–98. DOI 10.1016/j.compenvurb- sys.2014.07.005 ESRI (Environmental Systems Research Institute). 2017. ArcGIS De- sktop Help 10.5.1. Geostatistical Analyst. EEA (European environment agency). 2016. Corine land cover (CLC) 2012, 18.5.1. https://land.copernicus.eu/pan-european/corine- land-cover/clc-2012 (9. 7. 2017). Gunawardena K.R., Wells M.J., Kershaw T . 2017. Utilising green and bluespace to mitigate urban heat island intensity. Science of the Total Environment, 584–585, 1040–1055. DOI 10.1016/j.scito- tenv.2017.01.158 Hamada S., Ohta T . 2010. Seasonal variations in the cooling effect of urban green areas on surrounding urban areas. Urban Forestry & Urban Greening, 9, 1: 15–24. DOI 10.1016/j.ufug.2009.10.002 Imhoff M.L., Zhang P ., Wolfe R.E., Bounoua L. 2010. Remote sensing of the urban heat island effect across biomes in the continen- tal USA. Remote Sensing of Environment, 114, 3: 504–513. DOI 10.1016/j.rse.2009.10.008 Jamei E., Rajagopalan P ., Seyedmahmoudian M., Jamei Y. 2016. Review on the impact of urban geometry and pedestrian level greening on outdoor thermal comfort. Renewable and Sustainable Energy Reviews, 54: 1002–1017. DOI 10.1016/j.rser.2015.10.104 Jusuf S.K., Wong N.H., Hagen E., Anggoro R., Hong Y. 2007. The in- fluence of land use on the urban heat island in Singapore. Habitat International, 31, 2: 232–242. DOI 10.1016/j.habi- tatint.2007.02.006 Kim J., Donghwan G., Sohn W., Kil S., Kim H., Lee D. 2016. Neighbour- hood landscape spatial patterns and land surface temperature: an empirical study on single-family residential areas in Austin, Texas. International Journal of Environmental Research and Pu- blic Health, 13, 9: 808. DOI 10.3390/ijerph13090880 Kobal M., Bertoncelj I., Pirotti F., Dakskobler I., Kutnar L. 2015. Using lidar data to analyse sinkhole characteristics relevant for under- story vegetation under forest cover-case study of a high karst area in the Dinaric Mountains. PLoS One, 10, 3, e0122070. DOI 10.1371/journal.pone.0122070 Komac B., Ciglič R., Pavšek M., Kokalj Ž. 2017. Naravne nesreče v me- stih: primer mestnega toplotnega otoka. V: Zorn M., Komac B., Ciglič R., Tičar J. (eds.). Trajnostni razvoj mest in naravne nesre- če. Knjižna zbirka Naravne nesreče, 4. Ljubljana, Založba ZRC: 51–67. Kong F., Sun C., Liu F., Yin H., Jiang F., Pu Y., Cavan G., Skelhorn C., Middel A., Dronova I. 2016. Energy saving potential of fragmen- ted green spaces due to their temperature regulating ecosy- stem services in the summer. Applied Energy, 183: 1428–1440. DOI 10.1016/j.apenergy.2016.09.070 Kong F., Yin H., Wang C., Cavan G., James P . 2014a. A satellite image- based analysis of factors contributing to the green-space cool island intensity on a city scale. Urban forestry & Urban gree- ning, 13, 4: 846–853. DOI 10.1016/j.ufug.2014.09.009 Kong F., Yin H., James P ., Hutyra L.R., He H.S. 2014b. Effects of spati- al pattern of greenspace on urban cooling in a large metropoli- tan area of Eastern China. Landscape and Urban Planning, 128: 35–47. Kralj M. 2017. Vročina je predvsem v mestih že prava naravna nesre- ča. Dnevnik. https://www.dnevnik.si/1042777497 (13.7.2020) Li J., Song C., Cao L., Zhu F., Meng X., Wu J. 2011. Impacts of landscape structure on surface urban heat islands: a case study of Shang- hai, China. Remote Sensing of Environment, 115, 12: 3249–3263. Li X., Li W., Middel A., Harlan S.L., Brazel A.J., Turner II B.L. 2016. Re- mote sensing of the surface urban heat island and land architec- ture in Phoenix, Arizona: combined effects of land composition and configuration and cadastral-demographic-economic factors. Remote sensing of environment, 174: 233–243. DOI 10.1016/j. rse.2015.12.022 Lin B.B., Meyers J., Beaty R.M., Barnett G.B. 2016. Urban green infra- structure impacts on climate regulation services in Sydney, Au- stralia. Sustainability, 8, 8, 788. DOI 10.3390/su8080788 24 Simč ič A., P ečan P ., Nastr an M., K obal M.: The infl uenc e o f land use on the spa t ial d istr ibut ion and int ensi t y o f hea t ... Luo Z., Sun O.J., Ge Q., Xu W., Zheng J. 2007. Phenological responses of plants to climate change in an urban environment. Ecological Research, 22, 3: 507–514. DOI 10.1007/s11284-006-0044-6 Meerow S., Newell J.P . 2017. Spatial planning for multifunctional green infrastructure: growing resilience in Detroit. Landsca- pe and Urban Planning, 159: 62–75. DOI 10.1016/j.landur- bplan.2016.10.005 Metz M., Rocchini D., Neteler M. 2014. Surface temperatures at the continental scale: tracking changes with remote sensing at unprecedented detail. Remote Sensing, 6, 5: 3822–3840. DOI 10.3390/rs6053822 Nastran M., Kobal M., Eler K. 2019. Urban heat island in relation to green land use in European cities. Urban Forestry & Urban Gre- ening, 37: 33–41. Norton B.A., Coutts A.M., Livesley S.J., Harris R.J., Hunter A.M., Willi- ams N.S. 2015. Planning for cooler cities: a framework to prio- ritise green infrastructure to mitigate high temperatures in ur- ban landscapes. Landscape and Urban Planning, 134: 127–138. DOI 10.1016/j.landurbplan.2014.10.018 Obu J., Podobnikar T . 2013. Algoritem za prepoznavanje kraških ko- tanj na podlagi digitalnega modela reliefa. Geodetski vestnik, 57, 2: 260–270. Oliveira S., Andrade H., Vaz T . 2011. The cooling effect of green spaces as a contribution to the mitigation of urban heat: a case study in Lisbon. Building and Environment, 46: 2186–2194. Park J., Kim J., Lee D.K., Park C.Y., Jeoing S.G. 2017. The influence of small green space type and structure at the street level on urban heat island mitigation. Urban Forestry & Urban Greening, 21: 203–212. DOI 10.1016/j.ufug.2016.12.005 Rizwan A.M., Leung D.Y.C., Chuno L. 2008. A review on the genera- tion, determination and mitigation of Urban Heat Island. Jour- nal of Environmental Sciences, 20, 1: 120–128. DOI 10.1016/ S1001-0742(08)60019-4 Schwarz N., Lautenbach S., Seppelt R. 2011. Exploring indicators for quantifying surface urban heat islands of European cities with MODIS land surface temperatures. Remote Sensing of Enviro- nment, 115, 12: 3175–3186. Tran D.X., Pla F., Latorre-Carmona P ., Myint S.W., Caetano M., Kieu H.V. 2017. Characterizing the relationship between land use land cover change and land surface temperature. Journal of Photo- grammetry and Remote Sensing, 124: 119–132. DOI 10.1016/j. isprsjprs.2017.01.001 Yang P ., Ren G., Liu W. 2013. Spatial and temporal characteristics of Beijing urban heat island intensity. Journal of Applied Mete- orology and Climatology, 52: 1803–1816. DOI 10.1175/JAMC- D-12-0125.1 Zalar M., Pogačar T ., Črepinšek Z., Kajfež Bogataj L. 2017. Vročinski valovi kot naravna nesreča v mestih. V: Zorn M., Komac B., Ciglič R., Tičar J. (ur.) Trajnostni razvoj mest in naravne nesreče. Knji- žna zbirka Naravne nesreče, 4. Ljubljana, Založba ZRC: 41–49.