engleski

Seasonal variations of land surface temperature in the city of Dubrovnik, Croatia

Land cover properties are the key driver of differences in land surface temperature (LST), what is especially important in temporal analysis of LST. Moreover, the heterogeneity of land cover in urban areas indicates a specific spatial distribution of LST thus allowing areas with high heat load to be identified. This study investigates spatial and temporal variation of LST in Dubrovnik for land use/land cover (LULC) classes in four seasons of years 2020 and 2021 using remote sensing techniques, GIS, and statistical analysis. Each season is represented by one cloud-free Landsat-8 satellite image taken as close as possible to the middle of the season. LST and spectral indices for selected images were derived using Google Earth Engine, a cloud-based geospatial analysis platform. LST was retrieved from the Landsat-8 Thermal Infrared Sensor (TIRS) Band 10 data using a mono-window algorithm, while spectral indices were calculated from bands of the Operational Land Imager (OLI) sensor. LULC classes were grouped based on the dominance of urban or natural morphological features in order to assess the impact of urban form and natural properties on LST. For natural LULC classes, seasonal changes in land cover were quantified using Landsat-8 spectral indices, such as normalized difference vegetation index (NDVI), while urban ones were considered unchanged during analyzed seasons. Additionally, Copernicus data was used to signify differences between natural LULC classes. The results for all seasons show highest LST in urban LULC classes, emphasizing the surface urban heat island (SUHI) of Dubrovnik, and in barren land outside of the city, mostly on slopes of hills where growth of vegetation is prevented by strong bora winds. Lowest LST is recorded for natural LULC classes: forests, urban parks, and other green surfaces. Dominant karst topography in the city’s hinterlands also shows a profound influence on spatial variation of LST. A strong correlation is found between LST and NDVI, indicating that seasonal change in vegetation is influencing LST, e.g. seasonal change in leaf cover for broad-leaved forests.

Land Surface Temperature, Land Use/Land Cover, Landsat-8, Google Earth Engine, remote sensing, Dubrovnik

nije evidentirano