Analyzing the effects of 2D and 3D urban structures on LST changes using remotely sensed data

2D and 3D urban structures play an important role in characterizing Land Surface Temperature (LST). The effects of 2D urban structures on changes of LST have been studied in many studies, but less attention has been paid to the effects of 3D urban structures. Therefore, in this study, we aimed at employing LiDAR point clouds and a Landsat-8 image to investigate the effects of 2D and 3D structures on LST changes in Tartu, Estonia. To this end, LST retrieval and extraction of 2D structures were carried out using a Landsat-8 image. A 3D land cover of the study area was generated through employing random forest (RF) classifier on a joint set of features extracted from LiDAR point clouds and an orthophoto image, as a basis to identify 3D structures. The results of this study indicate a high overall accuracy (97.33%) of RF in land cover classification. Among the 2D structures, Normalized Difference Built-up Index (NDBI) had a higher correlation than Fractional Vegetation Cover (FVC) with LST. The results also indicate the significant role of height in impervious surfaces and vegetation in increasing and decreasing LST. Finally, the results showed that the hot spots emerged in areas where the mean height of impervious surfaces was higher, while the cold spots were linked to areas of higher mean vegetation heights.