Assessing the impacts of land use and land cover patterns on summer urban heat island effect.

摘要

Remotely sensed thermal infrared data have been widely used to retrieve land surface temperature. Land Surface Temperature (LST) and emissivity data are used in urban climate and environmental studies, mainly for analyzing LST patterns and the LST relationship with surface characteristics, to assess surface urban heat island (SUHI) effect. This research re-examines the relationship between urban LST and land surface characteristics. While reliable air temperature data could be used to assess the nightime UHI effect, reliable, high-quality air temperature data are lacking due to the few number of weather observation stations. Further, the air temperature data observed at these weather stations does not represent the area-wide pattern of air temperature at the pixel-scale. Therefore, the second objective of this thesis, to estimate continuous air temperature, at pixel level from weather station data. We developed a Random Forest Regression Model based on different land cover parameters, including the original reflectance values of six Landsat 8 multispectral bands, Normalized Difference Vegetation Index (NDVI), Biophysical Composition Index (BCI), Percentage of Impervious Surface (%ISA), Percentage of Tree Canopy (%Tree), and the first three Tasseled Cap (TC) components. This research found that high LST mainly occurs on impervious areas, comparatively, vegetation covered areas occupies relatively low LST. In addition, the Random Forests Regression Model appears useful for estimating minimum air temperature and mapping the distribution the minimum air temperature in the study area.