Abstract:
Efforts to combine satellite images from different sources are particularly needed in Land Surface Temperature-based (LST) studies. This research proposes for the first time, to our knowledge, a Google Earth Engine-based (GEE) 10-m LST system, named Ten-ST-GEE. It is based on both Landsat-8 and Sentinel-2 bands. Ten-ST-GEE has the ability to automatically transform 30-m to 10-m LST at Landsat-8 overpass time. Machine learning and regression methods (i.e., OLS, RLS, DisTrad, RF, and SVM) are embedded within this system. Ten-ST-GEE was applied over two agricultural lands and two urban regions in the United States of America and in Lebanon. OLS and RLS showed an RMSE of ∼1.1 °C compared to ∼2.4 °C for DisTrad and ∼2.5 °C for RF and SVM. The open-source and automated Ten-ST-GEE can generate information at the building-level and within the agricultural parcels. It has the potential to be portable to any region across the Globe, aiming at better management of environmental resources. © 2022 Elsevier Ltd
