Enhancing the spatial resolution of satellite-derived land surface temperature mapping for urban areas
- Publisher: Elsevier
Land surface temperature (LST) is an important environmental variable for urban studies such as those focused on the urban heat island (UHI). Though satellite-derived LST could be a useful complement to traditional LST data sources, the spatial resolution of the thermal sensors limits the utility of remotely sensed thermal data. Here, a thermal sharpening technique is proposed which could enhance the spatial resolution of satellite-derived LST based on super-resolution mapping (SRM) and super-resolution reconstruction (SRR). This method overcomes the limitation of traditional thermal image sharpeners that require fine spatial resolution images for resolution enhancement. Furthermore, environmental studies such as UHI modelling typically use statistical methods which require the input variables to be independent, which means the input LST and other indices should be uncorrelated. The proposed Super-Resolution Thermal Sharpener (SRTS) does not rely on any surface index, ensuring the independence of the derived LST to be as independent as possible from the other variables that UHI modelling often requires. To validate the SRTS, its performance is compared against that of four popular thermal sharpeners: the thermal sharpening algorithm (TsHARP), adjusted stratified stepwise regression method (Stepwise), pixel block intensity modulation (PBIM), and emissivity modulation (EM). The privilege of using the combination of SRR and SRM was also verified by comparing the accuracy of SRTS with sharpening process only based on SRM or SRR. The results show that the SRTS can enhance the spatial resolution of LST with a magnitude of accuracy that is equal or even superior to other thermal sharpeners, even without requiring fine spatial resolution input. This shows the potential of SRTS for application in conditions where only limited meteorological data sources are available yet where fine spatial resolution LST is desirable.