Disaggregation of remotely sensed land surface temperature: a generalized paradigm


Land surface temperature (LST) is an important parameter highly responsive to surface energy fluxes and has become valuable to many disciplines. Prior to the advent of satellites, it was difficult to obtain LSTs over extensive areas. Even today, as a result of the resolution tradeoffs involved in using satellite data, it is difficult, and sometimes impossible, to acquire satellite LSTs with high spatial and temporal resolutions. This low resolution results in a thermal mixture effect, where the resolution cells are larger than the thermal elements.

The disaggregation of remotely sensed land surface temperature (DLST), a research field that focuses on decomposing pixel-based temperatures, has been critical in related fields such as the surface flux downscaling, forest fire detection, and urban heat island monitoring and it is now growing rapidly as one of the thriving subbranches of thermal remote sensing. Various methods have been independently proposed for DLST in recent decades. However, this field is suffering a disorderly development. We thus critically investigate the interdisciplinary literature on DLST and identify the terms used to denote DLST in different disciplines. Two subtopics of DLST, thermal sharpening (TSP) and temperature unmixing (TUM), are identified as a dual pair of DLST because of their parallel areas of interest. Previous studies are classified into different categories in chronological and taxonomic order. We formulate definitions of TSP, TUM, and DLST, and we then examine how TSP and TUM are connected to related fields in remote sensing. Based on the literature, we present the key issues related to DLST, the recommended DLST methods in different applications, and the caveats that must be considered in future work, including (1) four predetermined assumptions (i.e., additivity, separability, connectivity, and convertibility), (2) the utilization of diurnal thermal observations, and (3) the complication of aggregation. This overview will provide a generalization of TSP and TUM, promote the understanding of DLST, and guide future research.

Fulltext Link