Permafrost estimation model in Upper Indus Basin

Abstract

Remotely sensed topo-climatic factors, potential incoming solar radiation (PISR), land surface temperature (LST), topographic wetness index (TWI), Surface emissivity, and elevation, and machine learning techniques are used for mapping the spatial distribution of permafrost in the Tso Kar, a sub-basin of Upper Indus Basin (UIB) in Leh, Ladakh (UT). This schematic model is employed to identify remotely sensed parameters which are crucial in assessing permafrost extent over the study region. It is followed by the application and tuning of several machine learning models to deliver an expected accuracy in terms of permafrost classes demarcated over the study region based on literature. Results show that the PISR, LST and TWI are the most significant remotely sensed parameters affecting the permafrost and associated processes. Above 5000 m a.s.l., the proportion of permafrost in the study catchment is higher. Synergistic use of remote sensing image processing and machine learning techniques together provide mapping of permafrost over the region, which is elusive so far