A neural-network technique for retrieving land surface temperature from AMSR-E passive microwave data

Abstract

It is very difficult to retrieve the land surface temperature (LST) from passive microwave remote sensing because a single multi-frequency thermal measurement with N bands owns n equations in N+1unknowns (N emissivities and LST) which is a typical ill-posed inversion problem. However, the emissivity is mainly influenced by dielectric constant which is a function of physical temperature, salinity, water content, soil texture, and other factors (the structure and types of vegetation). These make it very difficult to develop a general physical algorithm. This paper intends to utilize the multiple- sensor/resolution and neural network to retrieve land surface temperature from AMSR-E data. MODIS LST product is made as ground data which overcomes the difficulty of obtaining large scale land surface temperature data. The retrieval result and analysis indicate that the neural network can be used to accurately retrieve land surface temperature from AMSR-E data.