AMSR-E soil moisture inversion based on GRNN neural network

Abstract

Passive microwave remote sensing data and measured soil moisture content in AMSR-E (Advanced Microwave scanning Radiometer - Earth Observing System) had been adopted for hilly region of Central Sichuan province from 2006 to 2010. Using Tbv6.9, Tbv10.7, Tbv18.7 and Tbh36.5 as input factors, A GRNN neural network was established to retrieve soil water content, and the correlation coefficient between the simulated value and the spatial and temporal sequence of measured soil water was calculated. The results showed that the GRNN neural network's correlation coefficient inversion of soil moisture and the measured soil moisture is relatively higher than the one of soil moisture retrieved by the equation inversion of standard atmospheric transmission and the measured soil moisture. Thus using AMSR-E data based on GRNN neural network model to simulate the soil moisture is feasible.