Improvement of the AMSR-E algorithm for soil moisture estimation by introducing a fractional vegetation coverage dataset derived from MODIS data

Soil moisture is an important component of the hydrology of land surfaces. Accurate monitoring of soil moisture is essential in understanding energy and water cycles and ecological system processes. Microwave remote sensing using satellites is an effective method for collecting global information on land surface hydrology. In this study, the soil moisture algorithm of Koike et al. was revised by focusing on the vegetation component, with the goal of improving the accuracy of the soil moisture product of the Advanced Microwave Scanning Radiometer for the Earth Observing System mounted on the satellite Aqua. The water content of vegetation affects the sensitivity of the microwave remote sensing of soil moisture. In the Koike algorithm, a semi-empirical vegetation model with the assumption of uniform vegetation coverage was used to evaluate the vegetation effects on the retrieval of soil moisture data. However, satellite microwave radiometer observations have large footprints of several tens of kilometers. There are few land surface regions in the world that are uniformly covered with vegetation at this scale. The results of ground-based experiments demonstrated that non-uniformities in the vegetation coverage have very large effects on horizontally polarized waves. We therefore created a global fractional vegetation coverage dataset from the data gathered by the Moderate Resolution Imaging Spectroradiometer, and attempted to incorporate this into the algorithm. In addition, model parameters in the semi-empirical vegetation model were replaced on the basis of a ground-based experiment. The results were verified by the comparison of estimated and measured data for three locations with differing vegetation coverage conditions. Compared with results estimated by the Japan Aerospace Exploration Agency standard product version 5 (created by the algorithm before the current revision), the results estimated by the revised algorithm showed a significant improvement in accuracy and reduction in the number of erroneous estimations.