A two-step fusion framework for quality improvement of a remotely sensed soil moisture product: A case study for the ECV product over the Tibetan Plateau

Abstract

High quality remotely sensed soil moisture (SM) with high temporal coverage, high variation capturing ability and high accuracy, is very important to agricultural, hydrological and environmental applications. The Essential Climate Variable (ECV) SM dataset is the first product attempting to address the continuity and reliability issues of individual SM sensors by using multi-satellite observations, but limitations in quality still persist. This study proposes a two-step fusion framework to obtain high temporal coverage, high variation capturing ability and high accuracy remotely sensed SM. The first step aims to improve the temporal coverage by filling the gaps of remotely sensed SM with a machine learning method. The second step aims to improve the variation capturing ability and accuracy of the high temporal coverage SM from first step using a cumulative distribution function (CDF) matching based fusion method. We chose the ECV and Fengyun (FY) SM products as an algorithm test to improve the quality of ECV SM over the Tibetan Plateau (TP). Results suggest that the two-step SM fusion framework improved the original ECV SM temporal coverage from 30.84% to 78.12% for the entire TP. The variation capturing ability, here expressed in terms of correlation coefficient (R), was also improved by 6% and 49% at Naqu and Maqu, respectively, compared with the high temporal coverage ECV SM. The accuracy, measured in terms of unbiased root mean square error (ubRMSE), was also improved. Hence, the fusion framework and the FY SM could effectively improve the quality of ECV SM. The high quality fused SM is expected to help us better understand the role of SM in the water and energy cycles under global change.