Remote sensing error assessment for subsequent use in model confrontation

Abstract

Remote sensing provides an unprecedented opportunity to evaluate land surface models (LSMs) run at distributed scales. Yet, with so many remote sensing products available, it is unclear which product is the best, and how they should be validated due to the scaling issues between in situ measurements and remote sensing product grids. Therefore, a three-fold approach has been demonstrated at the Yanco core validation site in Australia (Yanco), whereby i) the representativeness of in situ soil moisture and evapotranspiration (ET) measurements within remote sensing product grids were investigated; ii) these representative measurements were used to validate remote sensing soil moisture and ET products, and iii) the validated products were used to evaluate distributed simulations of soil moisture and ET from the Joint UK Land Environment Simulator (JULES). In this research, the soil moisture stations within Yanco which provided representative measurements were identified based on geostatistical and temporal stability analysis of long-term soil moisture measurements and intensive measurements from three extensive field campaigns. Measurements from these stations were then used to validate soil moisture products from the Advanced Microwave Scanning Radiometer - 2 (AMSR-2) and Soil Moisture and Ocean Salinity (SMOS). Of these two sensors, soil moisture products from SMOS were found to perform best. In the case of ET, measurements from the same footprint derived using optical (LAS) and microwave (MWS) scintillometers, and an eddy covariance (EC) system were firstly inter-compared to understand their performances relative to each other. Subsequently, scintillometers were placed across different areas of a single 4 km Multi-functional Transport SATellites (MTSAT) ET grid established the representativeness of measurements from an EC system of the grid. EC measurements were consequently used to validate the performance of MTSAT 4 km ET products based on the Surface Energy Balance System (SEBS), Modified Priestley Taylor (PT-JPL) and Modified Penman Monteith (PM-Mu) models, whereby the PT-JPL model was found to perform the best. The importance of having a good understanding of satellite data was demonstrated by using both the best and poor products in a model intercomparison study showing that wrong conclusions can easily be reached. These results confirmed the utility of the rigorous and systematic methodology developed in this research.