Reducing Structural Uncertainty in AMSR2 Soil Moisture Using a Model Combination Approach

Wednesday, 17 December 2014
Seokhyeon Kim1, Yi Liu2, Fiona Johnson3, Robert Parinussa4 and Ashish Sharma1, (1)University of New South Wales, School of Civil and Environmental Engineering, Sydney, NSW, Australia, (2)ARC Centre of Excellence for Climate Systems Science & Climate Change Research Centre, University of New South Wales, Sydney, Australia, (3)University of New South Wales, School of Civil and Environmental Engineering, Sydney, Australia, (4)VU University Amsterdam, Amsterdam, Netherlands
Soil moisture is an important variable in hydrological systems affecting the water cycle in the atmosphere, land surface and subsurface. In past decades, a number of passive microwave based soil moisture products have been used in various fields of the earth sciences. While passive microwave can provide near-real time soil moisture (global coverage every 1-3 days), its direct applications have been limited due to the coarse spatial resolution (>100 km2) and uncertainties resulting from a number of complex factors that affects the radiative transfer model. In this aspect, it is essential to validate the accuracy prior to actual applications and to improve the dataset itself and the retrieval algorithms. As a first step to do this, two remotely sensed soil moisture products from the Advanced Microwave Scanning Radiometer 2 (AMSR2), retrieved by the Japan Aerospace Exploration Agency (JAXA) algorithm and the Land Parameter Retrieval Model (LPRM) are assessed and structural errors noted. The main findings are: 1) LPRM estimates are generally higher than JAXA except for in arid regions. 2) Comparisons with field measurements showed that JAXA has relatively better performance for locations with moderate vegetation density or dry conditions but the retrieved values are generally much lower than field measurements with little variance. 3) The advantage of LPRM is its ability to represent the relationship of soil moisture with surface temperature. 4) The performance of both products is strongly affected by the mean soil moisture. As it is found that the two products are complementary under the various conditions, a combinatorial approach is presented for improving the accuracy of soil moisture dataset. The approach is a linear combination technique which applies a spatio-temporal weighting, calculated based on error statistics of the products, to each product. This combinatorial approach is applied to a year of global dataset and generally shows better performances than the original products