Disaggregating satellite soil moisture products based on soil thermal inertia: A comparison of a downscaling model built at two spatial scales

Abstract

Lack of high spatial resolution soil moisture data is a major limitation in many regional scale hydrologic, climatic and agricultural applications. The available satellite soil moisture data products are too coarse and unable to cater for this resolution requirement. Downscaling coarse spatial resolution satellite soil moisture retrievals is a feasible option to meet the required level of spatial resolution for those applications. The main focus of this study is to compare two soil thermal inertia-based downscaling models, built by using long-term time records of (i) point scale in-situ data and, (ii) 25 km resolution Global Land Data Assimilation System (GLDAS) land surface model outputs. The developed models were tested over Goulburn River catchment in the Upper Hunter Region of NSW, Australia to downscale Soil Moisture Active Passive (SMAP) 36 km radiometric product into 1 km resolution.The downscaled SMAP product from both models produced encouraging results with unbiased root mean square errors (ubRMSEs) of 0.07–0.10 cm3/cm3, against in-situ field data, and an average ubRMSEs of 0.07 cm3/cm3 when compared to the National Airborne Field Experiment 2005 (NAFE'05) soil moisture retrievals. Both models showed promising results over semi-arid regions in estimating soil moisture at a high spatial resolution, but with their own strengths and weaknesses. The findings here provide useful insights on the robustness of the soil thermal inertia relationship across scales and the effects of the model resolution to the downscaled soil moisture estimates. The approach demonstrated encouraging results over semi-arid regions in estimating soil moisture at a high spatial resolution.