Satellite surface soil moisture from SMAP, SMOS, AMSR2 and ESA CCI: a comprehensive assessment using global ground-based soil moisture observations

Authors: MA, H. - Wuhan University; ZENG, J. - Beijing Normal University; CHEN, N. - Wuhan University; ZHANG, X. - Wuhan University; Cosh, Michael; WANG, W. - Wuhan University

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/30/2019
Publication Date: 9/15/2019
Citation: Ma, H., Zeng, J., Chen, N., Zhang, X., Cosh, M.H., Wang, W. 2019. Satellite surface soil moisture from SMAP, SMOS, AMSR2 and ESA CCI: a comprehensive assessment using global ground-based soil moisture observations. Remote Sensing of Environment. 231:111215. https://doi.org/10.1016/j.rse.2019.111215.
DOI: https://doi.org/10.1016/j.rse.2019.111215

Interpretive Summary: Soil moisture estimation from satellites require regular validation and calibration between the different satellite platforms, because they have different measurement methodologies. A comprehensive assessment of the currently available soil moisture products, including a composite product is presented with general comments made about the effectiveness across different land covers and biomass regimes. Discussion of potential future areas of improvement are identified including regions of high or low biomass, significant surface roughness and complex topography, as well as regions with high degrees of heterogeneity. These are the next challenges for these low resolution satellite products.

Technical Abstract: Comprehensive assessments on the reliability of remotely sensed soil moisture products are undeniably essential for their advancement and application. With the establishment of extensive dense networks across the globe, mismatches between satellite footprints and ground single-point observations can be feasibly relieved. In this study, five remotely sensed soil moisture products, namely, the Soil Moisture Active Passive (SMAP), two Soil Moisture and Ocean Salinity (SMOS) products, the Land Parameter Retrieval Model (LPRM) Advanced Microwave Scanning Radiometer 2 (AMSR2), and the European Space Agency (ESA) Climate Change Initiative (CCI), were systematically investigated by utilizing global in situ observations. Distinguished from previous studies, several perturbing factors comprising the surface temperature, vegetation optical depth (VOD), surface roughness and spatial heterogeneity were taken into account in this investigation. Furthermore, products’ skill under various climate types were also evaluated. Through the results, the SMAP product captures the temporal trends of ground soil moisture, exhibiting an averaged R of 0.728, whereas for overall accuracy, ESA CCI outperformed the other products with a slightly smaller ubRMSE of 0.041 m3m-3 and a bias of -0.005 m3m-3. This complementarity between SMAP and ESA CCI was further demonstrated under different climate conditions and can afford the reference of their integration for a more reliable global soil moisture product. The newly developed SMOS- INRA-CESBIO (SMOS-IC) was illustrated to gain considerable upgrades with regard to R and ubRMSE compared to the conventional SMOS CATDS product. Nevertheless, more serious underestimations were also found for SMOS-IC showing a Bias of -0.077 m3m-3. Generally, the underestimations of SMOS surface temperature under moderate or high VOD, heterogeneity, and most surface roughness conditions were consistent with the underestimations of its soil moisture product and provide the directions of product promotions. As for LPRM surface temperature, the worse skills can partially explain its unsatisfactory performance for LPRM soil moisture products. In spite of the relatively acceptable skills of SMAP and SMOS-IC soil moisture products under moderate VOD, small surface roughness, low heterogeneity conditions and temperate and cold climate types, advances in soil moisture products under high or even slightly low VOD, high roughness or topography complexity and heterogeneity, as well as in tropical or desert regions, remain challenging. It is expected that these findings can contribute to algorithm refinements, product enhancements (e.g., fusion and disaggregation) and usages.