Comparison of microwave remote sensing and land surface modeling for surface soil moisture climatology estimation

Authors: DONG, J - US Department Of Agriculture (USDA); Crow, Wade; TOBIN, K. - Texas A&M University; Cosh, Michael; Bosch, David - Dave; Starks, Patrick; Seyfried, Mark; Holifield Collins, Chandra

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/15/2020
Publication Date: 6/1/2020
Citation: Dong, J., Crow, W.T., Tobin, K., Cosh, M.H., Bosch, D.D., Starks, P.J., Seyfried, M.S., Holifield Collins, C.D. 2020. Comparison of microwave remote sensing and land surface modeling for surface soil moisture climatology estimation. Remote Sensing of Environment. 242:111756 . https://doi.org/10.1016/j.rse.2020.111756.
DOI: https://doi.org/10.1016/j.rse.2020.111756

Interpretive Summary: Accurate soil moisture information is critical for effectively monitoring the extent, duration and severity of agricultural drought events. To be of greatest value, such information should be provided relative to expectations for a given month of the year - since a soil moisture value in, for example, April has very different implications than the same soil moisture level occurring in August. Put another way - what often matters most is how much actual soil moisture has deviated from its typical value for a given point within the calendar year. Unfortunately, we currently lack a complete understanding of how expected soil moisture levels (referred to as the "soil moisture climatology") vary throughout the year. In this paper, we cross-compare soil moisture climatologies obtained from multiple modelling and remote sensing sources and evaluate them using long-term soil moisture measurements from ground-based instrumentation. Relative to existing older products, we find a substantial improvement in the quality of soil moisture climatology values obtained from new satellite-based remote sensing datasets. This suggests that the soil moisture climatology problem will soon be solved via the forthcoming availability of improved, long-term data sets from these new sources. Agricultural drought monitors will eventually use these results to enhance the value of their root-zone soil moisture estimates for mitigating the impact of drought on food and forage production.

Technical Abstract: Soil moisture climatology error dominates the overall inconsistency of different products. However, relatively little has been done to standardize - or even inter-compare - soil moisture climatology (SMC) information acquired from different sources. Therefore, this study evaluates surface (0 - 10 cm) and root-zone (0 - 40 cm) SMC obtained from four land surface models (LSMs) and two (C- and L-band) remote-sensing (RS) products using in-situ validations sites within the contiguous United States. Comparisons against in-situ observations demonstrate the spatial and inter-model variability of LSM surface SMC errors. Nonetheless, all the LSMs outperform surface SMC estimates obtained from C-band Advanced Microwave Scanning Radiometer for EOS (AMSR-E) retrievals. Relative to AMSR-E SMC retrievals, L-band Soil Moisture Ocean and Salinity (SMOS) demonstrates strongly increased skill in capturing observed surface SMC. In fact, SMOS-retrieved surface SMC intraseasonal variability and dynamic range estimates are generally more accurate than LSM-based SMCs. The feasibility of extending RS retrievals for root-zone SMC using simple exponential filtering is also investigated. It is notable that SMOS-derived root-zone SMC surprisingly captures the observed intraseasonal variability of SMC over most land covers with a skill comparable to the LSMs. These results suggest that the typical practice of neglecting remote-sensing SMC information in land data assimilation should be reconsidered for L-band retrievals. Although SMOS has made significant progresses towards retrieving absolute soil moisture values, a temporally constant dry bias is found in SMOS surface and root-zone SMCs over all land covers. Addressing this dry bias should be a priority for future generations of SMOS retrieval algorithms.