Can typical land surface model parameterizations support the expected soil moisture assimilation efficiency?

Authors: ZHOU, J - Tianjin University; DONG, J - Tianjin University; FENG, H - Hohai University; YANG, K - Tsinghua University; Crow, Wade; WU, Z - Hohai University; TIAN, X - Tianjin University; TIAN, J - Tsinghua University; MA, X - Tsinghua University; JIANG, Y - University Of Electronic Science And Technology Of China

Submitted to: Water Resources Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/14/2025
Publication Date: 3/29/2025
Citation: Zhou, J., Dong, J., Feng, H., Yang, K., Crow, W.T., Wu, Z., Tian, X., Tian, J., Ma, X., Jiang, Y. 2025. Can typical land surface model parameterizations support the expected soil moisture assimilation efficiency?. Water Resources Research. 61(4). Article e2024WR038702. https://doi.org/10.1029/2024WR038702.
DOI: https://doi.org/10.1029/2024WR038702

Interpretive Summary: State-of-the-art methods for monitoring agricultural drought are based on the application of complex land surface data assimilation systems to merge remotely sensed surface soil moisture retrievals with root-zone soil moisture estimates derived from a soil water balance model. These systems contain three separate components: i) a retrieval component to convert remotely-sensed observations into surface soil moisture estimates, ii) a soil water balance modeling component to convert observed precipitation into root-zone soil moisture predictions, and iii) a data assimilation component to combine information gleaned from both remote sensing and modelling. This paper derives and applies a novel benchmarking evaluation strategy to examine the performance of these components relative to a much simpler approach based on the direct averaging of remote sensing and modelling estimates. Results demonstrate that the limiting factor in the performance of complex land data assimilation systems is the accuracy of the soil water balance model - particularly its ability to accurately describe soil moisture dry-downs immediately following precipitation events. In this way, the paper provides valuable new insight into improving the ability of existing land data assimilation systems to capture the extent and severity of agricultural drought.

Technical Abstract: Remote sensing (RS) soil moisture retrievals are frequently assimilated into land surface models (LSMs) to enhance model estimates. However, soil moisture DA efficiency is highly model-dependent, making it imperative to investigate potential model limitations on DA for the advancement of land DA sys-tems. In this study, we examine soil moisture DA efficiency by benchmarking it against a reference soil moisture merging scheme based on identical error estimation theory and the same RS soil moisture datasets. In theory, the DA soil moisture estimates should be superior to the merged results – since DA can character-ize the temporal variation of model error and propagate DA benefits into subsequent forecast steps. Howev-er, ground-based validation results indicate that DA soil moisture estimates are found to be even worse than simply merged results in regions where the LSM is less precise than RS retrievals. Further combing synthet-ic experiment, we confirm that the unexpected DA results are primarily attributable to model parameter un-certainty, which leads to unrealistic soil moisture events (e.g., soil moisture dry-downs) and significantly hampers the DA application. Given this, soil moisture DA is likely to remain sub-optimal in achieving its desired goals due to existing model parameter uncertainty. Therefore, this study emphasizes the urgency and necessity of reducing LSM parameter uncertainty in land DA systems.