Bridging spatio-temporal discontinuities in global soil moisture mapping by coupling physics in deep learning

Authors: WEI, Z - Wuhan University; MIAO, L - Henan University; ZHAO, T - Chinese Academy Of Sciences; MENG, L - Wuhan University; LU, H - Key Laboratory Of Textile Science & Technology; PENG, Z - Chinese Academy Of Sciences; Cosh, Michael; FANG, B - University Of Virginia; LAKSHMI, V - University Of Virginia; SHI, J - Chinese Academy Of Sciences

Submitted to: IEEE Transactions on Geoscience and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/13/2024
Publication Date: 8/20/2024
Citation: Wei, Z., Miao, L., Zhao, T., Meng, L., Lu, H., Peng, Z., Cosh, M.H., Fang, B., Lakshmi, V., Shi, J.C. 2024. Bridging spatio-temporal discontinuities in global soil moisture mapping by coupling physics in deep learning. IEEE Transactions on Geoscience and Remote Sensing. 313. https://doi.org/https://doi.org/10.1016/j.rse.2024.114371.
DOI: https://doi.org/10.1016/j.rse.2024.114371

Interpretive Summary: Satellite remote sensing of soil moisture can produce daily soil moisture maps with gaps, because most satellites used to produce them require approximately three days to cover the earth. Furthermore, there are coverage gaps for various reasons like interference and model errors. A gap filling process was tested to address this issue with the Soil Moisture Active Passive mission data products. The gap filled datasets were found to be of comparable accuracy as the original data which provides valuable improvement to the data record for when gaps occur. This research will aid in producing more comprehensive retrospective datasets for future modeling work on global soil moisture monitoring.

Technical Abstract: The launch of Soil Moisture Active Passive (SMAP) satellite in 2015 has resulted in significant achievements in global soil moisture mapping. Nonetheless, spatiotemporal discontinuities in the soil moisture products have arisen due to the limitations of its orbit scanning gap and retrieval algorithms. To address these issues, this paper presents a physics-constrained gap-filling method, PhyFill for short. The PhyFill method employs a partial convolutional neural network technique to explore spatial domain features of the original SMAP soil moisture data. Then, it incorporates variations in soil moisture induced by precipitation events and dry-down events as penalty terms in the loss function, thereby accounting for monotonicity and boundary constraints in the physical processes governing the dynamic fluctuations of soil moisture. The PhyFill model was applied to SMAP soil moisture data, resulting in continuous spatially daily soil moisture data on a global scale. Three validation strategies are employed: visual inspection through global pattern, simulated missing-region validation, and soil moisture validation with in situ measurements. The results indicated that the reconstructed soil moisture achieved a higher spatial coverage with satisfactory spatial continuity with neighbouring pixels. The simulated validation of the missing regions revealed that the averaged unbiased root mean square difference (ubRMSD) and correlation coefficient (R) were 0.0102 m3/m3 and 0.9919, respectively versus the gap filled SMAP product. The core validation sites demonstrated that the reconstructed soil moisture data has a consistent ubRMSD compared with the original SMAP soil moisture data (0.041 m3/m3 vs. 0.040 m3/m3). The PhyFill method can generate globally continuous, high accurate soil moisture estimates, providing remarkable support for advanced hydrological applications, e.g., global soil moisture dry-down events and patterns.