High-resolution surface and rootzone soil moisture over the US cropland: a novel framework assimilating multi-source remote sensing data, machine learning, and the Layered Green and Ampt Infiltration with Redistribution model

Authors: CAI, S - University Of Wisconsin; XU, Y - University Of Wisconsin; YANG, Z - National Agricultural Statistical Service (NASS, USDA); Crow, Wade; ZHOU, Z - University Of Wisconsin; SHANG, J - International Development Research Centre; LA FOLLETTE, P - Lynker; REBERG-HORTON, C - North Carolina State University College Of Veterinary Medicine; Schomberg, Harry; Mirsky, Steven; DAVIS, B - North Carolina State University College Of Veterinary Medicine; SEEHAVER, S - North Carolina State University College Of Veterinary Medicine; CORREIRA, A - Kansas State University; BASCHE, A - University Of Nebraska; WAGGONER, A - University Of Wisconsin; ELLIS, C - University Of Missouri; PARK, D - Clemson University; CAMPBELL, D - University Of Florida; PRESLEY, D - Kansas State University; HENRIQUEZ, I - North Carolina State University; DARBY, H - University Of Vermont; ADAM, J - Montana State University; MILLER, J - University Of Delaware; HAYMAKER, J - Virginia Tech; WALLACE, J - Pennsylvania State University; GASKIN, J - University Of Georgia; Balkcom, Kipling; RUHL, L - University Of Vermont; REITER, M - Virginia Tech; RUARK, M - University Of Wisconsin; FLESSNER, M - Virginia Tech; SIAS, C - Virginia Tech; DAVIS, P - Virginia Tech; TOMLINSON, P - Kansas State University; SMITH, R - University Of New Hampshire; WARREN, N - University Of New Hampshire; DIERKING, R - Indigo Ag; ARMSTRONG, S - Purdue University; ALMEIDA, T - University Of Nebraska; HUANG, J - University Of Wisconsin

Submitted to: Remote Sensing of Environment
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 11/30/2025
Publication Date: 12/6/2025
Citation: Cai, S., Xu, Y., Yang, Z., Crow, W.T., Zhou, Z., Shang, J., La Follette, P., Reberg-Horton, C., Schomberg, H.H., Mirsky, S.B., Davis, B., Seehaver, S., Correira, A., Basche, A., Waggoner, A., Ellis, C., Park, D., Campbell, D., Presley, D., Henriquez, I.E., Darby, H., Adam, J., Miller, J., Haymaker, J., Wallace, J., Gaskin, J., Balkcom, K.S., Ruhl, L., Reiter, M., Ruark, M., Flessner, M., Sias, C., Davis, P., Tomlinson, P., Smith, R., Warren, N.D., Dierking, R., Armstrong, S., Almeida, T., Huang, J. 2025. High-resolution surface and rootzone soil moisture over the US cropland: a novel framework assimilating multi-source remote sensing data, machine learning, and the Layered Green and Ampt Infiltration with Redistribution model. Remote Sensing of Environment. 334. Article e115167. https://doi.org/10.1016/j.rse.2025.115167.
DOI: https://doi.org/10.1016/j.rse.2025.115167

Interpretive Summary: Timely and accurate soil moisture data is valuable for a range of agricultural applications including: yield forecasting, irrigation scheduling, fertilizer application, and the monitoring of nutrient transport for sustainable crop production. Large advances have been made recently in the remote estimation of soil moisture using satellite observations; however, these approaches are generally restricted to very coarse spatial resolutions (typically >10 km). Here, we develop a novel machine learning approach to downscale satellite-based soil moisture information to a 100-m resolution that is much more consistent with the data use requirements for agricultural applications. This approach will eventually be used by the USDA National Agricultural Statistics Service for the field-scale monitoring of crop condition and status – as well as individual producers looking to minimize production costs.

Technical Abstract: Accurate and timely soil moisture (SM) monitoring in cropland is important for water resource management, drought forecasting, and nutrient transport estimation for sustainable crop production. Although recent research has applied machine learning (ML) to downscale coarse-resolution satellite SM products, most of this past work has focused on surface SM estimation, and the performance of SM products has not been intensively evaluated in cropland. This study introduces a novel framework that integrates multi-source satellite-based ML models with the Layered Green and Ampt Infiltration with Redistribution (LGAR) model to produce high-resolution (100 m, hourly) soil moisture products for both the surface layer (0-5 cm) and rootzone (0-100 cm) across cropland in the contiguous United States (CONUS). First, six ML models are trained combining SMAP-L4 SM products using multiple high-resolution remote sensing datasets (Sentinel-1, Sentinel-2, and Landsat) to predict surface and rootzone SM, respectively. These ML predictions are then assimilated into the LGAR model using the ensemble Kalman filter (EnKF). The framework is developed and validated using an eight-fold cross-validation scheme with in-situ data from 431 cropland sites across CONUS, sourced from three networks (SCAN, USCRN, and PSA). The 100-m hourly soil moisture (SM) from this framework surpasses existing products (9-km SMAP L4, 1-km ThySM) in spatial and temporal resolution and captures rootzone SM better than SMAP-HydroBlocks surface SM. It achieves good performance, with median ubRMSE of 0.054 m³/m³ and median KGE of 0.359 at the surface layer, outperforming SMAP L4, and median ubRMSE of 0.025 m³/m³ and median KGE of 0.201 at the rootzone. While the framework demonstrates strong performance, its accuracy varies across climatic regimes, with surface SM performing better in non-humid areas (median KGE=0.339 versus median KGE=0.390) and rootzone SM in humid regions (median KGE=0.255 versus median KGE=0.138). The high-resolution cropland SM products from this data assimilation framework potentially benefit multiple agricultural applications and provide valuable insights to support farmers and land managers in decision-making processes.