Wang et al. (2026) A novel adaptive soil moisture retrieval method via stacked ensemble learning and a local Bayesian dynamic algorithm

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-03-19
• Authors: Fuqiang Wang, Ruiping Li, Sinan Wang, Xiaoming Ma, Jinming Yang, Jialu Dai, Xiaohui Lian, Jun Zhao, Liying Zhou, Yanxin Wang
• DOI: 10.1016/j.jhydrol.2026.135335

Research Groups

• College of Water Conservancy and Civil Engineering, Inner Mongolia Agricultural University, Hohhot, China
• Institute of Water Resources for Pastoral Area, China Institute of Water Resources and Hydropower Research, Hohhot, China
• Water Resources Research Institute of Inner Mongolia Autonomous Region, Hohhot, China
• State Key Laboratory of Ecological Safety and Sustainable Development in Arid Lands, Xinjiang Institute of Ecology and Geography, Chinese Academy of Sciences, Urumqi, China
• Center for Agricultural Water Research in China, China Agricultural University, Beijing, China
• Key Laboratory of Plant Nutrition and the Agri-environment in Northwest China, Ministry of Agriculture and Rural Affairs, College of Resources and Environment, Northwest A&F University, Yangling, China
• State Key Laboratory of Water Engineering Ecology and Environment in Arid Area, Inner Mongolia Agricultural University, Hohhot, China
• Inner Mongolia Hydrology and Water Resources Center, Hohhot, China

Short Summary

This study introduces a novel local Bayesian dynamically weighted stacking ensemble learning model (Stacking-BO) and a high-resolution spatiotemporal multilayer soil moisture simulation framework to enhance the accuracy and stability of soil moisture retrieval, demonstrating superior performance over existing methods.

Objective

• To develop and evaluate a novel adaptive soil moisture retrieval method, Stacking-BO, which integrates stacked ensemble learning with a local Bayesian dynamic algorithm, aiming to overcome limitations of current soil moisture forecasting approaches and improve inversion methodology.

Study Configuration

• Spatial Scale: Regional to global context for water management, with specific analysis conducted in a defined research region (D, S, T, M) characterized by varying aridity. High-resolution spatiotemporal framework.
• Temporal Scale: July 2023 and July 2024, analyzing soil moisture at depths of 0–0.1 m, 0.1–0.2 m, and 0.2–0.3 m.

Methodology and Data

• Models used: Stacking-BO (local Bayesian dynamically weighted stacking ensemble learning model), OL-ESTARFM, Stacking1.
• Data sources: Remote sensing data (implied from the use of reflectance, Land Surface Temperature (LST), Albedo, Evapotranspiration (ET), Leaf Area Index (LAI), and Digital Elevation Model (DEM) as input features for soil moisture inversion).

Main Results

• The OL-ESTARFM model achieved a maximum reflectance R² of 0.962 for the fusion band and an R² of 0.95 for LST.
• Major drivers of soil moisture at different depths (0–0.1 m, 0.1–0.2 m, 0.2–0.3 m) in July 2023 and 2024 included Albedo (r₁ = −0.61), ET (r₂ = 0.65; r₃ = 0.63), LAI (r₁ = r₂ = 0.48), and DEM (r₃ = −0.44).
• The Stacking-BO model demonstrated improved performance over the Stacking1 model, showing a 0.04 greater R² and a 1.14% lower Root Mean Square Error (RMSE) in July 2023, and a 0.48% lower RMSE in July 2024 on the validation set.
• The fraction of soil with a moisture content less than 15% in the research region was 85%.

Contributions

• Introduction of a novel local Bayesian dynamically weighted stacking ensemble learning model (Stacking-BO) for enhanced soil moisture inversion.
• Development of a high-resolution spatiotemporal multilayer soil moisture simulation framework.
• Construction of three stacking model-based predictive value weight-related objective functions, optimized by a Bayesian algorithm to determine optimal weight combinations for each sample.
• Demonstrated significant improvements in accuracy and stability of soil moisture retrieval compared to traditional stacking models.

Funding

• Not explicitly mentioned in the provided paper text.

Citation

@article{Wang2026novel,
  author = {Wang, Fuqiang and Li, Ruiping and Wang, Sinan and Ma, Xiaoming and Yang, Jinming and Dai, Jialu and Lian, Xiaohui and Zhao, Jun and Zhou, Liying and Wang, Yanxin},
  title = {A novel adaptive soil moisture retrieval method via stacked ensemble learning and a local Bayesian dynamic algorithm},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135335},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135335}
}