Chao et al. (2025) Evaluation of satellite soil moisture products (AMSR2, SMAP L3/L4) across mainland China using in situ data (2020–2024)

Identification

• Journal: Agricultural Water Management
• Year: 2025
• Date: 2025-12-02
• Authors: Lijun Chao, Siying Li, Sheng Wang, Guoqing Wang, Jianbin Su, Ke Zhang
• DOI: 10.1016/j.agwat.2025.110040

Research Groups

• College of Hydrology and Water Resources, Hohai University, Nanjing, Jiangsu, China
• China Meteorological Administration Hydro-Meteorology Key Laboratory, Hohai University, Nanjing, Jiangsu, China
• College of Computer Science and Software Engineering, Hohai University, Nanjing, Jiangsu, China
• Research Center for Climate Change, Ministry of Water Resources, Nanjing, Jiangsu, China
• Key Laboratory of Virtual Geographic Environment (Ministry of Education of PRC), Nanjing Normal University, Nanjing, Jiangsu, China

Short Summary

This study evaluated the soil moisture monitoring performance of three satellite products (AMSR2, SMAP L3, and SMAP L4) across mainland China using 3293 in-situ stations and Monte Carlo simulations. It found that SMAP L4 consistently performed best, followed by SMAP L3, while AMSR2 exhibited the largest errors, with performance differences primarily driven by sensor properties, algorithm complexity, and land cover heterogeneity.

Objective

• To evaluate the soil moisture monitoring performance (accuracy, consistency, and stability) of three satellite products (AMSR2, SMAP L3, and SMAP L4) across mainland China, considering temporal scales, geographical regions, and land cover types.

Study Configuration

• Spatial Scale: Mainland China (approximately 9.6 million km²), divided into seven geographical zones. Satellite product resolutions: AMSR2 L3 (25 km x 25 km), SMAP L3 (36 km x 36 km), SMAP L4 (9 km x 9 km). Validation used 3293 in-situ stations.
• Temporal Scale: January 1, 2020 to June 30, 2024 (4.5 years). Evaluations were conducted on annual, seasonal (spring, summer, autumn, winter), and monthly scales. SMAP L4 3-hour data were ensemble averaged to daily.

Methodology and Data

• Models used:
  • SMAP L4: Ensemble Kalman Filter (EnKF) assimilation method, NASA Catchment Land Surface Model, microwave radiation transmission model.
  • AMSR2 L3: Radiative transfer models (e.g., LPRM algorithm).
  • SMAP L3: Dual Channel Algorithm (DCA), Vertical Polarization Single Channel Algorithm (SCA-V), Horizontal Polarization Single Channel Algorithm (SCA-H).
  • Monte Carlo simulations (10,000 iterations) for uncertainty quantification.
• Data sources:
  • In-situ: "Automatic Soil Moisture Observation Dataset" from the China Meteorological Administration (CMA), comprising 3293 stations. Provides daily soil volumetric water content (SVWC) at 0-10 cm depth, measured using Frequency Domain Reflectometry (FDR) devices.
  • Satellite:
    • AMSR2 L3 (Advanced Microwave Scanning Radiometer 2) from JAXA/NASA Earthdata.
    • SMAP L3 (Soil Moisture Active Passive) from NASA.
    • SMAP L4 (Soil Moisture Active Passive) from NASA.
  • Auxiliary: 2020 CCI-LC global land cover product from the European Space Agency (ESA) at 300 m resolution, reclassified into six major types: agricultural, forest, grassland, wetland, settlement, and other.
  • Statistical Metrics: Bias, Root Mean Square Error (RMSE), unbiased Root Mean Square Error (ubRMSE), and correlation coefficient (R).

Main Results

• Overall Performance Ranking: SMAP L4 consistently demonstrated the best performance, followed by SMAP L3, while AMSR2 exhibited the largest errors across all temporal and spatial scales. This ranking was robust and not affected by site distribution.
• SMAP L4 Performance: Showed excellent accuracy, stability, and adaptability. Achieved the lowest Bias in winter (0.0049 m³/m³) and the strongest correlation in humid areas like Central China (R = 0.4891). It performed well on homogeneous surfaces (e.g., agricultural land R = 0.4763) and maintained good accuracy in forest areas. Monte Carlo simulations confirmed its stable error distributions (R-value fluctuation of 0.0002).
• SMAP L3 Performance: Displayed moderate performance with good accuracy (median Bias of 0.0043 m³/m³) but suffered from timing continuity shortcomings due to data gaps (e.g., RFI sensitivity). It achieved the lowest ubRMSE in the Southwest region.
• AMSR2 Performance: Consistently showed the largest errors and significantly underestimated soil moisture. It exhibited a dry bias across all temporal and spatial scales (RMSE = 0.2075–0.2265 m³/m³), with pronounced vegetation interference (summer ubRMSE increased to 0.1548 m³/m³). Errors significantly increased in densely vegetated areas (e.g., forest R = 0.1764 in South China).
• Impact of Land Cover: Product performance varied significantly with land cover type. High-precision areas (agriculture, grassland, other) yielded the best results for all products. Challenging medium-precision zones (forests) led to significant performance decline for AMSR2 and SMAP L3, while SMAP L4 maintained robustness due to its assimilation system. Low precision/high-risk areas (settlements, wetlands) posed challenges to all products due to high pixel heterogeneity.
• Seasonal Dynamics: All products displayed distinct seasonal rhythms. AMSR2 errors were highest in summer due to vegetation. SMAP L4 performed best in winter, effectively handling freeze-thaw processes. All products showed decreased correlation in August (China's rainy season) due to rapidly changing surface conditions.

Contributions

• Conducted the first nationwide, large-scale quality assessment of AMSR2 and SMAP (L3 and L4) soil moisture products across mainland China, utilizing data from 3293 in-situ stations.
• Revealed the underlying mechanisms driving performance differences in satellite soil moisture products from the perspective of land cover types, providing a more universal reference than traditional geographical zoning.
• Developed and applied a multi-scale and multi-dimensional verification framework to capture product response characteristics under various environmental pressures, enhancing understanding of their temporal stability and spatial adaptability.
• Provided a scientific basis for the application and algorithm optimization of soil moisture remote sensing products in China, particularly for water resource management and drought monitoring.

Funding

• National Key Research and Development Program of China (2023YFC3006500 and 2023YFC3006503–1)
• National Natural Sciences Foundation of China (U2243228)
• National Natural Science Foundation of China (52121006)
• Fundamental Research Funds for the Central Universities of China (B240201085)
• Open Foundation of the China Meteorological Administration Hydro-Meteorology Key Laboratory (23SWQXM044)

Citation

@article{Chao2025Evaluation,
  author = {Chao, Lijun and Li, Siying and Wang, Sheng and Wang, Guoqing and Su, Jianbin and Zhang, Ke},
  title = {Evaluation of satellite soil moisture products (AMSR2, SMAP L3/L4) across mainland China using in situ data (2020–2024)},
  journal = {Agricultural Water Management},
  year = {2025},
  doi = {10.1016/j.agwat.2025.110040},
  url = {https://doi.org/10.1016/j.agwat.2025.110040}
}