Wang et al. (2026) Fusing ERA5-Land and SMAP L4 for an improved global soil moisture product (1950–2025)

Identification

• Journal: Earth system science data
• Year: 2026
• Authors: Wenhong Wang, Shiao Feng, Yonggen Zhang, Zhongwang Wei, Jianzhi Dong, Lutz Weihermüller, Cong-Qiang Liu, Harry Vereecken
• DOI: 10.5194/essd-18-1061-2026

Research Groups

• Institute of Surface-Earth System Science, School of Earth System Science, Tianjin University, Tianjin, China.
• School of Atmospheric Sciences, Sun Yat-sen University, Guangzhou, China.
• Agrosphere Institute IBG-3, Forschungszentrum Jülich GmbH, Jülich, Germany.

Short Summary

This study develops a bias-corrected global soil moisture dataset (1950–2025) by fusing ERA5-Land reanalysis with SMAP L4 satellite-based data using a mean-variance rescaling method. The resulting product significantly reduces systematic errors and improves accuracy across diverse climate zones, providing a seamless 75-year record for climate and hydrological research.

Objective

• To create an improved, high-resolution global soil moisture product that overcomes the biases of reanalysis data (ERA5-Land), the temporal limitations of high-accuracy satellite data (SMAP L4), and the spatial gaps in multi-sensor products (ESA-CCI).

Study Configuration

• Spatial Scale: Global land surface at 0.1° spatial resolution.
• Temporal Scale: 1950 to 2025 at daily temporal resolution.

Methodology and Data

• Models used: CHTESSEL (underlying ERA5-Land), NASA Catchment land surface model (underlying SMAP L4), and a mean-variance rescaling fusion method.
• Data sources:
  • Reanalysis: ERA5-Land (1950–present).
  • Satellite/Assimilation: SMAP L4 (2015–2025) and ESA-CCI v09.1 Combined.
  • In situ Validation: A compilation of ~3.8 million records from five networks: International Soil Moisture Network (ISMN), China Meteorological Administration (CMA), Cemaden (Brazil), COSMOS-Europe, and SONTE-China.
• Fusion Technique: Grid-by-grid adjustment of ERA5-Land's mean and variance to match the statistical properties of SMAP L4 during a calibration period (2015–2025), with parameters applied backward to 1950.

Main Results

• Accuracy Improvement: During the primary validation period (2015–2020), the adjusted ERA5-Land product reduced the Root Mean Square Error (RMSE) by 24.6% and improved the normalized Nash-Sutcliffe Efficiency (NNSE) by 30.6% compared to the original ERA5-Land.
• Quantitative Performance: The adjusted product achieved a mean correlation coefficient ($r$) of 0.687, a mean RMSE of 0.087 $m^3 m^{-3}$, and a near-zero mean bias (−0.001 $m^3 m^{-3}$).
• Historical Robustness: Validation against independent historical data (1960–2015) showed a 19.7% reduction in RMSE and a 26.6% increase in NNSE, confirming the stability of the adjustment parameters over decades.
• Coverage: Unlike ESA-CCI, which has a mean data gap of 24.4% at station locations, the adjusted ERA5-Land provides seamless global coverage.

Contributions

• Long-term Seamless Dataset: Provides the first 75-year global soil moisture product that combines the temporal depth of reanalysis with the high accuracy of L-band satellite assimilation.
• Methodological Validation: Demonstrates the temporal transferability of mean-variance rescaling parameters by verifying them against over 60 years of in situ observations.
• Benchmark Evaluation: Conducts the first global-scale comparative analysis of the ESA-CCI v09.1 Combined product, identifying its specific regional and seasonal data gaps.

Funding

• National Natural Science Foundation of China (Grant nos. 42472327, 42077168, and 42293260).

Citation

@article{Wang2026Fusing,
  author = {Wang, Wenhong and Feng, Shiao and Zhang, Yonggen and Wei, Zhongwang and Dong, Jianzhi and Weihermüller, Lutz and Liu, Cong-Qiang and Vereecken, Harry},
  title = {Fusing ERA5-Land and SMAP L4 for an improved global soil moisture product (1950–2025)},
  journal = {Earth system science data},
  year = {2026},
  doi = {10.5194/essd-18-1061-2026},
  url = {https://doi.org/10.5194/essd-18-1061-2026}
}

Generated by BiblioAssistant using gemini-3-flash-preview (Google API)