Fu et al. (2026) Effectively assimilate satellite land surface temperature into offline land surface models within ensemble-based assimilation frameworks

Identification

• Journal: Geoscientific model development
• Year: 2026
• Date: 2026-09-28
• Authors: Yunhao Fu, Yongjun Zheng, Jingjia Luo
• DOI: 10.5194/gmd-19-9177-2026

Research Groups

• State Key Laboratory of Climate System Prediction and Risk Management (CPRM)/Institute of Climate Application Research (ICAR), Nanjing University of Information Science and Technology, Nanjing 210044, China
• School of Atmospheric Physics, Nanjing University of Information Science and Technology, Nanjing 210044, China
• School of Artificial Intelligence, Nanjing University of Information Science and Technology, Nanjing 210044, China

Short Summary

This study proposes a novel scheme to effectively assimilate satellite land surface temperature (LST) into offline land surface models by jointly updating soil temperature and soil moisture in upper layers. This approach significantly improves soil temperature, soil moisture, and snow variables, outperforming conventional temperature-only updates.

Objective

• To propose and evaluate a new data assimilation scheme that effectively assimilates satellite land surface temperature (LST) into offline land surface models by simultaneously updating both soil temperature and soil moisture in the upper soil layers within an ensemble-based framework.

Study Configuration

• Spatial Scale: Global, with a grid resolution of 0.5° × 0.5°. The localization radius for data assimilation varies linearly from 73 km at 80° N/S to 365 km at the equator.
• Temporal Scale: A 1-year assimilation period (1 January to 31 December 2001) followed by a 2-month free forecast (1 January to 28 February 2002). LST observations are assimilated every 3 hours, and the Common Land Model (CoLM) runs with an 1800-second time step.

Methodology and Data

• Models used:
  • Land Surface Model (LSM): Common Land Model (CoLM) version 2014.
  • Data Assimilation (DA) method: Local Ensemble Transform Kalman Filter (LETKF).
• Data sources:
  • Satellite observations: Moderate Resolution Imaging Spectroradiometer (MODIS) derived Land Surface Temperature (LST) from the Terra satellite.
  • Atmospheric forcing: WFDE5 forcing dataset (bias-adjusted ERA5 reanalysis).
  • Evaluation datasets:
    • Gridded reanalysis/model products: ERA5-Land, Global Land Data Assimilation System (GLDAS-2.1 Noah LSM), and Modern-Era Retrospective analysis for Research and Applications version 2 (MERRA2).
    • Independent satellite observations: ATSR-2, AVHRR, and MODIS (for snow cover).
    • In-situ station data: FLUXNET2015 and AmeriFlux networks (for sensible and latent heat fluxes).

Main Results

• Land Surface Temperature (LST): Showed marginal enhancement after assimilation (maximum absolute difference of approximately 0.1 K), consistent with its fast-varying nature and strong atmospheric forcing. Over Asia, a maximum RMSD reduction of 1.43% was observed.
• Soil Temperature: Significant reduction in BIAS over Northeast Asia, with magnitudes of 1.0 K (0–10 cm layer), 1.5 K (40–100 cm layer), and 2.0 K (100–200 cm layer). Global RMSD reductions of approximately 0.1 K (0–10 cm) and 0.2 K (100–200 cm) were observed, with improvements increasing with soil depth.
• Snow Temperature and Depth: Prominent improvements over Northeast Asia, with RMSD reductions of approximately 4 K for snow temperature and 150 mm for snow depth. The most significant enhancements in snow cover and depth occurred during spring and autumn (snow melting and freezing periods).
• Soil Moisture: Notable improvements in soil water content, particularly over humid tropical regions. The largest reductions in unbiased RMSD over the Amazon Rainforest were approximately 0.06 kg m⁻² (0–10 cm), 0.12 kg m⁻² (10–40 cm), 0.15 kg m⁻² (40–100 cm), and 6.00 kg m⁻² (100–200 cm), with improvements increasing with soil depth in humid areas.
• Surface Heat Fluxes: Improvements in evaporative heat flux were observed over western North America, the Tibetan Plateau, and Northeast Asia. Latent heat flux showed considerable improvements in humid regions (e.g., Amazon Rainforest, central Africa), while sensible heat flux exhibited more variable responses. Site-specific evaluations against FLUXNET/AmeriFlux data showed improvements in at least one, and often both, heat fluxes at most stations.

Contributions

• Introduces a novel LST assimilation scheme that jointly updates both soil temperature and soil moisture in the upper soil layers, departing from conventional temperature-only or separate updates.
• Demonstrates that this joint update strategy effectively prolongs the influence of LST assimilation and produces more physically consistent energy and water fields, particularly in regions experiencing freeze/thaw cycles.
• Provides a comprehensive evaluation of the scheme's performance using multiple third-party gridded datasets, independent satellite observations, and in-situ flux tower data, enhancing the robustness of the findings.
• Highlights the critical role of LST assimilation in land-surface-process modeling, showing its impact on deeper soil layers and snow variables through vertical diffusion and phase changes within the model.

Funding

• National Natural Science Foundation of China (grant nos. 12241104 and 42275161)
• State Key Laboratory of Climate System Prediction and Risk Management (CPRM) initiative project (grant no. CPRM-2025-NUIST-012)

Citation

@article{Fu2026Effectively,
  author = {Fu, Yunhao and Zheng, Yongjun and Luo, Jingjia},
  title = {Effectively assimilate satellite land surface temperature into offline land surface models within ensemble-based assimilation frameworks},
  journal = {Geoscientific model development},
  year = {2026},
  doi = {10.5194/gmd-19-9177-2026},
  url = {https://doi.org/10.5194/gmd-19-9177-2026}
}