Luo et al. (2025) Assessing uncertainties in modeling the climate of the Siberian frozen soils by contrasting CMIP6 and LS3MIP

Identification

• Journal: ˜The œcryosphere
• Year: 2025
• Date: 2025-12-05
• Authors: Zhicheng Luo, Danny Risto, Bodo Ahrens
• DOI: 10.5194/tc-19-6547-2025

Research Groups

• Institute for Atmospheric and Environmental Sciences, Goethe University Frankfurt, Frankfurt a.M., Germany

Short Summary

This study quantifies the contributions of land surface parameterization and atmospheric forcing to discrepancies in frozen soil simulations in Siberia by contrasting CMIP6 and LS3MIP models. It finds that land-only models (LS3MIP) exhibit larger biases and spread in frozen soil temperatures than coupled models (CMIP6), indicating significant error compensation in coupled systems and underscoring the need for improved snow insulation and soil hydrothermal dynamics in land surface models.

Objective

• To quantify the contributions of land surface parameterization schemes and atmospheric forcing to discrepancies in modeling the climate of Siberian frozen soils.
• To analyze biases and uncertainties in frozen soil regions by comparing CMIP6 (coupled) and LS3MIP (land-only) historical climate simulations.
• To explore inter-model variations within LS3MIP and assess how specific structural features (e.g., bottom boundary conditions, snow thermal conductivity parameterizations) relate to model performance in frozen soil regions.

Study Configuration

• Spatial Scale: Siberia, focusing on 152 observational sites (excluding sites west of 60° E, east of 120° E, and south of 45° N to target colder regions).
• Temporal Scale: 30-year period from 1985 to 2014, using monthly averaged data.

Methodology and Data

• Models used:
  • CMIP6 (Coupled Model Intercomparison Project Phase 6): Historical simulations from seven climate models (CESM2, CNRM-CM6.1, CNRM-ESM2.1, IPSL-CM6A-LR, HadGEM3-GC31-LL, UKESM1.0-LL, MIROC6).
  • LS3MIP (Land Surface, Snow, and Soil Moisture Model Intercomparison Project): Land-only simulations using the land components of the corresponding CMIP6 models (CLM5.0, Surfex 8.0c, ORCHIDEE v2.0, JULES-HadGEM3-GL7.1, JULES-ES-1.0, MATSIRO6.0).
• Data sources:
  • Observational data: Daily meteorological data (2 m air temperature (tas), 0.2 m soil temperature (tsl), snow depth (snd), precipitation (pr)) from 152 quality-controlled sites in Siberia provided by the All-Russian Scientific Research Institute of Hydrometeorological Information-World Data Center (RIHMI-WDC).
  • Reanalysis data: ERA5-Land monthly averaged reanalysis data (0.1° × 0.1° horizontal resolution) from the European Centre for Medium-Range Weather Forecasts (ECMWF), and a coarsened version (E5LC) remapped to 100 km resolution.
  • Atmospheric forcing for LS3MIP: Global Soil Wetness Project Phase 3 (GSWP3), based on the 20th Century Reanalysis (20CR) and bias-corrected with observational datasets.

Main Results

• In winter months (December, January, February), the LS3MIP ensemble bias in 0.2 m soil temperature was larger than the CMIP6 bias (−3.6 °C vs. −2.7 °C).
• The spread of winter 0.2 m soil temperatures was also larger in the LS3MIP ensemble (4.6 °C) than in the CMIP6 ensemble (3.0 °C).
• For permafrost sites, CMIP6 simulations showed correlations below 0.6 for winter soil temperatures and below 0.8 for spring/autumn snow depth with observations.
• When simulated soil temperature dropped below −5 °C, the median 0.2 m soil temperature in CMIP6 simulations was 0.3 °C warmer than observations, while LS3MIP simulations were colder with a median cold bias of 0.7 °C.
• Biases of 2 m air temperature in coupled simulations had an opposite sign and amplified magnitude compared to their soil temperature biases, particularly below 0 °C.
• Land-only models demonstrated limited capability in reproducing soil temperatures and snow depth under severe cold conditions (surface air temperature below −15 °C).
• Four climate models and their land components underestimated the insulating role of snow; in cases with shallow snow depth (0–0.2 m), models simulated air-soil temperature differences up to 10 °C, whereas in situ measurements indicated even larger differences.
• CMIP6 models tended to compensate for errors in their land component with errors in the atmospheric model component.
• Cold biases were present in deeper soil layers (0.8 m and 1.6 m) in all thermal regimes for both ensembles, suggesting deficiencies in soil heat conductivity and capacity representation.

Contributions

• Quantified the distinct contributions of land surface parameterization schemes and atmospheric forcing to biases and uncertainties in frozen soil simulations in Siberia.
• Revealed significant error compensation mechanisms within coupled CMIP6 models, where atmospheric components can mask underlying deficiencies in land surface models.
• Emphasized the critical need for improved representation of surface-soil insulation (including snow thermal conductivity, snow density, snow depth, snow cover fraction, and surface organic matter) and soil hydrothermal dynamics in land surface models.
• Provided insights into how specific model features (e.g., snow thermal conductivity formulations, bottom boundary conditions) influence model performance in permafrost regions, guiding future model refinements.

Funding

• China Scholarship Council (CSC) (No. 202006040064)
• DWD IDEA S4S – project FS-SF (4823IDEAP2)
• Deutsches Klimarechenzentrum (DKRZ) (project ID bb1064)
• Goethe-HLR
• World Climate Research Programme (WCRP) for CMIP6
• Earth System Grid Federation (ESGF)
• Goethe University Frankfurt (open-access publication)

Citation

@article{Luo2025Assessing,
  author = {Luo, Zhicheng and Risto, Danny and Ahrens, Bodo},
  title = {Assessing uncertainties in modeling the climate of the Siberian frozen soils by contrasting CMIP6 and LS3MIP},
  journal = {˜The œcryosphere},
  year = {2025},
  doi = {10.5194/tc-19-6547-2025},
  url = {https://doi.org/10.5194/tc-19-6547-2025}
}