Improved Process Understanding Using Stand-alone Land Surface Models in Simulating Permafrost

Land surface models (LSMs) still exhibit widespread deficiencies in frozen soil regions, particularly in overestimating soil temperature responses to air temperature under snow cover and in simulating soil moisture dynamics (see, e.g., Luo et al (2025)). Analysis of CMIP6 simulations reveals that, in frozen soil areas, the influence of LSMs on coupled climate model results is comparable in magnitude to that of atmospheric forcing. Moreover, compensating effects between land and atmosphere components often lead to apparently better performance in coupled simulations than in offline LSM experiments. This compensation poses a risk that structural deficiencies in LSMs may remain obscured when evaluating coupled model performance.

To identify specific weaknesses in current LSM formulations, we conduct offline simulations using three land surface models—CLM5, TERRA standalone, and JSBACH—at permafrost observation sites, including Bayelva (Svalbard) and Samoylov (Lena River Delta). The models are driven by meteorological forcing derived from the WATCH Forcing Data methodology applied to ERA5 (WFDE5), which is further bias-corrected to the sites using in situ observations.

We evaluate simulated snow water equivalent, soil temperature, soil moisture, and surface energy fluxes against observations, complemented by targeted sensitivity experiments. This approach aims to diagnose the key processes responsible for model biases in permafrost regions and to assess potential pathways for improving land surface model performance under cold-region conditions.

 

Zhicheng Luo, Risto, D., B. Ahrens (2025) Assessing Climate Modeling Uncertainties in the Siberian Frozen Soil Regions by Contrasting CMIP6 and LS3MIP. The Cryosphere, 19, 6547–6576. https://doi.org/10.5194/tc-19-6547-2025