Active-Layer Thickness across Alaska: Comparing Observation-Based Estimates with CMIP5 Earth System Model Predictions

Abstract

Predicted active-layer (AL) thicknesses of permafrost-affected soils influence earth system model predictions of C-climate feedbacks; yet, only a few observation-based studies have estimated AL thicknesses across large regions and at the spatial scale at which they vary. We used spatially referenced soil profile description data (n = 153) and environmental variables (topography, climate, and land cover) in a geographically weighted regression approach to predict the spatial variability of AL thickness across Alaska at a 60-m spatial resolution. The predicted AL thickness across Alaska ranged from 0.14 to 0.93 m, with a spatial average of 0.46 m and a coefficient of variation of 30%. The average prediction error and ratio of performance to deviation were 0.11 m and 1.8, respectively. Our study showed mean annual surface air temperature, land cover type, and slope gradient were primary controllers of AL thickness spatial variability. We compared our estimates with Coupled Model Intercomparison Project Phase 5 (CMIP5) earth system model predictions; those predictions showed large interquartile ranges in predicted AL thicknesses (0.35–4.4 m) indicating substantial overestimate of current AL thickness in Alaska, which might result in higher positive permafrost C feedback under future warming scenarios. The CMIP5 predictions of AL thicknesses spatial heterogeneity were unrealistic when compared with observations, and prediction errors were several times larger in comparison to errors from our observation-based approach. The coefficient of variability of AL thickness was substantially lower in CMIP5 predictions compared to our estimates when gridded at similar spatial resolutions. These results indicate the need for better process representations and representation of natural spatial heterogeneity due to local environment (topography, vegetation, and soil properties) in earth system models to generate a realistic variation of regional scale AL thickness, which could reduce the existing uncertainty in predicting permafrost C-climate feedbacks.