Correcting Regional Climate Bias in Reanalysis Forcing Products for The Warming Permafrost Model Intercomparison Project

Arctic and subarctic regions are warming faster than the global average, leading to poorly understood biophysical, biogeochemical, and ecohydrological changes in high-latitude ecosystems. To better understand these ecosystem changes and their impacts on permafrost thaw and subsequent feedbacks to the earth system, we have launched the Warming Permafrost Model Intercomparison Project (WarPMIP) to intercompare and benchmark model performance at experimental permafrost warming sites over the pan-Arctic. Considering that climate forcing is an important determining factor in model performance, it is critical to identify and correct for regional climate biases in forcing products for model intercomparisons within WarPMIP. However, climate observations in high-latitude regions are often limited, especially over winter, and current regional coarse-resolution reanalysis products that drive soil biogeochemistry models are typically biased compared to site-level meteorological measurements. To more accurately represent climate to force our model simulations we used both multiplicative and additive bias-correction on five reanalysis products (CRUNCEP, CRUJRA, GSWP3, MERRA-2, ERA5) compared to long-term climate observations at sites where experimental warming manipulations have been performed. We further report on the effect sizes of these climate uncertainties using three land models participating in the WarPMIP (ELM, CLM, ecosys). Preliminary results indicate that snow water equivalent, rainfall, wind speed, and short-wave incoming radiation were most biased compared to site-level observations and that the availability and quality of snow depth and snow water equivalent measurements are major limiting factors in the fidelity of climate product creation. Ultimately, identifying and correcting biases in forcing datasets is an important first step to more realistically drive WarPMIP simulations of permafrost warming and will help to better evaluate model structural and parametric uncertainties.