Greenland Monthly Accumulation Maps (1960–2022): A Statistical Semi-Empirical Bias-Adjustment Model

Abstract. Accurate estimates of snow accumulation over the Greenland ice-sheet are essential for reliable projections of sea-level rise. These are typically obtained from regional climate models, which carry systematic temporal and spatially variable biases, contributing to substantial uncertainties in sea-level rise projections. Here we present a novel statistical-semi-empirical model for bias-correcting gridded accumulation output from any regional climate model or reanalysis product, utilising the SUMup dataset, which provides the most comprehensive spatial and temporal coverage of surface mass balance observations to date. The method employs Empirical Orthogonal Function analysis to decompose the model accumulation output into the dominant patterns of spatial variability and their temporal evolution. Adjustment coefficients derived by fitting SUMup data enable the reconstruction of spatially complete, bias-corrected accumulation fields. We apply this approach to monthly accumulation output from the HIRLAM–ECHAM Regional Climate Model (HIRHAM5; 1960–2022), the Modèle Atmosphérique Régional (MAR3.14; 1960–2022), the Regional Atmospheric Climate Model (RACMO 2.4p1; 1980–2022), and the Copernicus Arctic Regional Reanalysis (CARRA; 1991–2022). Initial mean point-wise biases of −7.4 % (HIRHAM), −0.5 % (MAR), 0.0 % (RACMO) and +10.1 % (CARRA) (1991–2022), statistically significant for all models except RACMO, are reduced to ± 0.3 % following adjustment. Resulting bias-corrected mean annual accumulation rates over the ice sheet are estimated at 469 mm yr−1 (HIRHAM), 412 mm yr−1 (MAR), 435 mm yr−1 (RACMO) and 408 mm yr−1 (CARRA) between 1991–2022. Inter-model agreement improves significantly in the observation-rich accumulation zone, with a 68 % reduction in standard deviation of mean accumulation estimates, but deteriorates by 27 % in the sparsely sampled ablation zone, highlighting the need for additional observational constraints. Model bias is dominated by the southern ice sheet, with the largest statistically significant contributions from the south-east for HIRHAM (−39 to −54 Gt yr−1) and MAR (+30 to +33 Gt yr−1), and the south-west for RACMO (+20 to +26 Gt yr−1) and CARRA (+34 Gt yr−1). Temporal trends and temperature sensitivities exhibit a pronounced east-west contrast, with the east dominated by strong positive responses and negative responses in the west. The framework outlined in this study offers a scalable, transferable solution to improve accumulation estimates through enhanced integration of observational data, providing an improved input to ice-sheet models, with the potential to reduce uncertainties in future sea-level rise projections.