Model weighting for ISMIP6-Greenland based on observations and similarity among models

Abstract The Ice Sheet Model Intercomparison Project for CMIP6 (ISMIP6) resulted in many ice-sheet simulations from multiple ice-sheet models. To date, no model weighting studies have analyzed or quantified the model performance, possible duplication of the ISMIP6 ice-sheet models and the effect on mass loss projections. In this study, we adopt a model weighting scheme for the ISMIP6-Greenland that accounts for both model performance compared to observation and model similarity due to possible duplication. We choose ice velocity and thickness for the measurement of model performance, and we use all suitable variables to compute similarity indexes. We update the sea level rise contribution from ISMIP6-Greenland by the end of this century with the weights, and we find that, although the multi-model mean is not considerably shifted (mostly within $ \pm 1{\text{cm}}$ ), the model spreads are reduced by 10–30% after applying the model weights. The magnitude of reduction varies largely among experiments and types of model weights applied. In general, we find that the model weighting scheme is skillful in producing model weights that effectively and reasonably quantify the model performance and inter-dependency, which can potentially benefit the future phase of the Ice Sheet Model Intercomparison Project, i.e. ISMIP7.