CMIP6 models overestimate sea ice melt, growth and conduction relative to ice mass balance buoy estimates

Abstract

With the ongoing decline in Arctic sea ice extent, the accurate simulation of Arctic sea ice in coupled models remains an important problem in climate modelling. In this study, the substantial Coupled Model Intercomparison Project Phase 6 (CMIP6) model spread in Arctic sea ice extent and volume is investigated using a novel, process-based approach. An observational dataset derived from the Arctic ice mass balance buoy (IMB) network is used to evaluate fluxes of melt, growth and conduction produced by a subset of CMIP6 models, to better understand the model processes that underlie the large-scale sea ice states. Due to the sparse nature of the IMB observations, the evaluation is performed by comparing distributions of modelled and observed fluxes in the densely sampled regions of the North Pole and Beaufort Sea. We find that all fluxes are routinely biased high in magnitude with respect to the IMB measurements by nearly all models, with too much melt in summer and too much conduction and growth in winter, even as a function of ice thickness. We also show that fluxes vary in ways which are physically consistent with the thermodynamic parameterisations used and that these effects likely modulate the large-scale relationship between ice thickness and ice growth and melt in the CMIP6 models.