Emergent constraints on CMIP6 climate warming projections using cloud diagnostics

Abstract:

The latest Sixth Coupled Model Intercomparison Project (CMIP6) multi-model ensemble includes more models forced by a new set of emissions and land use scenarios and simulates a larger spread of projected climate warming than CMIP5. Here, we show that the projected warming is well-correlated with historical metrics including the extratropical low cloud fraction (LCF) sensitivity to the seasonal cycle of sea surface temperature (SST), interannual variability of LCF with SST variation in the tropics, the global near surface air temperature (GSAT) trend and low cloud shallowness. These physically-meaningful relationships enable us to constrain future warming with models that can reproduce observations well. We first select the two LCF related metrics as most influential constraints across four metrics based on stepwise linear regression method. The multiple diagnostic regression with selected metrics performs better on warming projections in cross-validation than a regression model which applies GSAT trend as constraint. Application of the LCF metrics results in narrower 5-95% ranges of warming, and higher mean warming, compared to the unconstrained CMIP6 ensemble. This result differs from constrained projections based on the GSAT trend alone, which result in lower mean warming.