Contextualizing GCM Biases in Low‐Cloud Coverage: The Role of Large‐Scale Conditions

Hamish Lewis, Gilles Bellon

Geophysical Research Letters · 24 July 2025 · Météo-France (incl. CNRM)

In one sentence. A machine-learning model trained on observed large-scale predictors attributes CMIP6 low-cloud biases about equally to large-scale conditions and to parameterizations.

Abstract

Abstract Large biases in subtropical marine low‐cloud cover (LCC) simulated by general circulation models (GCMs) are often attributed to their low‐cloud macrophysical parameterizations. There has been less investigation into other aspects of the simulated climate which may also contribute to biases in GCM LCC. Here, we quantify the contribution of GCM LCC biases from both the macrophysical low‐cloud parameterizations, and the GCM simulation of large‐scale meteorological conditions, which control a large portion of observed LCC variability. We use a machine‐learning model trained to predict observed relationship between LCC and large‐scale conditions to perform the bias decomposition within nine CMIP6 GCMs in both coupled and atmosphere‐only configurations. We find that the biases attributed to the GCM large‐scale conditions are of the same order as those attributed to low‐cloud parameterizations, in contradiction with previous assertions. This has implications for the use of statistical models of low‐clouds to assess their feedback to climate change.

Labels

Labels and summary are generated by Claude and reviewed weekly. Wrong? Edit this entry.