Climate models with moderate climate sensitivity best simulate the magnitude of Earth's energy imbalance

Abstract. Recent studies have highlighted that state-of-the-art climate models are not able to simulate the large observed trend in Earth's energy imbalance. Here we evaluate climate models' ability to represent both the trend and the magnitude of the imbalance, while accounting for model energy leakage and remnant drift. As reference we use satellite observations and we find that every observed annual mean energy imbalance is within the range simulated by models, including the record year 2023, and when averaged over the 2001–2025 period, 13 out of 30 models simulate magnitudes of the imbalance that are statistically consistent with the observations. Models, however, generally underestimate the positive trend in the energy imbalance, albeit barely within the range of uncertainty. We suspected that a discontinuity in volcanic forcing between the historical and future scenario in 2014–2015 could have caused the underestimated trend, but only found evidence of such artifacts for a few models. Finally, we find a weak correlation between short-term decadal warming and energy imbalance, but a surprisingly close relationship between energy imbalance and equilibrium climate sensitivity. Based on observational constraints, the relationship suggests that models with climate sensitivities of 3 to 5 K best simulate the observed energy imbalance.