Compensating biases in CMIP6 TOA radiative-flux trends and their implications for EEI-based constraints on equilibrium climate sensitivity
Abstract. Earth's energy imbalance (EEI) is a key metric for monitoring climate. Decomposing the all-sky net EEI trend into shortwave (SW) and longwave (LW) components and separating clear-sky and cloud radiative effect (CRE) contributions provide insights into factors contributing to radiative forcing and feedbacks. Herein, we quantify trends in nine TOA radiativeflux components using CERES satellite data and 34 CMIP6 models. Decomposed radiative-flux trends, particularly under clear-sky conditions, are useful for constraining equilibrium climate sensitivity (ECS), implying a moderate ECS range of 2.0–4.5 K. However, biases in clear-sky and CRE trends compensate for each other, yielding an apparently consistent net EEI trend relative to CERES. These compensating biases may complicate the physical interpretation of EEI-based constraints on ECS. To understand the mechanisms underlying these biases, the combined use of cloud cover trends provides a useful approach. Cloud cover and EEI trends covary, and some CMIP6 models even misrepresent the sign relative to the negative trend observed. Robust constraints therefore require consistency between the net EEI trend and its SW, LW, clear-sky, and CRE components, and cloud cover trends provide an additional physical diagnostic. Temporal variations in CERES radiative-flux trends likely reflect a combination of time-dependent radiative forcing, feedback responses, and internal variability. Therefore, the inferred ECS range should be interpreted as indicative rather than definitive. Our findings highlight the importance of evaluating component-level agreement and sensitivity to the analysis period.