Top-of-atmosphere non-cloud radiative kernels based on the CAS-ESM2-0 under 1×, 2×, and 4×CO2 forcing levels

Accurate quantification of radiative feedbacks is essential for constraining climate sensitivity, and radiative kernels have become an important tool for diagnosing these feedbacks. However, most existing radiative kernels are constructed under present-day or pre-industrial conditions, potentially introducing biases when applied to strong CO2 forcing scenarios. Here, we develop new sets of non-cloud radiative kernels using the Chinese Academy of Sciences Earth System Model version 2.0 (CAS-ESM2-0) under 1×, 2× and 4×CO2 climate states, and assess their state dependence and performance in feedback diagnostics. Clear-sky linearity tests show that, in the abrupt-4×CO2 experiment, the 2×CO2 kernels better approximate the mean radiative sensitivity over climate responses and provide the best closure of radiation budget. In contrast, the 4×CO2 kernels produce the smallest residuals in feedback decomposition based on the 150-year Gregory regression method, because diagnosed feedbacks are more strongly influenced by radiative sensitivities sampled under substantially warmed conditions. These results highlight the limitations of applying present-day or pre-industrial radiative kernels to high-CO2 climates and the importance of accounting for kernel state dependence.