As the atmosphere warms, part of the cloud population shifts from ice and mixed-phase (‘cold’) to liquid (‘warm’) clouds. Because warm clouds are more reflective and longer-lived, this phase change reduces the solar flux absorbed by the Earth and constitutes a negative radiative feedback. This cooling feedback is weaker in the sixth phase of the Coupled Model Intercomparison Project (CMIP6) than in the fifth phase (CMIP5), contributing to greater greenhouse warming. Although this change is often attributed to improvements in the simulated cloud phase, another model bias persists: warm clouds precipitate too readily, potentially leading to underestimated negative lifetime feedbacks. In this study we modified a climate model to better simulate warm-rain probability and found that it exhibits a cloud lifetime feedback nearly three times larger than the default model. This suggests that model errors in cloud-precipitation processes may bias cloud feedbacks by as much as the CMIP5-to-CMIP6 climate sensitivity difference. Reliable climate model projections therefore require improved cloud process realism guided by process-oriented observations and observational constraints.

随着大气变暖，部分云群从冰和混合相（“冷”）转变为液态（“暖”）云。由于温暖的云层反射性更强且寿命更长，因此这种相变会减少地球吸收的太阳通量并构成负辐射反馈。这种冷却反馈在耦合模型比对项目 (CMIP6) 的第六阶段比第五阶段 (CMIP5) 更弱，从而导致更大的温室变暖。尽管这种变化通常归因于模拟云阶段的改进，但另一个模型偏差仍然存在：温暖的云太容易沉淀，可能导致低估的负寿命反馈。在这项研究中，我们修改了一个气候模型以更好地模拟暖雨概率，并发现它的云寿命反馈比默认模型大近三倍。这表明云降水过程中的模型误差可能会使云反馈产生与 CMIP5 到 CMIP6 气候敏感性差异一样多的偏差。因此，可靠的气候模型预测需要在面向过程的观测和观测约束的指导下改进云过程的真实性。