Increased frequency and intensification of convective downdrafts with warming in km-scale simulations

Cloud responses to warming represent a substantial source of uncertainty in future climate projections, in part due to uncertain convective parameterizations in the global climate models (GCMs) used to estimate cloud feedbacks. A new generation of convection-permitting simulations at km-scale horizontal resolution offers new insight into clouds in a changing climate: Recent work has demonstrated a reduction in the thick ice clouds associated with convective cores relative to the optically thin anvil in idealized simulations. Here, we use multi-year simulations of the western Pacific in present-day and warmer climates to demonstrate fundamental differences in cloud-circulation interactions between convection-permitting models and GCMs. In particular, we find that strong descent in the mid-troposphere in our simulations is associated with both downdrafts in convective cores and radiatively-driven subsidence, in contrast to GCMs, which only represent the latter. We separate these influences on descending air using column relative humidity and demonstrate that the presence of downdrafts reduces the cooling effect of deep convective clouds. We show that our km-scale simulations exhibit the previously reported thinning of deep convective clouds with warming. Additionally, we show that this thinning is driven by a reduction in thick clouds associated with ascent, while that associated with descent – downdrafts – expands, and mean descent rates increase. We hypothesize that this increase in descent speeds is driven by increases in ascent, thus increasing the condensate loading and negative buoyancy in downdrafts, indicating a potential role for microphysical-dynamical pathways in cloud feedbacks largely absent in GCMs.