How Do Biases in the Simulation of Present-Day Clouds Affect Cloud Feedbacks?

Cloud radiative feedbacks are currently the largest source of spread in estimates of climate sensitivity. Here we examine how cloud feedbacks relate to biases in model simulations of present-day clouds relative to NASA Multiangle Imaging SpectroRadiometer (MISR) satellite observations. Specifically, we examine relationships between bias in cloud fraction, cloud-top-height, and cloud optical depth with tropical high cloud, midlatitude low cloud, global high-cloud altitude, and tropical low cloud feedbacks for CMIP5 and CMIP6 models that have produced MISR simulator output. We find that the strength of all four of these simulated cloud feedbacks have statistically significant relationships with biases in present-day simulations of clouds. We use these relationships in an emergent constraint analysis to narrow the spread in the strength of each feedback. This suggests that future expert assessments of cloud feedbacks (and climate sensitivity) should consider present day biases in clouds and the climate modeling community might consider undertaking simulations where models are systematically tuned to eliminate (or at least reduce) biases relative to satellite observations.