Snow Reconciles Observed and Simulated Phase Partitioning and Increases Cloud Feedback.

The surprising increase of Earth's climate sensitivity in the most recent coupled model intercomparison project (CMIP) models has been largely attributed to extratropical cloud feedback, which is thought to be driven by greater supercooled water in present‐day cloud phase partitioning (CPP). Here, we report that accounting for precipitation in the Goddard Institute for Space Studies ModelE3 radiation scheme, neglected in more than 60% of CMIP6 and 90% of CMIP5 models, systematically changes its apparent CPP and substantially increases its cloud feedback, consistent with results using CMIP models. Including precipitation in the comparison with cloud–aerosol lidar and infrared pathfinder satellite observations (CALIPSO) measurements and in model radiation schemes is essential to faithfully constrain cloud amount and phase partitioning, and simulate cloud feedbacks. Our findings suggest that making radiation schemes precipitation‐aware (missing in most CMIP6 models) should strengthen their positive cloud feedback and further increase their already high mean climate sensitivity. Plain Language Summary: The surprising increase of Earth's climate sensitivity–a proxy for future global warming–in the most recent climate models (CMIP6) has been largely attributed to the response of extratropical low clouds to warming. This cloud‐climate feedback is thought to be driven by greater supercooled water in present‐day cloud phase partitioning. Here we report that accounting for precipitation in climate model radiation schemes–neglected in more than 60% of CMIP6 and 90% of CMIP5 models–profoundly changes their apparent cloud phase partitioning and substantially increases their cloud‐climate feedbacks, which has not been reported before. Including precipitation in the comparison with observations and in model radiation schemes is essential to faithfully constrain cloud amount and phase partitioning and simulate cloud‐climate feedbacks. Our novel findings suggest that making radiation schemes precipitation‐aware, which is missing in most CMIP6 models, should strengthen their positive cloud feedback and further increase their already high mean climate sensitivity Key Points: Accounting for snow in a lidar simulator to compare with observations systematically reduces the simulated apparent supercooled liquidAllowing radiative schemes to be snow‐aware in climate models greatly increases their net cloud feedbackThe mean climate sensitivity is greater for recently coupled model intercomparison project models with snow‐aware radiative schemes compared to those without [ABSTRACT FROM AUTHOR]