Readers may remember the fun we had with a couple of papers at the start of the year. Cai et al found that climate models that simulated extreme rainfall well predicted more frequent El Ninos. Meanwhile Sherwood et al found that climate models that simulated clouds well had high climate sensitivity, a position that I characterised as "the best cloud simulators are the worst temperature simulators". Much amusement ensued when it emerged that the intersection of "best cloud simulators" with "best rainfall simulators" was in fact the empty set.
Leo Hickman now points us to a new paper by Su et al, which examines some climate models and concludes:
New model performance metrics proposed in this work, which emphasize how models reproduce satellite observed spatial variations of zonal-mean cloud fraction and relative humidity associated with the Hadley Circulation, indicate that the models closer to the satellite observations tend to have equilibrium climate sensitivity higher than the multi-model-mean.
This is an admirable confirmation of Sherwood's findings. Like I said: the best cloud simulators are the worst temperature simulators.