Fig. 3: CMIP-based predictions of past variability in tropical lower stratospheric water vapour.
From: Response of stratospheric water vapour to warming constrained by satellite observations
Black: monthly mean predictions of deseasonalized Δqstrat anomalies using the 27 CMIP-based functions provided with ERA5 reanalysis temperature data. We also show SWOOSH observational data for the same period (red). The blue line indicates average predictions conducted with the cross-validated functions learned from SWOOSH and ERA5, if ERA5 temperatures are used again as input. These predictions are highly correlated with SWOOSH (r2 score = 0.90; Pearson’s r = 0.96). The CMIP-based predictions also correlate well with SWOOSH but typically overestimate the amplitude of the undulations in line with their too-large sensitivities under climate change. Predictions with other reanalysis temperatures provide similar results (Extended Data Fig. 2).
