Using Climate Model Simulations to Constrain Observations

We compare atmospheric temperature changes in satellite data and in older and newer multi-model and single-model ensembles performed under phases 5 and 6 of the Coupled Model Intercomparison Project (CMIP5 and CMIP6). In the lower stratosphere, multi-decadal stratospheric cooling during the period of strong ozone depletion is smaller in newer CMIP6 simulations than in CMIP5 or satellite data. In the troposphere, however, despite differences in the forcings and climate sensitivity of the CMIP5 and CMIP6 ensembles, their ensemble-average global warming over the satellite era is remarkably similar. We also examine four well-understood properties of tropical behavior governed by basic physical processes. The first three properties are ratios between trends in water vapor (WV) and trends in sea surface temperature (SST), the temperature of the lower troposphere (TLT), and the temperature of the mid- to upper troposphere (TMT). The fourth property is the ratio between TMT and SST trends. All four trend ratios are tightly constrained in CMIP simulations. Observed ratios diverge markedly when calculated with SST, TLT, and TMT trends produced by different groups. Observed data sets with larger warming of the tropical ocean surface and tropical troposphere yield atmospheric moistening that is closer to model results. For the TMT/SST ratio, model-data consistency depends on the selected combination of observed data sets used to estimate TMT and SST trends. If model expectations of these four covariance relationships are realistic, one interpretation of our findings is that they reflect a systematic low bias in satellite tropospheric temperature trends. Alternately, the observed atmospheric moistening signal may be overestimated. Given the large structural uncertainties in observed tropical TMT and SST trends, and because satellite WV data are available from one group only, it is difficult to determine which interpretation is more credible. Nevertheless, our analysis illustrates the diagnostic power of simultaneously considering multiple complementary variables and points towards possible problems with certain observed data sets.