When the trend proportions between the variables were put into the simulation model, they were more in line with the proportions observed in reality when calculated with data sets that show greater tropical warming of both the ocean surface and the troposphere. . If the climate model expectations of these tropical temperature-humidity relationships are realistic, The findings would then be reflecting a systematic downward bias in the tropospheric temperature trends of the satellites, or an overestimation of the observed atmospheric humidity signal.
“It is currently difficult to determine which interpretation is more credible,” explains climate scientist Ben Santer, lead author of the paper. “But our analysis reveals that several observational data sets, particularly those with the smallest values for ocean surface warming and tropospheric warming, appear to disagree with other independently measured complementary variables.”
To reach these conclusions, the relationships between tropical temperature and tropical water vapor were considered. Water vapor trends were compared with trends in sea surface temperature, lower troposphere temperature, and upper middle troposphere temperature. The fourth property was the relationship between the trends ‘upper mean troposphere temperature’ and ‘sea surface temperature’. All four proportions are strictly restricted in climate model simulations, despite model differences in climate sensitivity, external forcings, and natural variability. In contrast, each ratio exhibits a wide range when calculated from observations. “This work shows that careful intercomparison of different geophysical fields can help us determine historical changes in climate with greater precision,” explains Stephen Po-Chedley, another of the authors.
Reference: Santer et al. 2021. Using Climate Model Simulations to Constrain Observations. Journal of Climate. https://doi.org/10.1175/JCLI-D-20-0768.1