Lapse rate deviations from the moist adiabat in the tropical upper troposphere in climate models

The tropospheric lapse rate in the tropics follows a moist adiabat quite closely and is mainly set by surface temperature and humidity in the convecting regions. Therefore, warming or biases at the surface are transferred via the moist adiabat to the upper troposphere. However, climate models show large discrepancies in the upper troposphere and recent observed upper tropospheric warming is around 0.5K weaker than predicted by the moist adiabat theory. Here we use the control simulations of the CMIP5 ensemble to show that large differences in the upper troposphere exist in the mean state that are unrelated to inter-model differences in the lower troposphere. In fact, CMIP5 models diverge (positively and negatively) from the moist pseudoadiabat by up to 2K at 300hPa. Precipitation weighted SSTs have recently been used to resolve the discrepancy between models and observations in upper tropospheric warming, but we show that they are not able to explain the differences in the mean state. While it is difficult to exactly depict the reasons for the inter-model spread, we demonstrate how the upper tropospheric lapse rate can deviate from the moist adiabat for the same lower tropospheric state with AMIP experiments. For this we use the ICON-A model, in which we tune convective and microphysical parameters. An improved understanding of the effect of different parameterisations on the models' lapse rates may help to better understand differences in the response to global warming.