Modelled effect of changes in the CO2 concentration on the middle and upper atmosphere: Sensitivity to gravity wave parameterization

We have studied the temperature response to changes in the CO2 concentration in the middle and upper atmosphere using the Coupled Middle Atmosphere-Thermosphere Model 2 (CMAT2). We have performed simulations with a range of CO2 concentrations and three different ways of accounting for the effects of gravity waves, to allow for comparison with previous studies and sensitivity analyses. We initially find that the response of the model to the changes in CO2 concentration which took place between 1965 and 1995 (320-360 ppm) is strongly dependent on the gravity wave parameterization that is used, but this is to a large degree due to steps or kinks in an otherwise nearly linear curve describing the temperature as a function of CO2 concentration. We have not been able to identify the cause of these steps as part of the present study, which is a limitation and must be studied in future work. Here we treated the steps as model noise and rather focused on correcting for their effects by fitting straight lines to the temperature-CO2 curves to estimate the overall slope of the curves. From these slopes we were able to obtain more robust trend estimates than can be obtained by comparing only two model simulations, as is normally done in other, similar studies. The corrected temperature responses to a 40 ppm change in CO2 concentration still show up to 15-17% sensitivity to the gravity wave parameterization in the mesosphere and thermosphere. This remaining sensitivity is likely to be related to the fundamental differences in the way a change in temperature modifies the propagation and dissipation characteristics of gravity waves in each parameterization, which is particularly different for linear and non-linear schemes. The corrected trends we find are largely in agreement with other modelling studies, and therefore do not fully explain observed trends, which are typically larger than those predicted by modelling studies. However, modelling results could be similarly sensitive to other model parameters and settings, for example to gravity wave characteristics or solar activity level, and this should be further investigated as well. (C) 2008 Elsevier Ltd. All rights reserved.