A comparison of model-simulated trends in stratospheric temperatures

Estimates of annual-mean stratospheric temperature trends over the past twenty years, from a wide variety of models, are compared both with each other and with the observed cooling seen in trend analyses using radiosonde and satellite observations. The modelled temperature trends are driven by changes in ozone (either imposed from observations or calculated by the model), carbon dioxide and other relatively well-mixed greenhouse gases, and stratospheric water vapour. The comparison shows that whilst models generally simulate similar patterns in the vertical profile of annual-and global-mean temperature trends, there is a significant divergence in the size of the modelled trends, even when similar trace gas perturbations are imposed. Coupled-chemistry models are in as good agreement as models using imposed observed ozone trends, despite the extra degree of freedom that the coupled models possess. The modelled annual- and global-mean cooling of the upper stratosphere (near 1 hPa) is dominated by ozone and carbon dioxide changes, and is in reasonable agreement with observations. At about 5 hPa, the mean cooling from the models is systematically greater than that seen in the satellite data; however, for some models, depending on the size of the temperature trend due to stratospheric water vapour changes, the uncertainty estimates of the model and observations just overlap. Near 10 hPa there is good agreement with observations. In the lower stratosphere (20-70 hPa), ozone appears to be the dominant contributor to the observed cooling, although it does not, on its own, seem to explain the entire cooling. Annual- and zonal-mean temperature trends at 100 hPa and 50 hPa are also examined. At 100 hPa, the modelled cooling due to ozone depletion alone is in reasonable agreement with the observed cooling at all latitudes. At 50 hPa, however, the observed cooling at midlatitudes of the northern hemisphere significantly exceeds the modelled cooling due to ozone depletion alone. There is an indication of a similar effect in high northern latitudes, but the greater variability in both models and observations precludes a firm conclusion. The discrepancies between modelled and observed temperature trends in the lower stratosphere are reduced if the cooling effects of increased stratospheric water vapour concentration are included, and could be largely removed if certain assumptions were made regarding the size and distribution of the water vapour increase. However, given the uncertainties in the geographical extent of water vapour changes in the lower stratosphere, and the time period over which such changes have been sustained, other reasons for the discrepancy between modelled and observed temperature trends cannot be ruled out.