THE DEPENDENCE OF THE CLIMATE SENSITIVITY ON CONVECTIVE PARAMETRIZATION - STATISTICAL EVALUATION

Two sensitivity experiments, in which CO2 is instantaneously doubled, have been performed with a general circulation model to determine the influence of the convective parametrization on simulated climate chang e. We have examined the spatial structure of changes in the annual mea n and annual cycle for surface temperature and precipitation for both experiments; similarly we have examined changes in the variance for th ese two fields. We have also computed a range of test statistics in or der to obtain reliable measures of the signal-to-noise ratio in the cl imate change signal from each experiment. We have computed test statis tics for the entire globe and for five different region and we contras t the global response with the response in the Australian region taken as a representative sample. We find that the highest signal-to-noise ratios in the change from 1CO2 to 2*CO2 are for the change in surface temperature for both experiments with little difference in the global averages between the experiments. Globally averaged precipitation sho ws a greater noise level but perhaps the greatest contrast between exp eriments. There are generally significant increases in the temporal an d spatial variability of precipitation in the change from the 1CO2 to 2CO2 and with some differences apparent between the two experiments. The temporal variability of surface temperature does not change signi ficantly in any of the 2CO2 cases, and there is little difference bet ween the experiments. There is a significant decrease in the spatial v ariability of surface temperature in all 2CO2 experiments in all case s and with significant differences in the seasonal variations between different experiments. The spatial variability of precipitation increa ses in all 2CO2 cases and also with substantial differences in the se asonal variations between the experiments. There are accompanying sign ificantly different spatial pattern correlations for both surface temp erature and precipitation. In general we find that the global changes are fairly robust with the differences associated with convective para metrization schemes being very small. However, at the regional level, there are marked differences between experiments with changes both in the means and in the spatial and temporal variances but often with low levels of significance.