Improving Estimates of Future Antarctic Ozone Change by Process Oriented Multiple Diagnostic Ensemble Regression

Future changes in stratospheric ozone will respond to the decline in ozone-depleting substances and greenhouse gas increases. They are expected to influence exposures to ultraviolet radiation and tropospheric climate. Chemistry-climate models provide projections of stratospheric ozone but the spread of the projections introduces an uncertainty. For practical applications, multi-model ensemble projections are often presented as an ensemble mean projection (ensemble mean method). Here we address the question whether stratospheric ozone projections over Antarctica in spring can be improved by using an ensemble regression based on a dependence of the simulated ozone changes on simulation of some key processes relevant for stratospheric ozone. A stepwise algorithm is used to select those process-oriented diagnostics which explain a significant fraction of the spread in the projected stratospheric ozone changes among the models over Antarctica. The results of the method are then cross-validated, i.e. validated against another model rather than against observations. Cross-validation shows that the ensemble regression method has a higher precision than the ensemble mean method, suggesting an improvement in the estimate of future Antarctic ozone change.