Abstract

We perform a formal attribution study of upperand lower-stratospheric ozone changes using observations together with simulations from the Whole Atmosphere Community Climate Model. Historical model simulations were used to estimate the zonal-mean response patterns ("finger-prints") to combined for... Show moreWe perform a formal attribution study of upperand lower-stratospheric ozone changes using observations together with simulations from the Whole Atmosphere Community Climate Model. Historical model simulations were used to estimate the zonal-mean response patterns ("finger-prints") to combined forcing by ozone-depleting substances (ODSs) and well-mixed greenhouse gases (GHGs), as well as to the individual forcing by each factor. Trends in the similarity between the searched-for fingerprints and homogenized observations of stratospheric ozone were compared to trends in pattern similarity between the fingerprints and the internally and naturally generated variability inferred from long control runs. This yields estimated signal-to-noise (S/N) ratios for each of the three fingerprints (ODS, GHG, and ODS + GHG). In both the upper stratosphere (defined in this paper as 1 to 10 hPa) and lower stratosphere (40 to 100 hPa), the spatial fingerprints of the ODS + GHG and ODS-only patterns were consistently detectable not only during the era of maximum ozone depletion but also throughout the observational record (1984-2016). We also develop a fingerprint attribution method to account for forcings whose time evolutions are markedly nonlinear over the observational record. When the nonlinearity of the time evolution of the ODS and ODS + GHG signals is accounted for, we find that the S/N ratios obtained with the stratospheric ODS and ODS + GHG fingerprints are enhanced relative to standard linear trend analysis. Use of the nonlinear signal detection method also reduces the detection time - the estimate of the date at which ODS and GHG impacts on ozone can be formally identified. Furthermore, by explicitly considering nonlinear signal evolution, the complete observational record can be used in the S/N analysis, without applying piecewise linear regression and introducing arbitrary break points. The GHG-driven fingerprint of ozone changes was not statistically identifiable in either the upper-or lower-stratospheric SWOOSH data, irrespective of the signal detection method used. In the WACCM simulations of future climate change, the GHG signal is statistically identifiable between 2020 and 2030. Our findings demonstrate the importance of continued stratospheric ozone monitoring to improve estimates of the contributions of ODS and GHG forcing to global changes in stratospheric ozone. Show less