Abstract

Estimates of the ozone layer future evolution must consider both climate dependencies and interannual variability. These considerations imply analyses of transient-scenario realisations with chemistry-climate models (CCM) under realistic boundary conditions. In this context, investigations of ozone variability usually involve multiple regression analysis (MRA), a statistically efficient albeit complicated tool. However, a careful use of advanced regression approaches may improve the variability assessment considerably. The present study addresses climate dependencies in ozone transport, and adopts an advanced regression approach to both quantify deterministic ozone variability and trace it back to the scenario boundary conditions; the investigations refer to transient output of the CCM E39/C. Recent observations show a cooling of the tropical lower stratosphere, and CCMs suggest a spatial coincidence of the cooling with a stronger upward advection of ozone-poor tropospheric air. This advection increase appears to result from a currently unexplained strengthening of the planetary-wave driven mean meridional transport, arguably relating to the anthropogenic climate signal. The present study explores the strengthening by comparing realisations of two different scenarios. Both share the same boundary conditions including concentrations of ozone-depleting substances (ODS), but differ in their climate forcing via sea surface temperatures (SSTs) and well-mixed greenhouse gas concentrations (GHG). In the summer hemisphere tropics, higher SSTs for the warmer scenario amplify deep convection and hence the convective excitation of internal planetary waves. These waves travel upward through easterly winds while dissipating, but still carry enough of the signal into the lower stratosphere to intensify the mean meridional transport. The transport change in turn strengthens the input rate into the tropical lower stratosphere of ozone-poor tropospheric air, ultimately weakening lower-stratospheric ozone concentrations via higher tropical SSTs. The ozone variability assessment relates to monthly-mean total columns from three independent realisations of a 60-year transient scenario with realistic boundary conditions. It focuses on three latitudinal bands: southern/northern midlatitudes (SH/NH) and tropics. Common ozone MRAs are linear and iterate to account for auto-regression-induced nonlinearity. The present MRA is nonlinear and the first to demonstrate the validity of such iterations with respect to the least-squares surface: it detects only a weak distortion of the surface associated with autocorrelation, at least for the ozone time series examined. Also, the present MRA is among the few to demonstrate sufficient compliance with the regression requirements, particularly with that of independent residuals. Additionally, the new approach of response confidence bands permits a correct attribution of individual anomalies to the scenario boundary conditions. As a consequence, the present MRA is the first to explain the year 1985 SH low-ozone event, here reproduced by E39/C. The MRA further captures, e.g., a similar anomaly for the year 1997, and verifies the total-ozone response to stratospheric-transport modulating boundary conditions: tropical-SST anomalies (ENSO) affect the tropics and NH, but not the SH; or, the quasi-biennial oscillation (QBO) causes a seasonally synchronised ozone response at SH and more weakly at NH, but not in the tropics. While these features have already been reported for E39/C data, the present study establishes a firm statistical framework and discusses the physical background. Other responses refer to the 11-year solar cycle (SSC), to sulfate aerosols, and to ODS concentrations. The present nonlinear regression approach provides ample potential for further development. For instance, nonlinear deterministic regression terms may examine the existence of interactions between the NH ENSO response with long-term changes in the probability for northern polar heterogenous ozone depletion. Last, accounting for moving-average regression parameters may improve the compliance with the inference requirements even further. In conclusion, the E39/C boundary conditions modulate the ozone layer as well as stratospheric mean meridional mass transport on long and short time scales. In this respect, the most important result is the universal significance of tropical SSTs controlling stratospheric transport by governing the deep-convective production of internal and, probably, external planetary waves. An important future research task is whether increasing tropical SSTs can cause ENSO-like changes in wintertime mid- and polar-latitude stratospheric planetary-wave activity; such changes could disturb the northern polar vortex against the effect of radiatively induced stabilisation by higher GHG concentrations. E39/C and other CCMS have certain weaknesses, one of which is an unrealistically consistent QBO-related modulation of the northern polar vortex. Keeping these weaknesses in mind, MRA may represent a helpful tool as it improves the statistical efficiency.