Figure 1 shows the seasonal cycle of total ozone in the SH from both observations and from CMAM, averaged over 35 oS–60 oS, 60 oS–80 oS and 35 oS–80 oS for each year in the relevant data sets. The values are the anomalies from the long term trend added to the EESC fit for the year 2000. In this way, the absolute values of the two data sets are directly comparable even though the data sets cover different time periods. The seasonal cycle of total ozone shows a late spring maximum in midlatitudes and a late spring minimum in polar regions, the latter reflecting the springtime polar ozone depletion. In autumn CMAM midlatitude ozone levels exhibit a positive bias of about 10DU relative to observations. Furthermore, the seasonal cycle of total ozone for the entire extratropical region is slightly stronger in the observations. As noted in F&S 2003, the anomalies established in spring appear to persist through the summer. This is illustrated by Fig. 2, which shows the year to year variability of total ozone anomalies averaged over the SH extratropics for the four spring/early summer months for both observations and model. The interannual anomalies are highly correlated from month to month, and the anomalies established in October decrease continuously through January. Since the seasonal cycle of the CMAM ozone is somewhat too weak the overall absolute decrease in ozone values during the four months is smaller in CMAM than in the observations, but the magnitude of the year to year variability of each month is in reasonably good agreement between the two data sets.
Figure 3 shows the correlation coefficients between ozone values at a given month of the year with ozone values at subsequent months for the SH extratropics. Colours other than grey denote statistically significant correlations at the 95% confidence level assuming each year is independent. The correlations are extremely high (above 0.8) in summer (November until February) with any later month up to March, and statistically significant up to April in the observations and June in CMAM. Thus, it is evident that the anomalies persist from the end of spring until early autumn in both observations and the CMAM. This high predictability of total ozone reflects the fact that there is not a great deal of dynamical variability in the summer stratosphere, and so the time evolution of ozone (integrated over the extratropics) is controlled by photochemical relaxation.
The relationship between ozone anomalies in November and in subsequent months can be estimated by linear regression, and is shown in Fig. 4. The regression coefficients illustrate that the amplitude of the ozone anomalies decays on a timescale of a few months through photochemical ozone loss. Again the CMAM results are very similar to the observations, indicating that the overall summertime ozone photochemistry in the model is reasonable.
Since the diagnostic reflects transport and summertime photochemistry, it should not depend on how the springtime ozone anomaly is created. To verify this hypothesis we examine results from a different version of CMAM which includes no heterogeneous chemistry and thus has no ozone hole. Two 15-year simulations were used here to obtain a combined 30-year ozone data set. The two model runs were performed under fixed external forcings corresponding to conditions in 2000 with annually repeating climatological sea surface temperatures. Since there is no springtime polar ozone depletion, the seasonal cycle of ozone has a late spring maximum in polar regions and hence a strong seasonal cycle (in fact approaching that of the NH) in the entire SH extratropical region. However, the persistence of springtime ozone anomalies and the photochemical decay is very similar to that in the observations as well as in the other CMAM run, as shown in Fig. 4. This confirms that the diagnostic only reflects the transport and summertime photochemistry, and not the nature of the springtime ozone anomaly.
We now return to the 45-year transient run and examine the behaviour in the NH. Figure 5 shows the seasonal cycle of total ozone over 35 oN–60 oN, 60 oN–80 oN, and 35 oN– 80 oN for the observations and CMAM and as before, the values are the anomalies from the long term trend added to the EESC fit for the year 2000. The seasonal cycles are in good agreement, but CMAM has comparatively limited interannual variability in midlatitudes and thus over 35 oN–80 oN as a whole. This deficiency is evident in other diagnostics using an earlier version of CMAM (Austin et al., 2003). Figure 6 shows the correlation coefficients between ozone anomalies in different months for the NH extratropics. In the observations (Fig. 6a) the anomalies persist from the end of spring during the whole summer until early autumn. The CMAM anomalies (Fig. 6b), although small in magnitude, show the same characteristics. The CMAM correlation coefficients are very similar to the observations, except for March where the correlations are somewhat smaller in CMAM. The decay of the anomalies is shown in Fig. 7 and is virtually identical between CMAM and the observations.
Since the diagnostic reflects transport and summertime photochemistry, it should depend on transport characteristics like the vertical diffusion coefficient. To verify this hypothesis we examine results from an older version of CMAM which has a much stronger and therefore less realistic (WMO, 1999) vertical diffusion coefficient (1.0m2/s) compared to the 45-year transient CMAM run (0.1m2/s). The strong diffusion run was performed under fixed external forcings corresponding to conditions in 2000 with annually repeating climatological sea surface temperatures. Figure 6c shows the correlation coefficients between monthly ozone anomalies for the strong diffusion run. These coefficients are smaller than for the observations or the transient CMAM run, especially in spring and early summer. It is most likely that the rapid decay of the anomalies is due to the stronger vertical diffusion. Consistently, the regression coefficients for the strong diffusion run are smaller than in the observations or the transient CMAM run as illustrated by Fig. 7. This confirms that the diagnostic is sensitive to summertime vertical transport, and can detect an unrealistically strong vertical diffusion coefficient.