The effect of interactive ozone chemistry on weak and strong stratospheric polar vortex events

Modeling and observational studies have reported effects of stratospheric ozone extremes on Northern Hemisphere spring climate. Recent work has further suggested that the coupling of ozone chemistry and dynamics amplifies the surface response to midwinter sudden stratospheric warmings (SSWs). Here, we study the importance of interactive ozone chemistry in representing the stratospheric polar vortex and Northern Hemisphere winter surface climate variability. We contrast two simulations from the interactive and specified chemistry (and thus ozone) versions of the Whole Atmosphere Community 5 Climate Model, designed to isolate the impact of interactive ozone on polar vortex variability. In particular, we analyze the response with and without interactive chemistry to midwinter SSWs, March SSWs, and strong polar vortex events (SPVs). With interactive chemistry, the stratospheric polar vortex is stronger, and more SPVs occur, but we find little effect on the frequency of midwinter SSWs. At the surface, interactive chemistry results in a pattern resembling a more negative North Atlantic Oscillation following midwinter SSWs, but with little impact on the surface signatures of late winter SSWs and SPVs. 10 These results suggest that including interactive ozone chemistry is important for representing North Atlantic and European winter climate variability.

and in March. We consider March events separately from December-February events due both to different shortwave heating behavior and to model bias in March SSW frequency.
To the best of our knowledge, there is no standard definition of an SPV. Different methods have been used in the literature (Limpasuvan et al., 2004;Tripathi et al., 2015;Scaife et al., 2016;Beerli and Grams, 2019). We here follow the definition used 95 in Scaife et al. (2016) and Smith et al. (2018), designed to be analagous to the Charlton and Polvani (2007a) SSW definition and to result in a similar number of events in reanalysis. We define an SPV as zonal mean zonal wind at 60 • N and 10 hPa reaching 48 m/s or higher (westerly) during November through March, with the central date being the first day of zonal mean zonal winds above 48 m/s. No later date can be a central date until the winds return below 48 m/s for at least 20 consecutive days. We focus on SPVs occurring in December-February, due to low event frequency in November and March.

100
The results we present here are based on composites of daily model output for climate variables, with composites centered around SSW or SPV central dates.
For composites from either CHEM or NOCHEM simulations, we calculate significance using a Monte Carlo test based on 5000 randomly chosen central dates. We also consider the difference in CHEM or NOCHEM composites, denoted CHEM-NOCHEM; for these, we calculate significance from a two-sided two-sample t-test.

105
3 Impact of interactive chemistry

Stratospheric mean state and extreme events
We first consider the effect of interactive chemistry on the mean state of the stratosphere by examining the climatological Northern Hemisphere 10 hPa zonal mean zonal wind ( Figure 1). We find stronger westerlies in CHEM than in NOCHEM in the vortex formation stage (September and early October) and in the latter half of winter (January-April), between 60 • − 80 • 110 N. This relative strength in CHEM in late winter also corresponds to a delayed final warming by 7 days on average (not shown). These results are consistent with those found by Haase and Matthes (2019). The same is not the case in Smith et al.  (2019)).

115
Because we identify extreme stratospheric events using zonal mean zonal winds at 10 hPa and 60 • N (U1060) (Charlton and Polvani, 2007a; Butler and Gerber, 2018), we next examine the mean state and variability of this quantity in CHEM and NOCHEM. Figure 2 shows the two distributions of U1060 in December through March. The average difference in DJFM between CHEM and NOCHEM is about 1.7 m/s, a small but statistically significant difference; Figure 1 indicates that this difference is larger in January through March. The CHEM distribution also has a longer right tail. This is consistent with the 120 polar vortex being stronger overall with interactive chemistry. It also indicates that we should expect more SPVs in CHEM than in NOCHEM. While there are fewer days of weak westerlies (0-20 m/s) in CHEM than in NOCHEM, the numbers of days of easterlies are similar, so we expect less of a difference in SSW frequency between the two simulations.
Indeed, this is what we find when we calculate the frequencies of weak and strong vortex events in the CHEM and NOCHEM simulations (Table 1). We consider December-February (DJF, midwinter) and March (late winter) separately for two reasons. The stronger vortex in midwinter in the CHEM simulation might lead us to expect fewer DJF SSWs in CHEM than in 130 NOCHEM. We do see a decrease of about 10% in DJF SSWs with interactive chemistry compared to specified chemistry, but this decrease is far from being statistically significant. In contrast, in March, we see more SSWs in CHEM than in NOCHEM, potentially related to the later breakdown of the vortex. (2019) consider the overall (November-March) number of SSWs. They report a decrease in overall SSWs of around 30%. In contrast, for November-March, we find a slight increase in SSWs of about 2% (from 109 events to 111, not 135 shown) from prescribed chemistry to interactive chemistry.

Haase and Matthes
We now consider SPV frequency. The increase in DJF SPV frequency from NOCHEM to CHEM is about 29%. This is unsurprising given the stronger vortex in CHEM overall. With our definition of SPVs, the number of March strong vortex events (in either simulation) is too small for a robust statistical analysis. This is because of the weaker vortex in March compared to DJF; a much larger anomalous vortex strength would be necessary to reach 48 m/s. Because of the low number of such events, 140 we do not further study March SPVs and thus discard them from the analysis.
We now examine DJF SSWs, March SSWs, and DJF SPVs separately in each of the following three sections.

Midwinter sudden stratospheric warmings
We start by focusing on the surface impacts of SSWs, seeking to document any differences between the CHEM and NOCHEM simulations. After noting the impact of the events on the surface, we then consider how any differences in those impacts arise 145 aloft. Figure 3 shows composite surface level pressure anomalies in the first and second months (top and bottom respectively) following December-February SSWs in CHEM (left, 75 events) and NOCHEM (middle, 67 events), as well as the difference between the two (right). We see a strong and significant pattern resembling a negative North Atlantic Oscillation (NAO) in the first month following SSWs in both CHEM and NOCHEM, and in both cases this negative annular mode persists through 150 the second month following the event. There is minimal difference between the two simulations in the first 30 days, with the CHEM simulation having only a slightly stronger signal. However, the difference is statistically significant and strongly projects onto the NAO 30-60 days after the central date. This indicates that the surface signature of SSWs is stronger and more persistent in CHEM than in NOCHEM.
To determine whether these anomalies originate in the stratosphere, we calculate the Northern Annular Mode (NAM) for 155 CHEM and NOCHEM. We use a method similar to that of Gerber et al. (2010) and Gerber and Martineau (2018); the detailed procedure is in Appendix A. We show the results of the NAM calculations in Figure 4. The CHEM and NOCHEM composites 1. We average model output to find a time series of daily, zonal mean geopotential height Z(t, λ, p) as a function of time t, latitude λ, and pressure p.
2. For every day and pressure level, we remove the global mean geopotential heightZ global (t, p). This helps to remove the global changes so that the index instead mainly captures meridional differences or shifts (Gerber et al. 3. For each day, latitude, and pressure level, we remove the average for that calendar day over the whole period; that is, we remove the climatology to find an anomalous height. 4. For each day, latitude, and pressure, we remove the linear trend over the period. 5. For each day and pressure level, we compute a polar cap average. Here we are interested in the NAM, and we take the 295 average from 65-90 • N. This is a proxy for the annular mode as shown in Baldwin and Thompson (2009). 6. We multiply by -1 so that a positive polar cap geopotential height anomaly yields a negative NAM, for consistency with the convention of Thompson and Wallace (1998). 7. We normalize the index by its standard deviation at each pressure level.