Dynamical mechanisms for the recent ozone depletion in the Arctic stratosphere linked to North Pacific sea surface temperatures

The stratospheric ozone layer, which prevents solar ultraviolet radiation from reaching the surface and thereby protects life on earth, is expected to recover from past depletion during this century due to the impact of the Montreal Protocol. However, how the ozone column over the Arctic will evolve over the next few decades is still under debate. In this study, we found that the ozone level in the Arctic stratosphere at 100–150 hPa during 1998–2018 exhibits a decreasing trend of − 0.12 ± 0.07 ppmv decade–1 from MERRA2, suggesting a continued depletion during this century. About 30% of this ozone depletion is contributed by the second leading mode of sea surface temperature anomalies (SSTAs) over the North Pacific with one month leading and therefore is dynamical in origin. The North Pacific SSTAs associated with this mode tend to result in a weakened Aleutian low, a strengthened Western Pacific pattern and a weakened Pacific–North American pattern, which impede the upward propagation of wavenumber-1 waves into the lower stratosphere. The changes in the stratospheric wave activity may result in decreased ozone in the Arctic lower stratosphere through weakening the Brewer-Dobson circulation. Our findings uniquely linked the recent ozone depletion in the Arctic stratosphere to the North Pacific SSTs and might provide new understanding of how dynamical processes control Arctic stratospheric ozone.


Introduction
Stratospheric ozone, which comprises about 90% of the total amounts present in the Earth's atmosphere, is a radiatively and chemically active gas that shields the Earth from harmful solar ultraviolet radiation (WMO 2018). In the stratosphere, ozone changes can alter the temperature and its gradient via radiative effects (Ramaswamy 2001) and modify the circulation and wave activity via radiative-dynamical feedbacks (Hu and Tung 2003;Eyring et al. 2007;Hu et al. 2015). Some studies have shown that depletion of stratospheric ozone during the austral summer may result in the poleward shift of the mid-latitude jet (e.g., Thompson et al. 2011;Son et al. 2018), widening of the Hadley circulation (Sonet al. 2010), an increase in subtropical precipitation (Kang et al. 2011) and the poleward extension of the subtropical dry zones  in the Southern Hemisphere. Ozone depletion over the Arctic may also affect sea-level pressure (SLP), temperature, and precipitation in most parts of the Northern Hemisphere (NH) (Calvo et al. 2015;Ivy et al. 2017), even the sea surface temperature anomalies (SSTAs) over tropical Pacific SSTs including El Niño-Southern Oscillation (ENSO) (Xie et al. 2017), though Harari et al. (2019) suggested that the Arctic stratospheric 1 3 ozone may be not the proximate cause of the impacts on the surface over polar and tropical latitudes.
As the rapid increase in anthropogenic emissions of ozone depleting substances (ODSs) peaked in the mid-1990s (Weatherhead and Andersen 2006), the globally averaged column ozone showed a negative trend from the late 1970s to the late 1990s (WMO 2007). With the observed decrease in ODSs in the atmosphere from the 1990s under the impact of the Montreal Protocol and its amendments (Chipperfield et al. 2015), numerical studies indicated that ozone concentrations in the upper stratosphere will recover due to the decreased ODSs (WMO 2018). Chemistry-climate models predicted that ozone will recover to the levels of pre-1980 around 2050 (e.g., Weatherhead and Andersen 2006). Bednarz et al. (2016) further reported that the ozone in the NH may recover to 1980 levels by about 2030-2040. Results from Chemistry-Climate Model Initiative (CCMI) simulations project under a Representative Concentration Pathway (RCP) of 6.0 showed that the column ozone will return to 1980 values in 2032 (2020-2044) at mid-latitudes but in 2034 (2025-2043) at high-latitudes in the NH (Dhomse et al. 2018).
Datasets from National Aeronautics and Space Administration (NASA) and National Oceanic and Atmospheric Administration (NOAA) satellites show that ozone in the mid-and upper stratosphere increased slowly during 2000-2016 (Steinbrecht et al. 2017). However, some studies have suggested that there was no significant trend in the ozone levels in the lower stratosphere from 1984 to 2011 (Tummon et al. 2015) or from 1995 to 2013 (Cohen et al. 2018). Some other studies reported that the ozone concentrations derived from merged datasets in the lower stratosphere between 40° S and 40° N after 1997 (Bourassa et al. 2014) and between 60° S and 60° N after 1998 (e.g., Ball et al. 2018;Ball et al. 2020;Wargan et al. 2018) were still decreasing. Given the declining ODS concentrations, extensive research, vigorous debate and a number of papers tried to refine the results and propose potential mechanisms after the continuing decline of the lower stratospheric ozone in the twenty-first century was first found by Ball et al. (2018). While these above-mentioned studies focused on tropical and midlatitudinal ozone trends, the result on the ozone over the Arctic is still unclear. Note that there has been a significant chemical depletion of ozone during some Arctic cold stratospheric winters during the past two decades (Tilmes et al. 2004;Manney et al. 2015). For example, the magnitude of the reduction in ozone concentrations over the Arctic observed during the late winter and early spring in 2011 was comparable with that over the Antarctic (e.g., Manney et al. 2011;Hurwitz et al. 2011). The lowest observed ozone levels in the Arctic occurred in 2020, which covered an area about three times the size of Greenland (e.g., Witze 2020; Dameris et al. 2021;Inness et al. 2020;Lawrence et al. 2020;Manney et al. 2020;Wohltmann et al. 2020;Xia et al. 2021). The above mentioned numerical and observational results point to two elements: the apparent negative trends over the past two decades constitute a new and intriguing result and large variability is a confounding factor in trend estimation.
Stratospheric ozone is not only affected by chemical processes related to ODSs (Rex et al. 2004), but is also modulated by SSTs via dynamical processes (e.g., Hu et al. 2014). Some studies suggested that SSTs in the North Pacific have significant impacts on the stratospheric Arctic vortex (e.g., Hurwitz et al. 2012;Hu et al. 2018). Hu et al. (2018) reported that the warming over the central North Pacific may lead to a strengthening of the stratospheric vortex over the Arctic during the boreal winter. Other studies revealed that the Arctic vortex in the stratosphere is related to the concentrations of ozone there (e.g., Hu et al. 2015). Polar vortices in cold years would have increased polar stratospheric clouds (PSCs) occurrence, on the surface of which chlorine-activating heterogeneous reactions occur, further reducing the ozone (Solomon et al. 1994;Chipperfield and Jones 1999;Daniel et al. 1999). Strength of the polar vortex during boreal winter is partly controlled by wave driving (e.g., Newman et al. 2001;Hu et al. 2018). The stronger and more variable wave driving can affect the ozone concentrations by both ozone transport (i.e., dynamical resupply) and chemical depletion (e.g., Strahan et al. 2016), i.e., stronger (weaker) wave driving is closely associated with increased (decreased) ozone by dynamical resupply and increased (decreased) ozone by reducing (increasing) ozone loss. A question therefore arises about whether the ozone concentrations in the stratosphere over the Arctic are affected by the SSTs over the North Pacific and how can these SSTs affect stratospheric ozone.
To answer the above questions, the reanalysis, observational datasets and a chemical transport model (CTM) are used to investigate the trends in ozone concentrations over the Arctic in the lower stratosphere during 1998-2018 and a dynamical mechanism is provided. Our results show that the ozone has declined during this period, which can be ascribed to the second leading mode of the SSTAs over the North Pacific or the Victoria mode, the low-frequency variability in the SSTAs over the North Pacific that cannot be explained by the Pacific decadal oscillation alone (Bond et al. 2003;Ding et al. 2015). The SSTAs over the North Pacific associated with the Victoria mode influence stratospheric ozone through reducing the upward propagation of the wavenumber-1 wave in the extratropical stratosphere, weakening the Brewer-Dobson circulation (BDC). The recent depletion of ozone in the Arctic lower stratosphere and its links to the North Pacific SSTs suggest that some potential dynamical processes play a key role in the stratospheric ozone variations over the Arctic, not only the ODSs controlled by the Montreal Protocol and the associated chemical processes. It is worth clarifying that a trend of two decades could reflect decadal variability that is likely to reverse going forward.

Datasets
The monthly mean datasets of temperature, winds, geopotential height, SLP, and ozone during 1980-2018 from Modern-Era Retrospective Analysis for Research and Applications version 2 (MERRA2) (Gelaro et al. 2017) are used in this study. Because the observations assimilated in reanalysis over the course of many decades are highly nonhomogeneous, changes in input data can and do lead to significant discontinuities in all assimilated fields (e.g., Davis et al. 2017;Wargan et al. 2018). Following the method in Wargan et al. (2018), we have removed the discontinuities of MERRA2 reanalysis by step changes using the results from the reference run in TOMCAT/SLIMCAT (hereafter TOMCAT, described in Sect. 2.2) (Chipperfield 2006) as a transfer function standard. And the results suggested that the discontinuities of MERRA2 are not an issue in the regions of 65°-90° N and 100-150 hPa we focused on (figure not shown). Wargan et al. (2018) have demonstrated that the ozone record from MERRA2 can be homogenized allowing reliable trend calculations. We also used the monthly mean ozone datasets from European Centre for Medium-Range Weather Forecasts fifth generation atmospheric reanalyses (ERA5) (Hersbach et al. 2019), Global OZone Chemistry And Related trace gas Data records for the Stratosphere (GOZCARDS) (Froidevaux et al. 2015), partial column ozone field from Solar Backscattered Ultraviolet (SBUV) (Kramarova et al. 2013;Bhartia et al. 2013), Stratospheric Water and OzOne Satellite Homogenized (SWOOSH) (Davis et al. 2016), Microwave Limb Sounder (MLS) (Schwartz et al. 2021). The SST data from the Extended Reconstructed Sea Surface Temperature V5 (Huang et al. 2017) was used. The description of above data sources is listed in Table 1.

Model and simulations
We also used a three dimensional (3D) chemical transport model TOMCAT/SLIMCAT (hereafter TOMCAT) (Chipperfield 2006). This model has a good description of chemistry for both the troposphere and stratosphere, which include the heterogeneous reactions on sulfate aerosols and liquid/ solid polar stratospheric clouds (Chipperfield et al. 2018a) as well as chemistry reactions of the oxygen, nitrogen, hydrogen, chlorine and bromine families (Grooss et al. 2018). More details about TOMCAT can be found in Chipperfield et al. (2018a).
Two experiments have been designed. Both simulations were forced by the temperature and winds fields from European Centre for Medium-Range Weather Forecasts ERA-Interim reanalysis (Dee et al. 2011). The model has a horizontal resolution of 2.8° × 2.8° and the vertical levels from the surface up to ~ 60 km (Chipperfield 2006). The only difference between the reference run and sensitivity run (ODSfix) (Feng et al. 2021) is that the ODSs after year 1995 are fixed in the sensitivity run but are time-varying in the reference run. The ODSs in the experiments are obtained from WMO (2018).

Methods
As the BDC is a Lagrangian mean circulation, approximated by the residual mean meridional circulation of the transformed Eulerian-mean equations (Dunkerton 1978), the various processes that can influence the ozone could be separated into the ozone advection by the BDC or mean ozone transport, the large-scale eddy transport, and the chemical net production term (Garcia and Solomon 1983). The zonal mean ozone tracer continuity equation in the transformed Eulerian-mean formulation in spherical geometry following Garcia and Solomon (1983), is as follows:  Huang et al. (2017) where is the zonal mean ozone concentration, * and w * calculated as Eq.
(2) are the BDC's meridional and vertical velocities, respectively, defined by Andrews et al. (1987). S is the chemical net production of ozone. The variables v and w are the meridional and vertical winds, respectively, is the potential temperature, a is the Earth's radius, 0 is air density, t , and z are time, latitude, and height, respectively. The overbars represent the zonal mean and the primes denote the departure from the zonal mean. The first and second terms on the right-hand side of Eq. (1) represent the advection of ozone by the BDC and the mean ozone transport. The eddy flux M is defined in Eq. (3) by Garcia and Solomon (1983) and represents the ozone flux related to the eddies caused by the wave component. ∇ ⋅ M is the eddy flux ( M ) divergence. So the third term in Eq. (1) represents the ozone caused by the large-scale eddy transport. The fourth term S in Eq. (1) represents the chemical net production of ozone.
The linear trends and their statistical significance were estimated with the Sen median slope (Sen 1968) and the Mann-Kendall (Kendall 1975) method, respectively, since the nonparametric methods are less sensitive to outliers. In addition, the two-tailed Student's t test is used to test the statistical significance of the regression and correlation coefficients between two auto-correlated time series. The effective number of degrees of freedom N eff is expressed below as Pyper and Peterman (1998): where N is the sample size and XX (j) and YY (j) are the autocorrelations of the two sampled time series X and Y at time lag j , respectively.
3 Decreasing trend in the ozone over the Arctic in the lower stratosphere Ivy et al. (2017) suggested that changes in the stratospheric ozone in March is a useful indicator for the climate anomalies in the troposphere in the NH. Xie et al. (2017) further revealed that changes in the stratospheric ozone over the Arctic in March has strongest linkage between the SSTs over the North Pacific in April. These previous studies suggested the importance of the stratospheric ozone over the Arctic in March. Figure 1 displays the trends in the zonal mean ozone concentrations in March derived from MERRA2 reanalysis and reference simulation in TOMCAT during 1998-2018. Downward trends in the March zonal mean ozone mixing ratios are observed in the lower stratosphere over the Arctic during the period 1998-2018 in MERRA2 (Fig. 1a), ERA5 (Fig. 1b), and TOMCAT (Chipperfield 2006) (Fig. 1c), with the largest negative trends occurring in the subpolar regions 50°-70° N at 100-150 hPa. The negative trends during 1998-2018 from MERRA2 and ERA5 can also be observed during different periods with the start year shifted several points earlier or later (figure not shown). Therefore, the negative ozone trend in the lower stratosphere (100-150 hPa) over the Arctic is not much influenced by the samples in some unusual years and is persistent during 1998-2018. The time series of ozone averaged over 65°-90° N from 100 to 150 hPa (hereafter O 3_ALS ) during 1998-2018 ( Fig. 1d) also shows statistically significant negative trends of − 0.12 ± 0.07 ppmv decade -1 from MERRA2 and − 0.07 ± 0.06 ppmv decade -1 from TOMCAT, respectively. Also, the year-to-year variability of ozone in the lower stratosphere over the Arctic from MERRA2 and TOMCAT ( Fig. 1d) can be observed clearly and is highly consistent, with a correlation coefficient r equals to 0.87, significant above the 95% confidence level. Moreover, the levels of ozone from MERRA2 are highly correlated with those from SWOOSH (r = 0.91), GOZCARDS (r = 0.82), and SBUV (r = 0.92) with these three correlation coefficients all significant above the 95% confidence level. These results suggest that the downward trend of ozone in the stratosphere over the Arctic is reliable. Note that the pattern of the negative trend in ozone from MERRA2 during 1998-2018 excluding 2011 (figure not shown) is similar to that includes 2011 (Fig. 1a). Trends in O 3_ALS during 1998-2018 excluding 2011 is − 0.11 ± 0.07 ppmv decade -1 , which is slightly smaller than the results including 2011. This implies that the year 2011 makes a contribution to the negative trend in ozone during 1998-2018, but the downward trend in Arctic ozone in the stratosphere during this period is not dominated by the large 2011 Arctic winter/spring ozone loss event.
Note that the negative trends in the lower stratospheric ozone over the Arctic from MERRA2 and TOMCAT are also observed during the period 1980-1997, which is shown in Fig. 2. The statistically significant decreasing ozone trends at high-latitude before 1980-1997 indicate a depletion of Arctic stratospheric ozone, consistent with previous studies (WMO 2018). However, the negative ozone trends at high-latitude in the lower stratosphere during 1980-1997 are larger than those during 1998-2018 (Fig. 1), which is possibly because of the decreased ODSs during the latter period (WMO 2018). Previous studies revealed that the concentration of stratospheric ozone from 1979 to mid-1990s exhibits a significant decreasing trend, and it is expected to recover to the level of pre-1980 around the middle of this century under the impacts of Montreal Protocol and its Amendments (e.g., Weatherhead and Andersen 2006;WMO 2018). However, the observations and simulation ( Fig. 1) presented here all show a continued decreasing trend in the levels of ozone in the lower stratosphere over the Arctic after the 2000s, which suggests that the levels of ozone in this region have not started to recover as expected, but the downward trend after the 2000s is slightly smaller because of the deceasing ODS levels. This is consistent with the results in Garfinkel et al. (2015).
To further verify the role of ODSs played in the ozone trends after the 2000s, the sensitivity experiment in which ODSs are fixed after the 1995 has been designed. More details are provided in Sect. 2.2. Figure 3 gives the trends in the zonal mean ozone concentrations in March from reference run and ODSfix run in TOMCAT during 1998-2018. The trends in ozone concentration in the stratosphere over the Arctic in two simulations are both statistically significantly negative, with smaller negative trends in reference run ( Fig. 3a) but larger negative trends in ozone in ODSfix run (Fig. 3b). This smaller negative trend in ozone in the reference simulation compared to that in the sensitivity  (Fig. 3) not only confirms the role of decreased ODSs after mid 1990s, but also suggests that some other processes might also influence the trends in ozone over the Arctic in the stratosphere.

Connections between the Arctic ozone and North Pacific SSTAs
The factors that affect the stratospheric ozone concentrations include the ODSs through chemical reactions (e.g., Rex et al. 2004) and the SSTs via dynamical processes (e.g., García-Herrera et al. 2006;Manzini et al. 2006;Hu et al. 2014). Previous studies suggested the delayed impacts of tropical SSTs on the stratosphere (e.g., García-Herrera et al. 2006;Manzini et al. 2006) and a significant impact of the North Pacific SSTAs on the stratospheric polar vortex (e.g., Hurwitz et al. 2012;Hu et al. 2018). However, the connection between the lower stratospheric ozone over the Arctic and the SSTAs over the North Pacific is still unclear. Figure 4 shows the regressed SSTAs over the North Pacific in February based on the normalized O 3_ALS index during 1980-2018 in March. From Fig. 4a, the SSTAs exhibit a northeast-southwest-oriented dipole pattern, i.e., the band of positive anomalous values that extends from the coast of California across the Pacific to the western Bering Sea, and the band of negative anomalous values from the central North Pacific to the coast of Asia. This pattern resembles the spatial pattern of the second leading mode of the SSTAs over the North Pacific (Bond et al. 2003;Ding et al. 2015). Following Bond et al. (2003), we adopted the monthly North Pacific (100° E-100.5° W, 20.5°-65.5° N) SSTAs during 1980-2018 in February to perform an Empirical Orthogonal Function (EOF) reanalysis. Its second EOF mode (EOF2) (Fig. 4b) resembles the pattern of the second leading mode of the North Pacific SSTAs or Victoria mode (Bond et al. 2003;Ding et al. 2015), accounting for 18.7% of the total variance. As expected, the pattern of the regressed SSTAs over the North Pacific in February based on the normalized O 3_ALS index in March (Fig. 4a) is similar to that of the EOF2 of North Pacific SSTAs (Fig. 4b), appearing as a Victoria-like mode. This suggests that ozone levels over the Arctic in the lower stratosphere in March are possibly related to the SSTAs over the North Pacific associated with the Victoria mode in February. It would be interesting to understand the role of ENSO, because of its impacts on the stratospheric polar vortex (e.g., Sassi et al. 2004;Manzini et al. 2006;Garfinkel and Hartmann 2008;Xie et al. 2012Xie et al. , 2014Rao and Ren 2016  An in-phase relationship between the PC2 SST in February and O 3_ALS ×(-1) (here the negative O 3_ALS is used for purposes of visualization) (Fig. 4c) can clearly be seen, and the correlation coefficient between PC2 SST and O 3_ALS is -0.40 during 1980-2018 and −0.46 during 1998-2018, respectively, with both values significant at/above the 95% confidence level. Note that the correlation coefficient between these two indices is only − 0.27 during 1980-1997, which is insignificant at the 90% confidence level. Similar results can be seen in TOMCAT data (figure not shown). This implies that there is an out-of-phase linkage between the lower-stratospheric ozone over the Arctic and SSTAs associated with the Victoria mode, but that this out-of-phase relationship is much stronger during 1998-2018. The interannual correlation between lower-stratospheric ozone over the Arctic and North Pacific SSTAs suggests that the decreasing Arctic lower stratospheric ozone trends during 1998-2018 ( Fig. 1) are connected to the trends in the North Pacific SSTAs associated with the Victoria mode. The linear trend in PC2 SST during 1998-2018 in February is consistent with the trend in O 3_ALS ×(-1) during 1998-2018 in March (Fig. 4c).
To quantify the contributions from different factors to ozone trends in the Arctic lower stratosphere, a multiple linear regression (MLR) was considered as follows: where Ozone( , p, t) is the interannual variability of zonal mean ozone concentration in March. Here the interannual variability of one variable is obtained by subjecting it to a seven-year high-pass Lanczos filter (Duchon 1979). , p, and t represent the latitude, level, and time, respectively. Variables F 1 , F 2 , F 3 , F 4 , F 5 , F 6 , F 7 and F 8 denote the interannual variabilities of stratospheric aerosol depth at 550 nm (SAD; the SAD before 1990 is downloaded from https:// data. giss. nasa. gov/ model force/ strat aer/ tau. line_ 2012. 12. txt and after 1990 is from http:// dx. doi. org/ 10. 5065/ D6S18 0JM), NPSST (represented by the PC2 SST index), multivariate ENSO index (NINO) (https:// www. esrl. noaa. gov/ psd/ data/ corre lation/ nina34. data), quasi-biennial oscillation (QBO) index at 30 hPa (https:// www. esrl. noaa. gov/ psd/ data/ corre lation/ qbo. data), solar cycle (SC, represented by the 10.7 cm solar flux; https:// www. esrl. noaa. gov/ psd/ data/ corre lation/  , and the tripole-like SSTAs over the North Atlantic (NASST), respectively. Some studies revealed that the North Atlantic also plays a role in the stratospheric Arctic vortex (e.g., Garfinkel et al. 2015;Hu et al. 2019). Figure 5 displays the trends in the SSTAs over the North Atlantic during 1998-2018. The trends in the North Atlantic SSTAs during this period exhibit a tripole-like pattern, with significant positive anomalies in the subtropical western North Atlantic and negative values in the tropical and subpolar eastern North Atlantic, which resembles the SSTAs in previous studies (e.g., Rodwell et al. 1999;Sutton et al. 2000;Czaja and Marshall 2001;Peng et al. 2003). Therefore, a NASST index is defined as the SSTAs averaged over a southern box (35°-45° N, 50° W-70° W) minus that in a northern box (50°-60° N, 20° W-40° W) according to Fig. 5. The trends in the ozone concentrations contributed from different factors are estimated by the linear trends of i ( , p) ⋅ Trend F i ∕Trend ozone ( i = 1, 2, 3, 4, 5,6, 7,8 ). As we focused on the trends in the ozone averaged over 65°-90° N from 100 to 150 hPa during 1998-2018, the contributions of different factors to the lower stratospheric ozone over the Arctic are calculated by the coefficient i ( , p) averaged over 65°-90°N from 100 to 150 hPa multiplied by the trends in different factors over ozone during 1998-2018. Figure 6 shows the contributions of the various factors including solar cycle, QBO, ENSO, North Pacific SSTAs, CO 2 , sea ice, stratospheric aerosol, and North Atlantic SSTAs to the recent decreasing trend in ozone. A key point here is that the North Pacific SSTAs associated with Victoria mode is the largest contributor to the decreased ozone over the lower stratospheric Arctic after the 2000s, which contributes ~ 30% to the decreased trend in the lower stratospheric ozone over the Arctic.

Dynamic mechanisms
We will now provide evidence for a causal mechanism linking the SSTAs associated with the Victoria mode to the concentrations of lower-stratospheric ozone over the Arctic. The variability of the ozone in the upper stratosphere was shown to be dominated by chemical processes, while ozone in the lower stratosphere is strongly affected by dynamical processes (e.g., Douglass et al. 1985;Hartmann 1981;Wargan et al. 2018;Ball et al. 2020;Orbe et al. 2020). And the SSTAs over the North Pacific were suggested to have significant effects on the stratospheric Arctic vortex via dynamical processes (e.g., Hurwitz et al. 2012;Hu et al. 2018). Therefore, it is worthwhile to investigate the possible dynamical mechanisms affecting ozone concentrations in the lower stratosphere over the Arctic in response to the North Pacific SSTAs. Figure 7 gives the trends in the geopotential height and horizontal winds at 200 and 500 hPa in March during  1998-2018. The geopotential height at both 200 and 500 hPa exhibits statistically significant positive trends north of 35° N in the North Pacific, along with anticyclonic trends in the horizontal winds (Fig. 7a, b). The regressed anomalies in the geopotential height and horizontal winds at 200 and 500 hPa in March from MERRA2 based on the PC2 SST in February during 1980-2018 (Fig. 7c, d) are similar to the pattern of tropospheric circulation trends during 1998-2018 (Fig. 7a,  b), but the magnitudes of the anomalies in the geopotential height and horizontal winds related to the North Pacific SSTAs are smaller than those of the trends. In response to the second leading mode of North Pacific SSTAs, there are statistically significant positive anomalies in the geopotential height at 200 hPa occurring in the north of 35° N in the North Pacific, accompanied by anticyclonic anomalies in the horizontal winds (Fig. 7c). The PC2 SST -related geopotential height over the southwestern North Pacific exhibits negative anomalies accompanied with cyclonic horizontal wind anomalies. The pattern of geopotential height over the North Pacific is consistent with that at 500 hPa (Fig. 7d), also similar to that of SST (Fig. 4b), which indicates a weakened Aleutian low in response to PC2 SST . A previous study has revealed that the warming in the central North Pacific corresponds to a weakened Aleutian low (Hu et al. 2018), consistent with our result here. Tropospheric teleconnection patterns, such as the Western Pacific (WP) and Pacific-North American (PNA) patterns, can be characterized by a deep Aleutian low (Wallance and Gutzler 1981). The correlation coefficients between the PC2 SST and WP, PNA teleconnection patterns at 200 hPa following the definitions in Wallace and Gutzler (1981) are 0.43 and − 0.37, respectively, both above the 95% confidence level. This implies that in response to the positive Victoria mode, the WP teleconnection pattern strengthens but the PNA teleconnection pattern weakens.
The weakened Aleutian low, accompanied by the strengthened WP and weakened PNA patterns, may affect the wave activity in the stratosphere (Hu et al. 2018). Therefore, trends in the longitudinal and vertical structures of the wavenumber-1 and -2 components of geopotential height averaged over 45° N-75° N during 1998-2018 are shown in Fig. 8a, b. Trends in the zonal wavenumber-1 component are out-of-phase with its climatologies, i.e., the positive (negative) trends are co-located with the negative (positive) climatologies (Fig. 8a). Whereas the trends in the wavenumber-2 component are in-phase with its climatologies, exhibiting the positive trends co-located with the positive climatolgoies and negative trends co-located with the negative climatologies (Fig. 8b). This suggests that the wavenumber-1 wave intensity during 1998-2018 weakens but the wavenumber-2 wave intensity during this period strengthens, consistent with the results in Hu et al. (2019). Similar to the trend results, anomalies in the longitudinal and vertical structure of the wavenumber-1 and -2 components of geopotential height in response to PC2 SST (Fig. 8c, d) exhibit positive (negative) anomalies in the zonal wavenumber-1 component of geopotential height that co-locates with the negative (positive) climatologies (Fig. 8c), suggesting a weakened wavenumber-1 planetary wave in response to the North Pacific SSTAs. However, anomalies in the wavenumber-2 component of geopotential height are in-phase with its climatologies (Fig. 8d), implying a strengthened wavenumber-2 planetary wave in response to the positive Victoria mode phases.
The details of the weakened wavenumber-1 and strengthened wavenumber-2 wave intensity during 1998-2018 can be seen more clearly in Fig. 9a, b. Meanwhile, Fig. 9c, d give the regressed anomalies in the wavenumber-1 and -2 components of geopotential height at 200 hPa based on PC2 SST . The out-of-phase (in-phase) between the anomalies and climatologies in the wavenumber-1 (-2) components of geopotential height in response to the North Pacific SSTAs (Fig. 8) can clearly be seen in the maps at 200 hPa (Fig. 9c, d). Above results suggest that the weakened WP and strengthened PNA patterns in response to the positive Victoria mode phases are consistent with the weakened wavenumber-1 component in the wave activity over the upper troposphere and lower stratosphere, which plays a dominant role in the weakening of the stratospheric wave flux in response to the Victoria mode. But the strengthening of the wavenumber-2 components associated with the Victoria mode counteracts the weakening of wavenumber-1 to some extent.
The quasi-geostrophic Eliassen-Palm (EP) flux (Edmon et al. 1980) is chosen to diagnose the propagation of planetary waves. During 1998-2018, there are weakened trends in the wavenumber-1 wave propagation in the lower stratosphere ( Fig. 10a) but strengthened trends in the wavenumber-2 wave propagation (Fig. 10b), which are consistent with the weakened wavenumber-1 wave intensity and Fig. 8 a, b Trends (shading, m decade -1 ) in the longitudinal and vertical structure of the wavenumber-1 and -2 components of geopotential height averaged over 45° N-75° N in March during 199845° N-75° N in March during -2018. c, d Same as (a, b), but for the geopotential height anomalies regressed on PC2 SST in February during 1980-2018. The contours represent the climatological mean of wavenumber-1 (left panels) and -2 (right panels) components of geopotential height averaged over 45° N-75° N. The values over the stippled regions are statistically significant at the 90% confidence level strengthened wavenumber-2 wave intensity during this period (Fig. 8a, b). In response to PC2 SST , there are weakened upward planetary wavenumber-1 waves in the lower stratosphere over the Arctic region (Fig. 10c), with slightly strengthened meridional propagation at mid-latitude in the upper troposphere. However, the planetary wavenumber-2 waves in response to PC2 SST exhibit strengthened upward propagation in the lower stratosphere with weakened equatorward propagation at mid-latitude in the upper troposphere (Fig. 10d). The weakened wavenumber-1 upward propagation and strengthened wavenumber-2 upward propagation (Fig. 10) are in accord with the weakened wavenumber-1 component but strengthened wavenumber-2 component in the wave activity over the upper troposphere and lower stratosphere shown in Fig. 8. Note that the weakened upward planetary wavenumber-1 wave propagation is accompanied with positive zonal wind anomalies over the Arctic and negative anomalies at mid-latitudes. This indicates that the subtropical westerly jet weakens in response to the positive PC2 SST phases, which may not favor the planetary wave upward propagation according to the wave-mean flow interaction theory (Andrews et al. 1987).
As the BDC is closely related to the stratospheric planetary wave activity (Butchart 2014 (Fig. 11a), along with the negative trends in the meridional velocity of BDC at extratropics in the stratosphere (Fig. 11b). It seems that the extratropical downwelling in the NH after 1998 does not totally weaken, but with some regional characteristics, i.e., the BDC weakens over the Arctic but strengthens at subpolar regions, which need more investigation. Because we focused on the decreasing trends in the stratospheric ozone over the Arctic, the anomalies in the BDC velocities over the Arctic related to the North Pacific SSTAs are paid more Fig. 9 a, b Trends (shading, m decade -1 ) in the wavenumber-1 and -2 components of geopotential height at 200 hPa in March during 1998-2018. c, d Same as (a, b), but for the geopotential height anomalies regressed on PC2 SST in February during 1980-2018. The contours represent the climatological mean of wavenumber-1 (left panels) and -2 (right panels) components of geopotential height at 200 hPa. Dotted regions represent the values statistically significant at the 90% confidence level attention to. As expected, there are weakened anomalies in the downwelling velocity over the Arctic compared to its climatology in response to the warmed North Pacific SSTAs (Fig. 11c), which implies a weakened BDC to the North Pacific SSTAs.
Changes in the BDC could modulate concentrations of ozone in the stratosphere (e.g., Hu et al. 2014Hu et al. , 2015. Anomalies in the lower-stratospheric ozone over the Arctic caused by the BDC and eddy transport can be examined according to the Transformed Eulerian-Mean formulation of the zonal-mean ozone tracer continuity equation (Garcia and Solomon 1983) (more details in the Sect 2.3). Figure 12 further shows the trends in the March ozone produced by the BDC and eddy during 1998-2018 and the associated anomalies regressed on PC2 SST in February during 1980-2018. The ozone caused by changes in the meridional and vertical velocities of BDC during 1998-2018 exhibits positive trends at mid-latitudes, negative trends at high-latitudes ( Fig. 12a, b). However, the ozone trends caused by changes in the eddy during this period are different, i.e., positive trends at high-latitudes but negative trends at mid-latitudes (Fig. 12c). These results imply that the ozone trends over the Arctic in the stratosphere are related to both the BDC and the eddy transport. Looking at Figs. 1 and 12 together, it seems that the vertical transport from BDC might play the dominant role in the decreasing trend in the ozone concentration during this period. In response to the North Pacific SSTAs, there are positive ozone anomalies caused by changes in the meridional BDC velocity (Fig. 12d) and eddy transport (Fig. 12f) in the Arctic lower stratosphere, but statistically significant negative ozone anomalies caused by the vertical transport of BDC there (Fig. 12e). These imply that the lower-stratospheric ozone anomalies over the Arctic in response to PC2 SST are mainly caused by vertical   transport of the BDC, and not by the eddy transport. However, the eddy transports in response to PC2 SST can result in negative anomalies of lower-stratospheric ozone at midlatitudes. The weakened BDC downwelling velocity over the Arctic (Fig. 11) may result in negative anomalies in the lower-stratospheric ozone over the Arctic via weakening the ozone transport from the ozone-rich middle stratosphere to the ozone-poor lower stratosphere (Fig. 12).
Besides changes in the BDC and eddy transport, the temperatures in the Arctic stratosphere can be also controlled by the anomalous planetary wave activity associated with the North Pacific SSTs. Figure 13 shows the trends in the temperature and zonal winds in March during 1998-2018 and their anomalies regressed on PC2 SST in February during 1980. During 1998, the temperature over the Arctic exhibits negative trends (Fig. 13a) along with the positive trends in the zonal winds there (Fig. 13b), but most of these trends are insignificant. In response to the warmed North Pacific SSTAs, there are cooling anomalies in the lower-stratospheric temperature over the Arctic (Fig. 13c) and strengthened anomalies in the zonal winds (Fig. 13d). These anomalies are in accord with the decreased ozone anomalies. The stronger and more variable wave driving can affect the ozone concentrations by both ozone transport (dynamical resupply) and chemical depletion (e.g., Strahan et al. 2016), i.e., stronger (weaker) wave driving is closely associated with increased (decreased) ozone by dynamical resupply and increased (decreased) ozone by reducing (increasing) chemical loss. In addition to the ozone decrease caused by the weakened BDC in response to the Victoria mode (Figs. 11 and 12), the cooler Arctic stratosphere (Fig. 13) can increase polar stratospheric cloud occurrence, on whose surface chlorine-activating heterogeneous reactions occur, further reducing the ozone (Solomon et al. 1994;Chipperfield and Jones 1999;Daniel et al. 1999). If the temperatures are low enough and active chlorine is present during boreal spring, particularly following cold winters, such as 1997 and 2011 (Chipperfield et al. 2015), and 2020 (Rao and Garfinkel 2020), photochemical ozone loss may depress the temperature, which in turn enhances the chemical reactions and leads to more ozone loss.

Conclusions and discussion
Using meteorological reanalysis, several observational datasets and a chemical transport model, trends in the concentrations of lower-stratospheric ozone over the Arctic and its links to the SSTAs over the North Pacific are examined in this study. Our results show a decreasing trend in the concentrations of ozone in March of − 0.12 ± 0.07 ppmv decade -1 from MERRA2 and − 0.09 ± 0.07 ppmv decade -1 from TOMCAT after 1998, in the period following the turnaround in the atmospheric ODS levels.
Further analysis suggested that the North Pacific SSTAs associated with the second leading mode in February appear to have impacts on the lower-stratospheric ozone over the Arctic in March with a contribution of about 30%. Ozone concentrations decrease with the warm phases of Victoria mode-related North Pacific SSTAs, and increase with the North Pacific SSTAs associated with its cold phases. The decrease in ozone over the lower stratospheric Arctic during 1998-2018 is consistent with an increase in the PC2 of the North Pacific SSTAs. The Victoria-mode-related SSTAs tend to result in a weakened Aleutian low accompanied by a strengthening in the WP pattern and a weakening in the PNA pattern, which impede the upward propagation of wavenumber-1 waves into the subpolar lower stratosphere. In response to the Victoria mode, the BDC is weakened via weakening the wave propagation, which results in the negative anomalies in the lower-stratospheric ozone over the Arctic via weakening the ozone transport from the middle stratosphere of ozone-rich to the ozone-poor lower stratosphere. Besides these dynamical processes, the cooler and stronger Arctic stratosphere in response to the North Pacific SSTAs related to the Victoria mode may also affect the ozone concentrations through chemical depletion, which needs further investigation. It is also worth clarifying that a trend of two decades could reflect decadal variability that is likely to reverse going forward, rather than, say, an anthropogenically forced signal.
Some previous studies investigated the connections between the stratospheric Arctic vortex and North Pacific SSTs associated with Pacific decadal oscillation (PDO) (e.g., Hurwitz et al. 2012;Woo et al. 2015;Kren et al. 2016;Hu et al. 2018). These studies showed that the warming in the North Pacific associated with the positive PDO phases could result in a stronger stratospheric Arctic vortex. While some other studies investigated the potential linkage between the Victoria mode-related North Pacific SSTAs with stratosphere (e.g., Xie et al. 2017;Li et al. 2018). Li et al. (2018) revealed that the positive phases of PC2 of North Pacific SSTAs tend to result in more frequent stratospheric sudden warming (SSW) events and longer SSW duration than their negative phases. That is, the warming in the North Pacific SSTs associated with the second leading mode of North Pacific SSTAs might lead to less SSW events (Li et al. 2018), suggesting more strong stratospheric vortex events. From this view, it seems that the stratospheric polar vortex might be not sensitive to the pattern of SSTAs over the North Pacific, but to the warming somewhere over the North Pacific. But it does not mean that the warming anywhere in the North Pacific will lead to a stronger vortex, because the warming related to the PDO and Victoria mode both are not uniform all over the North Pacific Ocean. A recent study suggested that a warming over the central North Pacific could lead to a stronger stratospheric polar vortex (Hu et al. 2018). The central North Pacific is the overlapping region of the PDO and Victoria mode of North Pacific SSTAs, we infer that the stratospheric polar vortex might be more sensitive to the warming over the central North Pacific. The connections between the stratospheric ozone over the Arctic with the warming in the North Pacific over different regions are still unclear, which are worthy of further investigation.
Recall that the ozone trends in the tropics and NH midlatitudes, and the potential mechanism, are under wide debate (e.g., Ball et al. 2018Ball et al. , 2019Wargan et al. 2018;Chipperfield et al. 2018b;Orbe et al. 2020). Wargan et al. (2018) provided the evidence for a dynamical origin of the observed decreased trend in the ozone in the extratropical lower stratosphere, which corroborated the results of Ball et al. (2018). Chipperfield et al. (2018b) argued that these trends resulted from natural variability. That met with a response from Ball et al. (2019) who demonstrated robustness of the trends through 2018. Orbe et al. (2020) demonstrated that the trends in ozone in the lower stratosphere in the NH midlatitudes result from trends in the residual circulation. In this paper, we link the polar ozone in the stratosphere to the BDC. Furthermore, Ball et al. (2020) suggests changes in mixing as a mechanism underpinning these trends, consistent with Wargan et al (2018), and points to an apparent inability of free-running models to reproduce the observed the lowerstratospheric ozone behavior. The latter point is also elaborated on extensively by Dietmüller et al. (2021). This present work explored the trends in the stratospheric ozone over the Arctic and uniquely linked the recent ozone depletion in the stratosphere over the Arctic to the North Pacific SSTs, which might provide another important element to the debate.