OMI TCNO2 was sampled under the wintertime (November–February) NAO-high (Fig. a) and NAO-low (Fig. b) for 2005–2015. Peak TCNO2 concentrations (over 15 × 1015 molecules cm-2) in both phases are over the Po Valley and the Benelux region. Over the UK, source-region TCNO2 ranges between 7–13 and 6–10 × 1015 molecules cm-2 in NAO-low and NAO-high. We hypothesise that NAO-high-enhanced westerly flow transports NO2 off the UK mainland, as seen by , who investigated the impacts of cyclonic conditions on UK TCNO2. Figure c supports this, highlighting significant negative anomalies of -4 to -2 × 1015 molecules cm-2 between the TCNO2 NAO-high composite and 11-year wintertime average. Significant anomalies, shown in the green polygon-outlined regions, are based on the WRT at the 95 % confidence level and where composite and wintertime averages ± their respective random errors do not overlap. Systematic errors will cancel when differencing the two TCNO2 composites. NAO-low reduces westerly flow across Europe and might be expected to aid TCNO2 accumulation, but there is actually little change in the anomaly field (Fig. d). Only the Benelux region (3–5 × 1015 molecules cm-2) and North Sea (-2.0 to -1.0 × 1015 molecules cm-2) show significant anomalies linked to NO2 accumulation and reduced transport off the UK mainland.

The MIPAS PAN 200–100 hPa average volume mixing ratio, sampled under NAO-high (NDJF) between 2002 and 2012 (Fig. a), shows peak (minimum) PAN concentrations of 50–55 (10–20) pptv in the subtropical North Atlantic (over Newfoundland and the Canadian Arctic). During NAO-low (Fig. b), PAN concentrations are lower over the subtropical Atlantic but slightly larger over Newfoundland and the Canadian Arctic between 25 and 40 pptv. PAN concentrations are also larger (approximately 40 pptv) over Iceland, southern Greenland and the Denmark Strait, leaving a spatially prominent feature. MIPAS-derived tropopause height (see Supplement) peaks at 11 km in this region, while it is only 9–10 km in the surrounding area (excluding the subtropical North Atlantic). There is also an increase in pressure and convergence of winds over this region (Fig. d), potentially highlighting the impact of NAO-low vertical transport of PAN into the UTLS; this is investigated further using TOMCAT (see Sect. ). In Fig. c under NAO-high, peak significant anomalies of 5–15 pptv occur over the subtropical Atlantic and north-eastern Arctic region (top right of the domain). There are also significant negative anomalies (-15 to -5 pptv) over the Québec region. Significant anomalies are based on the WRT (95 % confidence level) and where the NAO composite and the wintertime averages plus or minus their uncertainty ranges do not overlap. Over Iceland and Greenland (subtropical North Atlantic and Europe), there are positive (negative) anomalies of 5–15 (-5 to -1) pptv in NAO-low.

Figure  shows the zonally averaged (90–20∘ E) vertical profiles of MIPAS PAN under the two NAO phases. Peak PAN concentrations at 300–250 hPa range between 100 and 130 pptv in both phases but are larger in NAO-low between 30 and 50∘ N by 10–20 pptv. However, northwards of 70∘ N, PAN concentrations between 225 and 100 hPa tend to be larger under NAO-high conditions. Figure c shows the NAO-high zonal anomalies relative to the wintertime average (hatched anomalies are insignificant based on the WRT – 95 % confidence level). Northwards of 50∘ N, significant positive anomalies (5–15 pptv) exist between 300 and 125 hPa. MIPAS-derived tropopause height under NAO-high (see Supplement) is typically higher than in NAO-low over the North Atlantic and Europe by 1–2 km. The higher tropopause signifies enhanced vertical transport, which in this case is the propagation of polluted air masses (i.e. large PAN content) from further down in the troposphere into the UTLS. Southwards of 50∘ N, positive anomalies occur between 200 and 100 hPa, while negative anomalies are found between 300 and 250 hPa. Under NAO-low conditions (Fig. d), there are significant positive anomalies (5–15 pptv) at 300–275 hPa between 30 and 90∘ N, and they reach up to 125 hPa at 60–80∘ N coinciding with peak PAN concentrations over Iceland in Fig. b. Significant negative anomalies (-5 to -1 pptv) exist at 225–50 hPa, which coincides with the negative anomalies in Fig. d over the subtropical Atlantic. Between 30 and 50∘ N, there is an altitude anomaly dipole reversal with NAO-high showing significant positive (negative) anomalies at 200–100 hPa (300–250 hPa) and NAO-low highlighting significant positive (negative) anomalies at 300–275 hPa (225–50 hPa). These patterns are linked to changes to regional circulation patterns under the different NAO phases and will be explored further using TOMCAT in Sect. .

Figure  shows TES vertical profiles averaged over four regions (Zones 1–4) covering the North Atlantic between 2005 and 2011, which were sampled during significant wintertime (November–February) NAO events. These four domains are selected because TES has infrequent spatial sampling meaning spatial ozone distributions are often noisy or unclear. In Zone 1 (UK), TES ozone sampled under NAO-high (red line) is significantly larger (90 % – squares and 95 % – diamonds) than in NAO-low (blue line) by 3–4 ppbv throughout the region of peak sensitivity (900–650 hPa – yellow box). Similar patterns exist in surface ozone measurements from the UK Automatic Urban and Rural Network (AURN, ). Under NAO-high, surface ozone concentrations were significantly higher than in NAO-low by 5–10 µg m-3 across the UK (see Supplement). The opposite is true for AURN surface NO2 where concentrations across the UK are lower by 5–10 µg m-3 in NAO-high. This supports the hypothesis that NAO-high increases (decreases) ozone concentrations over western Europe (western Atlantic) through enhanced westerly transport of ozone and dispersion of other species involved in its removal (e.g. NO) over Europe. Zone 2 (Newfoundland) has the opposite signal whereby the NAO-low ozone profile is significantly (95 %) larger than the NAO-high profile by 2–4 ppbv. Again, this is potentially linked to enhanced westerly ozone transport across the Atlantic towards Europe during NAO-high. In Zone 3 (North Atlantic), there are insignificant differences at approximately 900 hPa. However, ozone is significantly greater under NAO-low between 875 and 350 hPa. There are insignificant differences in Zone 4 ozone up to 600 hPa, but NAO-high ozone is significantly larger above this altitude.

In NAO-high and NAO-low (Fig. a and b), where TOMCAT has been sampled under the NAO phases in Fig. a, the model TCNO2 over western Europe ranges between 3 and 9 × 1015 molecules cm-2 and 6 to over 10 × 1015 molecules cm-2. Over the UK, NAO-high-enhanced westerly flow transports NO2 off the mainland (Fig. c) with significant negative anomalies of -2.0 to -0.5 × 1015 molecules cm-2 relative to the wintertime average. OMI TCNO2 has a similar NAO-high UK signal (c), but it is less spatially extensive and does not cover as much of continental Europe. In NAO-low, OMI (Fig. d) only shows accumulation of TCNO2 over the Benelux region, while TOMCAT (positive anomalies over 1.5 × 1015 molecules cm-2) accumulates TCNO2 over all of continental Europe (Fig. d). Potential reasons for model-satellite NAO-low anomaly differences (Fig. d and d) included the following. (1) As OMI has peak retrieval sensitivity in the middle–upper troposphere it potentially underestimates the full TCNO2 under NAO-low conditions when the more stable conditions trap NO2 in the boundary layer. (2) The model NO2 lifetime in the NAO-low composite is potentially longer than the satellite equivalent as it represents all-sky conditions, while the satellite composite represents clear-sky conditions only (i.e. more photochemical loss of NO2).

At the surface, TOMCAT surface NO2 ranges between 0–6 and 2–8 ppbv in NAO-high and NAO-low. TOMCAT does have a systematic surface NO2 low bias against surface observations (see Supplement), but this systematic offset is removed when considering anomalies (Fig. c and d) relative to the wintertime average. TOMCAT surface anomalies typically have similar spatial patterns to the TOMCAT TCNO2, but they are less spatially extensive. Under the NAO-high, there are significant negative (positive) anomalies of -0.5 (0.2) ppbv over the UK (North Sea), highlighting the westerly transport of NO2 off the UK mainland. Under NAO-low, significant positive anomalies (0.0 to 1.0 ppbv) highlight the accumulation of NO2 from reduced westerly flow across the UK. This is consistent with the AURN results presented in the Supplement. The model also shows a significant anomaly dipole over Scandinavia which reverses between phases. This, in combination with the reduced spatial impact on surface NO2 compared with the tropospheric pattern, implies that processes above the surface also influence the response of the tropospheric NO2 distribution to the NAO.

Figure  shows the TOMCAT NO2 meridional vertical cross section at 0∘ E. Between 35 and 60∘ N TOMCAT simulates NO2 concentrations above 1.0 ppbv from 1000 hPa to 900 (850) hPa in NAO-high (NAO-low). Negative anomalies (under -0.05 ppbv) relative to the wintertime average from 1000 to 900 hPa at 50∘ N (Fig. c), show the enhanced NAO-high westerly flow transporting NO2 throughout the boundary layer away from UK source regions. This NO2 is transported into the North Sea, yielding positive anomalies of 0.02 ppbv northwards of 55∘ N with vertical ascent into the mid-troposphere (approximately 600–700 hPa at 60–70∘ N). Under NAO-low (Fig. d), there are positive (above 0.05 pptv) anomalies between 35 and 65∘ N as the weakened meridional winds have a southerly flow with ascent (descent) at 65 (40)∘ N. This highlights reduced NO2 transport from the climatological westerlies aiding accumulation in the lower troposphere (1000–700 hPa). Therefore, processes throughout the lower troposphere over the UK are important in governing the tropospheric column burden during the two NAO phases.

At the surface, although PAN has lower concentrations than NO2 in source regions, it has a longer lifetime resulting in more significant responses to the seasonal average under the different NAO phases. Under NAO-high (Fig. a), TOMCAT surface PAN peaks between 200 and 220 pptv over the western Atlantic. Over Europe, PAN ranges between 150 and 170 pptv as, like NO2, enhanced westerly flow transports PAN away from western European source regions, replacing it with cleaner subtropical North Atlantic air (100–150 pptv). Through reduced transport, NAO-low conditions aid pollutant accumulation over continental Europe with PAN concentrations of 190 to over 300 pptv. The NAO-high TOMCAT PAN anomalies (Fig. c) relative to the wintertime average highlight reduced concentrations of -50 to -20 pptv over continental Europe, while in the western North Atlantic there are no significant anomalies. This infers similar transport processes to the wintertime average, resulting in minimal PAN changes, yet NAO-low (Fig. d) weakens or reverses the winds, yielding significant negative anomalies of -20 to -10 pptv. Therefore, westerly flow, similar under NAO-high and average wintertime conditions, aids the long-range transport of PAN from North America. As NAO-low interrupts this transport pathway, there is a significant decrease in background PAN.

We now investigate whether TOMCAT reproduces the MIPAS UTLS PAN patterns under the NAO phases, despite the slightly different time periods. Previous studies (e.g. ) have shown that TOMCAT PAN compares reasonably well with aircraft observations, but there is a systematic difference between TOMCAT and KIT MIPAS PAN (see Supplement). Therefore, we primarily focus on the anomalies relative to the wintertime average under the NAO phases, as this systemic difference is removed. TOMCAT PAN 200–100 hPa average peaks (over 50 pptv) in NAO-high over the western subtropical North Atlantic (Fig. a). The south-westerly flow (approximately 30 m s-1) at this altitude transports PAN across the Atlantic, reaching 35 pptv over Iberia. However, at approximately 0∘ E, a southerly shift in the winds over the Mediterranean leads to lower continental Europe PAN concentrations (20–30 pptv). Northwards of 70∘ N, the flow (20–30 m s-1 ) accumulates PAN over the Arctic region (20–24 pptv). Similar spatial patterns are seen in MIPAS with peak PAN concentrations over the western subtropical Atlantic, minimum PAN over Canada and Hudson Bay and elevated PAN in the Arctic region. However, MIPAS PAN absolute concentrations are systematically higher than TOMCAT (see Supplement). Vertical transport will also have an important impact on NAO-high, as signified by the higher MIPAS-derived tropopause height (see Supplement), with propagation of polluted air masses from the lower troposphere into the UTLS. In NAO-low, peak PAN (over 40 ppbv) occurs in the subtropical Atlantic where the winds are predominately zonal (westerly), yielding lower PAN concentrations (15–25 pptv) over the mid-North Atlantic. Continental Europe PAN concentrations decrease (10–20 pptv), as north-westerly flow transports cleaner Arctic air masses into the region. PAN accumulation over Iceland and southern Greenland (25 ppbv) correlates with the large UTLS pressure increase shown in Fig. d. Figure d highlights the significant enhancement of PAN over Iceland and southern Greenland with positive anomalies relative to the wintertime average of 5–10 pptv. Again, the MIPAS-derived tropopause height in NAO-low peaks in this region (approximately 11 km – see Supplement) suggesting sufficiently strong vertical transport of tropospheric air masses. In NAO-low, as the strong westerly flow in NAO-high (Fig. b) has shifted equatorwards, there are significant negative anomalies under -15 pptv across the North Atlantic, which match the MIPAS equivalent in Fig. d. There are some similarities between the TOMCAT (Fig. c) and MIPAS (Fig. c) NAO-high PAN anomalies with increased PAN in the eastern Arctic. However, TOMCAT simulates positive anomalies (0–5 pptv) over the North Atlantic between 35 and 45∘ N, while MIPAS has significant positive anomalies of 10 pptv. TOMCAT also simulates significant negative anomalies over the UK and the eastern North Atlantic, which are not observed by MIPAS. Therefore, the model results only allow for limited assessment of the NAO influence of UTLS PAN in NAO-high over these regions.

Figure  shows the zonal average (90∘ W–20∘ E) meridional-vertical TOMCAT PAN distribution under both NAO phases. Between 1000 and 600 hPa, PAN concentrations are above 200 pptv apart from in the region 30–40∘ N. From 400 to 200 hPa, there is a sharp PAN decrease to less than 30 pptv. The vertical PAN profile between 30 and 40∘ N differs from other latitude bands, with PAN peaking at 150–200 pptv from 1000 to 700 hPa and then from 60–100 pptv up to 200 hPa. Between 200 and 100 hPa, PAN concentrations (30–60 pptv) are larger than other latitude bands linked to the higher tropopause (also observed by MIPAS – see Supplement). The significant decreases in surface PAN over Europe from NAO-high enhanced westerly flow (Fig. c) occur throughout the troposphere, with negative zonal anomalies of -10 to -3 pptv (Fig. c). Above the tropopause (dashed green line), strong vertical (winds are scaled by 104 for clarity) meridional transport accumulates PAN (positive anomalies over 10 pptv) in the Arctic UTLS. Under NAO-low (Fig. d), there are positive (5–10 pptv) and negative (-10 to -5 pptv) anomalies between 30–50 and 60–90∘ N throughout the troposphere. The surface patterns (Fig. d), where reduced transport aids PAN accumulation over Europe, appear to account for this zonal tropospheric pattern. Between 30 and 50∘ N, there is limited meridional flow aiding PAN accumulation over Europe in NAO-low. The vertical flow contributes to positive anomalies (3–5 pptv) propagating into the UTLS, which is consistent with the PAN accumulation shown in Fig. d between 50 and 70∘ N.

TOMCAT surface ozone under NAO-high (Fig. a) peaks at approximately 28–30 ppbv over the subtropical and western North Atlantic co-located with the enhanced westerlies. Over continental Europe, ozone concentrations are significantly larger (1–2 ppbv – Fig. c) than the wintertime average, ranging between 16 and 25 ppbv. This matches a similar pattern in the observations: AURN surface ozone was significantly higher over the UK under NAO-high than NAO-low (see Supplement), and TES lower-tropospheric ozone (Zone 1, Fig. ) was larger under NAO-high. In Zone 2, TES lower-tropospheric ozone was higher under NAO-low, which correlates with the surface TOMCAT pattern. found similar patterns with significant positive (negative) correlations over Europe (western North Atlantic) between surface ozone and the NAOI in DJF. Under NAO-low conditions, TOMCAT ozone concentrations are consistent across the North Atlantic (28–30 ppbv) as the weakened or reversed westerlies limit the transport of ozone-enriched Atlantic into Europe yielding lower concentrations of 13–20 ppbv. Over the western North Atlantic (Europe), ozone concentrations (Fig. d) have increased (decreased) with significant positive (negative) anomalies of 2–3 ppbv (-3 to -1 pptv). Again, presented similar results which also match TES and AURN ozone observations (see Supplement). TOMCAT tropospheric column ozone (not shown here) also showed similar anomalies.

At the surface, the PAN and NO2 spatial anomalies are anti-correlated with ozone, so UTLS ozone (Fig. ) was investigated to see if this relationship was consistent at higher altitudes. TOMCAT 200–100 hPa average ozone, sampled under NAO-high, ranges from 800 to 1000 ppbv northwards of 60∘ N but decreases towards the subtropical North Atlantic with minimum concentrations of 150–200 ppbv. A similar pattern occurs under NAO-low except for the ozone-reduced air mass (500–700 pptv), stretching from approximately 45 to 65∘ N along 15–45∘ W. Higher ozone concentrations (800–1000 ppbv) also propagate further south in NAO-low on either side of the Atlantic, surrounding the reduced ozone limb. The UTLS ozone anomalies (Fig. c and d) are also anti-correlated with the PAN. Whereas PAN has positive anomalies across the Atlantic basin in NAO-high, there are significant negative ozone anomalies (under -100 ppbv). This anti-correlation is also prominent under NAO-low, where significant TOMCAT ozone anomalies (50–200 ppbv) exist over the mid-North Atlantic and Europe but are significantly negative for PAN. As shown in Figs.  and , tropospheric positive PAN anomalies propagate into the UTLS over Iceland and southern Greenland, but the ozone anomalies are significantly negative (-150 to -50 ppbv). Potential reasons for the PAN-ozone anti-correlation include the air mass origin or the PAN (NOx)-ozone chemistry. The thermal decomposition of PAN forms the peroxyacetyl radical and NO2, which is an UTLS ozone sink (i.e. conversion of NO2 to NO and then reaction with ozone), while a tropospheric ozone source is in the presence of volatile organic compounds . However, lower UTLS temperatures (i.e. around 250 K) yield a PAN lifetime of several months and are a less likely factor in the PAN NOx-ozone anomaly anti-correlations. show that UTLS ozone-HOx chemistry is a more significant sink pathway for ozone; however, there is no clear correlation between the NAO HO2 and ozone anomalies. Methane, a good air mass tracer due to its approximate 9-year lifetime (e.g. ) and anthropogenic source, was sampled under the NAO phases (not shown) and highlighted similar anomaly patterns to PAN, again anti-correlated with ozone. Therefore, PAN and methane (ozone) act as signatures for the transport of polluted (clean) air masses in the troposphere for the different NAO phases.

Figure  shows the TOMCAT ozone cross section at 0∘ E, similar to NO2 in Fig. . Under both NAO phases, lower-tropospheric (UTLS – above 300 hPa) ozone ranges between 25 and 35 (above 100) ppbv. Meridionally, there is a decreasing poleward lower-tropospheric ozone gradient, while in the UTLS peak (minimum) concentrations are at the pole (30∘ N). The anomalies, as discussed above, are anti-correlated with NO2. In NAO-high, there are small positive (negative) anomalies over the UK (North Sea) consistent with ozone-enriched air transported into the UK from the North Atlantic and ozone loss downwind due to source-region NOx transport, which propagates up to approximately 600 hPa. The positive anomalies in the UTLS at 60∘ N are consistent with ozone accumulation seen in Fig. . Under NAO-low conditions, the negative anomalies (approximately -3 ppbv) between 45–65∘ N and 1000–700 hPa are linked to UK NOx accumulation (Fig. d). Atmospheric downwelling leads to UTLS ozone (positive anomalies over 3 ppbv) propagation into the mid-troposphere at 40–50∘ N. At high latitudes, small positive anomalies throughout the troposphere, which are anti-correlated with PAN, show clean-air transport around the UTLS Icelandic–southern Greenland anticyclone in which PAN accumulates.

A second TOMCAT ozone cross section at 56.25∘ W (Fig. ) has similar absolute ozone concentrations to the 0∘ E cross sections, but the anomalies highlight important differences. In NAO-high, both cross sections have similar lower-tropospheric ozone anomalies except at 50∘ N (Figs. c and c) with positive anomalies (approximately 1 ppbv) over the UK region and near 0 ppbv over the western North Atlantic. Under NAO-low conditions, there are positive anomalies (Fig. d, 1–3 ppbv) between 1000–600 hPa and 50–70∘ N, but the eastern cross section (Fig. d) highlights negative anomalies in this region (-3 to -1 pptv). While there is downwelling of stratospheric ozone in the eastern cross section into the middle–upper troposphere during NAO-low, the western cross section has an upwelling of ozone-reduced air into the UTLS with negative anomalies of less than -5 ppbv. Overall, in the lower troposphere, the TOMCAT cross section anomalies support the signals in the TES data. Over the UK (Zone 1, Fig. , and eastern cross section, Fig. ), lower-tropospheric ozone is larger (lower) than the wintertime average under NAO-high (NAO-low), while the opposite occurs in the western North Atlantic (Zone 2 and western cross section, Fig. ).