NOx mixing ratios were generally low in the Mediterranean Sea (Fig. 2a). One-minute mean and median mixing ratios of NOx as detected by the TD-CRDS were 1.3 and 0.3 ppbv, respectively. For the CLD measurements of NOx, the equivalent values are 1.1 and 0.2 ppbv, respectively. For both instruments, the difference between mean and median values stems from the frequent occurrence of NOx plumes resulting from emissions of nearby ships. The NOx mixing ratios measured by TD-CRDS and CLD were in good agreement (see Friedrich et al., 2020) and the bias of the TD-CRDS to higher values reflects the exclusion of data below the detection limit. A histogram of the NOx measurements made by the CLD is displayed in Fig. 2b), which indicates that 33 % of the NOx data were between 100 and 250 pptv and 24 % were above 1 ppbv. The maximum mixing ratio of NOx in the Mediterranean Sea of 84.7 ppbv was measured in the narrowest part of the Strait of Messina, which is a busy corridor for international shipping with ferry traffic between Italy and Sicily crossing the Kommandor Iona's ship track. This observation highlights the importance of NOx shipping emissions in some parts of the Mediterranean Sea, which we return to later.

Potential non-shipping sources of NOx in this region can be identified via the back-trajectories plotted in Fig. 1. In the eastern part of the Mediterranean Sea, the air masses were influenced by emissions from the heavily populated and industrialised western Turkish coastal area, the island of Crete, and mainland Greece. However, as we show below, the lifetime of NOx is generally less than 6 h, and the greater fraction of any land-based NOx emissions would have undergone oxidation to NOz during the 48 h transport time of the back-trajectory. In the western Mediterranean Sea, the 2 d back-trajectories end above the open ocean.

Our data can be compared to results from previous measurements of NOx in the Mediterranean area. Excluding pollution events, Mallik et al. (2018) report NO and NO2 levels below 0.05 and 0.25 ppbv, respectively, during the 2014 Cyprus-based CYPHEX (CYprus PHotochemical EXperiment 2014) campaign in the eastern Mediterranean Sea. Plume-like increases in NOx were associated with enhanced SO2 and related to emissions from shipping (Eger et al., 2019b). During the MINOS (Mediterranean INtensive Oxidant Study) campaign on the island of Crete, median NO2 mixing ratios between 0.3 ppbv and 0.7 ppbv were reported (Berresheim et al., 2003). The lower mixing ratios were associated with air masses arriving from the western European free troposphere, whereas the higher values were air masses impacted by biomass burning in eastern Europe. In contrast, higher NO2 mixing ratios (typically between 4 and 6 ppbv excluding plumes) were reported from shipboard measurements in the Aegean Sea (Večeřa et al., 2008). Satellite-based observations of NO2 vertical column densities over Crete and in the region between Crete and Sicily, were used to derive near-surface NO2 mixing ratios of up to ∼ 0.4 ppbv (Ladstätter-Weißenmayer et al., 2003, 2007).

Our NOx measurements are thus broadly consistent with previous measurements in the Mediterranean Sea which indicate mixing ratios of less than 1 ppbv in the absence of recent emissions from ships. The higher mixing ratios reported by Večeřa et al. (2008) are likely to be related to the close proximity of their ship to NOx sources on the European continent and denser ship traffic compared to the more southerly AQABA route through the eastern Mediterranean Sea.

Figure 2c and d show a time series and histogram, respectively, of NOz for the Mediterranean Sea. The shape of the distribution indicates that NOz mixing ratios close to the detection limit were rarely measured. The mean (0.8 ppbv), median (0.7 ppbv), maximum (2.8 ppbv), and minimum NOz mixing ratios (< 0.1 ppbv) along with the narrower distribution indicate that, as expected, NOz is significantly less variable than NOx. The ratio of the median mixing ratios NOz / NOy in the Mediterranean Sea is ∼ 0.8, and, concomitantly, that of NOx / NOy is ∼ 0.2. A more detailed analysis of the relative contributions of NOx and NOz to NOy, in which we divide the Mediterranean Sea into seven subregions, is presented in the following paragraphs. The choice of these subregions was based on the existence of homogeneous NOz / NOy ratios over periods of hours to days. This approach enabled us to compare subregions with substantially different chemical regimes along the ship track but does not lend itself to the derivation of a representative NOy budget for the entire region.

The pie charts in Fig. 3 indicate the regional average contributions (in subregions M1 to M7) of reactive nitrogen species to NOy. The fractional contributions are based on measurements of NOx, NOy, gas-phase organic nitrates (ON), particulate nitrate (pNit), ClNO2, and HONO. HNO3 was not measured directly but calculated from HNO3= NOy – (NOx+ ON + pNit + ClNO2+ HONO), where pNit refers to submicron particulate nitrate as measured by the HR-ToF-AMS. Detection of coarse-mode pNit by the TD-CRDS (see Friedrich et al., 2020) would lead to an overestimation of HNO3. However, given that the thermal dissociation to NO2 of NaNO3 particles with 300 nm diameter is inefficient (∼ 20 %) with this instrument, a significant bias by coarse-mode nitrate (e.g. associated with sea salt or mineral dust) appears unlikely. The data available in each subregion did not always cover an entire diurnal cycle, which will have an impact on the fractional contributions of individual NOy species (see differences between day- and night-time chemistry in Sect. 1). We argue, however, that diurnal patterns in NOy are likely overshadowed by the variability of air mass sources. The NOy compositions presented are thus to be considered coarse estimates.

In all subregions, HNO3 is the dominant component of NOy besides NOx and therefore the most important NOz species. Contrastingly, submicron pNit only contributes between 5.4 % (M6) and 15.5 % (M2) to NOy and ONs between 7.1 % (M4) and 16.9 % (M2). ClNO2 only constitutes a minor part of NOy with ca. 1 % contribution in all regions where ClNO2 was measured (M1–M6). The low mixing ratios of ClNO2 have been attributed to high night-time temperatures and high reactivity of NO3 which reduce the interaction of N2O5 with chloride-containing particles (Eger et al., 2019a). Elevated HONO mixing ratios (up to 0.3 ppbv) were observed in regions M3 and M6 where its contributions to NOz was 3.8 % and 4.2 %, respectively. As the daytime lifetime of HONO is short (a few minutes) due to its rapid photolysis (Platt et al., 1980), HONO levels up to 0.3 ppbv imply strong sources. Elevated HONO mixing ratios in ship plumes have been observed in previous field measurements (Večeřa et al., 2008; Sun et al., 2020) and could explain the presence of HONO in subregions M3 and M6. Other sources of HONO, summarised in Elshorbany et al. (2012), include heterogeneous or photochemical reactions of NOx and NOz on various surfaces and also the photolysis of particulate nitrate (Meusel et al., 2018).

Figure 3 also plots the NOz / NOy ratio along the ship's track. The highest values with median NOz / NOy > 0.68 were found in regions M2, M4, and M7, reflecting a lack of local NOx sources as confirmed by the back-trajectories. For visual clarity, only the back-trajectories starting at the geographical centres of the respective subregions are displayed in Fig. 3. Back-trajectories starting at the ship's location 4 h before or after confirmed that the air mass origin was very similar. In contrast, the regions designated M3, M5, and M6 are influenced by fresh emissions from land-based sources and are characterised by low NOz / NOy ratios (medians < 0.55), reflecting the higher levels of NOx which contributed 52 % (M3 and M5) and 43 % (M6) to total NOy.

In the following section, the observations of NOx in the Mediterranean Sea are analysed in terms of its production and loss. Following the considerations in Sect. 1, we compare the daytime loss of NOx via the reaction between NO2 and OH (Reaction R3; expected to dominate over other daytime NOx loss processes in the marine environment) with night-time losses via the reaction between NO2 and O3 (Reaction R8): 1kNO2=k3OH+k8O3, where kNO2 represents the total loss rate constant (in s-1) for NO2 and is the inverse of the NO2 lifetime (τNO2). The first term on the right-hand side of this expression is most important during the day when OH levels were high (up to 1.4 ×  107 molec. cm-3 at local noon) but relatively unimportant at night. In contrast, the second term on the right is only important at night as the NO3 product of R8 is rapidly photolysed back to NOx during daytime so that NOx is conserved.

By using Eq. (1) to approximate the NO2 loss rate constant, we neglect two further processes which can, under some conditions, influence the lifetime of NO2. Our approach assumes that the night-time formation of NO3 leads to the removal of one NO2 molecule. This approach would be invalid if a significant fraction of NO3 would be lost via formation (and subsequent heterogeneous loss) of N2O5. Firstly, we note that formation of N2O5 was hindered during AQABA by the high gas-phase reactivity of NO3 towards VOCs (Eger et al., 2019a) and that the transfer of N2O5 to the particle phase was hindered by high temperatures. For example, taking an N2O5 uptake coefficient γN2O5 of 0.03 (as found for polluted marine environments by Aldener et al., 2006) and the median night-time aerosol surface area (ASA) in the Mediterranean Sea of 1.78×10-6 cm2 cm-3 (Eger et al., 2019a), we estimated a loss rate constant for uptake of N2O5 of 3.5×10-4 s-1, which is 2 orders of magnitude lower than the rate constant (4.9×10-2 s-1) for thermal decomposition at 25.7 ∘C (the mean, minimum night-time temperature in the Mediterranean Sea).

We also neglect the loss of NOx via uptake of NO2 onto black carbon (BC) particles. Using a literature uptake coefficient γNO2 of ca. 1×10-4 (Longfellow et al., 1999) and the aforementioned ASA, the first-order loss rate constant for the heterogeneous uptake would be 1.8 × 10-6 s-1. Using an O3 mixing ratio of 63.4 ppbv (equal to night-time median mixing ratio in the Mediterranean Sea), we calculate a first-order loss rate constant for the reaction of NO2 and O3 of 5.5 × 10-5 s-1, which implies that > 95 % of total NO2 loss at night-time NO2 is due to O3. Uptake of NO2 might, therefore, be relevant for HONO formation (see Sect. 3.4) but does not constitute a relevant loss process for NOx.

In order to fill gaps in the OH data set (daytime data coverage of 71 %), complete diel cycles of OH were generated by scaling measurements of JO1D to the OH noon-time maxima. Figure S4 compares the measured OH concentrations with the interpolated trace and shows that the thereby derived OH levels can be considered upper limits. Inserting these values and the measured O3 concentration into Eq. (1) and using preferred rate coefficients for k3 and k8 (IUPAC, 2020) we derive lifetimes (Fig. 4a) of ∼ 2 h at local noon (largest OH levels) and 5–6 h at night. Loss of NOx by deposition may be important in forested regions (Delaria et al., 2018; Delaria and Cohen, 2020) but is expected to be insignificant in a marine environment. The relative importance of day- and night-time losses of NO2 in the Mediterranean Sea during AQABA was estimated by integrating the two loss terms using the available NO2, O3, and OH data. Averaged over the 6 d of measurements, 3.71 ppbv of NOx was lost per 12 h day and 1.51 ppbv was lost per 12 h night (Fig. 4c).

Although our conclusion is based on a limited data set, we calculate that the OH-induced daytime loss of NOx is most important in the Mediterranean Sea, reflecting the high levels of OH encountered during AQABA, but we note that night-time losses make a significant contribution. It is very likely that in other seasons with reduced photochemical activity and lower temperatures (which favour the formation of N2O5 which can remove two NO2 molecules via heterogeneous processes), the night-time losses gain in relative importance. Averaged over the entire data set obtained in the Mediterranean Sea, we calculate a lifetime of NO2 of 3.9 h. Chemical sources of NOx in the Mediterranean Sea, i.e. from the photolysis of HONO and pNit, as well the reaction of OH and HNO3, are discussed in detail in Sect. 3.4.

In the following, we examine the contribution of ship emissions to the NOx budget in the Mediterranean region and especially along the track taken by the Kommandor Iona during the AQABA campaign. In Fig. S5 we plot a time series of NOx and SO2 data for the transit through the Mediterranean Sea. It is immediately apparent that large plume-like features in NOx coincide with similar features in SO2. We now separate the data set into two regimes in which the NOz and NOy measurements indicate either relatively “fresh” emissions (NOz / NOy ratio < 0.4) or relatively “aged” emissions (NOz / NOy ratio > 0.8). In Fig. 5a we show that, for fresh emissions, SO2 and NOx are highly correlated (Pearson's R = 0.84) with a slope of 4 ± 0.1 ppbv NOx per ppbv SO2 and an intercept (at zero SO2) of -1.9 ± 0.3 ppbv. This strongly suggests that fresh NOx emissions are generally accompanied by SO2 and thus indicates that either ships or power plants, e.g. in coastal locations, are the likely sources of a large fraction of the NOx. The slope is similar to that derived by Celik et al. (2020) (2.7 ± 0.8), who examined single ship plumes in a more detailed analysis and with literature values that range from 6.8 ±  6.3 near the coast of Texas (Williams et al., 2009) to 11.2 ± 10.9 (Diesch et al., 2013) at the Elbe river near Hamburg/Germany. In comparison to Celik et al. (2020), however, the two other literature studies only sampled very fresh and unprocessed ship plumes, from a distance of less than ca. 5 km to the emission source. Fig. 5a shows that NOz and SO2 are not correlated (Pearson's R= 0.38) in air masses impacted by fresh emissions.

In more aged air masses (Fig. 5b) the slope of NOx per SO2 is, as expected, much smaller (0.16 ± 0.01 ppbv NO2 per ppbv SO2) which reflects the significantly longer lifetime of SO2 (∼ 10 d) compared to NOx. After a few days of transport an air mass containing co-emitted NOx and SO2 will still contain SO2 but the initially emitted NOx will, to a large extent, have been converted to NOz. The intercept (NOx= 0.049 ± 0.005 ppbv at zero SO2) is consistent with the re-generation of NOx from NOz (see above) but is also in the area of the detection limit of the NOx measurement.

The plot of NOz versus SO2 for aged emissions indicates a significant intercept (at zero SO2) of 0.4 ppbv NOz. As the lifetime of SO2 (∼ 10 d) is longer than of NOz (∼ half a day) (Dickerson et al., 1999; Romer et al., 2016), the residual NOz at zero SO2 cannot stem directly from ship emissions (or combustion sources that generate both NOx and SO2) but represents the background level of NOz in the Mediterranean Sea in aged air masses and is consistent with an average HNO3 mixing ratio of 0.48 ppbv observed during the MINOS campaign at Finokalia on Crete (Metzger et al., 2006).

The analysis above, when combined with back-trajectory information, provides clear evidence that shipping emissions are responsible for a large fraction of NOx in the Mediterranean Sea. The impact of shipping emissions on the atmospheric sulfur budget has been assessed in numerous studies which identify coastal areas and international shipping lanes as important hotspots for SO2 emissions (Capaldo et al., 1999; Dalsøren et al., 2009; Eyring et al., 2010), with emissions of SO2 severely impacting air quality in port regions (Isakson et al., 2001; Cooper, 2003; Saxe and Larsen, 2004; Marmer and Langmann, 2005; Ledoux et al., 2018). A detailed analysis of SO2 data with regard to ship emissions during AQABA is provided by Celik et al. (2020), who analysed emission factors from individual ship plumes during the AQABA campaign.

NOx mixing ratios in the Red Sea (excluding the 3 d layover in the port of Jeddah) as measured by the TD-CRDS and the CLD instruments are displayed in Fig. 6a. NOx mixing ratios were highly variable and there were only short periods free of NOx plumes > 10 ppbv (e.g. during the second leg on 19 and 20 August 2017). The mean NOx mixing ratios (2.8 ppbv measured by the TD-CRDS and 3.2 ppbv measured by the CLD) were therefore significantly higher than the median values of 1.0 ppbv. Figure 6b indicates that the NOx mixing ratios are broadly distributed around the median of 1.0 ppbv with 21 % of all data points >  3 ppbv. The highest NOx levels during AQABA were found in narrow shipping corridors of the Suez region and the Strait of Bab al-Mandab. When excluding the Suez and Bab al-Mandab regions, a median NOx mixing ratio of 0.7 ppbv can be derived for the maritime central part of the Red Sea.

To the best of our knowledge, in situ measurements in the Red Sea area are not available for comparison with our NOx data. Satellite-based modelling studies show that high NO2 column densities above the Red Sea are associated with shipping emissions (Richter et al., 2004; Alahmadi et al., 2019), which is consistent with our observation of a strong correlation between NOx and SO2 (see below). Johansson et al. (2017) have estimated a NOx emission rate of 0.70 t km-2 yr-1for the Red Sea (including the Suez region).

The mean mixing ratio of NOz over the Red Sea was 1.0 ppbv, with a maximum value of 8.0 ppbv measured in the Gulf of Suez on the first leg. NOz mixing ratios are narrowly distributed (see Fig. 6d) around a median value of 0.7 ppbv, with 53 % of the measurements between 0.4 and 1.0 ppbv and 41 % between 1.0 and 4.0 ppbv.

The NOz / NOy ratios along the ship's track are plotted in Fig. 7: values > 0.6 were mostly observed over the northern Red Sea on the first leg, after leaving the Gulf of Suez. On the second leg, the NOz / NOy ratio was higher in the southern Red Sea. NOz data coverage was limited in the Red Sea on both legs and the NOz / NOy ratio was more variable than values found in the Mediterranean Sea and the Arabian Gulf. The high variability in the NOz / NOy ratios is caused by the route of the Kommandor Iona along the main shipping lane connecting the Suez Canal and the Gulf of Aden and the frequent sampling of plumes from nearby ships. The observed NOz / NOy ratios of < 0.6 in the Red Sea highlight the impact of NOx emissions from shipping on the reactive nitrogen budget and the air quality in the Red Sea region (as discussed in Sect. 3.2.1).

For the Red Sea, we have defined four subregions in which we calculate the contributions of NOx and various NOz species to NOy: these are RS1 on the first leg and RS2, RS3, and RS4 on the second leg. Note that RS1 and RS4 are both located in the northern Red Sea, but the measurements (∼ 5 weeks apart) revealed different chemical characteristics; hence the separate treatment.

Due to poor data coverage, mainly of organic nitrates, we were not able to perform this calculation in further subregions on the first leg. In all four regions, NOx was the largest component of NOy which results from continuous NOx input from onshore and shipping emissions.

In RS1 we observed the lowest contribution (36.4 %) of NOx to NOy and the largest contribution of ONs (23.8 %) to NOy, over the Red Sea. The latter value is the highest found during the entire AQABA campaign and is comparable to the contribution of HNO3 (30.0 %). In roughly co-located RS4, but 5 weeks later, the NOx contribution was much larger (69.5 %). The divergent median NOx / NOy and NOz / NOy for subregions RS1 and RS4 can be understood when one examines the air mass back-trajectories for the two legs. On the second leg, strong northerly winds transported NOx from the highly polluted southern end of the Gulf of Suez to RS4, whereas during the first leg the back-trajectory for RS1 passed (with lower wind speeds) mainly over eastern Egyptian deserts, with emissions from Cairo requiring 36 h to reach RS1 during which a significant fraction of NOx was converted to NOz. We expect that the large contribution of ONs in RS1 is a result of the unique chemical environment at the southern end of the Gulf of Suez and in the northern Red Sea. A large coherent oil field is located south of the Gulf of Suez and the coast of eastern Egypt (Alsharhan, 2003), and the numerous facilities for oil extraction result in abundant emissions of VOCs, while the proximity to the Gulf of Suez and the narrowing shipping corridor on the approach to Suez provides the NOx required for formation of organic nitrates (ONs). Meteorological conditions additionally favoured a build-up of ONs during our passage through RS1: elevated wind speeds of up to 11 m s-1 coincided with temperatures below 30 ∘C, which slowed down the thermal decomposition of PAN (peroxyacetyl nitrate) compared to the ca. 35 ∘C regime in the Arabian Gulf. Average PAN mixing ratios, as measured by chemical ionisation mass spectroscopy (CIMS), were 190 pptv in this area, which constitutes ca. 20 % of the total ONs signal. On the second leg in RS4, the fractional contribution of ONs was overshadowed by the stronger impact of NOx pollution from the Suez region (see above).

In RS1 and RS4 the contributions of HONO and ClNO2 to NOy were minor (≤ 3 %). RS2 and RS3 are both located in the southern half of the Red Sea. For RS3 we observed the highest contribution (15 %) of AMS-measured particulate nitrate to NOy, and RS3 was characterised in large parts by coarse-mode OPC fractions > 85 % (i.e. (PM10-PM1) / PM10; see 18 and 19 August 2017 in Fig. S1). It is reasonable to assume that the coarse-mode particle mass concentrations in this area was due to sea salt, which reacts heterogeneously with HNO3 to form particle-phase nitrates (Mamane and Gottlieb, 1990). Refractory sea salt aerosol particles in the PM1 size range are, however, not expected to be detectable via AMS (Jimenez et al., 2003) or with only very low efficiency (ca. 1 %) (Zorn et al., 2008).

Region RS2 shows a intermediate behaviour, as NOz / NOy increases after leaving Bab al-Mandab and transported air only came from the surrounding southern Red Sea without being influenced by shore-side anthropogenic activities. Here, NOx and HNO3 contribute 57 % and 27 %, respectively. The relatively high NOx contribution, considering the remote area, can be explained by sampling ship plumes on the departure from Bab el-Mandab, which led to several NOx peaks above 10 ppbv (see Fig. 6a). Consequently, background NOx levels also did not fall below ca. 1.5 ppbv on the night from 17 to 18 August 2017. Overall, the fractional contributions of NOx were positively biased by short-term spikes in NOx mixing ratios caused by ship plumes in all Red Sea subregions. The use of mean values to assess the fractional contributions of NOy species in certain subregions is thus a caveat of this analysis, as NOz signals exhibit less variability during pollution events (see Fig. 6c). Employing the median values, however, would not allow the relative contributions to NOy to be assessed.

Analogous to Sect. 3.1.3, we now investigate the day- and night-time chemical losses of NO2 in the Red Sea (see Fig. 8). As described previously, we used an interpolated OH data set based on a scaling factor between the available OH data and JO1D.

As OH was not measured over the Red Sea on the first leg, our analysis is restricted to the second leg only. Daytime NO2 lifetimes with respect to loss by reaction with OH were usually in a range between 2 and 4 h, with a minimum of 1.7 h on 21 August 2017, where the noon-time OH concentration peaked at 1.1 × 107 molec. cm-3. Night-time NO2 lifetimes (determined by O3 levels) exhibited a larger variability but were mostly between 5 and 10 h. The average (day and night) NO2 lifetime in the Red Sea was 5.0 h.

Over the entire period of measurements in the Red Sea (8 d and 8 nights) we calculate that a cumulative total of 62 ppbv of NO2 were lost (Fig. 8c). Despite the shorter lifetime of NO2 at noon, the greater integrated loss of NO2 occurred during night-time (5.7 ppbv per night on average) when continually high O3 levels (median 54 ppbv) were available. At midday, NO2 mixing ratios are reduced due to the shift in the NO2 / NO ratio caused by the rapid photolysis of NO2 and also because the OH levels are highest then. On average, daytime loss rates were 2.0 ppbv per day.

In order to assess the contribution of shipping on NOx emissions, we correlated NOx and SO2 mixing ratios for freshly emitted (NOz / NOy < 0.4) and chemically more aged (NOz / NOy > 0.8) air masses. The results are illustrated in Fig. 9 and summarised in Table 2, which reveal a positive correlation (slope of 3.7 ± 0.1 and a regression coefficient R of 0.61) between NOx and SO2 in air masses containing freshly emitted pollutants. Six data points far above 20 ppbv (range 43–128 ppbv SO2) were excluded, as they would bias the linear regression result. Including these data points lowers the slope to 1.26 ± 0.04 and the correlation coefficient R to 0.40. The NOx / SO2 ratio is thus highly variable throughout the Red Sea, potentially reflecting variable NOx / SO2 emission ratios of different vessels, using various fuels, as well as the impact (on NOx) of offshore oil-drilling rigs and shore-side oil refineries. The latter are most important in the northern Red Sea, whereas shipping emissions dominate in the narrow shipping lanes of the Suez Canal.

For chemically aged air masses, the NOx / SO2 ratio is 0.20 ± 0.01 with R= 0.61 and the reduction in slope reflecting the shorter lifetime of NOx compared to SO2. We find, however, that in chemically aged air masses, NOz and SO2 are highly correlated (Fig. 9b) with a slope NOz / SO2 of 1.25 ± 0.04 and R=0.85. The intercept (see Fig. 9b) at an SO2 mixing ratio of zero is 0.40 ± 0.03 ppbv, which can be taken to be the regional NOz background mixing ratio (i.e. NOz formed from NOx which was not emitted from SO2-containing fuels).

Elevated NOx mixing ratios were detected by both TD-CRDS and CLD throughout the Arabian Gulf (see Fig. 10a). The TD-CRDS measured mean and median NOx mixing ratios of 3.3 and 1.6 ppbv, respectively. By comparison, the CLD measured an average of 4.1 ppbv and a median of 1.8 ppbv. The large difference between median and mean reflects the numerous plumes of high NOx detected by both instruments (Fig. 10a). The deviation of the TD-CRDS and the CLD data is caused by different data coverage as the CLD continued measuring in the most polluted areas close to Fujairah and Kuwait, while the TD-CRDS was switched to zeroing mode, in order to avoid contamination of the inlet lines. When limiting the comparison to periods where both instruments were operating, very similar median values are obtained, with 1.6 ppbv from the TD-CRDS and 1.5 ppbv from the CLD. A histogram of the NOx measurements (CLD data only) made in the Arabian Gulf (Fig. 10b) shows a broad distribution, reflecting high variability in the region, with 77 % of the data points falling into a range between 0.4 and 10 ppbv and a broad maximum at 1–3 ppbv. The highest NOx daily maxima were observed near Fujairah (up to 34 ppbv on the first and 153 ppbv on the second leg), in the Strait of Hormuz (26 and 30 ppbv), and when approaching and departing Kuwait (43 and 90 ppbv). The locations of these maxima close to the shore or in narrow shipping corridors and the plume-dominated time series suggest the influence of mostly local pollution sources of NOx, i.e. from ship traffic or from industrial activities in the shore-side areas of the neighbouring cities. NOx mixing ratios < 0.5 ppbv were found exclusively in the central part of the Arabian Gulf, which is the widest part (least influence from onshore activity) with the largest spread of the shipping lanes.

The generally very high levels of NOx in the Arabian Gulf are consistent with results from satellite measurements which have identified high NO2 tropospheric vertical column densities over the Gulf of Oman, the Strait of Hormuz, and the south-eastern Arabian Gulf (Beirle et al., 2004). Model studies estimate a NOx emission rate of 1.13 t km-2 yr-1 for the Arabian Gulf (Johansson et al., 2017). With a NOx lifetime of 4.0 h (see Sect. 3.3.3) and a boundary layer height of 1 km (Wu et al., 2008), this emission rate translates to a NOx mixing ratio of 0.3 ppbv. The lower mixing ratio, compared to the median NOx observed on AQABA (see above), is likely caused by the averaging of the model over the entire Arabian Gulf water surface area, whereas the Kommandor Iona followed common shipping routes with larger NOx emissions. To the best of our knowledge, there are no in situ measurements of NOx over the Arabian Gulf with which to compare our data.

The Arabian Gulf featured the highest NOz levels during the AQABA campaign (see Fig. 10c), with mixing ratios from < 0.1 ppbv up to 6.9 ppbv (mean 2.0 ± 1.5 ppbv (standard deviation) and median of 1.5 ± 0.7 ppbv (median absolute deviation)). The histogram of NOz mixing ratios (Fig. 10d) shows a maximum in the frequency distribution at 1–3 ppbv, with 73 % of all data above 1 ppbv and 15 % above 4 ppbv. Our results thus indicate that the Arabian Gulf is a hotspot for NOz formation, which is a result of high levels of the NOx and VOCs precursors and also O3. The spatial distribution of the NOz / NOy ratio for both legs is presented in Fig. 11. On both legs, NOz / NOy ratios above 0.8 were found in the central part of the Arabian Gulf, which results from the processing of NOx emissions during transport from the shore to the centre of the Arabian Gulf.

We now examine the partitioning of NOy into its various components in the four subregions (A1–A4) defined above for the Arabian Gulf (Fig. 11). On the approach to Kuwait (A2), winds from the north transported fresh NOx emissions from cities in Kuwait and Iraq to the ship and NOx accounted for 81 % of NOy. More aged air masses were found in other regions (A1, A3, and A4) with a roughly equal split between NOx and HNO3 (both 45 %–50 %) observed in A3 and A4. The major component of NOz was HNO3 in all regions, with significant but very variable contribution from organic nitrates, especially in A1 (13 ± 16 %) where the air masses originated from the eastern coast of Saudi Arabia, which accommodates numerous facilities for oil and gas extraction and processing, resulting in high levels of organic trace gases including alkanes, alkenes, and aromatics (Bourtsoukidis et al., 2019). Particulate nitrate contributed only minor amounts to NOz in the Arabian Gulf, which reflects the high temperatures and resultant partitioning of nitrate into the gas phase. Other NOz species contributed only weakly to the NOz as indicated in Fig. 11.

Analogously to Sect. 3.1.3, we also determined NO2 lifetimes and the cumulative loss of NO2 in the Arabian Gulf. The results are presented in Fig. 12. Limited by the availability of OH data, these calculations include only the time period after 29 August 2017 on the first leg. In the same way as in Sect. 3.1.3, we used an interpolated OH data set in the following calculations.

In the Arabian Gulf, daytime NO2 lifetimes (considering loss by OH) were generally between 2 and 4 h. Night-time lifetimes were in a similar range but also occasionally exceeded 10 h, e.g. when leaving the Arabian Gulf towards the Gulf of Oman and the Arabian Sea on the second leg, where O3 mixing ratios fell below 20 ppbv. The average NO2 lifetime was calculated to be 4.0 h.

Figure 12c shows that 50 ppbv of NO2 was lost cumulatively throughout the period of measurements over the Arabian Gulf, with night-time losses (black data points) being more important than daytime losses (red data points). On average 6.0 ppbv NO2 were lost per night and only 2.1 ppbv was lost per day. Large night-time compared to day-time losses are related to moderate OH levels in large parts of the Arabian Gulf (see Fig. 12b). The daytime average OH concentration was 2.4 × 106 molec. cm-3, while on average 73 ppbv O3 was present. The measured OH concentrations were generally low, given the NOx and O3 levels in the Arabian Gulf, which may have resulted from its reactions with VOCs. With a loss rate constant of 11.6 s-1, the Arabian Gulf was the AQABA region with the largest median OH reactivity (Pfannerstill et al., 2019), with 61 % of the total OH reactivity attributed to various measured VOCs. The daytime losses of NO2 are therefore indirectly limited by the availability of VOCs from the oil and gas production (see above).

Via analysis of correlation between SO2 and NO2 (Fig. 13 and Table 2), we can assess the influence of shipping emissions on NOx mixing ratios in the Arabian Gulf. In air masses recently influenced by NOx emissions (NOz /  NOy < 0.4), NOx and SO2 are only weakly correlated (slope = 4.1 ± 0.2, R= 0.41), indicating that many different NOx sources (i.e. not only shipping emissions) contribute. These might include vehicular traffic and industrial activity (e.g. production of nitrogen-based fertilisers Khan et al., 2016) in Kuwait City, the Iraqi city of Basra, and in Iranian harbours and offshore oil and gas terminals. Considering the limited NOx lifetime, the land-based emission sources of NOx from urban/industrialised areas gain in importance over plumes from nearby ships, when approaching the coast. In aged air masses, the slope of the NOx versus SO2 correlation is 0.11 ± 0.01 with a large correlation coefficient (R= 0.72). This indicates that in aged air masses, the NOx levels are linked to SO2 emissions, which is consistent with the photolysis of HONO being a major source of NOx in the region. From the intercept (SO2 mixing ratio = zero = 0.0 ± 0.3 ppbv), we would expect negligible background levels of NOz. Overall, shipping was an important source of NOx in the Arabian Gulf, both through direct emissions and via photolysis of ship-related HONO.