A high-flow inlet (stainless steel tubing, 0.2 m diameter, 5.5 m tall, and 3 m above the top of the containers and the front deck) was installed at the front of the ship where the laboratory containers were located. A high flow of air (approximately 10 m3 min-1) was drawn through the inlet, which provided a common attachment point for sub-sampling lines for all gas-phase measurement instruments. An air flow of 5 standard L min-1 for the first leg and 3.5 standard L min-1 for the second leg was pumped into the onboard lab container through an 1/2′′ (O.D. = 1.27 cm) FEP (fluorinated ethylene propylene) tubing (about 10 m long) insulated and heated to 50–60 ∘C. A PTFE (polytetrafluoroethylene) filter was placed at the beginning of the inlet to prevent insects, dust, and particles from entering the instruments. Every 2–5 d, the filter was replaced depending on the degree of pollution encountered. Inside the volatile organic compound (VOC) instrument container, the PTR-ToF-MS (8000, Ionicon Analytik GmbH Innsbruck, Austria) sampled a sub-flow at 80–100 sccm through 1/8′′ (0.3175 cm) FEP tubing (∼10 m in length, insulated and heated to 60 ∘C) from the main fast air flow and then to the instrument's PEEK (polyether ether ketone) inlet which was likewise heated to 60 ∘C. The inlet system was shared with total OH reactivity measurement (Pfannerstill et al., 2019).

The working principle of PTR-MS has been described in detail in previous studies (Lindinger et al., 1998; Ellis and Mayhew, 2013; Yuan et al., 2017). In brief, H3O+ primary ions are generated in the ion source and then drawn into the drift tube where they interact with sampled ambient air. Inside the drift tube, VOCs with a proton affinity greater than that of H2O (691 kJ mol-1) are protonated by proton transfer from H3O+. The resulting secondary ions are transferred to the detector, in this case a time-of-flight mass spectrometer with mass resolution around 3500 for the first leg and 4500 for the second leg at mass 96 amu. An internal standard of trichlorobenzene (C6H3Cl3) was continuously introduced into the instrument to ensure accurate mass calibration. Every minute a spectrum with mass range (m/z) 0–450 was generated. The data reported in this study are all at 1 min resolution unless otherwise specified.

The instrument background was determined every 3 h for 10 min with synthetic air. Four-point calibrations were performed five times during the whole campaign using a standard gas mixture (Apel-Riemer Environmental inc., Broomfield, USA) containing 14 compounds (methanol, acetonitrile, acetaldehyde, acetone, dimethyl sulfide, isoprene, methyl vinyl ketone, methacrolein, methyl ethyl ketone, benzene, toluene, xylene, 1,3,5-trimethylbenzene, and α-pinene). It has been previously reported that the sensitivities of some compounds measured by PTR-MS are humidity dependent (de Gouw and Warneke, 2007). As the relative humidity (RH) was expected to be high and varying (marine boundary layer with occasional desert air influence), humidity calibration was combined with four-point calibration by humidifying the gas mixture at different levels from 0 % to 100 % RH.

The data were initially processed by the PTR Analyzer software (Müller et al., 2013) to identify and integrate the peaks. After obtaining the raw data (counts per second for each mass identified), a custom-developed python-based program was used to further process the data to final mixing ratios. For compounds present in the standard gas cylinder, interpolated sensitivities based on the five in-campaign calibrations were applied to derive the mixing ratios, while mixing ratios of the other masses were calculated by using a proton transfer reaction rate constant (kPTR) of 2.0×10-9 cm3 s-1. The uncertainty associated with the mixing ratios of the calibrated compounds was around 6 %–17 % (see Table S1). For the mixing ratios derived by assuming kPTR, the accuracy was around ±50 % (Zhao and Zhang, 2004). The detection limit (LOD) was calculated from the background measurement with 3 times the standard deviation (3σ), 52±26 ppt for acetaldehyde, 22±9 ppt for acetone, and 9±6 ppt for methyl ethyl ketone (MEK) (Table S1 in the Supplement). Data below LOD were kept as determined for further statistical analysis (Fig. 2 and Table 1).

In this study, we have interpreted ion masses with the exact masses corresponding to CnH2nO, CnH2n-2O, and CnH2n-8O as aliphatic, unsaturated, and aromatic carbonyls, respectively (see the exact protonated m/z in Table S2). Carbonyl compounds with a carbon number of three and above can be either aldehydes or ketones, which are not distinguishable with PTR-ToF-MS using H3O+ as the primary ion. However, laboratory experiments have shown that protonated aldehydic ions with carbon atoms more than three tend to lose a H2O molecule and fragment to other masses (Buhr et al., 2002; Spanel et al., 2002). Moreover, although both ketones and aldehydes can be produced via atmospheric oxidation processes, ketones tend to have longer atmospheric lifetimes and higher photochemical yields than aldehydes, as mentioned in the introduction. The ratio of measured propanal to acetone was 0.07 in the western Pacific coastal region (Schlundt et al., 2017), 0.06 in urban Los Angeles (Borbon et al., 2013), and 0.17–0.22 in oil and gas production regions (summarized by Koss et al., 2017). Therefore, signals on the exact mass of carbonyl compounds from the PTR-ToF-MS are expected to be dominated by ketones, particularly in regions remote from the sources.

Relatively high mean mixing ratios of aliphatic carbonyls were observed over the Arabian Gulf, the highest being acetone (C3 carbonyl compound) at 4.50±2.40 ppb (median: 3.77 ppb), followed by formaldehyde at 3.83±2.55 ppb (median: 3.02 ppb), acetaldehyde at 1.73±1.61 ppb (median: 1.02 ppb), and MEK (C4 carbonyl compound) at 0.87±0.71 ppb (median: 0.56 ppb). As the Arabian Gulf is highly impacted by the oil and gas industry, we compared the measurements of the four aforementioned carbonyl compounds with those measured in the oil and gas region (Table 2). Except for formaldehyde, acetaldehyde, acetone, and MEK were lower than the mixing ratios measured in the Uintah Basin, which was influenced by intensive oil and natural gas activities (Koss et al., 2015). The general distribution of the aliphatic carbonyls in the Uintah Basin is similar to the Arabian Gulf, with acetone levels being approximately twice those of acetaldehyde. The carbonyl mixing ratios in the Arabian Gulf were comparable to those measured in Hickory (PA, USA) surrounded by natural gas wells (Swarthout et al., 2015). Koss et al. (2017) reported the maximum boundary layer enhancement of carbonyl compounds (C2–C7) measured during an aircraft measurement above the most productive oil field in the United States (Permian Basin). Within the boundary layer of the Permian Basin, C5–C7 aliphatic carbonyls had mixing ratios of 0.34, 0.08, and 0.03 ppb, which are of the same magnitude but lower than the levels measured over the Arabian Gulf for C5 (0.52±0.48 ppb), C6 (0.19±0.25 ppb), and C7 (0.04±0.04 ppb) carbonyl compounds. The sources of the major carbonyls in the Arabian Gulf will be discussed in detail in Sect. 3.1.2 and 3.4.3.

In contrast, aliphatic carbonyls had much lower average mixing ratios over the Arabian Sea and the Gulf of Aden, especially for C7–C9 carbonyls with mean mixing ratios below the detection limit for most of the time. During the summertime AQABA campaign, the prevailing wind direction over the Arabian Sea was south-west (Fig. S1 in the Supplement). Four-day back trajectories indicate the air was transported from the Arabian Sea (north-western Indian Ocean), passing the eastern Africa coast, which brought relatively clean, photochemically aged air masses (Bourtsoukidis et al., 2019). The mean level of acetone over the Arabian Sea (0.43±0.18 ppb, median: 0.34 ppb) is close to the level measured in the marine boundary layer of the western Indian Ocean (0.49 ppb) (Warneke and de Gouw, 2001) and comparable to other reported values from open-sea air measurement (see Table 2). Acetaldehyde was measured at relatively low mixing ratios over the Arabian Sea (0.13±0.12 ppb, median: 0.09 ppb), which is comparable than the levels reported by the measurements done in the Northern Hemisphere open ocean (see Table 2). Over the Gulf of Aden, acetaldehyde, acetone, and MEK had slightly higher mixing ratios than those over the Arabian Sea.

The Mediterranean Sea had somewhat higher levels of aliphatic carbonyls than the clean regions (the Arabian Sea and the Gulf of Aden) but with acetone (above 2 ppb) dominating the distribution. A much higher acetone level than the acetaldehyde level was also observed for some coastal site measurements, which were impacted by continental air (White et al., 2008; Schlundt et al., 2017; see Table 2). Larger aliphatic carbonyls (C6–C9) were below the detection limit most of the time. The aliphatic carbonyl levels over the Gulf of Oman were higher than the clean regions, while C1–C5 carbonyls were more variable over the Gulf of Oman compared to those over the Mediterranean Sea. This is probably because the Gulf of Oman connects to the Arabian Gulf, where intense oil and gas industrial activities are located. Over the Gulf of Oman, polluted air from the nearby sources of the Arabian Gulf is occasionally mixed with the clean air from the open sea (the Arabian Sea) under south-easterly wind conditions (Fig. S1).

Another region where abundant aliphatic carbonyls were observed was the Suez region. The air in this region was mainly influenced by nearby cities and marine transportation (ship emissions within the Suez Canal) (Bourtsoukidis et al., 2019; Pfannerstill et al., 2019). Therefore abundant precursors were available in the Suez region, producing more carbonyls regionally, especially for shorter-lived compounds (formaldehyde and acetaldehyde). Besides the local-scale emissions and photochemical production contribution to the carbonyls over the Suez, the longer-lived carbonyls (e.g. acetone) could also be transported from the Mediterranean Sea (where acetone was high). Four-day back trajectories indicate the air reaching the Suez region mostly originated from the European continent, passing over the Mediterranean Sea (Bourtsoukidis et al., 2019). Meanwhile, ocean uptake of acetone from the air due to polluted continental outflow (Marandino et al., 2005) as well as dilution and mixing with free tropospheric air during transport can modulate acetone mixing ratios. Although the mean mixing ratios of aliphatic carbonyls over the Suez were much lower than those over the Arabian Gulf, the variations were still more significant than other regions (not including the Arabian Gulf; see Table 1).

Over the Red Sea, acetone was the most abundant aliphatic carbonyl, followed by formaldehyde and acetaldehyde. The mixing ratios of acetaldehyde and acetone over the northern part of the Red Sea were similar to those levels measured in western Pacific coastal regions (South China Sea, Table 2). It is worth noticing that the levels of aliphatic carbonyls in the northern part of the Red Sea were almost 2 times higher than the southern part of the Red Sea. According to the 4 d back trajectories reported by Bourtsoukidis et al. (2019), the measured air masses that travelled to the northern part were from southern Europe and north-eastern Africa, while the southern part was more influenced by air from the northern part of the Red Sea mixed with the air masses from desertic areas of central Africa. Therefore, fewer primary precursors as well as carbonyls were transported to the southern part of the Red Sea compared to the northern part. Moreover, the unexpected sources of hydrocarbons (ethane and propane) from northern Red Sea deep water reported by Bourtsoukidis et al. (2020) would lead to higher carbonyl levels in the northern part compared with the southern part due to the additional precursors in the Red Sea North. However, acetaldehyde was still found to be significantly underestimated compared to the model results, even taking the deep-water source into consideration (Sect. 3.3). This indicates that extra sources of acetaldehyde may exist, which will be discussed in detail in Sect. 3.4.

The primary emission sources in the Arabian Gulf and Suez regions are quite different. While the Arabian Gulf is dominated by oil and gas operations, the Suez is more influenced by ship emissions and urban areas (Bourtsoukidis et al., 2019). Carbonyl compounds were most abundant in these two areas. For further insight, we focused on a time series of selected trace gases and their inter-correlations to better identify the sources of the major aliphatic carbonyls. Meanwhile, we calculated the OH exposure (OHΔt) based on hydrocarbon ratios (Roberts et al., 1984; de Gouw et al., 2005; Yuan et al., 2012) for the polluted regions (Arabian Gulf and Suez) where primary emissions have been identified (Bourtsoukidis et al., 2019; Bourtsoukidis et al., 2020), to better understand the photochemical aging of the major carbonyls using the following equation: 1OHΔt=1kX-kY⋅ln⁡XYt=0-ln⁡XY, where X and Y refer to two hydrocarbon compounds with different rates of reaction with the OH radical (k). For this study, we chose toluene (kOH+toluene: 5.63×10-12 cm3 molec.-1 s-1) and benzene (kOH+benzene: 1.22×10-12 cm3 molec.-1 s-1) (Atkinson and Arey, 2003), because both compounds were measured by PTR-ToF-MS at high frequency and these values showed good agreement with values measured by GC-FID (Fig. S2). The approach detailed by Yuan et al. (2012) was applied to determine the initial emission ratio XYt=0 in those two regions by only including nighttime data of benzene and toluene. We obtained initial emission ratios (toluene-to-benzene ratios) of 1.38 for the Arabian Gulf and 2.12 for the Suez region. Koss et al. (2017) summarized the toluene-to-benzene ratios observed in various locations and showed that urban and vehicle sources tend to have higher toluene-to-benzene ratios (mean ∼2.5) than the ratios of oil and gas sources (mean ∼1.2). Therefore, the toluene-to-benzene ratios obtained for those two regions agreed well with other studies done with similar emission sources. The corresponding correlation plots of toluene and benzene for those two regions can be found in Fig. S3.

Figure 3a shows the time series of acetaldehyde and acetone over the Arabian Gulf along with OH exposure (OHt) and ozone. We further separated the data into daytime and nighttime and calculated correlations among the carbonyls and other selected species (see Fig. 4b and c). Aliphatic carbonyls were well correlated with each other during the daytime, and ozone had a generally good correlation with C2–C7 carbonyls (r>0.7) during the daytime but a much lower correlation during the night, indicating ozone and carbonyls were co-produced via photochemical oxidation. Tadic et al. (2020) reported that the net ozone production rate over the Arabian Gulf (32 ppb d-1) was greatest over the Arabian Peninsula. They show that strong ozone-forming photochemistry occurred in this region, which would lead to abundant secondary photochemically produced products (including carbonyls). However, it should be noted that the good correlation between ozone and carbonyls could in part be due to carbonyls co-emitted with ozone precursors (hydrocarbons) as primary emissions. In Fig. 3a, the calculated OH exposure was high during the first night in leg 1, where an elevation of the acetone mixing ratio was observed, while the mixing ratio of acetaldehyde remained relatively constant. With limited OH radical abundance during the nighttime, the increased OH exposure indicates that the air reaching the ship was photochemically processed (aged). Therefore, the increase in acetone was mainly from long-distance transport as acetone has a much longer atmospheric lifetime than acetaldehyde. As the ship approached Kuwait, the calculated OH exposure was low (starting from 30 July 2017, 00:00 UTC), which is an indicator of nearby emission sources. The lifetime of the OH radical derived from the measured OH reactivity also decreased from ∼0.1 to ∼0.04 s during the same period (Pfannerstill et al., 2019). Oil fields and associated refineries are densely distributed in the north-west of the Arabian Gulf region (United States Central Intelligence Agency). The air reaching the ship when mixing ratios of acetone and acetaldehyde were highest was mainly from the north-west (Iraq oil field region) according to the back trajectories (Bourtsoukidis et al., 2019). This suggests that the air masses encountered in the north-western Arabian Gulf were a combination of fresh emissions from nearby sources and photochemically processed air transported from elsewhere. During the second leg, relatively low mixing ratios were identified in the same region (north-western Arabian Gulf), which was mainly due to a greater influence of air masses originating from less populated desert regions of north-eastern Iran (Bourtsoukidis et al., 2019), with much less influence from the oil field emissions, meaning fewer precursors were available for carbonyl production. Several plumes (extending over 2–3 h) of elevated carbonyls with increased ozone were observed during the nighttime for both legs (Fig. 4a), indicating transport of highly polluted air.

For the Suez region (Gulf of Suez and Suez Canal), data were only available for the second leg. A significant increase in acetonitrile (over 400 ppt) was observed just before entering the Great Bitter Lake (see Fig. 4a), indicating an increasing influence of biomass burning on the air composition (Lobert et al., 1990). Carbonyl compounds are important primary emissions in fresh biomass burning plumes (Holzinger et al., 1999, 2001; Schauer et al., 2001; Koss et al., 2018) as well as being formed as secondary products in more aged plumes (Holzinger et al., 2005). We further investigated the correlation coefficient among carbonyls during the biomass burning plume (Fig. 4b) in the Suez. Carbonyls had a high correlation with acetonitrile, benzene, and themselves, particularly for smaller carbonyls (acetaldehyde, C3–C5 carbonyls). The biomass burning emissions were probably transported by the prevailing northerly wind (Fig. S1) above north-eastern Egypt, where crop residues, especially rice straw, are often directly burned in the open fields (Abdelhady et al., 2014; Said et al., 2013; Youssef et al., 2009). Besides the direct biomass burning emission, the high mixing ratios and the good correlations of carbonyls could also have resulted from other sources such as hydrocarbons (alkanes, alkenes, and aromatics) which were elevated at the same time. Similar to conditions identified over the Arabian Gulf, elevated OH exposure accompanied by an increasing acetone mixing ratio was observed during the first night over the Gulf of Suez, indicating aged air-mass transportation. The OH exposure was then significantly lower during the daytime, when mixing ratios of carbonyls and alkanes increased as well. This indicates the presence of emission sources nearby. Oil refineries located on the coastal side of the Suez and oil tank terminals located in the northern part of the Gulf of Suez are likely sources.

The mixing ratios of unsaturated carbonyls were generally ∼10 ppt or lower than the LOD over the Mediterranean Sea and the clean regions (the Arabian Sea and the Gulf of Aden). The Red Sea region and the Gulf of Oman had slightly higher levels (LOD–40 ppt). The highest values were again observed in the Arabian Gulf (20–110 ppt), followed by the Suez (LOD–60 ppt). The numbers represent the range of the mean mixing ratios of unsaturated carbonyls in each region. In terms of the mixing ratio distribution (Fig. 2), the peak value was usually observed at C5 or C6 unsaturated carbonyls over most regions except for the Suez, where C4 carbonyl had the highest mixing ratio. Based on chemical formulas, unsaturated carbonyls can be either cyclic carbonyl compounds or carbonyls containing a carbon–carbon double bond. Therefore, the air chemistry could differ considerably depending on the compound assignment. A detailed analysis of the chemistry of the unsaturated carbonyls measured will be given in the following Sect. 3.2.2.

Regional variability was also observed for aromatic carbonyls, with the highest levels observed over the Arabian Gulf and Suez and much lower mixing ratios over the Arabian Sea, Mediterranean Sea, and Gulf of Aden (Table 1). Several studies using PTR-MS have reported values for m/z 107.049 (C7 aromatic carbonyls) attributed to benzaldehyde (Brilli et al., 2014; Koss et al., 2017, 2018), m/z 121.065 (C8 aromatic carbonyls) attributed to tolualdehyde (Koss et al., 2018) or acetophenone (Brilli et al., 2014), and m/z 135.080 (C9 aromatic carbonyls) attributed to methyl acetophenone (Koss et al., 2018) or benzyl methyl ketone (Brilli et al., 2014) or 3,5-dimethylbenzaldehyde (Müller et al., 2012). Atmospheric aromatic carbonyls are produced via photochemical oxidation of aromatic hydrocarbons (Finlayson-Pitts and Pitts, 1999; Wyche et al., 2009; Müller et al., 2012), and benzaldehyde was reported as having primary sources from biomass burning and anthropogenic emissions (Cabrera-Perez et al., 2016). Around the Arabian Peninsula, the level of aromatic carbonyls declined with increasing carbon number over most of the regions except in the Red Sea South, Gulf of Oman, and Arabian Gulf, where C7 carbonyls were comparable to C8 carbonyls (Fig. 2). Interestingly, only in the Suez region were the C7 aromatic carbonyls more abundant than other aromatic carbonyls, whereby the mean value (90±200 ppt) was much higher than the median value (20 ppt), indicating strong primary sources of benzaldehyde in the Suez. Otherwise, toluene was found to be more abundant over the Suez, with mean mixing ratios of 271±459 ppt than over other regions (the mean over the Arabian Gulf: 130±160 ppt), which would also lead to higher benzaldehyde as it is one of the OH-induced oxidation products of toluene via H abstraction (Ji et al., 2017).

Unsaturated carbonyls measured by PTR-MS have only rarely been reported in the atmosphere, with the exception of methyl vinyl ketone and methacrolein (C4 carbonyls), which are frequently reported as the oxidation products of isoprene (Williams et al., 2001; Fan and Zhang, 2004; Wennberg et al., 2018). According to the GC-FID measurement, isoprene was below the detection limit for most of the time during the AQABA cruise, with the highest values observed in the Suez (10–350 ppt). This shows that the AQABA campaign was little influenced by either terrestrial or marine isoprene emissions. However, we observed unexpected high levels on mass 69.070, which is usually interpreted as isoprene for PTR-MS measurements. Significant enhancements were even identified while sampling our own ship exhaust (in PTR-MS but not GC-FID), suggesting the presence of an anthropogenic interference at that mass under these extremely polluted conditions. Several studies have reported possible fragmentations of cyclic alkanes giving mass (m/z) 69.070. These include a laboratory study on gasoline hydrocarbon measurements by PTR-MS (Gueneron et al., 2015), a GC-PTR-MS study of an oil spill site combined with analysis of crude oil samples (Yuan et al., 2014), and an inter-comparison of PTR-MS and GC in an O&G industrial site (Warneke et al., 2014). From those studies, other fragmentations from C5 to C9 cycloalkanes, including m/z 43, m/z 57, m/z 83, m/z 111, and m/z 125, were identified together with m/z 69. Cyclic alkanes were directly measured in oil and gas fields (Simpson et al., 2010; Gilman et al., 2013; Li et al., 2017; Aklilu et al., 2018), vehicle exhaust (Gentner et al., 2012; Erickson et al., 2014), and vessel exhaust (Xiao et al., 2018), accounting for a non-negligible amount of the total VOC mass depending on the fuel type. Koss et al. (2017) reported enhancement of cyclic alkane fragment signals and increased levels of unsaturated carbonyls measured by PTR-ToF-MS over the O&G region in the US. The unsaturated carbonyls (C5–C9) were assigned as oxidation products of cycloalkanes. Therefore, we examined the correlations between m/z 69.070 and other cycloalkane fragments over the Arabian Gulf and Suez, where anthropogenic primary emissions were significant. As shown in Fig. 5, m/z 83 was the most abundant fragment, and it correlated better with m/z 69 than the other two masses, strongly supporting the presence of C6 cycloalkanes (methylcyclopentane and cyclohexane). The other two masses are distributed in two or three clusters, suggesting compositions of different cycloalkanes. m/z 43 and m/z 57 (fragments of C5 cycloalkanes) had lower correlations with other fragments (not shown in the graph) as they are also fragments of other higher hydrocarbons. Thereby we could assign those unsaturated carbonyls as photochemical oxidation products (i.e. cyclic ketones or aldehydes) from their precursor cycloalkanes.

As shown in Fig. 2 and Table 1, C6 unsaturated carbonyls displayed higher mixing ratios than any other unsaturated carbonyls over the Arabian Gulf, while C5 unsaturated carbonyl was slightly higher than C6 in the Suez. Bourtsoukidis et al. (2019) derived enhancement ratio slopes from pentane isomers and established that the Arabian Gulf is dominated by oil and gas operations and that the Suez is more influenced by ship emissions. Therefore, as the Arabian Gulf had much more active O&G activities than the Suez, our findings agree with Koss et al. (2017), who showed that C6 unsaturated carbonyls should be more abundant than C5 carbonyls since more precursors for C6 unsaturated carbonyls are emitted from active oil fields. It is worth mentioning that in Fig. 5b one cluster at the bottom showed m/z 69.070 had no correlation with the other three masses. Those points correspond to the time when the GC measured significantly elevated isoprene while passing through the narrow Suez Canal where some vegetation (e.g. palms and some agriculture) was present close to the shore, meaning m/z 69.070 during this period was isoprene. At the same time, m/z 71.049 (C4 unsaturated carbonyl) increased from 20 to 220 ppt. Isoprene oxidation products (MVK and methacrolein) were probably the major contribution to the C4 unsaturated carbonyls in this period. This also explains why C4 carbonyl dominated the distribution of unsaturated carbonyls over the Suez.

In the other regions (especially more remote areas), the cyclic alkane fragmentation masses had much lower abundance, leading to much less unsaturated carbonyls due to lack of precursors. Meanwhile, m/z 69.070 (C5H8H+), m/z 83.086 (C6H10H+), and m/z 97.101 (C7H12H+) could also be fragmentations from corresponding aldehydes losing one water molecule as mentioned in Sect. 2.3.3. Missing information on the chemical structure of unsaturated carbonyls and knowledge of their precursors preclude detailed investigation of the sources of large unsaturated carbonyls in these areas.

Northway et al. (2004) and Apel et al. (2008) reported that heterogeneous reactions of unsaturated organic species with ozone on the wall of the Teflon inlet can cause artifact signals of acetaldehyde but not of acetone. During AQABA, the highest and most variable ozone mixing ratios were observed during the campaign over the Arabian Gulf (mean: 80±34 ppb) and the Red Sea North (66±12 ppb), where a modest correlation was found between acetaldehyde and ozone over the Arabian Gulf (r2=0.54) and no significant correlation over the Red Sea North (r2=0.40). However, larger correlation coefficients were identified between ozone and other carbonyls over the Arabian Gulf (see Fig. S5), which suggests that the correlation was due to atmospheric photochemical production rather than artifacts. Moreover, acetaldehyde was found to have a much worse correlation with ozone during the nighttime compared to the correlation during the daytime over the Arabian Gulf (Fig. 3b and c), which also indicates that inlet generation of acetaldehyde was insignificant. Over other regions, especially the remote area (the Arabian Sea and Gulf of Aden), ozone was relatively constant and low, with poor correlation with acetaldehyde mixing ratios. Although we cannot completely exclude the possible existence of artifacts, the interference is likely to be insignificant in this dataset.

A bias between measured acetaldehyde and global model simulations has been observed in previous studies conducted in the remote troposphere (Singh et al., 2003; Singh, 2004; Wang et al., 2019) and in the marine boundary layer (Read et al., 2012). The aforementioned studies emphasized the potential importance of the seawater acting as a source of acetaldehyde emission via air–sea exchange. No significant correlation was found between acetaldehyde and DMS (dimethyl sulfide), a marker of marine biogenic emission which is produced by phytoplankton in seawater (Bates et al., 1992) (see Fig. S6). This indicates that the direct biogenic acetaldehyde emissions from the ocean are probably insufficient to explain the measured acetaldehyde. More likely, acetaldehyde and other small carbonyl compounds can be formed in the sea, especially in the surface microlayer (SML) via photodegradation of coloured dissolved organic matter (CDOM) (Kieber et al., 1990; Zhou and Mopper, 1997; Ciuraru et al., 2015). Zhou and Mopper (1997) calculated the exchange direction of small carbonyls based on measurement results and identified that the net flux of acetaldehyde was from the sea to the air, whereas formaldehyde was taken up by the sea. Sinha et al. (2007) characterized air–sea flux of several VOCs in a mesocosm experiment and found that acetaldehyde emissions were in close correlation with light intensity (r=0.7). By using a 3-D model, Millet et al. (2010) estimated the net oceanic emission of acetaldehyde to be as high as 57 Tg a-1 (in a global total budget: 213 Tg a-1), being the second largest global source. A similar approach was applied in a recent study done by Wang et al. (2019), reporting the upper limit of the net ocean emission of acetaldehyde to be 34 Tg a-1. Yang et al. (2014) quantified the air–sea fluxes of several OVOCs (oxygenated volatile organic compounds) over the Atlantic Ocean by eddy covariance measurements, showing the ocean is a net source of acetaldehyde. Although Schlundt et al. (2017) reported uptake of acetaldehyde by the ocean from measurement-inferred fluxes in western Pacific coastal regions, to our knowledge, there is no direct experimental evidence showing the ocean to be a sink for acetaldehyde.

In order to test the importance of the oceanic emission of acetaldehyde, we implemented this source in the EMAC model. The measured seawater concentration of acetaldehyde was not available for the water area around the Arabian Peninsula. Wang et al. (2019) estimated the global average acetaldehyde surface seawater concentrations of the ocean mixed layer using a satellite-based approach similar to Millet et al. (2010), where the model estimation agreed well with limited reported measurements. From the Wang et al. (2019) results, the averaged seawater concentration of acetaldehyde around the Arabian Peninsula was generally much higher from June to August. As the photodegradation of CDOM is highly dependent on sunlight, the air–sea submodel (Pozzer et al., 2006) was augmented to include throughout the campaign a scaled acetaldehyde seawater concentration in the range of 0 ∼50 nM according to the solar radiation (Fig. S7). With this approach, the average of acetaldehyde seawater concentration estimated by the model is 13.4 nM, a reasonable level compared to the predicted level by Wang et al. (2019).

After adding the oceanic source of acetaldehyde, the model estimation was significantly improved (Fig. 7). As the oceanic source in the model is scaled according to the solar radiation, the measurement-to-model ratios were more strongly reduced during the day compared to the night. With oceanic emission included, the model underestimation was less significant, within a factor of 3 during the day and 4 during the night over the Mediterranean Sea, Red Sea, and Gulf of Aden. The most significant improvement was identified over the Red Sea North. As shown in Fig. 8, the model had much better agreement with the measurement after adding the oceanic source. The scatter plots for other regions can be found in Fig. S8. Over the Arabian Sea, the model significantly overestimated acetaldehyde mixing ratios, indicating the input seawater concentration of acetaldehyde might be too high. The SML layer starts to be effectively destroyed by the wave breaking when the wind speed exceeds 8 m s-1 (Gantt et al., 2011). As the average wind speed over the Arabian Sea was highest among the cruised areas (8.1±2.4 m s-1, Fig. S1), less contribution from the CDOM photodegradation to acetaldehyde in the surface seawater would be expected. For the Suez region, due to the limited model resolution (1.1∘×1.1∘), little seawater was identified in the model, leading to negligible influence from the oceanic source.

Model underestimation of acetaldehyde, especially over the Suez, Red Sea, and Arabian Gulf, is also likely to be related to the coarse model resolution (∼1.1∘×1.1∘) (Fischer et al., 2015). Where model grid points contain areas of land, the higher and more variable terrestrial boundary layer height impacts the model prediction, whereas the measurements may only by influenced by a shallower and more stable marine boundary layer.

Over the Arabian Gulf and Suez, the intensive photochemical production of carbonyls is apparent. Bourtsoukidis et al. (2020) compared measured hydrocarbons (ethane, propane, and butane) with the results from model simulations (the same model used in this study with the newly discovered deep water source implemented). The model was able to reproduce the measurement over most regions expect for some significant model underestimations in the Suez and Arabian Gulf, in which local and small-scale emissions were difficult for the model to capture. Therefore, an underestimation of the precursor hydrocarbons, as well as those large alkanes, alkenes, and cyclic hydrocarbons which were not measured (> C8) or included in the model (> C5), could be a reason for the model underestimation of acetaldehyde, especially in polluted regions. In addition, as mentioned in the previous case studies, high-ozone mixing ratios were observed over the Arabian Gulf, especially during the nighttime. Ethene and propene were found to be significantly underestimated during the nighttime high-ozone period by a factor over 10 (Fig. S9), which indicates that the nighttime ozonolysis of alkenes could be another important source of acetaldehyde, formaldehyde, and other carbonyls (Atkinson et al., 1995; Altshuller, 1993) in the Arabian Gulf.

Acetaldehyde, an oxygenated VOC, is not generally considered an important primary emission from oil and gas fields, but instead a photochemical product of hydrocarbon oxidation (Yuan et al., 2014; Koss et al., 2015, 2017). In contrast, primary sources of formaldehyde from oil and gas production processes, including both combustion and non-combustion processes, have been ascertained (Vaught, 1991). Le Baron and Stoeckenius (2015) concluded in their report on the Uinta Basin winter ozone study that besides formaldehyde, the other carbonyls were poorly understood in terms of their primary sources. Acetaldehyde and other carbonyls (aldehydes and ketones) have been reported as primary emissions from fossil fuel combustion, including ship emissions (Reda et al., 2014; Xiao et al., 2018; Huang et al., 2018) and vehicle emissions (Nogueira et al., 2014; Erickson et al., 2014; Dong et al., 2014). A possible explanation for the measurement–model discrepancy is that the active petroleum industry located in the Arabian Gulf and intensive marine transportation in the Suez are primary sources of acetaldehyde and other carbonyls which were not well constrained in the model. The Suez region, where the largest acetaldehyde discrepancy was identified, had a significant influence from biomass burning (see Sect. 3.2.2). Biomass burning emissions are notoriously difficult to model as they are highly variable in both time and space. In this study, the model failed to reproduce the acetonitrile level, with a range of only 40–50 ppt rather than 100–550 ppt measured over the Suez. Thus, besides the possibility of seawater emission from the Gulf of Suez and the Suez Canal, the underestimated biomass burning source in the model over the Suez will lead to an underestimation of acetaldehyde as well as other carbonyl compounds in this region.

Although the model estimation was generally improved with the addition of an oceanic source, the model to measured ratios still varied over a wide range. As mentioned above, photodegradation of CDOM on the surface of seawater is a known source of acetaldehyde, although some studies focusing on real seawater samples did not observe clear diel cycles of seawater acetaldehyde (Beale et al., 2013; Yang et al., 2014). Fast microbial oxidation could be a reason (Dixon et al., 2013), while other non-light-driven sources of acetaldehyde could be an alternative explanation. In a recent study, Zhou et al. (2014) reported enhanced gas-phase carbonyl compounds including acetaldehyde during a laboratory experiment of ozone reacting with SML samples, indicating acetaldehyde could also be produced under non-light-driven heterogeneous oxidation. Wang et al. (2019) ventured a hypothetical source that organic aerosol can be an extra source of unattributed acetaldehyde in the free troposphere through light-driven production and ozonolysis. However, since the yield of acetaldehyde from such reactions is unknown, large uncertainties remain. Previous studies have shown that the organic matter fraction was highest in smaller sea spray aerosols and that the aerosols contain both saturated and unsaturated fatty acids originating from the seawater surface (i.e. SML) (Mochida et al., 2002; Cochran et al., 2016). Thus, for the AQABA campaign, both photodegradation and heterogeneous oxidation could occur on the surface of sea spray and pollution-associated aerosols, even over the remote open ocean, therefore being an extra source of acetaldehyde and other carbonyl compounds.

Another acetaldehyde formation pathway reported is gas-phase photolysis of pyruvic acid (Eger et al., 2020; Reed Harris et al., 2016), a compound mainly of biogenic origin. Pyruvic acid has been also observed in seawater (Kieber and Mopper, 1987; Zhou and Mopper, 1997) and was found up to 50 nM in the surface water of the eastern Pacific Ocean (Steinberg and Bada, 1984), while acetaldehyde was not the major product of aqueous-phase photolysis of pyruvic acid (Griffith et al., 2013). Zhou and Mopper (1997) pointed out that the net exchange direction for pyruvic acid is expected to be from the air to the sea due to high solubility, with a Henry's law constant of 3.1×103 mol m-3 Pa-1 (Sander, 2015). Moreover, partitioning to aerosols could be an important sink for pyruvic acid (Reed et al., 2014; Griffith et al., 2013): an increasing concentration trend of pyruvic acid was observed in marine aerosols over the western North Pacific Ocean (Boreddy et al., 2017). Therefore, due to limited terrestrial biogenic sources of pyruvic acid for the AQABA campaign, the gas-phase level of pyruvic acid was expected to be low. Limited studies reported pyruvic acid level in the marine boundary layer and Baboukas et al. (2000) measured 1.1±1.0 ppt of pyruvic acid above the Atlantic Ocean. Pyruvic acid was measured by Jardine et al. (2010) using a PTR-MS at m/z 89 in a forested environment. For the AQABA PTR-ToF-MS dataset, enhanced signals were observed at m/z 89.024, with a mean mixing ratio of 35–110 ppt over different regions (Table S4), which is much more abundant than reported pyruvic acid levels by Baboukas et al. (2000). This might be due to the uncertainty associated with the theoretical methods of quantification used here or the presence of isomeric compounds on that mass, since pyruvic acid was not calibrated with the standard. Even if we assume the m/z 89.024 to be entirely pyruvic acid, with a 60 % yield of acetaldehyde via photolysis (IUPAC, 2019), it gave a maximum of 13 ppt of acetaldehyde over the Arabian Gulf and 5–9 ppt over other regions, which were only 0.8 %–6 % of the mean mixing ratios (Table S4). Detailed information on the calculation can be found in the Supplement. Therefore, we conclude that the contribution from the photolysis of pyruvic acid is not an important source of the unattributed acetaldehyde during the AQABA campaign.