In this study, besides NO2 conversion onto clouds and aqueous particles (Reaction R4), the losses of HONO onto the ice and liquid clouds (Reaction R6) are also included. Therefore, for accurate simulations of HRs, we need to examine the cloud distribution. The CHASER model applied the common cloud maximum–random overlap assumptions (MRAN) in the radiation and cloud microphysics schemes as other general circulation models to estimate the distribution of the cloud fraction. The verification by using the satellite observation data ISCCP D2, CALIPSO-GOCCP, and reanalysis data JRA55 generally revealed good correlation, whereas notable (10 %–20 %) underestimation for the entire troposphere was yet salient. During June–July–August (JJA), CHASER's cloud fraction was likely overestimated for the lower troposphere of the North Pacific (NP) region (10 %–20 % compared to JRA55 reanalysis data). This finding indicated that thorough scrutiny of any impacts in this region is highly required (see the discussion in Sect. 3.2). Note that more detailed information for cloud verification for CHASER has been provided by Ha et al. (2021).

The heterogeneous processes by clouds and aerosol particles were parameterized by using surface area density (SAD) estimations alongside the cloud fraction and aerosol concentration. During DJF, the simulated total SAD was attributed to all types of aerosols. However, for JJA, liquid clouds and sulfate aerosols were the principal SAD sources. This was a peculiarly visible pattern for the northern polar and mid-latitude maritime regions. The performance for aerosol SAD in our model was in line with the earlier report by Thornton et al. (2008), except for sea salt density, which was very low in our model (up to 2 µm2 cm-3) compared to their work (up to 75 µm2 cm-3). This disagreement might be ascribed to the two models' different size distributions for sea salt. The calculated SAD for the liquid cloud was 2 orders of magnitude higher than SAD for ice cloud and total aerosols. Liquid cloud SAD maximized at ∼800 hPa in the tropical convective systems and over the mid-latitude storm tracks, reaching ∼50000µm2 cm-3 at the surface of the North Pacific region in JJA. Sulfate aerosols dominated above 600 hPa for the Northern Hemisphere (∼ 20 µm2 cm-3) among the total aerosol surface area, followed by organic carbons and soil dust (∼10µm2 cm-3 in JJA). At the surface layer, sulfate aerosols were prevalent in DJF for the Chinese region (>1000µm2 cm-3), the northeastern US (∼500µm2 cm-3), and North Pacific region in JJA (∼250µm2 cm-3). SAD for soil dust dominated in desert regions, with annual average values >100µm2 cm-3. Organic carbon (OC) was dominant in winter over biomass burning regions such as China (up to 1000µm2 cm-3) and South Africa (up to 800µm2 cm-3). For the Chinese region, SAD for black carbon (BC) could reach 600µm2 cm-3 in DJF and 75µm2 cm-3 in India. The total-aerosol SAD for the northern high-latitude and mid-latitude oceans was ∼75µm2 cm-3, consistent with the estimation by Thornton et al. (2008).

This section evaluated CHASER-based HONO estimates using the HONO measurements collected during the EMeRGe campaign off the coast of eastern Asia in spring 2018 (Andrés Hernández et al., 2021). This is the first global HONO modelling work using EMeRGe as the validation source. The HONO measurements in the free troposphere could provide essential information on the underlying gas-phase and heterogeneous HONO formation mechanisms as most current HONO measurements were conducted in the surface air. The daytime HONO concentration was retrieved from the aircraft-borne limb measurements using the HALO mini-DOAS (differential optical absorption spectroscopy) instrument, in which the absorbed UV light (310–440 nm) by HONO was detected (Hüneke et al., 2017). The mini-DOAS's measurement method relies on near-UV/VIS/IR skylight spectroscopy in nadir and limb geometry. Data evaluation consists of three steps: (1) retrieval of slant column densities (SCDs) of trace gases by the DOAS method (Platt and Stutz, 2008), (2) forward radiative transfer modelling for each measurement using McArtim (Deutschmann et al., 2011), and (3) retrieval of concentration through a new scaling method for UV/VIS data (Stutz et al., 2017; Hüneke et al., 2017; Werner et al., 2017; Kluge et al., 2020; Rotermund et al., 2021).

Additional sensitivity runs were conducted to explore potential HONO sources during the daytime (Table 5). The ratR4+CLD case is run in an attempt to produce more HONO from heterogeneous sources by altering the HONO:HNO3 yield ratio in Reaction (R4) to 0.9:0.1, and γliq. increased 100-fold (10-4 → 10-2). The main idea here is to evaluate whether the missing HONO source was sensitive to cloud uptake in this region or not. The maxST case maximized the uptake coefficients (γ values) of NO2 on organic and black carbons to 0.1 (Reactions R4, R5), to estimate the separate role of soot uptake under daytime conditions (George et al., 2005; Monge et al., 2010; Ndour et al., 2008), which could achieve an unrealistically high γ value of 10-1 (Ammann et al., 1998; Kalberer et al., 1999). In three other runs (JANO3-A, JANO3-B, JANO3-C), the photolysis of aerosol nitrate/adsorbed HNO3 on the ground and other surfaces (NO3-/HNO3) was examined, simply as HNO3 + hν → HONO (Reaction R7). These heterogeneous photolyses of HNO3 were previously proposed as potential HONO sources during the day (Lee et al., 2016; Scharko et al., 2014; Zhou et al., 2011). Because aerosol nitrate and aqueous surfaces are ubiquitous in the atmosphere, the photolysis (Reaction R7) was simply set for the gaseous HNO3 species to occur in particular model spatial grids exposing ground surfaces and sufficient surface area density for aerosols and clouds. The photolysis (Reaction R7) was taken at a rate 2 orders of magnitude faster than the gas-phase photolysis rate of HNO3 (HNO3 + hν → OH + NO2) (Zhou et al., 2011) and presumably yields 100 % HONO to access the maximum effects by this photolysis (Lee et al., 2016). This setting allows Reaction (R7) not only to occur at the surfaces of particles but also in the gas and bulk phases. However, in this test, Reaction (R7) generally refers to surface-catalysed photolysis or heterogeneous photolysis of HNO3. The JANO3-A case investigated the photolysis of adsorbed HNO3 on ground surfaces by implementing Reaction (R7) for the first vertical layer (z=1). The JANO3-B case explored photolysis of nitrate particles and adsorbed HNO3 gas on both ground surfaces and aerosol surfaces, applying Reaction (R7) for model grid cells with the SAD of 10-6–10-4 cm2 cm-3 (100 to 10 000 µm2 cm-3) to use the 10-4 cm2 cm-3 threshold to exclude cloud surfaces (Sect. 3.1.1). The JANO3-C case examined Reaction (R7) for regions present of all particles with SAD ≥10-7 cm2 cm-3 (10µm2 cm3). The SAD of 10-6 and 10-7 cm2 cm-3 was supposed to be the threshold for continental aerosols. The specifications of HONO chemistry in the maxST case and JANO3-B/JANO3-C cases were also combined in two additional cases (maxST+JANO3-B and maxST+JANO3-C, respectively), with the expectation that the opposite effects of the separate cases on NO2–O3–CO chemistry could compensate for each other in the combined cases; this is discussed in the following. Other tests examined the possible HONO sources from aviation crafts (AIRC), amplified emissions (EM×8), and amplified homogeneous HONO formation Reaction (R2) (GR×8), descriptions of which were listed in Table S3 in the Supplement.

The correlation coefficient (R) and model biases against EMeRGe for HONO are shown in Table S4. As seen for the STD run, general underestimations of HONO simulations were identified, in which better correlations were found at 1000–2000 m (R=0.31–0.49). Vertical profiles for HONO and other species (NO2, O3, CO), retrieved from the EMeRGe flights, were applied for the measurement-based model evaluation (Fig. 2). The model discrepancies for the measurement for HONO (ΔHONO) and NO2 (ΔNO2) in each flight trajectory, i.e. from Taiwan to South Korea, Japan, and the Philippines, were separated into bins of altitude ranges 0–1000–3000–5000–6000 m (Fig. 3). The frequency distributions of ΔHONO,ΔNO2,ΔO3, and ΔCO are shown in Figs. S7 and S8. Figure 2a shows the vertical average score (cruising altitudes ±500 m) for the measured (black) and simulated HONO concentrations in STD (reds) and those results of sensitivity cases. The measured daytime HONO concentration was close to the boundary layer (below 1000 m) over Taiwan, averaged at 115 ppt, and peaked at ∼250 ppt. Also, the HONO concentration decreased up to 9000 m (±500 m), with mean values dropping from 70 ppt (2000±500 m) to <20 ppt (5000±500 m) and <10 ppt above. These measured HONO values for this Asian coastal region were surprisingly high, which range from 10–115 ppt for 2000±500 m altitudes, compared to Wang's report of <100 ppt (maximum) and <30 ppt (4 daytime hour means) for 1500–2000 m altitudes measured by a MAX-DOAS at a station near the HONO source (Wang et al., 2019). This indicates that the source of HONO during EMeRGe might relate to mechanisms other than emission sources. In this study, the simulated HONO concentration in the STD case significantly underestimated the observations. They reached only 30–70 ppt at 1000 m and nearly zero from 2000 m upward (Fig. 2a: red versus black triangles for the simulation and the observations, respectively). These discrepancies indicate a significant unknown HONO source during the daytime, although the proposed heterogeneous HONO formation mechanisms were incorporated in our model. This finding adds another instance of evidence about missing HONO sources in the polluted boundary layer and free troposphere (e.g. Kleffmann et al., 2003; Li et al., 2014; VandenBoer et al., 2013; Xue et al., 2022a; Ye et al., 2018).

In Fig. 3, which shows model discrepancies, the measured NO2 below 3000 m (±500 m) close to land was well captured in the model (Fig. 3a, e, i: magenta and green), with 34 % of the data being quite close for NO2 (±70 ppt) (Fig. S7d). However, the modelling still underestimated the simulated HONO mixing ratio by up to 250 ppt (Fig. 3a: green). Over the region off the coast of Taiwan bound to Japan, NO2 was overestimated by up to 600 ppt in the model, corresponding to 20–70 ppt missing HONO (Fig. 3a, e: small orange area left of vertical line). The missing HONO can be driven by low HONO emission from land and low uptake of NO2 on organic carbon and soot, as the amplified EM×8 and maxST cases could alleviate the model underestimates for HONO (Fig. 3a vs. b: orange; Fig. 3a vs. c and e vs. f: green and magenta). The model also underestimated O3 and CO, usually by 25 ppb O3 (freq. 79 %) and 100 ppb CO (freq. 60 %) (Figs. S10, S11), which were larger than the model biases against ATom-1 observations (Sect. 3.1.3; Fig. 5) because of possible inland influence. A more accurate and detailed emission inventory for substances such as HONO, NOx, and CO was thus sensible as this region is the outflow of the Pearl River Delta and Yangtze River Delta regions. Besides the uptake on organic and black carbon, identified in the maxST simulations, we identified the NO2 uptake on sulfate nearly as important through a parallel test (not shown). In particular, the heterogeneous photolysis of HNO3 could not provide a significant HONO amount near Taiwan, South Korea, and Japan (in the JANO3-B and JANO3-C cases) and a small HONO amount for the Philippines bounding route (in JANO3-C case) for the altitudes below 2500 m (Fig. 3d, h: green, orange, blue).

For the middle troposphere (5000–6000±500 m) over the island of Taiwan, too abundant NO2 was predicted by the model during the cruises bounding to South Korea (up to 40 ppt) and the Philippines (up to 20 ppt) (Fig. 3m, p: small green and blue areas left of the vertical line). These overabundances might hint on the deep stratospheric intrusion in springtime that caused imperfect downward mixing fluxes (Lin et al., 2012; Stohl et al., 2003; Trickl et al., 2014). This excessive NO2 and the corresponding missing HONO were also sensitive to the AIRC and GR×8 cases (Fig. 3n, q), indicating that aircraft exhaust of HONO could adjust the HONO:NO2 ratio and more homogeneous HONO production might contribute more, given the high abundances of oxidizing substances at these altitudes. The possibility of emission of aviation-induced particles on which NOx to HONO conversion could reach 45 % (Meilinger et al., 2005) could support the need for NO2 reduction and HONO formation for this height across EMeRGe's near-land domains. Moreover, the surface-catalysed photolysis of HNO3 in the JANO3-C runs could serve as an efficient source and greatly reduced the model negative bias for HONO at 6000 m. However, the model overestimated NO2 levels in this case (ΔNO2 negative in Fig. 3s) because HNO3 was more photolysed to HONO and served less as a NOx removal process.

The model underestimation for HONO was also associated with the concurrent underestimation of NO2, observed more often at the altitudes of >1000 m. The erroneous NO2 concentrations of ∼1.8 ppb (1000 m) and ∼220 ppt (3000 m) across Taiwan were linked with a lack of HONO of as high as 290 ppt (1000 m) and 140 ppt (3000 m) (Fig. 3e, i). These likely inadequate NO2 abundances could be partially alleviated through the enhanced HONO:NOx emission ratio and more efficient NOx-recycling process in the ratR4+CLD cases, respectively (Fig. 3g, j). Here, the missing HONO was largely supplemented only at ∼1000 m over the marine environment (Taiwan–Japan cruise), when we identified more products for HONO on cloud in the ratR4+CLD case (Fig. 3e, g: red-orange diamonds). At ∼6000 m, small deficits of 60 ppt NO2 corresponding to ∼10 ppt HONO (Fig. 3p: orange and magenta) might correspond to lightning NOx emissions (Sudo et al., 2002) and stratospheric sources. Some homogeneous mechanisms at ∼6000 m as in the GR×8 could be effective (Fig. 3n, o). Moreover, the heterogeneous photolysis of HNO3 in the JANO3-C case could be an effective HONO supplement above 5500 m (Fig. 3s), while this photolysis acted as a NO2 production mechanism at any altitudes.

In general, the upper limit for the aerosol-uptake coefficients (maxST case) may be applicable for the lowest cruising altitudes, which induced the increase of modelled HONO levels during both daytime and nighttime (Fig. S10). The photolysis of adsorbed HNO3 on ground surfaces implemented in the JANO3-A case was impractical to be a source for HONO during EMeRGe, as this case only provided a mild HONO amount at a thin surface layer (<500 m; not shown). Fortunately, the surface-catalysed photolysis of HNO3 in JANO3-B and JANO3-C cases could remedy the model-measurement discrepancies; i.e. RHONO>0.6 and the model bias for HONO was reduced from -112 ppt (STD) to -22 ppt (JANO3-B) and -18 ppt (JANO3-C) for 0–500 m altitudes (Table S4). The HONO source from this photolysis of HNO3 was sufficient for continental and near-land regions. In particular, the photolysis of HNO3 adsorbed on particles with smaller SAD (JANO3-C case) was responsible for the 500–3000 m atmosphere around Philippines and South Korea and at higher altitudes where robust solar radiance might enhance the HNO3 photolysis. In the combined cases (maxST+JANO3-B and maxST+JANO3-C), HONO production was boosted, and the estimated NO2/O3 concentrations were best captured for 2000–5000 m (±500 m) altitudes (Fig. 2a, b, d: black diamonds vs. orange circles). Furthermore, the sensitivity cases including combined cases changed the global tropospheric effects differently, as discussed in Sect. 3.2.3.

The remaining drawbacks in reproducing HONO and other atmospheric species (NO2, O3, CO) by model urges further elucidation of efficient HONO formation mechanisms. To this end, one needs (1) to elaborate the combined HONO production mechanisms from enhanced NO2 aerosol uptakes and HNO3 photolysis alongside testification other potential HONO formation mechanisms and NOx-recycling processes; (2) to simulate the lower and upper limits for the uptake coefficients of NO2 on aerosols and clouds; (3) to provide better emission inventories for anthropogenic sources of pollutants from Southeast Asia and East Asia, lightning-produced NOx and HOx, and aviation-induced aerosols; and (4) to improve the vertical mixing and air mass transport from the stratosphere.

The model performance of the free troposphere was evaluated through the atmospheric tomography (ATom-1) aviation in August 2016 for NO2, OH, CO, HO2, and O3. The STD run reconstructed the chemical field observed in ATom-1 with moderate or strong positive correlations for NO2, OH, CO, and O3 (RNO2=0.730, RO3=0.751, ROH=0.579, RCO=0.659; Table S2). For the NP region, the model correlations for these species were slightly lower (RNO2=0.621, RO3=0.609, ROH=0.407, RCO=0.596; Table S2). The R values for NO2 and CO were consistently higher in the STD run than those in the OLD run, while for OH and O3, the R values are only improved for the NP region (Table S2).

Figure 4 shows measured (grey) and simulated (red and black) NO2, O3, OH, and CO concentrations and the effects of including HONO in the simulation for the NP region (flight no. 2 on 3 August). Figure 5 displays vertical profiles of the model biases in STD vs. OLD cases and photochemical effects by each HONO formation mechanism. Here, the data in all flights or in the NP region were classified based on the air pressure from 1000–200 hPa (±50 hPa) and separated into nine bins. In the NP region, the OLD run (black lines) tended to overestimate NO2, O3, OH, and HO2, but it underestimated CO at the lower troposphere, whereas the unsteady discrepancies at the upper layer were visible (Fig. 4a, b, e, f). All five species tended to be underestimated near the tropopause (300–400 hPa) and to be overestimated in the lower stratosphere (Fig. 5e–h, u). The HONO inclusion in the STD run (red lines) reduced NO2, OH, and O3 and increased CO levels, thereby dwindling the model biases for NO2, OH, and CO in the NP region except near the tropopause (Fig. 5e, g, h). HO2 was reduced near the surface layer and increased from the middle troposphere (Fig. 5v, w), reducing model bias for most parts of HO2's vertical profile (Fig. 5t, u). The reduction in the HO2 level near the surface might follow a similar cloud effect as that for OH, which turned into minor increases in the HO2 level at the middle and high atmosphere, given a lower HO2 level at these altitudes.

In the NP region, the surface NO2 level was reduced under the effects of HONO uptake on clouds (Fig. 4c: orange bars in vertical grey columns and Fig. 5p). Hence, O3 and OH were correspondingly reduced as their formations are presumably limited in the absence of sufficient NOx, that is, lacking atomic oxygen from NO2 photolysis and OH formation via HO2 + NO → OH + NO2 (Fig. 4d, g: orange bars in grey columns and Fig. 5q, r). Near the surface, aerosol HRs only slightly affected atmospheric species, whereas at high altitudes, the aerosol uptake was more relevant, especially for O3 concentrations (Fig. 4c, d, g, h and Fig. 5q–s: green bars), due to the contribution of aerosol direct effects to the O3 level (Xing et al., 2017). The dominant cloud effects near the surface appeared plausible for an ocean region with high cloud fractions at the lowest layer (Fig. S2). GRs also affected OH, O3, HO2, and CO, whereas the effects manifest in the upper troposphere rather than in the lower troposphere. This was likely the most influential factor that increases OH, HO2, and O3 levels at these high altitudes (Fig. 4d, g and Fig. 5: blue bars). The additional HONO from direct emissions had minor effects on NO2 and OH but contributed to the reductions of O3 and CO at high altitudes (Fig. 4c, d, g and Fig. 5q, s: red bars). At 900 hPa, the HONO emissions significantly reduced NO2 near the continental areas (Fig. 5l: red bars) due to its uptake by particles. These effects of the HONO chemistry in the STD simulation somewhat reduce the model biases for NO2, O3, OH, HO2, and CO (Fig. 5a–h, t, u: red numerical texts are the percentage reduction in model bias). Note that these biases were very pronounced near the surface (∼1000 hPa) in the NP region (51.7 % for NO2 and 77.3 % for OH). To capture the patterns identified by observations in the upper troposphere, except the NP region, more robust increases for NO2, OH, and HO2 levels were still required (Fig. 5a, c, t). At these altitudes, NOx and HOx sources from lightning (Brune et al., 2021) or aviation could also be relevant, as discussed in Sect. 3.1.2.

The simulations was also compared with the research vessel (R/V) Mirai's observation in the western Pacific Ocean for O3 and CO. The interpolation of model results for six cruises was provided, with four cruises across the Japan–Alaska region (40–75∘ N, 140∘ E–150∘ W) in July, August, and September 2015–2017 (summer), as well as one cruise for the Indonesia–Australia region (5–25∘ S, 105–115∘ E) and one cruise for the Indonesia–Japan region (10–35∘ N, 129–140∘ E) in December 2015 and January 2016 (winter). All the measured and simulated data were provided, whereas the data for the NP region (40–60∘ N) were analysed separately, as discussed in Sect. 3.2. More detailed information about the R/V Mirai can be found in Kanaya et al. (2019). Furthermore, the model evaluation with Mirai for the OLD run can be found in Ha et al. (2021).

Table 6 shows correlation coefficients, which indicated that the STD simulation for CO and O3 agreed well with Mirai (R=∼0.6). However, these correlation coefficients were slightly worsened compared with the OLD case. Although the HONO inclusion mostly reduced the model bias for CO, especially in the NP region (-16.158 to -4.948 ppb), the model bias for O3 was increased. The model biases exhibited a negative trend for both CO and O3 in the OLD case. This simulation pattern for O3 in the NP region was in line with the OMI comparison (Sect. 3.1.1). This finding seemingly indicated an insufficient downward mixing process of O3 in the free troposphere or inconsistent surface deposition (Ha et al., 2021; Kanaya et al., 2019). However, the CO underestimations in the NP region might mark the inadequate CO emission in the HTAP inventory in CHASER (Ha et al., 2021). In Fig. 6a and c, overestimations of CO and O3 were visible along Japan–Indonesia–Australia (Track-2) during the low episodes in December/January. Here, the larger model biases might account for the model's insufficient halogen chemistry (Kanaya et al., 2019; Ha et al., 2021). Figure 6 shows the model's percentage discrepancies for O3 from Mirai's data, except those from HONO concentrations interpolated for these regions. The underestimated simulations of O3 were enlarged, especially in the Japan–Alaska region, being driven by the reduction effects in the STD case. In another way, these effects weakened the O3 overestimates across the land areas, namely, over the region near Japan and Indonesia–Australia. Moreover, the higher HONO levels were identified for these offshore data with up to 1.4 ppb abundances (Fig. 6c: red numbers). This high HONO level might underestimate an accurate level as a stronger reduction for O3 was still required for the STD run (Fig. 6a: red marks).

The effects of the HONO chemistry along the R/V Mirai tracks exhibited various trends for each mechanism. Figure 7b and d illustrate O3 and CO changes triggered by the HONO gas reactions (GRs), uptakes (HRs), and emission (EM). The gaseous reactions (blue bars) had mostly increased CO levels due to the reduced OH and O3 levels (Fig. S14). The gaseous mechanisms caused some reductions at the peak CO level because the higher OH level from HONO photolysis near the land domain or extra OH flux from the stratosphere near 60∘ N latitude could dilute CO. Furthermore, the O3 level was slightly increased due to GRs north of 60∘ N, as GRs was a source for NO2 and thus enhanced O3 formation at these high latitudes (Fig. S14). The O3 level was often decreased in the NP region since minor NO2 increased and stronger OH reduction was seen for this region (Fig. S14). The change tendencies in O3 near land areas were varied (T2, T3) because the vertical effects to NOx and OH were stronger during DJF for this region (Fig. S14). HRs, largely consuming NO2, reduced O3 (as large as 8 ppb) and increased CO (∼10 ppb) levels (Fig. 7b, d: orange+green). HRs (particularly HRs on cloud surfaces, shown by orange bars), exerted the strongest contribution to the calculated changes in O3 and CO among the three HONO pathways. This predominant cloud effect was also prominent in the previous comparisons, especially EMEP (Fig. 10ii, jj: blue), thereby indicating substantial effects of clouds at the mid-latitudes where the cloud SAD is higher (Fig. S1). HRs on aerosols (green bars) had minor contributions during all cruises, despite causing a marked increase in the O3 concentrations off the coast of Japan (track no. 3). It should be noted that this is not enough to explain the simulation bias with regard to the measurements. The additional HONO from direct emission (red bars) mainly increased O3 and reduced CO concentrations, especially near land (latitude <50∘ N). This finding resonated with the comparison for continental stations (Fig. 10cc, ff, ii, jj: orange). The overall effects of the HONO chemistry along Mirai cruises tended to reduce O3 and increase CO levels. For the NP region, the CO level increased and the OH level reduction also ameliorated the model performance. The improved model performance is evidenced from the comparison of the simulation with ATom-1 aircraft data as well (Fig. 5g, h). Thus, the strengthened underestimation of O3 concentration in the NP region was not likely driven by HR on cloud particles (Sect. 3.1.1 and Ha et al., 2021). It was rather related to the inconsistencies in the surface deposition of ozone. These inconsistencies were supported by empirical evidence as the negative bias in this comparison turns neutral or positive for the aircraft measurements in the same region (Fig. 5f: at 1000 hPa).

Overall, the comparisons between the model and ATom-1 Mirai might indicate that the HRs on cloud surfaces were the main contributing factor to the marine boundary's photochemistry, whose effects emerged during the ATom-1 flights in the marine atmosphere. GRs and aerosol HRs had a stronger impact on atmospheric chemistry at higher altitudes than the near-surface layer. Also, their effects should be enhanced through the additional NOx and HONO sources to reconcile the model simulations with the observations.

In this part, HONO concentration and HONO-related species measured for summer 2018 at the leg (150 m) and the summit (1534 m) of Mt Tai (Shandong province, China) by Xue et al. (2022a) were reproduced using a data extraction tool from images (Fig. 5 in Xue et al., 2022a). We compared our model's additional sensitivity simulations (Table 5) with Xue's measurements (Fig. 8). The HONO estimated in the STD case for the foot station was rapidly reduced after 4:00 (from ∼1100 pptv), while the observed HONO level peaks at ∼6:00 and remains about 0.5 ppbv at noon (Fig. 8a). The HONO level produced by the sole NO3- photolysis on ground, aerosols, and cloud particles (JANO3-A/B/C cases – respectively in solid blue, dashed cyan, and dashed orange lines in Fig. 8a) can append moderately the simulated daytime HONO to reach the ground-based observatory levels comparing to the STD case. This addition indicated a partially important role of NO3- photolysis on all surfaces to HONO sources since early morning. Also, NO3- photolysis adjusted the ratio of HONO:NOx more analogue to the observed daytime ratios at the ground (Fig. 8c, d). The lack of NO2 sources, especially during nighttime (Fig. S15), even in the OLD case (without HONO chemistry), was one of the reasons for the remaining unknown HONO sources existing during ∼5:00–11:00. The combined cases (maxST+JANO3-A/B/C cases) produced too much HONO at the leg compared to observations (not shown in Fig. 8), indicating the improper mechanism of enhanced NO2 aerosol uptakes for ground-based stations. In contrast, the enhanced aerosol uptakes were more compatible with observation at the summit, where these combined cases provided the best agreement to observation (Fig. 8a, b: solid cyan and orange lines). However, the best simulation for HONO at the summit station only reached the lower line of the averaged daytime HONO level (Fig. 8b). Xue suggested that the high HONO level at the summit of Mt Tai was dominated by the rapid upward transport from the ground and the in situ heterogeneous formation on the mountain surfaces (Xue et al., 2022a), which the mismatching between the actual locations and the coarse model grid (2.8∘) of our model, including vertical layer, might not provide. Similar to NO2, the simulated CO concentrations at Mt Tai were very low (∼200 ppb at the foot station in the STD case versus ∼400–600 ppb measured CO; not shown), even in the OLD case. In our model, the inadequate emission inventory of CO and NO2 for the Asian region using HTAP-II-2008 and the coarse model resolution (2.8∘×2.8∘) were the reasons for the low ground-based emissions and vertical transport (Ha et al., 2021). Such emission inventory and vertical transportation improvement could further close the HONO observation gap and reduce the unrealistically high HONO:NOx ratios. Such a study could better show the validity of the HONO production mechanisms from enhanced NO2 aerosol uptake and NO3- photolysis (combined cases) for the summit station.

The above discussion confirmed that NO3- photolysis (Reaction R7) on the ground, aerosols, and clouds surfaces (JANO3-A/B/C cases) enhanced daytime HONO but being ignorant of the NO2 level (Fig. S15) due to an absent NOx recovery process (Ha et al., 2021). However, HONO sources via Reaction (R7) in the JANO3-A/B/C cases still increased daytime O3 levels at the foot of Mt Tai (Fig. 9a) as a result of rapid NOx cycling. Either the JANO3-A/B/C or the combined maxST+JANO3-B/C case was closest to O3 observation during cleaner (50–75 ppb O3 on 16–22 July) or dirtier episodes (<50 ppb O3), suggesting an enhanced role of aerosol uptakes during the polluted episode (indicated by the corresponding increases for CO; Fig. 7 in Xue et al., 2022a). At the summit station, NO3- photolysis in the JANO3-C and maxST+JANO3-C cases (all surfaces including clouds) boosted O3 up to the observational level, indicating the contribution of cloud surface at the summit (∼1500 m).

The effects of HONO chemistry in the continental near-surface layer of East Asia and Europe were also investigated. To this end, we conducted model comparisons versus EANET and EMEP stationary observations for mass and gaseous concentrations of PM2.5, SO42-, NO3-, HNO3, NOx, O3, and CO (CO for EMEP only). Table 7 shows the correlation coefficients (R) and model biases for each species in the OLD and STD cases. The OLD simulation had its fair correlations and RMSEs with observation for SO42- (R(EANET) = 0.56, R(EMEP) = 0.63), NO3- (R(EANET) = 0.36, R(EMEP) = 0.71), and HNO3 (R(EANET) = 0.18, R(EMEP) = 0.12), which were in line with other atmospheric chemistry models' R and RMSE values against EANET and EMEP (Bian et al., 2017), as also discussed in Ha et al. (2021).

Figure 10 compares the measured versus simulated HNO3, NOx, HONO, O3, and CO concentrations for the EANET and EMEP stations. The stations were divided into three groups: (1) high-NOx EANET stations, including Jinyunshan (China), Kanghwa, Imsil, Jeju (South Korea), Bangkok, Nai Mueang, Samut Prakan, Si Phum (Thailand), Metro Manila (Philippines), and Ulaanbaatar (Mongolia); (2) other EANET stations (39 for HNO3, 22 for O3, and 15 for NOx); and (3) all EMEP stations. The ground-based observations in the period 2010–2016 revealed the slightly decreasing NOx for moderate NOx concentrations, as well as PM2.5, and aerosols (Figs. 10e, h and S4e, g, h, i). These decreasing trends were not captured by our simulations, which used the high-emission scenario for the EDGAR/HTAP-2008 inventory. Note that NOx and PM2.5 concentrations were generally underestimated in the model (OLD), especially in high-NOx regions (Figs. 10b, e, h and S4a, d, g) with the model's averaged biases of -0.8 ppb NOx for EMEP and -4 ppb NOx for EANET (Table 7). These underestimations were stronger during winter, particularly for the high-NOx regions. It was possible that complex domestic sources could lead to diluted emissions for the simulations' moderate horizontal resolution (∼2.8∘). Higher model resolutions, such as 1.1∘, 0.56∘, or even higher, could remedy such effects (Sekiya et al., 2018).

HONO chemistry in the STD case increased HNO3, NO3-, SO42-, and PM2.5 for EANET and EMEP stations compared to the OLD case (Figs. 10 and S4: red vs. blue lines). HNO3 and NO3- were increased as the products of NO2 conversion (Reaction R4); thus, the model underprediction for NO3- in EANET stations was mitigated (bias OLD → STD: -0.439→-0.223µm m-3). As a result of the increased OH level at the surface of these ground-based stations, SO42- was also increased (Li et al., 2015; Lu et al., 2018) (Fig. S4j, k, l), although this effect enlarged the model overestimation for SO42- species at EANET and EMEP stations (Table 7). The consequent increase in PM2.5, though minor, remedied the model underestimate for PM2.5, e.g. model bias in OLD → STD: -3.044→-2.494µm m-3 (EMEP). Unfortunately, the model overestimate for HNO3 in the OLD case was enlarged with the inclusion of HONO.

In the STD case, including HONO photochemistry, the negative biases of NOx in the model had been adversely enhanced due to the NO2 loss processes (bias OLD → STD: -3.997→-4.358 ppb for EANET; Table 7). These processes also suppressed the NOx seasonality observed at most sites (Fig. 10b, h, red lines). The lack of seasonality was driven by the substantial loss of NO2 on the surfaces of atmospheric particles during winter. For EANET's low-NOx and EMEP stations, this huge NO2 loss was attributed to cloud surfaces (Fig. 10ee, hh: blue bars). However, NO2 uptake by aerosols has a comparable contribution effect to the cloud effect in high-NOx environments such as Jinyunshan (Fig. 10bb, grey bars). Namely, nearly half of the NO2 was converted to HNO3 in Reaction R4; Fig. 10aa, dd, gg) without an efficient recycling process, leading to an overall removal of NOx. This lack of NOx could be the main driver for the seasonal NOx deterioration and the exacerbated overestimations of HNO3 by simulations.

The STD O3 simulation exhibited moderate and strong positive correlations with EANET and EMEP observations, 0.595 and 0.707, respectively (Table 7). The model improvements for SO42-, NO3-, PM2.5, and HNO3 were minor. However, the model improvement for O3 was considerable, with a bias reduction of ∼67 % for EMEP and ∼74 % for EANET (Table 4). In the STD case, too little NOx was left from its heterogeneous loss, causing a net O3 chemical destruction (because lacking atomic oxygen from NO2 photolysis), which in turn reduced the model overestimates for O3 in the OLD case (Table 7; Fig. 10c, f, i, red versus blue lines). However, further improvements in the chemical scheme were necessary to reproduce the O3 measurements better; namely, a larger O3 reduction for the summer and a reduced effect in simulated O3 for the winter might alleviate the undesired effects. A delayed minimum from summer (as observed) to early winter (calculated in OLD and STD runs) causing opposite seasonality for O3 was prominent for the low-NOx EANET stations (Fig. 10f). The effects of HONO chemistry on the mean OH levels were small, although it showed slight increases for OH's minima (Fig. S4j, k, l). Thus, due to the apparent O3 reduction for EMEP stations, CO was increased. Despite the reductions in NOx and O3 levels being exaggerated during winter, the increment in CO reconciled the model's underestimation of CO high peaks in spring (Fig. 10j), thereby strongly dwindling the bias for CO by ∼59 % (Table 4). However, the CO concentrations during summer should be reduced in the STD case to capture the measurement. This finding might indicate inadequate HONO emissions for the EMEP stations (Fig. 10jj: orange), which otherwise had reducing effects on NOx, O3, and CO levels during summertime.

The breakdown scrutinies for aerosols and clouds effects for the ground-based stations (EANET/EMEP) also revealed the vast role of cloud uptakes in the HONO impacts on NO3- aerosols, NOx, O3, and CO (Figs. 10 and S16: blue bars), while the HONO impacts on HNO3, PM2.5, and SO42- aerosols were governed by aerosol uptakes and HONO emission (grey and yellow bars).

The existing ill reproduction in NOx's seasonality and overestimations for HNO3 might be amended by an explicit inventory for direct NOx emissions and an efficient NOx-recycling process. Such a mechanism via HNO3 uptakes on soot surfaces (HNO3 → NO2) was also tested in this study using the uptake coefficient range from 3×10-5-4.6×10-3 (Lary et al., 1997; Akimoto et al., 2019). Unfortunately, this heterogeneous HNO3 conversion could not solely serve as a productive NOx-recycling process in the EANET/EMEP stations (not shown). Among the additional cases described in Table 5, the alternated HONO:HNO3 (0.9:0.1) product ratio of Reaction (R4) (ratR4 case) showed a good remedy for NOx at EMEP stations (Fig. S11g: brown vs. black). For the EANET sites, the photolysis of adsorbed HNO3 on ground surfaces (JANO3-A case) avoiding NOx removal via HNO3 could remedy the NOx seasonality issue for these ground-based stations (Fig. S11a, d: green vs. black). However, the photolysis of adsorbed HNO3 on ground surfaces and aerosol/cloud surfaces (JANO3-B and JANO3-C cases) was not an effective NOx-recycling process for EANET/EMEP measurements, leaving only slight differences in surface NOx levels compared to the STD case (Fig. S11a, d, g: cyan and orange vs. reds). However, the heterogeneous photolysis of HNO3 increased O3 and OH at the high-NOx regions instead of O3 reduction in the STD case (Fig. S11b, c: cyan and orange vs. blue), which brought reconciliation to the underestimates for O3 peak in springs (Fig. S11e: dotted cyan vs. dotted black), although the runs with HNO3 photolysis still did not capture the O3 minimum in summer. Only the ratR4 case could capture the O3 minimum in summer among the sensitivity cases, which indicated the need for stronger NOx-recycling processes for these ground-based stations.

This section sheds light on the global HONO distribution computed for the STD case. The surface HONO concentration peaked over the geographical region that includes China, with seasonal mean levelled up to 2.8 ppbv during summer and 7.8 ppbv during winter (Fig. 11a, b). The winter peak agreed with observations for a large industrial region in the Yangtze River Delta of China (Zheng et al., 2020). The high concentrations of HONO were also identified in other industrial regional clusters: the northeastern US (seasonal mean up to 0.5–1 ppbv); India (up to 1–3 ppbv); forest regions, especially the extratropical evergreen forest in Europe (up to 1–3 ppbv); and Africa (up to 0.5–1 ppbv). Over the ocean, HONO levels remained at 10–30 pptv in the coastal regions and below 10 pptv far off the coast. The simulated HONO distribution was in line with a previous study (Elshorbany et al., 2012) despite the peaks over polluted Chinese areas being markedly higher in our model (10-fold). The overestimation associated with the soot uptake in our model has been previously neglected. The highest HONO concentrations (10–30 pptv) in the free troposphere (at 2500 m) were simulated over Africa's biomass burning region during wet months (JJA) (Fig. 11c, d), which could arise due to the NO2 uptake on aerosols, originated from this wildfire source.

In the model, HRs and EM were the main contributors to HONO at the surface layer (Fig. 12) by providing efficient HONO formation and promoting gas Reaction (R2). Of the various surfaces provided for HRs in our model, liquid/ice cloud particle surfaces were supposed to catalyse significant photochemical effects in remote regions. This phenomenon has not been previously addressed in other studies in detail. The uptake of liquid/ice cloud particles either increased HONO formation via Reaction (R4) for the tropical and southern oceans or reduced it via Reaction (R6) along 60∘ S in DJF and the Arctic in JJA (Fig. 12a, b). Besides cloud particles, HRs on aqueous aerosols also produced HONO in a continental atmosphere rich in sulfate, dust, and soot particles (Fig. 12c, d). EM included in the model has sharply increased the HONO level over deserts (Sahara, Arabian), grasslands (South Africa, South America), and boreal and agricultural land (western Europe, Australia). This finding agreed well with another study, based on spaceborne observations for HONO in wildfire plumes (Theys et al., 2020) and along-ship tracks in the marine boundary layer (Fig. 12e, f).

Table 8 summarizes the global sources and sinks of tropospheric HONO quantified by CHASER. The simulations indicated that GRs contribute only 11 % of the HONO net production. HRs and EM produced more significant HONO (63 % and 26 % HONO net production, respectively). The pyrogenic HONO emission estimated in this study might be underestimated as the HONO:NOx emission ratio could be enhanced by up to 1 at extratropical evergreen forests (universally 0.1 in this study (STD)) (Theys et al., 2020). For large metropolitan areas such as those in China, HRs and EM had also been reported as the two most significant contributors to HONO formation, at ∼59 % and 26 %–29 %, respectively (Li et al., 2011; Zhang et al., 2016). Of the various surfaces provided for HRs, aerosols represent a more effective HONO formation site (∼51.2 %) compared with ice and clouds, as they are contributing only 11.8 % to HONO production. Moreover, the HONO loss through photolysis Reactions (R1) and (R3) was equivalent to its uptake onto the particle Reaction (R6). In equilibrium, the tropospheric abundance of HONO averaged over the globe was estimated to be 1.4 TgN in our model.

HONO production calculated in sensitivity cases (Sect. 3.1.2) is recorded in Table S5 (last column), and the spatial distributions are plotted in Fig. S12. The small supplement by the photolysis of adsorbed HNO3 on ground surfaces (JANO3-A case) to surface HONO concentration and tropospheric HONO burden (1.40→1.45 TgN) was consistent with the discussion for the EMeRGe campaign. In the JANO3-B and JANO3-C cases, tropospheric HONO burden was increased to 2.02 and 2.93 TgN, respectively, mainly remaining for the lower troposphere (∼600 hPa) (Fig. S12g, h). Compared to HNO3 photolysis, the enhanced aerosol uptake (maxST case) produced HONO more extensively over the source region and in the winter hemisphere where there was no photolysis (Fig. S12a–d). Therefore, the maxST case did not produce enough HONO during EMeRGe (worse than the JANO3 cases). However, the global HONO burden in the maxST case was added to 7.79 TgN, which might be because we set the enhanced aerosol uptake of NO2 for all environments. The combined cases (maxST+JANO3-B or maxST+JANO3-C), although appropriately approaching daytime HONO production as well as NO2 and O3 levels during EMeRGe, incredibly escalated the global HONO burden to 12.64 and 17.13 TgN, respectively (Table S5), more via the enhanced aerosol-uptake setting that could reach the upper troposphere (Fig. S12k, l). However, it might be more realistic if HONO production stayed at the lower troposphere (Eshorbany et al., 2012). In future work for the combined cases as standard cases, the amplified NO2 conversion on aerosols should be confined to high SAD regions (Kalberer et al., 1999; Stadler and Rossi, 2000) or to the daytime only (Notholt et al., 1992; Stemmler, 2007).

P denotes chemical production, S denotes source (emission + chemical production), and L denotes loss. The numbers in parentheses represent the portion of each pathway to the total HONO net production. Bold lines show the most significant contributing mechanisms to the HONO burden.

The comparison between the simulation and OMI spaceborne observations for tropospheric column ozone (TCO) can be examined as a global effect for ozone. In Fig. 13, the STD run with HONO inclusion improved the overall tropospheric column ozone (TCO) distribution observed by the OMI, especially at the mid-latitudes. Figure S3 indicated a TCO reduction when HONO chemistry was included in the STD case (red lines vs. green lines). Although HONO photolysis Reaction (R1) was a source of OH, supposedly increasing the tropospheric oxidizing capacity, the calculation in the STD case showed the OH and O3 increased only occur at the surface of polluted sites (Fig. 14a, e). The NO2 conversion to HONO and HNO3 (Reaction R4) became a NOx's removal pathway at remote regions, thus restricting the formation of O3 and OH for the larger part of the troposphere (lacking atomic oxygen from NO2 photolysis). Figure 13 shows that the O3-reducing effects of HONO chemistry greatly reduced the model overestimates in the OLD simulation for the general Northern Hemisphere and polluted regions such as China. However, in the NP region, the inclusion of HONO only reduced the model overestimates during the insignificant episodes of TCO (autumn to early winter) while extending the underestimates for TCO for the rest of the year (Fig. S3b). These O3 underestimates in the NP region are also visible for the modelled surface air versus the measurements during the Mirai cruises (Fig. 6a, b). Notably, these underestimates for O3 could hold up to 400 hPa, as seen in comparison with the ATom-1 flights (Fig. 5f: 400–900 hPa). This phenomenon could be related to the stratospheric downward transport and insufficient vertical mixing, as discussed in Sects. 3.1.3 and 3.1.4, for comparisons in the NP region's surface air and free troposphere. Although the HONO level in STD remained <10 ppt for this area (Fig. 11), the O3-reducing effects exacerbated the model discrepancy. The HONO photochemistry was unlikely to be the primary driver of this phenomenon as the ozone simulation was improved over the continents when the HONO photochemistry is included.

In the combined sensitivity case maxST+JANO3-B (Sect. 3.1.2), O3 was further reduced than the STD case. The reduction in TCO might be due to the enhanced NO2 uptake on aerosols, leading to more substantial O3 formation restriction. The maxST+JANO3-B case showed better harmony with OMI for the regions of TCO overestimations (Fig. 13g, h, k), especially the annual mean (g panel). However, the underestimates of TCO, including the NP region, were worsened. These results indicated that the reduction for O3 by HONO chemistry was reasonable, and the combined cases such as maxST+JANO3-B could be plausible, although the estimated reduction degree should be reduced by elaborating the maxST's reactive conditions.

In this section, the global impact of HONO photochemistry is elucidated. To this end, Table 8 summarizes the HONO budget and the contribution of each pathway to the HONO photochemical cycle. Table 9 describes its consequences for the lifetime of CH4 and the budgets of NOx, O3, and CO. The gaseous reactions of HONO tended to increase the abundance of NOx, O3, and CO (+1.01 %, +0.15 %, +0.44 %, respectively) and CH4 lifetime (+0.36 %) in the troposphere. Without heterogeneous and direct emissions, the relatively low HONO formation by gaseous reactions (11 % of the total net HONO production; Table 8) did not cause any significant effects on NOx, O3, and CO in the troposphere.

Heterogeneous reactions that produce HONO were the most salient contributing factors to tropospheric chemistry, thereby decreasing the tropospheric oxidizing capacity and increasing the CH4 lifetime by 15 % and CO abundance by 10 %. HRs also reduced the NOx level by 23 % and the O3 level by 6 %, respectively (Table 9). The global HONO distribution from Fig. 14 was mainly caused by the HR formation of HONO. Here, the reducing effects for NOx levels, with consequences for OH and O3 level reductions by heterogeneous reactions, were significant at middle to high latitudes during summer – more specifically, in DJF along 60∘ S and the Arctic and NP oceans during JJA, which amounted to about a -100 % reduction in the NOx level at the surface (-60 % reduction in OH and -40 % reduction in O3) (Fig. 14a–f: blue areas). These reductions in NOx, OH, and O3 levels extended up to 400 hPa at high N/S latitudes (Fig. 14k, l). All these reduction effects for NOx, OH, and O3 levels were due to the removal of HONO on ice and cloud particles (Reaction R6) (Fig. 12a, b: blue fields). On the one hand, it accelerated the conversion of NO2 to HONO and ultimately strengthened its deposition by particulate nitrate (Reaction R4) (Fig. S13a). On the other hand, HRs occurring on aerosol surfaces led to increments in OH and O3 near the surface of polluted regions during winter. These were the main contributors to the regional photochemical effects over China, western Europe, and eastern US regions in winter (up to -74 % reduction in NOx, +1500 % increase in OH, and +48 % increase in O3; Figure 14a, c, e). However, these OH and O3 level increases were only accumulated in the surface layer (only small red areas at ∼1000 hPa in Fig. 14i, m). Compared to HRs on aerosol, HRs on clouds exhibited twice the effectiveness when reducing the tropospheric NOx level (-15 % versus -8 %) and caused 3 times the effects on the tropospheric oxidation capacity (+11.5 % in CH4 lifetime, -4.6 % in O3, +7 % in CO), compared with HRs on aerosol (+3.5 % CH4 lifetime, -1.5 % O3, +2.6 % CO) (Table 9).

Given the direct emissions of HONO (∼10 % of NOx emission inventory), the surface NOx, O3, and OH concentrations were generally enhanced in the STD case compared to the GR+HR case. They induced the concentration modification for NOx (+1.77 %), O3 (+0.97 %), and CO (-1.63 %), as well as a significant reduction (-2.3 %) in the CH4 lifetime (Table 9). Remarkable enhancements for NOx (up to +198 %), OH (+243 %), and O3 (+24 %) (Fig. 14b, d, f: red fields) were identified for the cropland and shrubland/forest regions in Australia, South America, and South Africa during JJA, as well as the boreal vegetation prevailing at middle to high latitudes in Europe, North America, and the polluted Chinese region in DJF (up to +748 % OH) (Fig. 14a: red fields). NOx and OH were elevated in these mid-latitude regions because of the enhanced HONO photolysis (Reaction R1) by the additional HONO source. However, OH, NOx, and O3 levels were reduced near the surface of the Northern Hemisphere's land during summer (up to -47 % reduction in OH, -82 % reduction in NOx, and -15 % reduction in O3) (Fig. 14b, d, f). The latter phenomenon was similar to the heterogeneous cloud effects for the high latitudes discussed above.

Overall, the inclusion of the three HONO processes (gas phase, aerosol and cloud uptakes, direct emission) caused changes of -20 % in NOx, -5 % in O3, and +8 % in CO, as well as a significant increase of +13 % CH4 lifetime in the troposphere (Table 9). Figure 15 highlights the consequences of HOx, NOx, and O3 for the Chinese and NP regions. The NOx level reduction accumulated in the Arctic and Antarctic during summer, especially over the NP ocean (reducing the NOx level by 60 %–90 %; Fig. 15i). These reductions in NO2 and HONO concentrations were due to their uptake onto ice and clouds in these regions. However, these reducing effects caused further reductions in OH and O3 levels for a larger part of the troposphere. As NOx was essential in regulating O3 and OH in the troposphere, a reduction of the NOx level increased the HO2:OH ratio (due to the HO2 + NO → OH + NO2 reaction), which restrained the formation of OH and ultimately of O3. Moreover, a NO2-deficit environment directly affected the O3 level as NO2 was a primary source of an oxygen atom that engages in the formation of O3. Thus, in summer, both OH and O3 levels were drastically reduced over the NP region (35 %–67 % for OH, 30 %–43 % for O3; Fig. 15g, k), and the CO level was increased by 18 % in this region (Fig. 14h).

The significant impacts of HONO photochemistry were especially relevant over eastern China in winter, which might reduce the NOx level by 48 %–78 % (Fig. 12c) due to the uptake of NO2 onto aqueous aerosols. At the surface, the OH level was enormously increased as a result of HONO photolysis (Reaction R1), heterogeneous NO2 conversions (Reactions R4, R5), and additional direct emissions (Fig. 15a, b). The corresponding increase in the O3 level was only identified at the surface of the Beijing region during winter, with +28.8 % caused by HRs on aerosols (Fig. 15e). For Beijing with high NOx emissions, VOC-limited O3 chemistry was likely the driving mechanism (Liu et al., 2010). The vast increases in OH and O3 levels over Beijing in winter were basically in line with the present knowledge of HONO photochemistry (e.g. Lu et al., 2018). Elshorbany et al. (2012) also reported an increase in OH (2–5×106 molecules cm-3) and O3 (0.3–0.5 ppbv) concentrations over polluted regions in China during winter. Compared with Elshorbany's work, the increases in OH and O3 concentrations in our model were higher due to the different HONO mechanisms applied in the two models, simply an averaged HONO:NOx ratio (0.02) in their model. In particular, our newly added heterogeneous reactions on cloud particles (Reaction R4) caused significant reductions in OH, NOx, and O3 levels in the NP region during summer, which their model did not cover. The overall reductions in tropospheric oxidizing capacity due to HONO photochemistry were in line with the expected response to heterogeneous processes (Liao et al., 2003; Martin et al., 2003) and agreed with those previously reported for other HRs (HO2, N2O5, and RO2; Ha et al., 2021; Table 9). Our findings indicate that a global model without heterogeneous processes for HONO would neglect the significant changes in OH and O3 concentrations in remote areas and, thus, will underestimate the potential effects in polluted regions.

As mentioned above, the relative importance of ice and cloud surfaces to the oxidant chemistry was negative for Arctic and NP regions. NO2's uptakes on ice and cloud surfaces (Reaction R4) were the main reason for the reductions in surface NOx, OH, and O3 concentrations in these regions during JJA (Fig. S13a). These reductions also occurred for the free troposphere, which generally improved the model comparison with ATom (Fig. 5) and partially improved in CO simulation by cloud effect (Fig. 7). To HONO formation, enhancement of NO2's uptake coefficient on cloud surface (γliq. (Reaction R4) = 10-4→10-2), along with changing the HONO:HNO3 yield ratio from 0.5:0.5 to 0.9:0.1 in Reaction (R4) in the ratR4+CLD case compared to the STD case, helped preserve more NOx during EMeRGe's flights. Here, a supplement for HONO production is only seen in the ratR4+CLD case for the marine environment ∼1000 m near Japan (Fig. 3g: red-bordered orange diamonds). The ratR4+CLD case introducing an approach to recycle NOx led to lowered global effects of HONO (only -8.57 % NOx globally, CH4 lifetime=9.6 years; Fig. 16). For the ground-based station in comparison with Xue's data (Xue et al., 2022a), the NO3- photolysis in JANO3-C and maxST+JANO3-C cases (all surfaces including clouds) boosted the O3 level, which improved the agreement with observed O3 at the Mt Tai summit station, indicating the contribution of cloud surface to O3 formation at the altitudes ∼1500 m over a mountainous area (Fig. 9b). The NO3- photolysis on cloud surface in the JANO3-C case could also be an effective HONO supplement for the tropospheric part above 2000 m over the Asian coastal region compared with EMeRGe's data as compared to the JANO3-B case, which excluded clouds (Fig. 2a). The simulated concentrations of NO2 and O3 also agreed better with the measured data in this comparison.

Contrary to cloud surfaces, the aerosol effect was only crucial for regional photochemistry at the surface layer of polluted regions, such as China, western Europe, and the eastern US in winter time (Fig. 14). As discussed above, aerosol uptakes reduce NOx but increase regional OH and O3 levels. For the sensitivity of HONO formation to aerosol effect, the cases JANO3-B and maxST+JANO3-B, which only included ground and aerosol surfaces for NO3- photolysis, also remedied the discrepancies for the daytime HONO level across various altitudes during EMeRGe flights (Fig. 2a). For the comparison with Mt Tai station, enhanced uptakes of NO2 onto aerosol surfaces in the combined cases (maxST+JANO3-B/C) provided more HONO production at the summit (Fig. 8b), as well as adjusting O3 levels at both foot and summit stations during the polluted episode (Fig. 9).

The estimated global effects of HONO chemistry in the STD case was the abatement of global tropospheric oxidizing power, despite surface OH and O3 levels being increased at polluted sites. The reduction tendency in global OH and O3 contrasted with other modelling studies (e.g. Elshorbany et al., 2012; Jorba et al., 2012; Lee et al., 2016; Zhang et al., 2021). Some discussions on the tendency of HONO's global effects are addressed here. The positive or negative impacts on oxidizing species (OH and O3) were constrained to HONO formation mechanisms rather than NOx concentration. For high-NOx regions such as EANET stations with 6-month-averaged NOx higher than 20 ppb, O3 and OH were reduced due to NOx removal via NO2 uptakes (Fig. 8b, c), especially at night. If NOx was highly underestimated in the model for these high-NOx regions and an efficient NOx-recycling process was still absent, OH and O3 might be reduced daily. The calculation for daytime only in comparison with Xue's data showed that O3 could be increased when a complementary HONO source was provided via NO3- photolysis (Fig. 9a) and sole enhancement of the aerosol effect rather than both aerosols and clouds (JANO3-C versus JANO3-B; Fig. 9b), even though NO2 was still underestimated with a large extent in the model (Fig. S15). At higher altitudes over remote regions in the ATom (NO2<10 ppb) and EMeRGe comparisons, OH and O3 can be increased due to more potent gas-phase chemistry of HONO in the STD case, including HONO photolysis (Fig. 4) and particle NO3- photolysis (Fig. 2).

The amplified aerosol uptake of NO2 (maxST case) further reduced an unrealistic degree of global NOx abundance (-55.4 %) and tropospheric oxidizing capacity, leading to 14.5 years for a global CH4 lifetime. The ratR4+CLD case introducing an approach to recycle NOx led to lowered global effects of HONO (only 8.57 % of NOx was reduced globally, CH4 lifetime=9.6 years). The photolysis of adsorbed HNO3 on ground surfaces (JANO3-A case) still showed reductions in global OH and O3 abundances (Table S5). The ground-surface HNO3 photolysis in the JANO3-A case caused only minor changes for a thin surface layer, which is in line with other studies (Ye et al., 2018; Zhang et al., 2009). In JANO3-B and JANO3-C cases, a recycling process for NOx via HNO3 photolysis was expected. However, only the JANO3-C case showed an increment in global NOx and O3 (+29 % and +16.1 %, respectively), leading to only 5.4 years for the global CH4 lifetime, which was impractical. This was because that simplified approach and maximum thresholds for the phase HNO3 photolysis were used. The combined case maxST+JANO3-B led to more convincing effects (CH4 lifetime was 10.2 years; Table S5), which held the same tendencies as those calculated in the STD case. However, validating this combined case was only conducted for the daytime environment during EMeRGe (Sect. 3.1.2) and for TCO at northern mid-latitudes with OMI (Fig. 13).

Figure 16 illustrates the calculated global-mean changes of tropospheric abundances in additional simulations from those in the OLD case (without HONO chemistry). The simulations of the largest negative to largest positive magnitudes of changes (%) in CH4 lifetime (purple bars) are shown from left to right. Other bars show percentage changes in NOx (red), O3 (blue), and CO (green). In simulations including ratR4, ratR4+CLD, JANO3-A, maxST+JANO3-B, and maxST cases, HONO's impacts on tropospheric CH4 lifetime and abundances of NOx, O3, and CO showed similar tendencies to those impacts in the STD simulation. These similarities indicated that the heterogeneous chemistry of HONO has a general tendency to reduce tropospheric oxidizing capacity (OH and O3) as a result of NOx removal globally via NO2's uptakes on aerosols and clouds. In these settings, more substantial reductions in oxidizing species and NOx were seen in maxST and maxST+JANO3-B cases via enhanced aerosol uptakes of NO2. The particle-phase NO3- photolysis can solely compensate for NOx removal processes and act as an efficient NOx-recycling mechanism (on a global scale), which can be seen in the JANO3-B and JANO3-C cases. The enhanced aerosol uptake in the maxST setting of the combined maxST+JANO3-C case neutralized this compensation. However, the tropospheric oxidizing capacity (OH and O3) was increased in the cases configured with NO3- photolysis (JANO3-B/C, maxST+JANO3-C cases), leading to a reduction in global CH4 lifetime. The quantitative impact was unrealistic in some cases, e.g. 30 % of CH4 lifetime in the maxST+JANO3-C case or -50 % of global NOx in the maxST+JANO3-B case, proven as the proper mechanism for a particular environment (along EMeRGe-Asia-2018 flights) and not globally. Thus, these changes merely provided the tendencies of impact sensitivity for different pathways of HONO formation, still with high uncertainty in their magnitudes.

In conclusion, we suggest the global effect tendency was towards tropospheric oxidizing capacity reduction, although further elaboration for enhanced aerosol uptakes of NO2 and surface-catalysed photolysis of HNO3 could drive the effect magnitude. The implication of HONO chemistry in a bottom-up-approached global model such as CHASER needs an intense examination of possible HONO sources and profound evaluations with observed HONO in the troposphere.