Ozone production was calculated for time periods in 2021 and 2022. A field campaign to measure VOCs in Santiago was conducted in March 2021 (description in Sect. 2.4.2 and Appendix A). Therefore, we performed detailed modeling of the chemistry of ozone production for sunny days during the month of March. We filtered out overcast days because photolysis frequencies were not measured and, thus, we rely on modeled actinic flux without cloud cover correction for the photolysis component of the model. March 2021 was mostly sunny in Santiago, so only 5 d were filtered out. Fig. S1 in the Supplement shows the time series of solar radiation overlapping mostly sunny days to all days.

To further explore ozone production under additional precursor conditions that are typically observed within a year, we ran simulations for average conditions per season in 2022. Thus, we run the model for a typical sunny day in austral summer (February and March, data in January was not included because of poor quality), fall (April, May), winter (June to August), and spring (September to November). As in the previous case, we filtered out overcast days from the time series. Mean conditions were found by overlapping meteorological and air quality data (1 h resolution) in 24 h plots and finding the average for every hour, namely the mean diurnal variation (MDV). These mean conditions were used as inputs to run the model. Model input variables are described in Sect. 3.

As in the case of Santiago, we performed detailed photochemical modeling for sunny days in March 2021 (11 overcast days were filtered out, Fig. S1) and for typical sunny conditions in 2022. Thus, we ran the model for typical conditions in January, often a sunny month; June–July, a time with low precursors due to summer vacation; September, an equinox month with increased precursors due to the return from summer vacation; and October–November, a time of transitioning into the rainy season. Mean conditions (sunny days only) for each period were found in a similar way as it was done for Santiago.

Air quality data and meteorological conditions (1 h) measured at Parque O'Higgins station were obtained from the Air Quality National Information System maintained by the Ministry of Environment (SINCA, Spanish acronym, https://sinca.mma.gob.cl/, last access: 1 October 2024). From network information, ozone is measured with UV photometry, NO_x with chemiluminescence, CO with infrared photometry, and published data complies with quality control standards according to national legislation (https://sinca.mma.gob.cl/index.php/documentos, last access: 1 October 2024). Figure S2 in the Supplement depicts the time series of measurements in March 2021. For comparisons with Quito, the time coordinates in all figures in this work are presented in UTC-5 (local and solar time in Quito and roughly solar time in Santiago).

All data described in this section are available as complete input files for the F0AM model for Quito (Cazorla et al., 2025a) and Santiago (Cazorla et al., 2025b), see also the Data Availability statement.

Ozone is measured at EMA USFQ with a Thermo 49i UV photometer. Measurements are periodically intercompared against a 2B-Technologies UV photometer. The agreement between measurements is better than 5 %. A Teledyne 400 chemiluminescence instrument is used to measure NO-NO₂-NO_x. Calibration is done with a certified NO standard and zero air to prepare calibration mixtures. Uncertainty in measurements is better than 5 % (Cazorla, 2016). The original rate of acquisition of ozone and NO_x measurements is 1 s. Meteorological measurements are also available at EMA USFQ with an original rate of acquisition of 30 s. CO is not measured at EMA USFQ. However, the car fleet is similar in the entire city. Hence, following previous work (Cazorla et al., 2021), we used an average of CO measurements from three stations (Belisario, Centro and Tumbaco) run by the Quito Air Quality Network (Secretariat of the Environment, Quito, Ecuador) (https://aireambiente.quito.gob.ec/, last access: 15 August 2024). Figure S3 depicts the time series of Quito observations.

VOC measurements in Santiago were taken from a fourth story window at the Geophysics Department of Universidad de Chile (33.457° S, 70.662° W, 535 m a.s.l.) from 17 to 28 March 2021, corresponding to the late summer. The site is located at campus Beauchef, in downtown Santiago, 770 m away from Parque O'Higgins station, where ozone, carbon monoxide, nitrogen oxides, and meteorological parameters are measured.

VOCs were measured using PTR-TOF-MS (proton transfer reaction time of flight mass spectrometry) at 1 s resolution. The instrument used was an Ionicon Analytik GmbH, Model 1000. A detailed description of the sampling operation of the PTR-TOF-MS instrument, calibration and mass discrimination methodology as well as the limit of detection of measured compounds is given in Appendix A. Table 1 shows the VOCs utilized in this study. Grouping was done for the practical purpose of organizing model runs, as explained in Sect. 2.5. The ISO nomenclature contains isoprene and monoterpenes also for practical reasons (isoprene contributes to P(O₃) formation but the monoterpene contribution was found to be marginal). In Appendix B (Fig. B1), we present the time series of measured VOCs. Based on details in Appendix A, we estimate a conservative error of 30 %–40 % in all VOC observations.

Previous studies conducted in numerous cities demonstrate that ambient measurements of VOCs, especially those whose source is fuel in vehicles, linearly correlate with CO, an anthropogenic tracer of combustion (Baker et al., 2008; Bon et al., 2011; Borbon et al., 2013; Brito et al., 2015). These correlations are useful as they can be used as emission ratios (Borbon et al., 2013). Thus, we obtained linear regressions between VOC and CO observations in Santiago for the campaign period. To this end, we linearly correlated the diurnal cycle of each VOC with the diurnal cycle of CO. We checked the results by correlating nighttime VOC and CO measurements and subtracting the background CO (noontime concentration) from the time series (Borbon et al., 2013; Brito et al., 2015). Both approaches yielded about similar slope values, but the determination coefficients were generally lower for the second approach. However, the determination coefficients improved for cresol, acetaldehyde, acetone, and isoprene. Thus, only for the latter compounds we used the linear fits obtained with the second approach. Table 1 shows the linear fit parameters used in this work. Table S1 contains the parameter regressions obtained with both approaches. Graphs for the VOC vs. CO regressions for benzene and toluene can be found in Figs. S4 and S5. The entire compendium of regression figures (both methods) can be found in the link indicated in the Data availability statement.

Compared to VOC/CO ratios (slope of linear fit in pptv ppbv⁻¹) reported for other cities in Latin America, namely Mexico City (Bon et al., 2011) and São Paulo (Brito et al., 2015), the slope we found for benzene (2.57) is about twice the value reported for Mexico City (1.21) and 2.5 times the value for São Paulo (1.03). For toluene, the slope in this study (7.54) is also higher than in Mexico City (4.2) and São Paulo (3.1). For C9 aromatics, we found a value of 3.64, close to what was reported in Mexico City (2.8). The Mexico City and São Paulo studies were done about a decade ago. Further work is needed regionally to better compare these ratios across cities, preferably with more recent measurements.

Linear fits for compounds associated with traffic (10 cases) show R2>0.83 and correspond to alkenes, aromatics, and some oxygenated compounds (methanol, ethanol, and phenol). Some of the latter compounds could also have a biogenic origin, but finding exact proportions from each source was not within the scope of this study. The determination coefficient is less good for other oxygenated compounds, namely acetaldehyde, acetone, and acetic acid (R2∼0.61), which can be attributed to the secondary nature of these compounds. As expected, the linear fit for isoprene is less good (R2=0.56) due to its combined anthropogenic and biogenic origin. In contrast, monoterpenes correlate well (R2=0.82), for which we associate these compounds with traffic sources in addition to biogenic sources as observed in other cities (Borbon et al., 2023).

Direct measurements of VOCs were not available in Santiago on all days of March 2021 or on typical days per season in 2022. In addition, the available measurements were subject to methodological limitations, as previously discussed. Given the error in measurements and the uncertainty of applying regressions in Table 1 to derive VOCs at times other than the campaign period, we ran 10 000 Monte Carlo (MC) simulations within ±35 % of the known values to find an array of inputs to the F0AM model. Thus, we ran individual MC simulations for 22 compounds (Table 1). From the bulk of each MC simulation, we extracted percentiles 10 to 90, which covered the simulation range. Figure S6 depicts the flow diagram followed to generate MC simulations. In Fig. S7, we present the MC simulations for benzene (time series and percentile distribution as diurnal cycles). The entire compendium of MC simulation figures for all VOCs are accessible through the link provided under Data Availability.

Measurements of VOCs are unavailable in Quito. For this reason, we used a Monte Carlo approach to simulate VOCs and to find an array of inputs to the model. VOCs in Quito could be derived from CO provided suitable emission ratios were available (Borbon et al., 2013). Given the lack of experimental data, we generated VOCs using CO measurements in Quito and the regressions in Table 1. Subsequently, we perturbed these data 10 000 times within ±50 % through Monte Carlo simulations. This conservative factor was applied based on the observation that CO in Quito varies from CO in Santiago by about ±25 % at noon, but during rush hours this variability could be up to ±50 %. From the bulk of the simulations, we extracted percentiles 10 to 90. This process was applied to all time periods analyzed in this study. In Fig. S8, we present the flow diagram followed to generate VOCs.

In Quito, 98 % of ambient CO is a primary emission that comes from on-road traffic (Hernandez and Mendez, 2020; Parra, 2017; Vega et al., 2015). Hence, alkene and aromatic compounds such as propene, butene, benzene, toluene, and xylenes, as well as C8 and C9 aromatics, are expected in the Quito environment. Fuel options such as gasoline mixtures with ethanol are not currently commercialized in Quito, but they are available in some other cities, mainly in the coast of Ecuador. Thus, we included ethanol and methanol to investigate the potential effect if these types of fuels were available. In addition, methanol can also be present from biogenic sources. Aldehydes and ketones are usually present in an urban environment as they are byproducts of oxidation and photolysis reactions.

We present MC simulations for benzene in Fig. S9. The entire compendium of MC simulation figures for all VOCs for Quito can be found in the link provided under Data availability.

The time step for the March 2021 model runs for Santiago and Quito was 10 min. To this end, the original Quito data were integrated into 10 min resolution. The same was done with Santiago 1 min VOC data. Regarding air quality and meteorological observations in Santiago, 1 h data were interpolated to generate 10 min time series. In contrast, the 2022 runs were done using 1 h data for average conditions in a season (Santiago) or month-grouping (Quito).

We constrained the model with ozone, CO, and meteorological observations from both cities. VOCs (obtained as described in the previous section) were also used to constrain the model. NO and NO₂ were not constrained but the NO_x sum was conserved (constrained). Following model recommendations, background concentrations were set to zero to avoid buildup of secondary species (Wolfe, 2023).

For boundary layer depth (PBLh) we used results from previous work done in Quito and Santiago. For Quito, we used the empirical model published by Cazorla and Juncosa (2018) that was inferred from balloon-borne measurements of PBLh and surface observations between 2014 and 2017. This approach was used to obtain PBLh in March 2021 and for the 2022 runs. For Santiago, we used results from the EMEP MSC-W model, that is an offline chemical transport model (Simpson et al., 2012), as recently applied in Santiago and other South American cities (Pachón et al., 2024). For the seasonal runs in 2022, we scaled the March 2021 PBLh with an empirical factor using, as reference, previous work (Muñoz and Undurraga, 2010). Thus, we used 1/3 of the summertime values for winter PBLh and 2/3 for fall. Figure S10 shows PBLh for Quito and Santiago in March 2021.

As stated in the F0AM model documentation, dilution is treated as a first order sink (Wolfe, 2023). Thus, we calculated a first order constant of dilution (kdil in the model) for each city by relating the time evolution of the PBL (m s⁻¹) to PBLh (m) at every step of the run. The bottom panels in Fig. S10 show kdil for both cities.

For the VOC input, compounds were distributed among explicit VOCs available in the MCM using weighing factors indicated in Table S2 in the Supplement. For example, compounds measured as butanone/butanal were distributed as 50 % butanone and 50 % butanal. These correspond to MCM notations MEK and C₃H₇CHO, respectively. Thus, a total of 36 VOCs were input to the model resulting in 9189 reactions. Compound-grouping and notation used in this study, along with all input species are depicted in Table 2.

Frequencies of photolysis were modeled using the MCM option in the F0AM for both cities. In the Results section we compare the mean diurnal variation of JNO₂ and JO¹D between both cities and with previous studies. For the ozone column and albedo, we used 1 h area-averaged MERRA-2 (Modern-Era Retrospective analysis for Research and Applications, Version 2) data sets for Quito and Santiago (Global Modeling and Assimilation Office (GMAO), 2015a, b). This data selection was based on previous work that showed good performance of MERRA-2 products for total column ozone in the region (Cazorla and Herrera, 2022). Table S3 contains a summary of options chosen in the F0AM for model runs.

We ran a series of experiments to investigate the individual contribution of VOC groups to P(O₃) as well as the sensitivity of P(O₃) to different combinations of VOC and NO_x levels.

For determining the contribution of VOC groups to P(O₃), we ran the model starting with alkenes (ALK) and adding one VOC group at the time until completion of all groups (ALK, ARO, OXY, ALD, and ISO). From these runs, we determined the percentage contribution of each group to P(O₃) in each city. For the last group (ISO), the percentage contribution refers to isoprene because adding monoterpenes essentially did not cause change.

We tested the sensitivity of P(O₃) to VOC levels P10, P50, and P90 (percentiles 10, 50 and 90 from MC simulations) under observed NO_x concentrations, 25 % NO_x reductions, and 75 % NO_x reductions in each city. VOC level P10 (P90) corresponds to a 30 % decrease (increase) from P50 in Santiago and 40 % in Quito. The group contribution runs were done using P50. In addition, we performed seasonal runs in 2022 for the entire VOC range (P10, P50, and P90) under observed NO_x levels in each city. Thus, we present the results from a total of 50 model runs (25 per city).

Ozone maxima in Santiago ranged between 30 to below 70 ppbv in March 2021, as shown by percentiles 10 and 90 of diurnal profiles obtained with 1 h observations in Fig. 2a. Looking at the entire data set, half of the days in the month had ozone higher than 50 ppbv and there was a day with a maximum of 100 ppbv on 4 March (the school year started on 1 March). Unlike Santiago, Quito does not have a well-established ozone problem or season, despite equatorial solar radiation at high altitude and urban emissions. Usually, 1 h data remain below 50 ppbv. For example, in March 2021, ozone maxima during mostly sunny days ranged between 30 to 50 ppbv for percentiles 10 and 90 as depicted in Fig. 2a. In previous years, high ozone in Quito (up to 80 ppbv) has been recorded episodically and has been associated with wildfires in the surrounding forests, usually during the month of September (Cadena et al., 2021). During the study period, PM_2.5 levels (24 h mean) were 17.6 µgm-3 in Quito and 22 µgm-3 in Santiago.

Regarding levels of NO_x, ambient concentrations of NO and NO₂ in Santiago in March 2021 were greater than in Quito by a factor of 3 for the maximum at the 50th percentile, as depicted in Fig. 2b and c. However, levels of CO at the 50th percentile from the mid-morning to noon were about similar (Fig. 2d), although data dispersion is within ±25 %. During the rush hours CO can differ by ±50 %. The CO and NO diurnal profiles show strong traffic signatures in the morning and evening rush hours over Quito. In contrast, the morning rush hour in Santiago stands out more, while the evening signature is delayed and less prominent. These features are associated with the urban activity and work habits of citizens in both cities. In Santiago, work hours extend into the evening and night, often past the work schedule, and citizens take a long time to return to their homes. In Quito, citizens usually leave work before dark and usually do not extend the 8 h work schedule.

The diurnal NO2/NOx ratio is an indicator of photochemical activity. It involves nitric oxide (NO) reacting with hydroperoxyl (HO₂) and alkyl peroxy (RO₂) radicals to form NO₂. The shape of the diurnal cycle of this ratio is similar between Quito and Santiago. However, Santiago's photochemical activity starts earlier than in Quito as given by an earlier increase of NO2/NOx (Fig. 3) right after the rush hour dip (due to increased NO_x from traffic emissions). This is related to the earlier start in traffic in Santiago as seen in NO and CO (Fig. 2b and d), which in turn leads to a distinct NO₂ morning maximum that is not present in Quito (Fig. 2c). At noon and in the afternoon, NO2/NOx at the 50th percentile (observations) is about similar in both cities, which indicates similar photochemical activity.

Photolysis reactions are key to radical and ozone production during daylight hours. Being that Quito and Santiago are located on the Equator and in the subtropics, respectively, the magnitude of frequencies of photolysis are different due to the intensity of solar radiation at both locations. From model output, the diurnal variation at the 50th percentile of frequencies of photolysis of ozone (towards the production of O¹D) and NO₂ in March 2021 are presented in Fig. 4. As expected, both quantities are greater in Quito than in Santiago due to its equatorial location and high altitude. Previous work on the oxidation capacity of the Santiago air reported direct measurements of JO¹D and JNO₂ taken during a field campaign in March 2005 (Elshorbany et al., 2009b). These measurements for diurnal maxima in Santiago were JNO2∼0.008 and JO¹D ∼2.5×10-5 s⁻¹, which are similar to mean values obtained through modeling in this work.

Ozone production rates were modeled for a range of VOC inputs and so are discussed in terms of ranges bounded by percentiles P10 and P90 obtained from the VOC distribution in each city (Fig. 5a). For Santiago, the range represents the uncertainty in the P(O₃) simulation given the uncertainty in VOC measurements. Thus, the best estimation of P(O₃) for Santiago is represented by P50 in Fig. 5a. The uncertainty was determined from the mean percentage difference of the upper and lower boundaries from percentile 50 and corresponds to ±32 % (Fig. S11). For Quito, P10 and P90 set the boundaries of the possible range of P(O₃) magnitudes expected for Quito calculated to the best of our knowledge from a distribution of VOC inputs. Due to a lack of in situ measurements we deal with the entire range when discussing Quito and we only present P50 curves in a referential way. The mean variability from percentile 50 is ±42 % (Fig. S11).

The range of net ozone production rates at both cities at noon are comparable (22–45 ppbv h⁻¹), but in the mid-morning values could be higher for the upper boundary in Quito (up to 55 ppbv h⁻¹) (Fig. 5a). Further below we show that in Quito the HO₂ and RO₂ abundance is greater when compared to Santiago (Fig. 9), but NO levels are lower (Fig. 2b). Thus, there is a compensating effect in the levels of HO₂ and NO in both cities in a way that application of Eq. (1) yields about similar ozone production rates. Previous work that determined net ozone production rates in Santiago from measurements and model calculations in 2005, reported substantially higher values for the mean P(O₃) diurnal maximum (160 ppbv h⁻¹) (Elshorbany et al., 2009a). Such high value is outside of our calculation range, although it is possible that after almost 20 years of reassessing P(O₃) in Santiago, conditions changed. P(O₃) maxima presented in Fig. 5a for percentiles 50 of the distributions in both cities are consistent with values obtained in previous studies. For example, direct measurements of P(O₃) in Houston in 2009 were about 30 ppbv h⁻¹ on average (Cazorla et al., 2012). Another example is a recent sensitivity study of ozone production to NO_x and VOCs in New York city that found P(O₃) within 26–37 ppbv h⁻¹ from modeling work with the F0AM (Sebol et al., 2024).

Losses to nitric acid are higher in Santiago than in Quito (Fig. 5b). This is consistent with higher NO_x levels present in the Santiago ambient air. Meanwhile, losses to hydrogen peroxide are higher in Quito but the magnitude of this loss is one order of magnitude lower that the loss to nitric acid (Fig. 5c). Other losses are presented in Fig. S12. The loss due to ozone photolysis and the reaction of ozone with HO₂ are two and one orders of magnitude lower than the loss to nitric acid, respectively. The reaction of ozone with alkenes is also an order of magnitude lower than the loss to nitric acid, but this is higher than the loss to hydrogen peroxide. The upper boundaries of the loss to alkyl nitrates (P(RONO₂)) approach a digit and become important along with the loss to nitric acid in both cities.

NO and NO₂ were unconstrained in the model, but the NO_x sum was constrained. A comparison of measured and modeled NO/NO2 ratio is presented in Fig. S13. From the mid-morning to the early afternoon values are about similar.

An interesting point to investigate in the future is the connection between ozone production rates and ambient ozone, whose budget not only depends on atmospheric chemistry but also on meteorological and circulation factors, including long-range transport. From our calculations, the range of P(O₃) in both cities is similar but ambient ozone is higher in Santiago. This city has a year-round influence of the Subtropical Pacific High that determines a strong subsidence inversion, which is often strengthened by low-level lows and occasionally disrupted by passing synoptic disturbances (Gallardo et al., 2002; Garreaud et al., 2002; Huneeus et al., 2006; Muñoz and Undurraga, 2010). Such meteorological condition is known to be the cause of pollutant accumulation in the boundary layer in Santiago. In contrast, Quito is not affected by a similar high-pressure system. In our simulations, we used information of the boundary layer evolution in both cities. The boundary layer in Quito can often grow up to 2200 m a.g.l. (Fig. S10) (Cazorla and Juncosa, 2018). Under such conditions, the first order dilution constant in the model reached up to 2.5×10-4–5.8×10-4 s⁻¹. Meanwhile in Santiago, the boundary layer can reach up to 1500 m and dilution constants usually stay below 2.5×10-4 s⁻¹ (Fig. S10). Additional research that uses a chemical transport model is needed to explain ambient ozone levels.

The chemical regime of ozone production in Santiago is strongly VOC-limited as indicated by the ratio LNOx/(LNOx+LROx) being very close to one for the entire range of the model output as depicted in Fig. 6a. This indicator is consistent with the formaldehyde to reactive nitrogen ratio (HCHO/NOy) being well below 1 (Fig. 6b), and with the losses to nitric acid being substantial (Fig. 5b).

For the Quito runs, LNOx/(LNOx+LROx) points to a VOC-limited regime as this indicator is above suggested thresholds of 0.5 (Kleinman, 2005c; Kleinman et al., 2001) or revisited 0.74 (Schroeder et al., 2017). Meanwhile, the HCHO/NOy ratio is still below 1 for the entire range (VOC-limited) even though the upper boundary approaches one digit. This indicates that with a higher VOC content, the Quito environment is more prone to transitioning towards a more NO_x-limited regime. At the moment, a lack of in situ measurements constrains our ability to define how close the Quito environment is from the transition region. Even though from our calculations both cities are in the VOC-limited chemical regime, this character is noticeably more intense in Santiago from the indicators in Fig. 6. This finding illustrates that within the VOC-limited and NO_x-limited categories of ozone production there is a range of intensities that depend on the unique conditions that surround every urban area.

The percentage contribution to P(O₃) caused by every chemical group (Table 2) during daylight hours in both cities is depicted in Fig. 7a (Quito) and 7b (Santiago). For the Quito simulations, alkenes and aromatics combined (blue and cyan in Fig. 7a) contribute 45 % of ozone formation around the morning rush-hour, 70 % at noon, and up to 90 % in the afternoon. Meanwhile, the contribution of these two groups to P(O₃) in Santiago is under 50 % throughout the day. The contribution of aldehydes and ketones is substantial over Santiago and, when combined with alkenes and aromatics, adds up to 80 %–90 % of P(O₃) during daylight hours (magenta in Fig. 7b). In Quito, aldehydes and ketones are important in the morning, between 07:00–08:00, when the percentage contribution could be up to an additional 40 %–50 %, while at noon and in the afternoon this contribution is about 20 %. Isoprene contributes an additional 10 %–20 % in Santiago and becomes increasingly important towards the afternoon. In Quito, we estimate that this contribution is about 10 %. It must be noted that Central and Southern Chile suffered a long-standing drought between 2010 and 2022 (Garreaud et al., 2020) that likely caused an increase in isoprene emissions by vegetation due to hydrological stress (Jiang et al., 2018). The contribution of oxygenated compounds in both cities is marginal (black in Fig. 7). Future work needs to evaluate the contribution of alkanes in both cities as liquid petroleum gas is used for heating and cooking (Chen et al., 2001).

The main sources of radical production are depicted in Fig. 8 and correspond to the photolysis of ozone followed by the reaction of O¹D with water vapor (Fig. 8a), the photolysis of formaldehyde (Fig. 8b), the photolysis of HONO (Fig. 8c), and the reaction of ozone with alkenes (Fig. 8d). The first contribution has the largest magnitude in both cities. Computing this rate does not depend on VOC abundance for which percentile boundaries are absent in Fig. 8a. Thus, the curves depict the mean diurnal cycles whose maxima are about 1 pptv s⁻¹ in both cities, on average. However, within the month, variability in Santiago (0.5–1.75 pptv s⁻¹) is somewhat greater than in Quito (0.75–1.5 pptv s⁻¹). From meteorological data, the water vapor content at solar noon in Quito and Santiago is similar (10 vs. 9 g kg⁻¹) but solar radiation does not vary much on the Equator while in Santiago daylight progressively lowers as the season advances towards the fall.

P(HO_x) production from formaldehyde photolysis in Quito and Santiago is comparable (0.18–0.35 pptv s⁻¹) (Fig. 8b). Overall, this contribution is 2–3 times lower than the first source. In the morning, radical production from the photolysis of HONO is equally important in Santiago and in Quito (0.35–0.5 pptv s⁻¹), as indicated in Fig. 8c, which is expected given the high abundance of NO_x in the morning. However, the contribution from this radical source gradually decreases from the mid-morning into the afternoon. The contribution from the reaction of alkenes and ozone is higher in Santiago (Fig. 8d), which is consistent with higher ozone levels. Previous work determined P(HO_x) from different sources in Santiago (Elshorbany et al., 2009b). They found rates of HO₂ production from formaldehyde photolysis of 0.15 ppt s⁻¹, on average, and rates of OH production from HONO photolysis of 0.4 ppt s⁻¹. Our calculations are consistent with these findings. However, rates of OH production reported by Elshorbany et al. (2009b) from ozone photolysis are substantially lower (0.27 ppt s⁻¹, on average). It is important to remark that this study only considers gas phase chemistry and urban emissions following previous work (Elshorbany et al., 2009a, b; Ren, 2003; Ren et al., 2013).

Radical abundances in both cities during the study time period are illustrated in Fig. 9. With higher frequencies of photolysis, the OH abundance in Quito for the entire simulation range (0.33–0.38 pptv) is overall greater than in Santiago (0.23 pptv) (Fig. 9a). The order of magnitude of radical abundances is similar to those found in other studies (Cazorla et al., 2021; Dusanter et al., 2009; Ren et al., 2013). Given the greater OH availability in Quito, reactions with VOCs yield HO₂ and RO₂ abundances that at noon in Quito are about three times greater than in Santiago (Fig. 9b and c). From the magnitudes in the latter figures, HO₂ and RO₂ are equally important, which is consistent with previous work (Bottorff et al., 2023; Dusanter et al., 2009; Elshorbany et al., 2009b).

In this section we explore the effect on P(O₃) due to changes in NO_x and VOCs that could arise from potential actions intended to improve air quality and help reduce SLCF. As stated before, diesel-based public transportation in South American cities is an important source of PM_2.5 and NO_x. It follows that a relevant scenario to consider is the effect on ozone production if this type of transportation were replaced, as it is often included in air quality and climate mitigation planning. Notice that Santiago's public transport bus fleet is roughly 10 % electric buses, a fraction that is expected to increase in the future. This type of action would significantly lower ambient NO_x (and PM_2.5), although VOCs would also decrease. Another relevant scenario is exploring changes in VOCs. For example, applying changes in the quality of fuels to reduce the content of aromatic compounds would influence P(O₃). Here, we explore some sensitivity scenarios, but additional research is needed to precisely determine the proportion in precursor reduction associated with specific environmental actions.

The effect on P(O₃) with different combinations of NO_x and VOCs is quantified in Fig. 10. The scenarios investigated were those of current NO_x levels, NO_x reductions by 25 %, and NO_x reductions by 75 % at VOC levels P10, P50 and P90. For the latter, the percentage decrease (or increase) from P50 to P10 (or to P90) is 30 % in Santiago and 40 % in Quito. In Santiago and Quito, lowering the levels of VOCs would result in lower ozone production rates at all tested levels of NO_x as presented by the progression of panels in Fig. 10a to c (Quito) and 10d to f (Santiago). This feature is expected in VOC-limited environments. For example, at the current NO_x Quito levels, reducing VOCs by 40 % from the P50 level reduces the maximum P(O₃) by 37 %. In Santiago, a 30 % decrease in VOCs would result in a 32 % decrease in the maximum P(O₃).

At lower levels of VOCs in Quito (P10), a 25 % decrease in NO_x causes a slight P(O₃) increase in the morning and almost no change in the afternoon (Fig. 10a), while a 75 % NO_x reduction modestly raises the morning P(O₃) but causes a drop at noon. As VOC levels increase (P50 and P90), a 25 % reduction in NO_x causes a modest decrease in P(O₃), but a 75 % decrease causes the noon P(O₃) to decrease by half. This result indicates that at higher VOC levels in Quito the ozone production chemistry is prone to transitioning towards a more NO_x-limited regime, especially at noon and in the afternoon. This aspect underscores the necessity of implementing in situ measurements of VOCs in Quito to accurately determine how close the Quito environment is to the transition region.

In Santiago, at the current levels of VOCs and NO_x (P50), lowering NO_x increases ozone production rates, especially in the morning and mainly when the decrease is substantial (75 %), but the change at noon and in the afternoon is modest. Thus, at 09:00 a.m. P(O₃) grows from 25 to 38 ppbv h⁻¹ (Fig. 10e). Similar effects are observed at the lowest (P10) and highest (P90) VOC levels (Fig. 10d and f). Thus, P(O₃) increases in the morning and either there is a modest change or a decrease from noon to the afternoon. Therefore, NO_x decreases cause the peak P(O₃) to shift from noon to morning.