Oxidation capacity of the city air of Santiago , Chile

The oxidation capacity of the highly polluted urban area of Santiago, Chile has been evaluated during a summer measurement campaign carried out from 8–20 March 2005. The hydroxyl (OH) radical budget was evaluated employing a simple quasi-photostationary-state model (PSS) constrained with simultaneous measurements of HONO, HCHO, O3, NO, NO2, j (O1D), j (NO2), 13 alkenes and meteorological parameters. In addition, a zero dimensional photochemical box model based on the Master Chemical Mechanism (MCMv3.1) has been used to estimate production rates and total free radical budgets, including OH, HO 2 and RO2. Besides the above parameters, the MCM model has been constrained by the measured CO and volatile organic compounds (VOCs) including alkanes and aromatics. Both models simulate the same OH concentration during daytime indicating that the primary OH sources and sinks included in the simple PSS model predominate. Mixing ratios of the main OH radical precursors were found to be in the range 0.8– 7 ppbv (HONO), 0.9–11 ppbv (HCHO) and 0–125 ppbv (O 3). The alkenes average mixing ratio was ∼58 ppbC accounting for ∼12% of the total identified non-methane hydrocarbons (NMHCs). During the daytime (08:00 h–19:00 h), HONO photolysis was shown to be the most important primary OH radical source comprising alone ∼55% of the total initial production rate, followed by alkene ozonolysis ( ∼24%) and Correspondence to: J. Kleffmann (kleffman@uni-wuppertal.de) photolysis of HCHO ( ∼16%) and O3 (∼5%). The calculated average and maximum daytime OH production rates from HONO photolysis was 1.7 ppbv h −1 and 3.1 ppbv h −1, respectively. Based on the experimental results a strong photochemical daytime source of HONO is proposed. A detailed analysis of the sources of OH radical precursors has also been carried out.

The oxidising capacity of the atmosphere determines the rate of their removal, and hence controls the abundance of these trace gases. Understanding the processes and rates by which species are oxidized in the atmosphere is thus crucial to our knowledge of the atmospheric composition of harmful and climate forcing species. The term "oxidation capacity", OC is defined as the sum of the respective oxidation rates of the 5 molecules Y i (VOCs, CO, CH 4 ) by the oxidant X (X =OH, O 3 , NO 3 ) (Geyer et al., 2001): where k Y i is the bi-molecular rate constant for the reaction of Y i with X . The hydroxyl radical (OH) is the primary oxidant in the atmosphere, responsible for the oxidation and removal of most natural and anthropogenic trace gases. In addition, 10 initiating oxidation by reaction with the OH radical leads to the formation of harmful oxidants, such as ozone (O 3 ) and peroxyacetylnitrate (PAN). Thus, the identification and quantification of the different atmospheric OH radical sources and sinks is of paramount importance. Primary sources of the OH radical include the photolysis of ozone followed by the subsequent reaction of the excited O 1 D atom with water, photolysis of formalde-15 hyde (HCHO) in the presence of nitrogen oxide (NO), direct photolysis of nitrous acid (HONO) and the reactions of unsaturated hydrocarbons with O 3 . Ren et al. (2003) recently calculated the relative importance of the above sources of OH in New York and estimated HONO photolysis contributed up to ∼60%. In other field work studies, unexpected high daytime values of HONO were observed (e.g. Zhou et al., 2002;Kleffmann 20 et al., 2002Kleffmann 20 et al., , 2005Acker et al., 2006a, b) and new photochemical HONO sources have been proposed (Kleffmann, 2007), some of which have recently been identified in the laboratory George et al., 2005;Stemmler et al., 2006Stemmler et al., , 2007Bejan et al., 2006;Li et al., 2008). Summertime urban OH and HO 2 radical budgets have been evaluated in several 25 field campaigns (e.g., George et al., 1999;Holland et al., 2003;Ren et al., 2003;Heard et al., 2004;Volkamer et al., 2007;Emmerson et al., 2007;Kanaya et al., 2007).
In most of these studies, the experimental measurements were complemented with 19125 model simulations in order to understand the chemical mechanisms that control tropospheric urban chemistry. Interestingly, the urban daytime OH and HO 2 radical budgets have been shown to be better simulated during the summer rather than winter, especially for high NO x environments. Ren et al. (2006) used a box model incorporating the Regional Atmospheric Chemistry Mechanism, (RACM; Stockwell et al., 1997), which 5 is based on the lumping technique to simulate radical budgets in New York during a winter campaign carried out in 2004 and obtained a median measured to model ratio of 0.98 for OH. However, the RACM model significantly underestimated HO 2 , both during day and at night, with median measurement to model ratio of 6.0. Similarly, during the IMPACT campaign in Tokyo the RACM model reproduced wintertime OH well but 10 underestimated the HO 2 by a median factor of 2. However, during the summer, the RACM model generally reproduced the daytime OH and HO 2 reasonably well (Kanaya et al., 2007). For Mexico City, Shirley et al. (2006) reported a median measured to model OH ratio of 1.07 during the morning and night and 0.77 during the rush hour using the RACM model. For HO 2 , median measured to model ratios of 1.17, 0.79 and 15 1.27 were determined during the morning rush hour, midday and night, respectively. Besides lumped mechanisms, the more explicit Master Chemical Mechanism, MCM (http://mcm.leeds.ac.uk/MCM/; Jenkin et al., 1997;Saunders et al., 2003;Bloss et al., 2005) has been used extensively to interpret field measurements, carried out under a variety of conditions, including urban environments (e.g. Mihelcic et al., 2003;Emmer-20 son et al., 2005aEmmer-20 son et al., , b, 2007. During the BERLIOZ campaign, which took place in Berlin in August 1998 (Mihelcic et al., 2003), the hydroxyl and peroxy radical (RO 2 ) budgets have been measured and compared to those calculated by a photochemical box model containing the MCM. The modelled OH concentrations were found to be in excellent agreement with the measurements under high-NO x conditions (NO x >10 ppbv). The 25 measured RO 2 /HO 2 ratio was also well reproduced by the model. The MCM modelled radical concentrations during the TORCH campaign, which took place ∼40 km NE of central London in the summer of 2003 also agreed well with measurements with only a 24% and 7% over prediction for OH and HO 2 , respectively (Emmerson et al., 2007).

19126
During the majority of the summer campaign studies reported in the literature the daytime peak in OH is well simulated, in the range of (3-10)×10 6 molecule cm −3 (Kanaya et al., 2007 and references therein). However, model OH production rate analysis has suffered from high uncertainties due to the use of estimated HONO concentrations rather than accurate direct simultaneous measurements (Heard et al., 2004;Emmer-5 son et al., 2005bEmmer-5 son et al., , 2007Kanaya et al., 2007). Using MCM constrained box model with estimated HONO concentrations, the diurnally averaged OH concentrations during the summer of 1999 PUMA field campaigns in Birmingham city centre was underestimated by a factor of ∼2 during the day especially under high NO x conditions (Emmerson et al., 2005a). This could be due to an underestimation of daytime HONO concentra-10 tions from using only known gas phase chemistry (Kleffmann et al., 2005). Thus, other photochemical sources have been proposed and recently identified in the laboratory, e.g. by the photochemical heterogeneous conversion of NO 2 on natural surfaces (George et al., 2005;Stemmler et al., 2006Stemmler et al., , 2007. It is worth noting that net average HONO contribution to radical budgets (defined as the HONO photolysis rate minus the radical loss rate due to the reaction OH+NO) of as low as 0% (Heard et al., 2004), and 3% (Emmerson et al., 2007;Kanaya et al., 2007) have been reported when HONO concentrations were only estimated. When the reaction of NO+OH was assumed as the unique HONO source, HONO was not a net source of OH radicals in the atmosphere (Heard et al., 2004). In other studies (Emmerson et al., 2005a(Emmerson et al., , 2007  It is clear, therefore, from results reported from several recent field studies (e.g. Ren et al., 2003;Kleffmann et al., 2003;Acker et al., 2006a, b), that the simultaneous measurement of HONO, along side other major radical precursors, is crucial in the analysis of atmospheric radical budgets. Several studies focusing on air quality issues in Santiago de Chile have shown that severe air quality problems, including the photochemical formation of large amounts of ozone, PAN and related photooxidants, have a significant impact on health problems in the city (Rappenglück et al., 2000;Rubio et al., 2004;Rappenglück et al., 2005). However, none of these studies have observed the diurnal variation of the important OH radical precursor HONO.

15
The work reported here focuses on the analysis of a comprehensive suite of data taken during a summertime field campaign carried out in the city of Santiago de Chile from 8-20 March 2005. This work constitutes the first detailed evaluation of photochemistry in Santiago, Chile that takes into account all the major primary OH radical sources, namely the photolysis of HONO, formaldehyde (HCHO) and ozone (O 3 ) and 20 the dark reactions of ozone with alkenes, in addition to peroxy radical (HO 2 and RO 2 ) recycling reactions. Under the high NO x conditions often experienced in Santiago, a constrained photochemical box model based around the MCM and a simple photostationary steady state (PSS) model were used to evaluate radical budgets and their source apportionment during late summer in order to understand the photochemistry 25 occurring in such a highly polluted urban environment as Santiago.

Measurement site
The measurements were performed downtown of the city of Santiago, Chile, on the third floor of the Physics Department of the University of Santiago (USACH) and in the Park O'Higgins station (POH), situated ∼1.8 km southeast of the main USACH 5 measurement site. The city of Santiago de Chile is located at -33.45 • latitude and 70.67 • longitude ∼550 m above sea level and surrounded by two mountain ranges, the Andes and the Cordillera de la Costa.

Measurement techniques
The techniques used to measure the different parameters are listed in Table 1 , 2006), in contrast to other intercomparison studies (e.g., Appel et al., 1990;Spindler et al., 2003). The excellent agreement can be explained by the active correction of interferences and the use of an external sampling unit, minimizing artefacts in sampling lines. Potential heterogeneous HONO formation on the walls of the USACH building on which the sampling unit was fixed (ca. 130 cm distance), was also investigated.
C 10 NMHCs were sampled at the USACH site on adsorption tubes and analyzed by GC-FID analysis following the US EPA Compendium Method TO-17 (see Table 2). The detailed analytical procedure is published elsewhere (Niedojadlo et al., 2007). The ambient NMHCs have been sampled using an automatic system equipped with calibrated regulated flow controllers and applying an air flow of 20 ml/min on the adsorbing tubes.

5
After sampling, the adsorption tubes were capped with Parafilm, stored in air sealed glass tubes in the refrigerator and returned to Germany for GC-FID analysis. Potential ozone interferences have been tested in the laboratory by sampling a standard VOC mixture over the same type of adsorption tubes with and without addition of ozone at a mixing ratio of 135 ppbv. Sampling periods of three hours were chosen using NMHCs 10 mixing ratios corresponding to the minimum observed NMHCs mixing ratios during the measurement campaign. Ozone was prepared by passing a regulated flow of pure synthetic air through a mercury UV-lamp based ozoniser followed by a reaction vessel with glass rings cooled with dry ice to 203 K in order to trap the HO x radicals from the ozonised air. Ozone has been monitored by a commercial UV absorption based 15 monitor (Table 1). Only reductions of as low as -8.8% for trans-2-butene and as high as -29.4% for cis-2-pentene were observed. However, since average and maximum ozone mixing ratios of 20 ppbv and 126 ppbv were observed during the measurement campaign, we exclude significant negative interferences from ozone. This result is in agreement with the study of Koppmann et al. (1995) who found no significant interferences from ozone up to mixing ratios of 100 ppbv either using pressurized air samples or cryogenically collected air samples even at very low VOC concentrations.

Modelling approach
2.3.1 Simple quasi-photostationary state model, PSS OH concentrations were calculated with the steady-state approximation using the rad- 25 ical production rates from HONO, HCHO and ozone photolysis, alkenes ozonolysis and the radical loss rate. Under the prevailing high NO x conditions radical loss is 19130 mainly governed by the reactions of OH with NO x (cf. George et al., 1999;Ren et al., 2006;Emmerson et al., 2005bEmmerson et al., , 2007Kanaya et al., 2007). During the day, formation of HONO by reaction of OH with NO is essentially balanced by photolysis of HONO formed from this reaction. Radical removal by peroxy-peroxy radical reactions is unimportant under high NO x conditions (see Sect. 3.3). Thus, the net radical loss rate can 5 be estimated from the rate of reaction of OH with NO 2 : The applied steady state approximation can be summarized as follows: The total rate of radical initiation, P R , is given by: For ozone photolysis Φ OH (defined here as the fraction of O 1 D produced that will react with H 2 O rather than is quenched to ground state O 3 P) was calculated using known rate constants for O 1 D quenching and reaction with water in addition to the measured water concentration. For the alkene ozonolysis reactions Φ OH represents 15 the OH yield from the respective reactions (e.g. Rickard et al., 1999). Therefore, the steady state OH concentration is given by: budgets. MCMv3.1 is a near-explicit chemical mechanism describing the detailed gas phase tropospheric degradation of methane and 135 primary emitted NMHCs, which leads to a mechanism containing ca. 5900 species and 13 500 reactions. The mechanism is constructed according to a set of rules as defined in the latest mechanism development protocols (Jenkin et al., 1997;Saunders et al., 2003;Bloss et al., 2005).

5
The MCM photochemical box model's system of simultaneous stiff ordinary differential equations (ODEs) was integrated with a variable order Gear's backward differentiation method (FACSIMILE; Curtis and Sweetenham, 1987). The model was constrained with average 10 min values of the following measured parameters: j (NO 2 ), j (O 1 D), relative humidity, pressure, temperature, NO, NO 2 , HONO, CO, HCHO, O 3 , PAN and 10 31 NMHCs (see Table 2). j (HONO) and j (HCHO radical ) where parameterized from the measured j (NO 2 ) and j (O 1 D) (Holland et al., 2003) and their values have been constrained in the model. The other photolysis frequencies are parameterized within the model using a two stream isotropic scattering model under clear sky summertime conditions (Hayman, 1997;Saunders et al., 2003). The photolysis rates are calculated as a function of solar zenith angle and normalized by a scaling factor, calculated from the ratio of measured and model calculated j (NO 2 ) values, which takes into account the effects of varying cloud cover. A series of rate of production analyses (ROPA) were carried out in order to identify the most important photochemical processes driving the formation and loss of OH and HO 2 . The MCM photochemical model was run for a pe-20 riod of five days, with the model being constrained with the same measured campaign average parameters each day, in order to generate realistic concentrations for the unmeasured intermediate species. By the fifth day the free radicals in the model have reached a photostationary state, which has been used for the data evaluation.
19132 3 Results and discussion

Measurements results analysis
For the data evaluation, all measurements were averaged over 10 min time intervals. The trace gases data of the whole campaign are shown in Fig. 1 while the 10 min average diurnal variation profiles are shown Fig. 2a. During the campaign sunny weather 5 conditions were prevailing with temperatures ranging from 285 K to 305 K during the daytime. The wind speed was relatively low ranging from 0.2 m s −1 to 4.1 m s −1 , and the average relative humidity was 49%, reaching up to 100% during the night. The maximum HONO mixing ratio during rush hour reached ∼7 ppbv on the 10 March at ∼9 h. For the campaign averaged data maximum and minimum HONO mixing ratios 10 of 3.7 ppbv at around 8 h and 1.5 ppbv around 17 h were obtained. For CO and NO a similar rush hour peak at ∼9 h on the 10 March was also observed with maximum concentrations of 3.6 ppmv and 480 ppbv, respectively. The average daytime rush hour maxima for CO and NO were 1.38 ppmv and 180 ppbv, respectively (see Fig. 2a). The NO 2 maximum was shifted later owing to 15 small direct emissions and formation by the reaction of NO with peroxy radicals and O 3 . From the slope of the correlation plot of HONO against NO x a mean HONO/NO x ratio of 0.008 was estimated during the rush hour peaks, which is in excellent agreement with direct tunnel measurements in Europe (Kurtenbach et al., 2001).
PAN, HCHO and O 3 showed typical diurnal variations with average daytime maxima 20 at about 14 h of 3 ppbv, 7 ppbv and 65 ppbv, respectively, demonstrating their photochemical formation. However, from the fast increase of HCHO in the early morning, when the O x (NO 2 +O 3 ) increase was still small, a significant contribution from direct emissions was also identified (see Sect. 3.8.1). In addition to the maximum at ca. 14 h, the ozone diurnal variation profile is characterized with an afternoon shoulder at 18 h, 25 which has become a typical feature under photochemical smog conditions in Santiago (Rappenglück et al., 2000(Rappenglück et al., , 2005 (Ren et al., 2003;Kleffmann et al., 2006;Acker et al., 2006b). The high mixing ratios and the daytime maximum of the HONO/NO x ratio (see Fig. 2b) in Santiago points to a very strong daytime HONO source. 53 measured NMHCs have been identified (see Table 2). The remaining 127 uniden-5 tified NMHCs represents in average about 43% of the total measured NMHCs. Total average measured NMHCs of ∼900 ppbC and known average measured NMHCs of ∼490 ppbC were determined, which correspond to average diurnal VOC/NO x ratios of 14 and 7, respectively. According to the VOC/NO x ratio rule (Sillman, 1999) the first value corresponds to a NO x -sensitive photochemical regime while the second corre-10 sponds to a VOC-sensitive photochemical regime. However, the VOC/NO x ratio may not correctly represent the sensitivity of a photochemical regime. An explicit VOC-NO x -O 3 sensitivity analysis showed that the photochemical regime in Santiago is clearly VOC sensitive (Elshorbany et al., 2008). Alkanes have the highest contribution (ppbC) to NMHCs (56%) followed by aromatic hydrocarbons (32%) and finally alkenes (12%).

Oxidation capacity
The loss rate of the VOCs and CO due to reactions with OH, O 3 and NO 3 has been calculated using the MCM model. The modelled NO 3 concentrations showed two peaks of 1.0×10 6 molecules cm −3 at about 13 h and of 8.4×10 5 molecules cm −3 at 19 h. The total number of the depleted molecules per day due to oxidation by OH, O 3 and NO 3 were 6.4×10 12 , 7.4×10 11 and 2.0×10 10 molecules cm −3 , respectively. Accordingly, the OH radical is the driving force of the oxidation capacity of the atmosphere in Santiago and thus, only the sources and 5 sinks of the OH radical are further considered in this study.

Radical production and destruction rates
The total production and destruction rates of OH and HO 2 calculated by the MCM model constrained to campaign averaged data are shown in Fig. 3a with the ratios of the radical production/destruction shown in Fig. 3b. The ratio was around unity 10 throughout the day for the hydroperoxy radical whilst the ratio for the hydroxyl radical reaches a maximum of ∼1.7 during the morning similar to that observed in other urban studies (Mihelcic et al., 2003;Shirley et al., 2006;Sheehy et al., 2008), which is caused by the photolysis of night time accumulated HONO. The high total production and destruction rates are dominated by the recycling re-15 actions of peroxyradicals (RO 2 +NO and HO 2 +NO). The main loss of RO 2 is due to its reaction with NO with an average daytime loss rate of ∼34.6 ppbv h −1 , which accounts for most of the HO 2 production. The next most important HO 2 sources are the reactions of OH with CO and HCHO with average daytime production rates of ∼0.5 and ∼1.  (Mihelcic et al., 2003). The loss rates due to the HO 2 self-reaction and its cross-reactions with RO 2 are very small with daytime averages of <0.01 and 0.02 ppbv h −1 , respectively, in agreement with other urban studies (e.g., George et al., 1999;Ren et al., 2006). The main OH loss route is through its reaction with hydrocarbons, followed by reactions with NO and NO 2 . The rates of OH destruction due 5 to hydrocarbons oxidation depend on the detailed chemical mechanism and can be estimated using the following relationships: or where L OH (total) is the total loss rate as calculated by the MCM model and k i represents the bimolecular rate constant for OH reaction with the corresponding VOC. If Eq.
(2) is used to calculate L OH due to reactions with the measured VOCs only, the OH loss rate will be underestimated since reactions with secondary VOC products are not included. Consequently, HO 2 as a source of OH will be over estimated. Rela-15 tionship (1) takes into account the detailed degradation of the VOCs due to reactions with OH, as calculated by the MCM photochemical box model, which includes the secondary VOC oxidation products. The average daytime (8 h-19 h) loss of OH radicals by VOC reaction calculated employing Eq.
(2) is about 6 ppbv h −1 while that obtained using Eq. (1) is about 28.4 ppbv h −1 , representing about 79% of the total OH loss. The 20 fraction of OH loss by VOC reactions is similar to that of Berlin, 50-70% (Mihelcic et al., 2003) and Mexico City, 72% (Shirley et al., 2006). OH production is dominated by the recycling reaction of HO 2 with NO, P OH (HO 2 →OH) recycled for which: Introduction  (Fig. 4a). The P OH (HO 2 →OH) route accounts for ∼80% of the total OH radical production. This value is comparable to that estimated during 5 TORCH, 80% (Emmerson et al., 2007) and BERLIOZ, >70% (Mihelcic et al., 2003) and Mexico City, >80% (Shirley et al., 2006 andSheehy et al., 2008). The oxidation of hydrocarbons results however in the production of other radical precursors namely, O 3 and HCHO as by-products in addition to alkene ozonolysis as a subsequent process. These processes, in addition to HONO photolysis, constitute the net radical produc-10 tion term, P R (Fig. 4a). The rest of the OH production term (1.1 ppbv h −1 ) is mainly due to the ozonolysis of the secondary alkenes produced from the oxidation process which are not constrained by the measurements (see Fig. 5). The balance between P OH (HO 2 →OH) recycled and L OH (OH+VOC) (see Sect. 3.5), results in the NO 2 +OH (termination) reaction becoming the net dominant sink for OH with a maximum loss 15 rate of 6.4 ppbv h −1 and a daytime average loss rate of ∼3.4 ppbv h −1 (see Fig. 5).
An accompanying sensitivity analysis showed that only under very low NO x conditions reaching three orders of magnitude lower than the current levels (i.e. ∼0.1 ppbv NO) HO 2 recycling through its reaction with NO could be a limiting factor (Elshorbany et al., 2008). Under these conditions hydrocarbon oxidation could be a net sink for OH rad-20 icals, which in turn will also lead to a reduction in the OH sources, i.e. O 3 and HCHO photolysis as well as alkenes ozonolysis.

OH reactivity
The OH reactivity defined as the reciprocal of the OH radical lifetime has been calculated as L OH (total)/[OH]. The mean day average modelled OH reactivity is about 42 s −1 25 reaching a maximum of 105 s −1 during rush hour (Fig. 4c) and a night-time peak of 60 s −1 . These numbers are slightly higher than the average and night-time peaks measured in Mexico City of 25 and 35 s −1 , respectively, while the maximum measured OH reactivity in Mexico City of 120 s −1 exceeded that of Santiago (Shirley et al., 2006). Sheehy et al. (2008) have also reported a modelled total reactivity of 110 s −1 during the morning rush hour and 45-50 s −1 at night in Mexico City. Both, Ren et al. (2006) 5 and Yoshino et al. (2006) reported OH reactivities in the range of 10-100 s −1 in New York City and Tokyo, respectively. The diurnal variation of the modelled OH reactivity (Fig. 4c) is characterized by morning rush hour and night peaks in agreement with studies of Ren et al. (2006) and Shirley et al. (2006). Underestimation of the OH reactivity using relationship (2) has been previously observed when compared with measured 10 OH reactivity in different field measurements (Di Carlo et al., 2004;Yoshino et al., 2006;Ren et al., 2006 and references therein). It is worth mentioning that the OH uptake on aerosol surfaces and the uncertainty of the rate coefficient of (k (NO2+OH) ) could not account for the missing OH reactivity in previous field measurements (Yoshino et al., 2006;Ren et al., 2006).

Radical propagation
Although hydrocarbon oxidation consumes most of the OH radicals (L OH (OH+VOC)=28.4 ppbv h −1 on average), it also regenerates these radicals through the secondary production of OH, P OH (sec.), (28.9 ppbv h −1 ) given by the sum and HO 2 →OH were not balanced (Emmerson et al., 2007). While all the measured hydrocarbons were quantified, not all could be defined (see Sect. 3.1). In addition, not all defined hydrocarbons could be included in the MCM model because either some of these compounds were measured as a mixture of two compounds (or more) or not defined in the MCM (see Table 2). Thus, to further investigate the recycling process, an additional MCM model scenario has been run, in which the concentrations of all aromatic hydrocarbons and alkanes in addition to iso-5 prene and propene have been increased by a factor of 2 while the rest of alkenes have been left unchanged. The reason for including only isoprene and propene is because of their relatively high reactivity with OH but their low potential for OH production through ozonolysis (see Sect. 3.8.2). Only ∼1% increase in the modelled OH concentration was observed for this additional scenario. In addition, although the fluxes 10 P OH (HO 2 →OH) recycled and L OH (OH+VOC) increased by almost a factor of 2, they were still balanced. These results clearly demonstrate that the main net radical sources and sinks were not affected by the VOC level and that the secondary radical sources (e.g. OVOC photolysis) and sinks (e.g. RONO 2 ) are included in the recycling process, i.e. do not add to the net initiation sources or termination reactions. In the main, this 15 can be explained by the high NO concentrations during daytime in Santiago and the fast recycling through the reactions RO 2 +NO and HO 2 +NO. The high recycling efficiency of the peroxy radicals can be demonstrated by the relatively low HO 2 /OH ratio evaluated by the MCM model (see Fig. 6a). The low maximum in the HO 2 /OH ratio of ∼11 is typical for highly polluted conditions (e.g. Mihelcic et al.,20 2003) and implies a high recycling efficiency towards OH. The RO 2 /HO 2 ratio (Fig. 6b) of 1-1.5 is similar to that reported in Berlin with a maximum modelled ratio of 1.3 (Mihelcic et al., 2003) but much lower than that of 3.9 calculated for the TORCH campaign (Emmerson et al., 2007). While the RO 2 /HO 2 and HO 2 /OH ratios both reach a minimum in the morning at about 9 h, the HO 2 /OH ratio reaches its afternoon maximum at 25 about 14:30 h when the NO levels reach a minimum. The average daytime maximum HO 2 radical concentration of 6.3 pptv (see Fig. 7a) is very similar to that measured in Tokyo, 2004(Kanaya et al., 2007. The average daytime maximum total peroxy radical concentration of 15 pptv is relatively low when compared with other studies (Mihelcic 19139 et al., 2003;Shirley et al., 2006) and can be explained by the high NO concentrations in the city of Santiago. This is also in agreement with the expected anti-correlation between the HO 2 /OH ratio and NO as shown in Fig. 7b in agreement with other studies (e.g. Emmerson et al., 2007 and references therein).

Net radical sources 5
Evaluation of the total rates of radical initiation and termination required a simple steady state approach (see Sect. 2.2.1) that takes into account only the net radical sources and sinks. The net photolysis of HONO, HCHO, ozone and the reactions of ozone with alkenes are considered as initiation reactions while reaction of the OH radical with NO 2 is the main termination reaction. According to this assumption, the radical produc-10 tion rates, P R , of the main corresponding species were evaluated with the same rate constants used in MCMv3.1. The average absolute and relative diurnal contributions to radical production are shown in Fig. 8a and b respectively. For daytime conditions (8 h-19 h) HONO photolysis has by far the highest contribution of ∼55% followed by alkenes ozonolysis (∼24%), HCHO photolysis (∼16%) and ozone photolysis (∼5%).

15
The high relative contribution of HONO is in excellent agreement with other recent studies (Ren et al., 2003(Ren et al., , 2006Kleffmann et al., 2005;Acker et al., 2006a, b), in which an integrated contribution of up to 56% was reported. For average daytime conditions (8 h-19 h), high net mean and maximum OH production rates by HONO photolysis of 1.7 ppbv h −1 and 3.1 ppbv h −1 , respectively, have been determined, the latter be-20 ing even higher than the ∼2 ppbv h −1 reported by Ren et al. (2003) for New York City.
Only in the study of Acker et al. (2006b) was a higher maximum OH production rate by HONO photolysis of up to 6 ppbv h −1 reported for the city of Rome. However, this number is an upper limit since in their estimations the back reaction of NO+OH was not considered. During the morning, for which the maximum production rate was re-25 ported by Acker et al. (2006b), high NO concentrations can especially lead to a strong overestimation of net OH production rates (see Sect. 3.8.3). On a 24-h basis, HONO photolysis was also the dominant radical source contributing ∼52% to P R followed by 19140 alkene ozonolysis, ∼29%, HCHO photolysis ∼15% and ozone photolysis ∼4%. During almost the entire daytime the HONO photolysis contribution was higher than any other primary source except in the early evening when the contribution from alkene ozonolysis starts to dominate. This is caused by the decreasing light intensity with the ozone concentrations remaining high. In the early morning, the photolysis of HONO 5 is the dominant source representing ∼80% of the total radical budget. This is due to its low dissociation energy threshold and the high concentrations accumulated during night-time.
A high morning peak production rate that slows down during the day has been previously observed in Los Angeles, Milan, Pabstthum (downwind of Berlin) and Mexico 10 City (George et al., 1999;Alicke et al., 2002Alicke et al., , 2003Volkamer et al., 2007, respectively). However, in contrast to these studies, where the net OH production was very low in the afternoon, the relative contribution of the OH production by HONO photolysis never falls below 40% for Santiago (see Fig. 8b). This high daytime contribution of HONO is in good agreement with other recent studies under urban conditions (Ren et al., 2003;15 Acker et al., 2006b). The reason for the difference between the two sets of studies in which the contribution of HONO to afternoon radical production is either significant or negligible is still unclear. One potential explanation would be an overestimation of HONO due to interferences and sampling artefacts for all studies, in which wet chemical instruments were used (see Kleffmann and Wiesen, 2008). However, the LOPAP 20 instrument used in the present study corrects for interferences and was successfully validated against the DOAS technique in a recent urban study in Milan (Kleffmann et al., 2006). In addition, a simple PSS analysis of the HONO data from the Milan campaign showed that HONO was also a strong net source of OH radicals during daytime, a result confirmed by the parallel co-located DOAS measurements (Kleffmann et al., Interactive Discussion altitudes and strong vertical gradients during daytime. However, in the study of Alicke et al. (2002) the light path of the DOAS was even lower than the sampling height during the present study and no gradients were observed during daytime . In addition, in the present study no horizontal gradients were observed towards the wall of the building on which the external sampling unit was fixed, excluding strong local wall 5 sources. In conclusion, the reason for the different daytime contribution of HONO to the OH production remains unclear. The high contribution of HONO observed in the present study may be explained by the unique geographical situation of Santiago under very high pollution levels.
The average diurnal variation of the OH concentration calculated by both the MCM 10 and PSS models are shown in Fig. 9. The maximum estimated OH concentrations of 1.4×10 7 molecules cm −3 occurs approximately one hour after the maximum in j (O 1 D). Using different simplified photo-stationary state approaches, Rappenglück et al. (2000) and Rubio et al. (2005) Rubio et al. (2005) did not consider alkenes ozonolysis. The excellent agreement between the OH concentration profiles evaluated by both the MCM and PSS models shows that the major OH radical sources and sinks are included in the PSS model and that the sinks OH→RO 2 are balanced with the sources 20 RO 2 →OH.

Correlation of OH with j (O 1 D) and j (NO 2 )
In spite of the complexity of the mechanisms controlling OH concentrations, the OH correlation with j (O 1 D) has shown to have a linear pattern in both urban and rural environments and for long and short time periods (Rohrer et al., 2006  . In addition, a better correlation between the total rate of radical initiation, P R , and j (NO 2 ) compared to the correlation with j (O 1 D) was observed in the present study, especially for low j -values in the morning and evening (see Fig. 10a and b). This can be explained by the much broader diurnal profile of j (NO 2 ) compared to j (O 1 D). The results demonstrate the importance of the UV-A rather than UV-B region for the 5 production of OH during daytime which is dominated by the daytime production of HONO.
3.8 Source apportionments of the main OH radical precursors

Formaldehyde (HCHO) contribution
HCHO is a main photochemical oxidation precursor contributing ∼16% of the total primary radical sources, P R , during the day time in Santiago. HCHO is both primarily emitted and produced photochemically from the oxidation of VOCs (Friedfeld et al., 2002;Garcia et al., 2006). In this study, we have used O 3 and NO x as HCHO tracers for which NO x has been assumed as an indicator for primary HCHO resulting from direct emissions and O 3 as a photochemical indicator. The measured HCHO was de-15 scribed by: where β o is the background HCHO (BKG), and the factors β 1 and β 2 are the average weighted slopes of HCHO to O 3 and NO x , respectively. For the whole campaign, values of β 1 =0.062 and β 2 =0.018 ppbv/ppbv, respectively were determined. The pho-20 tochemically formed HCHO (PHOT) comprises up to >70% of the observed HCHO in the afternoon (Fig. 11a). In contrast, during the early morning rush hour the primary HCHO (traffic) comprised up to 90% (Fig. 11a). Averaged on a daily basis, ∼34% of the measured HCHO is due to direct emissions while photochemical and background HCHO account for ∼28 and ∼38%, respectively. The value of the direct emitted fraction of the photochemical and background fractions is similar to the secondary fraction reported during the summer in Santiago, 79±23% (Rubio et al., 2006) and London, 74% (Harrison et al., 2006). Since only 28% of the HCHO is photochemically formed as a result of hydrocarbon oxidation, HCHO was considered as a net source of HO 2 (new OH 2 ) in the present study.

5
Photochemical HCHO production has also been simulated using the MCMv3.1 photochemical box model constrained with all measured trace gases including the NMHCs except the measured HCHO. The photochemical HCHO calculated using O 3 as tracer matched well that calculated by the MCM model with a gap in the late afternoon (see Fig. 11b). This gap however, is due to the afternoon ozone shoulder, which has be-10 come a typical feature during photochemical smog episodes in Santiago de Chile (Rappenglück, 2000(Rappenglück, , 2005. Primary HCHO starts to build up in the early morning at about 6:30 h, nearly one hour before sunrise, and becomes the dominant source until ∼9 h. The photochemical formation of HCHO follows the light intensity, and starts to increase nearly an hour after the sunrise, becoming dominant at around ∼13 h and reaching a 15 maximum at ∼16 h nearly 3 h after the maximum in j (NO 2 ). The photochemical HCHO contribution starts to decline at ∼19 h, about 3 h after the j (NO 2 ) starts decreasing, while the primary HCHO turns again to be the dominant source until 2 h due to night time emissions. The average background baseline of HCHO is less than 2 ppbv representing about 20% of the total HCHO throughout the day (Fig. 11a), which is in agree-(see Fig. 11a), which is unreasonable. Finally, the high background HCHO could also be caused by mixing of surface air masses with the residual layer in the morning when the boundary layer height is increasing. The concentration of HCHO in the residual layer could remain high from the previous day. Rappenglück et al. (2005) has also observed a similar background carbonyl peak at noontime in Santiago.

5
The contribution of each of the VOC classes (alkenes, alkanes, aromatics) to the photochemically formed formaldehyde has been determined by the MCM model. As expected, the alkenes are the dominant photochemical precursor contributing alone by more than 70%, followed by aromatics, 18%, and alkanes, 12%. These contributions are in good agreement with those reported in Mexico (Volkamer et al., 2007). Of the alkenes, oxidation of isoprene contributes alone about 23% to the photochemical produced HCHO, propene 11% and α-pinene 9%. From the aromatics class, 1,3,5trimethylbenzene represents 6% followed by ortho-xylol, 4%, and toluene, 3%. Of the alkanes, 2-metylbutane, decane and 3-methylpentane are the major sources contributing to about 3%, 2% and 1.6% respectively. OH is the dominant oxidant responsible 15 for nearly 85% of the total HCHO produced by the oxidation of hydrocarbons followed by alkene ozonolysis, 14%. The contribution of NO 3 was found to be negligible.

Alkenes ozonolysis contributions
Unlike the other OH radical sources, alkene ozonolysis can occur at night as well as during the day (Paulson and Orlando, 1996;Johnson and Marston, 2008). In this 20 study, the ozonolysis of alkenes was found to be the second most important radical initiation source after HONO photolysis, accounting for 29% of the OH formed on 24-h basis. Although their total concentrations are only ∼19% of the total measured alkenes, internal alkenes contribute 86% to the total alkene OH radical production and nearly 21% to the total primary radical production, P R , as shown in Fig. 12a. The order of 25 efficiency in OH production from the reactions of ozone with alkenes is: internal alkenes > cycloalkenes > terminal alkenes.

19145
Among the internal alkenes, 2-methyl-2-butene and 2,3-dimethyl-2-butene have the highest contributions to the alkenes OH radical production with 37% and 33%, respectively (see Fig. 12b). Cycloalkenes are represented by α-pinene alone and contribute about 6.6% to the total alkene concentration, ∼9% to total alkene OH production and ∼2% to P R . The other measured cycloalkenes are not yet included in the MCM. Termi-5 nal alkenes, while representing 75% of the alkenes concentration, contribute only ∼5% to the total alkene OH production rate and about 1% to P R (Fig. 12a).

Contribution of HONO during daytime
As already discussed, over the last few years it has been demonstrated that the contribution of nitrous acid to the primary radical production, P R , has been frequently underestimated (e.g. Ren et al., 2003 ;Kleffmann et al., 2005;Acker et al., 2006a). High measured daytime concentrations point to an additional strong HONO source (Kleffmann, 2007), for which several photochemical reactions have recently been proposed from laboratory studies George et al., 2005;Bejan et al., 2006;Stemmler et al., 2006Stemmler et al., , 2007.

15
The photostationary state concentration of HONO, [HONO] PSS , was calculated from the known gas phase chemistry by the following equation: On average, [HONO] PSS was found to account for about 69% of the observed HONO concentration reaching its maximum contribution during the rush hour peak time at 20 ∼10 h coinciding with the NO peak. During the early afternoon (12:30 h-15 h), when the absolute production rate of OH by HONO photolysis was highest, the PSS contributed on average ∼66% of the measured HONO. Thus, one reason for the extreme high HONO daytime concentrations observed is the daytime production of HONO by the gas phase reaction of NO+OH caused by the very high levels of OH and NO. However, this reaction and the uncertainty in the PSS concentration by only gas phase chemistry (see below) cannot explain the measured daytime values of HONO alone. 19146 If the heterogeneous dark conversion of NO 2 (see Sect.  (Mihelcic et al., 2003;Sheehy et al., 2008). Recently, Sheehy et al. (2008) reported maximum OH over prediction by the MCM of 20% during afternoon.

10
In contrast, for Santiago an under prediction of the modelled OH level by ca. 55% would be necessary to explain the daytime concentrations of HONO. Therefore, additional average daytime HONO sources of 1.7 ppbv h −1 are necessary. These additional daytime HONO sources become obvious from the diurnal variation of the HONO/NO x ratio (Fig. 2b). While the night-time behaviour, with a linear increase of the HONO/NO x ratio 15 from 2-5%, is typical for urban conditions and can be explained by known emission and heterogeneous conversion of NO 2 on ground surfaces Kleffmann et al., 2002Kleffmann et al., , 2003Vogel et al., 2003), the second daytime maximum, reaching almost 8%, has not been observed in our previous urban studies in such a pronounced manner. A daytime maximum under urban conditions was however observed for the daytime source is correlated with the light intensity, confirming former assumptions. However, since a better correlation was obtained when j (NO 2 ) was used, especially for low j -values, the heterogeneous conversion of NO 2 on photosensitized organics (George et al., 2005;Stemmler et al., 2006Stemmler et al., , 2007 and gas phase photolysis of organic 5 nitrogen compounds (e.g. nitrophenols, Bejan et al., 2006) may be of higher importance compared to the nitric acid photolysis in Santiago. Similar results were obtained when plotting P R against j (NO 2 ) and j (O 1 D) (see Fig. 10 and Sect. 3.7).

HONO dark sources
Besides photochemical daytime sources of HONO, formation of HONO during the night 10 by heterogeneous conversion of NO 2 on humid surfaces is well known . The dark heterogeneous rate constant of HONO formation, k het , due to the first order conversion of NO 2 on humid surfaces (NO 2 +X →HONO) has been estimated from the increase of the HONO/NO 2 ratio during the night (see also Alicke et al., 2002). An average k het of (3.5±1.9)×10 −6 s −1 has been obtained, which is similar to that of 15 (3.3±1.4)×10 −6 s −1 obtained by Alicke et al. (2002). This heterogeneous rate constant has been found to correlate inversely with the wind speed (R 2 =0.65) confirming heterogeneous formation on ground surfaces during the night (Kleffmann et al., 2003). However, almost no correlation of k het with relative humidity was observed (R 2 =0.086) in contrast to the study by Stutz et al. (2004). The lack of water dependence can be 20 explained by the heterogeneous conversion of NO 2 into HONO on adsorbed organics (Arens et al., 2002;Gutzwiller et al., 2002;Ammann et al., 2005), which are persistent on any urban surfaces. For this type of reactions only a moderate humidity dependence was observed in the laboratory (Arens et al., 2002) for a humidity range comparable to the present study. In addition, NO 2 conversion on organic surfaces is much faster than the typical proposed reaction of NO 2 with water on surfaces (Finlayson-Pitts et al., 2003) at atmospheric NO 2 levels and thus is a more reasonable source for night-time formation of HONO in the atmosphere.

Conclusions
The oxidising capacity of the atmosphere over the urban area of Santiago, Chile, has been studied for the first time during an extensive measurement campaign in the summer 2005. A zero dimensional photochemical box model containing the detailed gas 5 phase mechanism MCMv3.1 was constrained with a suite of ancillary measurements including HONO, HCHO, O 3 , NO x , PAN, VOCs, j (O 1 D), j (NO 2 ) and meteorological parameters. The average ratio of total production/destruction rates of the hydroperoxy radical (HO 2 ) was around unity throughout the day, whilst the production/destruction ratio for the hydroxyl radical (OH) reaches a maximum of ∼1.7 during the morning.

10
HO 2 radical production was dominated by the RO 2 +NO reaction while HO 2 destruction was dominated by its reaction with NO, which was also the strongest OH source (∼80%). OH loss was dominated with its reaction with hydrocarbons (∼79% source apportionment has revealed that alkenes contribute most by 70% followed by aromatics, 18%, and alkanes, 12%. The major contribution of HONO to the direct OH radical production is in good agreement with several recent studies and highlights the importance of HONO measurements in studies which focus on the radical chemistry of the atmosphere.