Production of peroxy radicals at night via reactions of ozone and the nitrate radical in the marine boundary layer

In this paper, a substantial set of simultaneous measurements of the sum of peroxy radicals, [HO2+RO2], NO3, hydrocarbons (HCs), and ozone, taken at Mace Head on the Atlantic coast of Ireland in spring 1997, is presented. Conditions encountered uring the experiment ranged from semipolluted air masses advected from Britain and continental Europe to clean air masses off the North and mid-Atlantic, where mixing ratios of pollution tracers approached Northern Hemispheric background mixing ratios. Average mixing ratios of peroxy radicals varied from 2.5 to 5.5 parts per trillion by volume (pptv) at night depending on wind sector, and were observed to decay only very slowly from late afternoon toearly morning (0.1-0.5 pptv h-l). Measurements of OH and HO2 on two nights using the Fluorescence Assay by Gas Expansion (FAGE) technique give an upper limit for [OH] of 2.5xl O s molecules cm -3 and an average upper limit [HO2]/[HO2+RO2] ratio of 0.27. A modeling study of the •/e lifetimes of the peroxy radicals, assuming o radical production at night, yielded mean lifetimes of between 8-23 min for HO2 and 3-18 min for CH302. Given these lifetimes, it may be concluded that the peroxy-radical mixing ratios observed could not be maintained without substantial production at night. No significant correlation is observed between measured [HO•+RO•] and [NO3] under any conditions. Calculation of the reaction rates for ozone and NO3 with hydrocarbons (HCs) shows that the ozone-initiated oxidation routes of HCs outweighed those of NO3 in the NE, SE and NW wind sectors. In the SW sector, however, the two mechanisms operated at similar rates on average, and oxidation by NO3 was the dominant route in the westerly sector. The oxidation of alkenes at night by ozone was greater by a factor of 4 than that by NO3 over the whole data set. The HC degradation rates from the three "westerly" sectors, where tracer mixing ratios were relatively low, may be representative ofthe nighttime oxidative capacity of the marine boundary layer throughout the background Northern Hemisphere. Further calculations using literature values for OH yields and inferred RO• yields from the ozone-alkene reactions how that peroxy radicals derived from the ozone reactions were likely to make up the major part of the peroxy-radical signal at night (mean value 66%). However, the NO3 source was of similar magnitude in the middle of the night, when [NO3] was generally at its maximum. The estimated total rates of formation of peroxy radicals are much higher under semipolluted conditions (mean 8.0xl 04 molecules cm -3 s -• in the SE wind sector) than under cleaner conditions (mean 2.4x104 molecules cm '3 s -• in the westerly wind sector). A model study using acampaign-tailored box model (CTBM) for semipolluted conditions hows that the major primary sources of OH, HO2, and CH30• (the most abundant organic peroxy radical) were the Criegee biradical intermediates formed in the reactions of ozone with alkenes.


Introduction
Although the hydroxyl radical (OH) is known to be the primary tropospheric oxidant in the daytime, there is mounting evidence in recent years that significant oxidation of non-methane hydrocarbons (NMHCs) can also occur at night in the lower atmosphere Figure 1.Schematic representation of the mechanism for the reactions of NO3 with alkenes.Atkinson, 1994] to form peroxy radicals, HO2 + RO2 (see Figure 1 [after Wayne et al., 1991]), where R is an alkyl or acyl group [e.g., Platt et al., 1990;Wayne et al., 1991;Jensen et al., 1992].Indeed, Penkerr et al. [1993] proposed that NO3 could have a significant impact on the atmospheric lifetimes of many unsaturated hydrocarbons throughout the Northern Hemispheric troposphere.
Peroxy radicals can also be formed at night via the reactions of ozone with alkenes [Hu and Stedman, 1995;Paulson and Orlando, 1996;Paulson et al., 1999aPaulson et al., , 1999b;;Rickard et al., 1999].Ozone adds to the double bond of the alkene in a concerted process, initially forming a primary ozonide (1,2,3-trioxolane), which rapidly decomposes to yield one or two carbonyl compounds and energy-rich Criegee intermediates.The vibrationally excited Criegee intermediate can then either be collisionally stabilized by a third body (M), or undergo unimolecular decomposition to products.It is now widely believed that alkyl-substituted Criegee intermediates can decompose via a vibrationally hot hydroperoxide intermediate to yield an OH radical, along with another radical species of the general form R1R2C'C(O)R3, which is expected to react rapidly with 02 to form a peroxy radical (RO2) [Martinez et al., 1981;Niki et al., 1987;Rathman et al., 1999].The general mechanism is shown in schematic form in Figure 2a  [ 1997] observed a positive correlation between NO3 and peroxy radicals measured at night at the coastal observatory in Weybourne on the east coast of England.NO3 reactions, particularly that with DMS, were again concluded to be the major routes to peroxy-radical formation in the nighttime.
Conversely, Hu and Stedman [ 1995] suggested that reactions of ozone with alkenes were largely responsible for peroxy radicals observed at night in Denver, Colorado, in the summer of 1993, although no NOx or HC measurements were undertaken in this study.Paulson and Orlando [ 1996]  In this paper, we report a substantial set of simultaneous observations of peroxy radicals and NO3 made at the Mace Head Atmospheric Research Station on the Atlantic coast of Ireland.These data, together with a comprehensive suite of hydrocarbon (HC), ozone, and ancillary measurements, are used to assess the relative importance of NO3 and 03 in the nighttime oxidation of HCs and production of peroxy radicals in the marine boundary layer.Mihele and Hastie [1998] and Mihele et al. [ 1999] has shown that the chain length of a chemical amplifier is reduced in the presence of atmospheric water vapor.This effect has also been observed in the instrument used in this work (G.Salisbury et al., unpublished results, 2000).However, the specific humidity observed during the night at Mace Head during EASE 97 was fairly uniform, with a mean value of 6.7 g kg -• and a standard deviation of only 0.1 g kg '• (<2%).It is likely therefore that any measurement error due to the CLN-humidity effect would be systematically present throughout the data set used in the present work.For a 1 hour average the detection limit of the PERCA instrument varied between 0.6 and 1.0 pptv, depending on the CLN (measured periodically throughout the campaign) and the stability of the background NO2 signal.The instrument accuracy (neglecting the humidity effect) was estimated to be +30%, with a precision of 15% [Clemitshaw et al., 1997].

Recent work by
In addition to the PERCA measurements, OH and HO2 were measured sequentially on two nights using the FAGE technique, with the instrument described and characterized by Creasey et al. [1997].Further details of the deployment of this instrument in EASE 97 are given by N. Carslaw et al. (Modeling the concentrations of OH, HO2, and RO2 during the EASE 97 campaign, 1, Comparison with measurements, submitted to dournal of Geophysical Research -Atmospheres, 2001, hereinafter referred to as Carslaw et al., submitted paper, 2001).During EASE 97 the detection limits for the FAGE instrument were 0.02 pptv for OH and 0.1 pptv for HO2 (150 s averages, signal-to-noise ratio equal to 1).The overall l c• uncertainty for a 150 s average measurement was estimated to be 20% for OH and 25% for HO2 (Carslaw et al., submitted paper, 2001).

Data Analysis
In order to examine the nighttime chemistry occurring in differing air masses observed at Mace Head during the campaign, the data from EASE 97 were divided up according to wind direction or more minutes' worth of data (75 + %) within that sector for a given hour.Hours in which there was no dominant sector by this criterion were rejected.The hourly averaged data points were then used to derive averaged nighttime profiles over the whole campaign for each wind sector.In this work, the nighttime period is defined using the criterion of nonzero [NO3], and hence usually extends from 1900 to 0600.This criterion was chosen in prefer-ence to a strict definition based on solar zenith angles reaching 85 ø or 90 ø because the possible sources of peroxy radicals in the early morning and late evening are also generally poorly understood.
Table 1 shows the average nighttime mixing ratios of selected species during EASE 97 by wind sector derived using this method.
Table 2 shows the average nighttime alkene mixing ratios by wind sector measured during EASE 97.The alkene measurements are discussed in detail by Lewis et al. [ 1999].Nevertheless, it is worth noting here that the alkene mixing ratios were generally lower in the "clean" wind sectors (SW, westerly and NW) than for the relatively polluted NE and SE sectors (Table 2), as would be expected from the oceanic origins of the air masses.Ethene, propene, and 2methylpropene levels, however, remained significant even in clean marine air.Lewis et al. [ 1999] proposed that these species had emission sources from coastal waters local to Mace Head, via photodegradation of organic carbon in seawater.

Modeling
Two modeling studies were performed in support of the experimental data analysis in the course of this work.First, a simple zero-dimensional analysis of nighttime peroxy-radical lifetimes was performed for the average conditions observed in each wind sector, using the reaction scheme shown in Table 3.This method was employed in preference to estimations of peroxy-radical lifetimes by direct calculation, as peroxy radicals can be destroyed by various routes, including reaction with NO, NO3, and 03 (HO2 only), as well as by the peroxy-radical self-and cross-reactions (compare with reactions (R2)-(Rll)).Model runs were constrained by average measurements of NO, NO2, 03, NO3, temperature, and humidity in each sector, but no hydrocarbons were included.Runs were initialized with various proportions of HO2 and CH302, so that their sum equaled the average [HO2+RO2] measured for that wind sector.[NO] and [NO3] were kept constant in each run.For the purposes of this work, RO2 radicals other than CH302 were assumed to behave as CH302.The aim was simply to calculate radical lifetimes for each set of conditions, assuming no radical production at night.The lifetimes so obtained (defined for the purposes of this work as the time taken for the radicals to fall to •/e of their initial mixing ratio) could be compared with the

Figures in parentheses are 1 o values.
Measurements were made using the PERCA technique.

Lifetimes
Figure 4 shows the averaged peroxy-radical data obtained for the SE wind sector in EASE 97.The daytime maximum was not at solar noon, as found under relatively unpolluted conditions [e.g., Monks et al., 2000], but rather midafternoon (Figure 4a).This was probably due to relatively elevated NO levels in the morning (not shown in Figure 4), which acted to suppress peroxy-radical mixing ratios (reactions (R2) and (Rll)), as well as likely enhanced peroxy-radical production from sources other than ozone photolysis in the afternoon, for example, formaldehyde photolysis.Space does not permit full discussion of these observations here, but case studies of a particular polluted day during the Eastern Atlantic Summer Experiment at Mace Head in 1996 (EASE 96) are given .As a result of this "delayed" maximum in photochemical activity, the peroxy-radical mixing ratios were still relatively high in the late afternoon (Figure 4b).After dropping off rapidly between 1600 and 1900 hours, the peroxy-radical mixing ratios stabilized, even increased slightly, between 1900 and 2200 hours, and then decreased only slowly through the night, at a mean rate of-•0.5 pptv h 4 (from Figure 4b).The average modeled peroxy-radical lifetimes at night for the SE sector were 5.4 min for HO2 (range 2.0-11.5 min, depending on HO2/CH302 ratio) and 3.0 (2.9-3.0)min for CH302.In the absence of nighttime production of peroxy radicals, therefore, significant mixing ratios of peroxy radicals could not be maintained throughout the night under these conditions, but would rapidly fall below the detection limit of the PERCA instrument.The combination of the sustained peroxy-radical mixing ratios measured and the calculated lifetimes can be taken as evidence that the peroxy radicals observed during the night were not simply those remaining from daytime photochemistry.In contrast, Figure 5 shows the analogous plots for the westerly wind sector. Figure 5a shows the full diel cycle for peroxy radicals and j(NO2), where the approximately symmetrical rise and fall of peroxy radicals about the solar maximum may be clearly observed.Figure 5b shows the composite nighttime profile obtained for this sector.Note first that the general peroxy-and nitrateradical mixing ratios were substantially lower than for the SE sector (see also Table 1).The mean nighttime ozone mixing ratio was also significantly lower.These results are not surprising given the likely mid-Atlantic origin of the air masses sampled, where NO,, Observed HO 2 loss rate for the night May 1-2.
Observed HO 2 loss rate for the night May 21-22.
and HC loading would be expected to be relatively low (see Tables 1 and 2). Figure 5a shows the same relatively sharp decrease in peroxy radicals up to around 1900 hours as seen for the SE wind sector, then a small upturn in midevening (-2000 hours).
Figure 5b shows an even slower gradual decrease through the night than observed for the SE (0.1 pptv h 'l, see Table 4).4 are the mean peroxy-radical decay rates observed from the five sets of experimental data.It is worth noting that the lowest modeled lifetimes were observed in the SE sector, where [NOx] was highest.The CH302 lifetimes for the three "westerly" sectors (SW, westerly, and NW) were all quite similar, at around 12-18 min.Setting NO to zero in the lifetime models generally left the HO2 lifetimes relatively unchanged, but greatly increased the CH302 lifetimes (Table 4).This was due to the relatively slow self-reaction of CH302, as well as the fact that HO 2 is also destroyed by reaction with ozone (reaction (R7)).

Also included in Table
Nevertheless, the CH302 •/e lifetimes were still all less than 3 hours (Table 4) In summary, slow net decay of peroxy-radical mixing ratios was observed during the night for the whole data set classified by wind sector: the observed decay rates (Table 4) imply •/e lifetimes of 34.7, 6.6, 11.4, 16.1, and 5.9 hours (mean 14.9 hours) for peroxy radicals in the five wind sectors NE, SE, SW, westerly, and NW, respectively.Conversely, the modeled peroxy-radical •/e lifetimes using average measured [NO], which take into account all the known loss reactions for peroxy radicals, were less than 1 hour for all wind sectors.Even with [NO] set to zero, the average HO2 lifetimes were only 8.0-22.6 min, while the average CH302 lifetimes were all <3 hours.In the absence of peroxy-radical production at night, therefore, the peroxy-radical mixing ratios would fall below the detection limit of the PERCA instrument in the course of the night.This shows that the peroxy radicals observed at night during EASE 97 were not simply those remaining from daytime photochemistry.It may be inferred, therefore, that peroxy radicals were produced in significant mixing ratios at night during EASE 97.In order to ascertain the sources of the observed peroxy radicals, total rates (cp) were calculated for the reactions of NO3 and 03 with those alkenes measured during EASE 97, namely, ethene, propene, 2-methylpropene, but-l-ene, cis-and trans-but-2-ene, cis-and trans-pent-2-ene, 1,3-butadiene, and isoprene (equations (1) and ( 2)).

alk cI) NO3 = Z kNo 3 +alk/[NO3 ][alki ]
(1) i The rate coefficients used in the NO3 calculations were taken from Atkinson [1994].Values for the pentenes were unavailable; averaged values were therefore calculated according to the procedure adopted by denkin et al. [1997].The rate coefficients for the 03 reactions were taken from Atkinson [ 1997] where available; the rate coefficients used for the pentenes were averages calculated after denkin et al. [ 1997].Uncertainties on the total rates for each mechanism were estimated using the measurement uncertainties given in section 2.2 and literature l c• uncertainties for the rate coefficients [Atkinson, 1991[Atkinson, , 1994[Atkinson, , 1997]], where available.Uncertainties for the remaining rate coefficients were taken as averages of the published cases.evidences production of peroxy radicals via NO3/O3 reactions as being important during EASE 97.
Table 5 gives the total mean rates (and 1 • uncertainties) for the two sets of reactions, (1) NO3 and (2) 03 with alkenes at night during EASE 97, averaged by wind sector, and the mean percentage contribution of ozone reactions to alkene oxidation.Table 5 shows that the ozone reactions dominated the nighttime oxidation of alkenes over the whole data set (average contribution -80%).
In terms of the overall oxidation of alkenes at Mace Head, the O3+alkene rates are significant, even when compared with daytime OH+alkene oxidation rates: for example, the SE value of 8.3x10 4 molecules cm -3 s -1 may be compared to an average daytime OH+alkene rate of 2.8x105 molecules cm -3 s 'l (for integrated [OH] = 1x106 molecules cm'3).The 24 hour averages (assuming zero [OH] at night; not shown in Table 5) are 1.1x105 and 1.5x10 s molecules cm '3 s -], respectively.Also included in the table are the mean rates for the NO3+CH4 and NO3+DMS reactions for comparison.The final two columns in the table give the percentage contributions of NO3 and 03 reactions to the total nighttime oxidation of HCs (including DMS) at Mace Head.Table 5 shows that the 03 reactions dominated the NO3 reactions in the NE, SE, and NW wind sectors.In the SW sector the contributions of the two mechanisms to HC oxidation were comparable, whereas in the westerly sector the NO3 reactions were dominant.From the data in Table 5 it can also be seen that total oxidation rates were higher for the NE and SE sectors than for the westerly, SW, and NW sectors.This would be expected, given the generally lower NOx and HC mixing ratios in the "westerly" air masses at Mace Head.This trend is in qualitative agreement with the peroxy-radical mixing ratios observed in the five wind sectors (Table 1).
Table 5 also shows that the mean rate through the NO3+DMS reaction was comparable to the mean total rate through the reac- b Upper limits using a maximum rate coefficient of lx10 '18 cm 3 molecule 'l s -I [Atkinson, 1994].
c Includes DMS.
12,681 tions of NO3 with the various alkenes in all wind sectors, and significantly greater than that total for the SW, westerly, and NW sectors, where [DMS] was highest relative to NMHC mixing ratios (see also Tables 1 and 2).The chemistry of NO3 during EASE 97 is dealt with in full by Allan et al. [2000], but these results show once more the importance of even small amounts of DMS in controlling NO3 levels in the marine boundary layer [Yvon et al., 1996;Carslaw et al., 1997;Allan et al., 1999Allan et al., , 2000]].The NO3+DMS reaction alone accounts for-14, 18, 33, 44, and 19% of the total HC oxidation at night in the NE, S E, SW, westerly, and NW wind sectors, respectively.It is also interesting to compare the rate through the NO3+CH4 and NO3+HCHO reactions with the total rate through the NO3+HC reactions.The methane reaction (see Table 5) is seen to have been insignificant in the NE and SE sectors, but represented up to 11% of the NO3+HC oxidation rate in the three "westerly" sectors, where mixing ratios of HCs other than DMS were generally lower.The NO3 reaction with }tCHO (not shown in Table 5) was found to be relatively unimportant in all sectors, with a maximum rate of only 9.1x102 molecules cm '3 s -• (SE sector).
In order to assess the relative peroxy-radical production from NO3 and 03 reactions with alkenes, it is necessary to consider the reaction mechanisms in more detail.It is assumed here that each molecule of alkene (and CH4) reacted with NO3 yields one peroxy radical (Figure 1  As already mentioned, the yield of OH (and hence RO2) froIn the reactions of ozone with alkenes varies according to the structure of the alkene [Paulson and Orlando, 1996;Paulson et al., 1997Paulson et al., , 1999aPaulson et al., , 1999b;;Rickard e! al., 1999].Table 6 shows the OH yields measured by various workers for the alkenes measured in EASE 97; the average values have been used in this study.Also included in Table 6 are the rate coefficients used for the reactions and their source.Since OH produces a second peroxy radical on reaction with the most abundant trace species, CO and CH 4 (as well as NMHCs), the effective peroxy-radical yield is estimated to be twice the OH yield for each alkene; that is, where YoH represents the OH yield from a given O3-alkene reaction.Figure 2b shows the full reaction scheme for the reaction between ozone and propene, with estimated branching ratios as given by Atkinson [ 1997], except for the stabilized Criegee biradical branching ratios [Rickard et al., 1999]; in addition, the synand anti-CH3CHO0* are assumed to be formed in equal amounts.
Figure 2b shows that the calculation used in this work is a simplification, since OH and peroxy radicals may be produced via pathways not considered in equation ( 4).The total HO2+RO2 yield estimated from Figure 2b, assuming all OH formed goes on to react with CO or CH 4 to produce a peroxy radical, is -96%.Nevertheless, 69% of the total yield is estimated to be due to the decay of the syn-CH3CHO0* Criegee biradical (species A in Figure 2b) via a vibrationally hot hydroperoxide (pathway 1 in Figure 2b).
Hence equation ( 4) may be taken to represent a reasonable lower limit fbr the peroxy-radical yield for the ozone-propene reaction, and, by analogy, the other ozone-alkene reactions considered here.

Other possible routes to peroxy radicals have not been considered
fhrther in the calculations presented in this section.
The assumption that all OH radicals formed in the reactions of ozone with alkenes react further to produce peroxy radicals was examined by calculating reactivity indices for the major loss reac- NMHCs were not included in this simple analysis.Equation ( 5) was then used to estimate a lower limit for the fraction of OH  Figure 9 shows the estimated peroxy-radical production rates (equations ( 3) and ( 4)) from alkene reactions for the SE sector (Figure 9a) and the westerly sector (Figure 9b).The error bars in rate (not included in Figure 9) was found to be insignificant in this sector (Table 5).The situation in the westerly sector was more complex.The high production from ozone reactions at the beginning of the composite night is because of high ozone mixing ratios on one particular night (day of year 137): the low level of chemical activity during the latter part of the night would appear to be more characteristic of clean air masses off the mid-Atlantic.Interestingly, for this sector a much higher relative contribution of NO3 chemistry is observed, particularly when the NO3 + CH4 reaction is also considered.This is reflected in Table 5, where the westerly sector shows the highest contribution of NO3 to HC oxidation over the whole night.
Figure 10 summarizes the estimated mean peroxy-radical production rates for the five wind sectors.The error bars in Figure 10 represent mean 1• uncertainties in the production rates for the whole data set (19% for the ozone reactions; 25% for the nitrateradical reactions).The total O3+alkene reaction rates are included for comparison with those using the projected RO2 yields for each reaction.The NO3 data include the reaction with methane, which was found to be important in the westerly, SW, and NW sectors.Nevertheless, over the whole nighttime period it is clear that the greater contribution to production of the peroxy radicals observed at night during EASE 97 was from the reactions of ozone with alkenes (---66% on average over the whole data set).Certainly, the contribution of the nitrate radical to the initiation of nighttime oxidation of alkenes would appear to be significantly smaller than that of ozone (on average, -20% versus 80%, respectively, see Table 5).The low HC reaction rates in the three "westerly" sectors, where pollution tracers were at relatively low mixing ratios, may be representative of the oxidative capacity of the nighttime marine boundary layer throughout the background Northern Hemisphere.Conversely, the rates calculated for the semipolluted sectors may be underestimates, since higher chainlength alkenes were not measured in EASE 97, but are believed to be present in significant mixing ratios in the semipolluted continental boundary layer [Lewis et al., 2000].

Campaign-Tailored Box Model Results
Qualitative supporting evidence for the conclusion that ozone chemistry was the dominant source for the nighttime peroxy radicals observed during EASE 97 was obtained from a model study The modeling results presented are for the period May 1-3, 1997, early in the campaign, when Mace Head experienced a semipolluted air mass originating on the continent [Jacobs, 2000].Even under these conditions, when NMHC mixing ratios were relatively high, the HO2 and CH302 radicals were found to be the dominant peroxy radicals observed at night (>65%), with CH302 invariably the most abundant species [Jacobs, 2000].The average measured NO level during the hours of complete darkness (-2030 -0445) during this period was -• 9 pptv.In order to assess the impact of different possible NO mixing ratios on the nighttime chemistry, a sensitivity analysis was carried out, running the model with 0, 1, 5, and 10 pptv of NO present.Figure 13 shows the percentage contributions to OH, CH302, and HO2 formation for [NO] = 10 pptv. Figure 13a shows that the main routes to OH formation were secondary processes, involving reactions of HO2, with the major source being the reaction of HO2 with 03.The only important primary source from HC reaction was via the decomposition of the Criegee biradical, CH3CHOO* (species A in Figure 2b), formed from the reactions of ozone with propene, cis-but-2ene, trans-but-2-ene, cis-pent-2-ene, and trans-pent-2-ene.Figure 13b shows that the highest contributions to CH302 formation were the decomposition of the Criegee biradical, CH3CHOO*, and the In terms of primary production from HC oxidation, it is clear that the ozone reactions predominated as sources of OH and the major peroxy radicals HO2 and CH302 during this period.The main qualitative difference observed in raising the NO level from 0 to 10 pptv was found to be the increased recycling of HO2 to OH.This in turn increased the relative contribution of the OH+CH4 reaction to CH302 formation.The relative contributions to HO2 formation were largely unchanged.In quantitative terms, the overall formation rates of the three radicals increased markedly for increased NO, but this was largely due to increased cycling between the three radical species; the primary formation routes were unaffected by the NO level.This was to be expected, since NO does not participate in the radical formation reactions.In order to assess the relative contributions of ozone and NO3 reactions to the peroxy-radical signal observed at night during EASE 97, reaction rate calculations were performed in two stages.
First, the total reaction rates were calculated for each mechanism.This showed that the ozone-initiated oxidation routes of HCs outweighed those of NO3 in the NE, SE, and NW sectors.However, in the SW sector the two mechanisms operated at similar rates on average, and oxidation by NO3 was the dominant route in the westerly sector.The oxidation of alkenes at night by ozone was greater by a factor of 4 than that by NO3 over the whole data set.
Next, peroxy-radical formation rates were estimated from both mechanisms (Figures 1 and 2).In the NO3 case it was assumed that one RO2 was formed from each NO3 + alkene reaction [Wayne et al., 1991].The NO 3 + CH 4 and NO3 + HCHO reactions were also considered, but not that with DMS, since the main product of this reaction is CH3SCH202, which is not believed to be detectable by the PERCA technique.In the 03 case, literature values of OH (and inferred RO2) yields were used to estimate total RO2 production.The calculations showed that ozone reactions produced more peroxy radicals over the whole night period (defined as where [NO3] :/: 0) than nitrate reactions (66 versus 34% on average).However, the two mechanisms were found to operate at a similar rate in the middle of the night, when [NO3] was highest.The main importance of these results is perhaps that there was a significant contribution from both mechanisms (Figure 12), since  Conversely, the rates calculated for the semipolluted sectors may be underestimates, since higher chain-length alkenes were not measured in EASE 97, but are believed to be present in significant mixing ratios in the semipolluted continental boundary layer [Lewis et al., 2000].
relatively few field studies of peroxy radicals at night.In the very clean, low-NOx atmosphere over the Southern Ocean, Monks et al. [ 1996] observed only very low nighttime mixing ratios (-1 pptv) of peroxy radicals.The data could be explained by the persistence into nighttime of CH302 radicals produced during the day, with a calculated lifetime of over 12 hours.Mihelcic et al. [1993] reported the first simultaneous measurements of peroxy radicals and NO3 from Schauinsland, Germany, using the matrix-isolation electron-spin resonance (MI-ESR) technique.NO3 and HO2 were found to be present in the range of 0-10 pptv, while the sum of organic peroxy radicals (RO2) reached mixing ratios of up to 40 pptv; the data suggested an anticorrelation between NO3 and RO2 radical mixing ratios on one night.The presence of unmeasured monoterpenes was proposed to explain the high RO2 and low NO3 mixing ratios, although the low NO3 lifetimes observed could have been due to heterogeneous losses or reaction with NO.Low nighttime mixing ratios of peroxy radicals were observed by Canttell et al. [ 1992] during the Rural Oxidants in the Southern Environment (ROSE) experiment, although no conclusion was reached regarding the source of the observed signal.In contrast, Cantrell et al. concluded that the reaction of NO3 with DMS could be largely responsible for the observed peroxy-radical signals at two sites where marine boundary layer air was frequently sampled, in Brittany [Canttell et al., 1996] and Mauna Loa [Canttell et al., 1997].Carslaw et al.

Figure 2 .
Figure 2. Schematic representation of the mechanisms for the reactions of O3 with (a) a generalized alkene and (b) propene.
Ireland, 88 km west of Galway.The station consists of two sets of buildings: one at 100 m from the shoreline with a 23 m tower for the air inlet of the ozone and carbon monoxide (CO) measurements; and another at 300 m from the shoreline at an altitude of 30 m above sea level [Carslaw et al., 1999a].During the campaign there was also a 10 m tower at the upper site, where the inlet units for the NOx, NOy and Peroxy-Radical Chemical Amplification (PERCA) measurements (see section 2.2) were situated, as well as a meteorological station at a height of 15 m above the ground.A 10 m tower near the shore buildings provided an inlet for hydrocarbon sampling, adjacent to the Fluorescence Assay by Gas Expansion (FAGE) instrument, which sampled hydroxyl (OH) and hydroperoxy (HO2) radicals at 4.5 m above the ground (see section 2.2).The Differential Optical Absorption Spectroscopy (DOAS) instrument for NO3 measurements was situated at the near-shore site, with the retro-reflector array placed on a small island 4.2 km to the west of the site [Allan et al.measured by the chemical amplification (PERCA) technique, pioneered by Cantrell et al. [1984], using the instrument described in detail by Clemitshaw et al. [1997] and Monks et al. [ 1998].Briefly, the method relies upon the HO2 and OH radical catalyzed conversion of NO and CO into NO2 and CO2chain length, is equal to the number of HO2/OH radical interconversion cycles that occur before radical termination.The NO2 yield of the PERCA instrument used during EASE 97 is measured using a commercial Scintrex LMA-3 detector.Since the ratio [HO2]/[OH] ranges from -50 to 200 in the atmosphere, and many organic peroxy radicals are readily converted into HO 2 in the presence of NO, for example, CH302 via reactions (R2) and (R5), the PERCA technique effectively measures the sum of peroxy radicals.
and Grenfell et al. [ 1999], with 1 c• uncertainties of +5-10% (1 hour average), depending on mixing ratio.Ozone measurements were made using a commercial UV ()• = 254 nm) spectrometer, as described by Derwent et al. [1994], with an estimated accuracy of +5%.NOx was measured using commercial CRANOX (Ecophysics) instrumentation, as described by Cc•rdenas et al. [1998]; in addition, more sensitive NO measurements were made using an instrument designed and built at the National Oceanic and Atmospheric Administration (NOAA) Aeronomy Laboratory and described by Carpenter et al. [2000].The latter instrument was also used to measure NO2 and NOy.The accuracy of the latter set of NOx measurements has been estimated by Carpenter et al. [2000] to be +13% for NO and +33% for NO2 (for ambient levels of 50 pptv and 150 pptv, respectively).Carbon monoxide, methane, and other tracers are measured routinely at the station, as described by Cvita$ and Kley [ 1994].

Figure 3 .
Figure 3. Map of Ireland, showing the location of the Mace Head Atmospheric Research Station, superimposed with the wind sectors used in the data analysis.
Figures in parentheses are 1 (• values.
, this represents indirect evidence for peroxyradical production at night, even under the relatively clean conditions of the westerly wind sector, albeit at much lower mixing ratios than under more polluted conditions.Similar results were obtained for the other three wind sectors.Table 4 gives the modeled peroxy-radical lifetimes obtained for each sector, using both the mean measured NO mixing ratios and with [NO] set to zero.This dual procedure was adopted because there is ongoing controversy on the feasibility of nonzero [NO] at night in the presence of ozone: since k(r=285 K)(NO+O3) = 1.5 x 10' •4 cm 3 molecule-• s-• [DeMote et al., 1997], the •/e lifetime of NO for [03] = 35 ppbv is calculated to be-77 s.The NO detection limit of the UEA NOxy instrument is estimated to be- . If [NO3] was also set to zero, the mean lifetimes for the NE sector became 8.5 (range 7.2-11.9)min for HO2 and 4+ (3.8-12+) hours for CH302.These results show the importance of NO3 to the processing of organic peroxy radicals at night, when NO is absent or at very low mixing ratios (conversion rates to be equal via the two routes.

Figure 7
Figure 7 shows a scatterplot of measured [HO2+RO2] versus [NO3], where each data-point is a composite wind sector hour,

Figure 8 Figure 8 .
Figure 8 shows a correlation plot of the total rate through the NO3 and 03 reactions with alkenes versus the measured sum of peroxy radicals, [HO2 + RO2], where the data for all wind sectors are plotted as a single series.The level of correlation (r 2 = 0.54) [after Wayne et al., 1991]); that is,

Figure 9 .
Figure 9.Estimated peroxy-radical production rates from the NO3+alkene and O3+alkene reactions for (a) the SE wind sector and (b) the westerly wind sector.The total rates are given for the NO3 reactions, since one RO2 is assumed to be formed from each reaction, whereas literature values of OH (and hence RO2) yields are factored into the 03 reaction rates.The error bars are estimated 1, uncertainties.

Figure 9 Figure 10 .
Figure 9 represent 1 • confidence limits based on the uncertainties in the total reaction rates, as well as the estimated error limits in the OH yields (Table6).Since there is a range of OH yields reported in the literature, the uncertainty in OH yield for each ozone-alkene reaction was estimated to be the standard deviation of the measured values given in Table6.For the pentenes, where only one value of the OH yield has been reported, the mean uncertainty for the other alkenes was used (18%).The uncertainty limits do not encompass possible errors due to the neglect of other OH, HO2, and RO2 formation routes.In the SE sector the 03 reactions appeared to dominate peroxy-radical production through most of the night, although the NO3 rate was of a similar magnitude around midnight (hour 6 of the night).The NO3 + CH4 reaction

Figure 10
Figure 10 highlights the difference between considering the overall HC oxidation at night, where O3 generally predominated, as opposed to the relative contributions to peroxy radicals observable by the PERCA technique, where the NO3 contribution was higher in comparison, although usually still outweighed by the O3 contribution.NO3 reactions made the highest relative contribution to the PERCA signal in the westerly sector (see Figure 9b).Figure 11 shows the estimated percentage contributions to the PERCA signal throughout the night for all wind sectors.Figure 1 la shows the values for each wind sector, while Figure 1 lb shows the overall average contributions by hour.Figure 11 shows that the O3 contri- Figure 11 shows the estimated percentage contributions to the PERCA signal throughout the night for all wind sectors.Figure 1 la shows the values for each wind sector, while Figure 1 lb shows the overall average contributions by hour.Figure 11 shows that the O3 contribution was always at least equal to that of NO3, and generally greater, apart from in the middle of the night, when NO3 mixing ratios were highest.It is worth noting that the two sets of reactions contributed almost equally to the PERCA signal in the central portion of the night, since previous workers have tended to conclude that either NO3 or O3 reactions are dominant (see Figure 12).

Figure 11 .
Figure 11.The estimated percentage contributions of NO3 reactions and O3 reactions to peroxy-radical formation at night during EASE 97.Data from all wind sectors are included: (a) all composite hours are shown as separate points for each wind sector; (b) averages of all wind sectors are plotted for each hour.

Figure 12 .
Figure 12.Schematic representation of nighttime radical initiation and propagation chemistry.The scheme excludes all radical-radical self-and cross-reaction termination processes.
east coast of England.The sustained mixing ratios of peroxy radicals observed at night (net decay rates of <0.5 pptv h '• in all wind sectors) demonstrated a requirement for substantial production processes in the absence of daytime photochemistry.In support of this finding, the average modeled lifetimes of HO2 and RO2 (where CH302 was taken to represent all RO2 present) were all less than 1 and 3 hours, respectively, even with [NO] set to zero.When NO3 was removed from the models, the lifetime of CH302 became considerably longer (>4 hours).This points to the importance of the nitrate radical in the processing of RO2 at night in the absence of NO (compare with reactions (R2) and (R3)).This route to HO2, and hence OH at night, is likely to be of importance in the nighttime troposphere, whatever the original source of RO2 (Figure12).Measurements of OH and HO2 on two nights using the FAGE technique gave an upper limit for [OH] of 2.5x105 molecules cm -3 and an average [HO2]/[HO2+RO2] ratio of 0.27 (upper limit).
previous studies have tended to emphasize one as dominant over the other.The HC oxidation rates from the three "westerly" sectors, where tracer mixing ratios were relatively low, may be representative of the nighttime oxidative capacity of the marine boundary layer throughout the background Northern Hemisphere.

Table 1 . Nighttime Averages (1900-0600 GMT) of Selected Trace Gases, Temperature, and Dew Point During EASE 97 a Quantity Wind Sector Means b
All values are in pptv except where indicated.

Table 5 . Average Nighttime Reaction Rates by Wind Sector During EASE 97 a
a All rates are in molecules cm '3 s 't.