Measurements of peroxycarboxylic nitric anhydrides (PANs) were made during the New England Air Quality Study 2002 cruise of the NOAA RV Ronald H Brown. The four compounds observed, PAN, peroxypropionic nitric anhydride (PPN), peroxymethacrylic nitric anhydride (MPAN), and peroxyisobutyric nitric anhydride (PiBN) were compared with results from other continental and Gulf of Maine sites. Systematic changes in PPN/PAN ratio, due to differential thermal decomposition rates, were related quantitatively to air mass aging. At least one early morning period was observed when O3 seemed to have been lost probably due to NO3 and N2O5 chemistry. The highest O3 episode was observed in the combined plume of isoprene sources and anthropogenic volatile organic compounds (VOCs) and NOx sources from the greater Boston area. A simple linear combination model showed that the organic precursors leading to elevated O3 were roughly half from the biogenic and half from anthropogenic VOC regimes. An explicit chemical box model confirmed that the chemistry in the Boston plume is well represented by the simple linear combination model. This degree of biogenic hydrocarbon involvement in the production of photochemical ozone has significant implications for air quality control strategies in this region.
 Peroxyacetic nitric anhydride (PAN), peroxypropionic nitric anhydride (PPN), and peroxymethacrylic nitric anhydride (MPAN), collectively known as peroxycarboxylic nitric anhydrides (PANs), have been recognized as important secondary tropospheric pollutants that are produced through the photochemical oxidation of volatile organic compounds (VOCs) in the presence of NOx [Stephens, 1969; Singh and Hanst, 1981]. Aside from having a detrimental effect on plants and human health, PANs often represent a large portion of the total oxidized nitrogen (NOy) in the troposphere [Singh et al., 1985; Singh et al., 1994; Roberts et al., 2004]. The relative abundances of the various PAN compounds can be useful indicators of the impact of individual VOCs or VOC classes on a given air mass [Williams et al., 1997; Roberts et al., 1998a, 1998b; Nouaime et al., 1998; Roberts et al., 2003]. PANs also serve as NOx reservoir species that can release NOx and an organic radical, allowing for ozone production downwind of emission sources [Moxim et al., 1996].
 While all the PANs are formed through similar chemistry they differ in relative amounts and in the nature of the source of their parent VOC. PAN is the family member usually present in the highest abundance. This is attributed to the wider range of hydrocarbon sources that can form PAN, both biogenic (acetaldehyde, ethanol, isoprene, certain monoterpenes) and anthropogenic (ethane, acetaldehyde, acetone, toluene, higher alkanes). PPN and MPAN are often observed at mixing ratios that are roughly an order of magnitude lower than PAN. PPN formation is almost entirely through oxidation of anthropogenic hydrocarbons such as propane, propanal, and 1-butene. The formation of MPAN occurs solely through the reaction of methacrolein, which comes almost exclusively from oxidation of isoprene.
 During the summers of 1999 and 2001 the northeastern US experienced numerous high pollution events marked by elevated levels of ozone [White et al., 2007]. The states in this region experience significant biogenic emissions [Geron et al., 1994]. Thus one of the main objectives of the NEAQS 2002 campaign was to determine what role biogenic hydrocarbons (BHC) and anthropogenic hydrocarbons (AHC) play in the photochemistry leading to these pollution events. Because of the differences in the nature of their source hydrocarbons, the measurement of PANs and ozone serve as useful indicators of the role of these two different photochemical regimes. These compounds were measured aboard the NOAA research vessel Ronald H. Brown during the summer of 2002, from the period 12 July to 10 August. The results of these measurements will be analyzed in this paper using a simple linear combination model (LCM) that has been used previously to estimate the contributions of AHC and BHC regimes to O3 production [Williams et al., 1997; Roberts et al., 1998a, 1998b]. This approach has not been tested with a full photochemical model in part due to the limitations inherent in current photochemical mechanisms used in regional photochemical transport models. The Master Chemical Mechanism MCM v3.1, developed and maintained by the University of Leeds [Jenkin et al., 2003; Saunders et al., 2003], contains explicit chemistry for all of the PAN species of interest here, and has been run using conditions appropriate to the Boston urban plume and compared with the LCM results from this NEAQS 2002 data.
2. Experimental Methods
 The measurements used in the analysis have been discussed in detail elsewhere and will be only briefly summarized here. Methods for the measurement of PANs in this study were described at length by Williams et al.  and Roberts et al. , and the calibration methods have also been discussed by Flocke et al. . Changes or additions to those methods are described below. The basis of the measurement was capillary gas chromatographic separation, followed by electron capture detection. The instrument used a continuously flushed sample loop which was injected every 10 min, on the even 10 min mark, i.e., XX:00, XX:10, XX:20 etc. The inlet consisted of an all-PFA Teflon manifold at high flow (10 SLPM). The inlet was filtered by a 1 μm pore size Teflon filter, which was changed every few days. We have tested filters of this type in the laboratory and found them to pass PAN with 100% efficiency. Moreover, no abrupt change in PANs, either up or down, was associated with the filter changes so we conclude that the filter did not interfere with the measurement. A smaller flow (1 SLPM) was taken from this manifold into the PAN instrument. The PAN thermal decomposition and calibration plumbing were fitted to the small flow inlet as described by Roberts et al. . The inlet equilibration time was estimated to be approximately 1 minute. The instrument response to PAN was calibrated routinely, every 5 h, using a modified acetone/CO/NO photolysis source which is based on the calibrated NO mixing ratio and known conversion efficiency (93 ± 3%) [Flocke et al., 2005]. The photosource modifications involved the replacement of the acetone addition system with a 20 ppmv acetone/air gas mixture and flow control apparatus as described by Flocke et al. . Prior to the cruise, the photolysis source was compared to a diffusion source consisting of a PAN/tridecane/pentadecane mixture in a pressure, temperature and flow controlled capillary cell. The output of the capillary cell was calibrated using an NOy instrument. The responses of the GC system to the two PAN sources were within 5% of each other. The system response to PPN was determined using the diffusion source and NOy instrument prior to the field experiment, and the relative response used during the cruise. Two of the other PAN compounds, PiBN and MPAN, were calibrated by using relative response factors that had been determined in the laboratory, and have been consistent within 5% over several different field campaigns. Detection limits for all of the compounds were 4pptv and the overall uncertainties were ±(4 pptv + 15%) for PAN and PPN, ±(4 pptv + 20%) for PiBN and MPAN.
 Measurements of the volatile organic compounds (VOC) were made either by gas chromatography/mass spectrometry (GC/MS) or by proton transfer reaction mass spectrometry (PTRMS). Those methods and aspects of the results of those measurements during this project have been discussed by de Gouw et al. , Goldan et al. , and Warneke et al. . Measurements of NO, NO2, NOy, and O3 were made using methods and procedures described previously by Williams et al. .
3. Chemical Model Description
 A model was constructed using a subset of the Master Chemical Mechanism (MCM), a fully explicit chemical mechanism based on the protocol by Jenkin et al.  and Saunders et al.  and freely available as a community resource at (http://mcm.leeds.ac.uk/). The model contained the complete oxidation mechanisms of 50 hydrocarbons, 14 oxygenated hydrocarbons, DMS, CH4 and CO plus a complete inorganic chemistry mechanism taken from the IUPAC evaluation, 2005. The mechanism was modified for this work to include the latest kinetic data on PANs chemistry. First, the reaction rate for MPAN + OH was changed from the MCM default value of 3.6 × 10−12 cm3 molecule−1 s−1 to the latest value of 3.2 × 10−11 cm3 molecule−1 s−1 [Orlando et al., 2002]. The thermal decomposition rate of all PANs in the MCM is the same as PAN, according to the protocol [Saunders et al., 2003]. For this work, the PPN decomposition rate was decreased by 15%, according to the results of Kirchner et al. . MPAN is also known to react with NO3 with a rate constant of 1.45 × 10−16 cm3 molecule−1 s−1 [Canosa-Mas et al., 1999], however for a NO3 mixing ratios of 60 pptv, which was representative of some of the higher episodes observed in NEAQS 2002 [Aldener et al., 2006], the MPAN lifetime is slightly longer than 5 days, so this process was considered insignificant.
 The model was initialized using emission ratios calculated from the measurements taken during the NEAQS 2002 and NEAQS 2004 campaigns. The initial concentrations of hydrocarbons (AHC and BHC) were calculated relative to CO according to Warneke et al. . The initial concentrations of oxygenated hydrocarbons were calculated relative to acetylene as described in de Gouw et al. , with the exception of HCHO which was estimated at 1.1 ppbv. Initial [CO] was 198.6 ppbv and initial NOx mixing ratios were 20 ppbv on the basis of the measurements taken during the NEAQS 2004 campaign. These conditions were assumed to be representative of Boston and other metropolitan areas of New England and therefore this model will be referred to in the rest of the paper as the ‘Boston’ model. The concentrations of the other species in the model (radicals and other short-lived or not-measured intermediates) were assumed to be initially zero. The model was then run for 5 days to calculate the concentrations of all the 3729 species in the chemical mechanism. The oxygenates and alkyl nitrates calculated for these conditions are the subjects of another manuscript in preparation [R. Sommariva personal communication].
 Five additional models were built for the expressed purpose of testing the PAN-PPN-MPAN linear combination model. In these five models (named ‘aa’, ‘aab’, ‘ab’, ‘abb’, ‘bb’) the initial concentrations of AHC and BHC (basically only isoprene) were varied according to the values listed in Table 1. Note that [α-pinene] = 18.8 pptv and [β-pinene] = 11.5 pptv in the ‘Boston’ model, but were set to 0 in the other five models to exclude the impact of monoterpenes and focus on isoprene chemistry.
Table 1. Initial Conditions for the MCM v3.1 Modeling
 Measurements of PAN compounds were made for the entire measurement period 12 July to 10 August 2002. A map of the ship track is shown in Figure 1 for the entire cruise, which started and ended in Charleston, SC. The first part of the cruise (12 July 22:00 to 15 July 04:00 GMT) was qualitatively and quantitatively different from the period when the ship was off the coast of New England and in the Gulf of Maine. This initial period was characterized by steady winds from the 0 to 180° sector, hence is representative of the clean Atlantic marine Boundary Layer (MBL), and will be analyzed separately. The bulk of the data was collected above 40°N latitude and was most representative of the New England coastal and Gulf of Maine environment that is impacted by emissions and photochemistry of urban North America, so will form the basis for most of the discussion in this paper. The remainder of the cruise from 40°N back down to SC was characterized by a mixture of North American and N. Atlantic air masses and will not be analyzed in this paper.
 The comparison of the initial clean N. Atlantic measurements with others is summarized in Figure 2. The values in this study are in general agreement with those of Gallagher et al.  and slightly lower than those of Muller and Rudolph  whose data had some direct continental contributions. The remote MBL is recognized to be a sink region for PANs [Roberts, 2007], due to relatively high temperatures, lack of NOx sources and slow VOC oxidation chemistry, hence fast destruction and slow production. These features are present in our measurements in this region and are further amplified by the fact that the other PAN compounds were not observed above their detection limits.
 The measurements of PANs above 40°N are summarized in Table 2 along with other data sets acquired in the New England/Gulf of Maine environment, and data sets acquired in the summertime continental U.S. The NEAQS 2002 data are similar in magnitude to the data collected in the Continental U.S. at sites generally recognized to be impacted by VOC and NOx sources. This impact carries into the near coastal and Gulf of Maine environments. The major difference between the NEAQS environment and those at the continental sites is that the air masses observed in the MBL had more defined, and in some respects longer, times since impact by fresh emissions. In contrast the continental sites were surrounded by sources and therefore almost always impacted by fresh emissions, Moreover, those times varied from a fraction of a day to many days. This permitted certain aspects of the PANs production and loss chemistry to be discerned in ways not possible at continental sites.
Table 2. Comparison of PANs Measured in NEAQS 2002 With Measurements in the New England, the Gulf of Maine Environment and Selected Na
 The relationship between PPN and PAN is shown in Figure 3a. The relative concentrations of PPN and PAN reflect the formation chemistry of the VOC-NOx mixtures characteristic of the NE environment, coupled with loss processes. The relative abundance of PPN varied from slightly less than 10% to over 30%. The 15% line is shown for comparison purposes because that is the ratio that has been observed in areas dominated by anthropogenic VOC-NOx photochemistry [Roberts et al., 1998a, 1998b, 2002, 2003]. Lower ratios are associated with significant contributions of isoprene photochemistry, which produces PAN but not PPN [Williams et al., 1997; Roberts et al., 2002]. Ratios of PPN to PAN significantly higher than the 15% line have been observed in the Houston area when the VOC mixture was perturbed by local petrochemical sources [Roberts et al., 2003], and in the MBL off the coast of California [Roberts et al., 2004] where thermal decomposition was starting to affect the ratio. The most intense case of high PPN/PAN during NEAQS 2002 was observed from 4 August, 1200 to 5 August, 0200 GMT, and will also be discussed in detail below.
 The relationship between MPAN and PAN is shown in Figure 3b. The presence of MPAN indicates active isoprene-NOx photochemical processing, since MPAN is a unique product of that chemistry. Moreover, since MPAN and its parent compounds, isoprene and methacrolein react relatively rapidly with OH, the presence of MPAN means that the impact of isoprene on the particular air mass was relatively recent (the past day or so). Transpositively, it must be noted that the absence of MPAN does not necessarily mean that there was no isoprene impact on a given air mass, as the marker of that impact may have been reacted away. The abundance of MPAN relative to PAN in the NEAQS 2002 data set varied from essentially 0, up to 25%. The highest MPAN/PAN is consistent with high isoprene impact time periods observed in previous studies, specifically the SOS Nashville 1999 study, in which MPAN/PAN as high as 25% was observed [Roberts et al., 2002]. The highest MPAN was observed on 23 July between 1400 and 2000 GMT and corresponded with the highest PAN and O3 observed during the study. This event will be discussed in detail below.
 The relationship between PiBN and PAN is shown in Figure 4. The bulk of the data is below the 2.5% line, which is roughly the mean ratio observed in the SOS 99 and TexAQS 2000 data sets [Roberts et al., 2002, 2003]. The median mixing ratio of PiBN was below the detection limit (4 pptv). This is because the precursors, isobutane and longer chain 2-methyl alkanes, while abundant in urban emissions, have relatively inefficient oxidation pathways to PiBN. Moreover, there are no recognized biogenic sources of compounds that yield PiBN. The previously identified time periods of interest are also plotted in Figure 4, with the 4 August–5 August period (open triangles) standing out because of high PiBN relative to PAN and the 23 July period (open circles) standing out due to the enhanced PAN source from isoprene chemistry.
 The plot of O3 versus PAN from NEAQS 2002 in shown in Figure 5. Also shown are the points from the previously identified time periods of interest, and points from the TRACE-P study that were identified as coming directly from the Asian continent [Russo et al., 2003]. The highest TRACE-P point was the maxima from the sector of highest anthropogenic impact, coastal Asia, while the others were medians of all three sectors. Several aspects of this plot are of interest. The O3-PAN relationship is non-linear, as observed in numerous data sets [Roberts et al., 1998a, 1998b], because of the non-linear dependence of O3 production on VOC-NOx photochemistry. The highest O3 observed in NEAQS was during the highest isoprene impact event. The time period of 4 August–5 August had lower PAN relative to the main cluster of the data. In addition the points from the 18 July period (open diamonds) were lower in O3 than the bulk of the data. The data from the TRACE-P study were roughly consistent with the NEAQS data, given the somewhat broad spread of points. The time periods of 23 July and 18 July will be discussed in detail below, in regards to O3 production and destruction, respectively.
5.1. The 4 August 2002 Event: Aged Continental Pollution
 The measurements made during the 4 August 1200 to 5 August 0200 time period are shown in timeline form in Figure 6. The top panel shows that the [PPN]/[PAN] ratio was generally high, above 0.20 for the time between 1500 and 2300, with a peak value of 0.51. The wind direction during this time was from the Southeast. The values of PAN (1000 pptv and above), NOx (1–10 ppbv), and O3 (over 100 ppbv) (bottom panel) are indicative of air that had been impacted heavily by VOC-NOx photochemistry. These chemical features, in concert with the local wind vector imply that this was an air mass heavily impacted by urban Northeastern sources, but more photochemically aged than air masses typically observed on the continent.
 There are several ways to use chemical measurements to indicate the “age” of an air mass [Roberts et al., 1984; Bertman et al., 1995; Stroud et al., 2001; de Gouw et al., 2005]. A photochemical age can be determined for a set of compounds if the emission time and photochemical loss rates are known or can be estimated. This has been done for simple aromatic HCs such as benzene and toluene [Roberts et al., 1984], which are generally co-emitted by motor vehicles and are lost primarily through reactions with OH radicals. Such an analysis has been done for the NEAQS 2002 data set as described by de Gouw et al. . In this case the analysis has been updated, so that an initial toluene/benzene ratio of 3.7 was used and an average [OH] of 3 × 106 molec/cm3 was assumed. The resulting times, shown in Figure 6 are as long as 30 h or more for the time period in which PPN/PAN reached 0.51.
 The ratio of PPN/PAN represents another potential indicator of age, due to a slight difference in the thermal decomposition rates of PPN and PAN [Kirchner et al., 1999]. The first step in thermal decomposition of the PAN-type compounds is the pressure-dependent reaction to form a peroxyacyl radical (e.g., CH3(O)OO, PA) and NO2 (shown here for PAN);
This reaction is about 15% faster for PAN relative to PPN. However, it should be recognized that net loss of PANs in the atmosphere is slower than the rate of Reaction (1), because of the reformation of PANs in the presence of NO2;
The loss of PA radical occurs primarily through the reaction with NO in NOx-impacted air masses.
The net rate can be expressed as that of Reaction (1), corrected for [NO] and [NO2] assuming those are the main reactions of PA radical;
The PA and PP radicals can also react with HO2 and other RO2 radical, however these are only important when [NOx] is below about 100 pptv. Measured NO2 averaged 3.9 ppbv for the period discussed below, so the radical reaction pathways were not important for that period. The change in [PPN]/[PAN] with time can therefore be calculated as a function of time, given the measured k1 as a function of temperature, measured values of k2 and k3, and average values for temperature, and [NO]/[NO2]. Figure 7 shows the results of the calculation assuming a starting ratio of [PPN]/[PAN] of 15%, the average temperature during the study, 293°K, and a series of [NO]/[NO2] ranging from 0 to 0.33, which span the ratios observed in this study (in the absence of power plant or ship plumes). Also shown are data points from the period from 17:00, 4 August to 02:00, on 5 August. The measurements are consistent with the notion that the air mass sampled during this time was impacted by North American emissions approximately 30 h previously.
5.2. The 18 July 2002 Event: Nighttime Loss of O3
 The measurements of PANs and associated NOx and O3 are shown in detail in Figure 8 for the period 0750 to 1130 UTC on 18 July 2002. These measurements were made fairly close to Boston, and the local wind vector indicated that air was coming from the Boston area for part of the time. This is corroborated by the presence of short-term perturbations in the NOx and O3 measurements; increases in NOx, and decreases in O3 corresponding to fresh plumes of NO, probably from power plants. In the middle of this period the local wind shifted direction from the southwest to the northeast. This corresponded to increases in PAN, MPAN, (and PPN) but not PPN/PAN nor the toluene/benzene photochemical age. This implies that the air observed during this time had been impacted by the Boston area the afternoon before [Warneke et al., 2004]. The presence of substantial PANs concentrations means that O3 was also produced in that air mass. However, O3 was suppressed relative to the levels usually associated with these kinds of air masses (see Figure 5).
 The loss of odd-oxygen through the action of NO3-N2O5 chemistry is proposed as the reason for the unusual PAN-O3 relationships observed during this period. The nighttime reaction of O3 with NO2 to form NO3, equilibration of NO2 and NO3 to form N2O5 and then chemical loss of NO3 and N2O5 due to reaction with VOCs and particles, respectively, serves as a sink for O3 in the lower troposphere [Brown et al., 2006]. This chemistry operates in an environment where PAN loss chemistry is very slow, since [NO]/[NO2] is 0 when NO3 chemistry is active, so the original PAN is preserved while O3 is lost. The production of PANs from NO3 reaction with aldehydes is possible, but under the conditions present on the night of 18 July accounts for less than 10pptv/hr of PAN, much too slow to account for the roughly 1 ppbv enhancement.
5.3. The 23 July 2002 Event: Strong O3 Production
 The most intense NOx-VOC photochemical event during the NEAQS 2002 study was observed on 23 July from 1400 to 2000 UTC. For this time period the Ronald H. Brown was located south of the Isles of Shoals (see Figure 1). This position allowed for sampling the extent of pollution that was entering the Gulf of Maine. The temperature for this time period was warm (high of 30°C) with clear skies. For almost the entire time, winds were out of the south to southwest allowing for transport of air into the sampling environment that had come from the east coast corridor of Boston-New York-Washington D.C., large urban/suburban complex. Figure 9 shows time series plots for the PANs and ozone for this sampling period. Elevated mixing ratios are reported for all the PANs and ozone. This pollution event also marks the highest observed levels of PAN (2790 pptv), MPAN (371 pptv) and ozone (123 ppbv) for the entire NEAQS campaign. Both PPN and MPAN levels were elevated for this sampling period.
 In addition to elevated PANs and O3, isoprene and its first-generation product, methacrolein (MACR) were also elevated. The initial isoprene peak in the early morning had ratios of [MACR]/[Isoprene] of about 0.33 while the noontime photochemical peak had ratios of 2. These are consistent with the results of Roberts et al.  for the entire 2002 data set, in which the systematic increase in that ratio is observed as air masses are transported and chemically processed over the Gulf of Maine.
6. Linear Combination Model
 Previous studies have illustrated how PANs data can be used to estimate the extent of photochemistry that is fueled by anthropogenic and/or biogenic VOCs of a particular region [Williams et al., 1997; Roberts et al., 1998a, 1998b, 2002]. This is achieved through a linear combination model that assigns PANs values to the two different VOC regimes. PAN is a “generic” secondary photochemical compound in that it can be formed through both biogenic and anthropogenic VOC photo-oxidation. The linear combination model therefore uses the PAN mixing ratio as the overall indicator of total photochemistry occurring. Meanwhile, PPN is largely formed through the photo-oxidation of anthropogenic VOCs (AHC). Therefore this linear combination model uses PPN mixing ratio as a marker of the extent of anthropogenic VOC driven photochemistry. MPAN is almost solely formed from biogenic VOCs (BHC), an aspect that has been confirmed in the NEAQS 2002 study by the source attribution work of de Gouw et al. , and therefore MPAN mixing ratio is used as a marker for biogenic VOC driven photochemistry. Using PPN and MPAN as markers of the different VOC-fueled photochemistry the PANs linear combination model predicts the overall amount of photochemistry with the measurement region through the following relationship:
The coefficient ‘a’ represents the ratio of PAN formed from isoprene chemistry in comparison to the formation of MPAN. Likewise, the coefficient ‘b’ represents the ratio of PAN formed from AHC chemistry compared to PPN.
 Aside from the assertion that PAN production can be described as a linear combination of the AHC and BHC chemical regimes as indicated by PPN and MPAN, respectively, the model makes a few other assumptions. First, it assumes that all of the PANs have the same loss rate. It is clear that this is not completely true; PAN and PPN have slightly different loss rates as described above, and MPAN has a faster loss rate that PAN and PPN due to the reaction with OH. However, the primary loss of all three of these compounds is largely driven by thermal decomposition, especially during the summertime, and as long as the times are kept reasonably short, a fraction of a day, the effect of this differential loss is minimized. The model also ignores mixing and surface deposition of PANs. Another assumption of the linear combination model is that the formation of the PANs occurs fast, again as long as times corresponding to a fraction of a day are used, this assumption should hold. The model will be applied to the limited case of 23 July, as it meets the above criteria. The following relationship was found for 23 July:
 The effectiveness of this model in predicting PAN from the AHC and BHC indicators can be assessed by calculating PAN from PPN, MPAN and the coefficients of fit, and comparing it with measured PAN. This is shown in Figure 10 for the 23 July data. With an r2 = 0.903 it appears that the linear combination model accurately estimates the amount of PAN during the 23 July event. The accuracy of this simple empirical model has not been previously tested with a full photochemical model, in part because of the lack of detailed chemical mechanisms. The Master Chemical Mechanism has a sufficiently explicit set of reactions that the PANs can be simulated accurately.
7. Master Chemical Mechanism Modeling
 The results of the MCM calculations for the cases listed in Table 1 can be used to examine the appropriateness of the above LCM analysis. The model was allowed to run for multiple days starting at 10:00 local time with the first day effectively serving as a “spin up” period. The results for MPAN showed a maximum concentration at 10:00 local time on the next model day. This was also a local maximum for PAN, PPN and O3 concentrations, and so is used to test the LCM for AHC-BHC chemistry. Figure 11a and 11b shows the PANs results for the different AHC-BHC combination cases. The maxima of MPAN, PPN and PAN at that time are plotted versus the initial Isoprene and AHC in the model. It is clear that the relationships of MPAN and PPN to initial conditions are somewhat non-linear, and that PAN has higher values at both AHC and BHC extremes.
 The linear combination calculation is easily accomplished in these cases, the PAN/PPN ratio for the pure AHC case, 5.0, is the coefficient for the PPN term, and given that value and the results for the high Isoprene case, a coefficient of 3.5 is calculated for the MPAN term. This then gives the modeled equation;
which is actually quite close (10% for PPN, 15% for MPAN) to the one derived above from measurements. Equation 7 can be used in, conjunction with MCM modeled PPN and MPAN, to calculate PAN and compare it to the MCM modeled PAN. This is shown for all six cases in Figure 12, wherein the data points lie quite close to the 1:1, with the Boston plume point approximately 15% off of the line. As with the measured data above, a good agreement between LCM-calculated PAN and the actual measured or fully modeled PAN implies that the linear combination model does a good job of attributing PAN production to the 2 different chemical regimes.
 The five next most abundant PAN compounds calculated by the model are APAN, PBzN, the butyl PANs, and pentyl PAN. The sum of these 5 other species was at most 3.8% of the sum of the major 3 PAN compounds. While this is a significant decrease in NOx compared to the starting conditions, NO2 was still fairly high, between 4 and 5 ppbv, at the PAN maximum on the second afternoon. This is far above the point at which other reactions (e.g., with RO2) can compete with the PA + NO2 reaction.
 The correlation of O3 and PAN to be verified using the concentrations calculated by the MCM model. Figure 13 shows a plot of the MCM O3 results versus MCM PAN for all of the six model cases. There is a reasonably linear correlation between the two species, confirming that the modeled atmosphere behaves in a manner similar to the observed atmosphere. It should also be noted that the Boston plume values, approximately 400 pptv PAN and 60 ppbv O3 lie in the middle of the distribution of points in the measured O3–PAN plot (Figure 5). This is further evidence that the MCM model provides a reasonable representation of the VOC-NOx photochemistry.
8. Estimation of Ozone From AHC and BHC
 The results from the PAN linear combination model can be used, along with the correlation of O3 and PAN, to estimate the amount of ozone formed from each VOC regime [Williams et al., 1997; Roberts et al., 1998a, 1998b, 2002]. The relationship between O3 and PAN for the 23 July period of interest can be reasonably fit with a straight line (Figure 5). The intercept of that fit, approximately 40 ppbv, can be thought of as a regional background, in other words, O3 that is there from the general influence of the continent, not as a result of production on that day. The following equations were used to estimate the amount of excess ozone produced from AHC and from BHC above the 40 ppbv background.
Figure 14 shows the estimated amounts of excess O3 produced from AHC and BHC chemical regimes. The model results show that the contributions to ozone production from AHC and BHC are roughly equal on 23 July. BHC chemistry, along with anthropogenic NOx, was sometimes responsible for up to 2/3 of the excess O3 on this day, and was a significant contributor to the O3 values in excess of the 84 ppbv national air quality standard (shown as line at 44 ppbv excess O3).
 Peroxycarboxylic nitric anhydrides (PANs) were measured during the NEAQS 2002 intensive experiment off the coast of New England and in the Gulf of Maine. The compounds observed were PAN, PPN, MPAN and PiBN, roughly in that order of abundance. Mostly clean North Atlantic air was observed during the first part of the cruise from South Carolina to 40°N and exhibited levels of PAN commensurate with other observations. The air north of 40°N was heavily impacted by North American photochemical oxidant chemistry. The relationships between PANs in this environment were similar to those observed at continental U.S. sites, with some differences. Those differences were related to the fact that the measurement platform was somewhat removed from the NOx and VOC sources that contributed to the PAN (and O3) formation, in contrast to ground sites that are surrounded by sources. The ratio of PPN/PAN was observed to change as a function of time due to differences in their thermal decomposition rates. The faster decomposition of PAN also shows up in the relationship between PiBN and PAN. The correlation of O3 and PAN revealed a population of points that appear to be low in O3 likely due to the nighttime loss of odd-oxygen driven by NO3 and N2O5 chemistry. The highest O3 observed during the cruise occurred on 23 July, and the linear combination model using PPN and MPAN as markers for AHC and BHC-driven chemistry, respectively, provided evidence that BHCs and isoprene in particular was responsible for roughly 50% of the O3 produced above background. This result was confirmed by explicit photochemical modeling using the Master Chemical Mechanism v3.1. It would be interesting to examine these PAN-PPN-MPAN relationships with regional models that include mixing and surface deposition because those are two processes that have very different impacts on these near-coastal MBL air masses relative to the continental studies done previously. Such a large and significant participation of isoprene in photochemical ozone production amplifies the need for a proper accounting of the role of BHCs in regional O3 chemistry in air quality control strategies to be applied to the Northeastern U.S.
 We thank the personnel of the NOAA RV Ronald H. Brown for their skill and professionalism during the 2002 field project. This work was conducted as part of the NOAA Air Quality Program. We thank Dr. J. W. Munger for the use of the Harvard Forest PAN data.