Environ. Sci. Technol. 2003, 37, 3522-3530

Past studies have shown that ON compounds are subject Conversion of Fogwater and Aerosol to chemical and photochemical transformations in the Organic to , troposphere, forming products that might potentially influ- ence the properties of atmospheric condensed phases and Nitrate, and NOx during Exposure to the bioavailability of N in deposition (1, 17-21). For instance, the more bioavailable, lower molecular weight compounds, Simulated Sunlight and such as amino acids and urea, usually account for less than 20% of the atmospheric ON pool (8, 9, 11, 22, 23), while QI ZHANG† AND CORT ANASTASIO* larger and less bioavailable species, such as humic substances, appear to be more abundant (24-26). Although biologically Atmospheric Science Program, Department of Land, Air, and refractory (27), humic substances are photochemically reac- Resources, University of California, One Shields Avenue, tive (28, 29), and exposure to sunlight might cause them to Davis, California 95616-8627 decompose into smaller and more bioavailable molecules. Indeed, sunlight illumination of organic matter from surface leads to the formation of amino acids and ammonium - Although organic nitrogen (ON) compounds are apparently (30 33). While similar reactions might occur in atmospheric ubiquitous in the troposphere, very little is known about drops and particles, no such studies have been performed, and little is known about the transformation rates and fates their fate and transformations. As one step in addressing of atmospheric organic nitrogen. this issue, we have studied the transformations of bulk The photochemical transformations of specific ON com- (uncharacterized) organic nitrogen in fogwaters and aerosol pounds, such as amino acids, nitrogen heterocycles, and aqueous extracts during exposure to simulated sunlight nitroaromatics, have been studied in atmospheric samples and O3. Our results show that over the course of several or under atmospherically relevant conditions (1, 19-21). hours of exposure a significant portion of condensed-phase However, studies of individual compounds are limited by organic nitrogen is transformed into ammonium, , the fact that the bulk of ON in atmospheric drops and particles nitrate, and NOx. For nitrite, there was both photochemical is uncharacterized (8, 9, 11, 12). Furthermore, because the formation and destruction, resulting in a slow net loss. reactivity of individual ON compounds varies widely (1, 19), Ammonium and nitrate were formed at initial rates on the it is currently infeasible to extrapolate from single-compound order of a few micromolar per hour in the bulk fogwaters, studies to transformation rates of bulk (and largely unchar- acterized) organic nitrogen in the atmosphere. Therefore, ∼ -3 -1 corresponding to formation rates of 10 and 40 ng m h , we initiated this study to examine photochemical transfor- respectively, in ambient fog. The average initial formation mations of bulk organic N in atmospheric fogwaters and -3 -1 + rate (expressed as ng (m of air) h )ofNH4 in the aerosol particles. We describe here a preliminary set of aqueous extracts of fine particles (PM2.5) was approximately experiments to examine whether bulk atmospheric organic one-half of the corresponding fogwater value. Initial N is transformed during exposure to sunlight and, if so, formation rates of NOx (i.e., NO + NO2) were equivalent whether these reactions form inorganic N. In addition, to ∼2-11 pptv h-1 in the three fogwaters tested. Although because ozone can be a major sink for some atmospheric the formation rates of ammonium and nitrate were organic nitrogen compounds (19, 22), the influence of ozone relatively small as compared to their initial concentrations on organic nitrogen transformations and inorganic N forma- tion was also investigated. in fogwaters (∼200-2000 µM) and aerosol particles (∼400- - 1500 ng m 3), this photochemical mineralization and 2. Experimental Methods “renoxification” from condensed-phase organic N is a 2.1. Samples, Materials, and Analyses. Since details of our previously uncharacterized source of inorganic N in the sampling dates and times, analytical methods, and sample atmosphere. This conversion also represents a new characteristics have been presented in prior reports (8, 9, component in the biogeochemical cycle of nitrogen that 22), only the main points are given here. Fogwaters were might have significant influences on atmospheric composition, collected at the Davis, CA, NADP site (CA88; 38°33′ N, 121°38′ condensed-phase properties, and the ecological impacts W) during winters from 1997 to 2001 using a Caltech Active of N deposition. Strand Cloudwater Collector (CASCC2). Immediately after collection, samples were filtered (0.22 µm Teflon) and stored frozen (-20 °C) in HDPE bottles. PM2.5 samples were collected 1. Introduction at the same location during August 1997-July 1998. Water- Organic nitrogen (ON) has been measured in precipitation soluble particle components were extracted into Milli-Q water (1-5), dry deposition (1, 5, 6), cloud waters (7), fogwaters (8), (g18.2 MΩ-cm) by sonication, followed by filtration (0.22 and atmospheric particles (4, 9-11). The ubiquity of these µm Teflon) and frozen storage (-20 °C) in HDPE bottles. + - - compounds suggests that they might play important roles in Concentrations of NH4 ,NO3 , and NO2 were analyzed atmospheric chemistry and in the biogeochemical cycling of using a Dionex DX-120 chromatograph (IC) with con- N. In addition, since organic forms often represent a ductivity detection. Dissolved organic nitrogen (DON) was significant and bioavailable portion of the total N in determined as the difference in inorganic N concentrations deposition (3-7, 12), atmospheric ON can be a nutrient in a given fogwater before and after adjustment to pH ≈3 burden to aquatic and terrestrial ecosystems (5, 13-16). and 24 h of illumination with 254 nm of light (to convert + - DON to inorganic forms): [DON] ) ([NH4 ] + [NO3 ] + * Corresponding author e-mail: [email protected]; tele- [NO -]) - ([NH +] + [NO -] + [NO -]) . phone: (530)754-6095; fax: (530)752-1552. 2 after 254 nm hν 4 3 2 before 254 nm hν † Present address: Cooperative Institute for Research in Envi- Concentrations of each N species in all samples were ronmental Sciences (CIRES), 216 UCB, University of Colorado, considerably larger than the corresponding detection limits + - - Boulder, CO 80309-0216. (i.e., ∼0.1 µM for NH4 ,NO3 , and NO2 and 1.0 µM for DON).

3522 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 16, 2003 10.1021/es034114x CCC: $25.00  2003 American Chemical Society Published on Web 07/19/2003 into the quartz cuvette (Figure 1). O3 from the “high- concentration” tube (∼1 ppmv) was used to oxidize NO to NO2 (see below). The air/ozone flow exiting the quartz cuvette was first filtered (5-µm unlaminated Teflon, Pall-Gelman) to remove any particles or drops generated during bubbling and then passed through a denuder coated with citric acid to collect FIGURE 1. Equipment used to measure the photoformation of NH3(g) and then a denuder coated with Na2CO3 to collect inorganic nitrogen in experiments with ozone purging. Chemical HNO2(g) and HNO3(g). Downstream of these two denuders formulas (e.g., NH and NO ) indicate the species collected on each was a 1-L Pyrex reaction flask where NO was oxidized to NO2 3 x ∼ component. (and more oxidized forms) by mixing with 1 ppmv O3 at a flow rate of 0.5 L min-1. The reaction flask was kept at ambient temperature and pressure and was covered in aluminum Concentrations of total N (TN) were determined from the foil to keep it dark. NO2 and other oxidized forms were + - sum of N in the sample solution (([NH4 ] + [NO3 ] + collected downstream of the reaction flask by two denuders - [NO2 ])after 254 nm hν, see section 2.2) plus, in the experiments coated with a solution of 10% guaiacol and 5% NaOH in + with ozone exposure, the concentrations of gas-derived NH4 , methanol (36). - - NO3 , and NO2 found on the denuders and reaction flask At measured time intervals, aliquots of the fog sample (section 2.3). Note that in the samples exposed to ozone, the were taken from the quartz cuvette; diluted 10-50 times + - - sum of nitrate and nitrite includes photoproduced NOx that with Milli-Q; and measured for NH4 ,NO3 ,NO2 , and DON. - - was converted to NO3 and NO2 in the reaction flask. At the same time, the exposed denuders and reaction flask 2.2. Dissolved Inorganic N (DIN) Formation during were replaced by a new set and were each extracted with 2.0 Simulated Sunlight Illumination. 2.2.1. Lyophilized Fog and mL of Milli-Q water. The aqueous extract from the citric + + PM2.5 Samples. Because concentrations of NH4 were gener- acid-coated denuder was analyzed for NH4 ; extracts from - - ally high in our samples (8, 9), we were initially concerned the other components were analyzed for NO3 and NO2 . + - - that it would be difficult to distinguish between the photo- Concentrations of NH4 ,NO2 , and NO3 formed in the formation of DIN and analytical variations. To increase our fogwaters were calculated by combining the amounts in the + ability to see the photochemical formation of NH4 , the first quartz cuvette with those collected on the citric acid-coated group of fogwaters and aerosol extracts were lyophilized (i.e., and Na2CO3-coated denuder, respectively. Concentrations + - - freeze-dried; 34) to reduce NH4 levels prior to illumination. of NOx were calculated from the sum of NO3 and NO2 To do this, NaOH (1 M in Milli-Q) was added to form NH4OH measured on the two guaiacol-coated denuders and in the stoichiometrically (based on the initial measurement of reaction flask. Controls were run in the same way as those + NH4 ); the sample was rapidly frozen in liquid nitrogen; and in section 2.2 (i.e., without O3 bubbling through the sample). then lyophilized to dryness on a Virtis 2SL freeze drier. To determine the relative amounts of NO and NO2 Afterward, the lyophilized residue was reconstituted with produced, in two experiments we modified the experimental Milli-Q water and H2SO4 (0.1 M in purified water) to the setup (Figure 1) so that one of the two guaiacol-coated original sample dilution and pH. denuders was placed before the reaction flask. Under this Irradiations were conducted using a solar simulator configuration, the NO2 concentration was calculated as the - - (Spectral Energy) (35) with a sample chamber maintained at sum of NO3 and NO2 measured on the first guaiacol-coated 20 ( 0.2 °C by a Neslab RTE 211 water bath. A total of 8.0 denuder (which should have collected NO2 and not NO). mL of the lyophilized and reconstituted sample was kept The NO concentration was calculated from the sum of nitrate within a 2.0-cm airtight quartz cuvette (Spectrocell) and and nitrite measured in the reaction flask and on the second stirred continuously. The light beam of the solar simulator guaiacol-coated denuder (after conversion of NO to NO2, - - was unfocused to illuminate the entire solution. At measured NO2 , and NO3 in the reaction flask). time intervals, aliquots of sample were taken from the quartz As a check on background concentrations, two complete + cuvette and analyzed for NH4 . With every illuminated sample sets of freshly coated but unexposed denuders and a clean ∼3 mL of identical sample was placed in a stirred 1.0-cm reaction flask were extracted and analyzed for inorganic N + - quartz cuvette kept in the dark at 20 °C to monitor changes species. The average concentrations of NH4 and NO3 + in [NH4 ] in the dark. detected on the citric acid- and Na2CO3-coated denuders, 2.2.2. Unaltered Fog Samples. For this set of experiments, respectively, were less than 10% of the corresponding lowest + unaltered fog samples with relatively low initial NH4 and sample values. N species on the other components were less - NO3 concentrations were tested using the conditions than detection limits. We also performed one procedural described above. At measured time intervals, aliquots of blank where Milli-Q water was used in place of a fog sample illuminated and dark samples were removed from the quartz and was purged with 70 ppb O3 and illuminated with cuvette; diluted 10-20 times with Milli-Q water (measured simulated sunlight for 2 h (the typical exposure time for a + - - gravimetrically); and analyzed for NH4 ,NO3 ,NO2 and, in given set of denuders and reaction flask during a kinetic + two samples, DON. run). Concentrations of NH4 in the quartz cuvette and on < 2.3. Formation of DIN and NOx during Exposure to the citric acid-coated denuders were 5% of the average Simulated Sunlight and O3. The apparatus for this set of values measured in illuminated fog samples; concentrations - experiments is illustrated in Figure 1. Unaltered fog samples of NO3 on the Na2CO3-coated denuder and the reaction (∼13 mL) were irradiated with simulated sunlight (see section flask were <10% of the average sample value. N species on 2.2) in a 5-cm quartz cuvette at 20 °C while O3 (78-93 ppbv) the other components were below detection limits. These was bubbled through the cuvette at a flow rate of 0.3 L min-1. blank levels of each species were subtracted from all sample ) O3 was generated by passing O2 (>99.6%, Matheson) through values obtained after t 0. one of two quartz tubes (GE 021, 5 mm i.d., 180 mm length) 2.4. Calculations and Tests of Statistical Significance. illuminated with an ozone-generating Mercury Analamp (620 Initial reaction rates (Ri, where i refers to a given nitrogen mm length, BHK, 81-1127-01). O3 from the “low-concentra- species) were determined from plots of concentration versus tion” tube (∼300 ppbv) was diluted with purified air (Aadco illumination time using an appropriate fit performed with + - 737-R) to a mixing ratio of ∼100 ppbv, saturated with water Sigma Plot 7.0 (SPSS Inc). Values of Ri for NH4 and NO3 vapor (by passing through a water bubbler at 20 °C), and fed formation were determined by fitting a three-parameter,

VOL. 37, NO. 16, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 3523 exponential rise to a maximum equation to the experimental have caused us to under-report rates of N transformation. data: On the basis of our calculations, light screening by the fogwater samples (i.e., the inner filter effect; 39) was relatively ) + - -bt < [i] [i]0 a(1 e ) (1) small ( 13%); therefore, our reported reaction rates were not increased to account for this effect. where [i] is the concentration of the species at time t, and To distinguish between actual changes in the concentra- [i]0, a, and b are parameters defined by the regression fit. tion of a given N species during an experiment and changes Conceptually, the variable [i]0 is the initial concentration, a due to analytical variations, we performed a student t-test is the product of concentration of the parent DON compound with Microsoft Excel assuming a two-tailed distribution to times the yield for formation of i, and b is the first-order rate compare the mean values of each pair of concentrations constant for formation of i. The initial rate (i.e., at t ) 0) of (i.e., before and after illumination). We set the error level at + - < NH4 or NO3 formation was calculated as 10% (i.e., p values 0.10 were considered significant while those >0.10 were considered insignificant). Changes that were ) Ri ab (2) not statistically significant are flagged as such in the tables of results. The p values for changes in the concentrations of + - - RNOx (i.e., the initial rate of NOx formation) was determined NH4 ,NO3 , and NO2 were usually calculated based on using the same equations except that the intercept was set standard deviations of replicate injections. When no replicate to zero in eq 1. measurements were performed (e.g., due to limited sample - Destruction rates of NO2 during illumination were volumes), p values were calculated using the average relative determined using a two-parameter, exponential decay fit: standard deviation for each N species from our experimental + - - data (1.0% for NH4 and NO3 and 2.3% for NO2 ). The p - - ) - values for changes in the concentrations of NOx, DON, and [NO2 ] [NO2 ]0 exp( jNO -t) (3) 2 total N (which were calculated from a linear combination of + - - - - independent measurements of NH4 ,NO3 , and NO2 ) were where [NO2 ]0 was a fitted parameter and jNO2 was the fitted - calculated using propagated standard deviations from the first-order rate constant for NO destruction. The initial 2 independent DIN measurements (40). destruction rate of nitrite was then calculated from

- 3. Results and Discussions R - ) [NO ] j - (4) NO2 2 0 NO2 + 3.1. Photoformation of NH4 in Lyophilized Fog and Aerosol + Samples. We first examined the formation of NH4 in 7 fog For DON loss, a single-exponential decay regression generally samples and 3 aerosol extracts that were lyophilized to reduce gave a poor fit, and so we used a double-exponential, four- + the high initial concentrations of NH4 so that we could more parameter fit: + sensitively determine changes in NH4 during experiments (section 2.2). Lyophilization successfully reduced NH + [DON] ) 4 concentrations in these samples to approximately 1% of the [DON ] exp(- j t) + [DON ] exp(- j t) (5) + 1 0 DON1 2 0 DON2 original values (Table 1). The photoformation of NH4 occurred in all 10 illuminated samples (Table 1). An average + where, conceptually, [DON1]0 and [DON2]0 are the sizes (µM) of 8.9 µMNH4 (range ) 1.2-21 µM) was formed in the of the two DON pools and jDON1 and jDON2 are the rate constants fogwaters during an average illumination time of 520 min, -1 + (min ) for their destruction. The initial rate of DON while 0.59-1.1 µMNH4 (average ) 0.83 µM) was formed in destruction was calculated using the aerosol extracts during 240 min of illumination (Table 1). All of these changes were significant at a 90% significance R ) [DON ] j + [DON ] j (6) DON 1 0 DON1 2 0 DON2 level. The photochemical and thermal (dark) formation of + NH4 in the lowest, median, and most reactive fog samples For species where plots of concentration versus time were and in all three aerosol extracts are shown in Figure 2. The linear (or where only two data points were available), the average ((1σ) initial formation rates in the illuminated fog + ( reaction rate was determined as the slope of a linear samples and aerosol extracts (RNH4 ,hν) were 3.0 4.0 and 0.4 regression to the data. Linear fits were used for most of the ( 0.1 µMh-1, respectively (Table 1). In part, rates in the dark control data. particle extracts were lower than in the fogwaters because Actinic flux values in each experiment were determined the extracts were much more dilute than the fogwater by measuring the rate constant for loss of 2-nitrobenzalde- samples: the average ((1σ) concentrations of ON prior to hyde (2NB) (j2NB,EXP) at the end of a given experiment (37, lyophilization were 522 ( 480 and 94 ( 43 µM in fogwater + 38). Cells were not air-purged during actinometry measure- and aerosol extracts, respectively; the average [NH4 ] prior ments since purging had a negligible effect on 2NB photolysis to lyophilization in the fogwaters was ∼7 times greater than rates. Values of j2NB,EXP measured during this study were ∼2-6 that in the particle extracts (8, 9). Converting the solution + times higher than the rate constant for 2NB loss under volume-based rates of NH4 formation to air volume-based midday, winter-solstice sunlight in Davis (j2NB,WIN ) 0.00697 rates (using measured fog liquid water contents, volumes of s-1; 19). However, we did not adjust reported rates of N air sampled, and volumes of aerosol extract solution) yields + -3 -1 transformation (e.g., Ri) for differences in actinometry since values of 4.0 ( 5.6 and 1.9 ( 0.95 ng of NH4 (m of air) h our initial evidence indicates that the photochemical forma- in fogwaters and PM2.5, respectively (Table 1). tion/destruction rates for N species are not linearly pro- The fact that the average air volume-based formation rate + portional to j2NB,EXP. In the one fogwater tested, increasing of NH4 in the fogwaters was ∼2 times larger than that in the the actinic flux (i.e., j2NB,EXP) by a factor of ∼4 only increased PM2.5 suggests that the fogwaters contained greater amounts - + the destruction rates of DON and NO2 by factors of ∼2.0 of photochemically labile DON precursor for NH4 . This might and 2.7, respectively, and increased the formation rates of be because (i) the fog samples contained DON from both + - NH4 ,NO3 , and NOx by factors of 1.3, 3.3, and 1.4, fine and coarse (>2.5 µm) particles (since both can act as respectively. Thus our N transformation rates might be over- condensation nuclei for fog drop formation) and/or (ii) most reported (relative to winter-solstice sunlight) in samples with of the fogwaters were collected during the night (when the high j2NB,EXP values. However, as discussed in sections 3.1 DON is protected from photochemical degradation) while and 3.3, there are also a few countervailing factors that could the aerosol samples were typically collected over a continuous

3524 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 16, 2003 + TABLE 1. Formation of NH4 in Lyophilized Fog Samples and Aerosol Extracts

+ + illumination conditions formation ratesc [NH4 ] prior to [NH4 ] after + b + + lyophilization lyophilization t j2NB ∆[NH4 ] RNH4 ,hν RNH4 ,hν a -1 -1 + -3 -1 d sample no. (µM) (µM) (min) (s ) (µM) (µMh ) RNH4 ,dark (ng m h ) Fog Samples DA97-05F 473 1.2 600 0.019 5.2 0.9 0.4 1.9 DA98-09F 1287 3.1 306 0.023 6.4 3.2 0.1 4.7 DA98-13F 719 1.9 360 0.020 1.2 0.2 0.0 0.27 DA99-01F 514 1.5 690 0.023 3.0 0.5 0.1 0.88 DA99-04F 3151 24 660 0.022 15.4 2.7 1.0 2.8 DA99-05F 1437 32 300 0.023 21 12 0.9 16 DA99-07F 1128 5.5 715 0.023 10 1.4 0.3 1.0 mean ( 1σ 1244 ( 921 9.9 ( 13 519 ( 188 0.022 ( 0.002 8.9 ( 7.1 3.0 ( 4.0 0.4 ( 0.4 4.0 ( 5.6 Aerosol Extracts ADA97-49/50e 87 3.1 240 0.024 0.59 0.4 0.02 0.91 ADA97-57/58f 200 1.9 240 0.024 1.1 0.5 0.03 2.8 ADA98-09 249 2.3 240 0.024 0.86 0.4 0.04 2.1 mean ( 1σ 178 ( 83 2.5 ( 0.8 240 ( 0 0.024 ( 0.000 0.8 ( 0.2 0.4 ( 0.1 0.03 ( 0.01 1.9 ( 0.95 a Information on sampling dates and times are given in refs 8 and 9. pH values of the samples ranged from 5.7 to 7.0 for the fogwaters and b + from 5.6 to 6.3 for the aerosol extracts. Changes in NH4 concentration during illumination. All changes were significant at p < 0.10 (see section c + ) + + ) + d + 2.4). RNH4 ,hν NH4 formation rate in illuminated cell; RNH4 ,dark NH4 formation rate in dark cell. Formation rate of NH4 in the ambient fog or aerosol, calculated using measured aqueous formation rates and ambient sample information (liquid water content for fog or the volumes of air sampled and aqueous extraction volumes for aerosol samples). e Mixture of equal amounts of extracts from ADA97-49 and ADA97-50. f Mixture of equal amounts of extracts from ADA97-57 and ADA97-58.

+ FIGURE 2. NH4 formation in lyophilized fog samples and fine particle extracts illuminated with simulated sunlight (plots a and c) and kept in the dark (b and d). Samples: 1, DA99-05F; 2, DA98-09F; 3, DA98-13F; 4, ADA97-57 + ADA97-58; 5, ADA98-09; and 6, ADA97-49 + ADA97-50. Symbols with error bars represent the average value ((1σ) from two replicate injections. Lines represent the regression fits to each set of experimental data.

2-day period (8, 9). A strong correlation (R 2 ) 0.83) between It should be noted that the measured rates described above + -3 -1 RNH4 ,hν (ng (m of air) h ) and the concentration (ng of N (and in subsequent sections) might overestimate or under- (m of air)-3) of combined amino N in these fog and PM estimate actual rates of inorganic N formation in ambient samples (8, 9, 22) suggests that combined amino compounds fog and PM2.5. There are several reasons for this. First, while such as proteins and peptides might be an important our actinometry indicates that the actinic flux in this set of + precursor of photochemically formed NH4 . In contrast, experiments was approximately 2.5 times higher than that + RNH4 ,hν was only very weakly correlated with initial concen- at midday on the winter solstice in Davis, initial data indicate + - - + trations of NH4 ,NO2 ,NO3 , total organic N, and free amino that rates of NH4 are only weakly dependent on actinic flux N(R 2 ) 0.13, 0.01, 0.07, 0.01, and 0.01, respectively). (section 2.4). In any case, this difference in actinic flux would + The formation of NH4 was observed in the dark in both be largely eliminated if we were to take into account + fogwaters and PM2.5 extracts, indicating that NH4 is formed numerical modeling results, which indicate that the actinic thermally as well as photochemically (Figure 2). However, flux in an aqueous drop is ∼2 times greater than in the + ∼ the thermal (i.e., dark) reaction rate (RNH4 ,dark) was only 10% surrounding gas phase (41). Second, although our samples of the corresponding rate in illuminated samples (Table 1). were stored frozen between collection and the experiments In addition, samples with higher photoformation rates of (a period of ∼1-2 yr), past studies have shown that ON can + + + NH4 (RNH4 ,hν) typically had higher dark rates (RNH4 ,dark), decay during this storage (8). This loss of labile ON likely although the correlation is relatively weak (R 2 ) 0.51). would cause our measured rates of IN formation to under-

VOL. 37, NO. 16, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 3525 , 0 ] estimate the actual rates in the atmosphere. The loss of ON ) 2 1 ν 3.1 (and accompanying underestimation of IN formation rates) - ( DON,h appears to have been somewhat exacerbated by lyophilization Mh R , [DON 52 54 56 µ 1 (

on the basis of the one sample tested where lyophilization . pH values - - -

9 DON reduced DON concentrations by 30% but did not significantly j ,

- 0 ] and 2.0 affect levels of NO3 . Finally, filtration of our samples might b 1

8 M) ( have also reduced sample reactivity since our filtered samples during illumination. µ [DON] - DON ( ∼ - ) - 2 44 45 47 typically have 30 40% less absorbance (λ 290 500 nm) ∆ - - - , formation or destruction as compared to unfiltered samples. However, a lowered ν h , i sample absorbance might not necessarily lead to a reduced R 0 16

sample reactivity, since according to a recent review (33), ( M) µ there is no correlation between light absorptivity and the ( [DON] 83 Lifetime of NO - photoformation rate of ammonium in aquatic samples. c (eq 5). Values of [DON

3.2. DIN and DON Dynamics during Simulated Sunlight 2 c 6.7 DON - j Illumination in Unaltered Fog Samples. Although the 2 (h) ). In the dark there were no detectable changes ( × NO - 2 lyophilization process allowed us to more sensitively measure τ 0 + ] NH4 formation rates, it also apparently reduced concentra- 2

tions of ON as described above. To avoid this possible bias, ) 1 ν 0.7 15 [DON h and NO , - - we examined the photochemical reactions of N species in + + 2 4 ( 1 Mh

five fogwater samples that were unaltered (i.e., without ΝÃ

R µ 1.1 13 1.1 2.10.6 6.8 19 0.6 22 94 DON ( lyophilization). To enhance our ability to measure DIN j 0.11). ------× ) 2

formation, we chose samples with relatively low initial 0 0.10 except as noted (see section 2.4). ] p a , respectively, for DA99-02F. ( 1 1 NO 0.19). Sampling dates and times are given in refs ]

∼ ant (for NH - <

concentrations of inorganic N (on average 5 times lower - 2.0 - 3 2

p ) (

than the typical value in Davis fog samples; 8). Three of the M)

p min [DON µ 4 [NO (

five unaltered samples showed a statistically significant - ) 7.4 6.0 5.5 3.4 7.7 dl 72 + ∆

< 10 - - - - - increase in [NH4 ](p 0.10), while the change in one sample < DON × was significant at a slightly larger value (p < 0.11, Table 2). R

+ 0 ] On average 15 µMNH4 was produced during ∼490 min of 1.5 - 2 M) (

irradiation (Table 2) and initial rates ranged from 1.8 to 8.4 µ + - ( dl

1 [NO µMNH4 h (average (1σ ) 5.7 ( 2.7; Table 2). This value 0.56); DA98-13F NO < M, and 5.0 ) was roughly two times higher than the average initial rate of µ p ) + ( 1 ν

NH formation in the lyophilized samples (Table 1), sug- ,65 h

4 + - 2.5 13 , 1 4 - - 3

gesting that lyophilization significantly reduced sample ( NO Mh

R reactivity. µ ( - 7.0 NO3 was formed in all five of the illuminated samples; ] -

in four samples the increase was statistically significant at 3 4.0 6.3 - 3 M)

< < M, 0.032 min p 0.10, while in the fifth was nearly so (p 0.11; Table 2). ( µ d µ [NO (

- NO

The initial rates of NO3 formation were very similar to those ∆ + -1 of NH4 , with an average ((1σ) value of 6.3 ( 2.5 µMh 0.11); DA99-01F NH ), but this value was statistically insignificant (

- + ) 1 1 ) - 0 ] (range 3.7 9.9 µMh ; Table 2). In contrast to NH4 and -

128 14 p - - - ( 3 M) + NO3 , we observed a net loss of NO2 in these fogwaters ( 4 µ Mh

- ( ( )- ( 1 µ during illumination (average 1σ 1.1 0.7 µMh , Table [NO - - 2). The average lifetime of NO2 (τΝÃ2 ) during these experi- 0.67

( ) - ) 1 ments was 15 6.7 h (range 6.8 22 h; Table 2), which is ) ν

h - 2.7 321 , + considerably larger than the lifetime of nitrite measured in 4 ,dark ( - NH

- Mh Milli-Q water (τΝÃ ) 5.8 h) under similar conditions (pH 3 2 R µ (

- NO ) 1 8.41.8 417 249 14 10 3.9 8.2 14 11 5.6; j2NB,EXP 0.022 s ). Since direct photolysis accounted R - ] ∼ 0.10: DA98-16F NH 9 5.7 for 90% of NO2 loss in the illuminated Milli-Q and since + 4 d < + 4 M) ( the inner-filter effect in the fogwaters was negligible (section , respectively, for DA98-13F and 30 d p 1 µ [NH

- ( - 2.4), the slower destruction rates of NO2 in the fog samples NH ∆ ], change of concentration after illumination. All listed changes were significant at - i [ min

indicate that NO2 was photochemically formed in the were obtained using a four-parameter, double-exponential fit to the data. ∆ 4 - fogwaters in addition to being photochemically destroyed. 0 ] 123 15 ,DON 10 - + ν 4 The net loss of NO2 in the illuminated fog samples indicates M) h ( ×

µ R (

that the photochemical destruction of nitrite was more rapid [NH than its photochemical formation. Although all five fog samples studied had significant Values of 155 574

( M, and 5.0 amounts of DON (average of 238 70 µM in the original b ν µ h

t (

samples; 8), due to limited sample volumes we followed DON (min) , initial concentration; 0 ,37 ] 1 loss in only two samples (DA98-13F and DA99-02F). Both i samples lost more than half of their initial DON over the - ∼ < changed in only one sample (DA99-02F; course of 10 h of illumination (p 0.10; Table 2). In both - 3 0.002 486 samples a single exponential regression gave a poor fit to the ) 1 - ( 2NB loss of DON during illumination but a double-exponential j (s M, 0.023 min fit (eq 5) worked well (e.g., Figure 3d). This suggests that the µ , while NO 0.019 DON compounds in fogwaters fall broadly into two pools: - one that is relatively rapidly mineralized and another that is 2 σ 1 more resistant to mineralization. On the basis of the curve- were 36 2 ( , duration of illumination; [ fitting parameters determined for these DON decays (Table ν

h DON t j

2), the sizes of the more reactive DON pools (i.e., [DON]1) sample a The change in concentration was statistically insignificant at DA97-01F 0.018 400 419 19 7.3 201 14 9.0 15 DA98-13F 0.020 600 730 496 14 DA98-16F 0.016 380 661 16 DA99-01FDA99-02F 0.021Mean 0.018 350 700 518 539 3.2 23 4.2 241 21 3.7 14 TABLE 2. Changes in Inorganic and Organic Nitrogen Concentrations in Unaltered Fogwaters during Simulated Sunlight Illumination in DON or NO rate in illuminated cell. Missing data were either not measured (for DON) or represented concentration changes that were not statistically signific of the samples ranged from 6.6 to 7.5. were up to 50% smaller than those of the less reactive pools and d

3526 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 16, 2003 experimentally measured values to determine actual reaction rates in ambient Davis fogwaters. In addition to the uncer- tainties described earlier, given that concentrations of inorganic N and DON are positively correlated in fogwaters (8) and that high DON levels suggest a larger reservoir of + - NH4 and NO3 precursors, our results in Table 2 might be biased to lower rates since we specifically chose samples with low initial concentrations of DIN. The loss of DON and formation of DIN in one fog sample (DA99-02F) are plotted in Figure 3. As seen in the other fog + - samples, the formation rates of NH4 and NO3 in DA99-02F - were similar, while the NO2 destruction rate was smaller and the DON destruction rate was much faster. In the dark cell, none of the species showed concentration changes distinguishable from analytical variations (i.e., p > 0.10; section 2.4). As shown in the bottom panel of Figure 3, the amount of total N in the quartz cuvette decreased by ∼10 µM during illumination, that is , the amount of DON lost was + - - ∼10 µM greater than the sum of NH4 ,NO3 , and NO2 formed (p < 0.10). In contrast, there was no statistically significant change of total N in the dark (p > 0.10). These results suggest that approximately 10 µM DON was converted into one or more volatile N species that escaped from the liquid phase during illumination and were not accounted for. This “missing” N species likely was produced from the more reactive DON pool since it was apparently produced during the initial stage of illumination (Figure 3d,e). As described in section 3.3, it appears that NOx was this missing inorganic N species. To examine the effects of solution pH on DON miner- alization, we characterized N photochemistry in DA99-02F at its original pH (6.4) and at pH 3.2 after adjustment with H2SO4. Acidification did not significantly influence the rate + of NH4 formation (Figure 4a), but it did accelerate the loss - - of N(III) (i.e., NO2 and HNO2) and the formation of NO3 - FIGURE 3. Concentration changes of nitrogen species in an unaltered (Figures 4b,c). The rates for N(III) destruction and NO3 fogwater (DA99-02F) during simulated sunlight illumination (open formation were 20 and 6 times faster, respectively, at pH 3.2 symbols) and in the dark (solid symbols). The value of j2NB,EXP/j2NB,WIN than at pH 6.4. The more rapid transformation of N(III) at for this experiment was 2.5. Each data point represents an average the lower pH was likely caused by the increase in the of two replicate injections, with error bars representing (1σ. Solid abundance of HNO2 at lower pH (pKa ) 3.25; 43), which is lines represent regression fits to the data (plots a-d). In plot e, much more photochemically reactive than its conjugate base + - - - - [Total N] is the sum of the concentrations of NH4 ,NO3 ,NO2 , and (NO2 ) in sunlight (44). The faster NO3 formation at lower DON in the fogwater in the quartz cuvette. The dashed line in plot pH was likely due to a more rapid transformation of N(III) e is the average of [Total N] in the dark and light solutions at rather than a faster DON mineralization, as suggested by the ) - t 0. fact that the sum of N(III) and NO3 was almost the same in the different pH solutions throughout the course of il- (i.e., [DON]2), while the lifetime of the more reactive pool lumination (Figure 4b). This experiment indicates that the ∼ ∼ (1/jDON1 1 h) was roughly 30 times shorter (1/jDON2 33 h). photochemical conversion of dissolved organic N species - + The net result was a quick depletion of the “more reactive” into NO3 (and NH4 ) was independent of pH and that a - ON compounds followed by a slower photodestruction of portion of the photoproduced NO3 resulted from N(III) the remaining DON (e.g., Figure 3d). Initial destruction rates photooxidation, while the remainder came from DON of DON in DA98-13F and DA99-02F were 52 and 56 µMh-1, mineralization. respectively, much higher than the sum of the initial 3.3. Photoformation of DIN and NOx in Unaltered Fog - + formation rates of NO3 and NH4 in these samples (Table Samples Exposed to Simulated Sunlight and O3. In an effort 2). It should also be noted that even those DON compounds to determine whether NOx was the missing product formed that were not mineralized likely underwent significant from DON destruction in DA99-02F, we modified our chemical transformations as a result of direct and/or indirect experimental apparatus to be able to collect NOx (section 2.3 photooxidation reactions to form products such as alkyl and Figure 1) and exposed the sample to simulated sunlight nitrates (42) and a suite of other compounds. during purging with 93 ppbv ozone. During 420 min of - + The average ((1σ) air volume-based initial formation rates exposure, 8.9 µMNO3 and 20 µMNH4 were formed while + - - for NH4 and NO3 in the unaltered fog samples were 12 ( 3.9 µMNO2 and 33 µM DON were destroyed (Figure 5). -3 -1 -3 -1 - + 8.4 ng m h (range ) 3.3-23) and 42 ( 12 ng m h Calculated initial rates of NO3 and NH4 formation in this (range ) 25-54), respectively. The initial destruction rate experiment were 7.4 and 11 µMh-1, respectively, while that - ( -3 -1 ) - - -1 - ) for NO2 was 6.4 5.6 ng m h (range 3.1 15). Assuming of NO2 destruction was 1.3 µMh (τΝÃ2 12 h). These an average molecular mass of DON of 100 Da per N atom values were ∼2-2.5 times higher than those measured in (9), we calculate that the destruction rates of DON in DA98- the same sample during simulated sunlight illumination 16F and DA99-02F were 490 and 610 ng m-3 h-1, respectively. without ozone (Figure 3, Table 2). Surprisingly, the initial Although these samples were not lyophilized, as discussed rate of DON loss in the presence of O3 was slower than in at the end of section 3.1 there are still a number of the experiment without ozone (Figures 3d and 5d). The uncertainties that prevent us from quantitatively using our kinetics of the DON decay were also different in the two

VOL. 37, NO. 16, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 3527 FIGURE 4. Influence of pH on inorganic nitrogen formation rates in (unaltered) sample DA99-02F during simulated sunlight illumina- tion (j2NB,EXP/j2NB,WIN ) 2.5). Filled symbols (and solid lines) represent results from the original, unaltered sample (pH 6.4), while open symbols (and dashed lines) represent results from the sample after adjustment to pH 3.2. In plot a, the data point at t ) 36 min was omitted from the regression fit for the pH 3.2 data. In plot b, the inverted triangles represent the changes in the total concentration - - of NO3 and N(III) [i.e., ∆([NO3 ] + [N(III)])] at pH 6.4, while the upright triangles represent changes at pH 3.2. The dotted curve is - an exponential regression fit to all of the ∆([NO3 ] + [N(III)]) data. - (Note that N(III) ) NO2 + HNO2.) In plot c, the regression fit to the pH 6.4 data was performed with the data point at t ) 0 min omitted. FIGURE 5. Changes in nitrogen speciation in (unaltered) fogwater experiments: in the presence of light only (no ozone) the DA99-02F during exposure to simulated sunlight and purging with DON decay was well described by a double-exponential fit + - - O3 (93 ppbv). Each data point for NH4 ,NO3 , and NO2 represents (Figure 3d), while the combination of light + ozone gave a the average ((1σ) of two replicate injections. Error bars on each DON decay that was well-described by a single-exponential DON, NOx, and total N data point represent (1σ calculated from + - regression (Figure 5d). These differences in rates and kinetics propagated standard deviations of the corresponding NH4 ,NO3 , - might be due to a change in the DON transformation and NO2 measurements. Open symbols represent samples exposed mechanism, possibly due to the presence of ozone, and/or to simulated sunlight and O3, while filled symbols represent samples because of the higher actinic flux employed in the second in dark, unpurged controls. In plot f, [Total N] is the sum of the -1 -1 + - - experiment (j2NB,EXP ) 0.044 s as compared to 0.018 s in concentrations of NH4 ,NO3 ,NO2 , and DON in the fogwater plus the previous one). In addition, although the initial DON the IN amounts from the denuders and reaction flask. For this ) concentrations in DA99-02F were nearly the same in the two experiment j2NB,EXP/j2NB,WIN 6.5. experiments (Figures 5d and 3d), the bulk reactivity of DON might also have changed because the second experiment actinic flux was higher than that of wintertime sunlight ∼ was performed 1 yr after the first (during which time the (j2NB,EXP ≈ 6.5 j2NB,WIN); (ii) this fog sample (DA99-02F) might sample was stored frozen). have suffered significant losses of reactive ON during storage; A total of 7.5 µMNOx (expressed as an equivalent aqueous and (iii) the mixing ratio of O3 used (93 ppbv) was much concentration in the fogwater sample) was formed in DA99- higher than typical values in the interstitial air of winter fogs 02F during the 420 min of illumination with O3 exposure at Davis (45). In addition, the characteristic time for -1 (Figure 5e). The initial formation rate of NOx was 3.7 µMh , evaporation of NOx in the bulk fog solution during the -1 corresponding to an air volume-based formation rate of 10 experiment (τevap ∼ 1 s at a purging rate of 0.3 l min ) was -1 -3 pptv NOx h (liquid water content ) 0.11 g (m of air) ). NO2 much longer as compared to that in ambient fog drops (τevap accounted for 31% of the NOx formed during this experiment. ∼ 0.022 s in a 20 µm diameter drop). One consequence of The inclusion of NOx in the N budget resulted in a better the slow evaporation in our experiments is that a significant - - + - balance of total N (i.e., [NO3 ] + [NO2 ] + [NH4 ] + [DON] portion of the aqueous NOx was likely converted into NO3 + [NOx]) throughout the experiment (Figure 5f). and other products (46) before it could be purged out of Extrapolation of this measured NOx formation rate to the solution. We calculate that the dominant sinks for aqueous • - ambient foggy air is complicated by several factors (in NOx were reactions with superoxide ( O2 ) and self-reactions - addition to those discussed earlier): (i) the experimental (e.g., NO + NO2), both of which eventually produce NO3

3528 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 16, 2003 + - - (46). Therefore, the upper bound for RNOx would be the sum phases are converted into NH4 ,NO3 ,NO2 and NOx during - of the measured formation rates of NOx and NO3 , that is, exposure to sunlight and O3. This transformation of ON to -1 -1 11 µMh (or 30 pptv NOx h on an air volume basis) in inorganic N is a previously unrecognized source of these DA99-02F. inorganic nitrogen species in the troposphere. Although the To understand the fraction of photoproduced NOx that formation rates of ammonium and nitrate were relatively came from DON reactions, we subtracted the rate of NOx small as compared to their initial concentrations, the absolute - formation due to the direct photolysis of NO3 and N(III) rates of formation were significant (i.e., up to nearly 10 µM using actinometry-normalized rate constants reported pre- h-1). As we note in several previous sections, quantitatively viously (38, 44). On the basis of these data and an estimated extrapolating from these laboratory-derived rates to rates in evaporation efficiency for NOx of 50%, we determined that the ambient atmosphere is not yet possible. Furthermore, approximately half of NOx production in DA99-02F during quantitatively describing or modeling the influence of organic - the light + ozone exposure was due to NO3 and N(III) nitrogen transformations as a source of inorganic N in the photolysis and that approximately half of the NOx was from atmosphere will require more extensive measurements of DON photodestruction. It is interesting to note that even ON in tropospheric particles and drops. though nitrate is photochemically reduced to NOx (47, 48), Despite these uncertainties, our results indicate that the - in our samples we generally saw an increase in NO3 during photochemical conversion of atmospheric organic N into illumination (although sometimes we could measure no net inorganic N is significant and could have important impacts. change), indicating that aqueous photochemical reactions For example, given the widespread occurrence and abun- were generally a net source of nitrate. dance of tropospheric ON (1, 5, 7-9, 11, 12) and the fact that The formation of NOx was also detected in the other two inorganic N is typically much more bioavailable than organic samples tested with simulated sunlight and ozone. Initial N(55), these transformations might significantly increase formation rates during the experiments were 1.5 µMh-1 the bioavailability of N in deposition. In addition, these same (DA99-04F, j2NB,EXP/j2NB,WIN ) 6.1, NO2:NOx ) 0.55) and 2.1 reactions likely occur after deposition (e.g., in surface waters), -1 µMh (DA01-01F, j2NB,EXP/j2NB,WIN ) 5.9), corresponding to thereby further increasing the ecological impacts of deposited -1 initial rates of NOx release of 2.1 and 5.1 pptv h , respectively, atmospheric organic N. Such effects might be relatively more in ambient foggy air. Due to high initial concentrations, important in remote areas where ON accounts for a larger - + neither sample had detectable changes in NO3 and NH4 fraction of the total N in the atmosphere (4, 12). during illumination. Unlike the case for DA99-02F, in these The conversion of ON into inorganic N species likely also experiments we calculate that all the NOx collected came influences the physical, chemical, and toxicological properties - from photolysis of NO3 and N(III), probably because the of atmospheric condensed phases. For example, due to the + - initial concentrations of these inorganic chromophores were high water solubility of NH4 and NO3 , formation of these very high. species from organic precursors likely alters the hygroscop- 3.4. Possible Reaction Mechanisms for Conversion of icity and surface tension of atmospheric particles and water Organic N to Inorganic Forms. Although a very incomplete drops. Since NOx is a key participant in the photochemical understanding of DON composition hampers our under- reactions that form ozone and thereby alters the oxidation standing of the mechanisms for the formation of inorganic capacity of the atmosphere, the photochemical formation of N in our experiments, we can make some suggestions for NOx from condensed-phase organic N is especially interesting possible mechanisms. For example, it has been reported that in terms of its potential influence on tropospheric chemistry. + transformations of amino compounds can produce NH4 , In addition, the observed rapid loss of organic nitrogen either through hydrolysis or through reactions with inter- compounds during exposure to light (or light + ozone) might mediates such as hydroxyl radicals (18, 19, 31, 32, 49, 50). profoundly influence the many important properties of Given that amino compounds typically accounted for 20% atmospheric condensed phases that are affected by organic of the organic N in the samples examined in this study (8, compounds (e.g., light absorption, acid-buffering capacity, 9, 22), it is possible that these compounds were a significant and complexation/speciation of transition metals). These + precursor for the photoproduced NH4 . The strong correla- same properties will also be altered by photochemically + tion between the initial formation rates of NH4 and the induced changes that occur in the organic N pool that remains concentrations of amino N in fogwaters (section 3.1) supports after exposure. Although this ON is not mineralized, it likely this hypothesis. In addition to amino compounds, humic undergoes alterations in composition, for example, to form substances are probably also present at significant concen- more oxidized forms of ON, including alkyl nitrates (42). trations in the Davis fogwaters and aerosol samples (37, 51) + and could be significant photochemical sources of NH4 (30- Acknowledgments 32). The authors thank Ingrid George and Jeff Chan for help with - The chemistry and formation of NO3 is intricately linked experiments and Keith McGregor and Mike Jimenez-Cruz - - with reactions of NO2 and NOx. For example, NO3 is both for assistance with fog and aerosol sampling. This work was - a product from oxidation reactions of NO2 and NOx as well supported by the U.S. EPA (R819658 and R825433) Center as a photochemical precursor of these two species (47, 48). for Ecological Health Research at the University of California Nonetheless, our results indicate that the mineralization of at Davis. Although the information in this document has organic nitrogen was primarily responsible for the formation been funded in part by the United States Environmental - - of NO3 and NO2 and could be a significant source of NOx. Protection Agency, it may not necessarily reflect the views - - One possible source of NOx,NO2 , and NO3 in our samples of the Agency, and no official endorsement should be inferred. is nitro compounds (e.g., nitrophenols), which can be Additional funding for this work was provided by a graduate photolyzed in particles and water drops to form NO and NO2 fellowship from the University of California Toxic Substances - - that can be further oxidized into NO2 and NO3 (20, 21, 52, Research and Teaching Program (ecotoxicology component), 53). In addition, Chang and Novakov (54) have suggested by the California Agricultural Experiment Station (Project that atmospheric ON can be formed by reactions of gaseous CA-D*-LAW-6403-RR), and by a Jastro-Shields Graduate NH3 and NOx with organics or soot; if these reactions are Research Award from the University of California at Davis. + reversed in the presence of light they might form NH4 , - NO3 and NOx. Literature Cited 3.5. Environmental Implications. Our results indicate (1) Anastasio, C.; McGregor, K. G. Aerosol Sci. Technol. 2000, 32, that organic nitrogen compounds in atmospheric condensed 106-119.

VOL. 37, NO. 16, 2003 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 9 3529 (2) Keene, W. C.; Montag, J. A.; Maben, J. R.; Southwell, M.; Leonard, (31) Tarr, M. A.; Wang, W. W.; Bianchi, T. S.; Engelhaupt, E. Water. J.; Church, T. M.; Moody, J. L.; Galloway, J. N. Atmos. Environ. Res. 2001, 35, 3688-3696. 2002, 36, 4529-4540. (32) Bushaw, K. L.; Zepp, R. G.; Tarr, M. A.; Schulzjander, D.; (3) Timperley, M. H.; Vigor-Brown, R. J.; Kawashima, M.; Ishigami, Bourbonniere, R. A.; Hodson, R. E.; Miller, W. L.; Bronk, D. A.; M. Can. J. Fish. Aquat. Sci. 1985, 42, 1171-1177. Moran, M. A. Nature 1996, 381, 404-407. (4) Cornell, S.; Rendell, A.; Jickells, T. Nature 1995, 376, 243-246. (33) Grzybowski, W. Chemosphere 2003, 52, 933-936. (5) Neff, J. C.; Holland, E. A.; Dentener, F. J.; McDowell, W. H.; (34) Day, J. G., McLellan, M. R., Eds. Cryopreservation and Freeze- Russell, K. M. Biogeochemistry 2002, 57,99-136. Drying Protocols; Humana Press: Totowa, NJ, 1995; Vol. 38. (6) Jassby, A. D.; Reuter, J. E.; Axler, R. P.; Goldman, C. R.; Hackley, (35) Faust, B. C. Rev. Sci. Instrum. 1993, 64, 577-578. S. H. Water Resour. Res. 1994, 30, 2207-2216. (36) Buttini, P.; Di Palo, V.; Possanzini, M. Sci. Total. Environ. 1987, (7) Weathers, K. C.; Lovett, G. M.; Likens, G. E.; Caraco, N. F. M. 61,59-72. Ecosystems 2000, 3, 590-595. (37) Anastasio, C. Ph.D. Thesis, Duke University, 1994. (8) Zhang, Q.; Anastasio, C. Atmos. Environ. 2001, 35, 5629-5643. (38) Anastasio, C.; McGregor, K. G. Atmos. Environ. 2001, 35, 1079- (9) Zhang, Q.; Anastasio, C.; Jimenez-Cruz, M. J. Geophys. Res. 1089. [Atmos.] 2002, 107, (DOI) 10.1029/2001JD000870. (39) Leifer, A. The Kinetics of Environmental Aquatic Photochem- (10) Zhang, Q.; Carroll, J. J.; Dixon, A. J.; Anastasio, C. Environ. Sci. istry: Theory and Practice; American Chemical Society: Wash- Technol. 2002, 36, 4981-4989. ington, DC, 1988. (11) Cornell, S.; Mace, K.; Coeppicus, S.; Duce, R.; Huebert, B.; Jickells, (40) Miller, J. C.; Miller, J. N. Statistics for Analytical Chemistry, 3rd T.; Zhuang, L. Z. J. Geophys. Res. [Atmos.] 2001, 106, 7973-7983. ed.; Ellis Horwood-Prentice Hall: New York, 1993. (12) Cornell, S.; Jickells, T. D.; Cape, J. N.; Rowland, A. P.; Duce, R. (41) Ruggaber, A.; Dlugi, R.; Bott, A.; Forkel, R.; Herrmann, H.; Jacobi, A. Atmos. Environ. 2003, 37, 2173-2191. H. W. Atmos. Environ. 1997, 31, 3135-3148. (13) Bytnerowicz, A.; Fenn, M. E. Environ. Pollut. 1996, 92, 127- (42) Chuck, A. L.; Turner, S. M.; Liss, P. S. Science 2002, 297, 1151- 146. 1154. (14) Moffat, A. S. Science 1998, 279, 988-989. (43) Lide, D. R. CRC Handbook of Chemistry and Physics, 73rd ed.; (15) Paerl, H. W.; Aguilar, C.; Fogel, M. L. In Atmospheric Deposition CRC Press: Boca Raton, FL, 1992. of Contaminants to the Great Lakes and Coastal Waters; Baker, (44) Arakaki, T.; Miyake, T.; Hirakawa, T.; Sakugawa, H. Environ. Sci. J. E., Ed.; SETAC Press: Pensacola, FL, 1997; pp 415-429. Technol. 1999, 33, 2561-2565. (16) Seitzinger, S. P.; Sanders, R. W. Limnol. Oceanogr. 1999, 44, (45) Reilly, J. E.; Rattigan, O. V.; Moore, K. F.; Judd, C.; Sherman, D. 721-730. E.; Dutkiewicz, V. A.; Kreidenweis, S. M.; Husain, L.; Collett, J. (17) Mopper, K.; Zika, R. G. Nature 1987, 325, 246-249. L., Jr. Atmos. Environ. 2001, 35, 5717-5728. (18) Milne, P. J.; Zika, R. G. J. Atmos. Chem. 1993, 16, 361-398. (46) Ross, A. B.; Mallard, W. G.; Helman, W. P.; Buxton, G. V.; Huie, (19) McGregor, K. G.; Anastasio, C. Atmos. Environ. 2001, 35, 1091- R. E.; Neta, P. NRDL-NIST Solution Kinetics Database-Version 1104. 2, 1994. (20) Larson, R. A.; Jafvert, C. T.; Bosca, F.; Marley, K. A.; Miller, P. (47) Jankowski, J. J.; Kieber, D. J.; Mopper, K. Photochem. Photobiol. L. Environ. Sci. Technol. 2000, 34, 505-508. 1999, 70, 319-328. (21) Feilberg, A.; Nielsen, T. Environ. Sci. Technol. 2000, 34, 789- (48) Fischer, M.; Warneck, P. J. Phys. Chem. 1996, 100, 18749-18756. 797. (49) Monig, J.; Chapman, R.; Asmus, K. D. J. Phys. Chem. 1985, 89, (22) Zhang, Q.; Anastasio, C. Atmos. Environ. 2003, 37, 2247-2258. 3139-3144. (23) Cornell, S. E.; Jickells, T. D.; Thornton, C. A. Atmos. Environ. (50) Fukushima, M.; Tatsumi, K.; Morimoto, K. Environ. Sci. Technol. 1998, 32, 1903-1910. 2000, 34, 2006-2013. (24) Facchini, M. C.; Fuzzi, S.; Zappoli, S.; Andracchio, A.; Gelencser, (51) McGregor, K., G. Masters Thesis, University of California, 2000. A.; Kiss, G.; Krivacsy, Z.; Meszaros, E.; Hansson, H. C.; Alsberg, (52) Dzengel, J.; Theurich, J.; Bahnemann, D. W. Environ. Sci. Technol. T.; Zebuhr, Y. J. Geophys. Res. [-Atmos.] 1999, 104, 26821-26832. 1999, 33, 294-300. (25) Havers, N.; Burba, P.; Lambert, J.; Klockow, D. J. Atmos. Chem. (53) Spanggord, R. J.; Yao, C. D.; Mill, T. Environ. Sci. Technol. 2000, 1998, 29,45-54. 34, 497-504. (26) Havers, N.; Burba, P.; Klockow, D.; Klockow-Beck, A. Chro- (54) Chang, S. G.; Novakov, T. Atmos. Environ. 1975, 9, 495-504. matographia 1998, 47, 619-624. (55) Antia, N. J.; Harrison, P. J.; Oliveira, L. Phycologia 1991, 30, (27) Thurman, E. M. Organic Geochemistry of Natural Waters; Kluwer 1-89. Academic: Boston, 1985. (28) Anastasio, C.; Faust, B. C.; Rao, C. J. Environ. Sci. Technol. 1997, 31, 218-232. Received for review February 7, 2003. Revised manuscript (29) Schulten, H. R.; Schnitzer, M. Biol. Fertil. Soils 1998, 26,1-15. received May 28, 2003. Accepted June 3, 2003. (30) Wang, W. W.; Tarr, M. A.; Bianchi, T. S.; Engelhaupt, E. Aquat. Geochem. 2000, 6, 275-292. ES034114X

3530 9 ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 37, NO. 16, 2003