Volatile organic compounds and isoprene oxidation products at a temperate deciduous forest

. Biogenic volatile organic compounds (BVOCs) and their role in atmospheric oxidant formation were investigated at a forest site near Oak Ridge, Tennessee, as part of the Nashville Southern Oxidants Study (SOS) in July 1995. Of 98 VOCs detected, a major fraction were anthropogenic VOCs such as chlorofluorocarbons (CFCs), alkanes, alkenes and aromatic compounds. Isoprene was the dominant BVOC during daytime. Primary products from BVOC oxidation were methylvinylketone, methacrolein and 3-methylfuran. Other compounds studied include the BVOCs a-pinene, camphene,/(cid:127)-pinene, p-cymene, limonene and cis-3-hexenyl acetate and a series of light alkanes, aromatic hydrocarbons and seven of the CFCs. The correlation of meteorological parameters, with the mixing ratios of these different compounds, reveals information on atmospheric oxidation processes and transport. Long-lived VOCs show very steady mixing ratio time series. Regionally and anthropogenically emitted VOCs display distinct diurnal cycles with a strong mixing ratio decrease in the morning from the breakup of the nocturnal boundary layer. Nighttime mixing ratio increases of CFCs and anthropogenic VOCs are suspected to derive from emissions within the Knoxville urban area into the shallow nocturnal boundary layer. In contrast, the time series of BVOCs and their oxidation products are determined by a combination of emission control, atmospheric oxidation and deposition, and boundary layer dynamics. Mixing ratio time series data for monoterpenes and cis-3-hexenyl acetate suggest a temporarily emission rate increase during and after heavy rain events. The isoprene oxidation products demonstrate differences in the oxidation pathways during night and day and in their dry and wet deposition observation infers a surface source and release of the shorter-lived


Introduction
The role of biogenic volatile organic compound (BVOC) emissions in atmospheric chemistry processes, in particular in the formation of tropospheric ozone, has been discussed extensively [Lloyd et al., 1983;Lopez et al., 1989;Atherton and Penner, 1990;MacKenzie et al., 1991;Fehsenfeld et al., 1992]. Regions where BVOC emissions have a particularly high impact on air quality are generally characterized by two features: Firstly, a high vegetation biomass density with a significant percentage of high BVOC emitting plant species, such as isoprene-emitting oak trees, and secondly, by sunny and warm climate conditions. BVOC emissions have been shown to increase exponentially with temperature [Guenther et al., 1991[Guenther et al., , 1993. Besides higher emissions in warm climates the formation of tropospheric pollutants such as ozone and other oxidants and aerosols is furthermore fostered by the high photochemical activity.
A problem region where BVOCs are suspected to play a major role in atmospheric processes and which recently has received

Sampling System
Ambient VOCs were analyzed by an in-situ technique, thereby eliminating sample storage time in sample containers or adsorbent cartridges and sample handling steps. This procedure also minimizes problems associated with storing labile compounds. Air was continuously pulled from the top of the sampling tower through an impregnated inlet filter for particulate and ozone removal. The scrubbing of ozone is essential to prevent potential losses of unsaturated compounds during the sampling step [Helrnig, 1997]. The filter material used was Teflon-impregnated glass fiber (Pallflex T60A20, Putnam, Connecticut) which had been soaked in a 5% solution of sodium thiosulfate and then dried for 3 hours in an oven at 120øC. Ozone scrubbing efficiency of this filter was > 95% and was checked regularly. Air was pulled through the filter and a 0.4 cm inner diameter Teflon tubing sampling line at 5 L min -I into the instrument trailer. Total length of the sampling line to the analytical instrument was approximately 12 m. Part of the sampling flow (2.5 L (for routine monitoring) to 10 L (for MS identification) sample volumes at 250 mL min '• sampling rate) was directed onto a temperature-controlled multistage solid adsorbent trap for enrichment of VOCs. The analytical range covered VOCs in the volatility range of approximately C3 to C•5. The adsorbent trap was filled with three layers of solid adsorbents arranged in order of increasing adsorbent strength (300 mg Carbotrap C, 200 mg Carbotrap, 100 mg Carbosieve SIII (all adsorbents from Supelco, Bellefonte, Pennsylvania). This adsorbent combination has been extensively tested and used in

Chemical Analysis
The concentrated VOCs were analyzed by thermal desorption and cryogenic freezeout with a custom-made inlet system [Helmig, 1996]. Schematics of the major components of this system have been given previously [Helmig and Greenberg, 1994; H98]. The whole analytical system was fully automated and computer controlled. The adsorbent trap temperature during the concentration step was • 30ø-35øC. After sampling, the adsorbent trap was moderately heated to 45ø-50øC and purged with 1 L dry helium in the sampling flow direction to remove atmospheric water concurrently enriched on the adsorbent [Helmig and Vierling, 1995;McClenny et al., 1995]. The adsorbent trap was then heated at a rate of 50øC min 'l to 250øC, and thermally desorbed VOCs were backflushed with 25 mL min -•. He onto a cryogenic freeze-out trap (10 cm of 0.125 outer diameter stainless steel tubing filled with 80/100 mesh glass beads) which was temperature controlled at -175øC by blowing liquid nitrogen blow-off onto the outside wall of the stainless steel tubing. Prior to the transfer of the sample VOCs, deuterated benzene (• 25 ng) was added to the freeze-out trap as an internal standard by filling and injecting a sample loop with a gas-phase standard from a compressed gas cylinder. For GC injection the freeze-out trap was rapidly heated to 80øC by flowing an electrical current directly through the stainless steel tubing and VOCs were backflushed onto a 30 m by 0.32 mm inner diameter, 1 •tm film thickness DB-1 column (J&W, Folsom, California). Separation and detection was achieved by temperatureprogramed gas chromatography with mass spectrometric (GC/MS) detection (Hewlett Packard 5890/5972). The MS was operated in the scan mode (m/z = 45-250 until July 17, and m/z = 33-250 from July 17 on) with electron impact ionization (70 eV). During the injection time the GC column was held at -60øC for 3 min and then programed to 150øC at a rate of 6øC min '•. Between runs the GC column and adsorbent trap were baked at 250øC and 3 00øC, respectively.

Calibration
Two pressurized gas standards were used to establish response factors for quantitative measurements. (1) A n-butane/benzene (10.53/10.30 ppb) standard from the National Institute of Standards and Technology (Boulder, Colorado) and (2) a multicomponent standard containing isoprene (8.02 ppb), neohexane (5.99), c•-pinene (4.63), 13-pinene (3.58), and limonene (3.07) (H. Westberg, Washington State University, Washington). The concentrations for this standard were determined at the site with a calibrated GC/atomic emission detector (AED) instrument using the same sample introduction technique as used for the GC/MS. Standards were analyzed by filling Teflon bags with about 5 L calibration gas and connecting the bag to the adsorbent trap inlet. In previous experiments we have found precision and accuracy to be better than • 5% for this procedure. For the multicomponent standard a comparison of the  Chromatographic peak areas were initially normalized by dividing through the peak area count of the internal standard and the sample volume. The data were then multiplied by the response factor from the calibration measurements. In cases where no calibration standard for the respective compound was available the average of all data was normalized to 1, and mixing ratio time series are reported as relative deviations from the mean. For the reasons detailed below, it became evident that the atmospheric levels of TCM deviated less than the readings for the internal standard.
Therefore all data were recalculated and normalized to ambient TCM levels rather than solely to the deuterated benzene signal.

Results and Discussion
A list of the identified compounds is given in Table 1. This table includes the compound retention index (RI), mass spectral fragmentation data (average of approximately six scans around peak maxima after background subtraction and normalization to the base peak), and literature reference data for compound confirmation. A total of 98 VOCs were detected, structural identification was accomplished for 89 compounds (approximately 98% of the overall peak area), nine compounds remained unidentified.
In a few instances, structural identification could not be achieved; however, characteristics of certain compound classes were evident. In those cases, tentative identifications are given.
A typical chromatogram obtained for an ambient air sample collected during noontime on July 17 is shown in Figure 2. An enlargement of a chromatogram section showing the monoterpene pattern is given in Figure 3.
The record of meteorological data (temperature, wind direction, wind speed, relative humidity, and rainfall (K. Birdwell, Oak Ridge National Laboratory, Oak Ridge, Tennessee), and ozone data during the study period from July 17 to 22 is illustrated in Figures 4a and 4b. Figure 5 shows the peak area counts for the internal standard over the 6 day measurement period. The sensitivity gain on July 17 resulted from a change of the MS tune parameters. An approximately 15% sensitivity decline over the last 4 days is observed. Relative standard deviations for the internal standard on the individual days were 6.6, 3.4, 4.8, 3.1, and 5.8%, respectively. A systematic decrease in the internal standard area counts during midday becomes evident during all days. We attribute this effect to the change in ambient temperature inside the trailer. Even though the trailer was air-conditioned, during daytimes, with three GC ovens operated inside the trailer and full sunlight exposure on the outside walls, temperatures inside the trailer increased from about 20øC to 35øC. The internal standard injection loop was neither temperature nor pressure controlled, and therefore these parameters varied with the conditions inside the trailer leading to lower internal standard injection amounts at the elevated temperatures. Since VOCs collected from ambient air were not   Figure   10. Another BVOC identified in numerous samples is cis-3-hexenyl acetate. Figure 11 shows the chromatogram section with the MS peak identification. The normalized relative mixing ratio time series of this compound is plotted in Figure 12.
The list of compound identifications (Table 1) and the sample chromatogram ( Figure 2) show typical characteristics for rural continental air with some anthropogenic influence. A major fraction of observed VOCs are anthropogenically emitted gases, such as the CFCs, alkanes, alkenes, and aromatic compounds. Identified BVOCs emitted from the forest vegetation include isoprene, alkenes, monoterpenes, and one oxygenated compound. Isoprene was the dominant VOC present in daytime air samples. A total of 11 terpenoid compounds were detected. The major monoterpenes were tricyclene, ot-pinene, camphene, 13-pinene, and limonene. Another abundant BVOC was p-cymene. Other monoterpenes identified include sabinene, 3-carene, transocimene, and ¾-terpinene. ß The major primary products from the oxidation of BVOCs identified were the isoprene products MACR, 3-methylfuran, and MVK. Other observed compounds that are potential oxidation products of both biogenic and anthropogenic VOCs include acetone and acetaldehyde.

Anthropogenic VOCs
Another group of anthropogenic VOCs that was looked at closely are light hydrocarbons and aromatic compounds. All three mixing ratio plots in Figure 8 show similar characteristics. Concentrations increase during night with maxima being reached in the early morning hours before sunrise. After sunrise a rapid concentration drop occurs. Mixing ratios decline to • 1/3 to 1/4 of their nighttime maxima within less than 3 to 5 hours. Similar This yields additional evidence that during these hours, more "aged" air from higher tropospheric layers is mixed into the surface layer.

Biogenic VOCs
Isoprene mixing ratios show the highest variability of all compounds analyzed (Figure 9) Pentane 1   ,  ,  I  ,  ,  ,  I  ,  ,  ,  I  ,  ,  ,  I  ,  ,  ,  I ,  ,  ,  I  ,  ,  ,  I  ,  ,  ,  I  ,  ,  ,  I  ,  ,  ,  I  Another VOC identified that can be assigned to biogenic sources is cis-3-hexenyl acetate (Figures 11 and 12) It should be noted that to date we have not experimentally proven quantitative recovery of this compound for our analytical system because of the lack of a stable quantitative standard. The data presented may therefore underestimate the overall abundance. Interestingly, the highest ambient levels of cis-3-hexenyl acetate were observed in a sample collected on July 17 (Julian day 198) about 3.5 hours after a thunder shower. Ambient levels of cis-3-hexenyl acetate more than doubled from before to after the rain. There are two possible explanations for this observation. Firsfly, after a heavy rain event a substantial decrease of the vertical mixing from the adherent surface cooling is expected. This lower vertical mixing will cause more stagnant air conditions and lead to enhanced ambient air BVOC concentrations near the surface. However, the concurrent temperature drop following the rainstorm (approximately 7ø-8øC, Figure 4) will lead to a decrease of the BVOC emission rates which will counteract the mixing effect. Using the temperature response algorithms developed by Guenther et al. [ 199 l, 1993], calculated decreases in the emission rates for isoprene and monoterpenes from this temperature change are 56% and 46%, respectively (this does not consider any emission rate changes, in particular for isoprene, as a response to the adherent light intensity changes in this kind of situation). The ambient isoprene mixing ratio increased 28% immediately after the first major rain shower and showed a steep decline afterward (from 8 to 2 ppb, Figure 9). Mixing ratio increases of monoterpenes are delayed by about 6 hours compared to isoprene and cis-3-hexenyl acetate. Assuming similar ground sources, wet deposition rates and atmospheric depletion rates for monoterpenes and cis-3-hexenyl acetate, the comparison of the mixing ratio time series infers that fluxes of cis-3-hexenyl acetate must have increased significantly more overall and also more rapidly than for the other BVOCs during this period. Therefore it appears plausible that physical stress caused by the heavy rainstorm must have triggered an instantaneous increase in the cis-3-hexenyl acetate emission rates.

Propane --.• I•Bt•tane --.-n-Butane --• -n-
The data are the first indication that the observations from enclosure measurements are at least to some degree referable to ambient conditions, where similar stress situations occur which may lead to enhanced cis-3-hexenyl acetate emissions. More research seems warranted for a further and more accurate study of these phenomena. It has previously been noted that cis-3-hexene-1-ol and cis-3-hexenyl acetate are major constituents in volatiles emitted from grasses and that emission rates may increase from mowing [Schulting et al., 1980;Ohta, 1984;Arey et al., 1991]. Since our measurement site was located within a forest clearing with mostly grassy ground cover, it is also possible that our observations stem from grass emissions rather than from emissions from the surrounding forests. With our currently available information we have no way to distinguish between these potential sources. Grass mowing of the clearing area did not occur during the measurement period.

Isoprene Oxidation Products
Even though isoprene oxidation has been thoroughly studied for many years, measurements on identification and concentrations of its atmospheric oxidation products in ambient air have only been accomplished very recently [Pierotti et al., 1990;Martin et al., 1991;Montzka et al., 1993Montzka et al., , 1995Riemer et al., 1994;Yokouchi, 1994;Goldan et al., 1995;Biesenthal et al., 1997]. Isoprene oxidation products that previously were detected in ambient air include MACR, MVK, and 3-methylfuran. All three of these compounds were identified in this study, and their relative mixing ratio time series are plotted in Figure 10. Despite a fairly high scatter in the data distinct diurnal cycles become evident. Gradual concentration increases occur throughout the day to around midnight followed by a concentration decline to the early morning hours. Previously, it was assumed that no other major primary or secondary sources of any of these compounds in ambient air are of importance. Furthermore, branch enclosure experiments reported in the literature do not indicate that any of these compounds is a major direct emission from vegetation. Howeve'r, a very recent study by Biesenthal et al. [1997] suggests that in the vicinity of large urban areas, anthropogenic sources may contribute to atmospheric levels of MACR Figure 14. The data presented here agree well with the field observations and model simulations by Montzka et al. [1993Montzka et al. [ , 1995 who found similar diurnal patterns and amplitudes in the MACR/MVK ratio.
It is remarkable that in the early morning hours of July 18 (Julian day 199) the MACR/MVK ratio reached a maximum that is about 3 times higher than values reached on the other days. This increase is preceded by a series of heavy rain showers in the late afternoon and evening of July 17 (between 15.00-16.00 and 23.00-24.00 hrs). Figure 15 shows the chromatograms of two samples collected right before (sampling time 12.55 to 13.35 hrs) and after the last of three rain showers (sampling time 06.32 to 06.42 hrs). From this comparison a change in the relative abundance of MACR and MVC becomes clearly visible with MVK demonstrating a much higher mixing ratio decline than MACR during this time. This effect is explainable by differences in the wet deposition rates. We were not able to find literature data on the Henry's constants of either of these two compounds. However, a comparison with data on analogous compounds allows us to estimate differences in the wet deposition rates. MACR Henry's constant is expected to be at the lower end of the range of the Henry's constants for acetaldehyde, propanal, butanal, and acrolein, which are 15, 13, 9, and 7, respectively [Sander, 1997]. MVK is expected to have a Henry's constant similar to acetone and 2-butanone, which are 30 and 20 respectively [Sander, 1997]. Consequently, Henry's constants for MACR and MVK are expected to differ by about a factor of • 2, with MVK being the compound with the higher Henry's constant and therefore having the higher washout ratio and wet deposition rate. This is reflected by the larger decrease in MVK ambient air mixing ratios yielding the steep increase in the MACR/MVC ratio observed on July 17/18. The data from the in-situ monitoring is in agreement with observations from tethered balloon measurements of VOCs conducted during this peri od (G98 ). Atmospheric levels of 3-methylfuran closely follow those of MACR and MVK, suggesting the same sources and formation routes (Figure 10). On most days 3-methylfuran mixing ratios decline sooner from their maximum values in the late evenings than for MACR and MVK. Similar observations were reported by Montzka et al. [ 1995]. Their mean mixing ratio measured was • 50 ppt and • 3-5% of the sum of MACR and MVK. The data presented here is in reasonable qualitative agreement with this  Figure 15. Selected ion current signals of isoprene oxidation products methacrolein, methylvinylketone and 3methylfuran in two samples collected before (upper window) and after (lower window) a series of three rain thunderstorms on July 17. reference with regard to the relative abundance and diurnal cycles.

Conclusion
The solid adsorbent trapping/cryogenic focusing and in situ GC/MS analysis technique applied here appears to be a viable alternative to the cryogenic sampling technique used by' others for the analysis of VOCs, BVOCs, and the biogenic oxidation products MACR, MVK, and 3-methylfuran. This method has advantages such as lower use of cryogen and no need for water and carbon dioxide removal in the sample airflow. A total of 98 VOCs and BVOCs were identified at a forest site near Oak Ridge, Tennessee.
Long-lived compounds, such as CFCs show very steady mixing ratios with little diurnal variations. In contrast, the mixing ratios of the hydrocarbons, BVOCs, and their oxidation products show distinct diurnal patterns. Observed concentration time series of anthropogenic VOCs, CFCs, and BVOCs can be attributed to a number of different factors, such as diurnal emission rate changes from light and temperature dependence of the BVOC emissions, boundary layer dynamics, and to the photochemical depletion. The correlation of the meteorological parameters and mixing ratio time series of anthropogenic compounds, biogenic emissions, and their oxidation products reveals information on the atmospheric oxidation processes and atmospheric transport of these organic compounds under ambient air conditions. At nighttime anthropogenic VOCs are trapped within the shallow nocturnal boundary layer and transported distances of the order of 50-100 km to remote sites without any significant mixing with the overlaying tropospheric air.
Mixing ratio time series of the isoprene oxidation products MACR and MVK confirm previous observations and are in accordance with the current understanding of their atmospheric formation and removal kinetics. Wet deposition of MVK is faster than for MACR and.in qualitative agreement with estimates of their wet deposition rates. Attention should be given to bursts of non-isoprene BVOC emissions, such as monoterpenes and cis-3-hexenyl acetate as a response to physical stress such as rain showers or possibly heavy winds. The data presented here suggest that emission rates may increase by several factors for a period of about half a day as a response to these events.