Eddy covariance measurement of biogenic oxygenated VOC emissions from hay harvesting

Biogenic oxygenated volatile organic compound (VOC) # uxes have been directly measured by eddy covariance using the combination of a fast response, real-time VOC sensor and an acoustic anemometer. VOC detection is based on proton-transfer reaction mass spectrometry which has currently a response time of ca. 0.8s and the system is suitable for making nearly unattended, long-term and continuous measurements of VOC # uxes. The eddy covariance system has a detection limit, for most VOCs, of less than 0.1 mg m ~2 h ~1 . The system was " eld tested above a hay " eld near St. Johann, Austria where cut and drying grasses released a variety of VOCs. High # uxes were observed for about 2 days after cutting and were dominated by methanol (1 } 8.4 mg m ~2 h ~1 ), acetaldehyde (0.5 } 3 mg m ~2 h ~1 ), hexenals (0.1 } 1.5 mg m ~2 h ~1 ) and acetone (0.1 } 1.5 mg m ~2 h ~1 ). The eddy covariance measurements generally agreed with # ux estimates based on enclosure measurements and surface layer gradients. The sensitivity and selectivity of the system make it suitable for quantifying the # uxes of the dominant biogenic VOCs from a variety of landscapes and sources. ( 2000 Published by Elsevier Science Ltd.


Introduction
Volatile organic compounds (VOCs) play an important role in determining the chemical composition of the atmosphere (Fehsenfeld et al., 1992). These compounds are emitted from diverse sources but are primarily of biogenic origin. The processes controlling these emissions must be understood in order to accurately describe their impact on the earth system. Past investigations of biogenic VOC emissions have focused on emissions of isoprene and monoterpenes but it has been suggested that about half of the global total are oxygenated VOCs including alcohols, aldehydes and ketones (Guenther et al., 1995). A major limitation in the investigation of these oxygenated VOCs has been the di$culty in measuring these compounds and, in particular, vertical #uxes in the turbulent boundary layer above a source. Methods that have been used to measure vertical turbulent #uxes, e.g., relaxed eddy accumulation and surface layer gradients, rely on assumptions that are reasonable in some cases but can result in large errors in estimated #uxes (Dabberdt et al., 1993). These methods are typically labor intensive, limiting the number of measurements that can be made, and face technical challenges that often limit the number of compounds that can be quanti"ed (especially oxygenated VOCs). Eddy covariance #ux measurements are directly determined by measuring the covariance of chemical concentration with vertical wind speed. It is the most direct #ux measurement method but has only been used for a selected number of compounds because it imposes two critical constraints on a chemical measurement system: a fast-response time and speci"city for the compound  (200 s) of methanol and acetaldehyde mixing ratios and the vertical wind speed measured at 4.5 m above a hay"eld in Austria and demonstrating the strong correlation between these variables. (b) Normalized methanol and vertical wind velocity (w) spectra and cospectra. whose #ux is being determined. The response time required for this study (with a measurement height of about 4.5 m and unstable conditions with Monin}Obukhov lengths between !3 and !15 m) is less than 1 s. These two constraints have limited the application of eddy covariance to about a dozen trace gases in total and only a few non-methane VOCs, isoprene (Guenther and Hills, 1998), acetone and formic acid (Shaw et al., 1998).
The application of a fast-response, proton-transfer reaction mass spectrometry (PTR-MS) system greatly expands the range of compounds that can be investigated using eddy covariance #ux measurements. In this paper we describe the capabilities of this system including limitations, and then compare eddy covariance "eld measurements with #ux estimates based on enclosure measurements extrapolation and estimated from surface layer gradients. The "eld test is an investigation of emissions from grass hay harvesting which is a potentially important source of oxygenated VOCs in the atmosphere (Kirstine et al., 1998;Fall et al., 1999;de Gouw et al., 1999de Gouw et al., , 2000.

PTR-MS
The measurements were performed using a PTR-MS system which allows for real-time monitoring of the concentrations of VOCs. The system and measuring procedure have been described earlier (Lindinger et al., 1998). The air to be investigated was sampled with a pumping speed of 0.5 l min\ through a Te#on line of 1 mm inner diameter, from which a small fraction of ca. 20 STP cm min\ was led into the PTR-MS system. The response time of the system is ca. 0.8 s.

Micrometeorological measurements
Eddy covariance and surface layer gradient systems were deployed near the center of a managed hay pasture near St. Johann (latitude 47.53N, longitude 12.43E), in Tyrol, Austria. The pasture grasses were dominated by Dactylis glomerata, Poa pratensis, Phleum pratense, Trifolium pratense and Poa trivialis. The sampling site was located near the center of the approximately 3 ha pasture and the upwind fetch was typically about 150 m. The footprint for typical measurement conditions (wind speeds between 1 and 3 m s\ and unstable conditions with Monin}Obukhov lengths between !3 and !15 m) suggest an underestimate of not more than 20% due to the contribution of the "eld further upwind, which was not mowed. The "eld was mowed in the morning of August 21, 1999 and #ux measurements were made from August 21 to 22. Daytime temperatures ranged from about 203C on the "rst day to about 243C on the second day. Gradient measurements were made at heights of 0.7 and 4.5 m above ground level and the eddy covariance system was at a height of 4.5 m. Estimates of eddy di!usivity used for gradient #ux estimates were based on the measured temperature gradient, sensible heat #ux and an iterative calculation procedure similar to that described by Fuentes et al. (1996).

Enclosure measurements
A dynamic (#ow-through) enclosure system was used to characterize two of the dominant grasses in the hay-"eld, Dactylis glomerata and Poa pratensis. The grasses were cut and immediately placed in the enclosure. A constant #ow rate of 135 STP cm min\ was maintained through the 350 ml enclosure and VOC concentrations were monitored over the whole period of drying. Measurements were repeated for drying temperatures of 25, 30 and 353C.

Sensitivity
The PTR-MS instrument noise is primarily high frequency, random noise, and is relatively independent of concentration. At a sampling rate of 10 Hz and average concentrations of 0.5 ppbv, the relative noise level is equivalent to about 40%. This is reduced to 4% at 0.1 Hz (equivalent to a 10 s time period). The high-frequency noise does not signi"cantly in#uence the eddy covariance #ux measurements, which were averaged over 15 min periods.

Response time
The response time was estimated by adding short pulses of air containing acetone to the system inlet and monitoring the respective ion signal in the PTR-MS. The time elapsed during the increase of the ion signal to 95% and the decline to 0% of the maximum ion signal during application of the acetone pulse was measured to be ca. 0.8 s. As it is notoriously di$cult to generate a true step change in concentration in laboratory #ows, we assume that the actual response time of the system is less than 0.8 s. Acetone is a typical non-sticky compound and thus can be taken as a representative of the VOCs investigated in this research, which are non-sticky as well. This would not apply for most of the volatile acids. The response time estimates indicate that this system should be ca-pable of resolving VOC #uctuations at rates as high as 1}2 Hz. Spectra and cospectra of vertical wind velocity and methanol mixing ratios are shown in Fig. 1b; the slopes of the wind}methanol cospectrum and the individual spectra are close to the theoretical slopes, !7/3 and !5/3, respectively. The attenuation due to passing through the sample tubing limits the response time of the system to approximately 8 Hz using a #ow rate of 0.5 l min\ in 450 cm of 0.1 cm diameter tubing. This value was calculated according to Massman (1991).

Selectivity
The compounds measured here appear at the following masses: methanol (M33), acetaldehyde (M45), acetone (M59), pentenols plus 2-methylbutanal (M69), hexenals (M57, M81 and M99), hexenols plus hexanal (M83), and butanone (M73). Details of these assignments are discussed by Fall et al. (1999) and con"rmation was also achieved by recent gas-chromatography}PTR-MS measurements (Karl, 2000). Characterizing and limiting interferences is a signi"cant challenge for a VOC eddy covariance system. An important component of the application of this system is the use of other tools, e.g. gas chromatography, for identifying the compound mix present at a given "eld site. In some cases, the #ux measured by this system will represent the sum of several compounds, for example the total monoterpene emission or the deposition of both methyl vinyl ketone and methacrolein. The resulting information will be useful as long as both deposition and emission #uxes of compounds do not occur with the same resulting mass ion.

Linearity and stability
The linearity of the system in the range of the VOC concentrations measured during the present work has been proven in several earlier investigations (Lindinger et al., 1998). The drift can be eliminated by standardizing the VOC concentrations by the H O> primary ions. Lack of signi"cant drift indicates that long-term unattended operation of the system is feasible.

VOC yuxes from hay harvesting
As illustrated in Fig. 1a, a positive correlation between vertical wind speed and VOC concentrations was observed above the hay"eld after mowing. The spectra and cospectra (Fig. 1b) demonstrate the expected frequency distribution for the vertical wind speed (w) in an unstable surface layer. Table 1 compares the #uxes measured by eddy covariance with estimates based on enclosure measurements and surface layer gradients obtained by using the Monin}Obhukhov similarity approach (Fuentes et al., 1996). The eddy covariance measurements generally agreed with #ux estimates based on enclosure measurements and surface layer gradients. For example, cuvette measurements of the methanol yielded 160 g gdw\. This corresponds to an estimated #ux of 3 mg m\ h\; this #ux compares to #uxes from chamber, surface gradient and eddy correlation ranging from 0.6 to 8.4 mg m\ h\. Summing individual VOC #uxes at midday gives 7}15 mg C m\ h\. The largest contributions to the total #ux were made by methanol (40%) and acetaldehyde (25%); smaller, but signi"cant, #uxes of C5 and C6 leaf wound VOCs were also detected. The highest, transient VOC releases occurred during hay cutting (Table 1) and when the hay was turned as part of the drying process (not shown).

Conclusions
Above-canopy VOC #uxes can be measured using the eddy covariance method with VOC detection based on PTR-MS. The resulting eddy covariance system provides a direct measurement of VOC #uxes and is suitable for long-term, nearly unattended, operation. The lower detection limit ((0.1 mg C m\ h\) makes the system useful for measuring #uxes from a wide range of locations. Care must be taken at locations where there may be downward and upward #uxes of VOCs yielding the same protonated ions by PTR-MS.
Field tests at a mowed hay"eld site demonstrate that the system is capable of accurately estimating VOC #uxes of methanol, acetaldehyde, acetone, pentenols plus 2-methylbutanal, hexenals and hexanal plus hexenols; these are typical VOCs released from wounded vegetation de Gouw et al., 1999de Gouw et al., , 2000. Although the mean temperatures were rather low, around 203C, the total VOC #ux for the entire drying period is estimated to be about 100}300 mg C m\. These results indicate that hay mowing can be a signi"cant source of VOC emission into the atmosphere, and should be included in VOC emission inventories.