Volatile organic compound emission rates from mixed deciduous and coniferous forests in Northern Wisconsin, USA

Biogenic emissions of volatile organic compounds (VOC) from forests play an important role in regulating the atmospheric trace gas composition including global tropospheric ozone concentrations. However, more information is needed on VOC emission rates from di ! erent forest regions of the world to understand regional and global impacts and to implement possible mitigation strategies. The mixed deciduous and coniferous forests of northern Wisconsin, USA, were predicted to have signi " cant VOC emission rates because they are comprised of many genera (i.e. Picea , Populus , Quercus , Salix ) known to be high VOC emitters. In July 1993, a study was conducted on the Chequamegon National Forest near Rhinelander, WI, to identify and quantify VOC emitted from major trees, shrubs, and understory herbs in the mixed northern forests of this region. Emission rates were measured at various scales } at the leaf level with cuvettes, the branch level with branch enclosures, the canopy level with a tower based system, and the landscape level with a tethered balloon air sampling system. Area-average emission rates were estimated by scaling, using biomass densities and species composition along transects representative of the study site. Isoprene (C 5 H 8 ) was the primary VOC emitted, although signi " cant quantities of monoterpenes (C 10 H 16 ) were also emitted. The highest emission rates of isoprene (at 30 3 in the Lake States region of the USA are a signi " cant source of reactive VOC to the atmosphere. Accurate estimates of these emissions are required for determining appropriate regulatory air pollution control strategies. Future studies are needed to extrapolate these estimates to other landscapes and to better understand the factors controlling observed variations in VOC emissions. ( 1999 Elsevier Science Ltd. All rights reserved.


Introduction
Tropospheric ozone (O ) concentrations are increasing globally as a result of increasing urban population, fossil fuel burning, and transportation (Firor, 1990), and they will increase further if rising anthropogenic atmospheric pollutants are not abated (Chameides et al., 1994(Chameides et al., , 1997. These increases may greatly in#uence future forest health because O is a widespread phytotoxin known to decrease productivity in crops and forests (MacKenzie and El-Ashry, 1989).
Natural emissions of volatile organic compounds (VOC) such as isoprene from forests are known to play an important role in in#uencing atmospheric chemistry related to regional O concentrations (Lamb et al., 1987;Guenther et al., 1994Guenther et al., , 1996a. However, more information on VOC identities and emission rates from trees and forests is needed to estimate regional biogenic VOC concentrations (Geron et al., 1994). Such information will help us to better understand atmosphere-biosphere interactions (Taylor et al., 1994) and global change scenarios, and to implement possible mitigation strategies .
The mixed deciduous and coniferous forests of northern Wisconsin are comprised of numerous genera which exhibit high VOC emission rates  including Picea (5% of basal area), Populus(14%), Quercus (3%), and Salix ((1%) (Hansen, 1984). These forests also contain a number of important plant genera that had not been previously studied for VOC emissions such as Alnus, ¹ilia, Rubus, and Sphagnum. In addition, the arboretum of the Forestry Sciences Laboratory, Rhinelander, WI contains numerous genetic sources of high isoprene emitting genera such as Picea, Populus, and Quercus planted in common garden experiments, providing access for the "eld study of genetic variation of isoprene emission rates reported to occur in tree genera (Tingey et al., 1991).
The objectives of our study were to: (1) survey VOC emission rates of important northern Wisconsin forest species in the "eld across di!erent scales (i.e. leaf, branch, canopy, and landscape) with qualitative and quantitative methods, and (2) quantify genetic variation of isoprene emission rates among selected genotypes of Picea and Populus not previously studied.

Materials and methods
In July 1993, measurements of VOC emissions were made in common garden plantations of Picea and Populus in the arboretum at the Forestry Sciences Laboratory, Rhinelander, WI, and on mixed deciduous and conifer forests of the Chequamegon National Forest nearby. Measurements were made at leaf, branch, canopy, and landscape scales. Some measurements were made on-site with portable equipment, and other samples were collected in stainless-steel canisters and subsequently analyzed in the laboratory. Details of methods are given by Guenther et al. (1996a, b). Preliminary isoprene emission measurements were made on selected poplar clones and Quercus rubra trees, both known high isoprene emitters , to assess possible e!ects of leaf position and temperature on isoprene emission in the "eld for the purposes of establishing sampling protocol. Position and temperature e!ects on isoprene emission rates have been reported for other species (Guenther et al., 1991(Guenther et al., , 1996aKuzma and Fall, 1993) and for Populus and Quercus by Sharkey et al. (1991b).

Enclosure yux systems
To assess isoprene emission rates at the leaf and branch level for a variety of species, three di!erent sampling schemes were used: two for assessing isoprene #uxes from individual leaves, and one for measuring branch level emissions. Emission rates for the poplar clones were measured with an open-path gas exchange system (MPH-1000, Campbell Scienti"c, Logan, UT) consisting of a temperature-controlled, fan-stirred cuvette, with an internal measurement and control system. Arti"cial light was also provided. Branches were cut from trees and then recut immediately under water before measurement. Details are provided by Harley et al. (1996).
A second leaf cuvette system, used for all leaf measurements except those on Populus, provided no control over leaf temperature. A portion of a leaf was enclosed in the 0.25 l cuvette of a commercially available portable photosynthesis system (Model 6200, Li-Cor, Inc., Lincoln, NE), operated in the open, #ow-through mode. Leaf temperature and incident photosynthetically active radiation (PAR) were measured using the thermocouple and PAR sensor supplied with the system. Immediately after isoprene was determined, the system was converted to the closed, recirculating mode, and CO uptake and water loss were measured. For both leaf cuvette systems, that portion of a leaf inside the cuvette was excised and its area measured with a leaf area meter (Model 3000A, Li-Cor, Inc, Lincoln, NE). All leaves were then ovendried (603C) for 48 h and weighed.
Isoprene concentration in the air entering and leaving both leaf cuvette systems was measured using a chemiluminescence detector (Hills and Zimmerman, 1990) which was calibrated hourly against a known isoprene standard.
Branch level emission rates of Picea were measured with a 24 l #ow-through branch enclosure consisting of 5 mm Te#on "lm over a stainless-steel support frame. Ambient air was pumped through the enclosure at a known rate of approximately 9 l min\. Leaf temperature and air temperature inside the enclosure were measured with shielded thermistors (YSI, Yellow Springs, OH), and incident PAR was measured with a Li-190SA quantum sensor (Li-Cor, Lincoln, NE) mounted adjacent to the enclosure. Replicate samples of air entering and leaving the enclosures were collected with 10 ml glass syringes, and the isoprene concentration was determined using gas chromatography with a reduction gas detector (RGD2; Trace Analytical, Menlo Park, CA). Details of this analytical system may be found in Greenberg et al. (1993). This system was also calibrated several times daily against a 71 ppbv isoprene standard. All leaves within each branch enclosure were removed, oven-dried at 603C for 48 h, and weighed.

Above canopy yux systems
For the canopy and landscape level portions of the study on the Chequamegon National Forest, estimation procedures were used for scaling isoprene emission rates from the leaf level to a ground area basis. These estimates require detailed information on species composition and leaf biomass . Speci"cally, two sites were chosen for study*one was a bog site dominated by Picea mariana,¸arix laricina, and Sphagnum, and the other a mixed deciduous forest dominated by Acer saccharum, Fraxinus americana, and Quercus rubra. On both sites, transects were used to characterize the composition, successional status, and environment of vegetation communities. For a general overall estimate of the composition of the major vegetation communities of the surrounding forests (50 km) on the Chequamegon, the Eastwide forest inventory data base was used (Hansen et al., 1992). Geron et al. (1994) have successfully used this approach for estimating regional VOC emissions.
Vertical gradients of isoprene concentration were measured in the surface layer above the canopy at the bog site on the Chequamegon National Forest. An 8 m tower provided access above the canopy. Isoprene #uxes (F) in the surface layer were estimated as follows: where dz is the di!erence in height, dC is the di!erence in isoprene concentration at di!erent heights, and K is eddy di!usivity (m\ s\). Estimates of eddy di!usivity were calculated using the Bowen ratio-energy balance technique (Guenther et al., 1996a, b). The Bowen ratio estimates of K were based on data from net radiometers (REBS Q6, Seattle, WA) and two temperature and humidity sensors positioned at the same heights as the air sampling systems. In addition, soil heat #ux was estimated from heat #ow transducers (REBS HFT-3, Seattle, WA) at a soil depth of 10 cm. Wind direction and speed were measured with a prop-vane anemometer (R.M. Young 05305-5, Traverse City, MI). PAR above and within the canopy was measured with quantum sensors (Li-190SA, Li-Cor, Inc., Lincoln, NE). Estimates of dC were obtained from ambient air samples pumped into Te#on bags during a 30 min sampling period at heights of 2.4 and 9 m above the mean canopy height. Eight vertical gradient replicates were measured. Samples were transferred into stainless steel canisters and stored for up to 3 weeks. Then they were analyzed for isoprene and monoterpenes by gas chromatography with cryogenic preconcentration using a #ame ionization detector (GC-FID) to quantify concentrations (Greenberg and Zimmerman, 1984).
Isoprene #uxes, F, were estimated from mixed layer measurements over the mixed hardwood site as follows: where z G is the height (m) of the mixed layer,¸is the estimated isoprene loss rate (s\) due to oxidation by OH and ozone, and C is the mixed layer average isoprene concentration (mg(C) m\). The calculated #ux is typically representative of the region within 10}15 km upwind of the sampling site. The assumptions used to develop Eq. (2) are described by Guenther et al. (1996a, b). Guenther et al. (1996c) have shown that midday (1100}1400) isoprene concentrations in the mixed layer change very slowly with time. The isoprene emission rate increases considerably during this time but is o!set by the increasing mixed layer height so that isoprene concentrations are nearly steady state. The estimates of z G were obtained from a LORAN atmospheric sounding system that used airsondes to measure temperature and humidity pro"les up to 5 km above the ground. The mixed layer height was identi"ed by an inversion layer that appears as a region of increasing temperature with height. To estimate the lifetime of isoprene,¸, we used the OH and ozone reaction rate coe$cients reported by Atkinson (1990) and an estimated midday ozone concentration of 40 ppb (variations in O concentration result in relatively small changes in¸) and OH concentration of 4;10 molecules cm\. Uncertainties in¸are about $50% and comprise the largest component of the overall uncertainty in the #ux estimate (Guenther et al., 1996a, b).
Our estimate of the mean isoprene concentration, C, was determined from ambient air samples collected in Te#on bags, using timer controlled, whole air samplers attached to the tether line of a helium-"lled balloon. Samples were collected simultaneously over a 30 min sampling period at 2}4 heights between 10 m and 1 km above ground level. Fourteen balloon #ights were made, but at times heavy winds made sampling di$cult. Samples were transferred into stainless-steel canisters and analyzed for isoprene by the GC-FID method described above.

Emission model
Emission model estimates of isoprene #uxes, F, for the bog and mixed hardwood sites were obtained as follows: where is a landscape average isoprene emission factor, is a temperature and PAR adjustment factor, and D is the total foliar density (g m\) of isoprene emitting species. The foliar density estimation techniques described by Geron et al. (1994) were used to calculate regional estimates of D for comparison with mixed layer measurements. Estimates of D around the tower, used for comparison with the surface layer measurements, were based on the "eld measurements described above. Temperature and PAR were measured at the "eld sites, and the algorithms developed by Guenther et al. (1993) were used to estimate . Landscape level estimates of isoprene emission rates from observed measurements at the bog and mixed hardwood sites were then compared to the model estimates for model veri"cation.

VOC compounds
Biogenic VOC are frequently classi"ed into three categories } isoprene, monoterpenes, and others, which contribute about 45, 10 and 45%, respectively, of the total annual global biogenic VOC emission (Guenther et al., 1995). Isoprene is categorized as a secondary metabolite (Fall, 1991) and is the basic chemical unit of monoterpenes and other terpenoid compounds (Lerdau, 1991). Secondary metabolites are known to play a role in allelopathy, thermal protection, chemical defense, attractants for pollination, and as phytopathic agents although no causal role has been found for isoprene per se (Fall, Table 1. Notes are leaf insertion points, starting with Node 1 (youngest leaf ) at tip of shoot. 1991). Secondary metabolites also commonly accumulate with stress and there are indications that some stresses enhance isoprene emissions (Sharkey et al., 1991a).

Leaf position ewects
Isoprene emission rates are known to vary with leaf developmental stage and position (Guenther et al., 1991;Sharkey et al., 1991b;Kuzma and Fall, 1993). Our survey of leaf developmental series of two hybrid poplar clones ( Fig. 1) showed the importance of sampling the same leaf developmental stage in the "eld. In a developing shoot of clone NC5260, isoprene emission rates (measured at 253C and PAR"1000 mol m\ s\) were measured on a series of successively older leaves, as indicated by their nodal position (i.e. leaf insertion point). Counting down from the newest leaf at the branch tip (Node 1), isoprene emission rates increased with leaf developmental age from Node 4 up to Node 8, a recently mature leaf (Fig. 1A). In an older shoot of clone DN5, isoprene emission rates increased with leaf development stage up to Node 8 and then decreased in older leaves (Fig. 1B). Isoprene emission rates of recently mature leaves were high in both poplar clones, peaking at 70 and 45 g(C) g\ h\ in NC5260 and DN5, respectively. Isoprene emission rates and net photosynthetic rates in Table 1 List of Populus clones surveyed for isoprene emission rates in an arboretum at Rhinelander, WI, in July 1993 (see Fig. 3 each of the two poplars were highly correlated as shown previously in Populus and Quercus by Sharkey et al. (1991b). Our results prompted us to select recently mature leaves as our sampling position for the remainder of the "eld study.

Temperature ewects
The e!ects of temperature on isoprene emission rates in plants are well known (Sharkey et al., 1991b;Tingey et al., 1991;Guenther et al., 1991). Isoprene emission rates of recently mature leaves of "eld-grown trees of two hybrid poplar clones (DN34, DN182) and two northern red oak trees (Quercus rubra) growing in the "eld in northern Wisconsin increased dramatically with increasing leaf temperature (Fig. 2). As expected, isoprene emission rates were very high (i.e., poplar' 150 g(C) g\ h\, oak '175 g(C) g\ h\), especially at temperatures above 303C. While isoprene emission rates increased exponentially up to 353C, net photosynthesis rates were #at or declined over the temperature range studied. As a result, the percentage of photosynthetically "xed carbon which is re-emitted as isoprene increases dramatically as temperature increases.

Genetic ewects
Populus. Eleven poplar clones which are widely planted in the North Central region of the USA (Hansen et al., 1994), and whose parentage is given in Table 1, were surveyed to assess genetic variation of isoprene emission rates in Populus. The selected clones represented a range of hybrid parentages among the Aegerios and ¹acamahaca sections of the genus Populus. They included "ve frequently planted P. deltoides x P. nigra clones and six other clones that have performed well in the "eld under short rotation culture (Hansen et al., 1994). Isoprene emission rates measured at PAR of 1000 mol m\ s\ and 253C were all high (clonal means ranging from 46 to 91 g(C) g\ h\; Fig. 3A) with a clonal average of 68 g(C) g\ h\. Using the Fig. 2. Leaf temperature e!ects on isoprene emission rates and net photosynthetic rates in (A) two hybrid poplar clones, DN34 and DN182; and B) two Quercus rubra trees growing in Rhinelander, WI. Open symbols are isoprene emission; "lled symbols are net photosynthesis. Circles in A are DN182; squares are DN34. Poplar clone parentage is given in Table 1. temperature response data shown in Fig. 2A, emission rates at 303C can be predicted using the algorithms of Guenther et al. (1993), in which case the clonal average is 129 g(C) g\ h\, a very high rate. There was no apparent trend in isoprene emission rates associated with the intersectional crosses. Notably three of the most widely planted clones*DN34, DN182, and NE308*had some of the highest isoprene emission rates. We found no signi"cant correlation between isoprene emission rates and net photosynthetic rates for the clones (Fig. 3) as we Fig. 3. Mean values ($S.E.; n"6) for (A) isoprene emission rates, (B) net photosynthesis rates, and (C) percentage of "xed carbon lost as isoprene, for recently mature leaves of 11 hybrid poplar clones growing in an arboretum in Rhinelander, WI in July 1993. Clone parentage is given in Table 1. had expected ; their correlation coe$cient was very low (r"0.04).
VOC emissions can make up a signi"cant portion of the carbon budget from trees (Gower et al., 1995). The percent of carbon (C) lost as isoprene at 253C in these clones ranged from 0.5 to 1.75% with a mean of 0.9% (Fig. 3C), and these values would nearly double at 303C (Harley et al., 1996). This range of percent C lost agrees with the 1}2% of photosynthetically "xed carbon lost as isoprene reported by Sharkey et al. (1991b). Again, three of the most widely planted poplar clones } DN34, DN182, and DN308 } had the highest percent C lost as isoprene. Our results suggest that poplar clones are very high isoprene emitters and that there is signi"cant clonal variation in the emission rates (i.e. two fold in our clones). In our study, speci"c causal genetic e!ects of isoprene emission rates could not be discerned. To detect a causal relationship between isoprene emission rates and heritability will require collaborative studies with molecular geneticists using pedigreed clonal material (Haissig et al., 1992).
Picea. Twelve Picea provenances of wide geographic origin were selected for survey of isoprene emission rates ( Table 2). The provenances included 11 Asian, European, and North American provenances and one hybrid between a European and Chinese seed source produced at Rhinelander, WI (Nienstaedt and Teich, 1971). The spruce provenances had much lower isoprene emission rates, when normalized to 303C and 1000 mol m\ s\, than the poplar clones. The range for branch level emissions was from 0.01 g(C) g\ h\ for Picea abies from Europe to 14.0 g(C) g\ h\ for Picea asperata from China. It should be noted that the hybrid between P. abies and P. asperata had an intermediate isoprene emission rate (2.6 g(C) g\ h\). These "ndings suggest that isoprene emission rates may be subject to genetic manipulation in breeding programs.
The mean emission rate for the spruce species was 4.2 g(C) g\ h\, which is in the &&low'' category for isoprene emission rates. The most common spruces in North America, Picea glauca (white spruce) and Picea mariana (black spruce), had branch level emission rates of 9.5 and 5.8 g(C) g\ h\, respectively. Picea mariana on a landscape scale is an important overall contributor to total VOC emissions from forests in the region because it makes up a large biomass component of the extensive boreal ecosystems in North America (Vitt et al., 1994).

VOC emission rates-leaf and branch level
Most of the uncertainties about existing isoprene emission rate inventories are associated with knowledge gaps in leaf level emission rates and leaf biomass estimates (Fall, 1991;Geron et al., 1994). Isoprene emission rates are classi"ed as &&high'' when they exceed 40 g(C) g\ h\, &&medium'' when they are between 10 and 40 g(C) g\ h\, &&low'' when they are between 1 and 10 g(C) g\ h\, and &&negligible'' when less than 1 g(C) g\ h\ (Guenther et al., 1996a). Monoterpene emission rates are classi"ed as &&high'' when they exceed 3 g(C) g\ h\, &&medium'' when they are between 1 and 3 g(C) g\ h\, &&low'' when they are between 0.2 and 1 g(C) g\ h\, and &&negligible'' when less than 0.2 g(C) g\ h\. Summaries of isoprene and monoterpene emissions from trees and other plants in the forest are given in Tables 3}5. Emissions of other VOC from this Wisconsin-based study are summarized by Helmig et al. (1998).  Table 3 Qualitative evaluation of isoprene emission rates of tree species from northern Wisconsin forests screened in the "eld near Rhinelander, WI in July 1993. Emission rates are compared to data base previously measured in other regions . Symbols: && "''"similar to data base; && ''"lower than data base; &&!''"higher than data base; && !''"not measured Previously Comparison to existing database measured species Isoprene Monoterpene emission rates emission rates Abies balsamea " Acer rubrum " Acer saccharum " Betula allegheniensis " Betula papynifera " " Cornus alternifolia " Fraxinus americana " " Fraxinus nigra " " Picea mariana " Pinus resinosa " Pinus strobus " " Populus grandidentata " ! Populus tremuloides " " Prunus pennsylvanica

Field surveys
In the mixed deciduous and coniferous forests of the Chequamegon National Forest, we conducted a qualitative evaluation of isoprene emission rates of 21 tree species that had been previously measured in other regions (Guenther et al., , 1996a. Our qualitative evaluation, done to validate the present data base, involved Table 4 Isoprene emission rates of previously unstudied plant species from northern Wisconsin forests screened in the "eld in  Table 5 Isoprene emission rates (mean$standard error) for known isoprene emitting species growing in the "eld at Rhinelander, WI, in July 1993. Emission rates normalized to 303C and 1000 mol m\ s\ PAR (Guenther et al., 1981). n"number of measurements screening two plants per species. Our results on isoprene and monoterpene emission rates in northern Wisconsin are compared to the existing database in Table 3. All species surveyed were within the range of expected isoprene emission rates, except for the Salix (willow) species where limited previous data were available. Three of the willow species were higher emitters than expected. As expected, our qualitative study showed that Picea, Populus, Quercus,and Salix were high emitters. These results prompted us to single out these species for further quantitative study of isoprene emission rates (reported below). The estimates of monoterpene emission rates were less consistent with the data base. For example, Abies, Acer, Betula, Cornus, Pinus, and Quercus species had lower emission rates than expected, while Populus, Prunus, ¹huja, and ¹suga species had higher monoterpene emissions than expected (Table 3). The inconsistencies with the monoterpene emission rates are probably related to the di$culties in "eld methodology. However, our results suggest that the aforementioned species deserve further study.
We also surveyed a number of plant species from the forest that have not previously been studied for isoprene and monoterpene emission rates (Table 4). Isoprene emission rates for all the species were negligible except for Sphagnum, which was within the &&low'' category (i.e. 1}10 g(C) g\ h\). Even though Sphagnum is a low emitter, it makes up a large portion of the overall biomass (Vitt et al., 1994) of the over 500 million hectares of boreal forest in North America. From a global perspective, it could be a signi"cant contributor to overall biogenic VOC emissions and deserves further study. Monoterpene emission rates ranged in these species from &&negligible'' to &&high''. Amelanchier, Ostrya, and ¹ilia had &&negligible'' rates; Sphagnum was &&low''; Alnus and¸arixwere &&medium'' emitters; and Chamaedaphne,¸edum, and Rubus were &&high'' emitters (Table 4). These "ndings are signi"cant from the standpoint of total regional VOC emissions in northern Wisconsin because the &&medium'' and &&high'' emitters make up a very large portion of the biomass in northern Wisconsin forests (Hansen, 1984) and other northern North American forest ecosystems (Whitney, 1986;Vitt et al., 1994). Chamaedaphne (leather leaf),¸edum (Laborador tea), and Rubus spp (raspberry) are important bog and understory plant species in northern Wisconsin, and Alnus and¸arix are important riparian and bog tree species in the region.
Based upon the results of the qualitative survey summarized in Table 3, we made a more intensive quantitative study of the prominent &&high'' isoprene emitters, Picea, Populus, Quercus, and Salix spp., from the forest (Table 5). Quercus rubra had a high isoprene emission rate (112 g(C) g\ h\) as found in other regions of the USA (Sharkey et al., 1991b;Geron et al., 1994;Guenther et al., 1994). Other high emitters were Salix petiolaris and Salix dicolor with 102 and 91 g(C) g\ h\, respectively.
The Populus species were also &&very high''. Populus euramericana (DN34), probably the most widely planted clone in the North Central region of the US (Hansen et al., 1994), emitted over 84 g(C) g\ h\of isoprene ( '150 g(C) g\ h\ at 303C); and Populus tremuloides (trembling aspen), which occupies over 15 million hectares of forest land in the northern USA, had isoprene emission rates of over 77 g(C) g\ h\. Aspen with its &&high'' isoprene emission rates is also a very sensitive species to ambient levels of tropospheric ozone (Coleman et al., 1995). The only Picea that we studied intensively was the native American white spruce that emitted isoprene in the &&medium'' category (i.e. 24 g(C) g\ h\). These results con"rmed our hypothesis that numerous tree species comprising our northern Wisconsin forest are &&high'' emitters of isoprene and that these forests are playing an important biogenic role in in#uencing tropospheric chemistry in the region (Geron et al., 1994).

VOC emission rates } landscape level
Landscape estimates of biogenic VOC are needed as inputs into regional models of atmospheric chemistry used by policy makers (Geron et al., 1994;Guenther et al., 1995). However, more information is needed about VOC emission rates for forested regions of the northern Lake States of the US for model inputs and model validations. The results of our surface layer (tower) estimates of isoprene emissions from the black spruce bog on the Chequamegon National Forest and mixed layer (tethered balloon) estimates from the mixed hardwood forest are given in Table 6. The observed mean value of isoprene #ux from the spruce bog was 1.25$0.48 mg(C) m\ h\ compared to a model estimate of 0.82$0.25 mg(C) m\ h\. The model estimate was derived by multiplying the isoprene emission factor for Picea (i.e., 14 g(C) g\ h\) from Guenther et al. (1994) by the measured Picea foliar mass (i.e. 136 g m\) times a temperature and PAR Table 6 Landscape level estimates of isoprene emission rates (mean$standard error) for forest types of the Chequamegon National Forest in northern Wisconsin in July 1993. n"number of replicates  Guenther et al. (1996a) (i.e., 0.43 for 243C at PAR"1300 mol m\ s\, the mean conditions during data collection). The model deviation was 0.43 mg(C) m\ h\, which means the predicted isoprene emission rates were within the limits of the standard error of the observed values. Our observed values were also within the range of 0}2 mg(C) m\ h\ reported by Geron et al. (1994) for the county-wide average of isoprene emission in northern Wisconsin. The observed mean value of the mixed layer estimates of the mixed hardwood site from the 14 tethered balloon #ights was 1.89$0.24 mg(C) m\ h\, which compared to the 1.27$0.10 mg(C) m\ h\model estimate. The model estimate assumed a regional species composition of three &&high'' isoprene emitters } 12.5% Picea, 11.3% Populus, and 2% Quercus from Hansen et al. (1992), and average isoprene emission rates of 14 g(C) g\ h\for Picea and 70 g(C) g\ h\ for Populus and Quercus  with a mean temperature and PAR adjustment factor of 0.39 for 233C and 1150 mol m\ s\ for PAR (Guenther et al., 1996a). Although the model deviation was 0.61 mg(C) m\ h\, both the observed and model estimates agree reasonably well with county-wide estimates for isoprene emission given by Geron et al. (1994). Based upon the results of both our extensive and intensive studies (Tables 2}5), it is not surprising that the observed isoprene emissions were higher than predictions based upon just the high three emitters and the existing emissions database. Based on our measurements, many northern Wisconsin plant species with signi"cant biomass in the tethered balloon footprint emitted isoprene at levels higher than found in the data base, e.g. Quercus rubra (112 g(C) g\ h\), Populus tremuloides (78 g(C) g\ h\), and the Salix spp. (55 g(C) g\ h\). In summary, numerous northern Wisconsin forest plant species emit high concentrations of isoprene and other VOC, and the mixed deciduous and coniferous forests of northern Wisconsin are a signi"cant source of VOC in the atmosphere. The highest isoprene emitters are species of Populus, Quercus, and Salix, with a lesser contribution from Picea. Genetic variation in isoprene emission was observed in Populus and Picea, but molecular genetic studies with pedigree material are needed to determine causal genetic relationships. Landscape level estimates of isoprene emission were between 1 and 2 mg(C) m\ h\, which agrees reasonably well with regional model predictions. Future studies are needed to extrapolate these estimates to other landscapes and to better understand the factors controlling observed variation in VOC emissions from forests.