JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 101, NO. D13, PAGES 18,555-18,567, AUGUST 20, 1996

Isoprene fluxes measured by enclosure, relaxed eddy

accumulation, surface layer gradient, mixed layer gradient, and

mixed layer mass balance techniques

Alex Guenther, William Baugh, Ken Davis, Gary Hampton,

Peter Harley, Lee Klinger, Lee Vierling, and Patrick Zimmerman

Atmospheric Chemistry Division, National Center for Atmospheric Research, Boulder, Colorado
[Hampton is now at Hampton Technologies, Inc. Colorado USA (gh@2HTI.com)]

Eugene Allwine, Steve Dilts, Brian Lamb, and Hal Westberg

Department of Civil and Environmental Engineering, Washington State University, Pullman

Dennis Baldocchi

Atmospheric Turbulence and Diffusion Division, NOAA, Oak Ridge, Tennessee

Chris Geron and Thomas Pierce

U.S. Environmental Protection Agency, Research Triangle Park, North Carolina

Abstract. Isoprene fluxes were estimated using eight different measurement techniques at a

forested site near Oak Ridge, Tennessee, during My and August 1992. Fluxes from individual

leaves and entire branches were estimated with four enclosure systems, including one system that

controls leaf temperature and light. Variations in isoprene emission with changes in light,

temperature, and canopy depth were investigated with leaf enclosure measurements. Representa-

tive emission rates for the dominant vegetation in the region were determined with branch

enclosure measurements. Species from six tree genera had negligible isoprene emissions, while

significant emissions were observed for Queráis, Liquidambar, and Nyssa species. Above-

canopy isoprene fluxes were estimated with surface layer gradients and relaxed eddy accumulation

measurements from a 44-m tower: Midday net emission fluxes from the canopy were typically 3

to 5 mg C m-2 h-l, although net isoprene deposition fluxes of -0.2 to -2 mg C nr2 h-1 were

occasionally observed in early morning and late afternoon. Above-canopy CO2 fluxes estimated

by eddy correlation using either an open path sensor or a closed path sensor agreed within ±5%.

Relaxed eddy accumulation estimates of CO2 fluxes were within 15% of the eddy correlation

estimates. Daytime isoprene mixing ratios in the mixed layer were investigated with a tethered

balloon sampling system and rangè^ from 0.2 to 5 ppbv, averaging 0.8 ppbv. The isoprene

mixing ratios in the mixed layer above the forested landscape were used to esumate isoprene

fluxes of 2 to 8 mg C nr2 h-l with mixed layer gradient and mixed layer mass balance tech-

niques. Total foliar density and dominant tree species composition for an approximately 8100

km2 region were estimated using high-resolution (30 m) satellite data with classifications super-

vised by ground measurements. A biogenic isoprene emission model used to compare flux

measurements, ranging from leaf scale (10 cm2) to landscape scale (102 km2), indicated agree-

ment to within ±25%, the uncertainty associated with these measurement techniques. Existing

biogenic emission models use isoprene emission rate capacities that range from 14.7 to 70 |ig C

g-l h"1 (leaf temperature of 30°C and photosynthetically active radiation of 1000 |imol m-2 s1)

for oak foliage. An isoprene emission rate capacity of 100 u.g C g-l h-l for oaks in this region is

more realistic and is recommended, based on these measurements.

1. Introduction control is difficult because tropospheric ozone production is

Elevated surface ozone concentrations are a persistent the result °f a complex set of photochemical reactions and pre-

pollution problem in many industrialized countries. Ozone cursor emissions> with significant contributions from natural

sources. Natural emissions of volatile organic compounds

' ^°C) ?** ."f1r°P°genic T™0™ 0n * g'°bal SCale'

Recognition of the important role of natural VOC has led to

Copyright 1996 by the American Geophysical Union. ^ incorPOTa«°ri of natural VOC emissions into the oxidant

control strategies developed to combat high-ozone levels.

Paper number 96JD00697. The Biogenic Emissions Inventory System (BEIS) developed

0148-0227/96/96JD-00697S09.00 by Pierce and Waldruff [1991] has been widely used to

18,555

18,556

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

incorporate natural VOC emission estimates into ozone

control strategies in both the United States and Europe. A

more recent version of this model (BEIS2) includes improved

methods for estimating landscape data and canopy

environment [Geron et al, 1994]. BEIS2 also utilizes

updated emission rate capacities [Guenther et al, 1994], and

experimentally verified relationships between emissions and

environmental conditions [Guenther et al., 1993]. The VOC

emission rates estimated by BEIS2 for most landscapes range

from slightly less than BEIS to as much as a factor of 5 higher

[Geron et al., 1995].

Geron et al. [1994] used the BEIS2 model to estimate natu-

ral VOC emissions from the eastern United States and found

that isoprene from oak (Quercus) trees dominates daytime

fluxes. A comprehensive field study of isoprene fluxes from a

forested region with a significant oak component was

conducted to evaluate the BEIS and BEIS2 models and the

uncertainties associated with each model component.

Representative isoprene emission characteristics for individ-

ual tree species were determined, based on isoprene flux meas-

urements from individual leaves and branches. Isoprene fluxes

in the surface layer above the forest canopy were used to

investigate net (emission minus deposition) canopy fluxes and

to evaluate emission models for a region where land cover and

environmental conditions were characterized in detail. The

ability of emission models to estimate fluxes on the scales used

in regional photochemical models (102 km2) was tested with

measurements that characterized the daytime mixed layer. Spe-

cific components of this study are described in detail else-

where [Baldocchi et al, 1995; P. Barley et al, Environmental

controls over isoprene emission from sun and shade leaves in a

deciduous oak canopy, submitted to Tree Physiology, 1996b;

Lamb et al, 1996].

2. Methods

Field experiments were conducted in July and August

1992 within the Walker Branch Watershed located on a U.S.

Department of Energy Reservation (35° 57' 30" N, 84° 17'

15" W, 365 m elevation) near Oak Ridge, Tennessee. Previous

biogeochemical cycling studies at this site and general site

characteristics are given by Johnson and van Hook [1989],

Fluxes were measured within three nested regions (Plate 1).

the surface layer (SL) flux region has a radius of 0.5 km and is

centered on a 44-m walkup tower that provided a platform for

leaf and branch enclosures and for above-canopy flux measure-

ment systems (relaxed eddy accumulation and surface layer

gradients). The mixed layer gradient (MLG) and mixed layer

mass balance (MB) regions are centered on a clearing, about

400 m north of the walkup tower, where a tethered balloon

sampling system was deployed. The MLG region has a 3-km

radius and includes the landscape that influences mixed layer

gradient flux estimates. The MB region has a 14-km radius and

includes the landscape influencing the mixed layer mass bal-

ance flux estimates. Previous studies have demonstrated that

meaningful above-canopy surface layer flux estimates can be

made with instruments on the 44-m tower even though it is

located on a ridge in moderately complex terrain [Baldocchi

and Harley, 1995; Verma et al, 1986]. The meteorological

conditions during the study were typical of summer in the

eastern United States. Maximum photosynthetically active

radiation (PAR) fluxes were over 1700 |imol nr2 s-1. Above-

canopy winds ranged from about 0.1 to 4.6 m s-',

predominately from the southwest (54%) and northeast (22%)

quadrants that correspond to flow along the ridge. Tempera-

tures during sampling periods ranged from 18° to 31 °C (mean =

24.5°C), and relative humidity ranged from 40% to 87% (mean

= 77%). The hot and humid afternoons were often accompa-

nied by thunderstorms.

2.1. Land Cover Characterization

Thirty-one circular 30 m diameter (707 m2) plots were

established along three transects located in representative

vegetation types within the MLG and MB region. Each plot

was sampled for species composition, diameter at breast height

(dbh, -1.5 m), tree height, seedling and sapling count, leaf area

index, and percent cover of the dominant growth forms

(evergreen scrub, deciduous shrub, ericaceous shrub, fprb,

graminoid, vine, lichen, pteridophyte, and moss). Individual

trees were assigned to one of three classes: trees, > 10 cm dbh;

saplings, dbh = 4 to 10 cm and height > 1.5 m; and seedlings,

dbh < 4 cm or height < 1.5 m In addition to the transects, an

area of 3500 m2 within the SL region was sampled for tree spe-

cies composition and diameter at breast height information.

A land cover database for the Oak Ridge region was devel-

oped using a Landsat satellite Thematic Mapper (TM) classifi-

cation image created from multidate imagery and supervised

classification techniques, referred to here as the Thematic

Mapper Land Cover (TMLC) database. The classification pro-

cedures and results will be described in detail elsewhere. The

TMLC database has 30 m spatial resolution and covers an area

of approximately 90 km X 90 km The database contains 11

land cover classes: loblolly pine, mixed pine, regrowing

(young) pine, high oak deciduous, medium oak deciduous,

low oak deciduous, shrubs and grasses, agriculture, water,

bare soil and urban. Four general categories are shown in

Plate 1. Total foliar density and the fraction contributed by

each tree species were estimated for each land cover class from

forest statistics data collected in the 30-m circular plots.

2.2. Flux Measurement Techniques

Isoprene fluxes were estimated using the eight measurement

systems listed in Table 1. The four systems used to estimate

fluxes from individual leaves or branches included a portable

leaf cuvette with environmental control (LEC), portable leaf

cuvettes with no environmental control (LNC), a tower-

mounted branch enclosure for investigating branches at

different canopy depths (BCS) and a tripod-mounted branch

enclosure for investigating lower branches of various tree

species throughout the region (BRS). Two systems were used

to estimate fluxes in the surface layer above the forest canopy:

a relaxed eddy accumulation (REA) system and a system that

estimated fluxes from isoprene gradient profiles and an eddy

diffusivity based on eddy correlation measurements of water

vapor (GPEC). Data collected with a tethered balloon system

were used to estimate landscape scale fluxes using both a

mixed layer gradient (MLG) technique and a mass balance

(MB) technique.

Enclosure methods. Three of the enclosure systems (LEC,

LNC, and BCS) are described in detail by P. Harley et al.

(1996b). The LEC system consisted of an open path gas

exchange system that provided control of temperature, light

intensity, water vapor, and C02 concentration within the

enclosure. The LNC system included three separate leaf

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

18,557

Nashville 250 kin

18,558

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

Table 1. Description of Isoprene Flux Measurement Techniques

Sample

Analytical

Flux Measurement Technique

Abbreviation

Collection

Method

Enclosures (Leaf/Branch Scale: 0.001 to 1 m2)

Leaf, with environmental control

LEC

glass syringe

Leaf, without environmental control

LNC

glass syringe

Branch, canopy survey

BCS

glass syringe

Branch, regional survey

BRS

SS canisters

GC-FID

Surface layer (Canopy Scale: 0.001 to l_km2)

Relaxed eddy accumulation

REA

Teflon bags

GC-FID

Gradient profile, eddy correlation

GPEC

Teflon bags

GC-FID

Mixed layer (Landscape Scale: 10 to 1000 km2)

Mixed-layer gradient

Teflon bags

GC-FID

Mass balance

MB

Teflon bags

GC-FID

Analytical methods include gas chromatography with reduction gas detector (GC-RGD) and gas

chromatography with flame ionization detector (GC-FID)

cuvettes, while the BCS system consisted of a single branch

enclosure. Samples of air exiting the enclosure were collected

into glass syringes and analyzed with the gas Chromatograph

(GC) with reduction gas detector (RGD) described in section

2.3. Isoprene emission rates, E (|Xg C g"1 h"1), for individual

leaves and branches were calculated as

(1)

where/is the flow rate (cubic meters per hour) into the enclo-

sure, C„ is the isoprene concentration (micrograms carbon per

cubic meters) of the outlet airstream, C¡ is the isoprene concen-

tration (micrograms carbon per Cubic meter) of the inlet air-

stream and b is the foliar mass (grams dry weight) within the

enclosure.

The 30-L cylindrical enclosure used for the BRS system

was constructed of Teflon film over a stainless steel support

frame. The enclosure was supported by an external PVC pipe

frame mounted on a camera tripod. Sweep air was supplied

from compressed gas cylinders at flow rates of 10 to 12 L min-1,

measured with a mass flow meter. Ambient humidity and C02

levels were approximated in the sweep air by mixing hydro-

carbon and C02 free air with hydrocarbon free air containing

2% C02 and then passing this air through distilled water.

Ancillary measurements included leaf temperature, relative

humidity, PAR, and C02 concentration. The entire apparatus

was battery powered and mounted on a cart for mobility.

Whole air samples from the enclosure were collected in Suma-

deactivated stainless steel canisters and analyzed by GC with

flame ionization detector (FID) described in section 2.3.

Surface layer methods. Isoprene fluxes in the surface

layer above the forest canopy were measured by relaxed eddy

accumulation and gradient profile methods. The data pre-

sented in this paper are the first measurements of isoprene

fluxes with an REA system. Tracer and gradient profile meth-

ods have previously been used to measure isoprene fluxes

[Lamb et ai, 1985; Lamb et al, 1986]. Previous applications

of the gradient profile method used indirect estimates of the

relationship between scalar fluxes and gradients. Fluxes esti-

mated by the REA and GPEC techniques are representative of

the area within 500 m of the tower (shown as the SL region in

Plate 1).

The 30-min average isoprene concentration gradients and

eddy diffusivity estimates were used to estimate isoprene

fluxes with the GPEC system. Whole air samples were

collected from a 44-m walkup tower at heights of 28.8 m, 32.3

m, and 38 m above ground level (AGL). Concentration gradi-

ents were estimated by a least squares best fit. In some cases,

concentrations were determined only at heights of 28.8 and 38

m. Samples were collected by pushing air with a Teflon dia-

phragm pump (KNF Neuberger, Princeton, N.J.) through 40-m

Teflon lines and into evacuated 15-L Teflon bags located at

the bottom of the tower. Samples were analyzed with the GC-

FID system described in section 2.3. No significant difference

in isoprene concentration was observed when air samples

were pushed directly into Teflon bags placed on the tower at

the28.8-mand 38-m heights instead of sampling through the

Teflon lines. The GPEC isoprene flux estimates assume simi-

larity between water vapor fluxes and isoprene fluxes. A

Lagrangian micrometeorological model was used to

demonstrate that eddy diffusivities based on water vapor pro-

vide better results for isoprene flux estimates than eddy

diffusivities based on CO2, wind speed, or temperature

[Baldocchi et al., 1995]. We expected water vapor to be more

representative of isoprene fluxes since both are unidirectional

gas fluxes, while the C02 flux contains both a strong emission

(from soils) and deposition (into the canopy) component.

Eddy diffusivity estimates for water vapor, K (m2 s_I), were

calculated from eddy covariance measurements of water vapor

fluxes, w'q', and the measured water vapor gradient, A<j/Az,

using the gradient flux relationship Khqlhz=w'q'.

Eddy covariance C02 and water fluxes were measured with

two systems. One system used a three-dimensional sonic

anemometer (Applied Technology SWS/3K, Boulder,

Colorado) and a custom-made open path infrared gas analyzer

(IRGA) [Baldocchi and Harley, 1995] and is referred to here

as the EC-0 system. The second system used a three-dimen-

sional sonic anemometer (Applied Technology SWS/3K,

Boulder, Colorado) with an analog signal digitizer and a

closed path IRGA (LI-COR 6262, Lincoln, Nebraska) and is

referred to here as the EC-C system. Isoprene fluxes, F, were

then estimated from K and the measured isoprene gradient as

F = KÍ¡.C¡lhz.

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

18,559

The relaxed eddy accumulation system consisted of a three-

dimensional sonic anemometer (Applied Technology, Inc.,

SWS/3K, Boulder, Colorado), a Teflon diaphragm pump (KNF

Neuberger, Princeton, N.J.), three-port Teflon isolatch valves

(General Valve Co., Fairfield, N.J.), a computer and Gontrol

program, and a custom-designed pump control and valve con-

trol board. The pump control system provided a constant air-

flow through the REA system. Air was pulled into the pump

through Teflon tubing (3 m length, 0.318 cm diameter) and

immediately directed through valves into either a vent line, or

one of two Teflon lines (40 m length, 0.635 cm diameter) lead-

ing to two 15-L evacuated Teflon bags located at the base of

the tower. The three-port valves were switched at a maximum

rate of 10 Hz to direct samples into the appropriate Teflon bag.

The time required for input air to reach the three-port valves

was matched to the processing time of the sonic anemometer,

computer, and valve control board. After an approximately

half-hour (1638.4 s) sampling period, air samples in the Teflon

bags were analyzed with the GC-FID system described in sec-

tion 2.3.

The REA system was deployed at a height of 30.5 -m AGL

on the walkup tower. Isoprene and C02 fluxes were estimated

using the relationship described by Businger and Onclev

[1990]

(2)

where ß is a nondimensional coefficient and C„ and Q are the

mean concentrations associated with updrafts and downdrafts,

respectively. A 4-min running mean for vertical wind speed,

w0 (meter per second), and standard deviation of vertical wind

velocity, aw (meter per second), were calculated in real time

and used to calculate the vertical wind speed threshold (wr).

This value was used to separate whole air samples into updraft

(w' >wT), downdraft (w' < -wT) and near zero {-wT < w" <wT)

components, where w' is the difference between the instanta-

neous vertical wind speed and w0. The threshold velocity, wT

= 0.5 aw, was selected to maximize the signal-to-noise ratio of

(Cu - Cd). Oncley et al. [1993] discuss the choice of wT in

some detail and conclude that a value of around 0.6 aw is

optimal.

Estimates of ß and cw were calculated for each 27-min sam-

pling period using wind and temperature data stored in digital

format at 10 Hz. Coordinate transformations were performed on

wind velocity data to set the mean vertical wind speed to zero.

Estimates of ß were calculated by assuming similarity between

isoprene and sensible heat fluxes. Eddy covariance, estimates

of sensible heat flux, w'T', and the mean temperatures associ-

ated with updrafts and downdrafts, Tu and Td, were used to

estimate ß by rearranging (2) and substituting w'T' for F T

toiCuaniTdfmCd.

Mixed layer. The tethered balloon profiling system

consists of a commercial helium-filled tethered balloon and

meteorological sounding system (AIR, Boulder, Colorado)

and a custom made whole air sampling unit that attaches to

any point on the tether line and pumps air into 15-L Teflon

bags. Automatic timers were used to collect 30-min samples

simultaneously at two to four heights between 50 and 800 m

AGL. Whole air samples were analyzed for isoprene by the

GC-FID method described in section 2.3. Isoprene fluxes were

calculated using the mixed layer mass balance (MB) and mixed

layer gradient (MLG) techniques.

The MB method assumptions and associated uncertainties

are discussed by Guenther et al. [1996]. MB fluxes are calcu-

lated as

F=ZiCmt-l (3)

where zt is the mixed layer height (m AGL), x is the estimated

lifetime of isoprene (s), and Cm (milligrams carbon per cubic

meter) is the mean mixed layer isoprene concentration. Esti-

mates of z; were obtained using airsondes (AIR, Boulder,

Colorado) that measure temperature and humidity profiles up

to heights of 5 km AGL. The mixed layer height was identified

by an inversion layer that appears as a region of increasing

potential temperature with height. To estimate the lifetime of

isoprene, T, we used the OH and ozone reaction rate coeffi-

cients reported by [Atkinson, 1990], the measured ozone con-

centration, the OH diurnal variation described by Lu and

Khalil [1991] and a maximum OH concentration of 4 x 10«

molecules cm-3 [Guenther et al, 1996]. We obtain Cm by fit-

ting the observations to a specified vertical profile shape

[Guenther et al, 1996] and then computing the vertical aver-

age of this profile. Fluxes estimated by the MB technique are

representative of an area about 14 km upwind of the measure-

ment site (shown as the MB region in Plate 1).

Fluxes estimated by the MLG technique are dependent on

z¡ and the convective velocity scale, w,. A major advantage of

this technique is that it does not depend directly on estimates

of OH concentrations. The MLG equations and assumptions

are described in detail elsewhere [X. Davis and D. Lenschow,

Scalar profiles and fluxes in the mixed layer, submitted to

Boundary Layer Meteorology, 1996; Guenther et al., 1996].

Fluxes estimated by the MLG technique are representative of

an area about 3 km upwind Of the measurement site (shown as

the MLG region in Plate 1).

2.3. Isoprene Analysis

The GC-RGD and GC-FID systems deployed at the field

site were intercalibrated using a compressed gas standard con-

taining 71-ppbv isoprene referenced to a National Institute of

Standards and Technology (NIST) propane standard on the

GC-FID. Compound identification was accomplished by

retention time comparison with known standards and by GC

with mass spectrometer (MS) analysis of samples transported

in stainless steel canisters to Boulder, Colorado. The isother-

mal GC-RGD system has a 2-mL sample loop, a stainless steel

column (1.3m long x 2mm ID) packed with Unibeads 3S,

60/80 mesh (Allteçh Assoc, Deerfield, Illinois), and an ROD2

(Trace Analytical, Menlo Park, California) detector. This sys-

tem does not require any preconcentration for isoprene mixing

ratios above 1 ppbv and is described in detail elsewhere

[Greenberg et al, 1993].

The Hewlett-Packard 5890 GC-FID system was equipped

with a 30-m DB-1 fused silica capillary column (J&W Scien-

tific). A 100-500 mL whole air sample was preconcentrated on

a cryogenically (liquid 02) cooled' stainless steel loop con-

taining 60-80 mesh silanized glass beads. The concentrated

samples were transferred to the GC column by immersing the

loop in an 80"-90''C water bath. The GC oven was temperature

programmed from -50°C to 80°C at 4°C min1.

2.4. Isoprene Emission Model

Isoprene emission rates, E (¡ig C g-1 h-1), for individual

leaves or branches are modeled as

18,560

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

(4)

where e is an isoprene emission capacity (|i.g C g_1 br1) that

represents the emission rate expected for a particular plant spe-

cies at specified conditions (e.g., sun leaf during peak growing

season, leaf temperature of 30°C and PAR of 1000 |imol nr2 s-

'). The nondimensional emission activity factor, % accounts

for variations in emissions due to changes in leaf temperature

and PAR and is estimated using the equations of Guenther et

al. [1993], Leaf level estimates of 7 can be estimated directly

from leaf level PAR and leaf temperature. Branch level esti-

mates are divided by a factor of 1.75 to account for decreases in

PAR due to self shading [Guenther et al., 1994]. A second

nondimensional emission activity factor, 8, accounts for the

longer-term variations in emissions due to season, phenology,

growth environment, and other factors. All sun leaves and

branches are assigned a value of 8 = 1 while shade leaves are

assigned a value of 8 = 0.8 [Harley et al., 1996a; P. Harley et

al., 1996b].

Model estimates of area-averaged isoprene fluxes (|lg C nr2

h-1) are calculated as

(5a)

where D is total foliar density (g dry weight nr2) and E,y, and 8

are all area-averaged estimates of corresponding leaf level

parameters. The area-averaged e represents the weighted aver-

age of all plant species within the area.

where Doakis me ra'i° °f oak f°uage t0 total foliage. Since

oak trees are estimated to. be responsible for over 95% of all

isoprene emissions in each of the regions shown in Plate 1, we

can simplify the following discussion by multiplying F by a

factor^, equal to the fraction of total isoprene emissions con-

tributed by oaks, and then neglecting isoprene emissions from

vegetation other than oaks:

(6a)

We can compare the results from each measurement technique

by inverting (6a) and solving for e0AK.

(6b)

Total foliar density, D, and the fraction of oaks, DOAK, for the

SL, MB, and MLG regions were estimated using the tech-

niques described in section 2.1. Canopy-averaged Y and 8

were estimated by dividing the canopy into "sun" and

"shade" leaf components and using the leaf level procedures

described above. Estimates of PAR for sun leaves and for

shade leaves were based on measured leaf area index (LAI) and

above-canopy PAR and calculated sun angle using the sun-

fleck radiative transfer model oí Norman [1982]. Relative

humidity, wind speed, and ambient temperature near sun and

. shade leaves were estimated from above-canopy relative

humidity, wind speed, and ambient temperature using vertical

profiles similar to those oí Lamb et al. [1993]. Leaf tempera-

tures for the shade and sun leaves were then calculated from

estimates of total radiation, ambient temperature, wind speed,

and relative humidity using a leaf energy balance model [Lamb

et al, 1993]. To convert between the LAI values used in the

radiative transfer model and the foliar density estimates used

in the emission model, we assume that shade leaves have a

specific leaf weight that is 70% [Geron et al, 1994; P. Harley

et al, 1996b] of the value used for sun leaves.

3. Flux Measurement Results

Isoprene fluxes measured with each of the systems listed in

Table 1 and described in section 2 are reported in this section.

The results are compared arid used to evaluate emission model

procedures in section 4.

3.1. Leaf and Branch Fluxes

Isoprene fluxes from tree species of nine different genera

were investigated with the BRS enclosure system. All of the

sampled trees were growing within the MB region (Plate 1).

Negligible isoprene emission rates were observed for six spe-

cies: Liriodendron tulipifera, Oxydendrum arboreum, Carya

tomentosa, Sassafras albidum, Cornus florida, and Prunus

serótina. Significant isoprène emission rates were observed

for species of Quercus, Nyssa, and Liquidambar. These

results agree with the emission database compiled by

Guenther et al. [1994]. An average emission rate of 1 |j.g C g"1

h-1 (n=10) determined for Nyssa sylvatica was representative

of very low light levels (mean PAR=85 umol nr2 s-1) and a

mean temperature of 24.5°C. Considerably higher mean iso-

prene emission rates were measured for Liquidambar

styraciflua (9.5 (lg g"1 h"1, «=14), Q. alba (14.1 ug C g"1 rr1,

«=16), Q. prinus (29.4 |Xg C g1 h"1, «=15) and Q. velutina

(49.0 ng C g-1 h-1, «=10). The mean PAR for each set of meas-

urements ranged from 180 to 462 |lmol nr2 s-1, while mean leaf

temperatures ranged from 24.2° to 30.0T. The oak emission

capacities listed in Table 2 range from 75 to 114, Hg C g"1 h-'.

These emission capacities are corrected for self shading and are

representative of a leaf temperature of 30°C and PAR of 1000

(imolm-2 j-1.

The LEC, LNC, and BCS systems were used to measure

emissions from a single mature Q. alba tree located adjacent to

the walkup tower. The BCS system averaged a large number of

leaves (15 to 30) with each measurement and was used to com-

pare leaf and branch measurements. Individual leaf emission

rates were estimated for a large number of leaves with the LNC

system. The LEC system was used to investigate emission rate

variations associated with changes in PAR and leaf tempera-

ture and to investigate leaf-to-leaf variation at constant PAR

and temperature. Leaves near the top of the canopy typically

had isoprene emission rates of less than 0.1 |j.g C g_1 h"1 before

700 local standard time (LST) and after 1700 LST with peak

emission rates of over 150 |Xg C g"1 h-1 in early afternoon,

associated with PAR fluxes over 1000 |xmol nr2 s4 and leaf

temperatures over 33°C. Peak emissions for leaves lower in

the canopy were as high as 100 |lg C g4 rr1 and occurred at

times ranging from morning to late afternoon depending on

their position relative to sunfleck gaps in the canopy.

3.2. Surface Layer Fluxes

Isoprene fluxes in the surface layer above the forest canopy

were measured by relaxed eddy accumulation and a gradient

profile method. The objectives of these surface layer flux meas-

urements included (1) testing the REA system, (2) comparing

REA and GPEC estimates of isoprene fluxes, and (3) evaluat-

ing the results of isoprene emission models with above-

canopy flux measurements.

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

Table 2. Comparison of Isoprene Flux Measurement Techniques and Emission Model Variables

18,561

Measurement

Cases

N

Measured

Flux

Emission Model Variables

Technique

6

DOAK

Leaf

LEC

LEC

LNC

LNC

Sunlit Leaves

Shaded Leaves

Sunlit Leaves

Shaded Leaves

49

9

99

111

1.0

1.0

0.86

0.59

1.0

0.8

1.0

0.8

Branch

BCS

BCS

BRS

BRS

BRS

Sunlit Branch

Shaded Branch

Q. alba

Q. prinus

Q. velutina

7

29

16

15

10

102

101

75

110

114

0.53

0.31

0.46

0.31

0.23

1.0

0.8

0.8

0.8

0.8

Surface layer

REA

GPEC

(PAR>250)

(PAR>250)

52

27

102

86

0.32

0.34

0.85

0.85

0.40

0.40

420

420

Mixed layer

All cases

Strong convection

Strong convection

29

8

8

102

122

148

0.39

0.41

0.41

0.86

0.86

0.86

0.20

0.20

0.24

380

380

400

Isoprene flux units include »ugCg-Uri and >> mg C nr* h-i. N is the number of flux measurements. Emission

model components are described by equations 5 and 6 and include oak (Quercus) emission rate capacity (en.r tie

C g-i hi), emission activity factors (y and 8), total foliar density (D, g dry weight m-*), and the oak fractionof'total

foliar density (Dqak)-

C02 and Water Vapor Fluxes. The accuracy of the C02

and water vapor fluxes estimated by direct eddy correlation

with the open path IRGA (EC-O) has been established in pre-

vious studies [Baldocchi and Harley, 1995; Verma et ai,

1986], Fluxes estimated with the EC-0 system are used here

to evaluate fluxes estimated by direct eddy correlation with a

closed path IRGA (EC-C) and by relaxed eddy accumulation.

The corrections proposed by Webb et al. [1980] to account for

density fluctuations were applied to the EC-0 flux estimates

but not to the EC-C measurements since heat transfer through

the sampling tube should dampen temperature fluctuations.

The EC-C and EC-0 systems were colocated for over 50 half-

hour sampling periods,. C02 fluxes estimated with the two

systems were within ±5% and were strongly correlated (r2 =

0.99). The cospectra and power spectra calculated from the

two analyzers agree for both C02 and water vapor.

^Slightly different sampling periods were used for the REA

(27 min) and EC-0 (30 min) systems. Air passed through a

relatively short length of tubing (3 m) prior to reaching the

valve that directed the sample into the up or down reservoir.

We assume that there is minimal dampening of temperature

fluctuations and so have applied the Webb et al. [1980] den-

sity corrections. This may result in an overestimate of C02

fluxes for late morning and midday when sensible heat fluxes

are highest. The REA and EC-O estimates of C02 fluxes

shown in Figure 1 follow the expected diurnal pattern of low

fluxes in early morning, maximum downward (negative) fluxes

at midday, followed by decreasing fluxes into the evening.

The mean REA C02 flux estimate (-0.45 ± 0.07 mg m-2 s-') is

13% greater than the mean flux (-0.40 ± 0.06 mg m-2 S"1) calcu-

lated with the EC-0 system. Thé mean REA flux is 43%

greater if density corrections [Webb et ai, 1980] are not

applied. The scatterplot shown in Figure 2 indicates that

there is only a moderate correlation (r2=0.45) between the

REA and EC-O flux estimates. The REA system slightly over-

estimates the lowest C02 fluxes and slightly underestimates

the highest fluxes.

REA Isoprene Flux Estimates. Isoprene fluxes were esti-

mated with the REA system for 59 sampling periods (-27 min).

The three variables that determine isoprene fluxes calculated

by the REA method, and the resulting isoprene flux, are shown

for August 5 m Figure 3. Variations in (C„ - Cd) dominate the

0.4

E-0.4

ai

x

3-0.8

-1.2

-1.6

o EC-0

» REA

—EC-O REA

18

21

12 15

TIME, LST (h)

Figure 1. Diurnal variations in above-canopy C02 fluxes

estimated with an eddy correlation (EC-O) system and a

relaxed eddy accumulation (REA) system. A best fit third-

order polynomial is shown as a dashed (EC-O) or solid (REA)

line. Times are local standard time.

18,562

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

0.4

-1.6

-1.6

-1.2

-0.8

-0.4

0.4

EC-0 C02 FLUX (mg m"2 s1)

Figure 2. Comparison of above-canopy CO2 fluxes estimated

with an eddy correlation (EC-O) system and a relaxed eddy

accumulation (REA) system.

ambient temperature of 26.TC, which is similar to the condi-

tions observed for the high-light sampling periods of the REA

system. For high-light conditions, the mean isoprene flux

estimated by the GPEC system is 20% lower than the mean

flux estimated by REA.

Negative isoprene gradients (increasing isoprene concen-

tration with height) greater than 0.05 mg C nr1 were observed

during four of the 30 sampling periods. The negative gradi-

ents were all observed in the afternoon between 1500 and

1700 h LST. The downward fluxes ranged from -0.2 to -2.0 mg

C nr2 h-l with an average deposition of-1.2 mg C nr2 h1.

3.3. Mixed Layer Fluxes

Isoprene mixing ratios in the atmospheric boundary layer

above the Walker Branch field site range from less than 0.2 to

greater than 5 ppbv isoprene. These data (Figure 5) represent

samples collected at heights up to 800 m AGL during 29 half-

hour sampling periods over a 2-week period. The mean mixed

layer (>160 m AGL) isoprene mixing ratio of 0.8 ppbv is 56%

less than the mean mixing ratio of 1.8 ppbv for the surface layer

(< 150 m AGL). Guenther et al [1996] observed a similar

mean isoprene mixing ratio (1.7 ppbv) in the surface layer

resulting diurnal pattern of isoprene fluxes (Figure 3). The

REA isoprene fluxes cover a period of 6 days (Julian date 213

to 218, Figure 4). The standard deviation of vertical wind

speed (ow) ranged from < 0.3 m s-1 (at night) to > 0.6 m S"1

(midday). The observed estimates of (C„ - Q) increased from

morning values of ± 0.5 jig C nr3, reaching > 15 |xg C nr3 at

midday and fell to < -1 ng C nr3 in the evening. The calculated

fluxes range from-1.4 to 15.8 mg C m-2 h-1, with a mean of

4.2±0.6 mg C nr2 h4, and follow the expected pattern of

increasing emissions with increasing PAR and temperature.

PAR fluxes above the canopy were < 450 (Xmol nr2 s-l during

14 of the measurement periods. The mean flux of 0.05 mg C nr2

h"1 observed for these low-light conditions is dominated by

deposition (downward) fluxes. Downward isoprene fluxes

were estimated for 11 (19%)-sampling periods. Isoprene depo-

sition fluxes of-0.02 to -1.4 (mean=-0.6) mg C m-2 h"1 were

estimated for these periods which typically occurred before

800 or after 1700 h LST. These results indicate that deposi-

tion fluxes may be significant but are much less than midday

emission fluxes. When only upward isoprene fluxes are con-

sidered, the mean flux of 0.9 mg C nr2 h_1 for low light condi-

tions (average PAR flux of 337 (Jmol nr2 s"1 and ambient

temperature of 23.7°C) is almost an order of magnitude lower

than tfie mean flux of 5.9 mg Cnr2 h"1 at-high light conditions

(average PAR of 1140 nmol nr2 s1 and ambient temperature of

25. rc).

GPEC Isoprene Flux Estimates. Half-hour average

isoprene fluxes were estimated during 30 sampling periods

over seven days with the GPEC system described above. The

GPEC system used water vapor gradients and direct eddy

correlation flux measurements to estimate eddy diffusivities

and should not be influenced by roughness sublayer effects.

Observed isoprene gradients ranged from -0.17 to 0.73

(mean=0.26) |xg C nr1 and were associated with eddy diffusivi-

ties of 4.0±0.6 m2 s-'. The GPEC isoprene flux estimates

(Julian date 208 to 214, Figure 4) range from < -1 mg C nr2 h-l

to > 10 rag C nr2 h1. The mean GPEC flux estimate for

high-light (>450 umol nr2 s1) conditions is 4.8±0.8 mg Cm-2

h-1. These fluxes represent a mean PAR of 850 ^mol nv2 s4 and

10 15

TIME, LST (h)

20

10 15

TIME, LST (h)

20

10 15

TIME, LST (h)

20

Figure 3. (bottom) Isoprene fluxes on August 5, 1992 esti-

mated by relaxed eddy accumulation (REA) as the product of

(top) the standard deviation of vertical wind speed (CTW) and

the empirical coefficient ß, and (middle) the isoprene concen-

trations associated with updrafts (Cu) and downdrafls (C<¡).

Times are local standard time.

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

18,563

208

214

JULIAN DATE

Figure 4. Above-canopy surface layer isoprene fluxes on

Julian dates 208 to 219 (July 26 to August 5) estimated with

the relaxed eddy accumulation (REA) and gradient profile with

eddy correlation (GPEC) systems and predicted with equation

5 (MODEL).

above a mixed forest site in the state of Georgia, but a much

higher (1.4 ppbv) mean mixing ratio in the mixed layer, equiva-

lent to an 18% decrease. Analysis of paired (surface layer and

mixed layer) samples at the Georgia field site showed a mean

decrease of 38% between surface and mixed layers. These

results demonstrate that there are considerable uncertainties

associated with using surface layer measurements to predict

isoprene mixing ratios in the mixed layer.

The diurnal patterns of the three variables required to esti-

mate fluxes by the MB method and the resulting isoprene

fluxes are shown in Figure 6. Estimates of the mixed layer cap-

1000

750

500

250

MIXED-LAYER

SURFACE-LAYER

ISOPRENE (ppbv)

Figure 5. Thirty-minute average isoprene mixing ratios meas-

ured with the tethered balloon sampling system. The mean and

standard deviations for altitude and mixing ratio are shown for

mixed layer and surface layer samples.

ping inversion height vary from under 300 m AGL before 900

h LST to over 1200 m AGL after 1500 h LST. Estimates of the

lifetime of isoprene (about 40 min) and the mean mixed layer

isoprene mixing ratio (about 0.8 ppbv) were nearly constant

between the hours of 900 and 1500 LST. The resulting iso-

prene flux estimates range from 0.5 to 7.6 mg C nr2 h1, with a

mean isoprene flux of 2.4±0.3 mg C nr2 lr1. The mean PAR of

1150 umol nr2 S"1 and ambient temperature of 26.0°C for these

29 sampling periods are slightly higher than the mean PAR

and temperature recorded during the REA and GPEC sampling

periods for high light (PAR > 450 umolnr2 s-1). The MB flux

source region (Plate 1) has a considerably lower density of

isoprene emitting foliage than the SL region that is the source

of the REA and GPEC fluxes. This is probably responsible for

the 56% lower mean flux estimated with the MB method than

for the REA system and 44% lower than the mean GPEC flux

estimate.

Isoprene mixing ratio profiles were used to estimate fluxes

using the mixed layer gradient technique described in section

2.2. Eight of the 29 sampling periods were suitable for calcu-

lating fluxes with the MLG method (i.e., significant surface

heat flux and minimal cumulus cloud cover). Isoprene fluxes

for these eight sampling periods were estimated to be 5.0±1.7

mg C nr2 lr1. The observed profiles and fits are shown in

Figure 7. The amount of isoprene emitting foliage in the MLG

region shown in Plate 1 is about 25% higher than in the MB

region. The mean MB flux estimate for the same eight sampling

periods was 2.8±0.7 mg C nr2 lr1 which is about 57% lower

than the MLG flux estimates. These findings of lower MB and

12 16

TIME, LST (h)

20

12 16

TIME, LST (h)

20

Figure 6. (bottom) Isoprene fluxes calculated with the mass

balance (MB) technique as the product of mixed layer capping

inversion height (z¡), isoprene lifetime (t), and the mean mixed

layer isoprene mixing ratio (Cm). Times are local standard time.

18,564

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

0 12 3 4

0 12 3 4

29JUI92 15:45

0 12 3 4

5Aug92 12:15

0 12 3 4

Isoprene Mixing Ratio (ppbv)

0 12 3 4

5Aug92 15:15

0 12 3 4

Figure 7. Isoprene profiles used to calculate fluxes using the MLG technique. The profile fit (solid line) is

calculated from the MLG-derived fluxes and large eddy simulation (LES) generated gradient functions.

Observed isoprene mixing ratios are marked with asterisks or crosses, where only asterisks were used to cal-

culate fluxes. Horizontal bars through these points are theoretical estimates of error variance. Times are local

standard time. SF and EF values to the right of each profile are the computed surface and entrainment fluxes,

respectively, in units of mg C nr2 h"1.

higher MLG flux estimates are consistent with the discussion

of Guentkcr et al. [1996] which predicts that the MB method

should systematically underestimate fluxes, while the MLG

technique may be prone to overestimates.

4. Emission Model Evaluation

The isoprene emission models described in section 2.4

require accurate estimates of oak emission capacity (eoak)>

emission activity factors (y and 8), total foliar density (D) and

the fraction of foliage that is oak leaves (DOAK). Table 2

includes estimates of each of these variables and the actual

fluxes estimated with each of the eight techniques (section 2,

Table 1).

4.1. Foliar Density, D, and Fraction of Oak Foliage, Z>oAK

Oak foliar densities range from 0 to 282 g m-2 within the 31

circular 30-ni diameter sample plots. These plots have an aver-

age tree foliar density of 366 g nr2 and a leaf area index of 4.9

m2 nr2 dominated by Queráis (95 g nr2) and Pin us (88 g m-2).

The remaining forest is dominated by deciduous trees, primar-

ily species of Cornus (26 g m-2), Acer (22 g nr2), Liriodendron

(22 g m-2), and Oxydendrum (15 g m-2).

The 12 sampling plots within the SL region are located at

distances ranging from 15 to 610 m from the tower. Oak foliar

densities were highly variable along the transect with values

ranging from 258 g m-2 at the plot centered at 15 m from the

tower to,less than 2 g m-2 in the sampling plots at 135 m from

the tower and between 300 and 500 m from the tower. Lamb et

al. [1996] report results of three methods of estimating oak

foliar density along this transect: an arithmetic average of 110

g m-2, a distance weighted average of 220 g nr2 and a footprint

weighted average of 203 g nr2. The footprint weighted aver-

age is assumed to be the best estimate.

The 30-m resolution Landsat TM land cover (TMLC) data-

base indicates that 98% of the SL region, 82% of the MLG

landscape region and 70% of the MB landscape region are

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

18,565

covered by one of six forest land cover types. The TMLC esti-

mates of total foliar density shown in Table 2 range from 420 g

nr2 in the SL region to 400 g nr2 in the MLG region and 380 g

m-2 in the MB region. The total foliar density predicted by the

BEIS2 model for thé three counties surrounding the field site

is about 40% less than the TMLC estimate for the MB region.

Table 2 shows that the TMLC estimates of D0AK include 0.20

in the MB region, 0.24 in the MLG region and 0.40 in thè SL

region. The BEIS2 estimates of D0AK for the three-county area

(0.42) is a factor of 2 higher than the TMLC estimate for the

MB region.

The TMLC estimate of oak foliar density in the SL region is

168 g m-2. The various estimates oí Lamb et al. [1996] are

within about ±50% of this value. Considerably lower oak

foliar densities are estimated by the TMLC database for the

MLG (96 g nr2) and MB (76 g nr2) regions. The TMLC esti-

mate for the MB region is within about 25% of the BE1S2

model estimate for the three-county area. These results indi-

cate that estimates of oak foliar densities are associated with

uncertainties of ±50% or less.

4.2. Emission Activity Factors, y and 8

The enclosure measurement systems were used to investi-

gate emission activity factors. The results summarized in this

section are described in detail by P. Harley et al., (1996b).

The LNC and LËC systems both estimate that shade leaf emis-

sions are about 10% less than sun leaf emissions. This is

equivalent to assigning shade leaves a value of S = 0.9. Meas-

urements from the BCS system result in a value of 8 = 0.79 for

shaded branches. The LEC, LNC, and BCS enclosure systems

were also used to evaluate the light and temperature depend-

ent algorithms developed by Guenther et al. [1993] to esti-

mate y. The results demonstrate that these algorithms can

simulate the observed isoprene emission rate variations in the

top of a mature tree canopy to within ±15% for midday condi-

tions.

A tree branch with a PAR flux of 1000 umol nr2 s-1 at the

top of the branch has PAR fluxes on individual leaves that

range from less than 100 to 1000 umol nr2 S"1 due to leaf angle

orientation and shading by upper leaves. In addition, the

temperature of various leaves on a branch may differ by several

degrees. Guenther et al. [1994] recognized that the emission

activity factor applied to a branch must be adjusted to account

for these effects. They noted that field comparisons of the ratio

between leaf and branch level emission rates range from 1.46 to

2.03 [Guenther et al., 1996] and recommended that y be

divided by a factor of 1.75 for branch measurements. We

evaluated this ratio by measuring isoprene emission rates for

whole branches and individual leaves within the canopy of a

mature oak (Q. alba) tree with the BCS and LNC systems. The

ratio of leaf-to-branch emission rates was 1.92 near the top of

the canopy and decreased to 1.69 in the middle of the canopy.

The radiative transfer model and emission algorithms

described in section 2.4 predict that this ratio will vary

between 1.3 and 1.9 depending on branch LAI, mean leaf

orientation angle, and solar elevation. This ratio increases

with decreasing mean leaf angle, increasing LAI and decreas-

ing solar elevation. These results support the Guenther et al.

[1994] recommendation to estimate y based on above branch

PAR and temperature and then divide by 1.75 to calculate a

branch average y. An overall uncertainty of about ±30% is

associated with estimates of y for branch enclosure measure-

ments.

Uncertainties in estimating y for an entire forest canopy are

greater than those associated with branch measurements.

Lamb et al. [1996] evaluated the canopy average estimates of y

predicted by five different canopy environment models for the

REA and GPEC estimates. The complexity of these canopy

environment models range from treating the canopy as a single

leaf to a detailed numerical model that accounts for leaf-sun

geometry, leaf energy balance, photosynthesis, transpiration,

respiration, and gas transport within the canopy. The results

indicate that the various models predict fluxes that are within

approximately ±20%. Additional uncertainties in the leaf

level relationships (about +15%) and in estimating ambient

temperature and above-canopy PAR (about ±25%) result in an

overall uncertainty of about ±35% for y.

The diurnal patterns of predicted and observed fluxes for

the entire study period (Figure 4) show positive fluxes

between 700 and 1800 LST with a maximum occurring in early

afternoon. The observed diurnal patterns are consistent with

emission model predictions.

4.3. Emission Capacity, E

Emission capacities for individual leaves or for an entire

branch are calculated from (6) using the measured emission

rate and an estimate of y calculated from the algorithms

described by Guenther et al. [1993], Measurements of indi-

vidual leaves at conditions where y = 1 result in estimates of e

that have relatively low uncertainties (±10%). Uncertainties

in e are considerably higher when y must be estimated from leaf

temperature and PAR conditions that deviate substantially

from standard conditions. Uncertainties in extrapolating E

obtained for a few leaves to all leaves in a forest belonging to a

particular plant genus are often about ±50% [Guenther et al,

1994]. Estimating emission capacities from area-averaged flux

measurements have uncertainties associated with the flux

measurements, the canopy environment model estimates of y,

and the requirement of foliar density estimates. Each of these

three variables has an uncertainty of about ±35% resulting in

an overall uncertainty of about ±60%.

The BRS enclosure system estimates of isoprene emission

capacities are 16±3 ug C g-l h-l (n=10) for Nyssa sylvatica and

64±7 |ig C g-1 h-1 (n=14) for Liquidambar styraciflua. These

results agree with published emission capacities [Harley et

al., 1996a; Guenther et al, 1994] which include 70±35 ug C

g-1 h"1 for Liquidambar species and 12±6 (ig C g-l h"1 for

Nyssa species. Estimates of e for the three oak species meas-

ured with the BRS system range from 75 ug C g-t h-l for Q.

alba to 114 jig C g-l h-l for Q. velutina. Emission capacities

estimated with the LEC, LNC, and BCS systems for a single Q.

alba tree next to the walkup tower range from 91 to 111 jxg C

g"1 h-l. The lowest uncertainty is associated with the isoprene

emission capacity of 99 ug C g-> h-l estimated from measure-

ments on sun leaves with a leaf temperature of 30°C and PAR

of 1000 umol m-2 s-i. Eight of the nine leaf and branch esti-

mates of e listed in Table 2 fall within the relatively narrow

range of 100±11 ug C g-l h-'.

Estimates of area-averaged oak emission capacities were

estimated from above-canopy surface layer and mixed layer

measurements and (6). The isoprene emission capacities esti-

mated for oaks in the SL flux region include 86 ug C g-l rr1

with the GPEC system and 102 ug C g-l h-l with the REA

system. The MLG estimate of 148 ug C g-l h-l is the highest

oak emission capacity listed in Table 2. The MB estimate for

the same eight sampling periods used for the MLG estimates is

18,566

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

122 \ig C g"1 h-1 and is about 20% higher than the mean emis-

sion capacity calculated for all 29 sampling periods. The sur-

face layer estimates of oak isoprene e (94±8 [ig C g-1 h-1) are

about 25% less than the mixed layer estimates of oak isoprene

e (124+22 ng C g-1 h-l).

The mean oak isoprene emission capacities for each meas-

urement scale shown in Table 2 include 102 ug C g"1 h-l for

leaf enclosure measurements, 100 Hg C g"1 h"1 for branch enclo-

sure measurements, 94 ng C g"1 Ir1 for above-canopy surface

layer measurements, and 124 |ig C g"1 rr1 for above-canopy

mixed layer measurements. These results demonstrate that the

emission model techniques, which enable us to estimate the

oak isoprene emission capacity associated with each measure-

ment, produce results within about ±25% which is within the

uncertainties associated with these flux measurements. Sig-

nificant deposition losses within the forest canopy would

result in lower emission capacities estimated by the above-

canopy flux measurement techniques. While one of the surface

layer techniques (GPEC) had a somewhat lower estimated e,

the other techniques (REA, MB, MLG) resulted in oak iso-

prene emission capacities that were about equal to or higher

than the emission capacities estimated from enclosure tech-

niques. This result suggests that although it is likely that

isoprene deposition losses occur, as indicated by the morning

and evening surface layer flux measurements, they are a rela-

tively minor component of the net flux. Thirteen of the fourteen

emission capacities estimated in Table 2 are within ±25% of

100 (tg C g-1 h"1 and the emission capacity estimated using the

mixed layer gradient technique (148 tig c g"1 h"1) Is even

greater. These results demonstrate that the isoprene emission

capacity for oak trees, in at least this region, is higher than the

value used in existing models (70 ug C g"1 h1 for BEIS2, 14.7

ugCg-lh-iforBEIS).

5. Summary

Eight flux measurement techniques were used to estimate

isoprene fluxes from a temperate forest. Fluxes from individual

leaves and branches were measured with enclosure systems,

above-canopy fluxes were measured from a tower by relaxed

eddy accumulation and surface layer gradients, and fluxes from

landscapes covering an area of up to a few hundred square

kilometers were estimated using a tethered balloon sampling

system. Each measurement system provided specific advan-

tages for investigating different aspects of biogenic emission

models. The results from all measurement techniques demon-

strate that existing emission models underestimate the

isoprene emission capacity for oaks in this region and a value

of 100 |j.g C g_1 h-1 is recommended. The results also demon-

strate that reasonable estimates of isoprene fluxes and bound-

ary layer isoprene mixing ratios can be predicted with existing

models if accurate data (e.g., emission capacities and land char-

acteristics data) are available for initializing the models. It is

clear that considerable uncertainties exist in field flux meas-

urements and in each of the main components of emission

models. Research directed at narrowing these uncertainties

and acquiring data for initializing models in other regions

remains a priority for biogenic emission studies.

Acknowledgments. This research was partially supported by the

Environmental Protection Agency, Research Triangle Park, North

Carolina, under Interagency Agreement Grant No. DW49934973-01-0,

and by the Southern Oxidants Study (SOS) under Assistance Agreement

No. CR817766 to North Carolina State University. The authors grate-

fully acknowledge Rick Lowe and Brett Taylor for assistance with field

measurements and data analysis and Mike Huston of the Oak Ridge

National Laboratory for assistance in characterizing the ecology of this

region. We thank Detlev Helmig and Chris Cantrell of NCAR and two

anonymous reviewers for helpful comments on this manuscript. The

National Center for Atmospheric Research is sponsored by the National

Science Foundation.

References

Atkinson, R., Gas-phase tropospheric chemistry of organic compounds:

A review, Amos. Environ., 24À, 1-41, 1990.

Baldocchi, D., A. Guenther, P. Harley, L. Klinger, P. Zimmerman, B.

Lamb, and H. Westberg, The fluxes and air chemistry of isoprene

above a deciduous hardwood forest. Philos. Trans. R. Soc. Lond., 350,

279-296, 1995.

Businger, J., and S. Oncley, Flux measurement with conditional sam-

pling, J. Atmos. Oceanic Technol, 7 (2), 349-352,1990.

Geron, C, A. Guenther, and T. Pierce, An improved model for estimat-

ing emissions of volatile organic compounds from forests in the east-

ern United States,/. Geophys. Res., 99, 12,773-12,792, 1994.

Geron, C, T. Pierce, and A. Guenther, Reassessment of biogenic vola-

tile organic compound emissions in the Atlanta area, Atmos. Environ.,

29,1569-1578,1995.

Greenberg, J., P. Zimmerman, B. Taylor, G. Silver, and R. Fall, Sub-

parts per billion detection of isoprene using a reduction gas detector

with a portable gas Chromatograph, Atmos. Environ., 27, 2689-2692,

1993.

Guenther, A., Seasonal and spatial variations in natural volatile organic

compound emissions, Ecol. Appi., in press, 1996.

Guenther, A., P. Zimmerman, P. Harley, R. Monson, and R. Fall, Iso-

prene and monoterpene emission rate variability: Model evaluation

and sensitivity analysis, J. Geophys. Res., 98,12,609-12,617,1993.

Guenther, A., P. Zimmerman, and M. Wildermuth, Natural volatile

organic compound emission rate estimates for U.S. woodland land-

scapes, Atmos. Environ., 28,1197-1210,1994.

Guenther, A., P. Zimmerman, L. Klinger, J. Greenberg, C. Ennis, K.

Davis, H. Westberg, and E. Allwine, Estimates of regional natural

volatile organic compound fluxes from enclosure and ambient meas-

urements, J. Geophys. Res., 101, 1345, 1996.

Harley, P., A. Guenther, and P. Zimmerman, Effects of light, tempera-

ture and canopy position on net photosynthesis and isoprene emission

from leaves of sweetgum (Liquidambar styraciflua L.), Tree Phys.,

16,25-32, 1996a.

Johnson, D., and R. van Hook, Analysis of Biogeochemical Cycling

Processes in Walker Branch Watershed, pp. 401, Springer-Verlag,

New York, 1989.

Lamb, B., et al., Evaluation of forest canopy models for estimating iso-

prene emissions, J. Geophys. Res., in press, 1996.

Lamb, B., D. Gay, H. Westberg, and T. Pierce, A biogenic hydrocarbon

emission inventory for the U.S. using a simple forest canopy model,

Atmos. Environ., 27,1673-1690,1993.

Lamb, B., H. Westberg, and G. Allwine, Biogenic hydrocarbon emis-

sions from deciduous and coniferous trees in the United States, J.

Geophys. Res., 90, 2380-2390, 1985.

Lamb, B., H. Westberg, and G. Allwine, Isoprene emission fluxes

determined by an atmospheric tracer technique, Atmos. Environ., 20,

1-8, 1986.

Lu, Y., and M. Khalil, Tropospheric OH: Model calculations of spatial,

temporal, and secular variations, Chemosphere, 23, 397-444, 1991.

Norman, J., Simulation of microclimates, in Biometeorology in Inte-

grated Pest Management, Academic, San Diego, Calif, 1982.

Oncley, S., A. Delany, T. Horst, and P. Tans, Verification of flux meas-

urement using relaxed eddy accumulation, Atmos. Environ., 27A,

mi-Uld, 1993.

Pierce, T., and P. Waldruff, PC-BEIS: A personal computer version of

the biogenic emissions inventory system, J. Air and Waste Manage.

Assoc, 41, 937-941, 1991.

Verma, S., D. Baldocchi, E. Anderson, D. Matt, and R. Clement, Eddy

fluxes of C02, water vapor and sensible heat over a deciduous

forest, Boundary Layer Meteorol., 36, 71-91, 1986.

GUENTHER ET AL.: ISOPRENE FLUX MEASUREMENTS TECHNIQUES

18,567

Webb, E., A. Pearra, and R. Leuning, Collection of flux measurements

for density effects due to heat and water vapor transfer, Q. J. R.

Meteorol. Soc, 106, 85-100, 1980.

A. Guenther, W. Baugh, K. Davis, G. Hampton,! P. Harley, L.

Klinger, L. Vierling, and P. Zimmerman, National Center for Atmos-

pheric Research, P.O. Box 3000, Boulder, CO 80307. (e-mail:

guenther@acd.ucar.edu)

G. Allwine, S. Dilts, B. Lamb, and H. Westberg, Department of Civil

and Environmental Engineering, Washington State University, Pullman,

WA 99164.

D. Baldocchi, Atmospheric Turbulence and Diffusion Division,

NOAA, Oak Ridge, TN 37830.

C. Geron, and T. Pierce, Office of Research and Development, U.S.

Environmental Protection Agency, Research Triangle Park, NC 27711.

(Received October Î6, 1995; revised January 29, 1996;

accepted February 27, 1996.)