goulden et al 96a

8/2/2019 Goulden Et Al 96a

1/14

Change Biology (1996) 2,169-182

of carbon sequestration by long-termcovariance: methods and a critical evaluation of

L. GOULDEN, ].WILLIAM MUNGER, SONG-MIAO FAN, BRUCE C. DAUBE,C. WOFSYof Applied Sciences, and Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA, USA

AbstractThe turbulent exchanges of CO2 and water vapour between an aggrading deciduousforest in the north-eastern United States (Harvard Forest) and the atmosphere weremeasured from 1990 to 1994 using the eddy covariance technique. We present a detaileddescription of the methods used and a rigorous evaluation of the precision and accuracyof these measurements. We partition the sources of error into three categories: (1) uniformsystematic errors are constant and independent of measurement conditions (2) selectivesystematic errors result when the accuracy of the exchange measurement varies as afunction of the physical environment, and (3) sampling uncertainty results when summingan incomplete data set to calculate long-term exchange.

Analysis of the surface energy budget indicates a uniform systematic error in theturbulent exchange measurements of -20 to 0%. A comparison of nocturnal eddy fluxwith chamber measurements indicates a selective systematic underestimation duringcalm (friction velocity < 0.17 m S-l) nocturnal periods. We describe an approach tocorrect for this error. The integrated carbon sequestration in 1994 was 2.1 t C ha-l y-lwith a 90% confidence interval due to sampling uncertainty of :!:0.3 t C ha-l y-ldetermined by Monte Carlo simulation. Sampling uncertainty may be reduced byestimating the flux as a function of the physical environment during periods whendirect observations are unavailable, and by minimizing the length of intervals withoutflux data. These analyses lead us to place an overall uncertainty on the annual carbonsequestration in 1994 of --0.3 to +0.8 t C ha-l y-l.Keywords: eciduous forest, eddy correlation, rnicrometeorology, NEp, net ecosystemproduction,photosynthesis, respirationReceived 5 June 1995; decision o author 8 September 995; revision accepted 0 October1995

of the exchangeof CO2betweenvegetation and the atmosphere have the potential

prove understanding of the role terrestrialplay in the global carbon cycle. Eddy covari-

grassland (Baldocchi et al. 1988).technical advances have made long-terrn eddy

s practical (Wofsy et al. 1993),possibility of a global network of field

stations for monitoring biosphere-atmosphere CO2exchange (Baldocchi et al. 1996). However, investigatorsmust first establish that the accuracy and precision ofeddy covariance s sufficient to allow a reliable assessmentof carbon sequestration over time scales ranging fromhours to decades.

We have used the eddy covariance technique through-out the last four years to monitor the net exchanges ofCO2 and H2O above the Harvard Forest, an aggradingdeciduous forest in the northeastern United States. Theobservations span a range of climatic conditions, allowinga quantitative assessmentof the physical and biologicalcontrols on whole-forest activity. In this paper we (i)describe the methods in use at Harvard Forest (ii) sum-Michael L. Goulden. e-mail [email protected],ax +1-617-495-9837, el +1-617-496-4571

1996 Blackwell Science Ltd 169


2/14

170 M.l GOULDEN et alurban areas are 100 km east (Boston) and 100 km south-west (Hartford). Additional site details are included inMunger et al. (1996), Moore et al. (1996), and Wofsyet al. (1993).

marize the observations of carbon flux, and (iii) evaluatethe precision and accuracy of these measurements.

We partition the sources of error into three categories,and proceed by assessing each category separately. (1Uniform systematic rrors are constant and independent ofmeasurement conditions (e.g. an error in the span of agas analyser). A uniform error of 15% in the hourlymeasurements results in a 15% error in the calculation oflong-term carbon sequestration. (2) Selectivesystematicerrors esult when the accuracy of the exchangemeasure-ment varies as a function of the physical environment.Because carbon sequestration reflects the differencebetween two larger fluxes, respiratory efflux during thenight and photosynthetic uptake during the day, a smallselective underestimation of nocturnal flux can causea large overestimation of long-term sequestration. (3)Samplinguncertaintyoccurs when summing an ncompletedata set to estimate ong-term exchange.The observationsat Harvard Forest are nterrupted for maintenance, equip-ment failure, and unsuitable atmospheric conditions.Day-to-day variability in carbon balance creates uncer-tainty in the calculation of long-term flux.

Tower measurementsThe eddy fluxes of sensible heat, latent heat, COu andmomentum are measured at 30 m on the tower. Windand temperature are measured with a 3-axis sonic anemo-meter (Applied Technologies, Boulder CO) pointed intothe prevailing wind direction (west). The mixing ratiosof CO2 and H2O are monitored by sampling 6--8standardlitre min-1 (slpm) through an inlet located 0.5 m behindthe vertical axis of the anemometer. The error due toseparation of the inlet from the anemometer should besmall at 30-m altitude (1-2% during unstable periods,Lee & Black 1994).The sample is drawn down the towerin a 50-m, 0.5-cm inner diameter (id) Teflon tube andthrough a CO2/H2O infrared gas analyser (IRGA; Model6262, LiCor, Lincoln NE) located in the instrument hut.

A 2-11nl,47-mm diameter Teflon filter at the inlet onthe tower is changed every 2-4 weeks, and a second filterimmediately before the IRGA is changed every 6 months.The nner surface of the sampling tube is cleaned periodic-ally using a moist cotton ball. The pressure immediatelyafter the IRGA is monitored and actively controlled at60 kPa using a variable valve (MKS Instruments, AndoverMA). The sample stream reaches an equilibrium temper-ature before entering the IRGA, removing the effects ofcoincident sensible heat flux (e.g. Webb et al. 1980). Theeffects of coincident water vapor fluctuations are removedby the IRGA software. The gain of the IRGA is automatic-ally determined every 5 h by addition to the main samplestream near the inlet of 1% CO2gas mixture at 30 standardml min-l (sccm). The response ime of the IRGA to a stepchange in CO2 s also determined during this procedure.

The data acquisition and control systems are fullyautomated, allowing extended periods of unattendedoperation. The system records outputs from the sonicanemometer and the IRGA at 4 Hz. A spectral analysisof data collected at 10 Hz revealed no appreciable under-estimation of flux due to 4 Hz sampling (Moore et al.1996). The data are stored on disk at the site andtransferred every two to four days for processing. EddyCO2 flux is calculated as the 30 minute covariance ofvertical wind velocity (w') and CO2 concentration (c').The CO2 record is detrended using a linear least squaresfit. The time lag required to draw air down the tower isdetermined by maximizing the correlation between w'and c' .The lag is extremely consistent due to the activepressure control at the IRGA. The flux is rotated to theplane where the mean vertical wind is zero (McMillen1988).

MethodsSite descriptionThe measurements are made on the Prospect Hill tractof the Harvard Forest, near Petersham, Massachusetts(4232' N, 7211' W, elevation 340 m), an area typicar ofrural New England (Foster 1992). The forest is 50-to-70-years old, dominated by red oak and red maple, withscattered stands of hemlock, and white and red pine. Thecanopy height is 20-24 m. Nearly continuous forestextends for several km to the south-west and the north-west of the site, the dominant wind directions. The siteis centred in a small level drainage, with a stream runningfrom the north-west to the east within 50 m of the tower,and a gentle upward slope for several hundred metersto the south-west (5 acing north). Soil type, soil drainage,forest age, and land-use history are patchy on the scalesof 50-200 m (Foster 1992).

The measurements are made from a 31-m-tall, 30-cm-cross-section tower of the type used to support radioantennas (Rohn 25G, Peoria IL ). The small diameter towerwas selected to minimize wind distortion. Most of theinstruments and the data acquisition system are housedin a hut 15 m east of the tower base. The hut is climatecontrolled and receives electrical and phone service. Thesite is accessibleby a dirt road which is closed to publictraffic. The area is sparsely populated with no occupieddwellings to the south-west and north-west closer than2 km. The nearest secondary road is 2 km to the westand the nearest highway is 5 km to the north. The nearest

@ 1996 Blackwell Science Ltd., Global Change Biology, 2, 169-182


3/14

MEASURING CARBON SEQUESTRATION BY EDDY COVARIANCE 171The closed-path IRGA and long sampling tube result

n a small underestimation of CO2 flux due to theamping of high-frequency fluctuations (Leuning & King

1992). The magnitude of underestimation varies withtmospheric stability, creating a possible selective system-tic error. We determine the magnitude of this under-stimation by recalculating the sensible heat flux afterumerically slowing the response of the temperature

detector to simulate the slower response of the CO2analyser (exponential time constant = 0.2 s as determinedy CO2 addition on the tower). We then increase the

calculated CO2 flux by the ratio of the fast-responseheatflux to the slow-response heat flux. The response-timecorrection is determined for each 30 minute run, and istypically 0-2% during the day and 0-5% during the night.We assure the quality of data by discarding periodswith water on the sonic transducers as indicated by anunreasonable emperature signal (low 0', too hot or cold),periods with spiking on any sonic axis as indicated bythe ratios between ow, au, (JV,and u*, or periods withunusual flow or pressure at the eddy IRGA.

We use a second IRGA (Binos, Hanau Germany) tosequentially measure the mixing ratio of CO2 at 8 levelsthrough the canopy (0.05, 0.85, 2.8, 6.2, 9.5, 18.2, and30.8 m). The hourly change n CO2 beneath 30 m (storage,Wofsy et al. 1993) s calculated by interpolation throughspace and time to synchronize the profile and eddyobservations. The eddy flux and storage observations areadded to calculate the hourly net ecosystem exchange(NEE, Wofsy et al. 1993). The flux of photosyntheticallyactive photons (PPFD) to the forest is measured with asilicon quantum sensor (LiCor, Lincoln NE). The netradiation at 30 m is measured with a thermopile netradiometer (REBSQ*6, Seattle WA). Air temperature andwater vapor content are measured at the top of the towerwith an aspirated thermistor and solid-state humidityprobe (Vaisala, Woburn MA). Soil temperature is meas-ured with an array of thermistors buried at the base ofthe litter layer. Soil water content is measured hourly bytime domain reflectometry using six 15-cm and two 50-cm sets of wave guides placed vertically (CampbellScientific, Logan UT).

a removable solid PVC lid. The collars were inserted toa depth of 2-4 cm during June 1992. Air was drawn at"" 5 slpm from a sample port on the chamber top throughthe sample cell of a differential infrared gas analyser(LiCor 6251,Lincoln NE). Ambient air entered the cham-ber through a 10-cm tall, 4-cm id chimney on the side ofthe lid opposite the sample port. A subsample from thechimney was drawn through the reference cell of theIRGA at"" 0.5 slpm. A measurement was made over a12 minute period by (i) selecting a chamber with asolenoid valve (Skinner. Valves, New Britain CT), (ii)measuring flow through the chamber with a mass flowmeter (MKS Instruments, Andover MA), and (iii) measur-ing the difference in CO2 between the inlet and outlet ofthe chamber.

The data were recorded and the system managed usinga data logger (Campbell Scientific, Logan UT). The IRGAand mass flow meter were zeroed every 2 h, and theambient concentration of CO2 entering a chamber deter-mined after each observation. Soil temperature at 2-cmand 5-cm depth was measured within each chamberusing type T thermocouples. The chambers remained inplace between measurements and flow was maintainedat 5 slpm. The system was moved to a new site everyone to four days, and the chamber tops removed whenthe system was deployed at other sites. Five soil collarswere measured at least once at each of the 10 sites.

We established the analytical accuracy of the systemwith a series of standard additions. When a small flowof 1% CO2 was metered into a chamber, the increase influx matched the rate of CO2 addition to within 5%.Errors caused by aspirating air from the soil are aparticular problem with open-type chambers (Mosier1989). We checked for this possibil ity by measuring theresistance to mass flow from the soil. When combinedwith the expected pressure drop through the chimney(10-3 Pa at 5 slpm) these observations indicate a forcedflow from the soil at the leakiest collars of a few sccm.Assuming a soil CO2 concentration of 4000 ~L L -1 themaximum overestimation due to aspiration is 0.2 ~ol-2 -1m s .

The stem respiration chambers consisted of 0.01-cmpolyethylene sheet wrapped around 500-3000 cm2 ofstem. The chambers were placed at a height of 1-2 m,and sealed around the trunks using caulking compoundand tape. The sample port consisted of a perforated 1-to-4-m length of tubing wrapped around the stem withineach chamber. Ambient air entered a chamber throughan opening in the sheet held open by an inlet tube (4-cmid). A subsample from the inlet tube was drawn throughthe reference cell of the IRGA. The respiration of fivestems representing various specieswas measured at eachof the 10 sites. The absolute rates of CO2 efflux from the50 stems were well correlated with both the volumes and

Chamber measurementsWe used a multiplexing gas exchange system to measurethe efflux of CO2 from soil, stems, and leaves over dielcourses during summer 1992. The system sequentiallysampled 10 open chambers (Field et al. 1989;Mosier 1989)deployed in a semi-circle, completing a circuit every 2 h.The system was moved among 10 sites around the tower,allowing measurements at 50 soil and 50 stem locations.

The soil chambers consisted of a 27-cm id by 10-cmlong by 0.5-cm hick polyvinyl chloride (PVC) collar with@ 1996 Blackwell ScienceLtd., Global ChangeBiology,2, 169-182


4/14

172 M.L. GOULDEN et al

c0)"6>""'"Qro(5a:

the areas enclosed within the chambers. The relationshipbetween enclosed area and CO2 efflux was linear with alarge zero offset such that small diameter stems had agreater flux per area, whereas the relationship betweenvolume and efflux was linear with a zero intercept suchthat flux per volume was constant over a range ofdiameters. Respiration was therefore calculated on avolume basis and converted to ground area based on asurvey of wood volume in 40 lO-rn radius plots within500 m of the tower. Most of the stem respiration measure'-ments were made after the period of maximum secondarygrowth {April through June), and the observed flux islikely dominated by maintenance respiration.

Leaf respiration-in the canopy was measured bydeploying branch chambers from a pair of 20-rn tallscaffolding towers. The branch chambers consisted of 50-cm long by 30-cm diameter by 0.01-cm hick polyethylenebags held open by a frame of O.64-cm outer diametercoated aluminium tube {Dekoron, Aurora OH). The cham-bers enclosed 20-50 leaves. Sample air was drawnthrough a series of perforations in the Dekoron frame.Measurements were made on a total of 10 branchesrepresenting 3 speciesduring two nights in August 1992.Respiration was calculated per leaf and converted toground area based on leaf litter collections in autumn1992.

Fig. 1 Sonic anemometer horizontal rotation angle as a functionof wind direction during 1992. Points are means =!=1 tandarddeviation at wind intervals of 4(n = 50-150 from the east, n =150-250 from the west). The rotation angle is the tilt between

the anemometer horizontal plane and the local streamlines,calculated over 30-minute periods.

provide a useful tool for assessing the reliability offlux measurements (Kaimal et al. 1972). Under idealcircumstances the shapes of the w'CO2', w'T', andw'H2O' cospectra should be similar (Ohtaki 1985). Anincomplete resolution of small eddies, a common errorwhen making eddy covariance measurements, is indi-cated by the loss of power at high frequencies (Leuning& King 1992). The power spectrum of the CO2 timeseries, he cospectrum of vertical wind and CO2, and thecospectrum of vertical wind and air temperature, indicatethat the closed-path IRGA records nearly all of thefluctuations in CO2 associated with turbulent transport(Fig. 2). The CO2 spectrum decreases hrough the sub-inertial range at the expected 2/3 power to a frequencyof 1 Hz (Fig. 2a). The cospectrum of w'CO2' is similar tothat of w'T' (Fig. 2b . Both cospectra indicate that largeeddies with frequency less than 0.1 Hz dominate flux (cf.Hollinger et al. 1994). The cospectrum of W'H20', alsomeasured with the closed-path IRGA, shows a nearlycomplete lack of flux at n > 0.2 Hz, and a modestunderestimation of flux at n > 0.01 Hz (data not shown).The damping of high frequency water vapor fluctuations,which presumably is due to adsorption and desorptionon the walls of the sample tube, results in an underestima-tion of evaporation by 20%.

Results: Measurements of turbulent exchangeTurbulence measurementsThe tilt of the mean wind at the top of the HarvardForest tower generally does not vary by more than a fewdegrees rom a fixed horizontal plane, indicating that flowdistortion due to local topography or tower shadowing isminor (Fig. 1 . The rotation angle varies as a function ofwind direction in a manner consistent with a simple 2offset between the plane through the u and v axes of theanemometer and the local topography (McMillen 1988).The relationship between rotation angle and wind direc-tion is consistent from day to night, summer to winter,and calm to windy periods. The rotation angle s relativelyvariable when wind is blowing from behind the tower(45-135), suggesting the possibility of modest distortiondespite the tower's small cross section. The ratio of owto u* is consistent as a function of direction with a modestincrease in variability from 45 to 135, also suggestingthe possibility of modest tower shadowing (Moore et al.1996). Fortunately wind from behind the tower is infre-quent (""' 14% of the time). We do not remove theseperiods from the main data set as hey are often associatedwith cloud cover and we do not want to introduce a bias.

Spectral analyses of the fluctuations in atmosphericCO2, T, and H2O associated with turbulent transport

CO2 exchange at Harvard ForestMeasurements of ca2 exchangewere made during 20 300of the 35 000 hours from act. 1990 o act. 1994 (Fig. 3),with interruptions for calibration, data transfer, mainten-ance, rain, and equipment failure. Notable gaps occurredwhen an IRGA failed repeatedly in spring 1991, when a@ 1996 Blackwell ScienceLtd., Global ChangeBiology,2, 169-182


5/14

MEASURING CARBON SEQUESTRATION BY EDDY COVARIANCE 173

~

.E-4 1.E-2n (hz)

1.E+O

(a) Spectrum of CO2 with dash line showing -2/3 slope.Cospectrum of vertical wind and CO2 (solid line), and

of vertical wind and T (dashed line). Curves are52 individual spectra calculated during

in September 1992.

drive failed in May 1991, and following lightningJune 1992, August 1993, and July 1994. Thecourse of CO2 exchange was similar each year,

a midday uptake of 20-30 ~mol m-2 s-l from Junea peak nighttime efflux of 5-10 ~mol

S-l from June through August, and an efflux duringA more detailed look reveals several

between years including a notable increase nfrom Dec. 1992 to Feb. 1993 (Fig. 3; Goulden

A typical summer day of turbulent fluxes is shown inRespiratory efflux of 3-5 !1Inol m-2 s-l during the

uptake of 14-19 ~ol m-2 s-l during the day ands-l during the second

(Fig. 4a). The difference in flux between nights waswith a difference in turbulence; the first night

windy whereas the second night was calm with avelocity that often approached zero (Fig. 4b .

was quite variable from hour toa general increase at night and a general

during the morning (Fig. 4a). Nocturnal storage

was not well correlated with turbulence, and the differ-ence in eddy flux between nights was not offset bystorage. The relationships between eddy flux, CO2 stor-age, and turbulence on calm nights are discussed furtherin the section on selective systematic errors. The sum ofsensible and latent heat was 100 W m-2 less than netradiation as the forest warmed in the morning, and 0-50 W m-2 less than net radiation in the afternoon andearly evening (Fig. 4c). The outgoing radiation at nightexceeded he influx of sensible heat by 25-75 W m-2, andsummed over the 24-hour period the turbulent fluxeswere within 10% of the net radiation.

Nocturnal NEE over a whole year was exponentiallyrelated to surface soil temperature with QI0 = 2.1 (Fig. 5;Wofsy et al. 1993; Hollinger et al. 1994; Fan et al. 1995).Similarly, soil CO2 efflux measured with an automatedchamber was tightly correlated with temperature at 2 cmdepth (Fig. 6a), and bole respiration measured with achamber was tightly correlated with temperature aver-aged over the outer 2 cm of stem (Fig. 6b . The completeset of soil chamber measurements was exponentiallyrelated to the soil temperature monitored at the towerwith QI0 = 2.2 (n = 2450, data not shown). The chambermeasurements were made over diel cycles, and most ofthe range in soil temperature was due to the differencebetween days and nights. The QI0 observed with thechambers s therefore appropriate for extrapolating night-time observations of respiration to daytime as a functionof soil temperature at the tower.

The relationship between nocturnal NEE and soil tem-perature allows a separation of the processes hat contrib-ute to daytime NEE. Net ecosystem exchange at nightshould equal the combined rates of autotrophic andheterotrophic respiration. During the day NEE shouldequal the combined rates of rubisco carboxylation andoxygenation (gross ecosystem production, GEP), andautotrophic respiration and heterotrophic respiration. Weseparate daytime NEE into respiration and GEP by firstdetermining the exponential fit with QI0 = 2.2 betweenNEE during well-mixed nocturnal periods (see sectionon selective systematic errors) and soil temperaturewithin time blocks that include 100 h of valid nocturnalobservations (2-4 weeks). We then calculate GEP as thedifference between NEE and the respiration estimatedfrom soil temperature. Our sign convention is that a netaddition of CO2 to the atmosphere is a positive flux andhence GEP s negative. However, we discuss all processes,including photosynthesis and carbon sequestration, aspositive.Hourly GEP during the summer was well correlatedwith incident light (Fig. 7; Wofsy et al. 1993). The slopeat low light indicated a quantum yield of 0.055 I1mol Cper I1mol ncident photon. The maximum photosyntheticrate was 20-25 I1mol tn-2 s-I, with moderate saturation

1996 Blackwell ScienceLtd., Global ChangeBiology,2, 169-182


6/14

174 M.L. GOULDEN et al-

':-(/)~

E""6E

~0

w "'o()"Q5z

Fig. 3 Hourly net CO2 exchange aboveHarvard Forest from October 1991 toOctober 1994. A net input of carbon tothe atmosphere is positive.

Jan July Jan July Jan July Jan July1991 1992 1993 1994

photosynthesis indicates that secondary limitations tophotosynthesis such as drought are typically of minorimportance at Harvard Forest. The relationship also sup-ports the use of 'big-leaf' and long-time-step models todescribe canopy carbon uptake (Monteith 1972; Sinclairet al. 1976;Jarvis & Leverenz 1983).

O:cn')'EOE.E,0"'U

50

-5.10.151.00.80.60.40.2

"Tcn

~

Effects of turbulence on CO2 exchange

'i'E~>.0)a;"LU

Local Time (h)Fig. 4 Typical diel course of (a) carbon exchange (b) frictionvelocity, and (c) energy exchange above Harvard Forest.Observations were made 31 Aug. 1992 to 2 Sept. 1992. Linesconnect hourly averages.

beginning around 400 J.Lmol hotons m-2 s-I. The relation-ship between photosynthesis and light was very tight,with an absolute scatter similar to that observed at night,approximately =!=5 lDlol m-2 S-I (Fig. 7, Fig. 5). Thisfidelity is remarkable since a range of phenomena, nclud-ing measurement variability due to the finite samplinginterval (Baldocchi et al. 1988), physiological processesthat modulate photosynthesis, variability in the respira-tion flux (Fig. 5), and spatial heterogeneity in photosyn-thesis, could lead to high variance in daytime flux.The simple and tight relationship between light and

Turbulence may affect CO2 efflux in several ways. Duringcalm periods turbulence limits the transport of CO2through the atmosphere as discussed in the sectionon selective systematic errors. During windy periodsturbulence appears to affect the movement of CO2 outof the soil as discussed in this section. The winterobservations of whole-forest exchange ndicate a positivecorrelation between efflux and friction velocity duringvery windy periods (Fig. 8, u* > 0.8m s-l ). This correlationlikely reflects the aspiration of CO2-rich air from soil andsnow pore space, rather than a short-term increase inCO2production. The increase n CO2 efflux during windyperiods was especially pronounced in winter 1993 (Fig. 3,9; Goulden et al. 1996).Rates of efflux exceeding 10 ~molm-2 s-l were repeatedly observed when the frictionvelocity exceeded0.8 m S-l (Fig. 9a). The enhanced effluxwas observed only when the tower sampled regions tothe north-west, a poorly drained area dominated by amaple bog and old stands of hemlock. Fluxes of CO2 tothe south-west, an upland area of oaks and maples, weresimilar to those in other winters. The fluxes of latent andsensible heat (Fig. 9b , the response and gain of the CO2analyser, and the flow through the CO2 analyser, did notgive any indication of experimental problems duringthese periods.

The soil chambers were designed to allow the entry ofstatic pressure fluctuations associated with overlying@ 1996 Blackwell ScienceLtd., Global ChangeBiology,2, 169-182

20

-20

400200

0


7/14

MEASURING CARBON SEQUESTRATION BY EDDY COVARIANCE 175-

';"00(\I'EaE~0)0)c:co.c:ux

UJ(\Iao0)E':8.c:.2>z

10

8

6

4

2

0Nocturnal CO2 efflux measured

as a function of soilduring 1992. Points are

averages during periods with0.17 m s-I. Flux (J.lmol m-2 S-I) =+ 0.073 * T (OC)), n = 1800.

-2-5 0 5 10 15 20 25

Soil Temperature (oC)

16141210864

and also to reports suggesting that fluctuations in staticpressure increase gas exchange from forest soils(Baldocchi & Meyers 1991), we found no consistentrelationship during the summer between CO2 efflux intothe soil or stem chambers and above-canopy turbulence(Fig. 6a, b, c). Similarly, we have not observed a tightcorrelation between eddy CO2 flux and friction velocityduring windy summer periods (u* > 0.17 m S-I). Thedifference from winter to summer in the sensitivity ofefflux to strong turbulence may be a consequenceof thecanopy. The canopy may reduce ground-level turbulencein summer, preventing it from reaching the intensitywhere aspiration occurs.

-iCD3"0CDQJCro--09

1210864

Discussion: accuracy 0 the measurementsLong-term precisionBased on instrument specifications and calibration proto-cols we estimate a long-term precision for the eddycovariance measurements of better than :!:5%. Similarinfrared gas analysers have been used for the flux meas-urements since April 1991.The performance of the IRGAis monitored closely with a calibration every 5 h todetermine both instrument gain and response time. TheCO2 standard used in the calibration has been replacedtwice during the study, with an intercomparison betweenstandards to within 1%. The zero offsets of the mass flowmeters used in the calibration are determined every 2-4days, and the meters are calibrated at least once everyyear. Similar sonic anemometershave been used through-out the study, with only a modest change in the on-lineshadowing correction, providing long-term stability of

Local Time

a 34-cm diametermeasured with an automated chamber, and (c) friction

measured at 30 m. Circles are temperature (a) at 2-cmin the chamber. Respiration curves are individual

at 2 hour intervals connected by lines, frictionis hourly average connected by l ines.

property that we confirmed by measuringwithin a chamber (data not shown; Sigmon

Contrary to the pattern observed in winter,1996 Blackwell Science Ltd., Global Change Biology, 2,169-182


8/14


';-In~Ea~(1)0)c(\1.c~w "'

o()InIne

"

10

0

-10

-20 Fig. 7 Gross exchange as a function ofincident photosynthetically activephoton flux density (PPFD). Points arehourly means from 10 June 1992 to 16Sept. 1992. Gross exchange wascalculated as the difference betweenmeasured net exchange and respirationestimated from soil temperature. Flux(~mol m-2 s-I) = 1.4 + (-32.9 + PPFD)/(587 + PPFD), n = 1350.

-30

0 500 1000 1500PPFD (~mol m-2 5-1,

Uniform systematic errors: Accuracy of daytimemeasurementsExperience has shown that eddy covariance works bestduring windy periods. Errors during these periods arepresumably uniformly systematic (e.g. naccurate concen-trations of calibration gases), and should apply equallyto all periods. An analysis of the surface energy budgetprovides a useful approach for evaluating the measure-ments of latent heat, sensible heat, and, based on spectralsimilarity (Fig. 2b , CO2 lux. A convincing closure of theenergy budget at Harvard Forest is difficult due tothe underestimation of latent heat flux revealed by thespectral analysis, and also due to uncertainty in the rateof heat storage. When these actors are taken into account(Verma et al. 1986; Moore 1986; Leuning & King 1992)good agreement s obtained between the loss and storageof energy at the surface, and the net flux of radiation tothe forest (Fig. 10). The comparison indicates a tendencyfor the turbulent fluxes to underestimate exchange by 5-10%. An additional :!:10% should be added to accountfor uncertainties in the measurement of net radiation, inthe calculation of heat storage, and in the correction oflatent heat flux. The confidence interval for the measure-ment of daytime turbulent exchange s -20 to 0%.

Additional analyses support this conclusion. The dailyevaporation at the site measured by eddy covarianceduring 20 summer days in the later stages of dryingcycles was 0.14cm day-l, while the evaporation estimatedfrom time domain reflectometry in the top 50 cm ofsoil was 0.16 cm day-l (data not shown). The meanphotosynthesis at the site measured by eddy covariance

0.2 0.4 0.6 0.8 1.0 1.2 1.4u* (m s"

Fig. 8 CO2 efflux during periods with air temperature less than-4 C as a function of friction velocity. Points are medians :!:1standard deviation measured by the tower during 1992. Datawere sorted by u* into 10 intervals, with each interval duringthe night (R < 0 W m-2) containing 30 observations, and eachinterval during the day (R > 0 W m-2) containing 16 observations.

better than 1% (Kaimal et al. 1990; H. A. Zimmerman,personal communication). The locations of the sonic andthe IRGA inlet, the flow and time lag, and the IRGAresponse ime have remained relatively constant over thestudy. We store and archive all the raw data, allowingrecalculation of the entire record of CO2 exchange n theevent of significant changes o the data analysis software.

@ 1996 Blackwell ScienceLtd., Global ChangeBiology,2, 169-182

~ 2.0

~C/)'E~ 1.5.:;!;Q)0)~ 1.0.1:=!3


9/14

MEASURING CARBON SEQUESTRATION BY EDDY COVARIANCE 177-

':-(/)C)lE

aE

.E:Q)0>c:~

.c:(.)xQ)

(\IO()

25 1.520

1.0 c .-3(1).:.-

1510 0.550 0.0

300200100

0

2-()~..-.0(/)

10(a) CO2 exchange (solid line) and

and (b) net radiation (fine

17-19 January 1993showingefflux during turbulent

-1

18.0 18.5 19.0 19.5 20.0January 1993

during three summer afternoons was 15.5 ~mol m-2 s-I,while the canopy photosynthesis estimated by aggregat-ing leaf~chamber observations was 12.5 ~ol m-2 S-I(Wofsy et al. 1993).

1000

800'E~


10/14

17R M.L. GOULDEN et a

0.0 0.2 0.4 0.6u* (m 5-1)

0.8 0 2 3 4 5Duration well mixed (h)

6

Fig. 11 (a) Fraction of expected respiration measured bynocturnal eddy covariance (filled circles) and by CO2 storagebeneath 30 m (open triangles) as a function of friction velocity.(b) Fraction measured by eddy coyariance and by storage as afunction of the recent history of mixing (hours since u* exceeded0.17m s-l with zero corresponding to periods that remain poorlymixed). (c) Fraction of expected soil respiration measured bythe chambers as a function of above-canopy friction velocity. (d)Fraction measured by the chambers as a function of the recenthistory of mixing. Expected respiration was calculated as afunction of soil temperature based on (a, b) NEE and (c, d)chamber observations. Fraction of expected respiration wasobserved efflux divided by expected respiration. Points aremeans =!=1 tandard deviation of (a) 60 observations (b) 200-{;0observations (c) 210 observations, and (d) 300-100 observations,recorded from 10 June 1992 o 16 Sept. 1992.

communication; P.G. Jarvis, personal communication).However, Grace 0. Grace, personal communication)observed that storage quantitatively offset the reductionin eddy flux during calm nocturnal periods in a trop-ical forest.

The causeof flux underestimation during stable periodshas not been dentified. One possibility is that CO2leavesthe forest in draining cool air that subsequently mixesupwards away from the tower. Alternatively, CO2 mayleave the forest in fluctuations that are too small or tooshort to be resolved with the available instrumentation.A third possibility is that the flux calculation is inadequatefor calm nocturnal periods, and a longer averaging timeor a different detrending algorithm is required due tothe dominance of sporadic mixing events.

The efflux of CO2 during summer nights becomesinsensitive to atmospheric turbulence at u* > 0.17 m S-l(Fig. 11a).A critical question is whether eddy covarianceprovides an accurate measure of ecosystem respirationduring these periods. The flux of CO2 during windynights is insensitive to net radiation (Fig. 12a), to thetemperature gradient beneath the canopy (Fig. 12b , andto the temperature gradient above the canopy (Fig. 12c),establishing that there is no apparent difference betweenperiods that are thermally stratified to those that areunstratified. Since there is no apparent selective errorbetween daytime periods that are neutral and those thatare unstable, we conclude that night-time observationsat u* > 0.17 m s-l are reliable. The contention that fluxmeasurements n windy .dark periods are not systematic-ally different from those in light periods is supported bythe observation that the relationship between u* andCO2 flux does not vary from day to night during thewinter (Fig. 8).

We correct for the selective underestimation of respira-tion at u* < 0.17 m s-l by substituting the respirationpredicted from soil temperature (Fig. 5) for the observedflux. The buildup of carbon dioxide beneath 30 m duringcalm periods is relatively small (Fig. 11b), minimizingproblems with double counting. The replacement of dataduring calm periods increases the calculated annualrespiration by 0.5-1.0 t C ha-l. This correction is largelyresponsible for a revision of our estimate of 1991 carbonsequestration from 3.7 t C ha-l to 2.8 t C ha-l (Gouldenet al. 1996),a somewhat larger effect than the 0.5 t C ha-lanticipated by Wofsy et al. (1993).

resolved by the sonic anemometer, and increased mpor-tance of storage.

The observations of nocturnal CO2exchangeat HarvardForest ndicate a selective underestimation of flux duringcalm periods. There is a reduction in the measuredvertical flux of CO2 at 30 m during poorly mixed periods(u* < 0.17 m s-I, Fig. 11a) hat is not due to a reductionin the flux of CO2 rom the soil (Fig. 11c).This discrepancycannot be explained entirely by increased storage. Therate of CO2 accumulation during calm periods is only20-300;0 f the eddy flux during windy intervals (Fig. 11a).The flux of CO2 at 30 m immediately following theresumption of mixing is not unusually high (Fig. 11b) aswould be expected with flushing of accumulated CO2.Carbon dioxide evidently escapes rom the forest duringpoorly mixed periods by an undetected route. Similarpatterns have been observed in boreal forests(M.L. Goulden, personal observation; T.A. Black, personal

Comparison of chamber and eddy-flux measurements ofecosystem respirationThe average respiration measured by eddy covarianceduring windy nights in summer 1992 was 4.2 ~molm-2 s-I. The chamber measurements during this periodindicated a rate of ecosystem espiration at a comparable@ 1996 Blackwell Science Ltd., Global ChangeBiology,2, 169-182


11/14

MEASURING CARBON SEQUESTRATION BY EDDY COVARIANCE 179ity of nocturnal eddy flux measurements during well-mixed periods. Additionally, a selective systematic under-estimation of NEE during windy periods is difficult toreconcile with observations of wood production. A 30%underestimation of respiration during mixed periodswould result in a revised annual net carbon sequestrationof less than 1 t C ha-l. Allometric observations in hard-wood stands at Harvard Forest ndicate wood productionof = 2.5 t C ha-l (Aber et al. 1993), mplying an unlikelyannual loss of more than a ton of Cha-l from the soil.

A possible explanation is that the chamber measure-ments overestimate respiration. The most likely candidateis, soil respiration; stem respiration is insufficient toaccount for the discrepancy (Table 1 , and the rate of leafrespiration is about 10% of the maximum rate of canopyphotosynthesis (Fig. 7), a pattern consistent with observa-tions on individual leaves (Aber et al. 1995). Measure-ments of soil trace-gas exchange are extremely difficultto verify. The rates of respiration measured with the openchambers are comparable to those measured around thetower using a closed chamber (LiCor 6000-09, LincolnNE; data not shown), but higher than those measuredwithin a few kIn of the tower using a static chamber(Kicklighter et al. 1994; 3.1 ~mol m-2 s-I at 17 C).

An alternative explanation is that both sets of respira-tion measurements are correct, but that different areaswere measured. A survey of mid-day soil respirationduring August and September 1993using a closed cham-ber (LiCor 6000-09) found considerable heterogeneityaround the tower (data not shown). For example, respira-tion at 19-21 C in an extensive poorly drained area tothe north-west of the tower averaged 3.8 ~mol m-2 S-I,while respiration in an upland area to the south-westaveraged 7.5 ~ol m-2 s-I. The depth and width ofthe nocturnal footprint from ground-level sources isunknown. Without this information it is not possible torule out spatial heterogeneity as a cause of the discrep-ancy. Ensuring that point measurements (e.g. chamberfluxes, foliar chemistry) are representative of the areasampled by micrometeorological observations representsa major challenge to interdisciplinary investigations. Theestablishment of a clear set of guidelines for distributingpoint measurements around eddy flux sites will be anecessary component of a successful flux network.

E

0)0>0:"ux

N

0)

0>

.100 -4080 -60Net Radiation (w m-1

net ecosystem exchange during well-mixed> 0.17 m s-I) as a function of (a) net radiation (b)

T at 15 m). Points are means =!=1 tandardof (a) 42 hourly observations (b) 26 hourly

and (c) 26 hourly observations, recorded from 1016 Sept. 1992.

Average nocturnal net ecosystemexchange during well-periods between 10 June 1992and 16 Sept. 1992 n = 660,soil and air temperature = 17.5C) compared to component

17 C. Soil respiration was calculated as Fluxm-2 s-I) = exp (0.14 + 0.079 " T), n = 2450. Stemwas calculated as Flux (l1Inol m-2 s-I) = exp (-0.062 " T), n = 965. Leaf respiration was based onservations on 10 eafy branches during August 1992.

Nocturnal NEEwell-mixed periods Aggregated openchamber

1.8 ~rnoO.3~o4.4 ~rno6.5 ~rno.2 I!mol m-2 5-1 Sampling uncertaintyOne of our goals s to determine the annual rate of carbon

sequestration (annual net ecosystem production, NEP).We assume hat NEP is equal to the annual net exchangeof CO2 with the atmosphere (integrated NEE) sincethe site has not burned during the study, and carbonexchanges in forms other than CO21and by processesother than turbulent transport, are likely small.

6.5 l1Inol m-2 S-l (Table 1). Both numbersre based on a large number of observations and thisiscrepancy s beyond the expected errors. The discussion

n the previous section provides evidence for the reliabil- 1996 Blackwell ScienceLtd., Global ChangeBiology,2, 169-182

I m-2 s-I m-2 s-I m-2 s-I m-2 s-


12/14


321O

-1-2

",;-o~Uu(1)->~

:.-(1)roC)"50:E ro~.c:

u~ w

0.30.2

~ ~ 0.1to()~ :::- 0.0tO ~o~() 0) -0.1

E -0.2

Fig, 13 (a) Cumulative net carbon exchange (points) and 90%confidence interval (shading) from Day 301, 1993 to Day 300,1994. Points are at intervals with 4 complete days of observations.(b) Increase through the year in 90% confidence interval due tosampling uncertainty. The annual net exchange during 1994was -2.1 t C ha-l, with an accumulated sampling uncertainty of:!:0.3 t C ha-l.

be collected and used to fill periods when flux observa-tions are unavailable. The uncertainty caused by fillingmissing periods as a function of the physical environmentis smaller than that causedby assuming that the days withobservations are representative. Second, the samplingstrategy should minimize the length of intervals withoutflux data. The correlation at Harvard Forest between thecarbon balance on days that are separated by less than aweek is high (i.e. the lag correlation, ,-2= 0.6-{!.7),whereasthe correlation between days that are separated by longerthan two weeks is low (,-2 = 0.2--0.3).The correlationbetween days separatedby more than a week is especiallylow in spring and fall when the carbon balance ischanging rapidly. Long gaps create disproportionatelylarge uncertainty, whereas short gaps (a few days) spreadthroughout the year are acceptable.This pattern supportsthe deployment of unattended monitoring systems, pro-vided that malfunctions can be repaired quickly. Periodicbreakdowns of unattended systems are inevitable, butthe uncertainty caused by these gaps is offset by theadvantage of year-round monitoring.

Conclusions1 Long-term eddy covariance provides an effective tech-nique for measuring the hourly; daily, monthly, andannual rates of carbon exchangeby terrestrial ecosystems.The long term precision of the approach is very good

Our current approach for calculating long-term carbonsequestration involves: (i) Estimating CO2 exchange as afunction of soil temperature (Fig. 5) for dark periodswhen u* < 0.17 m s-l or when flux measurements areunavailable. Estimating CO2 exchange as a function ofPPFD (Fig. 7) and soil temperature (Fig. 5) for summerlight periods when flux measurements are unavailable.(ii) Dividing the record of hourly CO2 exchange intointervals that encompass our days with complete obser-vations. (Most of the intervals are 4 days, but longerperiods may be required due to equipment malfunction.)(iii) Averaging by hour within each nterval. (iv) Summingto calculate the carbon balance of each interval. (v)Summing through the year assuming for each intervalthat the carbon balance calculated from the 4-days ofobservations is representative of the complete interval(Fig. 13a). Sampling uncertainty arises n step 1 when anempirical relationship is used to fill missing periods, andin step 5 when the carbon balance determined from 4days of observations is assumed representative of alonger interval.

We estimated this uncertainty using a Monte Carlomethod to simulate the sampling process.We assembleda total of eight populations of .daily carbon balances thatwere intended to represent the typical patterns of day-to-day variation. Each population consisted of the dailycarbon balances of 30 nearly consecutive days drawnfrom observations during 1994,with a pair of populationsrepresenting each season. One population within eachseason was composed of the actual measurements ofdaily carbon balance, and the other of the simultaneouscarbon balances calculated from the empirical relation-ships (Figs 5, 7). For each interval we sampled from theappropriate population one hundred randomly selectedsequencesof the interval length. The average of 4 ran-domly selected days within each sequencewas comparedwith the true sequenceaverage to simulate the samplingerror for each of the 100 cases.The sampling error foreach of the 100 caseswas accumulated through the yearand sorted to determine the 5th and 95th percentiles.

The integrated carbon sequestration from Day 301,1993 hrough Day 300, 1994was 2.1 t C ha-l y -1 (Fig. 13a,negative cumulative NEE refers to a loss of carbon fromthe atmosphere and a positive net sequestration of carbonby the forest), and the 90% confidence interval due tosampling uncertainty was :!:0.3 t C ha-l y-1 (Fig. 13b).About half of the annual sampling uncertainty resultedfrom missing flux and meteorological observations inJuly following lightning damage (Fig. 3). Missing observa-tions in winter did not cause appreciable uncertaintybecauseCO2 exchange was relatively low and consistentfrom day to day.

Two strategies are important for reducing samplinguncertainty. First, a continuous set of climate data may



13/14

MEASURING CARBON SEQUESTRATION BY EDDY COVARIANCE 181(:!:5%). Long-term eddy covariance is particularly wellsuited for quantifying the effects of stress, climate, andphenology on carbon exchange, and for developing andtesting mechanistic models and remote-sensing algo-rithms.2 We observe a selective systematic underestimation offlux during calm (u* < 0.17 m S-I) nocturnal periods atHarvard Forest, for which we compensate using anestimation of ecosystem respiration based on temper-ature. Additional work is needed to fully establish thereliability of eddy covariance during nocturnal periods. Inparticular, we are unable to fully account for a discrepancybetween the ecosystem respiration at Harvard Forestmeasured as NEE during mixed nocturnal periods, andthat measured using chambers.3 The uncertainty in the measurementof carbon exchangeat Harvard Forest due to uniform systematic errors is-20 to 0%. The annual carbon sequestration in 1994 was2.1 t C ha-l y -1 with a 90% confidence interval due tosampling uncertainty of :!:0.3 t C ha-l y-1. Samplinguncertainty is reduced by estimating the flux as a functionof the physical environment during periods when directobservations are unavailable, and by minimizing thelength of intervals without flux data~ especially duringspring and fall. The combined effects of uniform system-atic errors, sampling uncertainty, and the estimation ofrespiration during calm nocturnal periods leads to anoverall confidence interval for carbon sequestration in1994 of --0.3 o +0.8 t C ha-l y-1.

AcknowledgementsThis work was supported by grants to Harvard Universityfrom the U.S. National Science Foundation (BsR-89l9300, DEB-9411975), the U.S. National Aeronautics and Space Administra-tion (NAGW-3082, NAG5-2253), the U.S. Department of Energy(DOE) through the North-east Regional Centre of the NationalInstitute for Global Environmental Change (NIGEC, DOECooperative Agreement No. DE-FCO3-90ER6l0l0) and throughthe terrestrial carbon processes program (DE-FGO2-95ER62002),and by Harvard University (Harvard Forest and Division ofApplied Sciences). We thank A. Goldstein, M. Potosnak,R. Ditzion, D. sutton and s. Roy for technical support, K. Mooreand D. Fitzjarrald for helpful discussions, and D. Baldocchi andR. Valentini for organizing the La Thuile meeting.

References

Baldocchi DD, Hicks BB, Meyers TP (1988)Measuring biosphere-atmosphere exchanges of biologically related gases withmicrometeorological methods. Ecology, 9, 1331-1340.

Baldocchi DD, Meyers TP (1991) Trace gas exchange above thefloor of a deciduous forest. 1. Evaporation and CO2 flux.Journal of Geophysical esearch, 6D, 7271-7285.

Baldocchi DD, Valentini R, Running S, Oechel W, Dahlman R(1996)Strategies or measuring and modelling carbon dioxideand water vapor fluxes over terrestrial ecosystems. GlobalChangeBiology,2, 159-168.

Denmead OT, Bradley EF (1985) Flux-gradient relationships ina forest canopy. In. The Forest-Atmospherenteraction (edsHutchinson BA, Hicks BB), pp. 421-44. Reidel , Dordecht.

Fan S-M, Goulden ML, Munger JW, Daube BC, Bakwin PS,Wofsy SC, Amthor JS, Fitzjarrald DR, Moore KE, Moore TR(1995) Environmental controls on the photosynthesis andrespiration of a boreal lichen woodland: A growing seasonof whole-ecosystem exchange measurements by eddycorrelation. Oecologia, 02, 443-452.

Field CB, Ball JT, Berry, JA (1989)Photosynthesis: principles andfield techniques. In: Plant PhysiologicalEcology;Field methodsand instrumentation (Pearcy RW, Ehleringer J, Mooney HARundel PW), pp. 209-254. Chapman and Hall, London.

Fitzjarrald DR, Moore KE (1990) Mechanisms of nocturnalexchangebetween the rain forest and the atmosphere. Journalof Geophysical esearch, 5, 16,839-16,850.Foster D (1992) Land-use history (1730-1990) and vegetationdynamics in central New England. Journal of Ecology,76,135-151.

Goulden ML, Munger JW, Fan S-M, Daube BC, Wofsy SC (1996)Carbon dioxide exchange by a temperate deciduous forest:response to interann4al changes in climate. Science, (inreview).Hollinger DY, Kelliher FM, Byers IN, Hunt JE, McSeveny TM,Weir PL (1994) Carbon dioxide exchange between anundisturbed old-growth temperate forest and theatmosphere. Ecology, 5, 134-150.

Jarvis PG, Leverenz JW (1983) Productivity of temperate,deciduous and evergreen forests. In: Physiological lant ecologyIV Encyclopedian plant physiology,NS, Vol. 12D (eds LangeOL, Nobel PS, Osmond CB, Ziegler H), pp. 233-280.Springer, Berlin.

Kaimal JC, Wyngard JC, Izumi Y, Cote OR (1972) Spectralcharacteristics of surface-layer turbulence. Quarterly Journalof the Royal MeteorologicalSociety, 8, 563-589.

Kaimal JC, Gaynor JE, Zimmerman HA, Zimmerman GA (1990)Minimizing flow distortion errors in a sonic anemometer.Boundary-LayerMeteorology, 3, 103-115.

Kelliher FM, Kostner BMM, Hollinger DY, Byers IN, Hunt JE,McSeveny TM, Meserth R, Weir PL, Schulze E-D (1992)Evaporation, xylem sap flow, and tree transpiration in aNew Zealand broad-leaved forest. Agricultural and ForestMeteorology, 2, 53-73.

Kicklighter DW, Melillo JM, Peterjohn WT, Rastetter EB, McGuireAD, Steudler PA, Aber JD (1994) Aspects of spatial andtemporal aggregation in estimating regional carbon dioxidefluxes from temperate forest soils. Journal of GeophysicalResearch, 9, 1303-1315.

Aber]D, Magill A, Boone R, Melillo JM, Steudler PA, Bowden R(1993) Plant and soil responses to three years of chronicnitrogen additions at Harvard Forest. EcologicalApplications,3, 156-166.

Aber JD, Reich PB, Goulden ML (1995) Extrapolating leaf CO2exchange to the canopy: A generalized model of forestphotosynthesis validated by eddy correlation. Oecologia, inpress).

@ 1996 Blackwel\ Science Ltd., Global Change Biology, 2, 169-182


14/14

182 M.l GOULDEN et alreactive nitrogen oxides and ozone in a temperate deciduousforest and a sub-arctic spruce woodland. 1. Measurementsand mechanisms. Journal of Geophysical esearch,n review.

Ohtaki E (1985) On the similarity in atmospheric fluctuations ofcarbon dioxide, water vapor and temperature over vegetatedfields. Boundary-LayerMeteorology, 2, 15-37.

5igmon JT, Knoerr KR, 5haughnessy EJ (1983) Microscalepressure luctuations in a mature deciduous forest. Boundary-Layer Meteorology, 7, 345-358.

5inclair TR, Murphey CE, Knoerr KR (1976) Development andevaluation of simplified models for simulating canopyphotosynthesis and transpiration. Journal of Applied Ecology,13, 813-829.

Verma 58, 8aldocchi DD, Anderson DE, Matt DR, Clement RJ(1986) Eddy fluxes of CO2. water vapor, and sensible heatover a deciduous forest. Boundary-LayerMeteorology, 6, 71-91.

Webb EK, Pearman GI, Leuning R (1980) Correction of fluxmeasurements for density effects due to heat and watervapor transfer. Quarterly Journal Royal MeteorologicalSociety,106, 85-100.

Wofsy 5C, Goulden ML, Munger JW, Fan 5-M, 8akwin PS,Daube 8C, 8assow 5L, 8azzaz, FA (1993) Net exchange ofCO2 n a rnid-Iatitude forest. Science, 60, 1314-1317.

Lee X, Black TA (1994) Relating eddy correlation sensible heatflux to horizontal sensor separation in the unstableatmospheric surface layer. Journal Geophysical Research, 99,18545-18553.

Leuning R, King KM (1992) Comparison of eddy-covariancemeasurements of CO2 fluxes by open- and closed-path CO2analyzers. Boundary-Layer Meteorology, 59, 297-311.

McMillen RT (1988) An eddy correlation technique with extendedapplicability to non-simple terrain. Boundary-LayerMeteorology, 43, 231-245.

Monteith JL (1972) Solar radiation and productivity in tropicaiecosystems. Journal Applied Ecology, 9, 747-766.

Mosier AR (1989) Chamber and isotope techniques. In: Exchangeof trace gasesbetween ecosystemsand the atmosphere (eds AndreaeMO, Schimel DS).J. Wiley & Sons.

Moore CJ (1986) Frequency response corrections for eddycorrelation systems. Boundary-Layer Meteorology, 37, 17-35.

Moore KE, Fitzjarrald DR, Sakai RK, Goulden ML, Munger JW,Wofsy SC (1996) Seasonal variation in radiative and turbulentexchange at a deciduous forest in central Massachusetts.Journal of Applied Meteorology, in press.

Munger JW, Wofsy SC, Bakwin PS, Fan S-F, Goulden ML, DaubeBC, Moore, K, Fitzjarrald D (1996) Atmospheric deposition of


goulden et al 96a

Documents