University of Montana ScholarWorks at University of Montana

Numerical Terradynamic Simulation Group Publications Numerical Terradynamic Simulation Group

11-2001

FLUXNET: A New Tool to Study the Temporal and Spatial Variability of Ecosystem-Scale Carbon Dioxide, Water Vapor, and Energy Flux Densities

Dennis Baldocchi

Eva Falge

Lianhong Gu

Richard Olson

D. Y. Hollinger

See next page for additional authors

Follow this and additional works at: https://scholarworks.umt.edu/ntsg_pubs Let us know how access to this document benefits ou.y

Recommended Citation Baldocchi, D. D., Falge E., Gu L., Olson R., Hollinger D. Y., Running S. W., Anthoni P., Bernhofer Ch., Davis K. J., Evans R., Fuentes J., Goldstein A., Katul G., Law B. E., Lee X., Malhi Y., Meyers T. P., Munger J. W., Oechel W. C., Paw U K. T., Pilegaard K., Schmid H. P., Valentini R., Verma S., Vesala T., Wilson K. B., and Wofsy S. C. (2001). FLUXNET: A New Tool to Study the Temporal and Spatial Variability of Ecosystem- Scale Carbon Dioxide, Water Vapor, and Energy Flux Densities. Bulletin of the American Meteorological Society, 82: 2415-2434, doi: 10.1175/1520-0477(2001)082<2415:FANTTS>2.3.CO;2

This Article is brought to you for free and open access by the Numerical Terradynamic Simulation Group at ScholarWorks at University of Montana. It has been accepted for inclusion in Numerical Terradynamic Simulation Group Publications by an authorized administrator of ScholarWorks at University of Montana. For more information, please contact [email protected]. Authors Dennis Baldocchi, Eva Falge, Lianhong Gu, Richard Olson, D. Y. Hollinger, Steven W. Running, Peter Anthoni, Ch. Bernhofer, Kenneth Davis, Robert Evans, Jose Fuentes, Allen Goldstein, Gabriel Katul, Beverly Law, Xuhui Lee, Yadvinder Malhi, Tilden Meyers, William Munger, Walt Oechel, K. T. Paw U, Kim Pilegaard, H. P. Schmid, Riccardo Valentini, Shashi Verma, Timo Vesala, Kell Wilson, and Steve Wofsy

This article is available at ScholarWorks at University of Montana: https://scholarworks.umt.edu/ntsg_pubs/112 FLUXNET: A New Tool to Study the Temporal and Spatial Variability of Ecosystem-Scale Carbon Dioxide, Water Vapor, and Energy Flux Densities

Dennis Baldocchl,® Eva Falge,’’ Llanhong Gu,® Richard Olson,^ David Floiilnger,'’ Steve Running,® Peter Anthonl,f Ch. Bernhofer,^ Kenneth Davis,'' Robert Evans," Jose Fuentes; Allen Goldstein,® Gabriel KatuI,) Beverly Law,f Xuhul Lee," Yadvlnder Malhl,' Tllden Meyers,"' William Munger," Walt Oechel,° K. T. Paw U, p Kim Pllegaard,®i FI. P. Schmld,'^ RIccardo Valentlnl," Shashi Verma,* TImo Vesala," Kell Wilson,"' and Steve Wofsy" ABSTRACT

FLUXNET is a global network of micrometeorological flux measurement sites that measure the exchanges of car­ bon dioxide, water vapor, and energy between the biosphere and atmosphere. At present over 140 sites are operating on a long-term and continuous basis. Vegetation under study includes temperate conifer and broadleaved (deciduous and evergreen) forests, tropical and boreal forests, crops, grasslands, chaparral, wetlands, and tundra. Sites exist on five con­ tinents and their latitudinal distribution ranges from 70°N to 30°S. FLUXNET has several primary functions. First, it provides infrastructure for compiling, archiving, and distributing carbon, water, and energy flux measurement, and meteorological, plant, and soil data to the science community. (Data and site information are available online at the FLUXNET Web site, http://www-eosdis.oml.gov/FLUXNET/.) Second, the project supports calibration and flux intercomparison activities. This activity ensures that data from the regional networks are intercomparable. And third, FLUXNET supports the synthesis, discussion, and communication of ideas and data by supporting project scientists, workshops, and visiting scientists. The overarching goal is to provide infor­ mation for validating computations of net primary productivity, evaporation, and energy absorption that are being generated by sensors mounted on the NASATerra satellite. Data being compiled by FLUXNET are being used to quantify and compare magnitudes and dynamics of annual ecosystem carbon and water balances, to quantify the response of stand-scale carbon dioxide and water vapor flux densities to controlling biotic and abiotic factors, and to validate a hierarchy of soil-plant-atmosphere trace gas ex­ change models. Findings so far include 1) net CO^ exchange of temperate broadleaved forests increases by about 5.7 g C m“^ day^ for each additional day that the growing season is extended; 2) the sensitivity of net ecosystem CO^ exchange to sunlight doubles if the sky is cloudy rather than clear; 3) the spectrum of CO^ flux density exhibits peaks at timescales of days, weeks, and years, and a spectral gap exists at the month timescale; 4) the optimal temperature of net CO^ exchange varies with mean summer temperature; and 5) stand age affects carbon dioxide and water vapor flux densities.

^ESPM, University of California, Berkeley, Berkeley, California. "Department of Earth and Planetary Sciences, Harvard University, Cam­ *^Pflanzenokologie, Universitat Bayreuth, Bayreuth, Germany. bridge, Massachusetts. Environmental Science Division, Oak Ridge National Laboratory, Oak "Department of Biology, San Diego State University, San Diego, Califomia. Ridge, Tennessee. PLand, Air and Water Resources, University of California, Davis, Davis, ^USDA Forest Service, Durham, New Hampshire. Califomia. ^School of Forestry, University of Montana, Missoula, Montana. Elant Biology and Biogeochemistry Department, Risoe National Labo­ 4lichardson Hall, Oregon State University, Corvallis, Oregon. ratory, Roskilde, Denmark. *Technische Universitat Dresden, IHM Meteorologie, Tharandt, Germany. ‘‘Department of Geography, Indiana University, Bloomington, Indiana. ^Department of Meteorology, The Pennsylvania State University, Univer­ DISAERI, Universita de Tuscia, Viterbo, Italy. sity Park, Pennsylvania. ‘School of Natural Resource Sciences, University of Nebraska at Lincoln, ‘Department of Environmental Science, University of Virginia, Lincoln, Nebraska. Charlottesville, Virginia. "Department of Physics, University of Helsinki, Helsinki, Finland. ^School of the Environment, Duke University, Durham, North Carolina. Corresponding author address: Dennis Baldocchi, Department of Environ­ •^School of Forestry, Yale University, New Haven, Connecticut. mental Science, Policy and Management, Ecosystems Science Division, 151 Institute of Ecology and Resource Management, University of Edinburgh,Hilgard Hall, University of Califomia, Berkeley, Berkeley, CA 94720. Edinburgh, United Kingdom. E-mail: [email protected] ■"NOAA/Atmospheric Turbulence and Diffusion Division, Oak Ridge,In final form 27 Eebmary 2001. Tennessee. ©2001 American Meteorological Society

Bulletin of the American Meteorological Society 2415 1. Introduction growth of atmospheric CO^. Typical values of inter­ annual variability are on the order of 0.5-3.0 ppm y r f Large-scale, multi-investigator projects have been On a mass basis, these values correspond with a range the keystone of many scientific and technological ad­between 1 and 5 gigatons C yr '. Potential sources of vances in the twentieth century. Physicists have year-to-year changes in atmospheric CO^ remain a leamed much about the structure of an atom’s nucleustopic of debate. Studies of atmospheric and dis­ with particle accelerators. Astrophysicists now peer tributions imply that the terrestrial biosphere plays an deep into space with the Hubble Space Telescope andimportant role in this interannual variability (Ciais an array of radio telescopes. Molecular biologists are et al. 1995; Keeling et al. 1996). Sources of this vari­ decoding the structure of our DNA with the Humanability have been attributed to El Nino/Ea Nina events, Genome project. which cause regions of droughts or superabundant Ecosystem scientists need a tool that assesses the rainfall (Conway et al. 1994; Keeling et al. 1995), and flows of carbon, water, and energy to and from thealterations in the timing and length of the growing terrestrial biosphere across the spectrum of time- and season (Mynenietal. 1997a,b; Randersonetal. 1997). space scales over which the biosphere operates (see Rising levels of CO^, and other greenhouse gases, Running et al. 1999; Canadell et al. 2000). Similarly,are of concem to scientists and policy makers because atmospheric scientists need a tool to quantify surfacethey trap infrared radiation that is emitted by the energy fluxes at the land-atmosphere interface, asearth’s surface. Potential consequences of elevated these energy fluxes influence weather and climateCO 2 concentrations include a warming of the earth’s (Betts et al. 1996; Pielke et al. 1998). In this paper, wesurface (Hansen et al. 1998), melting of polar icecaps report on such a tool. It consists of a global array of and a rising sea level, and an alteration of plant and micrometeorological towers that are measuring fluxecosystem physiological functioning and plant com­ densities of carbon dioxide, water vapor, and energyposition (Amthor 1995; Norby et al. 1999). between vegetation and the atmosphere on a quasi- With regard to plants and ecosystems, short-term continuous and long-term (multiyear) basis. The ob­experiments with elevated CO^ show increased rates jectives of this paper are to introduce the FLUXNET of photosynthesis and plant growth and lowered sto- project and describe its rationale, goals, measurementmatal conductance (Drake et al. 1996). The long-term methods, and geographic distribution. We also presentsustainability of enhanced rates of growth by plants, a sampling of new results that are being generatedhowever, depends on nutrient and water availability, through the project. The intent is to show how infor­ temperature, and light competition (Ceulemans and mation from this network can aid ecologists, meteo­ Mousseau 1994; Norby etal. 1999). Warmer tempera­ rologists, hydrologists, and biogeochemists to tures, associated with elevated CO^, promote increased understand temporal and spatial variations that are as­rates of respiration when soil moisture is ample. But sociated with fluxes of carbon and water between theif climate warming is associated with drying, there can biosphere and atmosphere. be reduced assimilation and lower rates of soil/root respiration in temperate ecosystems. In the tundra and boreal forest, warming and drying can lower water 2. What is the problem? tables, exposing organic peat to air and increasing its rate of respiration (Goulden et al. 1998; Eindroth et al. Over the past century, the states of the earth’s at­ 1998; Oechel etal. 2000). mosphere and biosphere have experienced much How evaporation is impacted by elevated CO^ de­ change. Since the dawn of the industrial revolution, thepends on the scale of study and concurrent changes in mean global CO^ concentration has risen from aboutleaf area index, canopy surface conductance, and the 280 ppm to over 368 ppm (Keeling and Whorf 1994;depth of the planetary boundary layer. Theory indi­ Conway et al. 1994). The secular rise in atmosphericcates that complex feedbacks among stomatal conduc­ carbon dioxide concentrations is occurring due totance, im­ leaf temperature, and the air’s vapor pressure balances between the rates that anthropogenic anddeficit cause responses at the leaf scale to differ from natural sources emit CO^ and the rate that biosphericthat of the canopy scale. For instance, a potential re­ and oceanic sinks remove CO^ from the atmosphere. duction of leaf transpiration, by stomatal closure, is Superimposed on the secular trend of CO^ is a recordcompensated for in part by the entrainment of dry air of large interannual variability in the annual rate offrom above the planetary boundary layer and elevated

2416 Vol. 82, No. 1 1, November 2001 leaf temperatures and humidities (Jacobs and deBruinand O2 concentration and wind (Tans et al. 1990; Ciais 1992). Hence, canopy evaporation may not be reducedetal. 1995; Denning etal. 1996; Fan etal. 1998). This to the same degree as leaf transpiration in an elevatedapproach is subject to errors due to the sparseness of

CO2 world. the trace gas measurement network, their biased place­ In the meantime, the composition of the land sur­ment in the marine boundary layer, and the accuracy face has changed dramatically to meet the needs of theof the atmospheric transport models (Tans etal. 1990; growing human population. Many agricultural landsDenning et al. 1996; Fan et al. 1998). have been transformed into suburban and urban land­Instruments mounted on satellite platforms view scapes, wetlands have been drained, and many tropi­the earth in total. Consequently, satellite-based instru­ cal forests have been logged, bumt, and converted toments offer the potential to evaluate surface carbon pasture. In contrast, abandoned farmland in the north­fluxes on the basis of algorithms that can be driven by east United States and Europe is retuming to forestreflected and emitted radiation measurements as those societies become more urban. Changes in land(Running et al. 1999; Cramer et al. 1999). This ap­ use alter the earth’s radiation balance by changing itsproach, however, is inferential, so it is dependent on albedo, Bowen ratio (the ratio between the flux den­the accuracy of the model algorithms, the frequency sities of sensible and latent heat exchange), leaf areaof satellite images, and the spectral information con­ index, and physiological capacity to assimilate carbontained in the images. and evaporate water (Betts et al. 1996; Pielke et al. At the landscape to regional scale we can use in­ 1998). Changing a landscape from forest to agricul­struments mounted on aircraft (Crawford et al. 1996; tural crops, for instance, increases the surface’s albedoDesjardins et al. 1984, 1997) or boimdary layer bud­ and decreases the Bowen ratio (Betts et al. 1996);get methods (Denmead et al. 1996; Fevy et al. 1999; forests have a lower physiological capacity to assimi­ Yi et al. 2000) to assess carbon and water fluxes. late carbon and a lower ability to transpire water, as Aircraft-based eddy covariance flux density measure­ compared to crops (Kelliher et al. 1995; Baldocchi andments give good information on spatial pattems of Meyers 1998). A change in the age structure of for­carbon and water fluxes across transects to tens to ests due to direct (deforestation) or indirect (climate-hundreds of kilometers. However, they do not provide induced fires) disturbance alters its ability to acquireinformation that is continuous in time, nor do they carbon and transpire water (Amiro et al. 1999; Schulzeprovide insights on the physiological mechanisms that et al. 1999). govem carbon and water fluxes. Boundary layer bud­ The issues identified above all require informationget methods can be implemented using tall towers, on fluxes of carbon, water, and energy at the earth’stethered balloons, or small aircraft. This method pro­ surface and how these fluxes interact with the physi­vides information on spatially integrated fluxes of tens cal climate and physiological functioning of plants andto hundreds of square kilometers, but it can be applied ecosystems. In the next section we identify numerousonly during ideal meteorological conditions (Denmead ways to obtain this information and illustrate how aet al. 1996). Routine application of this method is ham­ network of long flux measurement sites can providepered by a lack of data on entrainment and horizontal particularly useful information to study these complexadvection. problems. One measure of carbon flux “ground truth” can be provided by biomass surveys (Kauppi et al. 1992; Gower et al. 1999). However, biomass surveys pro­ 3. How to study biosphere CO^ vide information on multiyear to decadal timescales, exchange? so they do not provide information on shorter-term physiological forcings and mechanisms. Furthermore, Study of the earth’s biogeochemistry and hydrol­forest inventory studies are labor intensive and are ogy involves quantifying the flows of matter in andinferential estimates of net carbon exchange. Such out of the atmosphere. Numerous techniques exist forstudies rarely measure growth of small trees and below- studying biosphere-atmosphere CO^ exchange, eachground allocation of carbon. Instead biomass surveys with distinct advantages and disadvantages. At thecommonly assume that a certain portion of carbon is continental and global scales, scientists assess carbonallocated below ground (Gower et al. 1999). Soil car­ dioxide sources and sinks using atmospheric inversionbon surveys can be conducted too but, like biomass models that ingest information on fields of CO surveys, they require long intervals to resolve de­

Bulletin of the American Mefeorological Society 2417 tectable differences in net carbon uptake or loss forest ecologists in their interpretation of forest bio­ (Amundson et al. 1998). mass inventories, by providing information on fluxes The eddy covariance method, a micrometeorologi­from carbon pools and how these fluxes vary on a sea­ cal technique, provides a direct measure of net carbonsonal and interannual basis. and water fluxes between vegetated canopies and the atmosphere (Baldocchi et al. 1988; Token and Wichura 1996; Aubinet et al. 2000). With current technology,4. History: What has been done? the eddy covariance method is able to measure mass and energy fluxes over short and long timescales (hour, Micrometeorologists have been measuring CO^ days, seasons, and years) with minimal disturbance toand water vapor exchange between vegetation and the the underlying vegetation. Another attribute of theatmosphere since the late 1950s and early 1960s. Yet, eddy covariance method is its ability to sample a rela­ only recently have we had the technology available tively large area of land. Typical footprints have lon­ that enables us to make continuous flux measurements gitudinal length scales of 100-2000 m (Schmid 1994).at numerous sites. If deployed as a coordinated network of measurements The earliest reported micrometeorological mea­ sites, the eddy covariance method has the potential ofsurements of CO^ exchange were conducted by Inoue quantifying how whole ecosystems respond to a spec­(1958), Lemon (1960), Monteith and Sziecz (1960), trum of climate regimes, thereby expanding the spa­and Denmead (1969). These studies employed the tial scope of this method. flux-gradient method over agricultural crops during The eddy covariance method is not without weak­the growing season. By the late 1960s and early 1970s, nesses. Its application is generally restricted to peri­ a few scientists started applying the flux-gradient ods when atmospheric conditions are steady and tomethod over natural landscapes. Several flux-gradient locations with relatively flat terrain and vegetation thatstudies of CO^ exchange were conducted over forests extends horizontally about 100 times the sampling(Denmead 1969; Baumgartner 1969; Jarvis etal. 1976), height. Consequently, the method is not suitable forand one team ventured to assess the annual carbon measuring fluxes in rough mountainous terrain or balancenear of a salt marsh (Houghton and Woodwell 1980). distinct landscape transitions such as lakes. As more data were collected it became evident that Nevertheless, the strengths of the eddy covariance the flux-gradient method was suffering from major method far outweigh its weaknesses, as it can provide deficiencies, when applied over tall forests. One flaw information that can aid in inferential carbon fluxesarose from large-scale transport in the roughness being determined by the global inversion model com­sublayer, which caused local eddy exchange coeffi­ munity, remote sensing scientists, and forest ecolo­cients to be enhanced relative to estimations based on gists. For example, data from an eddy covarianceMonin-Obukhov scaling theory (Rapauch and Legg measurement site have the potential to complement 1984; Kaimal and Finnigan 1994). The flux-gradient the global flask sampling network in a very direct way. method also failed to provide reliable means of evalu­ Historically, the global flask sampling network has ating eddy exchange coefficients at night. been concentrated at marine locations (Tans et al. Routine application of the eddy covariance method 1990). Recent studies using tall towers show that therewas delayed until the 1980s, when technological ad­ are times during the day, when either local surfacevances in sonic anemometry, infrared spectrometry, fluxes are small or convective mixing is vigorous, thatand digital computers were made. Initial studies CO^ concentrations measured in the atmospheric sur­were conducted over crops (Anderson et al. 1984; face layers are very close to the mean boundary layer Desjardins et al. 1984; Ohtaki 1984), forests (Verma concentration (Bakwin et al. 1998; Potosnak et al.etal. 1986), and native grasslands (Verma etal. 1989) 1999). High-precision CO^ concentration measure­for short intense periods during the peak of the grow­ ments and surface carbon flux measurements, madeing at season. By the late 1980s and early 1990s, the fur­ eddy covariance measurement sites, have the poten­ther technological developments, such as larger data tial to expand the density of the existing flask network.storage capacity and linear and nondrifting instru­ Data from a network of eddy covariance measurementments, enabled scientists to make defensible measure­ sites can also be used to improve and validate the al­ments of eddy fluxes for extended periods. Wofsy et al. gorithms being used by remote sensing scientists and(1993), at Harvard Forest, and Vermetten etal. (1994), ecosystem modelers. And finally, such data can aidin the Netherlands, were among the first investigators

2418 Vol. 82, No. 1 1, November 2001 to measure earbon dioxide and water vapor fluxes eon-ongoing tower studies and initiating many new stud­ tinuously over a forest over the eourse of a year with ies. With the sueeess of the European and Ameriean the eddy eovarianee method. Spurred by these tworegional networks and antieipation of the Earth Ob­ pioneering studies, a handful of other towers were soonservation Satellite {EOS/Terra), the National Aero- established and operating by 1993 in North Amerieanauties and Spaee Administration (NASA) deeided, [Oak Ridge, Tennessee (Greeo and Baldoeehi 1996);in 1998, to fund the global-seale FEUXNET projeet, Prinee Albert, Saskatehewan, Canada (Blaek et al. 1996)]as a means of validating EOS produets. and Japan (Yamamoto et al. 1999), and in Europe by 1994 (Valentini et al. 1996). This era also heralded the start of longer, quasi-eontinuous multi-investigator5. What is being done? experiments. Noted examples inelude the Boreal Eeosystem-Atmosphere Study (Sellers et al. 1997) The FEUXNET projeet serves as a meehanism for and the Northern Hemisphere Climate-Proeesses uniting the aetivities of several regional and eontinen- Eand-Surfaee Experiment (Halldin et al. 1999).tal networks into an integrated global network. At The next natural step in the evolutionary develop­present over 140 flux tower sites are registered on the ment of this field was to establish a global network of FEUXNET data arehive. Researeh sites are operating researeh sites. The eoneept of a global network of long­ aeross the globe (Fig. 1) in North, Central, and South term flux measurement sites had a genesis as early asAmeriea; Europe; Seandinavia; Siberia; Asia; and Affiea. 1993, as noted in the seienee plan of the InternationalThe regional networks inelude AmeriFlux [whieh in- Geosphere-Biosphere Program/Biospherie Aspeets of eludes Earge-seale Biosphere-Atmosphere Experi­ the Hydrologieal Cyele (BAHC Core Projeet Offiee ment (EBA) sites in Brazil], CarboEuroflux (whieh 1993). Formal diseussion of the eoneept among thehas subsumed Euroflux and Medeflu), AsiaFlux, and international seienee eommunity oeeurred at the 1995OzFlux (Australia, New Zealand). There are also dis­ Ea Thuile workshop (Baldoeehi et al. 1996). At thisparate sites in Botswana and South Afriea. meeting, the flux measurement eommunity diseussedThe global nature of FEUXNET extends the di­ the possibilities, problems, and pitfalls assoeiated with versity of biomes, elimate regions, and methods that are making long-term flux measurements (e.g., Gouldenassoeiated with the regional networks. For example, et al. 1996a,b; Monererff et al. 1996). After the Ea Thuilesites in the original Euroflux network eonsisted of meeting there was aeeeleration in the establishment of eonifer and deeiduous forests and Mediterranean flux tower sites and regional flux measurement net­shrubland. The European networks also used a stan­ works. The Euroflux projeet started in 1996 (Aubinetdard methodology, based on elosed-path infrared spee- et al. 2000; Yalentini et al. 2000). The AmeriFluxtrometers (Aubinet et al. 2000; Yalentini et al. 2000). projeet was eoneeived in 1997, subsuming the severalBy eontrast, the Ameriean network, AmeriFlux, has

^53*

Legend , ASIAFLUX AmeriFlu* CARBOEROFLUX EOS Volidotion I EUROFLUX EuroSiberiaPlux LBA I MEDEFLU • Other OzNet

Fig . 1. Global map of FLUXNET field sites.

Bulletin of the American Meteorological Society 2419 more diversity in terms of the number of biomes and2) constmcting and analyzing integrated datasets for climates studied and methods used. The American re­ synthesis of field data and for the development and gional network includes sites in temperate conifer and testing of soil-atmosphere-vegetation-transfer deciduous forests, tundra, cropland, grassland, chap­models, arral, boreal forests, and tropical forests. Regarding 3) organizing workshops for data synthesis and model methodology, open- and closed-path infrared spec­ testing activities, trometers are used to measure CO^ and water vapor4) preparing peer-reviewed research papers and re­ fluctuations. Another unique dimension of AmeriFluxports on FEUXNET activities and analyses, is its inclusion of a tall (400 m) tower (Bakwin et al. 5) funding the execution and analysis of site inter­ 1998; Yi et al. 2000) for study of CO^ flux and con­comparison studies, and centration profiles and boundary layer evolution. 6) providing scientific guidance to the FEUXNET The placement of field sites in FLUXNET has Data and Information System (DIS). been ad hoc, due to limited funding and investigator capabilities, rather than following a predetermined The data archive office is responsible for geostatistical design. So with the current network con­ figuration, we are not capable of measuring fluxes1) compiling and documenting data in consistent from every patch on earth, nor do we intend to. On formats, the other hand, this coordinated network of sites is able2) developing data guidelines and coordinating devel­ to deduce certain information on spatial pattems of opment of the FEUXNET DIS, fluxes by placing research sites across a spectrum of3) scrutinizing datasets with standard quality control climate and plant functional regions (Running et al. and assurance procedures, 1999). Spatially integrated fluxes of carbon and4) wa­ maintaining the FEUXNET Web page for com­ ter can only be constmcted by combining eddy flux munications and data exchange (http://www- measurements with ecosystem and biophysical mod­ eosdis.oml.gov/FEUXNET/), and els and satellite measurements to perform the spatial 5) transferring FEUXNET data and metadata to a integration (Running et al. 1999). long-term archive, currently designated as the Oak The scientific goals of FLUXNET are to Ridge National Laboratory Distributed Active Archive Center. 1) quantify the spatial differences in carbon dioxide and water vapor exchange rates that may be expe­Back up of data and long-term accessibility of the data, rienced within and across natural ecosystems andprovided by FEUXNET DIS, ensures the protection climatic gradients; and extended use of the data by project scientists, as 2) quantify temporal dynamics and variability (sea­well as students and citizens, well into the future. sonal, interannual) of carbon, water, and energy FEUXNET does not fund tower sites directly, but flux densities—such data allow us to examinedepends the upon institutional support associated with the influences of phenology, droughts, heat spells,funding of the AmeriFlux, CarboEuroflux, AsiaFlux, El Nino, length of growing season, and presenceand OzFlux networks (see acknowledgments). or absence of snow on canopy-scale fluxes; and 3) quantify the variations of carbon dioxide and waterMethodology vapor fluxes due to changes in insolation, tempera­ The eddy covariance method is used to assess trace ture, soil moisture, photosynthetic capacity, nutrition,gas fluxes between the biosphere and atmosphere at canopy stmcture, and ecosystem functional type.each site within the FEUXNET community (Aubinet et al. 2000; Valentini et al. 2000). Vertical flux densi­

FEUXNET has two operational components, a ties of CO2 (N) and latent{XE) and sensible heat{H) project office and a data archive office. The project of­ between vegetation and the atmosphere are propor­ fice houses the principal investigator and a postdoctoraltional to the mean covariance between vertical veloc­ scientist. Specific duties of the FEUXNET office includeity {w') and the respective scalar(c') fluctuations (e.g.,

CO2 , water vapor, and temperature). Positive flux den­ 1) communicating with participants to ensure thesities represent mass and energy transfer into the at­ timely submission of data and documentation mosphereto and away from the surface, and negative the data archive. values denote the reverse; ecologists use an opposite

2420 Vol. 82, No. 1 1, November 2001 sign convention where the uptake of carbon by the Aubinet et al. 2000). Ideally the field site should be biosphere is positive. Turbulent fluctuations wereflat, with an extensive fetch of uniform vegetation. In computed as the difference between instantaneous andpractice many of the FLUXNET sites are on undulat­ mean scalar quantities. ing or gently sloping terrain, as this is where native Our community is interested in assessing the netvegetation exists. Sites on extreme terrain, which may uptake of carbon dioxide by the biosphere, not the fluxforce flow separation and advection, are excluded. The across some arbitrary horizontal plane. When the ther­ degree of uniformity of the underlying vegetation var­ mal stratification of the atmosphere is stable or turbu­ ies across the network, too. Some sites consist of lent mixing is weak, material diffusing from leaves andmonospecific vegetation, others contain a mixture of the soil may not reach the reference height in a time species and a third grouping possesses different plant that is short compared to the averaging timeT, thereby functional types in different wind quadrants. All sites violating the assumption of steady-state conditions andhave sufficient fetch to generate an intemal boundary a constant flux layer. Under such conditions the stor­layer where fluxes are constant with height (Kaimal age term becomes nonzero, so it must be added to theand Finnigan 1994). eddy covariance measurement to represent the balance Agricultural scientists mount their sensors on small of material flowing into and out of the soil and veg­poles, while forest scientists use either walk-up scaf­ etation (Wofsy etal. 1993; Moncrieff etal. 1996). Withfolding or low-profile radio towers. The height of the respect to CO^, the storage term is small over short sensors depends on the height of the vegetation, the crops and is an important quantity over taller forests.extent of fetch, the range of wind velocity, and the fre­ The storage term value is greatest around sunrise andquency response of the instruments. To minimize sunset, when there is a transition between respirationtower interference on scaffold towers, investigators and photosynthesis and between the stable noctumalplace their instruments booms so that they point sev­ boundary layer and daytime convective turbulenceeral meters upwind or at the top of the tower. Spatial (Aubinet et al. 2000; Yi et al. 2000). Summed over separation between anemometry and gas analyzers 24 h, the storage term is nil (Baldocchi et al. 2000). depend on whether one uses a closed- or open-path gas Horizontal advection is possible when vegetationsensor. With the closed-path systems, the intake is is patchy or is on level terrain (Lee 1998; Finniganoften near or within the volume of the sonic anemom­ 1999). At present, routine corrections for advectioneters. A delay occurs as air flows through the tubing may require a three-dimensional array of towers to the sensor, which is compensated for with software (Finnigan 1999), rather than the application of a ver­during postprocessing. Some investigators place their tical velocity advection correction (see Lee 1998;gas transducer on the tower in a constant environment Baldocchi et al. 2000). box to minimize the lag time from the sample port and the sensor. Others draw air down long tubes to instru­ ments housed in an air-conditioned hut below the 6. Instrumentation, data acquisition, tower. In either circumstance, flow rates are high and processing (6 L m in ') to ensure turbulent flow and minimize the diffusive smearing of eddies (Aubinet et al. 2000). Typical instrumentation at FLUXNET field sites Open-path gas sensors are typically placed within includes a three-dimensional sonic anemometer, to0.5 m of a sonic anemometer, a distance that mini­ measure wind velocities and virtual temperature, andmizes flow distortion and lag effects (Baldocchi et al. a fast responding sensor to measure CO^ and water2000; Meyers 2001). vapor. Scalar concentration fluctuations are measured Sampling rates between 10 and 20 Hz ensure com­ with open- and closed-path infrared gas analyzers.plete sampling of the high-frequency portion of the Standardized data processing routines are used to com­flux cospectrum (Anderson et al. 1984). The sampling pute flux covariances (Token and Wichura 1996;duration must be long enough to capture low- Moncrieff et al. 1996; Aubinet et al. 2000). frequency contributions to flux covariances, but not Application of the eddy covariance methods in­too long to be affected by diumal changes in tempera­ volves issues relating to site selection, instrumentture, humidity, and CO^. Adequate sampling duration placement, sampling duration and frequency, calibra­and averaging period vary between 30 and 60 min tion and postprocessing (Moore 1986; Baldocchi et al.(Aubinetetal. 2000). Coordinate rotation calculations 1988; Token and Wichura 1996; Moncrieff etal. 1996;of the orthogonal wind vectors{w,u,v) are performed

Bulletin of the American Mefeorological Society 2421 to correct for instrument misalignment and nonlevelfound by comparing open- and closed-path CO^ sen­ terrain. The vertical velocity, w, is rotated to zero, sors. Side-by-side comparisons between open- and allowing flux covariances to be computed orthogonalclosed-path water vapor sensors, on the other hand, are to the mean streamlines. not as good. The absorption and desportion of water Calibration frequencies of gas instruments varyvapor on tubing walls can cause water uptake to be from team to team. With closed-path sensors, investi­ underestimated by 20%. gators are able to calibrate frequently and automati­ cally, such as hourly or once a day. Teams with a Gap filling open-path sensors calibrate less frequently, for ex­ Most clients of eddy flux data, for example mod­ ample, every few weeks. However, a body of accumu­ elers, require uninterrupted time series. It is the intent lating data indicates that calibrating coefficients ofof the micrometeorological community to collect eddy contemporary instruments remain steady within thatcovariance data 24 hours a day and 365 days a year. duration (±5%; Meyers 2001). Scientists using open-However, missing data in the archived records is a path sensors also compare their instrument responses common feature. Gaps in the data record are attributed to an independent measure of CO^ concentration andto system or sensor breakdown, periods when instru­ humidity. Members of this network do not use a uni­ments are off-scale, when the wind is blowing through form standard for calibration CO^, yet. But many of usa tower, when spikes occur in the raw data, when the use CO2 gas standards that are traceable to the standardsvertical angle of attack by the wind vector is too se­ at the Climate Monitoring and Diagnostics Laboratory vere, and when data are missing because of calibration of the National Oceanic and Atmospheric Administra­and maintenance. Other sources of missing data arise tion, the standards of the global flask network. from farming operations and other management activi­ To ensure network intercomparability, the AmeriFluxties (e.g., prescribed bum of grasslands). Rejection cri­ project circulates a set of reference sensors to mem­ teria applied to the data vary among the flux tower bers in the network, and the FLUXNET project spon­ groups. Data might be rejected, when stationarity tests sors the circulation of this set of instruments to sites or integral turbulence characteristics fail (Kaimal and in Europe, Asia, and Australia. Figure 2 shows a typi­Finnigan 1994; Foken and Wichura 1996), when pre­ cal comparison between the roving set of reference cipitation limits the performance of open-path sensors, instrumentation and data from one research group.during sensor calibration, or when spikes occur in in­ Carbon dioxide flux densities measured by the strumenttwo readings. Other criteria used to reject data systems agree within 5% of one another on an hour- include applications of biological or physical con­ by-hour basis. A similar level of agreement has been straints (lack of energy balance closure; Aubinet et al. 2000) and a meandering flux-footprint source area (e.g., Schmid 1994). If the wind is coming from a 10 nonpreferred direction as may occur over mixed stands, a certain portion of the data will need to be screened. b[0] -0.468 b[1] 0.994 In addition, rejection probability for some sites is higher 0 r " 0.961 o during nighttime because of calm wind conditions. .3.E The average data coverage during a year is between • • 65% and 75% due to system failures or data rejection -10 (Falge et al. 2001). Tests showthatthis observed level of data acceptance provides a statistically robust and oversampled estimate of the ensemble mean. So the -20 filling of missing data does not provide a significant source of bias error. -20 -10 0 10 F^,q2 AmeriFlux Reference (^imoi m'^ s'^) b. Accuracy assessment Errors associated with the application of the eddy F ig . 2. A comparison of CO^ flux densities between thecovariance method can be random or biased (Moncrieff FLUXNET reference set of instruments and instruments at the Italian CarboEuroflux site. Here b(0) is the zero intercept andet al. 1996). Random errors arise from sensor noise and b(l) is the slope of the regressive curve, and E is the coefficient the random nature of atmospheric turbulence. Fully of determination. systematic bias errors arise from errors due to sensor

2422 Vol. 82, No. I I November 2001 drift, limited spectral response of instruments, limitedporal dynamics, and response to the environment dur­ sampling duration and frequency, and calibration inger­ the day. We also note that the factors that cause rors. Selectively systematic bias errors arise from mis­ large bias errors associated with noctumal CO^ fluxes application of the eddy covariance method, as whenmay not cause significant errors to water vapor fluxes, advection or storage occurs. Formal discussions on er­since evaporative fluxes are near zero during the night. rors associated with long-term flux measurementsConsequently, studies comparing integrated annual were reported in previous papers by our colleaguesmeasurements of evaporation with lysimeters and (e.g., Goulden et al. 1996a; Moncrieff et al. 1996;watersheds show good agreement (Barr et al. 2000; Aubinet et al. 2000). For clarification, we provide a Wilson et al. 2000). On the other hand, tests of sur­ summary of their findings. Moncrieff et al. (1996)face energy balance closure suggest that turbulent demonstrate how random sampling errors diminishfluxes at some sites are systematically 10% to 30% too greatly as the number of samples increase. The over­ small to close the energy budget, raising the possibil­ all random error using a dataset of 44 days is aboutity that CO 2 fluxes are similarly underestimated 0.4 qmol m^^ s f This figure of merit is close to the (Aubinet et al. 2000). The solution to this problem “flux detection” level that many of us see with our fieldremains unresolved at the time of this writing. measurements. Accumulated over a year, this random error is ±53 g C m^^ y r', a value that agrees with a formal error analysis by Goulden et al. (1996a,b). 7. Recent findings Fully systematic bias errors associated with imper­ fect sensor response, sensor orientation, and sampling As we write this report, 69 site-years of carbon frequency and duration are typically compensated bydioxide and water vapor flux data have been published applying spectral transfer functions to the flux cova­in the literature. Table 1 presents a survey of the an­ riance calculations (Moore 1986; Massman 2000).nual sums of net carbon dioxide exchange. Most sites Depending on instrument configuration, wind speed,are net sinks of CO^ on an annual basis. Magnitudes and thermal stratification, corrections can range fromof net carbon uptake by whole ecosystems range from a few to 50 percent. near zero to up to over 700 g C m^^ y r f Largest val­ Selectively systematic bias errors are more diffi­ ues of carbon uptake are associated with temperate for­ cult to assess. A cohort of carbon flux researchers sus­ests (conifer and deciduous) growing near the southem pect that they may be underevaluating the nighttimeedge of that climate range. The substantial rate of net measure of CO^ fluxes, despite attempts to measurecarbon uptake by tropical and temperate forests sites, the storage term (Black et al. 1996; Goulden et al.under study, is an indicator that these sites have expe­ 1996a; Baldocchi et al. 2000; Aubinet et al. 2000).rienced substantial amounts of disturbances in the rela­ Potential explanations include insufficient turbulenttively near past (50-100 yr) and have not reached mixing, incorrect measurement of the storage term oflong-term steady state. CO^ in the air space and soil, and the drainage of CO^ On close inspection of Table 1, we note that sev­ out of the canopy volume at night. At present, severaleral sites are sources of CO^. Sites losing carbon di­ teams of investigators are applying an empirical cor­ oxide include boreal forests and tundra, which are rection to compensate for the underestimation of night­experiencing disturbance of long-term carbon storage time carbon flux measurements. This correction is basedpools (Goulden et al. 1998; Lindroth et al. 1998; on CO^ flux density measurements obtained duringOechel et al. 2000), and grasslands, which were bumt windy periods or by replacing data with a temperature-(Suyker and Verma 2000) or were suffering from dependent respiration function (Black et al. 1996;drought (Meyers 2001). The maximum carbon lost by Goulden et al. 1996a,b; Lindroth et al. 1998). Mostthese sites on an annual basis does not exceed researchers presume that data from windy periods rep­100 g C m-2. resent conditions when the storage and drainage of With the plethora of long-term carbon flux data CO^ is minimal (Aubinet et al. 2000). available, we can quantify seasonal pattems of carbon When using FLUXNET data, the user should rec­fluxes across a variety of climates, biomes, and plant ognize that some sites are more suitable for construct­ functional groups. Figure 3 shows the annual pattem ing annual sums than others. Yet good science can beof weekly CO^ exchange for several temperate broad­ produced from the nonideal sites, too. They can pro­leaved deciduous forests. Broadleaved forests lose vide information on physiological functioning, tem­carbon during the winter, when they are leafless and

Bulletin of the American Mefeorological Society 2423 T able 1. List of published yearlong studies of net biosphere-atmosphere CO^ exchange, W. Negative signs represent a loss of carbon from the atmosphere and a net gain by the ecosystem.

N Observation Principal Site Vegetation (g C yi^^) year investigator Citation

a. Temperate broadleaved deciduous forests

Italy Beech, Fagus -470 1994 R. Valentini Valentini et al. (1996)

France Beech, Fagus -218 1996 A. Granier Granier et al. (2000) -257 1997

Denmark Beech, Fagus -169 1997 N. O. Jensen Filegaard et al. (2001) -124 1998

Iceland Poplar -100 1997 H. Thoreirsson Valentini et al. (2000)

Saskatchwan, Aspen -160 1994 T. A. Black Black et al. (1996) Canada -144 1994 -80 1996 -116 1997 -290 1998

Oak Ridge, TN Acer, Quercus -525 1994 D. Baldocchi Greco and -610 1995 Baldocchi (1996); Maple, oak -597 1996 Wilson and -652 1997 Baldocchi (2000) -656 1998 -739 1999

Petersham, MA Acer, Quercus -280 1991 S. Wofsy W ofsy et al. (1993) -220 1992 Maple, oak -140 1993 Goulden et al. -210 1994 (1996a,b) -270 1995

Borden, ON, Canada Maple,Acer -80 1996 X. Lee/J. Fuentes Lee et al. (1999) -270 1997 -200 1998

Indiana Deciduous forest -240 1998 H. P. Schmid Schmid et al. (2000)

Takayama, Japan Broadleaf deciduous -120 1994 Yamomoto et al. -70 1995 (1999) -140 1996 -150 1997 -140 1998

h. Mixed deciduous and evergreen forests, temperate and Mediterranean

Belgium Conifer and broadleaf -430 1997 M. Aubinet Valentini et al. (2000)

Belgium Conifer and broadleaf -157 1997 R. Ceulemans Valentini et al. (2000)

c. Tropical forests

Amazon Tropical forest -102 1995 J. Grace Malhi et al. (1998)

d. Needleleaf evergreen, conifer forests

Sweden Pinus sylvestris 90 1995 A. Lindroth Lindroth et al. (1998) -5 1996 80 1997 Picea abies -190 1997

Italy Picea abies -450 1998 R. Valentini Valentini et al. (2000)

2424 Vol. 82, No. 1 1, November 2001 T a b l e 1. Continued.

V Observation Principal Site Vegetation (g C m-^ yr ') year investigator Citation

d Needleleaf, evergreen, conifer forests

France Pinus pinaster -430 1997 P. Berbigier Berbigier et al. (2001)

Germany Picea abies -7 7 1997 Valentini et al. (2000)

Germany Picea abies -330 1996 Ch. Bemhofer Valentini et al. (2000) -480 1997 -45 1998

Germany Conifer -310 1996 A. Ibrom Valentini et al. (2000) -490 1997

Germany Picea abies -77 1997 E.-D. Schulze Valentini et al. (2000)

Netherlands Pinus sylvestris -210 1997 H. Dolman Valentini et al. (2000)

United Kingdom Picea sitchensis -670 1997 J. M oncrieff Valentini et al. (2000) -570 1998 Valentini et al. (2000)

Finland Pinus sylvestris -230 1997 T. Vesala Valentini et al. (2000) -260 1998 Markkanen et al. -190 1999 (2001)

Howland, ME Conifer -210 1996 D. Hollinger Hollinger et al. (1999)

Florida Slash pine -740 1996 H. Gholz Clark et al. (1999) -610 1997

Florida Cypress -84 1996 H. Gholz Clarke et al. (1999) -3 7 1997

Manitoba, Canada Picea mariana 70 1995 Goulden et al. (1998) 20 1996 -10 1997

Metolius, OR Pinus ponderosa -320 1996 B. Law Anthoni et al. (1999) -270 1997

Wind River, WA Pseudotsuga menzieii -167 1998 K. T. Paw U Paw U et al. (2000, to -220 unpublished manuscript)

Prince Albert, Picea mariana -68 1995 P. Jarvis Malhi et al. (1999) SK, Canada

e. Broadleaved evergreen, Mediterranean forests

Italy Quercus ilex -660 1997 R. Valentini Valentini et al. (2000)

f Grassland

Ponca, OK Tall-grass prairie 0 1997 S. B. Verma Suyker and Verma (2001)

Little Washita, OK Grazed grassland 41 1997 Meyers (2001)

g. Tundra

Alaska Tundra 40 W. Oechel Oechel et al. (2000)

Bulletin of the American Mefeorological Society 2425 dormant, and gain carbon during the summer grow­system CO^ exchange versus length of growing sea­ ing season. What differs most among broadleaved for­son in days; for convenience, we define the growing est sites is the timing of the transition between gainingseason as the period over which mean daily CO^ ex­ and losing carbon. The site in the southeastern Unitedchange is negative, due to net uptake by the biosphere. States (Tennessee) is the first to experience springWe observe that net ecosystem CO^ exchange in­ growth and photosynthesis, followed by the site in creases 5.7 g C m~^ for each day that the length of the Denmark and one in the northeast United States (Mas­growing season increases. The predictive power of this sachusetts). The Harvard Forest in Massachusetts is result is very robust, as variations in the length of the on the eastern side of the North American continent.growing season account for 83% of the variance of

It experiences a later spring than the more northerlyCO 2 exchange that is observed. The coherent response Danish site because the climate of Denmark is mod­ of net ecosystem CO^ exchange of the deciduous for­ erated by the Gulf Stream. During the peak of the est biome reflects their broadly similar growing sea­ growing season, these three sites, separated by sev­ son climate, age, and disturbance history. It also gives eral thousand kilometers, experience similar dailyus confidence in the accuracy of independent research sums of carbon uptake (around weeks 23-24). Andsites across the network. On the basis of this figure, the two North American sites, separated by over one may conclude that site-to-site measurement errors 1000 km, experience similar weekly sums of net CO^ and biases may be relatively small compared to vari­ exchange from about week 23 to 35. In the autumn,ance caused by ecological and meteorological factors. the Danish site is the first to drop leaves and respire, In Fig. 4, we also added a data point from a tropi­ and the most southem site is last. Another feature tocal forest, which experiences a year-round growing be gathered from this figure is how ideal and nonidealseason. On the basis of data from the Tropics we ex­ sites affect the measurement of respiration. During the pect the regression curve to reach an asymptote as the winter the Tennessee site experiences extremely lowgrowing extends beyond 220 days. Consequently, one rates of respiration. This site is situated on rolling ter­ rain and near several power plants. Cold air drainage causes CO^ to drain out from under the flux tower, A Massachusetts, USA ■ Michigan, USA and plume impaction causes downward carbon fluxes ▼ Indiana, USA • Prince Albert, CANADA that counter the upward respiratory flux. ^ tropical forest, Brazil ▼ Tennessee, USA Based on data in Fig. 3, one may suspect that phe- □ Denmark □ Belgium A Italy A Ontario • Japan nological timing of spring and autumn would be a 100 ■ major factor causing site variability among net carbon Broad-Leaved Forests exchange of temperate deciduous forests. We quan­ 0 - tify this effect in Fig. 4 by plotting annual net eco- -100 -

-200 - b[0]: 603 40 b[1]: -5.72 Denmark, 1997, 55° N -300 “ : 0.82 Oak Ridge, TN, 36° N Harvard Forest, MA, 42° N 20 -400 -

-500 - N E -600 - 3 U -20 m -700 - z \ -800 ■ -40 T “T T “T 100 150 200 250 300 350

Length of Growing Season -60 0 510 15 20 25 30 35 40 45 50 55 Eig . 4. Annual net eeosystem CO^ exehange: Impaet of length

W eek of growing season (in days) on temperate broadleaved forests and one tropieal forest (Brazil). Here, is the zero intereept and F ig . 3. Annual course of weekly net eeosystem CO^ exehangeis the slope of the regression eurve, and is the eoeffieient of (N E E ) for three temperate deeiduous forests, loeated in Oak Ridge,determination. Dashed lines are the standard error of the regres­ Tennessee; Petersham, Massaehusetts; and Denmark. sion at the 95% level.

2426 Vol. 82, No. I I November 2001 100 temperate, humid conifer, Durham, NC western, semi-arid conifer, Metolius, OR boreal conifer, Hyytiala, Finland 50 !c c 10 o E E o -50 1992 D) 1993 X 1994 3 IL -100 1995 oN 1996 1997 o -150 1998 O) -20 1999 -200 1 3 5 7 9 11 -30 Month

F ig . 5. Year-to-year variation in the annual eyele of monthly -40 CO^ exehange at Harvard Forest, a broadleaved deeiduous forest. 0 10 20 30 40 50

W eek

F ig . 6. Seasonal course of net weekly CO^ exehange over con­ should not extrapolate the regression line out to 365 trasting eonifer forests during 1997. days. Nor do we advocate using this plot for estimat­ ing year-to-year variability in net ecosystem CO^ ex­ change at a site. To illustrate this point we show an With longer datasets, alternative analytical meth­ example of year-to-year variation of the annual cycleods can be used to interpret them. Application of Fou­ of CO^ exchange at Harvard Forest (Fig. 5), the site rier transforms to yearlong and multiyear flux with the world’s longest measurement record. Othermeasurement records provides a means of quantify­ factors, such as the presence or absence of snow cover, ing the timescales of flux variance (Baldocchi et al. degree of cloudiness (Goulden et al. 1996b; Lee et2001). al. The dominant time periods over which carbon 1999), and drought (Wilson and Baldocchi 2000)dioxide im­ fluxes vary can be days, weeks, seasons, and pose variance on the annual carbon budget, too. years. Figure 7 shows the power spectra of CO^ fluxes The seasonal trend of CO^ exchange experienced by conifer forests is affected by latitudinal differences, which shorten or lengthen the growing season, and Grassland, Little Washita, OK Conifer Forest, Hyytiala, Finland by continental position, which forces conifer ecosys­ Deciduous forest. Oak Ridge, TN tems to experience oceanic, Mediterranean, or conti­ Corn, Bondville, IL nental climates. To illustrate these points we present data on seasonal patterns of net CO^ exchange from three contrasting conifer forest systems (Fig. 6). The

North Carolina site is at a temperate and humid lo­ 0.1 cale in the southeastern United States. There, theN 0 c0) canopy is able to take up carbon all year. The Oregon b 0) c Vi site experiences cool, wet winters and dry, warm sum­C mers. It assimilates carbon during the late winter, 0.01 spring, and autumn, but experiences large reductions in carbon dioxide uptake during the dry summer growing season (there are instances when it loses car­ 0.001 bon, too). The Finnish site is in the boreal zone. The 0.001 0.01 0.1 1 trees are dormant during the cold winter, but the for­ cycles per day est stand loses carbon due to soil and microbial res­ piration under the snow pack. In the spring, the onset F ig . 7. Power spectrum of CO^ flux over the eourse of a year at sites of contrasting elimate, functionality, and structure. Power of photosynthesis is much delayed compared to the spectra for the other sites were derived from time series of 1 yr more southerly temperate sites, and its conclusion isin duration. The power spectra on they axis are multiplied by sooner in the autumn. natural frequency(n) and are normalized by variance.

Bulletin of the American Meteorological Society 2427 over a variety of field sites on a daily timescale. These bon cycling models to translate remotely sensed ra­ variations in CO^ and water vapor exchange are forceddiation measurements to an areawide estimate of car­ by daily rhythms in solar radiation, air and soil tem­bon uptake (Ruimy et al. 1995; Cramer et al. 1999). perature, humidity, CO^, stomatal aperture, photosyn­Figure 8 shows that the canopy-scale quantum yield thesis, and respiration. On a weekly timescale, depends on the sky conditions, as well as temperature. fluctuations in CO^ and water vapor exchange are in­The canopy-scale quantum yield is nearly double un­ duced at midlatitude sites from synoptic weatherder cloudy skies, as compared to clear skies. changes that are associated with the passage of high Combining data from the North American and and low pressure systems and fronts. Weather eventsEuropean networks gives us a wider breadth of infor­ cause distinct periods of clear sky, overcast, and partlymation on how CO2 flux densities respond to tempera­ cloudy conditions, which alter the amount of availableture. The temperate zone of Europe is displaced light to an ecosystem, how light is transmitted through northward, as compared to North America, due to the a plant canopy, and the efficiency by which it is usedpresence of the Gulf Stream. Summer temperatures to assimilate carbon (Gu et al. 1999). The passing ofin Europe are on the order of 15°-20°C, while in North weather fronts also changes air temperature, humid­America summer temperatures are in the range of 20°- ity deficits, and pressure. On seasonal timescales,30°C. Figure 9 indicates that the temperature at which changes in the sun’s position alter the amount of sun­peak gross CO^ exchange occurs is several degrees light received, the air, soil, and surface temperature oflower for many European sites than for their North the ecosystem, and its water balance. SuperimposedAmerican counterparts. These data suggest that glo­ upon these meteorological factors is a variance causedbal carbon flux models will need to vary CO^ flux- by phenology and seasonal changes in photosynthetictemperature response curves for North America and capacity (Wilson et al. 2000). The timing and occur­Europe, in support of findings by Niinemets et al. rence of leaf expansion, senescence, leaf fall, fine root(1999). growth and turnover, and seasonal changes in photo­ With regard to meteorology, FEUXNET can pro­ synthetic capacity and leaf area index are examples videof new information on how the biosphere affects phonological factors affecting low-frequency fluctua­the partitioning of net radiation into sensible(H) and tions in canopy CO^ exchange. Prolonged wet and drylatent heat (XE) exchange, as quantified by the Bowen spells are other factors that exert variance on the spec­ ratio. Figure 10 shows the yearly course in Bowen ra­ tral record. tio for several selected sites. During the winter, the Canopy-scale quantum yield represents the initialBowen ratio is variable, and in many instances large. slope of the relation between net ecosystem CO^ ex­ change and the photosynthetic photon flux density. Canopy-scale quantum yield is of particular interest, 35 as it is a parameter used by many biogeochemical car- O 30 C Q) 3 (Q 4.5 : I IL 25 • Diffuse PAR 4.0 : # Direct PAR ao o 20 3.5 - 3 >1 « “ ■ 5 o 15 E o % 0) C % ,*%v b[0] 3.192 2.5 1 0) b[1] 0.923 • • • 10 2.°: r " 0.830 £ o 1.54 ^ 5 10 15 20 25 30 1.0 : Mean Summer Temperature (C) 0.5 - 220 240 F ig . 9. The relation between the optimum temperature for

Day of Year eanopy-seale gross primary produetivity vs maximum mean monthly air temperature during the summer growing season. F ig . 8. Seasonal variation of initial quantum yield of a eoniferThese data show how photosynthesis adapts to elimate. Hereb(0) (Seots) pine forest in Finland. Quantum yield is the initial slopeis the zero intereept andb(l) is the slope of the regression eurve, of the response between eanopy CO^ exehange and photosynthetieand E is the eoeffieient of determination. Dashed lines are the stan­ photon flux density. PAR is photosynthetieally aetive radiation.dard error of the regression at the 95% level.

2428 Vol. 82, No. I I November 2001 with H being up to 6 times greater thanXE. casting models (Foley et al. 1998). The trend is war­ Occurrences of negative Bowen ratios are recordedranted because information on photosynthesis con­ as well and are attributed to radiative cooling and thestrains the information on stomatal conductance, a presence of water on the surface. In both instances,variable needed to compute sensible and latent heat stable thermal stratification is associated with the exchange. Figure 11 shows how well carbon and wa­ downward transfer of sensible heat. The timing ofter fluxes are coupled for two functional forest groups. leaf-out in the temperate broadleaved forest and grass­Water use efficiency can be defined as the regression land biomes affects the Bowen ratio markedly. The slope between NEE and EE in Fig. 11. We observe presence of leaves causes Bowen ratios to drop inthat broadleaved forests use water more efficiently magnitude, compared to their leafless states, andthan conifer forests, by a factor of almost 1.5. range between 0 and 1. The timing of leaf-out alsoFurthermore, variations in latent heat flux explain has a distinct impact on the humidity and tempera­more of the variance in CO^ exchange of broadleaved ture of the planetary boundary layer (Schwartz 1996)forests than it does for conifer forests; the coefficients and the seasonal pattem of its maximum height (Wil­of variation were 0.77 and 0.42, respectively. Another son and Baldocchi 2000; Yi et al. 2000). conclusion that can be drawn from this information There is a growing trend toward using coupledrelates to carbon sequestration and carbon storage. carbon-water flux models in global climate and fore-Greater carbon uptake cannot be achieved without the availability and use of water. Ecosystems with high rates of carbon uptake use more water. Deciduous Forests The Euroflux community published results show­ Oak Ridge, TN, 36 N ing that gross primary productivity is relatively con­ Harvard Forest, 42 N « stant (Valentini et al. 2000). Theoretical calculations also support the idea that gross primary productivity is relatively constant for the range site characteristics

conifer forests 0 10 b[0] 13.6 b[1] -0.56 Conifer Forests 0.428 Tharandt, Germany, 51 N Metolius, OR, 44 N Howland, ME, 45 N 5) -100 Ui U1 -150

LE (MJ m '' month'h

0 10 20

Grassland and Crops broadleaves 50 ■ ■ Little Washlta,_OK 0 “ ■ B Bondville, IL Ponca. OK -50 ■

-100 OO) lU -150 b[0] 55.7 UJ b[1] -0.82 ■ -200 r 2 0.77 -

-250 0 50 100 150 200 250 300 350

LE (MJ m'^ month'^^

Fig . 11. The linkage between earbon (NEE) and water fluxes F ig . 10. Seasonality of Bowen ratio, the ratio between sensible (EE) for eonifer and temperate forests (broadleaves). Here is and latent heat flux densities, at representative deeiduousthe for­ zero intereept, is the slope of the regression eurve, and E ests, eonifer forests, and grassland and erop sites. is the eoeffieient of determination.

Bulletin of the American Meteorological Society 2429 studied by the Euroflux network. The Euroflux data,Oregon (Metolius) and a young stand in northern however, come from a relatively narrow range of plant Califomia (Blodgett) are shown in Fig. 13. Peak rates architectural and functional types (closed conifer andof evaporation are almost triple for the young stand, deciduous forests) and climate conditions, so one canas compared to the old stand. Net daytime carbon as­ question how applicable the results may be across a similation fluxes of the young and old stands are out wider range of conditions. Using a biophysical modelof phase during the summer growing season (seen at (Baldocchi and Meyers 1998), we predict that gross18 months in Fig. 13). The old stand is most produc­ primary productivity (GPP) of plants via photosynthe­ tive during the spring and autumn periods, while soil sis is an asymptotic function of leaf area index, leafwater deficits restrict summertime rates of CO^ up­ nitrogen, and the fraction of photosynthetieally activetake. The young stand experiences greatest rates of radiation that the canopy intercepts (Fig. 12). uptake during the summer, and peak rates of the young There is a growing body of literature that is at­ stand can exceed peak rates of the old stand in the tempting to understand the impact of stand age onsummer by 14% (-140 g C m“^ month“^ near month carbon and water fluxes (Amiro et al. 1999; Schulze30 vs -120 near month 18). Simulation modeling, et al. 1999; Eaw et al. 2001, manuscript submittedwhich to includes the effects of tree age, suggested that Global Change B; K. iol T. Paw U et al. 2001, manu­ the milder winters and ample annual rainfall in the script submitted toEcosystems). Productivity is as­ young stand allow it to have a higher leaf area than sumed to diminish with age, and disturbances affectthe old forest, allowing more carbon uptake through the relative ratio between canopy carbon uptake andthe year (Eaw et al. 2001, manuscript submitted to soil respiration, and the ratio between canopy transpi­Global Change Biol). Site management also has a ration and soil evaporation. An attribute of a globalsubstantial impact on carbon and water fluxes at the network, such as FEUXNET, is the occurrence of sites with similar species and functionality but different Blodgett, < 10 years old ages. Based on the data surveyed in Table 1, part d, Metolius, > 200 years old Ponderosa Pine we can draw the conclusion that the net carbon ex­ -20 change of old growth conifer forests is not carbon _ -40 c neutral (e.g., K. T. Paw U et al. 2001, manuscript o submitted toEcosystems). A direct comparison of flux E .60 densities of carbon dioxide and water vapor that wereQ -80 8) measured over an old age Ponderosa pine stand B in -100 z I -120 shrub removal 2400 « -140 2200 0 CANVEG Q -160 2000 © EUROFLUX 0 6 12 18 24 30 36 42 48 • Walker Branch Watershed 1800 400 'h>1 i 1600 shrub removal E 1400 300 n 1200 Ic c o J 1000 E o 800 'i'g 200 600 Sn 400 UJ " 100 200 0 100 200 300 400 500 600

Vc^ax LAI/ffpar 0 6 12 18 24 30 36 42 48

F ig . 12. The relation between gross primary produetivity Months after Jan., 1996 (GPP) and plant physiologieal eapaeity using field data and a biophysieal model. The index of plant physiologieal eapaeity is F ig . 13. The impaet of stand age on daytime CO^ exehange a funetion of the maximum earboxylation veloeity of photosyn­(NEE) and latent heat exehange (EE). The data are from a young thesis (Eemax), leaf area index (LAI), and the fraetion of ab­(<10 yr; blaek eireles) and an old (> 200 yr; gray eireles) pon­ sorbed visible sunlight (/par). The Euroflux field data arederosa pine forest (see Eaw et al. 2001, manuseript submitted reported in Valentini et al. (2000). to Global Change Biol', Goldstein et al. 2000).

2430 Vol. 82, No. 11, Novem ber 2001 young site. The removal of understory shrubs in 1999domain. More physiological and stand stmcture stud­ reduced both CO^ uptake and evaporation. ies will be needed to aid in the interpretation of tower measurements and to provide contextual information to test ecophysiological models. 8. Conclusions As data accumulate, new questions are arising about how to interpret fluxes made over mixed stands. A global network of long-term measurement sites Frontal passages cause the annual sums of ecosystem has been established and is producing new informa­CO2 exchange to be biased, as associated changes in tion on how CO2 and water vapor between the terres­ wind direction and speed cause a tower site to view a trial biosphere and atmosphere vary across a broad different flux footprint (Amiro 1998) or expose the spectrum of biomes, climates, and timescales. The sci­ ecosystem to different temperature and humidity re­ ence generated so far indicates that rates of net car­gimes. Over sites with inadequate fetch or mixed veg­ bon uptake by tropical and temperate forests are etation in the flux footprint, flux densities that are substantial and exceed estimates previously generatedmeasured will vary with wind direction. As knowl­ with biospheric modeling systems. Gross primary edge evolves we may have to evaluate annual sums productivity of forest ecosystems may not be constant, by weighting field measurements on the basis of wind but may depend on plant architecture (e.g., on leafdirection and the appropriate flux footprint probabil­ area), foliage photosynthetie capacity, and the amount ity density functions (Schmid 1994). On the flip side, of sunlight absorbed. we can use information from different wind directions Carbon net ecosystem exchange of broadleavedto address questions relating to the role of biodiver­ deciduous forests strongly depends on length of grow­sity on mass and energy exchange of ecosystems. ing season, but is sensitive to perturbations such as droughts, clouds, winter snow cover, and early thaw­ Acknowledgments. The FLUXNET project office and Data ing, too. Conceming conifer forests, the conclusionsArchive is sponsored by NASA’s EOS Validation Program. The are not as unifying. Seasonal and annual sums of componentnet regional networks and individual field sites are spon­ carbon exchange by boreal, semiarid, temperate, andsored by an array of government agencies. The Americas net­ work, AmeriFlux, is sponsored by the United States Departments humid conifers differ among one another and forof Energy [Terrestrial Carbon Program, National Institutes of different physiological reasons. Global Environmental Change (NIGEC)], Commerce (NOAA), The response of canopy-scale CO^ exchange to and Agriculture (USDA/Forest Service), NASA, the National sunlight varies with cloud cover as clouds alter the di­Science Foundation, and the Smithsonian Institution. These fed­ rection of incoming sunlight and how it penetrates intoeral agencies sponsor the Global Change Research Program (USGCRP) to promote cooperation with each other, academia, a canopy. The response of canopy-scale CO^ exchange and the international community, in order to make it as easy as to temperature is sensitive to the local climate. The possible for researchers and others to access and use global temperature optima for canopy photosynthesis are dif­ change data and information. ferent for similar functional forest types in Europe and European sites in the Euroflux and Medeflu projects are North America. supported by the European Commission Directorate General Xll Environment and Climate Program. Canadian collabora­ tors are sponsored by the Natural Sciences and Engineering Research Council of Canada (NSERC). Japanese and Asian 9. Future plans sites are supported by the Ministry of Agriculture, Forest and Fisheries (MAFF), the Ministry of Industrial Trade and Indus­ The presence of sites on less than ideal terrain af­try (MlTl), and Ministry of Education, Science, Sports and fords the micrometeorological community to take the Culture (MESSC). We thank the numerous scientists, students, and technicians challenge and assess the impact of advection on the responsible for the day-to-day gathering of the flux data, and measurement of long-term fluxes. Given the currentthe agency representatives who fund the respective projects. tower site distribution, there is a need to developWithout the dedicated efforts of so many individuals, this glo­ transect studies at anchor sites, so we can study spa­bal network would not function or exist. Special appreciation tial pattems in greater detail and better assess the im­ is extended to Drs. Roger Dahlman, Diane Wickland, Ruth pact of forest stand age, biodiversity, and landReck, and David Starr, who have provided funds that supported initial planning workshops and that support many of the oper­ disturbances. Aircraft flux studies can be used to aug­ating science teams. ment and extrapolate tower information in the spatial

Bulletin of the American Mefeorological Society 2431 References Betts, A. K., J. FI. Ball, A. C. Beljaars, M. J. Miller, and P. A. Viterbo, 1996: The land surface-atmosphere interaction: A review based on observational and global modeling perspec­ Amiro, B. D., 1998: Footprint climatologies for evapotranspi- tives. J. Geophys. Res., 101, 7209-7225. ration in a boreal catchment.Agric. For. Meteor., 90, 195- Black, T. A., and Coauthors, 1996: Atmual cycles of CO^ and

201 . water vapor fluxes above and within a Boreal aspen stand. , J. 1. MacPherson, and R. L. Desjardins, 1999: BOREAS Global Change Biol, 2, 219-230. flight measurements of forest-frre effects on carbon dioxide , W. J. Chen, A. G. Barr, M. A. Arain, Z. Chen, Z. Nesic, and energy fluxes.Agric. For. Meteor., 96, 199-208. E. FI. Flogg, FI. FI. Neumatm, and P. C. Yang, 2000: Increased Amthor, J. S., 1995: Terrestrial higher-plant response to increas­ carbon sequestration by a boreal deciduous forest in years ing atmospheric [CO^] in relation to the global carbon cycle. with a warm spring. Geophys. Res. Lett, 27, 1271-1274. Global Change Biol, 1, 243-274. Canadell, J., and Coauthors, 2000: Carbon metabolism of the Amundson, R., L. Stem, T. Raisden, and Y. Wang, 1998: Theterrestrial biosphere. Ecosystems, 3, 115-130. isotopic composition of soil and soil-respired CO^. Ceulemans, R., and M. Mousseau, 1994: Tansley Review No. Geoderma, 82,83-114. 71. Effects of elevated atmospheric CO^ on woody plants. Anderson, D. E., S. B. Verma, andN. J. Rosenberg, 1984: Eddy New PhytoL,Xll, 425^46. correlation measurements of CO^, latent heat and sensible heatCiais, P., P. P. Tans, M. Trolier, J. W. C. White, and R. J. Francy, fluxes over a crop surface.Bound.-Layer Meteor., 29, 167- 1995: A large North Flemisphere terrestrial CO^ sink indi­ 183. cated by the '^C/'^C ratio of atmospheric CO^. Science, 269, Anthoni, P. M., B. E. Law, and M. El. Unsworth, 1999: Carbon 1098-1102. and water vapor exchange of an open-canopied ponderosa pineClark, K. L., FI. L. Gholtz, J. B. Moncrieff, F. Cropley, and FI. ecosystem. Agric. For. Meteor., 95, 115-168. W. Loescher, 1999: Environmental controls over net ex­ Aubinet, M., and Coauthors, 2000: Estimates of the atmual net changes of carbon dioxide from contrasting Florida ecosys­ carbon and water exchange of Europeran forests: The tems. Ecological AppL, 9, 936-948. EUROFLUX methodology. Adv. Ecol Res., 30, 113-175. Conway, T. J., P. P. Tans, L. S. Waterman, K. W. Thoning, D. R. BAFIC Core Project Office, Eds., 1993: Biospheric aspects of Kitzis, K. Masarie, andN. Zhang, 1994: Evidence for inter- the hydrological cycle. The operational plan. IGBP Rep. 27, atmual variability of the carbon cycle from NOAA/CMDL Stockholm, Sweden, 103 pp. global sampling network. J. Geophys. Res., 99, 22 831-22 855. Bakwin, P. S., P. P. Tans, D. F. Flurst, and C. L. Zhao, 1998:Cramer, W. J., D. W. Kicklighter, A. Bondeau, B. Moore111, Measurements of carbon dioxide on very tall towers: Results G. Churkina, B. Nemry, A. Ruimy, and A. L. Schloss, 1999: of the NOAA/CMDL program. Tellus, SOB, 401M15. Comparing global models of terrestrial primary productivity (NPP): Baldocchi, D. D., and T. P. Meyers, 1998: On using eco-physi- Overview and key results.Global Change Biol, 5, 1-15. ological, micrometeorological and biogeochemical theory to Crawford, T. L., R. J. Dobosy, R. T. McMillen, C. A. Vogel, and evaluate carbon dioxide, water vapor and gaseous depositionB. B. Flicks, 1996: Air-sinface exchange measurement in het­ fluxes over vegetation.Agric. For. Meteor., 90, 1-26. erogeneous regions: Extending tower observations with spa­ , B. B. Flicks, and T. P. Meyers, 1988: Measuring biosphere- tial structure observed from small aircraft.Global Change atmosphere exchanges of biologically related gases with mi­ Biol, 2, 275-285. crometeorological methods. Eco/ogy, 69, 1331-1340. Denmead, O. T., 1969: Comparative micrometeorology of a , R. Valentini, S. R. Running, W. Oechel, and R. Dahlman, wheat fleld and a forestof Pinus radiata. Agric. For. Meteor., 1996: Strategies for measuring and modeling CO^ and water6, 357-371. vapor fluxes over terrestrial ecosystems.Global Change Biol, , M. R. Raupach, F. X. Dimin, FI. A. Cleugh, andR. Leuning, 2, 159-168. 1996: Boundary layer budgets for regional estimates of scalar , J. J. Finnigan, K. W. Wilson, K. T. Paw U, and E. Falge, fluxes.Global Change Biol, 2, 255-264. 2000: On measuring net ecosystem carbon exchange in com­Denning, A. S., J. G. Collatz, C. Zhang, D. A. Randall, J. A. Berry, plex terrain over tall vegetation.Bound.-Layer Meteor., 96, P. J. Sellers, G. D. Colello, and D. A. Dazlich, 1996: Simula­ 257-291. tions of terrestrial carbon metabolism and atmospheric CO^ in , E. Falge, and K. Wilson, 2001: A spectral analysis of bio­ a general circulation model. Part 1: Surface carbon fluxes. sphere-atmosphere trace gas flux densities and meteorologi­ Tellus, 48B,521-542. cal variables across hour to multi-year time scales.Agric. Desjardins, R. L., D. J. Buckley, and G. St. Amour, 1984: Eddy For. Meteor., 107, 1-27. flux measurements of CO^ above com using a microcomputer Barr, A. G., G. van der Kamp, R. Schmidt, andT. A. Black, 2000: system. Agric. Meteor., 32, 257-265. Monitoring the moisture balance of a boreal aspen forest us­ , and Coauthors, 1997: Scaling up flux measurements for the ing a deep groundwater piezometer.Agric. For. Meteor., 102, boreal forest using aircraft-tower combinations.J. Geophys. 13-24. Res., 102, 29 125-29 134. Baumgartner, A., 1969: Meteorological approach to the exchangeDrake, B. G.,M. A. Gonzalez-Meler, and S. P. Long, 1996: More of CO^ between atmosphere and vegetation, particularly for­ efflcient plants: A consequence of rising atmospheric CO^? ests stands.Fhotosynthetica, 3, 127-149. Annu. Rev. Plant Physiol. Plant Mol. 48,Biol,609-639. Berbigier, P., J.-M. Botmefond, and P. Melhnan, 2001: CO^ andFalge, E., and Coauthors, 2001: Gap fllling strategies for defen­ water vapour fluxes for 2 years above Euroflux forest site. sible atmual sums of net ecosystem exchange.Agric. For. Agric. For. Meteor., 108, 183-197. Meteor., 107, 43-69.

2432 Vol. 82, No. 1 1, November 2001 Fan, S. M., M. Gloor, J. Mahlman, S. Pacala, J. Sarmiento, Jarvis, P. G., G. B. James, and J. J. Landsberg, 1976: Conifer­ T. Takahashi, and P. Tans, 1998: A large terrestrial carbon ous forest. Vegetation and the Atmosphere, J. L. Monteith, sink in North America implied by atmospheric and oceanic Ed., Vol. 2, Academic Press, 171-240. carbon dioxide data and models.Science, 282,442^46. Kaimal, J. C., and J. J. Finnigan, 1994:Atmospheric Boundary Finnigan, J. J., 1999: A comment on the paper by Lee (1998):Layer Flows: Their Structure and Measurement. Oxford Uni­ “On micrometeorological observations of surface-air ex­ versity Press, 289 pp. change over tall vegetation.” Agnc.For. Meteor., 97, 55-64. Kauppi, P. E., K. Mielikainen, and K. Kuuseia, 1992: Biomass Foken, P., and B. Wichura, 1996: Tools for quality assessment and carbon budget of European forests, 1971 to 1990.Sci­ of surface-based flux measurements.Agric. For. Meteor., ence, 256,70-74. 78, 83-105. Keeling, C. D., and T. P. Whorf, 1994: Atmospheric CO^ records Foley, J. A., S. Levis, 1. C. Prentice, D. Pollard, and S. L. Th­ from sites in the SIO air sampling network. Trends ’93: A ompson, 1998: Coupling dynamic models of climate and veg­compendium of data on global change. Oak Ridge National etation. Global Change Biol, 4, 561-580. Laboratory Rep. ORNL/CDlAC-65, 16-26. Goldstein, A. FI., and Coauthors, 2000: Effects of climate vari­ , ------, M. Wahlen, and J. van der Plicht, 1995: Interannual ability on the carbon dioxide, water, and sensible heat fluxesextremes in the rate of rise of atmospheric carbon dioxide above a ponderosa pine plantation in the Sierra Nevada (CA). since 1980.Nature, 375, 666-670. Agric. For. Meteor., 101, 113-129. , S. C. Piper, andM. Fleimann, 1996: Global and hemispheric Goulden, M. L., J. W. Munger, S. M. Fan, B. C. Daube, and S. C. CO^ sinks deduced from changes in atmospheric concen­ Wofsy, 1996a: Measurement of carbon storage by long-term tration. Nature, 381,218-221. eddy correlation: Methods and a critical assessment of ac­Kelliher, F. M., R. Leuning, M. R. Raupach, and E. D. Schulze, curacy. Global Change Biol, 2, 169-182. 1995: Maximum conductances for evaporation from global , ----- , ------, ------, and------, 1996b: Exchange of carbon vegetation types. Agric. For. Meteor., 73, 1-16. dioxide by a deciduous forest: Response to interannual climateLaw, B. E., A. FI. Goldstein, P. M. Anthoni, J. A. Panek, M. FI. variability. Science, 271, 1576-1578. Unsworth, M. R. Bauer, J. M. Fracheboud, and N. Flultman, , and Coauthors, 1998: Sensitivity of boreal forest carbon bal- 2000: CO^ and water vapor exchange by young and old pon­ ance to soil thaw. Science, 279, 214-217. derosa pine ecosystems during a dry summer.Tree Physiol, Gower, S. T., C. J. Kucharik, and J. M. Norman, 1999: Direct and 21, 299-308. indirect estimation of leaf area index, fpar and net primary pro­Lee, X., 1998: On micrometeorological observations of surface- duction of terrestrial ecosystems.Remote Sens. Environ., 70, air exchange over tall vegetation.Agric. For. Meteor., 91, 29-5 T 39-50. Granier, A., E. Ceschia, and C. Damesin, 2000: The carbon bal­ , J. D. Fuentes, R. M. Staebler, and FI. FI. Neumann, 1999: ance of a young beech forest.Fund Ecol., 14, 312-325. Long-term observation of the atmospheric exchange of CO^ Greco, S., andD. D. Baldocchi, 1996: Seasonal variations ofCO^ with a temperate deciduous forest in southem Ontario, Canada. and water vapor exchange rates over a temperate deciduous J. Geophys. Res., 104, 15 975-15 984. forest. Global Change Biol, 2, 183-198. Lemon, E. R., 1960: Photosynthesis under field conditions. 11. An Gu, L., J. D. Fuentes, FI. FI. Shugart, R. M. Staebler, and T. A. aerodynamic method for determining the turbulent carbon di­ Black, 1999: Responses of net ecosystem exchanges of car­ oxide exchange between the atmosphere and a com field. bon dioxide to changes in cloudiness: Results from two NorthAgron. J., 52,697-703. American deciduous forests.J. Geophys. Res., 104, 31 421- Levy, P. E., A. Grelle, A. Lindroth, M. Molder, P. G. Jarvis, 31 434. B. Kruijt, and J. B. Moncrieff, 1999: Regional scale CO^ fluxes Flalldin, S., S. E. Gryning, L. Gottschalk, A. Jochum, L.-C. Lundin, over central Sweden by a boundary layer budget method. and A. A. Van de Griend, 1999: Energy, water and carbon ex­Agric. For. Meteor., 98-99, 169-180. change in a boreal forest landscape—^NOPEX experiences.Lindroth, A., A. Grelle, and A. S. Moren, 1998: Long-term mea­ Agric. For. Meteor., 98-99, 5-29. surements of boreal forest carbon balance reveal large tempera­ Flanson, J. E., M. Sato, A. Lacias, R. Ruedy, 1. Tegen, and ture sensitivity.Global Change Biol, 4, 443^50. E. Matthews, 1998: Climate forcings in the industrial era.MaUii, Y., A. D. Nobre, J. Grace, B. Kruijt, M. Pereira, A. Gulf, Proc. Natl. Acad. Sci. USA. 95, 12 753-12 758. and S. Scott, 1998: Carbon dioxide transfer over a central Flollinger, D.Y., S. M. Goltz, E. A. Davidson, J. T. Lee, K. Tu, Amazonian rain forest. J.Geophys. Res., 103, 31 593-31 612. and FI. T. Valentine, 1999: Seasonal pattems and environ­ , D. D. Baldocchi, and P. G. Jarvis, 1999: The carbon bal­ mental control of carbon dioxide and water vapor exchange inance of tropical, temperate and boreal forests.Plant, Cell and an ecotonal boreal forest.Global Change Biol, 5, 891-902. Environ., 22,715-740. Floughton, R. A., and G. Woodwell, 1980: The flax pond eco­Markkanen, T., U. Rannik, P. Keronen, T. Surd, and T. Vesala, system study: Exchanges of CO^ between a salt marsh and 2001: Eddy covariance fluxes over a boreal Scots pine forest. the atmosphere. Ecology, 61, 1434-1445. Boreal Environ. Res., 6, 65-78. Inoue, T, 1958: An aerodynamic measurement of photosynthe­Massman, W. J., 2000: A simple method for estimating fre­ sis over a paddy field.Proc. Seventh Japan National Con­quency response corrections for eddy covariance systems. gress o f Applied Mechanics, 211-214. Agric. For. Meteor., 104, 185-198. Jacobs, C. M. J., and FI. A. R. De Bmin, 1992: The sensitivityMeyers, T. P., 2001: A comparison of summertime water and of transpiration to land-surface characteristics: Significance CO^ fluxes over rangeland for well-watered and drought con­ of feedback. J. Climate, 5, 683-698. ditions.Agric. For. Meteor., 106, 205-214.

Bulletin of the American Mefeorological Society 2433 Moncrieff, J. B., Y. Mahli, and R. Leuning, 1996: The prop­ , C. S. Grimmond, F. Cropley, B. Offerle, and H.-B. Su, agation of errors in long term measurements of land atmo­ 2000: Measurements of CO^ and energy fluxes over a mixed sphere fluxes of carbon and water.Global Change Biol, 2, hardwood forest in the mid-western United States.Agric. For. 231-240. Meteor., 103, 357-374. Monteith, J. L., and G. Szeicz, 1960: The CO^ flux over a fieldSchulze, E. D., and Coauthors, 1999: Productivity of forests of sugar beets. Quart. J. Roy. Meteor. Soa, 86, 205-214. in the Eurosiberian boreal region and their potential to act Moore, C. J., 1986: Frequency response corrections for eddy co- as a carbon sink—A synthesis.Global Change Biol, 5, 703- variance systems. Bound.-Layer Meteor., 37, 17-35. 722. Myneni, R. B., C. D. Keeling, C. J. Tucker, G. Asrar, and R. R. Schwartz, M. D., 1996: Examining the spring discontinuity in Nemani, 1997a: Increased plant growth in the northern high daily temperature ranges.J. Climate, 9, 803-808. latitudes from 1981-1991.Nature, 386,698-702. Sellers, P. J., and Coauthors, 1997: BOREAS in 1997: Scien­ , R. R. Nemani, and S. W. Rimning, 1997b: Estimation of tific results, experimental overview and future directions.J. global leaf area index and absorbed par using radiative trans­Geophys. Res., 102, 28 731-28 770. fer models. IEEE Trans. Geosci. Remote Sens., 35, 1380- Suyker, A. E., and S. B. Verma, 2001: Year-round observations 1393. of the net ecosystem exchange of carbon dioxide in a native Niinemets, U., V. Oja, and O. Kull, 1999: Shape of leaf pho­ tall grass prairie. Global Change Biol, 3, 279-290. tosynthetic electron transport versus temperature responseTans, P. P., 1. Y. Fung, and T. Takahashi, 1990: Observational curve is not constant along canopy light gradients in tem­ constraints on the global atmospheric CO^ budget.Science, perate deciduous trees.Plant, Cell Environ., 22,1497-1513. 247,1431-1438. Norby, R. J., S. D. Wullschleger, C. A. Gunderson, D. W. Valentini, R., P. de Angelis, G. Matteucci, R. Monaco, S. Dore, Johnson, and R. Ceulemans, 1999: Tree responses to risingand G. E. Scarascia-Mugnozza, 1996: Seasonal net carbon CO^ in field experiments: Implication for the future forest. dioxide exchange of a beech forest with the atmosphere.Glo­ Plant, Cell Environ., 22,683-714. bal Change Biol, 2, 199-208. Oechel, W. C., G. L. Vourlitis, S. J. Hastings, R. C. Zulueta, , G. Matteucci, and A. J. Dolman, 2000: Respiration as the L. Hinzman, and D. Kane, 2000: Acclimation of ecosystem main determinant of carbon balance in European forests.Na­ CO^ exchange in the Alaskan Arctic in response to decadal ture, 404,861-864. climate warming. Nature, 406, 978-981. Verma, S. B., D. D. Baldocchi, D. E. Anderson, D. R. Matt, and Ohtaki, E., 1984: Application of an infrared carbon dioxide and R. E. Clement, 1986: Eddy fluxes of CO^ water vapor, and himidity instnment to studies of turbulent transport.Bound.- sensible heat over a deciduous forest.Bound.-Layer Meteor., Layer Meteor., 29, 85-107. 36, 71-91. Pielke, R. A., R. Avissar, M. Raupach, A. J. Dolman, X. Zeng, , J. Kim, and R. J. Clement, 1989: Carbon dioxide, water andA. S.Detming, 1998: Interactions between the atmosphere vapor and sensible heat fluxes over a tall grass prairie. and terrestrial ecosystems: Influence on weather and climate. Bound.-Layer Meteor., 46, 53-67. Global Change Biol, 4, 461M75. Vermetten, A. W. M., L. Ganzeveld, A. Jeuken, P. Hofschreuder, Pilegaard, K., P. Hunnelshoj, N. O. Jensen, and Z. Chen, 2001: and G. M. J. Mohren, 1994: CO^ uptake by a stand of Douglas Two years of continuous CO^ eddy-fiux measurements over afir: Flux measurements compared with model calculations. Danish beech forest. Agric. For. Meteor., 107, 29M1. Agric. For. Meteor., 72, 57-80. Potosnak, M. J., S. C. Wofsy, A. S. Denning, T. J. Conway, J. W. Wilson, K. B., and D. D. Baldocchi, 2000: Esimating atmual net Munger, and D. H. Bames, 1999: Influence of biotic exchangeecosystem exchange of carbon over five years at a deciduous and combustion sources on atmospheric CO^ concentrations forest in the southem United States.Agric. For. Meteor., 100, in New England from observations at a forests flux tower.J. 1-18. Geophys. Res., 104, 9561-9569. , ------, and P. J. Hanson, 2000: Spatial and seasonal vari­ Randerson, J. T., M. V. Thompson, T. J. Conway, 1. Y. Fung, and ability of photosynthesis parameters and their relationship to C. B. Field, 1997: The contribution of terrestrial sources andleaf nitrogen in a deciduous forest.Tree Physiol, 20, 787- sinks to trends in the seasonal cycle of atmospheric carbon 797. dioxide.Global Biogeochem. Cycles, 11, 535-560. Wofsy, S. C., M. L. Goulden, J. W. Munger, S. M. Fan, P. S. Raupach, M. R., and B. J. Legg, 1984: The uses and limitations Bakwin, B. C. Daube, S. L. Bassow, and F. A. Bazzaz, 1993: of flux-gradient relationships in micrometeorology.Agric. Net exchange of CO^ in a mid-latitude forest.Science, 260, Water Manage., 8, 119-131. 1314-1317. Ruimy, A., P. G. Jarvis, D. D. Baldocchi, and B. Saugier, 1995:Yamamoto, S., S. Murayama, N. Saigusa, and H. Kondo, 1999: CO^ fluxes over plant canopies and solar radiation: A review. Seasonal and interarmual variation of CO^ flux between a tem­ Adv. Ecol. Res., 26, 1-53. perate forest and the atmosphere in Japan.Tellus, 51B, 402- Running, S. W., D. D. Baldocchi, D. Turner, S. T. Gower, 413. P. Bakwin, and K. Hibbard, 1999: A global terrestrial moni­Yi, C., K. J. Davis, P. S. Bakwin, B. W. Berger, and L. C. Marr, toring network, scaling tower fluxes with ecosystem model­ 2000: Influence of advection on measurements of the net eco- ing and EOS satellite data.Remote Sens. Environ., 70, 108- system-atmosphere exchange of CO^ from a very tall tower. 127. J. Geophys. Res., 105, 9991-9999. Schmid, H. P., 1994: Source areas for scalars and scalar fluxes. Bound.-Layer Meteor., 67, 293-318.

2434 Vol. 82, No. 1 1, November 2001