• 沒有找到結果。

Role of vegetation in determining carbon sequestration along ecological succession in the southeastern United States

N/A
N/A
Protected

Academic year: 2021

Share "Role of vegetation in determining carbon sequestration along ecological succession in the southeastern United States"

Copied!
19
0
0

加載中.... (立即查看全文)

全文

(1)

Role of vegetation in determining carbon sequestration

along ecological succession in the southeastern

United States

P A U L C . S T O Y*w, G A B R I E L G . K A T U L w z, M A R I O B . S . S I Q U E I R A w § , J E H N - Y I H J U A N G w },

K I M B E R L Y A . N O V I C K w, H E A T H E R R . MCC A R T H Y w k, A . C H R I S T O P H E R O I S H I w and R A M O R E N w

*Department of Atmospheric and Environmental Science, School of GeoSciences, University of Edinburgh, Edinburgh EH9 3JN, UK, wNicholas School of the Environment and Earth Sciences, Duke University, Box 90328, Durham, NC, USA, zDepartment of Civil and Environmental Engineering, Pratt School of Engineering, Duke University, Durham, NC, USA, §Departamento de Engenharia Mecaˆnica, Universidade de Brası´lia, Brazil, }Department of Geography, National Taiwan University, Taipei, Taiwan, kDepartment of Earth System Science, University of California, Irvine, CA 92697-3100, USA

Abstract

Vegetation plays a central role in controlling terrestrial carbon (C) exchange, but quantifying its impacts on C cycling on time scales of ecological succession is hindered by a lack of long-term observations. The net ecosystem exchange of carbon (NEE) was measured for several years in adjacent ecosystems that represent distinct phases of ecological succession in the southeastern USA. The experiment was designed to isolate the role of vegetation – apart from climate and soils – in controlling biosphere–atmo-sphere fluxes of CO2 and water vapor. NEE was near zero over 5 years at an early successional old-field ecosystem (OF). However, mean annual NEE was nearly equal, approximately 450 g C m2yr1, at an early successional planted pine forest (PP) and a late successional hardwood forest (HW) due to the sensitivity of the former to drought and ice storm damage. We hypothesize that these observations can be explained by the relationships between gross ecosystem productivity (GEP), ecosystem respiration (RE) and canopy conductance, and long-term shifts in ecosystem physiology in response to climate to maintain near-constant ecosystem-level water-use efficiency (EWUE). Data support our hypotheses, but future research should examine if GEP and RE are causally related or merely controlled by similar drivers. At successional time scales, GEP and RE observations generally followed predictions from E. P. Odum’s ‘Strategy of Ecosystem Development’, with the surprising exception that the relationship between GEP and RE resulted in large NEE at the late successional HW. A practical consequence of this research suggests that plantation forestry may confer no net benefit over the conservation of mature forests for C sequestration.

Keywords: ecosystem respiration, ecological succession, eddy covariance, grass field, gross ecosystem productivity, net ecosystem exchange, oak-hickory forest, Pinus taeda

Received 21 December 2006; revised version received 17 December 2007 and accepted 31 December 2007

Introduction

North American terrestrial ecosystems represent a net C sink of debatable magnitude (Pacala et al., 2001; Potter et al., 2006), primarily attributable to forest growth and reforestation (Delcourt & Harris, 1980; Caspersen et al., 2000) (but see Jackson et al., 2002), with strong

inter-annual variability (Keeling et al., 1996; Fan et al., 1998; Houghton, 2000) due largely to changes in climate (Houghton, 2000). Plot-level studies (Delcourt & Harris, 1980; Caspersen et al., 2000) and modeling approaches (Pacala et al., 2001; Potter et al., 2006) agree that the warm and moist southeastern (SE) region represents the strongest regional C sink in the USA.

Active land management (Wear & Greis, 2002) and pronounced interannual climatic variability (Peters et al., 2003) make SE ecosystems ideal case studies for Correspondence: Paul C. Stoy, tel. 1 44 131 650 7722, fax 1 44 131

662 0478, e-mail: paul.stoy@ed.ac.uk r2008 The Authors

(2)

quantifying the role of physical and biological factors in controlling C sequestration. The SE comprises primarily private land and is thus subject to dramatic shifts in land use that correspond to both economic and natural forces (Wear & Greis, 2002). Large-scale abandonment of agriculture after the US Civil War has continued to the present, and these old-field ecosystems have been largely replaced by forested ecosystems through ecolo-gical succession (Oosting, 1942; Johnston & Odum, 1956). Since the 1950s, both abandoned agricultural ecosystems and second-growth forests have been in-creasingly converted to intensive forest plantations, commonly composed of loblolly pine (Pinus taeda L.) or similar fast growing species (Wear & Greis, 2002). The result is a patchwork of vegetative types that represent different stages of ecological succession. This situation provides an opportunity to isolate the role of vegetation from that of climate and soils in controlling C flux in terrestrial ecosystems, and thus address one of the major research goals of the United States Global Change Research Program (USGCRP; Sarmiento & Wofsy, 1999).

We quantified net ecosystem exchange of carbon (NEE) and its components, gross ecosystem productiv-ity (GEP) and ecosystem respiration (RE), using long-term eddy-covariance measurements from old-field (OF), planted pine (PP), and hardwood forest (HW) ecosystems that represent distinct stages of ecological succession in the SE. The ecosystems are adjacent and share identical macroclimatic conditions, and identical soil type over much of their extent, such that any difference in observed biosphere–atmosphere flux is largely due to the role of vegetation (Stoy et al., 2006a). The experimental setup was designed to exam-ine ecological hypotheses regarding the roles of vegeta-tion and climate in controlling long-term C exchange.

We examine three hypotheses pertinent to ecosystem-scale processes:

(H1) At annual and growing season time scales, the variability in RE is primarily explained by variability in GEP.

(H2) Changes in annual GEP (GEPA1) are primarily determined by differences in the magnitude of an-nual canopy conductance (Gc,A) and its sensitivity to drought and disturbances.

(H3) After accounting for the effects of Gcon GEP, the remaining variation in observed GEPAis attributable to changes in ecosystem physiology in response to climatic variability, namely the response of the ratio

of leaf-internal to atmospheric CO2concentration (the Ci/Caratio) to vapor pressure deficit (D).

While current process-based models address some aspects of these three hypotheses, these models are known to fail in situations of drought or rapid distur-bances, as evidenced by recent modeling studies fo-cused on SE forested ecosystems (Hanson et al., 2004; Siqueira et al., 2006). It is envisioned that examining ecological hypotheses regarding the interplay between hydrology and C dynamics in a controlled experimental setting over long time scales will ultimately lead to model improvements.

Some preliminary experimental evidence for H1 fol-lows from recent studies demonstrating that GEP and RE are strongly coupled (Ho¨gberg et al., 2001; Ryan & Law, 2005) and also controlled by similar drivers (Reich-stein et al., 2007). With respect to H2, it is well under-stood that leaf-level photosynthesis is fundamentally coupled to water fluxes via leaf stomatal function. However, models of canopy-level photosynthesis are complicated by vertical gradients in intercellular CO2 concentrations and shifts in leaf-level physiology due to canopy nitrogen distribution and temperature acclima-tion that interact with complex light environments (Kull & Jarvis, 1995; Kull & Kruijt, 1998, 1999; Gu et al., 2002). Also, ecosystem-level water fluxes comprise both tran-spiration and evaporation. Despite these complications, canopy-scale flux measurements permit the exploration of whether changes in GEPA are dominated by Gc,A (Law et al., 2002). In addition to the strong role of hydrology via Gcin controlling GEP (and thus poten-tially RE), Brodribb & Feild (2000) and Katul et al. (2003) demonstrated that the parameters that determine cano-py photosynthesis may themselves vary with hydrolo-gic changes over longer time scales, which is examined in H3. The notion that H2 and H3 may interact such that their combined impact may be approximated by a near-constant EWUE is also explored at interannual time scales. Hence, hydrology plays a central role in deter-mining ecosystem C uptake via both canopy conduc-tance at multiple time scales and physiological adjustments to climate at longer (e.g. seasonal and annual) time scales, the latter of which has received less attention.

We focus on relationships between measured ecosys-tem-level fluxes and climatic drivers that emerge to become important at longer time scales, namely the annual and growing season time scales, and reserve discussion of short-term dynamics for cases in which they contribute to the interpretation of C flux or ecolo-gical succession on longer time scales. The discussion of growing season time scales is intended to demonstrate that the results are robust at multiple time scales and

1Throughout, flux variables with subscript A denote annual flux

sums and the subscript GS denotes April–September peak grow-ing season flux sums. The averaggrow-ing operator h i denotes annual or growing season averages.

(3)

insensitive to averaging over the wintertime period that generally has lower biological activity.

The measured flux results are discussed first in the context of the three hypotheses. We then place the results in the broader context of classic and contempor-ary ecological ideas regarding:

(E1) the role of assimilation and respiration in con-trolling the net C flux of terrestrial ecosystems (Va-lentini et al., 2000; Reichstein et al., 2007);

(E2) C exchange along ecological succession, with a focus on the ‘Strategy of Ecosystem Succession’ of E. P. Odum (1969); and correspondingly

(E3) the role of ecosystem resistance and resilience to disturbances such as droughts and ice storms in maintaining the ecosystem service of C sequestration.

Methods Site description

The study ecosystems are located in the Blackwood Division of the Duke Forest near Durham, NC (351980N, 79180W, 163 m a.s.l.). The long-term (111-year) mean annual temperature is 15.5 1C and long-term annual and April–September peak growing season pre-cipitation is 1145  180 and 632  130 mm, respectively. Climate during the 1998–2005 measurement period was variable and included late-season droughts in 2001 and 2005, a severe drought in 2002, a severe ice storm event in December 2002 (McCarthy et al., 2006), and wetter than average growing seasons in 1999, 2000, and 2003, with concomitant variability in photosynthetically ac-tive radiation (PAR) and D (Palmroth et al., 2005; Stoy et al., 2006a, b).

OF vegetation is harvested at least once a year and is dominated by the C3 grass Festuca arundinacea Schreb., with minor contributions from forbs and other C3 and C4 grass species (Novick et al., 2004). EC instrumenta-tion is at 2.8 m, and canopy height ranged from 0.1 to 1 m over the study period. PP is primarily composed of P. taeda L. with a diverse understory (Oren et al., 2001; Stoy et al., 2006a). Mean canopy height increased from 14 m in 1998 to 19 m in 2005 and EC instrumentation was raised from 15.5 to 20.2 m in January 2001. HW is an uneven-aged (80–100 years old) forest dominated by several Quercus (oak) and Carya (hickory) species (Pa-taki & Oren, 2003; Palmroth et al., 2005; Stoy et al., 2005), with a minor component of evergreen species in the overstory (P. taeda) and understory (Juniperus virginiana L.). Canopy height averaged 25 m with some treetops exceeding 30 m, and EC instrumentation is at 39.8 m.

OF, PP, and HW are adjacent and flux towers lie within 750 m of each other such that their macroclimatic

conditions are identical with minor seasonal differences in microclimate (e.g. D) due to vegetative activity (Stoy et al., 2006a). While rainfall, soil type, and rooting depth are similar across sites, soil moisture (y) need not be identical among sites. For example, earlier spring draw-down of y at PP compared with that at HW reflects differences in through-fall and transpiration, which is attributed to phenological differences between the two ecosystems (Stoy et al., 2005).

The dominant edaphic characteristics are similar among the adjacent ecosystems. All ecosystems lie on Enon silt loam, which transitions to Iredell gravelly loam in parts of OF and HW, and the soil profile of all ecosystems is dominated by a clay pan at a depth of ca. 30–50 cm. Roots were not observed below 45 cm at OF (Lai & Katul, 2000), and ET was well described by modeling root water capture in the upper 35 cm of soil. Likewise, water uptake in the upper 35 cm at PP bal-ances sap flux-measured transpiration, and a clay pan was also observed at this depth (Oren et al., 1998). Stoy et al. (2006a) used the time series of soil moisture measurements to demonstrate that the effective rooting depth at HW is of the order 50 cm, but direct soil core and pit measurements revealed few roots (o1% by mass) below 35 cm in all ecosystems (K. Johnsen, un-published data). Some edaphic differences among eco-systems cannot be ruled out (especially rooting depth), which may impact the results.

Leaf area index measurements

LAI at OF was estimated by calculating gap fractions from below-canopy PAR transmission measurements made using the 80 quantum sensor array on the Accu-PAR Accu-PAR-80 Ceptometer (Decagon Instruments, Pull-man, WA, USA). After 2001, LAI at OF was estimated by combining litterfall and LAI-2000 (Li-Cor, Lincoln, NE, USA) measurements. LAI at PP was calculated after McCarthy et al. (2007), who used a combination of needle elongation and litterfall measurements to mea-sure the contribution of overstory P. taeda trees to total LAI, and a combination of degree-day sums and litter-fall measurements to estimate LAI of the understory hardwood species. At HW, LAI was measured using a combination of LAI-2000 and litterfall measurements (Palmroth et al., 2005).

Micrometeorological methodology

Specific details regarding the EC and micrometeorolo-gical instrumentation at the OF, PP, and HW ecosystems can be found elsewhere (Katul et al., 2001; Novick et al., 2004; Stoy et al., 2006a). Briefly, fluxes were measured using open-path infrared gas analyzers (Li-Cor 7500) r2008 The Authors

(4)

coupled with sonic anemometers (CSAT3, Campbell Scientific, Logan, UT, USA). A closed-path IRGA (Li-Cor 6262) was employed before May 1, 2001, at PP. Corrections to fluxes made using the closed path system are detailed in Oren et al. (2006). A full suite of micro-meteorological measurements, including air tempera-ture (Ta), D, y, soil temperature, net radiation (Rn), and PAR, were made in conjunction with each EC system. The value of y was intensively measured throughout the active rooting depth at all ecosystems using a combination of CS615 sensors (Campbell Scientific) and type ML1 ThetaProbe sensors (Delta-T Devices, Cambridge, UK).

A detailed analysis of the source weight function (used to compute the footprint) of EC-measured turbu-lent fluxes was performed to ensure that the flux source area did not exceed ecosystem dimensions or receive contamination from the nearby elevated CO2 rings of the Duke Forest FACE experiment at PP, or the private-land clear cut that occurred in December 2002, ca. 200 m south of the HW tower (Stoy et al., 2006b). A semi-analytical footprint model originally developed by Hsieh et al. (2000) and extended to two dimensions by Detto et al. (2006) was used to quantify the dimensions of the flux footprint for half-hourly periods.

It is necessary to estimate transpiration (T) and there-by Gc(through unit conversion and measured D) for the present analysis, which seeks to describe EC-measured fluxes in an ecological and hydrological context. We followed the approach of Stoy et al. (2006a, b), who found a relationship between modeled below-canopy radiation (Campbell & Norman, 1998) and soil evapora-tion (E). In this approach, measured ET during periods when the canopies were inactive [i.e. during leaf off at OF and HW and when Ta was below 10 1C at PP (Scha¨fer et al., 2002)] and not wet were identified and related to modeled forest floor radiation. This relation-ship was applied during all periods to produce a time series of E estimates. T and Gcwere then estimated by difference. These EC-based estimates of T (500– 560 mm yr1 for 2001–2004 excluding severe drought) closely matched sap flux estimates from PP, which ranged from 520 to 530 mm yr1 (Scha¨fer et al., 2002; Stoy et al., 2006a).

Calibration strategy

A three-stage calibration procedure was adopted to ensure that the flux measurements were accurate and the flux sums are defensible. The calibration parameters of all three Li-Cor 7500 instruments were found to be stable over time, and each was calibrated biannually with minimal change in calibration coefficients (o3% in all cases). The EC measurement systems at PP and HW

were shown to be in good agreement with the Ameriflux roving system. To ensure that long-term flux sums are defensible, EC-measured CO2 flux time series were compared with against available independent measure-ments and model results from C budgeting approaches (Hamilton et al., 2002), inverse models (Lai et al., 2002b; Juang et al., 2006), physiology-based forward models (Luo et al., 2001), chamber respiration measurements (Palmroth et al., 2005), and constraints on assimilation based on sap flux and evapotranspiration (ET) measure-ments (Pataki & Oren, 2003; Scha¨fer et al., 2003; Stoy et al., 2006a), as described by Stoy et al. (2006b). This comparison was performed to minimize and identify potential sources of error and bias in the estimation of NEE, GEP, and RE. The nonrectangular hyperbolic (NRH) method (Gilmanov et al., 2003) resulted in the best ‘defensible’ EC-based GEP and RE estimates after employing an atmospheric stability filter to remove nighttime measurements taken under conditions of insufficient turbulence (Novick et al., 2004; Stoy et al., 2006b). As discussed in Cava et al. (2004), very stable nocturnal conditions decouple the local flux measure-ments above the canopy from the CO2 production inside the canopy volume. It should be noted that harvesting removes ca. 200 g C m2yr1on an average from OF, and that the disagreement between EC and harvesting measurements likely represents EC-mea-surement uncertainty rather than other losses of C from the ecosystem, for example via the export of dissolved organic C rather than chronic soil C loss. Jaksic et al. (2006) found that the uncertainty and interannual varia-bility of NEE in a grassland ecosystem were of similar magnitude.

Flux error

The present analysis does not focus on errors in mea-sured fluxes, as a full discussion of these errors has been presented elsewhere (Oren et al., 2006; Stoy et al., 2006a, b). However, previous findings are reviewed for completeness. Error in NEEA at PP estimated after Goulden et al. (1996) varied between 79 and 127 g C m2yr1, due largely to uncertainty in gapfilling missing data (47–93% of estimated error) and spatial variability in fluxes (6–49%) rather than instrument error (1–6%; Oren et al., 2006). Error values represent 1 SD about the estimated mean annual NEE. NEEAerror ranged from 42 to 68 g C m2yr1at OF and between 84 and 113 g CO2m2yr1at HW (Stoy et al., 2006b). Error in GEPA and REA was of the order of 9–30% and averaged 16% for the study period at all ecosystems (Stoy et al., 2006b). Missing data comprised 41%, 43%, and 40% of the OF, PP, and HW time series, respectively, somewhat higher than many other eddy-covariance r2008 The Authors

(5)

sites (Falge et al., 2001), in accordance with our strict nighttime data filter (Novick et al., 2004). ETAestimates contained instrument and gapfilling error of the order of 7–14% for all ecosystems (Stoy et al., 2006a).

We note that the 30-min sum of sensible and latent heat fluxes explained 66–75% of the measured Rnat the three sites (Stoy et al., 2006a). It was shown by Stoy et al. (2006a) that the daily energy balance was nearly closed during days dominated by near-neutral atmospheric stability, while the largest average imbalance was ob-served during days dominated by near-convective con-ditions. This implies that the 30-min flux-averaging period may be adequate for near-neutral conditions, but appears to be filtering out nontrivial contributions of low-frequency eddies, the length scales of which are comparable to the (deep) convective atmospheric boundary layer height. This low-frequency loss under near-convective conditions may be by far more severe for sensible heat flux, rather than latent heat flux, because entrainment fluxes at the capping inversion can be as large as 30% of the surface sensible heat fluxes (Kim & Entekhabi, 1998; Juang et al., 2007), and it is these entrainment fluxes that may be sampled under convective conditions. The broad agreement between ecosystem water balance studies and flux results at PP lends support to the notion that latent heat fluxes are not underestimated (Scha¨fer et al., 2002; Stoy et al., 2006a).

We additionally note that the EC system measures the turbulent flux crossing a horizontal plane above the given ecosystem, and internal recycling of C below the sensors cannot be measured. Some authors conse-quently prefer that ecosystem C uptake be called GEP rather than GPP (Goulden et al., 1997; Stoy et al., 2006b). The differences between these two terms are likely to be minor, and we assume that GEP and GPP are compar-able as per the FluxNet convention. These two terms are, thus, used interchangeably here.

Analysis

One of the goals of the present study is to analyze the coupled dynamics of carbon and water fluxes to ad-dress H2 and H3. To link the two fluxes mechanistically, we used Fick’s law of diffusion:

GEP ¼ eGcCa 1  Ci Ca

 

; ð1Þ

where Ca is the atmospheric CO2 concentration, e ( 0.625) corrects for the difference in molecular diffu-sivity between H2O and CO2, and Ci/Cais the ratio of internal to external [CO2] at the canopy scale and represents the ‘driving force’ for canopy C uptake.

Defining h i as the annual or seasonal averaging opera-tor, Eqn (1) can be used to explore mean annual GEP via

GEP h i  eCa hGci 1  Ci Ca       : ð2Þ

We also investigate the April–September peak grow-ing season period to demonstrate that the analysis is robust at multiple time scales. Note that the approxima-tion in the above formulaapproxima-tion is due to the fact that Gh cCi=Cai ¼ Gh ci Ch i=Cai þ rGc;Ci=CasGcsCi=Ca and not hGci hCi/Cai, where rGc;Ci=Ca is the correlation coefficient between Gcand Ci/Ca, and sxis the standard deviation of an arbitrary variable x. Katul et al. (2000) used a linearized analysis of the Farquhar et al. (1980) photo-synthesis model to demonstrate that rGc;Ci=CasGcsCi=Ca

Gc

h i Ch i=Cai

 1

when hGci6¼0. We used Gc and Ci/Ca estimates from Eqn (1) to determine that the value of rGc;Ci=CasGcsCi=Ca

Gc

h i Ch i=Cai is less than 0.015 for all ecosystems at the annual time scale. Hence, the observed variation in GEPA(or GEPGS) can be decomposed into the product of variation in both Gc,A and Ci/Ca,A (or corresponding growing season averages) with minimal error, noting also that annual Ca was relatively invariant over the 5–8-year periods analyzed here.

Combining Eqn (1) with the Fick’s law analogy for T ( 5 lGcD, where l is a unit conversion factor) gives an ecosystem-scale approximation of water-use efficiency (EWUE):

EWUEAGEPA TA

¼eCað1  Ch i=CaiÞ

lh iD ; ð3Þ

which upon re-arranging yields Ci=Ca h iA 1  l eCa EWUEA   D h i: ð4Þ

Growing season estimates for these variables are obtainable by substituting growing season sums or averages.

We also analyze a temporally averaged model for Gc,A after Oren et al. (1999):

Gc;A¼ gs;AhLAIi ¼ ah i PARh i 1  m ln Dð h iÞ LAIh i; ð5Þ where the stomatal sensitivity parameter m takes the theoretical value of 0.6 (Oren et al., 1999). We use this model to interpret changes in Gcin the analysis of H2 and H3.

Results

The major results of the long-term EC measurements are described first, followed by a presentation of results that relate to the experimental hypotheses, namely the relationships between GEP and RE and between Gcand r2008 The Authors

(6)

GEP, and the role of parameter variability in maintain-ing a near-constant EWUE.

At OF, GEPAand REAwere nearly in balance, produ-cing a near-zero mean NEEA(Table 1, Figs 1–3). How-ever, it is important to note that adding the ca. 200 g C m2yr1 removed by harvesting results in an imbalance that likely represents flux uncertainty as discussed in Methods. Although mean NEEA at PP and HW were almost identical (ca. 450 g C m2yr1), the variability (SD) in NEEAat PP was nearly fourfold that of HW due to large interannual variations in both GEP and RE (Figs 1–3). The magnitude of NEEAat PP was over twofold less during severe drought (2002) and immediately following the ice storm (2003) than its maximum of over 600 g C m2yr1, which occurred during the 2 years with late-season droughts (2001 and 2005) (Figs 1 and 2).

OF was a source of C to the atmosphere during the peak of the severe droughts of 2002 and 2005 (Figs 1 and 2). However, NEE switched sign rapidly after drought-breaking rains (indicated by the stars in Fig. 2; see Fig. 2a in Stoy et al., 2006a for monthly precipitation time series) such that NEEAat OF remained near zero as a result. In contrast, the magnitude of NEEA at PP de-creased dramatically in response to the severe drought from 610 g C m2yr1in 2001 to 270 g C m2yr1in 2002 (Figs 1 and 2). The magnitude of NEE remained low in 2003 due to the impacts of the December 2002 ice storm, and then took 2 years to recover to its previous maximum. This sequential decrease and increase in the magnitude of NEE is tracked by arrows in Fig. 2. NEE at HW was relatively resistant to the wide range of cli-matic variability observed.

When the annual mean and annual variability of GEP and RE are plotted against ecosystem age, the general shape of the relationship is similar to that given by Odum (1969) (Fig. 3), but NEEAdid not approach zero in our case because of the significant relationship be-tween REA and GEPA for all site-years of flux data

(Po0.0002). REAwas significantly related to GEPA at OF and PP (Po0.005), and this pattern also held at the growing season time scale (Po0.02). The relatively invariant annual and growing season fluxes at HW failed to produce a strong relationship. Using monthly averages, GEP explained 84% of the variation in RE at OF, 71% at PP, and 63% at HW (Fig. 4). Both linear and exponential models using monthly average Ta or soil temperature explained a lesser amount of the RE varia-tion, ca. 68%, 50%, and 56% for the case of Taat the three ecosystems, respectively.

GEPA, not REA, was significantly related to NEEA after pooling measurement years from all ecosystems (Po0.001), and this relationship also held at the grow-ing season time scale. However, REAexplained a larger degree of the variability in NEEAat HW (r250.6) than did GEPA(r250.1). The relationship between REGSand NEEGS at HW was significant (r250.94; P 5 0.0071) as opposed to the relationship between GEPGSand NEEGS at HW (P 5 0.83). We note that these relationships are a secondary issue given the small variations in NEEAat HW and the uncertainty in gap-filled REA.

GEPAand GEPGSwere closely related to T and there-by Gc2 at OF (A: r250.77, P 5 0.048; GS: r250.85, P 5 0.025) and PP (A: r25

0.62, P 5 0.021; GS: r250.84, P 5 0.012), but not HW (A: r250.0046, P 5 0.91; GS: r250.53, P 5 0.16) (Fig. 5). Again, the weak correlation at HW is primarily due to the small interannual varia-bility of GEPAand Gc,A.

Guided by various leaf-level studies including Leun-ing (1995), Farquhar et al. (1993), Lloyd & Farquhar (1994), Cowan & Farquhar (1977), and Wong & Dunin (1987), the notion that ecosystem-level Ci/Cafrom Eqn (1) is related to mean annual and growing season D at all ecosystems [Eqn (4)] is explored in Fig. 6 (A. OF: r250.59, P 5 0.02; PP: r250.84, P 5 0.001; HW: r250.97, P 5 0.002; GS. OF: r250.60, P 5 0.12; PP: r250.71, P 5 0.009; HW: r250.98, P 5 0.001). The line-arity of this dependence (Fig. 6) has important implica-tions to the near-constant EWUE, discussed later. In contrast to the strong relationships between Ci/Caand D at all ecosystems and time scales (except the growing season time scale at OF), well-known relationships between PAR and GEP and between Ta and RE were significant at short time scales using half-hourly data (Po0.05), but not at the annual or growing season time scales (P40.05) for all ecosystems.

Table 1 The mean and variability (SD in parentheses) of the annual net ecosystem exchange of CO2(NEE) and its

compo-nents – gross ecosystem productivity (GEP) and ecosystem respiration (RE) at old-field (OF), planted pine (PP) and hard-wood forest (HW) ecosystems in the Duke Forest, NC

Ecosystem NEE GEP RE GEP/RE

OF 10 (40) 1230 (210) 1240 (230) 0.99 (0.03) PP 460 (190) 1890 (390) 1440 (340) 1.33 (0.15) HW 440 (50) 1710 (30) 1260 (50) 1.35 (0.05) Signs follow the micrometeorological convention where flux from atmosphere to biosphere is denoted as negative. Flux units are in g C m2yr1.

2G

c(usually expressed in mol m2time1) can be converted to T

(usually expressed in mm time1) by considering the latent heat of vaporization and vapor pressure deficit, both functions of Ta.

Throughout, we use Gcwhen referring to Eqn (1), and T when

referring to annual or seasonal canopy water flux.

(7)

The parameters that describe Gc [Eqn (5)] varied at the annual time scale in response to canopy structure and diffuse radiation. The mean annual (or growing season) light sensitivity of canopy conductance (hai) increased with decreasing hLAIi owing to less self-shading of the canopies (Fig. 7a and b; A: r250.85, Po0.001; GS: r250.77, Po1.6  106). hai also in-creased with an increasing ratio of diffuse PAR in the forest ecosystems (Fig. 7c and d).

Discussion

The hypotheses are first discussed in the context of the experimental results. Next, the ecological implications

of the experimental findings are explored via E1–E3, and their relevance to ecosystem management is also presented.

Hypothesis 1: GEP is the primary determinant of RE In the ecosystems studied here and, importantly, at the time scales that we have measured, variability in RE at annual, growing season and monthly time scales ap-pears to be more related to GEP than to total biomass (see Table 2 in Stoy et al., 2006a) or extant environmental drivers such as temperature (see also Law et al., 2002) that act upon the entire ecosystem C pool. This is despite the fact that most short-term RE models are

200 OF PP HW –200 –400 –600 0 –500 –1000 –1500 –2000 –2500 1600 1200 800 400 0 0 NEE (g C m –2 ) GEP (g C m –2 ) RE (g C m –2 ) 0 100 200 300 0 100 Day of year 200 300 0 100 200 300 1998 1999 2000 2001 2002 2003 2004 2005

Fig. 1 Cumulative annual net ecosystem exchange of CO2(NEE), gross ecosystem productivity (GEP), and ecosystem respiration (RE)

at the adjacent old-field (OF), planted pine (PP), and broadleaf deciduous (HW) forests in the Duke Forest, NC. The measurement record began in 1998 at PP and in 2001 at OF and HW. Signs follow the micrometeorological convention where flux from atmosphere to biosphere is denoted as negative. Flux error is discussed in Oren et al. (2006) and Stoy et al. (2006a, b).

(8)

driven by temperature (Morgenstern et al., 2004), and temperature-driven models well describe daily Rsoilat PP and HW under conditions of adequate soil moisture (Palmroth et al., 2005). Although RE is likely to vary with ecosystem biomass, the variability in RE appears

to be dominated by C pools with short turnover times rather than recalcitrant C pools in the stem or soil (e.g. Taneva et al., 2006). Future work should disentangle the central or complementary role played by GEP in deter-mining the magnitude of RE (Janssens et al., 2001; Ryan & Law, 2005), and also investigate the temperature response of the different C pools (Bosatta & A˚ gren, 1999), noting that labile C pools from recent photoassi-milates may dominate RE in many ecosystems (David-son & Janssens, 2006).

The results here conceptually agree with a number of recent studies that have quantified the importance of GEP to Rsoil (Ho¨gberg et al., 2001; Ekblad et al., 2005; Tang et al., 2005) and RE (Janssens et al., 2001). Rsoil dominates RE at HW and comprises a large proportion of RE at PP (Scha¨fer et al., 2003; Mortazavi et al., 2005; Palmroth et al., 2005). However, based on this analysis with eddy-covariance data, we cannot determine if GEP and RE are fundamentally coupled (H1), or if they simply co-vary, because they are controlled by similar environmental drivers (Reichstein et al., 2007). There is some indirect evidence for the former given that both Rsoiland RE at PP and HW show the isotopic signature of recently assimilated C (Andrews et al., 1999; Morta-zavi et al., 2005), and some 70% of Rsoilat PP arises from pools with turnover times of 1 month or less (Taneva et al., 2006). However, Stoy et al. (2007) investigated the 2001 –800 –700 –600 –500 –400 –300 –200 –100 2001.5 2002 2002.5 2003 2003.5 2004 2004.5 2005 2005.5 2006 0 100 200 * * Year OF PP HW Cumulative NEE (g C m –2 )

Fig. 2 Same as Fig. 1a–c, detailing differences in the cumulative sum of net ecosystem exchange of C (NEE) for each year during 2001– 2005, the period over which measurements were available at all three study ecosystems. Asterisks denote the timing of drought-breaking rains in 2002 and 2005. Arrows indicate changes in annual NEE at the planted pine (PP) ecosystem. Error bars indicate  1 SD about the mean annual estimate calculated after Goulden et al. (1997) by Stoy et al. (2006b).

Fig. 3 Change in mean annual GEP and RE estimates (thick lines) from the initial condition (OF) to early successional (PP) and late successional (HW) forests with annual variability (stan-dard deviation) denoted as error bars. The relationships of Odum are drawn for comparison in thin lines, noting that gross primary productivity (GPP) rather than GEP was used in the original study. The original figure did not reference an ordinate; Fig. 1a of Odum (1969) was digitized and scaled to approxi-mately match the flux magnitude observed here. Years are approximate.

(9)

coupling between photosynthesis and soil respiration at relatively short time scales (of the order of days) at PP and HW and found little evidence of a strong coupling. It was argued that multiple turnover times in the labile C pool in both plant and soil obscured any obvious pulse-response dynamics between ecosystem C uptake and loss. Again, the important finding here is the strong correlation between GEP and RE, with respect to en-vironmental drivers, at seasonal and annual time scales, although the precise causation remains to be explored.

Hypothesis 2: Gcis the primary determinant of GEP Changes in GEPAare linearly related to differences in the magnitude of annual canopy conductance (Gc,A) and its sensitivity to drought and disturbances (Fig. 5), as predicted by the mechanistic model for diffusive flux, Fick’s law [Eqn (1)], irregardless of averaging via Eqn (2). Notwithstanding the small scatter for the HW site, it is interesting to note that the inverse of the slopes of the GEPA/T relationships for OF, PP, and HW were

(a) (b) (c)

Fig. 4 Relationship between the monthly sums of gross ecosystem productivity (GEP) and ecosystem respiration (RE) at the three study ecosystems. OF, old-field; PP, planted pine forest; HW, hardwood forest (P  0.05 in all cases).

200 300 400 500 600 −2600 −2400 −2200 −2000 −1800 −1600 −1400 −1200 −1000 −800 GEP A (g C m2 yr1 ) TA (mM yr−1) (a) OF PP HW 100 200 300 400 500 −1800 −1600 −1400 −1200 −1000 −800 −600 GEP GS (g C m2 GS1 ) TGS (mM GS−1) (b)

Fig. 5 (a) Relationship between annual transpiration (TA) estimated using the radiation attenuation approach of Stoy et al. (2006b) and

annual gross ecosystem productivity (GEPA) at the old-field (OF), planted pine (PP) and hardwood forest (HW) study ecosystems. Error

in GEP and T estimates is discussed elsewhere (Oren et al., 2006; Stoy et al., 2006a, b). (b) Same as part (a), but using April–September peak growing season (GS) sums (OF. A: r250.77, P 5 0.048; GS: r250.85, P 5 0.025. PP. A: r250.62, P 5 0.021; GS: r250.84, P 5 0.012. HW. A: r250.0046, P 5 0.91; GS: r250.53, P 5 0.16).

(10)

comparable (3.4, 3.8, and 3.8 g C kg1H2O, respec-tively), if we assume that GEPAis negligible for negli-gible T (i.e. if a point at zero flux is included in the computation). These values are 3.2, 3.4, and 3.3 g C kg1H2O for OF, PP, and HW, respectively, at the growing season time scale and collectively suggest that the long-term ecosystem water-use efficiencies are comparable across the three ecosystems, in further support of H2 and H3.

Hypothesis 3: parameter variability describes remaining variability in GEP

The relationship between Ci/Caand D [Eqn (4)] at short (e.g. half-hourly) time steps in C3 species has been demonstrated often (e.g. Leuning, 1995; Katul et al., 2000), and D is known to be one of the most important explanatory variables for modeling weekly to monthly GEP at PP (Stoy et al., 2005). The approximately linear relationship between hCi/Cai and hDi at the annual and growing season time scales (Fig. 6) can be analyzed in the context of constant water-use efficiency for the different canopies.

One might expect hCi/CaiA to vary linearly as a function of hDiA (Fig. 6) if EWUEA is near constant, as already suggested by the slopes of Fig. 5 (H2). The Ci/Ca ratio is related to canopy physiology in that it reflects the driving force for C uptake from Eqn (1). Equation (4) demonstrates that canopy physiology re-sponds linearly to atmospheric demand for water (i.e. D) at longer time scales if EWUE is near constant. Note also that the slopes in Fig. 6 support the comparable EWUE found in Fig. 5 for PP and OF. When compared with the inverse of the slopes in Fig. 5, the analysis in Fig. 6 is perhaps more revealing about the constant EWUEAfor HW, because the spread in hDiAand hDiGS at the three ecosystems is now comparable. Interest-ingly, the analysis in Fig. 6 suggests that the EWUEAis approximately equal at all three ecosystems, but slightly larger at HW and PP at the annual and growing season time scales, respectively. In short, while many internal leaf-level physiological parameters for each of the three ecosystems are not stationary across years (e.g. Ells-worth, 1999), it appears that their cumulative impact may still be captured by a near-constant EWUE for this semidecadal time-scale analysis. Furthermore, the

var-1.2 1.3 1.4 1.5 1.6 0.64 0.66 0.68 0.7 0.72 0.74 0.76 0.78 0.8 (a) OF PP HW 1.3 1.4 1.5 1.6 1.7 0.6 0.62 0.64 0.66 0.68 0.7 0.72 0.74 0.76 0.78 0.8 <D >GS (kPa) <D >A (kPa) Ci /C a ,GS Ci /C a ,A (b)

Fig. 6 (a) Relationship between mean annual vapor pressure deficit (hDiA) and the ratio of leaf-internal to atmospheric [CO2] at the

annual time scale [Ci/Ca,A, Eqn (1)] in the study ecosystems. OF, old-field (squares); PP, planted pine forest (triangles); HW, hardwood

forest (circles). Note that for constant water-use efficiency, the relationship between Ci/Caand D must be linear [Eqn (6)]. (b) Same as part

(a), but using April–September peak growing season (GS) averages (OF. A: r250.59, P 5 0.02; GS: r250.60, P 5 0.12. PP. A: r250.84,

P 5 0.001; GS: r250.71, P 5 0.009. HW. A: r250.97, P 5 0.002; GS: r250.98, P 5 0.001).

(11)

iation of long-term EWUE across the three ecosystems is surprisingly small, perhaps suggesting that in a first-order analysis, its bulk value is set by climate and soil type rather than ecosystem type.

This result also implies that simple mechanistic mod-els for GEP may be obtained by combining Eqn (1) with mechanistic models for Gcand the relationship between Ci/Caand D (Leuning, 1995). Models that employ light-response curves for estimating GEP should ensure that both the short-term and long-term dynamics of GEP are accurately estimated, as we found no significant relationship between PARA and GEPA (or PARGS and GEPGS) at our study ecosystems. It is important to reiterate that we also found no relationship between mean annual air and soil temperature and REA (Jans-sens et al., 2001; Law et al., 2002), despite the widespread use of temperature-based respiration models (Morgen-stern et al., 2004).

In addition to the case of Ci/Caand D, there are other examples in which ecosystem parameters vary with respect to climatic or biological conditions that have clear implications for ecological modeling. The

para-meters that describe Gcvaried at the annual time scale in response to canopy structure and diffuse radiation [Eqn (5)]; the mean annual (or growing season) hai increased with decreasing hLAIi owing to less self-shading of the canopies (Fig. 7a and b) and an increas-ing ratio of diffuse PAR in the forest ecosystems (Fig. 7c and d), consistent with studies on whole-canopy re-sponse to diffuse radiation (Gu et al., 2002). Diffuse radiation can penetrate plant canopies more readily and has been shown both theoretically and experimentally to increase canopy photosynthesis over direct radiation (Gu et al., 1999, 2003; Law et al., 2002). Thus, long-term ecosystem responses to hydrology and radiation quality controlled annual C fluxes in the study ecosystems, rather than the responses to temperature and radiation quantity.

E1: controls on NEE by component fluxes

If GEP and RE are related (H1), the question becomes: how do they interact to determine the C balance along successional time scales? Results here provide a

differ-0 1 2 3 4 0.5 1 1.5 2 2.5 3 3.5 a (mol µ mol1 photons) a (mol µ mol1 photons) a (mol µ mol1 photons) a (mol µ mol1 photons) (a) OF PP HW 0 2 4 6 0.5 1 1.5 2 2.5 3 3.5 <LAI>GS <LAI>A (b) 45 50 55 60 65 6 8 10 12 14×10 −5 ×10−5 ×10−4 ×10−4 (c) 45 50 55 60 65 6 8 10 12 14

Dir. / Max. PARGS (%) Dir. / Max. PARA (%)

(d)

Fig. 7 (a) Relationship between the light sensitivity of canopy conductance at the annual time scale (a) to mean annual leaf area index (hLAIi) at the old-field (OF), planted pine (PP), and hardwood forest (HW) ecosystems in the Duke Forest, NC. (b) Same as part (a), but for growing season (GS) averages (A: r250.85, Po0.001; GS: r250.77, Po1.6  106

). The relationship between haiAand the fraction of

direct (Dir.) to maximum photosynthetically active radiation (PAR) at the annual (c) and April–September peak growing season (d) time scales (A: r250.39, P 5 0.022; GS: r250.40, P 5 0.020).

(12)

ent picture than continental-scale studies of forested ecosystems, which suggest that variability in RE pri-marily controls the C balance across space (Valentini et al., 2000). Rather, because of the coupling between GEP and RE, the former may play a more important role in determining long-term C fluxes across different ecosystems (Janssens et al., 2001; Reichstein et al., 2007). Variability in GEPAwas in general more related to variability in NEEA at OF and PP, but NEEA was nearly constant at HW when compared with OF and PP, making it difficult to discern the relative importance of its component fluxes in determining its variability. Taken as a whole, these results emphasize the relative importance of GEP over RE in determining the medium (monthly) and long-term (interannual) variability in NEE, despite the strong relationships between instanta-neous (half-hourly) RE and temperature.

E2: C exchange along succession

NEEAat OF was near zero, PP was the most productive of the three ecosystems under ideal conditions, and NEEAat HW was highly resistant to climatic variability (Table 1, Figs 1–3). These results largely followed ex-pectations based on the ‘Strategy of Ecosystem Devel-opment’ of Odum (1969), which hypothesizes that GPP rapidly increases with ecological succession, then de-creases as forests age, while RE inde-creases monotonically due to the increase in autotrophic biomass. There is an important difference between flux results and the ‘Strat-egy of Ecosystem Development’; the covariance be-tween GEP and RE (H1) resulted in a decrease in RE along the forest succession studied here such that mean NEEA at PP and HW was almost identical over the respective 8-year and 5-year measurement periods (Table 1), as well as the commonly measured 5-year period.

It is important to note that Odum’s hypothesis was based on a single time series of temperate forest succes-sion (Kira & Shidei, 1967), as is the present study. Regardless, Odum’s theory is conceptually appealing given its simplicity, and is often presented in textbooks (e.g. Odum, 1971; Schlesinger, 1997), leading genera-tions of ecologists to acknowledge its basic tenets. Whereas our experimental evidence largely agrees with its assumptions, modifying these ‘classic’ ideas by acknowledging a covariance between GEP or RE (Ho¨g-berg et al., 2001; Janssens et al., 2001; Ryan & Law, 2005) may provide a realistic picture of C exchange along succession while retaining the simplicity of Odum’s hypothesis, although this assertion must be tested using many and various examples of ecological succession. ‘Modernizing’ the ‘Strategy of Ecosystem Development’ (Odum, 1969) to take into account the relationship

between GEP and RE may improve understanding of biosphere–atmosphere C exchange over successionary time scales and, combining our results with others, may help dispel the prevailing idea that mature forest ecosystems with large C pools must be small C sinks (see Carey et al., 2001; Ro¨ser et al., 2002; Knohl et al., 2003; Zhou et al., 2006; Urbanski et al., 2007; Baldocchi, 2008).

E3: ecological resistance and resilience

Odum (1969) additionally hypothesized that early suc-cessional ecosystems maximize productivity, while late successional ecosystems maximize ‘protection’ against (i.e. resistance to) environmental variation. Indirect evidence of this characteristic can be identified in the NEE measurements: NEEA at PP was highest under ideal conditions, but was also highly variable in re-sponse to environmental variability, while the interann-ual variability of NEEAat HW was small (Table 1, Figs 1–3).

We can indirectly explore these ideas in the context of ecological resistance and resilience for the case of bio-sphere–atmosphere interaction (Moorcroft, 2003). A logical first step follows from examining the sensitivity of C uptake to changes in Gcafter Eqn (1), which have been thoroughly quantified for the study ecosystems (Oren et al., 1998; Scha¨fer et al., 2002; Pataki & Oren, 2003; Stoy et al., 2006a).

GEPA and GEPGS were strongly related to Gc,A and Gc,GS, respectively (H2, Fig. 5), and the observed inter-annual changes in Gcare explainable by the sensitivity to and recovery from drought and disturbance (i.e. ecological resistance and resilience to perturbations) of canopy dominant species (Oren et al., 1998; Oren & Pataki, 2001; Pataki & Oren, 2003; Novick et al., 2004; Stoy et al., 2005, 2006a; Siqueira et al., 2006). Long-term changes in Gc,GSat OF occurred via a decrease in LAI in response to drought and an increase in the intrinsic mean stomatal conductance (gs) during wet years (Stoy et al., 2006a). At PP, Gc,GS was reduced due to low gs during drought and low LAI following a severe ice storm event in December 2002 (McCarthy et al., 2006; Stoy et al., 2006a). Oren & Pataki (2001) and Pataki & Oren (2003) demonstrated that most species at HW are drought tolerant; only Liriodendron tulipifera decreased Gcwith declining y. We add that soil water savings from later leaf-out phenology at HW resulted in Gc,GS that was relatively invariant among years (Stoy et al., 2005, 2006a), although access to deeper water sources cannot be entirely dismissed (as noted in the experimental setup). Consequently, the coefficient of variation (CV) of annual Gcwas 0.13 at HW compared with 0.29 at OF r2008 The Authors

(13)

and 0.23 at PP, consistent with Odum’s notion of ‘increasing protection’ as succession progresses.

It is clear that larger canopy investment at PP resulted in longer recovery times (less resilience) from distur-bance than at OF (Figs 1 and 2). Years with large NEEA followed years with ideal growing conditions, which resulted in large leaf area indices (McCarthy et al., 2007). The low soil moisture conditions observed late in the growing season in 2001 and 2005 are known to reduce the magnitude of soil respiration (Rsoil) at both PP and HW (Palmroth et al., 2005). Thus, a combination of drought, disturbance, and lagged recovery from ice storm damage were the primary sources of variability in NEEAat PP. NEEAat HW was comparatively resilient to the drought and the disturbances encountered over the measurement period.

The result that the hardwood forest (HW) was insen-sitive to drought compared with the adjacent coniferous forest (PP) contrasts with previous results in Canada and Europe (Kljun et al., 2006; Granier et al., 2007), which have found coniferous forests to be generally less drought sensitive (Baldocchi, 2008). Our results agree instead with flux research, demonstrating that more mature forests are less drought sensitive (Law et al., 2001). Again, the observed ecosystem-level mechanism for the drought resistance at HW is that later leaf-out, coupled with soil water savings from the underlying clay pan, allows HW to spend less time with low y during the growing season compared with PP (Palmroth et al., 2005), despite broad similarities in ETGSbetween the two forested ecosystems (Stoy et al., 2005, 2006a).

The resistance and resilience to disturbance of the study ecosystems may have important implications for ecosystem succession and thus long-term C flux (Odum, 1969; Moorcroft, 2003). Ecosystem stability, defined here as the tendency of an ecosystem to remain in its current state, can be formally analyzed by explor-ing whether the ordinary differential equation dB/ dt 5 f(B) has multiple equilibria, as well as the stability of these equilibria, where f(B) describes the carbon gains and losses from the autotrophic system. A symptom of instability is a hysteresis curve when ecosystem state via B is plotted against the rate of change in B (i.e. dB/ dt) in the phase plane. Hysteresis suggests that forward and backward shifts occur at different critical condi-tions, thereby signaling possible existence of multiple equilibria with an unstable zone between their basins of attraction (Scheffer et al., 2001). In our context, these state changes correspond to changes in ecosystem structure that signify ecosystem succession.

To simplify this analysis, consider the autotrophic biomass budget equation, given by

dB=dt ¼ GPP  Ra L; ð6Þ

where Rais the autotrophic respiration and L represents rate of C losses due to litterfall, exudation, or other factors such as harvesting. Noting that NEE 5 GEPRE and RE 5 Ra1Rh, where Rhis the heterotrophic respira-tion, results in

dB=dt ¼ GPP  ðRE  RhÞ  L ¼ NEE þ Rh L; ð7Þ and that dB/dt would be proportional to NEE if Rh–L is relatively invariant at the annual time scale in compar-ison to NEE [i.e. treated as an intercept in Eqn (7)].

If B 5 Ba1Bb, where subscripts a and b refer to the above and belowground autotrophic system, respec-tively, then Ba is proportional to LAI and woody bio-mass, and Bbis proportional to root biomass, the latter being proportional to the root area index (RAI). Many studies on plant hydraulics support the idea that RAI and LAI may not be independent for a given rooting zone depth (Jackson et al., 2000), nutrient input (Ewers et al., 2000), or soil texture (Sperry et al., 1998, Hacke et al., 2000). For example, Sperry et al. (1998) used a model based on Darcy’s law for hydraulic conductance to show that the ratio of root area to leaf area should fall within limits to avoid plant cavitation, and that this ratio varies as a function of soil properties. If so, B may be proportional to LAI. Hence, only for the purposes of exploring hysteresis in the phase plane at a given site – rather than proposing a prognostic model – we can assume that B is proportional to LAI, and that dB/dt is proportional to NEE at the annual and growing season time steps. We can then explore a surrogate phase plane for ecosystem stability by plotting hLAIiAand hLAIiGS vs. NEEAand NEEGS, respectively (Fig. 8).

A large hysteresis in the phase space is evident for OF and PP, and a smaller hysteresis was observed at HW at both annual and growing season time scales (Fig. 8). For OF, the hydro-climatic fluctuations produced large var-iations in hLAIi and induced a sign shift in NEEA, but had a minor impact on the magnitude of NEEAwhen compared with PP. The converse is true for PP; changes in hLAIi were relatively small compared with OF, but NEEA showed large variability in response to these changes. HW showed small responses in hLAIi and NEEA compared with the early successional ecosys-tems. This analysis suggests that earlier successional stages – represented by OF and PP – are relatively ‘unstable’ and hence less resistant to disturbances. The instability of these two ecosystems seems to originate from different attributes with potentially similar con-sequences: (1) the instability in OF (mainly along the abscissa) will directly allow other types of vegetation to invade during times in which LAI is low and can only recover slowly. (2) The instability in PP (mainly along the ordinate) is a consequence of loss of vigor in C accumulation (dB/dt), with an indirect effect on stand r2008 The Authors

(14)

vegetation dynamics. The drought and ice storm re-duced overstory pine LAI (McCarthy et al., 2006, 2007) and resulted in a decrease in NEEA(Figs 1 and 2). This loss of overstory LAI appeared to benefit the growth of understory hardwood species, which will eventually dominate this forest as it follows the typical succes-sional trajectory (Oosting, 1942; Johnston & Odum, 1956). The minor response of the HW state and rate variables in this simplified analysis indicates that only a major disturbance, for example from a long series of consecutive dry years or via a stand-replacing distur-bance such as a hurricane, could push the forest into a new state. Thus, classic ecological theories of succession can be related to modern ideas of ecological resistance and resilience via measurements of changes in ecosys-tem activity and state from flux and canopy leaf area measurements, respectively.

Following this analysis, it is apparent that the timing and rate of recovery from disturbance may determine the degree to which forest tree species succeed in colonizing OF. A disturbance that reduces LAI will increase exposure and thus light availability. Woody species may benefit if grass LAI replacement is slow. To further assess resilience to disturbance over short time

scales at OF and its potential consequences for succes-sion, the cumulative sum of NEE for 2 weeks before the annual harvest at OF vs. the cumulative sum of NEE for 3 weeks afterwards is plotted (Fig. 9), noting that LAI estimates immediately after the mow were not available for all measurement years. To simplify this analysis, only 2 years with similar early-season harvest dates but different hydrologic regimes, namely 2002 and 2004 (Table 2), are considered.

The response of NEE at OF to harvesting corre-sponded to hydrological conditions and plant growth before the harvest. The diurnal signal of NEE was notably absent after the 2002 harvest, which coincided with severe drought (Table 2, Fig. 9). It is clear that ecosystem function at OF did not readily recover from the combination of drought and harvest. In 2004, an optimal year for grass growth, the harvest had minimal impacts on daily NEE, and OF was a net C sink immediately after the mow. It is envisioned that tree recruitment is hindered by the rapidly re-growing grass canopy under this scenario, noting also the effects of mowing on seedling recruitment, and this hypothesis can be further tested using a combination of flux and canopy measurements at OF.

0 1 2 3 4 −800 −700 −600 −500 −400 −300 −200 −100 0 100 <LAI>A (m2 m−2) NEE (g C m2 yr1 ) (a) OF PP HW 0 2 4 6 −800 −700 −600 −500 −400 −300 −200 −100 0 100 <LAI>GS (m2 m−2) NEE (g C m2 GS1 ) (b)

Fig. 8 (a) Phase-space plot of annual net ecosystem exchange (NEEA) as surrogate for the change in ecosystem biomass over time (dB/

dt) vs. mean annual LAI (hLAIiA, as surrogate for B) for the old-field (OF), planted pine (PP), and hardwood forest (HW) ecosystems in

the Duke Forest, NC. (b) Same as part (a), but over the April–September peak growing season (GS) time scale.

(15)

Methodological issues

It is important to consider the inherent methodological issues that arise when making long-term ecological measurements in a field setting. The hay removed from OF and used for forage according to local practices (Novick et al., 2004) is likely respired at relatively short time scales off site. The observed NEE should approx-imate the amount of biomass removed (mean ca. 180 g C m2yr1; see Stoy et al., 2006b), if the C balance were to be near zero, the measurements would indicate a source of C of this magnitude, and a likely candidate is soil C loss. However, OF and PP underwent similar management practices over much of their extent before the establishment of the current vegetation, yet have similar stocks of soil C (ca. 4500 g C m2; K. Johnsen, unpublished data). Stoy et al. (2006b) thus attributed the C imbalance to a potential bias in the EC measurement

system, noting that temporal variability in soil C over annual time scales can be difficult to discern given its large spatial variability across many ecosystems. A recent review by Baldocchi (2008) demonstrated a po-sitive offset in the relationship between GEP and RE among FluxNet sites that are managed or have been disturbed, including OF. This analysis is consistent with the idea that there is a consistent loss of soil C from OF within the error and spatial variability of soil C mea-surements such that the EC measurement bias may not exist, lending indirect support to our methodology. We note that the conclusions regarding the experimental hypotheses would not change regardless of whether the EC measurements at OF were biased by ca. 180 g C m2yr1, but the relationship in Fig. 4a must be revised to include an offset.

Another methodological concern is the minor edaphic differences among study sites. The below-ground environment is heterogeneous both within and among ecosystems, and the dynamic flux footprint incorporates this spatial variability. The depth of the rooting zone and clay pan is inherently patchy, but numerous pit and core measurements confirm that the clay pan is consistently of the order of 30–35 cm across ecosystems. Despite the core observations, the water balance measurements at HW suggested that the effec-tive rooting depth may be of the order of 50 cm (Stoy et al., 2006a). Vegetation influences the depth and prop-erties of the soil profile, and it can be expected that more mature ecosystems have greater soil and rooting depth in some cases. This would enhance water availability at HW and may be a factor in the observed resilience NEE to disturbance. However, if drought periods were re-moved from the analysis, support for the experimental hypotheses would still hold, as PP was also sensitive to ice storm damage. NEEAwas still more variable at PP (175 g C m2yr1) than at HW (49 g C m2yr1) when removing 2002 from the analysis, and we see no reason why the ca. 2-year recovery of LAI after the ice storm would have differed had the ice storm not been pre-ceded by drought as the LAI disturbance was large (McCarthy et al., 2007). Interestingly, as mentioned, the years with the highest NEE at both PP and HW were impacted by the mild drought in the late season, in accordance with the chamber observation that Rsoil decreases at low y (Palmroth et al., 2005).

Management implications

Active land management in the SE adds a dynamic component to the successional trajectory represented by the three study ecosystems and provides the opportu-nity to link theoretical and measurement results with land-use practices. Current management trends favor Fig. 9 Cumulative net ecosystem exchange of carbon (NEE) at

the old-field ecosystem (OF) for 2 weeks before and 4 weeks after harvesting for the 2002 and 2004 harvests. The vertical bar denotes the day of harvest and is removed from the data record. Dates of harvesting are listed in Table 2.

Table 2 The day of year on which harvesting occurred in the old-field ecosystem (OF), with mean 10–25 cm soil moisture content and relative hydrologic signature during the time of harvest Year Day of harvest Soil moisture (m3m3 ) Hydrologic signature 2001 179 0.25 Average 2002 153 0.15 Severe drought 2003 206 0.35 Wet 2004 139 0.24 Average 2005 244 0.16 Severe drought r2008 The Authors

(16)

the transition of both OF and HW to PP-type ecosys-tems, which are projected to comprise 30% of SE forested area by 2040 (Wear & Greis, 2002). Water (Pataki & Oren, 2003; Stoy et al., 2006a) and carbon cycling at the ecosystems studied here was similar to other SE ecosystems (Clark et al., 2004; Hanson et al., 2004), as were ecosystem responses to radiation and drought (Wilson et al., 2001; Gholz & Clark, 2002; Hanson et al., 2004). Our analysis suggests that PP-type ecosystems may not significantly increase regional C sequestration if they replace HW-type forests, assuming similar future climatic variability. Actively managed PP-type ecosystems are smaller C pools than mature forests, and may also be smaller C sinks, especially when considering the strong atmospheric C source after clear-cutting (Lai et al., 2002a; Clark et al., 2004), the short rotation length of their management, and their sensitivity to drought and ice storm damage (Oren et al., 1998; McCarthy et al., 2006). The conservation of spe-cies-rich hardwood-type forests may be a sensible strat-egy for maintaining high C sequestration in the SE. These forests are already large pools of C, and C additions to these pools are less affected by climatic extremes, at least within the semidecadal time scales considered here. Future work should consider the ben-efits and consequences of converting OF and HW-type ecosystems to PP-type ecosystems on both the terres-trial carbon and water cycles at regional scales (Juang et al., 2007).

Acknowledgements

This research was supported by the Office of Science (BER), US Department of Energy, Grant No. DE-FG02-95ER82083, by the National Institute of Global Environmental Change (NIGEC) through the Southeast Regional Center at the University of Alaba-ma, Tuscaloosa (DOE cooperative agreement DE-FC030-90ER61010), and by the SERC-NIGEC RCIAP Research Program and the Na-tional Science Foundation (NSF-EAR-06-28432 and 06-35787).

We would like to thank F. Colchero, R. B. Jackson, A. Porpor-ato, and B. Poulter for helpful comments on the manuscript, and Y. Parashkevov, J. Edeburn of the office of the Duke Forest, and Brookhaven National Laboratory, for logistical support. All data are available publicly on the AmeriFlux website at public.ornl.gov/ameriflux, the FluxNet website, and Duke FACE website at http://face.env.duke.edu/main.cfm or by request from the first author.

References

Andrews JA, Harrison KG, Matamala R, Schlesinger WH (1999) Separation of root respiration from total soil respiration using carbon-13 labeling during free-air carbon dioxide enrichment (FACE). Soil Science Society of America Journal, 63, 1429–1435. Baldocchi DD (2008) ‘Breathing’ of the terrestrial biosphere:

lessons learned from a global network of carbon dioxide flux measurement systems. Australian Journal of Botany, 56, 1–26.

Bosatta E, A˚ gren GI (1999) Soil organic matter quality interpreted thermodynamically. Soil Biology and Biochemistry, 31, 1889–1891. Brodribb TJ, Feild TS (2000) Stem hydraulic supply is linked to leaf photosynthetic capacity: evidence from New Caledonian and Tasmanian rainforests. Plant, Cell and Environment, 23, 1381–1388.

Campbell GS, Norman JM (1998) An Introduction to Environmental Biophysics, 2nd edn. Springer, New York.

Carey EV, Sala A, Keane R, Callaway RM (2001) Are old forests underestimated as global carbon sinks? Global Change Biology, 7, 339–344.

Caspersen JP, Pacala SW, Jenkins JC, Hurtt GC, Moorcroft PR, Birdsey RA (2000) Contributions of land-use history to carbon accumulation in US forests. Science, 290, 1148–1151.

Cava D, Giostra U, Siqueira MBS, Katul GG (2004) Organized motion and radiative perturbations in the nocturnal canopy sublayer above an even-aged pine forest. Boundary-Layer Me-teorology, 112, 129–157.

Clark KL, Gholz HL, Castro MS (2004) Carbon dynamics along a chronosequence of slash pine plantations in north Florida. Ecological Applications, 14, 1154–1171.

Cowan IR, Farquhar JD (1977) Stomatal function in relation to leaf metabolism and environment: stomatal function in the regulation of gas exchange. In: Symposium of the Society for Experimental Botany (ed. Jennings DH), Cambridge University Press, Cambridge.

Davidson EA, Janssens IA (2006) Temperature sensitivity of soil carbon decomposition and feedbacks to climate. Nature, 440, 165–173.

Delcourt HR, Harris WF (1980) Carbon budget of the south-eastern U.S. biota: analysis of historical change in trend from source to sink. Science, 210, 321–323.

Detto M, Montaldo N, Albertson JD, Mancini M, Katul GG (2006) Soil moisture and vegetation controls on evapotranspiration in a heterogeneous Mediterranean ecosystem on Sardinia, Italy. Water Resources Research, 42, W08419, doi: 10.1029/ 2005WR004693.

Ekblad A, Bostro¨m B, Holm A, Comstedt D (2005) Forest soil respiration rate and d13C is regulated by recent above ground

weather conditions. Oecologia, 143, 136–142.

Ellsworth DS (1999) CO2enrichment in a maturing pine forest:

are CO2 exchange and water status in the canopy affected?

Plant, Cell and Environment, 22, 461–472.

Ewers BE, Oren R, Sperry JS (2000) Influence of nutrient versus water supply on hydraulic architecture and water balance in Pinus taeda. Plant, Cell and Environment, 23, 1055–1066. Falge E, Baldocchi DD, Olson R et al. (2001) Gap filling strategies

for defensible annual sums f net ecosystem exchange. Agri-cultural and Forest Meteorology, 107, 43–69.

Fan S, Gloor M, Mahlman J, Pacala S, Sarmiento J, Takahashi T, Tans P (1998) A large terrestrial carbon sink in North America implied by atmospheric and oceanic carbon dioxide data and models. Science, 282, 442–446.

Farquhar GD, Lloyd J, Taylor JA et al. (1993) Vegetation effects on the isotope composition of oxygen in atmospheric CO2. Nature,

363, 439–443.

Farquhar GD, von Caemmerer S, Berry JA (1980) A biochemical model of photosynthetic CO2 assimilation in leaves of C3

species. Planta, 149, 78–90.

(17)

Gholz HL, Clark KL (2002) Energy exchange across a chronose-quence of slash pine forests in Florida. Agricultural and Forest Meteorology, 112, 87–102.

Gilmanov TG, Verma SB, Sims PL, Meyers TP, Bradford JA, Burba GG, Suyker AE (2003) Gross primary production and light response parameters of four Southern Plains ecosystems estimated using long-term CO2-flux tower measurements.

Global Biogeochemical Cycles, 17, 1071.

Goulden ML, Daube BC, Fan S-M, Sutton DJ, Bazzaz A, Munger JW, Wofsy SA (1997) Physiological responses of a black spruce forest to weather. Journal of Geophysical Research, 102, 28987–28996. Goulden ML, Munger JW, Fan S-M, Daube BC, Wofsy SC (1996)

Measurements of carbon sequestration by long-term eddy covariance: methods and a critical evaluation of accuracy. Global Change Biology, 2, 169–182.

Granier A, Reichstein M, Bre´da N et al. (2007) Evidence for soil water control on carbon and water dynamics in European forests during the extremely dry year: 2003. Agricultural and Forest Meteorology, 143, 123–145.

Gu LH, Baldocchi D, Verma SB, Black TA, Vesala T, Falge EM, Dowty PR (2002) Advantages of diffuse radiation for terres-trial ecosystem productivity. Journal of Geophysical Research – Atmospheres, 107, doi: 10.1029/2001JD001242.

Gu LH, Baldocchi DD, Wofsy SC, Munger JW, Michalsky JJ, Urbanski SP, Boden TA (2003) Response of a deciduous forest to the Mount Pinatubo eruption: enhanced photosynthesis. Science, 299, 2035–2038.

Gu LH, Fuentes JD, Shugart HH, Staebler RM, Black TA (1999) Responses of net ecosystem exchanges of carbon dioxide to changes in cloudiness: results from two North American deciduous forests. Journal of Geophysical Research – Atmospheres, 104, 31421–31434.

Hacke UG, Sperry JS, Ewers BE, Scha¨fer KVR, Oren R, Ellsworth DS (2000) Influence of soil porosity on water use in Pinus taeda. Oecologia, 124, 495–505.

Hamilton JG, De Lucia EH, George K, Naidu SL, Finzi AC, Schlesinger WH (2002) Forest carbon balance under elevated CO2. Oecologia, 131, 250–260.

Hanson PJ, Amthor JS, Wullschleger SD et al. (2004) Oak forest carbon and water simulations: model intercomparisons and evaluations against independent data. Ecological Monographs, 74, 443–489.

Ho¨gberg P, Nordgren A, Buchmann N et al. (2001) Large-scale forest girdling shows that current photosynthesis drives soil respiration. Nature, 411, 789–792.

Houghton RA (2000) Interannual variability in the global carbon cycle. Journal of Geophysical Research – Atmospheres, 105, 20121– 20130.

Hsieh CI, Katul G, Chi T (2000) An approximate analytical model for footprint estimation of scalar fluxes in thermally stratified atmospheric flows. Advances in Water Resources, 23, 765–772.

Jackson RB, Banner JL, Jobbagy EG, Pockman WT, Wall DH (2002) Ecosystem carbon loss with woody plant invasion of grasslands. Nature, 418, 623–626.

Jackson RB, Sperry JS, Dawson TE (2000) Root water uptake and transport: using physiological processes in global predictions. Trends in Plant Science, 5, 482–488.

Jaksic V, Kiely G, Albertson J, Oren R, Katul G, Leahy P, Byrne K (2006) Net ecosystem exchange of grassland in contrasting wet and dry years. Agricultural and Forest Meteorology, 139, 323–334. Janssens IA, Lankreijer H, Matteucci G et al. (2001) Productivity overshadows temperature in determining soil and ecosystem respiration across European forests. Global Change Biology, 7, 269–278.

Johnston DW, Odum EP (1956) Breeding bird populations in relation to plant succession on the Piedmont of Georgia. Ecology, 37, 50–62.

Juang J-Y, Katul GG, Siqueira MBS et al. (2006) Modeling night-time ecosystem respiration from measured CO2concentration

and air temperature profiles using inverse methods. Journal of Geophysical Research – Atmospheres, 111, D08S05, doi: 10.1029/ 2005JD005976.

Juang J-Y, Porporato A, Stoy PC et al. (2007) Hydrologic and atmospheric controls on convective precipitation events in a southeaster US mosaic landscape. Water Resources Research, 43, W03421, doi: 10.1029/2006WR004954.

Katul GG, Ellsworth DS, Lai CT (2000) Modeling assimilation and intercellular CO2from measured conductance: a synthesis

of approaches. Plant, Cell and Environment, 23, 1313–1328. Katul GG, Lai C-T, Scha¨fer KVR, Vidakovic B, Albertson JD,

Ellsworth DS, Oren R (2001) Multiscale analysis of vegetation surface fluxes: from seconds to years. Advances in Water Resources, 24, 1119–1132.

Katul GG, Leuning R, Oren R (2003) Relationship between plant hydraulic and biochemical properties derived from a steady-state coupled water and carbon transport model. Plant, Cell and Environment, 26, 339–350.

Keeling CD, Chin JFS, Whorf TP (1996) Increased activity of northern vegetation inferred from atmospheric CO2

measure-ments. Nature, 382, 146–149.

Kim CP, Entekhabi D (1998) Feedbacks in the land-surface and mixed-layer energy budgets. Boundary-Layer Meteorology, 88, 1–21.

Kira T, Shidei T (1967) Primary production and turnover of organic matter in different forest ecosystems of the western Pacific. Japanese Journal of Ecology, 17, 70–87.

Kljun N, Black TA, Griffis TJ et al. (2006) Response of net ecosystem productivity of three boreal forest stands to drought. Ecosystems, 9, 1128–1144.

Knohl A, Schulze E-D, Kolle O, Buchmann N (2003) Large carbon uptake by an unmanaged 250-year-old deciduous forest in central Germany. Agricultural and Forest Meteorology, 118, 151–167. Kull O, Jarvis PG (1995) The role of nitrogen in a simple scheme

to scale-up photosynthesis from leaf to canopy. Plant, Cell and Environment, 18, 1174–1182.

Kull O, Kruijt B (1998) Leaf photosynthetic light response: a mechanistic model for scaling photosynthesis to leaves and canopies. Functional Ecology, 12, 767–777.

Kull O, Kruijt B (1999) Acclimation of photosynthesis to light: a mechanistic approach. Functional Ecology, 13, 24–36.

Lai C-T, Katul GG (2000) The dynamic role of root-water uptake in coupling potential to actual transpiration. Advances in Water Resources, 23, 427–439.

Lai C-T, Katul GG, Butnor J et al. (2002a) Modelling the limits on the response of net carbon exchange to fertilization in a south-eastern pine forest. Plant, Cell and Environment, 25, 1095–1119. r2008 The Authors

數據

Table 1 The mean and variability (SD in parentheses) of the annual net ecosystem exchange of CO 2 (NEE) and its  compo-nents – gross ecosystem productivity (GEP) and ecosystem respiration (RE) at old-field (OF), planted pine (PP) and  hard-wood forest (HW)
Fig. 1 Cumulative annual net ecosystem exchange of CO 2 (NEE), gross ecosystem productivity (GEP), and ecosystem respiration (RE) at the adjacent old-field (OF), planted pine (PP), and broadleaf deciduous (HW) forests in the Duke Forest, NC
Fig. 2 Same as Fig. 1a–c, detailing differences in the cumulative sum of net ecosystem exchange of C (NEE) for each year during 2001–
Fig. 5 (a) Relationship between annual transpiration (T A ) estimated using the radiation attenuation approach of Stoy et al
+3

參考文獻

相關文件

• One technique for determining empirical formulas in the laboratory is combustion analysis, commonly used for compounds containing principally carbon and

Students should also be able to appreciate the interrelation between bonding, structures and properties of substances by learning the properties of metals, giant ionic

Due to the important role played by σ -term in the analysis of second-order cone, before developing the sufficient conditions for the existence of local saddle points, we shall

– Change of ownership principle in recording trade in goods sent abroad for processing – The term Gross National Product (GNP) is... Capitalisation of research and development

• developing coherent short-term and long-term school development plan that aligns the school aims, the needs, interests and abilities of students in accordance with the

Microphone and 600 ohm line conduits shall be mechanically and electrically connected to receptacle boxes and electrically grounded to the audio system ground point.. Lines in

Along with this process, a critical component that must be realized in order to assist management in determining knowledge objective and strategies is the assessment of

To solve this problem, this study proposed a novel neural network model, Ecological Succession Neural Network (ESNN), which is inspired by the concept of ecological succession