Relative contributions of soil, foliar, and woody tissue respiration to total ecosystem respiration in four pine forests of different ages


  • Myroslava Khomik,

    1. School of Geography and Earth Sciences, McMaster University, Hamilton, Ontario, Canada
    2. Now at Biogeochemical Model-Data Integration Group, Max Planck Institute for Biogeochemistry, Jena, Germany.
    Search for more papers by this author
  • M. Altaf Arain,

    1. School of Geography and Earth Sciences, McMaster University, Hamilton, Ontario, Canada
    2. McMaster Centre for Climate Change, McMaster University, Hamilton, Ontario, Canada
    Search for more papers by this author
  • Jason J. Brodeur,

    1. School of Geography and Earth Sciences, McMaster University, Hamilton, Ontario, Canada
    Search for more papers by this author
  • Matthias Peichl,

    1. School of Geography and Earth Sciences, McMaster University, Hamilton, Ontario, Canada
    2. Now at Civil and Environmental Engineering Department, University College Cork, Cork, Ireland.
    Search for more papers by this author
  • Natalia Restrepo-Coupé,

    1. School of Geography and Earth Sciences, McMaster University, Hamilton, Ontario, Canada
    2. Now at Department of Ecology and Evolutionary Biology, University of Arizona, Tucson, Arizona, USA.
    Search for more papers by this author
  • Joshua D. McLaren

    1. School of Geography and Earth Sciences, McMaster University, Hamilton, Ontario, Canada
    2. Now at Harvard School of Engineering and Applied Sciences, Cambridge, Massachusetts, USA.
    Search for more papers by this author


[1] Carbon dioxide (CO2) emissions from soil, foliage, and live woody tissue were measured throughout the year in afforested, white pine (Pinus strobus L.) stands (67, 32, 17, and 4 years old as of 2006), growing in a northern temperate climate. The data were used to estimate annual ecosystem respiration (Re) and its component fluxes, including soil, foliar, and woody tissue respiration; to investigate major environmental factors causing intersite and temporal variability in the observed fluxes; and to compare chamber-based Re estimates with eddy covariance-based estimates. While temperature was the dominant driving factor of temporal variability in component fluxes, intersite variability in CO2 emissions was attributed to differences in stand physiological characteristics, such as the presence of the LFH soil horizon, its carbon-to-nitrogen ratio, and the amount of canopy cover. Additional factors that contributed to flux variability included the frequency of precipitation events, vapor pressure deficit and stem diameter, depending on the component considered. Estimated annual chamber-based totals of Re across the four stands were 1526 ± 137, 1278 ± 137, 1985 ± 293, and 773 ± 46 g C m−2 yr−1 for the 67-, 32-, 17-, and 4-year-old stands, respectively. Soil respiration dominated emissions at the 4-year-old stand, while foliar respiration dominated emissions at the 17-year-old stand. In contrast, at the two oldest stands, soil and foliar respirations were comparable. Soil respiration accounted for 44%, 44%, 26%, and 70% of annual Re, across the 67-, 32-, 17-, and 4-year-old stands, while foliar respiration accounted for 48%, 41%, 60%, and 30% of annual Re, across the respective sites. Wood respiration was the smallest component of annual Re across the stands (8%, 15%, 14%, and 0.1%, respectively). The chamber-based Re values were higher than tower-based eddy covariance Re estimates, on average by 18%, 70%, 18%, and 36% at the 67-, 32-, 17-, and 4-year-old stands, respectively. This study contributes to our general understanding of the age-related effects and the role of climate on carbon emissions from various components of afforested ecosystems. Our results suggest that foliar respiration could be comparable to or higher than soil respiration in its contribution to Re in young to mature, planted or afforested, ecosystems. They also suggest that site quality and stand age are important factors to be considered in future studies of carbon dynamics of afforested stands.

1. Introduction

[2] Afforestation, planting forests on abandoned agricultural and marginal lands, has been proposed as a means to help sequester anthropogenic carbon emissions [Nabuurs et al., 2007; Intergovernmental Panel on Climate Change, 2007]. On an annual basis, the net carbon balance of a forest ecosystem is determined by two major fluxes: uptake due to photosynthetic activity and emissions due to respiratory fluxes (Re). Arain and Restrepo-Coupé [2005] have shown that both fluxes are significantly higher for planted or afforested stands when compared to naturally regenerated forests, while others have shown that, of the two fluxes, Re determines the sink/source strength of forest ecosystems [Valentini et al., 2000; Kolari et al., 2004]. Therefore, if afforestation is to be used for atmospheric carbon mitigation purposes, it is necessary to understand the carbon dynamics of afforested stands, including the factors that drive the variability of Re.

[3] In the past 20 years, there have been many studies that assess the carbon dynamics and sink/source potential of forest ecosystems based on carbon dioxide (CO2) flux measurements (see review by Baldocchi [2008]). However, most of these studies focused on naturally regenerated stands or those planted after disturbances such as fire [Hermle et al., 2010; Amiro et al., 2006; Bond-Lamberty et al., 2004] or harvest [Zha et al., 2009; Humphreys et al., 2006; Kolari et al., 2004, Law et al., 2003]. The carbon dynamics of these stands are expected to be quite different compared to those of afforested stands, especially during the initial years after establishment. For example, after a fire, there often remains a significant amount of aboveground biomass and dead root biomass, which could contribute to increased ecosystem respiration due to decomposition in the years that follow [Amiro et al., 2006; Bond-Lamberty et al., 2004, Litvak et al., 2003]. Likewise, there may also be a significant amount of carbon left behind at a site after a forest harvest, in terms of logging residue, the LFH soil horizon, and dead roots, all of which can contribute to increased respiration, due to decomposition in the initial years after harvest [Humphreys et al., 2006; Kolari et al., 2004]. The presence of the LFH horizon and debris could moderate soil temperature and moisture dynamics [Bond-Lamberty et al., 2004], preventing extremes in soil temperatures and moisture conditions that afforested sites may experience before canopy closure. Soils in afforested stands are likely to have depleted organic carbon in the initial years after establishment [Hooker and Compton, 2003; Thuille and Schulze, 2006]. They also may lack extensive belowground biomass accumulation prevalent on former forested or grassland sites. Variable nutrient contents of soils on former agricultural or marginal lands are also expected, which can in turn affect site quality and consequently the carbon dynamics of newly established stands.

[4] The study of Re dynamics is further complicated by the fact that Re is a sum of component fluxes that differ in their response to environmental factors, such as temperature and moisture [Gaumont-Guay et al., 2006], and also to stand physiological characteristics, such as canopy cover, stand age, and nutrient contents [Bolstad et al., 2004; Vose and Ryan, 2002]. On annual basis, the major components of Re include carbon emissions from soils (from roots and microorganisms), foliage, and woody tissue respiration. Several studies have investigated the variability in annual Re composition within various forest ecosystems [Bolstad et al., 2004; Gaumont-Guay et al., 2006; Lavigne et al., 1997; Law et al., 1999; Tang et al., 2008; Vose and Ryan, 2002], but only two studies were of young to mature (0- to 70-year-old) planted forests in temperate climates [Bolstad et al., 2004 and Vose and Ryan, 2002], where afforestation since the 1950s is believed to have led to increased carbon sequestration [Van Minnen et al., 2009]. A young, recently established stand will be expected to have lower accumulated biomass and carbon stocks compared to a mature stand [Hooker and Compton, 2003], which in turn will impact Re composition and dynamics. For example, Tang et al. [2008] found that Re generally increased from young to mature forests and then declined from mature to old-growth stands. Young stands with open canopies have been shown to respire less compared to older stands [Lindroth et al., 2008; Noormets et al., 2007]. However, young stands may be growing more actively and so their growth and maintenance respiration may be higher than those of mature stands. Therefore, in addition to land use history, the age-related variability in forest carbon dynamics should also be considered in studies of afforested stands.

[5] Several techniques can be used to estimate Re in forests, with the most widely used one being the eddy covariance technique [Baldocchi, 2003]. Re can also be estimated as a sum of chamber-based measurements of various respiratory components that have been scaled-up to the ecosystem level [Gaumont-Guay et al., 2006; Lavigne et al., 1997; Law et al., 1999; Tang et al., 2008]. Re estimates using the eddy covariance technique are generally lower than chamber-based estimates, with differences ranging from 2% to 63% [Gaumont-Guay et al., 2006; Lavigne et al., 1997; Law et al., 1999; Tang et al., 2008]. Chamber methods have an advantage over the eddy covariance technique, because of their ability to partition CO2 emissions into various ecosystem components, such as soil, foliage, and woody tissue respiration. This, in turn, can allow researchers to determine the contribution of each component flux to the overall ecosystem respiration and improve our understanding of Re dynamics.

[6] The objectives of this study were (1) to investigate the driving factors of intersite and temporal variability of soil, foliage, and live woody tissue respiration across four afforested stands of different ages; (2) to quantify and compare the contribution of each respiration component to total ecosystem respiration; and (3) to compare total ecosystem respiration derived from scaled-up chamber measurements with that derived from eddy covariance measurements at the sites.

2. Methods

2.1. Study Sites

[7] This study was conducted at the Turkey Point Flux Station (TPFS), located on the north-western shore of Lake Erie, in southern Ontario, Canada. TPFS consists of an age sequence of four white pine (Pinus strobus L.) stands, which were 4, 17, 32, and 67 years old at the time of the study in 2006. The 67- and 32-year-old sites are located beside each other, whereas the 17-year-old and 4-year-old sites are located about 10 km northwest and 18 km west of the two older sites, respectively. The two oldest stands (67 and 32 years old) were planted, or afforested, to stabilize local sandy soils, while the two younger stands (17 and 4 years old) were planted on abandoned agricultural lands that were last cultivated 10 years prior to tree planting. In 1983, thinning was performed at the 67-year-old site, during which 104.76 m3 ha−1 of wood volume was removed from a 38.6 ha area (Ontario Ministry of Natural Resources records). The remaining sites have not been thinned yet. Further stand characteristics are given in Table 1. Hereafter, we refer to the four sites by their shortened code names: TP39, TP74, TP89, and TP02. The acronyms correspond to “Turkey Point”, followed by stand establishment year, i.e. 1939, 1974, 1989, and 2002, respectively.

Table 1. Site Characteristicsa
Site CharacteristicTP39TP74TP89TP02
  • a

    N/A, measurement unavailable.

  • b

    From Chen et al. [2006], measured in August 2005.

  • c

    Estimated from our seasonal measurements with LI-2000, not corrected for clumping.

  • d

    Mean values from Peichl and Arain [2006]; where applicable, data are for trees with diameter at breast height (1.3 m), DBH > 9 cm.

  • e

    Estimated (i.e., sum of stem and branch sapwood volume, assuming branches are 100% sapwood).

  • f

    Measured by Larry Flanagan, University of Lethbridge, AB (unpublished).

  • g

    Includes only white pine needles.

  • h

    Includes needles, leaves, and cones.

  • i

    Nutrients measured in the top 20 cm of the mineral soil (P, phosphorus; Mg, magnesium; K, potassium; Ca, calcium).

Location42°42′55″N 80°22′20″W42°42′34″N 80°21′05″W42°46′32″N 80°28′28″W42°39′49″N 80°34′24″W
Elevation (m)184184212265
Maximum LAI (m2/m2)b8.05.912.8N/A
Mean annual LAI (m2/m2)c4.
Tree height (m)d20.
DBH (cm)d34.615.615.8N/A
Stem density (trees/ha)d429 ± 1661492 ± 3221242 ± 2631683 ± 147
Stem volume (m3/ha)d3761601160.45
Live branch volume (m3/ha)5872101N/A
Total sapwood volume (m3/ha)e1781701760.45
Foliar biomass (kg/ha)d285546018727208
Foliar N (mg/g)f13.911.313.421.4
Foliar CNf38.
Litter-falld,g (kg/ha, Sept-Nov)172518643698N/A
Litter-falld,h (kg/ha, total annual)399029805190N/A
Litter thickness (cm)4.13 ± 1.093.63 ± 0.804.11 ± 1.270
Litter CN ratio17.4 ± 4.824.5 ± 5.616.1 ± 7.1N/A
Mineral soil carbon (top 55 cm)d36.730.133.937.2
Mineral soil available P (ppm)i139 ± 18117 ± 21188 ± 56169 ± 82
Mineral soil Mg (ppm)i10 ± 313 ± 633 ± 3444 ± 5
Mineral soil K (ppm)i12 ± 510 ± 332 ± 1848 ± 18
Mineral soil Ca (ppm)i109 ± 28153 ± 107827 ± 9941669 ± 753

[8] All four stands grow on well-drained sandy soils, classified as Brunisolic Gray Brown Luvisols, following the Canadian Soil Classification Scheme [Presant and Acton, 1984]. The climate in the region is cool temperate. On the basis of a 30-year record, from a World Meteorological Organization-accredited Environment Canada station, located 10 km north of the youngest site at Delhi, Ontario, the mean annual air temperature is 7.8°C, and mean annual precipitation is 1010 mm at TPFS [Environment Canada, 2008]. Normally, the precipitation is distributed evenly throughout the year, with 133 mm falling as snow. At the time of the study, TP39 had a well-developed understory of white pine seedlings (Pinus strobus L.), black cherry (Prunus serotina Ehrh.), white oak (Quercus alba L.), poison ivy (Rhus radicans L. ssp.), bracken ferns (Pteridium aquilinum L.), and blackberry (Rubus allegheniensis Porter). TP74 had minimal understory vegetation, patches of moss cover consisting mostly of Polytrichum spp., and occasional fungi. TP89 had no understory growth, only a layer of pine needles and occasional fungi. The youngest stand (TP02) had no effective litter layer accumulation. Seasonal herbaceous growth (grasses, weeds, etc.) occurred at TP02 from May to October. Further site characteristics are given in the study by Peichl and Arain [2006] and Peichl et al. [2010], but relevant site characteristics are also given in Table 1.

2.2. Eddy Covariance-Based Ecosystem Respiration and Meteorological Measurements

[9] Net ecosystem exchange (NEE) was measured at each site using the eddy covariance (EC) technique. At TP39, a permanent closed-path EC system has been operating since 2002, on top of a 28 m walk-up tower. For details regarding the closed-path EC system set-up, see Arain and Restrepo-Coupé [2005]. A roving open-path EC system was used to measure NEE at the three younger stands. The open-path system was rotated among the three younger sites on biweekly to monthly time intervals, from 2004 to 2006 [Peichl et al., 2010]. Meteorological and flux data were quality controlled following Fluxnet-Canada Research Network protocols. Further details of eddy covariance systems and flux and meteorological data analysis are given in Peichl et al. [2010].

[10] In brief, eddy covariance Re (Re _ec) was calculated using a nonlinear logistic relationship between nighttime NEE, when site-specific friction velocity (u*) was above its minimum threshold value (i.e., 0.325, 0.15, 0.1, and 0.1 m s−1 for the 67-, 32-, 17-, and 4-year-old sites, respectively) and soil temperature at 5 cm depth [Arain and Restrepo-Coupé, 2005]. The Re_ec versus soil temperature (Ts) relationship was calculated for each individual year at TP39, whereas at the three younger stands, NEE data were pooled for 2003–2007 to fit a single relationship because of large gaps in measured EC fluxes at these sites. An analysis of flux data at TP39 showed that, over the 5-year study, the difference between annual Re_ec derived from data pooled over 2003–2007 versus Re_ec derived for individual years was less than 5% [Peichl et al., 2010]. Missing nighttime NEE values (i.e., Re_ec) were filled using this relationship.

[11] Meteorological variables such as radiation, air temperature, humidity, wind speed, and direction, etc., were measured using automatic weather stations at all four sites throughout the year. Additionally, at TP39, precipitation was measured using a heated tipping bucket rain gauge mounted above the canopy, on the flux tower. The four sites experienced similar climate. For example, mean daily above-canopy air temperature (Tair) measurements across all four sites were comparable (for a 1:1 linear relationship, using Tair observations from 2003 to 2006; coefficient of determination, R2 = 0.995, 0.995, 0.993; and slopes = 1.010, 1.004, and 1.005 for TP74, TP89, and TP02, respectively; plots not shown). Therefore, the mean daily and annual climatic variables presented below are those from the TP39 site because this site had the most continuous record of all four sites.

[12] At each site, soil temperature was continuously measured at two locations at 2, 5, 10, 20, 50, and 100 cm depths. Similarly, volumetric soil water content (cm3 cm−3) was measured at the same two locations at 5, 10, 20, 50, and 100 cm depths at the three oldest stands, and at 5, 10, 20, and 50 cm depths at TP02 using water content reflectometers (CS615; Campbell Scientific Inc.). Meteorological and soil data were recorded at half-hour intervals. See further details on meteorological measurements in the study by Peichl et al. [2010].

2.3. Chamber-Based Ecosystem Respiration and Its Component Measurements

2.3.1. Soil Respiration

[13] Soil respiration (Rs) was measured on a monthly basis from 1 January 2006 to 31 December 2006, along 50 m transects, using a portable chamber system, the LI-6400 photosynthesis system (LI-COR Inc., NE, USA) that had a soil chamber and a 15 cm soil temperature probe attachments (models LI-COR 6400-09 and LI-COR 6400-013, respectively; LI-COR, Inc., Lincoln, NE, USA). In 2003, 12 PVC collars (10.16 cm in diameter, 7.5 cm long, inserted into the soil to a depth of about 5 cm) were installed at 4 m intervals along each transect, as part of a long-term study (2003–2006). Once installed, collars remained in the ground for the duration of the study. Herbaceous vegetation inside collars was avoided during initial installation. Any vegetation that grew up inside any collar, since installation, was trimmed back to the soil surface.

[14] At each sampling point, three replicate Rs measurements were recorded. At the same time, soil temperature (Ts_LI_COR) was also measured, within 20–30 cm of each collar, using the LI-COR temperature probe inserted vertically to its full length. The probe was not used during winter, when the top of the soil was frozen. In modeling analysis as described below, missing winter Ts_LI_COR measurements (3% of total) were supplemented with soil temperature measurements from each site′s weather station. Soil temperature from the LI_COR probe (at 15 cm depth) and the weather station′s soil temperature probes (within top 20 cm of soil surface) were comparable within 2% across all four sites (R2 = 0.99–0.98, data not shown).

[15] Measuring Rs over snowpack is a challenge [McDowell et al., 2000]. At TPFS, snow accumulation was sporadic throughout the winter. During our study, when large snow accumulations completely covered our permanent collars along transects, Rs measurements were made directly over snow, in the vicinity of the permanent collars, using a custom-made snow collar. This snow collar was simply one of our PVC collars, as described above, mounted onto a square, framed, metal mesh, which created a “snow-shoe” that prevented the collar from sinking into the snow once the chamber was set on it. The PVC collar was mounted halfway into the mesh. During measurements, the collar was gently inserted into the deep snow few minutes before the Li-6400 chamber was placed over it. This allowed for any flux of CO2 due to snowpack disturbance to vent off, before the chamber was placed over the collar. However, due to the possibility of horizontal advection of CO2 in the snowpack during measurements, because the PVC collar did not go down all the way to the ground when inserted, we did not use these observations in model parameterization. Overall, during 2006, only 1 out of the 5 days of winter measurements had enough snow accumulation to require the use of the snow collar. This measurement was excluded from model analysis below but is still shown in the figures as part of the observations.

2.3.2. Foliar Respiration

[16] Because of logistic constraints, we were unable to measure foliar respiration (Rf) during nighttime. However, we used dark foliar gas exchange measurements collected across TPFS as part of controlled light response curve measurements conducted in a separate photosynthesis study. Thus, Rf is defined here as net CO2 exchange of foliage at zero light level (i.e., photosynthetically active radiation, PAR = 0 μmol m−2 s−1). The LI-6400 instrument was used to generate the light response curves, using the 2 × 3 cm2 foliar chamber attachment (LI-COR Inc., NE, USA). An artificial light source attachment, 6200-02B LED and the 6400-01 CO2 mixer were used to control chamber light and CO2 conditions, respectively. Light response curves were measured under controlled chamber conditions: fixed air temperature, which was within 5°C of ambient temperature; CO2 concentration between 360 and 380 ppm; and changing chamber light conditions (i.e., stepwise reduction of PAR from saturation at 2000 to 0 μmol m−2 s−1). For this study, only measurements corresponding to net gas exchange measurements at PAR = 0 μmol m−2 s−1 were used. In each case the measured foliage was slowly acclimatized to low light (at least half an hour) before being subjected to complete darkness for the Rf measurement. During the low to zero PAR measurements (i.e., 200 to 0 μmol m−2 s−1), the foliage surrounding the chamber was also covered with a nontransparent cloth, to reduce light levels to the surrounding foliage on the measured branch, not just the needles in the chamber.

[17] Ten to fifteen white pine needles (2–3 whorls) were placed in a single flat layer into the chamber, such that the length of the needles inside the chamber was 3 cm. Chamber area was set to 1 cm2 in the instrument program that recorded the measurements. Later the measurements were corrected for the so called true half-surface area (HSA) of the needles in the chamber, which represented an estimate of the surface area of needles exposed to the light source in the chamber. The true HSA was determined using the volume displacement method described by Brand [1987]. Corrected respiration measurements were used for Rf model parameterization.

[18] Two trees were sampled at TP39 by accessing their middle canopy from the eddy covariance walk-up scaffolding tower. At TP74 and TP89, three trees were sampled at midcanopy, using scaffoldings to reach the midcanopy of the trees. At TP02, the trees were small enough to be accessed at midcanopy from the ground. Measurements were conducted on 1-year-old needles. One light response curve was measured at each of the sampled trees at each site, on a monthly basis from June to August in 2006, and Rf was extracted from each curve. Additional light-response curves were conducted at TP39 and TP02 in April, May, September, and November in 2007 to capture seasonal variability in foliar fluxes. Interannual variability in Rf was assumed to be small compared to seasonal and intersite variability of respiration across these different-age stands.

2.3.3. Woody Tissue Respiration

[19] We used the approach of Xu et al. [2000] to measure woody tissue respiration (Rw) at the three older stands, using the LI-6400 system with its soil chamber attachment. In autumn of 2005, four trees of variable diameter were selected around the EC towers at each stand. Collars (same make and diameter as those used for soil respiration) were attached vertically at 1.3 m height on the stem of each tree with silicone, following Xu et al. [2000]. Loose bark was removed around the circumference of the collar, as necessary. Woody tissue respiration was sampled on a monthly basis, from April to November 2006. At the end of the measurement campaign, increment cores were taken from the center of each collar to determine the sapwood volume under the collars as follows: the area enclosed by the collar (i.e., πr2, where r was the radius of the collar on the stem) was multiplied by the width of sapwood below the collar, which was measured from the tree core taken from the center of the collar. Measured respiration values were corrected from per surface area to per sapwood volume before being used in model parameterization.

[20] At the two oldest stands, tree bole temperature (Tb) was measured in several trees using thermocouples that were continuously sampled, every half hour, by the weather station data logger. These thermocouples were inserted into sapwood at 2–5 cm from the surface. Missing Tb values and those for TP89 were estimated from air temperature (Ta), separately for each site, using linear regressions developed between Ta and Tb at TP39 and TP74. For model parameterization, Tb values corresponding to the time of day when Rw was measured were used.

[21] At TP02, because of the small size of the tree stems, we were unable to implement our Rw chambers to measure Rw directly. However, we estimated Rw for TP02 by assuming that the rate of woody tissue respiration would be similar to that of the second youngest stand (TP89). Thus, we used simulated Rw values from TP89, upscaled them with an estimate of sapwood volume per square meter at TP02. Past biomass studies at TP02 showed that the wood in the seedlings was all sapwood because of the young age and tree size of the stand at the time of this study [Peichl and Arain, 2007].

2.4. Data Analysis

[22] We used the Gamma model [Khomik et al., 2009] to simulate component fluxes at our sites:

equation image

where Ri is respiration of component i in μmol of CO2 m−2 s−1; Ti is the temperature of component i in °C, but shifted by 40°C (i.e., Ti = measured temperature + 40°C, see Khomik et al. [2009] for more details); and α, β0, and β1 are the model parameter coefficients to be estimated. The model was linearized before being parameterized, by taking the natural logarithm of equation 1.

[23] To investigate intersite differences in fluxes, the model was parameterized, individually, for each component using all observations, with the exception of one snow-covered day described above. Intersite differences were initially tested using the categorical variable (i.e., “dummy variables”) approach [McClave and Sincich, 2003, p. 630]. This allowed the α, β0, and β1 (equation 1) to vary among sites of different ages, as follows:

equation image
equation image
equation image

where β01, β02, β03, β04, β11, β12, β13, β14, α1, α2, α3, and α4 are unknown coefficients to be estimated. In equations 2b and 2c, the terms involving the product of a dummy variable and Ti or LnTi are called “interaction terms. ” By considering the product of a dummy variable and Ti as a new explanatory variable, equation 1 is transformed into a multivariate linear regression model [Otomo and Liaw, 2003]. In this model, all of the unknown coefficients can be simultaneously estimated by using a linear regression procedure in any widely available statistical software, such as SAS (SAS Institute Inc., NC, USA) or SPSS (SPSS Inc., IL, USA).

[24] To establish which environmental factors other than temperature controlled variability in Re components, we tested the following additional explanatory factors in our models: the thickness of the LFH soil horizon (LFH) and its carbon-to-nitrogen ratio (CN), soil volumetric water content, air temperature, precipitation amount and frequency, photosynthetically active radiation, stem diameter, and vapor pressure deficit. These additional variables, depending on the respiration component considered, were added into the linearized form of the Gamma model (equation 1) to create a multivariable linear regression model:

equation image

where Ri, Ti, α, β0, and β1 are as in equation 1 above, X2Xn are additional explanatory variables and β2βn are their corresponding model parameter coefficients to be estimated.

[25] The expanded models were evaluated, to determine which of the added variables were statistically significant in improving the model′s explanatory power (i.e., P < 0.05 of their estimated coefficient). Only variables that were statistically significant were retained in the final models (Table 2).

Table 2. Models Used to Simulate Respiration of Individual Component Fluxes, Including the Models′ Coefficient of Determination and the Number of Observations Used for Model Parameterizationa
ComponentModelT-only R2R2n
  • a

    Here, Rs is soil respiration (μmol CO2 m−2 s−1); Rw is woody tissue respiration of sapwood (μmol CO2 m−3s−1); Rf is foliar respiration (μmol CO2 m−2 s−1; half-surface area, HSA); A3, A4 are categorical variables used to represent intersite variability (i.e., A3 = 1 for all observations belonging to TP89 and zero otherwise, whereas A4 = 1 for all observations belonging to TP02 and zero otherwise); PPTf and PPTf_1 are categorical variables that represent the frequency of precipitation occurrence on the day of Rf measurement and 1 day before Rf measurement, respectively; LFH is site mean thickness of the LFH horizon (cm); CN is site mean carbon-to-nitrogen (CN) ratio of the LFH; Ta is air temperature (above canopy, °C); Tb is tree bole temperature (2–5 cm, °C); Ts is soil temperature (2–20 cm, °C); VPD is vapor pressure deficit; and α, βx and βxx are model parameters to be estimated (see Table 3 for values).

RsRs = image0.850.881967
RwRw = image0.660.82336
RfRf = image0.440.61719

[26] During analysis, we observed that some of the environmental explanatory variables were able to replace the dummy variables representing intersite variability. This occurred if the two variables mutually excluded each other when both were included in the model simultaneously. Such instances provided insight into which physiological factors influenced intersite variability in respiration.

[27] Statistical analysis and model parameterization was performed using the SAS 9.1 software. The unknown parameter coefficients were estimated using the linear regression procedure, PROC REG, in the SAS software.

[28] The best specification of the model for each component respiration was used to simulate the component CO2 emissions on a daily timescale, using daily means and totals of meteorological variables as inputs. These values were then upscaled to the stand level, using biometric variables, and used to compute monthly and annual CO2 emissions of Re components. In turn, Re values were simulated as the sum of daily Rf, Rs, and Rw values:

equation image

Monthly and annual Re values were computed from the sum of the daily Re values.

[29] Uncertainty in simulated component respiration fluxes were estimated as the ratio between ±2 standard deviations (σr) about the predicted value and the total annual predicted flux (i.e., (2σr)/Ri, where Ri was the annual predicted sum of the component i). The σr values were computed as σr = 2equation image, where n is the sample size (i.e., 365 days of the year) and σr2 is the error mean square from the model output. Both values used to compute the ratio were in their original simulated units prior to upscaling. This ratio was then applied to upscaled sums of the individual foliar and woody tissue fluxes to obtain an estimate of the error on those values. This was done since for Rf and Rw fluxes, the error computed from simulated values was not directly transferable to upscaled values as in the case of Rs measurements that were already in units of square meter per ground area. We also report uncertainty in Re, which we calculated arithmetically from Rs, Rw, and Rf uncertainties (i.e., as square root of the sum of squared uncertainties of the individual Re components, for a given time period).

2.5. Up-Scaling to Ecosystem Level

[30] Because simulated Rw were in units of per sapwood volume and simulated Rf in units of per half-needle-surface area (HSA), we upscaled them to per ground surface area of the stand, using biometric indices from the individual stands in order to compare them with associated Rs values and to calculate Re values (g m−2).

[31] Foliar respiration was upscaled using seasonal leaf area indices (LAIs) for each site. Seasonal LAI for each site was determined using a combination of LAI measurements conducted by Chen et al. [2006] and our own seasonal measurements. Chen et al. [2006] reported LAI values for the three oldest stands that they measured once, in August 2005, using two different techniques (Li-2000 (Li-COR Inc. Lincoln, NE, USA) and TRAC (ThirdWave Engineering, Ottawa, Canada)). However, Vose and Swank [1990] reported that LAI in Pinus strobus L. forests varies considerably during the year, with peak LAI reported in late July for their site. Therefore, the LAI values presented in the study by Chen et al. [2006] were likely the maximum annual values for our sites. In order to determine the seasonal variation in LAI values, we used our own LAI measurements made during different seasons throughout the year at the sites, using the Li-2000. We determined the percent-relative contribution of these seasonal LAI values (i.e., spring, summer, and autumn of 2002) to that of the maximum LAI we measured at the end of the summer 2002. We then used these relative percent ratios to determine seasonal LAI values from the single measurement reported by Chen et al. [2006] for each site. We assumed Chen et al.'s measurements to be the more accurate estimates of the maximum seasonal LAI, compared to our measurements, because Chen et al. [2006] corrected their measurements for branch and needle clumping. No measurements of LAI for TP02 were available from Chen et al. [2006]; therefore for that site, we report our own estimates from the Li-2000 measurements (Table 1). We also assumed little interannual variability in LAI from years 2002 to 2006 in our estimates. These estimated seasonal LAI values were used to upscale modeled Rf (from g m−2 HSA per day) to per ground area (i.e., g CO2 m−2 ground area) for each site by multiplying simulated daily Rf values by the corresponding seasonal LAI value (i.e., we estimated one mean LAI value for each month of the year).

[32] Rf was measured on 1-year-old foliage. We assumed that respiration from 1-year-old foliage was a good approximation of the overall mean canopy respiration at our sites. In white pines, most foliage is shed at the end of its second growing season, but a small fraction of foliage survives for up to 4 years [Vose and Swank, 1990]. At our sites, we mostly observed 0- to 2-year-old needles. Our Rf measurements were conducted at midcanopy. Differences in Rf are expected along the vertical profile of the tree canopy because of the variable gas exchange dynamics of sunlit and shaded foliage [Givnish, 1988]. Because of logistical constraints, we were able to sample only at midcanopy. Our upscaled measurements could over- or underestimate total foliar gas exchange. However, we would like to point out that older white pine stands tend to have most of their needles on branches located several meters above the ground, at the top of the stem. Therefore, we assumed midcanopy measurements to be good representatives of overall canopy Rf.

2.6. Upscaling Rw to Stand Level

[33] Woody tissue respiration (simulated on per sapwood volume basis) was upscaled using mean stem sapwood volume per ground area of each stand. Sapwood volume was determined from a separate destructive sampling study at our sites [Peichl and Arain, 2007]. Branch sapwood volume was estimated by assuming branches were 100% sapwood and using branch volume per stand as determined by Peichl and Arain [2007]. Similarly, at TP02, the seedling trees were assumed to be 100% sapwood. One would expect Rw values, which have been upscaled to the stand level, to vary due to differences in stem density, while Rw values on per-sapwood-volume basis may also be affected by site quality, stand age, and thinning practices. Therefore, in our comparison of woody tissue respiration among TPFS sites, Rw rates on per-sapwood-volume basis were used.

3. Results

3.1. Meteorology and Site Microclimate During the Study Year

[34] The study year 2006 was relatively warm and wet, compared to the 30-year norm for the area (Figures 1a and 1d). Mean annual temperature was 1.9°C above the norm (about 25% higher), and total annual precipitation was 177 mm above the norm (about 18% higher). Precipitation during normal years is usually evenly distributed throughout the year in the region. However, in 2006, higher precipitation occurred in winter (February) and later around September (Figure 1d). The seasonal course of soil temperature (Ts) followed that of the mean daily air temperature with some lag (Figures 1a and 1b). Differences in soil temperature between sites were more pronounced than differences in soil moisture, with the youngest stand reaching some of the highest observed soil temperatures across all four sites (Figure 1b). Soil moisture (θs) at all four sites was relatively low, averaging about 0.12 cm3 cm−3 per year, despite increased precipitation during the study year (Figure 1c). This was due to the sandy, well-drained nature of the soils at the sites. The annual ranges in θs were lower (0.14, 0.14, 0.12, and 0.10 cm3 cm−3 for TP39, TP74, TP89, and TP02, respectively) than the annual ranges in Ts (24.1°C, 26.0°C, 21.2°C, and 30.7°C, respectively).

Figure 1.

Comparison of climatic and edaphic conditions across Turkey Point sites in 2006. (a) Daily mean air temperature, Tair, (b) daily mean soil temperature, Ts (mean of all sensors in top 20 cm of mineral soil), (c) daily mean soil moisture content, VMC (mean of all sensors in top 20 cm of mineral soil), and (d) daily total precipitation, Ppt. Also listed in top right corner of each plot are the annual mean or total values, as well as the annual minimum and maximum values.

3.2. Environmental and Physiological Controls on Respiration

3.2.1. Effect of Climate

[35] Multivariate analysis of observed data showed that temporal variability in temperature (i.e., soil temperature for Rs, tree bole temperature for Rw, and air temperature for Rf) was the dominant driving factor of temporal variability of the respective component fluxes and consequently Re (Table 3). Temperature-only component models had R2 vlaues of 0.85, 0.44, and 0.66 for Rs, Rf, and Rw, respectively.

Table 3. Estimated Coefficients of the Model Parameters in the Best Specification of the Gamma Models for Each Component Fluxa
Estimated Coefficientt ValueP ValueEstimated Coefficientt ValueP ValueEstimated Coefficientt ValueP Value
  • a

    Also shown are the associated statistics: t and P values. See Table 2 for model equations.

β03   2.684.8<.00011.2213.5<.0001
β12   −0.03−4.1<.0001   

[36] However, several additional factors were found to be statistically significant (P < 0.05) in explaining Rs variability: the thickness of the LHF horizon (LFH) and its carbon-to-nitrogen (CN) ratio, which accounted for intersite variability as discussed below; mean daily air temperature (Ta); and precipitation frequency (represented by categorical variables, PPTf and PPTf_1) (Table 2). Including these additional variables in the Rs model helped to improve the model′s R2 (from 0.85 to 0.88). Daily soil moisture (i.e., daily mean of measurements in the top 20 cm of the mineral soil) was also tested in the Rs model but was found to be insignificant (P > 0.05). However, moisture in the LFH horizon, which was not measured directly at TPFS, may have been more important to Rs variability across our sites, compared to the mineral soil moisture content. The LFH and PPTf variables were both found to be statistically significant in the model, with PPTf becoming more significant once LFH was included in the model (results not shown).

[37] The Rf model was first parameterized with observed Rf values and the estimated foliar temperatures, measured within the chamber. The model produced R2 of 0.44 (adjusted R2 = 0.44). However, because we did not have continuous measurements of foliar temperature at the sites to simulate daily Rf, we assumed that Ta was a good surrogate temperature for Rf. Therefore, we also fitted the model to observed Rf values and the corresponding mean daily Ta at each site and found that the fit with Ta was as good as with foliar temperatures (R2 remained 0.44). Additional variables that helped to explain Rf variability in the model included stand age, mean daily vapor pressure deficit (VPD), mean daily downwelling photosynthetically active radiation, and the frequency of precipitation 1 day before Rf measurements (Table 2). Including those variables in the model improved its explanatory power by 17% (R2 = 0.61). Of the additional variables, VPD was found to be negatively related to Rf.

[38] The additional variables that improved the Rw model′s explanatory power included stand age categorical variables, the frequency of precipitation on the day of Rw measurements, and tree diameter at breast height (Table 2). Including those additional variables in our final best specification of the Rw model improved the model fit to R2 = 0.82.

3.2.2. Effect of Stand Age and Resulting Physiology

[39] Intersite differences in the Rs-Ts model, and thus, Rs variability were observed between all four sites. In the initial analysis of observed data we used dummy variables to represent site differences. The estimated coefficients of all three of the site dummy variables were statistically significant (i.e., P < 0.05), and the magnitudes of their estimated coefficients were negative (i.e., -0.18, -0.10, and -0.34 for TP74, TP89, and TP02, respectively), suggesting that Rs at the three younger stands was lower compared to TP39. However, in subsequent analysis, we found that two of the site variables could be replaced with site physiological variables. For example, when the thickness of the soil LFH horizon (LFH) was added into the model (i.e., equation 1 modified with dummy variables, as outlined in equations 2a2c), the estimated coefficients of the dummy variables representing TP02 became insignificant, and the model did not converge. Similarly, the variable representing the carbon-to-nitrogen ratio of the LFH horizon and the dummy variable representing TP74 were mutually exclusive, if included together in the model, requiring one of the variables to be dropped. Therefore, we retained only LFH and CN in the final model and dropped the categorical variables for TP74 and TP02 (Tables 2). The model′s estimated R2 remained unchanged when the dummy variables were replaced with the more physically meaningful variables.

[40] We also evaluated intersite differences in Rf and found that, on per-leaf-area basis, there was no statistical difference in Rf between the two oldest stands (TP39 and TP74) because the dummy variable representing TP74 was found to be statistically insignificant (P > 0.05) in the model. In contrast, the dummy variables representing TP89 and TP02 were statistically significant, and their estimated coefficients were higher than the reference site (TP39). This suggested that Rf at the youngest two stands, on per-leaf-area basis, was statistically different and higher from Rf of the older stand (Table 3). However, unlike for the Rs model, we did not find environmental or physiological variables to replace the dummy variables in the Rf model.

[41] Intersite differences in Rw were also important to consider because the dummy variables representing TP74 and TP89 were found to be statistically significant (Table 3). The estimated coefficients of the two variables were positive and implied that Rw values, on per-sapwood-volume basis, at TP74 and TP89 were higher than those at TP39, with the highest rates at TP89. DBH and precipitation frequency also had a strong contribution to the model's R2. Similar to the Rf model, we were unable to find physiological or environmental factors that could statistically explain the intersite variability in Rw. Therefore, the dummy variables were retained in the final Rw model.

3.3. Contribution of Individual Component Fluxes to Ecosystem Respiration, Re

[42] Observed mean annual Rs values were 2.2, 1.9, 1.8, and 1.9 μmol CO2 m−2 s−1 for the TP39, TP74, TP89, and TP02 year-old stands, respectively, with ranges of 0.3–5.9, 0.3–4.5, 0.3–4.7, and 0.3–4.2 μmol CO2 m−2 s−1, respectively (Figures 2b–2e). Similarly, simulated mean annual Rs values were 1.8, 1.5, 1.4, and 1.5 g C m−2 d−1 for the TP39, TP74, TP89, and TP02 sites, respectively (Figure 2a). While the magnitude of simulated Rs was generally lower compared to observations, the temporal trend was well represented. Mean daily soil respiration was lowest during winter months and peaked in late July to early August (Figure 2a). Estimated annual Rs values were 671 ± 33, 558 ± 35, 511 ± 36, and 539 ± 32 g C m−2 yr−1 for the TP39, TP74, TP89, and TP02 year-old stands, respectively. The highest annual Rs value was observed at the oldest stand, TP39, and was different from the annual Rs at the three youngest stands (Figure 5). In contrast, annual Rs values among the three youngest stands of various ages were comparable (Figure 5).

Figure 2.

Comparison of (a) simulated daily mean soil respiration Rs in g C m−2, across Turkey Point sites in 2006, and (b–e) relationships between observed Rs (μmol CO2 m−2 s−1) versus soil temperature (Ts) across all four sites. Symbols represent measured values, whereas lines represent simulations using the Ts -only Gamma model. In the top right corner of Figure 2a, annual mean, minimum, and maximum simulated Rs values are listed as well as total annual emissions with their estimated errors in g C m−2.

[43] Observed annual mean foliar respiration, Rf values were 1.6, 2.1, 2.5, and 2.9 μmol CO2 (half-surface area of needles) m−2 s−1 at TP39, TP74, TP89, and TP02, respectively, with the corresponding Rf ranges of 0.2–3.0, 0.9–3.3, 1.0–4.3, and 0.7–5.8 μmol CO2 (half-surface area of needles) m−2 s−1, respectively (Figures 3b–3e). We note that measurements at TP74 and TP89 were made only for 3 months of the year (June through August, n = 9), whereas measurements at TP39 and TP02 spanned over spring and autumn months as well (i.e., April through November). Therefore, the observed temperature range is smaller at TP74 and TP89 in Figures 3c and 3d. Observed Rf rates decreased with increasing stand age, with the highest Rf rates observed at TP02. However, the upscaled annual Rf totals did not follow this age pattern and were 726 ± 182, 527 ± 132, 1203 ± 290, and 234 ± 33 g C m−2 yr−1 for the TP39, TP74, TP89, and TP02 year-old stands, respectively (Figures 3 and 5).

Figure 3.

Comparison of (a) simulated daily mean foliar respiration (Rf) in g C m−2, across Turkey Point sites during 2006, and (b–e) relationships between observed Rf (μmol CO2 (half-needle surface area in m−2) s−1) versus air temperature (Ta) across all four sites. Symbols represent measured values, whereas lines represent simulations using the Ts-only Gamma model. In the top right corner of Figure 3a, annual mean, minimum, and maximum simulated Rf values are listed as well as total annual emissions with their estimated errors in g C m−2.

[44] Observed annual mean Rw values were 38.1, 55.4, and 81.1 μmol CO2 (sapwood volume) m−3 s−1 at the TP39, TP74, and TP89 year-old stands, respectively, with the respective Rw ranges of 3.4–151.3, 4.8–162.6, and 5.4–187.5 μmol CO2 (sapwood volume) m−3 s−1 (Figures 4b–4d). The course of temporal variability in Rw followed closely that of air temperature variability, given that the bole temperatures were highly correlated with Ta (results not shown). Annual total values of Rw were 129 ± 13, 193 ± 14, 271 ± 14, and 0.7 ± 0.04 g C m−2 yr−1 for TP39, TP74, TP89, and TP02, respectively (Figures 3 and 5). Unlike for Rf and Rs, Rw values, both, on per-sapwood-volume and ground-area basis, followed a distinct linear age-related pattern across the three oldest TPFS stands, i.e., Rw decreased with increasing stand age (Figures 4 and 5).

Figure 4.

Comparison of (a) simulated daily mean woody tissue respiration (Rw; which included both branch and stem respiration) in g C m−2, across Turkey Point sites during 2006, and (b-d) the relationships between observed Rw (μmol CO2 (sapwood volume of stem in m−3 ) s−1) versus tree bole temperature (Tb) across all four sites. Symbols represent measured values, whereas lines represent simulations using the Ts-only Gamma model. In the top right corner of Figure 4a, annual mean, minimum, and maximum simulated Rw values are listed, as well as total annual emissions with their estimated errors in g C m−2.

Figure 5.

Intersite comparison of annual totals of the three major ecosystem respiration components, i.e., total soil (Rs), woody tissue (Rw), and foliar (Rf) respiration, in g C m−2 yr−1. Also included are estimated annual totals of ecosystem respiration (Re) in the legend. Estimated errors on each total are shown as ± error bars and numerically.

[45] Based on upscaled chamber measurements, annual total Re values were estimated to be 1526 ± 137, 1278 ± 137, 1985 ± 293, and 773 ± 46 g C m−2 yr−1 at TP39, TP74, TP89, and TP02, respectively (Figure 5). Annual totals at the two oldest stands, TP39 and TP74, were comparable, that of the TP89 stand was the highest of all four stands, while that of the youngest TP02 stand was the lowest of all. At the TP02, Rs accounted for 70% of Re, with Rf accounting for the remaining 30%. Rw was minimal (0.1%) at that site. This was in contrast to the TP89 stand, where Rf accounted for the majority of the annual Re (60%), with Rs accounting for an additional 26% and Rw for 14%. At the two oldest stands, TP39 and TP74, the contributions of Rs to annual Re were comparable (i.e., 44% each), whereas the contribution of Rf to Re at TP39 was higher than that at TP74 (i.e., 48% versus 41%, respectively). In contrast, Rw contribution at TP74 was higher compared to TP39 (i.e., 15% versus 8%, respectively). The relative percent contribution of the individual Re components to total annual Re was variable across all four stands throughout the year (Figure 6). During the growing season, Rf dominated Re, whereas in winter months, Rs dominated Re. Even at TP02, in August, Rs and Rf were comparable in their contribution to Re (Figure 6), highlighting the importance of foliar respiration in the carbon cycle of young to mature afforested stands.

Figure 6.

Comparison of percentage-relative contribution of monthly soil respiration (Rs), woody tissue respiration (Rw), and foliage respiration (Rf) to total monthly ecosystem respiration at (a) TP39, (b) TP74, (c) TP89, and (d) TP02 site.

3.4. Comparison of Re Derived by Chamber Versus Eddy Covariance Methods

[46] On a daily scale, chamber-based estimates of Re overestimated the eddy covariance, EC-based Re estimates (Re_ec) from April to November across all four stands (Figure 7), although Re from both methods was highly correlated (i.e., for 1:1 relationships, R2 = 0.94–0.96, plots not shown). The high correlation could be in part due to temperature being used as the main predictor of simulated fluxes in both methods. At TP02, upscaled chamber estimates of Re better matched observed nighttime NEE values in July (Figure 7d). Annual total Re_ec values were 1293, 751, 1678 and 569 g C m−2 yr−1 at the TP39, TP74, TP89 and TP02, respectively. At all four stands, Re values were higher than Re_ec by 18% at TP39 and TP89, by 70% at TP74, and 36%, at TP02.

Figure 7.

Comparison of simulated mean daily ecosystem respiration estimates using the models developed by the chamber-based (Re) and by the eddy covariance (Re_ec) measurements. Observed values of daily mean nighttime net ecosystem exchanges (NEEs) are also shown. (a) TP39, (b) TP74, (c) TP89, and (d) TP02. In the top right corner of each frame, annual totals estimated from each method in g C m−2 yr−1 and their ratios are also given.

4. Discussion

4.1. Meteorology and Site Microclimate

[47] The study year was relatively warmer and wetter compared to the norm for the area, and therefore, it is possible that some of the discrepancy between estimated Re values at our sites and those reported in relevant literature could be due to this more favorable climate for growth at our sites. The estimated component fluxes may have been higher this year because variability in both air temperature and precipitation was included in model simulations of individual component fluxes (Table 2). Despite that, our results support the idea that in young afforested stands, the lack of forest floor accumulation and canopy cover lead to a stronger coupling between the soil and the atmosphere. For example, at TP02, the site without LFH accumulation and an open canopy, soil temperatures reached the highest values. At TP02, Ts was more strongly coupled to Ta compared to the older three stands. In contrast, TP89 had the highest LAI and relatively thick litter layer accumulation, which resulted in the lowest Ts among sites and highest time lag between Ts and Ta. Therefore, the differences in the sites′ microclimates that resulted from age-related differences in stand physiology contributed to the variability in observed respiration fluxes.

4.2. Environmental and Physiological Controls on Respiration

[48] Because of the relatively open canopy and lack of litter layer, Rs increased about 1 month earlier in spring at TP02 as compared to three older stands (Figure 2, March–April). A similar phenomenon was reported previously by Noormets et al. [2007] but for Re. In our study, seasonal dynamics in Re were generally dominated by Rf across our stands.

[49] The highest annual Rf was obtained at TP89, the site with the highest LAI, whereas the lowest annual Rf was obtained at TP02, the site with the lowest LAI (Table 1). Estimated annual Rf values, on a per ground area basis, for all but the youngest Turkey Point stand (TP02), were much higher compared to those reported in similar studies in the literature. For example, Tang et al. [2008] reported Rf values of 69–121 g C m−2 yr−1, whereas Law et al. [1999] reported 157 g C m−2 yr−1 from their upscaling studies in old-growth forests. Similarly, Gaumont-Guay et al. [2006] reported total estimated Rf in the range of 173 ± 14 to 243 ± 21 g C m−2 yr−1 in their study of component fluxes of an 81-year-old boreal aspen forest. However, our measured Rf rates, on per leaf area basis, were close to the range and seasonality reported by Cooper et al. [2006] in a mixed conifer forest in Washington, USA (0–4.6 μmol CO2 m−2 s−1), which attained maximum values in June and minimum in December. Therefore, intersite variability of upscaled Rf values among our stands, as well as between our stands and those reported in literature, was driven by intersite differences in leaf area indices.

[50] At TP02, very low Rw was due to the low sapwood volume per hectare in this young seedling site (Table 1). Otherwise, across our sites the range of observed live woody tissue respiration was larger compared to Rw values reported in the literature but comparable to literature values once upscaled on per ground area basis. For example, Tang et al. [2008] reported Rw values of 4–40 μmol CO2 (sapwood volume) m−3 s−1 in the mixed-wood forest in Michigan, USA, but 130–209 g C per m2 yr−1. Similarly, Law et al. [1999] reported Rw of 2.5–19.5 μmol CO2 (sapwood volume) m−3 s−1 in their ponderosa pine forest in Oregon, USA, and 54 g C per m2 yr−1. Griffis et al. [2004] reported stem respiration of 155–198 g C per m2 yr−1 in their 74-year-old boreal aspen forest. The differences between Turkey Point stands and literature values of Rw could be due to physiological differences, such as differences in sapwood volume per ground area between our sites and those in literature. High sapwood production, and consequently high Rw, at our sites may be partly caused by greater water availability in deeper soil layer (∼1 m depth) at TP89. Peichl et al. [2010] discussed the unlimited/continuous soil water access for trees at TP89, which allows those trees to have exceptional amounts of productivity (both foliar, i.e., LAI, and sapwood), compared to the other three TPFS stands. Indeed, foliar biomass across our sites was highest for TP89, suggesting high productivity (Table 1) and estimated sapwood growth at DBH was also highest at TP89 in 2006 (i.e., 9.9 cm2 at TP89, versus 6.2 cm2 at TP39 and 3.3 cm2 at TP74, unpublished data).

[51] Other reasons for the intersite differences in CO2 emissions among our stands, as well as among our sites and those reported in literature, could be differences in site quality. For example, intersite variability in observed Rf values (on per HSA of needles) among our stands was likely due to intersite variability in foliar nitrogen content, with the highest foliar N and Rf values observed at TP02 (Figure 3 and Table 1). Foliar gas exchange has been shown to relate strongly and positively to foliar nitrogen content [Dang et al., 1997; Vose and Ryan, 2002]. The amount of soil nutrients such as phosphorus, magnesium, and potassium in the surface soils at TP89 support the idea that site quality at TP89 may have been better suited for tree growth compared to TP39 or TP74. Peichl et al. [2010] discussed in more detail the effects of site quality on intersite differences in ecosystem respiration across TPFS. Finally, the generally higher and more spatially variable amounts of soil nutrients at the younger stands (Table 1) are reflective of the sites′ past agricultural land use, suggesting that past land use history is an important factor to consider when studying carbon dynamics of young afforested stands.

4.3. Contribution of Individual Component Fluxes to Ecosystem Respiration, Re

[52] Annual Re values across our sites (Figure 7) were higher than literature-reported values, especially for TP89. For example, Law et al. [1999] reported Re of 894 g C per m2 yr−1 from upscaled chamber measurements in their old-growth ponderosa pine forest in Oregon, whereas Tang et al. [2008] reported a range of 600–742 g C per m2 yr−1 in their old-growth mixedwood site in Michigan, USA. For comparison, Gaumont-Guay et al. [2006] reported upscaled chamber-based Re of 1190 g C per m2 yr−1 for an 81-year-old aspen stand in Saskatchewan, Canada. Differences in LAI and stand age are the likely reasons for the discrepancy.

[53] In a recent study, Lindroth et al. [2008] have shown that intersite differences in Re across a number of coniferous forests in northern Europe were driven first by differences in LAI and second by differences in stand age. The generally high LAI at our Turkey Point sites was attributed to high values of white pine needle clumping [Chen et al., 2006], which could be species specific. Maximum LAI values across our three older stands varied from 5.9 to 12.8, with an estimated annual mean of 3.4–7.4 (Table 1). In contrast, LAI of the stands studied by Tang et al. [2008] averaged 3.8–4.1, whereas that of the ponderosa pine stand studied by Law et al. [1999] was only 1.5. The highest LAI at TP89 may have been due to more favorable soil water availability and nutrient conditions compared to the two older stands and/or because of the fact that TP89 was in an active stage of growth for white pine species.

[54] Noormets et al. [2007] studied the effect of stand age on total ecosystem carbon fluxes in managed forests (3 to 65 years old) in the Great Lakes region and reported higher Re in younger stands compared to older ones, explaining that the difference was due in part to the inherent greater biological activity of younger stands. Results presented in this study suggest that on annual basis, foliar respiration dominated ecosystem respiration at our older (17- to 67-year-old) stands, which was in contrast to the more widely reported dominance of Rs on Re in forested ecosystems [Bolstad et al., 2004; Law et al., 1999; Gaumont-Guay et al., 2006; Tang et al., 2008; Zha et al., 2009]. The unusually high Rf in our study, especially for the 17-year-old stand, may be due to the young age of the stand and its inherent high productivity. Recent studies have suggested a strong positive link between CO2 emissions in forests and their primary productivity, whereby increased productivity leads to increased respiration [Hibbard et al., 2005; Litton et al., 2007]. Furthermore, there is increasing evidence that across different forest biomes, productivity tends to increase with forest age up to an age of about 120 years [Pregitzer and Euskirchen, 2004; Noormets et al., 2007; Gough et al., 2008], with peak production occurring around 11–30 years in temperate forests [Pregitzer and Euskirchen, 2004]. Lancaster and Leak [1978] have shown that white pine productivity tends to peak around the age of 15 years, which is close to the age of TP89, the site with the highest Re.

4.4. Comparison of Re Derived by Chamber Versus Eddy Covariance Technique

[55] Our results agree with literature studies that generally report higher values of chamber-estimated Re as compared to eddy covariance-based estimates (Re_ec). For example, Griffis et al. [2004] reported chamber-based Re to be higher than Re_ec by 20%–37% at the boreal aspen forest in Canada and Gaumont-Guay et al. [2006] reported chamber-based Re to be higher by 25% at the same aspen site. Similarly, Lavigne et al. [1997] reported chamber-based Re to be higher by 20%–40% compared to Re_ec at the six boreal forest sites in Canada. Law et al. [1999] reported 50% higher chamber-based Re estimates compared to Re_ec in their study. In contrast, Tang et al. [2008] reported only a 2% difference between the two methods but with chamber-based Re values being higher. Our percent differences (18%–70% on average) were within those reported in the literature.

[56] The discrepancy between chamber- and eddy covariance-based estimates could be due to numerous reasons and are difficult to account for. For example, the eddy covariance technique may underestimate emissions during nighttime because of low turbulence conditions. For a more detailed account of possible causes of Re_ec underestimation, see a recent review by Aubinet [2008]. Alternatively, it may be possible that what was measured by the chambers was not within or representative of the overall tower footprint, thus causing discrepancies between the resulting estimated Re values [Lavigne et al., 1997]. There could also be a number of errors in chamber methods, related to inadequate estimates of biological indices, such as LAI (i.e., if too few measurements were taken during the year to account for seasonality in LAI as well as vertical difference in Rf within the canopy) and sapwood volume (i.e., compressing the tree core, when coring the tree to measure sapwood width in a stem sample) used for upscaling. Lavigne et al. [1997] discusses in more detail some of the challenges in upscaling chamber-based estimates of Re to those derived with the eddy covariance technique.

[57] The continued observed discrepancy between eddy covariance estimates of Re and chamber-based estimates, across a number of studies in different ecosystems, suggests that caution should be taken when one tries to estimate the relative percent contribution of Rs to Re at a site by using chamber-based estimates of Rs and eddy covariance-based estimates for Re. Conclusions about the ecosystem′s carbon cycle derived from such calculations could be flawed. On the basis of our study, we would caution against such a practice, until differences between the two methods have been resolved, or unless the researchers can demonstrate close agreement between Re estimated by both methods for their sites. Otherwise, the relative contribution of Rs to Re for a given site could be grossly overestimated.

5. Conclusions

[58] We measured CO2 emissions from soil (Rs), foliage (Rf), and live woody tissue (Rw) in four temperate white pine (Pinus strobus L.) ecosystems aged 67, 32, 17, and 4 years old at the time of the study. Temperature was the dominant environmental factor driving temporal variability of all three individual components of ecosystem respiration, Re. However, other environmental and physiological factors also showed statistical significance in their control on intersite and temporal variability in Re.

[59] Overall, our results suggest that young actively growing stands may have CO2 emissions comparable to mature stands and that in such stands Rf may dominate Re, especially during the growing season. Intersite variability in CO2 emissions was attributed to differences in stand physiological characteristics, such as the presence of the LFH horizon, canopy cover, foliar, and soil nutrient status. These differences in stand characteristics were reflective of site quality and stand age and highlight the importance of considering, both, stand age, and the knowledge of past land use history, when assessing carbon budgets of planted or afforested ecosystems. However, more chronosequence studies are required to confirm this trend because we were unable to separate differences in site quality due to our limited data set. Improved site quality, in terms of soil water availability and nutrients, may overshadow or confine any age-related trends.


[60] Funding for this study was provided by the Natural Sciences and Engineering Research Council (NSERC) of Canada Discovery and Strategic Project Grants and NSERC, the Canadian Foundation for Climate and Atmospheric Sciences (CFCAS), and BIOCAP Canada Foundation funded Fluxnet-Canada Research Network as well as CFCAS funded Canadian Carbon Program. Support from the Canadian Foundation of Innovation, the Ontario Innovation Trust, Ontario Ministry of Environment, and McMaster University, is also acknowledged. In-kind support from the Ontario Ministry of Natural Resources (OMNR), the Long Point Recreation and Conservation Authority, the Canadian Forest Service, and Ontario Power Generation is appreciated. We thank Steve Williams, from OMNR, for his assistance in site selection and maintenance of the oldest stands. We thank Frank Bahula and Bruce Whitside, and their families, for providing access to the forests on their properties (TP89 and TP02). We are grateful to Eugenia Aoucheva, Fauzia Arain, Rose Blair, Sven D′Souza, Lara Kujtan, Mahmoud Pejam, Olesia Peshko, Rao Polasam, Talar Sahsuvaroglu, Shuhua Yi, Fengming Yuan, Dali, and Jagadeesh Yeluripati for their help in the field. We thank anonymous reviewers for their prompt and constructive reviews.