Interannual variability of the carbon balance of three different-aged Douglas-fir stands in the Pacific Northwest

Authors

  • Praveena Krishnan,

    1. Biometeorology and Soil Physics Group, Faculty of Land and Food Systems, University of British Columbia, Vancouver, British Columbia, Canada
    2. Now at Atmospheric Turbulence and Diffusion Division, NOAA, Oak Ridge, Tennessee, USA.
    Search for more papers by this author
  • T. Andrew Black,

    1. Biometeorology and Soil Physics Group, Faculty of Land and Food Systems, University of British Columbia, Vancouver, British Columbia, Canada
    Search for more papers by this author
  • Rachhpal S. Jassal,

    1. Biometeorology and Soil Physics Group, Faculty of Land and Food Systems, University of British Columbia, Vancouver, British Columbia, Canada
    Search for more papers by this author
  • Baozhang Chen,

    1. Department of Forest Resources Management, Faculty of Forestry, University of British Columbia, Vancouver, British Columbia, Canada
    2. Now at Institute of Geographic Science and Natural Resources Research, Chinese Academy of Sciences, Beijing, China.
    Search for more papers by this author
  • Zoran Nesic

    1. Biometeorology and Soil Physics Group, Faculty of Land and Food Systems, University of British Columbia, Vancouver, British Columbia, Canada
    Search for more papers by this author

Abstract

[1] The seasonal and interannual variability of gross ecosystem photosynthesis (Pg) and ecosystem respiration (Re), and their relationships to environmental variables and stand characteristics were used to explain the variation of eddy-covariance-measured net ecosystem productivity (FNEP) of three different-aged Douglas-fir stands located on the east coast of Vancouver Island in British Columbia, Canada. During the 9-year period, 1998–2006, which included a strong El Niño/La Niña event, the near-end-of-rotation stand (DF49, 57 years old in 2006) was a moderate carbon (C) sink for CO2 with annual FNEP ranging from 267 to 410 g C m−2 yr−1 (mean ± SD, 357 ± 51 g C m−2 yr−1). The pole/sapling stand (HDF88, 18 years old in 2006) was a weak C source (FNEP = −64 ± 75 g C m−2 yr−1), and the recently harvested stand (HDF00, 6 years old in 2006) was a large C source (FNEP = −515 ± 88 g C m−2 yr−1) during 2002–2006. Irrespective of stand age, all sites responded quite similarly to changes in environmental variables during each year. Daily total values of Pg and Re were highest in July–August in all three stands, while daily FNEP peaked during April–June at DF49, May–June at HDF88, and June–July at HDF00. Reductions in root-zone soil water content decreased both Pg and Re especially during the dry period from May to September, and this effect was more pronounced in the younger stands. Evapotranspiration and dry-foliage surface conductance also decreased with decreasing root-zone soil water content whereas water use efficiency appeared to be conservative, especially at DF49. Increasing spring temperature had a positive effect on annual Pg and Re but caused a slight decrease in annual FNEP. During the summer to autumn transition period, increases in soil water content resulted in a greater increase in Re than Pg causing a reduction in FNEP. The interannual variation in the C balance was determined mainly by the interannual variation in Re for the near-end-of-rotation stand and Pg for the two younger stands. The results indicate that regardless of the stand age, interannual variability in the C balance was mainly determined by year-to-year variability in spring temperature and water availability in late summer.

1. Introduction

[2] Characterizing the effects of the interannual variability of climate on the carbon (C) balance of ecosystems is necessary for understanding the causes of the interannual variability of global CO2 budget and also predicting the consequences of projected climate change in response to increasing global temperature [Barford et al., 2001; Intergovernmental Panel on Climate Change (IPCC), 2007]. Even though the northern high-latitude boreal and temperate forest biomes are known to be sinks for atmospheric CO2 [Keeling et al., 1996; Myneni et al., 1997; Sellers et al., 1997; Schimel et al., 2001], ecophysiological factors can change an ecosystem from a C sink to a C source or vice versa [Goulden et al., 1996; Lindroth et al., 1998; Dunn et al., 2007]. In addition to climatic variability, land use change, changes in age distribution and species composition, recovery and response time of a forest ecosystem from a past disturbance like fire, harvesting and insect defoliation can alter the magnitude of C gain or loss by an ecosystem over diurnal, seasonal, annual and longer timescales. This suggests that for accurately determining the contribution of forested ecosystems to the global C budget and also for assessing the impact of climate change and alternative land use at landscape and regional scales, adequate understanding of the processes that control net CO2 exchange of different vegetation types over the course of stand development is required [Turner et al., 2000]. This information is also crucial to quantify and predict vegetation feedback on the climate system [Cramer et al., 2001; Heimann and Reichstein, 2008].

[3] The C balance at the ecosystem level (net ecosystem productivity, FNEP) is the difference between gross ecosystem photosynthesis (Pg) and ecosystem respiration (Re) and is typically almost an order of magnitude smaller than these nearly offsetting components. Assessing the factors controlling FNEP from diurnal to interannual timescales is challenging because of the complexity of the processes controlling its components. Previous studies have reported that both Pg and Re are highly sensitive to changes in solar irradiance, temperature, water availability and leaf area index (LAI) in temperate [Reichstein et al., 2002, 2007] and boreal [Barr et al., 2007; Krishnan et al., 2006, 2008] stands. There is an ongoing debate about which component largely controls the interannual variation in FNEP for various ecosystems. Studies on southern boreal ecosystems, where the growing season is short, show that an increase in spring temperature causes an increase in annual FNEP by causing earlier leaf emergence in deciduous forest and Pg to increase faster than Re in coniferous forests [Black et al., 2000, 2005; Barr et al., 2002]. However, Valentini et al. [2000] reported that Re was the main determinant of the variation in the C balance along a continental gradient in Europe. Studies have shown that the C balance of a forest varies dramatically during stand development by changing its status from a C source in the early stages of development to a C sink in the intermediate and mature stages, and can even maintain its status as a sink or turn into a C source as it ages [Schulze et al., 1999; Amiro et al., 2005; Chen et al., 2002; Clark et al., 2004; Bond-Lamberty et al., 2004; Humphreys et al., 2006; McMillan et al., 2008]. The above studies showed how stand age influences annual FNEP; however, the information on how climate causes interannual variation in ecosystem C exchange at different stages of development following disturbance is lacking. Long-term measurements are needed to quantify the interannual variability of C exchange at different ages of the stand, to relate these differences to environmental forcings, and to determine the presence of long-term trends. In this context, direct, long-term measurement of CO2 fluxes using the eddy-covariance (EC) technique permits the assessment of the C exchange at different stages of development.

[4] In this paper, we examine the interannual variation in the C balance of three Pacific Northwest (PNW) Douglas-fir (Pseudotsuga menziesii var. menziesii (Mirb.) Franco stands at different stages of development after harvesting. Coastal Douglas-fir occurs in a large geographical region (>20 million ha) in the PNW from central California to the midcoast of British Columbia (BC) [Hermann, 1985] and likely plays a significant role in the global C cycle. The forests of the PNW, which are often dominated by Douglas-fir, a commercially desirable timber species, store more C per unit area than any other forested area of North America [Turner et al., 1995] and with improved management are capable of storing more C than they do at present [Smithwick et al., 2002]. The climate in the PNW, which is characterized by wet mild winters and warm and dry summers, contrasts strikingly with other temperate forest regions in eastern United States, eastern Asia and Europe as they have a more even distribution of precipitation during the year with no reduction during the growing season [Waring and Franklin, 1979]. El Niño–Southern Oscillation (ENSO) with a periodicity of 2–7 years [Rasmussen and Wallace, 1983)] and Pacific decadal oscillations (PDO), primarily on decade-to-decade timescales, have been identified as key sources of interannual climate variations in the PNW. Warm phases of ENSO (El Niño) and PDO typically induce warmer and drier conditions in the PNW, while cool phases of ENSO (La Niña) and PDO bring cooler, wetter weather to the region [Rasmussen and Wallace, 1983; Case and Peterson, 2005]. In a recent analysis on trends in temperature and precipitation in the PNW, Mote et al. [1999] and Mote [2003] found that temperature and precipitation have increased more than their respective global averages and this is likely to continue with increased precipitation mostly occurring in winter. In this context, it is important to investigate the impact of climate variability on the C balance and how these ecosystems may respond to projected climate change.

[5] The study sites included a young plantation, a pole/sapling stand and a near-end-of-harvest-rotation stand (hereafter referred to as HDF00, HDF88 and DF49, respectively). Previous studies on this chronosequence have focused on the climatic sensitivity and uncertainty of the C balance [Morgenstern et al., 2004], soil and ecosystem respiration [Drewitt et al., 2002; Jassal et al., 2007, 2008a], and variability and controls of sensible heat, latent heat and CO2 fluxes [Humphreys et al., 2003, 2005, 2006; Jassal et al., 2008b]. In this paper, we focus on the interannual variability in the C balance of the chronosequence. The main objectives of this paper are (1) to examine the seasonal and interannual variation in FNEP, Pg and R of the three stands, (2) to determine the responses of Pg and Re to environmental variables at different stages of stand development, and (3) to understand the processes controlling the interannual variability of FNEP, Pg and Re.

2. Materials and Methods

2.1. Study Sites

[6] The three stands are located between Buckley Bay and Campbell River within 13 km of Georgia Strait on the east coast of Vancouver Island, BC, Canada. The oldest stand (DF49) was planted with Douglas-fir seedlings in 1949 after the original stand was logged in 1937 and slash-burned in 1943. In 2004, it was composed of 80% Douglas-fir, 17% western red cedar (Thuja plicata Donn) and 3% western hemlock (Tsuga heterophylla (Raf.) Sarg.) [Humphreys et al., 2006] and its leaf area index (LAI) was 7.3 [Chen et al., 2006]. DF49 is a 130-ha stand on a 5–10° NE facing slope located only 3 km NWN of the plantation (HDF00), which was clear-cut-harvested in the 1999/2000 winter and planted in the spring of 2000 with 1-year-old seedlings (93% Douglas-fir and 7% western red cedar). About 48 km SE of these two sites is the pole/sapling stand (HDF88) established in 1988. This stand was clear-cut in 1987, burned, planted in 1988 with 75% Douglas-fir, 25% western red cedar and 4% grand fir (Abies grandis var. grandis) and sprayed with herbicide to control competing brush in 1992. The chronosequence is part of the seasonal dry variety of the temperate rain forest that covers much of North America's PNW in the coastal western hemlock (CWH) biogeoclimatic zone [Meidinger and Pojar, 1991]. Even though the three stands were located in very similar ecosystems and experienced similar weather, there were differences in elevation and soil characteristics between the sites. Key stand and soil attributes, and flux-footprint areas at each site are given in Table 1 and further details are provided elsewhere [Morgenstern et al., 2004; Humphreys et al., 2006; Chen et al., 2009]. The prevailing wind directions at the sites were NE to E during daytime and SW to W during nighttime, and are attributed to the topography and land-sea circulation due to the sites' proximity to Georgia Strait.

Table 1. Sites Characteristics
 DF49HDF88HDF00
  • a

    Stand height and density is for live trees in 2002.

  • b

    Chen et al. [2009].

  • c

    Both woody debris and live biomasses are the sum of all aboveground tree components and understory vegetation [Humphreys et al., 2006].

  • d

    Total aboveground biomass is the sum of tree wood, tree foliage, shrub, nonwoody biomass while total belowground biomass is the sum of fine root and coarse root biomass.

  • e

    LAI at DF49 [Chen et al., 2006].

  • f

    Seasonal range of stand LAI values in 2002 at HDF88 and HDF00 [Humphreys et al., 2006].

  • g

    Mean values during 1998–2006 for DF49 and 2002–2006 for HDF88 and HDF00, respectively.

Location49°52′7.8″N, 125°20′6.3″W49°32′10.49″N, 124°54′7.18″W49°52′1.08″N, 125°16′43.80″W
Elevation (m)320173 188
Slope (deg)5–10 NE2–5 SE0–2
Stand heighta (m)34 7 0.8
Distance from coastline (km)15.73.111.6
Stand densitya (trees ha−1)1150 12401500
Stand area (ha)13011032
80% cumulative flux footprint area (annual average)b (ha)93178
Mean DBH (cm)28.37.31.02 (at base)
Woody debris biomassc (kg m−2)0.7–103.4–16.50.5–2.2
Live aboveground biomassc (kg m−2)8.5–23.00.8–2.1<0.1–0.4
Total aboveground biomassd (kg m−2)13.811.1390.165
Total belowground biomassd (kg m−2)1.25--
Tree cover type77% Douglas-fir, 18% western red cedar, 4.6% western hemlock75% Douglas-fir, 21% western red cedar, and 4% grand fir97% Douglas-fir, 3% western red cedar
UnderstoryVanilla leaf deerfoot, Oregon grape, mossesSalal, Oregon grape, fireweed and bracken fern, huckleberryGrasses, sword and bracken fern, herbs, woody shrubs: twinflower, honeysuckle rubrus spp., Oregon grape
Stand historyClear-cut, slash burned, planted stock mixed with natural regenerationClear-cut, broadcast burned, plantedClear-cut, planted
LAIe,f (m2 m−2)7.33.0–7.20.3–2.5
Soil classificationHumo-ferric podzolHumo-ferric podzolHumo-ferric podzol
Soil textureGravelly loamy sandGravelly loamGravelly loamy sand to sand
Soil organic layer (cm)34.13.6
Flux measurement height (m)42122.5
u* threshold (u*T) (m s−1)0.300.160.08
Mean annual temperatureg (°C)8.49.68.9
Mean annual rainfallg (mm)124515461227

2.2. Eddy-Covariance Flux Measurements

[7] Continuous, half-hourly turbulent fluxes of CO2, water vapor (E) and sensible heat (H) were measured above the canopy using the EC technique [Morgenstern et al., 2004; Humphreys et al., 2005, 2006]. EC fluxes were measured at all sites above the canopy (Table 1) using a three-dimensional sonic anemometer-thermometer (Model R2 or R3, Gill Instruments, Lymington, UK) and a closed-path, temperature-controlled infrared gas (CO2/H2O) analyzer (IRGA) (model LI-6262, LI-COR Inc., Lincoln, NE, USA) or open-path IRGA (HDF88 up to 2003, model LI-7500, LI-COR Inc.). Closed-path IRGA analog signals (CO2 and H2O mole fractions) were sampled at 125 Hz with a data acquisition system (Model DaqBook/200, Iotech Inc., Cleveland, OH, USA), digitally filtered and down sampled to 20.83 Hz and transferred every half-hour to a site computer for further flux calculations. Further details on the experimental setup, calibration and eddy flux calculations can be found in the works of Morgenstern et al. [2004] and Humphreys et al. [2006].

2.3. Supplementary Meteorological Measurements

[8] Meteorological variables continuously measured at all three sites included above-canopy downwelling and upwelling solar and longwave radiation (model CNR1, Kipp and Zonen B.V., Delft, The Netherlands), above-canopy downwelling and upwelling photosynthetically active radiation (PAR) (model 190SB, LI-COR Inc.), above- and within-canopy air temperature and relative humidity (model HMP35CF or HMP45CF, Vaisala Oyj, Helsinki, Finland), rainfall (two tipping-bucket rain gauges, model 2, Texas Electronics, Dallas, TX, USA or model 2501, Sierra Misco, Berkeley, CA, USA), soil temperature at the 2, 5, 10, 20 and 50 cm depths (with copper-constantan thermocouples), soil-heat flux (3 heat-flux plates, model F, Middleton Instruments, Melbourne, Australia), volumetric soil water content at the depths of 1–2, 10–12, 35–48 and 70–100 cm (with reflectometers model CS-615, Campbell Scientific Inc. (CSI), Logan, UT, USA) at two locations. A precipitation gauge (model I-200B, Geonor A.S., Oslo, Norway) for measuring the liquid-water equivalent of snowfall was installed at HDF00 near the end of 2003.

2.4. Calculation of Annual Totals, Water Use Efficiency, and Surface Conductance

[9] The half-hourly turbulent fluxes were calculated using the covariance of the fluctuations in the vertical wind velocity and scalar quantities, namely, air temperature for sensible heat, molar mixing ratio of CO2 for CO2 flux and the molar mixing ratio of water vapor for water vapor flux. The net ecosystem exchange (FNEE) of CO2 was calculated by adding the turbulent flux of CO2 to the time derivative of the air-column storage of CO2 below the EC sensor level (i.e., the storage flux). In this analysis, the storage flux was calculated following Morgenstern et al. [2004] using the CO2 mixing ratio measured at the EC sensor level. FNEE provides a direct measure of FNEP (i.e., FNEP = −FNEE). Positive values of FNEP correspond to C gain by the ecosystem and negative values to C loss. Humphreys et al. [2006] found that average half-hour energy balance closure was 0.80, 0.90 and 0.91 for DF49, HDF88 and HDF00, respectively, for 2002; however, closure corrections were not applied in this analysis.

[10] A common procedure was applied to all sites to fill half-hour data gaps and to obtain annual estimates of FNEP This included filling gaps caused by missing data and created by removing nighttime fluxes when turbulence was not fully developed, i.e., friction velocity (u*) was less than a threshold value (u*T). Values of u*T of 0.30, 0.16 and 0.08 m s−1 were used for DF49, HDF88 and HDF00, respectively [see Morgenstern et al., 2004; Humphreys et al., 2006]. This procedure removed ∼60–70% of the nighttime fluxes at each site. Small gaps (i.e., <2 h) in FNEP were filled by linear interpolation. Measured Re was obtained as Re = FNEE during periods when Pg was known to be zero, i.e., during night in the growing season and during night and day in the winter when both air (Ta) and shallow soil (Ts) temperatures were less than 0°C. An exponential relationship between measured Re and Ts at the 2-cm depth for each year was obtained as follows:

equation image

This logarithmic transformation of the data minimizes inhomogeneous scatter [Morgenstern et al., 2004] and allows the assumption of normality and homoscedasticity to be valid for linear least squares regression [Humphreys et al., 2005, 2006]. Pg was estimated as FNEP + daytime Re (calculated using equation (1) with daytime Ts) and assumed to be zero at night and during periods when both Ta and Ts < 0°C. Gaps in Pg were filled using the rectangular hyperbolic dependence of Pg on PAR (Q) (Michaelis-Menten light-response equation) given by

equation image

where α is the quantum yield and Px is the photosynthetic capacity (Pg at light saturation). To account for short-term changes in environmental conditions, a time-varying parameter was determined within a 100-point moving window as the slope of a linear regression, forced through the origin, of the modeled Pg (and Re) versus measured Pg (and Re) and was applied as a multiplier to the above relationships [Barr et al., 2004]. Gaps in FNEP > 2 h were filled using modeled PgRe. Gap-filled values of FNEP, Re for both nighttime and daytime, and Pg were used to calculate annual totals of FNEP, Re, and Pg. As mentioned above, the u*T criteria removed 60–70% of the nighttime and hence 30–40% of the total flux data while the original data (before applying u*T) loss was <10% when using the closed-path IRGA and 20–30% when using the open-path IRGA (at HDF88). Uncertainties in annual FNEP associated with the random error in half-hourly fluxes and the gap-filling procedure were obtained by resampling half-hourly fluxes after introducing a 20% random error and also by using Monte Carlo simulation [Griffis et al., 2003; Morgenstern et al., 2004; Krishnan et al., 2006, 2008] by artificially generating gaps (up to 40% of the total half hours in a year, with continuous gaps varying from a half hour to 10 days) using a uniformly distributed random-number generator. This procedure was repeated 1000 times and the 95% confidence levels in the annual estimates of FNEP, Pg and Re were calculated.

[11] To better understand the physiological response of the ecosystem to changes in soil water content, canopy-level water use efficiency and dry-foliage surface conductance were used. Water use efficiency (WUE), an important indicator of ecosystem function, adaptation and productivity of plants in water-limited areas [Xu and Hsiao, 2004], was calculated as the ratio of daily daytime (Q > 200 μmol m−2 s−1) average Pg to average evapotranspiration (E), i.e., WUE = Pg/E. The surface conductance to water vapor transfer (gs) was calculated by rearranging the Penman–Monteith equation [Monteith and Unsworth, 1990] as given by Krishnan et al. [2006]. The daytime mean value of dry-foliage gs was calculated by averaging daytime gs values for Q > 200 μmol m−2 s−1 for those days without precipitation.

3. Results

3.1. Seasonal and Interannual Variations in Environmental Variables

[12] Figure 1, which shows the annual cycles of monthly total Q, rainfall (R) and monthly mean above-canopy air temperature (Ta), soil temperature at the 2-cm depth (Ts), daytime (Q > 0) atmospheric saturation deficit (D), average soil water content in the 0–30 cm layer (θ) for 2002–2005, indicates that the three sites experienced similar weather with cool and wet winters and dry and warm summers. Q generally peaked during June and July with slightly higher values at HDF88. The variations in Ts and D closely followed those in Ta, all being highest in July and August with the highest values occurring at HDF00 as expected being a recently clear-cut site. However, winters were slightly warmer at HDF88 than at the other two sites. More than 75% of the annual rainfall occurred between October and March leading to high θ during these months. The low rainfall rates during May to September resulted in dry growing seasons with the lowest θ values occurring during July to September and August being the driest month at all three sites. R at HDF88 was higher than other sites especially during winter–spring months, but not during growing season. However, θ at HDF88 was generally higher than at the other two sites due to finer textured soil (Table 1).

Figure 1.

Average monthly values of (a) total downwelling PAR (Q), (b) mean above-canopy air temperature (Ta), (c) mean soil temperature at the 2-cm depth (Ts), (d) mean daytime (Q > 0) above-canopy atmospheric saturation deficit (D), (e) total rainfall (mm), and (f) mean soil water content (θ) in the 0–30 cm layer for 2002–2006 at DF49, HDF88, and HFD00. The dotted curve represents the 1971–2000 precipitation (snowfall plus rainfall) normals for Campbell River Airport, elevation 105 m, 10 km NE of DF49. Average monthly snowfall totals (mm liquid-water equivalent) for 2002–2006 were January, 24.3; February, 8.6; March, 27.7; April, 1.0; November, 22.1; December, 19.7.

[13] Interannual variations in annual total Q (QAnnual), annual mean above-canopy air temperature (TAnnual), spring (March to May) mean above-canopy air temperature (TSpring), annual total rainfall (RAnnual) and total May–September rainfall (RMay–Sept) at DF49, HDF88 and HDF00 are shown in Figure 2. All three sites showed similar interannual variation with HDF88 being somewhat warmer. Based on climate data from the Environment Canada meteorological station at the Campbell River Airport (49° 57′N, 125° 16.2′ W, elevation 105 m), 10 km NE of DF49, the 1971–2000 mean annual temperature was 8.4°C and the mean RAnnual was 1452 mm (snowfall 109 mm liquid-water equivalent). ENSO has been identified as a key source of interannual climatic variations for the PNW and has a periodicity of 2–7 years [Rasmussen and Wallace, 1983] whereas PDO fluctuations influence climate on a decadal scale. In the PNW, El Niño events typically occur from October through March or April, and induce mild winters followed by unusually warm springs and dry summers, while La Niña years are generally characterized by wet winters and relatively cool and cloudy conditions in the following spring and summer [Shabbar et al., 1997; Shabbar and Khandekar, 1996]. During the study period from 1998 to 2006, El Niño events occurred in the winters of 1997–1998 (strong), 2002–2003 (medium) and 2006–2007 (weak) while La Niña events occurred during the winters of 1998–1999 (moderate) and 1999–2000 (strong) (http://www.msc-smc.ec.gc.ca/education/elnino/comparing/enso1950_2002_e.html).

Figure 2.

Interannual variations in (a) annual total of PAR (QAnnual), (b) annual mean above-canopy temperature (TAnnual), (c) spring (March–May) mean above-canopy air temperature (Tspring), (d) annual total rainfall (RAnnual), and (e) May–September total rainfall (RMay–Sept) at DF49, HDF88, and HDF00. The dotted lines represent 1971–2000 climate normals for Campbell River Airport, elevation 105 m, 10 km NE of DF49. At the airport, annual rainfall was 89–97% of annual precipitation for 1998–2006. Snowfall (liquid-water equivalent) at HDF00 (measured using the Geonor precipitation gauge) and at the airport in 2004 were 43 and 86 mm and in 2005 were 63 and 52 mm, respectively.

[14] The impact of El Niño and La Niña events were pronounced in 1998 and 1999 as evident from temperature and RMay–Sept. The highest TAnnual (9.1°C) and second highest spring, summer and autumn temperatures at DF49 occurred in 1998 with dry summer periods as a result of below-normal RMay–Sept. The highest RAnnual occurred in 1998. The coldest spring and summer temperatures of the measurement period occurred in 1999 following the La Niña event in the previous winter (1998–1999). The 1999–2000 La Niña event did not significantly influenced temperatures and precipitation at the study sites and 2000 was a normal year in terms of TSpring and RMay–Sept. So was the case with 2001. 2002 experienced the second coldest spring, which resulted from two weeks of cold weather in March. A moderate El Niño event in 2002–2003 had no impact on TSpring but RMay–Sept was below normal. Of the other neutral years, 2004 and 2005 had warm springs and wetter than normal May–September periods. Summer and April–June temperatures (not shown) at DF49 were highest in 2004, even higher than that in 1998. Even though 2005 was the year with highest TSpring (8.7°C), annually, it was a cool year with the third (to 1999 and 2001) coolest summer (June–August) and the lowest Q of the 9 growing seasons. Figure 2 also suggests an increase in Tannual during the study period excluding 1998, the strong El Niño year. All three sites, especially HDF00, had low RMay–Sept in 2003, a relatively warm year following the moderate El Niño event in 2002. Even though, 2005 had slightly higher RMay–Sept, QAnnual and TAnnual were lower than those in 2004. 2006 had a normal TSpring with a below-normal RMay–Sept, especially at HDF00. Even though annual rainfall was slightly higher at HDF88 during most of the years, RMay–Sept was quite similar in magnitude to values for the other two sites. The results suggest that the main interannual variations in climate at these sites result from variation in TSpring and summer rainfall rather than annual means of the environmental variables.

3.2. Interannual Variation in Annual C Fluxes

3.2.1. Annual C Fluxes

[15] The patterns of the interannual variation of FNEP at the three sites were similar during 2002–2006. This was also true for Pg and Re (Figure 3). DF49 was a moderate C sink (mean ± SD = 328 ± 52 g C m−2 yr−1 during 2002–2006 and 357 ± 51 during 1998–2006) while HDF88 and HDF00, during 2002–2006, were a weak C source (−64 ± 75 g C m−2 yr−1) and a large C source (−515 ± 88 g C m−2 yr−1), respectively. The interannual variation of Re at DF49 (1830 ± 173 during 2002–2006 and 1767 ± 147 g C m−2 yr−1 during 1998–2006) was greater than that of Pg (2158 ± 163 during 2002–2006 and 2124 ± 125 g C m−2 yr−1 during 1998–2006). For HDF88 and HDF00 during 2002–2006, the interannual variation of Pg (1423 ± 193 and 781 ± 226 g C m−2 yr−1, respectively) was greater than that in Re (1487 ± 147 and 1297 ± 185 g C m−2 yr−1, respectively), while Pg for HDF88 was 642 g C m−2 yr−1 greater than that for HDF00 and Re was only 190 g C m−2 yr−1 greater (Table 2). HDF88 changed from a weak C source (−105 ± 10 g C m−2 yr−1) during 2002–2004 to a weak sink (21 ± 30 g C m−2 yr−1) during 2005–2006. As shown in Table 2, intersite differences were greater than interannual variation showing the dominating role of stand age [Humphreys et al., 2006] in determining the magnitude of FNEP. Regardless of stand age, the general pattern of interannual variations in FNEP, Pg and Re were roughly similar suggesting similar responses to environmental variables with few exceptions due to stand age.

Figure 3.

Interannual variation in annual (a) net ecosystem productivity (FNEP), (b) gross ecosystem photosynthesis (Pg), and (c) ecosystem respiration (Re) at the three sites. The inset in each plot shows relationships of the FNEP, Pg, and Re to stand age. Uncertainty in annual FNEP (see section 2.4) were ±44, ±22, and ±24 g C m−2 yr−1 during 1998–2006 at DF49 and 2002–2006 at HDF88 and HDF00, respectively.

Table 2. Interannual and Intersite Variation of FNEP, Pg, and Re During 2002–2006
 Variable (g C m−2 yr−1)Mean ± SD (g C m−2 yr−1)Range (g C m−2 yr−1)CVa
  • a

    Coefficient of variation (CV) is calculated as the ratio of standard deviation (SD) to the mean.

  • b

    At DF49 1998–2006 mean ± SD, values of range and CV, were 357 ± 51, 143, and 0.14 for FNEP, 2124 ± 125, 386, and 0.06 for Pg, and 1767 ± 146, 429, and 0.08 for Re, respectively.

Interannual Variationb
Site
   DF49FNEP328 ± 521190.16
 Pg2158 ± 1633860.08
 Re1830 ± 1733960.09
   HDF88FNEP−64 ± 751531.18
 Pg1423 ± 1934890.14
 Re1487 ± 1483680.10
   HDF00FNEP−516 ± 881750.17
 Pg781 ± 2265830.29
 Re1297 ± 1854350.14
 
Intersite Variation
Year
   2002FNEP−136 ± 4218423.09
   2003 −109 ± 4679334.29
   2004 −153 ± 4308602.82
   2005 −15 ± 38977725.57
   2006 −6 ± 40280468.18
   2002Pg1250 ± 71314250.57
   2003 1302 ± 73314560.56
   2004 1551 ± 74414790.48
   2005 1705 ± 60011990.35
   2006 1462 ± 66513290.45
   2002Re1389 ± 2895770.21
   2003 1411 ± 2775230.19
   2004 1704 ± 3256190.19
   2005 1721 ± 2154220.13
   2006 1468 ± 2625230.17

[16] In 1998, the warmest year of the record (Figure 2) with a warm spring and dry growing season, both Pg and Re at DF49 were higher than for all years except for 2004 and 2005 (Figure 3). In 2002, the lowest Pg of the record, mainly due to the cold period in March, led to the second lowest value of annual FNEP at DF49. The lowest FNEP values at all three sites were recorded in 2004, the second warmest year for DF49 and the warmest year for HDF88 and HDF00 with a wet and warm growing season. At DF49, Re and Pg in 2004 were higher than their 9-year mean by 17% and 10%, respectively, and were the highest of the record, while FNEP was lower than its 9-year mean by 32%. 2005 had the second highest Pg and Re of the record. At HDF88 and HDF00, Pg, Re and FNEP were highest in 2005. In 2006, a cool year with a dry growing season, Re was suppressed more than Pg resulting in the third highest FNEP of the record at DF49 and the second highest FNEP at HDF88 and HDF00. FNEP at DF49 was highest during the normal years, 2000 and 2001 (400 and 410 g C m−2 yr−1, respectively).

3.2.2. Climatic Controls of Interannual Variability of Annual C Fluxes

[17] To obtain better insight into the factors controlling the interannual variability of C fluxes, a linear regression analysis of the relationship between C balance components and environmental variables was carried out on monthly and seasonal scales (Tables 3 and 4). The monthly regression analysis suggests that temperature had a positive effect on Pg and Re at DF49 until July whereas it had a negative effect on FNEP from May onward. In the latter part of the year, the negative impact of temperature on FNEP was due to the relatively greater increase in Re than in Pg with increasing temperature at all three sites. The impact of θ on FNEP, Pg and Re at all sites was most pronounced in August. The interannual variation of seasonal weather, and its control on C balance components was discernible only for spring (Table 4) and was mainly a result of the effect of Ta on Pg and Re at all three sites although relationships between FNEP and Ta were not significant. It appears that environmental variables exerted a strong effect on the C balance at half hourly to daily (data not shown) and even monthly timescales. Progressing to the seasonal scale, it was difficult to distinguish the effect except for temperature and to some extent θ.

Table 3. Results of the Linear Regression Analysis (y = bx + c) of the Relationship of Monthly Total C Fluxes to Monthly Mean Environmental Variablesa
Sitey VariableJFMAMJJASOND
  • a

    Here y is the independent variable (FNEP, Pg, and Re), and x is the dependent variable (Ta, Ts, θ, and D). The bold letters indicate the x variables with highly significant relations with the y variable (p < 0.05), nonbold letters indicate relationships with 0.1 < p > 0.05, and letters given in parentheses are for relationships with 0.2 < p > 0.1. Relationships with p > 0.2 are not shown here. Negative signs indicate negative correlations.

DF49FNEPTa, DTsTa, Q−θTa, −Ts, θTa, −Ts, −D, −QQ, −Ta, −Dθ, Q Ts, (Q)equation image, −Ta, Qequation image, −Ta, −Ts
DF49PgTa, (D)Ta, TsTa, Ts, DTa, Ts, −θ, (Q)Ta, Ts, D, (θ)Ta, Ts, QTa, Ts, θ, (D)equation image, −D, −QTs, −TaD, Q, Ts, (−θ)Qθ
DF49ReTaTa, TsTa, TsTa, TsTa, Ts, (θ)Ta, TsTa, Ts, θθTsTaTa, −θTa, Ts
HDF88FNEPθTa, θD−θTa, −equation imageTa, −D, −QTa, −Ts  Q, TsQTa, D, −Ts, (θ)
HDF88PgTs, (θ) Ta, Ts Ta, (Ts)Ta, −D, (−Ts)Ta, −Tsθ, −DTa, (θ)Ts, D  
HDF88ReTsTaTa, TsTsTa, (Ts)TsTa, −TsθTa, θ TsTa, (Ts)
HDF00FNEPDTa, −TsTs, −Ta, θTa, θ, −D, −Q, (−Ts)Ta, Tsθ, (−Ts)Ts, −θTa, −Ts, −QQ, (−θ)Ts, D, QTa, −θD, −Ta, (−Ts)
HDF00Pg  Ta Ta, Tsequation image, Ts D, θTs, −Ta   
HDF00Re Ta, (Ts)Ta, Ts Ta, Tsθ θTa, −Ts, (θ) TsTa, (Ts)
Table 4. Results of the Linear Regression Analysis (y = bx + c) of Seasonal Total C Fluxes to Seasonal Mean Environmental Variablesa
y VariableSiteWinter (DJF)Spring (MAM)Summer (JJA)Autumn (SON)
  • a

    Here y is the independent variable (FNEP, Pg, and Re), and x is the dependent variable (Ta, Ts, θ, and D). The bold letters indicate the x variables with highly significant relations with the y variable (p < 0.05), nonbold letters indicate relationships with 0.1 < p > 0.05, and letters given in parentheses are for relations with 0.2 < p > 0.1. Relations with p > 0.2 are not shown here. Negative signs indicate negative correlations.

FNEPDF49    
HDF88Ts, −θ, (−Ta)(−θ)Ta, (θ), (−D)Ts, −equation image, (Q)
HDF00Ta−θ, D, (Ta)(−Ta), (−Ts), (−D) 
PgDF49 Ta, Ts, D, (−θ)(Ts)Ts
HDF88Ts, (−θ)Ta, (Ts)Q, −D, θ, (−Ta)Q, −D, (−Ta)
HDF00 Ta, Tsθ, −Q, (−D)(θ)
ReDF49 Ta, Ts, (−θ)(Ts)(Ts)
HDF88Ta, Ts, θTa, Ts Ts
HDF00(Ta)Ta, Tsθs)

[18] To examine the influence of Ta during the early part of the growing season on the annual C balance, we carried out linear regression analysis as shown in Figure 4. An increase in TSpring had a positive effect on annual values of Pg and Re at all three sites; however, its effect on annual FNEP was not significant. At DF49, annual FNEP showed a stronger correlation with April to June temperature (r = −0.40, not shown) than with TSpring (r = −0.09). Annual FNEP showed a nonsignificant decrease with an increase in summer and autumn temperature at all three sites; however, the correlation was much higher for HDF88 and HDF00 than DF49. As a result, FNEP had a relatively better negative correlation to TAnnual for HDF88 and HDF00 than for DF49. Both Pg and Re increased with the TAnnual, but the increase was not significant (not shown). The above analysis suggests that annual FNEP at the younger sites was more sensitive to changes in temperature than at DF49.

Figure 4.

Annual gross ecosystem photosynthesis (Pg) and ecosystem respiration (Re) at the three sites as a function of spring temperature, and net ecosystem productivity (FNEP), as a function of spring, summer, autumn, and annual mean air temperature. The numbers in parenthesis are values of slope, intercept, r2, and p for the respective regression relations.

[19] RMay–Sept had a strong positive effect on Pg and Re at all three sites (Figure 5), but its effect on FNEP was negative although not significant. To analyze it further, we examined the response of annual FNEP, Pg and Re to θ of the 0–30 cm soil layer. At DF49, annual FNEP, Pg and Re were more dependent on θ in the latter part of the growing season, mainly August to October. At HDF00, Pg and Re were better correlated with August–October θ (both positive) than RMay–Sept. At HDF88, Pg and Re were only significantly correlated to August θ. The small range of θ at this site made it difficult to assess its effect, so we carried out an analysis comparing the effect of θ of the 0–30 cm soil layer with that of the 0–100 cm soil layer. At DF49, the Pg and Re relationships were similar for both layers but for HDF88, both Pg and Re were better correlated to θ of the 0–30 cm soil layer.

Figure 5.

Annual gross ecosystem photosynthesis (Pg) and ecosystem respiration (Re) at the three sites as a function of May–September rainfall and August–October 0–30 cm soil water content (θ), for DF49, HDF88, and HDF00, respectively. The relationships of annual net ecosystem productivity (FNEP) to rainfall and θ were not significant. The numbers in parenthesis are values of slope, intercept, r2, and p for the respective regression relations.

[20] As shown above, TSpring and August–October θ explained most of the variation in Pg and Re at DF49; and, FNEP correlated better to April–June temperature. Therefore, we used the above environmental variables to develop a regression model to explain the effect of climate induced variability on the C balance at DF49, which is expected to have reached a nearly constant growth rates with little year-to-year change in LAI [Humphreys et al., 2006] and most of the interannual variability attributable to variations in climate [Schwalm et al., 2007; Jassal et al., 2008b]. We added March average temperature to account for the significant effect of frost events in late winter–early spring as in 2002 when it markedly reduced Pg.

[21] The empirical relations for DF49 (1998–2006) using air temperature and θ of the 0–30 cm layer are:

equation image

At the younger sites similar empirical relations using TSpring and late growing season θ (Figure 5) were able to explain only ∼40% of the variance in annual FNEP, Pg and Re. This was likely the effect of stand age on the annual totals of FNEP, Pg and Re at HDF00 and HDF88 (Figure 3 and Table 2). FNEP at these sites showed a better correlation to TAnnual than at DF49.

3.3. Seasonal Variation in C Fluxes

[22] The above analysis shows that the annual C balance is influenced by stand age and variation in seasonal environmental variables. To better understand ecosystem C exchange dynamics at the three sites, a close inspection of the seasonal fluctuation of FNEP, Pg and Re in relation to environmental variables at the three sites is required.

3.3.1. Seasonal C Fluxes

[23] At DF49, where the annual courses of cumulative FNEP showed a similar pattern for the 9 years, C gain started in February and continued to the end of July (Figure 6). The maximum rates achieved in March–April varied little among years with the exception of 2002, the year with the cold March. Pg generally exceeded 100 g C m−2 in March (Figure 7), except in 2002, in response to the increase in Q to values >500 mol m−2 mon−1, while Re in March (Figure 7) generally remained low due to low temperature (Figure 1). This resulted in a marked increase in cumulative FNEP early in the growing season. The interannual differences in C sequestration became visible mainly from May onward (Figure 6). During May–June, Re increased substantially with the increase in temperature (Figure 1) and served to offset the high values of Pg resulting in a decrease in monthly FNEP from May to July. From August to October, FNEP was close to zero resulting in almost no change in cumulative FNEP. In 2004, a year with a warm spring and wet growing season, FNEP was close to zero or slightly negative from August to October resulting in the lowest annual FNEP of the record. Similar variations were observed at HDF88 in 2004.

Figure 6.

Annual cycles of cumulative FNEP at DF49, HDF88, and HDF00.

Figure 7.

The 2002–2006 mean monthly FNEP, Pg, and Re at DF49, HDF88, and HDF00.

[24] In contrast to the interannual variations in cumulative FNEP for DF49, cumulative FNEP at HDF88 early in the year increased each year as stand age increased. However, interannual differences in FNEP observed early in the year changed in the latter part of the growing season (Figure 6). A remarkable increase in cumulative FNEP began in early April of 2005 and 2006 compared with the previous years. This C gain eventually helped HDF88 to slightly exceed C neutrality (i.e., FNEP = 0 annually) by the end of these two years. The relatively high loss of CO2 during most of the year resulted in HDF00 being a strong C source. However, the positive changes in FNEP during June to August 2005 and 2006 resulted in an increase in annual FNEP of 150 g C m−2, similar to those for HDF88 in these two years. While DF49 and HDF88 had significant photosynthesis throughout most of the year, Pg at HDF00 was significant only during April to October (Figure 7). During the peak growing season, monthly total Re at DF49 was ∼100 g C m−2 higher than that at HFD88 or HDF00. During January to March and October to December, Re values at HDF00 were similar to those at DF49 while at HDF88 they were higher than at the other two sites. Wintertime Re was nearly equal to soil respiration [Jassal et al., 2007, 2008a].

3.3.2. Effect of Seasonal Changes in Environmental Variables on the C Balance

[25] To examine the seasonal response of the C balance components to changes in Ta, θ, and to aid in our interpretation of annual FNEP, the time series of 5-day running averages of Ta, θ, FNEP, Pg, Re and E during 1998 (for DF49 only), and 2004, 2005 and 2006 (for all 3 sites) are shown in Figure 8. These years were chosen because 2004 and 2005 were years with a warm spring and a wet growing season whereas 2006 was a year with a normal spring with a warm and dry growing season. 1998 is included to show the pattern during a year with a warm spring, like 2004 and 2005, but with a warm and dry growing season. The main difference between 2004 and 2005 was the precipitation distribution and the variation in θ during the growing season. 2004 had a dry early growing season and a wet late growing season whereas 2005 had a wet early growing season while the rest of the growing season was relatively dry. E is shown here because of the close relationship between the CO2 and water vapor exchange processes. This coupled dependence is further examined using parameters describing physiological responses of the ecosystem, namely, canopy level gs and WUE as shown in Figure 9. At all the three sites, Pg was highly correlated to E (Pg (g C m−2 mon−1) = 5.78E (mm mon−1) + 9.6, r2 = 0.97, Pg = 3.25 E +18, r2 = 0.86 and Pg = 3.68 E −14, r2 = 0.91 for DF49 (1998–2006), HDF88 (2002–2006) and HDF00 (2002–2006), respectively). FNEP, Pg and Re values were highest at DF49. Even though peak FNEP was lowest at HDF00, magnitudes of Pg and Re at HDF00 were higher than those at HDF88 during some part of growing season in 2004, 2005 and 2006. Furthermore, the number of days when daily FNEP > 0 (C sink) was higher at DF49 than at the other two sites (233 ± 15, 167 ± 19, 42 ± 14 days for DF49, HDF88 and HDF00, respectively, during 2004, 2005 and 2006). Maximum daily E at all three sites reached a value of 2.5 mm d−1 indicating a relatively conservative response to stand age.

Figure 8.

Five-day running mean of air temperature (Ta), soil water content (θ) in the 0–30 cm layer, FNEP, Pg, and Re during 2004–2006 at DF49, HDF88, and HDF00. The years 2004 and 2005 were years with a warm spring and a wet growing season whereas 2006 was a year with a normal spring with a dry growing season. The year 1998 is included to show the pattern during a year with a warm spring, like 2004 and 2005, but with a warm and dry growing season. Of the 4 years, 2005 had the coldest growing season. Values of annual E for 2004–2006 were 420, 429, and 398 mm at DF49 (418 mm for 1998), 387, 382, and 361 mm at HDF88, and 302, 316, and 253 mm at HDF00. Annual C fluxes (g C m−2) at DF49 for 1998, 2004–2006 were 379, 267, 355, and 386 for FNEP, 2131, 2338, 2310, and 2112 for Pg, and1752, 2071, 1955, and 1726 for Re, respectively. The annual C fluxes at HDF88 for 2004–2006 were −132, 20 and 15 for FNEP, 1457, 1694, and 1490 for Pg, and 1589, 1694, and 1475 for Re, respectively, and at HDF00, annual values were −593, −421, and −418 for FNEP, 859, 1111, and 784 for Pg, 1452,1533, and 1204 for Re, respectively.

Figure 9.

Five-day running mean of dry-foliage surface conductance (gs) and water use efficiency (WUE) at DF49, HDF88, and HDF00 during May–September for the years considered in Figure 8. Mean May–September WUE values at DF49 for 1998–2006 were 5.95, 5.95, 5.76, 6.43, 5.94, 6.48, 6.29, 6.01, and 6.33 g C kg−1 H2O, and at HDF88 for 2002–2006 values were 3.24, 3.21, 3.74, 4.37, and 4.13 g C kg−1 H2O, and at HDF00 for 2002–2006 values were 2.42, 3.00, 3.30, 3.89, and 3.40 g C kg−1 H2O, respectively.

[26] In 1998, even though Pg, Re and E increased in response to the relatively warmer than normal February to May temperature, the relatively dry conditions during the growing season, especially in August and September, suppressed Pg, Re, E and gs. The greater decrease in Re than Pg in August and September resulted in an increase in FNEP. In 2004, the year with second warmest spring and a warm growing season with the warmest August, the values of θ up to July at all the three sites were similar to those of 2006, which had low RMay–Sept (Figure 8). The dry conditions in early summer had no effect on Pg and Re at DF49, with both being higher than in other years. At HDF88 on the other hand, the dry conditions during May end to July resulted in low values of Pg, Re, and gs. In August of 2004, the increase in θ increased Re more than Pg leading to the lowest values of FNEP for the four years at all the three sites. The warmer than normal spring in 2005 followed by a relatively cool summer, favored Pg and gs at all three sites leading to an increase in FNEP during the early part of the growing season with very discernible effects at the younger sites. Wetter growing conditions from June to September resulted in higher than normal Pg,Re, E and gs at all three sites, especially the younger sites. The cool and wet summer in 2005 suppressed annual Re more than Pg resulting in an increase in annual FNEP over that in 2004 (Figure 3). This effect on Pg was more pronounced at the younger sites resulting in the highest annual values of Pg and FNEP of the record. The dry growing season in 2006, a year with normal TSpring, led to an increase in FNEP at DF49 due to the relatively larger decrease in Re than Pg in response to the dry conditions. At the younger sites, the dry conditions in the latter part of the growing season decreased Pg more than Re resulting in low FNEP during September and October and slightly lower annual FNEP values than those in 2005. However, in a comparison of C fluxes at HDF88 for the years 2005 and 2006, Jassal et al. [2008a] found that the late summer–early fall (August–October) drought in 2006 resulted a greater decrease in Re than in Pg leading to an increase in FNEP over the same period in 2005.

3.3.3. Response of Canopy Level Physiological Parameters

[27] Both WUE and gs at these sites have been reported to decrease in response to increasing D at hourly timescales [Ponton et al., 2006; Humphreys et al., 2003]. As shown in Figure 1, D reached its highest values in August, the driest month at all three sites. Results indicated that high D and dry soil water conditions had a strong negative effect on Pg (Figure 8). The differences in the magnitude of gs with high values at DF49 and low values at HDF00 were as expected considering the ages of the stands and low LAI at HDF00 [Kelliher et al., 1993] (Figure 9). gs values at all the three sites were lowest during July and August, the period with lowest θ and highest D. The lowest values of gs occurred in 2006 due to its dry and warm weather. The changes in gs at these sites closely followed the changes in θ shown in Figure 8. For example, the increase in θ in mid-August 2004 increased gs, Pg and Re at all three sites. Time series of WUE (Figure 9) show that during May to September, it was slightly higher at HDF88 than at HDF00. At HDF88, mean May–September WUE increased from 3.2 to 4.4 g C kg−1 water and at HDF00 2.4 to 3.9 g C kg−1 water from 2002 to 2005. At the two younger sites, WUE was lower in 2006. At DF49, WUE in July 2006 was slightly higher than that in 2005. The interannual (1998–2006) variability in mean May–September WUE at DF49 was less than 4% of the mean (6.12 ± 0.24 g C kg−1 H2O) whereas at the younger sites it was more than 15% of the mean (3.21 ± 0.54 and 3.74 ± 0.52 g C kg−1 H2O at HDF00 and HDF88, respectively) likely due to their young age and less developed root systems. The small variation in WUE at DF49 indicates that it is a relatively conservative quantity. In general, WUE at HDF00 was low in May and September and reached almost a steady value in July and August. Similar values of WUE during the peak growing season at all three sites irrespective of changes in θ between the years suggest strong coupling of CO2 and water vapor exchange processes at these sites.

3.3.4. Response of Pg and Re to Wet and Dry Conditions

[28] In order to examine the responses of Pg and Re to wet and dry conditions, we examined the response of daytime Pg to Q and nighttime Re to Ts at the 2-cm depth during May to September for wet and dry conditions as shown in Figure 10. The parameters of the equations obtained from the regression analysis (equation (1) for Re and equation (2) for Pg) during dry and wet conditions are shown in Table 5. Here ‘wet conditions' refers to periods when relative available water content (θR) in the 0–30 cm soil layer was >0.40 and ‘dry conditions' refers to periods when it was <0.20. θR is defined as (θ − θWP)/(θFC − θWP), where θFC and θWP are θ at field capacity (soil water matric potential = −10 kPa) and the wilting point (matric potential = −1.5 MPa), respectively. θFC and θWP were 0.24 and 0.11 m3 m−3 at HDF88 [Jassal et al., 2008a], 0.21 and 0.06 m3 m−3 at DF49 and 0.24 and 0.11 m3 m−3 at HDF00 [Humphreys et al., 2006], respectively. Periods with θR < 0.40 have been reported to cause noticeable drought stress to plants [Campbell and Norman, 1998; Krishnan et al., 2006].

Figure 10.

Influence of wet (circles) and dry (triangles) conditions on the response of Pg to Q, and Re to Ts at the 2-cm depth. Here wet conditions refer to s θR > 0.40 in the 0–30 cm soil layer, and dry conditions refer to θR < 0.20 in the 0–30 cm soil layer. Symbols in Figure 10 (left) represent binned Pg using bin width of 50 μmol m−2 s−1, and symbols in Figure 10 (right) represent Re using bin widths of 0.2°C. Vertical bars represent standard deviations in each bin. Parameters for the fitted curves (solid lines for θR > 0.40 and dashed lines for θR < 0.20) are in Table 5.

Table 5. Parameters of Equations (1) and (2) Describing the May–September Relationships Between Re and Ts at the 2-cm Depth, and Pg and Q for Dry and Moist Conditionsa
SiteSoil ConditionAB (K−1)Q10R10 (μmol m−2 s−1)r2α (mol CO2 mol−1 photons)Pmax (μmol m−2 s−1)r2
  • a

    See Figure 8. Temperature coefficient of ecosystem respiration, Q10 = exp(10B), and ecosystem respiration at a reference temperature of 10°C, R10 = Q10 expA. All relationships are significant at p < 0.05. Moist conditions refer to periods when θR > 0.4, and dry conditions refer to periods when θR < 0.2. θR is the relative available water content in the 0–30 cm layer (see text).

DF49moist0.1340.5623.86.70.920.0924.470.96
dry0.0671.2341.96.80.600.0920.70.94
 
HDF88moist0.1180.3813.34.80.970.0619.840.96
dry0.0960.2782.63.40.880.088.080.85
 
HDF00moist0.0810.6822.04.50.920.0314.150.90
dry0.0460.8791.63.80.500.067.940.87

[29] During dry conditions both Pg and Re at all the three sites showed a decreased sensitivity to Q and Ts, respectively. The impact of dry conditions on the response of Pg to Q was less pronounced at DF49 than HDF00 and HDF88. Photosynthetic capacity in dry conditions decreased by 18%, 59% and 44% from their values in wet conditions at DF49, HDF88 and HDF00, respectively. Changes in the response of Re to Ts under dry conditions occurred for Ts > 11°C (Figure 10). Even though Q10 values decreased by 50% from those in wet conditions at DF49, R10 increased slightly. R10 at HDF88 and HDF00 decreased by 21% and 20%, respectively, whereas Q10 decreased by 29% and 16% at HDF88 and HDF00, respectively. For the same range of soil temperature (12–14°C) during July–September, the driest part of the growing season, mean Re during dry conditions (as shown in Figure 10) decreased by 15% (10.7 to 9.1 g C m−2 d−1) at DF49, 29% (7.5 to 5.3 g C m−2 d−1) at HDF88 and 26% (6.4 to 4.7 g C m−2 d−1) at HDF00 from values in moist conditions. This analysis suggests that the availability of soil water has a strong influence on Re and Pg at the younger sites, especially at HDF88, consistent with the findings of Jassal et al. [2008a].

4. Discussion

4.1. Effect of Stand Age

[30] Our study indicates that the intersite differences in FNEP, Pg and Re were much greater than the interannual variability. For the near-end-of rotation stand, there was greater control of the interannual variability in FNEP by Re whereas for the younger stands it was largely controlled by Pg (Table 2). The relatively greater control by Pg than Re has mainly been reported for boreal ecosystems [Barr et al., 2007; Krishnan et al., 2006; Zha et al., 2008] mainly because of the role of growing season length in those ecosystems in controlling the C balance. Based on long-term measurements of C exchange in a 75–110 year-old mixed-deciduous Harvard Forest in central Massachusetts, USA, Urbanski et al. [2007] reported that interannual variation in the annual C balance is mainly due to Pg. However, for a southern boreal black spruce forest, Re in years with a warm spring was found to control interannual variation [Krishnan et al., 2008]. Similar reports suggesting the dominant role of Re in controlling the annual C balance were made by Valentini et al. [2000], Bubier et al. [2003], and Dunn et al. [2007] for European forests, peatland in Ontario and a northern boreal black spruce forest, respectively.

[31] The intersite and interannual differences in FNEP, Pg and Re for these three temperate Douglas-fir stands were higher than those reported for a boreal jack pine chronosequence (2–90 years) by Zha et al. [2008]. In our study, HDF88 became C neutral at about an age of 17 years whereas in the jack pine chronosequence, the ecosystems were reported to switch from a C source to sink of C at an age of about 10 years [Zha et al., 2008] and about 11–19 years in a black spruce chronosequence [Bond-Lamberty et al., 2004]. Annual C loss (FNEP = −515 ± 88 g C m−2 yr−1) following harvest at HDF00 (2–6 years) was higher than that reported for a 4-year-old Scots pine stand (FNEP = −386 g C m−2 yr−1) [Kolari et al., 2004], a 2-year-old boreal jack pine stand (FNEP = −137 g C m−2 yr−1) [Zha et al., 2008], a 3-year-old black spruce stand (FNEP = −124 g C m−2 yr−1) [Bergeron et al., 2008].

[32] Harvesting influences the forest C balance components in different ways: because of the changes in aboveground and belowground biomass, it can induce soil warming and enrich soil organic matter that together with dead roots and logging debris can stimulate decomposition and enhance respiratory fluxes. Furthermore, harvesting can cause an increase in θ as a result of reduced E following harvest, thus preventing the occurrence of dry-soil conditions that can limit soil respiration. Furthermore, C uptake by young stands is limited by the low LAI [Humphreys et al., 2005]. Grant et al. [2007] hypothesized that changes in FNEP during the aging of coastal Douglas-fir stands can be explained by changing nutrient uptake caused by different timescales for decomposition of fine nonwoody and coarse woody litter left after harvesting, a drop in canopy water potential with lengthening of the water uptake pathway during bole and branch growth, and increases in the ratio of autotrophic respiration to Pg with phytomass accumulation.

4.2. Effect of Temperature

[33] In our study, the early growing season temperature had a significant positive effect on both annual Pg and Re at all sites. Because of the greater response of Re than Pg to temperature and an almost yearlong growing season, an increase in temperature during any part of the year could result in a decrease in FNEP. A similar response of Pg and Re to temperature and almost no dependence of FNEP on temperature was reported by Reichstein et al. [2007] based on an analysis of the determinants of the C balance using 93 site years of EC data from the EUROFLUX and CARBOEUROPE networks. Chen et al. [2002] reported that temperature had no significant influence on CO2 fluxes at the old-growth Douglas-fir–western hemlock forest stands in the Wind River Valley but it significantly affected CO2 exchange in the younger stands. The variations in FNEP, Pg and Re in Figure 8 suggest that regardless of age, Douglas-fir stands responded similarly to the changes in TSpring and growing season θ. A similar response was reported for the Douglas-fir stands at the Wind River site [Chen et al., 2002, 2004], which has a climate similar to that on the east coast of Vancouver Island, with the majority of C uptake occurring during early spring, followed by C loss in summer and little net C exchange in winter. In southern boreal aspen and black spruce forests an increase in spring temperature was found to increase annual FNEP,Pg and Re [Black et al., 2000, 2005; Krishnan et al., 2006, 2008]. We found that the increase in summer or autumn temperature resulted in a decrease in FNEP and this effect was more noticeable at the younger sites than at DF49. This agrees with the findings of Piao et al. [2008] who analyzed EC CO2 flux measurements from 28 northern ecosystems. Since the magnitude of C loss at the younger sites during the latter part of the year was higher than the near-zero values at DF49, changes in temperature had more effect on annual FNEP at the younger sites than that at DF49.

4.3. Effect of Changes in Water Availability

[34] Our analysis suggests that dry soil conditions during the growing season had a significant effect on the C balance. While Re was controlled by temperature, soil moisture strongly influenced its response to temperature. Soil respiration (Rs) measurements during 2003–2006 at DF49 suggested that more than 60% of Re is due to Rs [Jassal et al., 2007]. Mean Rs during May 2002 to December 2002 was found to be higher at HDF88 than that at DF49 and HDF00 [Humphreys et al., 2006]. The drying of the soil causes a large reduction in both heterotrophic respiration and autotrophic respiration [Jassal et al., 2008a]. Because of the large contribution of Rs, the decrease in Re could be largely due to the reduction in Rs. As shown in Figures 8, 9, and 10, HDF88 responded more to dry conditions than the other two sites. In a recent study using 2005–2006 soil chamber and EC measurement at HDF88, Jassal et al. [2008a] concluded that in extremely dry conditions, Rs is more influenced by θ than Ts. Figure 10 suggests that higher θ during the growing season has a strong effect on both Pg and Re at all sites. The slightly higher response of Re than Pg resulted in a decrease in annual FNEP during the very wet year 2004 and an increase in annual FNEP during the dry years 1998 and 2006 at DF49. This is similar to the earlier report by Krishnan et al. [2008] for the SOBS forest in Saskatchewan. Direct evidence of an increase in FNEP with increasing rainfall during the growing season is reported mainly for grass ecosystems [Flanagan et al., 2002; Ma et al., 2007]. We found that the increase in θ during the growing season is more likely to decrease FNEP at DF49 due to the greater response of Re than Pg to θ, although it was found to be insignificant.

[35] Dry conditions reduce long-term photosynthetic rates as soil water becomes depleted in the root zone [Goulden et al., 1996; Reichstein et al., 2002; Griffis et al., 2004]. The reduction in Pg could occur due to direct effects of soil water stress on carboxylation [Warren et al., 2003] or the decrease in stomatal conductance associated with soil water stress and high values of D [Farquhar and Sharkey, 1982; Baldocchi, 1997; Anthoni et al., 2002; Reichstein et al., 2002; Krishnan et al., 2006, 2008]. Because of the developed root system in mature stands, Pg can be maintained due to the available water deep in the root zone whereas it appears that the impact of a shallower root zone on Pg in the younger stands result in a greater reduction in FNEP (Figure 8 for 2004).

[36] Future climate projections forecast greater variability in precipitation and higher frequency of drought events in the mid and high latitudes [IPCC, 2007]. Earlier reports on drought effects focused on localized short-term and moderate drought [Law et al., 2001; Reichstein et al., 2002; Rambal et al., 2003; Ciais et al., 2005; Goulden et al., 1996; Baldocchi, 1997; Goldstein et al., 2000; Barford et al., 2001; Anthoni et al., 2002; Gu et al., 2006]. Recent reports are on regional-scale and long-term drought, e.g., the 2003 drought in Europe [Ciais et al., 2005; Granier et al., 2007; Reichstein et al., 2007] and the drought in Western North America in 2001–2003, which affected the southern boreal forest [Barr et al., 2007; Krishnan et al., 2006; Kljun et al., 2006]. All these studies suggested strong effects on the ecosystem C balance. Dry summers like those in 1998 and 2006, decreased Re more than Pg resulting in an enhancement in FNEP in mature sites similar to that observed in the first year of drought in 2001 at the SOA forest [Barr et al., 2007; Krishnan et al., 2006; Kljun et al., 2006]. Drought early in the growing season or in spring or during canopy development can more strongly affect Pg than Re resulting in a decreased annual FNEP as reported by Noormets et al. [2007] for a 50-year-old mixed-oak woodland in northern Ohio, USA and by Krishnan et al. [2006] for the third year of drought at the SOA forest. This effect will likely be greater for younger stands than mature stands as seen here at HDF88 and HDF00 in 2004.

[37] The impact of changes in water availability on Pg and Re likely depends on the water holding capacity of the soil, the vertical distribution of the amount of soil organic matter and roots in the soil and the general drought sensitivity of the vegetation [Heimann and Reichstein, 2008]. Warren et al. [2005], in a study of the vertical distribution of soil water in PNW old-growth ponderosa pine, young and old–growth Douglas fir stands, reported that the greater responsiveness of a young Douglas-fir stand to changes in θ was associated with the difference in the vertical distribution of root surface area in the young and old stands, which they attributed to developmental stage and stand density. They reported that total fine root biomass in the upper 1 m were higher for the young stand with 60% of the fine roots located in the upper 20 cm. Studies conducted on young and old-growth ponderosa pine stands [Irvine et al., 2002; Anthoni et al., 2002] also showed higher drought stress in young stands than in old-growth forest because soil moisture was being extracted from greater depths at the old-growth site [Irvine et al., 2002]. In our study, high E in younger stands with less developed root systems, similar in magnitude to that at DF49 during the growing season, could result in greater water uptake leading to water stress earlier in these stands than at DF49 for similar environmental conditions (Figure 2). Furthermore, the greater response of the younger stands (Figure 10), especially HDF88, to changes in θ could be due to the differences in soil physical and chemical characteristics at that site (Table 1). The greater response of HDF88 to θ is in contrast to the finding of McMillan et al. [2008] that younger stands were more susceptible to water or heat stress during a warm dry year. Even with the differences in stand density, LAI, basal area, soil characteristics, elevation, and proximity to the ocean, all sites showed similar responses when exposed to similar climate.

4.4. Effect of ENSO and Climate Change on the Long-Term C Balance

[38] As mentioned above, ENSO plays an important role in year-to-year climate variability and hence potentially influences long-term trends in temperature and precipitation in the PNW [Mote et al., 1999; Mote, 2003]. The globally averaged growth rate of atmospheric CO2 is also tightly correlated to ENSO climate variations [Heimann and Reichstein, 2008] and with a warming planet the frequency of occurrence of stronger El Niño events is likely to increase. In a warmer climate, the change in precipitation distribution in the growing season is an important factor in controlling the C balance of ecosystems. The warmer conditions can extend the length of the effective growing season. As seen in our analysis, the extreme climatic conditions include a warm and mild winter followed by unusually warm spring months and a dry summer during El Niño (e.g., 1998), a wet winter and relatively cool and cloudy conditions in the following spring and summer during La Niña (e.g., 1999), and a warm winter and spring followed by a wet summer and autumn in neutral years (e.g., 2004 and 2005). As seen above, a warm spring and a wet growing season usually has a more positive effect on Pg than Re at all the three sites. However, the increase in winter temperature can result in early enhancement of Re and reduce the number of days with positive FNEP in the early part of the growing season resulting in a reduction in annual FNEP at the near-end-of-rotation site. For the younger sites, this could increase the number of C uptake days in early part of the growing season and enhance annual FNEP. The effect of dry versus wet conditions and their times of occurrence (early versus late in the growing season) are important in this ecosystem as shown in our analysis. While wet soil conditions early in the growing season can favor Pg, Re and FNEP (as in 2005), dry conditions in the latter part can reduce Re more than Pg resulting in an enhancement of FNEP (as in 2006) at near-end-of-rotation site. However, wetter conditions in latter part of the growing season can enhance Re and decrease FNEP (e.g., 2004) at all three sites. The question of how each component (Pg or Re) responds to changes in soil water availability and the impact on FNEP depends on the time of occurrence of dry and wet conditions. Almost parallel variations of Pg and Re and the lack of a statistically significant relationship of FNEP to environmental variables suggest that we need to consider the impact of environmental variables on the C balance components rather than on FNEP when simulating the impact of a changing climate on the C balance. In addition, the increase in annual and autumn temperature had a negative effect on FNEP, especially at the younger sites. The C balance of this ecosystem in the long-term will likely be determined by how the temperature-driven enhancement of C uptake in the early part of the growing season compares with the effect of changes in the soil water conditions or temperature during the latter part of the growing season.

5. Conclusions

[39] 1. Stand-age effects on FNEP, Pg and Re were much higher than that of interannual variability. The near-end of-rotation Douglas-fir stand, DF49, was a C sink (357 ± 51 g C m−2 yr−1) during 1998–2006 while the young plantation, HDF00, was a large C source (−515 ± 88 g C m−2 yr−1) during 2002–2006. The pole-sapling stand, HDF88, was initially (2002–2004) a weak C source (−105 ± 10 g C m−2 yr−1) and in the last 2 years (2005–2006) was a weak sink (21 ± 30 g C m−2 yr−1).

[40] 2. Interannual variability in annual FNEP at DF49 was mainly due to interannual variability in annual Re whereas at the younger sites interannual variability in annual Pg largely determined interannual variability in annual FNEP.

[41] 3. Early growing season (spring) temperature and May–September (summer) soil water availability were the main factors determining the interannual variability in Pg and Re of the three Douglas–fir stands. The linear relationship between annual FNEP and the above variables was not significant because of the similar changes in Pg and Re. Even though not statistically significant, the decrease of FNEP with increasing spring, summer, autumn temperatures and growing season water content were due to a slightly greater response of Re than Pg to the above climatic variables.

[42] 4. The impact of extremely dry conditions on Pg and Re at the younger sites, especially at HDF88, was greater than at the near-end-of-rotation site, DF49. Drought conditions during the latter part of the growing season had a positive effect on FNEP because of the greater reduction in Re than Pg in response to the decrease in θ.

[43] 5. Empirical relationships using controlling environmental variables were effective in explaining more than 85% of the interannual variability in Pg, Re and FNEP at DF49; however, less than 40% could be explained using similar empirical relationships for the younger sites.

Acknowledgments

[44] This research was funded by the Natural Sciences and Engineering Research Council of Canada (NSERC), the Canadian Foundation for Climate and Atmospheric Science (CFCAS), and BIOCAP Canada Foundation through the Fluxnet Canada Research network (now the Canadian Carbon Program), and by a NSERC Discovery (operating) grant to T.A.B. We are grateful for assistance in laboratory and field from Andrew Sauter, Rick Ketler, Shawn O'Neill, Dominic Lessard, and Andrew Hum, and also for the data quality control by Elyn Humphreys, Christopher Schwalm, Nick Grant, Kai Morgenstern, Christian Bruemmer, and Adrian Leitch during the course of this research. Part of this research was performed while one of the authors (P.K.) held a National Research Council Associateship Award at NOAA/ATDD.

Ancillary