4.2. Seasonal and Interannual Behavior of NEE and SR
 Both the measured and modeled NEE data indicate moderate seasonal variability in this jack pine forest during the growing season (Table 1). However, our estimated values of NEE are a weak point of this study since we lost data during July 2003, and all the gap-filled values in the data set were likely sensitive to the postprocessing scheme [Falge et al., 2001]. Nevertheless, seasonal variation might be attributable to the behavior of the ecosystem during the spring recovery of jack pine photosynthesis at the onset of the growing season and the entrance into winter dormancy in the fall. Relatively high rates of early season C uptake, as indicated by high values of Pmax (−1.19 μmol CO2 m−2 s−1), were observed in the early spring. Pmax then declined in the late spring (−0.66 μmol CO2 m−2 s−1), rose again in summer (−0.79 μmol CO2 m−2 s−1) and fell off again in the fall (−0.61 μmol CO2 m−2 s−1; Table 2).
 Lower values of Pmax in late spring compared to early spring (Table 2) may have been caused by either nutrient or water limitations. Soil water matric potential was generally high (∼−0.5 bar) during the later spring months, particularly in 2002 when most of the heavier rains occurred in midsummer (Figure 3c). This increase in soil water matric potential may have contributed to water stress and a reduction in photosynthesis during the late spring in this ecosystem. It is also possible that while the trees have considerable photosynthetic capacity in the early spring, before bud break and the development of new shoots, they retranslocate nitrogen to the developing needles during the late spring, resulting in nutrient limitations for a period of time. We did observe appreciable needle yellowing in the late spring, a characteristic of water-stressed and/or nutrient-limited trees.
 It is also possible that during the early spring in April and May, bud break and full leaf out in the dense understory of ferns, blueberry bushes, and grasses would contribute to a steep rise in photosynthetic capacity and a greater net C uptake in this ecosystem. Although we did not measure LAI in early spring, we did note first bud break in both the deciduous understory and the jack pine overstory occurred about three weeks before full leaf out of the understory. The full leaf out of the understory occurred earliest in the warm year of 2002 (15 May), and latest in 2003 (23 May). Correspondingly, the cumulative net C uptake from 1 May to 15 May was greater in 2002 (∼0.11 g C m−2 d−1, or 1.7 g C m−2 over the 15 day period) than in 2003 (∼0.08 g C m−2 d−1, or 1.2 g C m−2 over the 15 day period), with cumulative net C uptake from April to May being 7.2 g C m−2 in 2002 and 6.4 g C m−2 in 2003. Future studies in this ecosystem would benefit from a comprehensive partitioning of NEE into overstory and understory fluxes.
 The effect of subzero air temperatures in the spring and early fall had a much different impact on net C uptake than they did in the late fall. The last frosts in the spring, which occurred during periods of long photoperiods in late May or early June, seemingly did not decrease the net C uptake during this time (Figures 3a and 3b and Table 2). While the first frost occurred earlier in 2003 (4 September; daily minimum of −0.6°C) than in 2001 (25 September; daily minimum of −2.9°C) or 2002 (5 October; daily minimum of −8.2°C), this early fall frost did not strongly decrease C uptake in the following weeks (Figure 3a). In fact, net C uptake was higher in September and October of 2003 than in September and October of 2001 and 2002 (Table 1), indicating that the plants easily recovered from early frosts if temperatures were not substantially below 0°C, but did not recover from colder frosts later in the season [Havranek and Tranquillini, 1995; Lamontagne et al., 1998; Monson et al., 2002]. That is, during the late fall there appear to be temperature thresholds that initiate a decline in stomatal conductance and gas exchange rates, effectively ending the growing season. These thresholds may also be related to the decreases in light intensity and photoperiod at this time of year [Havranek and Tranquillini, 1995].
 The overall low temperatures during September and October of 2002 that resulted in little net C uptake also resulted in generally low values of SR, except for a brief period in late September 2002 when soil temperatures and SR increased, and NEE decreased substantially (Figures 3b, 5, and 6). In 2001, the warmest of the three measurement years, C losses from soil respiration were high from June to August (Figure 6a), and consequently, total net carbon uptake was reduced below that observed during 2002 and 2003 (Table 1). Thus it appears that soil respiratory losses had a large impact on net C uptake in this ecosystem. However, to gain a more complete understanding of the influence of soil respiration in this system, it would be useful to obtain predictions of total soil respiration based not just on temperature, but also on moisture since moisture appeared to have an influence on soil respiration rates (Table 4).
Table 4. Parameters and R2 Values of the Exponential Models Fitted to the Soil Respiration Data Based on Soil Temperature and Soil Moisturea
|Year(s)||Modelb||Parameter (± Standard Error)||R2|
|2001||Ts||0.759 (0.257)||0.102 (0.017)||-||-||0.76|
|2001||Ts * Ms||0.330 (2.036)||0.088 (0.057)||−0.683 (5.885)||0.166 (0.841)||0.78|
|2002||Ts||0.868 (0.351)||0.085 (0.020)||-||-||0.68|
|2002||Ts * Ms||0.014 (0.0018)||0.098 (0.018)||−0.037 (0.088)||0.728 (0.284)||0.87|
|2003||Ts||1.554 (0.0493)||0.066 (0.010)||-||-||0.77|
|2003||Ts * Ms||0.091 (0.0198)||0.045 (0.013)||−0.504 (3.095)||0.613 (0.270)||0.88|
|All||Ts||1.114 (0.0313)||0.079 (0.009)||-||-||0.69|
|All||Ts * Ms||0.181 (0.0181)||0.075 (0.028)||−0.075 (0.593)||0.210 (0.431)||0.75|
 The SR rates we measured were comparable to those of others taken in coniferous ecosystems of similar age. For example, estimates of SR in jack pine forests during one growing season, with a length that was similar to the one in this study, were 415.2 g C m−2 at an 8-year-old ecosystem and 378 g C m−2 at a 20-year forest in Saskatchewan, Canada [Striegl and Wickland, 1998]. A decline in SR between the 8- and 20-year forests may be caused by a decrease in the amount of material available for decomposition as the large amounts of microbial substrate due to the previous disturbance are exhausted. In particular, the forest in this study contained large piles of slash left behind from logging practices that were probably a major source of substrate for heterotrophs, which tend to favor the less resilient organic mater fractions [Alexander, 1977], and a reason for relatively high SR rates. Older jack pine forests growing on outwash sands with low soil C content have been shown to exhibit lower SR (e.g., 300 g C m−2 over the growing season, with a length that was similar to the one in this study, at a 60–75-year-old jack pine ecosystem [Striegl and Wickland, 1998]) than younger forests: a finding that is generally attributable to an absence of large pools of labile litter that are associated with disturbance events.
 Nevertheless, the day-to-day activities of the soil microorganisms are highly temperature dependent and even with large amounts of labile substrate, their activities decline during low temperatures. Consequently, SR contributed less to NEE during the low temperatures in the spring (Figure 6). When the soils remained cool, (≈1°–3°C), SR stabilized at around 2.0 g C m−2 day−1, but the overall carbon balance of the ecosystem still fluctuated between ±0.05 g C m−2 day−1 in concert with fluctuations in air temperatures (Figures 3a, 3b, and 6). From roughly mid-October to November, SR and NEE were poorly correlated. We hypothesize that at this point, the soils were still warm and the microbes still responsive, but the trees began to enter winter dormancy (Figure 6).
4.3. Annual NEE and Nongrowing Season C Losses
 At the annual scale, it is possible that the 12–14-year-old ecosystem in this study has recently switched from a source to slight sink of CO2. All the same, the weak growing season sink strength measured in this young jack pine forest is likely an overestimation of the annual C uptake. For instance, Griffis et al.  found that nongrowing season C losses accounted for 46% of the summertime NEE in an old jack pine ecosystem in Saskatchewan, Canada.
 Moreover, although we did not consistently measure SR in the winter, we did find that even during periods of near freezing soil temperatures some carbon efflux was occurring, the sum of which could amount to significant carbon losses at the site. Empirically based studies of winter SR have measured highly temperature-dependent rates between 40 and 132 g C m−2, with soil moisture having little to no effect [McDowell et al., 2000; Winston et al., 1997]. Projections of climate change forecast warmer winters within the latitude of this forest. Such warming could elicit greater respiratory losses from the soil during the nongrowing season, and consequently affect the C balance of these young jack pine forests. Alternatively, warming may also lengthen the growing season and thereby increase the cumulative NEE in this forest [Myneni et al., 1997].
4.4. Comparison With Other Direct Measurements of Ecosystem C Flux
 Although we know of no studies of direct measurements of NEE in younger jack pine ecosystems over an entire growing season, Amiro  used eddy covariance techniques to measure NEE in a 1-year-old burned jack pine ecosystem for nine days in July 1998, during the height of the growing season. This ecosystem was a consistent C source at roughly 0.8 g C m−2 day−1. Independently, Pypker and Fredeen  measured C fluxes in a 5–6-year-old subboreal clear-cut composed of white spruce and lodgepole pine, with a net C loss of 1.0 to 1.4 Mg C ha−1 during the growing season (Figure 7).
Figure 7. Summary of NEE during the growing season (Mg C ha−1 growing season−1) for three comparable pine ecosystems of various age classes. The 5- to 6-year, recent clear-cut is located in British Columbia, Canada (54°N) [Pypker and Fredeen, 2002]. The 30- to 32-year forest is located in Manitoba, Canada (56°N) [Joiner et al., 1999], and the 65- to 71-year forest is in Saskatchewan, Canada (53°N) [Baldocchi et al., 1997; Griffis et al., 2003]. The solid line is drawn by hand to indicate a general trend in ecosystem carbon flux across the age classes.
Download figure to PowerPoint
 Mature jack pine ecosystems are likely to sequester more C than the young ecosystem described in the current study with variability in these ecosystems being generally attributable to prior land use and latitude. Prior land use differences may be related to soil type, stand density, reseeding, and the amount of slash left to decay following harvest [Pregitzer and Euskirchen, 2004]. It is also possible that jack pine growing in the harsher climates of the northern limit of the species range are less productive than those in the southern limits. Using a similar measurement period as the one reported in this study, Joiner et al.  reported a net uptake of 2.1 and 2.7 Mg C ha−1 growing season−1 for a 30–32-year-old jack pine ecosystem in Manitoba, Canada. These estimates and those from this study suggest that jack pine ecosystems switch from acting as a source to sink of C at around 10 to 20 years (Figure 7). Meanwhile, older (e.g., >50 year) jack pine ecosystems may sequester less carbon than mature jack pine ecosystems. For example, a jack pine forest measured during two growing seasons at 65 and 71 years took up −0.47 and −0.36 Mg C ha−1 growing season−1, respectively [Baldocchi et al., 1997; Griffis et al., 2003] (Figure 7).