Net ecosystem productivity of boreal jack pine stands regenerating from clearcutting under current and future climates

Energy exchange

Canopy energy exchange in ecosys is calculated from an hourly two-stage convergence solution for the transfer of water and heat through a multilayered multipopulation soil–root–canopy system. The first stage of this solution requires convergence to a value of canopy temperature T_c for each plant population at which the first-order closure of the canopy energy balance (net radiation, sensible heat flux, latent heat flux and change in heat storage) is achieved [Eqs. (A1–A15) in Grant et al., 1999b]. These fluxes are controlled by aerodynamic (g_a) and stomatal (g_c) conductances. Two controlling mechanisms are postulated for g_c:

Water relations

After convergence for T_c is achieved, the difference between canopy transpiration E from the energy balance and total water uptake U from all rooted layers in the soil is tested against the difference between canopy water content from the previous hour and that from the current hour (Grant et al., 1999b, 2006b). This difference is minimized by adjusting ψ_c which determines transpiration through ψ_T, and hence g_c [Eqs. (A24–A25) in Grant et al., 1999b; Eqs. (B1–B4) in Grant et al., 2006b], and uptake through the potential differences with root water potential ψ_r, and soil water potential ψ_s across soil–root radial hydraulic resistance Ω_s and root–canopy radial and axial hydraulic resistances Ω_r and Ω_a in each rooted soil layer [Eqs. (A32–A37) in Grant et al., 1999b; Eqs. (B5–B12) in Grant et al., 2006b).

GPP

After successful convergence for T_c and ψ_c, leaf carboxylation rates are adjusted from those calculated under full ψ_T to those under ambient ψ_T. This adjustment is required by the decrease in g_c from its maximum value (calculated in ‘Energy exchange’) to that at ambient ψ_T (calculated in ‘Water relations’). The adjustment is achieved through a convergence solution for C_i at which the diffusion rate of gaseous CO₂ between C_a and C_i through g_l [Eqs. (A48–A53) in Grant et al., 1999b] equals the carboxylation rate of aqueous CO₂ at C_m (described in ‘Energy exchange’) [Eqs. (A38–A47) in Grant et al., 1999b; Eqs. (C1–C11) in Grant et al., 2006b]. The CO₂ fixation rate of each leaf surface at convergence is added to arrive at a value for GPP by each tiller (or branch) of each plant population (i.e. species or cohort) in the model. The CO₂ fixation product is stored in a nonstructural C pool for translocation to growing organs.

Nutrient uptake and translocation

Uptake of nitrogen (N) and phosphorous (P) is calculated by solving for solution [NH₄⁺], [NO₃⁻] and [H₂PO₄⁻] at root and mycorrhizal surfaces at which radial transport by mass flow and diffusion from the soil solution to these surfaces equals active uptake by the surfaces [Eq. (A36) in Grant, 1998; Grant et al., 2006b]. Solution N and P concentrations are controlled by precipitation, adsorption and ion pairing reactions [Eqs. (A1–A55) in Grant & Heaney, 1997; Grant et al., 2004a], vertical and horizontal solute transport (Grant & Heaney, 1997) and microbial activity including mineralization, nitrification, denitrification, volatilization and N₂ fixation described in ‘Heterotrophic respiration’ below. Products of N and P uptake are stored in nonstructural pools for translocation to growing organs.

Nonstructural C is coupled with nonstructural N and P to drive growth of shoots, roots and mycorrhizae according to organ-specific C : N : P ratios. Coupling requires the transfer of nonstructural C, N and P among shoots, roots and mycorrhizae [Eqs. (18–24) in Grant, 1998). These transfers are driven by concentration differences in nonstructural C, N and P (Brugge & Thornley, 1985).

Low ratios of nonstructural N or P to nonstructural C in branch or tiller pools indicate excess CO₂ fixation with respect to N or P uptake for shoot growth. Such ratios in the model have two effects:

Autotrophic respiration

The nonstructural C pool in each branch is also used for autotrophic respiration R_a [Eqs. (A26–A31) in Grant et al., 1999b; Eqs. (C12–C16) in Grant et al., 2006b]. R_a is first used to meet requirements for maintenance respiration R_m then any excess is used for growth respiration R_g to drive biosynthesis according to organ-specific growth yields. Allocation of C to aboveground biosynthesis of foliage, twigs, branches, boles and reproductive material is directed by phenology-dependent partitioning coefficients (e.g. Weinstein et al., 1991) driven by temperature and photoperiod. Allocation of C to biosynthesis of roots is governed by a functional equilibrium between shoot and root nonstructural pools. Low nonstructural C or rapid R_m may cause R_a to become less than R_m, in which case the shortfall is made up through respiration of remobilizable protein C withdrawn from foliar and twig C at each node. The remaining structural C at the node is dropped from the branch and added to the soil surface as litter [Eqs. (C17–C19) in Grant et al., 2006b]. Environmental constraints such as nutrient, heat or water stress that lower nonstructural C and hence R_a, or that raise R_m, will therefore hasten litterfall from the plant.

Heterotrophic respiration

Organic transformations in ecosys occur in five organic matter – microbe complexes i [coarse woody litter, fine nonwoody litter, animal manure, particulate organic matter (POM), and humus] in each soil layer. Each complex consists of five organic states: solid organic matter, dissolved organic matter, sorbed organic matter, microbial biomass, and microbial residues, among which C, N and P are transformed. Coarse woody and fine nonwoody litterfall are partitioned into carbohydrate, protein, cellulose and lignin components (Trofymow et al., 1995), each of which is of differing vulnerability to hydrolysis by heterotrophic decomposers. Hydrolysis rates are controlled by soil temperature T_s through an Arrhenius function [Eq. (A5) in Grant et al., 2006b], and by soil water content θ through its effect on aqueous microbial concentrations and thereby on specific activity [Eqs. (A3–A4) in Grant et al., 2006b], in surface residue and in a spatially resolved soil profile.

Heterotrophic respiration R_h in each soil layer is the sum of that by all heterotrophic microbial populations n in each substrate-microbe complex [Eq. (A11) in Grant et al., 2006b]. Total R_h for all soil layers drives CO₂ emission from the soil surface through volatilization and diffusion in gaseous and aqueous phases. Heterotrophic oxidation rates may be constrained by microbial nutrient concentrations [Eq. (A12)], and by θ [Eq. (A3)], T_s [Eq. (A13, A18)], DOC [Eq. (A13)] and O₂ [Eqs. (A14–A16), all in Grant et al., 2006b]. Further details about the calculation of R_h may be found in Grant (2001), Grant et al. (1993a, b, 2006b).

Model evaluation

Ecosys was initialized with the soil and topographic properties given for SOJP in Tables 1 and 2, and with above- and belowground residues estimated to correspond to those left after a stand-replacing fire (e.g. Grant et al., 2006a). The rooting zone consisted of the soil horizons described in Table 2, with subdivision into nine soil layers to increase spatial resolution. Two additional soil layers with properties the same as those of the lowest rooted layer were modelled below the rooting zone. The lower soil boundary was set so that the water content in the lowest soil layer was maintained near field capacity. These additional layers enabled water transfer modelled in the rooting zone to be largely independent of assumptions about drainage through the lower boundary.

Ecosys was then run for three 126-year cycles, each of which ended with clearcutting in which 0.1, 0.1 and 0.6 of foliar, nonfoliar nonwoody, and coarse woody aboveground phytomass respectively above a height of 0.5 m was removed on 30 June. This cycle length was selected to allow modelling of forest productivity during regeneration, maturation and postmaturation growth stages. To simulate disturbance effects on surface litter during clearcutting, a small fraction (0.05) of surface woody and nonwoody litter was incorporated into the soil LFH horizon after simulated clearcutting. All fine nonwoody and coarse woody root phytomass in the model at clearcutting was added as litter to the soil layers in which they had been growing.

Jack pine was germinated at a density of 2 m⁻² on 1 April of the year after initialization and each clearcut, and nonvascular vegetation (e.g. moss as in Grant et al., 2001a) was germinated at a density of 100 m⁻² to simulate the surface vegetation that competed with jack pine for nutrients and water at most of the chronosequence sites (Table 1). These two plant populations used common attributes for CO₂ fixation and C allocation by C₃ perennial species (Appendices B and C of Grant et al., 2006b). In ecosys, these attributes include two key inputs for plant thermal adaptation: (1) one that defines thermal requirements to complete growth stages within the annual phenological cycle that controls aboveground C allocation among foliage, stem and reproductive material (e.g. Weinstein et al., 1991) and (2) one that determines values of temperature sensitivity functions for photosynthesis (Arrhenius) and maintenance respiration (exponential). These inputs were set to values for boreal ecotypes used in earlier testing of CO₂ and energy exchange over boreal forests (Grant et al., 1999a, 2001a, b). The rooting system of jack pine was modelled with large primary root axes (Grant, 1998), creating a strong primary root sink that enabled roots to reach the lowest horizon boundary at a depth of 3 m (Table 2) about 20 years after seeding. Both populations regenerated without further management in the model, except for annual self-thinnings of jack pine that allowed the model to simulate a gradual decline in tree density from 2.0 m⁻² at germination to 0.1 m⁻² after 50 years, consistent with declines observed in jack pine stands. All self-thinned phytomass was added to above- and belowground residues.

Each logging cycle was run under repeated 11-year sequences (1994–2004 in Table 3) of hourly averaged weather data (shortwave radiation, T_a, humidity, wind speed and precipitation). Concentrations of NH₄⁺ and NO₃⁻ in precipitation were set to 0.15 and 0.60 g N m⁻³, respectively to give average wet deposition rates of ca. 0.3 g N m⁻² yr⁻¹, and atmospheric NH₃ concentration was set at 0.0025 μmol mol⁻¹ to give average dry deposition rates of ca. 0.03 g N m⁻² yr⁻¹. Hourly CO₂ fluxes and daily and annual NEP modelled from site weather data were compared with measured half-hourly fluxes and with total daily and annual fluxes derived through gap filling (Barr et al., 2002). These comparisons were made in years 2–3 after the first clearcut, which corresponded to 2003 and 2004 at HJP02, in years 8–11 after the first clearcut, which corresponded to 2001–2004 at HJP94, in year 29 after the first clearcut, which corresponded to 2004 at HJP75, and in years 78–81 after model initialization following fire, which corresponded to 2001–2004 at SOJP. The weather data under which the model was run were recorded 10 m above the SOJP forest canopy, but were replaced by those recorded above the HJP02, HJP94 and HJP75 sites during years when model results were compared with EC fluxes recorded over the younger stands.

Model predictions

Climate represented by the hourly averaged weather data used in the Model Evaluation is likely to change during the next century. To examine the effects of climate change on jack pine productivity, the model run described above was also conducted under C_a, T_a and precipitation events that were incremented hourly after the first clearcut at rates corresponding to those in the IPCC SRES A1B scenario (IPCC, 2001), and in the average projections of the climate change scenarios in Berthelot et al. (2005) (Table 4). This model run was intended to demonstrate model sensitivity to climate change. A fuller examination of climate change effects on jack pine productivity should include a wider range of climate and disturbance scenarios.

There is concern that current tree varieties may not be well adapted to higher temperatures expected during climate change, and so should be replaced at planting by varieties adapted to warmer climates. To examine possible impacts of adaptation on productivity, the climate change run was also conducted with the attributes for thermal adaptation of the boreal jack pine ecotype replaced after the first clearcut by those of a cold temperate ecotype. This changed thermal adaptation represented an ecotype intermediate between the boreal ecotype and a cool temperate ecotype used in Grant et al. (2006b) to model CO₂ exchange and growth over a Douglas-fir forest in British Columbia. This change involved (1) increasing the thermal requirements to complete growth stages so that timing of these stages, and of associated changes in aboveground C allocation, during climate change coincided approximately with those modelled under current climate, and (2) displacing temperature functions for photosynthesis (Berry & Björkman, 1980) and respiration (Gifford, 1995) upwards by 0.75 °C, thereby reducing their values at lower temperatures, but increasing them at higher temperatures.

Diurnal CO₂ exchange

Differences in precipitation recorded at SOJP during 2001–2004 caused θ measured and modelled in the LFH and upper part of the mineral soil (0–15 cm) to remain near field capacity for most of 2001, 2002 and 2004 except during brief drying cycles, but declined to wilting point in 2003 from mid-July onwards (Fig. 1). Modelled θ remained within 0.02 m³ m⁻³ of measured values except when the soil was frozen, at which time modelled θ, which excluded ice, was lower. Both modelled and measured θ declined rapidly between rainfall events during all years because of the low water holding capacity and high hydraulic conductivity of the sandy soil at the jack pine sites (Table 2).

These soil attributes could cause productivity at these sites to be affected by hydraulic constraints if rainfall events are infrequent. These constraints were apparent during soil drying under high T_a and D from DOY 228–234 (16–22 August) 2003 (Fig. 2). During this period, rapid midday declines in CO₂ influxes were measured and modelled at HJP94 and SOJP (Fig. 3b and c), especially when T_a > 25 °C and D > 2 kPa on DOY 228, 230, 231 and 234. Hydraulic constraints on CO₂ fixation were alleviated by rainfall early in DOY 235 (Fig. 2), so that CO₂ influxes at HJP94 and SOJP were greater from DOY 235 to 237 (23–25 August) 2003 (Fig. 3). CO₂ influxes modelled and measured at HJP02 remained small and unaffected by rainfall (Fig. 3a).

Soil drying slowed CO₂ fixation and hence R_a (‘Autotrophic respiration’), and slowed decomposition and R_h (‘Heterotrophic respiration’). Ecosystem CO₂ effluxes declined during soil drying at SOJP (e.g. DOY 228 vs. 234 in Fig. 3c). Soil CO₂ effluxes also declined with θ during DOY 228–234 in 2003 (Fig. 4). These effluxes were partially offset by CO₂ uptake from surface vegetation during daytimes, although uptake also declined with soil drying. Soil CO₂ effluxes rose briefly following rainfall on DOY 235, as did CO₂ uptake, but declined thereafter as soil drying continued.

Hydraulic constraints on CO₂ fixation were further studied during soil drying under changing T_a and D from DOY 177 to 183 (25 June–1 July) 2004 (Fig. 5). Hydraulic constraints on CO₂ fixation were apparent in the midafternoon declines in CO₂ influxes measured and modelled during DOY 179 and 180 at HJP94, HJP75 and SOJP when T_a > 25 °C and D > 2 kPa (Fig. 6b–d). CO₂ influxes at HJP02 remained low and unaffected by T_a and D (Fig. 6a).

Some of the variation in EC measurements of CO₂ fluxes was not explained by the model. Root mean squares for differences (RMSD) between measured and modelled CO₂ fluxes during all of 2003 and 2004 rose with the magnitude of CO₂ fluxes from 0.65 μmol m⁻² s⁻¹ at HJP02 to 1.40 μmol m⁻² s⁻¹ at HJP94 to 1.75 μmol m⁻² s⁻¹ at HJP75 and SOJP. Richardson et al. (2006) estimated that random errors of EC measurements at a range of forest sites rose from ca. 0.6 μmol m⁻² s⁻¹ for CO₂ fluxes of 0 μmol m⁻² s⁻¹ to ca. 3.6 μmol m⁻² s⁻¹ for CO₂ fluxes of +10 (influxes) or −5 (effluxes) μmol m⁻² s⁻¹, the largest fluxes measured in this study. The RMSD at HJP02, where CO₂ fluxes were small, was comparable with random errors for near zero CO₂ fluxes estimated by Richardson et al. (2006). The RMSD at HJP75 and SOJP was comparable with the middle of the range in random errors estimated for the CO₂ fluxes measured in this study, and similar to an estimate of 2.4 and 1.5 μmol m⁻² s⁻¹ for random errors in annual CO₂ influxes and effluxes respectively over a boreal transition spruce/hemlock forest (Richardson et al., 2006). These comparisons indicate that much of the variation in measured CO₂ fluxes not explained by the model (e.g. 3, 6) may be attributed to uncertainty in EC measurements. This attribution is also suggested by the narrow range of 1.8–1.9 μmol m⁻² s⁻¹ in RMSDs reported by several other ecosystem models during a model intercomparison at SOJP (Grant et al., 2005).

Seasonal NEP

Daily NEP measured and modelled at HJP02 during 2004 remained below zero (net C source) during the entire year (Fig. 7b) because GPP remained smaller than R_a+R_h (Fig. 6a). Daily NEP measured and modelled at HJP94 during the same year was slightly above zero (net C sink) during late May and June, but remained near or slightly below zero during the rest of the growing season (Fig. 7c). NEP measured and modelled at HJP75 and SOJP during 2004 rose above zero during late April, reached maximum values during June, but declined gradually thereafter to zero by mid-October (Fig. 7d and e). Larger NEP at SOJP and HJP75 vs. HJP94 and HJP02 was caused by larger CO₂ influxes vs. effluxes (3, 6) modelled and measured at the older sites. Declines in NEP modelled after June at all sites were attributed to rising CO₂ effluxes from warming soils that increasingly offset CO₂ influxes as the growing season progressed. NEP at all sites was adversely affected by warm weather between DOY 195 and 205 (Fig. 7).

Annual NEP

During the first 2 years after clearcutting, NPP in the model was small in comparison with R_h so that annual NEP (= sum of all CO₂ influxes and effluxes) averaged −180 g C m⁻² yr⁻¹ (net C source) (Table 5; Fig. 8a). GPP in the model was similar to that derived from gap-filled EC fluxes during these 2 years, but R_e in the model rose with microbial growth on litter left after clearcutting more than did R_e derived from EC. Howard et al. (2004) calculated NPP and R_h of 90 and 270 g C m⁻² yr⁻¹ respectively during the first year after logging in this jack pine chronosequence.

Subsequent rises in GPP and declines in R_h caused NEP in the model to increase to ca. −50 g C m⁻² yr⁻¹ (net C source) between 7 and 10 years after clearcutting (Table 5; Fig. 8a). Howard et al. (2004) also calculated rises in NPP and declines in R_h to 130 and 170 g C m⁻² yr⁻¹ respectively after 5 years, and to 270 and 210 g C m⁻² yr⁻¹ respectively after 10 years. They estimated that boreal jack pine stands likely remained C sources for 7 years after clearcutting before becoming C sinks. The model run in this study indicated that jack pine stands would remain C sources for 15–20 years after clearcutting (Fig. 8a).

GPP in the model began to rise more rapidly after 10 years while R_h rose little, so that NEP approached mature values of 50 g C m⁻² yr⁻¹ after 29 years (Table 5; Fig. 8a). GPP rose more slowly thereafter, so that NEP remained at average values of 53±29 g C m⁻² yr⁻¹ interannual variability under 2000–2004 weather. NEP declined very gradually thereafter (Fig. 8a) from slight declines in stem axial conductance caused by slowly increasing bole height (Grant et al., 2006b), as observed experimentally in pine by Mencuccini & Grace (1996).

R _a rose with respect to GPP during modelled forest growth, so that ratios of NPP : GPP declined from 0.50 to 0.55 during the first decade of growth to 0.40–0.45 after 80 years. These ratios were within the range of those derived in diverse forest ecosystems by Waring et al. (1998), but were larger than one of 0.31 derived from chamber measurements at SOJP in 1994 by Ryan et al. (1997). In the mature jack pine, 0.40–0.45 of modelled NPP was attributed to trees above ground (approximated by Δwood C+ aboveground litterfall C in Table 5, both of which were corroborated by measurements) and the remainder was attributed to surface vegetation or belowground. Gower et al. (1997) estimated from biometric measurements that 0.46 of NPP was allocated below ground at SOJP during 1994, but these measurements did not account for exudation, which contributed to belowground NPP in the model. Aboveground NPP in the model drove annual wood growth (Table 5) that was consistent with mean annual increments derived from wood growth curves for boreal jack pine at medium-productivity sites (Alberta Forest Service, 1985). These increments reached peak values of ca. 75 g C m⁻² yr⁻¹ 60 years after stand establishment, and declined gradually thereafter.

Most modelled NEP was stored in wood C (Fig. 8b) and soil C (Fig. 8c) because losses of C as DOC (∼0.2 g C m⁻² yr⁻¹) and CH₄ (0 g C m⁻² yr⁻¹) were negligible. Wood C growth modeled after the first clearcut was similar to that estimated from biometric measurements or from allometric calculations used in wood inventory assessments of jack pine sites with site indices similar to that at SOJP (Fig. 8b). Soil+litter C declined during the first 20 years after clearcutting while litter left from the previous forest decomposed (Fig. 8c), during which time NEP was negative (Fig. 8a; Table 5). Soil+litter C stabilized ca. 50 years after clearcutting at values that rose by ca. 1.6 kg C m⁻² in successive logging cycles or 13 g C m⁻² yr⁻¹ (Fig. 8c). This rise was partly attributed to a small increase in ecosystem N stocks that slightly raised ecosystem productivity during successive logging cycles because N removals in wood during logging events were smaller than N inputs from atmospheric deposition and biological fixation during the intervening 126 years of growth in the absence of fire. This increase in modelled N stocks would be sensitive to logging frequency, and would be adversely affected by fire. Annualized NEP, C removal by clearcutting, and net biome productivity (NBP =∑NEP−C removal) averaged 47, 33 and 14 g C m⁻² yr⁻¹ respectively during the 126-year logging cycle.

Changes in NEP and C stocks with climate change

After 70 years of climate change, rising T_a (average+1.9 °C from Table 4) hastened thawing and evapotranspiration (‘Energy exchange’), so that θ rose and declined earlier than under current climate (Fig. 9a). Earlier declines in θ combined with higher C_a (Table 4) and D (Fig. 9b) to lower g_c during soil drying, especially during midafternoons under higher D (Fig. 9c). Lower g_c offset the effects of higher C_a on GPP so that CO₂ influxes modelled under climate change declined with respect to those under current climate as D rose (Fig. 9d). At the same time, soil warming raised CO₂ effluxes above those modelled under current climate, so that higher D between precipitation events had an adverse impact on NEP during climate change.

Annual NEP of boreal jack pine modelled under climate change gradually rose above that under current climate over time after clearcutting, especially during years with larger NEP (Fig. 10a). However, the adverse effects of rising T_a and D on GPP vs. R_e (e.g. Fig. 9d) hastened declines in NEP more than 75 years after clearcutting. These declines led to the dieback of boreal jack pine in the model after 100 years of climate change as GPP eventually failed to offset R_a, depleting nonstructural C and accelerating litterfall (‘Autotrophic respiration’). At this time jack pine boles were transferred to standing dead, the remains of which were harvested at the end of the modelled logging cycle, and jack pine was reseeded. However, substantial losses of ecosystem C were modelled during jack pine regeneration following dieback (Fig. 10a).

Replacing the boreal pine ecotype with the cold temperate pine ecotype in the model lowered NEP during the first 50 years of climate change, but raised it thereafter (Fig. 10a). This later gain occurred as the temperature sensitivity of GPP in the temperate ecotype became better adapted to rising T_a, enabling GPP to offset R_a and so avoiding dieback. As productivity of the temperate pine rose, productivity of the understory declined, causing further gains in modelled NEP as understory nutrients were released. Annualized NEP, C removal by clearcutting, and net biome productivity (NBP =∑NEP−C removal) averaged 81, 56 and 25 g C m⁻² yr⁻¹ respectively during the 126-year logging cycle under climate change.

Climate change accelerated wood growth of boreal jack pine until dieback after 100 years (Fig. 10b), following which standing dead wood contributed to rises in soil+litter C (Fig. 10c). Wood growth of the cold temperate pine ecotype was slower than that of the boreal pine ecotype during the first 75 years of climate change, but surpassed it thereafter. Gains in wood C under climate change were partially offset by losses in soil+litter C (Fig. 10c).

Hydraulic constraints on forest NEP

NEP of jack pine stands was affected at diurnal time scales by hydraulic constraints imposed by declines in θ and rises in D between rainfall events (3, 6) on this rapidly drained soil with low water holding capacity (Table 2). In the model, declines in θ forced declines in ψ_s, rises in Ω_s and Ω_a, and hence declines in ψ_c to meet transpirational demands (‘Water relations’), particularly when these demands were raised by large D (e.g. DOY 179–180 in Fig. 6). Values of ψ_c modeled in conifers were more sensitive to θ and D than were those in other plant functional types because of the larger root and stem axial resistances used to calculate Ω_a in conifers (Grant et al., 2006b). Strong sensitivity of g_c to D in conifers has often been observed in other studies (e.g. Saugier et al., 1997; Grelle et al., 1999; Ohta et al., 2001; Chen et al., 2002), and may explain the positive correlation between δ¹³C and potential evapotranspiration found at SOJP by Brooks et al. (1998).

Declines in ψ_c imposed by hydraulic constraints forced declines in ψ_T and eventually in g_c (e.g. Fig. 9c), which imposed nonstomatal (ψ_T) and stomatal (g_c) limitations on CO₂ fixation (‘GPP’). Declines in CO₂ fixation under higher D, combined with rises in R_a+R_h under higher T_a and T_s, caused earlier declines in daytime CO₂ influxes and rises in night-time CO₂ effluxes to be modelled and measured during warming events (e.g. DOY 228 in Fig. 3c and DOY 179–180 in Fig. 6c). Daytime CO₂ influxes rapidly returned to higher values with less-pronounced mid-day declines when hydraulic constraints on CO₂ fixation were alleviated by lower D during subsequent cooling (e.g. DOY 236–237 in Fig. 3b and c and DOY 181–183 in Fig. 6c and d). The imposition and alleviation of these constraints on CO₂ fixation was clearly apparent in EC measurements of CO₂ influxes when D rose above or fell below 2 kPa. Declines in CO₂ influxes with higher D, combined with rises in CO₂ effluxes with higher T_a, were also apparent as sharp declines in daily NEP from the model or from gap-filled EC fluxes whenever maximum daily T_a exceeded 25 °C (e.g. DOY 195–205 in Fig. 7). The adverse impacts of warming on CO₂ influxes caused by hydraulic constraints and accelerated respiration in conifers has been observed in several other studies (Griffis et al., 2003; Morgenstern et al., 2004; Grant et al., 2005, 2006b).

Hydraulic constraints in the model rose gradually with stand age because Ω_a increased with longer axial path lengths for water uptake as trees grew taller (Grant et al., 2006b), reaching 13 m in the 75-year-old stand at SOJP. Mencuccini & Grace (1996) found that hydraulic conductance and NPP of Scots pine attained maximum values in a stand of 7.7 m height, and declined in stands of greater heights. The gradual rise in hydraulic constraints, particularly on this rapidly drained soil with low water holding capacity, caused the very gradual decline in NEP and wood growth modelled more than 50 years after clearcutting (Fig. 8a and b). This decline was consistent with smaller wood growth increments for boreal jack pine stands older than 60 years calculated from allometric relationships (Alberta Forest Service, 1985).

Nutrient constraints on forest NEP

CO₂ influxes remained small during the first decade after clearcutting, even when hydraulic constraints were not present (3, 6). During this period, GPP was limited by root N uptake which in turn was limited by competition with microbial populations for soil mineral N (‘Heterotrophic respiration’) during decomposition of fine nonwoody and coarse woody litter with high C : N ratios (ca. 40 : 1 and 250 : 1, respectively) left after clearcutting. Ecosystem productivity is frequently found to be strongly limited by N during regeneration after logging (e.g. Schimel & Firestone, 1989; Kimmins, 2004). Consequently, annual NEP from the model and from EC measurements remained below zero for at least 10 years after harvesting (Table 5; Fig. 8a). In sensitivity tests of the model, NEP declined further and rose more slowly after clearcutting when larger amounts of litter remained on site. This model behaviour was consistent with the observation by Janisch & Harmon (2002) that the regeneration of disturbed coniferous sites in northwestern USA was delayed by large amounts of coarse woody debris left after disturbance. Although the amounts of litter left on site after clearcutting in the model were generally consistent with litter C stock measurements at the jack pine chronosequence, these amounts will depend on harvesting practices, for which better estimates of wood and foliar litter will be needed in the future.

After clearcutting, some of the litter decomposition products in the model gradually accumulated in microbial residue pools with lower C : N ratios (Grant et al., 1993a, b). The mineralization of these pools eventually alleviated competition for soil mineral N among root and microbial populations, allowing sustained rises in GPP that enabled NEP to reach maximum values between 25 and 50 years after clearcutting. These values were consistent with those estimated from EC fluxes and C stock measurements 29 and 76–80 years after clearcutting (Table 5; Fig. 8a). However, jack pine growth remained N-limited throughout the logging cycle, as indicated by foliar C : N ratios of 45–50 modelled and measured in mature stands. These ratios remained well above the threshold of 30 above which N is considered limiting in conifers (Erisman et al., 1998). Growth of boreal jack pine stands in Canada with soil and foliar C : N ratios similar to those at SOJP have been found to increase sharply with N fertilizer application (Morrison & Foster, 1977). The sensitivity of GPP to N indicates the importance of representing N constraints on GPP when modelling boreal forest response to climate change, especially because higher soil temperatures accelerate N mineralization in boreal forest ecosystems (Verburg, 2004).

Climate change effects on forest NEP: Hydraulic constraints

The gradual rise in C_a during climate change (Table 4) accelerated CO₂ fixation modelled in boreal jack pine by increasing C_i and hence carboxylation rates, thereby raising CO₂ influxes (e.g. DOY 221–222 in Fig. 9d). This rise also accelerated CO₂ fixation by reducing g_c (Fig. 9c) and hence transpiration, thereby raising ψ_c and lowering nonstomatal limitations to CO₂ fixation at a given ψ_s (Grant et al., 2004b). However, the rise in T_a during climate change (Table 4) offset the effects of reduced g_c on transpiration by raising D (Fig. 9b), so that evapotranspiration rose and θ declined (Fig. 9a). Rising T_a also increased the frequency and intensity of warming events and their adverse impacts on jack pine CO₂ fixation caused by hydraulic limitations under higher D (e.g. DOY 226–227 in Fig. 9d). These warming events lowered nonstructural C in the model by slowing contributions from CO₂ fixation and accelerating removals by R_m, thereby hastening litterfall and needle turnover and lowering LAI (‘Autotrophic respiration’). Aboveground litterfall rates were comparable with measured rates under current climate (Table 5), and gave an annual turnover of foliage and twigs of ∼0.3. The more rapid litterfall modelled during climate change was consistent with the frequent observation that accelerated litterfall in pine follows high temperatures (Kouki & Hokkanen, 1992) or high evapotranspiration rates (Berg & Meentemeyer, 2001).

During litterfall in the model, remobilized C from senescent material sustained nonstructural C and hence R_a, while litterfall reduced foliar C and hence R_m, thereby rebalancing contributions to and removals from nonstructural C. Litterfall was thus used in the model to lower foliar mass and area of jack pine under rising D so that respiration requirements remained in balance with primary productivity. The model thus simulated the lower relative foliar mass found in pine growing in climates with higher D (Callaway et al., 1994; DeLucia et al., 2000) or in fir growing under higher T_a (Olszyk et al., 1998).

As climate change progressed, C gains modelled under lower T_a and D gradually failed to offset C losses modelled under higher T_a and D (Fig. 9). This model behaviour was consistent with the findings from EC measurements that above-average summer T_a was correlated with decreased net C uptake (Hollinger et al., 2004), from tree ring analyses that above-average summer T_a adversely affected growth of spruce and pine (Dang & Lieffers, 1988; Oleksyn et al., 1998; Hofgaard et al., 1999), and from growth chambers that elevated T_a reduced height and foliar mass of conifer seedlings (Olszyk et al., 1998). The model thus combined a positive response of NEP to increases in lower T_a with a negative response of NEP to increases in higher T_a (Fig. 9). These model results may explain the findings of Wilmking et al. (2004) that growth of Alaskan conifers tended to respond positively to warmer springs and negatively to warmer summers.

Declines in foliar mass from accelerated litterfall under continually rising T_a eventually caused failure of the boreal jack pine ecotype after 100 years of climate change (Fig. 8a and b), and its replacement by other vegetation. Loss of conifer dominance during climate change in boreal ecosystems has been predicted by biogeochemical models (e.g. Potter, 2004) and by dynamic vegetation models (e.g. Kirilenko & Solomon, 1998). The ability to model rapid change in vegetation caused by mortality of dominant woodland species is an important attribute of models used to project climate change impacts on vegetation dynamics (Allen & Breshears, 1998).

Replacing the boreal ecotype with a cold temperate ecotype reduced jack pine productivity during the first 75 years of climate change, but greatly improved it thereafter (Fig. 10a and b). This improvement was attributed to:

Climate change effects on forest NEP: nutrient constraints

In N-limited ecosystems, gains in CO₂ fixation during climate change can only be sustained by commensurate gains in root N uptake. Under gradual climate change in the model, root N uptake rose almost proportionately with CO₂ fixation, so that foliar C : N ratios by which carboxylation rates were governed rose only marginally (e.g. from 42.7 in July of the 100th year under current climate to 44.1 in July of the 100th year under climate change). The model thus simulated only to a limited extent the downregulation of CO₂ fixation caused by lower foliar N contents frequently observed in trees under elevated vs. ambient C_a and common T_a (e.g. Medlyn et al., 1999). This rise in root N uptake was modelled because:

These three responses to rising C_a and T_a allowed the model to simulate a continuous rise in NEP and wood growth during climate change. These responses simulated the coupling of C and N cycles in soil–plant systems that was proposed by Eamus & Jarvis (1989) and Idso & Idso (1994) to moderate N limitations under long-term increases in C_a.