Scaling foliar respiration in two contrasting forest canopies

Authors


†Author to whom correspondence should be addressed. E-mail: m.turnbull@botn.canterbury.ac.nz

Summary

  • 1The responses of night-time dark respiration (Rd) to temperature and leaf characteristics were measured through the canopies of tree species from two distinct forests – an oak-dominated deciduous forest in north-eastern USA, and a conifer-dominated temperate rainforest in New Zealand. These were chosen to examine the extent to which canopy level changes in dark respiration can be applied across forest biomes, and the appropriateness of scaling rules to calculations of whole-canopy carbon efflux.
  • 2The response of respiration to temperature differed significantly between species and with height in the canopy. This involved changes in both Rd at a reference temperature, and the extent to which Rd increased with temperature (described by the energy of activation, Eo, or the change in Rd over a 10 °C range, Q10). Eo ranged from 25 (lower-canopy leaves) to 53·8 kJ mol−1 K−1 (upper-canopy leaves) in the deciduous forest, and from 24–37 kJ mol−1 K−1 in the temperate rainforest site.
  • 3Relationships between respiratory and leaf characteristics indicated that the instantaneous rate of respiration covaries with soluble sugar concentration and leaf nitrogen, but the temperature response of respiration (Eo or Q10) appears to be driven by leaf N.
  • 4Scaling leaf respiratory carbon loss to the whole-canopy level indicated that simplifying assumptions regarding the variation in respiration and its temperature response with canopy height tend to underestimate carbon loss if the assumptions are based on lower-canopy leaf physiology, but overestimate carbon loss if the assumptions are based on upper-canopy physiology. Thus, canopy-level differences in leaf respiratory characteristics should be considered in modelling efforts attempting to estimate whole-canopy respiration.

Introduction

Although there is debate concerning our ability to predict future temperature change, it is now widely accepted that global temperatures will be 1–6 °C warmer by the year 2100 (Hansen et al. 1999; IPCC 2001). This warming will be more pronounced at night when dark respiration (Rd) is the dominant physiological process in vegetation (Easterling et al. 1997; Alward et al. 1999; IPCC 2001). The degree to which respiration in leaves changes with temperature is highly variable, with Q10 values (the increase in respiration rate with a 10 °C increase in temperature) ranging from 1·4–4·0 (Azcón-Bieto 1992). The temperature response of respiration is influenced by measurement temperature (Tjoelker et al. 2001); species (Larigauderie & Körner 1995); season (Stockfors & Linder 1998; Atkin et al. 2000); growth temperature (e.g. Larigauderie & Körner 1995; Atkin et al. 2000); and leaf metabolic state (Berry & Raison 1981). Respiration rate under field conditions will also be a function of temperature and physiological history, as it is subject to acclimation and/or adaptation (Amthor 1989; Atkin et al. 2000). Because actual leaf and canopy respiratory carbon release will be influenced by complex interactions between physiological and environmental factors, global warming could have dramatic nonlinear effects on system-level carbon, nitrogen and water cycling, and therefore on carbon storage (Dewar et al. 1999; Melillo 1999).

The carbon balance of forest ecosystems is defined by the difference between two large fluxes: photosynthetic carbon gain and respiratory carbon loss. On an annual basis, respiration can consume between 30 and 70% of photosynthetic carbon fixation (Amthor 1989; Ryan 1991; Ryan et al. 1994; Ryan et al. 1996). The processes of photosynthesis and respiration respond independently to environmental variation, and thus have nonlinear effects on forest carbon gain (Ryan 1991; Dewar et al. 1999; Gunderson et al. 2000). Using data on net CO2 exchange for a network of sites in Europe, Valentini et al. (2000) concluded that the decrease in carbon uptake with increasing latitude was the result of differences in ecosystem respiration rather than differences in gross carbon uptake. Given that, at the global scale, plant respiration results in the release of 60 Gt of carbon annually (Amthor 1997), 80% of which is from forest trees (Houghton 1993), it is essential that respiratory responses of forest trees to environmental variables be better understood. In order to interpret correctly the existing experimental results at the global scale, and to apply this knowledge correctly to predictive models, the uncertainties regarding both short- and long-term responses of respiration to temperature (Grace & Rayment 2000) must be resolved.

While the distribution of respiratory activity (Bolstad et al. 1999; Griffin et al. 2001; Meir et al. 2001) and photosynthetic capacity (Field 1983; Field 1991; Evans 1993; Anten et al. 1995; Hollinger 1996) vary through the canopy, the temperature response of these processes at different canopy depths has not been well studied (Bolstad et al. 1999). Soluble carbohydrate concentrations may regulate the temperature response, and these carbohydrates in turn vary through the canopy (Atkin et al. 2000; Griffin et al. 2001). Similarly, leaf nitrogen concentration (N) tends to decrease with canopy depth (Field 1983; Evans 1989; Leuning et al. 1991; Ellsworth & Reich 1993; Hollinger 1996), and a general relationship between N and Rd has been reported (Ryan 1991; Ryan 1995; Reich et al. 1996, Ryan et al. 1996; Reich et al. 1998). Hence, the temperature response of respiration might be expected to vary through the canopy of large trees.

In order to scale foliar respiration from leaf to canopy level, we examine the potential for species-level and canopy-level differences in (i) the temperature response of Rd; and (ii) the relationships between Rd and leaf characteristics. Rd of leaves was measured throughout the canopies of three dominant tree species from a temperate deciduous forest in eastern North America and three tree species from a conifer-dominated forest in southern New Zealand. We hypothesized that the response of leaf respiration to temperature would differ with position in the canopy, and would be driven by leaf N and carbohydrate content. Our two research sites comprise two very different forest types. The North American site is a temperate deciduous forest which experiences significant anthropogenic N input (in excess of 4·7 kg N ha−1 year−1; National Atmospheric Deposition Programme, 2001, personal communication) and summer drought. The New Zealand site is a nutrient-poor conifer-dominated site that experiences high rainfall throughout the year. These distinct sites were chosen to examine the extent to which canopy-level changes in the response of dark respiration to temperature and leaf characteristics can be broadly applied across forest biomes, and to investigate the appropriateness of scaling rules to calculations of whole-canopy respiratory carbon efflux.

Materials and methods

study sites and species

The Black Rock Forest (BRF) is a 1500 ha scientific preserve within the Highlands Province of New York, USA and was established in 1927. The forest is located at a latitude of 41·3° N and a longitude of 74·0° W with elevations ranging from 110–450 m above sea level. Average annual precipitation is 1190 mm. Air temperature is strongly seasonal, with monthly averages ranging from −2·7 °C in January to 23·4 °C in July. The soils are typically shallow, brown forest soils of medium texture, with granite gneiss bedrock or glacial till parent material at depths ranging from 250 mm at elevated sites (Hollis soil) to 1 m at lower sites (Charlton/Paxton soil; Olsson 1981). The forest contains an average of 734 trees ha−1 and an average basal area of 21·0 m2 ha−1. The forest population structure has been described previously (Turnbull et al. 2002). A 0·1 ha study plot was established at the bottom of the Cascade Brook watershed at an elevation of 270 m and a slope of <5%. In this plot the canopy height ranges from 20–30 m. Basal area distribution of the three species studied is 41%Quercus rubra L., 6%Quercus prinus L., 32%Acer rubrum L. and 20% other species.

The New Zealand site is located in an extensive, mixed conifer–broadleaved forest at Okarito Forest, Westland (OFW) at latitude 43·2° S and longitude 170·3° E and 50 m elevation above sea level. Because of the prevailing, onshore westerly winds, and the rapid increase in altitude associated with the southern Alps in close proximity east of the forest, annual rainfall is high (≈3400 mm) and evenly distributed throughout the year. Mean annual temperature is 11·3 °C, with a small range between winter and summer of 8·6 °C. The landform at the site is glacial in origin, and the forest is established on terrace outwashes from moraines formed ≈20 000 years ago. The loess is poorly preserved because of erosion and acid dissolution from extreme leaching, resulting from high rainfall (Almond 1996). The soils have a high organic matter content, low permeability and porosity, and are frequently waterlogged. Soils are extremely acid (pH 3·8–4·4 to a depth of 500 mm) with medium total N concentrations (2·1 mol kg−1) in the upper 150 mm, falling to 0·14 mol kg−1 at a depth of 150 mm, and low concentrations of acid-extractable phosphorus and low phosphorus retention (Mew & Lee 1981). Forests of this kind at lowland sites are characterized by a dense, mixed understorey and large, emergent conifers, principally rimu (Dacrydium cupressinum Sol. ex Lamb.) (Ogden & Stewart 1995). Our study plot is dominated by 400–600-year-old rimu trees with a mean canopy top height of 20 m. Other conifers and angiosperm species are also present (James & Norton 2002). Miro (Podocarpus ferruginea (D. Don) de Laub.) and kamahi (Weinmannia racemosa L. f.) are common in the subcanopy, and silver pine (Manoao colensoi (Hook.) Molloy), Hall's totara (Podocarpus hallii Kirk), Westland quintinia (Quintinia acutifolia Kirk) and southern rata (Metrosideros umbellata Cav.) are present, but less common. The basal area distribution of the three species studied here is 73%D. cupressinum, 12%W. racemosa, 10%Q. acutifolia and 5% other species.

respiration measurements

Physiological measurements presented here were undertaken during summer (June 2000 at BRF; January 2001 at OFW). Measurements of leaf dark respiration were made on fully expanded leaves from three levels (upper, mid- and lower) within the canopies of each experimental tree. Canopy access was gained at BRF using a 20 m boom ‘cherry-picker’ and at OFW using a 22 m permanent tower. For Q. rubra and Q. prinus at BRF and D. cupressinum at OFW, the upper-canopy level consisted of fully sunlit leaves from the top of the forest canopy. For A. rubrum at BRF and Q. acutifolia and W. racemosa at OFW, upper-level leaves were partially sunlit in the mid-stratum of the forest. At least six leaves (from at least two separate branches on three individual trees) were measured for each species per canopy level. Leaf carbohydrate, total N and specific leaf area (SLA) analyses were performed on the same leaves used for gas-exchange measurements.

Dark respiration measurements were made using gas analysis systems (Li-Cor model 6400, Lincoln, NE, USA) equipped with CO2 control modules. Response curves were generated following measurements of steady-state responses of leaf dark respiration (Rd) to four different cuvette temperatures (≈12, 18, 24 and 28 °C) which were set using thermoelectric coolers. These experimental temperatures represented a range of approximately ±7 °C around the ambient minimum temperature during the study period. Responses were modelled according to actual leaf temperatures achieved at these temperature set points (≈13, 18, 22 and 25 °C, respectively). External CO2 (Ca) was maintained at ambient atmospheric partial pressure (36 Pa). Measurements were made at each temperature set point when respiratory gas exchange had equilibrated (taken to be when the coefficient of variation for the CO2 partial pressure differential between the sample and reference analysers was <1%). This condition was typically achieved ≈10 min after the stable temperature set point had been reached. Measurements were made at night (after 2200 h) in the dark using individual leaves on large branches excised under water from trees in the field. Previous studies have shown no differences in respiration between in situ leaves and leaves from detached branches in a range of deciduous species, including Q. rubra and A. rubrum (Mitchell et al. 1999). We also confirmed that this applies for the New Zealand forest species (unpublished results). Photosynthetic capacity (Amax) was determined under saturating photon flux density (1500–2000 µmol m−2 s−1) and ambient Ca (36 Pa) and 25 °C during the previous daytime period for in situ leaves, adjacent in the canopy to those subsequently used for respiration determinations. All gas-exchange measurements are presented on the basis of half leaf surface area.

Analysis of temperature response curves was performed using a modification of an Arrhenius function described by Lloyd & Taylor (1994), where respiration rate at a given temperature is given by:

image(eqn 1)

where R0 is the respiration rate at the base temperature T0 (here 10 °C), Ta is leaf temperature (K), Rg is the gas constant (8·314 J mol−1 K−1), and Eo is a parameter related to the energy of activation which describes the magnitude of the temperature response. Nonlinear curve fitting was performed using the Marquardt–Levenberg algorithm (sigmaplot ver. 2·0, 1994, Jandel Corp.). In order to facilitate comparisons with other investigations, the parameter Q10 was also calculated (the ratio of Rd at 25 °C divided by Rd at 15 °C). Q10 should be used with caution as the value changes according to the temperature range used (Amthor 1989; Ryan et al. 1994).

modelled stand foliar carbon loss

The calculated temperature response (equation 1) was used to predict leaf respiration from actual air temperatures during a 2-week period centred on when gas exchange measurements were made. The instantaneous rates were then scaled to the stand level (Rcan) by multiplying by the leaf area index for each forest (5·8 m2 m−2 for BRF and 5·4 m2 m−2 for OFW) and by total area (1 ha). Two separate model calculations were made in order to illustrate the impact of simplifying assumptions regarding the variation in respiration and its temperature response with canopy height. In the first, the distributed foliar characteristics model, the vertical distribution of the respiration rate and temperature response of respiration were considered explicitly, along with the vertical distribution of leaf area and contribution to leaf area made by the three major component species for each forest (based on stem basal area). For BRF, vertical leaf area distribution was considered to be the same as that determined previously for an oak-dominated deciduous forest (Hutchison et al. 1986). At OFW, vertical leaf area distribution was inferred from photographs of tree crowns (Whitehead et al. 2002). Within each 1 m segment of the canopy, the appropriate fraction of total leaf area was allocated to those species and leaf types present according to the contribution each species made to stem basal area. In the second model calculation, the constant foliar characteristics model (with the additional designation of lower, mid- or upper canopy), all leaves for a given species were assumed to have the respiration rate and temperature response measured for that species in only the lower, mid- or upper canopy, respectively. All calculations were made on a 20 min time scale and then summed for the duration of the night-time period (2100–0700 h). Canopy assimilation rate (Acan) during the day was also calculated for the two forests using a process-based model (Whitehead et al. 2002).

leaf analyses

All analyses were determined on leaf material harvested directly following respiratory measurements. Specific leaf area was calculated following determination of individual leaf area and dry weight. Soluble sugar content of leaves was determined colorimetrically using an ethanol extraction technique described by Griffin et al. (1999). Total N content of leaves was determined on dried and ground material using an NCS autoanalyser (Carlo Erba NCS 2500, Milan, Italy).

statistical analysis

Two-way anova was used to test for the main effects and interactions of species and canopy level on respiration parameters (s-plus ver. 3·3, MathSoft Inc., Seattle, USA). A nested model (individual leaves nested within trees) was used to account for leaf vs tree variation in parameters (Underwood 1981). Differences were considered significant if probabilities (P) were <0·05. Treatment means were compared by least significant difference to determine whether means of the dependent variable were significantly different at P = 0·05 (Sokal & Rohlf 1981).

Results

In both forests the response of respiration to temperature differed between species and with position in the canopy. At BRF, A. rubrum displayed a less pronounced increase in respiration with increasing temperature than did either Q. rubra or Q. prinus (Fig. 1). The effects of species and stratum on the temperature response of respiration were also exemplified in the parameters Eo and Q10 (Table 1). For both these parameters A. rubrum had lower values than the other two species, and leaves in the upper-canopy stratum had higher values than those in the mid- and lower strata. Eo ranged from 25·2 in lower-canopy leaves of A. rubrum to 53·9 kJ mol−1 K−1 in upper-canopy leaves in Q. prinus (Table 1). Leaf respiration differed significantly between canopy strata, with R10 (P < 0·01) and R25 (P < 0·001) significantly smaller in lower-canopy than in upper-canopy leaves (Table 1). There were significant differences in respiration between species at both high (R25, P < 0·001) and low temperature (R10, P = 0·02). At OFW, D. cupressinum displayed the most pronounced increase in respiration with increasing temperature, particularly in upper- and mid-canopy leaves (Fig. 1). Eo and Q10 both differed significantly among species (Table 1). Although there was a trend for both parameters to be greater in the upper canopy than the lower, this was much less pronounced than for BRF. Eo ranged from 23·9 in lower-canopy leaves of Q. acutifolia to 38·3 kJ mol−1 K−1 in mid- and upper canopy leaves in W. racemosa (Table 1). Leaf respiration differed significantly among strata, with both R10 (P < 0·001) and R25 (P < 0·001) significantly smaller in lower-canopy than upper-canopy leaves (Table 1).

Figure 1.

Response of respiration rate, Rd, to temperature for leaves at different canopy levels of Quercus rubra, Quercus prinus and Acer rubrum at Black Rock Forest, NY, USA and Dacrydium cupressinum, Weinmannia racemosa and Quintinia acutifolia at Okarito Forest, Westland, NZ. Data are shown for upper-canopy (□), mid-canopy (○) and lower-canopy (•) leaves. The curves describe modelled responses derived using the mean parameters from 6 to 7 replicate response curves at each canopy level (individual curves fitted using a modification of the Arrhenius equation described by Lloyd & Taylor 1994). For statistical comparison of respiration parameters, see Table 1.

Table 1.  Modelled dark respiration parameters calculated from fitted temperature response curves (Fig. 1) as a function of canopy height in three tree species from a deciduous forest (Black Rock Forest, NY, USA) and three species from a conifer-dominated temperate rainforest (Okarito Forest, Westland, NZ)
 CanopyCanopyCanopyanova statistics
UpperMidLowerUpperMidLowerUpperMid Lower
  1. R10 is respiration rate at the base temperature (10 °C) used to fit equation 1 to actual responses. R25 denotes respiration rates calculated from the fitted responses at 25 °C. Eo is a modelled parameter related to the energy of activation. Q10 denotes the relative change in respiration rate with a 10 °C increase in temperature (15–25 °C). Values shown are means (± SEM) where n = 6–7. Significance of treatment effect for species (Sp), canopy position (Pos) and the interaction between species and canopy position (Sp × P) are indicated as the P value or as nonsignificant (NS). Different letters within rows indicate statistically different values at P < 0·05 based on the least significant difference test of treatment means.

Black Rock Forest (USA)Quercus rubraQuercus prinusAcer rubrum 
R10 (µmol CO2 m−2 s−1) 0·49 (0·02)bc0·46 (0·03)b0·42 (0·04)b0·58 (0·08)c0·41 (0·05)b0·37 (0·03)ab0·42 (0·04)b0·31 (0·03)a0·39 (0·03)abSp 0·02 Pos 0·01 Sp × P NS
R25 (µmol CO2 m−2 s−1) 1·53 (0·11)d1·42 (0·16)d1·13 (0·18)c1·77 (0·13)e0·90 (0·06)bc0·95 (0·11)bc1·00 (0·09)c0·74 (0·05)ab0·66 (0·03)aSp <0·001 Pos <0·001 Sp × P NS
Eo (kJ mol−1 K−1) 53·7 (3·3)c51·6 (3·9)c44·2 (4·7)bc53·9 (5·2)c38·0 (3·3)b43·3 (6·3)bc39·9 (3·6)b42·0 (2·0)bc25·2 (3·1)aSp 0·01 Pos 0·01 Sp × P NS
Q102·13 (0·09)c2·07 (0·11)c1·87 (0·11)bc2·16 (0·15)c1·71 (0·08)b1·86 (0·15)bc1·76 (0·09)b1·80 (0·05)b1·43 (0·06)aSp 0·04 Pos 0·03 Sp × P NS
Okarito Forest (NZ)Dacrydium cupressinumWeinmannia racemosaQuintinia acutifolia 
R10 (µmol CO2 m−2 s−1) 0·73 (0·06)c0·67 (0·03)c0·38 (0·04)ab0·65 (0·08)c0·35 (0·08)ab0·46 (0·06)b0·44 (0·03)b0·31 (0·02)ab0·20 (0·03)aSp 0·05 Pos <0·001 Sp × P NS
R25 (µmol CO2 m−2 s−1) 1·46 (0·15)c1·19 (0·07)bc0·71 (0·12)ab1·39 (0·13)c0·81 (0·22)b0·88 (0·12)b0·89 (0·05)b0·52 (0·04)ab0·37 (0·05)aSp NS Pos <0·001 Sp × P NS
Eo (kJ mol−1 K−1) 31·8 (2·9)ab26·0 (2·6)a26·4 (4·6)a37·1 (4·2)b38·3 (6·3)b29·1 (5·1)ab33·3 (2·7)b23·9 (3·1)a32·4 (4·2)bSp 0·05 Pos NS Sp × P NS
Q101·57 (0·06)ab1·45 (0·05)a1·45 (0·09)a1·71 (0·11)b1·74 (0·16)b1·53 (0·10)ab1·60 (0·06)ab1·41 (0·06)a1·59 (0·10)abSp 0·05 Pos NS Sp × P NS

Leaf characteristics differed among species, and displayed significant responses to canopy position (Table 2). At BRF, SLA responded to canopy position (P < 0·0001) and was significantly larger in A. rubrum (15·2 in upper canopy compared to 17·6 m2 kg−1 in lower canopy) than in both Q. rubra (10·3–17·4 m2 kg−1) and Q. prinus (9·0–14·9 m2 kg−1). The SLA responded similarly to canopy position at OFW. Nitrogen concentration on both a mass and area basis differed significantly among species at both sites, and was greatest in Q. rubra at BRF and D. cupressinum at OFW. Narea decreased significantly with depth in the canopy in both forests (P < 0·001 at BRF; P < 0·0001 at OFW). In contrast, Nmass was much more constant through the canopy strata. At BRF and OFW, leaf soluble sugar content was greatest in the two Quercus species (BRF) and D. cupressinum (OFW), and decreased in lower-canopy leaves. At OFW, the range of values between canopy strata in SLA, Narea and soluble sugars was greatest in D. cupressinum and much less pronounced in W. racemosa and Q. acutifolia, as the latter were shaded subcanopy trees.

Table 2.  Leaf characteristics at different canopy levels of three tree species from a deciduous forest (Black Rock Forest, NY, USA) and three species from a conifer-dominated temperate rainforest (Okarito Forest, Westland, NZ)
 Canopy CanopyCanopyanova statistics
UpperMidLowerUpperMidLowerUpperMidLower
  1. Values shown are means (± SEM) where n = 6–7. Significance of treatment effect for species (Sp), canopy position (Pos) and the interaction between species and canopy position (Sp × P) are indicated as the P value or as nonsignificant (NS). Different letters within rows indicate statistically different values at P < 0·05 based on the least significant difference test of treatment means.

Black Rock Forest (USA)Quercus rubraQuercus prinusAcer rubrum 
SLA (m2 kg−1) 10·3 (0·31)a13·5 (0·85)b17·4 (2·15)cd9·0 (0·25)a14·7 (1·0)b14·9 (0·61)b15·2 (1·14)bc17·0 (0·58)cd17·6 (0·45)dSp <0·0001 Pos <0·0001 Sp × P 0·035
Nmass (mmol N g−1) 1·59 (0·03)bc1·73 (0·05)c1·75 (0·02)c1·33 (0·03)a1·29 (0·04)a1·38 (0·05)ab1·56 (0·04)bc1·55 (0·03)bc1·44 (0·04)abSp <0·01 Pos NS Sp × P 0·05
Narea (mmol N m−2)155 (3·5)d129 (4·3)cd107 (12·5)bc147 (5·5)d89 (4·6)ab93 (3·8)ab105 (8·2)b92 (3·3)ab72 (3·2)aSp <0·001 Pos <0·001 Sp × P NS
Sugars (g m−2)2·80 (0·27)cd2·07 (0·36)bc1·81 (0·18)ab3·64 (0·45)d3·04 (0·31)cd2·99 (0·13)cd1·15 (0·15)a1·67 (0·39)ab1·25 (0·15)aSp <0·01 Pos NS Sp × P NS
Okarito Forest (NZ)Dacrydium cupressinumWeinmannia racemosaQuintinia acutifolia 
SLA (m2 kg−1) 4·86 (0·32)a6·43 (0·35)ab7·55 (0·27)b7·90 (0·61)b7·90 (0·99)b7·91 (0·41)b10·8 (0·20)cd12·1 (0·38)d13·2 (0·53)eSp 0·013 Pos 0·022 Sp × P NS
Nmass (mmol N g−1)0·93 (0·03)e0·84 (0·02)d0·84 (0·05)d0·69 (0·04)b0·68 (0·02)ab0·66 (0·02)a0·79 (0·03)c0·72 (0·03)bc0·70 (0·02)bSp <0·0001 Pos 0·01 Sp × P NS
Narea (mmol N m−2)186 (7·9)f133 (6·1)e114 (7·8)de91 (5·4)cd90 (6·1)cd85 (4·2)bc73 (2·4)abc60 (2·9)ab54 (2·2)aSp <0·0001 Pos <0·001 Sp × P <0·01
Sugars (g m−2)9·76 (1·25)c5·67 (0·64)ab5·72 (0·84)ab6·76 (0·85)b7·33 (0·83)bc6·26 (0·52)b6·26 (1·30)b5·39 (0·75)ab3·12 (0·50)aSp NS Pos <0·01 Sp × P 0·03

Leaf Rd at the average minimum night-time temperature for each site (June average minimum at BRF was 17·2 ± 0·53 °C; January minimum at OFW was 11·4 ± 0·50 °C) was calculated from the respiration response characteristics (Fig. 1; Table 1) to reflect respiration under ambient conditions. Respiration rates expressed on a leaf area basis (Rarea) differed between species at both sites, and were significantly greater in upper-canopy leaves than in mid- and lower-canopy leaves (Table 3). At BRF, Rarea ranged from 0·98 µmol m−2 s−1 in upper-canopy leaves of Q. prinus to 0·49 µmol m−2 s−1 in lower-canopy leaves of A. rubrum. At OFW Rarea values were smaller, reflecting lower air temperature, and ranged from 0·80 in upper-canopy leaves of D. cupressinum to 0·22 µmol m−2 s−1 in lower-canopy leaves of Q. acutifolia. When expressed on a mass basis, patterns of variation in respiration rate (Rmass) were less clear, displaying a significant species effect (P = 0·031) at BRF and a significant canopy position effect at OFW (P < 0·001). At OFW, Rmass was generally smaller in lower-canopy leaves. The response of respiration on a nitrogen basis (RN) also differed between the two forest sites, with no canopy position effect in the deciduous BRF, but a significant canopy position effect at evergreen OFW. At OFW, RN tended to decrease in lower-canopy leaves.

Table 3.  Night respiration for leaves at different canopy levels of three tree species from a deciduous forest (Black Rock Forest, NY, USA) and three species from a conifer-dominated temperate rainforest (Okarito Forest, Westland, NZ)
 Canopy CanopyCanopyanova Statistics
UpperMidLowerUpperMidLowerUpperMidLower
  1. Respiration rates are calculated from modelled responses (Table 1) at the average minimum night-time temperature for each site during a 2-week period (17·2 °C for Black Rock Forest and 11·4 °C for Okarito Forest). Values shown are means (± SEM) where n = 6–7. Significance of treatment effect for species (Sp), canopy position (Pos) and the interaction between species and canopy position (Sp × P) are indicated as the P value or as nonsignificant (NS). Different letters within rows indicate statistically different values at P < 0·05 based on the least significant difference test of treatment means.

Black Rock Forest (USA)Quercus rubraQuercus prinusAcer rubrum 
Rarea (µmol CO2 m−2 s−1) 0·83 (0·04)d0·78 (0·06)cd0·67 (0·08)bc0·98 (0·10)e0·59 (0·05)ab0·58 (0·05)ab0·63 (0·06)bc0·46 (0·03)a0·49 (0·02)aSp <0·001 Pos 0·026 Sp × P NS
Rmass (µmol CO2 kg−1 s−1) 8·57 (0·41)ab10·5 (0·98)b12·7 (3·05)c8·87 (0·90)ab8·47 (0·38)ab8·55 (0·68)ab8·91 (0·63)ab7·93 (0·54)a8·84 (0·45)abSp 0·031 Pos NS Sp × P NS
RN (µmol CO2 mol N−1 s−1)5·39 (0·41)ab6·08 (0·52)bc5·45 (0·34)ab6·74 (0·78)c6·65 (0·41)c6·23 (0·48)bc5·65 (0·24)ab5·00 (0·24)a6·16 (0·33)bcSp NS Pos NS Sp × P NS
Okarito Forest (NZ)Dacrydium cupressinumWeinmannia racemosaQuintinia acutifolia 
Rarea (µmol CO2 m−2 s−1)0·80 (0·07)c0·73 (0·04)c0·41 (0·05)ab0·72 (0·08)c0·39 (0·09)ab0·50 (0·06)b0·48 (0·03)b0·37 (0·02)ab0·22 (0·03)aSp 0·05 Pos <0·0001 Sp × P NS
Rmass (µmol CO2 kg−1 s−1)3·93 (0·44)abc4·69 (0·34)cd3·07 (0·34)a5·73 (0·80)d2·91 (0·66)a4·11 (0·59)abc5·32 (0·42)c3·93 (0·22)abc3·08 (0·50)aSp NS Pos <0·001 Sp × P NS
RN (µmol CO2 mol N−1 s−1)4·31 (0·36)ab5·61 (0·35)bc3·77 (0·40)a7·96 (0·90)d4·22 (0·90)ab6·33 (0·98)cd6·93 (0·76)cd5·52 (0·31)bc4·50 (0·80)abSp <0·001Pos <0·01Sp × P NS

At BRF, the magnitude of the temperature response of leaf respiration (Eo) was positively correlated with Narea (Fig. 2a; r2 = 0·29–0·31, Table 4). The slope of these responses did not differ between species. In contrast, Eo was not related to leaf soluble sugar content (Fig. 2b). R25 was strongly correlated with Narea (Fig. 2c; r2 ranged from 0·28 in Q. rubra to 0·77 in Q. prinus and 0·83 in A. rubrum, Table 4), but weakly correlated with soluble sugar content in Q. prinus only (Fig. 2d). Relationships between respiration parameters and leaf characteristics were less clear at OFW. In contrast to findings for deciduous forest species at BRF, at OFW Eo was not significantly related to Narea (Fig. 2e), although a weak relationship was observed for D. cupressinum. Eo was not related to leaf soluble sugar content (Fig. 2f). R25 at OFW was correlated with Narea in D. cupressinum and Q. acutifolia (Fig. 2g; Table 4). R25 was not significantly related to leaf soluble sugar content at OFW (Fig. 2h).

Figure 2.

Relationships between respiratory (Eo and R25) and leaf characteristics (Narea and soluble sugar content) for leaves of Quercus rubra (▵), Quercus prinus (•) and Acer rubrum (□) from Black Rock Forest, NY, USA and Weinmannia racemosa (▵), Quintinia acutifolia (•) and Dacrydium cupressinum (□) trees from Okarito Forest, Westland, NZ. For statistical comparison of leaf characteristics and respiration parameters, see Tables 2 and 3, respectively.

Table 4.  Summary of regression statistics for significant relationships (P < 0·05) between respiratory (Eo and R25) and leaf characteristics (Narea and soluble sugar content) for leaves of Quercus rubra, Quercus prinus and Acer rubrum from Black Rock Forest, NY, USA (BRF) and Weinmannia racemosa, Quintinia acutifolia and Dacrydium cupressinum trees from Okarito Forest, Westland, NZ (OFW)
Dependent variableIndependent variableSpeciesSiteRegression relationshipr2
EoNareaQ. rubraBRF8300 + (302 × Narea)0·31
(kJ mol−1 K−1) Q. prinusBRF18210 + (245 × Narea)0·29
A. rubrumBRF1710 + (345 × Narea)0·31
R25NareaQ. rubraBRF−0·025 + (0·010 × Narea)0·27
(µmol CO2 m−2 × s−1) Q. prinusBRF−0·337 + (0·014 × Narea)0·77
A. rubrumBRF−0·236 + (0·011 × Narea)0·83
W. racemosaOFWNSNS
Q. acutifoliaOFW0·080 + (0·008 × Narea)0·17
D. cupressinumOFW−0·120 + (0·009 × Narea)0·47
R25Soluble sugarsQ. rubraBRFNSNS
(µmol CO2 m−2 s−1) Q. prinusBRF−0·003 + (0·375 × SolSug)0·42
A. rubrumBRFNSNS
W. racemosaOFWNSNS
Q. acutifoliaOFW0·420 + (0·040 × SolSug)0·25
D. cupressinumOFWNSNS

Leaf-level responses to ambient temperature conditions were scaled to reflect night-time respiratory carbon loss at the whole-canopy level (Fig. 3). The results are expressed as the difference between the constant physiology and distributed physiology models. At both sites there were clear differences in canopy respiration rates, depending on the model calculation. At BRF, assuming all leaves had the characteristics of those in the mid-canopy provided a respiratory loss estimate closest to that obtained using the distributed model. The lower-canopy constant physiology model consistently underestimated the distributed physiology model by as much as 39 mmol m−2 per night, while the upper-canopy constant physiology model consistently overestimated the distributed physiology model by as much as 32 mmol m−2 per night. Both upper- and lower-canopy constant physiology models displayed the strongest temperature sensitivity, with the greatest deviations from the distributed physiology model occurring at warmer temperatures. At OFW, respiratory loss was less variable due to more constant field temperature. Here, assuming all leaves had the characteristics of those in the upper canopy provided the respiratory loss estimate closest to that obtained using the distributed model. The mid- and lower-canopy constant physiology models consistently underestimated the distributed physiology model by as much as 29 and 87 mmol CO2 m−2 per night, respectively.

Figure 3.

Differences between two model calculations of canopy respiratory carbon release for a deciduous forest (Black Rock Forest, NY, USA) and a conifer-dominated temperate rainforest (Okarito Forest, Westland, NZ). Respiration rates are calculated from modelled responses (Table 1) using actual field temperatures (T) over the night period (2100–0700). The constant foliar characteristics models assume all leaves in the mixed canopies have the physiological properties of either lower-, mid- or upper-canopy leaves, respectively. In the distributed foliar characteristics model, leaves were assigned the appropriate response characteristics according to position in the canopy. For detailed modelling assumptions, see Methods. Note the different y-axis scales for each forest.

When the above responses to actual temperature variations are totalled for the 2-week period bracketing our measurement nights, it is possible to calculate total foliar carbon efflux and the extent to which simplifying assumptions affect calculated respiratory carbon loss under field conditions. Total canopy foliar carbon efflux (Rcan) over the 2-week period was 2·26 mol m−2 at BRF and 2·23 mol m−2 at OFW. This corresponded to total canopy foliar assimilation (Acan) during the same period of 10·3 mol m−2 at BRF (daily Acan/Rcan of 4·51 ± 0·34) and 5·73 mol m−2 at OFW (daily Acan/Rcan of 2·59 ± 0·09). At BRF, if all leaves in the canopy were assumed to have constant physiological characteristics, then total leaf carbon loss over the 2-week period varied from 1·93 (assuming the properties of the lower-canopy leaves) to 2·58 mol m−2 (assuming the properties of the upper-canopy leaves) (Table 5). These estimates were 14·7% less and 14·2% greater than that of the distributed physiology model. If the canopy was assumed to consist entirely of mid-canopy leaves, the estimate of total leaf carbon loss was not greatly different (−5·6%) from the distributed physiology model calculation. At the OFW site, stand foliar respiration rate was estimated at 2·23 mol m−2, similar to that at BRF, with the mid- and lower-canopy constant models underestimating the respiratory loss calculated by the distributed model by 14·5 and 46·9%, respectively. When expressed on a mass basis, canopy respiratory carbon efflux at BRF (32·0 mol kg−1) was twice that at OFW (16·4 mol kg−1; Table 5). On a mass basis, the constant physiology estimates of carbon efflux at BRF were much closer to the distributed physiology estimate throughout the canopy (+3·7 to −5·7% for the lower and upper canopy estimates, respectively). However, at OFW, mass-based estimates did little to bring the constant physiology model calculations together or reduce their deviation from the distributed model calculation.

Table 5.  Modelled canopy foliar respiratory carbon release (Rcan) for a deciduous forest (Black Rock Forest, NY, USA) and a conifer-dominated temperate rainforest (Okarito Forest, Westland, NZ)
 Foliar characteristics
DistributedConstant; canopy:
LowerMidUpper
  1. Respiration rates are calculated from modelled responses (Table 1) using actual field temperatures (Fig. 3) and are presented as the total carbon released per m2 ground area or per kg of leaf mass over a 2-week period during the growing season. For detailed modelling assumptions see caption for Fig. 3 and Methods. Percentage difference refers to difference between constant and distributed foliar characteristics model calculations.

Area-based
Black Rock Forest (USA)
 Rcan (mol m−2) 2·261·932·13   2·58
 % Difference  −14·7 −5·6+14·2
Okarito Forest (NZ)
 Rcan (mol m−2) 2·231·191·91   2·25
 % Difference  −46·9 −14·5 +0·70
Mass-based
Black Rock Forest (USA)
 Rcan (mol kg−1)32·033·231·1 30·2
 % Difference +3·7 −2·9 −5·7
Okarito Forest (NZ)
 Rcan (mol kg−1)16·49·4113·1 12·6
 % Difference  −42·7 −20·4 −23·0

Discussion

response of respiration to temperature

Our results show that the response of respiration to temperature may differ significantly in leaves of different canopy strata, and that these differences can alter calculated estimates of stand respiratory carbon efflux. The significant differences observed in leaf-level respiration rates between different canopy positions and in scaled canopy-level fluxes may have important mechanistic implications for whole-tree and ecosystem responses to variations in night-time temperature. Given that leaf respiratory response to temperature is a function of temperature and physiological history (Amthor 1989; Atkin et al. 2000) and varies with site (Turnbull et al. 2001), these findings indicate a need for further investigations into within-stand variation in the temperature responses of respiration in trees. Such investigations are essential to inform models that extrapolate from leaf-level to ecosystem- and landscape-level responses.

While the two oak species studied at BRF (Q. rubra and Q. prinus) showed similar responses to temperature, respiration in A. rubrum was considerably less sensitive (lower Eo and Q10). These values for Eo are consistent with those previously obtained at this site (Turnbull et al. 2001), but differ from the temperature responses of similar species in other studies (Bolstad et al. 1999; Amthor 2000) which found that Q10 rarely differed among species. Here we found significant differences between species, with values of Q10 ranging from 1·4–2·2. These values are at the low end of published values for Q10 (Azcón-Bieto 1992) and are likely to be due to seasonal differences. Our measurements were made when the active metabolic state of leaves and higher temperatures may reduce the temperature response of respiration (Stockfors & Linder 1998; Atkin et al. 2000). Temperature responses at OFW were generally more muted than at BRF, with Eo values in the range 24–38 kJ mol−1 K−1 (Q10 values in the range 1·4–1·7). There are few, if any, previous studies of equivalent species with which to assess these values. A review of Pinus spp. found Q10 values in the range 1·9–2·3 (Ryan et al. 1994). This places D. cupressinum (range of Q10 1·45–1·57) below other coniferous species.

An important finding of our investigation was the trend in all species, in both forests, toward a less accentuated response of respiration to temperature (lower Eo and Q10) with depth in the canopy. Although the canopy position effect on Eo and Q10 was statistically significant only at BRF, a clear trend was also displayed at OFW. This finding is of interest given its implications for modelling forest carbon budgets and appropriate sampling protocols. Few studies have investigated variation in the temperature response of Rd, but many more have investigated the effects of growth temperature (e.g. Larigauderie & Körner 1995; Atkin et al. 2000); geographical location (e.g. Gunderson et al. 2000); elevation (Mitchell et al. 1999); CO2 concentration (e.g. Amthor 2000; Tissue et al. 2002); and canopy position (e.g. Bolstad et al. 1999; Griffin et al. 2001) on respiration rate at a given temperature. We found that canopy-level differences in temperature response functions of respiration resulted from changes in both R10 and Eo. This is consistent with the findings of Griffin et al. (2002), but differs from those of Bolstad et al. (1999), who found that differences among species and canopy position were almost solely dependent on changes in respiration at the reference temperature (RTref) and not Q10.

respiration at ambient temperature

At BRF, canopy position differences in respiration rates were observed on a leaf area basis, but not when rates were scaled by either SLA or leaf N concentration. The consistency in Rmass and RN between canopy levels indicates a strong coupling between factors influencing respiration and those affecting leaf characteristics. We conclude from these results that differences in dark respiration are most probably based on the effects that canopy position has on demand for energy associated with leaf maintenance (Lambers et al. 1998). Higher respiration rates in upper-canopy leaves of Q. rubra and Q. prinus corresponded with greater foliar concentrations of soluble sugars and lower SLA (Table 2). This is consistent with previous findings (Azcón-Bieto & Osmond 1983; Tissue et al. 2002), although relationships between respiration and leaf carbohydrate content are often not clear (Atkin et al. 2000). In contrast, at OFW, canopy position differences in respiration rates were observed on a leaf area, leaf mass and leaf N basis. In all three species, Rarea, Rmass and RN decreased with depth in the canopy. This inconsistency in Rmass and RN with depth in the canopy is because SLA and Narea displayed modest changes with canopy depth, but Rarea decreased strongly in mid- and lower-canopy leaves.

relationships between respiration and leaf characteristics

Canopy depth differences in Rarea, Narea and soluble sugars allow us to investigate the likely physiological mechanisms underpinning respiration rate and its temperature response. We found that R25 was positively associated with Narea at both sites. This is consistent with previous studies (Reich et al. 1998; Bolstad et al. 1999; Mitchell et al. 1999), and is considered to reflect the close relationship between leaf N concentration and maintenance respiration (Amthor 1989; Ryan 1991). R25 was also positively correlated with leaf soluble sugar content in both forests. In contrast, the response of respiration to temperature (Eo, Q10) was associated with Narea only. This relationship was strong at BRF, but much less clear at OFW. At OFW, the dominant canopy tree, D. cupressinum, displayed only a weak relationship between Eo and Narea. The other two species at OFW were shaded subcanopy trees with little range in leaf N content. The temperature response of respiration is related to metabolic activities such as growth, the maintenance of ion gradients, protein turnover and cellular repair, phloem loading and, in some cases, excess carbohydrate consumption, all of which are likely to vary with canopy depth (Penning de Vries 1975; De Visser et al. 1992). Because all measurements here were made on fully expanded leaves, it seems unlikely that growth processes were affected significantly, and therefore the observed results are probably related to maintenance processes. We conclude from these relationships that, while the instantaneous rate of respiration is related to substrate availability (or metabolic activity that covaries with substrate availability) and leaf N, the temperature response of respiration (as indicated by Eo) appears to be driven almost solely by leaf N and maintenance processes.

scaling of respiration from leaf to canopy

Assessing the impact of the observed variation in leaf-level respiratory physiology requires the results to be scaled to the stand level. On a leaf area basis, Rcan was similar for the two forest stands (2·2 mol m−2 over a 2-week period during the growing season). This reflects the fact that the OFW trees displayed higher R10 but lower Eo values, and experienced cooler minimum temperatures in summer than BRF species. When expressed on a mass basis, the greater SLA of leaves at BRF compared with OFW resulted in an Rcan for BRF which was twice that at OFW (Table 4). These values for the foliar component of stand respiration at BRF and OFW are comparable with those published previously for deciduous (Vose & Bolstad 1999) and coniferous (Ryan et al. 1994; Law et al. 2001) forests.

Our results demonstrate that forest structure and the location of individual leaf measurements can have a significant effect on the model outcome. When scaling our results to the stand level at BRF, we find that simplifying assumptions regarding the variation in respiration rate and the temperature response of respiration with canopy height underestimate actual carbon loss if based on lower- or mid-canopy physiology. We overestimate the actual carbon loss if calculations are based on upper-canopy physiology. Furthermore, when the effects of the observed variation in respiration rate and its temperature response with canopy depth are taken into consideration, differences between the distributed and constant physiology model calculations are greatest at higher temperatures. In general, the difference between model calculations was minimized at BRF when the constant physiology model was based on the mid-canopy leaves. Although this proved to be the case in a mixed deciduous forest with three dominant species and a closed canopy, our conclusions are quite different for OFW, a temperate mixed rainforest dominated by a single conifer (D. cupressinum) and with much less defined canopy layering. At OFW, the constant physiology calculation based on upper-canopy leaves was not greatly different from the distributed model, as D. cupressinum represents almost 75% of total leaf area, and the open canopy structure means that differences between upper- and mid-canopy leaves are small.

SLA changes with canopy depth (Ellsworth & Reich 1993; Hollinger 1996), and N is related to respiration rate (Reich et al. 1996; Ryan et al. 1996; Reich et al. 1998; Tissue et al. 2002), so it should be possible to scale foliar respiration from leaf to canopy level using mass- or N-based measures. When we scaled from leaf to canopy respiration at BRF using Rmass rather than Rarea, the simplified constant-characteristics model calculations for the three canopy levels were similar to the calculation where the actual distribution of leaf characteristics was taken into account (a maximum deviation of −5·7% for the upper-canopy constant-characteristics calculation). However, this simplified scaling based on Rmass did little to improve the constant physiology calculations at OFW, as in this forest stand Rmass decreased with depth in the canopy. Therefore, while it may be possible to use simplifying assumptions regarding respiratory carbon loss in more complex models containing ecosystem physiology, some caution must be used when selecting representative leaves to generate empirical relationships between respiration and temperature. This has important implications for accurate modelling of respiratory carbon fluxes from forests, particularly for models that assume single temperature responses among species and with depth in canopies. Given the importance of respiration in influencing net carbon sequestration in forest communities, such calculations will be an important underpinning of our understanding of the ecology of forests in response to environmental change.

Acknowledgements

The authors gratefully acknowledge the A.W. Mellon Foundation for providing the principal financial support for this research, and the University of Canterbury Erskine Fund for travel support to M.H.T. We thank the Black Rock Forest Consortium for access to the BRF field sites and for providing seed funding, accommodation and logistical support. We are indebted to Timberlands West Coast Ltd for providing access to the OFW site and for logistical support in the construction of the canopy access tower. The excellent technical assistance of Ardis Thompson, Brent Wehner and Nicole Lauren is gratefully acknowledged. We also thank Ray Leuning for providing the modified Arrhenius function for temperature response analysis.

Ancillary