Regenerating temperate forests under elevated CO

Regenerating temperate forests under elevated CO Blackwell Science Ltd 2 and nitrogen deposition: comparing biochemical and stomatal limitation of photosynthesis G. A. Bauer 1, G. M. Berntson 2 and F. A. Bazzaz 1 1 Harvard University, Cambridge, Massachusetts, USA; 2 University of New Hampshire, Durham, New Hampshire, USA Summary Author for correspondence: G. A. Bauer Tel: +1 617 496 8931 Fax: +1 617 495 9300 Email: gbauer@oeb.harvard.edu Received: 26 April 2001 Accepted: 19 July 2001 Photosynthesis of temperate trees growing in a competitive environment was investigated here in a factorial design of community composition (deciduous, coniferous and mixed species), carbon dioxide and nitrogen treatments. This study included seedlings of three deciduous (Betula alleghaniensis, Quercus rubra, Acer rubrum) and three coniferous (Pinus strobus, Picea rubens, Tsuga canadensis) species. Nitrogen partitioning changed significantly in response to the treatments. Higher area-based nitrogen concentrations (N a ) in the conifer needles, however, did not induce higher growth rates. Analysis of biochemical limitations of photosynthesis revealed that deciduous trees invested more nitrogen into carboxylation (V cmax ), electron transport (J max ) and P i regeneration capacity, but at much lower absolute concentrations for N a than conifers; consequently conifers maintained much higher rates for all three parameters. Deciduous species showed a strong stomatal limitation, whereas conifers maintained higher stomatal conductance at increasing mesophyll internal carbon dioxide concentration, indicating a much stronger assimilatory response to elevated carbon dioxide. Differences between the biochemical and stomatal response to elevated carbon dioxide and nitrogen indicate that within mixed stands, individual plant responses do not fully characterize community response. Key words: photosynthesis, elevated, nitrogen partitioning, temperate forests, open top chambers. New Phytologist (2001) 152: 249 266 Introduction Recent theoretical and empirical analyses have suggested that among the major terrestrial biomes, temperate forest ecosystems are a globally important sink for atmospheric carbon dioxide ( ) (Tans et al., 1990; Ciais et al., 1995; Tans et al., 1995), and may drive important feedbacks to future global warming (Woodwell & Mackenzie, 1995). However, predictions of the future capacity of temperate forests to act as long-term sinks for atmospheric remain tentative in part due to uncertainty in predicted climatic change (Houghton et al., 1996). Other aspects of global change, specifically atmospheric levels and nitrogen (N) deposition rates may also play a critical role in altering forest dynamics and productivity. By contrast to changes in climate, elevated atmospheric and N deposition are welldocumented and inevitable components of our future environment (Galloway et al., 1994; Keeling & Whorf, 1994; Holland et al., 1997; Vitousek et al., 1997). Current evidence of a -fertilization of forest growth since pre-industrialized times is inconclusive, with evidence for and against a significant enhancement of wood production in forests (Kienast & Luxmoore, 1988; LaMarche et al., 1984; Becker, 1989). Several recent reviews have suggested that the intrinsic capacity of trees to increase biomass production and thus net primary productivity of the terrestrial biosphere could average approx. 30% or more above current levels (Ceulemans & Mousseau, 1994; Wullschleger et al., 1995). A more detailed inspection of -driven growth responses reveals conflicting results of the magnitude, direction, New Phytologist (2001) 152: 249 266 www.newphytologist.com 249

250 and contribution of individual species or tree genera to altered net primary productivity. In summarizing research on the effects of doubling on trees Ceulemans & Mousseau (1994) found an average biomass increment of +38% and +63%, with a photosynthetic enhancement of +40% and +61% for conifers and deciduous trees, respectively. Another recent review, on responses to doubling, reported an average increase in biomass for conifers of +130% and for deciduous of +49%, with photosynthesis responding with an increment of +62% and +53% for conifers and deciduous, respectively (Saxe et al., 1998). These predictions demonstrate the uncertainty regarding the capacity of forest trees to sustain enhanced long-term net primary productivity in an elevated world if experimental data are to be scaled to forest ecosystems. Given the importance of N limitation in constraining the growth responses of plants in general and of temperate forests in particular to rising atmospheric levels (McGuire et al., 1995; Stitt & Krapp, 1999), then the potential synergism between the two becomes critical. Laboratory studies of individually grown plants have documented that increasing N supplies can significantly increase relative growth responses to elevated (Patterson & Flint, 1982; Brown & Higginbotham, 1986; Wong et al., 1992; Crabtree & Bazzaz, 1993). Thus, we could expect that increases in the availability of N from atmospheric deposition may increase tree growth in elevated. However, it is uncertain if temperate forest tree species are able to withstand long-term N limitation without any loss in productivity (Oren et al., 2001). This is especially important for reforestation and afforestation projects on marginal lands with poor conditions following land use change. Under such circumstances the critical determinant will be the success with which temperate trees can compete with soil microorganisms for the additional N (Berntson & Bazzaz, 1997; Aber et al., 1998; Zak et al., 2000). The efficiency with which N is used to maintain plant growth can vary considerably (Sheriff et al., 1995). High N use efficiency through partitioning of organic N into the different photosynthetic components has to be maintained in the leaves (Sage, 1994), the location where increased atmospheric directly influences the plant growth response. Therefore changes in photosynthetic N partitioning will affect canopy photosynthesis and in turn will alter C and N cycling in forest ecosystems (Hungate, 1999). In leaves, N investment in Rubisco, thylakoid-dependent RuBP regeneration and carbohydrate-dependent P i regeneration are the three major component processes of photosynthetic N use efficiency (PNUE). Several studies with individually grown plants have shown that these three processes scale with total leaf N (Harley et al., 1992; Cheng & Fuchigami, 2000; De Lucia & Thomas, 2000; Makino et al., 2000), which highlights the importance of N partitioning as a regulatory mechanism. However, no information is available on N partitioning for different species grown in competition for available soil N. Also, it has emerged that if elevated treatments significantly change photosynthetic N partitioning, leaf internal adjustments between the N pools occur proportionally (Medlyn et al., 1999). Consequently the resulting rates of carboxylation capacity ( Vc max ), electron transport capacity (J max ) and triosephosphate utilization (TPU), changed little with due to highly efficient N re-allocation. This could mean that photosynthetic N partitioning is independent of changes in atmospheric because plants can adjust their photosynthetic N investment. In addition it is not known how a competitive environment might affect the N partitioning of individual species. Reports of limited changes in N chemistry and acclimation of photosynthesis have emphasised the importance of stomatal vs biochemical acclimation to elevated (Morison, 1998; Medlyn et al., 2001). The main unresolved question is whether there is an independent stomatal response from photosynthesis under elevated (Sage, 1994; Drake et al., 1997). Leaf stomatal conductance (g s ) adjusts to changes in photosynthesis (Lodge et al., 2001; Šantrucek & Sage, 1996). If stomata adjust independently of photosynthesis, then any changes in photosynthetic N partitioning might not necessarily translate into larger effects on water relations. However, if stomatal acclimation under elevated is linked to photosynthetic acclimation, then there may be a strong nutritional control on leaf conductance, which in turn links leaf level processes to canopy and regional scale changes to the basic mechanism of photosynthetic N partitioning (Field et al., 1995). To investigate possible interactions between and N we have established an open top chamber study with regenerating communities consisting of red maple (Acer saccharum), yellow birch (Betula alleghaniensis), red oak (Quercus rubra), red spruce (Picea rubens), white pine (Pinus strobus) and hemlock (Tsuga canadensis). These species were subjected to a factorial treatment of, N and community composition. This study aims to investigate photosynthetic N partitioning and its effects on community productivity in a competitive environment. Materials and Methods Plant material and experimental design The study was located at Harvard University s Concord Field Station, about 25 km east of the Boston urban area in Bedford (MA, USA), in a cleared plot within a mixed forest dominated by Quercus rubra and Pinus strobus adjacent to a National Wildlife Refuge. There is also a secondary but significant amount of Acer rubrum, Betula lenta and Populus tremuloides present. All trees and topsoil (Oe, Oa, and A horizons) were removed 6 7 yr before the start of the experiment. During summer of 1997, all regenerating vegetation (mostly birch and sweet fern) was removed and the field was leveled with a bulldozer to minimize variation in local topography. The soil in the adjacent intact forest is a typical haplothord. The www.newphytologist.com New Phytologist (2001) 152: 249 266

251 remaining soil at the site is a well-drained, largely sandy deposit of glacial till with a characteristically high volume of large rocks and boulders. Beyond leveling the field, we have made no additional preparations, other than weeding, before transplanting the seedlings. Seeds of the six species (Betula alleghaniensis, Quercus rubra, Acer saccharum, Pinus strobus, Tsuga canadensis, Picea rubens) were obtained from a local seed supplier and were germinated in the open top chambers at their growth concentration (started on April 28, 1998). The seedlings were transplanted into the open top chambers with unrestricted rooting volume by mid-july of the same year. We have selected these species because the Concord Field Station is located in the Transition Hardwood zone between the Central Hardwood zone to the south and the Northern Hardwood/Spruce-Fir zone to the north ( Westveld et al., 1956), and they represent the most common indigenous species in the area and also promised the highest potential for successful germination. Nonchamber control plants were provided in four plots of the same diameter as the chambers and were planted randomly between the chambers. Two of these plots received the identical N amendment. Mortality in the nonchamber control plots was very high due to rodent herbivory. Therefore seedlings had to be replaced with backup plants from a glasshouse for 2 consecutive yr, which is the reason why no results are reported from these trees. Seedlings were planted in the chambers at a minimum density of 7.6 trees m 2. For the conifer and deciduous communities, this corresponds to eight plants per species and chamber, while for the mixed communities this corresponds to four plants per species and chamber. Given the small stature of the spruce and hemlock seedlings at the time of transplanting, entire patches of six to ten individuals instead of individual seedlings were transplanted. In total, the experiment began with 3456 trees transplanted into the open soil of the chambers. The chambers are based on standard open top chamber designs used in a number of studies on the effects of atmospheric pollution on plants grown in the field (Davis & Rogers, 1980; Drake et al., 1989; Leadley & Drake, 1993). The design was modified in order to build a large number of chambers with relative ease, using PVC piping for the frames and a UV-resistant plastic for cover, which absorbs about 25% of the incoming photosynthetic photon flux density (PPFD). The experiment consists of 36 open top chambers, each measuring 3.14 m 2 in basal area, and 2.5 m in height. The chambers are distributed in three blocks of 12 chambers within a 60 m 40 m clearing with a southern exposure, which provides equal irradiance for all the chambers between 09 : 00 am to 05 : 00 pm. Half the chambers were maintained at elevated atmospheric concentrations (a constant offset of 350 µl l 1 relative to the ambient chambers), and the other half at ambient (c. 370 µl l 1 ). In a factorial combination, half the chambers receive a N amendment of NH 4 NO 3, equivalent to a deposition rate of 50 kg ha 1 yr 1, the other half receives only ambient N deposition of about 8 kg ha 1 yr 1 in natural precipitation (Ollinger et al., 1993). The waterdissolved fertilizer is applied six times over the course of the growing season on the soil surface to simulate a rain event. The N treatment is combined with three different community types, which include pure conifer chambers, pure deciduous chambers and chambers with mixed (conifers & deciduous) communities. Thus, there are a total of three replicate chambers per N Community treatment. is supplied via a 13-ton liquid tank on site, and monitored via an IRGA (Horiba, PIR-2000, Horiba Instruments Inc., Irvine, CA, USA). The concentration in the elevated chambers is maintained passively by manual adjustment of flow controllers and injected into the blower air stream, which ventilates the chambers. By continuously evacuating sample lines, we independently measured concentrations from all 36 chambers once every 360 s which are logged with a computer. The IRGA is automatically recalibrated every 8 h. The control system is able to maintain a control band of ± c. 50 µl l 1 around the set point for the elevated chambers. Chamber temperature is typically < 2.0 C above ambient air temperatures. All 36 chambers were watered once a day during the peak dry season using a computer controlled irrigation system consisting of one overhead sprinkler per chamber. During the winter months the fumigation is turned off from mid-november to end of March the following spring. Plant growth measurements Because we wanted to investigate plant performance in a competitive environment, we abandoned any destructive growth or biomass measurements other than leaf level sampling. Therefore plant height was the only whole plant assessment of growth under the experimental treatments and was determineded 4, 8 and 13 months after transplanting. Leaf gas exchange and parameter estimation for the biochemical model of leaf photosynthesis Photosynthesis measurements were carried out for the deciduous seedlings at the time they had fully expanded leaves, beginning in mid-june of 2000. Needle development in conifers was slower, and their photosynthetic measurements were made in mid-july on recent foliage. Leaf gas exchange was measured with a Li-Cor 6400 photosynthesis system with a red-blue light source and a mixer (Li-Cor Inc., Lincoln, NB, USA). In each chamber we measured at least one leaf per species from the upper part of the seedling. Steady state light response measurements (P max ) at growth in combination with response curves (A/C i -curves) at saturating photosynthetic photon flux density (PPFD > 1500 µmol m 2 s 1 ) were carried out on all needles and leaves between 08 : 00 and 11 : 30 am. This sequence of measurements had the advantage that stomatal conductance, which generally responds much New Phytologist (2001) 152: 249 266 www.newphytologist.com

252 slower to changes in environmental conditions than photosynthesis, could be assessed independently of photosynthetic responses to elevated. This also allowed the A/C i -curves to be carried out in as short a time period as possible in order to avoid -induced changes in Rubisco activation state during the measurements. During the measurements, temperature in the leaf cuvette was kept between 27 and 30 C and rh was kept between 45 and 60%. Conifer needles used for gas exchange measurements were kept on ice and projected needle area was determined using a flatbed scanner to recalculate area-based assimilation rates. The needles and additional leaf punches from the deciduous trees were dried at 75 C and used to calculate specific leaf area (SLA, m 2 kg 1 ). The response curves were fitted using the commercially available software Photosyn Assistant (Dundee Scientific, Dundee, Scotland, UK). The program uses algorithms based on the biochemical model of leaf photosynthesis by Farquhar et al. (1980), with subsequent modifications by von Caemmerer & Farquhar (1981), Harley & Sharkey (1991) and Harley et al. (1992). According to the model the assimilation in a plant leaf can be mathematically described by saturation type kinetics, which account for the change in assimilation rate at increasing mesophyll internal concentration (C i ). The software provides initial estimates for the maximum rate of carboxylation by Rubisco (V cmax ) at low C i, the PPFD saturated rate of maximum electron transport ( J max ), and the P i limited rate of triose phosphate utilization (TPU). The model calculations are based on the assumption that one of these parameters is at a minimum and limits photosynthesis. The program then uses an iterative procedure to calculate rates for V cmax, J max and TPU from A/C i curves obtained from gas exchange measurements in the field. Limitation by TPU was ignored for those A/C i -curves as there was no obvious trend towards saturation at high C i. For each of the individual curves we used the leaf temperature as recorded in the field. Other parameters used in the model were fixed at the values listed by Harley et al. (1992) and Wullschleger (1993). For the calculation of maximum net photosynthesis at high C i and saturating light (A max ) we used a rectangular hyperbola following (Olsson & Leverenz, 1994). In the program a leastsquares fit is used to estimate carboxylation efficiency (CE), light and dark respiration (R) and A max. For conifers with no apparent curvature, CE and R were estimated from the lower part of the A/C i -curves and the maximum net photosynthesis rate measured in the field was used as A max. Biochemical analyses All the chemical analyses were carried out on a second set of conifer needles taken either from the same twig used for gas exchange measurements, or from the nearest twig of the same age class. For the deciduous trees subsamples for biochemical analysis were taken from the gas exchange leaves with a leaf corer. All leaf material for chemical analysis was immediately shock-frozen and kept in liquid N until analysed on either the same day or the following morning. The frozen leaf material was ground with a pre-chilled mortar and pestle to a fine powder and transferred immediately into a pre-weighed test tube. The frozen leaf powder (50 100 mg f. wt) was mixed with 1.5 ml of extraction buffer consisting of 100 mm Tris- HCl, 20 mm MgCl 2, 10 mm NaHCO 3, 1 mm NaEDTA, 5 mm DTT and 10% (v/v) glycerol (ph = 8.0 with NaOH). The test tube was mixed vigorously and the crude extract was immediately centrifuged at 15 000 g for 10 min. The supernatant was transferred into a new vial and used for analysis of soluble protein and Rubisco quantification. Soluble protein was determined with a microplate reader (Bio-Tek EL-309, Winooski, VT, USA) based on the method of Bradford, 1976) using Bovine Serum Albumine (BSA) as the standard ( Jones et al., 1989). Rubisco was analysed in the supernatant by SDS gel electrophoresis (4.75% stacking gel and 14% running gel). On each gel a prestained molecular weight standard (BIO RAD, Hercules, CA, USA) and a known amount of BSA were run parallel to the leaf samples. All gels were documented with a flat bed scanner. The amount of Rubisco was quantified by converting the pixel density of the leaf sample into mg of protein relative to the amount of BSA standard from the same gel. Chlorophyll was extracted from fresh leaf material with dimethylformamide (DMF) as the solvent (Porra et al., 1989), and chlorophyll concentrations were calculated according to the equations given by Wellburn (1994). Total N concentration was determined from dried leaf material with an elemental analyser (Fison 1500 N/C). Chlorophyll, Rubisco and soluble protein were converted to mass-based concentrations of N using their average N content of 6.31, 16.67 and 16.00%, respectively, as conversion factors. Carbohydrates were analysed from oven dried leaf material (75 C) following extraction of soluble sugars with ethanol (75%) and hydrolization of starch from the remaining pellet using HCl (1.1%) following the procedure of Oren et al. (1988). Both soluble sugars and starch were analysed colourimetrically as glucose equivalents with the anthrone reaction (Allen, 1989) using a microplate reader. Statistical analysis The experiment was designed as a randomized block design, with each of the three blocks containing one replicate chamber per community N treatment for a total of 12 chambers per block. Statistical analysis was performed using SPSS for Windows ( Version 9.0, SPSS Inc., Chicago, IL, USA). Main treatment effects were tested using multivariate analysis (GLM procedure) with community type, level, N level and block as main factors. Significant differences between species were further tested using ANOVA with the Scheffe posthoc test to account for unequal sample sizes. Data were transformed if necessary to comply with normal distribution and homogeneity of variances. www.newphytologist.com New Phytologist (2001) 152: 249 266

253 Fig. 1 Average height growth rate (cm d 1 ) for regenerating temperate tree species grown under elevated with either low N (a, c, e) or high N (b, d, f ) availability. Data in panels (a) and (b) represent the growth rates during the first 8 months, panels (c) and (d) represent months 9 through 12, and panels (e) and (f) represent the months 13 through 25 after transplanting the seedlings into the chambers. Closed columns, ambient ; open columns, elevated. YB, yellow birch; RO, red oak; RM, red maple; WP, white pine; RS, red spruce; HM, hemlock. Results Plant growth Plant growth changed significantly throughout the entire study period (Fig. 1). During the first 8 months plant height was more an indication of initial seedling height at the time of transplanting and average growth rates showed no significant effect of either (P = 0.176) or N (P = 0.511), with growth rates below 1 mm d 1. Average rates of height increment changed by more than an order of magnitude during the second growing season after transplanting with significant effects due to and N (P < 0.001). The N treatment seemed to have a particular effect. The difference in plant height between low N and high N for the deciduous trees was much larger than for the conifer species. Until the end of the third growing season (Fig. 1e,f ) growth rate remained in the same range. However, differences in plant height due to the N treatment had increased. While no significant main effect of the treatment was evident (Table 1), there were highly significant interaction terms for N species (P = 0.005) and species (P < 0.001), indicating that the overall and N response of height growth is different between the six tree species. Even though there was no significant effect on average height increment due to community composition over the entire study period (P = 0.261), several significant higher order interaction terms (Table 1) indicate first signs of competition in the experimental communities, mainly due to variability in the response of individual species to N. Carbon and nitrogen components in the leaves The leaf C : N ratio was significantly higher at low N supply regardless of concentration (Fig. 2, Table 2; P = 0.006), New Phytologist (2001) 152: 249 266 www.newphytologist.com

254 Table 1 ANOVA for the log-transformed averages of height growth rates (cm d 1 ) between the second and third census (August 2000), which corresponds to the data shown in Fig. 1(e,f). The r 2 for the overall model was 0.962 df MS F P Species 5 8.496 120.84 < 0.001*** 1 0.222 3.154 0.076 Nitrogen 1 9.13 129.86 < 0.001*** Block 2 0.419 5.965 0.003** Nitrogen 1 0.134 1.91 0.167 Species 5 0.333 4.743 < 0.001*** Block 2 0.441 6.267 0.002** Nitrogen Species 5 0.237 3.365 0.005** Nitrogen Block 2 0.008 1.196 0.303 Species Block 10 0.415 5.899 < 0.001*** Nitrogen Species 5 0.307 4.364 0.001** Nitrogen Block 2 0.107 1.515 0.22 Species Block 10 0.118 1.678 0.08 Nitrogen Species Block 10 0.337 4.798 < 0.001*** Error 1802 0.007 ***, sign. at 0.1% level; **, sign. at 1% level; *, sign. at 5% level. and was higher under elevated compared with ambient. Even though significant differences in the C : N ratio between species occurred, there was an overlap in the C : N ratio between deciduous and coniferous trees at both N supply levels. This also holds true for mass-based N concentrations (N m ), which changed significantly in response to and N. However, area-based leaf N concentrations (N a ) differed more strongly between deciduous and coniferous species, largely as a result of a strong effect on soluble and structural carbon, and large differences in specific leaf area (SLA) (Tables 2 and 3). By contrast, the N treatment did not show any significant effect on SLA or on carbon constituents, and like N m, there was no clear differentiation between the two functional groups with regard to soluble or structural carbon. Soluble protein, chlorophyll and Rubisco were the three main constituents of the leaf N pool, which were analysed separately (Table 3). In contrast to total leaf N there was no significant effect of the N treatment, and only chlorophyll decreased significantly due to increased (Table 2). Soluble protein Fig. 2 Leaf and needle averages for C : N ratio (a, d), mass based N concentration (b, e) and area based nitrogen concentration (c, f) of the six temperate tree species grown for 2.5 yr in either ambient (panels a c) or elevated (panels d f) and altered N availability. Closed columns, low N; open columns, high N. YB, yellow birch; RO, red oak; RM, red maple; WP, white pine; RS, red spruce; HM, hemlock. www.newphytologist.com New Phytologist (2001) 152: 249 266

255 Table 2 Multivariate analysis of variance for foliar C : N ratio, log-transformed mass-based nitrogen concentration (g N g 1 DW), logtransformed area-based nitrogen concentration (g N m 2 leaf area), specific leaf area (m 2 kg 1 ), soluble carbon (mg C g 1 d. wt) and structural carbon (mg C g 1 d. wt). All parameters were tested using a GLM procedure with species, community type, concentration, N level and block as fixed factors. All tested parameters had an r 2 between 0.879 and 0.997 for the overall model C : N ratio N m N a df MS F P MS F P MS F P Species 5 1233.24 9.31 < 0.0001*** 0.09 6.24 0.001** 0.12 6.27 < 0.0001*** Community 2 70.58 0.53 0.593 0.00 0.28 0.760 0.00 0.21 0.812 1 3262.57 24.64 < 0.0001*** 0.29 20.30 < 0.001*** 0.17 9.35 0.005** Nitrogen 1 1181.15 8.92 0.006** 0.10 6.93 0.014* 0.07 3.61 0.068 Block 2 329.66 2.49 0.102 0.02 1.30 0.289* 0.01 0.70 0.503 Block 2 491.86 3.71 0.038* 0.05 3.35 0.050 0.03 1.42 0.260 Error 27 132.42 0.01 0.02 SLA Soluble carbon Structural carbon df MS F P MS F P MS F P Species 5 174.13 39.43 < 0.001*** 4121.82 14.76 < 0.001*** 8134.93 4.71 0.003** Community 2 0.01 0.00 0.997 426.08 1.53 0.236 251.61 0.15 0.864 1 13.66 3.09 0.090 13736.11 49.20 0.001*** 25592.45 14.92 0.001** Nitrogen 1 8.34 1.89 0.181 875.15 3.13 0.088 1678.59 0.98 0.331 Block 2 1.12 0.25 0.777 477.96 1.71 0.200 1245.96 0.73 0.493 Species 4 14.20 3.22 0.028* 252.92 0.91 0.474 976.59 0.57 0.687 Block 2 10.04 2.27 0.122 1327.43 4.75 0.017* 2164.10 1.26 0.299 Species Community 3 10.85 2.46 0.085 978.43 3.50 0.029* 738.92 0.43 0.733 Species Nitrogen 4 16.36 3.71 0.016* 547.11 1.96 0.129 2204.66 1.29 0.300 Speciies Block 8 11.90 2.70 0.025* 147.85 0.53 0.824 947.91 0.55 0.806 Community Block 4 14.10 3.19 0.029* 257.76 0.92 0.465 493.07 0.29 0.883 Error 27 4.42 279.20 1715.01 Chlorophyll Rubisco Soluble protein df MS F P MS F P MS F P Species 5 0.07 4.12 0.011* 0.34 4.36 0.009** 0.03 1.40 0.270 Community 2 0.00 0.16 0.854 0.04 0.50 0.613 0.02 1.14 0.342 1 0.08 5.06 0.037* 0.04 0.48 0.497 0.06 2.86 0.108 Nitrogen 1 0.01 0.55 0.469 0.02 0.21 0.649 0.01 0.37 0.548 Block 2 0.00 0.13 0.876 0.32 4.11 0.034* 0.00 0.03 0.967 Species Block 9 0.01 0.53 0.836 0.30 3.90 0.007** 0.06 2.85 0.028* Error 18 0.02 0.08 0.02 Asterisks indicate significance values. Table 3 Average values for specific leaf area (SLA in m 2 kg 1 ), soluble carbon and structural carbon (both in mg C g 1 d. wt), soluble protein (g N m 2 ), chlorophyll and Rubisco (both in mg N m 2 ). Individual tree species were grown for 2.5 yr in either ambient or elevated concentrations. Values in the same column followed by a different letter are significantly different (Scheffe, P = 0.05) SLA (m 2 kg 1 ) Soluble carbon (mg C g 1 d. wt) Structural carbon (mg C g 1 d. wt) Ambient Elevated Ambient Elevated Ambient Elevated Yellow birch 15.90 ± 2.04a 15.29 ± 4.45a 116.8 ± 18.2c 144.1 ± 12.1a 346.0 ± 23.5a 308.5 ± 19.8a Red maple 15.49 ± 3.29a 14.43 ± 3.97a 72.2 ± 17.4a 100.3 ± 24.1b 394.5 ± 52.3a 381.6 ± 31.8c Red oak 17.42 ± 2.69a 13.84 ± 2.63a 107.4 ± 25.9bc 127.5 ± 34.1ab 362.3 ± 37.0a 329.5 ± 34.1ab Hemlock n.a. 1 6.70 ± 0.06b n.a. 126.5 ± 20.5ab n.a. 370.2 ± 34.4bc Red spruce 6.79 ± 0.13b 6.81 ± 0.15b 87.8 ± 14.3ab 129.2 ± 14.8ab 404.8 ± 53.6a 334.2 ± 18.3ab White pine 6.85 ± 0.15b 6.71 ± 0.10b 79.1 ± 9.5a 105.9 ± 18.2ab 389.5 ± 47.6a 355.2 ± 36.1abc Soluble protein (g N m 2 ) Chlorophyll (mg N m 2 ) Rubisco (mg N m 2 ) Ambient Elevated Ambient Elevated Ambient Elevated Yellow birch 0.51 ± 0.21a 0.48 ± 0.19a 31.85 ± 6.33a 29.49 ± 9.40a 31.59 ± 23.95a 24.16 ± 17.95a Red maple 0.55 ± 0.10a 0.51 ± 0.18ab 26.49 ± 8.80ab 21.47 ± 8.56ab 11.29 ± 8.75ab 7.79 ± 2.85a Red oak 0.41 ± 0.20a 0.51 ± 0.24ab 25.08 ± 8.96abc 22.00 ± 7.85ab 9.32 ± 8.90b 11.10 ± 20.75a Hemlock n.a. 0.78 ± 0.14b n.a. 23.10 ± 7.33ab n.a. 25.47 ± 24.48a Red spruce 0.57 ± 0.22a 0.34 ± 0.15a 19.70 ± 3.79bc 15.21 ± 5.00b 19.86 ± 10.31ab 14.58 ± 8.26a White pine 0.60 ± 0.29a 0.44 ± 0.25a 16.64 ± 3.43c 15.34 ± 3.69b 18.43 ± 6.97ab 13.68 ± 5.04a New Phytologist (2001) 152: 249 266 www.newphytologist.com

256 Fig. 3 Area based concentrations of soluble protein (a), Rubisco (b), and chlorophyll (c) and percentage of total N invested into soluble protein (d), Rubisco (e) and chlorophyll (f) in relation to total N on area basis. Solid lines in panels a, c and f represent the only significant linear regressions (r 2 = 0.172 0.615). Closed circles, conifers; open circles, deciduous. concentrations were very conservative due to the large variation in SLA, which obscured significant species differences. Mass-based concentrations are influenced by the accumulation of soluble carbohydrates (Table 3) so photosynthetic N partitioning was expressed as area-based concentrations. The significantly higher N a in conifers did not result in higher concentrations of soluble protein, chlorophyll or Rubisco. In absolute terms, the three major N constituents varied over the same range but at much lower total leaf N concentrations in the deciduous species (Fig. 3a c). Both soluble protein and chlorophyll increased with increasing N a. The relationships in Fig. 3(a) (c) also indicate significant differences in the slopes, thus providing evidence for quantitative as well as qualitative differences in photosynthetic N investment among plant functional types. This is also a strong indication, that protein : N, chlorophyll : N and Rubisco : N ratios may not be constant between C 3 species (Evans, 1989). Rubisco concentration was not related to total N a (Fig. 3b,e). In particular, for some of the deciduous trees, Rubisco concentration increased up to about 40 mg N m 2 at almost constant leaf N. The investment of N into Rubisco, protein and chlorophyll was significantly different between the species (Table 4). The species-specific differences in leaf N also caused N investment in Rubisco to be significantly affected by the N treatment. However, this effect was driven largely by high Rubisco concentrations in yellow birch and very low concentrations in red oak (Table 5). Based on protein nitrogen, red spruce and white pine allocated more N into Rubisco than red oak and red maple. For all the species, N investment into soluble protein, Rubisco and chlorophyll showed a strong negative relationship with increasing N a (Fig. 3d f ). Photosynthetic response The response curves showed no common pattern with regard to the experimental treatments or among species www.newphytologist.com New Phytologist (2001) 152: 249 266

257 Table 4 Multivariate analysis of variance for leaf area-based concentrations of soluble leaf protein (g N m 2 ), Rubisco (mg N m 2 ) and chlorophyll (mg N m 2 ). All parameters were log-transformed to achieve normality and tested using a GLM procedure with species, community type, concentration, nitrogen level and block as fixed factors. All tested parameters had an r 2 between 0.980 and 0.998 for the overall model. Only the statistically significant higher order interaction terms are listed N in Rubisco (%) N in Soluble protein (%) N in Chlorophyll (%) df MS F P MS F I MS F P Species 5 0.42 8.08 0.001*** 761.0 7.61 0.001*** 0.53 26.76 < 0.001*** Community 2 0.14 2.64 0.102 121.5 1.22 0.323 0.00 0.08 0.922 1 0.04 0.72 0.408 63.4 0.63 0.437 0.00 0.05 0.827 Nitrogen 1 0.29 5.59 0.031* 32.2 0.32 0.578 0.00 0.02 0.884 Block 2 0.24 4.61 0.026* 102.9 1.03 0.38 0.01 0.61 0.555 Species Block 9 0.16 3.11 0.023* 682.4 6.83 < 0.001*** 0.03 1.29 0.314 Community Nitrogen 2 0.02 0.44 0.649 47.2 0.47 0.632 0.08 3.99 0.039* Nitrogen Block 2 0.02 0.32 0.733 506.5 5.07 0.02* 0.07 3.31 0.063 Species Nitrogen 3 0.07 1.33 0.299 734.3 7.34 0.00** 0.05 2.60 0.088 Species Block 5 0.06 1.12 0.388 170.0 1.70 0.192 0.06 3.22 0.034* Species Nitrogen Block 4 0.03 0.55 0.699 166.2 1.66 0.208 0.08 3.84 0.023* Error 16 0.05 100.0 0.02 Asterisks indicate significance values. Table 5 Average percent N investment into soluble protein, chlorophyll and Rubisco, and average percent of soluble protein allocated to Rubisco. Individual tree species were grown for 2.5 yr in either ambient or elevated concentrations. Values in the same column followed by a different letter are significantly different (Scheffe, P = 0.05) N in soluble protein (%) N in chlorophyll (%) Low N High N Low N High N Yellow birch 61.24 ± 5.55ab 52.22 ± 4.17a 3.31 ± 0.14b 3.60 ± 0.17a Red maple 75.17 ± 5.08a 66.97 ± 6.94a 3.77 ± 0.32b 3.52 ± 0.24a Red oak 49.48 ± 4.28b 60.07 ± 6.91a 2.81 ± 0.14b 3.05 ± 0.18a Hemlock na 40.94 ± 9.01a na 1.74 ± 0.19b Red spruce 37.14 ± 4.59a 35.51 ± 3.17a 1.53 ± 0.11a 1.43 ± 0.14b White pine 36.38 ± 5.30a 40.83 ± 8.03a 1.20 ± 0.10a 1.17 ± 0.09b N in Rubisco (%) Protein in Rubisco (%) Low N High N Low N High N Yellow birch 3.03 ± 0.71b 2.72 ± 0.52b 5.29 ± 1.25a 5.48 ± 1.11b Red maple 1.38 ± 0.40ab 1.09 ± 0.20ab 1.45 ± 0.40a 1.50 ± 0.24ab Red oak 0.97 ± 0.30a 0.82 ± 0.35a 1.88 ± 0.53a 0.98 ± 0.39a Hemlock na 1.19 ± 0.60ab na 2.91 ± 1.11ab Red spruce 1.61 ± 0.35ab 1.34 ± 0.17ab 3.94 ± 0.48a 3.82 ± 0.54b White pine 1.25 ± 0.13ab 1.08 ± 0.09ab 3.57 ± 0.42a 3.63 ± 0.76b na, not available. (Fig. 4). In yellow birch net photosynthesis under elevated was intermediate to the two ambient treatments, with maximum rates in the high N trees. Red oak and red maple had higher net photosynthetic rates under elevated compared with ambient, but with no effect of the two N treatments. All three deciduous species differed strongly in terms of initial slope and degree of saturation at high intercellular concentration (C i ). The two conifers reached similar rates of net photosynthesis at high C i to deciduous species, but showed no trend for net photosynthesis to saturate. At low values of C i (< 100 ppm) the ratio of net photosynthesis at elevated to ambient for the deciduous species was close to unity, and remained below one at increasing C i. The only exception was yellow birch, with photosynthetic enhancement greater than one under enrichment in the low N treatment. The ratio for the two conifers generally remained close to unity. Following the interpretation of Sage (1994) the change in this ratio with increasing C i in the deciduous species indicates that there is little or no decrease in Rubisco activity, while values less than unity at high C i indicate that there is no adjustment in the P i regeneration capacity. In conifers the ratio is less than unity or low C i but close to unity at high C i suggesting a decrease in Rubisco and an increase in P i regeneration to maximize the photosynthetic nitrogen use efficiency. New Phytologist (2001) 152: 249 266 www.newphytologist.com

258 Fig. 4 Average A/C i -curves for five temperate tree species grown for 2.5 yr in a factorial combination of altered N and concentration. For each species the lower part of the figure shows the ratio of apparent net photosynthesis at elevated vs ambient. Closed symbols, low N; open symbols, high N. Closed squares, elevated low N; open squares, ambient low N; closed circles, elevated high N; open circles, ambient high N. Biochemical limitations There was no significant effect of N treatment (Table 6) on photosynthetic parameters, which were calculated using the biochemical model of C 3 photosynthesis (Farquhar et al., 1980; Harley et al., 1992). There was a significant species N interaction term for J max (P = 0.049) as well as for community N (P = 0.004) and species community (P = 0.031) for CE. This indicated that any effect of N supply on leaf N partitioning is species specific, and that partitioning within a species changes with neighbouring plants. Although conifers had higher rates of net photosynthesis (Fig. 4), their limited curvature in the A/C i response curves did not justify calculating an asymptote to A at high C i using a nonlinear approach, while the more saturating curves for the deciduous species were extrapolated until A reached a steady state. In contrast, net photosynthesis under saturating light and growth C i (P max ) was significantly different between species and higher under elevated (Fig. 5b). Carboxylation efficiency (CE, i.e. the initial slope of the curve up to a C i of c. 200 ppm in Fig. 4), Vc max as well as www.newphytologist.com New Phytologist (2001) 152: 249 266

259 Table 6 Multivariate analysis of variance for photosynthetic rates and the limiting parameters fitted with the biochemical model of photosynthesis in C 3 plants. All parameters, except for triosephosphate utilization (TPU), were log-transformed to achieve normality and tested using a GLM procedure with species, community type, concentration, nitrogen level and block as fixed factors. All tested parameters had an r 2 between 0.868 and 0.998 for the overall model. Only the statistically significant higher order interaction terms are listed A max P max CE df MS F P MS F P MS F P Species 4 0.028 1.02 0.413 0.155 4.80 0.004** 0.059 2.21 0.092 Community 2 0.004 0.14 0.869 0.023 0.72 0.496 0.003 0.10 0.903 1 0.039 1.45 0.238 0.108 3.34 0.078 0.388 14.44 0.001** Nitrogen 1 0.013 0.50 0.485 0.002 0.08 0.786 0.069 2.57 0.119 Block 2 0.034 1.27 0.297 0.227 7.01 0.003** 0.041 1.55 0.230 Species Community 3 0.002 0.09 0.965 0.010 0.32 0.813 0.091 3.39 0.031* Community Nitrogen 1 0.042 1.55 0.223 0.004 0.13 0.720 0.265 9.87 0.004** Community Block 3 0.085 3.14 0.040* 0.048 1.49 0.237 0.047 1.75 0.177 Error 30 0.027 0.032 0.027 J max Vc max TPU df MS F P MS F P MS F P Species 4 0.124 7.12 < 0.001*** 0.111 3.63 0.016* 0.17 6.02 0.001** Community 2 0.005 0.31 0.735 0.006 0.20 0.819 0.015 0.52 0.6 1 0.2 11.46 0.002** 0.144 4.72 0.038* 0.103 3.65 0.066 Nitrogen 1 0.033 1.87 0.181 0.005 0.17 0.681 0.015 0.55 0.466 Block 2 0.014 0.83 0.447 0.013 0.43 0.657 0.003 0.09 0.911 Species Nitrogen 4 0.047 2.71 0.049* 0.060 1.95 0.128 0.071 2.51 0.063 Community Block 3 0.05 2.89 0.052 0.050 1.62 0.206 0.141 5.00 0.006** Error 30 0.017 0.031 0.028 Asterisks indicate significance values. Fig. 5 Average rate of saturated (A max ) and light saturated (P max ) rate of net photosynthesis and the results of the curve fit analysis for CE, J max, Vc max and triosephosphate utilization (TPU). Tree species were grown for 2.5 yr in a factorial combination of altered and N availability. Closed columns, ambient ; open columns, elevated. YB, yellow birch; RO, red oak; RM, red maple; WP, white pine; RS, red spruce; HM, hemlock. New Phytologist (2001) 152: 249 266 www.newphytologist.com

260 J max were significantly affected by the concentration (Table 6). A significant species effect was only apparent for Vc max, J max and the rate of triose phosphate utilization (TPU). The results, summarized in Fig. 5 indicate some photosynthetic acclimation to elevated following significant changes in N allocation. The two conifers sustain similar rates of net photosynthesis to the deciduous species, by maintaining much higher rates for Vc max, J max and TPU (Fig. 6a c). This represents a significant difference in the way N is partitioned between photosynthetic processes (Medlyn et al., 1999). There is a significant increasing trend in V cmax, J max and TPU with N a (Fig. 7a c). Even though conifers follow this trend, they show a much larger variation in the three rates at high N. When related to protein nitrogen however, the deciduous species show an entirely different relationship, with Vc max, J max and TPU slightly decreasing as protein N increases, whereas the conifers maintain a positive slope (Fig. 7d f ) and sustain higher turnover rates for protein nitrogen. Stomatal response At low values of C i (50 200 ppm) the C i : C a ratio in the deciduous species decreased slightly from about 1.2 to about 0.7 (Fig. 8a). In the conifer needles the decrease in the C i : C a ratio was stronger for the same change in external with an average decline from about 2.3 to the same value of 0.7 as in the deciduous leaves (Fig. 8b). This indicates that in both groups the C i : C a ratio is maintained conservatively at a level similar to other species (Wong et al., 1979). The limited change in the C i : C a ratio of deciduous leaves is accompanied by a decline in stomatal conductance with C i (Fig. 9a c). This indicates that the initial change in A (Fig. 4a c) is a result of stomatal closure, rather than increased assimilation. In the conifers, stomatal conductance remains more or less constant with C i, indicating no stomatal limitation, but a strong assimilatory effect as C i increases. The differences in stomatal sensitivity between the two groups coincide with higher rates of J max, Vc max and TPU in the conifers, while the deciduous trees maintain a higher carboxylation efficiency (Fig. 5c) which balances the demand for inorganic C. Fig. 6 Relationship between Vc max, J max and triosephosphate utilization (TPU) with A max for coniferous (closed circles) and deciduous (open circles) tree species grown for 2.5 yr in a factorial combination of altered and N availability. The solid lines represent linear regression lines (all regressions significant at P 0.001) for the two functional groups (conifers and deciduous) with r 2 between 0.645 and 0.900 for the conifers and between 0.454 and 0.599 for the deciduous species. Discussion This experiment on multispecies assemblages of temperate trees provided a factorial combination of community composition, and N treatments and is the first experiment of its kind in which tree responses to predicted environmental changes are tested at community level (Körner, 1996). The communities contained relatively slow-growing (e.g. hemlock, red spruce) and fast-growing trees (yellow birch, red oak), both at low N and high N supply. In general there was little evidence for a relationship between growth rate (based on height measurements) and N use efficiency (Pons et al., 1994; Poorter & Evans, 1998). After 2.5 yr of exposure, www.newphytologist.com New Phytologist (2001) 152: 249 266

261 Fig. 7 Relationship between the three major biochemical limitations (V cmax, J max, triosephosphate utilization (TPU) ) and area based total N (panels a c) and area based protein nitrogen (panels e f) for coniferous (closed circles) and deciduous (open circles) temperate tree species. The solid lines represent those linear regressions which showed a significant relationship between two parameters for each of the two functional groups (P 0.05). The dashed lines in panels a c represent linear regressions (P 0.001) over the entire data set with r 2 of 0.572 (V cmax ), 0.678 (J max ) and 0.624 (TPU). Fig. 8 Change in the C i : C a ratio with increasing external concentration for deciduous trees (yellow birch, red oak and red maple; panel a) and coniferous trees (white pine, red spruce; panel b). New Phytologist (2001) 152: 249 266 www.newphytologist.com

262 Fig. 9 Relationship between stomatal conductance and mesophyll internal concentration (C i ) for temperate tree species grown for 2.5 yr in a factorial combination of altered and N availability. Each data point represents the average of 3 5 measurements on individual leaves. Closed squares, elevated low N; open squares, ambient low N; closed circles, elevated high N; open circles, ambient high N. Deciduous Conifers Growth Heigth increase * High Low SLA High Low Nitrogen N m + Identical N a + Low High Carbon Soluble carbon + Identical Structural carbon + Identical C : N ratio */+ Low High % N partitioning into Protein Identical Chlorophyll High Low Rubisco * High Low Biochemical limitations V cmax + Low High J max + Low High TPU Low High Photosynthesis (A max ) Identical Change in stom. conductance with increasing C i Strong Decrease Constant or Increase Table 7 Summary of the analysed plant responses after long-term exposure (> 2.5 yr) to elevated atmospheric and N fertilization based on functional groups (deciduous and conifers). Symxbols following each physiological trait represent either a stronger response to N fertilization (*), a stronger response to the treatment (+), or no significant change ( ). The arrow pointing from low to high represents a gradual increase in a physiological parameter across the six different tree species with hemlock < red spruce < white pine < red maple < red oak < yellow birch www.newphytologist.com New Phytologist (2001) 152: 249 266

263 it was evident that the growth response of the trees was more affected by N than by (Table 7), with deciduous trees showing a greater difference in plant height between low and high N than the conifers. The analysis of leaf C and N constituents confirms recent findings of altered chemical composition of tree foliage under elevated (Tissue et al., 1993; Rey & Jarvis, 1998). The results presented here are opposite to reports in recent meta-analyses (Curtis, 1996; Curtis & Wang, 1998; Medlyn et al., 1999), which reported that under enrichment a compensatory decrease in SLA leads to unchanging N a and chlorophyll concentration. In our study rooting volume is unrestricted and trees can use the surplus of carbohydrates to promote below-ground C allocation, thus avoiding a dilution effect of -induced growth on N m. Despite unrestricted rooting volume, nonstructural carbohydrates accumulated in the foliage. The widely accepted view is that elevated leads to decreased N m, which is counterbalanced by changes in SLA thus maintaining N a constant. By contrast we found that both N m and N a changed significantly due to but SLA did not. The general interpretation of lower N m and constant N a is that it is a result of altered leaf morphology (leaf density in terms of area per mass) and not a result of altered plant internal allocation of N (Medlyn, 1996; Curtis & Wang, 1998). Our study shows that when grown in a competitive environment with unrestricted rooting volume, significant changes in N partitioning occur and exert a strong influence on photosynthetic performance (Stitt & Krapp, 1999). Even though soluble protein, Rubisco and chlorophyll increased proportionally with N a, deciduous trees had a significantly different ratio of protein : N, Rubisco : N and chlorophyll : N than conifers. In absolute terms, deciduous trees invested more N into each individual component at a lower N a, while conifers with a higher N a had less N allocated to all three N pools (Table 7). The results for N partitioning provide evidence for qualitative as well as quantitative differences in N allocation in individual tree species. The fraction of N partitioning into the major components decreased across all species with increasing N a (Fig. 3d f ). This negative relationship is a further indication of altered plant N investment into photosynthesis at increased N availability. It also appears that high leaf N concentrations are not necessarily conducive for a higher N investment into major photosynthetic components, nor do higher N concentrations in the conifer needles coincide with higher growth rates (Pons et al., 1994; Poorter & Evans, 1998). Nitrogen investment into soluble protein ranged between 35 and 75% of total leaf N (Table 5), indicating that in trees other organic N components such as amino acids could be an important medium for storing N (Stitt & Krapp, 1999; Bauer et al., 2000). In addition, the species-specific variability prohibited clear evidence for a lower N investment in soluble protein and in Rubisco under elevated. These results are similar to observations by De Lucia & Thomas (2000), who also found a significant species effect but no significant treatment effect on Rubisco or soluble protein. Two important observations emerge. First, the N treatment caused a significant reduction of N investment into Rubisco with trees at high N investing less N into Rubisco (Table 5). Second and in contrast to De Lucia & Thomas (2000), a high amount of N allocated to Rubisco did not necessarily coincide with a higher proportion of soluble protein invested in Rubisco. This was evident in particular for the conifers, which on average invested about 1% of their total leaf N in Rubisco, but allocated almost 4% of their protein N to Rubisco, while this ratio was much lower for red maple and red oak. It also appears, therefore, that in contrast to other observations, Rubisco does not constitute the major enzyme in terms of N costs in the foliage of trees. Based on this study and the analysis by Tissue et al. (1993), it rather seems that in trees Rubisco constitutes less than 5% of total N, which is about 2 3 times lower than for warm-temperate forest species (Hikosaka & Hirose, 2000) and about 5 7 times lower compared to annual and crop plants (Evans, 1989; Makino et al., 2000). The model of biochemical limitations showed only a limited adjustment between carboxylation capacity, electron transport capacity and P i regeneration capacity in response to elevated (Medlyn et al., 2001). In general, any photosynthetic adjustments occurred in response to higher N availability (Table 7). However, the limited -induced changes in biochemical limitations were offset by a higher partitioning of N into either Rubisco or chlorophyll. Therefore the degree of protein N allocated to one particular biochemical component appears more important than their turnover rate. For a given A max, deciduous trees had much lower rates for Vc max, J max and TPU than the conifers, but deciduous trees had significantly more N allocated to chlorophyll. This was also confirmed by the relationship of the respective rates with either N a or soluble protein N (Fig. 7). While there was a positive slope of Vc max, J max and TPU with N a, only in conifers was there a positive slope against protein. Increased atmospheric exerts a strong influence on stomatal conductance and water use efficiency due to a direct effect of C i on stomatal opening (Assmann, 1999). Changes in net photosynthesis under elevated could be due either to a true increase in assimilation or to a change in stomatal conductance, limiting the diffusion of into the substomatal cavities. Therefore knowledge of the change in the C i : C a ratio is necessary in order to separate a stomatal response from an assimilatory response under elevated. In addition, even with no apparent photosynthetic acclimation in response to elevated or N, the stomatal response was much more dramatic and seems to be a key component in plant response to climate change (Medlyn et al., 2001). We found evidence for a stomatal acclimation, but it remains unknown whether the stomatal responses are independent of the biochemical responses. The deciduous trees, but not the conifers, showed a variable decrease in stomatal conductance New Phytologist (2001) 152: 249 266 www.newphytologist.com