1. To test hypotheses concerning adaptation and acclimation of tree species to shaded habitats we determined the growth, biomass partitioning and morphology of seedlings of nine near-boreal tree species in high- and low-light greenhouse environment (25 and 5% of full sunlight, respectively), comparable to sunlit gap and shaded microsites in boreal forests. The species differ widely in shade tolerance, seed size and leaf life span.
2. In low light, all species allocated proportionally more biomass to stems and less to roots, but the same to foliage, compared with the high-light environment. At a common size, all species had finer leaf morphology (higher specific leaf area, SLA) but coarser root morphology (lower specific root length, SRL) in low than high light. From a whole plant perspective, all species enhanced leaf area per unit plant mass (leaf area ratio, LAR) in low light and root length per unit plant mass (root length ratio, RLR) in high light.
3. Shade-intolerant deciduous species had higher RGR, SLA and SRL than larger seeded evergreens: ranking from Populus, Betula and Larix spp., then to five evergreen Pinus, Picea and Thuja spp., which were generally comparable in these traits. There were no changes in growth rankings of species between high- and low-light environments, nor consistent differences among species in biomass partitioning. Hence, species differences in leaf and root morphology (SLA, SRL) drove whole plant patterns, such as Populus, Betula and Larix had greater total leaf area and root length per unit plant mass (LAR and RLR, respectively) than the evergreens. Interspecific variation in RGR in both high and low light was positively correlated (r≈ 0·9) with SLA, SRL, LAR and RLR, and negatively correlated (r≈–0·9) to seed mass and leaf life span.
4. These data suggest that SLA, SRL, NAR and RGR are closely associated with variation in life-history traits and that variation in leaf and root structure more strongly influences patterns of RGR among species and light environments than does biomass partitioning.
Differences among species in intrinsic traits and their plasticity may reflect interspecific microhabitat differentiation. Variation in physiological and morphological traits among tree species has been related to their differences in regeneration habitat conditions, including light availability (e.g. Loach 1970; Bazzaz 1979; Pompa & Bongers 1988; Walters, Kruger & Reich 1993a; Jones et al. 1994; Kitajima 1994; Walters & Reich 1996). One facet of such work asks which combination of traits (metabolic, morphological and/or allocational) enhances relative growth rate (RGR) in high light and/or deep shade, and how these traits vary in plants acclimated to contrasting light conditions. A related issue concerns whether rankings of RGR reverse between shade intolerant and tolerant species when compared at high vs low light levels (e.g. Kitajima 1994; Walters & Reich 1996); i.e. do shade-tolerant species outgrow intolerants in deep shade? Another question is whether a set of leaf and plant level attributes, such as leaf life span, RGR, leaf weight ratio (LWR), specific leaf area (SLA) and leaf area ratio (LAR) (Poorter, Remkes & Lambers 1990; Lambers & Poorter 1992; Reich, Walters & Ellsworth 1992; Walters et al. 1993a) vary systematically among species adapted to high- vs low-resource habitats.
In this paper we compare growth, tissue structure and biomass distribution among young seedlings of nine common North American cold-temperate and boreal species (hereafter boreal) which vary in habitat affinity. The nine species overlap in their range in the northern Great Lakes States, northern New England and in southern Ontario and Quebec. Species are listed in Table 1 in the general order of successional niche, seed size, shade tolerance and leaf life span. These species rank differences also correspond roughly to hardwood vs conifer and deciduous vs evergreen groupings. Successional niche and/or shade tolerance (often based upon the habitat distribution of species) have been examined in relation to variation in ecophysiological and life-history traits (Grime & Hunt 1975; Bazzaz 1979; Chapin & Tryon 1983; Reich, Ellsworth & Uhl 1995; Walters & Reich 1996). Two other traits often related to successional habit, shade tolerance and growth rate, are seed size and leaf life span (Reich et al. 1992; Westoby, Jurado & Leishman 1992; Jones et al. 1994). Seedling RGR is often (but not always) inversely related to seed size (Shipley & Peters 1990b; Huante, Rincon & Gavito 1992; Westoby et al. 1992). Chapin (1993) hypothesized that seedlings of large-seeded species have been observed to have low RGR compared with fast-growing species because comparisons are often made of plants differing in size and RGR generally decreases with plant size in all species, but to our knowledge there are insufficient data available to test that question. Based on a literature survey, it seems that species with short leaf life spans tend to have higher SLA, LAR and RGR than those with long leaf life spans (Reich et al. 1992). Seed size and leaf life span do not directly affect whole-plant carbon gain (especially for seedlings during a period where leaf turnover is nil) but may covary among species in relation with physiological and structural traits that influence RGR, such as LAR and photosynthetic rate (Reich et al. 1992). Based on the above set of patterns, we hypothesized that interspecific variation in RGR and its determinates should be related to variation in successional niche, shade tolerance, seed size and leaf life span.
Table 1. . Study species, seed mass, leaf life span and shade tolerance. Mean seed mass determined from mass of 100 seeds. Leaf life span based on field observations of trees (Reich, Walters & Ellsworth 1997 and unpublished data). Shade tolerance rankings based on USDA (1990). Because the combined rankings of seed mass, leaf life span and shade tolerance are comparable for the evergreen conifers except P. banksiana, they are listed in alphabetical order
The influence of light on growth, allocation and tissue structure was examined by growing seedlings of the nine species in contrasting conditions of relatively high vs low light (25 vs 5% of full sun light). By sequentially harvesting subsets of seedlings in all treatments, we were better able to separate ontogenetic from species and treatment effects on measured plant traits than in single harvest studies (Coleman, McConnaughay & Ackerly 1994). Old-growth sub-boreal transition forests in north-eastern MN are noted for a relatively sparse canopy with numerous small gaps caused by tree-falls and/or patches of thin soil on rocks. Light levels of 25 and 5% are well within the range of light microhabitats that commonly occur in that region, which vary typically from 2 to 45% (J. L. Machado & P. B. Reich, unpublished data). Thus light levels used in this study are representative of integrated long-term averages for understorey and small gap conditions, respectively, in thin-soil, sparse-canopied boreal forests, where understorey light values are often greater than beneath temperate mixed mesophytic forest (Canham et al. 1990; Ellsworth & Reich 1992).
This study builds on and extends a prior study by our group (Walters et al. 1993a,b). The present study contrasted nine rather than three species in high vs low light, included species with a broader array of plant traits, utilized lower light regimes, better accounted for ontogenetic drift and included measures of specific root length not made in the earlier study. Moreover, as will be shown, the results and conclusions were different in a number of ways, not surprisingly given the differences in overall design. Thus, the study complements and extends the prior work.
In this paper we ask several broad questions. How do species differ in biomass partitioning, tissue structure and growth rate in high and low light? What traits are correlated with RGR in high or low light and across species? Moreover, how do RGR and its determinants correspond with variation among these species in habitat affinities, leaf life span, seed mass and shade tolerance? In addressing these issues we tested the following specific hypotheses: (1) RGR is negatively related to increasing shade tolerance, seed size and leaf life span; (2) species differing in shade tolerance and/or growth rate differ in fine tissue morphology, with faster-growing, shade-intolerant species having finer leaves and coarser roots (higher SLA and lower SRL) than slower-growing, shade-tolerant species; (3) species varying in shade tolerance and/or growth rate differ in proportional biomass allocation to leaves, with faster-growing, shade-intolerant species having proportionally greater biomass in leaves than slower-growing, shade-tolerant species, when compared under comparable light conditions; (4) plants (all species) adjust their biomass allocation preferentially towards leaves and produce finer leaves (high SLA) when grown in shade, but shade-tolerant and -intolerant species will differ in their plasticity; (5) seedlings of large-seeded and small-seeded species have similar RGR when comparisons are made of plants similar in size.
Materials and methods
The nine species used in this study were (Table 1): Populus tremuloides Michx., a very small-seeded, extremely shade-intolerant and widely distributed pioneer species (Burns & Honkala 1990); Betula papyrifera Marsh., a small-seeded, shade-intolerant pioneer of disturbed sites (Perala & Alm 1990); Betula alleghaniensis Britton, a small-seeded species that grows well in early-successional, high-light environments as well as in partial shade (Perala & Alm 1990); Larix laricina (Du Roi) K. Koch, a shade-intolerant deciduous conifer with intermediate-sized seeds; Pinus banksiana Lamb., a shade-intolerant, fire- and drought-adapted evergreen conifer with intermediate- to large-sized seeds (Burns & Honkala 1990); Picea glauca (Moench) Voss, an intermediate to shade-tolerant evergreen conifer (Burns & Honkala 1990) with intermediate-sized seeds; Picea mariana (Mill.) B.S.P., a shade-tolerant evergreen conifer (Burns & Honkala 1990) with intermediate to small seeds; Pinus strobus L. an intermediate to shade-tolerant evergreen conifer (Burns & Honkala 1990) with large seeds; Thuja occidentalis L., a medium-seeded, late-successional species that is very shade-tolerant (Burns & Honkala 1990). Most of these species have very wide geographic distribution across eastern and northern North America. Species are listed in Table 1 and Figs 1–5 in order of their (additive combined) rank order of increasing seed mass, leaf life span and shade tolerance. Differences in the combined rank order of the four shade-tolerant evergreen conifers were minimal and thus they are arbitrarily displayed in alphabetic order. Seeds of all species were collected in northern MN and were obtained from the University of Minnesota Aspen-Larch Cooperative (Populus) and the Minnesota Department of Natural Resources (all others). For brevity, species will be denoted by genus hereafter, except where necessary to differentiate among species within a genus.
Seeds were germinated in late April 1992. Germinants were planted in 2·7l plastic pots in a temperature-controlled greenhouse at the University of Minnesota, St Paul, MN, USA. Seedlings were grown in a 60/40% mixture of silica sand and field soil mix, and irrigated daily with half-strength Hoagland’s solution. Pots were periodically flushed with larger volumes of water to prevent the build up of nutrient salts. A fivefold contrast in light level was produced by comparing unshaded blocks with those covered with neutral-density woven polypropylene cloth attached to wooden frames. Total daily photosynthetic photon flux density (PPFD, mol m–2 d–1) in the unshaded (other than by the greenhouse itself) and shaded treatments were 25% and 5% of full sunlight (based on comparisons of treatment vs outdoor daily integrated PPFD), similar to medium gap size and shaded understorey microsites, respectively, in boreal forests in northern MN (J. L. Machado & P. B. Reich, unpublished data; M. B. Walters et al., unpublished data). Over the course of the experiment, day/night temperatures averaged 25/20°C.
EXPERIMENTAL DESIGN AND GROWTH MEASUREMENTS
The experiment was arranged as a randomized complete block design, with four blocks containing the two light levels. The nine species and two light combinations comprised a complete factorial arranged as a split plot, with species as sub-plots within light environment whole plots. An average of 16 germinants per species were harvested for dry mass determination at the start of experimental treatments (i.e. harvest 0 at time0). All plants were harvested and oven dried (70°C for 48h) masses of leaves, stems and roots were determined over five additional harvests (at approximately 12-day intervals). One randomly chosen plant per species–light-block combination was harvested at each harvest. There were no visible signs of root necrosis or turnover. Projected area (one-sided) of individually sorted leaves and total length of root systems, were assessed at each harvest with a video-imaging system (AgVision, Decagon Devices, Inc., Pullman, WA, USA). Based on the harvest data, a number of plant traits were calculated. Acronyms and units for these parameters are summarized in Table 2.
Table 2. . Abbreviation, full name and units for growth analysis and tissue morphology terms used in this paper. All areas presented for projected areas
Relative growth rate (RGR) (and net assimilation rate, NAR) were determined using both the classical and several alternative curve-fitting approaches, describing the change in natural log-transformed plant mass against time (Evans 1972; Hunt 1982). Different techniques resulted in only slight differences in temporal patterns or in time-averaged RGR and did not markedly affect treatment or species comparisons. All species had maximum RGR early in the experiment. Differences in early RGR among species and treatments largely determined eventual differences in total plant mass later in the study. RGR, SLA, SRL and biomass distribution all varied with time and/or plant size in at least some (and often many) species-treatment combinations. Plant mass of the different species had substantial overlap during the experimental time interval, thus providing a useful comparison of plants of similar age and size. For instance, P. tremuloides and B. papyrifera were smaller than the other species after 14 days of treatments but were larger by the 61st day. The average RGR values at the first four harvests for the 61-day time interval following the onset of treatments, based on analyses for all plants from harvests 0–5, are presented and used in this paper. Comparison among species and treatments of RGR for plants at a common mass were similar. Most other plant traits (e.g. LAR) are shown using the mean values of measurements made at each of four approximately evenly spaced harvests (after 14, 27, 36 and 48days of treatment) during that 61-day period.
Values at a common mass were also estimated for several measures from regressions of plant traits on ln-transformed dry mass using individual plant data from all blocks and harvests within a treatment (sensuWalters et al. 1993a,b). For instance, SRL and RLR were estimated for plants of a common root mass (7mg, which corresponds to an average plant mass of 30–40mg), based on significant (P<0·001) linear regressions of SRL and RLR vs ln root mass for each species–light combination. Species at a common root size are compared for these traits because they varied enormously with root size.
Block and treatment (species, light) effects and species × light interaction on growth, morphology and physiology variables with analysis of variance were tested for using JMP statistical software (SAS Institute, Cary, NC, USA). Blocking had no significant effect on measured variables. Light treatment and species effects were considered to be statistically significant if P 0·05 for appropriate F-tests. Correlation analyses were used to address the hypotheses listed above.
GROWTH AND ALLOCATION
Absolute and relative growth responses were strongly related to species differences in seed mass, leaf life span and reported shade tolerances. For instance, dry mass differences among species initially followed seed mass differences, such as after 14 days of treatments: small-seeded Betula and Populus initially had smaller plants than larger seeded conifers, and large-seeded P. strobus seedlings were the largest. However, by 48 days of treatments, plant mass in 25% and 5% light was inversely related to seed mass (data not shown) owing to differences in RGR for species varying in seed mass (Table 3).
Table 3. . Correlation matrix for relative growth rate (RGR) and other measured variables. The upper right half of the matrix shows correlation coefficients in high light (25%) and the lower left half those in low light (5%). Correlations significant at P<0·05 are shown in bold. Average RGR, net assimilation rate (NAR), leaf weight ratio (LWR), stem weight ratio (SWR), root weight ratio (RWR), leaf area ratio (LAR) and specific leaf area (SLA) were for data averaged over four approximately equally spaced harvests. Correlation coefficients for RGR at a common plant size were similar to those for RGR averaged over the harvest intervals and are not shown. Root length ratio (RLR) and specific root length (SRL) were averages at a common plant size. Seed mass is the logarithm of dry seed mass and leaf life span the inverse (i.e. turnover rate) observed in field studies
Mean RGR over the 61-day treatment period varied fourfold among species in both the 25% and 5% light environments (Fig. 1). In both high and low light, RGR was highest in the shade-intolerant species: Populus, followed in decreasing rank by the two Betula spp., Larix and the five evergreen conifers. This ranking of RGR among species follows the patterns reported (Table 1) for increasing shade tolerance of the species, leaf life span and seed mass (all P<0·01, each tested using Spearman rank correlation). Populus is extremely shade intolerant, has the lightest seeds and shortest leaf life span of the nine species, closely followed in these respects by B.papyrifera. Betula alleghaniensis and L. laricina were the next fastest-growing species and have heavier seeds and slightly longer leaf life spans than Populus or B. papyrifera. All of the evergreen coniferous species have longer leaf life spans, heavier seeds and/or a greater degree of reported shade tolerance. Pinus strobus has the heaviest seed mass of all species and the lowest RGR observed over the 61-day period. In both light environments, lowest RGR was displayed by P. strobus and Thuja, with slightly higher rates displayed by the two Picea spp. and P. banksiana. Differences in RGR among the five evergreen conifers disappear when their RGR is compared at similar total plant mass (data not shown).
Mean RGR for all species was significantly (P<0·001) lower in low than in high light and there were significant species × light interactions (ANOVA, P<0·01): Populus and Betula had a greater difference in growth between light environments than the slower-growing evergreen conifers. As a result of differences in RGR across light environments, Populus and B. papyrifera plants in high light grew about 20× as large by the end of the study than those in low light, followed in this ranking by B. alleghaniensis and Larix, and then the evergreen conifers, all of the latter of which were about 5× as large in the high- than low-light treatment.
The relative partitioning of biomass to foliage, leaf weight ratio (LWR, g leaf/g plant), was generally no higher in low than high light (Fig. 2) (P>0·10) and there were no species × light interactions in this variable. Pinus strobus had the lowest LWR of all species (P<0·05). Differences in LWR among other species were not significant. In all species and both light treatments, plants allocated roughly 60% of plant dry mass to foliage. Proportional biomass partitioned to foliage did not differ by species or light environments across a wide range of plant mass (data not shown). Moreover, LWR did not differ among species in any pattern related to physiological or ecological species attributes, in marked contrast to the differences in RGR or root and leaf structure (see Table 3 and Figs).
All species allocated more biomass to stems and less to roots in low than high light (Fig. 2, P<0·001). On average, all species allocated proportionally one-third to one-quarter more dry mass to stems in low than high light. Allocation to roots averaged 22% in high light and 16% in low light. Significant differences existed among some species in stem weight ratio (SWR) and RWR within light levels, and in the magnitude of the effects of light level. However, these differences did not follow any pattern with respect to species differences in seed mass, leaf life span, or shade tolerance, such as were observed for RGR, LAR and most other responses (see below).
MORPHOLOGY AND METABOLISM
In contrast to the low variation in the proportion of biomass allocated to leaves among species and light environments, the amount of leaf area displayed per unit plant dry mass, leaf area ratio [LAR, cm2 leaf (gplant)–1], was consistently different across light levels (P<0·001) and varied among species (P<0·001) in a pattern parallel to that for RGR (Fig.3). In all species, LAR was greater in low than high light (by one-third to one-half) and in both light treatments LAR was greatest in Populus, followed in decreasing order by Betula, Larix and the evergreen conifers. The magnitude of species differences in LAR was large: Populus had fourfold greater LAR than P. strobus, roughly proportional with their differences in RGR. Although the general response of LAR among species to the contrasting light regimes was similar, there were significant species × light interactions; for example, B. papyrifera had the greatest difference in LAR across light treatments and P. strobus the smallest.
Because the proportion of dry mass per plant allocated to foliage did not differ markedly among species or light regimes (Fig. 2), and given that LAR = LWR × SLA; differences in LAR (Fig. 3) resulted primarily from differences among species and light regimes in SLA (Fig. 4). Thus, at a whole plant scale, leaf area displayed per unit plant mass is driven largely by differences in individual leaf morphology. In all species, SLA was significantly greater in low than high light (by 25 to 100%). In both light treatments SLA was greatest in Populus, followed in decreasing order by Betula and the conifers (with P. strobus having lowest SLA), a similar order as for LAR. Again, the magnitude of these differences was large (almost fourfold between Populus and P. strobus) and of a similar magnitude as species differences in LAR and RGR.
Differences among species in root structure, assessed by the ratio of root length (of entire root system) to dry mass (specific root length, SRL), closely paralleled those for foliage. In both light regimes SRL (Fig. 4) for plants of common size were greatest in Populus and B. papyrifera, followed in decreasing order by B. alleghaniensis and the conifers (with P.strobus lowest), a similar order as for SLA or LAR. Again, the magnitude of these differences was large (fourfold between Populus and P. strobus) and of a similar magnitude as species differences in LAR and RGR. Hence, species with high total surface areas (per unit plant mass) dedicated to light harvesting also had high total root-system lengths dedicated to acquisition of water and nutrients.
Rank order among species was generally similar for SRL at a common time (averaged at days 14, 27, 36 and 48) as for SRL at a common mass (data not shown). Moreover, SRL at a common size was significantly (P<0·05) correlated with SRL at a common time in high and low light (data not shown). In all species, SRL was greater in high than low light for plants of comparable size (Fig. 4). However, SRL for plants of common age (averaged over the first 2months of growth) was higher on average for low-light grown (than high light) plants because higher SRL is strongly associated with smaller plant and root mass (data not shown).
The significantly greater (≈35–40% more) root biomass allocation and SRL in high light resulted in substantially greater root-length ratios (root length per unit plant mass, RLR) for all species in high than low light (Fig. 3). Thus, all species enhanced leaf area (LAR) in low light and root length (RLR) in high light. High LAR in low light was accomplished by plasticity in leaf morphology (high SLA) while high RLR in high light was the result of both altered biomass allocation (high RWR in high light) and root structure (high SRL). Rank order among species was generally similar for mean RLR during the first 2months as at a common size (data not shown) and these two measures of RLR were significantly correlated in high (r = 0·85) and low (r = 0·89) light (data not shown).
Net assimilation rate (NAR) expressed on both mass (NARmass) and area (NARarea) bases were consistently different across light levels and varied among species in a pattern parallel to that for RGR (Fig. 5). In all species, NARmass was greater in high than low light, and in both light treatments NAR was greatest in Populus, followed in decreasing order by Betula, Larix and the evergreen conifers. The magnitude of species differences in mass-based NAR was large: Populus had fourfold greater NAR than P.strobus, roughly proportional with their differences in RGR.
IS INTERSPECIFIC VARIATION IN RGR RELATED TO VARIATION IN ALLOCATION, PLANT STRUCTURE OR OTHER PLANT TRAITS?
Variation in RGR among species was highly correlated (P<0·001) with both root and leaf morphology (SLA, SRL) and the resulting whole-plant measures LAR and RLR (r≈0·90 in high light and 0·95 in low light), but not to LWR, SWR or RWR (P<0·05) (Fig.6, Table 3 and data not shown). Thus, plants with high surface areas for light, CO2, water and nutrient acquisition had higher growth rates. Biologically, RGR at any point in time is the instantaneous product of LWR × NARmass, so given that there was little observed variation in LWR (and no LWR–RGR relationship), RGR should be a strict function of NARmass. The correlation between the two measures was extremely tight (r=0·99) in both high and low light (Fig. 7). Hence, mass-based metabolic differences among species exerted almost total control over differences in RGR, with hypothetical differences in biomass allocation apparently irrelevant, because they did not occur. RGR is also the instantaneous product of NAR expressed on an area basis (NARarea) multiplied by LAR. RGR was positively related (P=0·05) to NARarea in both high (r=0·71) and low light (r=0·68), but interspecific variation in LAR explains a greater share of variation in RGR than NARarea. Can these area vs mass-based explanations of control of RGR be reconciled? The area-based measures indicate that development of high foliar surface area, and secondarily, high productivity per unit foliar area, both are important determinants of RGR. The mass-based measurements indicate that biomass partitioning to leaves plays no role (at least for these species) but that metabolism per mass invested in leaves (NARmass) is a critical factor controlling RGR. Variation in SLA provides the direct linkage between these two ideas. Plants have high LAR because they have high SLA, and high SLA is also strongly related to high mass-based photosynthetic rate (Reich et al. 1992; Reich, Kloeppel et al. 1995; Reich, Ellsworth etal. 1997; Reich, Walters et al. 1998) and hence high NARmass.
RGR was well correlated with seed mass in both high (r=–0·86) and low light (r=–0·84), respectively (Fig. 8, Table 3). RGR was also well correlated in high (r=0·94) and low light (r=0·95), respectively, with the leaf turnover rate (inverse of leaf life span) of the species as assessed under field conditions.
In both high- and low-light growth environments there was fourfold variation in RGR among tree species that co-occur on southern boreal sites in North America. Variation in RGR and associated tissue structure was closely associated with species differences in successional niche, seed size, shade tolerance and leaf life span, supporting hypothesis (1). The five evergreen conifers all possess at least several traits predicted to be associated with low RGR (long leaf life span, large seeds and/or shade tolerance). Although the nine species comparison includes contrasting deciduous hardwood with evergreen conifers, the general patterns are similar within smaller groupings. For instance, in limiting a comparison to deciduous species, the traits displayed by the deciduous conifer, Larix, vs the three deciduous hardwoods are consistent with its larger seed mass and similar leaf life span and shade intolerance. Similarly, in comparing only needle-leaved conifers, Larix’s traits (higher RGR, SLA, etc., than the other conifers) are consistent with its lesser shade tolerance, shorter leaf life span and small seed size. Thus, the general patterns observed in this study hold both within and across broad species groupings for the dominant trees in the sub-boreal biome.
GROWTH AND PLANT TRAITS OF SPECIES DIFFERING IN HABITATS AND INTRINSIC TRAITS
However, data for the nine species also differ in several ways from related studies. Differences include the role (or lack thereof) of differing biomass allocation among species vis-a-vis inherent variation in RGR, the relationship between variation in SRL and RGR among species and the shift in biomass allocation in low light to stems but not to leaves.
First, refuting part of hypothesis (2), the pattern of variation among tree species in SRL (higher in fast growers, lower in slow growers, with a strong SRL–RGR correlation) differs from prior studies. For grass species, slower growing species from nutrient-poor habitats are considered to generally have a higher SRL (Boot 1989; Lambers & Poorter 1992). Poorter & Remkes (1990) found no correlation between SRL and RGR in a comparison of 24 herbaceous species. Huante et al. (1992) found higher SRL in early than late secondary successional tropical dry forest seedlings but also saw no relationship between RGR and SRL. We argue that it is physiologically logical that fast-growing species with high resource acquiring surface areas above ground should have ‘mirror-image’ root systems with high total root system lengths that potentially enhance the water and nutrient acquisition that is required to support a high leaf area and a high RGR. Differences in biomass allocation could override this: a species could have proportionally twice the mass of roots, with half the length per root mass (and in theory, half the rate of resource uptake). However, species in our seedling study showed no differences in proportional biomass allocation.
Second, contrary to hypothesis 3, there was no tendency for species arrayed along any of the above ‘life-history’ continua (shade tolerance, leaf life span, etc.) to allocate greater biomass to roots, stems or foliage, in contrast to the results of Walters et al. (1993a) and Kitajima (1994). Both of those studies found shade-tolerant species to allocate proportionally more to roots and less to leaves than shade-intolerant species. Data in the present study do not support the hypotheses that species adapted for high RGR have greater allocation to leaf mass and that RGR is closely related to biomass allocation. Contrasts among herbaceous species have shown positive, neutral and negative relationships (e.g. Hunt & Lloyd 1987; Poorter & Remkes 1990; Shipley & Peters 1990a; Fichtner & Schulze 1993) between RGR and biomass allocation (e.g. LWR). In a broad review, Lambers & Poorter (1992) suggested that LWR is less important than SLA in explaining inherent variation in RGR. Our data are consistent with their view.
Given the general paucity of SRL data, it is useful to search widely for relevant studies with which to compare our data. Gross, Maruca & Pregitzer (1992) reported no significant correlation between seed mass and SRL for 12 herbaceous species, although small-seeded species tended to have higher SRL, as was also true for seven tropical dry-forest species (Huante et al. 1992). Fahey & Hughes (1994) found no consistent or marked differences in SRL of three northern hardwood tree species in NH, USA, for fine roots harvested from mature groves. These three species do not differ as widely in shade tolerance, growth rate, leaf life span and other traits as the nine species in the present study. Thus, it is impossible to determine whether the different results [a lack of species differences in Fahey & Hughes (1994) vs strong differences in ours] are a result of the specific species contrasts, the differences in number of species compared, differences in seedlings vs mature trees, and/or greenhouse vs field conditions. At the same NH site, Mou, Fahey & Hughes (1993) found higher SRL in a young stand of a faster-growing pioneer species (Prunus) than in the mature stand of the more shade-tolerant, later successional species. This difference is consistent with our finding that SRL in trees is negatively related to shade tolerance, growth rate and successional niche, but differences among species in these NH field plots could be confounded by differences in stand age or soil. We are unaware of any other multiple species comparison of SRL among tree species with which to compare our data.
GROWTH AND ITS COMPONENTS AS INFLUENCED BY LIGHT LEVELS
Changes in SLA, LAR and RGR from high to low light were consistent with many prior reports. However, a uniform lack of any change in biomass allocation to leaves (zero of nine species) [refuting hypothesis (4)], but a consistent (nine of nine) shift in allocation away from roots and towards stems in low light, differs somewhat from other studies with tree seedlings (Walters et al. 1993a; Kitajima 1994; Walters & Reich 1996). Those prior studies found some species to shift allocation preferentially towards leaves and other species that did not shift allocation away from roots. More generally, it is most frequently reported that plants grown at low light show a shift in biomass allocation from roots and stems to leaves (Lambers & Poorter 1992), whereas in this study the shift was uniformly from roots to stems, with leaves unaffected. Is this difference related to plant type, i.e. a ‘tree’vs‘herb’ contrast? Or, is it possible that many earlier studies in the literature were confounded by plant size differences (plants in high light were larger and thus had low LWR mostly owing to ontogeny and secondarily to allocation, see Walters et al. 1993b)? The answers to these questions are important, as the extent to which a paradigm shift is necessary, from root:shoot biomass partitioning as a central focus of plant traits and their importance, to root and shoot morphology/structure (see Lambers & Poorter 1992; Reich et al. 1992; Walters et al. 1993a), hinges on such results.
RESULTS IN RELATION TO VARIATION IN SPECIES ECOLOGY
The question remains as to whether success in deeply shaded habitats depends, in part, on the ability of tolerant species to outgrow less-tolerant competitors in low light. Contrasting data exist. In some instances, but not others, shade-tolerant species had higher RGR than intolerant species in very low light (2–3% light) (Loach 1970; Kitajima 1994; Walters & Reich 1996). In this study, Populus and Betula had the highest RGR in both high and low light. Although the ranking of species growth rates was similar in 25 and 5% light intensities, it is possible that even lower light intensities might have resulted in a nullification or reversal of these rankings (contrast Walters & Reich 1996 with Walters et al. 1993a). Moreover, in field conditions which are temporally and spatially heterogeneous, the ability of intolerants to grow fast and survive at an average 5% of full light (rather than a constant 5%) might be constrained by temporal and spatial heterogeneity in light or other resources and stresses, which might lead to greater mortality from drought, herbivory, pathogens and other natural stresses.
Is variation in seed mass related to species differences in RGR? Observations in this study and elsewhere (Westoby et al. 1992) of higher RGR in smaller- than larger-seeded plants could result from differences in plant size, given the negative relationship between size and RGR, rather than differences in growth rate potential, per se. Chapin (1993) suggested that large-seeded species have lower reported RGR than small-seeded species because large seeds produce large seedlings, which have a low RGR because of their larger size at the time of measurement, given the negative relationship between plant size and RGR that holds within species during development. Data from the present study, however, do not support that idea [refuting hypothesis (5)], because RGR across a range of comparable plant masses, or standardized to a common plant mass, was still strongly negatively correlated with seed mass. Thus, at least for these nine species, seed mass varies in co-ordination with RGR and the traits that control it.
In general, results of this study are consistent with the notion that species adapted to high-resource environments have high potential rates of resource capture and growth relative to species characteristic of low-resource environments. With their small, well-dispersed seeds, Populus and B. papyrifera colonize high-light early-successional sites where a high RGR and associated traits, such as high SLA, SRL, LAR, RLR and NAR (plus high photosynthetic and nutrient uptake rates, see Reich, Ellsworth et al. 1997; Reich, Walters et al. 1998) may be advantageous. Betula alleghaniensis, which is ecologically similar to B.papyrifera but is less of a pioneer and apparently more successful at colonizing small treefall gaps rather than large openings (Perala & Alm 1990), possesses growth and structural attributes intermediate between B. papyrifera and the fastest growing conifer, deciduous Larix. On the other end of the spectrum, low potential growth rate and tissue turnover rates are presumed to be adaptations to habitats characterized by chronically low resource availability (Grime & Hunt 1975; Chapin 1980; Reich etal. 1992), such as the often shaded and/or nutrient-poor environments in which Pinus, Picea and Thuja are common. These ideas are consistent with the low RGR, SLA, SRL, LAR and RLR of these five evergreen species in this study; with their low photosynthetic and N uptake rates as reported in a companion paper (Reich, Walters et al. 1998); and with their low mass-based photosynthetic rate as mature trees in the field (Reich, Kloeppel et al. 1995). In conclusion, a host of plant traits (including leaf life span, shade tolerance, NAR, LAR, RLR, SLA and SRL) appear to be mutually co-ordinated with RGR in these tree species, supporting the idea of a predictable plant trait syndrome across broad species groups (Grime & Hunt 1975; Reich et al. 1992).