• This study addresses the question of whether a trade-off between the annual nitrogen (N) productivity (aNP) and the mean residence time (MRT), an index of N conservation, occurs both among and within species. We hypothesized that a trade-off between aNP and MRT observed among species might not hold within species.
• Seven perennial graminoids and herbs (including one evergreen) were subjected to two fertilizer concentrations and light intensities. The effect of these treatments was examined on plant traits determining aNP and MRT.
• Except for the evergreen Tofieldia pusilla, fertilization had an overriding effect on most variables related to the species’ nitrogen use and the response to fertilization among and within species closely matched. By contrast, shade had minor effects on plant nitrogen use. Across all species, aNP varied inversely with MRT, but within species aNP and MRT were not related.
• We suggest that there might not be a trade-off between aNP and MRT within species because small intraspecific changes in leaf life span may not involve significant consequences for leaf morphology, photosynthesis and growth.
On the other hand, within a given species or among congeners under different environmental conditions it is not self-evident that an adjustment of leaf longevity should be associated with a decrease in photosynthetic rate and productivity. Garnier & Aronson (1998) concluded that the observed relationships between nutrient productivity and nutrient conservation are inconsistent. Indeed, studies focusing on interspecific differences (including evergreens) found the proposed trade-off (Aerts, 1990; Eckstein & Karlsson, 1997), while those dealing with congeneric or within-species comparisons did not (Aerts & De Caluwe, 1994; Weih et al., 1998). A better knowledge of how MRT and aNP vary and relate to each other, both among and within species, therefore appears to be essential for the understanding of plant nutrient use strategies.
It should be noted that among deciduous species (woody and herbaceous) there might be little variation in biomass loss rate, leaf life span and, consequently, MRT, but a large variation in aNP. By contrast, evergreens, that is species that keep their leaves for more than one year, show the reverse pattern. This is illustrated in our in situ comparison of some subarctic plants (Eckstein & Karlsson, 1997).
In this paper we test the hypothesis that MRT and aNP should vary inversely among species with a wide range of biomass loss rates and leaf life spans, but within species or among species with a similar leaf life span no such relationship should emerge. To this end we studied the response of seven perennial graminoids and herbs (including one evergreen species) originating from contrasting habitat types with respect to a factorial combination of two nutrient and two light levels. The aim of the present study was to compare the effect of nutrient availability and light on variation in nutrient use strategy among and within species (that is plasticity). We specifically refer to nitrogen in this study, since this is the limiting mineral nutrient in lowland mire and heath ecosystems of the study area (Aerts et al., 1992; Jonasson et al., 1993).
Materials and Methods
Species and habitats
In August 1994, 320 ramets of each of the seven investigated species – consisting of one or a few aerial shoots, rhizome and roots – were collected from two field sites described by Eckstein & Karlsson (1997). These were a nutrient-poor, open mire area (Empetro-Sphagnetum fusci, Dierßen, 1996) on a subalpine heath (the ‘Torneträsk heath’ in Josefsson, 1990), and a nutrient-rich, shady meadow birch forest (Geranio-Betuletum, Dierßen, 1996) at the northern slope of Mt. Njulla close to the Abisko Scientific Research Station (68°21′ N, 18°49′ E). Single ramets were planted in 1 dm3 plastic pots containing peat of local origin as substrate and were kept outdoors in an experimental garden at Abisko during the whole experiment (1994–96).
The graminoids Calamagrostis lapponica (Wahlenb.) Hartm. and Carex vaginata Tausch were collected on the nutrient-poor, open subalpine heath. These species occur on sandy or peaty substrates in mires, heaths and open, dry forests (Roweck, 1981) and have a relatively wide ecological range with respect to nutrient availability (Sonesson & Lundberg, 1974). The third species from the heath area was Tofieldia pusilla (Michx.) Pers., a small herb with long-lived evergreen leaves, which is characteristic of open, disturbed or nutrient-poor habitats on peaty or sandy soil (Roweck, 1981). It occurs in heaths and on frost-heaved ground (Jonasson, 1986).
The wintergreen graminoid Deschampsia flexuosa (L.) Trin. and the broad-leaved graminoid Milium effusum L. were collected from the nutrient-rich, shady birch forest. Deschampsia has a wide ecological amplitude (Roweck, 1981), but in the study area this species is confined to places with good snow protection during winter (Sonesson & Lundberg, 1974) and is thus characteristic of rich forest sites. The herbs Rubus saxatilis L. and Stellaria nemorum L., which are also characteristic elements of the rich birch forests of the meadow type in the Torneträsk area (Sonesson & Lundberg, 1974), were also collected from the same location. While Rubus also occurs on the forest margins and in screes, Stellaria and Milium are confined to moist, shady habitats in the forest interior (Roweck, 1981; R. L. Eckstein, pers. obs.).
The experimental treatments consisted of a factorial combination of two fertiliser levels, that is 1 and 10 g nitrogen m−2 yr−1 (hereafter named FL and FH, respectively), and two light levels, that is ambient light and a light reduction by c. 64% (LH and LL, respectively). The treatments FL LH and FH LL mimic the nitrogen and light conditions of the bog and the meadow birch forest. The treatments FL LL and FH LH were assumed to represent these factors for plant growth in an adverse and an optimal combination, respectively. The nutrient solutions were prepared using a complete, commercial fertiliser (Superba S, Hydro Agri, Landskrona, Sweden) with a N : P : K ratio of 7 : 1 : 5. Each pot was fertilized 12 times during each of the two growing seasons, that is from mid-June to early August in both 1995 and 1996. Nitrogen concentrations in the solutions were 1.85 (FL) and 18.5 mmol N (FH). Plants of the low and high fertiliser treatment received 0.86 and 8.6 mmol N per pot per year, respectively. The interval between fertilizations, which was six days at the beginning of the season, was reduced to about three days later on, when the plants had grown larger. We assume that the plant response to fertiliser treatment is solely due to nitrogen since the subalpine heath- and forest-ecosystems of the study area are strongly nitrogen limited (Jonasson et al., 1993; Parsons et al., 1994, 1995; Weih, 1998a).
For the shade treatments, four wooden frames of 3 m × 3 m were erected. Two of those were covered with clear 0.05-mm polyethylene plastic sheets on top (high light, LH) and two carried plastic sheet plus green plastic cloth (shade treatment, LL). The green cloth covered the wooden constructions to the ground to shade the plants also from the sides. The shade treatment was thus replicated twice and pots of both fertilization treatments were placed under each frame.
Radiation spectra of the light treatments were measured with a spectroradiometer (Li-1800, Li-Cor Inc., Lincoln, NE, USA). On a clear day, transmittance in the range of photosynthetically active radiation (400–700 nm, PAR) was c. 82% and 36% in the high light and shade treatment, respectively, and the corresponding red to far-red ratios were 1.03 and 0.92 (M. Mohr, pers. comm.). Representative values for a natural mountain birch canopy close to Abisko are a mean transmittance (PAR) of 38%, and a red to far-red ratio of 1.12 and 0.82 above and below the canopy, respectively (Fig. 5 in Ovhed & Holmgren, 1995).
Harvesting procedure and nitrogen analysis
After planting in August 1994, all pots received fertiliser corresponding to the low fertiliser treatment. Thereafter, the plants were kept outdoors during the winter. During the growing season of 1995, the plants were subject to the treatments from June to September to acclimatize to the treatment conditions. The harvests covered the different phases of plant development after snowmelt in the early season (June 1996), after the main development of leaves (July 1996), at peak above-ground biomass (August 1996) and in winter (November 1996). At each harvest (except November 1996), 10 replicates per species and treatment were randomly selected (five pots of each fertiliser level from each of the frames). In November 1996 the number of replicates was three to six, and pots were dug out from under a 30-cm snow cover.
The plant material in each pot was divided into green and brown leaves (leaf-blades in case of graminoids), stems (including leaf sheaths in graminoids), inflorescences, roots and rhizomes. All plant material was then dried at 70°C for at least 24 h and weighed.
Plant material of six replicates of each species, harvest and treatment were analysed photometrically for Kjeldahl nitrogen using a flow injection analyser (Fia star 5012, Foss Tecator AB, Höganäs, Sweden). In total, this study is based on 1260 harvested plants and about 3400 N analyses. All N pools and losses in this study are expressed on a per plant basis.
Total net primary productivity (NPP) was calculated as the sum of differences in dry weight between June and July 1996, July and August 1996 and August and November 1996. During the growing season from June to August, above-ground tissues senescing between two harvests could be taken into account since these remained attached to the plant. However, some dead material was lost during the last harvest in November 1996
Nitrogen productivity (NP) was calculated separately for three harvest periods with positive productivity: (1) June to July; (2) July to August; and (3) August to November. During the rest of the year productivity and NP was zero.
To calculate NP we applied the following equation adopted from Evans (1972: 262):
(where W and N pool are the plant d. wt and N pool size, respectively, at two consecutive harvests conducted at times T1 and T2 (indexed by the subscripts).)
This equation is essentially analogous to the formula developed by Evans (1972: 262) for the computation of unit leaf rate. It is based on the assumption that d. wt is linearly related to plant N pool size but does not require that NP is constant between the harvest periods (Evans, 1972). An average aNP was calculated across a whole year, that is including months with zero productivity and aNP. The average aNP was then expressed as g dry matter per mol N and year.
The mean residence time of N (MRT) was calculated as the ratio between the average N pool size and the annual N losses. Nutrient losses were calculated based on the N concentration and d. wt of all above-ground dead material found at the November harvest plus the difference in above-ground d. wt between August and November 1996, since some dead material was lost during the latter harvest due to heavy snow covering the plants.
The NUE approach of Berendse & Aerts (1987) builds on the assumption that the system is in a ‘steady-state’ (Frissel, 1981) with respect to biomass production and N content (Garnier & Aronson, 1998). Since this was not the case in our study, we calculated N productivity as the average NP of shorter time intervals, as proposed by Vázquez de Aldana & Berendse (1997) (see previously). This estimate of N productivity was closely correlated with N productivity calculated using the annual NPP and the annual average pool size (Pearsons r = 0.932, P < 0.0001).
With respect to mean residence time steady state conditions are generally assumed for in situ studies without fertiliser application, but, to our knowledge, have never been explicitly tested. Furthermore, a steady state with respect to nutrient flux or dry matter is much easier to anticipate at the population level (Garnier & Aronson, 1998) where birth and death of plant individuals are important components of the N and biomass flow. At the level of the individual plant, nutrient uptake and growth and nutrient and dry matter losses occur during different times of the season. These processes may thus be controlled by different environmental factors (Weih & Karlsson, 1997). Therefore, the steady state concept appears to be difficult to apply at the level of individual plants and more research into the consequences of violation of the steady state assumption is needed. The main flaw of calculating MRT under nonsteady state using growth analysis techniques, as done in the present study, appears to be that the estimate obtained is representative only for the study period. However, for reasons of clarity we refrain from introducing new labels to our estimates. The above problems should not limit the meaningfulness of comparative studies dealing with the variation of MRT among species or the interrelationships among MRT and different plant traits as in the present study.
To test for the effect of species, fertiliser and light on parameters related to nutrient use efficiency, a three-way fixed factor analysis of variance (ANOVA) was calculated. To avoid problems with the standard ANOVA assumptions, in this analysis P-values were obtained by a permutation test (Manly, 1991) using an ‘approximate randomization’ procedure (Noreen, 1989), by calculating 5000 permutations (Potvin & Roff, 1993). Hence, an error probability of 0.01 means that in 1% (50) of the 5000 permutations the obtained F-value was equal to or higher than the original F-value. By definition, the original data are one of the ‘permutations’; thus P can not be lower than 1/number of permutations. The analyses were done on the original, untransformed data, since permutation analysis is independent on normal distribution and homoscedasticity of variances (Potvin & Roff, 1993).
Derived variables like NPP and the average N pool are not appropriate for ANOVA analyses since these are computed from harvest means. Therefore we used the mean total plant d. wt at above-ground peak season (August 1996) as a surrogate for NPP (Pearson’s r = 0.981), and the N pool size at this harvest as surrogate for the average N pool (r = 0.930). To test only for intraspecific response we carried out a two-way permutation ANOVA as described above for each species with fertiliser and light as factors (Aerts & De Caluwe, 1995).
The average pool size of N varied by two orders of magnitude and ranged from 0.06 mmol per plant in Tofieldia to 6.0 mmol per plant in Milium. The N pool of living parts in August 1996, which was closely correlated to the average N pool size, varied significantly among species (P < 0.001, ANOVA) and between fertiliser treatments (P < 0.001), but not between light levels. Averaged across all species and light levels, those pots receiving a high fertiliser dose had a N pool of 8.2 mmol per plant at peak above-ground biomass, while plants in the low fertiliser treatment contained an average of 1.4 mmol. However, since the fertiliser response was not equal across all species there was a significant species × fertiliser interaction (P < 0.001).
Averaged across treatments dry matter productivity (NPP) varied between 0.04 g yr−1 in Tofieldia and about 7.3 g yr−1 in Milium and Stellaria (Table 1). Total plant d. wt in August 1996, as a surrogate for NPP, varied significantly with species (P < 0.001, ANOVA), fertilization (P < 0.001) and light (P < 0.01). Averaged across all species, total d. wt increased with increasing N availability (2.1 g vs 10.0 g in FL and FH treatment, respectively) and with increasing illumination (5.5 g vs 6.7 g in LL and LH treatment, respectively). Analysis of variance revealed significant interaction effects of species × fertiliser (P < 0.001), species × light (P < 0.001) and fertiliser × light (P < 0.01) on total plant d. wt at peak above-ground biomass.
Table 1. Net primary productivity (NPP, g per plant per year), N losses (mmol N per plant per year), mean residence time (MRT, yr) of nitrogen (N) mean N productivity (aNP, g mol N−1 yr−1) and N resorption efficiency (REEFF,%) of seven subarctic perennials subject to two levels of fertiliser (FL = 1 g N m−2 yr−1; FH = 10 g N m−2 yr−1) and two levels of light (LL = c. 64% light reduction; LH= ambient light)
The amount of dry matter found in November 1996, which is an estimate of litter production, varied between 0.003 (Tofieldia) and about 5 g per plant (Calamagrostis) (Table 2). Actual annual litter production was slightly higher since some dead dry matter was lost during the November harvest due to heavy snow cover. Averaged across treatments, N concentration in litter ranged between 0.5 (Deschampsia) and 1.2 mmol N g−1 (Tofieldia) (Table 2). In many species plants subject to the shade treatment showed higher litter N concentrations than plants under ambient light (Table 2). However, this relationship was not consistent across all species and thus not statistically significant. Averaged across treatments, resorption ranged between 43 (Stellaria) and 67% (Calamagrostis) (Table 1) and was thus within the range known for perennial plants (Eckstein et al., 1999).
Table 2. Litter production (g plant −1 years−1) estimated as the amount of dry matter present in November 1996 and litter N concentration (mmol N g−1) of seven subarctic perennials subject to two levels of fertiliser (FL = 1 g N m−2 years−1; FH = 10 g N m−2 yr−1) and two levels of light (LL = c. 64% light reduction; LH= ambient light). Data are means with s.e. in brackets, n = 3–6
Litter N conc.
Species CL, Calamagrostis lapponica; CV, Carex vaginata; DF, Deschampsia flexuosa; ME, Milium effusum; RS, Rubus saxatilis; SN, Stellaria nemorum; TP, Tofieldia pusilla. *no measure of variance can be given because samples had to be pooled (n = 1).
Accordingly, annual N losses per plant ranged from 0.003 mmol per plant in Tofieldia to 5.6 mmol in Stellaria (Table 1). Both N pool and d. wt of above-ground dead material (litter) in November 1996 varied significantly among species (P < 0.001, ANOVA) and increased with fertiliser application (P < 0.001) but not with light level. There were significant interaction effects of species × fertiliser (P < 0.001 in both cases) and species × light (P < 0.05 for litter N pool and P < 0.01 for litter d. wt).
Among- vs within-species variation of N losses and dry matter productivity
Comparisons within species may comprise at least two sublevels, viz. distinct ecotypes (provenances) of a species and plasticity of a single genotype. In the present study all species were clonal plants regenerating mainly vegetatively and all specimens were collected from a single field population (see Species and habitats section). Therefore we assume that the within-species variation measured refers to plasticity.
Averaged across all treatments, that is considering interspecific differences only, species with a higher average N pool generally lost more N per year than species with low N content (Fig. 1a). Similarly, dry matter productivity and average N pool size were closely linked (Fig. 1b).
Within-species responses with respect to fertiliser and light treatments are presented in Figs 2, 3. Within the same light treatment (LL or LH), an increased N availability led to an increased average N pool and increased annual N losses (Fig. 2a, Table 1). An exception to this pattern was Tofieldia (Fig. 2a, inset). Among- and within-species responses thus corresponded well and had the same direction. However, within the same fertiliser treatment (FL and FH), there was no uniform response to a change in light conditions (Fig. 2b).
Within each light treatment, dry matter productivity increased uniformly with fertilization (Fig. 3a, Table 1), again except for Tofieldia, which showed the reverse trend (Fig. 3a, inset). Within the fertilization treatments dry matter productivity increased with increasing illumination in most species (Fig. 3b).
However, in the forest species Milium and Stellaria plants of the shade treatment produced more biomass than plants subject to high light conditions. For none of these species was the fertiliser × light interaction significant (P > 0.05) when the response was analysed separately for each species by a two-way analysis of variance.
MRT and aNP
Averaged across species MRT tended to decrease with increasing nutrient availability, within both the LL and the LH treatment (Table 1), though the differences among the four treatments were not significant. Within the low fertiliser treatment, MRT was similar for both light treatments (3.2 vs 2.9 yr, n = 6 species, Tofieldia excluded; 5.2 vs 6.1 yr, n = 7, Tofieldia included). By contrast, within the high fertiliser treatment MRT increased by about 50% with increasing light conditions (1.8 vs 2.7 yr, n = 6 species, Tofieldia excluded; 3.0 vs 4.6 yr, n = 7, Tofieldia included).
Values for aNP did not differ much among species, with the exception of Tofieldia, which had much lower values of aNP than all other species. Thus, at the interspecific level, there appeared to be an inverse relationship between MRT and aNP indicating a trade-off between those two parameters (Fig. 4a). However, at the intraspecific level an increased MRT was not associated with a decrease in aNP (Fig. 4b).
Species adapted to nutrient-poor habitats are generally assumed to be slow growing (Chapin et al., 1993) and to show less phenotypic plasticity (Grime, 1979; Chapin, 1980) and a slower response to environmental changes than species of nutrient-rich habitats. However, this depends largely on the plant trait considered. For example, grasses with an inherent high growth rate were more plastic than species with a low growth rate in terms of growth and N concentration but not in terms of biomass allocation (Van der Vijver et al., 1993; Garnier, 1998). Owing to the great variation among plant species occurring in nutrient-poor habitats, Grubb (1998) proposed to refine the ‘stress-tolerance’ strategy of Grime (1977, 1979) by dividing it into three different strategies:
• The ‘low-flexibility’ strategy. This strategy accords to Grime’s stress-tolerance strategy and comprises plants with a long leaf life span and low rates of growth, both in the juvenile and the adult phase of the life cycle.
• The ‘gearing-down’ strategy. Plants adopting this strategy show a short leaf life span. They tolerate resource shortage, but show a strong and fast growth response when the resource shortage is relieved through, for example fertilization or disturbance.
• The ‘switching’ strategy. Plants belonging to this strategy type are stress-tolerators as seedlings but not as adults. Consequently, there occurs a switch in growth rate between these stages of the life cycle.
Two of those refer to plants that show contrasting traits, viz. a long leaf life span (‘low-flexibility’ strategy) and short leaf life span (‘gearing-down’ strategy), while the third involves a major change in growth rate between the seedling and the adult (‘switching’ strategy) (Grubb, 1998). In our experiment two of the three species collected from a nutrient-poor bog (Calamagrostis and Carex) responded in a plastic way to the imposed treatments (Figs 2, 3). Compared with the forest species these two produced more biomass than the relatively slow growing Deschampsia and Rubus (Fig. 1, Table 1). Calamagrostis and Carex would thus correspond to Grubb’s ‘gearing-down’ strategy, which is exemplified by grasses from nutrient-poor heath systems, which respond quickly when the nutrient limitation is relieved (Aerts et al., 1990; Parsons et al., 1995). The persistence of these species under nutrient-poor conditions is assumed to depend on their ability to gear-down their metabolism (Grubb, 1998). The only species from the infertile habitat showing the low-flexibility strategy was the stress tolerant herb Tofieldia pusilla (Figs 1, 2, 3), which is a slow-growing nonwoody evergreen. In comparison with other life-forms evergreens appear to be least responsive to changes in nutrient availability and temperature, while graminoids are most responsive (Shaver & Chapin, 1986; Parsons et al., 1994; Aerts, 1995, Hartley et al., 1999).
The effects of fertilization and shade
Fertilization had an overriding effect on most of the parameters determining MRT and aNP. This could be expected since either N or P limits plant productivity in most ecosystems in the study area (Aerts et al., 1992; Jonasson et al., 1993; Parsons et al., 1994, 1995; Weih, 1998a,b). As in field fertilization experiments (Parsons et al., 1995) increased nutrient availability lead to higher biomass, an increased nutrient pool size, a higher shoot number and increased flowering frequency also in the present experiment.
There is a whole suite of traits related to infertile habitats, including scleromorphic leaves, slow growth, low photosynthetic rate and low leaf turnover rate. It has been suggested that this ‘stress resistance syndrome’ (SRS) may evolve by a relatively simple genetic change in a switch or underlying trait, such as a hormone, which may turn on the SRS (Chapin et al., 1993). Since there are clear physiological links between leaf structure, growth rate and stress resistance (Reich et al., 1992; Chapin et al., 1993) we may ask why no trade-off between aNP (a measure of productivity) and MRT (a measure of nutrient conservation) has been found within-species. However, the apparent lack of a functional link between MRT and aNP within-species may have several reasons.
Secondly, and probably most importantly, MRT and aNP are not intimately/directly linked with each other, but their relationship is mediated through other interdependent traits. MRT is closely related to the leaf life span (Aerts, 1990; Escudero et al., 1992; Garnier & Aronson, 1998; Eckstein et al., 1999). The large variation in leaf life span found at the macro-evolutionary (interspecific) level, for example between life-forms, invariably involves large changes in leaf function. For example, the longevity of a leaf has implications for leaf structure and morphology (Turner, 1994; Ryser, 1996), photosynthesis (Chabot & Hicks, 1982; Field & Mooney, 1986) and growth (Reich et al., 1998). Furthermore, a long life span often motivates larger investments in defence against herbivores and pathogens (Fagerström et al., 1987; Herms & Mattson, 1992) which may have negative effects on productivity. By contrast, within the relatively narrow intraspecific range of leaf life spans (Eckstein et al., 1999) nutrient conservation may be improved by increasing leaf life span without necessarily involving associated changes in leaf structure and productivity. For deciduous species, for example, a change in leaf life span of some days or weeks does probably not motivate increased investments in leaf morphology or defence. Rather, increases in leaf life span probably require changes in characteristics such as frost tolerance since the leaves then merge earlier in the spring and/or are maintained longer into the autumn. Such an increase in leaf longevity probably does not result in any costs as reduced productivity in the short term. However, if a change in leaf longevity is not fully matched by a corresponding change in frost tolerance, it may involve a larger risk of productivity losses due to frost damage at a longer perspective.
We therefore suggest that one may not expect a trade-off between aNP and MRT at the intraspecific level since the relatively small variation in traits possibly improving nutrient conservation may have no negative effects for growth and production. The negative relationship between MRT and aNP appears to reveal an evolutionary trade-off, that is in adaptation to habitats with different nutrient availability species have evolved nutrient use strategies, which consist in having either a long MRT and a low aNP or vice versa. Leaf life span appears to play an important role in this trade-off. The variation of leaf life span within species is low (Eckstein et al., 1999). Therefore, the proposed trade-off between MRT and aNP may be found among life-forms and species, if the variation in leaf life span is large (Aerts, 1990; Eckstein & Karlsson, 1997). By contrast, within species no such trade-off can be expected owing to the small variation in biomass loss rate, biomass life span, MRT or aNP.
We would like to thank J. Lindeberg for carefully nursing the plants and T. Westin, who built the shade frames and a protecting enclosure for the experiment. J. Lindeberg, M. Kardefelt, L. Ericsson and K. Rudfeldt-Tjus kindly helped harvesting and analysing the plant material. R. Aerts, E. Garnier, B. Vazquéz de Aldana, M. Weih and two anonymous referees gave valuable comments on earlier versions of this manuscript. We are much indebted to the director and staff of the Abisko Scientific Research Station for help and hospitality during the years of this study. Financial support was obtained from the Swedish Natural Science Council and the Abisko Scientific Research Station. Richard Hopkins kindly corrected the English.