SEARCH

SEARCH BY CITATION

Keywords:

  • environmental variability;
  • genotypic variability;
  • mineral nutrients;
  • Salix;
  • small-scale variability;
  • within-individual variability

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  • All plant species require at least 16 elements for their growth and survival but the relative requirements and the variability at different organizational scales is not well understood.
  • We use a fertiliser experiment with six willow (Salix spp.) genotypes to evaluate a methodology based on Euclidian distances for stoichiometric analysis of the variability in leaf nutrient relations of twelve of those (C, N, P, K, Ca, Mg, Mn, S, Fe, Zn, B, Cu) plus Na and Al.
  • Differences in availability of the elements in the environment was the major driver of variation. Variability between leaves within a plant or between individuals of the same genotype growing in close proximity was as large as variability between genotypes.
  • Elements could be grouped by influence on growth: N, P, S and Mn concentrations follow each other and increase with growth rate; K, Ca and Mg uptake follow the increase in biomass; but uptake of Fe, B, Zn and Al seems to be limited. The position of Cu lies between the first two groups. Only for Na is there a difference in element concentrations between genotypes. The three groups of elements can be associated with different biochemical functions.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

All plant species require at least 16 elements for their growth and survival, and some species may need an additional four (Mengel & Kirkby, 2001). Because N and P generally are the elements limiting plant growth rates, most studies of plant stoichiometry have focused on these elements, and it is clearly established how the requirements of these two elements vary. The point of departure for this paper is, therefore, to explore the possibilities and methodologies for including additional elements in stoichiometric analyses. Plant stoichiometry varies with growth rate and the environmental conditions affecting it (Ågren, 2004, 2008; Cernusak et al., 2010); it might therefore reflect plant adaptation to variable element needs over space and time (Alonso & Herrera, 2001, 2003; Penuelas et al., 2008). For example, increasing growth rates require higher concentrations of elements relative to C and mostly higher concentrations of P relative to N – the growth rate hypothesis (Sterner & Elser, 2002). To our knowledge, no studies have investigated the relative requirements of other elements under varying growth conditions and only few studies have investigated how other elements vary in relation to each other (e.g. Garten, 1978; Thompson et al., 1997; Wright et al., 2005; Watanabe et al., 2007; Ladanai et al., 2010; Zhang et al., 2011). Moreover, it is not well known, except for N and P, to what extent observed element concentrations reflect physiological needs of plants, as determined by genotype, or their plasticity by means of environmental availability and excess uptake (Ågren, 2008). It has even been speculated that each plant species has developed a specific element composition, and plant stoichiometry reflects first of all plant nutrient requirements rather than nutrient availability in the soil (Garten, 1978; Markert, 1989). In other words, we know little about the causes of variability in plant stoichiometry, although elemental concentration pattern is possibly adaptive and related to plant fitness in a changing environment. If elemental concentration patterns are adaptive and related to plant fitness, plant stoichiometry could be an opportunity for use as a target in plant breeding.

The relationship between plant growth rate and plant nutrient concentration is for all elements approximately linear between a minimum concentration and an optimum concentration at which the plant reaches its maximum growth rate (Ågren, 2008). Increases in plant nutrient concentrations beyond the optimum (excess uptake) are not accompanied by any increase in growth rate. The plasticity in excess uptake varies greatly between species and elements (Ågren, 2008). Figure 1 illustrates the plasticity in excess uptake between nine essential elements for birch seedlings (Betula pendula) grown under similar conditions; except for the identity of the growth-limiting nutrient; these are to our knowledge the only data available allowing a systematic comparison to growth responses of so many elements under strictly controlled conditions. Similar patterns in excess uptake can also be expected for the Salix plants investigated further here. All elements, but N, seem to have the potential of being taken up in great excess, which has to be considered when interpreting plant stoichiometry.

image

Figure 1. Relationship between relative growth rate and whole plant nutrient concentrations for nine elements in experiments with birch seedlings (Betula pendula). Units mg g−1for N, P, K, S and Mg; μg g−1for Zn, Mn; Fe and Cu. Data for N, Ingestad et al. (1994): P, Ericsson & Ingestad (1988), K, Ericsson & Kähr (1993): Mg, Ericsson & Kähr (1995); Mn, Göransson (1994): S, T. Ericsson & M. Kähr (pers. comm.); Fe, Göransson (1993); Zn, Göransson (1999); Cu, Göransson (1998).

Download figure to PowerPoint

Variability in plant nutrient concentrations occurs on at least six different organizational scales and is caused by different factors:

  • 1
    Plastic variation between environments caused by the large-scale variability in element availability.
  • 2
    Heritable variation between genotypes grown in the same environment and caused by physiological differences between genotypes.
  • 3
    Plastic or heritable variation between individuals of the same genotype in a given environment and caused not only by small-scale variability in element availability (plastic variation), but also by small differences between individual plant genotypes.
  • 4
    Plastic variation within a plant (e.g. acclimation) and caused by translocations within a plant and growth conditions when tissues were formed.
  • 5
    Plastic or heritable variation within tissues and caused by inhomogeneous distribution of functions within them (Conn & Gilliham, 2010).
  • 6
    Variability in chemical composition of different compounds with similar functioning, e.g. S-rich vs. S-poor amino acids.

We consider the last two scales to be too detailed with respect to ecological functioning and will not consider these any further, but see Elser et al. (2011).

Based on previous results, we expect that a major driver of variability in nutrient concentration pattern is variation between environments (i.e. growth sites) caused by factors such as soil parent material and site history. Differences between genotypes in their differential requirements of nutrients and efficiencies in uptake should be the next most important factor causing variability. Identical genotypes should behave similarly with respect to nutrients (cf. Markert, 1989); individuals at a given site are expected to experience small-scale variability in nutrient availability, but contribution to overall variability should be small. Finally, tissues formed at closely similar times within a plant individual should be the most similar. We hypothesize, therefore, that the magnitude of the variability will follow (1) environmental > (2) genotype > (3) individual > (4) within-individual variation. A second hypothesis is that different elements will respond differently at the different scales.

A test of the hypotheses requires a methodology that allows stoichiometric analyses of variation in element relationships at various organizational scales. Here we aim to evaluate an appropriate methodology by analysing plant nutrient concentration data from a relevant ecological context. We have found a suitable dataset, which covers the relevant scales, in a fertilization experiment with taxonomically and evolutionary closely related Salix genotypes (i.e. natural or hybrid willows selected or produced during the last 10 to 30 years). These genotypes differ significantly in growth and other fitness-related characteristics (Weih & Nordh, 2005; Weih & Rönnberg-Wästljung, 2007). The plants were grown in a fertilizer trial with six different genotypes, and we will here investigate the importance of the first four of the above-mentioned causes for variability in leaf nutrient concentrations. Besides C, N and P, we include eleven other elements (K, Ca, Mg, Mn, S, Na, Fe, Zn, Al, B, Cu); of these Na and Al are not essential in the genotypes studied. These fourteen elements define a multidimensional space within which leaves will be positioned. We will consider element composition on three different bases: dry weight (DW) (14 dimensions), element :  C ratios (13 dimensions), and element : N ratios (12 dimensions). The standard variables for element concentrations are DW and element:C ratios are. However, Ladanai et al. (2010) found that element : N ratios was a more relevant variable for expressing stoichiometric relationships than the other two, probably better reflecting the physiological use of elements. This may also be a consequence of N being taken up less when in excess.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Plant material

The plant material was randomly selected from the mature leaves sampled from the upper, middle and lower part of the leader shoots of 35 Salix individuals (in all 114 leaves with 1–4 leaves per plant individual) grown with no treatment (0) or with irrigation and fertilization (W + F) in a field trial established in spring 2001 in Ultuna, Uppsala, Central Sweden (59º36′N, 18º27′E). The plant individuals are grown from cuttings of one natural willow clone (L78183) and five Salix hybrid varieties (Gudrun, Jorr, Loden, Tora and Tordis), each represented by three individuals (except L78183 with only two individuals in the 0 treatment) for each fertilizer treatment. The Salix varieties used have previously been shown to differ greatly in terms of growth and physiology (e.g. Weih & Nordh, 2005; Weih & Rönnberg-Wästljung, 2007). Plants in the W + F treatment were fertilized with a mineral nutrient solution (N, P, K, Ca, Mg, S, Fe, Mn, B, Zn, Cu, Mo) in 2002 and with a solid NPK fertilizer in 2003, and irrigated from May to September to supplement precipitation in both years. Soil pH(CaCl2) was c. 6.3 in 2002 and similar between treatments (M. Weih, pers. obs.). Leaves were sampled in August 2006. The W + F treatment resulted in more than doubling of the leaf area. For further experimental details, see Weih & Rönnberg-Wästljung (2007; leaf sampling) and Weih & Nordh (2005; field trial and specification of plant material).

Chemical analysis

Total C and total N in the samples were determined with a Carlo-Erba NA 1500 Analyzer. For the analysis of the other elements the leaf samples were weighed after drying at 105°C and wet-combusted in a mixture of nitric acid/perchloric acid (ratio 2.5 : 1.0) and element concentrations determined using an ICP emission spectrometry (PerkinElmer Optima 7300, Waltham, Massachusetts, USA).

Statistics

Effects of clones (genotypes) were analysed using a multivariate model:

  • image( Eqn 1)

(where yijkl = measured values; μ = average value; τi = effect of treatment or ‘site’ (0, W + F); cj = effect of clone (genotypes L78183, Gudrun, …); (τc)ij = interaction between treatment and clone; pijk = N(0,θ2), randomly and normally distributed plant effects; eijkl = N(0,σ2), randomly and normally distributed effects of individual measurements).

Effects of plants were analysed with a simplified model where clone effects were excluded

  • image( Eqn 2)

In both cases both homoscedastic and heteroscedastic models were used. As there were almost no differences between the two models, only results from former will be reported.

We have also calculated distances between leaves to see if there are factors that cause a clustering of leaves with similar stoichiometry. Let xi be a vector of k element concentrations for leaf i, where concentrations have been scaled such that the average for each element over all leaves is 0 and the standard deviation 1. We then define the average distance, d, between two leaves i and j from their Euclidian distance:

  • image( Eqn 3)

If the leaves are randomly distributed in the nutrient concentration space, the average distance between them is 0.371 (= 14) when expressed on a DW basis and 0.399 (= 12) when expressed as element : N.

We have used the prcomp module of the R statistical package (R 2.14.0, The R Foundation for Statistical Computing) to perform principal component analysis (PCA).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

The distributions for all element concentrations on a DW basis in all sampled leaves are shown in Fig. 2. Most of the distributions fit rather well to normal distributions and those that do not, do so after log-transformation. When element concentrations are expressed relative to C the distributions are almost identical to those in Fig. 2 (data not shown). Because there are only small differences between variables when DW or C is used as basis, we will only report results on a DW basis. However, when concentrations are expressed relative to N, all distributions are skewed to the left and those that in Fig. 2 followed a normal distribution now deviate slightly from it (data not shown).

image

Figure 2. Distribution of element concentrations (DW) in all leaves in a field experiment with irrigation and fertilization of six Salix clones. The solid line is the normal distribution fitted with mean and variance to the observed distribution. Four outliers of high concentrations have been excluded. Horizontal thick lines indicate the range where the element has been shown to be limiting; horizontal thin, broken lines indicate the range of excess uptake in a laboratory experiment with small birch plants (Fig. 1).

Download figure to PowerPoint

Figure 2 also shows the ranges of element concentrations within which the growth rate of small birch plants (Betula pendula) responded when the element was limiting, as well as the concentration ranges over which the elements were taken up when supplied in free access (excess uptake); no data are available for Ca and B, while Na and Al are not essential and ranges cannot be defined. For N and P the observed ranges of concentrations fall within the response ranges for birch and no uptake is observed beyond that where a growth response can be expected. For all the other elements the observed concentrations are all in the excess range of birch; S and Zn are taken up over wider and Cu and Mn over narrower concentration ranges by Salix than by birch.

Concentrations were significantly different between W + F and 0 (environment effect) for all elements but Na (Tables 1 and 2). However, it matters whether concentrations are expressed on a DW basis or as element : N ratios. Thus, there are significant differences for P, S and Mn when expressed on a DW basis but no differences with N as a basis. On the other hand, K, Ca and Mg differ significantly when expressed as element : N but not on a DW basis. The remaining elements (Fe, B, Zn, Cu, Al) differ significantly as a result of treatment/environment with both ways of expressing concentrations. The distributions of concentrations (DW) are modified accordingly, with distributions shifted towards higher concentrations in the W + F treatment for N, P, S, Mn and Cu (data not shown; see, however, Fig. 5) but in the opposite directions for Fe, B, Zn and Al. For the remaining elements the distributions do not differ between the two treatments/environments.

Table 1.   Average and standard deviations (in parentheses) of element concentrations in leaves from untreated (0) and irrigated and fertilized (W + F) plants
Element0W + F
  1. mg g−1 for C, N, P, K, Mg, S.

  2. μg g−1 for Fe, Mn, B, Zn, Cu, Na, Al.

  3. Significant differences: *, 95% probability; **, 99% probability.

C484 (13)488 (6)
N23 (6) 32** (5)
P2.5 (0.8)  3.3** (0.7)
K16.4 (2.2) 16.9 (3.2)
Ca13.4 (7.7) 11.8 (4.3)
Mg2.4 (0.9)  2.4 (0.7)
S3.9 (1.1)  5.0** (1.1)
Fe147 (65)107** (21)
Mn37 (13) 49** (21)
B30 (8) 23* (6)
Zn65 (40) 44** (18)
Cu7.6 (2.0)  8.6* (2.0)
Na49 (35) 41 (33)
Al47 (26) 29** (13)
Table 2.   Probabilities of significant differences in element concentrations (expressed as DW and element : N) for environment (treatments), individual (all plants, only untreated (0) plants, only irrigated and fertilized (W + F) plants), and genotype (all plants, only 0 plants, only W + F plants)
ElementDWElement : N
EnvironmentIndividualGenotypeEnvironmentIndividualGenotype
0/W + F0 + W + F0W + F0 + W + F0W + F0/W + F0 + W + F0W + F0 + W + F0W + F
  1. –, < 95% probability; *, 95% probability; **, 99% probability.

C******       
N******       
P***************
K***********
Ca******
Mg************
S**********
Fe**********
Mn********
B****************
Zn***************
Cu**********
Na**********************
Al******

There are almost no significant differences in concentrations between genotypes whether we use DW or N as base or whether we consider all plants or just effects within a treatment (Table 2). The noteworthy exception is Na, where varieties differ independently of how we express concentrations. Boron differs also between varieties when expressed on a DW basis but barely when N is the basis.

Concentrations of N and P increased whereas Ca, Mn, Al and B decreased towards the top of the plant; for the other elements there were no clear trends. These within-plant differences were consistent between treatments and the differences in concentrations between the top leaf and the bottom leaf were in general as large as differences between treatments. These differences are also consistent with the general patterns of allocation of N and P towards the most active leaves and precipitation of Ca in older tissues.

There are significant differences in concentrations (DW) between individual plants for all elements when considered over the entire set of plants (Table 2). When considered within a treatment there are fewer significant differences between plants and some are significant only within one of the treatments. The differences tend to be somewhat less pronounced when N rather than DW is used as basis.

Distances between leaves in the multidimensional element space varied with scale; shorter distances imply that leaves are more similar in stoichiometry (Fig. 3). When expressed on a DW basis, the distances between all the leaves do not differ from the distances between randomly distributed leaves. When split between treatments, there is a tendency for W + F leaves to be somewhat closer together than 0 leaves; W + F leaves appear more similar compared to 0 leaves. The leaves on the same plant are closer to each other than the average distance between leaves; the distance between leaves on the same plant is 67% of the average distance between all leaves (Fig. 3). When calculated as element : N, the distances between W + F leaves are closer than random leaves while 0 leaves behave as random leaves. The leaves on a single plant are situated at 76% of the average distance between two arbitrary leaves and thus closer than random leaves. Separating the leaves on genotypes seems not to bring the leaves closer than all the leaves within a treatment.

image

Figure 3. Average distances (bars) with standard deviations (whiskers) between Salix leaves when plants are grouped in different ways. The horizontal, broken line is the average distance if leaves were randomly distributed in the nutrient concentration space. (a) Concentrations expressed on a DW basis. (b) Concentrations expressed relative to N.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

Fertilization plus irrigation causes significant changes in stoichiometry with some elements increasing in concert (N, P, S and Mn) and others decreasing (Fe, B, Zn and Al), Table 1. The overall fertility of an environment (site) is therefore important because element concentrations will shift in a systematic way when fertility changes and, hence, plants growing under different fertility conditions will be separated from each other. Uptake of P, S and Mn match uptake of N resulting in concentration differences on the basis of DW but not of N. Uptake of K, Ca and Mg follow the increase in leaf biomass as a result of the W + F treatment and, hence, there is no effect on a DW basis, but element : N ratios decrease with the increasing N concentration. Uptake of Fe, B, Zn and Al did not follow the increase in growth caused by the W + F treatment, so concentrations decrease as a result of increased growth; the effect becomes still more pronounced when concentrations are expressed on an N basis because biomass increased less than N uptake to leaves (Weih & Nordh, 2005). This is expected as the N concentration should increase with growth rate. The uptake of Cu increased enough in the W + F treatment to cause a significant difference between treatments on a DW basis, but not enough to follow the increase in N; thus, on an N basis the Cu concentration is significantly lower in the W + F treatment.

Our results suggest that nutrients can be grouped stoichiometrically in different clusters. The elements N, P, S and Mn define one group with concentrations changing in concert. Another group is formed by K, Ca and Mg which follow plant size. The uptake of these elements is probably restricted by a maximal concentration in the biomass. The third group of Fe, B, Zn and Al probably did not increase in uptake in the W + F treatment. This suggests that these elements are limited by availability in the soil and/or interactions with other elements reduced uptake. The mechanisms determining the extent of excess levels are poorly known. Are the lower leaf concentrations of Fe, B, Zn and Al a result of decreased uptake and hence processes at the root surface, or instead a result of decreased translocation to leaves and, hence, an internal plant process? The position of Cu in this classification is unclear as it behaves somewhere in-between the first and second cluster, whereas the uptake of Na varies strongly between genotypes. The existence of distinct clusters of nutrients behaving differently across different scales does not support the view that plant stoichiometry reflects primarily plant nutrient requirements rather than nutrient concentrations in the soil (Markert, 1989).

Our grouping of elements according to Euclidian distances agrees with the findings in studies using PCA (Garten, 1978; Thompson et al., 1997; Wright et al., 2005). Also, the PCA of our biomass-based concentrations gives similar results (Fig. 4); similar clusters are obtained also when the PCA is carried out for leaves in different positions (data not shown). The three groups have been associated with different biological functions (Garten, 1978). Nitrogen, P and S, which all are mostly covalently bound, are all key elements for nucleic acids and proteins. Calcium and Mg are coupled to structure and photosynthetic activity while Fe, Zn, B are important for enzyme activities. The other elements are classified differentially between studies. This may be a result of some elements having multiple roles but also of arbitrariness in forming groups, in particular when using PCA, as well as scatter in data. However, if we perform a PCA on element : N ratios in our data, no groups can be identified in contrast to the results when using the distance measure. The PCA and our distance measure, Eqn 3, describe different aspects of plant stoichiometry. The PCA identifies correlations between elements without regard to the identity of the leaf, whereas the distance measure identifies regulations within a leaf in relation to the organizational context of the sample. The fact that these two different approaches, but not the PCA-approach using element : N ratios, lead to similar groups suggests that the biochemical regulation of elements is based on element concentrations rather than element : N ratios.

image

Figure 4. Grouping of elements according to their biological functioning found in different studies. Black squares, principal component analysis (PCA), Garten (1978); red triangles, PCA, Zhang et al. (2011); blue triangles, PCA, Wright et al. (2005); green circles, PCA, this paper; yellow circles, Euclidian distances, this paper. Note that Mn and Mg in one classification are attributed to two groups.

Download figure to PowerPoint

In general, the heritable variation in stoichiometry appeared small in our data, because there was little difference between genotypes as long as the essential elements are considered. Uptake of Na, which is not essential for most plants (Welch, 1995), varies greatly between genotypes, possibly indicating genotypic differences in salt tolerance (Hangs et al., 2011). It is possible that the differences between genotypes would have been larger if we had used taxonomically less related plant material. However, the connections between elements that we observe correspond to the correlations observed by Knecht & Göransson (2004) over a large range of species. Based on our results, element concentration pattern therefore seems to reflect plastic responses due to environment more than genotype differences.

The different behaviour of element uptake in response to availability explains also why leaves in the W + F treatment are closer together than leaves in the 0 treatment. Figure 5 shows the projection of the 14-dimensional element concentration space on a 2-dimensional plane with N on one axis and an element from each of the three clusters identified above (P, K and Fe) on the other. In the N–S plane the W + F and 0 leaves are clearly separated along both axes but the distances between the leaves within either of the clusters have not changed. In the N–K plane the elements are separated along the N-axes but not along the K-axes and again the distances between the leaves within either of the clusters have not changed. In the N–Fe plane there is a separation along the N-axes but now also along the Fe-axis. Moreover, this separation has pushed together the W + F leaves towards the lower end of the Fe-axis and along this dimension the W + F leaves are closer together. When the leaves are separated into genotypes (in Fig. 5), the genotypes are more or less evenly distributed within the clouds of points from the W+F respectively 0 treatments, again reflecting the poor genetic variation in plant stoichiometry. It is, therefore, the compression of variability in concentrations in the Fe-B-Zn-Al cluster that has brought the leaves in the W + F treatment closer together than in the 0 treatment.

image

Figure 5. Relationship between scaled element concentrations in the Salix leaves. (a) S vs. N; (b) K vs. N; (c) Fe vs. N. Green symbols = 0 treatment. Red symbols = with irrigation and fertilization (W + F) treatment. Salix genotypes are represented by: diamond, 78183; star, Gudrun; upright triangle, Jorr; inverted triangle, Loden; circle, Tora; square, Tordis.

Download figure to PowerPoint

The variability within a plant or between individuals of the same genotype growing in close proximity is as large as between different genotypes growing in the same area (Fig. 3), and this causes problems when sampling for identification of species (cf. Markert, 1989) and species properties. It might not be enough to sample just the top leaves, because from a functional point of view all leaves are operating and it is necessary to properly average over all leaves. This is a major obstacle when applying stoichiometric theory to large plants. The functional relationship between nutrient proportions and growth suggest that stoichiometric data could be useful in plant breeding. However, the low variability between genotypes along with the great variability among individuals of the same genotype makes it questionable whether plant stoichiometry is a meaningful breeding target or not.

Our observation that all elements except N and P seem to be taken up in quantities well above those required for growth also suggests that it is not meaningful to define a niche space containing the other elements because there should be no competition for those elements, cf. Garten (1978). A deeper analysis of excess uptake requires also that organs other than leaves are included; for example, stems and roots can be important storage organs especially in perennial plants, in which stored nutrients can lead to great delay in growth response to environmental change (e.g. Weih, 2000; Garrish et al., 2010). Our results also suggest that the functional diversity with respect to stoichiometry within individuals (between leaves) or between individuals is as large as the one between genotypes, making it questionable whether stoichiometry can be used as a functional trait to describe community structure (McGill et al., 2006; Cianciaruso et al., 2009).

In conclusion, the method using Euclidian distances appears appropriate for evaluating variation in element relationships, within and between plants, across different organizational scales. In the dataset used here, the magnitude in variability of elemental composition did not follow the hypothesized pattern of (1) environmental > (2) genotype > (3) individual > (4) within-individual variation, mainly because the variation between genotypes was small compared with environmental and individual variation. The results confirm, however, our second hypothesis: different elements responded differently at the different scales. Based on our data, plant stoichiometry therefore appears to reflect the outcome of individual elements’ responses (here in three groups) to the environmental conditions; the genetic component of variation is relatively low, although these genotypes have been previously reported to considerably vary in growth and other fitness-related traits.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References

We thank Lennart Norell for help with the statistical analyses and Tomas Grönkvist for doing the chemical analyses.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References