1Nitrogen or phosphorus limits plant growth in many wetlands. If specific mechanisms reducing losses of the growth-limiting nutrient have been favoured by selection, the N and P resorption efficiency (RE) during leaf senescence (NRE, PRE: the fraction of N or P resorbed) might depend on the type of nutrient limitation.
2The size, mass, and N and P concentrations of green and senesced leaves were determined for 10 graminoid species at Dutch and Swiss wetland sites, with N : P ratios in leaves (6–27 by mass) indicating N or P limitation.
3During senescence, leaf area decreased by 8–19%, and leaf mass by 8–38%; NRE ranged from 0 to 87%, and PRE from 30 to 96%. PRE correlated strongly with NRE (r = 0·91) but was, on average, 17% higher. Within the Swiss or Dutch sites, NRE and PRE did not correlate with foliar N : P ratios, indicating that RE was not directly adjusted to the type of nutrient limitation.
4NRE and PRE were, on average, higher at the P-limited Swiss sites than at the N-limited Dutch sites. Because PRE exceeded NRE, high RE would be most beneficial when P limits plant growth. This may have contributed to the dominance of graminoids with high RE in P-limited wetlands.
Other evidence does suggest that efficient nutrient resorption is important as a nutrient-conserving strategy. A meta-analysis of leaf-level data revealed phosphorus resorption efficiency (PRE) to be the main determinant of P residence time in plant foliage (Aerts & Chapin 2000). A review of changes in leaf area and mass during senescence suggested that high RE (= 80%) is more common than previously thought, especially for P (van Heerwaarden, Toet & Aerts 2003). Wright & Westoby (2003) argued that nutritional control of RE does not necessarily imply negative correlations with nutrient availability, and is therefore not refuted by inconsistent changes along nutrient gradients.
The question remains as to whether and how RE is optimized in response to variation in nutrient availability. The type of nutrient limitation might be important. Either N or P (and sometimes K) can be limiting for plant growth in terrestrial ecosystems (Aerts & Chapin 2000; Olde Venterink et al. 2003). The type of nutrient limitation is approximately indicated by the N : P ratios (ratio between N and P concentration) of plant biomass: N : P mass ratios above 16 suggest P limitation, whereas N : P mass ratios below 13 suggest N limitation (Güsewell & Koerselman 2002; Tessier & Raynal 2003). Plants are likely to benefit most from an efficient use of the growth-limiting nutrient (Koide, Dickie & Goff 1999). If selection has optimized the benefits of RE (Wright & Westoby 2003), high nitrogen resorption efficiency (NRE) should have been favoured under N-limited conditions, and high PRE under P-limited conditions, even at relatively productive sites. As a result, PRE would increase in relation to NRE as plants become more P-limited, that is, as their foliar N : P ratios increase (Güsewell & Koerselman 2002). Alternatively, if nutrient resorption is not regulated by nutrition, but only by the flow of solutes from senescing leaves (Chapin & Kedrowski 1983), NRE and PRE should strongly correlate with each other. N and P would then be resorbed in direct proportion to foliar N : P.
Patterns and determinants of nutrient resorption have been investigated mostly for woody plants, and much less is known for herbaceous plants (Chapin & Kedrowski 1983; Aerts 1996; Eckstein et al. 1999; van Heerwaarden et al. 2003). This scarcity of data contrasts with the fact that nutrient resorption can be very important in the nutrient economy of herbaceous plants, especially graminoids, because of their short leaf life span, sequential leaf growth, and potentially high RE (Jonasson & Chapin 1985; Aerts 1996; Aerts & Chapin 2000; Bausenwein, Millard & Raven 2001). Several studies revealed plastic adjustments of RE to nutrient availability in graminoids (e.g. Rejmánková 2001). In a growth experiment with five Carex species from wetlands, Güsewell (2005) found that, within each species, NRE was greatest in plants with low to intermediate N : P ratios, while PRE was greatest in plants with intermediate to high N : P ratios (Güsewell 2005). Relationships between RE and biomass N : P ratios (as indicators of N or P limitation) have not yet been tested in field experiments.
This study investigates how leaf size, leaf mass and foliar nutrient concentrations change during leaf senescence in graminoids from N- and P-limited wetlands. Two main questions are addressed: first, do patterns of nutrient resorption in graminoids parallel those in woody plants? Second, is there evidence that plants adjust NRE and PRE to the type of nutrient limitation? While most previous studies focused either on interspecific differences (e.g. Wright & Westoby 2003) or on intraspecific variation (e.g. Vitousek 1998), this study considered both by sampling 10 species at several locations in Swiss fen meadows and in Dutch floating fens, to include plants with a broad range of biomass N : P ratios.
The four Swiss sites are fen meadows near Zürich (47°25′ N, 8°40′ E) at 430–550 m a.s.l. The long-term mean annual temperature is 7·9 °C, and the mean annual precipitation is 1144 mm. Soils of the study sites are base-rich loamy gleysols; they are waterlogged or flooded in winter but relatively dry in summer; the soil pH is about 7 (Brülisauer & Klötzli 1998; S.G., unpublished data). All sites are mown annually in autumn. The vegetation belongs to the phytosociological alliance Molinion (Ellenberg 1996). It is moderately species-rich and reaches a peak above-ground biomass of 300–600 g dry matter m−2 in August (Güsewell & Klötzli 1998). Five Carex species and the grass Molinia caerulea were included in this study (nomenclature, Halliday & Beadle 1983). Three of the species occurred at all four sites, and the three other species at only two or three sites. A given species at a given site is hereafter called a ‘plant population’.
The three Dutch sites are floating fens near Utrecht at sea level (52°9′ N, 5°7′ E). The long-term mean annual temperature of the area is 9·0 °C, and the mean annual precipitation is 732 mm. Floating fens have developed through the terrestrialization of ponds created by peat excavation (van Wirdum, den Held & Schmitz 1992). The substrate is a floating mat of partly decomposed plant roots and rhizomes and Sphagnum mosses (Bakker, Jasperse & Verhoeven 1997). The pH of the sites ranges from 4 to 6, and the vegetation belongs to the phytosociological alliances Caricion davallianae and Caricion nigrae (Schaminée, Stortelder & Westhoff 1995). All sites have been mown every year in July or August for ≈30 years. Peak above-ground biomass ranges from 300 to 700 g m−2, of which up to 70% may be bryophyte mass.
The Dutch sites were fertilized experimentally in 1999 and 2000 (Güsewell, Koerselman & Verhoeven 2002). At each site 16 plots of 50 × 50 cm2 were assigned to 16 different treatments resulting from the factorial combination of: +N or –N in 1999; +N or –N in 2000; +P or –P in 1999; +P or –P in 2000. Thus some plots received the same element in both years, while others received different elements (Güsewell et al. 2002). In the +N treatments 20 g N m−2 year−1 were supplied as NO3NH4, and in the +P treatments 5 g P m−2 year−1 were supplied as NaH2PO4. The main purpose of the experiment was to investigate the effects of fertilization on vegetation biomass; therefore the vegetation was clipped within the central 40 × 40 cm2 of each plot in July 2000 (Güsewell et al. 2002). For this study the 10-cm-wide strips bordering the central quadrat were sampled in November 2000; this marginal zone was visibly influenced by fertilizer treatments. Study species were Anthoxanthum odoratum, Carex acutiformis, Carex curta, Carex diandra and Eriophorum angustifolium, the five most abundant graminoid species at these sites.
On 24–25 May 2000, a line 3–6 m long was laid out within each site at a spot where the studied species co-occurred, and 24 shoots of each species were selected along the line. Twelve shoots were tagged with numbered plastic labels, and their youngest fully expanded leaf (usually the third visible leaf from the top of the shoot) was tagged with a small piece of red or yellow plastic foil. The length of this leaf (from the tip of the ligule to the tip of the leaf) was measured to the nearest millimetre, and the maximum width was measured to the nearest 0·2 mm. On each of the 12 other shoots, the youngest fully expanded leaf was measured similarly, then removed and taken to the laboratory. Leaf mass was determined after 48 h drying at 70 °C, and leaves were kept for nutrient analyses.
Tagged leaves were monitored after 4, 8 and 13 weeks. Once they had fully senesced up to the ligule (grey-brown colour and dry appearance), they were cut off, taken to the laboratory, dried and weighed. The length and width of those senesced leaves harvested at week 13 (end of August) was measured again in the field to determine whether leaf area changed during senescence.
For nutrient analyses of green and senesced leaves, four to six leaves of the same species and site were always pooled to obtain a sufficient amount of material. They were digested with a modified Kjeldahl procedure (1 h at 200 °C and 2 h at 340 °C in a mixture of concentrated sulphuric acid, salicylic acid, copper and selenium). Concentrations of total N and P in the digests were determined colorimetrically on a continuous-flow analyser (Skalar SA-40, Skalar, Breda, the Netherlands).
A second set of leaves were measured, weighed and analysed in July 2000 (same sample size and procedure as in May), but only few tagged leaves senesced before the end of August, when the study had to stop due to the mowing of sites. To obtain a measure of nutrient concentrations in senesced leaves for the July cohort, the youngest fully senesced leaf blade from five shoots per species and site was also collected at the end of August and analysed for N and P as described above; these leaves had approximately the same age as the tagged ones.
Sampling at the Dutch sites was restricted to a single harvest in November 2000, after completion of the fertilization experiment. Despite the late sampling date, nutrient concentrations in mature leaves did not differ significantly from those of the same plots in July 2000 (Güsewell et al. 2002). Plant shoots were divided into mature green and entirely senesced leaves or leaf parts; young (expanding) leaves and material in an intermediate state were discarded. The length and width of 15 mature and 15 recently senesced leaves of each species were measured. The mass of these leaves was determined after 48 h drying at 70 °C. The other material was dried and analysed for N and P.
Changes in leaf size during senescence were estimated for individual tagged leaves as the percentage difference between the length, width and area of the leaf when it was green (LG, WG, AG) and after senescence (LS, WS, AS). In these calculations, leaf area was estimated by the product of length and width (AG ≈ PG = LGWG; AS ≈ PS = LSWS), after checking that the two variables were proportional to each other on a separate set of leaves (PG = 1·1AG). As the size of senesced leaves was measured only for some of the tagged leaves (those senesced in August), data from the four sites were pooled for each species, and two-sided paired t-tests were used to test whether the average size changes differed significantly from zero.
Changes in leaf mass during senescence were determined for each species and site as the percentage difference between the mean mass of the 12 leaves harvested when they were green (MG) and of those harvested after senescence (MS). To obtain accurate means despite variation in leaf size, the mass of each leaf (MG or MS, in g) was first standardized by the product of its length and width (PG). MS was divided by PG and not by PS so that mass loss would not be underestimated if leaf size decreased during senescence (van Heerwaarden et al. 2003). For a given species and site, both MG and MS were proportional to PG (r2 = 0·90), so that the ratios MG/PG and MS/PG were independent of leaf size. Thus, for each species and site, mass loss was estimated accurately as:
(eqn 1 )
where horizontal lines above symbols indicate means per species and site. Differences among species and sites were tested with two-way anova.
Leaf-level nutrient resorption efficiencies (NRE, PRE) were estimated for each species and site as the relative changes in the mean N and P contents of a leaf (NC, PC, in mg leaf−1):
(eqn 2 )
As NCG = [N]GMG and NCS = [N]SMS, with [N] (in mg g−1) being the mean N concentration per species and site, the formula for NRE can be rewritten as:
(eqn 3 )
Again, the ratio MS/MG was estimated for each species and site from means of leaf mass standardized by leaf size: . The mean NRE per species and site was therefore given by:
(eqn 4 )
Calculations were identical for PRE, replacing [N] with [P]. As the ratio MS/PG could not be determined accurately for the July cohort (too few tagged leaves senesced before the last sampling), values for MG/PG and MS/PG from the May cohort were used to calculate NRE and PRE for both leaf cohorts. As RE was generally ≥ 80%, possible differences in mass loss between May and July would hardly have affected the results (van Heerwaarden et al. 2003). Finally, the N : P resorption ratio was calculated for each species and site as the ratio of the mean amounts of N and P resorbed per leaf:
( eqn 5 )
Differences among species, sites and leaf cohorts in nutrient concentrations (log-transformed) as well as NRE and PRE were tested with repeated-measures anova.
At the Dutch sites the length and width of senesced leaves were measured only after senescence. The change in leaf size during senescence therefore could not be determined. Mass loss was estimated using equation 1 and NRE and PRE using equation 4, but MS/PG was replaced with MS/PS. This corresponds to the usual calculations of NRE and PRE based on N or P concentrations per leaf area, which slightly underestimate the true values if leaf area decreases during senescence (van Heerwaarden et al. 2003).
Relationships among nutrient concentrations in green and senesced leaves and nutrient resorption were represented by scatter plots and quantified with Pearson's correlation coefficients (r), based on log-transformed data for nutrient concentrations and N : P ratios. Four types of relationship were considered. (1) Overall correlations were calculated from individual data points, separately for Swiss sites (two cohorts) and Dutch sites. (2) Interspecific correlations were calculated from species means. For the Dutch sites, species means across all plots were always closely similar to those across unfertilized plots, so that results were not affected by including fertilized plots. For C. acutiformis, the only species sampled in both countries, means for Swiss and Dutch sites were calculated separately. (3) Intraspecific correlations were calculated after verifying that, for all relevant pairs of variables, model-I regression slopes of one variable against the other did not differ significantly among species (i.e. no interaction in an analysis of covariance). All variables were then adjusted by subtracting species means, and Pearson's correlation coefficients were calculated from the adjusted data. (4) The relationship between NRE and PRE was further described with a model-II regression, which corresponds to the principal axis of the distribution of the two variables (Sokal & Rohlf 1995). Model-II regressions minimize error variation in both x and y dimensions, and therefore describe bivariate data more accurately than ordinary regressions. Data from Swiss sites (July cohort) and Dutch sites were included in the calculation. All analyses were carried out with the statistical package jmp ver. 3·2·2 (SAS Institute 1993–2000). Pairwise significance levels are reported for correlations because each tested for a different type of relationship.
changes in leaf size and leaf mass during senescence
The six species investigated in Swiss fen meadows differed considerably in leaf size and mass, whereas differences among the four sites were small (data not shown). During senescence, leaf length (May cohort) changed by <6% in all species, whereas leaf width decreased by 3–17%, and estimated leaf area by 6–19% (Table 1). The decrease in leaf area was significant (P < 0·05) in five of six species.
Table 1. Changes in size and mass of leaf blades during senescence, as a percentage of initial values, for six graminoid species in Swiss fen meadows
Figures are means ± SD of nL leaves (for size) or of nS sites (for mass) and the significance of the change (two-sided paired t-tests; ***, P < 0·001; **, P < 0·01; *, P < 0·05; °, P < 0·1).
Leaves from the May cohort lost 17–38% of their initial dry mass during senescence. Species means (20–34%; Table 1) differed (two-way anova, P < 0·05), whereas site means did not (P > 0·05). In Dutch fens, leaf mass loss was 33·5% in A. odoratum, 17% in C. curta, and 8·5–9% in the three other species.
nutrient concentrations and resorption efficiency
In Swiss fen meadows, the N concentration of green leaves ([N]G) differed between May and July cohorts (on average 45·5% lower in July), but not among species or sites (Table 2). The P concentration ([P]G) differed among species, sites and cohorts (18·5% smaller in July; Table 2). Nutrient concentrations of senesced leaves differed among species but not among sites or cohorts (Table 2). As a result, NRE and PRE were higher in the May cohort than in the July cohort. NRE differed among species while PRE did not, and there were no consistent differences among sites (Table 2).
Table 2. N and P concentrations (mg g−1) in green and recently senesced leaf blades of six graminoid species, determined for two leaf cohorts (May and July 2000) in Swiss fen meadows
Nutrient concentrations (mg g−1)
Figures are means ± SD of nS sites (see Table 1 for nS), F ratios and significance levels from repeated-measures anova, and correlations (Pearson's r) between the two leaf cohorts, calculated across species and sites. Significance levels are as in Table 1.
17·2 ± 2·4
5·5 ± 0·4
0·78 ± 0·08
0·12 ± 0·07
78·0 ± 3·9
88·0 ± 3·2
18·3 ± 2·4
7·1 ± 1·4
0·78 ± 0·09
0·09 ± 0·10
72·3 ± 7·1
88·3 ± 9·3
16·1 ± 3·1
5·1 ± 0·4
1·10 ± 0·41
0·19 ± 0·02
77·9 ± 5·8
88·4 ± 3·7
19·9 ± 2·5
7·1 ± 1·3
1·18 ± 0·12
0·25 ± 0·02
75·1 ± 2·7
84·4 ± 1·2
17·2 ± 2·0
4·7 ± 1·8
0·81 ± 0·21
0·12 ± 0·05
82·2 ± 3·9
90·6 ± 4·1
18·5 ± 0·7
3·9 ± 1·2
0·79 ± 0·05
0·06 ± 0·03
82·6 ± 3·4
93·5 ± 2·9
11·0 ± 0·8
5·2 ± 1·1
0·61 ± 0·12
0·14 ± 0·06
65·9 ± 3·3
84·7 ± 2·7
10·3 ± 0·4
6·3 ± 1·0
0·40 ± 0·07
0·11 ± 0·07
50·3 ± 11·1
74·3 ± 15·2
12·4 ± 0·1
4·3 ± 0·1
1·11 ± 0·02
0·15 ± 0·01
71·9 ± 2·7
88·7 ± 0·2
12·2 ± 1·8
6·7 ± 0·3
0·96 ± 0·09
0·26 ± 0·05
59·5 ± 1·2
80·7 ± 2·2
11·7 ± 0·8
4·4 ± 0·7
0·71 ± 0·09
0·10 ± 0·08
74·1 ± 5·3
89·3 ± 4·6
12·6 ± 0·6
4·0 ± 1·2
0·64 ± 0·05
0·06 ± 0·07
74·7 ± 3·6
92·1 ± 3·2
Species (df = 5)
Site (df = 3)
Month (df = 1)
Month × species
Month × site
In Dutch fens, nutrient concentrations of green and senesced leaves varied more among species than at the Swiss sites, and the P concentration was generally higher (Table 3). The estimated NRE and PRE were more variable and, on average, lower than at the Swiss sites (Table 3).
Table 3. N and P concentrations (mg g−1) in green and in recently senesced leaf blades of five graminoid species in Dutch fens
Nutrient concentrations (mg g−1)
Figures are means ± SD of n plots which had been partly fertilized with N and/or P during the two preceding years.
18·8 ± 3·4
12·0 ± 2·3
2·82 ± 0·91
1·36 ± 0·48
56·9 ± 8·8
67·9 ± 9·1
11·8 ± 1·7
10·0 ± 1·9
1·20 ± 0·52
0·59 ± 0·24
20·9 ± 18·1
52·0 ± 10·5
15·3 ± 2·9
8·9 ± 2·2
2·18 ± 0·66
1·04 ± 0·56
49·2 ± 8·0
60·8 ± 11·9
12·5 ± 1·6
8·8 ± 1·1
1·77 ± 0·77
1·08 ± 0·55
36·0 ± 5·5
47·2 ± 8·6
15·7 ± 2·4
8·6 ± 3·6
1·97 ± 0·96
0·52 ± 0·36
51·3 ± 16·8
73·7 ± 17·0
correlations among variables
Nutrient concentrations of green and senesced leaves correlated positively with each other, except for [N]G and [N]S in the Swiss fen meadows (Fig. 1a,b). Inter- and intraspecific relationships were similar: both were stronger for P than for N (Table 4a). NRE correlated positively with [N]G in the Dutch fens but not in the Swiss ones (Fig. 1c). PRE and [P]G did not correlate with each other within the data sets (Fig. 1d), but a weak negative interspecific correlation (Table 4a) reflected the above-mentioned differences in PRE and [P]G between Swiss and Dutch sites.
Table 4. Inter- and intraspecific correlations among nutrient concentrations in green and senesced leaves and nutrient resorption, as shown in Figs 1–3
Pairs of variables correlated are indicated in the first column. Correlations (Pearson's r) were calculated for Swiss sites (July cohort) and Dutch sites together, first from species means (interspecific correlations, n = 11), then from data adjusted for species means (intraspecific correlations, n = 94). Abbreviations: [N], [P] = N or P concentrations in mg g−1; G = green leaves; S = senesced leaves; NRE, PRE = N or P resorption efficiency in percentage; N : P = N : P ratio; N : Pres = ratio of amounts of N and P resorbed. Pairwise significance levels are indicated by symbols as in Table 1.
Resorption efficiency correlated positively with leaf mass loss in the Dutch fens (r = 0·52 for NRE; 0·23 for PRE). In the Swiss fen meadows there was no correlation across species, but within species NRE and PRE were also positively related to mass loss (ancova, P < 0·01).
N and P concentrations correlated positively with each other in green leaves (Fig. 2a) and senesced leaves (Fig. 2b). An even stronger correlation occurred between NRE and PRE (Fig. 2c). This was most apparent at the intraspecific level: although [N]G and [P]G correlated poorly within species, NRE and PRE were still strongly correlated (Table 4b). However, PRE was always greater than NRE, as the relationship between NRE and PRE (model-II regression line) was given by PRE = 1·003NRE + 17·1%, r = 0·91.
The N : P ratios of green leaves, senesced leaves and nutrient resorption correlated strongly with each other in all types of comparison (Fig. 3; Table 4c). Because PRE generally exceeded NRE, N : P was mostly greater in senesced leaves than in green leaves (Fig. 3a), whereas the N : P resorption ratio (equation 5) was slightly smaller than the N : P ratio of green leaves (Fig. 3b).
Nutrient resorption efficiency (NRE and PRE) did not generally correlate with the N : P ratios of green leaves within the Swiss or Dutch data sets (Fig. 3c,d), nor did the ratio of NRE to PRE (P > 0·05, not shown). However, positive interspecific correlations between N : PG and NRE or PRE (Table 4c) reflected the fact that all three variables had larger values in Swiss fen meadows than in Dutch fens. The absence of intraspecific correlations (Table 4c) indicates that individual species did not directly adjust their NRE and PRE to the relative degree of N vs P limitation.
nutrient resorption in graminoids and woody plants
Nutrient resorption during senescence has been studied mainly for woody plants. This study has shown a close similarity between graminoids and woody plants regarding changes during leaf senescence. Leaf mass loss in the graminoids ranged from 8 to 38%, which is close to the range of 3–37% reported for woody species by van Heerwaarden et al. (2003). An even broader range of mass loss (0–60%) was found by Chapin & Kedrowski (1983) for boreal trees, but only few values exceeded 40%. The 5–19% decrease in leaf area found here for graminoids compares well with the 4–22% area loss reported by van Heerwaarden et al. (2003) and Peng & Wang (2001) for woody species. Leaf area decreased mainly through a shrinkage of leaf width, whereas leaf length changed little. This supports the proposal of van Heerwaarden et al. (2003) that, in graminoids, nutrient concentrations per leaf length may be a good basis for calculating RE if the change in leaf size during senescence cannot be monitored. Nutrient concentrations per needle length have also been used to calculate nutrient resorption in coniferous trees (Enoki & Kawaguchi 1999).
Nutrient resorption efficiency varied widely among and within species, with a range from 0 to nearly 100%. The uppermost values exceeded the maximal RE (about 80% of N and 90% of P) reported for woody plants (Walbridge 1991; Wright & Westoby 2003). However, only leaf blades were sampled here for practical reasons (part of the sheaths were in the soil or waterlogged moss layer, and senesced later than blades). Leaf sheaths have a lower RE than blades (Aerts 1989). Based on entire leaves, Güsewell (2005) found ≈10% lower RE than in the present study for five of the Carex species. This suggests that maximal RE from entire leaves does not differ between graminoids and woody plants. Nevertheless, the average RE of graminoids is higher (Aerts 1996), which may reflect contrasting selective forces. The main adaptation of woody plants to low nutrient availability is an increased leaf life span (Escudero et al. 1992; Wright & Westoby 2003). In contrast, most graminoids have short-lived leaves (Aerts & de Caluwe 1995; Ryser & Urbas 2000), so that RE plays a greater role than life span for their nutrient conservation (Aerts & Chapin 2000).
resorption efficiency not determined by leaf nutrient concentrations
Resorption efficiency was not consistently related to nutrient concentrations in green leaves, as has often been demonstrated for trees (Del Arco et al. 1991; Wright & Westoby 2003). In particular, high N or P concentrations in green leaves did not result in low NRE or PRE, as should be expected if plants adjusted their RE to their nutrient status. Downregulation of PRE from P-rich leaves has often been observed in trees grown on very P-rich soils or after P fertilization (Chapin & Moilanen 1991; Uliassi & Ruess 2002). In a glasshouse experiment with five wetland Carex species, PRE was also strongly reduced when foliar P concentrations exceeded 2 mg g−1 (Güsewell 2005). The same did not hold here for the Dutch fens, where even plants with P concentrations of 3–4 mg g−1 had a relatively high PRE. Likewise, the forest herb Claytonia virginica had a PRE of 87% with a foliar P concentration of nearly 6% (Anderson & Eickmeier 2000). Thus even extremely high P concentrations do not necessarily downregulate PRE.
Although N and P concentrations in green leaves correlated only moderately with each other, there was a strong positive correlation between NRE and PRE (Fig. 2). An equally strong correlation between NRE and PRE was obtained by Aerts (1996) in a meta-analysis of published data. The present study shows that this relationship holds among and within species (Table 4b) and for N- and P-limited plants, as indicated by the broad range of biomass N : P (Fig. 3). Contrary to my initial hypothesis, N-limited plants (with low N : P) did not specifically increase their NRE (in absolute terms or relative to PRE), nor did P-limited plants (with high N : P) specifically increase their PRE. Apparently, RE was not regulated by the type of nutrient limitation as reflected by foliar N : P ratios. Instead, nutrient concentrations of senesced leaves correlated positively with those of green leaves. This nutritional control was stronger for P than for N (to judge from correlation coefficients), but not stronger at the P-limited Swiss sites than at the N-limited Dutch sites, so that it was not specific to the type of nutrient limitation either.
The positive correlation between leaf N concentration and NRE at the Dutch sites might be interpreted as indicating that NRE was determined by the fraction of the leaf's N allocated to labile, metabolically active compounds such as photosynthetic enzymes and pigments (Nordell & Karlsson 1995). However, there was no correlation at the Swiss sites, and other studies that considered the fractions of N involved in resorption concluded that most of the leaf's N and P can be solubilized, so that RE is generally not limited by the availability of labile compounds (Chapin & Kedrowski 1983; Cartaxana & Catarino 2002; Yasumura et al. 2005). In those studies RE rather appeared to depend on the flow of solutes from senescing biomass, and thus on the duration of the senescent stage (Del Arco et al. 1991; Silla & Escudero 2004), or on the strength of sinks for assimilates (growing parts or storage tissues: Chapin & Moilanen 1991; Yasumura et al. 2005). The high RE and low nutrient concentrations in litter found at the Swiss sites further indicate that most N and P contained in green leaves of graminoids can potentially be resorbed. Moreover, RE correlated positively with leaf mass loss, supporting control by the total outflow of solutes (Chapin & Kedrowski 1983).
differences between swiss and dutch sites
Resorption efficiency was generally greater at the Swiss than at the Dutch sites. Despite similar above-ground biomass production, plants at the Swiss sites had lower P concentrations and higher N : P than plants at the Dutch sites, suggesting that they were more P-limited. Fertilization experiments directly demonstrated N or N + P limitation at the Dutch sites (Güsewell, Koerselman & Verhoeven 2003) and P limitation at Swiss sites similar to those investigated here (Egloff 1983). Aerts & Chapin (2000) hypothesized that nutrient resorption is more important for nutrient conservation at P-limited than at N-limited sites. Indeed, reports of very high RE (>80%) mainly concern vegetation with high N : P ratios, suggesting growth limitation by P (Walbridge 1991; Wright & Westoby 2003; McGroddy, Daufresne & Hedina 2004). Could this explain the difference in RE between Swiss and Dutch sites?
It is first necessary to examine whether the difference in RE might simply reflect the different sampling methods. Changes in leaf area during senescence were not accounted for at the Dutch sites. If leaf area decreased by 5–20% at the Dutch sites, similar to the Swiss ones, RE would have been underestimated by 5–20% in plant populations with an apparent RE of 0%, but only by 1–4% in plant populations with an apparent RE of 80% (van Heerwaarden et al. 2003). Thus accounting for leaf area loss would only marginally reduce differences between Swiss and Dutch sites. Plant material was sampled later in the year at the Dutch sites, but nutrient concentrations in biomass did not differ between July and November (see Methods). The lower RE of Dutch plants therefore could not be explained by export of nutrients from leaves during the autumn while they were still green. The oceanic climate of the Netherlands enables plants to grow during most of the winter, but photosynthesis is certainly restricted by the low irradiation. A seasonal reduction in photosynthesis and growth would affect RE if the current carbon assimilation determines the sink effect of growing plant parts (Chapin & Moilanen 1991). Insufficient data are available to assess how important this effect might be. Shading of birch (Betula papyrifera) leaves to prevent photosynthesis caused a 20–30% decrease in RE (Chapin & Moilanen 1991), whereas RE did not differ between sun and shade leaves in a beech forest (Yasumura et al. 2004).
Despite a possible seasonal bias, differences between Swiss and Dutch sites probably reflected inherent differences among the species or populations sampled in each region. Indeed, similar interspecific differences were also found in other studies without differences in sampling time. In a glasshouse experiment, RE was greatest in Carex panicea and Carex flacca, intermediate in Carex flava, and lowest in C. curta and Carex elata, which corresponds to their ranking in the present study (Güsewell 2005). In a Dutch heathland, RE of M. caerulea (foliar N : P = 33) between September and November (NRE = 74–75%, PRE = 80–87%) resembled that between July and August in the Swiss fen meadows (Aerts 1989). Conversely, in a Swiss fen meadow the RE of A. odoratum (foliar N : P = 19) estimated from nutrient concentrations in August 2004 (NRE = 73%, PRE = 54%) resembled that found in November 2000 at the Dutch sites (unpublished data). Among these species, those that dominate in P-limited wetlands (e.g. Carex panicea, Molinia caerulea) consistently had a higher RE than those that dominate in N-limited wetlands (C. curta, C. elata, A. odoratum).
Carex acutiformis had a much higher RE at the Swiss sites than at the Dutch ones, in apparent contradiction with the suggestion of inherent differences among species. It is possible that genetic differences exist between populations from Swiss and Dutch sites. A comparable difference was observed for C. flacca: at the P-limited Swiss sites: its N and P concentrations decreased by 65 and 86%, respectively, during senescence (Table 2). In a Dutch dune slack, where biomass production is generally N-limited (Güsewell et al. 2003), N and P concentrations decreased by only 33 and 50%, respectively, indicating lower NRE and PRE (unpublished data). For several evergreen woody species, genetically determined differences in NRE were found among populations from sites with contrasting types of nutrient limitation (Treseder & Vitousek 2001) or with differing productivity (Oleksyn et al. 2003). In outcrossing species such as many graminoids, populations from geographically distant areas may be required for genetic differentiation to become apparent (Vellend & Waterway 1999; Stenström et al. 2001).
Even if NRE and PRE are not directly regulated by the type of limitation (no correlation with foliar N : P), differences in RE may have had contrasting implications for plant fitness at N- and P-limited sites. PRE exceeded NRE by, on average, 17% throughout the range of foliar N : P. Given the nonlinear relationship between RE and nutrient retention by plants with a given leaf life span (assuming that resorbed nutrients are effectively recycled; Escudero et al. 1992), the difference between NRE and PRE suggests that a parallel increase of NRE and PRE (e.g. by extending the senescent stage) would enhance P retention more than N retention (Fig. 4). Accordingly, plants at Swiss and Dutch sites differed only slightly in N retention, but differed greatly in P retention (Fig. 4). Graminoids that dominate in P-limited wetlands typically have long-lived below-ground storage organs (rhizomes, basal internodes), which act as sinks for solutes during leaf senescence (Aerts 1989; Güsewell 2005). In contrast, graminoid species typical of N-limited wetlands produce more fine roots below ground, and their leaves senesce rapidly when the plants are P-deficient (Güsewell 2005 and unpublished data), reducing sink strength and the duration of nutrient resorption. These contrasting traits may give a selective advantage to different graminoid species (or ecotypes) at N- and P-limited sites, as suggested by Aerts & Chapin (2000).
This study of nutrient resorption in wetland graminoids addressed two questions. First, are patterns of nutrient resorption found in woody plants also valid for graminoids? The results show that changes in leaf area, leaf mass and nutrient concentrations during senescence vary over a similar range in graminoids to that in woody plants. Second, is there evidence that plants adjust their RE to the type of nutrient limitation? At first sight, the answer is negative: NRE and PRE correlated strongly with each other, but not with foliar N : P. However, high RE appeared to improve P retention more than N retention, suggesting that plants benefit most from high RE when P limits their growth. This may have contributed to the dominance of graminoids with high RE in P-limited wetlands.
I thank I. Wright, J. Oleksyn, D. Uliassi and one anonymous referee for their helpful comments on a draft of the manuscript, J.T.A. Verhoeven and W. Koerselman for advice on the fertilization experiments, G. Rouwenhorst and P. van der Ven for assistance with nutrient analyses, P. van Bodegom for sampling C. flacca in a dune slack, and the landowners and conservation authorities (Amt für Natur und Landschaft Zürich, Staatsbosbeheer Utrecht) for access to the study sites. The research was funded by TMR grant ENV4-CT97-5075 from the Commission of the European Communities.