Nine wetland plant species were included in the experiment (nomenclature: Lauber & Wagner 1996): five grass species (Agrostis canina, Anthoxanthum odoratum, Alopecurus pratensis, Molinia caerulea, Phalaris arundinacea), two sedge species (Carex flava, Carex panicea) and two forb species (Centaurea angustifolia, Mentha aquatica). Further forb species were included initially but did not survive. In addition to representing three functional groups, the nine species were also selected because their typical habitats range from nutrient-rich, N-limited sites (A. pratensis, P. arundinacea) to nutrient-poor, P-limited sites (M. caerulea, C. flava, C. panicea). Plants were grown in spring 2003 from seeds (A. canina, A. odoratum, A. pratensis, C. angustifolia) or cuttings (M. aquatica, P. arundinacea, C. flava, C. panicea, M. caerulea) collected in wet areas near Zurich. When seedlings had two to three leaves or when cuttings started to form new roots, they were planted in 0·4 l pots (1 l pots for the tall species P. arundinacea) filled with quartz sand (0·1–0·7 mm; Carlo Bernasconi, Zürich) and placed on trays in a garden in Zurich at 450 m a.s.l.
Nutrient treatments were applied weekly from May until August 2003 and fortnightly in September 2003. There were 12 treatments, which combined six N : P supply ratios (N : Ps) with two supply levels in a factorial design. Treatments were defined by the total amounts of N and P (mg) supplied per plant during the growing season, which were calculated as:
with L, the supply level (geometric mean of N and P supply) being either 18 mg (high supply) or 6 mg (low supply), and N : Ps, the mass ratio of N to P supplied, forming a geometric series from 5 to 160. There were six replicates per species and treatment at high supply, and four replicates at low supply. Nitrogen was supplied as solutions of KNO3, and P as solutions of KH2PO4; KCl was added as required so that all plants received at least 250 mg K; the maximal K supply was 632 mg (as KNO3, at high supply with N : P = 160). Other essential nutrients (Ca, Mg, Fe, S, microelements) were supplied every 3 weeks to provide non-limiting amounts. Plants were watered daily or as required to keep the sand moist.
All dead leaves were removed a the end of June. Subsequently, freshly senesced leaves were collected every 2 or 3 weeks and stored air-dry in paper envelopes. As the amount of litter produced by individual plants was often small, the litter from replicate plants was pooled for the decomposition experiments, yielding 108 litter types (nine species × 12 treatments). The air-dry material was cut in pieces ≈ 0·5 cm long to obtain homogeneous samples for incubations; the length of pieces corresponded to the (unchanged) width of leaves (mostly 0·2–0·5 cm).
We harvested two replicates of each plant grown at high supply in October 2003 to determine their total biomass and nutrient concentrations of the above-ground living tissues (leaves and culms). The other plants were kept for further measurements in a second growing season. Agrostis canina was removed from the experiment during the summer because its stolons grew into neighbouring pots while its tiny leaves were difficult to harvest. The litter already collected was still used for enzyme assays.
For nutrient analyses, a 150 mg subsample of each litter type was dried at 75 °C for 24 h to determine the mean water content of the air-dry material (6%). This subsample was then digested with a modified Kjeldahl procedure (1 h digestion at 420 °C with 98% H2SO4 and a copper sulphate-titanium oxide catalyst). Concentrations of N and P in digests were determined colorimetrically on a flow-injection analyser (FIA, Tecator, Höganäs, Sweden). The same procedure was used for living tissues.
Litter was incubated in Petri dishes (6 cm diameter) on 18 g quartz sand covered with a disk of polyethylene mesh (300 µm). Eleven Petri dishes, each containing 150 mg air-dry litter, were prepared with each litter type if enough material was available. Two Petri dishes served for enzyme assays (see below), and nine to determine the initial rate of mass loss as well as its response to N or P addition (three treatments with three replicates). If the amount of material was insufficient, we prepared as many Petri dishes as possible and attributed them to incubations in the following sequence: (1,2) enzyme assays; (3) mass loss without nutrient addition; (4,5) mass loss with addition of N or P; (6,7) replicated mass loss without nutrient addition; (8–11) replicated mass loss with nutrient addition. For example, if seven Petri dishes could be prepared with a litter type, two were used for enzyme assays, three for mass loss without nutrient addition, and two for mass loss with addition of N or P, respectively. Enzyme assays were carried out with all litter types (no missing data). Mass loss could be determined with 45 of the 54 litter types from plants grown at the high supply level; only these were included in the data analysis as there were few data from plants grown at the low supply level.
The litter in Petri dishes was wetted with 8 ml of a microbial inoculum prepared by mixing 1 kg fresh topsoil from a wetland where seven of the species co-occurred with 3 l of deionized water. After 12 h with repeated stirring, the slurry was filtered (coarse filter paper, LS171/2, Schleicher & Schuell, Dassel, Germany), and the filtrate (pH 7) was used to inoculate the samples. Nutrient amendment consisted of either 8 mg N (as NH4NO3) or 1·5 mg P (as KH2PO4) per Petri dish; these nutrients were dissolved in the soil filtrate before wetting the litter. To ensure equal K supply in all treatments, 1·5 mg K as KCl was added to the control and +N treatments. Control incubations with cellulose as substrate confirmed that a good microbial inoculation was achieved with the filtrate (rapid decomposition when N + P were added), while the nutrient concentrations in the filtrate were negligible (almost no decomposition without nutrient addition, or when N or P was added alone).
Petri dishes were incubated in boxes at 22 °C. The litter was sprinkled with deionized water every 2 weeks to replace the evaporated moisture and to simulate leaching by rainfall. After 5 weeks (half the period over which mass loss was determined), the litter samples intended for microbial assays were removed from the Petri dishes with tweezers and kept in aluminium foil at 5 °C. After 10 weeks the litter was removed from the other Petri dishes and dried for 24 h at 70 °C to determine the remaining dry mass and thus percentage mass loss. The artificial conditions meant that mass loss was a measure of litter decomposability rather than actual decomposition; they were chosen here because we needed a nutrient-free control treatment and small error variation to be able to assess nutrient limitation for decomposition.
microbial respiration and enzyme assays
Microbial respiration and enzyme activities were assayed within 6 days of removing the litter from Petri dishes. To measure microbial respiration, litter samples were placed in 10 ml glass vials in darkness at 20 °C. After 24 h a 5 ml gas sample was drawn, and its CO2 concentration was analysed on a gas chromatograph (AMS Model 92, Analytical Measuring Systems, Cambridge, UK). The CO2 concentration of the gas was converted into CO2 production per initial g litter dry mass and hour of incubation. Using mass loss data (if available) we also estimated CO2 production per g litter present at the time of the measurement, but do not report these values here as they showed virtually the same species and treatment effects.
Enzyme activities were determined following Sinsabaugh et al. (1991); Kang & Freeman (1999); Kourtev et al. (2002). After respiration measurements, the litter was mixed for 60 s with 10 ml deionized water in a laboratory blender (Stomacher Type 80, Seward, UK). No buffer was used because species- or treatment-related differences in pH and their possible influence on enzyme activities were of interest here. Immediately after mixing, six 1 ml aliquots of the extracts (without large litter pieces) were transferred to six Eppendorf reaction vials. Three enzymes were assayed in duplicate by adding 0·5 ml of the appropriate substrate solutions to each of two vials. Substrate solutions were 400 µm 4-methylumbelliferyl (MUF)-β-d-glucoside (Sigma M-3633) for the activity of β-glucosidase; 400 µm 4-MUF-N-acetyl-β-d-glucosaminide (Sigma M-2133) for the activity of chitobiase; and 200 µm 4-MUF-phosphate (Sigma M-8883) for the activity of acidic and alkaline phosphatases (Sigma-Aldrich Chemie, Steinheim, Germany); each substrate was first dissolved in 20 ml cellosolve (2-ethoxyethanol) and then diluted with water to obtain a final cellosolve concentration of 2%. After 2 h incubation at 20 °C, the vials were centrifuged at 10 000 r.p.m. for 4 min, and the fluorescence of the supernatant was measured immediately on a microplate reader (FLUOstar Galaxy, BMG Labtechnologies, Offenburg, Germany) at 460 nm emission and 355 nm excitation wavelength. For the calibration curve, 0·5 ml extract from each litter sample in the current assay (24 samples per batch) were pooled and centrifuged for 4 min at 10 000 r.p.m. The supernatant was used to dilute a stock solution of 1000 µm 4-methylumbelliferone in 2% cellosolve to six concentrations between 0 and 80 µm, so that the calibration curve would account for possible interactions between MUF and other compounds in the litter extracts. The incubation period and substrate concentrations were based on preliminary assays; control assays with enzyme substrates added to deionized water showed no fluorescence. Enzyme activities were expressed as µmol of substrate converted per h and per g initial dry mass of litter. The pH of all litter extracts was measured with a glass electrode (Model 720A, Orion Research, Boston, MA, USA).
Most variables were log-transformed before statistical analysis to obtain normally distributed error terms with homogeneous variance (plant biomass, pH and percentage mass loss were analysed untransformed). The effects of plant species, N : P supply ratios and supply level on microbial activity after 5 weeks’ decomposition were analysed with three-way fixed-factor anova. To obtain a balanced design, means of the two replicates were used in data analysis, and the three-way interaction was used as error term. For response variables that were determined only at the high supply level (plant biomass, litter mass loss), two-way anova (factors species and N : P supply ratios) was used. As some species–treatment combinations were missing for mass loss after 10 weeks, the interaction term was removed; the analysis used type III sums of squares, which account for an unbalanced design.
To determine if decomposition was limited by N or P, we calculated for each litter type the difference between the mean dry mass loss with N or P addition and the mean dry mass loss without nutrient addition. For some litter types differences could not be tested statistically as there were only one or two replicates per treatment. We therefore regarded as significant a difference in mass loss of 5%, which corresponds to the least significant difference for litter types that had two or three replicates per treatment (Dunnett test; α = 0·10). This threshold was a compromise between power (ability to detect nutrient limitation when it existed) and significance (avoid inferring nutrient limitation from random differences), as both were equally important to assess if nutrient limitation was related to N : P ratios.
Correlations among enzyme activities, microbial respiration and pH were quantified with Pearson's correlation coefficients (r) in three ways: (1) overall correlations based on individual samples; (2) interspecific correlations based on species means across treatments; (3) intraspecific correlations based on individual samples after adjusting for species means. Correlations between microbial activity after 5 weeks and mass loss after 10 weeks or initial litter chemistry were calculated in the same way, but based on means per species and treatment, as only these could be matched. All analyses were carried out with the statistical package jmp ver. 3·2·2 (SAS Institute, Cary, NC, USA).