soil analysis: field sites
Soil parameters were tested to determine if they were correlated; five pairs of soil parameters were significantly correlated while three were not. Despite the lack of complete correlation, manova was used to analyse the soil parameters because they were determined from the same soil samples and therefore could not be considered independent. Potassium, total nitrogen, organic matter, nitrate-nitrogen and magnesium were transformed to meet the assumption of multivariate normality. The manova showed significant differences among sites (Wilks’ lambda, λ= 0·022, P < 0·0006, d.f. = 16,16). Univariate protected anovas (Goldberg & Scheiner 1993) were used to determine which response variables contributed to the significance of the model. Potassium (F = 18·6, P < 0·001, d.f. = 2,15), phosphorus (F = 5·7, P < 0·015, d.f. = 2,15), total nitrogen (F = 6·6, P < 0·009, d.f. = 2,15), organic matter (F = 3·8, P < 0·046, 2,15), magnesium (F = 7·14, P < 0·006, d.f. = 2,15) and nitrate-nitrogen (F = 4·59, P < 0·028, d.f. = 2,15) were significantly different among sites. A Ryan–Einot–Gabriel–Welsch (REGWQ) (SAS Institute 1997) multiple range test was used to determine which sites accounted for the differences within soil variables. Banner Marsh had lower total nitrogen, phosphorus and potassium than Lincoln and Herrin (Table 1). Magnesium levels were significantly lower for Banner Marsh and Herrin than they were for Lincoln.
Table 1. Mean (± 1 SE) soil parameters for the three study sites. Within a row means with the same letters are not significantly different
|Organic matter (%)|| 4·1 ± 1·1a|| 3·8 ± 0·6a|| 2·1 ± 0·1a|
|Total nitrogen (%)|| 0·2 ± 0·06a|| 0·2 ± 0·04a|| 0·08 ± 0·01b|
|Nitrate-nitrogen (p.p.m.)|| 3·7 ± 2·9a|| 3·2 ± 1·4a|| 0·2 ± 0·02a|
|Phosphorus (p.p.m.)|| 38·3 ± 12·5a|| 55·5 ± 12·6a|| 7·3 ± 0·7b|
|Magnesium (p.p.m.)|| 267·5 ± 73·0a|| 534·2 ± 41·8|| 321·7 ± 14·1a|
|Potassium (p.p.m.)|| 134·2 ± 15·9a|| 169·1 ± 25·5a|| 67·8 ± 1·9b|
|Calcium (p.p.m.)||3170·0 ± 360·8a||2975·0 ± 188·4a||3517·5 ± 26·1a|
|pH|| 7·7 ± 0·04a|| 7·5 ± 0·06a|| 7·8 ± 0·09a|
above-ground and below-ground biomass
To compare the relative response of P. australis to S. pectinata under high- and low-nitrogen conditions, a ratio of P. australis above-ground biomass (stem plus flower biomass) to S. pectinata biomass was made. A two-way anova was used to compare the ratio under the different nitrogen treatments and populations. The data were log transformed, although normality could not be achieved (W = 0·97, P < 0·012). The anova model was significant (F = 4·60, P < 0·0007, d.f. = 5,113) and the sources of nitrogen treatment (F = 11·02, P < 0·0012, d.f. = 1,113) and population (F = 5·24, P < 0·006, d.f. = 2,113) were significant, but the interaction was non-significant (F = 0·75, P < 0·47, d.f. = 2,113). The mean ratio of P. australis to S. pectinata above-ground biomass was 2·72 ± 0·499 for the high-nitrogen treatment and 1·83 ± 0·42 for the low-nitrogen treatment. The two-way anova model used to compare the below-ground biomass ratio of P. australis to S. pectinata between nitrogen treatments and populations was non-significant (F = 1·94, P < 0·12, d.f. = 5,25).
Two mancovas were used to determine if there were differences in P. australis and S. pectinata stem and inflorescence biomasses between nitrogen treatments (Scheiner 1993). Stem and inflorescence biomasses were correlated for P. australis (Pearson correlation, r= 0·21, P < 0·019) and S. pectinata (Pearson correlation, r= 0·52, P < 0·0001). Phragmites australis stem biomass and S. pectinata stem biomass were correlated (Pearson correlation, r=−0·25, P < 0·0055); however, P. australis inflorescence biomass and S. pectinata inflorescence biomass were not correlated. Due to the incomplete correlation between P. australis and S. pectinata response variables, they were analysed with separate mancovas. The P. australismancova used P. australis stem and flower biomasses as main response variables and P. australis initial rhizome biomass as a covariate. The mancova for S. pectinata used S. pectinata stem and inflorescence biomasses as response variables and initial height of longest leaf as a covariate. Nitrogen treatment and population were used as fixed-effect independent variables for both mancovas. Although the S. pectinata plants were purchased from a nursery rather than being collected from different populations, population was included as a factor in the analysis because S. pectinata was grown with P. australis from different populations. The homogeneity of slopes assumption of ancova was met for both analyses. A Bonferroni correction was used to compensate for the multiple (two) mancovas; the accepted alpha level was 0·025 (Sokal & Rohlf 1995). Protected univariate ancovas were used as follow-up tests for significant mancovas and a REGWQ test was used to test for differences in means for the significant univariate ancovas.
The overall model mancova on P. australis stem and flower biomasses was significant for population (Wilks’ lambda, λ= 0·840, P < 0·0005, d.f. = 4,228), and nitrogen treatment (Wilks’ lambda, λ= 0·541, P < 0·0001, d.f. = 2,114), and the interaction between population and nitrogen was non-significant (Wilks’ lambda, λ= 0·936, P < 0·1109, d.f. = 2,228); therefore univariate ancovas were run. The overall univariate model for the P. australis stem biomass ancova was significant (F = 21·9(α=0·025), P < 0·0001, d.f. = 6,115). There were significant differences in P. australis stem biomass with respect to population and nitrogen treatment, but there was no interaction between site and nitrogen treatment (Table 2). The REGWQ test revealed that mean stem biomass for P. australis from the Herrin site (60·7 ± 3·8 g) was significantly higher than P. australis from the Lincoln (50·4 ± 3·9 g) and Banner Marsh (47·4 ± 3·2 g) sites. Lincoln and Banner Marsh were not significantly different from each other (Fig. 2). The REGWQ test also revealed significant differences between nitrogen treatments: P. australis stem biomass was significantly heavier in the high-nitrogen treatment (68·4 ± 2·6 g) than the low-nitrogen treatment (37·3 ± 2·0 g) (Fig. 2). The overall model for the univariate ancova showed that there were marginal differences in P. australis inflorescence biomass (F = 2·43(α=0·025), P < 0·03, d.f. = 6,115). The mean inflorescence biomasses for the high- and low-nitrogen treatments were 2·7 ± 0·43 g and 2·0 ± 0·33 g, respectively. The inflorescence biomasses for Lincoln (1·4 ± 0·30 g), Banner (2·9 ± 0·54 g) and Herrin (2·8 ± 0·53 g) were not significantly different. The Banner and Lincoln populations had greater biomasses in the high-nitrogen treatments than low-nitrogen treatments (Fig. 3), although the interaction between nitrogen treatment and population was non-significant (F = 2·96(α=0·025), P < 0·055, d.f. = 2,115).
Table 2. ancova table for P. australis stem biomass, with initial rhizome weight as the covariate and population and nitrogen treatment as the main effects. A Bonferroni-corrected alpha value of 0·025 was used because two mancovas were used to analyse P. australis and S. pectinata stem and flower biomasses
|Model|| 5|| 6 189||21·60||< 0·001|
|Population|| 2|| 2 086|| 6·2|| 0·003|
|Nitrogen|| 1||27 510||93·2||< 0·001|
|Population × nitrogen|| 2|| 140|| 0·5|| 0·601|
|Initial rhizome biomass|| 1|| 3 585||12·7||< 0·001|
|Error||116|| 282|| || |
Figure 3. Mean inflorescence biomass ± 1 SE for Phragmites australis (Pa) and Spartina pectinata (Sp). Capital letters represent significant differences among nitrogen treatments within species. Lower case letters represent significant differences among populations within species. Spartina pectinata means are not significantly different.
Download figure to PowerPoint
The mancova on S. pectinata stem (Fig. 2) and flower biomass (Fig. 3) showed no significant differences in the overall models for site (Wilks’ lambda, λ= 0·94, P < 0·113, d.f. = 4,228), nitrogen treatment (Wilks’ lambda, λ= 0·98, P < 0·310, d.f. = 2,114) and the interaction term (Wilks’ lambda, λ= 0·94, P < 0·166, d.f. = 4,228). Spartina pectinata stems and flowers were not heavier with respect to nitrogen treatment and population.
To determine if there were differences in P. australis and S. pectinata below-ground biomass among nitrogen treatments, two ancovas were run. P. australis root biomass and S. pectinata biomass were not correlated (Pearson correlation, r=−0·32, P < 0·08), and were therefore analysed with separate ancovas. However, because these measurements came from plants grown in the same pot, a Bonferroni correction was made and the accepted alpha level was 0·025.
The overall ancova model for P. australis below-ground biomass was marginally different (F = 2·81(α=0·025), P < 0·032, d.f. = 6,24). Nitrogen was significant but population, the interaction between population, nitrogen and the initial rhizome weight were not significant (Table 3). REGWQ revealed that below-ground biomass was significantly higher in the high-nitrogen treatment (39·0 ± 4·5 g) than in the low-nitrogen treatment (25·5 ± 4·5 g) (Fig. 4).
Table 3. ancova table for P. australis below-ground biomass, with initial rhizome weight as the covariate and population and nitrogen treatment as the main effects. A Bonferroni-corrected alpha value of 0·025 was used because two ancovas were used to analyse P. australis and S. pectinata below-ground biomasses
|Population × nitrogen|| 2||0·06||0·06||0·55|
|Initial rhizome biomass|| 1||0·059||0·59||0·44|
|Error||24||0·1|| || |
Figure 4. Mean below-ground biomass ± 1 SE for Phragmites australis (Pa) and Spartina pectinata (Sp). Capital letters represent significant differences among nitrogen treatments within species. Lower case letters represent significant differences among populations within species. Spartina pectinata means are not significantly different.
Download figure to PowerPoint
The overall ancova model for S. pectinata below-ground biomass was non-significant (F = 1·48(0·025), P < 0·49, d.f. = 6,24). Nitrogen was the closest source to significance in the ancova model (F = 3·31, P < 0·08, d.f. = 1,24). Although not significantly different, S. pectinata plants grown together with Banner Marsh P. australis had the greatest biomass (Fig. 4). Spartina pectinata plants grown under the high-nitrogen treatment had greater biomass than those grown in the low-nitrogen treatment, although not significantly higher.