|Home | About | Journals | Submit | Contact Us | Français|
Understanding the role of different components of hydrology in structuring wetland communities is not well developed. A sequence of adjacent wetlands located on a catenary sequence of soils and receiving the same sources and qualities of water is used to examine specifically the role of water-table median position and variability in affecting plant and microbial community composition and soil properties.
Two replicates of three types of wetland found adjacent to each other along a hydrological gradient in the New Jersey Pinelands (USA) were studied. Plant-community and water-table data were obtained within a 100-m2 plot in each community (pine swamp, maple swamp and Atlantic-white-cedar swamp). Monthly soil samples from each plot were analysed for soil moisture, organic matter, extractable nitrogen fractions, N mineralization rate and microbial community composition. Multivariate ordination methods were used to compare patterns among sites within and between data sets.
The maple and pine wetlands were more similar to each other in plant community composition, soil properties and microbial community composition than either was to the cedar swamps. However, maple and pine wetlands differed from each other in water-table descriptors as much as they differed from the cedar swamps. All microbial communities were dominated by Gram-positive bacteria despite hydrologic differences among the sites. Water-table variability was as important as water-table level in affecting microbial communities.
Water tables affect wetland communities through both median level and variability. Differentiation of both plant and microbial communities are not simple transforms of differences in water-table position, even when other hydrologic factors are kept constant. Rather, soil genesis, a result of both water-table position and geologic history, appears to be the main factor affecting plant and microbial community similarities.
Wetland plant communities are structured by both hydrological and soil factors, which together are considered the primary factors in wetland ecology (NRC, 1995; Mitsch and Gosselink, 2000; Cronk and Fennessy, 2001; Pezeshki, 2001; Nilsson and Svedmark, 2002). In the long term, hydrology drives the evolution of diverse adaptations to anoxia and the presence of moving or stagnant water; in the short term, hydrology acts as a filter determining which species can survive and persist under a given set of wetland conditions (Cronk and Fennessy, 2001; Pezeshki, 2001). Hydrology is similarly a primary driving factor in the development of wetland soils (Richardson and Vepraskas, 2000). It controls such basic features as the accumulation of organic matter, the creation of horizons and profile structure, and especially the chemistry of nutrients and toxic materials (Wachinger et al., 2000; Vepraskas and Faulkner, 2001; Vepraskas et al., 2004; Beumer et al., 2008). The plant communities and soil properties are linked by their dependence on the microbial transformations of the redox sequence (Vepraskas and Faulkner, 2001), which affect both the availability of nutrients and the presence of toxic substances such as Fe(II), sulfide and methane (Loeb et al., 2008). While many studies examine the association between hydrology and plant communities, few have explicitly attempted to relate hydrology and plant communities to soil microbial communities. Such a study, in which links between components of hydrology, plant community composition, soil properties and microbial communities are examined for a series of wetlands occurring along water-table gradients in the New Jersey Pinelands, has been undertaken.
The ways in which hydrology acts as a driving factor for plants, soils and microbial communities are not clear, however. Wheeler (1999), in a comprehensive review of the role of hydrology in structuring wetland plant communities, showed that many different components of hydrology, such as seasonality, duration, frequency, intensity and extremes of flooding or saturation, may affect plant communities, but that there are surprisingly few studies that clearly differentiate or specifically test the role of each of these different aspects of hydrology. Wheeler (1999) further emphasizes that the role of water-table fluctuations in affecting soil moisture and soil redox chemistry is also not well defined, nor are its effects on biotic interactions such as competition well known. Moreover, differences in hydrology based on differences in source waters (i.e. ground vs. surface vs. precipitation inputs) are confounded with differences in water chemistry, and effects of water quality differences may be confounded with water quantity effects. In order to isolate clearly particular aspects of hydrology as a primary driving factor for vegetation and for the linked properties and communities of the soil, it is necessary to examine a set of wetlands in which as many confounding environmental factors are held constant as possible, leaving only particular elements of hydrology as potentially differentiating factors. A catenary sequence of adjacent wetlands in the New Jersey Pinelands (USA) was used as such a test set. These wetlands receive primarily groundwater inputs of very similar origin and chemistry (Lord et al., 1990), and have annual hydrographs that are also very similar in seasonality, durations, and frequencies of extremes (Zampella, 1994; Zampella et al., 1992, 2001). The adjacent wetlands differ only in their locations along a shallow and short (100 m) topographic gradient, so that the main differences in hydrology are the position and variability of the water table. This enables us to isolate the role of water table dynamics as a single hydrologic factor potentially driving differences in plant communities and soil properties.
While numerous studies have shown that plants assort themselves along hydrological gradients as a function of their suites of adaptations to both anoxia and flood-generated physical disturbance (Silvertown et al., 1999; Visser et al., 2000; van Eck et al., 2004), the amount of change in the plant community relative to the amount of difference in hydrology is rarely considered (Wheeler and Shaw, 1995; Wheeler, 1999). In riverine wetlands, large differences in flood regime generate different communities (Nilsson and Svedmark, 2002; Renofalt et al., 2007), while in depressional wetlands, small differences in elevation are known to be associated with differences in plant community structure (Seabloom and van der Valk, 2003; Battaglia and Collins, 2006). However, differences in plant communities that are larger than differences in hydrology have also been documented; in these cases, other factors, such as dispersal limitation, disturbance and water chemistry interact with hydrology to determine community structure (Mulhouse and Galatowitsch, 2003; Seabloom and van der Valk, 2003; Luyssaert et al., 2007; Rheinhardt, 2007). Wheeler and Shaw (1995) suggest that, with respect to fens, which are primarily ground water driven, it is difficult to predict vegetation responses to changes in hydrology because of the lack of detailed understanding of how the components of hydrology interact with individual species and species groups.
Similarly, soil properties, including the structure and composition of soil microbial communities, are well known to vary with hydrologic patterns (Bruland and Richardson, 2004; Mitchell and Branfireun, 2005; Jordan et al., 2007). But, as with vegetation, there is considerable variability in the degree to which hydric soil properties develop under a given hydrologic regime (e.g. Karathanasis et al., 2003). Temperature regime, plant communities and prior land use can modify the development of soil properties, including microbial community structure and function, in response to a given hydrologic regime.
The relationship of soil microbial communities to plant communities has also been a topic of considerable research. Plant communities of upland and wetland habitats are linked to soil microbial communities and processes by numerous direct and indirect pathways of interaction (van den Putten, 1997; Reynolds et al., 2003; Ehrenfeld et al. 2005; Kulmatiski et al., 2008), as has been long recognized (Jenny, 1958). Plant communities affect soil properties through a range of mechanisms, including amounts and qualities of organic inputs to the soil from the plant community, the effects of plants on the physical soil environment, and the impact of nutrient availability on plant community interactions and plant population growth. Many studies, particularly in upland environments, have shown that microbial community structure often varies with plant community composition, and between individual plant species (Steenwerth et al., 2002; Kourtev et al., 2003b; Hawkes et al., 2005; Batten et al., 2006). However, other studies have documented contradictory patterns, in which plant community composition has little or no effect on microbial communities, in comparison with other factors, including soil anoxia in wetlands (Ravit et al., 2003), and soil physical properties (Grayston and Prescott, 2005; Potthoff et al., 2006). In wetlands, microbial communities vary among wetland types (e.g. ombrotrophic vs. minerotrophic peatlands; Borga et al., 1994; Jaatinen et al., 2007) and with respect to hydrology (Boon et al., 1996; Sundh et al., 1997; Bossio and Scow, 1998), but the relative importance of hydrology and plant community structure is not clear (Gutknecht et al., 2006). Moveover, as with the plant community, the role of the different components of hydrology in generating differences among communities is not well known.
In addition to the potentially complex network of relationships among hydrology, plant communities and soil properties in wetlands, there is the potential for temporal variation among these factors to confound the understanding of causative patterns. While some studies of microbial community dynamics find significant variations among seasons within a given plant community (e.g. Rogers and Tate, 2001; Habekost et al., 2008), others have failed to find systematic changes in community composition with time (Sundh et al., 1997). Seasonal changes in hydrology and temperature are likely to affect microbial communities, but these effects are occurring within the context of invariant plant communities (at least in wetlands dominated by perennial and woody plants) and static soil properties (e.g. soil organic matter, texture). Thus, the role of seasonal hydrologic changes in driving microbial community change needs to be separated from other time-variable factors, and assessed relative to time-invariant factors.
The uniformity of water source and water chemistry in a series of adjacent wetlands was used to isolate and test the role of water-table depths and fluctuations in affecting plant communities, soil physical properties and soil microbial communities, and to test the inter-relationships among these components of the wetland ecosystem. Specifically it was questioned whether the amount of difference among wetlands along the toposequence in water-table depths (medians, extremes) and variability is correlated with the amount of difference in plant and microbial communities and soil properties (response variables). This approach is based on the hypothesis that if water-table characteristics are the driving factor for the response variables, the amount of difference among the wetlands in each response variable will be highly correlated with the amount of difference among the wetlands in water-table characteristics. As part of this analysis, the relative importance of metrics of water-table position (median, extremes) or variability [coefficient of variation (CV)] in driving plant community and soil properties are investigated. It is tested further whether such relationships are constant over seasonal time. These tests thus respond to the need to isolate the role of particular aspects of hydrology in understanding the connections between hydrology and wetland ecology.
The study was carried out at two sets of sites in Brendan Byrne State Forest, NJ (74°30′20″W, 39°53′05″N) in the Pinelands, a region of flat topography, sandy soils and an extensive ground water aquifer that together create extensive wetlands (Ehrenfeld, 1986; Tedrow, 1986; Forman, 1998). The region receives approx. 1100 mm of precipitation annually, evenly distributed through the year.
Wetland communities are arrayed along a shallow topographic gradient (maximum elevation change across the wetland gradient of <2 m) and catena of soils (Fig. 1). All of the wetlands are ground water driven (Forman, 1998), reflecting shallow groundwater flow paths from adjacent upland pine forests through the extremely sandy soils of the Cohansey formation. Previous studies have demonstrated very similar ground-water chemical characteristics (low concentrations of base cations, very low concentrations of nitrate and phosphate) among the wetlands along the gradient (Lord et al., 1990). Previous studies have also shown that water-table fluctuations among the wetlands are highly correlated with each other (Zampella et al., 2001)
The uppermost wetland community is referred to as a ‘pine wetland’ (referred to below as ‘PW’); it occurs on a hydric mineral soil (Atsion series; sandy, siliceous, mesic aeric alaquods). This community is dominated by pitch pine (Pinus rigida); co-occurring species include small amounts of red maple (Acer rubrum var. tribolum), and a dense shrub understorey including highbush blueberry (Vaccinium corymbosum), huckleberry (Gaylussicia frondosa), fetterbush (Eubotrys racemosa), sweet pepperbush (Clethra alnifolia) and wintergreen (Gaultheria procumbens) (taxonomy following USDA, NRCS, 2009). Herbaceous species are infrequent, but include patches of Sphagnum spp. Pine wetlands are never flooded, but are saturated within the rooting zone throughout the winter and spring (Zampella et al., 2001). This community grades into a ‘maple swamp’ (referred to as ‘MS’ below), in which hardwoods (Acer rubrum and, to a lesser extent, black gum (Nyssa sylvatica) and sweetbay magnolia (Magnolia virginiana) are usually dominant, but some pitch pines are present together with a more open understorey of a similar suite of shrub species. Maple swamps are wetter than PW, with saturation present for a longer period of time, although they rarely experience flooding; soils are mapped in the Atsion or Berryland series. This community grades, at the lowest elevations, into cedar swamps (referred to as ‘CS’ below) in which the canopy dominant, and often sole canopy tree, is Atlantic white-cedar (Chamaecyparis thyoides). Small numbers of the same hardwood tree species may be present. The shrub layer contains many of the same species of shrub, plus several not found at the higher elevation (e.g. swamp azalea, Rhododendron viscosum), and there is normally a continuous layer of Sphagnum mosses on the ground layer. The forest floor in the cedar swamps is characterized by a pronounced hummock–hollow microtopography (Ehrenfeld, 1995). The herb community is sparse, but includes several sedges (e.g. Carex collinsii). The cedar swamps are found on organic soils (Manahawkin Muck) that may have up to 2 m of sapric peat over alluvial sands and gravels. Standing water is often present within the hollows throughout the late autumn, winter and into early summer. These wetland communities have been previously described (Ehrenfeld and Gulick, 1981; Ehrenfeld, 1986, 1995; Ehrenfeld and Schneider, 1991; Zampella et al., 1992; Laidig and Zampella, 1999).
The two sets of sites were located above and below a road crossing the McDonald's Branch, a first-order stream within the state forest. The set of sites above the road (referred to below as ‘upper’) was located in the headwaters area of the stream, and may have experienced wetter hydrology because of the road; the set of sites below the road (referred to below as ‘lower’) was located approx. 0·6 km from the upper set, and may have experienced drier hydrology because of a damming effect of the road. Other than the road, there are no other anthropogenic disturbances present.
At each location a 10 × 10 m plot was established within each community type. Plant community data (percentage cover of all species) was collected for a related study within each plot (A. Brown, Pinelands Commission, NJ, USA, pers. comm.). A manually read well (perforated PVC pipe to 2 m depth) was installed in the centre of each plot, and water tables were determined on a biweekly basis. A continuously recording well located in a nearby part of the McDonald's Branch basin (also in the headwaters area, about 0·5 km from the lower set of sites) was used to extrapolate daily water levels for each of the observational wells in the study plots (R. Nicholson, US Geological Survey, NJ, USA, pers. comm.).
Soil samples were obtained from the perimeter of each 100 m2 plot once a month during the period May 2005 to April 2006. Sampling was done on three of the plot boundaries, in order to ensure that any spatial variability in the conditions at each of the sites was fully represented in the study. Each month, the sampling location on each plot boundary was moved about 1 m, so that no soil area was sampled after disturbance from a previous month's sampling. Thus, there were three replicate samples from each site per month. In the PW and MS sites, which are located on mineral soils, the overlying organic horizon was removed prior to sampling in order to ensure that variable thicknesses of organic horizon material did not confound the results. The top 10 cm of mineral soil was obtained using a hammer corer with a plastic insert (3 cm diameter). The depth of the organic horizon was measured prior to removing it before sampling. In the two cedar swamps, any poorly decomposed material (usually Sphagnum spp.) was removed before sampling the highly decomposed muck (saprist) soil material. Bulk density of all sampled soils was determined with separate cores obtained with a bulk density corer.
On each sample date, duplicate cores were obtained in close proximity to each other. One core was placed in a cooler, and returned to the laboratory for analysis; the other core was capped at one end with a rubber stopper and at the top end with aluminium foil, and replaced in the hole for in situ incubation to determine nitrogen mineralization rates (following standard methods in Robertson et al., 1999). Aliquots of the core returned to the laboratory were analysed for pH (1 : 1 ratio of soil to deionized water after shaking suspensions for 30 min), and gravimetric soil moisture (oven-dried at 105 °C for 8 h), and organic matter by loss-on-ignition (combusted at 550 °C for 6 h). Extractable ammonium (NH4+) and extractable nitrate (as NO2− + NO3−) were extracted at a 1 : 4 ratio of soil to 2 mol KCl L−1 solution and measured after shaking suspensions for 1 h. Concentrations in extracts were analysed using a flow injection analyser (Lachat® QuickChem® FIA + 8000; Milwaukee, WI, USA). Soils from the incubated core were similarly subsampled to determine moisture and KCl-extractable NH4 + and NO3− + NO2−. Net nitrogen mineralization was calculated as the difference in both N fractions between the incubated and the initial cores. Nitrate concentrations were almost always below detection, and so are not reported below.
A separate aliquot of soil from the initial core was frozen at −20 °C and freeze-dried, and subsequently used for analysis of the microbial community using phospholipid fatty acid (PLFA) analysis (White et al., 1979; Frostegard and Baath, 1996; Bardgett and McAlister, 1999; Potthoff et al., 2006). This method is based on the differential production of PLFAs, components of the living cell membrane, by different species and groups of microbes. The total amount of PLFA provides information about the size of the living microbiota (microbial biomass); the relative abundance of specific sets of PLFAs can be used to measure the abundance of different groups of actively growing microbes (e.g. gram-negative and gram-positive bacteria, fungi, actinomycetes); and the total number of PLFAs is an indicator of the diversity and structure of the microbiota. Current practice was followed in assigning PLFAs to microbial groups, following Bossio and Scow (1998), White and Ringelberg (1998), Olsson (1999), Zelles (1999), and Waldrop and Firestone (2004). Briefly, soils samples were extracted in a chloroform : methanol (0·8 : 1·2) solution twice, the chloroform fraction was then evaporated and concentrated (close to 0·5 mL) in a vacuum evaporator, and the lipids separated on a silicic acid column using chloroform–acetone–methanol. The phospholipid fraction (methanol eluate) was then evaporated, concentrated to about 0·5 mL, saponified and methylated following a procedure given for the Sherlock Microbial Identification System (MIDI Inc., Newark, DE, USA). The fatty acid methyl esters were extracted into hexane and analysed using the Sherlock Microbial Identification System (MIDI Inc.). This consisted of a gas chromatograph system (Hewlett Packard 5890 Series II; Palo Alto, CA, USA) equipped with an HP Ultra 2 phenyl methyl silicone fused capillary column (25 m × 0·2 mm i.d., film thickness 0·33 µm), a flame ionization detector (GC-FID) and an automatic sampler (HP 7673). The temperature programme was ramped from 170 °C to 250 °C at 5 °C min−1 and hydrogen was the carrier gas. The fatty acid methyl esters in the samples were identified by the Sherlock Microbial Identification System on the basis of the retention time and quantified by peak area. The amount of each fatty acid methyl ester identified was standardized to the peak area and amount (5 µg) of the internal standard (19 : 0) and expressed as mg kg−1 soil. A total of 72 PLFAs were detected, with, on average, 36–44 compounds detected per sample.
Altogether, 216 soil samples were analysed for basic properties and for microbial community composition.
Plant community data (percentage cover) were ordinated by detrended correspondence analysis (McCune and Grace, 2002), downweighting rare species (and deleting species occurring at <5 % cover in only one plot). The first axis of the ordination had a length of 3·72 standard deviation units, supporting the use of DCA (detrended correspondence analysis) rather than an ordination method assuming linear changes in vegetation. Water-level data from the partial-series wells were summarized over the year of sampling (medians, interquartile range, maximum, minimum) to provide a descriptor for each plot. A principal components ordination was used to describe the overall relationship among the sites based on the four hydrological descriptors above. The data for each site (one or two dates per month) were also regressed against the continuously recording well, and the regressions used to provide estimates of the median water-level position per month and CV of water level, a metric of variability allowing comparison among sites, as data matrices for the temporal analyses. Soil variables (pH, gravimetric moisture, organic matter, organic horizon depth, extractable NH4, net mineralization rate) were summarized over the 12 months of sampling and ordinated by principal components analysis to provide overall descriptors of the plots; separate matrices of soil properties by month were used to examine changes over time. For these ordinations, the three replicate measurements per site were used (total of 18 observations per matrix).
In order to compare how much difference there is among sites with respect to vegetation, hydrology, soil properties and microbial communities, Euclidean distances (McCune and Grace, 2002) were calculated between all pairs of sites. Distances for vegetation, hydrology, soil properties and microbial communities were based on site scores on the first two axes of each ordination.
All ordinations were conducted in PC-ORD ver. 5·0; all other statistical analyses were conducted in SigmaPlot ver. 11·0.
In the ordination of the plant community data (Fig. 2A), the two pine wetlands and the two maple swamps were each grouped together on the first axis, which accounted for 59 % of the variance in the data, but the two cedar swamps were more distinct from each other, as shown by the Euclidean distance between points (Table 1). The two types of wetlands on mineral soils (MS and PW) were more different from the cedar swamps than from each other (Table 1).
Water levels in the two sets of sites showed clear differences in hydrology along the topographic gradient (Fig. 3). Water levels in the cedar swamps were at or above the ground surface through the winter and spring, fell below the surface but not below plant rooting depth (about 30 cm) during the summer to mid-autumn, and then rose to near or at the surface in response to autumn rain events. However, although the pattern of water level change in the two cedar swamps was similar, the two sites differed in that the upper site was flooded for much of the year, whereas the lower site was saturated but rarely flooded. Water levels in the two mineral soil sites followed very similar temporal dynamics, but were lower. Notably, there was a larger difference between the lower PW (LPW) and lower MS (LMS) plots than between the upper PW (UPW) and upper MS (UMS) sites (Table 1 and Fig. 3), and the UPW site and the LPW site were considerably more different from each other than either MS or CS was from each other (Table 1). The different patterns of relationship among the sites based on hydrology and vegetation are apparent in the PCA (principal component analysis) ordinations of hydrological properties (Fig. 2B)
Soil properties among the sites displayed similar patterns among the sites as did the vegetation. Organic horizon thickness, organic matter content of the top 10 cm of mineral soil, soil moisture, extractable NH4-N and the N mineralization rate were very similar in the two types of mineral-soil wetlands (MS and PW), and quite distinct from properties of the cedar swamp organic soils (Table 2 and Fig. 4A). All the soils were highly acidic (Table 2). Distances among replicate sites of each type (Table 1 and Fig. 4A) calculated from axis scores of the principal components ordination of the soils data show that the sites of each type were very similar to each other, notably more similar to each other than they were found to be for vegetation and hydrological properties.
Despite seasonal variations in soil moisture, extractable NH4-N, and net N mineralization rates (data not shown), these properties were surprisingly poorly related to each other (Table 3). Soil moisture varies with water tables in the mineral-soil wetlands, changing by similar amounts (30–40 %) per unit of water level change, but there is no corresponding relationship in the cedar swamps. N mineralization rates are only related to soil moisture in the LPW site, increasing slightly as soil moisture increases (Table 3). In contrast, the amount of extractable NH4-N in the soil is significantly related to soil moisture in all but the LCS site, albeit with shallow, similar slopes and much variability around the regression relationship (low r2 values).
Microbial communities in all sites were dominated by bacteria, in an approximately equal mix of Gram-positive and Gram-negative species (Fig. 5). The UCS site had significantly higher total microbial biomass than all other sites, while the LPW site had a lower microbial biomass than either the LMS or LCS sites (two-way ANOVA of mean biomass for each sample at each site, interaction F2,66 = 25·64, P < 0·001; Holm–Sidak post-hoc tests of wetland types in each site), while the LCS site had a higher total microbial biomass. The two maple swamps had the lowest ratios of bacteria : fungi (1·98 and 2·19, respectively, for the lower and upper sites), while the cedar swamps had the highest ratios (4·03 and 3·70, respectively, for the lower and upper sites; overall ANOVA F5,12 = 6·43, P = 0·004, LCS significantly different from LMS and UMS, and UCS significantly different from LMS by post-hoc tests). The two cedar swamps had higher amounts of the fatty acid 16:1ω5, an indicator of vesicular–arbuscular mycorrhizae than the other sites (one-way ANOVA F5,12 = 10·33, P < 0·001, Holm–Sidak tests showing both cedar swamps significantly higher than other sites but not different from each other). This result may reflect the dominance of the endomycorrhizal C. thyoides (Cantelmo and Ehrenfeld, 1999), in contrast to the mixture of endo- and ecto-mycorrhizal tree species in the maple swamps (maples and pines), and dominantly ectomycorhizal tree species in the pine swamps.
Two-way ANOVAs of two measures of microbial stress, the ratio of cyclopropyl PLFAs : precursor PLFAs (17:0cy + 19:0cy/16:1ω7c + 18:1ω7c), and the ratio of saturated to monounsaturated PLFAs (McKinley et al., 2005; Moore-Kucera and Dick, 2008) did not reveal any difference among the sites or wetland types in physiological status of the microbial populations. However, a principal components analysis (Fig. 4B) showed that microbial communities in the two cedar swamps are quite distinct from those of the maple and pine wetlands (post-hoc Tukey's tests showing that the UCS site is significant different from both maple and pine sites, and the LCS site is significantly different from the LMS site on axis 1 and from the UPW site on axis 2). Sulfate-reducing bacteria (identified here by the concentrations of 16:0 10Me and 17:lω7c (Boon et al., 1996; Sundh et al., 1997; Kourtev et al., 2003a) were present in small amounts (0·57–1·54 g m−3 soil), with significant higher concentrations in the UCS site than all others [Holm–Sidak post-hoc test in a one-way ANOVA of the six sites, using monthly values (n = 12) for each site, F5,66 = 18·177, P < 0·001)].
While the plant community and some soil properties (organic matter, organic horizon thickness, pH) are constant over seasons, both hydrology and other soil properties, particularly nitrogen availability and soil moisture, are variable among months. Several approaches were used to determine whether temporal variations in these properties led to different controls on microbial community composition at different times, and whether there were temporal differences in the patterns of microbial community similarity among the sites and wetland types.
Multiple response permutation procedures (McCune and Grace, 2002) were first used to test when differences among community types occurred (Table 4). Analyses were conducted using a matrix of microbial groups [Gram-positive, Gram-negative, fungi, actinomycetes, protozoa, ubiquitous PLFAs, and indices of relative structure (bacterial : fungal PLFAs, cyclic PLFAs : precursors and saturated : unsaturated PLFAs) as indicators of microbial stress (Bossio and Scow, 1998; Moore-Kucera and Dick, 2008)]. Communities in the CS wetlands were significantly different from those of the pine and maple wetlands during autumn and winter months, but not during spring and summer (except for the month of June; Table 4). Total microbial biomass and the bacteria : fungi ratio showed similar, but not identical patterns (Table 4). Mantel tests were then used to compare a matrix of total microbial biomass in the 18 samples (three replicates from each of the six sites) in each month with matrices of soil and hydrological variables. A similar matrix of microbial community composition, using the first axis scores for each sample from a PCA analysis of microbial groups in each month, was compared with the matrices of monthly environmental variables. These tests showed that the temporal variation in both the total microbial biomass and in microbial community structure varied in ways that were similar to the temporal variation in soil moisture and nitrogen dynamics, and also to the temporal variation in water-table level.
In order to explore the relative importance and relationship of soil and environmental factors in differentiating microbial communities, and to explore the similarities of communities in plots with the same vegetation type, canonical correspondence analyses were applied to matrices of microbial community composition (biomass, g PLFAs m−3) of microbial groups for each month, analysed with respect to a matrix of environmental variables (soil properties, water-table medians, and water-table CV, as a measure of hydrologic variability) assembled for each month. The loadings of the environmental factors on the first axis of the ordinations (which accounted for 50–70 % of the variance in each analysis, and which were all significant in Monte Carlo tests) were examined for each month. In addition, the axis 1 scores for the 18 samples were tested by analysis of variance first comparing the three vegetation types (pooling samples from the two locations), and also testing differences among the six samples to determine whether axis scores differed significantly between the two replicate samples of the same vegetation type. Table 5 shows that, as in previous analyses unconstrained by environmental factors, the cedar swamps had significantly different microbial community structure from the pine and maple swamps in most but not all months. Some sets of sites within each vegetation type differed from each other in some months, particularly during the autumn (September to December), when water tables were both at their lowest levels and undergoing the largest amount of change. Notably, variability in water-table level (as indexed by the CV of the daily values) was as important as median water level in structuring microbial communities (Table 5), and often formed a gradient in opposition to the water-level median gradient. There was also considerable variation over time in the importance of different soil factors in structuring the microbial communities. Soil moisture was of high importance in most months (9 months), while net N mineralization rate (1 month) and pH (3 months) were the factors least frequently correlated with axis 1. Table 5 also shows that the soil and water-table factors varied among months in whether they were acting in concert (same correlation sign), or in opposition to each other; e.g. in July, soil moisture and water-table CV formed opposing gradients, whereas in August, they formed parallel gradients.
The results demonstrate that water tables affect plant and soil microbial communities in wetlands in complex ways, such that differentiation of both types of community are not simple transforms of differences in water-table position. Thus, even among wetlands that differ solely in water level (with no differences in hydrograph seasonality, water source, or water chemistry), both plant and microbial communities are affected by factors other than water-table position. Moreover, water-table variability is an important a factor in structuring both plant and microbial communities as are measures of central tendency of water-table position. These results demonstrate that water-table position, while an important factor affecting wetland biotic communities, acts in concert with other factors to affect community composition. In other studies, the role of different water sources and water chemistries have been emphasized as factors that interact with water level in structuring wetland communities (de Mars et al., 1997; Boomer and Bedford, 2008; Beumer et al., 2008); in the present study the role of water-table levels has been isolated from these other factors, so that its role can be more clearly assessed (Wheeler, 1999). This assessment suggests that water tables alone are not, by themselves, the primary factor differentiating wetland communities.
The disparities in the similarities of water-table descriptors between the sites and of both the plant and microbial communities (Table 1) clearly demonstrates the lack of close correspondence between water-table levels and communities of both plants and microbes. Furthermore, the identity and relationship of explanatory variables for the microbial community vary over time, also supporting the conclusion that water-table levels are not the dominant influence on microbial community variation over time. In many cases, plant communities are much more similar to each other than are their hydrologies, and similarly, patterns of microbial community similarities also are often quite different from the patterns of hydrological similarity. The data suggest that the main difference differentiating both plant and microbial communities is the nature of the soil. Mineral soils support plant and microbial communities that are quite similar to each other, despite substantial differences in water-level dynamics, in contrast to the communities associated with organic soils. Thus, hydrology is interacting with other landscape factors to drive the composition of plant and microbial communities largely through its effects on the type of hydric soil which develops, rather than as a result of the regular, monotonic changes in water-table position along a topographic gradient. While the complex interrelationships of hydrology and vegetation in affecting soil pedogenesis has long been recognized (Jenny, 1958), the results show that important non-linearities can occur in these interactions along topographic gradients. Notably, water levels are only poorly correlated with the soil properties – soil moisture and nitrogen availability – which are likely to be most important in driving plant community composition. Indeed the seasonal pattern of water-table drawdown, evident in all sites, is only poorly correlated, if at all, with both soil properties and microbial communities. The variation in the water-table level within months is at least as important a factor in driving microbial community composition as is the median level, as shown by the CCA (canonical correspondence analyses).
Water tables influenced microbial communities primarily during the summer and autumn; during these time periods, fluctuations in water level can create alternating oxic and anoxic conditions (Faulkner and Patrick, 1992; Vepraskas and Faulkner, 2001). The data suggest that it is this temporal alternation which affects microbial communities, rather than the overall level of the water table. Balasooriya et al. (2008) compared microbial communities between wetter and drier wetland grassland communities, and found that there were consistent differences in composition associated with these two states, particularly among Gram-positive and Gram-negative bacteria. Other investigators have also documented higher abundances of Gram-positive bacteria in wet or anaerobic soils (Sundh et al., 1997; Bossio and Scow, 1998; Jaatinen et al., 2007). In the samples, Gram-positive bacteria were usually more abundant than Gram-negatives (ratios >1, ranging to >2·0 ), but there were no significant differences in their ratio, despite the differences in hydrology. These patterns suggest that despite the water-table drawdown in August to September, soil moisture conditions maintain an environment favourable to the more anaerobic Gram-positive bacteria.
Studies of microbial communities in other wetlands also show that hydrological differences interact with a number of other factors. For example, Jaatinen et al. (2007) found that the response of different microbial groups to changes in water-table levels differed between ombrotrophic bogs, mesotrophic peatlands and fens. Responses of microbial communities in wetlands to hydrology has been found to be contingent on nutrient availability, either added experimentally (Mentzer et al., 2006) or intrinsic to different types of wetlands (e.g. Jaatinen et al., 2007), on land-use history, on land management (Jaatinen et al., 2007), and on the types of carbon compounds available for microbial consumption (Bossio et al., 2006). These studies support the present finding that water-table dynamics alone do not drive microbial community structure.
The other major factor thought to affect microbial communities is the composition of the plant community. As with hydrology, the present results support a partial role for the plant community as a determinant of the microbiota. In most, but not all, comparisons of the Euclidean distance between plant and microbial communities (Table 1), sites that had very different plant communities also have very different microbial communities. Balasooriya et al. (2008), using a 13CO2 tracer, showed that some components of the microbial community in wetland soils rapidly assimilated carbon inputs from photosynthesizing plants, but other components, including the Gram-positive bacteria, the actinomycetes and the fungi, make little use of fresh plant assimilate. This result suggests that the responsiveness of microbial communities to plant community composition may itself vary with the composition of the microbiota. In our study sites, which are dominated by Gram-positive bacteria, the responsiveness of the microbiota to variations in plant communities thus may be less than in sites that are dominated by Gram-negative bacteria. This may explain why it was found that sites with very different plant communities had microbial communities that were not very different (e.g. UPW and LCS, or LPW and LMS; Table 1). This would also explain why the role of the plant community in structuring soil microbial communities is often unclear (Gutknecht et al., 2006; Balasooriya et al., 2008). When there are clear associations of microbial communities with different plant communities, the communities are often associated with very different environments. This has been observed, for example in comparing sedge fens vs. ombrotrophic bogs (Borga et al., 1994), permanently flooded versus ephemeral wetland sediments (Boon et al., 1996), native tropical forest versus plantation (Waldrop et al., 2000), and eroded wasteland compared with orchard soils or pine forests (Yao et al., 2000). Indeed, small changes in composition, e.g. such as a change in relative abundance of species within grasslands, may be associated with little or no difference in microbiota (Bardgett and McAlister, 1999). Our gradient of plant community composition clearly is in line with Bardgett and McAlister's results.
The most consistent contrast in plant and microbial communities among our sites paralleled the contrast in properties of the mineral and organic hydric soils. This result suggests that both groups of species may be ultimately structured by the long-term processes of geomorphic change that generate different soil types (Huggett, 1998). In the Pinelands, organic soils occupy riverine depressions created during erosional phases during the Last Glacial Maximum, and which subsequently filled in with peat as the rise in sea level during the Holocene resulted in ever-higher water-table levels (Sanford, 2000; Stanford et al., 2002). These large-scale geological processes result in adjacent histosols and mineral wetland soils across a gradient of low topographic relief, that in turn drives regular patterns of hydrological differences among the sites. Although the interplay of vegetation and hydrology has long been identified as important in soil pedogenesis (Jenny, 1958), the present results show that topography and geological history can have an overriding effect on the interactions among soil, vegetation and hydrology. The large differences in properties between the two types of soils result in larger differences in plant and microbial communities than is directly attributable just to the water-table differences.
Thus, it is concluded that these data suggest that water-table levels, as a distinct component of hydrology as a primary driving factor in wetlands, operates within a complex of other factors acting over the geologic time-scales of soil genesis, as well as shorter time-scales of human land management and short time-scales of seasonal change. These conclusions are admittedly based on only two wetlands of each type; a larger sample of wetlands within the Pinelands landscape might reveal different patterns. However, water tables and plant communities at the study sites are similar to those described in previous work, as cited above, lending support to the present conclusions.
We thank Jodi Messina for essential assistance with both field and laboratory work, and Dr Max Haggblom for assistance with the PLFA analyses. We also thank other members of the K-C project team for access to data (Dr Allison Brown, The Pinelands Commission, for access to the plant community data and the partial series ground water level data, and Mr Robert Nicholson, US Geological Survey, for access to the continuously recording well data). This work was supported by The Pinelands Commission, NJ Department of Environmental Protection.