| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Dep. of Biological Sciences, P.O. Box 369, Univ. of Notre Dame, Notre Dame, IN 46556-0369
b Natural Resources Research Inst., Univ. of Minnesota, 5013 Miller Trunk Highway, Duluth, MN 55811
c National Center for Environmental Assessment, U.S. Environmental Protection Agency, 26 W. Martin Luther King Drive, Cincinnati, OH 45268
d Dep. of Natural Resources, Fernow Hall, Cornell Univ., Ithaca, NY 14853
Corresponding author (bridgham.1{at}nd.edu)
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
32 µmol P L-1), indicating extreme soil buffering capacity of porewater P concentrations. Spatial variation in soil P pools across each wetland were predicted very well in stepwise regressions, particularly in the summer (R2 = 0.49–1.00). Variables that were important in explaining this variation included the amount of P sorbed at equilibrium, maximum P sorption capacity, percentage of P sorption sites occupied at equilibrium, organic matter content, bulk density, and oxalate-extractable Fe and Al content. Phosphorus concentrations in surface water were predicted less well by stepwise regression (R2 = 0.04–0.46), suggesting only weak-to-moderate spatial coupling between soils and surface-water P dynamics. Plant P pools were predicted poorly. Our results indicate the importance of geochemical sorption in controlling P dynamics in riverine soils. We suggest that nutrient studies in spatially diverse wetlands must be designed in a manner that adequately captures the rich spatial dynamics of the system.
Abbreviations: ANOVA, analysis of variance • EPCo, equilibrium P concentration at ambient conditions • K, linear adsorption coefficient, or buffer intensity at EPCo • PS, P sorbed (µmol P g-1 dry mass soil) • PSD, P saturation degree • PS-EPCo, P sorbed at EPCo • PSmax, maximum P sorption potential • SRP, soluble-reactive P • SRPf, final SRP concentration • SRPi, initial SRP concentration • 
SRP, SRPi - SRPf
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Phosphorus movement in sediments and soils is controlled by both geochemical and biological phenomena. Plants, microbes, and soil organic matter form the three primary biological P pools in wetlands. Organic matter has important effects on P flux through accumulation of peat (Richardson and Marshall, 1986; Reddy et al., 1998), biological mineralization–immobilization dynamics (Diaz et al., 1993; Bridgham et al., 1998), and interactions with mineral compounds (Sah et al., 1989; Lockaby and Walbridge, 1998; Axt and Walbridge, 1999). Microbial uptake dominates short-term P dynamics in many wetlands, but microbial biomass forms a transient pool of limited potential size that can be quickly overwhelmed at high P inputs (Richardson and Marshall, 1986; Walbridge, 1991). Similarly, while plants can form a substantial P pool, this pool is often seasonally released due to the die-back of vegetation and subsequent mineralization of the litter (Richardson and Marshall, 1986; Johnston, 1991; Reddy et al., 1995). However, P storage in young, aggrading forested wetlands may be substantial (Lockaby and Walbridge, 1998). Additionally, algal mats affect the diel dynamics of P flux from sediments due to oxygenation of surficial sediments during the day (Carlton and Wetzel, 1988).
The most important geochemical P removal processes in wetlands are adsorption and sedimentation (Lockaby and Walbridge, 1998). Studies from 17 wetlands showed that storage of P from sedimentation is an order of magnitude greater than that from soil organic matter accumulation, with highest sedimentation rates in riverine floodplains (Johnston, 1991). Most P retention in several constructed riverine wetlands was also by sedimentation (Mitsch et al., 1995). In most of these studies, P removal via direct physical sedimentation and adsorption of dissolved phosphate to the deposited minerals were not distinguished. However, several studies suggest that adsorption to soils may be the dominant P removal pathway in riverine wetlands (reviewed in Lockaby and Walbridge, 1998).
Sorption reactions are driven by the formation of Fe and Al phosphates and oxyhydroxides at low pH and Ca phosphates at alkaline pHs (Stevenson, 1986). Redox potential has a dramatic effect on P dynamics through its interactions with Fe, with Fe+3 forming insoluble complexes that are released upon reduction to Fe+2 (Patrick et al., 1973). However, anaerobic conditions can enhance P sorption because of the formation of amorphous ferrous hydroxides (Holford and Patrick, 1979; Sah and Mikkelsen, 1989; Sah et al., 1989). High S levels may enhance P flux from soils, due to the binding of Fe by sulfides (Caraco et al., 1989).
Most studies of P dynamics have either fractionated various soil pools (Cross and Schlesinger, 1995; Qualls and Richardson, 1995; Reddy et al., 1998) or studied P sorption behavior. Sorption studies have tended to focus on either the maximum P sorption potential of soils (Richardson, 1985; Sah and Mikkelsen, 1989; Walbridge and Struthers, 1993) or dynamics at low phosphate concentrations typical of most natural waters (Mayer and Gloss, 1980; Klotz, 1985; Froelich, 1988; Lyons et al., 1998). However, fewer studies have combined both approaches to examine P sorption behavior across a broad range of P concentrations in a diverse group of wetland soils (Reddy et al., 1995; Sallade and Sims, 1997).
Soluble-reactive P (SRP) concentrations in many rivers, streams, and estuaries appear to be maintained at an equilibrium by the adsorption–desorption behavior of suspended sediments (Froelich, 1988; Fox, 1993). Defining these equilibrium dynamics has been the objective of many sorption studies done at low P concentrations. In contrast, sorption experiments done at high P concentrations are typically designed to determine the maximum P removal capacity of the soils. Since long-term P removal is predominantly a geochemical phenomenon (Richardson and Marshall, 1986; Reddy et al., 1995), ecosystems will have a finite number of sorption sites that can be exhausted with continued P loading. Defining this sorption capacity has important implications for the ability of different wetlands to ameliorate water quality (Richardson, 1985; Walbridge and Struthers, 1993).
The objectives of this study were threefold:
1. We quantified P sorption behavior at ambient concentrations and determined P sorption maxima in two diverse, 2-ha riverine wetlands in northern Minnesota and Wisconsin.
2. We correlated P sorption parameters with the physical and chemical characteristics of the soils and then used these correlations to extrapolate P sorption across each wetland. We had previously established an extensive spatial database from each wetland of physical, chemical, and biological variables in both soils and overlying water (Johnston et al., 1997, 2001).
3. We then used this spatial database and the derived P sorption variables to relate soil physical and chemical characteristics with P concentrations in soil, water, and plants in the wetlands.
We hypothesized that (i) geochemical sorption maintains porewater P concentrations at low equilibrium concentrations under low-to-moderate P loading, (ii) spatial variation in P sorption maxima primarily reflect differences in amorphous Fe and Al pool sizes, and (iii) P sorption is a dominant control over porewater P and plant P concentrations but is weakly related to P concentrations in overlying water.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
13 km downstream in Superior, WI. The St. Louis River has an average annual discharge of 56 m3 s-1, with extremes of 902 and 2 m3 s-1 during the 20th century (Mitton et al., 1994). Both sites are within the freshwater estuarine reach of Lake Superior with regular seiche-driven tides of 3 to 4 cm (C. Johnston, 1992–1993, unpublished data). Each 2.5-ha site was chosen to have a similar geomorphic position within the river, with a low levee forming a backwater area with an open mouth allowing free water exchange with the river. However, the Pokegama site is bounded by upland, whereas the Fond du Lac site adjoins a shallow bay. Soils at both sites are quite diverse relating to their geomorphic position within the wetland, but clay content at Pokegama generally exceeds 45%, reflecting the glaciolacustrine clayey soils of its watershed, whereas clay content is <30% at Fond du Lac, reflecting the silty alluvium soils of the St. Louis River watershed (Johnston et al., 2001). Additionally, Fond du Lac soils have lower pH, greater organic matter, silt, and oxalate-extractable Fe content, and lower total N concentrations than Pokegama soils (Johnston et al., 2001). Water depth varied in a similar manner at both sites, with an average depth of 30 to 36 cm in fringing wetlands between the river and the levee (riverbed) and 36 to 46 cm in backwater areas. Backwater areas also included sites that were saturated, but without standing water. The water table was below the surface in most of the levee area except during major flood events. Emergent macrophytes dominated the riverbed and most backwater areas, with floating and submerged aquatics in deeper parts of the backwater areas, and tall shrubs on the levees (Johnston et al., 1996).
Phosphorus Sorption-Theoretical Considerations
In P sorption experiments, one adds varying amounts of phosphate to a sterilized soil–water slurry and then calculates the difference (SRP, in µmol P L-1) between the initial (SRPi) and final (SRPf) soluble-reactive P concentrations. This difference represents the amount of P adsorbed or desorbed by the solid phase and is usually expressed per dry-mass equivalent of soil (PS, in µmol P g-1 dry-mass soil). The wide range of initial P concentrations used in our experiment allowed us to estimate adsorption–desorption reactions at approximately ambient concentration, as well as to calculate maximum adsorption potentials. However, chloroform (used in this study), as well as other sterilants, lyze P from microbial cells, so that geochemical sorption is probably underestimated, particularly at low initial P concentrations (SRPi).
Buffer or crossover plots (Fig. 1) provide a useful graphical approach for examining P sorption behavior at near-ambient concentrations (Froelich, 1988). When PS is graphed on the y-axis against soluble-reactive P concentrations (SRPi or SRPf) on the x-axis, the x-intercept is termed the equilibrium P concentration (EPCo). There is no net adsorption or desorption from the solid phase at this solution concentration, and thus it represents the equilibrium P concentration of the aqueous phase. The EPCo is also the solution concentration at which soils display their maximum buffering capacity for solution P. Small changes in solution P will be compensated for by adsorption or desorption of P from the solid phase to return the system to its EPCo.
|
SRP is graphed on the y-axis vs. SRPi on the x-axis, a slope of one indicates a perfect buffering capacity; that is, any perturbation from EPCo will be completely compensated for by adsorption–desorption reactions from the solid phase (Mayer and Gloss, 1980). The Langmuir model is often used to estimate the maximum P sorption potential of soils (PSmax), which is calculated as the inverse of the slope of SRPf/PS vs. SRPf (Harter and Smith, 1981). The Bache–Williams index is another frequently used comparative measure of P sorption potential of soils at a single SRPi concentration (Bache and Williams, 1971). The index is calculated as PS/log SRPf at a prescribed SRPi concentration; we used a SRPi concentration of 4101 µmol P L-1 to compute this index to allow comparison to previous studies done in wetlands (Walbridge and Struthers, 1993, and references therein).
Phosphorus Sorption Methodology
We chose six sampling points at each 2.5-ha site, based upon maximum diversity of physical and chemical soil characteristics, to do intensive P sorption experiments. These sampling points were a subset of a larger sampling grid established to characterize spatial variation in the biogeochemistry of both wetlands (Johnston et al., 2001). We took a 15-cm-deep core from each sampling point in August 1994 using 5-cm-i.d. plexiglass tubing with a sampling device similar to that described by Cooper et al. (1991).
To determine P sorption potentials, 4-g wet mass of soil from each core were added to 60-mL centrifuge tubes. To each tube, we added 25 mL of one of the following initial P concentrations (SRPi) as KH2PO4 in 0.02 M KCl: 0.5, 0.7, 1.0, 1.3, 1.9, 3.2, 32, 1033, 4101, 6681, and 8395 µmol P L-1. Additionally, two drops of chloroform were added to each tube to sterilize the samples. The initial P concentrations were divided into a low range (0.5–32 µmol P L-1) and a high range (32–8395 µmol P L-1). The low-range concentrations reflected ambient porewater SRP concentrations at the sites, which averaged between 0.4 and 2.3 µmol P L-1, depending on season and site, with a range of 0.1 to 16.0 µmol P L-1 (Johnston et al., 2001). There were three replicates of all treatments.
The tubes were continuously mixed on a shaker table for 24 h in the dark, after which the samples were centrifuged and the supernatant filtered with acid-washed Fisherbrand Q2 filter paper (1-µm pore size; Fisher Scientific, Pittsburgh, PA). An aliquot of the filtrate was frozen for later analysis of the final SRP concentration (SRPf). For the low-range samples, SRPf concentrations were measured with the ascorbic-acid method (Murphy and Riley, 1962) and a spectrophotometer (Spectronic 501, Spectronic Instruments, Inc., Rochester, NY) with a 5-cm path length to increase sensitivity, whereas a Lachat QuikChem 8000 (Lachat Instruments, Milwaukee, WI) autoanalyzer was used to measure SRPf concentrations on the high-range samples.
Environmental Characterization
We also examined correlations of our P sorption variables with a large number of biogeochemical variables in soils, surface water, and plant tissue in a sampling grid that encompassed each 2.5-ha wetland. Our 12 sampling points for the P sorption experiment were a subset of this grid. The sampling points in the grid were 20 m apart in the backwater areas, but were only 4 to 10 m apart in the riverbed and levee areas to capture large environmental changes over short distances. There was a total of 73 and 84 sampling points in the Fond du Lac and Pokegama sites, respectively. The data for the entire grid are described in Johnston et al. (2001), but we only summarize the methods here.
Soil cores and surface-water (when present) samples were taken at all points in both wetlands in May and August 1993. Cores were taken as described above for the P sorption experiment. Cores were immediately put on ice in the field and stored at 4°C in the laboratory. We homogenized each core by hand, removing roots and stones. Bulk density was determined by measuring the wet weight of each core and converting to dry-mass equivalent per unit volume of soil with percentage moisture data (see below) and the known volume of the cores. A subsample was centrifuged to remove porewater, from which an aliquot was frozen for later analysis. Samples were simultaneously taken of surface water, put on ice until return to the laboratory, and suctioned through an acid-washed 0.45-µm filter. Soluble-reactive P concentrations were determined in both porewater and surface water by the ascorbic acid method (Murphy and Riley, 1962).
Two subsamples were taken from each core to determine percentage moisture gravimetrically by drying at 105°C. Organic matter content was estimated by loss-on-ignition at 550°C for 3 h. Soil pH and sulfide concentrations were determined in a 1:1 soil/deionized water slurry using a pH/ion-specific meter (Orion model 290A, Orion Research Inc., Beverly, MA). Particle-size distribution (percentage sand, silt, and clay) was determined with the pipette method (Gee and Bauder, 1986). Total soil P was determined by digestion of dried soil with 50% H2SO4 at 300°C for 1 h and analysis by inductively coupled plasma optical emission spectrometry (Perkin Elmer Optima 3000, Perkin-Elmer, Norwalk, CT) by Dr. Glen Guntenspergen (U.S. Geologic Survey, Laurel, MD). Extractable soil P was estimated with dilute acid fluoride (Olsen and Sommers, 1982), and P concentrations were quantified on a Lachat QuikChem 4 autoanalyzer (Lachat Instruments, Milwaukee, WI). Amorphous Fe and Al were determined by acid-ammonium-oxalate extraction (Jackson et al., 1986) and analyzed by atomic adsorption spectrophotometry (Varian Spectra-30A, Varian Associates Inc., Walnut Creek, CA).
Vegetation was clipped from 0.25-m2 plots at each sample point in August 1993, dried at 105°C, and ground with a Udy grinder (Udy Corp., Fort Collins, CO). Total P concentrations in vegetation were determined as described above for total soil P.
Statistical Analyses
We performed two sets of statistical analyses. First, we used analyses of variance (ANOVA) and regressions to examine controls over P sorption dynamics at the 12 sample points where P sorption was directly quantified. The effects of site (Pokegama and Fond du Lac) and individual sample points were tested on the P sorption variables (SRPf, PS, EPCo, PS-EPCo, PSmax, Bache–Williams index) using ANOVAs. Specifics of the ANOVA design are described for each analysis in the Results and Discussion section below. Additionally, we used Pearson correlations to examine the individual relationship between each P sorption variable and each biogeochemical variable measured at the respective sampling points. We then used forward stepwise regressions, with a P value to enter and remove of 0.15, to develop the best set of predictive multiple regression equations to describe P sorption at the 12 sampling points.
Secondly, these equations allowed us to extrapolate the P sorption parameters throughout both 2.5-ha wetlands and, thus, to examine spatial controls over P dynamics. Each multiple regression equation derived from the 12 intensively studied sample points was used to extrapolate P sorption parameters for all 157 points in both wetlands using the biogeochemical data in Johnston et al. (2001). Subsequently, P concentrations in surface water, soils, and plant biomass were compared with calculated P sorption variables and measured biogeochemical variables for all sampling points with Pearson correlations to examine potential mechanisms controlling P dynamics across the two wetlands. Similarly, P concentrations in these pools were evaluated with a forward stepwise regression. When samples were measured in both May and August, only the corresponding seasonal data were used in the regression. The number of sample points varied in the regressions somewhat because of missing data at some points. Systat was used for statistical analyses (Wilkinson et al., 1992).
|
RESULTS AND DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
|
|
When SRPi and P sorbed are expressed in the same units (µmol P L-1), a slope of one indicates a "perfect buffering capacity" (Mayer and Gloss, 1980); that is, any change in P concentration is perfectly compensated for by soil adsorption–desorption, maintaining a constant SRP concentration. Within the entire low range SRPi of 0.5 to 32 µmol P L-1, this slope was not significantly different from one for any sample point (slope = 0.996–1.004, r2 
0.997). Because there was a 10-fold increase between the 32 µmol P L-1 treatment and the next lowest treatment, we also did this analysis without the 32 µmol P L-1 treatment. Similar results were obtained as before, except the slope of the outlier Fond du Lac sample point (C9), a levee point, was 1.53. The graph for C9 indicated that it was beginning to reach saturation at a SRPi of 3.2 µmol P L-1 (Fig. 3). The increase in linearity at 32 µmol P L-1 suggests a new set of binding sites with higher activation energy were being accessed, a commonly observed phenomenon (Froelich, 1988).
Our results indicate an extreme buffering capacity of the soils for porewater P concentrations. The fact that addition of chloroform as a sterilant probably lysed microbial cells, causing a release of P to solution and consequently underestimating SRPi and P sorption, suggests an even greater buffer capacity of the soils. Consequently, moderate perturbations of the P cycle (e.g., increased P inputs) will have little effect on P concentrations in porewater. Suspended solids in surface waters may have a similar buffering effect in these sites, although we did not measure it.
Graphically derived equilibrium porewater P concentrations (EPCo) were remarkably similar to empirically determined SRPf values (Fig. 4
, r2 = 0.99, P < 0.001), adding confidence to the graphical interpretation of P sorption dynamics (Fig. 1). Equilibrium porewater P concentrations were <1 µmol L-1, except for the levee point C9 with its EPCo of 1.74 µmol L-1 (Table 2). Ambient porewater SRP was maintained at similar low concentration throughout both wetlands (
= 0.4–2.3 µmol L-1), depending on season and site (Johnston et al., 2001). Our EPCo values are much lower than those reported in a Rhode Island riparian forest ( = 17–126 µmol L-1; Lyons et al., 1998), but a North Carolina riparian forest had EPCo values (0.3–2.0 µmol L-1; Cooper and Gilliam, 1987) that were similar to ours. Stream sediments and wetland soils in the Lake Okeechobee Basin, Florida, had EPCo values ranging from 1 to 292 µmol L-1 (Reddy et al., 1995).
|
|
10–132 µmol P g-1, Walbridge and Struthers, 1993). A number of Florida stream sediments and wetland soils had lower values of PSmax (1–19 µmol P g-1; Reddy et al., 1995). Thus, variability in P sorption within a single riverine wetland complex can rival that found in wetlands throughout the USA.
|
0–51 µmol P g-1; Walbridge and Struthers, 1993) with experiments performed at a similar SRPi concentration. The P saturation degree (PSD) can be defined as the percentage of potential soil P sorption sites that are occupied under ambient conditions (Brookes et al., 1997), which we calculated as (PS - EPCo/PSmax)100. The PSD was <0.04% in all the soils tested (Table 2), indicating that these wetlands have an essentially untapped capacity to remove P from the water column. The PSD of the C9 levee point, 0.033%, was 3.7 times higher than that of any other point. Thus, it is suggestive that the higher PSD at C9, although still very low, explains the large observed desorption of P in the low-range sorption experiment (Fig. 3). The same calculation was done for Florida stream sediments and wetland soils enriched with P from agricultural runoff (Reddy et al., 1995). Phosphorus-saturation degree values of 0.2 to 5% occurred in most sites, but one stream site had a value of 20%, and a drained marsh had a value of 204%, emphasizing the very low degree of P saturation for our two wetlands.
Correlations among Phosphorus Sorption and Site Variables
Low-Range Phosphorus Samples
Pearson correlations and forward stepwise regressions were used to determine predictive relationships between variables describing P sorption dynamics and relevant physical and chemical variables for each of the 12 sample points (Tables 3 and 4). Equilibrium porewater P concentrations (EPCo) and P sorbed at equilibrium (PS-EPCo) were predicted extremely well in the stepwise regressions (R2 = 0.99). Extractable P, porewater P, sulfide, Fe, Al, and water depth entered into the predictive equation for EPCo, whereas porewater P, pH, sulfide, bulk density, and percentage sand entered into the equation for PS-EPCo (Table 4). However, the relationship was heavily influenced by the sample point C9 in Fond du Lac and its very low sulfide concentration (Table 1). Sulfide remained important though when C9 was excluded from the analysis, and the R2 remained high (0.92, data not shown).
|
|
The strong effect of sulfide concentration on P dynamics is more difficult to explain. There can be numerous linkages between the S and P cycles (Caraco et al., 1989, 1993; Curtis, 1989), but most results suggest that higher S levels lead to higher P concentrations in solution. In particular, sulfide can bind Fe in soils, resulting in the release of P, and evidence in lakes suggests that this can be a dominant control over SRP concentrations in the water column (Caraco et al., 1989, 1993). In contrast, in our study sulfide concentrations were negatively correlated with SRPf, EPCo, and PS-EPCo (Table 3), with the levee point C9 in Fond du Lac having very high values for SRPf, EPCo, and PS-EPCo, a water-table depth that varied between 0 and less than -32 cm, and very low sulfide concentrations in both spring and summer (Table 1, 2). The low sulfide concentrations result from aerobic surface soil conditions during much of the growing season. Without the C9 sample point (data not shown), the relationship remained moderately strong between summer sulfide concentrations and SRPf (r = -0.54, P = 0.088) or EPCo (r = -0.51, P = 0.106).
Johnston et al. (1995) found that porewater sulfate concentrations decreased, whereas Fe+2 and SRP concentrations increased with increasing soil wetness in Minnesota beaver (Castor canadensis) meadows, which they interpreted as release of ferric phosphates to solution upon the reduction of Fe. In contrast, in our study SRPf, EPCo, and PS-EPCo were negatively correlated with water-table depth (Table 3), but this effect was solely due to the inclusion of the C9 levee point at Fond du Lac. Such results suggest the interactions between the S, P, and Fe cycles in wetland soils need to be more closely examined with mechanistic studies.
High-Range Phosphorus Samples
Stepwise regression analysis yielded excellent predictive correlations between PSmax, the Bache–Williams index, and PSD and the physical and chemical characteristics at each sample point (R2 = 0.94–1.00, Table 4). Extractable P, porewater SRP, sulfide, and bulk density entered into the equation for PSmax, whereas porewater SRP, surface-water SRP, pH, sulfide, bulk density, and water depth entered into the equation for the Bache–Williams index. Porewater SRP, sulfide, Fe, and water depth entered in the equation for PSD.
Bulk density alone was a good predictor of PSmax and the Bache–Williams index (r2 0.69; Table 3), with higher bulk density leading to lower maximum P sorption potential. The effect of bulk density is perplexing, as P sorption maxima were expressed on a mass basis. This effect may be due to differences in particle density (mass of a unit volume of soil solids) between sampling points. However, bulk density is determined both by particle density and pore volume of the soil. Particle density tends to be rather uniform among mineral soils (2.60–2.75 Mg m-3), but is much less in organic matter (1.1–1.4 Mg m-3; Brady, 1984). The correlation of PSmax with percentage organic matter content (r2 = 0.66) is almost as good as with bulk density (r2 0.73; Table 3). However, bulk density and percentage organic matter were only moderately correlated (r2 = 0.50–0.74), suggesting that they only partially act as surrogates for each other in explaining P sorption maxima. Bulk density may act as an integrating factor for other important variables, as it was also negatively correlated with Fe (r2 = 0.61) and Al (r2 = 0.56) and positively correlated with percentage sand (r2 0.46), all of which could have significant effects on P sorption capacity.
Predictive Regressions for Porewater and Surface-Water Phosphorus
We examined the variables that best predicted surface-water SRP and total P, porewater SRP, acid-fluoride extractable soil P, total soil P, and plant P in the 157 sample points in the two wetlands (Tables 5 and 6). We calculated P sorption parameters as potential explanatory variables for these P pools at all 157 sample points by using the predictive equations determined in Table 4 and the biogeochemical data in Johnston et al. (2001).
|
|
Surface-water total P in the spring was reasonably well explained by the stepwise regression (R2 = 0.46), and there was a particularly strong positive effect of total suspended solids in the individual correlations (r = 0.63). Surface-water total P was negatively affected by organic matter content of the soils, sulfide concentrations, and total soil N. It is reasonable to surmise that more suspended solids in the water column would lead to greater total P, apart from any other effects on P concentrations in the solid or aqueous phases, and this appears to have been the dominant control over total P in the surface water in spring. However, significant correlations of many soil variables with surface water total P (Tables 5 and 6) suggest a secondary effect of soil water exchange.
Surface-water total P in the summer was moderately well predicted by the stepwise regression (R2 = 0.45), but the effects were distributed over numerous variables that entered the equation. Total suspended solids still entered into the equation (Pearson r = 0.23), but their effect was much less dominant than in the spring. Total suspended solids varied in complex ways among seasons, sites, and geomorphic units within sites (i.e., levee, river, mineral and organic soil backwater areas), but they were several times higher in Pokegama in spring than at Fond du Lac in either season or at Pokegama in the summer (Johnston et al., 2001). The other variables that entered into the equation (PS-EPCo, PSD, plant N and P, soil N and P, soil pH, percentage organic matter, and percentage silt) suggest at least moderate soil and plant control over total P concentrations in surface water during the summer.
Extractable soil P during the spring was reasonably well predicted by the stepwise regression (R2 = 0.50). Numerous soil variables had a moderately high Pearson r, but only water depth, bulk density, organic matter content, and percentage clay entered into the stepwise equation. The stepwise regression explained essentially all of the variation in extractable P in the summer (R2 = 1.00), with points with low bulk density and high Al and Fe content, EPCo, PSD, and PSmax having higher values. In particular, PSmax was strongly positively correlated with extractable P in the summer, and factors that were related to P sorption capacity, such as bulk density, Fe, Al, and organic matter content, and PSD, entered into either the stepwise equation or had a high Pearson r.
Porewater SRP was moderately well predicted by the stepwise regression in the spring (R2 = 0.49), with PSD and EPCo having the strongest effects (Pearson r = 0.35–0.48). Again, essentially all of the variation in porewater SRP in the summer could be explained by the stepwise regression (R2 = 1.00). Numerous variables entered into the stepwise regression and were individually significant, but PS-EPCo accounted alone for 94% of the variation in summertime porewater SRP. In contrast, PS-EPCo was not significantly correlated with porewater SRP in the spring (Pearson r = 0.08). This discrepancy between seasons may be partially explained by the much lower porewater SRP concentrations in spring (<2.3 µmol P L-1) than in the summer (16.0 µmol P L-1), suggesting a strong seasonal effect on porewater SRP concentrations that cannot be explained by P sorption in soils. However, PS-EPCo is clearly a strong predictor of porewater SRP concentrations in the summer.
Total soil P was reasonably predicted by the stepwise regression using summer variables (R2 = 0.59). Numerous variables were individually highly correlated with total soil P, but PSmax, organic matter content, and percentage silt were significant in both the Pearson and stepwise regressions. All three variables were positively correlated with total soil P. When stepwise regression was performed with the spring data, similar results were obtained, with an R2 of 0.52 and only organic matter content and percentage silt entering in the equation (data not shown).
Plant P content was poorly predicted in the stepwise regression (R2 = 0.26). The only variables to enter into both the Pearson and stepwise regressions were plant N content, surface-water total P concentration, and water depth. It is perhaps not surprising that plants that maintain a higher P content also have a high N content, indicating overall higher nutrient assimilation. A similarly low stepwise correlation was found between plant P content and spring data (R2 = 0.28). The low predictive power between plant P content and the large number of measured and estimated variables in this study suggests that plant availability in riverine wetlands is controlled by a complex suite of factors, which are only partly soil dependent.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
Porewater P concentrations were effectively buffered by geochemical sorption reactions, such that final P concentrations in solution were held constant and were unrelated to the amount of added P when 32 µmol P L-1. Although this study was conducted in soils, we suggest that similar dynamics may occur in turbid surface waters typical of many rivers.
The levee point in Fond du Lac (C9) had distinctive P dynamics. The majority of the levees in the two wetlands had a water-table level well below the surface during most of the growing season, which was the likely cause of the low sulfide concentrations. The large net P desorption in the low-range experiment (Fig. 3) and the much higher P saturation degree (PSD, Table 2) at C9 suggest fundamentally different P dynamics along the levees.
The variability in the maximum P sorption potentials within these two sites was equal to that measured in the literature in a wide variety of wetlands, demonstrating the importance of intrasite variation. Few studies explicitly consider such variation; most studies do everything in their power to reduce intrasite variation. Our results suggest that intrasite variation should be included in the design of any study of patchily distributed wetland soils.
A strong seasonal component was evident in that surface-water SRP, extractable soil P, and porewater SRP were predicted much better in the summer than in the spring by the large suite of environmental variables used in this study (Table 6). In particular, spatial variation in extractable soil P and porewater SRP was almost totally explained by the stepwise regressions in the summer, with R2 of 1.00, suggesting tight geochemical controls over soil P cycling and availability during this season. The most dramatic single effect was the excellent correlation between PS-EPCo and summer porewater SRP concentrations (Pearson r = 0.97). Other particularly strong soil effects were found between spring porewater SRP concentrations and PSD; extractable soil P concentrations and percentage organic matter, bulk density, amorphous Fe and Al concentrations, and PSmax; and total soil P concentrations and percentage organic matter and percentage silt (Tables 5 and 6).
In contrast, surface-water SRP and surface-water total P concentrations had poor-to-moderate stepwise correlations (R2 = 0.04–0.46), and few variables individually explained large amounts of the observed spatial variation (Tables 5 and 6). The exception was the effect of high spring suspended solid loads at Pokegama on surface-water total P concentrations (Pearson r = 0.63). Overall, our results suggest only moderate spatial coupling between surface-water and soil P cycling. Frequent hydrologic flushing may preclude establishment of equilibrium conditions between soils and the surface water, such that strong spatial correlations in P chemistry do not develop.
Few studies have performed sorption experiments at such a wide range of initial P concentrations, so that both control over P dynamics at ambient concentrations and maximum P sorption potentials could be compared in the same sites. We have shown that both types of sorption experiments are useful in explaining P dynamics in natural aquatic ecosystems. Of particular importance was the dominating effect of PS-EPCo in predicting summer porewater SRP concentrations. Additionally, these experimentally derived variables were extremely well correlated with physical and chemical variables measured at the two sites.
Overall, our results show the dominance of geochemical sorption reactions on P dynamics at even moderate P loading rates. There appears to be only weak-to-moderate coupling between soil and water P cycling, and plant P concentrations appear to be controlled by a complicated set of environmental conditions that were not captured in this study. Our results indicate that nutrient studies in spatially diverse wetlands must design their experiment in a manner that adequately captures the rich spatial dynamics of the system.
|
ACKNOWLEDGMENTS |
|---|
Received for publication August 13, 1999.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONCONCLUSIONSREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. Grunwald and K. R. Reddy Spatial Behavior of Phosphorus and Nitrogen in a Subtropical Wetland Soil Sci. Soc. Am. J., June 18, 2008; 72(4): 1174 - 1183. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. L. Bruland and C. J. Richardson Spatial Variability of Soil Properties in Created, Restored, and Paired Natural Wetlands Soil Sci. Soc. Am. J., January 1, 2005; 69(1): 273 - 284. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. L. Bruland and C. J. Richardson Hydrologic Gradients and Topsoil Additions Affect Soil Properties of Virginia Created Wetlands Soil Sci. Soc. Am. J., November 1, 2004; 68(6): 2069 - 2077. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. L. Bruland and C. J. Richardson A Spatially Explicit Investigation of Phosphorus Sorption and Related Soil Properties in Two Riparian Wetlands J. Environ. Qual., March 1, 2004; 33(2): 785 - 794. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. A. Johnston, S. D. Bridgham, and J. P. Schubauer-Berigan Nutrient Dynamics in Relation to Geomorphology of Riverine Wetlands Soil Sci. Soc. Am. J., March 1, 2001; 65(2): 557 - 577. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||
