| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Duke University Wetland Center, Nicholas School of the Environment and Earth Sciences, Box 90333, Durham, NC 27708-0333
* Corresponding author (glb5{at}duke.edu).
Received for publication May 20, 2003.
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Abbreviations: Alox, oxalate-extractable aluminum • Feox, oxalate-extractable iron • GrCr, Grindle Creek • PSI, phosphorus sorption index • RoBr, Rowel Branch • SOM, soil organic matter
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
In forested riparian wetlands in the southeastern United States, long-term P retention is generally believed to be controlled by two main processes: (i) deposition of sediment and particulate organic P (a physical process) and (ii) sorption of dissolved phosphate (a geochemical process) (Axt and Walbridge, 1999). While significant amounts of P can be stored by sedimentation (Johnston, 1991), these sediments may be resuspended in future hydrologic events. Consequently, sorption is believed to represent the most important long-term P retention pathway in wetland soils (Richardson, 1985; Richardson et al., 1988; Walbridge and Struthers, 1993; Lockaby and Walbridge, 1998; Axt and Walbridge, 1999).
In the acid wetland soils of the southeastern Coastal Plain, P sorption has been shown to be significantly correlated with amorphous Al and Fe content (Richardson, 1985; Lockaby and Walbridge, 1998), and to a lesser degree soil organic matter (SOM) (Axt and Walbridge, 1999), and particle size (Axt and Walbridge, 1999). While iron phosphates are solubilized when Fe(III) is reduced to Fe(II) under anaerobic conditions, aluminum phosphates are unaffected by changes in redox potential (Ponnamperuma, 1972). Thus, soluble iron and phosphate may be lost from wetland soils in reducing conditions while aluminum phosphates appear to persist in acid wetland soils of the southeast (Darke and Walbridge, 2000).
Another layer of complexity is added when we consider that amorphous Al and Fe, SOM, and texture exhibit significant spatial and temporal variability in riparian wetlands (Lyons et al., 1998; Darke and Walbridge, 2000; Johnston et al., 2001). Beyond a locally random aspect, this spatial variability may be related to the combined action of physical, chemical, or biological processes that operate at different spatial scales (Goovaerts, 1998). In natural riparian forested wetlands, these processes might include overbank flooding, sediment deposition, surface runoff, erosion, ground water inputs, fire, tree-throw, root activity, litter production, and activity of macro and micro soil fauna. Each of these processes may influence particular locations of the riparian zone with varying degrees of intensity. Thus it has been recommended that any study of patchily distributed wetland soils should attempt to quantify not only the soil properties or processes under investigation but also their spatial variability (Johnston et al., 2001; Stolt et al., 2000). Improving our understanding of the spatial variability of soil properties in wetlands is important because such variability influences both the structure and function of soil ecosystems (May, 1974; Legendre and Fortin, 1989; Reynolds and Houle, 2002). Unfortunately, due to the combined difficulties of establishing spatially explicit sampling designs at remote, wet, and densely vegetated sites, and processing the large number of samples needed for adequate spatial coverage, sampling designs that quantify spatial variability are seldom used in studies of wetland soils (Bridgham et al., 2001).
Thus, the complex relationships among soil properties and their associated spatial variabilities render attempts to establish causality between individual soil properties and P sorption highly difficult. These intertwined factors have confounded traditional parametric analyses of the relationships between soil properties and P sorption, and, in some cases, they may have even lead to identification of spurious correlations between dependent and independent variables. In response to these complexities, we established a spatially explicit sampling design that enabled us to use Mantel tests (Mantel, 1967) to ascertain the importance of edaphic and spatial variables in controlling the P sorption capacity of soil cores taken from two natural forested riparian wetlands in the North Carolina Coastal Plain. The objectives of this study were to (i) quantify patterns of spatial variability of P sorption and related soil properties in the wetlands and (ii) determine which soil properties best explained the variability in P sorption after accounting for the effects of spatial autocorrelation.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
Sampling Design
At each site, we identified a relatively flat section of the floodplain with no major elevation gradients in directions perpendicular or parallel to the flow of water. Microtopography in these areas was generally within ±0.5 m from the mean elevation of the plot. As close to the stream as possible, we established a 32 by 32 m plot (0.1 ha). We chose this plot size for two reasons: (i) soil properties such as amorphous aluminum and iron have been shown to be spatially autocorreated at scales captured by our plots (Lyons et al., 1998), and (ii) it was the largest plot that could be fit onto the floodplain of each site. To allow for a robust Mantel analysis of the data we established sampling designs at both RoBr and GrCr with four transects, separated by 8 m, which ran perpendicular to the direction of water flow (see Fig. 2)
. Each transect consisted of four centroids, allowing us to establish clusters of samples separated by a wide range of distances. This was advantageous because both uniform and random sampling designs fail to provide adequate sample points separated by short distances, which in turn, does not permit analysis of fine-scale spatial variability. Furthermore, random sampling is often very difficult to carry out in the field and uniform designs may have inadequate lag distances or may be out of phase with the existing spatial structure (Fortin et al., 1989). Three of the four centroids on each transect were selected at random for sampling, while the fourth was skipped to keep the sample size from becoming too large. Also, to limit the sample size, we took three cores at two of the centroids, while we took only two cores at the other centroid. Cores were collected at random directions and distances from the centroids (all <4 m) (see Fig. 2). We generated unique random directions and distances for each plot. The location of each of the cores was determined in reference to a datum with a tape measure and compass.
|
Laboratory Analyses of Soil Properties
Upon arrival at the laboratory, the cores were extruded from the plastic sleeves and split in half vertically with a sharp knife. Half of the core was oven-dried at 105°C for 24 h to determine the moisture content and bulk density. The dried soils were then ground and passed through a 2-mm sieve to remove pebbles and macroorganic matter. The sieved soil was then used to determine the percent SOM by loss on ignition (Campbell et al., 2002) and the percent clay, silt, and sand by the pipette method (Sheldrick and Wang, 1993).
The other half of the core was kept at field moisture and passed through a 2-mm sieve. The sieved field-moist soil was analyzed for pH (Hendershot et al., 1993) and oxalate-extractable aluminum (Alox) and iron (Feox) (Richardson, 1985). We used the phosphorus sorption index (PSI) to quantify the P sorption capacity of the soils in this study (Richardson, 1985). Previous studies have established that the PSI (i) serves as a reliable gauge of a wetland soil's P sorption potential, (ii) is less time-consuming to measure than multiple-point P sorption isotherms, and (iii) facilitates comparison with related soil properties (Richardson, 1985; Axt and Walbridge, 1999; Bridgham et al., 2001). The PSI was determined by shaking a sterilized soil sample with a solution of 130 mg P L–1 for 24 h. The difference in concentration of inorganic P between the initial (130) and final concentration represents the amount of P sorbed. The index is then calculated as X(log C)–1, where X is the amount of P sorbed (mg P 100 g soil–1) and C is the final inorganic P concentration in solution (mg P L–1).
Statistical Analyses
Means and standard errors for each soil property (n = 32) were calculated for both RoBr and GrCr. These values were based on the actual soil cores we collected, not on the additional values generated from the geostatistical analyses. Due to the nonindependent nature of the data, we did not compare mean values of the two sites with t tests or analysis of variance. Pearson correlations were also calculated to determine collinearities among soil properties so that redundant variables would not be included in the Mantel analysis.
Geostatistical Analyses
Semivariance analysis (GS+ Version 5.0; Gamma Design Software, 2000) was used to quantify the degree of spatial autocorrelation that exists among soil cores taken from each plot and to facilitate subsequent mapping of soil properties (Boerner et al., 1998). This analysis calculates an index of autocorrelation (the semivariance) among groups of paired samples separated by increasing distances. We calculated empirical semivariance values for eight equal (2 m) distance classes from 0 to 16 m. For both sites, the average number of sample pairs per distance class was 28. The empirical semivariance data were then fit with the spherical semivariogram model (Webster, 1985). A number of parameters were extracted from the fitted spherical model including the following: the nugget (Co, the semivariance at distance zero); the sill (Co + C, the y value at which the semivariance reaches an asymptote); and the range (Ao, the distance [x value] at which this leveling occurs). Finally, we determined Q, which is the ratio of the sill to the sum of the sill and nugget variance (Lyons et al., 1998). This Q parameter represents the proportion of the sample variance explained by the fitted model. In a plot with strong spatial structure, the Q value will approach 1.0; whereas in a plot without spatial structure, the Q value will approach 0.0 (Morris, 1999). We also used a system proposed by Cambardella et al. (1994) to define different classes of spatial dependence (S, strong spatial dependence; M, moderate spatial dependence; W, weak of spatial dependence) for the soil properties measured in this study that are based on the ratio of the nugget to the sill. If the nugget to sill ratio was 
25%, the soil property was considered to be strongly spatially dependent, or distributed in patches; if the ratio was between 26 and 75%, the soil property was considered to be moderately spatially dependent; and if the ratio was >75% the soil property was considered to be weakly spatially dependent (Cambardella et al., 1994).
As the sample plots at each site were chosen for their relatively homogeneous topography and lack of major elevation gradients, we expected that isotropic semivariograms would be acceptable for modeling spatial variability of soil properties and P sorption. Furthermore, it appeared that isotropic semivariogram models fit the empirical semivariogram data better than did anisotropic semivariograms for most soil properties. The best-fit isotropic semivariogram model was then used to map the distribution of soil properties across the study plots by ordinary point kriging (with GS+), a form of nonlinear interpolation that provides optimal, unbiased estimates of points not sampled (Boerner et al., 1998; Ettema and Wardle, 2002).
Nonparametric Statistical Analyses: Simple and Partial Mantel Tests
Moisture, bulk density, and SOM were significantly collinear at both sites as were percent clay and percent silt with percent sand. Thus, we chose to omit moisture, bulk density, and percent sand from the Mantel analysis. While collinearities were found in some of the other soil properties, each was considered sufficiently unique to be included in the Mantel tests. Simple and partial Mantel tests were calculated for each site to analyze the effects of soil properties and spatial variability on P sorption. Mantel tests are a form of partial regression that measures the correlation between two distance or dissimilarity matrices (Mantel, 1967; Sanderson et al., 1995). By using distance matrices, the Mantel approach allowed us to extract variation in soil properties and P sorption that were due to spatial autocorrelation (King et al., 2004). As the elements of these matrices are not independent, standard parametric methods for determining significance of the correlation coefficient are not appropriate. Thus, the significance of the Mantel correlation was assessed by permutation (Manly, 1997). As Mantel correlation coefficients do not behave like product–moment correlation coefficients, they do not have to be large in absolute value to be statistically significant (Legendre and Fortin, 1989).
Partial Mantel tests were also used to calculate the partial correlation (
) between individual soil properties and P sorption, controlling for the effects of spatial autocorrelation (Sanderson et al., 1995). Finally, pure-partial Mantel correlations were estimated. Here, the strength of correlation between a soil property and P sorption was assessed after variation attributed to all other soil properties and spatial autocorrelation had been removed (King et al., 2004). The pure-partial relationship of space and P sorption was also estimated, which would indicate a residual spatial pattern in P sorption that could only be explained by spatial factors (King et al., 2004).
For this study, as the soil properties were in different measurement units, they were standardized into z scores. This standardized data was then converted to a Euclidean distance matrix. Mantel tests were then conducted to examine the relationships of soil properties and space on P sorption capacity. The Mantel test procedure was performed in S+ (Version 6.1) using code developed by Goslee (Urban et al., 2002). Significance of the Mantel coefficients was estimated by bootstrapping with 10000 random permutations. As a visual framework for examining the correlations among P sorption, soil properties, and space, path diagrams were created to depict significant relationships among variables (Leduc et al., 1992).
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
|
As shown in the kriged map (Fig. 3a) , percent clay at RoBr exhibited a patchy distribution. Patches were generally 4 to 8 m in diameter and dispersed across the plot. Grindle Creek followed a different pattern, with the plot capturing only part of what appeared to be a single large patch with the high values near the lower left corner (Fig. 3b). The distribution of Alox at RoBr was similar to that of percent clay, except for the fact that patches of Alox were slightly larger than patches of clay (Fig. 3c). Compared with RoBr, there was much less variability in Alox content across the GrCr plot. The downstream areas (left side) of the GrCr plot had slightly higher Alox values than upstream areas (Fig. 3d). In terms of PSI, the RoBr plot had not only a higher but also a more heterogeneous distribution of values across the plot (Fig. 3e). At GrCr, on the other hand, PSI had a much more homogeneous distribution and increased slightly from the upstream to the downstream areas of the plot (Fig. 3f).
|
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Our mean SOM values of 21.8% at RoBr and 12.0% at GrCr were also similar to mean SOM values of 23.4 and 9.8% reported for two forested riparian wetlands in the Virginia Coastal Plain (Axt and Walbridge, 1999). These results suggested that there is considerable variability in mean SOM content among natural forested riparian wetlands in the southeastern Coastal Plain. The higher SOM content in RoBr may have contributed to its higher PSI values, as another study of Coastal Plain riparian wetlands reported a positive relationship between SOM and P sorption (Axt and Walbridge, 1999). Unlike SOM values, mean soil pH values at RoBr and GrCr were very similar. In addition, mean pH values for RoBr and GrCr were comparable with mean pH values of 4.1, 4.8, and 5.0 reported in other studies of forested riparian wetlands in the Coastal Plain (Axt and Walbridge, 1999; Darke and Walbridge, 2000).
The major difference in the texture of the soils from RoBr and GrCr was that mean percent silt at RoBr was more than double that of GrCr. Consequently, the soil of RoBr would be classified as a silt loam while the soil of GrCr would be classified as a sandy loam (Brady and Weil, 1999). Our results supported the generally accepted idea that soils dominated by fine-textured clays and silts will have a higher P sorption capacity than soils dominated by coarse-textured sands (Sample et al., 1980; Poach and Faulkner, 1998). Our data for mean percent clay, percent silt, and percent sand values of 11.4, 18.4, and 70.2% for GrCr were similar to values of 17.4, 15.1, and 67.5% reported for a floodplain forest in the Georgia Coastal Plain (Darke and Walbridge, 2000).
Oxalate-extractable Al concentrations of RoBr were almost double those of GrCr. Similarly, Feox concentrations of RoBr were much higher than GrCr. Our mean Alox values in RoBr and GrCr were similar both to mean values (1.97, 2.19, and 5.16 mg g–1) reported for a drainage catena spanning a forested riparian wetland in Rhode Island (Lyons et al., 1998), and to the mid-range values (4.50–6.00 mg g–1) reported for 11 swamp forests in Maryland (Richardson, 1985). Our Feox values were similar to mean values (2.16, 1.63, and 3.51 mg g–1) reported for the forested riparian wetland catena in Rhode Island (Lyons et al., 1998).
As previous studies have shown that Alox and Feox are highly correlated with the PSI (Richardson, 1985; Axt and Walbridge, 1999), it was not surprising that RoBr, which had almost twice as much Alox and Feox than GrCr, had mean PSI values that were almost double those of GrCr. Furthermore, the PSI has been shown to be positively correlated to SOM and silt (Axt and Walbridge, 1999), for which RoBr had significantly higher amounts than GrCr. Our mean PSI values for RoBr and GrCr (166.4 and 85.8) were similar to the range of PSI values for swamps in Maryland (approximately 20–160) (Richardson, 1985). These results illustrated that forested riparian wetlands in the southeastern Coastal Plain with similar vegetative communities and similar hydrogeomorphic settings may have very different P sorption capacities.
Geostatistical Analyses of Soil Properties and Phosphorus Sorption
Semivariance analysis indicated that at RoBr, pH, percent Alox, and Feox exhibited strong spatial dependencies, while at GrCr, Alox, Feox, and PSI exhibited strong spatial dependencies. At both sites, percent clay exhibited moderate spatial structuring. As nugget values, which can be viewed as indicators of spatial continuity at close distances, were lower at RoBr than at GrCr for percent clay, Alox, and Feox, it appeared that RoBr exhibited greater fine-scale heterogeneity in soil properties than GrCr. In other words, variance among soil cores taken at short distances from each other was greater at RoBr than at GrCr for the majority of soil properties. A number of the measured soil properties at both sites exhibited high relative spatial structure values, indicating that the spherical model explained the spatial structure of these soil properties quite effectively. For example, sample variances for Alox and Feox at both sites were 90 to 100% spatially structured. Ranges of autocorrelation for soil properties commonly related to P sorption tended to be smaller at RoBr than at GrCr.
The high degree of spatial variability of Alox and Feox observed in this study corroborated with previous studies that reported substantial spatial variability of these soil properties in riparian zones (Lyons et al., 1998; Darke and Walbridge, 2000; Johnston et al., 2001). For example, the study in Rhode Island found that Alox exhibited a range of autocorrelation of 8.8 m and Feox a range of 4.3 m (Lyons et al., 1998). These were similar to the values we reported in this study for Alox at RoBr (13.2 m) and for Feox at RoBr (7.6 m). Unlike the study in Rhode Island, in which P sorption did not exhibit spatial autocorrelation (Lyons et al., 1998), our semivariance data for both RoBr and GrCr suggested that P sorption was spatially autocorrelated.
The tendency of soil properties at RoBr to vary at smaller scales than soil properties at GrCr was further illustrated by the kriged maps. These maps presented a striking illustration of the spatial distributions of percent clay, Alox, and PSI across the study plots. For example, percent clay in the RoBr plot was particularly heterogeneous with patches of high and low clay content. Patch size for percent clay in the GrCr plot, on the other hand, was much larger, as the sample plot captured only part of a larger patch of high clay soil. The spatial distribution of Alox across the two study plots followed a similar pattern to percent clay, with greater heterogeneity in the RoBr plot. High PSI areas in both plots typically were located in areas that had high percent clay and high Alox content. The differences in spatial distributions of soil properties between the two sites may be due to the fact that RoBr was adjacent to a second-order stream while GrCr was adjacent to a third-order stream. Differential frequencies, durations, and intensities of flooding may cause soil properties of headwater streams to be more heterogeneous than those of larger-order streams. In addition, areas within the plots that were further downstream or at slightly lower elevations may have been subject to greater deposition of fine-textured materials that have an affinity for P sorption.
If the spatial heterogeneity of soil properties and P sorption observed in this study is representative of natural forested riparian wetlands in the southeastern Coastal Plain, there are a number of important implications. First, it appears that due to the inherent variability of soil properties in riparian zones, attempts to characterize soil properties in these zones by taking a single soil sample or even by random sampling may be inappropriate. For example, if we had randomly taken three cores from each plot, we might not have found a significant difference between PSI values at RoBr and GL, let alone have any idea about the spatial distribution of P sorption across the two study plots. As noted in other recent studies of wetland soils (Johnston et al., 2001), we recommend that sampling schemes in riparian zones be designed to capture spatial variability of soil properties to extrapolate soil properties across sites with greater accuracy. Second, water quality models may also inaccurately extrapolate homogeneous P sorption values across riparian zones, when in fact P sorption in these zones is highly heterogeneous. Third, our results indicate that assessment of wetland functions with geographic information systems (GIS) analyses that rely on vegetative cover data may be unreliable, as we have shown that two forested riparian wetlands with similar vegetation and hydrogeomorphic settings had very different P sorption capacities and associated patterns of spatial variability. Lastly, relationships of soil properties to P sorption may continue to be oversimplified, if we continue to ignore the fact that these properties are spatially autocorrelated and should be measured with spatial sampling designs. Therefore, despite the additional sampling and labwork required for spatially explicit research, we not only believe it to be worthwhile, but also necessary to further our understanding of P sorption dynamics. Improving our understanding of the spatial variability of soil properties in forested riparian wetlands may also provide insights on how to better reproduce patterns of natural variation in restored or created wetlands (Mummey et al., 2002).
Nonparametric Statistical Analyses: Simple and Partial Mantel Tests
There are some important caveats that should be made before discussing the results of the Mantel tests. First, while "space," in and of itself, does not control the distribution of soil properties at these sites, a significant Mantel correlation with space indicates that soil properties at RoBr and GrCr were influenced by biotic or abiotic processes having spatial components (Leduc et al., 1992). Second, although these Mantel analyses do not establish causation (Petraitis et al., 1996), they do offer strong evidence as to which soil properties are most likely to influence P sorption in natural forested riparian wetlands and suggest the best variables for use in future studies. Third, differences in P sorption at RoBr and GrCr may partially be the result of soil properties that were not measured in this study, such as redox potential or exchangeable calcium. Finally, the measured values for the soil properties may have been noisier than the precise distance measurements, causing statistical relationships to be degraded (King et al., 2004).
Despite these caveats, the strong spatial dependencies of soil properties at RoBr and GrCr as well as the strong Pearson correlations for almost all measured soil properties with PSI, made it necessary to employ a spatially explicit approach to determine which soil properties best explained P sorption. At both RoBr and GrCr, a number of soil properties exhibited significant simple Mantel correlations with PSI. The partial Mantel tests allowed us to examine the relationship of soil properties to P sorption after removing the effects of spatial autocorrelation. At both RoBr and GrCr, none of the soil properties that exhibited significant simple Mantel correlations with PSI ceased to be significantly related to PSI, after partialling out the effects of space. However, at RoBr, after removing the effects of all the other soil properties and space, only Alox and percent silt continued to have significant correlations with P sorption, and percent clay actually emerged with a significant correlation to PSI. Had we not employed the Mantel tests, we would have concluded that, at RoBr, SOM and pH were significant predictors of P sorption. In actuality, once we accounted for collinearities with other soil properties and spatial autocorrelation, SOM and pH were no longer significantly related to P sorption. Furthermore, we might have concluded that at RoBr, percent clay was not a useful predictor of PSI. However, according to the pure-partial Mantel test, percent clay accounted for a unique part of the variability in P sorption that was not accounted for by any other soil property. While previous studies have shown Feox to be an excellent predictor of P sorption (Darke and Walbridge, 2000; Bridgham et al., 2001), at RoBr, Feox did not exhibit a significant Mantel correlation with P sorption. One explanation for this may be that acid ammonium oxalate extracts appreciable amounts of magnetite Fe, which does not bind with phosphate (Walbridge and Struthers, 1993). The RoBr site may have had higher pools of magnetite Fe than did the GrCr site. At GrCr, after removing the effects of all other soil properties and space, SOM, pH, percent clay, Alox, and Feox all remained significant predictors of P sorption.
We determined the soil property with the strongest pure-partial Mantel correlation with PSI at both RoBr and GrCr to be Alox. While Alox was the best predictor of PSI at both sites, the different suites of soil properties that had significant pure-partial Mantel correlations with PSI at RoBr and GrCr suggested that P sorption dynamics may vary from site to site, depending not only upon the amounts but also upon the spatial distributions of soil properties across the sites. However, if Alox is the master variable controlling PSI in natural forested riparian wetlands with acid soils, as suggested by previous studies (Richardson, 1985), we recommend that future research of P sorption dynamics in riparian wetlands should focus on Alox, as well as factors that control development and spatial distribution of amorphous aluminum pools including parent material, weathering, and organic matter accumulation. In future studies, if financial or logistical constraints limit the number of soil properties to be analyzed, it appears that the measurement of Alox alone will give a strong indication of the P sorption capacity. Further research is also needed to determine whether the quantity and spatial distribution of Alox found in natural wetlands can be found in similar degrees in mitigation wetlands.
|
CONCLUSIONS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONCONCLUSIONSREFERENCES |
|---|
Related articles in JEQ:
This article has been cited by other articles:
|
|
 |
|
  R. E. Vaughan, B. A. Needelman, P. J. A. Kleinman, and A. L. Allen Spatial Variation of Soil Phosphorus within a Drainage Ditch Network J. Environ. Qual., May 25, 2007; 36(4): 1096 - 1104. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. L. Bruland, S. Grunwald, T. Z. Osborne, K. R. Reddy, and S. Newman Spatial Distribution of Soil Properties in Water Conservation Area 3 of the Everglades Soil Sci. Soc. Am. J., August 3, 2006; 70(5): 1662 - 1676. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. J. Anderson, W. J. Mitsch, and R. W. Nairn Temporal and Spatial Development of Surface Soil Conditions at Two Created Riverine Marshes J. Environ. Qual., October 12, 2005; 34(6): 2072 - 2081. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. J. Cohen, J. P. Prenger, and W. F. DeBusk Visible-Near Infrared Reflectance Spectroscopy for Rapid, Nondestructive Assessment of Wetland Soil Quality J. Environ. Qual., July 5, 2005; 34(4): 1422 - 1434. [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]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
