SSSAJ
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 18 June 2008
Published in Soil Sci Soc Am J 72:1174-1183 (2008)
DOI: 10.2136/sssaj2007.0354
© 2008 Soil Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Grunwald, S.
Right arrow Articles by Reddy, K. R.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Grunwald, S.
Right arrow Articles by Reddy, K. R.
Agricola
Right arrow Articles by Grunwald, S.
Right arrow Articles by Reddy, K. R.
Related Collections
Right arrow Wetlands and Aquatic Processes
Right arrow Nitrogen
Right arrow Phosphorus
Right arrow Spatial Distribution

WETLAND SOILS

Spatial Behavior of Phosphorus and Nitrogen in a Subtropical Wetland

S. Grunwald* and K. R. Reddy

Soil and Water Science Dep., Univ. of Florida, 2169 McCarty Hall, P.O. Box 110290, Gainesville, FL 32611

* Corresponding author (sabgru{at}ufl.edu).


   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Complex spatial patterns of biogeochemical properties such as soils are generated in wetlands as a result of internal cycling of nutrients, natural forcing functions (e.g., hydrologic flow paths), and anthropogenic impacts (e.g., nutrient inputs). To preserve ecosystem health and the functions of a wetland ecosystem, it is critical to maintain a soil nutrient status that resembles natural reference conditions and to preserve biogeochemical variability. We investigated the spatial behavior of total P (TP) and total N (TN) in a subtropical wetland in Florida that historically has been impacted by nutrient influx. Our specific objectives were to: (i) assess the spatial autocorrelation structure and spatial variability of TP and TN across a subtropical wetland given a dense observation set (n = 266); and (ii) assess the information loss associated with underlying spatial autocorrelations and spatial variability in TP and TN under various scenarios of reduced sampling densities (n = 175–50). Both properties showed contrasting spatial metrics with long spatial autocorrelations for TP (7240 m) and much shorter ones for TN (1007 m). The contrasting spatial metrics of TP and TN along trajectories of sparser observation sets translated into different responses in spatial patterns that disintegrated dramatically for TN and less so for TP when compared with the reference set (n = 266). In this subtropical wetland, a sampling set of n > 75 (0.036 samples ha–1) for TP and n > 125 (0.0260 samples ha–1) for TN yielded satisfactory results to resemble spatial patterns of the reference set. This study provided insight into the spatial behavior and patterns of two important soil nutrients in wetlands.

Abbreviations: BCMCA, Blue Cypress Marsh Conservation Area • MPE, mean prediction error • TN, total nitrogen • TP, total phosphorus


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Wetlands and aquatic systems are complex in their biogeochemical structure and composition. This determines their resilience and sensitivity to respond to internal and external stresses. Anthropogenic forces, such as nutrient influx, and natural events, such as hurricanes and fire, may impact wetlands at local or regional spatial scales. Biogeochemical cycling (e.g., N, P, C, and S cycles) and transport processes continuously change pedological and biological patterns of aquatic ecosystems. Noe et al. (2001) presented in a meta-analysis of how spatial patterns of soil P are interlinked to periphyton, vegetation, hydrologic manipulation, nutrient influx, and internal biogeochemical cycling in the Greater Everglades (~8500 km2), one unique and large subtropical wetland in Florida. Barbiéro et al. (2002) illustrated the complexity of the geochemistry of surface and groundwater in the Pantanal of Mato Grosso, Brazil. The health and functions of the Pantanal in the heart of South America (~165,000 km2), which is the world largest freshwater wetland, has been threatened by anthropogenic impacts.

A large number of wetland studies have focused on the characterization of biogeochemical soil properties along expected gradients (transects) of impact (Qualls and Richardson, 1995; Reddy et al., 1998; Richardson and Qian, 1999; White and Reddy, 2000; Fisher and Reddy, 2001; Craft and Chiang, 2002). Other studies characterized spatial patterns of soil and water properties across wetlands at various spatial scales in different geographic regions (Table 1 ). Grunwald et al. (2007b) pointed out that biogeochemical cycling within wetlands and aquatic systems can only be understood within a spatially explicit context considering the spatial autocorrelation and co-variation among different environmental properties. The spatial autocorrelation (or range) describes the relatedness among measurements, indicating that places close to one another tend to have similar values, whereas ones that are farther apart differ more (Webster and Oliver, 2001). In a subtropical freshwater marsh in the Everglades it was demonstrated how sparse, site-specific observations were too simplistic to characterize the complex spatial patterns of soil P (Grunwald et al., 2007b) that were generated by various ecosystem processes (P influx, transformations, resuspension, transport, and chemical precipitation). Complex interactions between P and other constituents (organic matter, metals) and environmental factors (vegetation, hydrology) and external forces (e.g., fire) continuously caused the co-evolution of soil patterns across this subtropical marsh (Grunwald et al., 2007b).


View this table:
[in this window]
[in a new window]

  		Table 1. Overview studies that characterized the spatial distribution and variability of soil and water properties in wetlands.

 
Questions remain that address the population size and density of observations that are required to characterize the spatial variability and distribution patterns of biogeochemical properties within wetlands. Reddy and DeLaune (2008) argued that spatially explicit modeling of biogeochemical properties in wetlands at an appropriate spatial scale is critical for our understanding of ecosystem processes. They suggested using biogeochemical indicators (Levels I, II, and III) that behave differently at micro, meso, and macro spatial scales to characterize wetland health. Knowledge of continuous spatial distributions and scaling behavior of biogeochemical soil properties in wetlands and aquatic systems, however, is still limited. Investigations that aimed at describing the spatial variability of biogeochemical soil properties have ranged from paddy fields (0.5 ha) to large wetlands (825,000 ha) (Table 1). High-intensity studies used sampling densities of 182.0 samples ha–1 (n = 91, Yanai et al., 2000), 8333.3 samples ha–1 (n = 300, Lyons et al., 1998), and 541 samples ha–1 (n = 541, Gallardo, 2003), while other studies have relied on fewer observations to characterize spatial patterns (Mitsch et al., 1995; DeBusk et al., 2001; Bruland and Richardson, 2005). Regional investigations of biogeochemical patterns in soils have used coarse sampling densities, with 0.0025 samples ha–1 (n = 111, Rivero et al., 2007a,b), 0.0019 samples ha–1 (n = 388, Bruland et al., 2006a), and 0.0016 samples ha–1 (n = 1341, Grunwald et al., 2008). Often costs and labor are prohibitive to use high sampling densities for sampling across larger wetlands. Coarser spatial studies often aim to elucidate on regional spatial patterns but fail to reveal fine-scale spatial patterns (compare Table 1).

Grunwald et al. (2007a) suggested distinguishing among three major groups of different biogeochemical indicator variables in wetlands: (i) labile, fast-response properties with fine-scale spatial autocorrelation (e.g., total labile organic N and total labile organic C); (ii) stable, slow-response properties with regional spatial autocorrelation (e.g., total P); and (iii) properties showing intermediate response (e.g., total C). Interestingly, the documented spatial autocorrelation range of biogeochemical properties in wetlands and aquatic systems differed widely among studies. The range was relatively short, with values as low as 2 m for pH, 37 m for available P, and 49 m for K in paddy fields (Yanai et al., 2000), suggesting that observations should be taken at 2 m or finer intervals to capture the underlying spatial biogeochemical patterns. Fennessy and Mitsch (2001) identified a range of 300 m for Bray-extractable P in a floodplain in Illinois. In contrast, DeBusk et al. (2001) identified long-range (regional) spatial autocorrelation structure for total P (TP) with 9.3 (1990) and 11.3 km (1998) in a subtropical wetland in Florida. The long range for TP in the topsoil was confirmed in other studies at 6400 (1990) and 7550 m in Water Conservation Area (WCA)-2A (Grunwald et al., 2004), 7270 m in WCA-3AS and 13,750 m in WCA-3AN (Bruland et al., 2006a), double ranges with 3290 and 9978 m in WCA-1 (Corstanje et al., 2006), and 7468 m in WCA-2A (2003) (Rivero et al., 2007b). These findings suggest that soil TP shows large (regional) spatial autocorrelations in subtropical wetlands that requires less dense sampling to characterize spatial distribution patterns.

Other biogeochemical properties seem to vary across intermediate spatial scales, with spatial autocorrelations on the order of <1500 m (e.g., microbial biomass P, NH4–N, peptidase activity, and total labile organic C) (Grunwald et al., 2007a). To avoid oversampling and excessive costs and labor, the aim is to identify an optimal spatial scale and sampling density for accurate characterization of biogeochemical patterns across aquatic and wetland ecosystems. Geostatistical analysis provides metrics to quantify spatial variability and distribution patterns (Goovaerts, 1997; Webster and Oliver, 2001). This study underpins the importance of observing biogeochemical properties at an appropriate spatial scale and density.

Our specific objectives were to: (i) assess the spatial autocorrelation structure and spatial variability of TP and total N (TN) across a subtropical wetland given a dense observation set; and (ii) assess the information loss associated with underlying spatial autocorrelations and spatial variability in TP and TN under various scenarios of reduced sampling densities.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Study Area
The study area comprised 4800 ha of freshwater wetland within the Blue Cypress Marsh Conservation Area (BCMCA), located in the headwater region of the St. Johns River in east-central Florida. The BCMCA is a subtropical marsh that has been impacted by nonpoint-source pollution, which began to be reduced in the mid-1990s. Since then, this ecosystem has undergone a natural succession of recovery. Ollila et al. (1995) and D'Angelo et al. (1999) documented the occurrence of nutrient-enriched and unimpacted zones within the marsh. Native vegetation in the study area is predominately a mosaic of sawgrass (Cladium jamaicense Crantz) and maidencane (Panicum hemitomon Schult.) flats, areas of scrub-shrub vegetation (e.g., the coastal plain willow, Salix caroliniana Michx.), cattail (Typha ssp.) marshes, and deep-water slough communities (e.g., white water lily, Nymphaea ssp.). Drying of the marsh permitted the expansion of woody vegetation (e.g., coastal plain willow) into areas previously occupied by herbaceous, wetland marsh plants. Since the early 1970s, coastal plain willow has been expanding from the southern to the northern part of the study area. Soils in the BCMCA are in the Histosol soil order (NRCS, 2007).

Field Data Set
We used a data set (n = 266) collected within the BCMCA in March to April 2002 by the staff of the Wetland Biogeochemical Laboratory, Soil and Water Science Department, University of Florida. Each soil sample was the composite of two soil cores (0–10-cm depth) taken with a stainless steel core tube. Samples were sealed in plastic bags and stored on ice until received at the laboratory, where they were stored at 4°C until analysis. A targeted sampling design was used with irregular spacing on an approximate grid, with spacings of 31 and 420 m between sampling locations. Total soil P was determined using the ashing method (Anderson, 1976). Phosphorus content was analyzed using an automated colorimetric analysis (USEPA, 1993, Method 365.1). Total N was measured using a Carlo-Erba NA-1500 CNS Analyzer (Haak-Buchler Instruments, Saddlebrook, NJ). All analyses for this study followed National Environmental Laboratory Accreditation Conference quality control and quality assurance protocols.

Geospatial Analysis
Semivariograms were derived for TP and TN, respectively, using the total data set collected at 266 observation sites. Variables were log transformed to better comply with stationarity assumptions and to stabilize variances, as suggested by Webster and Oliver (2001). Log-normal point ordinary kriging (LN-OK) (Goovaerts, 1997) was used to derive estimates of each soil variable at unsampled locations across a grid, using a 100-m resolution. The geostatistical analyses were performed in ISATIS (Geovariances Inc., Houston, TX). To evaluate the effect of using a reduced data set to represent the underlying spatial variability and distribution of TP and TN, we considered the following criteria: (i) random selection of observations might be biased toward specific geographic regions, resulting in clustered or clumped subsets; and (ii) a one-time selection of observations from the total data set ignores the impact of sampling fluctuations on prediction performances. Thus, subsets of 175, 150, 125, 100, 75, and 50 observation sites from the total soil population (n = 266) were selected in jackknife mode using a random number generator. Jackknifing is a statistical method of numerical resampling based on deleting a portion of the original observations in subsequent samples. The average sampling density was 0.0365 samples ha–1 (n = 175), 0.0312 samples ha–1 (n = 150), 0.0260 samples ha–1 (n = 125), 0.0208 samples ha–1 (n = 100), 0.0156 samples ha–1 (n = 75), and 0.0104 samples ha–1 (n = 50). The average sampling density for the total data set of n = 266 was 0.0554 samples ha–1. The random selection procedure to identify each subset was repeated 50 times to account for sampling fluctuations. For each selected subset, the semivariogram was computed and TP and TN predicted using LN-OK (Goovaerts, 1997). The 50 simulations of TP and TN at each subset level generated cumulative distribution functions that were characterized by their mean and CV for each individual wetland pixel (100 by 100 m2). A validation set of 91 samples was used to assess model performance for different subsets of 175, 150, 125, 100, 75, and 50 observation sites, using 50 iterations for each subset. Model performance was evaluated in validation mode for TP and TN using the mean prediction error (MPE) and RMSE (Webster and Oliver, 2001).

For TP, a cutoff value of 550 mg kg–1 was assumed, resembling historic conditions, and for each subset (n = 175, 150, 125, 100, 75, and 50), probabilities to exceed the cutoff value were calculated. We further assumed that the dense observation model (n = 266), called the reference model, was the most accurate among the considered soil populations. A misclassification rate was calculated as the deviation between model and reference probabilities for each subset (n = 175, 150, 125, 100, 75, and 50) using ArcGIS 9.1 (Environmental Systems Research Inst., Redlands, CA).


   RESULTS AND DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Total P ranged from 349.6 to 1013.7 mg kg–1 with a median of 602.8 mg kg–1. Total N showed a minimum of 17.6, maximum of 82.4, and median of 26.7 g kg–1 (Table 2 ). Semivariograms and estimated TP and TN are shown in Fig. 1 . A spherical model was used to fit the experimental semivariogram of TP with a nugget variance of 0.154, sill variance of 1.327, and range of 7240 m. While long-range spatial autocorrelation dominated TP, short-range spatial autocorrelation was found for TN. The experimental semivariogram for TN was fitted using a spherical model with a nugget variance of 0.044, sill variance of 0.808, and range of 1007 m. Interestingly, the strength of the spatial relationships was stronger for TN than for TP, as indicated by the nugget/sill ratio. Total P and TN showed contrasting spatial patterns, with high TP in the south and east of the wetland and concentrated TN hotspots in the west and northeast (Fig. 1). A crescent-shaped area in the northwest of the BCMCA showed the lowest TP. Historical P inputs from agricultural activities entered the marsh in the south, which is consistent with the observed spatial patterns of TP.


View this table:
[in this window]
[in a new window]

  		Table 2. Descriptive statistics of total P (TP) and total N (TN).

 

Figure 1
View larger version (37K):
[in this window]
[in a new window]

  		Fig. 1. Variography and predictions based on log-normal kriging using the dense soil population (n = 266) (experimental semivariogram, thin line, D1; fitted semivariogram, thick line, M1): (a) semivariogram of log total P (TP); (b) prediction of TP; (c) semivariogram of log total N (TN); (d) predictions of TN.

 
Spatial patterns of TP may be linked to N cycling (White and Reddy, 2000), microbial community structure (Drake et al., 1996), and the productivity and community structure of macrophytes (Rivero et al., 2007b). Other processes that may play a major role for P retention include uptake and release by periphytons and microorganisms, sorption and exchange reactions with soils and sediments (Richardson, 1985), chemical precipitation in the water column, and sedimentation and entrainment (Reddy et al., 1999). In the BCMCA, it was found that TP was positively correlated with labile inorganic P (r = 0.53) and labile organic P (r = 0.52), but less so with microbial biomass P (r = 0.21) (Grunwald et al., 2006). These findings are unusual because eutrophication in marsh systems has generally been associated with increases in biomass content (Qualls and Richardson, 2000). In the BCMCA, however, the level of microbial biomass seemed to primarily respond to factors other than TP enrichment. The findings of Grunwald et al. (2006) in the BCMCA suggested that, in TP-enriched areas, the primary productivity is enhanced, which increased the sizes and turnover rates associated with different P pools.

Mean spatial patterns of TP generated using various subsets (n = 175–50) based on 50 iterations are shown in Fig. 2 . Spatial patterns disintegrated slightly under reduced sample densities, although regional high and low TP values were prominent across all six maps. The spatial autocorrelation range for TP was relatively persistent across various sample subsets (n = 175, 150, 125, 100, 75, and 50) at 7293, 7267, 7296, 6080, 5730, and 5710 m, respectively (Table 3 ). Similarly, the TP nugget/sill ratio, which describes the strength of the spatial relationship, was preserved across various sample sets and ranged from 0.22 to 0.26. This suggests that TP was relatively insensitive to different sampling densities. Even the sparser data sets of n = 75 and 100 seemed able to preserve the spatial patterns of TP across the wetland. Only very low sample densities (n = 50) were less able to resemble the spatial patterns found with denser samplings.


Figure 2
View larger version (40K):
[in this window]
[in a new window]

  		Fig. 2. Mean total P (TP) computed for subsets of 175, 150, 125, 100, 75, and 50 observation sites with 50 iterations.

 

View this table:
[in this window]
[in a new window]

  		Table 3. Pooled semivariogram metrics based on 50 iterations of log(total P) and log(total N) for different sets of observation sites.

 
The disintegration of spatial patterns under reduced sample sets was more severe for TN than for TP (compare Fig. 2 and 3 ). Total N showed heterogeneous spatial behavior based on denser sampling (e.g., n = 266 and 175) due to the short spatial autocorrelations of 1007 and 1006 m, respectively. The spatial autocorrelation range for TN for different sampling densities (175, 150, 125, 100, 75, and 50 m) changed dramatically from 1006 to 913, 999, 1425, 5026, and 6210 m, respectively (Table 3). The large spatial autocorrelation range of 6210 m (n = 50) generated the more muted, homogeneous spatial TN patterns when compared with the heterogeneous, fine-scale TN patterns shown for the range of 1006 m (n = 175). The nugget/sill ratio for TN was relative stable for observation sets between 266 and 125, with 0.05 to 0.17 indicating strong spatial relationships, but started to increase dramatically for n < 125. This suggests that spatial relationships for TN were not represented adequately using smaller observations sets of n < 125. This was confirmed by a dramatic increase in the TN range for n < 125 compared with denser TN sample sets.


Figure 3
View larger version (54K):
[in this window]
[in a new window]

  		Fig. 3. Mean total N (TN) computed for subsets of 175, 150, 125, 100, 75, and 50 observation sites with 50 iterations.

 
The processes that may have contributed to generate TN spatial patterns include transport and transformations of N including enzyme hydrolysis, mineralization, nitrification, NH4–N adsorption and desorption, NH3–N volatilization, denitrification, and vegetative assimilation and decay (Reddy and DeLaune, 2008). Based on mechanistic N simulations in wetlands by Martin and Reddy (1997) under low-N-input conditions (51 g N yr–1), vegetative change accounted for –35.0 g N yr–1, denitrification for 27.2 g N yr–1, N soil change for –13.9 g N yr–1, and volatilization for 1.8 g N yr–1, resulting in a total N output of the wetland of 69.7 g N yr–1. In contrast, under high-N-loading conditions (2453 g N yr–1), vegetative change accounted for 44.7 g N yr–1, denitrification for 651.5 g N yr–1, N soil change for 707.4 g N yr–1, and volatilization for 400.9 g N yr–1, resulting in total N output of 652.3 g N yr–1 for the wetland. Similarly, in the BCMCA, spatial TN patterns in areas impacted by N influx may have been controlled by soil N change, whereas areas low in N were probably dominated by denitrification and vegetative change.

The spatial behavior of TP and TN was very different across the BCMCA, which suggests that biogeochemical indicator variables be separated based on their spatial variability and behavior (Grunwald et al., 2007a). This has implications for the observation density and population size required to map a given biogeochemical indicator variable. For TN, a much higher sampling density was required to capture spatial patterns than for TP. Webster and Oliver (2001, p. 90) pinpointed that "100 observations might be acceptable in some circumstances and 144 are likely adequate (to characterize soil spatial variability), at least with normally distributed isotropic data. Semivariograms computed from samples of 225 will almost certainly be reliable, and samples over 400 are extravagant." Chilès and Delfiner (1999) considered experimental semivariograms not reliable if calculated from <50 observation data pairs. The small number of observations (n < 50) used in numerous wetlands studies (e.g., Mitsch et al., 1995; Ettema et al., 1998; Bruland and Richardson, 2005) may be of concern; however; the researchers acknowledged the considerable difficulties associated with collecting large numbers of samples in spatially explicit designs at remote, wet, and densely vegetated sites (Bridgham et al., 2001).

As expected, the validation showed larger MPE and RMSE values with decreasing sampling density (n = 266–50) for TP and TN (Table 4 ); however, the RMSE increased only 18% for TP but 390% for TN for the reduced sample set (n = 50) when compared with the dense reference set (n = 266). Such large RMSE and MPE errors for TN for small sample sets of n < 125 are neither desirable nor acceptable due to large uncertainties in predictions. This was confirmed by the large CVs for TN for various subset scenarios (n = 175–50) when compared with CVs for TP that appear much lower (compare Fig. 4 and 5 ). The CV maps describe the uncertainty of predictions based on 50 iterations for various subset samples (n = 175–50).


View this table:
[in this window]
[in a new window]

  		Table 4. Validation results: a validation set of 91 samples was used to assess the mean prediction error (MPE) and root mean square prediction error (RMSE) for the dense soil population (n = 266) and subsets of 175, 150, 125, 100, 75, and 50 observation sites based on 50 iterations.

 

Figure 4
View larger version (77K):
[in this window]
[in a new window]

  		Fig. 4. Coefficients of variation for total P (TP) for subsets of 175, 150, 125, 100, 75, and 50 observations with 50 iterations.

 

Figure 5
View larger version (70K):
[in this window]
[in a new window]

  		Fig. 5. Coefficients of variation for total N (TN) for subsets of 175, 150, 125, 100, 75, and 50 observations with 50 iterations.

 
To restore wetlands, often threshold values are used to indicate nutrient enrichment (Reddy and DeLaune, 2008). In particular, oligotrophic subtropical wetlands are P limited and respond to P enrichment (Rivero et al., 2007b). This is also true for the BCMCA, where nutrient influx has dramatically changed vegetation patterns, with non-native species dominating nutrient-enriched areas. To assess the impact of biased assessment of TP spatial patterns, the probabilities of exceeding a cutoff TP value of 550 mg kg–1 were evaluated using various sample subsets (n = 175–50) (Fig. 6 ). Overall, probabilities with small sample densities (n < 100) showed smaller coverage of low probabilities (<0.25), larger coverage of intermediate probabilities (0.25–0.75), and smaller coverage of large probabilities (>0.75) compared with the dense sample set (n = 266). These under- and overestimations in exceedence probabilities from subsets (n = 175–50) to the reference set (n = 266) are shown in Fig. 7 . These maps illustrate the magnitude of deviations between the reference set, which was assumed most accurate, and sample subsets (n = 175–50). Some of the over- and underestimations were dramatic, as indicated in dark red and dark green (Fig. 7).


Figure 6
View larger version (62K):
[in this window]
[in a new window]

  		Fig. 6. Probabilities to exceed total P (TP) of 550 mg kg–1 computed for subsets of 175, 150, 125, 100, 75, and 50 observation sites with 50 iterations.

 

Figure 7
View larger version (24K):
[in this window]
[in a new window]

  		Fig. 7. The dense soil population of total P (TP) was used as reference and model probabilities for subsets of 175, 150, 125, 100, 75, and 50 observation sites with 50 iterations subtracted to derive deviation maps.

 
The implications for biased TP predictions (i.e., large overestimations) are noteworthy. Misleading spatially explicit predictions of TP or TN that provide base data for wetland restoration programs involve much risk. For successful restoration and impact assessment, it is pertinent to aim at preserving spatial biogeochemical patterns and diversity across wetlands and aquatic systems. Thus, it is not only important to assess the magnitude (e.g., maximum values of TP and TN) but to assess the spatial distribution patterns and spatial diversity. Optimized observation schemes for wetlands and aquatic systems are needed that provide insight into the spatial behavior of various biogeochemical patterns that take into account short- and long-range variability of properties. One-dimensional investigations that analyze biogeochemical properties at a few sites may be too simplistic and fail to incorporate lateral biogeochemical interactions with adjacent locations.


   CONCLUSIONS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
This study characterized the spatial variability of soil TP and TN across a subtropical freshwater marsh at different sampling densities. The disintegration of spatial patterns of soil TP and TN across a subtropical wetland in Florida was illustrated along trajectories of dense (n = 266) to sparse (n = 50) observations. The two properties showed contrasting spatial metrics, with long spatial autocorrelations for TP (7240 m, n = 266) and much shorter ones for TN (1007 m, n = 266). Total P and TN also different in terms of the strength of the spatial relationship, as indicated by the nugget/sill ratio, which was twice as high for TP as for TN. Spatial patterns of TP were much more resilient across different observation densities compared with TN. The contrasting spatial metrics of TP and TN along trajectories of sparser observation sets (n = 175–50) translated into different responses in spatial patterns that disintegrated dramatically for TN and less so for TP when compared with the reference set (dense set, n = 266 or 0.0554 samples ha–1). In this subtropical wetland, a sampling set of n > 75 (or 0.0156 samples ha–1) for TP and n > 125 (or 0.0260 samples ha–1) for TN yielded satisfactory results to resemble the spatial patterns of the reference set. Yet larger observation densities are desirable to quantify spatial biogeochemical patterns across wetlands.

There are numerous implications of our findings for successful assessment of nutrient enrichment, functionality, and restoration of an impaired wetland ecosystem. Besides accurate assessment of nutrient status (TP, TN, and various other biogeochemical response variables), it is pertinent to characterize spatial behavior and distribution patterns of biogeochemical properties. It is critical to preserve biogeochemical variability (or pedodiversity) to maintain ecosystem health and functions of a wetland system. Observation schemes that measure the spatial variability of biogeochemical properties at a suitable spatial scale and density of measurements are needed. Geostatistical metrics such as spatial autocorrelations for various biogeochemical properties can guide this endeavor. This study provided insight into the spatial behavior and patterns of two important soil nutrients in wetlands that may guide future investigations. More research is needed to gain insight into the spatial behavior of various other biogeochemical indicator variables.


   ACKNOWLEDGMENTS
 
The data collection was supported by a grant from the USEPA, Grant no. R-827641-01. We thank J. Prenger, Matt M. Fisher, R. Corstanje, and Y. Wang for assistance in sampling and analytical analysis of soil samples.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.

Received for publication September 27, 2007.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Grunwald, S.
Right arrow Articles by Reddy, K. R.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Grunwald, S.
Right arrow Articles by Reddy, K. R.
Agricola
Right arrow Articles by Grunwald, S.
Right arrow Articles by Reddy, K. R.
Related Collections
Right arrow Wetlands and Aquatic Processes
Right arrow Nitrogen
Right arrow Phosphorus
Right arrow Spatial Distribution

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Crop Science
Journal of Natural Resources
and Life Sciences Education		 Vadose Zone Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome