Linking Historic, Present and Future Spatial Variability of Soil Attributes in the Greater Everglades Ecosystem
S. Grunwald, T.F.A. Bishop, and K. R. Reddy
Soil and Water Science Department, University of Florida, Gainesville, FL and
S. Newman
South Florida Water Management District
Introduction
Wetlands are known to accrete nutrients (N and P) and other contaminants. The Greater Everglades ecosystem has been impacted by agricultural and urban activities over several decades leading to phosphorus enrichment (DeBusk et al., 1994; DeBusk et al., 2001; Newman et al., 1997). The degree of nutrient enrichment depends both on nutrient loading and hydraulic retention time. This effect is distinct in many wetlands, and most notably in several hydrologic units of the Everglades, including water conservation areas, and the Everglades National Park.
Some variables, such as vegetation have visible patterns and their spatial scales are obvious. Remote sensing techniques can be employed to capture these patterns. However, many other biogeochemical attributes in the soil and water column are invisible, hence challenging to assess the spatial scales at which they vary without first sampling exhaustively. Attributes can also vary at scales that differ by several orders of magnitude simultaneously. Historic data are valuable to assess the spatial variability of floc/detrital plant tissue, surface water and soils attributes and guide future sampling designs and the assessment of environmental quality. The identified spatial scale and autocorrelation from historic surveys are beneficial to target future sampling locations reducing costs and labor. Our objectives were to assess the spatial variability of selected soil quality indicators in the Greater Everglades ecosystem using previous observations and to develop optimized sampling designs for future studies.
Methodology
The variogram is the cornerstone of geostatistics, and it is therefore vital to estimate it and model it correctly. Kriging requires the calculation of an experimental semivariogram to which a theoretical model is fitted. This provides a description of the spatial structure of the attribute (Fig. 1). The key features are the nugget semivariance which is representative of the measurement error and unmeasured variation at distances shorter than the smallest sampling interval. The other important feature is the range at which the spatial autocorreclation becomes 0. The range marks the limit of spatial dependence, i.e., the distance at which there is no spatial relationship between sampling points. Observations further apart than the range are spatially independent. The sill is the semivariance at the range and is the a priori variance of the process (Webster and Oliver, 2001). The formula to calculate the semivariance is presented below (Eq. 1).
Eq. 1:
where
: estimated semivariance
z(xi): data values
h: lag vector (distance)
m(h): number of pairs of data points separated by the
particular lag vector Fig. 1. An experimental semi- variogram and model parameters.
In this study we used previous observations of chemical soil attributes shown in Fig. 2 collected by the Wetland Biogeochemistry Laboratory (WBL), Soil and Water Science Dept. University of Florida and U.S. Environmental Protection Agency.
Survey / Date / # SitesWCA-1 (WBL) / 9/1991 / 103
WCA-21 (WBL) / 7/1990 / 74
WCA-3 (WBL) / 2/1992 / 100
WCA-3 (WBL) / 6/1992 / 74
EPA / 4/1995 / 120
EPA / 9/1995 / 123
EPA / 5/1996 / 123
EPA / 9/1996 / 119
EPA / 5/1999 / 121
EPA / 9/1999 / 119
Fig. 2. Available soil attribute datasets used in this study.
Results
When designing a sampling scheme it is crucial to sample at distances smaller than the range, and also to sample at very small distances to adequately characterize the nugget semivariance. Therefore, the semivariogram parameters for soil properties from previous soil surveys were used to provide a guide to the required sampling density. A stratified random sampling design was chosen to identify sample locations. Zones within each hydrological unit were identified with k-means clustering of the kriged soil data (Hartigan & Wong, 1979). The number of zones within each hydrological unit was chosen subjectively based on prior expert knowledge of the study areas. Samples were randomly allocated within each zone where the proportion of samples per zone is equal to the product of the area and the within-zone standard deviation (SD) in total phosphorus (TP) 0-10 cm. This methodology was chosen to ensure that large zones with low variability were not over-sampled and small zones with high variability were not under-sampled. TP was chosen as it is the soil property of most interest in this study. In each hydrological unit approximately 80% of the sampling stations were identified using stratified random sampling. The remaining stations were allocated to characterize the short range variability. Results for Water Conservation Area 1 (WCA-1) are given below (Table 1 and 2). The same methodology was applied to the other hydrological units of Fig. 2.
Table 1. Semivariogram parameters for WCA-1 dataset (number of samples: 103; time of data collection: September 1991) (Reddy et al., 1994a).
Attribute / Model / Nugget / Sill / Range (m) / Sample No.*Total P 0-10cm / Spherical / 92,662 / 71,892 / 21,675 / 19
Total N 0-10cm / Spherical / 14,164,940 / 21,982,981 / 21,298 / 20
Total C 0-10cm / Spherical / 8.94 / 12.91 / 20,591 / 21
*Sample number is based on a grid sampling scheme where the grid spacing is equal to one quarter of the range parameter in the semivariogram. Therefore, a sample number between 19 and 21 was needed to characterize the spatial variability of Total N, P and C. This is for a grid sampling scheme which have been found to be inefficient for characterizing spatial variability. Instead a stratified random sampling scheme is suggested.
Table 2. Cluster statistics for WCA-1.
Cluster / Area (ha) / Mean TP 0-10cm / SD TP 0-10cm / Sample No.*1 / 21,916 / 465.0 / 78.2 / 35/4
2 / 10,844 / 717.9 / 123.1 / 28/4
3 / 23,944 / 334.3 / 34.7 / 17/2
* Number to the left of the dash is number of samples randomly allocated within the cluster; number to the right is the number of locations where short range variability will be sampled.
Discussion
Spatially explicit modeling of chemical, physical and biological attributes is essential to understanding the structure and function of biodiversity at the soil/water interface of wetlands. Characterization of these patterns is pivotal to improve our understanding of factors that drive phosphorus retention and mobilization across spatial and temporal scales. The suggested methodology facilitates to improve the documentation of the ongoing restoration efforts in the Everglades ecosystem.
Sabine Grunwald, Soil and Water Science Department, University of Florida, Institute of Food and Agricultural Sciences, 2169 McCarty Hall, PO Box 110290, Gainesville, FL 32611-0290, Phone: 352-392-4508, Fax: 352-392-3902, Email:
References
DeBusk, W.F., K.R. Reddy, M.S. Koch, and Y. Wang. 1994. Spatial distribution of soil nutrients in a northern Everglades marsh: Water Conservation Area 2A. Soil Sci. Soc. Am. J. 58:543-552.
DeBusk, W.F., S. Newman, and K.R. Reddy. 2001. Spatio-temporal patterns of soil phosphorus enrichment in Everglades WCA-2A. J. Environ. Qual. (30:1438).
Hartigan, J.A., Wong, M.A. 1979. A k-means clustering algorithm. Applied Statistics, 28, 100-108.
Newman S., K.R. Reddy, W.F. DeBusk, Y. Wang, G. Shih, and M.M. Fisher. 1997. Spatial distribution of soil nutrients in a Northern Everglades Marsh: Water Conservation Area 1. Soil Sci. Soc. Am. J. 61:1275-1283.
Reddy, K.R., DeBusk, W.F., Wang, Y., Newman, S. 1994a. Physico-Chemical Properties of Soils in the Water Conservationa Area 1 (WCA-1) of the Everglades.
Webster, R., Oliver, M.A. 1992. Sample adequately to estimate variograms of soil properties. Journal of Soil Science, 43, 177-192.
Webster, R., and Oliver, M.A. 2001. Geostatistics for Environmental Scientists. John Wiley & Sons, Ltd., New York.