Select Page

Stormwater Ponds in Coastal South Carolina Appendix A1 – Detailed Inventory Methodology

A1.1 Pond Digitization Methods

The identification and geolocation of existing ponds for the eight coastal counties of SC (Horry, Georgetown, Charleston, Dorchester, Berkeley, Colleton, Beaufort, and Jasper) was based on United States Department of Agriculture (USDA) National Agriculture Imagery Program (NAIP) 2013-Natural Color 1 meter resolution Imagery, the most recent high-resolution imagery available for coastal SC (USDA 2013a). Heads-up digitization was conducted using a systematic grid review approach at a consistent scale of 1:3000. This approach was necessary both to identify the numerous small ponds that dominate the total population of ponds and to delineate the perimeter and open water area of each pond with a sufficient degree of precision (Fig A1.1). Five individuals conducted the digitization. To ensure consistency among digitizers, a series of selection criteria were established for determining which water bodies to include in the inventory (detailed below) as well as a coding scheme for identifying questionable selections. Questionable water bodies were marked by the digitizers in an additional layer then individually reviewed by the GIS Manager as a first pass QA/QC. Any ponds that could not be resolved in this first pass QA/QC were elevated to the primary project team (all of whom are listed as authors on this report) for a second pass QA/QC to collectively determine the outcome of questionable selections. Geodatabase creation, digitization, and subsequent spatial analyses were performed in ArcGIS 10.2.

As ponds were identified and selected for digitization, the surrounding land use was visually interpreted to first ensure that the water body met the requirements for inclusion in the inventory and then to place the digitized ponds in one of seven visually defined land-use classifications (Table A1.1). Identified water bodies that were intentionally excluded from the pond inventory were those interpreted as any of the following: 1) marsh or riverine impoundments, including former rice fields; 2) ponds or lagoons associated with wastewater treatment plants; 3) any pond or lagoon associated with industrial facilities or power plants; 4) open water within forested wetlands (swamps); and 5) small water features created for fountains or other visual aesthetics, such as those associated with miniature golf sites, or similar.

CategoryDescription
RuralPonds that are either associated with agricultural practices or associated with in rural homesteads and which have presumably been created for irrigation or fishing and other recreational uses.
ForestPonds in the middle of woods with no structures in the relative vicinity.
MiningPonds located adjacent to borrow pits or sand mining operations.
ResidentialPonds located in residential neighborhoods and not adjacent to golf course land.
GolfPonds located on golf courses with no adjacent houses.
CommercialPonds located in or adjacent to shopping areas, office complexes, government properties, school properties, or downtown urban areas.
MixedPonds adjacent to either residential or commercial properties combined with golf course, which in most cases were residential developments that included golf courses.

Table A1.1 Visually defined land-use categories used to classify ponds.  The first three categories are not considered to be associated with coastal development, while the last four classes are assumed to be the direct result of development and thus likely constructed as BMPs for stormwater control.  Examples of the development related pond categories are shown in Figure A1.2.

A1.2 Associating Ponds with Location and Landscape Attributes

The 2013 pond inventory was intersected with a series of geographic, environmental, and demographic datasets to develop a series of spatially related attributes for each pond. One objective for this effort was to attempt to create an independent data-driven pond classification scheme that would enhance and expand the subjective, visually interpreted pond classification scheme.  To incorporate surrounding land-use characteristics into this classification, a buffer area was defined around each individual pond such that the pond area comprised 8% of the total area of the buffer. Buffer metrics were then calculated based on the buffer area, excluding the pond. The 8% value used to define a pond’s buffer area was based on prior manual areal measurements made for 30 commercial and residential developments (10 each in upper coast, mid coast, and lower coast). The percentage of each development represented by the area of the stormwater ponds within the development averaged 7.91 ± 0.74 % (µ ± S.E.). To apply this to all digitized ponds, percentages were calculated based on the assumption that all ponds and their respective buffers were circular according to the equation:   where B is buffer distance from pond edge, in meters, and A is pond area in square meters. Due to variation in pond shape, actual buffered amounts differed slightly, and the average pond comprised 8.4% of the total buffer area. Results for buffer metrics are presented as averages or percentages to account for this variability between ponds. Because larger developments were associated with greater pond area, variable buffer width based on pond size was determined to be the best approach when relating the geographic, environmental, and demographic datasets. 

The buffered feature class was intersected with a series of data layers including land cover, impervious cover, soils, and elevation (Fig A1.3). Ponds were also associated with US Census population data, hydrologic unit (HUCs) watersheds, counties, and SC Department of Health and Environmental Control-Office of Ocean and Coastal Resource Management (SCDHEC-OCRM) Critical Area delineation as well as distance to major surface waters, as defined by the National Wetland Inventory. The specific data layers are described below.

Coastal Change Analysis Program (C-CAP) Southeast Region land cover data from 2010 was used to identify the land cover data near the ponds (e.g., forested, developed, wetlands, etc.) at 30 m resolution (NOAA 2013). The 21 land cover categories were simplified into eleven categories: developed high, developed medium, developed low, developed open space, agriculture, pasture (pasture/hay, grassland/herbaceous), forest (deciduous, evergreen, mixed forest, and shrub/scrub), freshwater wetland (three palustrine wetland types), estuarine wetland (three estuarine wetland types), open water, and other (unconsolidated shore, bare land, palustrine aquatic bed). For each pond, areal coverage of each land class was divided by the total area of all land classes for that buffer to yield percent coverage for each land class.

USGS National Land Cover Database – Percent Developed Imperviousness data from 2011 provided an estimate of the impervious cover surrounding each pond at 30m resolution (MRLC 2014). The Zonal Statistics tool in ArcGIS 10.2 was used to calculate the mean, minimum and maximum impervious cover values for each pond buffer. This algorithm excluded edge pixels when a majority of the pixel fell outside the buffer. Because commercial ponds are often placed along the edge of the developed parcel next to potentially unrelated land use types in an adjacent parcel, mean impervious cover statistics may underestimate the impervious cover associated with that pond. In this case, the maximum impervious cover observed may provide additional information for classification.

The USDA Soil Survey Geographic SSURGO Database provided soil drainage classes (e.g., moderately drained, poorly drained) surrounding each pond (USDA 2013b). The data were merged into two classes: poorly drained (sum of very poorly drained, somewhat poorly drained, and poorly drained classes) and well drained (sum of moderately well drained, well drained, excessively drained, and somewhat excessively drained classes). The percent coverage of each soil class was then calculated for each buffer area.

Elevation data referenced to the North American Vertical Datum of 1988 (NAVD88, feet) was associated with each pond buffer area, as well as the pond surface itself, using the most recent LIDAR-derived digital elevation model (DEM) available, which was 2007 or 2009 for all counties except Beaufort which was collected in 2013 (SCDNR 2016). To avoid computational limitations, it was necessary to resample the original 10 foot raster to a 10 m raster using mean pixel values. Mean, minimum, and maximum elevations for each buffer were calculated using zonal statistics on the set of 10 m pixels falling within a pond buffer polygon. As the flat water surface of ponds is easily distinguished and often quite different from the adjacent landscape, pond elevation was calculated at the centroid of each pond polygon as a separate metric. Because exact pond construction date is unknown, these values were calculated for all ponds that existed in 2013, and may not accurately represent ponds constructed between the LIDAR collection date and the imagery date.

To provide another metric to assess pond location within the landscape, distance to nearest surface waters was calculated. A surface waters layer was isolated from US Fish and Wildlife Service National Wetlands Inventory (NWI) data by exporting the “Estuarine and Marine Deepwater” and “Riverine” categories (USFWS 1979). This dataset delineates freshwater and marine rivers, tidal creeks, and estuaries to a minimum width of approximately 20 m (66 feet). The planar distance to the nearest feature was calculated for each pond.

Each pond was also assigned to a United States Geological Survey – Watershed Boundary Dataset – Hydrologic Unit Code (HUC) 12-digit watershed (USGS 2005), and each pond was also coded with respect to whether or not it occurs within the boundary of the SCDHEC-OCRM Critical Area (DHEC 2008), which delineates the boundaries of coastal wetland systems where SCDHEC-OCRM has direct permitting authority and has established additional regulatory criteria to protect sensitive coastal waters.

It is noted here that comparisons between mean proportions of land-use classifications within pond buffer areas and pond categorizations based on visual determination during digitization process yielded decidedly mixed results. Overall, attempts to derive a data-driven pond categorization scheme based on proportions of CCAP land-use classifications failed to perform any better than the original simple visual assessments. Similarly, combining land use classifications with other environmental, geographic, or demographic attributes of the pond buffer area (e.g., soil data, elevation data, census data etc.) also failed to yield any clear potential for categorizing ponds beyond what was achieved with simple visual assessments of surrounding land use types (data not shown) and were thus not further considered. Comparisons between visually determined pond categories and % impervious cover within pond buffer did, however, show significant differences in average % impervious surface coverage. There was also a clear tendency for the range in % impervious cover within a pond category to increase from the non-developed pond categories to the development-related pond categories, and within this latter category for the commercial ponds to have the largest mean and range in % impervious surface coverage, suggesting that the visually determined pond classification reflected meaningful differences in the degree of development surrounding the ponds.

A1.3 Determining Rate of Increase in Pond Number and Cumulative Area

An assessment of change over time in pond number and cumulative area was conducted for two pilot areas: (1) The “Grand Strand” area of Horry and Georgetown counties, and (2) the greater Charleston area that comprises portions of Charleston, Berkeley, Dorchester counties (Fig A1.4). The first pilot area is roughly the extent of greater Myrtle Beach area covering approximately 421 square miles from the Waccamaw River east toward the Atlantic Ocean. This area was chosen for a pilot study because of its relatively high rate of development during the past two decades. The second pilot area is roughly the extent of the Charleston metropolitan area, covering about 595 square miles, from the North Edisto River to just south of Bulls Bay and west into portions of Dorchester and Berkeley Counties.  This area was also chosen because of its relatively high rate of development in recent decades. To conduct the change analysis, ponds were identified on the 1994, 1999, and 2006 aerial imagery in addition to the 2013 imagery.

Figure A1.1 Examples of minimum scale that digitizing occurred.

Figure A1.2 Examples of the four development-related land use classifications.

Figure A1.3 Ponds shown with buffers on a) 2013 NAIP Aerial Imagery, b) C-CAP 2010 Land Classifications, and c) NLCD 2011 Impervious Cover.

Figure A1.4 2013 Pond inventory layer for the eight coastal counties of South Carolina with the two pilot areas used for change over time analysis outlined in red.