Water Quality Methods

This page describes the methods and procedures for the Phase I and II water quality studies conducted by the South Dakota Mines and funded by the West Dakota Water District in 2013-2015.  The purpose of the study was to conduct voluntary testing of residential wells to determine baseline water quality parameters and to identify areas of concern for water problems in the crystalline aquifer of the central Black Hills in western Pennington County.  The cost of testing was covered by the project funding, and each participating homeowner received a copy of the test results for his/her residence.

Location of Study
Sampling Strategy
Sampling Procedures
Location of sampling sites
Sampling results
Area of Concern maps
Hardness IDW map

Location of Study

Phase I of the study, sampled between May and November in 2013, targeted the Hill City and Mount Rushmore 1:24,000 quadrangles in western Pennington County, SD (2013 in Figure 1).  This area was chosen because of its relation to a previous aquifer study and because 1:24,000 scale geologic maps are already published or in press for those two quadrangles, providing an understanding of the rock units from which the well water is taken.  Additional sampling extended the study area to the Pennington County boundary in the north and south, across the Rochford and Medicine Mountain quadrangles to the west, and to the limit of the PreCambrian metamorphic units to the east (2014 and 2015 in Figure 1).  

Figure 1.  Quadrangles, sampling schedule, and home sites for the project area Study Plan Map

Sampling Strategy

The density of residences in the study area is not uniform (Figure 1).  Most of the study area is part of the Black Hills National Forest, and development is constrained to private inholdings scattered through the federal lands and strongly associated with historic settlements and mining districts.  In Phase I (2013), for example, most homes clustered along the major roads within the rough triangle made by Hill City, Keystone, and Three Forks.  Moreover we had to rely on voluntary participation of homeowners.  Thus the sampling sites could not be spatially uniform nor randomly selected.  We did not sample water from residences using city-supplied water, thus excluding homes within the city limits of Hill City and Keystone.  Within these limitations we attempted to provide a broad spatial coverage by sampling every cluster of houses in the study area at least once and distributing the samples within the larger housing clusters as evenly as possible. The research team went door to door on Sunday afternoons, asking homeowners who were available whether they would like to participate in the study.  If they agreed, they signed a consent form (pdf) and the samples were taken immediately by the research team and delivered to the lab on Monday morning.

Sampling Procedures

Sampling directions provided by the testing laboratory, MidContinent Laboratories in Rapid City, SD, were followed.  The lab provided two sealed plastic bottles for each test: a 500 ml bottle for mineral samples and a 100 ml bottle for bacteria samples.

The ideal sampling source was an indoor cold water faucet with a non-rotating spigot that pulled water from the well before it went through any water softening units or other treatment systems.  Not every house had spigots which met this condition, and some accommodations had to be made on a case by case basis.  The primary goal was to be sure that the water was untreated; provided this constraint was met the sample might have been collected from a fixed, non-rotating kitchen or bathroom spigot.  In every case, the water was allowed to run for a minimum of 2 minutes prior to sampling.  The mineral sample was collected first.  A butane lighter was used to flame the faucet for a minimum of 10 seconds before the bacteria sample was collected.  Bottles were immediately sealed and labeled with the South Dakota Mines contact person, date, time, and the residence address, and the samples were delivered to the testing laboratory the following morning (Monday) to fulfill the requirement that bacteria samples be delivered within 24 hours of testing.  Samples testing positive for fecal coliform were retested within a few weeks to rule out possible contamination due to sampling methods; in all cases the positive result was confirmed.

Location of Sampling Sites

Sampling locations were also entered into an Excel spreadsheet with full information on each locality, including the owner name, contact information, sampling date, and whether the well was believed to serve single or multiple residences.
The sampling sites were plotted in a geographic information system (GIS) based on the residence address using the City of Rapid City/Pennington County RapidMap website ( URL:  http://www.rcgov.org/GIS/rapidmap.html ), which shows address points and parcel boundaries and allows the user to search for address points.  Each sampling point was placed close to the address point, which will generally be offset from the actual well location; however, this offset was considered inconsequential to the analysis results.  Each sampling location is assigned a unique identification number in the database (IDNum), and lab results, receipts, and consent forms are all indexed to the IDNum.  The sampling localities are maintained in a private, password-protected user group in an organizational subscription to ArcGIS Online, a cloud-based GIS provided by ESRI, so that all research personnel on the team have access to the same data, but it is excluded from outside access.  Only the homeowner and the South Dakota Mines research team have access to site-specific information regarding the private wells.

Sampling Results

Sampling results from the private wells were entered in a spreadsheet, indexed by IDNum and address.  In some cases, multiple tests were made from the same well and entered on separate lines in the spreadsheet so that every test value was maintained, although this created a one-to-many cardinality between the sample point locations and the results table. Test values below the detection limit for any constituent were entered as zero in the table.  All values are reported in milligrams per liter (mg/L).

Water quality data from public wells was compiled to add to the data set.  The two available sources are the National Water Information System (NWIS) maintained by the U.S. Geological Survey and public water reports maintained by the South Dakota Department of Environment and Natural Resources (SDDENR).  We found data on NWIS for 38 wells and data from SDDENR for 41 wells, with test dates ranging from 1967 to 2013.  Public water reports vary in the frequency of testing and the panel of constituents tested, in the end providing an incomplete picture of water quality over time and space.
After the Phase I, II and III samplings were completed, the public data were merged with the private well information and imported to a geodatabase standalone table file in ArcGIS Desktop (ESRI, Inc.).  

Area of Concern Maps

Our agreement with the homeowners stipulates that results from individual wells cannot be shown on maps presented to the public.  Therefore we present the results as density maps showing colored areas indicating a greater frequency of elevated values of the constituents.

The maps were constructed using both public and private well data.  Most of the private wells only have one well analysis; however, a few private wells and nearly all public wells had multiple analyses per well.  For arsenic, iron, nitrate, and sulfate, we determined the maximum constituent value measured at each well, selected the wells with values greater than or equal to 80% of the EPA limit, and used these selected points to create a density map in ArcGIS Spatial Analyst.  We used a kernel density function with a 2000 meter radius to provide smooth result that remained faithful to the local density values; no weighting based on the constituent concentration was employed, such that each well point represents merely the presence of an elevated constituent value and not its magnitude.  For total and fecal coliform, the wells were selected if at least one test of the well showed the presence of bacteria, and the density maps were calculated as previously described.

The output maps for all constituents were symbolized using the same density ranges of 0 to 1.5 wells/km2.  The value of 0.2 wells/km was selected as the threshold of the shaded areas; this value produced shaded regions that corresponded visually with clusters of elevated values, and worked for all constituents.  The shaded areas represent regions where a greater frequency, and therefore a higher risk, of elevated values is present.  It is important to understand that subsurface conditions can change rapidly from place to place, and not all wells in the shaded areas will have problems.  The spatial autocorrelation of constituent concentrations is generally low, with high values often adjacent to lower ones.  Water flow in metamorphic rocks is often strongly anisotropic and controlled by regional fractures and joint sets.  The only way to tell if a particular well has a problem is to test it.

Because the density of home sites in the study area is not uniform, one should take care in interpreting these maps.  It is true that higher densities of elevated values are often spatially associated with higher densities of home sites, where more samples were taken.  However, this phenomenon does not imply that the higher density of elevated values is the result of the higher density of sampling, but rather reflects underlying geological or anthropological factors.  For example, sulfate problems are shown to be minor and restricted in extent compared to problems with arsenic or iron, yet all three sets of data have the same sampling density.  Conversely, problems may still exist in areas with a low home densities, but might have been missed in the sampling.  The strongest conclusions may be drawn from areas with a high density of homes and no problems detected.

Hardness IDW Map

There is no EPA guideline or standard for hardness, so the hardness data could not be used to develop Area of Concern maps.  Instead, we selected all public and private wells with at least one hardness measurement (n = 120) and calculated the average hardness.  Seven wells had a hardness value of 0, which would be highly unusual and was interpreted to mean that the sampling location was, in spite of our efforts, downstream of a water softener installed in the home and therefore unreliable.  These values were excluded.  The remaining values were used to perform an Inverse Distance Weighted (IDW) interpolation.  Because the hardness values can vary significantly over short distances in response to subsurface heterogeneity, the parameters for the IDW were chosen to ensure that only the four closest measurements would be used and that no measurements further than 4 km from the interpolation point were used.  An inverse distance weight power of 3 ensured that closest neighbors were more influential in the interpolation.  The resulting map appears blocky and abrupt compared to most IDW maps, but remains more true to the local data than would occur with more typical IDW settings.