The use of a GIS to compare the land areas captured by very basic and complex wellhead protection area models.

Introduction

Methods of Delineating Wellhead Protection Areas

A wellhead protection area is commonly defined as the surface and subsurface area that surrounds a well field, well, or spring that supplies a public drinking-water system, and through which potential contaminants can reach the sources (Jennings, 1995). These areas are delineated and displayed on the basis of the amount of time it takes for water to make the journey from the ground surface to the well. The time-of-travel (TOT) areas are typically displayed as 1-, 5-, and 10-year land areas.

The Washington State Department of Health (WSDOH) accepts four methods for delineating wellhead protection areas around public supply wells. These methods, in ascending order of complexity and cost, are the calculated-fixed-radius (CFR) method, the analytical method, hydrogeologic mapping, and numerical flow/transport flow model.

In the past, Wellhead Protection Areas (WHPAs) were limited to paper maps found within the written plans. Maps in this format are difficult to reference in environmental health, land use planning, and technical research processes. Recent research (Miller, Chudek, & Babcock, 2003) and subsequent conference proceedings (Miller, 2004) demonstrate the capabilities of a GIS for compiling, displaying, recognizing, and comparing wellhead protection areas for a given region.

GIS and Wellhead Protection

The 1986 Amendments to the Safe Drinking Water Act set forth a program to protect groundwater resources used for public water supplies from all potential sources of contamination. The U.S. Environmental Protection Agency (U.S. EPA) national wellhead protection program was designed to limit, within the jurisdiction of wellhead protection areas, contaminants that may have an adverse health effect for humans. Wellhead protection areas are a main component from which potential contaminant assessments are addressed (U.S. EPA, 1987). More recently, in an effort to reduce disease risks from groundwater, U.S. EPA proposed requirements for sanitary surveys, hydrogeologic assessments, and disinfection (U.S. EPA 1999).

Although the spatial characteristics of GIS lend themselves to delineating WHPAs and identifying contamination sources (Vieux, Mubaraki, & Brown, 1998), Levy and Ludy (2000) note that most WHPA delineation studies lack a measure of the uncertainty associated with their model predictions and that a quantitative representation of uncertainty could be useful for regulators implementing different degrees of protection in areas with differing degrees of certainty. Forster, Lachmar, and Oliver (1997) would like to see a series of sensitivity studies with modeling projects involving WHPA delineation approximating the uncertainty, especially when the capture zones are submitted to government agencies. Kinzelbach, Vassolo, and Li (1996) have found that although different wellhead protection areas can be arrived at through model calibrations, each model calibration could provide acceptable results, mainly because of the uncertainty associated with the temporal variability and averaging techniques of the system. Most important, they feel that assessing the total uncertainty of WHPA delineation is important since regulators and managers need to make intelligent decisions when implementing source control strategies and contingency plans, and monitoring network design. Miller and co-authors (2003) found that prior to organizing a countywide wellhead protection program it was important to conduct a comparative analysis, with a GIS, of the methods used for their delineation. With this analysis they were able to better protect all public sources through interim wellhead protection areas, which are recognized through various land use control mechanisms.

[FIGURE 1 OMITTED]

[FIGURE 2 OMITTED]

Research Hypothesis

Following up on a GIS-based study of the effectiveness of the CFR method in delineating WHPAs (Miller et al., 2003), it was determined that it was important see how the CFR model compared specifically to each of the more advanced models. The 2003 study addressed the question of which land area (1-, 5-, or 10-year) the CFR more closely resembled when it was compared with results from more complex groundwater-modeling methods utilizing GIS distance, buffering, spatial query, and polygon overlay methods (Miller et al., 2003). The analysis outlined in the study reported here utilized similar methods to make an even more important distinction: a distinction as to which specific model the CFR model more closely resembles when it is simply overlaid and compared in GIS--the analytical model, the hydrogeologic model, or the numerical flow/transport model (Figure 2).

The operating hypothesis for this study was that the volumetric CFR method for delineating groundwater flow around a public supply well captures land areas increasingly similar to those found by the more complex, less linear, models--either the hydrogeologic or numerical flow/transport model. In other words, the author hypothesized that the CFR method may provide protection that more closely resembles that provided by the most complex groundwater mapping methods (hydrogeologic and numerical/transport models), as opposed to the analytical model.

Methods

Project Study Area

The project study area, Whatcom County, is located in the northwestern corner of Washington State. During the study reported here, the approximate population was about 174,500 people who largely resided in the western third of the county (this part of the country also represents the bulk of the research study area).

Existing Public Water System Wellhead Protection Area Data

The initial digital wellhead protection area data and the methods used for this study were available from a previous study (Miller et al., 2003). ArcView GIS version 3.2a software and a GeoExplorer II GPS (with Pathfinder Office software) were utilized to complete the research.

Further Overlay Data--CFR Versus Each of the More Complex Methods

The 12 wellhead protection areas (16 separate areas total) used for the 2003 study (Miller et al., 2003) were also used for the initial comparisons made in the research reported here. Each existing water system wellhead protection plan was reviewed to determine which delineation method had been used for the map. If a wellhead protection plan stated that it utilized multiple methods to create the capture areas, the most complex method utilized was used for the comparisons, unless the methods combined were clearly based on or could not be separated from a CFR. In that case, the WHPA was not utilized for this study (Figure 1).

A spreadsheet was created for each of the 1-, 5-, and 10-year overlay areas for the CFR method versus the analytical method, the hydrogeologic method, and the numerical method, on the basis of the data tables created in the 2003 study (Table 3 from the 2003 study gives an example of the initial 1-year data [Miller et al., 2003]). Table 1 displays the results of the following comparisons:

* CFR versus analytical methods -- six areas compared;

* CFR versus hydrogeologic methods -- five areas compared (Figure 3 and Figure 4 from the 2003 study display the GIS overlays for some of the CFR and hydrogeologic models [Miller et al., 2003]); and

* CFR versus numerical flow/transport methods -- five areas compared.

New Public-Water-System WHPA Data

Some more current WPHA plans and maps were on file; the author located others by contacting WHPA plan engineers, hydrogeologists, the Washington State Department of Health, Evergreen Rural Water of Washington, the Nooksack Indian Tribe, and related agencies. Twenty-four new wellhead protection plans and maps had been compiled by the spring of 2003, making a total of 41 formal plans and maps. At the conclusion of the study (fall 2003) approximately 30 percent of Group A public water systems with groundwater sources had formal wellhead protection plans--up from only 5 percent in 1999.

Of the 24 new wellhead protection areas, only eight could be used for the second phase of this research. As was found in previous research (Miller et al., 2003), some wellhead protection areas cannot be used for the GIS overlays, for a variety of reasons. Some of the more notable reasons were as follows:

* Most of the new formal wellhead protection plans utilized the CFR method since the systems that created these plans were smaller and could not afford to delineate according to a more complex model.

* Some wellhead protection plans begin with a simple CFR model then progress to an analytical or more complex method. In some cases, the CFR and the more complex model are not delineated separately; in those cases, the WHPA could not be used. The Anderson Creek wellhead protection plan stated that "due to the uncertainty involved with both of the methods used to determine wellhead protection areas [for this wellhead protection plan report] and the lack of field determined local aquifer conditions, the community has chosen to combine the results of the two methods. The community accepts a composite of the analytical and CFR WHP areas as a conservative estimate of time of travel zones." (Lane, Thomas, & Becker, 2001, p. 5).

* Occasionally, a wellhead protection plan will delineate and combine the 1-, 5-, and 10-year areas for two separate wells (that are not close together [as in a wellfield]): "Interpretation of the field data and development of a plausible conceptual model of the Isle Aire Beach Water System flow was not possible, because of the complex fractured bedrock and the lack of an immediately apparent logic in the data. Consequently, a combination of methods was used. A computer-based numerical modeling approach was adopted" (Willing, 2000, p. 4). Again, since the CFR could not be separated out of these WHPAs, it was not used for the study reported here.

* Five new separate public water systems had wellhead protection areas that were utilized for the same mapping and GIS overlay methods outlined in the 2003 study (Miller et al., 2003). These new 1-, 5-, and 10-year areas capture and overlay percentages were organized into a spread-sheet and compared according to the same overlay and averaging methods outlined in the 2003 study (Miller et al., 2003).

* Table 1 shows the initial comparisons and overall averages of the data that were used in the 2003 study.

* Table 2 gives the final averaged results when the comparison percentages for the new data are averaged in with the results of Table 1. Table 2 also gives the overall averaged results from the existing and new comparison percentages.

* Table 3 displays the final averaged time-of-travel percentages for the comparisons as a whole from the 2003 study (Miller et al., 2003) along with the final averaged time-of-travel percentages for the comparisons as they were broken down and finalized at the base of Table 2 in the study reported here. Table 3 allows the reader to view the final results of the 2003 study and the current study in one table.

Results

Initial Data Overlay Results -- Existing Data

Figure 2 from the 2003 study (Miller et al., 2003) and Figure 2 from this paper both illustrate the importance of the overall averaged results for all three categories. For example, an overlay may do very well in one area, as when a CFR captures all of a noncircular area, but very poorly in another area, as when the CFR captures a large area outside of the noncircular area. Again, the CFR result is most similar to results from a more complex method when the comparison shows the highest percentage in Area 1 (area captured), the lowest percentage in Area 2 (area not captured), and the lowest percentage in Area 3 (area overcaptured).

The following results from Table 1 are notable:

* The most similar overlays (indicated by a superscript "a" in the tables) occurred in all three comparisons of the CFR and hydrogeologic models and in all three of the one-year overlays (Area 1: 89.70 percent; Area 2: 10.30 percent; and Area 3: 40.10 percent). In other words, the CFR model matched up best with the hydrogeologic model.

* By contrast, the comparisons of the CFR and analytical overlays never produced the greatest similarities (indicated by a superscript "b" in the tables) in any of the 1-, 5-, and 10-year comparisons. In other words, results from the CFR model had the least similarity with results from the analytical model.

* The least similar overlays all occurred in the three 10-year comparisons (Area 1: 47.79 percent--10-year analytical; Area 2: 48.72 percent--10-year numerical flow/transport; Area 3: 84.57 percent--10-year numerical flow/transport). Basically, the CFR 10-year areas do not match up very well with the 10-year areas for the more complex models.

Data Overlay Results -- Existing and New Data Combined

All of the newly available WHPAs utilized for the final portion of the study were based on analytical models. No other method had been used by a water system to delineate its WHPA since the 2003 study. Table 2 displays the existing data from Table 1 averaged with the new data. The bottom of Table 2 displays the overall averaged results for all existing and new data. These overall averages show a clear distinction between each of the areas:

* The CFR and hydrogeologic-model overlays were the most similar in each of the three categories.

* The CFR and analytical-model overlays were least similar in each of the three categories.

The base results given in Table 2 and the final results given in Table 4 from the 2003 study (Miller et al., 2003) are shown in Table 3. These combined results indicate that the 1-year CFR more closely resembles results from more advanced methods of groundwater modeling in areas closer to the well (Miller et al., 2003), with the hydrogeologic method having the most similarity to the CFR method (results from the study reported here).

Discussion

The combination of the 2003 study results and the results of the study reported here has allowed the county to have even more confidence in the CFR method, especially for areas closer to the well. Local health and planning departments can provide public wells with a level of protection by simply using the CFR method, either as an interim measure or as a final model. This higher confidence level is evident in the fact that WHPAs have made their way further into local land use and emergency planning processes, as noted in the 2003 study (Miller, et al., 2003). During this second study, WHPAs were making their way even deeper into local rules and regulations--such as the Whatcom County Draft Critical Areas Ordinance and Comprehensive Plan.

Further Research

Further research is combining the GIS overlay and comparison methods outlined here and in the previous article (Miller et al., 2003) in order to compare CFR land areas with existing wellhead protection areas on a statewide level. WSDOH has been accumulating GIS data on WHPAs for several years. A comparison at this level will provide a much larger volume of data, potentially more conclusive results, and an even broader scope of varying hydrogeologic conditions.

Conclusion

It is difficult to say why the overall CFR-versus-hydrogeologic-model overlays were most similar, or why the CFR-versus-analytical-model overlays were least similar. The author's initial hypothesis--that the results from the CFR method would be most similar to results from the more advanced models--was simply based on looking at the higher models available in GIS after the 2003 study (Miller et al, 2003) and how they tend to capture land areas. The analytical models seem to all head in almost direct narrow lines as they initially move away from the wellhead, whereas the more advanced models seem to compensate for more land area closer to the wellhead in their less linear delineations.

The results of the study reported here indicate that the areas delineated by the CFR method for groundwater public drinking-water supplies most closely resemble those delineated by the more complex hydrogeologic model. Furthermore, the overlays indicate that the results from the less sophisticated analytical-model comparisons were least similar. Based on these results and the results of the 2003 study, the author feels that it would be wise for any local health department to work towards providing permanent or interim protection for public drinking-water-system wellheads by simply

1. establishing a GIS with standard base maps,

2. mapping the location of any public supply wells with a global positioning system (GPS),

3. buffering these wells with the volumetric CFR method in GIS,

4. adding any noncircular (more-advanced-model) wellhead protection areas to the GIS, and

5. working with the local water systems and any local or state agencies to recognize these wellhead protection areas in land use decision-making processes.

Acknowledgements: The author thanks Scott Babcock, Ph.D., professor of geology at Western Washington University, for his ongoing interest, his motivation in groundwater protection, and his review of the final drafts of this research.

Corresponding Author: Chris Miller, Environmental Health Specialist II, Whatcom County Health Department, 509 Girard St., Bellingham, WA 98225-4005. E-mail: [e-mail].

REFERENCES

Forster, C.B., Lachmar, T.E., & Oliver, D.S. (1997). Comparison of models for delineating wellhead protection areas in confined to semiconfined aquifers in alluvial basins. Ground Water, 35(4), 689.

Jennings, D.G. (1995). Washington state wellhead protection program guidance (Publication #331-018). Olympia, WA: Washington State Department of Health.

Kinzelbach, W.M., Vassolo, S., & Li, G.M. (1996). Determination of capture zones of wells by Monte Carlo simulation. In K. Kovar & P. van der Heijde (Eds.). ModelCARE Conference on Calibration and Reliability in Groundwater Modeling: Vol. 237. Calibration and reliability in groundwater modeling (pp. 543-550). Wallingford, UK: International Association of Hydrological Sciences Press.

Lane, D., Thomas, E., & Becker, D. (2001). Anderson Creek Water Association: Wellhead protection plan, Chapter 3: Identification of the wellhead protection areas. Ellensburg, WA: Evergreen Rural Water of Washington.

Levy, J., & Ludy, E.E. (2000). Uncertainty quantification for delineation of wellhead protection areas using the Gauss-Hermite quadrature approach. Ground Water, 38(1), 63.

Miller, C., Chudek, P., & Babcock, S. (2003). A comparison of well-head protection area delineation methods for public drinking water systems in Whatcom County, Washington. Journal of Environmental Health, 66(2), 17.

Miller, C. (2004, May). The use of a GIS to compare and recognize wellhead protection areas for public drinking water systems in Whatcom County, WA. Paper presented at the National Environmental Health Association 2004 Annual Educational Conference, Anchorage, AK.

U. S. Environmental Protection Agency. (1987). Guidelines for delineation of wellhead protection areas. Washington, DC: U.S. Environmental Protection Agency, Office of Water, Office of Ground-Water Protection.

U. S. Environmental Protection Agency (1999). Ground water rule deliberative document, November 23. Washington, DC: Author.

Vieux, E.B., Mubaraki, A.M., & Brown, D. (1998). Wellhead protection area delineation using a coupled GIS and groundwater model. Journal of Environmental Management, 54(3), 205.

Willing, P. (2000). Isle Aire Beach Water Association Wellhead Protection Program. Bellingham, WA: Water Resources Consulting, LLC.

Chris Miller, M.S., R.E.H.S.

TABLE 1 Data from the 2003 Study: Overlap of 1-, 5-, and 10-Year CFRs
with Results from Each of the More Complex Methods*

                                 Area 1:     Area 2:
                                 Percentage  Percentage    Area 3:
                                 Captured    Not Captured  Percentage
                                 (Best If    (Best If      Overcaptured
Year and Method                  High)       Low)          (Best If Low)

1-year CFR vs. analytical        74.7%       25.30%         68.3%
1-year CFR vs. hydrogeologic     89.70% (a)  10.30% (a)     40.10% (a)
1-year CFR vs. numerical         75.15%      24.84%         61.44%
5-year CFR vs. analytical        53.53%      46.46%         78.36%
5-year CFR vs. hydrogeologic     64.95%      35.04%         53.51% (a)
5-year CFR vs. numerical         71.19% (a)  28.80% (a)     74.67%
10-year CFR vs. analytical       47.79%      32.21%         67.13%
10-year CFR vs. hydrogeologic    68.44% (a)  31.55% (a)     56.73% (a)
10-year CFR vs. numerical        51.27%      48.72%         84.57%
Overall averages
  Analytical average 1-, 5-,     58.67% (b)  34.66% (b)     71.26%
    10-year
  Hydrogeologic average 1-, 5-,  74.36% (a)  25.63% (a)     50.11% (a)
    10-year
  Numerical average 1-, 5-,      65.87%      34.12%         73.56% (b)
    10-year

(a) most similar.
(b) least similar.
*The overall averages given at the bottom of this table show that for
each column, the CFR had the most similarity with results from the
hydrogeologic model and the least similarity with results from the
analytical model.

TABLE 2 Existing and New Data Combined to Display CFR vs. Analytical,
Hydrogeologic, and Numerical Average Percentages, and Overall Averages*

                                 Area 1:     Area 2:
                                 Percentage  Percentage    Area 3:
                                 Captured    Not Captured  Percentage
                                 (Best If    (Best If      Overcaptured
Year and Method                  High)       Low)          (Best If Low)

1-year CFR vs. analytical        76.3%       23.96%         69.5% (b)
1-year CFR vs. hydrogeologic     89.70% (a)  10.30% (a)     40.10% (a)
1-year CFR vs. numerical         75.15% (b)  24.84% (b)     61.44%
5-year CFR vs. analytical        55.35% (b)  74.60% (b)     81.20% (b)
5-year CFR vs. hydrogeologic     64.95%      35.04%         53.51% (a)
5-year CFR vs. numerical         71.19% (a)  28.80% (a)     74.67%
10-year CFR vs. analytical       51.78%      48.16%         88.63% (b)
10-year CFR vs. hydrogeologic    68.44% (a)  31.55% (a)     56.73% (a)
10-year CFR vs. numerical        51.27% (b)  48.72% (b)     84.57%
1-, 5-, 10-year overall
    averages
  Analytical average 1-, 5-,     61.14% (b)  48.91% (b)     79.76% (b)
    10-year
  Hydrogeologic average 1-, 5-,  74.36% (a)  25.63% (a)     50.11% (a)
    10-year
  Numerical average 1-, 5-,      65.87%      34.12%         73.56%
    10-year

(a) most similar.
(b) least similar.
*The overall averages given at the bottom of this table show that for
each column, the CFR had the most similarity with results from the
hydrogeologic model and the least similarity with results from the
analytical model.

TABLE 3 Final Averaged Percentages for the 2003 Study and the New Study

                        Area 1:         Area 2:        Area 3:
                        Percentage      Percentage     Percentage
                        Captured (Best  Not Captured   Overcaptured
Time-of-Travel Area     If High)        (Best If Low)  (Best If Low)

2003 study results
  1-year                79.52% (a)      20.48% (a)     57.35% (a)
  5-year                62.62%          37.38%         69.45%
  10-year               57.55% (b)      42.45% (b)     74.83% (b)
Results from new study (by method)
  Analytical            61.14% (b)      48.90% (b)     79.76% (b)
  Hydrogeologic         74.36% (a)      25.63% (a)     50.11% (a)
  Numerical             65.87%          34.12%         73.56%

(a) most similar. (b) least similar.