CONTENTS Acknowledgements Executive Summary

" Language skills will be important in achieving European policy goals, particularly against a background of increasing global competition " A new framework strategy for multilingualism


iii Executive Summary
In this study, a screening level cost assessment was conducted to evaluate the national cost implications of five potential regulatory levels for perchlorate in drinking water -4, 6, 12, 18, and 24 µg/L. Initially, the Unregulated Contaminants Monitoring Rule (UCMR) database was reviewed to identify all source waters/entry points contaminated with perchlorate. Ninetieth percentile and median perchlorate concentrations were calculated for each contaminated source. Capital and operating costs to treat the contaminated sources were determined for each potential maximum contaminant level (MCL).
Results from the study reveal that perchlorate contamination is national in scope, with detectable perchlorate concentrations observed in source waters in 26 different states. However, only 4.1% of all source waters sampled under the UCMR exhibited detectable levels of perchlorate. Further, most perchlorate detections were at concentrations ranging from 4 to 12 µg/L, indicating that only a very few PWSs would be required to treat for perchlorate at higher regulatory levels. Although perchlorate contamination was detected in a number of different states, perchlorate occurrence is geographically focused in southern California, the southern states, and the northeast. Approximately one third of the PWSs affected are located in California.
Capital and operations and maintenance (O&M) costs to treat source waters contaminated with perchlorate were estimated based on the assumption that all contaminated sources would be treated (i.e., no blending or abandonment of sources) and that single pass ion exchange would be implemented for treatment at all sites. Full scale capital and O&M cost information was used to develop cost curves as a function of system size. All contaminated sources were then assigned capital and O&M costs based on the estimated design and average flow rates, respectively, for each site.
Figure ES-1 displays the total annualized national compliance costs associated with each potential regulatory level based on 20 years life of service at a 3% discount rate, using both the calculated median and the 90 th percentile perchlorate concentrations for each source. At the most stringent potential MCL evaluated (4 µg/L), the national compliance cost is estimated to be $140 million per year using the 90 th percentile and a 3% discount rate.
The national compliance cost to meet a 4 µg/L perchlorate MCL is smaller than estimated compliance costs for other drinking water regulations (e.g., $585 million/year for the Arsenic Rule at 10 µg/L; USEPA, 2001a). The relatively low national compliance costs

Introduction and Background
Perchlorate is a persistent, inorganic anion known to disrupt thyroid function if ingested in significant quantity. Perchlorate salts have been used in a number of applications, including as an oxidizer in solid rocket fuel and as a component in fireworks and other explosives. Perchlorate has also been found as a contaminant in Chilean fertilizer and it has been used in some medical and analytical applications. The U.S. Environmental Protection Agency (USEPA) has identified over 100 potential perchlorate releases from governmental and non-governmental sites in 26 states, mostly associated with the use of perchlorate in solid rocket fuel (USEPA, 2003). USEPA is required under the Safe Drinking Water Act (SDWA) to make regulatory determinations on a five-year cycle for contaminants included on the CCL. In May 2007, USEPA determined that insufficient information was available to make a decision of whether or not to regulate perchlorate, primarily due to the lack of complete information on perchlorate in food as opposed to water (USEPA, 2007). Since then, data from the Food and Drug Adminstration's (FDA's) Total Diet Study has been published (Murray et al, 2008). USEPA has also re-stated its intention to complete a regulatory determination for perchlorate by the end of 2008 (Grumbles, 2008). USEPA may be pressured by Congress via the Solis Bill (H.R. 1747) to establish a NPDWR for perchlorate if the currently proposed bill is enacted into law.
USEPA takes into account a number of factors when making a determination whether or not to regulate a drinking water contaminant, including the health impacts from exposure to the contaminant, the number of people impacted, the degree of contaminant occurrence, and whether or not a national drinking water regulation would provide an opportunity for significant risk reduction as required by the SDWA. The American

Section 1 Introduction and Background
American Water Works Association National Cost Implications of a Potential Perchlorate Regulation 1571028 2 Water Works Association (AWWA) has recognized the lack of available information on projected national costs associated with treatment of perchlorate to various levels. Based on this recognized data gap, AWWA requested that Malcolm Pirnie conduct a study to estimate the national cost implications of setting a federal maximum contaminant level (MCL) for perchlorate at different levels between 4 and 24 µg/L. This report presents the results from the study. The approach followed to identify contaminated sources and to assign costs to treat water from those sources is presented in Section 2. Results from the analysis of perchlorate occurrence and the cost evaluation are presented in Section 3. Section 4 provides a discussion of trends and implications of those costs. Potential limitations in the methods used to identify the treatment costs are also presented in Section 4.

1.
American Water Works Association National Cost Implications of a Potential Perchlorate Regulation 1571028 3

Approach
Five potential perchlorate MCLs were evaluated -4, 6, 12, 18, and 24 µg/L. The lower end of the range is based on the 4 µg/L detection limit for reporting (DLR) associated with EPA Method 314 at the time that UCMR samples were analyzed. Although many laboratories are now able to measure perchlorate to concentrations as low as 0.5 µg/L, the sensitivity of the analytical method at the time samples were collected for the UCMR only allowed detection to a concentration of 4 µg/L or greater. The upper end of the range is based on the 24.5 µg/L concentration associated with the previously discussed, EPA-adopted Reference Dose (RfD) of 0.007 milligram perchlorate per kilogram body weight (USEPA, 2005b).
Table 2-1 presents the general approach used to estimate national costs associated with treating source waters containing perchlorate at concentrations exceeding each proposed MCL. The approach was developed based on the guidelines described in Raucher et al (1995) for estimating the cost of compliance with drinking water standards. Various aspects of the economic analysis conducted for the Arsenic Rule (US EPA, 2000) and the Stage 2 Disinfectants and Disinfection By-Product Rule (DDBPR; US EPA 2005c) were also used as examples.

. Steps to Identify the Compliance Costs for a Given Regulatory Level
Step 1 Identify source waters and public water systems (PWSs) contaminated with perchlorate Step 2 Determine perchlorate concentration and flow rate for each contaminated source Step 3 Identify a likely treatment strategy for the contaminated sources Step 4 Assign capital costs associated with treating each contaminated source Step 5 Assign operations and maintenance (O&M) costs associated with treating each contaminated source Step 6 Tally capital and O&M costs to treat each contaminated source with a perchlorate concentration exceeding a given value (e.g., 4, 6, 12, 18, or 24 µg/L)

Occurrence Data
As a first step, public water systems (PWSs) with detectable concentrations of perchlorate were identified using the Unregulated Contaminant Monitoring Rule (UCMR) database. A relatively complete set of occurrence data is available for perchlorate as a result of the UCMR data collection effort. However, the data set does have some limitations which are discussed in this report. Several states also have A randomly selected sample of 800 CWSs and NTNCWSs serving less than 10,000 people (small systems) were also assessed for perchlorate contamination (US EPA, 2001c). The small systems were required to monitor all entry points to their distribution system once during one year between January 1, 2001 and December 31, 2003.
The UCMR database was queried for all entry points/source waters with a detectable perchlorate concentration. As mentioned, multiple samples were collected from each large system sample point during a 12-month period. To obtain a single perchlorate concentration associated with each sample point, non-detects were assigned zero values and the 90 th percentile and median values were calculated for the given sample point. Both 90 th percentile and median values were assessed to obtain a range of the expected extent of perchlorate contamination. 1 Several additional data processing steps were required to enable assignment of perchlorate treatment costs for each contaminated source water. Specifically a design and average daily flow needed to be identified for each source water in order to estimate the capital and operations and maintenance (O&M) costs, respectively, to treat the water. Flow rates were first calculated for each PWS with a perchlorate detection using the following regression equations developed by US EPA (2005c). 1 Both 90 th percentile and median perchlorate concentrations were calculated for a given source water/point of entry. In the absence of information on the history of the sample points, it is difficult to ascertain whether the median or 90 th percentile values is more representative. As an example, one contaminated sample point that was sampled two times over the course of one year had one detectable perchlorate concentration of 11.9 µg/L and one non-detect. The collection of two samples suggests the sample point correlates to a groundwater system. If a well was taken offline after the perchlorate "hit" was detected in the first sample, the subsequent sample may not be representative because water drawn from an inactive well may likely have different quality than water drawn from a well that is continuously pumping. On the other hand, the first sample could have exhibited measurable perchlorate concentrations due to analytical error. In the first case, the 90 th percentile value may be more representative; in the latter case, the median value may be more accurate. Since it is impossible to discern which scenario is more likely without additional information for each of the 387 sample points, the occurrence and cost results are reported for both the median and 90 th percentile values to provide a likely range of the expected number of contaminated PWSs and associated treatment costs. where X is the population for the associated PWS.
Recent population data for each PWS was retrieved from US EPA's SDWIS database and cross-checked with the population size assignment for the PWS in the UCMR database. Design and average daily flow rates were then estimated for each contaminated source water by dividing the PWS flow rates by the total number of sources in the PWS under consideration. The number of sources for each PWS were tallied based on the total number of sampling points included for that system during the UCMR sampling effort.

Treatment Strategy
Several treatment technologies are available for perchlorate removal -regenerable ion exchange, single pass ion exchange, biological treatment through fixed or fluidized bed reactors, and reverse osmosis. Table 2-2 lists advantages and disadvantages associated with each treatment technology and Appendix B provides additional information on each of the technologies. Regenerable ion exchange and reverse osmosis are more costly than single pass ion exchange and generate a waste brine stream, creating disposal issues. The effectiveness of biological treatment has been demonstrated; however, due to potential public acceptance issues and additional post-treatment costs to meet SWTR requirements, no water utilities have adopted biological treatment for perchlorate removal in the U.S. to date. Based on the advantages and disadvantages mentioned above and also based on current trends in treatment selection in southern California, single pass ion exchange treatment was considered to be the reasonable treatment technology choice for the purpose of this national cost evaluation.
The costs associated with treating each contaminated source water were estimated by assuming that each source would be treated; blending and source abandonment were not considered as potential contamination abatement options. Capital and O&M costs were then assigned assuming that all contaminated sources would use single pass ion exchange systems for perchlorate removal.

Cost Data
The following paragraphs describe Steps 4 -6 in Table 2-1: assignment of capital and O&M costs for each contaminated source and calculation of the total resulting treatment costs. Capital and O&M costs were assumed to be independent of the influent perchlorate concentration based on experience that system size and operation (and associated capital and O&M costs) are primarily dictated by plant capacity and concentrations of competing anions (e.g., nitrate). Additionally, the costs were calculated based on the assumption that the treatment goal would be non-detect since all but one queried utilities with single pass ion exchange systems for perchlorate removal treat to non-detect. This approach neglects the option of treating partial flow and blending to meet the MCL. A detailed list of assumptions made to designate capital and O&M treatment costs for each contaminated source water is provided in Appendix C.

Capital Costs
Capital costs to install single pass ion exchange systems were obtained from seven different water utilities in southern California. The ENR Cost Indices for Los Angeles were used to adjust capital costs to 2008 dollars for systems installed in previous years. Los Angeles Cost Indices were used since more of the impacted utilities are located in California than any other individual state and since the baseline cost data was obtained from California utilities that have already installed perchlorate treatment. Use of Los Angeles Cost Indices (and baseline cost data) is considered a conservative approach due to the generally higher construction costs in southern California relative to other parts of the nation.
Baseline capital costs for each of the participating utilities are depicted in Figure 2-1 and include the first fill of resin, ion exchange vessels, foundation and site work, installation of the vessels and resin, electrical, process controls, and engineering services. Cost data from a perchlorate cost study conducted by Kennedy/Jenks Consultants (2004) is also included in the graph.

Figure 2-1. Baseline Capital Costs as a Function of System Flow Rate
As indicated by the R-squared value for the linear trendline depicted in the graph, the correlation between the full-scale baseline capital costs and system flow rate shows a very consistent trend. The linear regression equation was therefore used to assign baseline capital costs for all identified source waters/entry points from the UCMR database with a perchlorate detection. The minimum and maximum flow rates for the contaminated source waters identified in the UCMR database -3 to 9,320 gpmgenerally fall within the range of full-scale data obtained for this study.
Some water systems are expected to incur additional costs for installation of perchlorate treatment. For example, additional land may be required to accommodate the single pass ion exchange vessels, as demonstrated by a water utility in southern California that is currently procuring land to enable installation of their single pass ion exchange system. Several recent perchlorate treatment installations have included pre-filtration to protect the resin from clogging with suspended solids in the source water. Acid addition may also be added to protect against scaling. Additionally, walls or buildings may be required for aesthetic purposes at some facilities, building for process controls, etc. may be required for sites that currently do not have any treatment installed, and piping may be required for systems that are one mile or more from the well due to space limitations.
The following approach was used to assign these additional costs to a portion of the identified contaminated source waters: 1. Land and demolition costs were added to 35% of the contaminated sources. This percentage (35%) was selected based on trends observed in sourthern California (e.g., one of three utilities in the San Gabriel Balley installing single pass ion exchange required additional land purchase). 2 A 7,800 gpm system and above is assumed to require three 0.2-acre parcels of land; a 2,500 to 5,000 gpm system is assumed to require two 0.2-acre parcels of land; and a 2,500 gpm system or smaller is assumed to require one 0.2-acre parcel of land. The median price for a 0.2-acre parcel of land is assumed to be $220,000 based on the National Association of Realtors median home prices in the U.S. between 2005 and 2008. Demolition costs are assumed to be $50,000 per lot.
2. Pre-treatment costs were added to 40% of the contaminated sources at 20% of the baseline capital costs.
3. Wall/building/piping costs were added to 50% of the contaminated sources at 15% of the baseline capital costs.
For each additional cost category, the Excel random number generator function was used to randomly assign the additional costs for land/demolition, pre-treatment, and wall/building/piping to the designated percentages of sources. Total capital costs for each contaminated source were then calculated by summing the baseline capital costs and any additional costs associated with land requirements, pre-treatment, etc.
Several PWSs in southern California have already installed (or are currently installing) single pass ion exchange systems to treat their contaminated source waters in compliance with the California Code of Regulations Title 22 perchlorate standard of 6 µg/L. Capital and O&M costs for these systems (approximately 6 out of 387 sources with perchlorate detections) were considered in estimating the national compliance costs.

Operations and Maintenance Costs
Operations and maintenance costs were obtained from five PWSs that currently operate (or previously operated) single pass ion exchange systems for perchlorate removal. Figure 2-2 shows the O&M costs (as cost per 1,000 gallons of water treated) for the fullscale single pass ion exchange systems by system flow rate. O&M costs that were estimated in a previous study (MP, 2008) for one 2,500 gpm single pass ion exchange treatment system and two 7,800 gpm systems are also included in the graph.

Figure 2-2. Operations and Maintenance Costs as a Function of System Flow Rate
A best fit polynomial was calculated using the Excel trendline function to correlate known operating costs to system flow rate. The polynomial equation was then used to estimate O&M costs for each contaminated source water/entry point identified in the UCMR database. Due to limitations in the data and the ability to fit a trendline enabling accurate estimates of O&M costs for all system flow rates under consideration, a lower bound cost of $0.33 per 1,000 gallons water treated was established. This assumption is reasonable given that economies of scale do not completely apply to larger flow rates as more vessels are simply added to increase the treatment capacity.

Total Costs
After assigning capital and O&M costs for each contaminated source water, the costs were tallied to identify total national costs for perchlorate treatment to meet a given MCL. The following steps were followed to tally the total costs. (1) y = -9e-13x 3 + 2e-08x 2 -0.0002x + 0.7139 (2) y > $0.33/1,000 gal

Ground Truthing
Ground truthing of several parameters was conducted for quality assurance/control. Specifically, the following parameters were checked for a subset of the contaminated water sources and PWSs to verify the data: • number of sources for a given PWS, • population size for a given PWS, • estimated design and average flow rate for a given contaminated source water, and • estimated perchlorate concentration for all contaminated sources with design flow rates greater than 10,000 gpm.
For most of the listed parameters, the values were checked for PWSs that the project team was familiar with through previous work. However in some cases, the PWSs were contacted directly to verify the estimated values. For example, estimated perchlorate concentrations for all contaminated sources with design flow rates greater than 10,000 gpm were verified by contacting the PWS to inquire about the validity of the UCMR data.

1.
American Water Works Association National Cost Implications of a Potential Perchlorate Regulation 1571028 11 3. Results

Perchlorate Occurrence
The UCMR database was queried to identify all sample points with perchlorate detections to determine the distribution of PWSs that would be affected by a perchlorate MCL.   12 occurrence in small PWSs. Therefore, when compared to the available data for small systems, large PWSs offer the best opportunity to estimate national economic impacts, since the data set for large PWSs is more complete.  (2007) also observed a higher percent (15%) of source waters throughout California with perchlorate detections, compared to the estimated 2.6% of source water/entry points sampled nationwide; however, the California data set analyzed in that study is biased towards source waters expected to be at high risk for perchlorate contamination. The majority of the remaining contaminated PWSs identified from the UCMR database are distributed across Nevada and Arizona, south central U.S. (e.g., Texas, Louisiana), the southeast (e.g., Florida, Georgia, and North Carolina) and the northeast (Figure 3-2). Even at the most stringent MCL evaluated (i.e., 4 µg/L), the percent of PWSs expected to be affected is relatively low; only 2.2 to 3.4% are estimated to require perchlorate treatment based on the calculated median and 90 th percentile perchlorate concentrations, respectively. This estimate suggests that the costs for PWSs to comply with a potential perchlorate regulation may also be relatively low.
The calculated percentages of PWSs affected for a given perchlorate MCL shown in

Capital Costs
Total capital costs were estimated for each potential perchlorate MCL based on the calculated 90 th percentile and median perchlorate concentrations in all contaminated sources (Figure 3-4). As expected, capital costs are higher if the 90 th percentile perchlorate concentration is used. While a higher influent perchlorate concentration is not expected to affect capital costs to install a treatment system, a greater number of contaminated sources are estimated for a given perchlorate MCL if the 90 th percentile value is considered.
Generally, capital costs for perchlorate treatment are estimated to be low when compared to other regulations, even at the most stringent regulatory level evaluated. As a comparison, estimated capital costs to treat for arsenic, another inorganic anion known to contaminate some source waters, were estimated to be $4. of systems are carrying this cost burden and the cost impacts to an individual system installing perchlorate treatment would likely be significant, particularly when O&M costs are also taken into account.  Baseline capital costs to install this 5,000 gpm system include the ion exchange vessels, first fill of resin, foundation, piping and valves, telemetry, electrical, and engineering services, for a total of $2.8M in 2008 dollars. As the picture illustrates, the vessels can be installed uncovered in regions with moderate temperatures, cutting down costs.

O&M Costs
Annual operation and maintenance costs for a 5,000 gpm single pass ion exchange system are listed in Table 3-3. As indicated by the data, the major annual costs to operate the system are for power, labor, resin replacement, and contractor labor, with resin replacement costs being the most expensive line item.  Table 3-4 lists total O&M costs for the 5,000 gpm system discussed above (Cal Domestic, Figure 3-5) and the seven additional PWSs plotted in Figure 2-2. Several trends in the data warrant further discussion. The higher O&M costs per 1,000 gallons of water treated for the smaller systems (< 1,000 gpm) reflect economies of scale (e.g., a minimum level of staff hours are required despite system size). Additionally, several of the queried small water systems have entered into lease agreements with vendors of the single pass ion exchange systems. O&M costs for these systems tend to be higher due to the type of contract and service fees to the vendors for oversight and maintenance of the systems. The difference in estimated O&M costs for the two 7,800 gpm treatment systems ($0.31 and $0.40 per 1,000 gallons) reflects the differences in water quality between the two facilities. Nitrate concentrations in the source water for the Valley County Water District (Valley County) ranges between 11 and 13 mg/L NO 3 -N (mg/L as nitrogen) compared with an average nitrate concentration of 7.3 mg/L NO 3 -N for the San Gabriel Valley Water Company (San Gabriel) B6 plant. Despite higher selectivity of the perchlorateselective resins for perchlorate removal, nitrate in the water can significantly reduce resin capacity due to orders of magnitude higher concentrations of nitrate than perchlorate (mg/L of nitrate as opposed to µg/L of perchlorate). The higher estimated operating cost ($0.40/1,000 gallons) at Valley County (compared to $0.31/1,000 gallons at San Gabriel's B6 plant) reflects the impact of nitrate co-occurrence on resin and total O&M costs. These factors known to influence O&M costs were inherently included in the nation-wide compliance cost estimates since the O&M cost curve (Figure 2-2) was developed using cost data for PWSs covering a range of different operations agreements (e.g., leasing versus operation by utility staff) and water quality.
Figure 3-6 shows the nationwide costs to operate single pass ion exchange treatment systems for perchlorate treatment for 20 years at a 3% discount rate. The costs shown are based on the 90 th percentile perchlorate concentrations. Capital costs are also included in the graph as a reference point. As the data shows, O&M costs account for a larger portion of the total net present value (NPV) costs to treat for perchlorate than the capital costs. Similar to liquid-phase granular active carbon (GAC), the cost to operate single pass ion exchange is high relative to the capital costs since a significant component of the system must be replaced on a continual basis (i.e., the resin). O&M costs also continue in perpetuity. Although

Total Costs
The total cost of compliance for an MCL of 4 µg/L is estimated to be $2.1 billion dollars ($0.85 billion in capital and $1.28 billion total NPV in operating costs) based on the 90 th percentile perchlorate concentrations and operation of the systems for 20 years at a 3% discount rate (Table 3-5). In comparison, the estimated compliance cost for an MCL of 24 µg/L is much lower at approximately $0.1 billion or 4% of the cost at the most stringent MCL evaluated (4 µg/L). The significantly lower cost for the higher perchlorate concentration reflects the small number of PWSs that would be affected at that regulatory level (Figure 3-3). Capital and O&M costs to remove perchlorate from contaminated sources for large PWSs account for a greater percent of the nationwide compliance costs associated with each potential perchlorate MCL (Figure 3-7). The higher portion of costs for large PWS compliance is attributed to several factors: (1) the higher percentage of large PWSs with perchlorate contamination (Table 3-2); (2) the higher capital costs for a given system to meet the higher design flow; and, (3) higher operating costs associated with the greater quantity of water requiring treatment for a given system.

Discussion
As with any attempt to assess the national costs associated with a potential drinking water regulation, the accuracy of the cost estimate is dependent on the information available to develop those costs (e.g., contaminant occurrence data, PWS size and type, capital and O&M costs for a given treatment process, etc.). Inevitably, assumptions must be made due to the magnitude of the studies (i.e., contaminated sources cannot be evaluated on a case-by-case basis) and the likely absence of data required for precise evaluation of costs.
The following paragraphs discuss these limitations and the potential ramifications they have on the accuracy of the compliance costs cited in the previous Section.

Limitations in Data Sources
A number of limitations in the UCMR data were identified during the assessment of perchlorate occurrence: • The UCMR sampling effort was collected when the analytical limit for perchlorate was 4 µg/L. 3 All samples with perchlorate concentrations below 4 µg/L are therefore reported as non-detect in the UCMR database. All non-detect samples were assigned zero values in order to analyze the data. Therefore, any samples with perchlorate concentrations less than 4 µg/L (but above 0 µg/L) are misrepresented due to the analytical limitations. However, the low concentrations should not have significantly affected the calculated 90 th percentile and median values for a given source.
If EPA sets the perchlorate MCL below 4 µg/L, this limitation in the UCMR data could significantly impact calculation of the national compliance costs. However, at the potential perchlorate regulatory levels evaluated for this study, the calculated compliance costs are not expected to be significantly affected by this limitation in the UCMR data.
• Only 800 small PWSs (CWSs and NTNCWSs serving less than 10,000 people) were sampled. The compliance costs were scaled up to account for the limited sample set. While USEPA and the State regulatory agencies carefully selected the 800 small PWSs to provide a representative distribution of samples, only approximate costs can be determined in the absence of a full data set. The 800 PWSs sampled account for less than 2% of the total number of CWSs and NTNCWSs throughout the United States. Nevertheless, the greater importance of large than small PWSs (Table 3-2) is promising from the standpoint of estimating realistic compliance costs since the data set for large PWSs is more complete.
• All systems using surface water for a portion of their water are classified as "surface water" systems in the UCMR database. The US EPA SDWIS database also characterizes systems under any influence of surface water as "surface water" systems. A number of the contaminated water sources identified from a review of the UCMR database are classified as surface water, but are actually groundwater. Since the design and average flow rates for each contaminated source water were calculated using the US EPA (2005) regression equations for surface and ground water systems, any discrepancies in the source water classification may have resulted in slight inaccuracies in the estimated flow rates. However, for this order of magnitude cost evaluation, the effect of this limitation is expected to be relatively small.
• Perchlorate was not detected during UCMR sampling for several PWSs known to be contaminated with perchlorate. For example, one well operated by the California Domestic Water Company is known to be contaminated with perchlorate. Since perchlorate treatment is already installed at this site and the costs to operate the perchlorate treatment system are covered primarily by the PRPs, the actual impact on nationwide costs to treat water from this well is negligible. However, the omission in the UCMR database suggests potential limitations in the database.
A comparison of contaminated source waters in California identified through the UCMR sampling effort and statewide sampling further reveals potential omissions in the UCMR database. Approximately twice as many of the source waters detected during state-wide sampling in California do not appear in the list of contaminated sources from the UCMR database. Several factors may be attributed to the discrepancy between the state and nationwide datasets. First, some (or all) of the contaminated sources in the California data set that are not in the UCMR list may be small PWSs that were not included in the UCMR sampling effort. Second, the California data set includes some samples for which perchlorate concentrations below 4 µg/L were reported. These perchlorate detections would not have appeared in the UCMR database. Third, the California data set (1997 -2003) includes samples collected prior to the UCMR sampling effort (2001 -2003). Although unlikely, it is possible that source waters sampled prior to the UCMR sampling effort were remediated, reducing concentrations below 4 µg/L before samples were collected for analysis under UCMR. It is also possible that some wells were taken out of service. Assuming the number of contaminated water sources detected during UCMR sampling is off by 50% (i.e., the discrepancy observed in the California data is consistent nationwide), then the estimated compliance costs (Table 3-3) should be increased by a factor of two. Although more accurate costs are desirable, this analysis of the UCMR data limitation still suggests that the costs are accurate within an order of magnitude.

Limitations in Assumptions and Approach
Limitations in the approach used and assumptions made to estimate the national compliance costs could also result in inaccuracies in the cited costs. The approach and assumptions were developed using best engineering judgment; however, two limitations in the scope warrant further discussion.
Monitoring Costs. Source water monitoring costs associated with a federal regulatory determination were not included in the cost evaluation. If USEPA establishes an MCL for perchlorate, the regulation will most certainly require that PWSs conduct an initial round of monitoring to determine if their source waters are contaminated. Subsequently, PWSs may be required to monitor on an annual or triennial basis. For example, the Arsenic Rule required that surface water systems monitor annually for arsenic contamination; ground water systems are required to collect triennial samples. CDPH estimated monitoring costs associated with their determination to regulate perchlorate at a 6 µg/L MCL (CDPH, 2007). The estimated annual monitoring costs were 2% of the total annualized treatment costs (capital and O&M). Assuming a similar proportioning of monitoring to treatment costs at a national level, the omission of monitoring costs in this study is not expected to significantly affect the accuracy of the calculated compliance costs.
Effect of Nitrate on Perchlorate Treatment Costs. The presence of nitrate is known to substantially impact resin capacity. The effect of higher source water nitrate concentrations and anion loading on O&M costs were inherently considered via the distribution of water qualities for the systems considered in the cost analysis. However, depending on the nationwide co-occurrence of nitrate with perchlorate, the operating costs for perchlorate treatment may be underestimated in this study. Kimbrough and Parekh (2007) observed a weak, but statistically significant, correlation between perchlorate and nitrate occurrence in California water sources. To more accurately account for the effect of nitrate co-occurrence on perchlorate treatment costs, it may be beneficial in subsequent studies to evaluate the distribution of nitrate contamination in the United States. The nationwide distribution of nitrate in surface and ground water could then be used to make reasonable assumptions on nitrate concentrations in source waters contaminated with perchlorate. The effect of nitrate on resin capacity and associated annual operating costs could then be incorporated in the compliance costs by assuming a

Ground Truthing
Several parameters with significant importance in the determination of compliance costs were evaluated for accuracy via an assessment of those parameters for a subset of PWSs for which reasonable information was available. Specifically, values for the following parameters were checked for accuracy for a subset of the contaminated PWSs or contaminated source waters: • number of sources for a given PWS, • population size for a given PWS, • estimated design and average flow rate for a given contaminated source water, and • estimated perchlorate concentration for all contaminated sources with design flow rates greater than 10,000 gpm.
The estimated number of source waters for several of the PWSs with perchlorate detections were in the hundreds, whereas the nationwide median number of entry points per ground water system ranges from 1 to 9 depending on systems size (US EPA, 1999). The majority of surface water systems report only one or two entry points PWS. Based on the number of different sampling points incorporated in the UCMR database for the City of Tucson, that water utility has over 148 different water sources. Suffolk County Water Authority was estimated to have 502 different water sources/entry points and the Coachella Valley Water District was estimated to have 89 different water sources. Based on previous experience with these PWSs, all three serve groundwater from at least as many wells as indicated based on the review of the UCMR data. The number of source waters estimated to serve the remaining 157 affected PWSs were within range of expected values. Further, at lower estimated numbers of source waters for a given PWS, a slight discrepancy in the number of sources would not significantly impact the calculated flow rate for each contaminated source for that utility.
Population data for each PWS was obtained from the US EPA SDWIS database. All population data obtained from SDWIS was checked to ensure the numbers fell within the range of expected values based on the cited PWS size in the UCMR database (i.e., the population for a large PWS should fall between 10,001 and 100,000 people). No discrepancies were found.
The estimated design and average flow rates were evaluated for a subset of PWSs for which flow rate data was available. the wells are lower, ranging from 290 to 2,110 and encompass the estimated average flow rate calculated for this cost study. Therefore, the estimated flow rates calculated from the USEPA regression equations (2005) relatively accurately predicted known flow rates for source waters for the Cal Water Dominguez water system. Similar assessments were conducted for other systems with known flow rates.
The PWSs for all contaminated sources with design flow rates greater than 10,000 gpm were contacted to inquire about the validity of the estimated perchlorate concentration. Detailed evaluation of these PWSs was conducted since the costs associated with treating large contaminated sources could significantly skew the national cost estimate. Table 4-1 lists the PWSs for which one or more contaminated sources were identified with a design flow rate greater than 10,000 gpm. All listed PWSs are classified as very large utilities that serve more than 100,000 people.

Section 4 Discussion
American Water Works Association National Cost Implications of a Potential Perchlorate Regulation 1571028 28 was included in the national cost estimates for potential perchlorate MCLs of 4 and 6 µg/L. Based on the perchlorate data, the utility would need to reduce perchlorate concentrations at those regulatory levels if perchlorate concentrations in their source water remain at current levels. The CCR data was used rather than UCMR measurements to assign perchlorate concentrations for this site. Specifically, a maximum concentration of 9.7 µg/L is used in place of the 90 th percentile value. The 7.1 µg/L average concentration is used in place of the median value. The more current perchlorate measurements are expected to be more representative to current levels in the City of Henderson source water intake than the measurements collected in the early 2000s for the UCMR sampling effort.
The authors were unable to reach Montgomery Water Works. Therefore, the two contaminated sources identified from the UCMR database for this PWS were included in the estimated national compliance costs.
Suburban Water System has a well that contains perchlorate at concentrations between 9 and 10 µg/L in their San Jose system. 5 They blend water from this well with other source water to achieve perchlorate concentrations below the California MCL of 6 µg/L. This system is not included in the cost assessment; it is assumed that Suburban will continue the blending scenario regardless of the establishment of a national perchlorate MCL.
Metropolitan Water District of Southern California was not included in the cost assessment. Recently measured perchlorate concentrations in Colorado River Water are below 2 µg/L due to the success of upstream remediation efforts. The City of Yuma was also excluded from the cost assessment since it is also served by Colorado River water.
The detectable perchlorate concentration (13.8 µg/L) measured in one of eight samples collected at an entry point for the City of High Point, North Carolina was confirmed by the contract laboratory as a false positive. 6 Bill Frazier, a lab supervisor at the City of High Point, North Carolina, indicated that subsequent source water sampling conducted by the U.S. Geological Survey and the City also confirms the absence of perchlorate in their source water. Therefore, the City of High Point was not included in the cost assessment.
During UCMR sampling, one sample collected for the Manatee County Utilities Operations Department had a detectable concentration of perchlorate. However, the other three quarterly samples were collected by a different lab and were non-detect. Mark Simpson at the Manatee County Public Works Department believes that the positive hit is attributable to analytical errors. For all other UCMR sampling for their

Section 4 Discussion
American Water Works Association National Cost Implications of a Potential Perchlorate Regulation 1571028 29 system, perchlorate concentrations were non-detect. Additionally, all subsequent sampling also indicated an absence of perchlorate in the water. Therefore, Manatee County was not included in the cost assessment.
The City of Midland, TX had perchlorate contamination in a 5 MGD well field that was used to provide water during peak summer demands. The well field was abandoned five years ago after the perchlorate contamination was discovered. The well field is mostly dry; prior to five years ago, they would inject water from a more distant groundwater source into the 5 MGD well field during the winter. The stored water in the 5 MGD well field would then be withdrawn to meet peak summer demands for the City of Midland. This system was not included in the cost assessment.

Comparison to Other Cost Studies
The total annualized national compliance costs shown in Figure 4-1 were compared to costs developed in two previous studies -Kennedy/Jenks (2004) and CDPH (2007). The two previous studies estimated total perchlorate treatment costs for utilities in California to respond to a state regulation. Based on the trends shown in Figure 3-1, the estimated national compliance costs should be approximately three times the calculated California costs, assuming that all three studies produced fairly accurate cost information. . The perchlorate occurrence data was then used to estimate national costs to treat contaminated water sources to meet five potential regulatory levels. The following conclusions can be made from the evaluation: • Only 4.1% of all PWSs sampled under the UCMR had detectable levels of perchlorate in one or more of their source waters/entry points. Further, measured perchlorate concentrations at most locations were relatively low (12 µg/L or less). • Only 3.4% of PWSs would be affected by a perchlorate MCL of 4 µg/L; less than 1% of PWSs would be required to treat their water at an MCL of 24 µg/L. • While perchlorate contamination has been detected in source waters in 26 different states, one third of the PWSs affected are located in California. Most of the affected PWSs in California are already required to treat to remove perchlorate to meet the 6 µg/L MCL for the State of California. • Most PWSs required to treat for perchlorate are expected to install single pass ion exchange systems given the simplicity and relatively low costs and based on current trends in Southern California. The advent of perchlorate-selective resins has made single pass ion exchange an economically competitive treatment option for perchlorate removal. • Compared to other regulatory determinations, cost implications of a perchlorate MCL are relatively low due to the limited occurrence in source waters throughout the U.S. At an MCL of 4 µg/L, total compliance costs are estimated to be $2.1 billion. 7 The estimated nationwide compliance cost drops to approximately $0.1 billion at an MCL of 24 µg/L due to the small number of PWSs contaminated with perchlorate at that level. However, a small number of systems are carrying this cost burden and the cost impacts to an individual system installing perchlorate treatment would likely be significant. • Costs to treat large PWSs account for the majority of the estimated nationwide compliance costs due to the higher percentage of large PWSs with perchlorate contamination (Table 3-2) and the higher capital and O&M costs to treat the greater quantity of water requiring treatment for a large system. • Capital costs for single pass ion exchange are relatively low due to the simplicity of the treatment system. Capital costs to install single pass ion exchange systems 7 Capital plus total operating costs (NPV) based on 20 years life of service at a 3% discount rate.

Section 5 Conclusions
American Water Works Association National Cost Implications of a Potential Perchlorate Regulation 1571028 32 for all PWSs with perchlorate concentrations exceeding 4 µg/L are estimated to be $0.8 billion. Costs to operate the treatment systems for 20 years account for a larger percent of the total costs at $1.2 billion (NPV). A significant portion of the O&M costs for single pass ion exchange systems is the cost to periodically replace the spent resin.
• The presence of nitrate is known to substantially affect resin capacity and thus O&M costs. The effect of nitrate co-occurrence on costs was implicitly included in the cost evaluation by basing the O&M cost equation on known full-scale operating costs for systems with a range of water quality characteristics (i.e., nitrate concentrations ranging from 5 to 13 mg/L as nitrogen. Nevertheless, it may be beneficial in subsequent studies to consider the distribution of nitrate cooccurrence in the United States and then make reasonable assumptions of treatment process selection for the impacted utilities and the associated treatment costs.