Full opinion text
MEMORANDUM AND ORDER ON A MOTION BY THE UNITED STATES FOR AN ORDER OF IN-JUNCTIVE RELIEF STEARNS, District Judge. On February 12, 1998, the United States, on behalf of the federal Environmental Protection Agency (“EPA”), brought this enforcement action against the Massachusetts Water Resources Authority (“MWRA”) and the Metropolitan District Commission (“MDC”), alleging violations of the Safe Drinking Water Act (“SDWA”), 42 U.S.C. §§ 300f, et seq., and EPA’s Surface Water Treatment Rule (“SWTR”), 40 C.F.R. Part 141. The United States seeks injunctive relief in the form of an order requiring the MWRA to build a filtration plant to treat the water that it draws from the Wachusett Reservoir to supply the metropolitan Boston area. The MWRA initially maintained that because the Massachusetts Department of Environmental Protection (“DEP”), the primary agency responsible for enforcement of the SWTR, had determined that it was in compliance with the SWTR’s filtration avoidance criteria, it could not be compelled by the EPA to filter its water. The MWRA proposed instead to treat its water with ozone, which coupled with aggressive watershed protection and an accelerated program to replace aging pipes, the MWRA believed to be a more cost-efficient alternative to filtration. The MWRA conceded that subsequent to the DEP’s determination (and after the filing of its initial brief), it fell, albeit narrowly, out of compliance with the fecal coliform avoidance criterion (one of the eleven filtration avoidance criteria specified by the SWTR). The EPA immediately renewed its request for a filtration order, arguing that, the SWTR admits only a filtration remedy for a compliance violation, no matter what its magnitude. The MWRA took the position that because the SDWA, 42 U.S.C. § 300g-3(b), authorizes a district court to enforce compliance with the SWTR by entering “such judgment as protection of public health may require,” the court’s power to fashion a remedy for a compliance violation is more flexible than the enforcement mandate conferred by Congress on the EPA. In a written opinion, the court agreed with the MWRA that “the SDWA does not deprive a court of discretion in fashioning remedies- for a violation of the SWTR.” See United States v. MWRA, 48 F.Supp.2d 65, 72 (D.Mass.1999). After the Court of Appeals rejected the EPA’s petition for interlocutory review of the court’s determination, twenty-four days of eviden-tiary hearings were held to consider the EPA’s request for injunctive relief. Twenty-three witnesses, mostly experts, testified and 524 exhibits, were entered in evidence. Final arguments were held on April 14, 2000. The court agreed to the parties’ request that it expedite its decision for release on May 5, 2000, so that there would be no delay in the construction of the planned new treatment facility. This self-imposed deadline has aspects both good and bad. On the positive side, this opinion is much shorter than it would otherwise have been. There is, however, a lingering fear, that in reviewing the mass of testimony and exhibits offered during the trial, I may have missed something truly important. As a prophylaxis, I have used the very thorough suggested findings submitted by the parties as a cross-check on my evaluation of the evidence. I have read the transcripts of the witness testimony and, to the extent humanly possibly in so short a time, the tens of thousand pages of trial exhibits. What follows is not a conventional finding of facts. I have not selected one version of a contested fact over another based on any assignment of the burden of proof. Burdens of proof, while they work well in resolving most legal disputes, do not easily lend themselves to the resolution of scientific controversies. Science, by and large, rejects binary decision making in favor of a more nuanced quest for understanding. While a scientist might testify that a supposed fact has been proven to be false, the same scientist, when asked about conflicting data, will say only that an asserted fact has not been disproven or “falsified,” and could therefore “possibly” be true. In this decision, I relate those facts, including those that are in dispute, that fall within what I consider to be a reasonable range of possibility, indicating where appropriate the facts that I believe were shown to enjoy the greater empirical support or reflected the thinking of witnesses whom I found especially credible. I will incorporate the rulings of law made in United States v. MWRA, supra. No subsequent decision of a higher court has caused me to doubt their essential correctness. While there is no doubt that Congress, in enacting a statute, “may intervene and guide or control the exercise of the courts’ discretion,” its decision to do so is not to be “lightly assume[d],” especially in the absence of a clear legislative command. Weinberger v. Romero-Barcelo, 456 U.S. 305, 313, 102 S.Ct. 1798, 72 L.Ed.2d 91 (1982). “Unless a statute in so many words, or by a necessary and inescapable inference, restricts the court’s jurisdiction in equity, the full scope of that jurisdiction is to be recognized and applied. ‘The great principles of equity, securing complete justice, should not be yielded to light inferences or doubtful construction.’ ” Id., quoting from Porter v. Warner Holding Co., 328 U.S. 395, 398, 66 S.Ct. 1086, 90 L.Ed. 1332 (1946). The most explicit Congressional statement clarifying the intent of § 300g-3(b) [providing for judicial review of regulatory orders of the EPA Administrator] appears in the House Conference Committee Report on the 1974 enactment of the SDWA. The Conference Report states that: [t]he Committee intends that courts which are considering remedies in enforcement actions under this section are not to apply traditional balancing principles used by equity courts. Rather they are directed to give utmost weight to the Committee’s paramount objective of providing maximum feasible protection of the public health. H.R.Conf. Rep. No. 93-1185, at 23 (1974), 1974 U.S.C.C.A.N. 6454. In emphasizing its overriding goal of protecting the public health, Congress did not, however, say that a court was to limit itself to mechanical enforcement of EPA compliance orders. Had it been Congress’s intent to strip the courts of their equitable powers, one would think that it would have drafted § 300g-3(b) to say so, for example, by imposing the same narrow mandate on the courts that it imposed on the EPA in § 300g-1(b)(7)(C)®. Instead Congress used language descriptive of the traditional powers of a court of chancery. Why Congress might not have wanted to eliminate judicial discretion in ordering compliance with the SDWA is not difficult to imagine. Technology evolves more rapidly than typically does legislation, and there is an inherent danger in attempting to legislate today’s science as the foreordained solution for tomorrow’s problems. Congress may also have been concerned that an overly rigid application of the filtration mandate by the EPA might result in a wasteful expenditure of finite public funds to correct de minimis problems, or even exacerbate problems that the legislators had not foreseen. Cf. United States v. City of San Diego, 1994 WL 521216, at 8 (S.D.Cal.1994); 33 U.S.C. § 1311(j)(5). In sum, while the issue is by no means open and shut, I agree with the MWRA that the SDWA does not deprive a court of discretion in fashioning remedies for a violation of the SWTR. 48 F.Supp.2d at 71-72. The opinion will proceed as follows. I begin with a brief history of Boston’s water supply, followed by a discourse on the pathogenic threats that influence contemporary thinking about the safety of the nation’s drinking water. I then describe the legal and regulatory framework intended by Congress and the EPA to insure the health of public water supplies. I follow with a discussion of the MWRA distribution system and the watersheds from which it draws its water. Finally, I assess the current quality of MWRA water and the differing approaches of the MWRA and the EPA to the issue of preserving its safety. I. HISTORICAL BACKGROUND Colonial Boston drew its water from underground wells and rain-fed cisterns. Ex. 291, at 3-2. By the end of the eighteenth century, consumption began to outstrip the increasingly contaminated supply of natural water. Fern L. Nesson, Great Waters: A History of Boston’s Water Supply 112 (1983) (“Nesson”). In 1796, a privately chartered company, the Aqueduct Corporation, sought to profit from the demand for clean water by building a network of gravity-fed, underground wooden pipes connecting Boston to Jamaica Pond. The company’s efforts, however, did little to slake a rapacious public thirst. Ex. 291, at 3-1. Public officials ineffectually debated Boston’s water problem for several decades without achieving a consensus. In 1845, a frustrated Boston Water Committee turned to John Jervis, the engineer who built New York City’s Croton aqueducts, for advice. Ex. 291, at 3-1. Jervis recommended that an aqueduct be built to carry water from Long Pond (Lake Cochi-tuate)' in Natick to a holding reservoir in Brookline. The City Council endorsed Jervis’ proposal, and in 1846, the General Court passed the Boston Water Act. The Act established a three-member Cochituate Water Board, and authorized the issuance of $3,000,000 in public bonds. In 1848, the Cochituate water system, capable of delivering 18 million gallons daily of fresh water, was opened. As indoor plumbing became more commonplace, the demand for water increased accordingly. In 1851, the Cochituate Water Board purchased the Aqueduct Corporation and connected the Jamaica Pond waterworks to the Cochituate system. In 1865, the Board began construction of a 731 million gallon reservoir at Chestnut Hill to serve a population now in excess of 175;000. Ex. 316, Att. 3, at 114. In 1878, the Board added six small reservoirs fed by the Sudbury River to the system. Ex. 291, at 3-1. By the 1890’s, Boston and its burgeoning suburbs were again experiencing severe water shortages. In 1893, the Legislature ordered the State Board of Public Health to explore the feasability of a permanent solution. In 1895, Frederic Pike Stearns, the Board’s chief engineer, saw such a solution in the pristine watersheds to the west of Boston. He recommended that the south branch of the Nashua River be dammed to create a 63 billion gallon reservoir in Clinton, Massachusetts (the Wachusett Reservoir). Stearns based his recommendations on three contemporary factors: the sparsity of settlement and industry in the Nashua watershed, the relative purity of the water (which would improve through long storage in a large reservoir), and the availability of a supply propelled by gravity rather than pumping. Nesson, at 21. Stearns also urged that a unified water district encompássing the greater Boston metropolitan area be created. After devising a fee-sharing formula based on real estate values and population size, the Legislature adopted Stearns’ proposals, and in 1895 created the Metropolitan Water District. Interestingly enough, the political resolve that led to the adoption of Stearns’ plan was heavily influenced by a distrust of filtered drinking water. The virtues of avoiding filtration seemed self-evident in 1895. Filtration had worked under experimental conditions, but it was too new and involved technology that could malfunction. Disruptions in water flow and the serious consequences of polluted water supply were thought best avoided altogether. Id., at 32. The newly-established Metropolitan Water Board purchased 4,100 acres of land in West Boylston and Clinton as the site for the new reservoir, together with 5,600 acres of the adjacent watershed. The Wa-chusett Reservoir, in its day the largest man-made reservoir in the world, was completed in 1906 under Stearns’ oversight. Ex. 395, at 4-2. The Wachusett Reservoir was connected by two massive aqueducts, deliberately over-engineered to accommodate future expansion, to the Sud-bury system and the Chestnut Hill Reservoir. Ex. 291, at 3-2. In 1922, Henry Goodnough, Stearns’ successor, recommended the construction of a reservoir on the Swift River to collect its flood flows. He also proposed to channel flood water from the Ware River through a gravity-operated aqueduct to the Wachusett Reservoir. (Both of these projects had been originally conceived by Stearns in his master plan). X.H. Good-nough, Proposed Extension of the Metropolitan Water District, Journal of N.E. Water Works Ass’n, June 1922, at 254. In 1926, the State Board of Public Health and the rechristened Metropolitan Water and Sewerage Board embarked on the second stage of Stearns’ visionary scheme. Ex. 291, at 3-2. Despite fierce opposition from the four towns that were to be flooded, the Ware River Supply Act was passed on May 29, 1926, authorizing the construction of the Wachusett-Coldbrook Tunnel. See 1926 Mass. Acts, ch. 35. The Swift River Act followed on April 26, 1927, extending the tunnel to the Swift River. See 1927 Mass. Acts, ch. 111. The Ware River aqueduct was completed in 1931, and the Swift River Reservoir in 1939 (later rebap-tized as the Quabbin Reservoir). Because of its size, the Quabbin Reservoir took seven years to fill. Eighteen miles long, with a holding capacity of 412 billion gallons of water, the Quabbin remains one of the largest man-made reservoirs in the world. Ex. 291, at 3-3. After years of intervening neglect, the Legislature in 1985 created the MWRA. MacDonald, 1:25. The enabling statute established an MWRA Advisory Board consisting of representatives, of each of the cities and towns in the MWRA’s service area. MacDonald, 2:12. The MWRA is responsible for maintaining 130 miles of aqueducts and 265 miles of water mains. Ex. 291, at 3-15. The constituent cities and towns are in turn responsible for maintaining the 6,700 miles of service pipes within their boundaries. Id. The MWRA is funded by annual charges assessed to the member communities based on water use. Id. The member communities, however, set water rates for then-residents. The MDC is responsible for monitoring the quality of water entering the MWRA system, and for managing the Wachusett, Quabbin and Ware watersheds. Ex. 291, at 3-15. The MWRA reimburses the MDC for the costs of watershed protection, and services the debt incurred by the MDC’s watershed land acquisition program. Estes-Smargiassi, 2:103. II. PATHOGENS AND TESTING METHODS The earliest recognized microbiological contaminants of drinking water were the Rickettsia and Vibro cholerae bacteria responsible for outbreaks of typhoid and cholera in the Dickensian urban conditions associated with the nascent Industrial Revolution. Because most bacteria thrive in the intestinal tract, they are often spread, by fecal-oral contamination. Bacteria have relatively short lives and are highly susceptible to oxidizing disinfectants like chlorine and ozone. Some pathogenic bacteria like Legionella (associated with Legionnaire’s Disease) and Mycobacterium avi-um (associated with opportunistic infections in immunocompromised individuals) occur: naturally in the environment and breed prolifically' in plumbing systems. They can also grow in water distribution systems. Rose, 12:129. Viruses are a second microbiological contaminant that pose a threat to the public drinking water supply. The smallest of the pathogens, viruses have no independent metabolism and are only able to reproduce by parasitieally invading a host cell and using its genetic material to replicate. Those that are known to cause waterborne disease in humans are the so-called enteric (intestinal) viruses associated with acute gastroenteritis. Although more resistant than bacteria, viruses are vulnerable to disinfectants. Of the varieties of waterborne microbial organisms that pose a potential danger to the public water supply, two protozoans, Giardia lamblia and Cryptosporidium parvum cause the greatest current concern. This is because of their resistance to disinfection, prolonged life cycles, and high infectivity. Giardia was first identified as a disease-causing organism in the late nineteenth century. Giardia is an intestinal parasite and the cause of the disease giardiasis, the common symptoms of which are diarrhea and dyspepsia. Giardiasis is easily treatable. Giardia is transmitted fecally in a protective cyst that opens (excystates) when it becomes attached to the intestinal wall of an animal or human host. The shell-like structure of the Giardia cyst offers protection from disinfectants but is pervious to chlorine and ozone. Cryptosporidium parvum was recognized as a water contaminant in the early twentieth century, but was not identified as a human pathogen until the 1980’s. Daniel, 4:132; Rose, 12:119. Like Giar-dia, Cryptosporidium is common in surface water sources, including bodies of water generally thought to be pristine. In the human body, a Cryptosporidium par-vum infection can lead to the disease eryp-tosporidiosis, which manifests itself in symptoms of chronic fatigue, gastric disturbance, nausea, weight loss diarrhea, and fever. The symptoms can be fatal to persons with compromised immune systems, particularly those suffering from AIDS, cancer patients, and the very young or old. There is no effective treatment or cure for eryptosporidiosis, Rose, 12:121, although in healthy individuals the disease is self-limiting and usually runs its course in 7 to 14 days. Fecal-oral ingestion is a common form of transmission of the disease, but it may also be transmitted by direct or indirect contact with an infected person or animal. In 1993, after a mechanical malfunction in Milwaukee’s water filtration plant, a Cryptosporidium release caused an outbreak of eryptosporidiosis which infected over half of the system’s 800,000 consumers. At least 50, and perhaps as many as 100 immunocompromised individuals are believed to have died from the illness. Ex. 112, 4-2; Hass, 8:85. According to the Centers for Disease Control, there were ten outbreaks of eryptosporidiosis in the United States between 1984 and 1995 (none of which were associated with unfiltered water systems). Rutherford, 8:42. The Milwaukee outbreak accounted for 93 percent of the recorded cases of eryptospo-ridiosis, although there is a consensus among health professionals that the great majority of cases go unreported. Ex. 112, at 4-3. Cryptosporidium (only the parvum species is infective to humans) is endemic in the animal kingdom. The Cryptosporidi-um oocyst excystates in the intestines of its animal host and is shed into the environment in fecal waste. The thick-walled oocyst is capable of surviving outside the host in a fully infective stage for weeks, and even longer in cold water. Rose, 12.TS0. The oocyst, because of the thickness of its shell, is highly resistant to chlorine. Daniel, 4:133; 5:58. Farm animals, particularly cattle, are prolific ex-creters of the Cryptosporidium parasite, a concern in the Wachusett watershed because of several local dairy farms. Humans and mammalian wildlife are also sources, as are birds, although avian Cryp-tosporidium is not infective of humans. The dose-response relationship of the Cryptosporidium parvum oocyst is largely a matter of conjecture, although it appears to vary depending on the strain involved and the relative health and resistance of the exposed subject. An often-cited study involving human volunteers was conducted at the University of Texas at Houston. Rutherford, 8:38. The study, which focused on the Iowa strain of Cryptosporidi-um parvum, estimated the ID60 of Cryp-tosporidium to be 132, that is, that the average person would contract cryptospo-ridiosis after ingesting 132 oocysts. Id., at 8:39. Because fecal deposits are the most common means by which pathogens are introduced into public water supplies, science for nearly a century has used fecal coliform counts as a predictor of pathogenic risk. Rose, 12:123. (Fecal coliform itself is not a risk to human health, nor are all bacteria that respond positively to the thermal test for fecal coliform necessarily fecal in origin). While fecal coliform is a reasonably good indicator of gross bacterial contamination, scientists have come to understand that there is no statistical correlation between fecal coliform counts and the presence or concentration of Cryptos-poridium and Giardia in water. Rose, 13:72. Attempts to develop a precise method of monitoring for the presence of Cryptospo-ridium have been largely unsuccessful. While laboratory testing methods for fecal coliform are very reliable, the Cryptospori-dium oocyst, some 5 to 10 microns in diameter, can only be detected by sophisticated microscopic analysis. The testing method sanctioned by the EPA throughout the 1990’s depended on a separation technique and the use of a fluorescent stain to identify the oocyst. Clancy, 15:95. The immunoflourescence assay, however, has proven to be a poor oocyst detector, because the stain tends to identify with other types of biological matter, causing false positives. The separation technique, which involves the reduction of the water sample by filtration, centrifugation, and flotation, if done improperly, can also lead to false negatives. Id., at 15:110-111. The reliability of Cryptosporidium oo-cyst screening was called into serious question by a 1997 EPA study supervised by Dr. Jennifer Clancy. Ex. 445. Independent laboratories participating in a performance evaluation succeeded in recovering only 23 percent of the oocysts in spiked samples, with numerous reports of false negatives and false positives. In response to the Clancy study, the EPA implemented a number of changes intended to give greater rigor to the testing protocol, although with disappointing results. According to Dr. Clancy, even with the new protocol, the oocyst recovery rate improved to only .35 percent. Clancy, 15:115. A subsequently revised EPA protocol (developed by Dr. Clancy) was promulgated in draft form in 1999 as Method 1622. It uses magnetic beading to concentrate the oocysts, Ex. 131, 2-7, yielding a further improved recovery rate in the range of 40 to 50 percent. Clancy, 15:119. Despite the absence of a regulatory mandate, the MWRA has monitored its water for Cryptosporidium since 1994. The testing method used by the MWRA is the one sanctioned for data collection under the EPA’s 1996 Information Collection Rule (“ICR”), although the MWRA has used water samples 10 to 22 times larger than called for by the ICR and has tested larger portions of the samples collected. Under the ICR protocol, sampling was required on a monthly basis. From 1996 through 1998, the MWRA sampled two to four times a month, and in January of 1999, began sampling on a weekly basis. Even with the enhanced sampling method, there has never been a confirmed Cryptos-poridium oocyst detected (that is, an oo-cyst identifiable by its internal structure) at the Cosgrove Intake or in the Wachu-sett Reservoir, and no presumptive oocyst since 1995. Estes-Smargiassi, 2:138-139; Clancy, 16:58-59; Ex. 47. See also Aieta, 24:29; Ex. 519. EPA’s Revised Draft Unfiltered Water System Guidance Manual, issued February 4, 1999, recommends either filtration or a minimum of 2-log Cryptosporidium inactivation when the number of oocysts in baseline samples exceeds 1 per 100/1. Ex. 398, at 4-4. According to EPA, the Draft Guidance was issued without appropriate review by senior EPA management. King, 22:91-92, 109-110. EPA has there-fore concluded “that it should not finalize the Draft Guidance, at least until after evaluation of the spiking study, the full ICR data set,'and completion of ongoing Federal Advisory Committee Act deliberations.” EPA Proposed Finding # 162. III. STATUTORY AND REGULATORY FRAMEWORK The SDWA, enacted in 1974, charges the EPA with overall responsibility for insuring the safety of the nation’s public water supply. Congress directed the EPA to promulgate Maximum Contaminant Levels (“MCLs”) specifying the upper bound of contaminants permissible in finished water, or, if that was not feasible for economic or technological reasons, to mandate treatment techniques to insure the public health. 42 U.S.C. §§ 300f(l)(C); 300g-l(a); 300g-l(b)(7)(A). In 1986, frustrated by an apparent lack of rule-making progress, Congress amended the SDWA to require the EPA to mandate disinfection for all public water systems. 42 U.S.C. § 300g-l(b)(8). The 1986 amendments entailed a Congressional judgment that filtration is a superior technology for treating contaminated water supplies. Congress directed the EPA to specify criteria for requiring filtration as a treatment technique after considering “the quality of source waters, protection afforded by watershed management, treatment practices (such as disinfection and length of water storage) and other factors relevant to protection of health.” 42 U.S.C. § 300g-l(b)(7)(C)(i). Congress, in other words, stopped short of ordering filtration as an all-encompassing preventive. While the 1986 amendments strengthened EPA’s oversight of the regulatory process, the amendments preserved the role of the states in enforcing the SDWA. States whose drinking water regulations are determined by the EPA to be at least as strict as those mandated by federal regulations have “primary enforcement responsibility” for the safety of public water systems within their jurisdiction, including the decision whether to order filtration. 42 U.S.C. § 300g-2(a); 42 U.S.C. § 300g-l(b)(7)(C)(ii). The DEP was granted enforcement responsibility by EPA in 1993. In response to the Congressional prodding, on June 29, 1989, the EPA promulgated the SWTR, which applies to all public water systems using surface water or ground water sources affected by surface water. See 40 C.F.R. § 141.70 et seq. The EPA concluded that it was not feasible to establish MCLs for Giardia larnb-lia, viruses, heterotrophic bacteria, and Legionella, the contaminants that had been identified as presenting the most significant risks to the public health. Consequently, the SWTR mandated that both filtered and unfiltered systems achieve: (1) at least 99.9 percent (3-log) removal and/or inactivation of Giardia lamblia cysts between a point where the raw water is not subject to recontamination by surface water runoff and a point downstream before or at the first customer; and (2) [a]t least 99.99 percent (4-log) removal and/or inactivation of viruses between a point where the raw water is not subject to recontamination by surface water runoff and a point downstream before or at the first customer. 40 C.F.R. § 141.70(a). The stated goal of the SWTR is to reduce the risk of illness from waterborne pathogens to one occurrence yearly among every 10,000 consumers of public water. Ex. 115, SWTR, 54 FecLReg. at 27490; Ex. 114, IESWTR, 62 Fed.Reg. at 59489. The SWTR was intended to be “self-implementing,” in the sense that the SDWA required nonconforming water systems to begin filtration within 30 months of the SWTR’s promulgation (that is, by June 29, 1993), unless within 18 months (by December 30, 1991), the system could demonstrate that it met the filtration avoidance criteria. (These are set out at 40 C.F.R. § 141.71(a) and (b)). Public water systems that met the avoidance criteria but later fell out of compliance were given eighteen months from the date of noncompliance to begin filtration. 40 C.F.R. § 141.73. Although the deadlines are couched in categorical terms, an internal guidance issued by the EPA in 1992 gave state enforcement agencies discretion to defer a final filtration determination if it appeared that a water system through intermediate measures could bring itself into compliance. The SWTR established eleven avoidance criteria, all of which a water system must meet to be exempt from the filtration requirement. Two of the criteria concern the quality of a system’s source water. (1): In relevant part, no more than 10 percent of samples taken prior to the first point of disinfection may contain fecal coli-form concentrations in excess of 20 colony forming units (“cfu”) per 100 ml during any six month period. 40 C.F.R. § 141.71(a)(1). Sampling must be done by EPA-approved methods. 40 C.F.R. § 141.74(a). (2): Turbidity cannot exceed 5 nephelometric turbidity units (“NTU”) in samples taken prior to the first point of disinfection (with an exception for unusual and unpredictable events). 40 C.F.R. § 141.71(a). Four criteria establish minimum levels of disinfection. (1): The system must meet the 3-log (99.9 percent) requirement for inactivation of Giardia lamblia cysts in at least 11 out of any preceding 12 months, and the 4-log (99.99 percent) inactivation requirement for viruses every day but one during any given month. 40 C.F.R. § 141.71(b)(l)(i). (Log removal is measured as a function of contact time (“CT”), a value obtained by multiplying the amount of residual disinfectant by the time it is in contact with treated water. 40 C.F.R. § 141.72(a)(1)). (2): The system must either be redundant in design or provide for the automatic shut-off of flow if the concentration of residual disinfectant falls below 0.2 mg/1. 40 C.F.R. § 141.71(b)(1)(h). (3): The system must not permit the residual disinfectant concentration in water entering the distribution system to fall below 0.2 mgd for more than four continuous hours. 40 C.F.R. § 141.71(b)(l)(iii). (4): The residual disinfectant concentration must not be undetectable in 5 percent of the samples taken during any month for two consecutive months. 40 C.F.R. § 141.71(b)(l)(iv). Five criteria involve watershed protection and systems operations. (1): The system must have a comprehensive watershed control program that meets mandated standards designed to minimize the infiltration of the source water by Giardia lamblia and viruses. 40 C.F.R. § 141.71(b)(2). (2): The system must be inspected annually by the state enforcement authority to insure the efficacy of the watershed control program and disinfection procedures. 40 C.F.R. § 141.71(b)(3). (3): The system must not have been identified as responsible for an outbreak of waterborne disease, or if it has, it must have implemented corrective measures adequate to prevent a recurrence. 40 C.F.R. § 141.71(b)(4). (4): The system must be in compliance with the M.C.L. for total coliform concentrations in the distribution system. No more than 5 percent of samples in any eleven of twelve months may exceed the total coliform standard of 100 cfu per 100 ml. 40 C.F.R. § 141.71(b)(5). (5): The system must meet the M.C.L. for disinfection by-products (“DBPs”) in the distribution system (currently 0.10 mg trihalomethanes per liter or 100 parts per billion (“ppb”)). 40 C.F.R. § 141.71(b)(6). In 1996, Congress amended the SDWA a second time, by directing the EPA to promulgate an Interim (“IESWTR”) and Final Enhanced Surface Water Treatment Rule addressing the threat of Cryp-tosporidium and DBPs to the safety of drinking water supplies. 42 U.S.C. § 300g-l (b)(2)(C). The 1996 amendments loosened the filtration avoidance criteria for water systems “having uninhabited, undeveloped watersheds in consolidated ownership, and having control over, access to, and activities in, those watersheds” where a state determines that alternative treatment methods will achieve a greater removal of pathogens from drinking water than will filtration. 42 U.S.C. § 300g-l(b)(7)(C)(v). The amendments also directed the EPA Administrator to use a cost-benefit analysis in assessing the health risk reduction benefits expected from any new national primary drinking water regulation that includes an M.C.L. or proposed treatment technique. 42 U.S.C. § 300g-l(b)(3)(C)® & (ii). Cost-benefit analysis may not, however, be used to establish MCLs for DBPs “or to establish a maximum contamination level or treatment technique requirement for the control of Cryptosporidium.” 42 U.S.C. § 300g-l(b)(6)(C). Apart from these modifications, the 1996 amendments left the essential structure of the SDWA intact. The IESWTR was published on December 16, 1998, and will take effect on December 17, 2001. Consistent with the 1996 amendments, the IESWTR focuses' on Cryptosporidium. For • filtered systems, the IESWTR requires a 2-log (99 percent) reduction of Cryptosporidium oocysts. It also requires unfiltered systems to extend the existing Giardia lamblia watershed controls to cover Cryptosporidium. Unfiltered systems are not, however, required to monitor their treated water for Cryptos-poridium. Much of the empirical data on which the IESWTR is based was gathered by EPA under the ICR assaying the prevalence of Cryptosporidium in source water. See 61 Fed.Reg. 24354 (1996). Water systems serving in excess of 100,000 consumers were required to test monthly for 18 months for the presence of Cryptosporidi-um and to submit water samples for independent analysis by EPA-approved laboratories. IV. JUDICIAL ENFORCEMENT The EPA Administrator is authorized to seek enforcement of the SDWA’s requirements in the district court if a state, after being given notice of a violation in a regulated water system, does not within 30 days commence an appropriate enforcement action. 42 U.S.C. § 300g — 3(a)(1)(B); 300g-3(b). The district court may enter, in an action brought under this subsection, such judgment as protection of public health may require, taking into consideration the time necessary to comply and the availability of alternative water supplies; and, if the court determines that there has been a violation of the regulation or schedule or other requirements with respect to which the action was brought, the court may, taking into account the seriousness of the violation, the population at risk, and other appropriate factors, impose on the violator a civil penalty of [sic] not to exceed $25,000 for each day in which such violation occurs. 42 U.S.C. § 300g-3(b). V. THE DISTRIBUTION SYSTEM Quabbin water enters the Wachusett Reservón- at its western end near the mouth of the Quinapoxet River. The amount of water withdrawn from the Quabbin Reservoir varies seasonally depending on demand, which is highest in the dry months of May to December. Ex. 16; Estes-Smargiassi, 2:63. At any given time, roughly half the water in the Wachu-sett Reservoir is contributed by the Quab-bin. Ex. 395, at 1-5 (Table 1-1). Most of the remaining water is collected from the Wachusett watershed. Ex. 127, at 2-7. At the eastern end of the Reservoir, water flows into the Cosgrove Tunnel at the Cosgrove Intake. The tunnel is 8 miles long, 14 feet in diameter, and has a carrying capacity of up to 600 million gallons of water a day. Estes-Smargiassi, 2:66. At the terminus of the Cosgrove Tunnel, at Shaft C in Marlborough, the water enters the Hultman Aqueduct. At Framingham, the Hultman Aqueduct branches in two directions. The smaller branch, the Weston Aqueduct, empties into the Weston Reservoir. The main branch of the Hultman Aqueduct flows beneath the Norumbega Reservoir in Weston. A portion of the water is drawn into the Norumbega Reservoir (which supplies backup water during periods of peak demand). The remainder flows through various connecting tunnels to consumers. The MWRA treats its drinking water with three general techniques: primary disinfection, corrosion control, and residual disinfection. Primary disinfection is the use of chemicals, primarily chlorine, to kill microorganisms in water. Estes-Smar-giassi, 2:125-126. Corrosion control involves the adjustment of the chemistry of water to reduce the leaching of metals (such as lead) from pipe casings. Id., at 127. Residual disinfection maintains low doses of disinfectant in the water to prevent recontamination as the water moves through the distribution system. Id., at 128. Prior to June of 1996, MWRA water received primary and residual treatment at Weston. The water was adjusted for pH and fluoridized at Shaft 4 in Southborough. In September of 1997, the primary disinfection point was moved to the Cosgrove Intake and the disinfectant changed from chloramine to more powerful free chlorine. The corrosion control point was moved to Marlborough and the chemical mix was changed to regulate alkalinity as well as pH. Residual disinfection treatment was also modified in August of 1997 by injecting ammonia into the water at a point downstream of the Norumbega station to generate a more durable chloramine residual. Id., at 127-128. Most of the 265 miles of mainline pipes in the MWRA system were installed in the 1800’s and early 1900’s and only intermittently replaced or refurbished. Estes-Smargiassi, 2:74; Ex. 127, at 2-12 to 2-13. Some eighty percent of the present-day pipes are unlined cast iron or steel. Most are corroded and “prone to leaks ... [and] water quality problems” caused by intrusion (outside contaminants seeping into the pipe and the water supply). Id., at 3-10; Daniel, 5:74-76. Many of the pipes are severely tuberculated (incrusted with metal deposits) to the point that the flow of water is virtually occluded. Ex. 17; Ex. 18 and Ex. 19. The 6,700 miles of pipe owned by the MWRA’s constituent cities and towns have suffered from even greater neglect. Although a few communities (Brookline, for example), have done an admirable job in rehabilitating their delivery systems, nearly half the pipes supplying water to consumers are of the antiquated unlined cast iron type (in some communities the figure rises to 70 percent). Ex. 20. In the mid-1990’s, the MWRA established a rehabilitation target of 2 % percent per year for its own pipes, a goal that it has generally exceeded. In 1997, it instituted a two year pilot project, the Local Water Infrastructure Rehabilitation Assistance Program, offering $30 million in grants and interest free loans to member communities to encourage pipeline improvements. Ex. 301. In 1999, the MWRA Advisory Board extended the program for ten years, with a projected annual allocation of $25 million. The MWRA presently has four capital improvement projects under construction that will impact the distribution system. These are the 17.6 mile MetroWest Water Supply Tunnel ($728 million) which will carry water from Marlborough to Weston; the construction of covered facilities to store finished water, permitting the closure of the two remaining open reservoirs ($205 million); the construction of a new disinfection facility at Walnut Hill in Marlborough ($309 million); and the ongoing rehabilitation of the water mains ($460 million). Ex. 5. VI. THE WATERSHEDS There are three contiguous watersheds within the MWRA system, the Quabbin, the Ware, and the Wachusett, collectively covering some 400 square miles of land inhabited by some 44,000 humans and 3,600 farm animals. Estes-Smargiassi, 2:54; Reilly, 21:7-8; Ex. 131, at 4-2. Wachusett Watershed The Wachusett is the most developed of the three watersheds. Its 117 square miles of surface area contain all or parts of twelve towns with watershed populations ranging from single numbers into the thousands. Ex. 204. The largest population centers are concentrated in the southern portion of the watershed (primarily in the Towns of West Boylston and Holden). Population density in the watershed as a whole is 290 persons per square mile. The present-day human population numbers approximately 34,000. The watershed contains 118 miles of roads and 17.5 miles of railroad track. Ex. 149. The topography of the watershed is hilly, sloping upward from the Reservoir at 395 feet above sea level, in a northwesterly direction towards the 2,006 foot peak of Mount Wachusett. Estes-Smargiassi, 2:59. Approximately 75 percent of the watershed is forested or covered by wetlands. Twenty-six percent of the land area is owned outright by the MDC, while an equivalent amount is owned by other government agencies and conservation groups. Eight percent of the land is used for agricultural pursuits, 9 percent is settled urban or residential, and 1 percent is dedicated to industrial and commercial uses. The Nashua, Quinapoxet and Stillwater Rivers contribute roughly 40 percent of the water collected by the Reservoir. Estes-Smargiassi, 2:62; Ex. 147, at 38; Ex. 395, at 2-33. A very small contribution is made by Malagasco Brook, which empties into the Reservoir at South Bay, and by the Boylston, French and Hasting Cove Brooks, which enter the Reservoir on the southeast shore. Nearly half of the Wachusett water arrives from the Quabbin watershed. Id. The balance comes from run-off and direct precipitation. Id. Over ninety percent of Wachusett water enters at the Thomas Basin, a narrow, elongated appendage to the northwest of the main body of the Reservoir. Ex. 13. The mouth at the southern end of the Thomas Basin is artificially constricted by the Route 12 causeway. “The constriction at the Route 12 bridge narrows the reservoir from approximately 1,000 feet to 50 feet, and makes Thomas Basin an effective detention and sedimentation basin helping to maintain the high quality of water in the main body of the reservoir.” Ex. 395, at ES-5. See also Ex. 13. The average time taken for water entering the Thomas Basin to migrate to the Cosgrove Intake is six months. Ex. 129. Water released from the Reservoir at the Wachusett Dam is drained by the Nashua and Merrimac Rivers into the Atlantic Ocean. Estes-Smargiassi, 2:54. The watershed is an important wildlife habitat and a major recreational area. The MDC permits hiking, cycling, seasonal shore fishing and cross-country skiing on much of the land under its control. Canoeing is permitted on West Waushacum Pond and on the upper reaches of the Quinapoxet and Stillwater Rivers. Seasonal hunting is also allowed in some areas. Recreational uses of MDC land are regulated by a Public Access Plan promulgated in 1996. See Ex. 147. The Massachusetts Department of Environmental Management (“DEM”) owns 2,052 acres of the watershed, including portions of the Leominster State Forest and the Wachusett Mountain Reservation. Id., at 131. The Massachusetts Division of Fisheries and Wildlife (“DFW”) manages 580 acres. DEM and DFW allow a variety of recreational activities. Sporting clubs own 1,450 acres of open space. The clubs allow hunting, trapping and target practice, and permit dogs. There are two private land trusts in the watershed, the White Oak Land Conservation Society (122 acres) and the Princeton Land Trust (4 acres). The trusts permit hiking, hunting, snowshoeing and skiing on their land. The Massachusetts Audubon Society owns three sanctuaries totaling 1,257 acres. There are also several municipal parks, six country clubs, and several public golf courses. The Trout Brook Reservation (660 acres) and Town Forest (124 acres) permit horseback riding, dogs, hunting, fishing and camping. Id., at 134 -135. The watershed contains twelve “significant” farms (with 10 or more cows), although six of these are located in the Worcester/Quinapoxet Basin. Ex. 35; Estes-Smargiassi, 2:122. There are also numerous “hobby” farms that stable horses and other animals. Id. In total, the farms in the watershed house some 2,250 domestic animals (principally dairy cows, horses and pigs). The Quabbin and Ware Watersheds The Quabbin, the most westerly of the three watersheds, covers an area of 187 square miles, and is 93 percent forested. Id., at 55. More than half of the land surface is owned by the MDC. Ex. 207. Approximately 3,000 persons live within the watershed (a population density of 16 persons per square mile). Less than 3 percent of the land area is dedicated to agricultural use, involving fewer than 450 farm animals. Ex. 131, at 3. The Quabbin Reservoir is a pristine water source, with very low turbidity, and extremely low levels of contaminants. There is no dispute that the Quabbin amply meets the filtration avoidance criteria. Ex. 127, at 2-3; Ex. 390, at 2-3, n. 1. The Ware River watershed, to the east of the Quabbin, contains 97 square miles of surface area, 85 percent of which consists of forests and wetlands. The population density is 77 inhabitants per square mile. More than one-third of the land area is owned by the MDC, while one-half of the remaining watershed is protected open space. Ex. 207. Very little water is presently diverted from the Ware watershed into the Wachusett Reservoir. There have been no diversions during the past five years. Estes-Smargiassi, 2:56-57. MDC Management Practices Within the Watershed The MDC, through its Division of Watershed Management, (“DWM”) is mandated by statute “to assure pure water for future generations.” M.G.L. c. 92, § 105. The DWM and the MWRA collaborated on the development of the 1991 Watershed Protection Plan (‘WPP”). The WPP identified livestock, unsewered septic systems, wildlife (principally birds, beaver and muskrat), human recreation, urban run-off, and transportation spills, as the major potential threats to the watershed’s integrity. The WPP was updated in 1995 and again in 1998. It has six principal components: (1) staffing and management goals and objectives; (2) a bird control program; (3) land acquisition; (4) implementation of the Watershed Protection Act; (5) the elimination of unsewered septic systems; and (6) mitigation of the threat posed by farm animal excretions and farm operations. Ex. 395, at 3-5 to 3-15; Estes-Smargiassi, 2:103-104. Some of the more significant initiatives taken pursuant to the WPP are described below. The Land Acquisition Program The 1992 Watershed Protection Act (“WPA”) established a $135 million fund to purchase development rights to environmentally sensitive property in the Wachu-sett and Quabbin watersheds. The goal established by the MDC was to raise its total watershed holdings to 25 percent of the land area, giving priority to purchases that would mitigate development and farming activity in close proximity to the Reservoir or its tributaries. See M.G.L. c. 92, § 107(A), inserting act § 6. In designing the program, [t]he MDC and MWRA developed a prioritization mechanism to establish what are the most important parcels of land to be purchased.... The MDC and MWRA staff, planners, hatural — environmental scientists and natural resource folks and others, ranked a series of factors as to how important they were to water quality. They included things such as steep slopes, the type of development which could be built on land, the proximity to tributaries and a number of other issues, including aquifer which allowed us to say that this piece of land is more important than another piece. And then we have, in fact, ranked every parcel in the watershed [Wje’re concentrating on purchases in the portion of the watershed which is more directly tributary to the reservoir. Estes-Smargiassi, 2:109-111. See also, Ex. 395, at 4-7. Since 1985, the MDC has purchased approximately 17,000 acres within the two principal watersheds. With respect to the Wachusett watershed, the MDC has surpassed its goal of 25 percent ownership. Ex. 395, at 3-14. The MDC expects to purchase 5,000 additional acres from private owners over the next five years, principally in the Wachusett watershed. Estes-Smargiassi, 2:37. Sewering The WPA identified leaking septic systems within the watershed as “the most significant potential source of pathogens and other pollutants of concern”. Ex. 395, at 6-29. This concern is exacerbated by sandy soil conditions that do a poor job of filtering wastewater. Walker, 16:103. In 1930, the MDC built sewers to evacuate wastewater from Holden and Rutland for treatment outside of the watershed. In 1991, however, the watershed still had 6,558 unsewered septic systems. Ex. 139, at 3-5. In 1995, the MDC inaugurated a Wastewater Facilities Plan to provide sewer connections for septic systems in and around Holden and West Boylston that had been identified as a source of fecal coliform polluting Gates Brook and the West Boylston Brook. Ex. 41; Ex. 395, at ES-20. When the Plan is completed in 2004, more than 40 percent of the septic systems in the watershed will be connected to the Upper Blackstone treatment plant in Worcester. Estes-Smargiassi, 2:120; Ex. 395, at 6-31. Bird Harassment Program In 1991, the MWRA concluded that roosting gulls and other birds were the probable source of seasonally high fecal coliform concentrations detected in water samples taken at the Cosgrove Intake. Estes-Smargiassi, 2:106.; Ex. 218, at 1. In 1992, the MDC instituted a campaign of harassment to discourage birds from roosting near the Intake. Scannel, 9:76-77. In 1993, the MDC intensified the harassment program in the late fall and winter when the bird population reaches its peak. Id., at 78-79. The MDC scatters the birds with noise makers, pyrotechnical devices, propane cannons, and distress tapes, and (most effectively) by launching boats in areas favored by the birds. Id., at 81. The MDC has also deployed aquatic nets and has erected scaring devices on islands and along the shoreline. Id., at 83; Ex. 218, at 4. In 1994, the MDC acquired two small Model 600 hovercraft to permit boats to be launched in winter when the Reservoir begins to ice over. Scannel, 9:84. In 1994, after experiencing a number of days when choppy conditions or ice made it impossible to launch the smaller hovercraft, the MDC purchased a more powerful all-weather Model 800 and built a de-icing dock. Id., at 92-94. VIL WACHUSETT/MWRA WATER QUALITY Despite deficiencies in the various methods used to test for the presence of contaminants, the filtration avoidance criteria of the SWTR provide a useful benchmark for measuring water quality. Of the two source water criteria, turbidity has not affected the quality of Wachusett water. Samples taken at the Cosgrove Intake have never exceeded the SWTR limit of 5 NTU, nor since 1991 have they exceeded the more stringent Massachusetts standard of 1 NTU, even during intense storm events. Aieta 11:42-43; Ex. 53. Algae are a potential threat to disinfection that are often associated with turbidity. However, algae levels in the Wachusett Reservoir are extremely low, as would be expected from the low turbidity. Edzwald, 14:59,106-107; Hiltebrand, 23:93-94. Fecal coliform is another matter. As the MWRA admits in its Proposed Finding # 189(a)(iii), the inability to satisfy the fecal coliform standard in 1991 was a principal reason why its Board of Directors voted not to seek a filtration waiver. The term “fecal coliform,” as previously noted, is somewhat misleading, as it is a generic description encompassing all coliform bacteria that respond positively to thermal testing. For the most part, the presence of fecal coliform in water is a poor marker for fecal contamination. The specific indicator for the presence of fecal matter, the bacterium E. coli, is in fact a small subset of the total fecal coliform population. Edberg, 7:27. Nor is fecal coliform a reliable indicator of the presence of Giardia or Cryptosporidium in water. Id., at 7:28. The source water avoidance criterion for fecal coliform requires that no more than 10 percent of samples taken prior to the first point of disinfection contain fecal eoli-form concentrations in excess of 20 cfu per 100 ml during any six continuous months. In 1991, 1992 and 1993, water samples taken at the Cosgrove Intake often exceeded the 10 percent limit, particularly during the winter months. (Fecal coliform levels are generally higher in winter because co-liform bacteria survive longer in cold water. Rose, 12: 130). The sharp drop in levels of fecal coliform recorded at the Intake after the full implementation of the gull harassment program in 1993 strongly supports the MWRA’s determination that roosting gulls were the principal coliform source. That determination is further corroborated by the temporary spikes in fecal coliform levels that were observed on occasions when the harassment program was disrupted by severe winter weather. The EPA presented evidence, through the testimony of Dr. William Walker, that a number of the tributaries in the Wachu-sett watershed do not meet the Massachusetts Class A Water Quality Standard with respect to fecal coliform concentrations. Walker, 16:109, 111; Ex. 449. The excee-dance is especially acute in areas with higher population densities and those impacted by agricultural activity. As Dr. Walker’s analysis of MDC data showed, Justice Brook, which is the cleanest of the streams for which data was gathered, is in the least developed area of the watershed, while Gates Brook, the dirtiest, flows through the area that is most urbanized. Walker, 16:114-115. Dr. Walker hypothesized that because of the gradient and soil composition of the watershed, storm (wet weather) events could cause large concentrations of fecal coliform to leach into the western end of the Reservoir where, under the right hydraulic and wind conditions, they could be transported to the Cosgrove Intake in “less than a day.” Walker, 16:184. The weak statistical association developed by Dr. Walker between antecedent rain events and fluctuations in fecal colifprm counts at the Cosgrove Intake (1 percent) and the virtual absence of detectable Giardia or Cryptosporidium in the samples taken (Ex. 438), strongly suggest that the hypothesis is flawed. (According to Dr. Walker Giardia or Cryptosporidi-um have settling rates one magnitude slower than fecal coliform which would lead one to expect them to be more readily transportable en masse). As the MWRA pointed out, the correlation between spikes in fecal coliform counts and the numbers of gulls roosting near the Cosgrove Intake is far stronger than any association with storm events identified by Dr. Walker’s models. This litigation was triggered by the MWRA’s admission that in January of 1999, it had fallen out of compliance with the fecal coliform avoidance criterion. More specifically, in December of 1998 and January of 1999, 14 samples were taken at the Cosgrove Intake in which fecal coli-form concentrations exceeded 20 cfu, that is, one more than the 10 percent of samples permitted. It was this admission that caused the court on May 3, 1999, to enter partial summary judgment for the EPA. The argument is now made by the MWRA that the facts developed at trial disprove its prior admission and that (inferentially) partial summary judgment was improvidently granted. The crux of the dispute involves the MWRA’s testing method for fecal coliform. In 1989, when promulgating the avoidance criteria, EPA required that fecal coliform levels be measured by EPA-approved methods. Ex. 115, SWTR, Fed.Reg. at 27530. Among these’ were the ■ MPN Method 908C and the Membrane Filter Procedure Method 909C, which while differing in format, involve incubation of a lactose-based solution at a temperature of 44.5°C. Id., at 27531 The MWRA, on the other hand (for reasons that are unclear), chose to use instead a non-EPA-approved enhanced recovery method, which is far more sensitive than either of the approved methods. Edberg, 7:36. A split-sample study of water samples taken at the Cos-grove Inlet during the first three months of 1999, showed a site-specific 100 percent increase in average recoveries using the MWRA’s enhanced method. See Ex. 401. The MWRA argues that had it used “the analytical method to which the avoidance criterion was calibrated, it would not have detected or reported a violation.” MWRA Proposed Finding # 206(d)(v). From a different perspective, the MWRA argues that, even if the enhanced method results are considered, the violation is de minimis, that is, “had the results shown 1 less bacterium, on one day, the MWRA would have reported compliance.” Id., at (d)(vi)(6). As to the' first argument, EPA cites a Ninth Circuit decision holding that a defendant in an environmental case cannot “challenge [its] own sampling results as a means of avoiding liability.” Sierra Club v. Union Oil Co. of Cal., 813 F.2d 1480, 1491-1492 (9th Cir.1987), vacated on other grounds, 485 U.S. 931, 108 S.Ct. 1102, 99 L.Ed.2d 264 (1988). The case stands for something less than EPA contends, emphasizing as it does the unfairness of permitting a defendant to impeach its own reported excursions by claiming sampling error, thereby “creating] the perverse result of rewarding permittees for sloppy laboratory practices.” Id., at 1492. Here the issue is not whether the results-are bad because of sampling error, but whether they are better than they should have been because the testing method used was more accurate than what the regulations require. The short answer to this (not -by any means specious) argument is that the issue was not raised (at least in a developed form) by the MWRA in its opposition to partial summary judgment. As to the suggestion' that 'any violation established using the enhanced recovery method is de minimis, EPA makes a convoluted argument (that I do not fully follow) that the SWTR’s “historical standard” for unfiltered drinking water is 10 rather than 20 cfu, and that the figure of' 20 cfu was written into the SWTR as an upper 90 percent confidence interval to account for variations in the results of MPN testing. Thus, “[i]f more than 10 percent of a system’s source water samples exceed 20 fecal coliform [cfu] ..., it provides a high degree of confidence that the source water' frequently exceeds the historical standard of 10[cfu].... ” EPA Proposed Finding # 231. Whatever one is supposed to make of this “historical standard,” the fact remains that the SWTR’s fecal coliform avoidance criterion is set at 20, not 10 cfu. EPA also makes the more inviting argument that if fecal coliform concentrations are disregarded, the avoidance criterion defaults to a total coliform count which is not permitted to exceed 100 cfu per 100 ml.- 40 C.F.R. § 141.71(l)(a). That Wa-chusett water failed this standard several times between 1997 and 1999 (using results obtained by the enhanced recovery method) is not disputed by the MWRA. Perhaps associated with the problem of fecal coliform concentrations at the Cos-grove Intake have been numerous instances in which the water reaching the MWRA’s constituent communities has exceeded the Total Coliform Rule (“TCR”) (no more than 5%' of samples may exceed 100 cfu per 100 ml). The data are somewhat difficult to interpret because they are collected on a community-by-community rather than on an aggregate basis. But it is clear (and the MWRA does not suggest otherwise) that one or more communities (and as many as twelve in 1995-1996) have exceeded the TCR threshold on an episodic basis, although compliance has improved substantially since the mid-1990’s as open storage reservoirs have been taken offline. Estes-Smargiassi, 2:147-148; Ex. 391, at 2-8. In most other respects, the MWRA system and its finished water either are presently, or have historically been in compliance with the filtration avoidance criteria. The system has never been identified as the source of an outbreak of waterborne disease. It has for at least five years met the requirement that its water carry a residual disinfectant of at least 0.2 mg/1 that is detectable in at least 95 percent of samples taken from the distribution system. Levels of DBPs (measured in total trihalomethanes) are well below the permissible maximum of 100 ppb despite increases in the amount of chlorine used to treat MWRA water. The efficacy of the MDC’s watershed protection plan, the state’s inspection and reporting requirements, and the system’s redundant capacity are not matters of dispute. Finally, the system provides sufficient chlorination to achieve the required 2-log inactivation of Giardia and 3-log inactivation of viruses. VIII. THE ROAD TO LITIGATION The EPA granted primary enforcement responsibility to the DEP on June 28, 1993. 58 Fed.Reg. 34,583 (1993). The DEP’s drinking water regulations, like the SWTR, require filtration if a water system fails to meet one or more of the avoidance crite