United States
        Environmental Protection
        Agency
Office of Water
(WH-550)
EPA811/R-92-008
November 1992
&EPA ANALYSIS OF POTENTIAL
       TRADE-OFFS IN REGULATION
        OF DISINFECTION BY-PRODUCTS
                                        Recycled/Recyclable
                                        Printed on paper that contains
                                        at least 50% recycled fiber

-------

-------
                             Analysis of Potential Trade-offs In
                          Regulation of Disinfection By-Products
                                John E. Cromwell, III & Xin Zhang
                                    Wade Miller Associates, Inc.

                                         Frank Letkiewicz
                                        Abt Associates, Inc.

                                     Stig Regli & Brace Macler
                               U.S. Environmental Protection Agency
 1.0     Introduction

        Executive Order 12291 requires the preparation of a Regulatory Impact Analysis (RIA) on all new
 major federal regulations.  The U.S. Environmental Protection Agency (EPA) guidelines for performing
 regulatory impact analysis state:

               The goal of regulatory impact analysis is to develop and organize information on benefits,
               costs, and economic impacts so  as to clarify trade-offs among alternative regulatory
               options.1

        Potential trade-offs between microbial and disinfection by-product risks are technically complex
 and fraught with uncertainties.  In addition to meeting the requirement for a Regulatory Impact Analysis,
 the research effort reported in this  paper is intended to establish a systematic analytical framework for
 appreciating these factors.  Explicit methodology for assessing the technical potential for risk-risk trade-
 offs is developed (termed, the Disinfection By-Products Regulatory Analysis Model,  [DBF-RAM]).  The
 strategies used to cope with complexities and uncertainties in developing the DBP-RAM are explained.
 Results are presented and discussed in light of uncertainties, and in light of the analytical requirements
 for regulatory impact analysis.

        Technical  complexities stem from  several competing  relationships between microbial and by-
 product treatment objectives that are inherent  in  water treatment processes.  These relationships  are
 described in Section 2.

        There is current  exposure to both microbial and disinfection by-product risks. Neither of these
 is well-characterized in  baseline data.   Moreover,  the baseline  is changing.  Section 3 presents an
 assessment of available baseline data.

        Section 4  presents a statement  of the specific research objectives and of the overall approach
pursued in this analysis.   The overall objective is to bound the  range  of possibilities implied by  the
competing treatment relationships in terms of the potential for trade-offs between microbial and by-product
risks under different regulatory scenarios.

-------
        Section 5 describes the analytical approach that has been adopted in developing the DBF-RAM
 and the supporting rationale.  Profiles of the changes in microbial and by-product risks associated with
 different regulatory scenarios are generated through Monte Carlo simulation of the competing treatment
 relationships. Specific methodology and assumptions of the DBF-RAM are presented. Section 6 presents
 results and discussion. Section 7 presents summary observations relevant to the research questions posed.
 2.0     Competing Relationships In Microbial and By-Product Control

        Exhibit 1 presents an overview of the key factors involved in the competing relationships which
 result in exposure to microbial contaminants and disinfection  by-products at the consumer's tap.  As
 illustrated in the diagram, these factors can be sorted into four groups:
        o
        o
        o
        o
raw water quality;
treatment combinations applied;
treatment process control strategies applied; and
distribution system characteristics.
2.1.1 Raw Water Quality

        The effectiveness of treatment technologies to control pathogens and disinfection by-products is
affected by the interaction of multiple influent water quality parameters.  Many water quality parameters
affect both microbial and by-product treatment simultaneously (e.g., TOC, UV absorbance, temperature,
pH, alkalinity, and hardness).  Other parameters such as bromide and pH  can strongly influence the
formation of one type of disinfection by-product versus another.

        Chemical disinfectants are oxidants which  destroy or inactivate  microorganisms by oxidation.
They also react with naturally occuring dissolved organics (grossly represented by TOC ~ Total Organic
Carbon) and with inorganics (such as bromide) to produce disinfection by-products, some of which may
be carcinogenic. Higher doses of disinfectants are required in high TOC waters to achieve a desired level
of microbial kill because some of the disinfectant is consumed by organic compounds comprising the
TOC.  But, higher disinfectant dosages lead to higher disinfection  by-product formation unless TOC  is
removed before the disinfectant is applied.  Coagulation and filtration processes, granular activated carbon,
and membrane filtration remove TOC.  However, such processes  may also  shift the chemical balance
towards more brominated by-products, depending upon the bromide concentration  and the amount of
disinfectant applied.

        The chemical reactions that kill microorganisms and those that form halogenated by-products will
proceed to  greater or lesser extent of completion depending upon  the pH (often correlated to alkalinity and
hardness);  and, at a faster or slower rate depending upon the temperature of the water.  Microbial kill  is
increased at higher temperature but so is disinfection by-product formation.   At different levels of  pH,
different groups of by-products are favored. Trihalomethane formation is favored at  high pH, while
trichloroacetic acid formation is favored at low pH.

-------
 2,1.2 Treatment Processes Applied

         Of the possible treatment processes indicated in the diagram in Exhibit 1, some (e.g., coagulation
 and filtration) are appropriate mainly in surface water systems. While there are some groundwaters that
 also use these treatments (e.g., in lime softening), surface waters generally present a far more complicated
 treatment problem where the conflict between the competing objectives (controlling for both pathogens
 and disinfection by-products) is much more pronounced.

        Surface waters contain protozoa whereas groundwaters (that are not under the direct influence of
 surface  waters)  may  only contain viruses  or bacteria.   Since protozoa are much more  resistant to
 disinfection than bacteria or viruses, higher levels of disinfection are needed  in surface waters than in
 groundwaters. Because of this feature, the analysis presented in this paper focuses exclusively on surface
 water systems.  The number of factors involved is so great, that an excessive amount of analysis would
 be required  to assess every  conceivable type of  treatment situation.  By focusing on the surface water
 situation, which is likely  to be the most highly  constrained, alternative regulatory scenarios can be
 subjected to the toughest test. Whatever by-product control levels are achievable in surface water systems
 without unacceptable trade-offs are probably achievable in generally less constrained groundwaters without
 unacceptable trade-offs.

        The nature of  treatment for microbial and disinfection by-product  control is complex in that
 several treatment steps may be added in succession to achieve successively greater levels of control.  Thus,
 there is not a single solution to these treatment problems, but rather a menu of choices. This menu feature
 adds  a considerable degree of complication to the evaluation of regulatory alternatives because impact
 assessment requires a method for  forecasting these complex treatment choices.

 2.1.3  Process Control Strategies Applied

       The box labelled "treatment process control strategies" in Exhibit  1 is perhaps the most important
 part of the diagram.  In striving to meet any  set of microbial and by-product treatment targets, the plant
 operator  has a broad array  of choices.  The following control options are available:

                              control TOC
                       •       control pH
                              select type of disinfectant
                              control disinfectant dose
                       •       control contact time

       In addition to influent water quality parameters such as TOC  and pH,  another major  factor
 influencing the extent of microbial kill and by-product  formation is the amount of time the disinfectant
 is in contact with the water.  For any given disinfectant dose, both the level of microbial kill and the level
of by-product formation will increase as the contact time is increased.  Thus, the treatment process may
be optimized through a combination of strategies intended to: 1) change the water quality parameters that
affect the relevant reactions,  2) change the disinfectant or disinfectant dose, or  3) change the amount of
contact time.

       Temperature is one key variable  that is  beyond the operator's direct  control.  However, it is
possible to vary winter and summer strategies to take advantage of the temperature effect.  For example,
a higher disinfectant residual might be needed within the plant in the winter to maintain  the same level

-------
of inactivalion as in the summer. This may be possible without significant increased by-product formation
since by-product formation is slowed at colder temperatures.  Additional flexibility would  be made
available if compliance with by-product MCLs is determined on the basis of the running average of the
last four quarterly samples, as under the current Tl'HM standard.

2.2.4 Distribution System Characteristics

        A final key variable that is generally beyond the plant operator's direct control is the  residence
time in the distribution system — the amount of time elapsing from the moment the water leaves the plant
to when it is drawn from  the consumer's tap.  The residence time may  vary enormously between the
customer nearest to the plant and the customer farthest from the plant.  The resulting microbial and by-
product risk may also vary greatly between first and last customer.  The plant operator must account for
this additional contact time at the plant, adjusting the available controls to simultaneously assure: 1) an
acceptable level of microbial kill at  the first customer's tap, 2) additional microbial kill sufficient to
protect the distribution system from bacterial growth and external contamination  all the way to the last
customer's tap, and 3) compliance with by-product MCLs on average across the total distribution system
(i.e., at the average customer).

        Thus, the  competition  between  microbial and  by-product control  strategies is driven by
performance  standards  that must be  met in the distribution  system  rather than at the plant effluent.
Because there are so many control options available to the plant operator, there is no unique solution to
the control problem — i.e., within each plant, there may be a number of different control strategies that
can meet a given set of performance targets.  Viewing the variability between plants, there is an even
broader range of possible solutions since raw water and distribution system characteristics vary greatly.

2.7-5 Implications of Variability

        The inherent variability in possible treatment and control configurations has two profound effects
on the assessment of prospective impacts of regulatory alternatives.  First, it is extraordinarily difficult to
fully define the baseline, or pre-regulatory condition. This difficulty is a combined result of the fact that
current microbial performance requirements are in the process of being changed through implementation
of the Surface Water Treatment Rule (SWTR) and the Total Coliform Rule (TCR),  and the fact that plants
can meet performance standards with vastly different treatment and control strategies, depending on their
individual circumstances. Also, the recent promulgation of the Lead rule will result in still further changes
in water treatment that influence microbial control  and disinfection  by-product  formation.   Even if a
complete characterization of baseline process settings were available, it would still be extremely difficult
to predict impacts in terms of the increment of change likely to be  induced  by  alternative by-product
MCLs because the process control options are so numerous.  Baseline issues are discussed  further in
Section 3.

        The second major  analytical implication is that for any  set of performance targets, some plants
will be driven to more extreme treatment and control strategies than others due to  the inherent variability
in raw water conditions.  This range of strategies is significant because the competition between treatment
objectives results from side-effects —  inadvertent increases in either microbial or by-product risks ~ that
are inherent in the underlying raw water chemistry.  Many of these side-effects become more pronounced
as control strategies become more extreme.  Thus a regulatory alternative could appear to represent a
happy medium between microbial and carcinogenic risks, on average (i.e., across all plants), but in fact
entail significant trade-offs due to side-effects induced in the extreme  cases.  As a result, a complete

-------
  assessment of the trade-offs entailed in any set of microbial and by-product performance standards must
  incorporate an analysis of the downside risks that could result from side-effects at the extremes. The
  approach to coping with this analytical requirement is via the technique of Monte Carlo simulation as
  described in Section 4. In the remainder of this section, treatment and control strategies are individually
  characterized, including  an  identification of the side-effects potentially associated with  each of them.

  2.2     Microbial Control

         There are two distinct objectives in controlling microbial contamination in water supplies:  1)
  removing or killing pathogens present in the source water, and 2) preventing contamination of the treated
  water during storage and distribution. The Surface Water Treatment Rule (SWTR) specifies performance
  targets for both of these objectives which must be met simultaneously. The Total Coliform Rule (TCR)
  specifies additional  performance requirements that must be  met in the  distribution system.  Both the
  SWTR and the TCR are  currently being implemented.

 2.2.1  Source Water Treatment

        Microbial contamination in water sources presents a significant treatment challenge because the
 level of contamination can  change sporadically and logarithmically (by several  orders of magnitude).
 Available monitoring protocols are subject to significant sources of sampling error and measurement error,
 making reliable real-time process control impossible.  The only effective strategy is to assure a level of
 over-kill sufficient to meet the most extreme condition at all times.  However,  this  strategy conflicts with
 by-product control objectives.

        In general, individual  unit microbial treatment processes are not 100 percent effective.  While
 coagulation/filtration processes and disinfection can each  remove and inactivate a high proportion of
 organisms present,  neither treatment is completely successful  by itself (unless the source water is
 exceptional).  The best means  of assuring the necessary high  degree of over-kill in microbial treatment
 is through the multiple barriers approach, combining watershed protection, coagulation/filtration processes,
 and disinfection.  In the  SWTR,  the multiple barriers approach is advocated with focus on a  "target
 organism," Giardia lamblia.

        Giardia is used as a target organism in the SWTR because it has been implicated as causing the
 most waterborne disease outbreaks of any specific organism and Giardia  cysts are much more resistant
 to disinfection than bacteria and viruses.  Control of Giardia therefore implies significant risk reduction
 for bacteria and viruses that are responsible for the most notorious and lethal forms  of waterborne disease
 (typhoid, cholera, hepatitis, etc.).  Control of Giardia. however, may not provide adequate protection from
 Cryptosporidium.  Crvptosporidium is also transmitted in the form of cysts,  but  these are  much more
 resistant to disinfection than  Giardia cysts.

       The SWTR  imposes a filtration requirement  for  all  surface waters unless the  system has an
effective watershed control program and the source water is of exceptional quality. Associated with the
filtration requirement are design and operating conditions specified by the state, and performance standards
for effluent turbidity.   In combination,  these requirements are believed  to represent reduction  of the
influent concentration of Giardia cysts by two-and-one-half orders of magnitude (2.5 "logs"). (Note: plants
that do not employ a sedimentation step are assumed to achieve only a 2.0 log reduction of Giardia cysts.")

-------
        In addition to the filtration and turbidity requirements, the SWTR also specifies the need for plants
with sedimentation and filtration to provide an additional half an order of magnitude reduction (0.5 "log")
in Giardia cyst concentration  through  the provision of adequate disinfection.   The adequacy of this
additional disinfection is determined via equations, as follows:

        Total Log Reduction = filtration removal credit + X logs of inactivation

        Where: X = /{disinfectant dose, contact time, temp, pH}

        Methods to  predict the level of inactivation of Giardia achieved through disinfection  (C-T
calculations) have been developed. Also, a dose response function exists for Giardia (derived from human
data) which has received some empirical validation through study of actual waterborne disease outbreak
events.2 Knowledge sufficient to develop comparable risk assessment methodology for Crvptosporidium
does not exist, so it remains an unknown at present.

        The filtration and disinfection requirements of the SWTR provide for a minimum 3-log total
reduction in Giardia cyst concentrations in surface waters by the time the water reaches the first customer
in the  distribution system.  Based on risk assessment,  this level of reduction was intended to provide
protection equivalent to a giardiasis incidence rate (endemic rate) of less than 1 infection per 10,000
persons per year in  systems with good source water quality. The SWTR Guidance Document3 also
stipulated that systems with source waters containing higher than average levels of contamination should
provide an additional log of inactivation through disinfection for each additional  log of Giardia present
in the  source water.  This guidance is difficult to implement because practical methods for measuring
Giardia cysts and for determining  whether they are viable or infectious to humans are not yet  available.
It is significant that, at this time, the criterion for achieving higher levels of treatment for poorer source
water quality is not a requirement, but stipulated only as guidance.   Recent Giardia occurrence data4
indicate that many more systems may require such additional inactivation than was believed necessary at
the time the SWTR was promulgated. These data are discussed further in the baseline analysis  presented
in section 3.

        Surface water systems meeting  the  requirements of the SWTR are typically not expected to
experience epidemic outbreaks of giardiasis or of any other waterbome disease. However, recent data on
Giardia occurrence5 indicate that plants having extremely poor source water quality, and only meeting
the minimal requirements of the SWTR (i.e., not providing treatment according to the guidance) may still
be vulnerable to outbreaks or significant levels of endemic  illness.  Another key  point is that  provision
of additional logs of disinfection  in  accordance with  the SWTR guidance (e.g., replacing chloramine
contact time with chlorine contact time) can result  in significant by-product formation when there is also
a high level of TOC in the water.

2.22  Protection from Contamination During Distribution and Storage

        The SWTR  specifies  that at least a trace level of disinfectant residual be maintained  in the
distribution system to contain "colonization" and "regrowth" of bacteria which might have survived the
multiple treatment barriers. Even if less than one-in-a-million bacteria survives, it  can conceivably attach
to the wall of a pipe in an area where deposits -- potentially containing trace nutrients — have  collected.
Such colonization can ultimately  result in the presence of  significant numbers of bacteria in localized
areas.  The disinfectant residual is intended,  in-part, to limit such growth.

-------
        The other function of the  disinfectant residual is to kill microorganisms that may enter the
distribution system through breaks, leaks, cross connections, or other violations of its integrity.  The
absence of a residual indicates when such contamination may be occurring. This requirement to maintain
a disinfectant residual complements the Total Coliform Rule (TCR) which specifies a monitoring protocol
that must be followed to regularly sample the distribution system for coliform bacteria — an indicator of
either  inadequate disinfection or a violation of system  integrity.  Under the SWTR, monitoring for
disinfectant residual is  required at  the same locations where coliform  samples are taken.  The TCR,
promulgated in 1989 and currently being implemented, imposes more stringent coliform monitoring than
previous requirements and a more  sensitive threshold for determining a violation.  Many systems are
expected to compensate by increasing the level of disinfectant residual they maintain in the distribution
system, with a likely side effect of some increased by-product formation.

        Like source water microbial treatment, protection  from contamination during distribution and
storage is not  an exact science.  Several competing relationships enter into the picture.  First, the extent
to which the disinfectant can remain active all the way to the ends of the distribution system is, in part,
a function  of  the amount of TOC in the water.  The higher  the TOC, the higher the disinfectant dose
required. The  higher the disinfectant dose, in the presence of high TOC, the greater will be the by-product
formation.  Secondly, as described  earlier, chlorine — the most commonly used disinfectant — is more
effective at low pH, whereas corrosion control treatment required to comply with the lead rule results in
higher pH  in distributed water.  Higher disinfectant dosages are required to compensate for higher pH,
increasing  formation of certain by-products such as trihalomethanes.

        Adding to the significance of these potential side effects, there is some evidence which suggests
that levels  of over-kill implied by the SWTR and TCR distribution system requirements may not be as
protective as had previously been believed.  Recent epidemiological evidence  on this topic is reviewed
in the baseline discussion presented in section 3.

2.3 Disinfection By-Product Control

        The chemistry of disinfection by-product formation has been theoretically defined and empirically
described for trihalomethanes and haloacetic acids with equations6 of the following form:

        DBFs  =/{[Cy, contact  time,  [TOC], UV absorbance, [Bf], temp, pH, alkalinity, hardness}

        All disinfection by-product control strategies have potential side effects which pose certain risks,
some more than others. The most practical strategy is to limit formation through precursor removal and/or
through the use of alternatives to chlorine disinfection.

23.1  Adjusting Pre-Disinfection Practices

        A first instinct in attempting to lower DBF formation is to relocate the point of disinfection, or
to discontinue the practice of pre-disinfection — disinfection prior to settling or filtration. Such relocation
permits the coagulation and filtration steps of the treatment process to remove a substantial amount of the
TOC in the water before the introduction of chlorine.  The extent of such removal and its significance for
reducing by-product formation will  largely depend upon the water quality characteristic to the plant and
the extent to which clarification  processes are optimized.

-------
        There are compromises in moving the first point of disinfection further into the treatment train.
Abandonment of pre-disinfection may reduce the disinfectant contact time prior to the first customer and
make it necessary  to add storage  to the end of the treatment process as a means of making up the lost
contact time under the SWTR, especially if the guidance formula is followed.  Plant operators also may
be reluctant to give up pre-disinfection because of its benefit of sometimes improving coagulation or taste
and odor control, and of keeping the filter from becoming fouled with microbial growths.

233, Use of Chloramines

        In complying with the present Total Trihalomethanes MCL of 100 ug/1, it appears that many plant
operators  opted to  switch to chloramines in lieu of abandoning pre-disinfection.  It is estimated that as
many as one-third of the large surface water systems (> 10,000 population) that are subject to the TTHM
standard use chloramines.  Prior to  the TTHM standard as few as five percent of these  systems are
believed to have used chloramines.   The low cost of making such a change, the  effectiveness of
chloramines in reducing TTHMs and their ability to persist as a residual in the distribution system makes
this an attractive option.

        Chloramines are formed through the  addition of ammonia which suppresses many of the by-
product formation  reactions while preserving some, albeit weaker, disinfecting capability.  When by-
product control problems are only moderately constraining,  ammonia may be added after a brief period
of contact with the free chlorine.  Allowing the free contact time preserves some of the benefits of pre-
disinfection.  In more severely constrained cases, ammonia is added from the start. A side effect of the
switch to  chloramines  is that the weaker disinfectant may not be capable of providing the level of log
reduction  required  at the first customer under the  SWTR and/or recommended in the SWTR guidance.
Some plants may have to re-evaluate their use of chloramines in complying with the  SWTR.

        Use of chloramines also entails potential side effects on microbial protection of distributed water.
Through chemical  interactions,  the ammonia may eventually become a source of nitrogen for bacterial
growth and may promote nitrification which can lead to poor tasting water. Chloramines may also offset
the need to remove TOC to some extent, leaving some available carbon in the water. If corrosion control
happens to involve iron pipes  and the addition  of phosphate  inhibitors, the biochemical setting for
microbial growth in the distribution system may be further established. Coupled with an inherently weaker
disinfectant, there is potential for negative side effects of a chloramines DBF control strategy on microbial
control  in distributed water.  However, the fact that the chloramine residual may be more persistent in
some systems holds the possibility that chloramines may  provide more bacterial protection than chlorine
where maintenance of residual is a problem.  Microbial regrowth is a known problem in  many systems
using chloramines presently.  To compensate, some systems have increased  chloramine residuals to the
point where they may not meet potential MCLs for chloramines.

233  Optimization of Precursor Removal

        In  the succession of likely treatments available  for by-product control,  the plant operator will
probably turn next to optimization of TOC removal through fine-tuning of the  coagulation and filtration
processes.  Removal of TOC directly reduces the potential  for by-product formation.  Coagulation and
filtration are very effective, and depending  on source  water quality, may be relatively  inexpensive
compared to other treatments such as addition of granular activated carbon (GAC) or membrane filtration.
Much of the added expense is due to increased sludge disposal. In some plants, high alkalinity  could
make enhanced precursor removal very expensive due to the need for pH reduction.

-------
        Optimized precursor removal has two potentially significant side effects.  First, the lower TOC
 makes  it possible to meet  the disinfectant residual requirement of the SWTR with much lower free
 chlorine concentrations leaving the plant.  However, the reduction in TOC may have a perverse effect of
 allowing the system to have less C-T (less inactivation) prior to the first customer, conceivably increasing
 the risk from Giardia in poorer source waters, unless additional inactivation is provided as recommended
 in the SWTR guidance document.  Under the minimum requirements of the SWTR, attainment  of a level
 of log-reduction at the first customer greater than or equal to the 1/10,000 risk level may be resulting
 incidentally from the need to overcome the competing effect of TOC in meeting the residual requirement.

        The second potential side effect of optimizing TOC removal is a change in the chemical balances
 with  respect to the different classes of by-products.  Removal of organic precursors does not reduce
 bromide concentrations.  As TOC  removal increases, the formation of brominated by-products, some of
 which may be more carcinogenic, may be favored depending upon  the molar ratios of bromide, free
 chlorine, and organic precursors.

 23.4 Use of Ozone

        Modifying pre-disinfection practice, adding ammonia, and optimizing TOC removal are relatively
 inexpensive, conventional technologies that are likely to be tried first.  The next threshold is switching
 to alternative technologies that are relatively novel and relatively more  expensive.  The first of these is
 the use of ozone  as a primary disinfectant.  Ozone is an  extremely powerful oxidant.  From a process
 perspective, it can be very beneficial (depending on source water quality) in achieving a  high level of
 microbial kill  and in enhancing the performance of coagulation processes, sometimes with enhanced
 precursor removal.  Because ozone is so reactive, however, it does not  leave a detectable trace residual
 in the distributed  water.  To provide  a measurable residual, chloramines or chlorine are generally added
 at the end of the plant.

        There  are numerous potential side effects associated with the use of ozone. First, ozone is so
 powerful an oxidant that it produces very small organic fragments that can evade coagulation and filtration
 processes, ending up  in the distribution system in a form that is readily assimilable to bacteria.  The
 increase in  assimilable organic carbon (AOC) can lead  to significant baterial growth in distribution
 systems.  Some of these bacteria may  be pathogenic  and especially harmful to immune compromised
 individuals. The biochemical setting for bacterial growth may be further enhanced with the  addition of
 ammonia as a  nitrogen source.

        In  addition to  the microbial side  effects, the oxidizing properties of ozone also  result in the
 formation of entirely  different mixes  of other chemical by-products which could be carcinogenic.
Toxicological  research on ozone by-products  is in infancy; many unknowns remain in this  area.   A
 potential remedy to minimize these side effects, employed in Europe, is to follow ozone treatment with
 a biological activated carbon filter. This provides a means of biologically removing AOC while simulta-
 neously removing organic precursors to by-product formation.  It represents another technology shift,
 however, as well as another level of expense.

        Another concern of ozonation is the extent to which bromate is  formed.  Bromate appears to be
 a potent carcinogen.  At the  moment, there is insufficient understanding  of the rate at which it is formed
during ozonation and the extent to which its formation can be controlled  by chemical treatment processes
such as pH adjustment or ammonia addition. Occurrence data are also slight due to limitations in present
 analytical methods of detection.

-------
  233 Use of Chlorine Dioxide

         Another alternative disinfectant is chlorine dioxide. Although there is concern about health effects
  caused by exposure to chlorine dioxide and its by-products (chlorite and chlorate), its use in preventing
  or limiting other disinfection by-products may make it attractive in certain water quality situations.  Unlike
  chlorine,  chlorine dioxide is a very effective  disinfectant at high pH.  It does not react by itself with
  organics to form trihalomethanes or haloacetic acids.  Also, it apparently does not react with bromide to
  form bromate.  Systems with high levels of pathogens, TOC, and bromide in their source water which also
  have high hardness and alkalinity and use lime  softening, may have difficulty in simultaneously achieving
  low by-product levels and compliance with the SWTR guidance. Use of reducing agents such  as ferrous
  chloride,  to convert chlorine dioxide to chloride may make use of chlorine dioxide a possible means of
  controlling disinfection by-products in such systems.

  23.6 Use of Granular Activated Carbon or Membrane Filtration

         Depending upon the source water quality, attainment of low levels of disinfection by-products may
  ultimately require deep bed Granular Activated Carbon (GAC) (following conventional filtration processes
  and the use of alternate disinfectants) or membrane filtration.  Due to the expense of GAC and membrane
  filtration,  these treatments are likely to be selected as a last resort compliance option.

         When GAC is used, application of chlorine as a residual disinfectant for the distribution system
 should be delayed until after the water's passage through  the GAC filter. This  strategy optimizes the
 efficiency of biological removal of by-product precursors and  prevents desorption of chlorinated by-
 products that might otherwise be fqrmed if chlorine were added prior  to GAC. Although GAC can
 remove by-products by adsorption, the frequency of carbon regeneration required would not be  practical.
 Chlorine, if used, should therefore be added after GAC.  Ozone, on the other hand, can be used prior to
 GAC to enhance removal of by-product precursors and AOC,  and to provide substantial disinfection
 (depending upon the ozone demand in the water).

        There are, however, several side  effects of GAC.   Deep bed carbon following conventional
 filtration processes may not remove significant levels of pathogens still remaining.  Thus plants with high
 pathogen levels in their source water and currently predisinfecting prior to sedimentation may need to
 install substantial disinfectant contact time following GAC to prevent significant increases in risks from
 protozoa to populations near the first customer.  Since GAC will significantly reduce disinfectant demand,
 aslant would be able to have very low levels of disinfectant dose and  still  maintain a residual in the
 distribution system. While use of GAC would likely decrease risk from pathogenic  bacteria that could
 grow  in the distribution system,  it could  increase risk from pathogens  in the source water if less
 disinfectant or contact time is used prior to the first customer.

        Membrane technologies can remove greater than 90 percent of disinfection by-product precursors
 and essentially all pathogens which might be present  in the source water.  Thus, there is little or no
 downside risk concerning pathogens with this by-product control technology. It is however, the most
 expensive approach. Solid waste and water wastage could increase costs  even more.  In surface waters,
 membrane  technology would have to follow other filtration processes in most cases to prevent  clogging
 of the membranes.

        Both  GAC and membranes are subject to the  same uncertainties as  previously mentioned for
optimized coagulation/filtration precursor removal.  Depending largely upon the bromide concentration,

                                               10

-------
there is potential for a shift in the mix of resulting by-products towards more brominated and more
carcinogenic compounds.  The shift is likely to be considerably less significant in terms of overall risk
for most waters, however, because of the much greater levels of precursor removal achieved  by these
technologies, especially membrane filtration.
3.0    Baseline Assessment ofMicrobiat and By-Product Exposures And Risks

       The existing exposure to microbial and by-product risks is shaped by microbial performance
requirements (applicable to all systems) and an MCL of 100 ug/1 for Total Trihalomethanes (applicable
to systems serving more than 10,000 persons) which have both been in place for over a decade.  In
addition, as noted in the foregoing, the SWTR and TCR, promulgated in 1989, have begun to set changes
in motion in the level of microbial control being achieved from the first customer to the last. Inescapable
changes in by-product exposures are also occurring as a result of these changes.

       As discussed  above, it  is  virtually impossible to predict  the  precise nature of the process
adaptations that are presently taking place  due  to the enormous variability from one case to the next and
the broad range of process control strategies available. It is possible to develop some characterization of
the baseline that existed before these changes were put in motion, however, from data collected prior to
these latest adaptations.

3.1    Microbial Contaminant Exposure And Risk

       As mentioned in Section  2, the analysis presented in this paper focuses on surface water systems,
since these systems are believed  to be the  group most highly constrained by the competing relationships
between  microbial and by-product control.  More specifically, the analysis in this paper will focus on
filtered surface water systems for which there are more data with which to characterize influent levels of
contamination with the target organism, Giardia.

       In the SWTR regulatory impact analysis  (RIA) prepared in 1989,7 data available at that time8
were used to derive an assumption of 3.3 Giardia cysts per  100 liters as an average influent concentration
in filtered surface water systems.   Based in part on data collected by the  American  Water Works
Association,9 an average of 3.33 logs of total  reduction at the first customer was assumed to represent
pre-SWTR treatment practices in filtered systems.   Combining these two crude assumptions,  via the
dose/response function for Giardia  infection,10 with  an estimated 133,560,000 population exposure in
filtered surface water systems yielded an estimated pre-SWTR baseline incidence rate of 27,839 infections
per year  (2  per 10,000 population), assuming no additional inactivation is  achieved in the distribution
system.  It was then projected that  the  SWTR would increase the total  log  reduction achieved  in such
systems to an average of 4.5 logs at the first customer, due mainly to improved efficiency of filtration
(since average turbidities were expected  to drop from 0.7 MTU to 0.3 NTU), but also due to some systems
increasing their level of disinfection. EPA estimated a post-SWTR incidence rate of 1,882 infections per
year in filtered systems (1 per 100,000 population).

       The minimum 3-logs of total reduction required in the SWTR was believed to represent less than
a 1/10,000 giardiasis risk level at the first  customer for systems with good source water quality (i.e., <  1
cyst/100  liters).  It is important to note  that predicted  infections  are  a worst case representation of
predicted illness since many infected people are  asymptomatic. The rationale supporting the use of  a
1/10,000 giardiasis risk level as  a regulatory target is further explored by Macler and Regli.11

                                               11

-------
        The SWTR Guidance Document recommends an additional log of inactivation be applied for every
 additional log of influent Giardia in order to maintain  the 1/10,000 risk level at the first customer.  As
 indicated by the assumptions used in the SWTR RIA, it was believed that most plants would exceed the
 minimum requirements through unproved filtration and additional disinfection.   It was further believed
 that systems meeting a 1/10,000 endemic incidence rate would be below the threshold at which epidemic
 outbreaks occur. The true extent of outbreaks of giardiasis and of other forms of waterborne disease is
 subject to great uncertainty.  The common gastrointestinal symptoms are so easily associated with other
 causes that it is suspected that much waterborne disease goes unreported.   It is known, however, that
 outbreaks are not reported in systems which both filter  and disinfect except when there is an identifiable
 failure in one of these two barriers.

        In a study published  in  1991, LeChevallier,  et.al.12 collected comprehensive data including
 measurements of influent Giardia concentrations and filter effluent  Giardia concentrations, as well as
 computations (via CxT equations) of additional inactivation achieved through  disinfection at the first
 customer.  Eliminating results where  the Giardia concentrations were estimated on the basis of the
 detection limit of the monitoring technique, the LeChevallier study provides 62 complete sets of data,
 representing 46 different plants.

        The LeChevallier study  employed new,  state-of-the-art  monitoring  techniques for Giardia.
 Corrected for recovery efficiency (.48) and conservatively estimated cyst viability (.13) (and, dropping data
 points computed on the basis of the  method detection  limit rather than actually measured), the results
 indicate an average influent  concentration of 203 cysts per 100 liters.  The median is 89 cysts per 100
 liters. These results are considerably higher than the average contamination levels of 3.3 cysts/100 liters
 assumed to exist in such systems  in the SWTR regulatory impact  analysis.

        In addition to noting the  generally higher  level of influent Giardia  implied by  the results,  the
 LeChevallier study presents an analysis computing the minimum total log reduction necessary at each plant
 in order to produce an effluent concentration of 0.0007 cysts/100 liters, equivalent (based on the same
 dose/response function employed here and in the SWTR RIA) to the 1/10,000 population risk level at the
 first customer sought by the SWTR. The conclusion is  that the plants studied would have to achieve, on
 average, a 5-log total reduction — compared to the 3-log minimum required in the SWTR to achieve, on
 average,  the 1/10,000 risk level at the first customer.  This difference is consistent  with  the two-log
 difference between the average influent Giardia concentration found by LeChevallier and  the average
 influent concentration assumed in the SWTR RIA.  An  apparent contradiction is raised, however, by the
 fact that the average of the total (winter) log-reductions calculated to be actually  achieved across the 62
 data sets is 7.0 logs, whereas the average Giardia concentration at the first customer is 0.0016 cysts per
 100 liters. Despite having a  very  high average log-reduction,  the average Giardia concentration remains
 significantly higher than the  1/10,000 risk level.

        Exhibit  2 provides some further illumination through a plot of influent Giardia versus total log
 reduction achieved for the 62 data sets. A critical underlying feature of the pre-SWTR baseline, made
 clear by the random scatter pattern in  this plot, is the fact that  there appears to be  no correlation between
 the influent Giardia concentration and  the total log reduction achieved. The diagonal line  through the
points represents the formula prescribed in the SWTR Guidance Document; points to the right are plants
 achieving more  total log reduction than needed to meet  the 1/10,000 risk target and points to the left are
plants achieving less  total log reduction than needed to  meet the 1/10,000 risk target.
                                                12

-------
         The lack of correlation between the influent Giardia concentration and the total log reduction
 achieved could be attributable to numerous factors: 1) the lack of monitoring capability for Giardia and
 appropriately adjusted treatment; 2) attempts to comply with the current TTHM MCL (Significantly, many
 of the plants to the left of the line are using chloramines.); 3) attempts to comply with current standards
 for coliforms and state chlorine residual requirements (e.g., Ten States Standards); or 4) attempts to meet
 other treatment objectives served by pre-disinfection.  Whatever the reason, much of the 7.0 logs of total
 reduction that is achieved, on average, does not translate into reduced risk of microbial contamination.

         The LeChevallier study reveals that it is the unique relationship between the influent Giardia level
 and the total log reduction achieved at each individual plant which matters rather than the average level
 across all plants. Indeed, it is made clear that  such simple arithmetic as was performed  in the SWTR RIA
 (in the absence of data such as that provided  by LeChevallier) is subject to significant error.

         Performing simple arithmetic based  upon assumptions of average influent concentrations  and
 average total  log reductions may produce very deceiving results.  For example, if the average influent
 concentration (203 cysts/lOOL) and average total log reduction (7.0) of the 62 data sets in the LeChevallier
 study are applied to the same exposed population used in the SWTR RIA, a pre-SWTR endemic incidence
 rate of 353 infections per year is  the result (3 per 100,000 population).

         If, in  contrast, the 133,560,000 persons served by filtered surface water plants are distributed
 equally across  the 62 pairs of influent Giardia  and total log-reduction estimates, the results indicate a pre-
 SWTR disease incidence rate of 2,801,396 infections  per year (2 per 100 population).  While this rate of
 endemic incidence is not beyond belief, given the problems of reporting and the similarity of symptoms
 to those of other diseases,  it  is probably higher than what most knowledgeable observers would offer as
 a best estimate.  If it is  instead  assumed that only  the first  10  percent of customers are exposed to
 minimally treated water - the rest receiving water having substantially more logs of inactivation through
 contact with residual disinfectant in the distribution system — then the estimate of pre-SWTR incidence
 of infection would be 280,000 per year.

        The clear implication of the LeCevallier data is that the minimum 3-log reduction requirement of
 the SWTR does not appear sufficient to reliably achieve the target 1/10,000 risk level in all systems. The
 higher levels of total reduction specified in the  SWTR Guidance Document appear to be necessary in order
 to assure meeting the target risk level.

        Epidemiological evidence confirms the suspicion  that the extent of waterborne disease may be
 greater  than previously realized.  Bennett, et al present analysis of data  collected by the Centers for
 Disease Control (CDC) which suggest that about 70,000 of 940,000 cases of waterbome disease per year
 are attributable to giardiasis.13

        Another study published in 1991 provides evidence suggesting significantly greater incidence of
 waterborne  disease  than previously  suspected.  Payment14 conducted  a  15-month tracking study  in
 Montreal comparing large samplings of families using reverse osmosis treatment units and families using
 tap water from a water system using very poor source water but apparently meeting the SWTR criteria.
Treatment included coagulation, sedimentation, filtration, and residual disinfection.  His findings indicate
a 35  percent greater incidence of gastrointestinal illness among tap water users.   In  interpreting these
results,  it is impossible to determine if the increased  incidence is  attributable to inadequacies in source
water treatment or due to inadequacies in the protection provided from contamination during storage and
distribution.  In either case,  the results indicate more residual microbial risk than previously believed,

                                                13

-------
 although the significance of these findings as they pertain to systems with different source water qualities
 and levels of treatment still needs to be addressed.

 3.2     Disinfection By-Product Exposure And Risk

         Despite the fact that an MCL for Total Trihalomethanes (TTHMs) has been in effect for over a
 decade, there is no resulting  national data base with which to characterize exposure because the EPA
 Federal Reporting Data System records only compliance status, not monitoring results.  Many of the large
 water utilities (those serving more than  10,000 persons) do have records, however, and they have been
 polled on  several occasions by the American Water Works Association (AWWA).

         In a 1987 effort sponsored by the AWWA Research Foundation, the Metropolitan Water District
 of Southern  California (MWDSC) conducted a national  mail survey of water systems serving more than
 10,000 persons.15 Completed questionnaires were received from 910 respondents.  This consisted of 286
 surface water systems, 243 groundwater systems,  127 purchased water systems, and 254 mixed source
 systems. Quarterly sampling data on TTHMs were received from over half of the respondents covering
 1984, 1985,  and  1986. MWDSC summarized the data in terms of the cross-utility medians (i.e., half of
 all utilities higher; half lower) of the maximum, minimum, and mean results obtained by individual utili-
 ties  in each quarter. The summary, presented in Exhibit 3, shows that while utilities are achieving levels
 in a 30 to 44 ug/1 range, on average, half of them have maximum values higher than 65 ug/1 in the third
 quarter. Five percent of utilities reported having been required to make at least one public notification
 regarding violation of the TTHM standard.  Groundwater  systems had  markedly lower TTHM levels,
 averaging around 10 ug/1.

        Thirteen  percent of respondents to the MWDSC  survey indicated they had changed chlorine
 dosages to comply with the TTHM standard. Sixteen  percent indicated they had moved the point of
 disinfection.  Eleven percent of all respondents indicated use of chloramines. Among users of large lake
 and flowing surface sources, 25 percent reported use of chloramines.  Improved clarification was reported
 by  five percent  of respondents.  Very  few other TTHM compliance adaptations were reported by
 significant proportions of the respondents.

        In  1989,  the AWWA Research Foundation sponsored a survey to establish the first phase of a
 standing Water Industry Data Base (WIDE).16 The effort covered all utilities serving more than 50,000
 persons. The Phase I WIDE contains data for 442 surface water treatment plants (a second phase of the
 WIDE project has extended the data base down to systems serving more than 10,000 persons and includes
 over a thousand records).

        The Phase I WIDE data for plants that presently filter shows that 65 percent of softening plants
 make at least some  use of alternate disinfectants  (56 of the 65 percent use chloramines) compared to 37
 percent for plants that do not soften (25 of the 37 percent  use chloramines).  For softening plants, the
 median of the average annual TTHM concentrations was 26 ug/1 for 27 plants using chloramines versus
 60 ug/1 for 19 plants using chlorine. For plants that do not soften, the median of the average annual
TTHM concentrations was 36 ug/1 for 43 plants using chloramines versus 45  ug/1 for 113 plants using
chlorine.  These  data suggest  that softening plants are more tightly constrained by the current TTHM
standard than non-softening plants.

        Among  the filtered surface water plants  which practice softening,  those using chloramines
answered yes to the  practice of pre-disinfection 91 percent of the time while those using chlorine answered

                                                14

-------
yes only 54 percent of the time.  Similarly, in filtered surface water plants which do not soften, those
using chloramines answered yes to the practice of pre-disinfection 91 percent of the time compared to 80
percent for those using chlorine.  These data suggest that chloramines may be being used as a means of
clinging to some of the benefits of pre-disinfection while also maintaining lower levels of TTHMs than
if chlorine were used.

        Other noteworthy relationships in the Phase I WIDB data are that plants using chloramines have
higher influent TOC concentrations, higher average temperatures, and longer distribution system residence
times than plants using chlorine. All of these factors, which cause higher TTHM formation, have probably
influenced the shift to chloramines.

        In 1989-90, the AWWA Disinfection Committee performed a survey of disinfection practices17
which drew responses from 283 respondents, mostly (96%) systems serving more  than 10,000 persons.
About half were predominantly surface water systems and the other half were predominantly groundwater
systems. Of 186 respondents (surface and ground) describing measures taken to comply with the TTHM
MCL of 100 ug/1, the top categories  of responses were:

               44%    no changes needed
               24%    moved first point of chlorination downstream
               23%    ceased prechlorinating
               20%    decreased prechlorination dose
               19%    add ammonia after some contact time
               16%    improve coagulation by increasing pH
               11%    modify coagulation by other method

        Of these respondents, 39 utilities reported problems as a result of modifications made to attain
TTHM compliance. The top categories of modifications producing operational problems were:

               56%    add ammonia after some contact time
               44%    moved first point of chlorination downstream
               36%    improve coagulation by decreasing pH
               36%    ceased prechlorinating
               28%    decreased prechlorination dose
               18%    modify coagulation by other method

        Of the systems reporting problems, the categories of problems  reported were as follows:

               15%    increased heterotrophic plate counts in treated water
               21%    increased heterotrophic plate counts in tap water
               13%    coliforms in  distribution system

        For  the 134 predominantly  surface water systems  that responded to the question regarding
treatment modifications to meet the TTHM MCL of 100 ug/1, the responses were as follows:

               20%    moved chlorination point
               19%    ceased prechlorination
               16%    decrease prechlorination
               16%    add ammonia after some contact time
                                              15

-------
               16%   improve coagulation
               13%   change preoxidation

       The results of the AWWA Disinfection Survey appear to confirm that compliance with the current
TTHM  standard  can be achieved  with relatively simple and  inexpensive  treatment  modifications.
However, evidence of side effects on microbial treatment objectives are also apparent.

       Although the three studies discussed above provide evidence of progress made in complying with
the TTHM MCL of 100 ug/1, assessment of the baseline cancer risk posed by the remaining by-product
exposure requires data on the concentrations of the individual compounds, encompassing not only  the
individual trihalomethanes but the individual haloacetic acids as well. Such data does not exist in great
abundance.  The  two such data sets available have been utilized to produce high and low estimates of
baseline cancer incidence (computed from a lexicological perspective; i.e., assumed exposure factored by
estimated dose/response relationships), summarized in  Exhibit 4.

       The first data set has been collected by the EPA Technical Support Division (TSD)18  and is
probably biased towards worst cases which were the original focus of the EPA investigations.  Results are
summarized in Exhibit 4 in the column  labelled "High Case Estimate." As a point of comparison,  the
TSD data indicate an average TTHM concentration in systems serving more than 10,000 persons of 84
ug/1, roughly twice the average levels reported by such systems in the AWWARF survey and the  WIDE
survey discussed above. The high case cancer incidence estimates are therefore considered a high case,
being based on this data set.

       Cancer incidence estimates are presented in two forms, based on maximum likelihood estimates
of the dose-response relationship and based on a 95th percentile estimate of the dose-response relationship.
WhUe the 95th percentile is the basis that has been used in most previous assessments of drinking water
cancer risks, the  microbial risk assessment methodology provided for assessing incidence of giardiasis
infections is based upon a maximum likelihood procedure.  Thus, the maximum likelihood estimates are
a better basis for comparison. As shown in Exhibit 4, the total baseline cancer incidence in the high case,
based on the maximum likelihood estimates, is on the order of 159 cases per year (0.4+2.4+20+136-
=158.8). Several significant footnotes must accompany this result.  First, it is significant that all but three
of these cases  are attributable not to trihalomethanes, but to haloacetic acids.  The greater impact of
haloacetic acids is a result of the use of the maximum likelihood procedure and the higher cancer potency
factors attributed to dichloroacetic acid and trichloroacetic acid than to the THMs.   On a maximum
likelihood basis, the cancer  risk  from trihalomethanes becomes  less significant.  A second footnote
concerns the fact that cancer risk estimates are presently available for only two of the haloacetic acids.
No risk factors are available for the brominated haloacetic acids.  A third footnote is that TCAA  and
dibromochloromethane are tentatively being classified by EPA as Class C (possible) carcinogens, rather
than  Class B  (probable) carcinogens.    As a  conservative assumption, this  analysis  has  assigned
carcinogenic potential equal to that of Group B.  Finally, it is noted that of the total incidence, 20 cases
are projected in systems serving fewer than 10,000 persons.

        The Low Case Estimates presented in Exhibit 4 are based on a study by Krasner performed by
the Metropolitan Water District of Southern California (MWDSC)19 in which data on concentrations of
individual compounds were collected at the clearwell prior to entry into the distribution system.  Higher
concentration levels could be expected if the samples had been collected in the distribution system.  The
median TTHM concentration of the plants surveyed in this study was 25 ug/1, somewhat below the levels
typifying the other surveys reviewed above.  If these  data were assumed to represent a lower bound of

                                                16

-------
 exposure, the total baseline cancer incidence, as indicated in Exhibit 4, would be 40 cases per year, on
 a maximum likelihood basis.

        A recent "meta" analysis of epidemiological data20 has produced estimates that as many as 10,000
 cases of cancer per year may be attributable to chlorinated drinking water.  There is considerable
 controversy among epidemiologists on the appropriateness of applying "meta" analysis on the data that
 were used, and on the interpretation of the results. The results of the study differ significantly from the
 estimates yielded by the above toxicological approach (exposure/dose-response) to estimating baseline
 cancer risk. The analysis in this paper has been developed using the toxicological approach.


 4.0     Research Objective and Approach

        Clearly, assessing the impacts of alternative regulatory options for disinfection by-products is not
 straightforward. Additional difficulty results because the  objective of this research is to devise a means
 of analyzing the problem in a summary form that  permits drawing generalized conclusions to serve as a
 basis for national  level analysis. At least  four features of the problem contribute significant additional
 complications in elevating analysis to an aggregate level:

        The baseline is in motion. Water systems are presently in the process of complying with SWTR
        and TCR requirements subject to the existing by-product constraint embodied in the TTHM MCL
        of 100  ug/1.  Analysis of microbial and by-product exposures under the currently defined
        regulatory regime indicates that ~ under various interpretations — both types of exposure have a
        wide range of possible risk.

        The two types of exposures are dynamically  interrelated in a manner which is not completely
        revealed  through simple  intuition.   Inadvertent side-effects are  accentuated under various
        combinations of extreme conditions, making the average condition a deceiving benchmark. The
        central tendencies are not as important as the interactions between the tails in assessing potential
        trade-offs.

•       There is a factorial number  of combinations  of raw water,  treatment, process control, and
        distribution conditions that present an unmanageable number of conceivable "extremes" to be
        considered in analyzing potential interactions  of the exposures.  There are, in fact, many times
        more combinations than there are public water systems. There is also a factorial number of con-
        ceivable combinations of regulatory options to consider (e.g., TTHMs 100, SWTR 3-logs; TTHMs
        50, SWTR 4-logs; etc.).

•       Finally, there are significant uncertainties inherent  in the current state-of-the-art of both exposure
        and risk assessment pertaining to both types of health risks. Many of the key areas of uncertainty
        can be addressed through "what-if" analysis to test the sensitivity of results to uncertain estimating
        procedures.   However, such extensive  "what-ifs"  add  to the factorial burden  of  analysis.
        Moreover, in addition  to areas of uncertainty, there are several areas where so little is known of
        potential side-effects that even provisional quantification of exposure and risk is infeasible (e.g.,
        formation  and health  risk from brominated haloacetic acids, ozone by-products,  changes in
        bacterial populations in the distribution system, Cryptosporidium. etc.), requiring them to be
        treated as complete unknowns at this point.
                                               17

-------
       Analysis of by-product regulatory options is a unique problem. Most drinking water regulations
present much simpler analytical problems, where national impacts can be fully characterized in a way that
is representative of the overall national exposure and risk. Because of the factorial dimensions of the by-
products problem, and the extensive range of uncertainties and unknowns characterizing the available
information and understanding, analysis of by-products regulation cannot be approached  in the same
comprehensive manner.  Even if the outstanding uncertainties and unknowns were resolved, the factorial
dimensions would still place practical limits on the extent of the analysis.

4.1 Defining A Manageable And Meaningful "Core Problem "

       Recognizing the practical constraints, the specific research approach adopted in this analysis is to
define the nature of the trade-offs by focusing on a core problem; a manageable segment of the problem
where the most conflicting interactions and the most  significant population  exposures  are likely to be
found. A manageable and meaningful core problem can be defined by two  means. First, some of the
factorial dimensions may be collapsed into a smaller number by focusing on those relationships which are
perceived to be most influential in, causing trade-offs to occur.  Fortunately, this approach is feasible
because, although there  are many dimensions conceivably involved, there is sufficient understanding to
suggest which are the strongest relationships (i.e., where extreme treatments, and therefore trade-offs, are
most likely).

       Surface water systems are generally much more likely to have to resort to extreme treatments than
groundwater systems. They also  serve a much larger  segment of the overall population. Though there
are locations (e.g., FL, CA) characterized by  high TOC or  high bromide groundwaters, unless they are
groundwater under the direct influence of surface water, they are  not vulnerable to protozoa and therefore
do not need as much disinfection. Any set  of regulatory requirements achievable by most surface water
systems without adverse trade-offs Will probably also be achievable by most groundwater systems. Exhibit
5 illustrates the relative proportions of population exposure  between surface and ground water systems.

       In addition  to collapsing factorial dimensions  to  define a manageable and meaningful core
problem, there must also be a strategy to maintain an accurate summary representation of the variability
introduced by site-specific conditions. Site-specific variability affects the analysis in two ways. First, the
multidimensional baseline configuration of factors affecting microbial and by-product risks is unique in
every local situation. Second, the baseline condition is changing in every local circumstance in response
to implementation  of the  SWTR and TCR.  The various conceivable  combinations of site-specific
conditions must be factored into the analysis in a way  that is somehow representative. This site-specific
variability can be analyzed statistically through  the technique of Monte Carlo simulation, described in
section 5.0.

4.2 Coping With Uncertainties and Unknowns: Scenarios And Place-Holders

        The second part of the research strategy for defining a manageable and meaningful core problem
is to collapse additional dimensions of complexity  through the use of scenarios and place-holder
assumptions.  If  the above-described complex of interrelated factors and processes that simultaneously
contribute to by-product and microbial risks is envisioned as one great  multi-dimensional matrix, the
scenarios and place-holder assumptions provide a limited selection of cross-sections through it. A factorial
number of other cross-sections are feasible, but the selected views are intended to offer a meaningful
starting point — a benchmark from which other "what-if' excursions  can be launched.
                                                18

-------
        Even within the core problem category of surface water systems, the number of conceivable
 combinations of microbial and by-product risk targets that could be considered is factorially large. The
 approach to coping with this large number of combinations is to bracket the range of microbial protection
 through evaluation of two scenarios: a SWTR scenario, and an Enhanced SWTR scenario (ESWTR). The
 SWTR scenario assumes that all surface water systems will be constrained to meet a total reduction target
 for Giardia of only 3-logs at the first customer.  Additional log-reduction for poorer quality source water
 that is recommended in the SWTR Guidance Document is assumed to be ignored. The ESWTR scenario
 presumes an enhanced SWTR is enacted which  would convert the prescription for  additional total
 reduction presently made in guidance into a mandatory requirement. Due to lack of adequate monitoring
 techniques, an easily implementable enhanced SWTR does not appear to be a near-term possibility, but
 is nonetheless  an  important  benchmark.  Both  scenarios assume the SWTR  disinfectant residual
 requirements, the coliform rule, and the lead rule are met in the distribution system.

       Another significant challenge in defining a core problem is presented by the fact that quantitative
 exposure and risk assessment is feasible for some of the microbial and by-product risks (albeit, subject
 to significant uncertainties), but is not feasible for others, leaving them as pure unknowns. The estimable
 aspects of the problem include risks associated with Giardia. trihalomethanes, dichloroacetic acid and
 trichloroacetic  acid.   Exposure and risk  assessment models  have  been pieced together for  these
 contaminants.  The assumptions required are regarded as place-holders. As implied by the name, place-
 holder assumptions are intended to be continually refined as better information becomes available.  They
 can  also  be systematically varied as the object of "what-if" analysis to examine the sensitivity of the
 results to uncertainties associated with them.

       As a result of its use as  a  target organism  in the SWTR, methods to  predict  the level  of
 inactivation of Giardia  achieved through disinfection (C-T calculations) have been developed.  Also as a
 result of its recent popularity, a dose response function exists for Giardia.

       Chemical equations with which to predict the formation of disinfection by-products have been the
focus of much recent research effort.  As a result, equations are available with which to predict formation
of trihalomethanes and  haloacetic acids as a function of the concentrations of organic precursors and other
raw water and treatment conditions. The research in this area is far from complete. The equations utilized
in the present analysis are clearly  place holders.   Although the equations have been validated within
numerous individual plants, their validity when extrapolated outside the boundary conditions of those cases
remains a question mark which can only be removed through broader validation studies. The equations
used to predict  the brominated species are particularly prone to such error. The equations utilized for
haloacetic acids base predictions on concentrations of THMs rather than directly as a function of water
treatment and precursor parameters. The caveats associated with all  the  formation equations have been
documented in the users manual which accompanies the USEPA Water Treatment Plant Model.

       The cancer risks attributable to three  of the four trihalomethanes (chloroform,  bromodichloro-
methane (BDCM), and  bromoform) have been estimated according to standard protocols used by EPA in
establishing MCLGs. The final carcinogenic status of dibromochloromethane (DBCM) is unclear at this
time, however.  As a conservative place holder assumption, the cancer risk estimate for BDCM has been
used for DBCM. (NB:  the estimated risk for BDCM has been increased since this analysis was finalized.)

       Cancer risk estimates have been developed for the chlorinated haloacetic acids (dichloroacetic acid
(DCAA)  and trichloroacetic acid (TCAA)) following  standard protocols used  by  EPA in establishing
MCLGs. Although a cancer risk estimate has been  developed for TCAA. its carcinogenic status has not

                                               19

-------
 been finally established. As a conservative place holder assumption, the cancer risk estimate for TCAA
 has been used in the analysis as though it were a B2 carcinogen.  While formation equations exist for five
 of the haloacetic acids (mono-, di-, and trichloroacetic acids and mono- and dibromoacetic acid), cancer
 risk estimates  have been  developed only for  DCAA and TCAA (MCAA does  not appear to  be a
 carcinogen). Thus the present analysis predicts concentrations of two brominated haloacetic acids, but
 does not quantify associated risks. Finally, to further condense dimensions, summary results are presented
 in  terms of hypothetical  Total Trihalomethane  (TTHM) MCL  alternatives and  hypothetical Total
 Haloacetic Acid (THAA) MCLs, despite the fact that calculations are performed for individual compounds.
 Results for  individual compounds are provided in  appendices.

        In focusing on those contaminants for which exposures and risks are estimable, it is possible to
 begin to quantify the core problem and learn as much about the  potential risk-risk trade-offs as possible
 given current information.  Insights regarding the potential significance of yet inestimable, or unknown,
 risks  can be obtained by further segmenting the  analysis into  two additional scenarios.  Since more
 unknown risks  are associated with the use of alternate disinfectants than with the use of chlorine, the
 analysts has been structured into two scenarios:  with and without alternate disinfectants.

        The without scenario permits evaluation of the limits of risk reduction possibilities for Giardia.
 trihalomethanes, and chlorinated acetic acids without introducing other unknown by-product or microbial
 risks associated with alternate disinfectants.  The  with scenario permits evaluation of the potential for
 further  reductions in  Giardia,  trihalomethane,  and chlorinated acetic  acid  risks at the expense of
 introducing  other unknown risks. Results for the with scenario can be further subdivided to differentiate
 between a scenario allowing only chloramines  and a  scenario  allowing both  chloramines and ozone.
 Comparison of the results of with versus without scenarios should  make  it possible to evaluate the
 potential significance of the unknowns (i.e., how many plants might switch  to alternate disinfectants as
 a means for meeting TTHM and THAA regulatory requirements).  Cost estimates can also be  made to
 compare complying with and without alternate disinfectants.

        The guiding principle in this attempt to develop exposure and  risk assessment models is the
 scientific method of inquiry. Models are constructed using the best information available as place-holders.
 Hypotheses  are substituted as place-holders  for missing information  or understanding.  Scenarios are
 structured to bound the range of possibilities introduced by complete unknowns. The models are exercised
 under a range of conditions to examine the plausibility of results and to explore sensitivity to the major
 uncertainties and unknowns. Using the exposure and risk assessment models in this way, significant areas
of research need can be identified and prioritized in a systematic manner. The models can be refined as
 research results provide new information and understanding.


5.0    Analytical Methodology

       There are four steps to the overall methodology for analyzing potential trade-offs.

        •       step 1: model the mix of raw water and  distribution system characteristics in a manner
               that is representative of the variability in site-specific conditions;

        •       step 2: predict the compliance choices that plants  will make in attempting to meet a given
               set of regulatory targets, given their raw water and distribution system characteristics;
                                               20

-------
         •       step 3: predict the plant effluent (1st customer) and distribution system exposures that will
                result from the given characteristics and compliance choices;

         •       step 4: predict the microbial and by-product risks that will result from these exposures.

         Exhibit 6 illustrates these steps and summarizes the different analytical strategies that have been
 applied to each. These are discussed below.

 5.1 Step 1: Characterize Raw Water and Site-Specific Variability

         Step 1 requires a means of modelling influent variability that is representative of conditions nation-
 wide. As described in earlier sections, both microbial and by-product control techniques have side effects
 which are exacerbated under extreme combinations of influent water quality conditions.  It is these side
 effects that are responsible for the existence of trade-offs between microbial and by-product treatment
 objectives.  The analytical objective for a national exposure assessment is therefore to build a model that
 permits analysis of side effects resulting from  treating extreme raw water conditions.

        Analyzing a core problem focusing on surface waters provides a practical strategy for reducing
 the scope of this analytical need to just one category of raw waters. The most fundamental relationship
 in the core problem is that between influent TOC, influent bromide, and influent Giardia concentrations.
 Although many other  factors are included in the model, as described below, these three are assumed to
 be  the fundamental influences  - the "drivers."  High-TOC/high-Bromide/high-Giardia source waters are
 likely to be much more susceptible to adverse trade-offs than low-TOC/low-bromide/low-Giardia source
 waters.  In addition, for any given combination  of influent concentrations, the extent of the trade-offs will
 be  a  function  of the interplay between influent conditions and  alternative combinations of regulatory
 constraints.  There are three different regulatory targets: a microbial target at the plant effluent (or 1st
 customer); a microbial target in the distribution system (the disinfectant residual requirement); and a by-
 products target in  the distribution system.  Significant trade-offs can result  depending  upon how the
 regulatory regime specified at these three points is constrained. Modelling the core problem under a range
 of assumptions about influent conditions and regulatory alternatives permits an  assessment of the possible
 extent to which trade-offs may be encountered.

        Even when limiting analysis to focus on the core problem, a difficulty still arises in coping with
 the enormous variability in site-specific conditions. A first instinct in  approaching  an analysis  of by-
 product regulatory options might be to perform case studies of plants representing  different types of raw
 water conditions.  Upon closer scrutiny of the factors involved, however, it becomes  apparent that an
 unmanageable number of different cases would be required in order to capture the diversity that exists in
 water supplies.  Moreover, significant leaps of faith would be needed for determining the influence that
 individual cases should have in the overall analysis and in the ultimate selection of appropriate regulatory
 options.

        Another approach would be to convene a panel of "experts" to render a judgement regarding the
effects of different regulatory options.  Experts provide a concentrated source of case studies, drawn from
their collective experiences. While this feature is very efficient, it is likely that such a panel would also
have difficulty  weighting different cases in arriving at the type of generalized conclusion required in this
unusual rulemaking.  The panel of experts approach has been used in development of EPA regulatory
impact analyses involving straightforward contamination problems such as Flouride, VOCs, lOCs, SOCs,
and Radionuclides.  It  was also used - with much greater difficulty - in analyses of the Surface Water

                                               21

-------
 Treatment Rule and the Lead and Copper Rule, since these rulemakings involved optimization of entire
 treatment and distribution systems as a function of variations in raw water quality.  The treatment and
 distribution optimizations implied by the by-products regulatory options are several times more complex
 than previous rulemakings.

        Another analytical option is to characterize site-specific variability through statistical modelling
 (simulation) of the variability in raw water and distribution system characteristics. The diversity of site-
 specific conditions is represented through the technique of Monte Carlo simulation applied to the core
 problem.

        The key question is: what percent of the time can the most extreme combinations of circumstances
 be expected to co-occur?  Monte Carlo simulation provides an explicit methodology through which the
 frequency of such co-occurrences can be  assessed. In this procedure  (fully described in a companion
 paper),21 numerous parameters that are critical to predicting microbial and by-product treatment perfor-
 mance are characterized as probability distributions. Next, interdependencies between related parameters
 (e.g., TOC and UV; pH and alkalinity) are examined  and joint occurrence relationships are statistically
 defined. A significant assumption made in this procedure is that influent Giardia occurrence is indepen-
 dent of all other influent parameters,  (Since the approach to simulating  influent Giardia levels is integral
 to the approach taken to the Giardia risk assessment, it is discussed  under step 4, below).

        In the actual simulation step, individual and joint distributions are sampled randomly to create 100
 different simulated raw water quality profiles which, depending on the quality of the data and the statisti-
 cal analysis, should be representative of the variability that will confront surface water systems. In other
 words,  the critical analytical step of weighting, or assigning a frequency to different  combinations of
 extreme raw water and distribution system conditions to form a  representative composite picture is handled
 by this  procedure, based on an explicit statistical analysis. In effect, the statistical approach provides a
 means of considering 100 case studies at a time,  that are  randomly selected via a procedure that should
 be broadly representative of the inherent variability between  individual sites, and thus of the extent to
 which extreme conditions co-occur.

        The characterization of influent raw water quality and distribution system residence time, was
 made possible through the extensiveness of the AWWA Water  Industry Data Base (WIDB) for fundamen-
 tal parameters such as pH,  hardness, alkalinity, temperature, and many others.  There are important
 weaknesses, however, affecting the three driving variables in the core problem. The WIDB contains only
just enough data on TOC (84 observations), to permit a statistical approach.  Furthermore, the WIDB does
 not provide data on bromide or Giardia. Thirty bromide observations from a national study by Krasner,
 et al were utilized to define a statistical distribution for influent bromide concentrations.22  (An ongoing
 study by the AWWA Research Foundation is developing new data on TOC and bromide which may be
 used to improve current place holder assumptions.)  The LeChevallier study of Giardia concentrations
 (representing  46 plants), discussed above, was  utilized to model  the variability in  influent  Giardia
 concentrations between plants. Another data base was used to  model  the day-to-day variability of Giardia
 cysts within individual plants  (Hibler23).  A  companion paper describes these  statistical  analyses
 developed to characterize Giardia occurrence as a function of both these considerations.24

        The provision of explicit methodology  for assessing site-specific variability  enables  explicit
 inspection of the uncertainties inherent in  this key contributor to  the overall exposure assessment. All
distributional assumptions are regarded as place-holders in the same  sense discussed earlier. The Monte
 Carlo framework  can support sensitivity  and what-if analyses  that  can shed light on  the potential

-------
 significance of major areas of uncertainty.  Exhibit 7 presents a three-panel summary of the simulated
 values TOC, bromide and Giardia as compared to the respective observations in the data sets on which
 the simulations are based. Exhibit 8 presents plots of simulated influent TOC versus simulated influent
 bromide and versus simulated influent Giardia.  The assumption of total  independence between these
 "driving" variables is evident in the  scatter patterns.

 5.2 Step 2: Predict Compliance Choices

        The second step, predicting the compliance response to individual site specific circumstances in
 terms of treatments deployed and process control settings, is far more troublesome. It represents another
 source  of factorial possibilities that  must  somehow be contained within  a  manageable analytical
 framework.  As mentioned earlier, there is often no unique solution to the treatment and process control
 problem.  There may be several different combinations of strategies that will serve to achieve compliance
 in any given circumstance. As noted in Section 2.0, this feature has two profound analytical implications:
 1) it is virtually impossible to construct a model that can precisely replicate the existing baseline condition;
 and 2) the extent to which the most extreme compliance options (e.g., GAC or membrane filtration) are
 relied upon depends on the extent to which less extreme options (e.g., use of alternate disinfectants) are
 adequate, given raw water conditions.

        A baseline "calibration" of the Monte Carlo simulation procedure was undertaken during prototype
 development of the modelling methods employed here.  Results of that effort, summarized in the next few
 paragraphs, are documented in more detail in  a separate paper.25  The AWWA WIDE  is unique  in that
 it enables a correlation of existing treatment practices with  data on influent water quality, plant effluent
 quality, and distributed water quality.  The uniquely comprehensive coverage of the WIDE data across
 the treatment train suggests the potential to obtain a baseline "calibration" of the Monte Carlo procedure.
 Such calibration might be achieved by adjusting treatment and  process control assumptions to produce
 model results that replicate the statistical  distributions exhibited in the  WIDB  finished water and
 distributed water data.  In  the end, it became clear that process control assumptions regarding factors not
 included hi the WIDB (e.g., chlorine dosages, basin contact times, and points of chlorination) are more
 important to matching  the  observed distributions of  effluent  and distributed water quality than the
 distributional assumptions applied to the raw water parameters.  It was shown to be possible to achieve
a number of different calibrations through a number of different combinations of these fixed settings —
i.e., since  different process control  assumptions could be  used  to achieve  the same target used  for
calibration, one could not  conclude which assumptions would be most appropriate.

        It was possible to match the means of the predicted total trihalomethane and chlorine residual
distributions for 100 randomly synthesized plants to the means of the observed distributions in the WIDB
data under a number of different settings.  The predicted and observed distributions for chlorine residual
were shown not to be statistically different as measured by a Kolomogorov-Smirnoff test.  The predicted
and observed distributions for total  trihalomethanes did not compare favorably, however. Since the
chlorine dosages were fixed by calibrating to the chlorine residual (via a chlorine decay equation), they
reflected the log-normal distribution of the chlorine residual data. The total trihalomethane levels predict-
ed  from these dosages were then  also  log-normal.   However, the observed  distribution for total
trihalomethanes in the WIDB data is normal — a reflection of  the fact  it is a regulated parameter.  In
reality, the plants in the upper tail of the log-normal TTHM distribution predicted by the model would
have undertaken various process refinements to meet the current TTHM standard. Such refinements were
not accounted for in the model.
                                               23

-------
       It was concluded that true model "calibration" could eventually require process data at such a level
of detail that the exercise becomes indistinguishable from that  required for case-by-case "validation"
studies. The rule development process might stimulate the development of additional validation through
the release of EPA's Water Treatment Plant Model and accompanying users manual.26  Despite the
inherent  difficulties, some measure of calibration was achieved.  The revised version of  the model
produces predictions that appear to approximate national occurrence data on average and that, starting from
a pre-TTHM standard condition, reflect treatment changes comparable to those that actually occurred to
meet the TTHM MCL of 100 ug/1.  These results are discussed in section 6.0 and confirm that the Monte
Carlo procedure involving  distributional representation of the variability in  raw water conditions is
reasonable, on average.  But, the imperfect match  to observed by-product distributions is an important
source of error inherent in the process control assumptions applied to the simulated raw waters. The tails
of the distributions of chlorine dosages, chlorine residuals, and by-products are ultimately the determining
influences in characterizing what percent  of plants can meet a given microbial or by-product risk target.
Results show that process control assumptions and model equations are over-predicting TTHMs in the
upper tail of the distribution and under-predicting  in the lower tail.  Thus conclusions regarding the
attainability of relatively less stringent TTHM MCLs are likely to be overly pessimistic while conclusions
regarding the attainability of relatively  more stringent MCLs are  likely to be overly optimistic.

       Without a baseline characterization of the configuration and performance of treatments and process
controls, it is not possible to approach regulatory impact analysis in the conventional manner of comparing
pre- and post-regulatory conditions.  The  approach adopted therefore is one of assessing the attainability
of different treatment objectives as if working from a "clean slate."  Microbial treatment objectives are
evaluated in terms of two scenarios, as described  above: 1) the SWTR scenario, assuming  minimalist
adherence to the existing SWTR (i.e.,  ignoring the  advice of the  Guidance Document for poorer quality
source waters);  and 2) the  Enhanced  SWTR scenario  (ESWTR), assuming adherence to the SWTR
Guidance as if it were a firm requirement.  Both the SWTR and the ESWTR scenarios assume disinfectant
dosages sufficient  to meet the distribution system  residual requirements  of the SWTR, as well as pH
conditions assumed sufficient to meet the requirements of the Lead Rule. A full range of alternative by-
product MCLs are evaluated under each microbial scenario.

       The answer to the question of what is ultimately attainable  should  be the same regardless of
whether the baseline is understood.  Similarly, the  trade-offs between the different treatment objectives
can be revealed just as clearly through the  "clean slate" approach to assessing treatment performance under
different regulatory scenarios. At some point it will become necessary to "net out" the treatments already
applied to comply with current standards from the treatments required to comply with alternative standards
in order to pinpoint the regulatory impact.   Fortunately, studies  such as those reviewed in Section 3.0
indicate that most of the adaptations undertaken in response to the current TTHM standard have involved
simple low cost process control modifications (moving points of chlorination, adjusting dosages, etc.) that
would not amount to significant costs  at an aggregate level.  The largest adaptation in evidence is the
indication from WIDE data that perhaps as many as one-third of surface water plants are using  chloramin-
es, which is also relatively inexpensive.

       The "clean slate" approach is analyzed in terms of a With Alternate Disinfectants Scenario  and a
Without Alternate Disinfectants Scenario  (described in Section 4.0). The Without Scenario would not be
expected to bear any resemblance to the baseline treatment configuration in any case, since chioramines
and other alternate disinfectants are already in use.  In both the With and Without Scenarios, a progression
of successively  more extreme treatment  and process control adaptations is  assumed with increasingly
stringent MCLs.  The  progression of treatments  follows a least-cost criteria; the less costly compliance

                                               24

-------
options being tried first before progressing to more extreme and more costly measures. This sequence is
illustrated in the decision tree illustrated in Exhibit 9.  As noted in Section 3.0, there appears to be a
tendency of some plant operators to add ammonia as a means of preserving the  use of pre-chlorination
in the face of by-product control requirements. The least-cost criteria of the decision tree produces results
(Section 6.0) that indicate less extensive use of chloramines in the With Alternate Disinfectants Scenario
than the proportion shown in the WIDE and AWWARF data. Despite this divergence from reality, the
least-cost  sequence, as illustrated in Exhibit 9 is believed  to be an intuitively sound place holder
assumption to assess likely treatment adaptations.  The decision tree is discussed further in a separate
paper
      27
        A related problem arises in the "clean slate" approach regarding the need to make an assumption
about the presence or absence of pre-chlorination  in the starting condition.  As shown in Exhibit 9,
elimination of pre-chlorination is the first step in the compliance decision tree.  It has the potential to
greatly  reduce by-product  formation, but  can also compromise  microbial  treatment under certain
conditions.  For this analysis, the presence  or  absence of pre-chlorination was simulated  as a random
variable based on the prevalence of the practice indicated in the AWWA WIDB which shows about 80
percent  of surface water plants still practicing pre-disinfection as of 1988 (Pre-SWTR). This proportion
seems high by comparison to the Disinfection Survey conducted by the AWWA Disinfection Committee
in 1990 which shows, at most, two-thirds. However, the wording of the WIDB question was somewhat
vague since the term "predisinfection" was not more precisely defined.  Ideally,  pre-TTHM standard
baseline data on the extensiveness of pre-chlorination would be required in order to truly start with a clean
slate. The WIDB estimate is used with the knowledge that it may be on the  high side. Alternatively, it
may be  an accurate reflection of the fact that many plants have followed a different decision tree in their
compliance choices, favoring the introduction of chloramines as a strategy for preserving the benefits of
pre-chlorination.

        In addition to  defining the  likely sequence of treatments resulting from alternative regulatory
scenarios, it is also necessary to specify a wide range of specific process control assumptions.  A long list
of place holder assumptions is required  here, as discussed at length in  a separate paper.28 In general,
the assumptions fall into three categories: best estimates, constraints, and bounding assumptions. Best
estimates are employed in places (e.g., initial dosages and unit process contact times) where the range of
variability is considered reasonably estimable based on engineering judgement and experience. Constraints
are performance requirements that relate to treatment objectives other than microbial or by-product objec-
tives. The need  to simultaneously meet these  other objectives forces more  extreme microbial and by-
product treatment solutions in some instances. The most significant constraints are pH conditions assumed
necessary to comply with the Lead Rule, and maximum disinfectant residual concentrations required to
satisfy the taste constraint for chlorine and a possible MCL of 3.0 mg/1 for chloramines.

        Bounding assumptions were applied at several  key  junctures where it was deemed prudent to
estimate conservatively in order to minimize the chances of under-estimating the potential  for trade-offs
between the competing treatment objectives.  An  important  set of bounding assumptions concerns the
engineering design margins assumed in making treatment choices (described more fully in a companion
paper29).  The CxT equations used to assess compliance with SWTR requirements for disinfection are
applied with a 20 percent over-design factor. Similarly, compliance with by-product  MCLs is  evaluated
with a 20 percent over-design factor. In combination, these  two bounding assumptions are expected to
reflect the upperbound of possibilities for conflicts between treatment objectives, minimizing the possibility
that the conflicts between them will be under-estimated.
                                                25

-------
        Similarly, the effect of seasonal temperature variations on compliance choices and resulting
 exposure estimates has been incorporated into the model via a set of bounding assumptions.  Partially to
 avoid the necessity  of building a much more complex  model capable of simulating seasonal treatment
 adaptations,  and partially to minimize the possibility of under-estimating the potential for trade-offs
 between treatment objectives, the model equations are structured to predict Giardia concentrations on the
 basis of  average winter  temperatures and  by-product  concentrations on the basis  of average annual
 temperatures.  The  two  temperatures are simulated based on the joint  distribution they exhibit in the
 AWWA WIDE data.

 S3  Step 3: Predict Resulting Microbial &. By-Product Exposures

        Step 3, predicting microbial and by-product exposures resulting from the different combinations
 of raw water, distribution system, and treatment characteristics, is accomplished through application of a
 number of chemical equations that have been brought together in a comprehensive modelling framework
 in the USEPA Water Treatment Plant Model.30 Among these are the CxT equations utilized to compute
 disinfection requirements, trihalomethane formation equations, haloacetic acid formation equations, and
 chlorine decay equations.

        All of these  relationships are considered place holder assumptions.  They have all received some
 measure of validation and yet many uncertainties remain.  The major ones are as follows:

 •       The CxT equations were developed to assess inactivation at the first customer.  In the model
        employed in this analysis, they are linearly extrapolated  assuming first order kinetics to estimate
        total log inactivation through the distribution system. Thus, for a given concentration of chlorine,
        pH, and water temperature, if a CT of 300 achieves a 3-log inactivation of Giardia cysts, a CT
        of 900 (which may occur through  a  point in the  distribution system) would be  estimated  to
        achieve a 9-log inactivation. No data are available to assess the validity  of this assumption for
        Giardia inactivation, although  for viruses there appears to be reduced inactivation efficiency
        relative to CT, as the level of inactivation increases.

 •       Giardia has been selected as the target organism for analysis. The rationale for this is described
        in more detail in a companion  paper.31  Since Cryptosporidium is much  more resistant  to
        disinfection than Giardia, and is ubiquitous in surface waters, it may be a more appropria*" *arget
        organism, but it is not sufficiently understood to support all the analytical needs.

•      The trihalomethane formation equations  have been the focus of considerable research and have
       been empirically validated to some extent. A major source of potential error concerns use of these
       equations to extrapolate outside the boundaries  of the raw water and treatment conditions that
       characterized the plants in which  they were originally developed and tested. Although predicted
        results for total THMs seem to fall  within a plausible range of occurrence, results for individual
       THMs, especially regarding the brominated compounds, may have large uncertainties.

•      The haloacetic acid formation equations are at a less refined stage of development. The equations
       predict haloacetic acids as a function of predicted trihalomethane concentrations rather than as a
       direct function of  raw water parameters and treatment conditions.  Much less validation research
       has been performed on  haloacetic acid formation equations. Although results for THAAs seem
       to fall within a plausible range of occurrence, there could be large uncertainties in the predictive
       capability for individual compounds, especially the brominated compounds.  New equations have

                                               26

-------
        recently been developed and are being incorporated into the DBF-RAM for future iterations of the
        analysis.

        While the collection of equations being utilized produces results in plausible ranges, all equations
 are subject to error.  Significantly, extrapolations outside the boundaries of the conditions where equations
 were estimated and verified may be most susceptible to error.  Unfortunately, such results are also most
 likely to represent  the  types  of extreme conditions that are critical to the analysis.  Until  further
 characterization of the error in the selected model equations or replacement with better equations, the
 results presented herein represent best estimates.

        Presently, trihalomethane and haloacetic acid formation equations have been developed only for
 surface water plants that do not soften.  Owing to the different chemical relationships that characterize a
 softening plant, a different set of by-product formation equations is required. EPA is currently developing
 such equations for incorporation  into the Water Treatment Plant Model and DBP-RAM.

        The initial analytical effort has focused only on large (>10,000 population) surface water systems
 that presently filter but do not soften. This portion of the universe represents a substantial proportion of
 the total population exposed to the highest microbial and by-product risk levels. They serve a total of 103
 million persons, equivalent to  64 percent of  the total population served by community surface water
 systems (162 million) and 43 percent of the total population served by all community water supplies (242
 million; not including non-transient, non-community systems).  Under some conditions, surface water
 plants which  soften could be more highly constrained.  This question bears further  investigation.  By
 comparison, the total population served by large (> 10,000) surface water softening plants  is about 20
 million.

        Although  the initial modelling  work has focused on  large (> 10,000 population) surface water
 systems that do not soften, it is believed that results of this analysis can be applied to smaller systems of
 the same type. The raw water characteristics of the large systems represented in the principal data source -
 - the AWWA WIDE — should not be vastly different from raw water characteristics facing  small water
 systems of the same type.

5.4  Step 4: Predict Resulting Microbial and By-Product Risks

        Step  4,  predicting microbial and  by-product  risks resulting from the projected exposures, is
performed through the application of risk assessment models for Giardia and carcinogenic by-products,
respectively.   These models are of the nature of best estimates.  In more precise statistical terms, the risk
models  are based upon  "maximum likelihood" estimates of the dose-response  relationships.   Use  of
maximum likelihood estimates is a departure from  EPA's  usual practice in evaluating carcinogens.
Carcinogenic  risks in drinking  water have more typically been evaluated on the  basis of the upper 95-
percent confidence bound of the  dose-response relationship.  Since risk assessment methods for Giardia
are developed on a maximum likelihood basis, however, the cancer risk has been evaluated on a similar
basis to facilitate comparison. Cancer risk estimates are also presented in summary print-outs for the 95-
percent confidence interval  in order to provide a point of comparison to other drinking water rules that
have been assessed by that  procedure.   However, the maximum likelihood estimates  are considered  the
best estimates. The methods employed in derivation of cancer risk estimates are summarized in Technical
Note 1.
                                                27

-------
        Risk assessment methodology for Giardia has been previously defined32 and has received some
 empirical validation.33 Exhibit 10 presents the results of previous attempts to validate the Giardia risk
 assessment procedure using data from epidemic outbreaks of the disease.  In the analysis supporting
 Exhibit 10, the following assumptions were made:

 •      the dose-response function (developed from data on human  subjects)  is representative for all
        populations;

 •      all cysts recovered by the monitoring technique  are viable;

 •      all viable cysts are infective in humans;

 •      a single cyst  can cause an infection  (the  dose-response function  estimates that only  about 2
        percent will cause an infection upon ingestion, but the point is it only takes one);

 •      infection = illness (i.e., all infections are symptomatic);

 •      cysts recovered by the analytical  technique represent all cysts  which are present;  and

 •      the error  caused by assuming all cysts are recoverable and  infectious to  humans (an under-
        estimate) is  equivalent to the error in assuming that there is 100 percent recovery of cysts by the
        analytical method (an over-estimate).

        These are the same assumptions that were used in the Surface Water Treatment Rule Regulatory
 Impact Analysis prepared in  1989.34 At that time, they were considered to be conservative assumptions.
 The results of the validation studies suggest that the assumptions may not be significantly conservative
 since the  results they generate replicate the empirical data fairly well.

        In approaching the assessment of by-product regulatory alternatives, a number of refinements were
 made in the existing risk assessment methodology for Giardia.  These refinements are fully described in
 a companion paper.35  In considering analysis of waterborne disease risk within a simulation framework,
 there are two important dimensions of variability in influent concentrations: 1) variation between different
 plants, and 2) temporal (day-to-day) variation within individual plants.  The first dimension is important
 for accurately characterizing the risk at an aggregate or national level.  The  second dimension is critical
 to assessing the risk of epidemic occurrence of giardiasis.

        The SWTR  endemic risk target  of one case per 10,000 persons per year  was  believed to be
 equivalent to a level of outbreak risk that would be trivially small.  The recent findings of LeChevallier
 discussed in Section 3.0, indicate that the minimum treatment requirements specified in the SWTR would
 not be sufficient to achieve the target risk level for many systems.  If by-product regulatory alternatives
 further erode the level of microbial treatment provided, the threshold of significant outbreak risk may be
 crossed. The refinement of predictive methods for assessing outbreak risk is therefore important.

        There are two primary sources of data regarding the raw water occurrence of Giardia. Both were
used to support the analysis.   Hibler collected data for 73 plants which includes multiple samples at the
same plants to give an indication of the variation from day to day within the same plant. These data have
been utilized (adjusted by a recovery efficiency factor of two) to develop a statistical characterization of
the day-to-day within-plant variation in terms of a delta negative binomial distribution which is described

                                               28

-------
 in an accompanying paper.   LeChevallier collected data for 46 plants using state-of-the-art monitoring
 methods.  These data have been used to simulate the raw water variability in influent Giardia between
 different plants. Since LeChevallier provides empirical estimates of the recovery efficiency (48%) and
 percentage of viable cysts (13%), these estimates were adopted in the analysis.  All the other pertinent
 assumptions, listed above, from previous Giardia risk modelling efforts are kept the same.

        A few  other best  estimate, place-holder assumptions  have been  added to  the Giardia risk
 assessment methodology. To reflect the day-to-day variability in operations, the level of Giardia present
 in the simulated effluents is assumed to drop by 1-log five  percent of the time.37  In addition, to  the
 primary infections estimated by the dose-response function, a secondary infection rate of 25 percent is
 assumed.  Finally, the outbreak threshold, the level at which outbreaks are considered to be detected and
 reported, resulting in a chain reaction of societal coping costs, is assumed  to be the. point  at which 1
 percent of the population is ill (defined, in  this analysis,  as infection of 1 percent of the population
 exposed in the vicinity of the first customer)  within any 30-day period.   The overall Giardia risk
 assessment methodology is  illustrated in Exhibit 11.

 5.5 Summary of Analytical Methodology

        The overall analytical methodology is summarized in  the diagram in Exhibit  12. The modelling
 apparatus is structured in a "batch mode." A batch of 100 simulated input profiles, characterizing influent,
 distribution system, and some treatment parameters are run through the EPA treatment model as a separate
 batch run for an initial (clean slate status quo) treatment benchmark and again  for each treatment
 configuration indicated by the successive nodes of the decision tree (Exhibit 9). In this way, the same 100
 input files are run as a batch through every one of the conceivable treatment configurations.

        These batch results are then interpreted by a compliance sorting routine.  Compliance with either
 the present SWTR or an Enhanced SWTR is assured by the structure of the scenarios.  The compliance
 sorting routine evaluates  the extent to  which more extreme  treatments are required in order to  meet
 alternative  by-product MCLs.   (For economy  of  presentation, these are summarized  in terms  of
 hypothetical total THM MCLs and  in terms  of  total haloacetic acid MCLs.  Detailed results are also
 generated for each individual compound, however.)

        The compliance sorting routine begins by reading the  100 effluent by-product concentrations for
 the initial, status quo, treatment configuration. For each alternative MCL being evaluated, it selects the
 records for those plants which meet the MCL and saves their output files to the master output file for that
 MCL. It then reads the model results at the next treatment node on the decision tree and performs the
 same selection routine, but  working only on  the remainder of the  100 plants ~ i.e., those that did not
 achieve compliance at the previous  node.  Exhibit 13 presents a two-panel summary of the procedure.
 The flow diagram in  the top panel summarizes the logic.  The bottom panel summarizes the form of the
 output.

        As illustrated in the bottom panel, the sorting routine produces a complete assessment  of the core
problem.  For any given combination of microbial treatment targets and by-product  MCLs, it produces
 an assessment of the extent to which relatively more extreme treatments will be required. These treatment
 deployment profiles are matched to  the appropriate  unit cost vectors  to produce estimates of the total
 national treatment cost of alternative  by-product MCLs for the  subset of the universe being  studied (large
 surface water plants that filter but do  not soften).  In addition to the treatment percentages, the output files
 assembled by the sorting routine provide a  quantitative profile of the exposures to  Giardia. trihalo-

                                               29

-------
 methanes, and baloacetic acids at each benchmark.  The exposures are then converted into estimates of
 disease incidence via the risk models described above.  Taken together, these elements of methodology
 provide a complete basis from which to characterize the trade-offs between regulatory alternatives — i.e.,
 a complete basis for regulatory impact analysis.
 6,0 Results & Discussion

        The results generated by the above-described DBF-RAM methodology are presented in a series
 of tables in Appendix A.  These tables present results in a form similar to the bottom panel of Exhibit 13,
 showing the degree of treatment required to meet a specific combination of regulatory targets. In addition,
 the appendix tables also present quantitative estimates of the microbial and cancer risks associated with
 each combination of regulatory targets.

        Several different tabulations are provided, representing different scenarios. Results are summarized
 in terms of the SWTR Scenario With Alternate Disinfectants versus the SWTR Scenario Without Alternate
 Disinfectants.  Equivalent summaries are provided for the Enhanced SWTR Scenario.

        For each of the scenarios evaluated, two complete sets of tables are provided; one in which by-
 product control is dimensioned in terms of total trihalomethane (TTHM)  MCLs and the other in terms of
 total haloacctic acid (THAA) MCLs.  To facilitate comparisons between hypothetical TTHM MCLs and
 comparable THAA MCLs, tables comparing the two are provided in Appendix B.  Appendix C presents
 tables which document  the associated concentrations of  individual  trihalomethanes  and individual
 baloacetic acids under the different scenarios.

        The results are evaluated and discussed at a summary level in the remainder of this section. There
 are two broad questions which must be  addressed in reviewing these results:

        1)     Are the DBF-RAM outputs a plausible representation of reality?

        2)     What is implied about the trade-offs between microbial and by-product  risks?

 The first question is covered in Section  6.1; the second in Section 6.2.

 6.1 Assessment of Model Performance

        Before examining model results in terms of the potential trade-offs between microbial and by-
 product control objectives, it is  necessary to first compare model outputs to available studies of actual
 conditions in order to assess the realism of model predictions. Such comparisons can be made  in terms
 of: 1) by-product occurrence,  2) treatment adaptations to  meet the  current  trihalomethane standard, 3)
 cancer incidence, and 4) Giardia incidence.  Given the "clean slate" approach outlined in Section 5, the
SWTR With Alternate Disinfectants Scenario, evaluated at a TTHM MCL of 100 ug/1, should demonstrate
a reasonable degree of replication of the pre-SWTR/TTHM MCL=100 conditions reflected in a number
of available studies.
                                               30

-------
 6.1.1  Comparison of DBF-RAM Results to By-Product Occurrence Data

        In the  SWTR Scenario  (with alternate disinfectants) at a TTHM MCL of 100 ug/1, the 100
 simulated plants exhibit a mean TTHM concentration at the average customer of 40 ug/1, with a 95th
 percentile of 78 ug/1.  These results compare favorably with data reported by Krasner et al38 which shows
 a median of 25 ug/1 and a 75th percentile of 59 ug/1, measured at the plant effluent.  It also compares
 favorably with  the AWWARF TTHM survey39 which reported the mean of quarterly distribution system
 samples, averaged across  all plants,  to be in a range of 30 to 45 ug/1.  Similarly, the AWWA Water
 Industry Data Base shows annual means of distribution system samples to  lie within a range of 30 to 60
 ug/1.  In contrast, data collected by EPA at  plants selected to reflect more extreme conditions exhibited
 a mean of 84 ug/1.40

        As  noted  in section 5..3,  the place  holder  assumptions used to predict  haloacetic  acid
 concentrations are based on a relationship to TTHMs rather than on chemical formation relationships.  It
 is important therefore to check the realism of THAA predictions also. Based on only dichloroactic acid
 and  trichloroacetic acid,  average  customer levels predicted in the SWTR  scenario (with alternate
 disinfectants) for a TTHM MCL of  100  ug/1 exhibit means of 10 ug/1 and 8 ug/1, respectively.  95th
 percentiles are 24 ug/1 and 20 ug/1, respectively.  These model results compare favorably to data reported
 by Krasner et al41 which have medians of 6 ug/1 and  75th percentiles of 12 to 15 ug/1, measured at the
 plant effluent.  In contrast, data collected by EPA at  plants selected to reflect more extreme conditions
 exhibits a mean of 39 ug/1.42

 6.1 J Comparison of DBF-RAM Compliance Predictions to Actual Compliance Patterns

        While it is good that mean values are replicated by the modeling apparatus, the true test of validity
 is in the prediction of performance at the  extreme ends of the range ~ the  interactions between the tails
 of the different random variables.  Some indication of this critical aspect of model performance is provided
by a comparison of the  treatment deployment projected by the "clean slate" model results at a TTHM
 MCL of 100 ug/1 versus the treatment deployment that  actually resulted from the current TTHM standard.
The  need to  resort  to higher levels of treatment to meet a TTHM MCL  is driven by the tails of the
distributions  ~  by  the more extreme conditions.  The model results show  the  following treatment
deployment required  to  meet a TTHM MCL  of 100 ug/1 in the SWTR  Scenario  with alternate
disinfectants:

               No Further Treatment                61%
               Eliminate Pre-chlor                   21%
               Eliminate Pre-chlor & Add NH3      18%
       Data on treatment changes made to comply with the TTHM MCL of 100 ug/1 were collected in
the AWWA Disinfection Survey43 and in the AWWARF TTHM Survey44.  Unfortunately, both studies
tabulate the data in a manner which does not permit direct comparison because the treatment percentages
(presented  in  section 3.2) do  not  sum  to  100 percent.   It is nonetheless  possible to make some
comparisons.

       Compared to the model prediction that 61 percent of plants require no treatment changes to meet
a TTHM MCL of 100 ug/1, the AWWA Disinfection Survey reported 44 percent.  The survey included
the option of  moving  the  first  point  of chlorination as well  as  a  second option of eliminating

                                               31

-------
 prechlorination, versus the model algorithm which offers only elimination of prechlorination as the first
 option.  Both surveys indicate that compliance with the present TTHM standard was achieved predomi-
 nantly through changes in chlorination practice; and, at the most extreme, addition of ammonia.  As in
 the model results, more significant treatment modifications were generally not  rquired to meet TTHM
 MCL of 100 ug/1. The model indicates 18 percent of systems had to resort to chloramines versus survey
 estimates of 25 percent (AWWARF) and 16 percent (AWWA).  The AWWA Water Industry Data Base
 provides another data point, showing 25 percent of surface water plants that filter, but do not soften, using
 chloramines.  As discussed in Section 3.2, some plants may be using chloramines as a means of clinging
 to the benefits of pre-disinfection. This may account for survey estimates of chloramines usage exceeding
 estimates produced by the DBF-RAM algorithm.

 6.13 Comparison of DBF-RAM Estimates of Cancer Incidence to Other Estimates

        As  the tables  in  Appendix A indicate, the model results  suggest that both cancer risks and
 microbial risks lie in significant ranges.  The potential for significant trade-offs to occur appears to be
 confirmed.  On a maximum likelihood basis, however, the risk posed by trihalomethanes becomes very
 small while the cancer risk of dichloroacetic acid becomes the predominant concern. DCAA is shown to
 account for most of the total by-product cancer risk.  In the period since the input assumptions used in
 the current model were fixed, the cancer risk factors that EPA believes should be used for the brominated
 THMs have been increased, so this result may change somewhat in the next iteration.

        Within the 103 million population covered by the current  model, DCAA would be responsible for
 38 cases of cancer per year.  Because of the widespread exposure,  this would make DCAA one of the
 most significant carcinogens ever encountered in the SDWA regulatory program. As discussed in Section
5, the DBF-RAM predictions of cancer incidence are based on dose-response relationships derived from
 toxicological research.  There is debate in the epidemiological  literature over suggestions that cancer
incidence attributable  to  disinfection by-products may be  on  the order of thousands of  cases per
year/5'46'47  EPA does not believe that it is appropriate to derive quantitative cancer risk estimates from
the available studies because the weight of evidence is not strong enough to indicate a causal relationship
between exposure to chlorinated drinking water and cancer incidence.

6.1.4  Comparison of DBF-RAM Results to Estimates of Waterborne Disease Incidence

       The Appendix A tables show significant incidence of Giardia infections at the benchmark TTHM
MCL of 100 ug/1 in the SWTR Scenario (with alternate disinfectants).  Within the 103 million population
represented  by the model, 340,000  infections per year are indicated.  These results are two orders of
magnitude higher than  the post-SWTR predictions presented in the SWTR Regulatory Impact Analysis.
As discussed  in  section 3.1, the difference results  from use  of  Giardia  data collected recently by
LeChevallier18 which indicate occurrence levels about two orders  of magnitude higher than that estimated
in the SWTR RIA.  As the TTHM MCL is changed from 100 ug/1 to  25 ug/1, the DBP-RAM predicts a
near doubling over this baseline level of Giardia infections.

       In the present  analysis, all  predicted Giardia infections occur at the first customer which is
assumed to  represent 10 percent of the  total 103 million population exposed hi the  model.  The C-T
equations have been extrapolated to predict additional log-kill achieved in the distribution system.  The
resulting log-kills are sufficiently high that Giardia infections are indicated only at the first customer.  As
described in an accompanying paper by Grubbs. et.  al.,49 use of  the Giardia dose/response model
employed in this analysis includes several new aspects of analysis beyond that developed originally by

                                               32

-------
Rose et. al.50 The same dose/response relationship is utilized, but the daily variation in the influent
concentration is simulated over 365 days rather than based on a single annual average.  In addition, a
treatment failure rate is incorporated by assuming a 1-log drop in the total reduction achieved five percent
of the time.  Also, a secondary infection rate of 25 percent  is assumed.  The collective result of these
refinements is a model which predicts approximately twice the level of infections estimated by using only
the arithmetic mean with the Rose equation.

        In addition to predicting endemic incidence of giardiasis, the modelling framework developed by
Grubbs et. al. also includes an outbreak simulation feature. As shown in the results in Appendix A, the
outbreak risk is  shown to be significant at the current TTHM  MCL, a finding that does not appear to
match reality. This prediction is based, however, on the conservative assumption that infection equals
illness and that infection of one percent of the population at the first customer (versus the total population)
would lead to community awareness of waterborne disease and  formal recognition of an outbreak event.

        On  balance, both the  endemic and epidemic predictions for  Giardia risk are believed to be
plausible considering that there are great uncertainties regarding accurate reporting of disease and outbreak
incidence and considering the uncertainty in the relationship between infection and symptomatic illness.
Reconciling the  model results against  epidemiological evidence on endemic incidence of waterborne
disease31, it  is possible that the target organism methodology  used  in DBF-RAM may be a conservative
estimate of the total microbial risk.
       A flaw in the current Giardia modelling exists in the  evaluation of the  more extreme  MCL
scenarios where ozone is employed to obtain compliance.  The present treatment model assumes that if
ozone is used for primary disinfection, then Giardia inactivation will be achieved to meet  only  the
minimum SWTR  requirements in the SWTR Scenario, or the 1/10,000 risk level in the Enhanced SWTR
Scenario.  This must  be refined in future development of the model.  Due to concerns over unknown
dimensions of ozone by-products and ozone decay, the current framework focuses more keenly on chlorine
and chloramines.  Ultimately, analysis of the "ozone paradigm"52 will have to be fully developed to  the
same level of sophistication as the "chlorine paradigm."

6.2 Assessment of Potential Trade-offs

       As discussed in Section 4, the analysis of trade-offs between competing risk reduction objectives
presents a problem that has infinite dimensions, making it necessary to focus on a smaller core problem
for purposes of analysis.  Analysis of the core problem is intended to illuminate both the nature of the
trade-offs involved and the potential significance of uncertainties  and unknowns.   There  are three
conceivable types of trade-offs to  be evaluated:

•      As by-product MCLs are imposed, the level of microbial protection may be compromised. As a
       result, the reduction in known cancer risks  (THMs and HAAs) may be offset by  increased
       microbial risk.

•      If the level of microbial control is first enhanced  to prevent adverse side effects on the level of
       microbial protection, the increased level of microbial control could result in greater by-product
       formation and higher costs for by-product control.

•      If by-product control  is pursued with the use of alternate disinfectants, the incidence of known
       cancer risks (THMs and HAAs) will be reduced, but unknown cancer risks (e.g., bromate, other

                                               33

-------
         DBFs) and unknown microbial risks (e.g., bacterial growth from increased AOC with use of
         ozone, or reduced inactivation with use of chloramines) may be introduced.  Weighing against
         these concerns, alternative disinfectants may  be less expensive than other advanced by-product
         control strategies.

         A structured  approach to evaluating these three  types of trade-offs is provided through the
 consideration of four scenarios:
 •      The SWTR Scenario evaluates the potential for increased microbial risk when by-product MCLs
        are imposed with the currently required level of microbial control. The decrease in known cancer
        risks may be compared to the increase in microbial risk.

 •      The Enhanced SWTR (ESWTR) Scenario assumes the microbial risk level is fixed at a 1/10,000
        endemic incidence rate for giardiasis.  Alternative by-product MCLs are  evaluated with this
        microbial constraint. Results can be compared to the SWTR Scenario to determine 1) how much
        additional known cancer risk results from the higher level of microbial control; and, 2) how much
        additional treatment cost is entailed in by-product control to compensate for the higher level of
        microbial protection.

 •      Both the SWTR Scenario and the ESWTR Scenario  are evaluated in two variations: 1) with
        alternate disinfectants;  and 2) without alternate disinfectants.  Comparing the results of the two
        variants permits an assessment of the treatment cost savings made possible by the use of alternate
        disinfectants.  This cost savings must then be weighed against the unknown cancer risks and
        unknown microbial risks that may be introduced through the use of alternate disinfectants.  The
        unknown risks cannot be quantified, but the comparison  to cost savings provides a framework
        within  which to consider these unknowns.

 6.2.1 Potential for Increased Microbial Risks

        Exhibit 14 presents a graphic illustration of the process through which by-product controls might
 result in inadvertent increases  in microbial risks.   Coagulation and filtration  have the joint effect of
 removing Giardia and reducing TOC.  Reducing TOC decreases the overall potential for by-product forma-
 tion.  It also decreases the chlorine demand in the water and permits the SWTR disinfectant residual
 requirements in the distribution system to be met with lower chlorine dosages.  While the diminished use
 of chlorine may reduce the overall potential for by-product formation, it may also decrease the total level
 of Giardia inactivation achieved at the first customer, unless compensating treatment is added.

        Reducing TOC  may not seem like a problem since the SWTR requires a minimum of 3-logs of
 inactivation at  the first customer.  New Giardia occurrence data collected by  LeChevallier, however,
 indicate that an average of at least 5-logs  may be required by many utilities in order to meet the 1/10,000
 endemic incidence goal  sought by the SWTR (as discussed in Section 3). Many plants may be achieving
 adequate log-reduction at the first customer as an accidental result of the need to meet the distribution
 system residual requirement in the  presence of high chlorine demand. Removal of TOC to meet by-
product constraints may allow the log-reduction at the first customer to drop back to  the 3-log minimum
which now appears inadequate.  This phenomenon is observed in  the modelling results.

       Exhibit 2 summarizes the LeChevallier data in terms of a plot of influent Giardia versus the total
 log reduction achieved.  The pattern of the data indicate  a complete lack of correlation between the

                                               34

-------
influent Giardia  concentration and the log reduction achieved.  This lack of correlation is perhaps the
result of the dominant influence of the residual requirement, in addition to other factors.  Model results
appear to support this explanation.  The Monte Carlo simulation framework established in the DBF-RAM
is well-suited to analyzing the relationship  between influent Giardia  levels and total log-reductions
achieved at each individual plant. Since the influent Giardia distribution used to drive the simulation was
developed  from the LeChevallier data, some direct similarities to the  LeChevallier findings are to be
expected in the model results.  Unfortunately  a complete calibration  check of the model against the
LeChevallier results  is not possible because  the LeChevallier  data  was  collected in  a pre-SWTR
compliance environment whereas the model is constructed to include the SWTR requirements as initial
constraints. Some similarities are evident, nonetheless, as discussed further below.

       There are many possible reasons why the LeChevallier data indicate a lack of correlation between
the influent Giardia level and the total log reduction achieved. In the over-simplified framework of the
DBF-RAM, a few assumptions dominate this relationship.  Exhibit 15 presents a plot similar to that of
Exhibit 2 showing  simulation results for a scenario that assumes implementation of the SWTR (with
alternate disinfectants)  and a  TTHM standard  of 100 ug/1.   Of 100 simulated plants, only 39 were
determined by the DBF-RAM algorithm to require treatment adjustments in order to meet a TTHM MCL
of 100 ug/1. These 39 points are plotted in Exhibit 15. The plotted points reflect the relationship between
influent Giardia and total log reduction after compliance with these scenario constraints.

       Many of the 39 points in Exhibit 15 appear in a vertical line reflecting the SWTR requirement for
a  minimum  of 3-logs  of  total reduction.   In contrast to the pre-SWTR baseline reflected in the
LeChevallier data in Exhibit 2, some of the points along the 3-log line in Exhibit 15 represent a higher
level of total reduction than in  the pre-SWTR  condition while others represent  a lower level.  In the
simplified model logic, those plants that were not previously achieving a 3-log level would have increased
their level of total reduction to equal 3-logs, while some of those achieving more than a 3-log reduction
would have dropped their level of total reduction to the 3-log minimum, where they could still meet the
residual requirement.  Notably, all the points  aligned vertically along the 3-log minimum are to  the left
of the diagonal line that defines the 1/10,000 risk threshold that the SWTR sought to achieve, indicating
that they are  incurring risks greater than this target level.

       The remaining points plotted in Exhibit 15 show more than  3-logs of total reduction.  Some of
them are above the  1/10,000 risk threshold (to the right of the diagonal)  while others are below it (to the
left of the diagonal). In tracing the source of this result through the model logic, it appears that the reason
these plants exhibit more than the 3-log minimum is due to the additional  SWTR constraint of maintaining
a disinfectant residual in the distribution system. The simulated plants that are achieving more than a 3-
log total reduction at the first customer in Exhibit 15 represent conditions where there  is enough TOC to
require elevated chlorine dosages to overcome the chlorine demand of the TOC in the distribution system,
but yet where the resulting contribution to TTHM formation is not significant enough to require additional
TOC removal.

       Exhibit 16 illustrates the effect of a tighter by-product MCL of 25 ug/1. In this scenario, 83 plants
would be required to change treatment in order to comply.  As shown in the graph, the result is that all
the points are aligned  along the 3-log minimum line.  Those plants  that had more than 3-logs  of total
reduction in Exhibit 15 have been guided by the treatment algorithm into TOC removal strategies which
in turn reduced the chlorine demands that had to be overcome  in order to meet the  SWTR residual
requirement in the distribution system.  As a result, the model predicts that all plants required to change
                                               35

-------
 treatment to meet a  TTHM standard  of 25 ug/1 would" reduce their total log reduction to  the 3-log
 minimum.

        As shown in Exhibit 16, the convergence to the 3-log minimum results in a risk level higher than
 the 1/10,000 target for all of the 83 plants compelled to modify treatment in meeting a TTHM MCL of
 25 ug/1. The resulting change in endemic incidence of Giardia infections will be larger for some plants
 than for others. In particular, there may be some plants for which the co-occurrence of high influent TOC
 and high influent Giardia result in a very large increase in incidence of Giardia infections. This increase
 results from significant drops in chlorine dosages for compliance with SWTR residual requirement after
 TOC removal has been triggered by the need to meet a TTHM MCL.  The increase in incidence of
 Giardia infections has been computed based on the change in  total log-reduction between one MCL and
 another, as computed  by the DBF-RAM model. Results of this analysis are presented in Exhibits 17 and
 18.

        Exhibit 17 plots the cumulative distribution of the 100 simulated plants in terms of the increase
 in annual Giardia infections per 1 million persons when moving from a TTHM MCL of 100 ug/1 to an
 MCL of 75 ug/1.  The plot confirms that  there are  a  few  plants that would face extreme conditions,
 resulting in significant increases in Giardia infection rates.  Nearly 10 percent of plants could experience
 increases of between  1,000  and 10,000 infections per 1 million population.  The top end of this range
 approaches outbreak proportions.  Exhibit 18 shows that the impact of a shift from a TTHM MCL of 100
 ug/1 to a TTHM MCL of 25 ug/1 would produce increases in  Giardia infection rates on this same order
 of magnitude for over 30 percent of plants.  Exhibits 19 and 20  present the same results that are plotted
 in Exhibits 17 and 18 in terms of bar charts that emphasize the impacts at the extreme ends of the  range.

        Ultimately, a  comparison must be made between the increase in incidence of Giardia infections
 and the corresponding decrease in cancer incidence  that might result from alternative MCLs. This is
 achieved by  combining the above analysis  of predicted Giardia infections with predictions of cancer
 incidence resulting from exposure to by-product concentrations that are simultaneously projected by the
 DBF-RAM.  Results  are summarized graphically in  the log-log plots presented in  Exhibits 21 and 22.
 Exhibit 21 plots increased incidence of Giardia infections per 1 million persons per year versus decreased
 cancer incidence per 1 million persons per year for the plants having to change treatment when the TTHM
 MCL is changed from 100 ug/l to 75 ug/1 in a SWTR scenario (with alternate disinfectants).

        Exhibit 22 presents  the same picture for the plants that would have to change  treatment if the
TTHM MCL is changed from 100 ug/1 to 25 ug/1.  The pattern of the points plotted in these diagrams
 indicates that the degree of change in cancer incidence generally spans only one order of magnitude (less
 than one case per 1 million persons per year) while the degree of change in incidence  of Giardia infections
spans several orders of magnitude (from one to more than  10,000 infections per 1 million persons per
year).  This points to  a finding that changes in microbial risk  are more sensitive than changes hi cancer
 risk to treatment changes needed to meet TTHM MCLs.

        Exhibits 23 through 30 present the same series of graphs as Exhibits 15 through 22 except that
 the entire analysis is structured in terms of alternative MCLs for total haloacetic acids. The interpretation
 is identical to that described above.
                                               36

-------
622 Potential to Avoid Increased Microbial Risk

       One obvious direction indicated by the above results would be to strengthen the SWTR before
proceeding with requirements for additional by-product controls. An enhanced SWTR has been evaluated
using the DBF-RAM.  It was  assumed  that the portion of the SWTR  Guidance Document which
recommends an additional log of inactivation for each additional  log of influent Giardia is converted to
a mandatory requirement. If an  enhanced SWTR were in place,  the graphs presented in Exhibits 2,  16
and  17, would  show plants lined up along the diagonal line representing the 1/10,000 target risk level
rather than along the vertical line representing the 3-log minimum required by the present SWTR.  By
definition of the 1/10,000 risk level as the target, there would therefore be  no adverse risk trade-offs in
terms of Giardia. (Trade-offs with respect to other pathogens such as Cryptosporidium may still occur,
depending on the efficiency of using Giardia as a target organism for defining adequate treatment.) Such
an Enhanced SWTR may not be feasible due to the lack of practical monitoring protocols available for
Giardia. but there is promising research in this area.53

       The evaluation of the Enhanced SWTR with the DBF-RAM showed that adherence to the 1/10,000
risk level resulted in very little additional cancer incidence, very little modification in treatment and, there-
fore, very little additional cost. This minimal cost impact resulted from the model assumption that the
additional inactivation requirements at the first customer would be met with additional contact time at the
plant effluent.  Additional contact time allows inactivation  requirements to be met without significant
increased chlorine dosages. While additional contact time contributes to inactivation, it does not contribute
very much to additional by-product formation, because the additional increment of contact time is trivial
in proportion to the total residence time in the distribution system.

6.2 3 Potential Trade-Offs Involved In The  Use of Alternate Disinfectants

       The most striking aspect of the cancer incidence predicted  by the DBF-RAM  is the order  of
magnitude of the results (tabulated in Exhibits A-l through A-8).  Exhibit A-l presents an evaluation of
the SWTR Scenario in the presence of alternate disinfectants. This  exhibit most closely resembles the
status quo. Using maximum likelihood estimates of the dose response functions, results indicate 45.5
cases of  cancer per year (trihalomethanes: 0.9 cases; di- and tri- chloroacetic acids: 44.6 cases) at the
current TTHM  standard of 100 ug/1.  Using the upper 95 percent confidence bound of the dose response
functions, results indicate 163  cases of cancer  per  year  (trihalomethanes: 24.6 cases; di- and  tri-
chloroacetic acids: 138.7 cases) at  the current TTHM standard of 100 ug/1.  About 70-85% of the total
projected cancer incidence is attributable  to dichloroacetic acid.

       The most promising aspect of the cancer incidence results  is the small difference between SWTR
Scenarios and Enhanced SWTR Scenarios. Comparison of results in Exhibits A-l and A-2 indicate that
an Enhanced SWTR, versus the SWTR, would produce about 1 additional case of cancer per year at the
current TTHM  standard of 100 ug/1.

       Exhibits 31  and 32 present an analysis of the total and incremental costs of cancer risk reduction.
The  analysis is  presented in terms of different Total Haloacetic Acid MCLs in recognition of the fact that
dichloroacetic acid  accounts  for most of the projected cancer risk.  The tables in Appendix B permit
various comparisons between THAA MCLs and  comparable TTHM MCLs.  The total and incremental
cost  analysis presented in Exhibits 31 and 32 is developed  from  the treatment  profiles assigned by the
DBF-RAM to meet different regulatory targets and from unit cost  estimates derived from the Water Cost
Model.54

                                               37

-------
        Exhibits 33 and 34 present vectors of compliance percentages for the with and without alternate
 disinfectants variants of the SWTR and the ESWTR scenarios. Each column in each of the tables contains
 a compliance vector for a given THAA MCL. The percentages in each vector total 100 percent. Taken
 together, the percentages comprising each vector indicate the mix of treatments that are projected to be
 required to comply with each MCL alternative.  For example, in Exhibit 34, compliance with an MCL of
 60 ug/1 in the without alternate disinfectants variant of the SWTR scenario is projected to require no
 further treatment by 70 percent of plants; elimination  of pre-chlorination by 12 percent of plants;
 elimination of pre-chlorination and optimization of alum  coagulation by  11  percent of plants;  and
 elimination of pre-chlorination, optimization of alum coagulation, and installation of granular activated
 carbon (GAC) by 7 percent of plants.

        These compliance vectors are an input to the computation of the aggregate treatment cost of each
 MCL alternative.  The percentages are used to split the total number of plants into treatment categories.
 The corresponding unit costs for each treatment are then multiplied by the number of plants assigned to
 the category.  The resulting products in each category are summed to produce the estimate of the total
 aggregate treatment cost. (Because of the clean slate approach adopted in the DBF-RAM (see Section
 5), some of the costs counted by this procedure may have already been incurred by systems complying
 with the current TTHM standard. These costs are trivially small, however, because they involve low cost
 techniques such as moving the point of chlorination and adding ammonia.)

        It is important to note that although the vectors of compliance percentages total 100 percent, a
 portion of the plants assigned to the final treatment category, involving GAC, may not actually be able
 to comply with the MCL.  This is particularly  true at  the more stringent MCLs.  For purposes of
 estimating aggregate treatment costs, it makes no difference.  Once GAC is installed, the plant would
 either be in compliance or would be eligible for a variance.  The cost computed by the DBF-RAM is the
 same either way.  The inability to attain compliance even with GAC is taken into account, however, in
 the analysis of the change in by-product exposures. Appendix A presents a more detailed print-out of the
 modelling results which contains a graphical summary of the extent of variances that would be implied
 with increasing stringency of regulation.

       Results in Exhibits 33 and 34 indicate that (assuming some proportion of variances is accepted)
 stringent by-product MCLs are technically feasible if the use of alternate disinfectants and/or GAC is
 accepted.  In without alternate disinfectants scenarios, 57 to 58 percent of plants are projected to require
 GAC to comply with a THAA MCL of 10 ug/1.  Nearly one-third of the plants driven to  GAC at this
 MCL would  not  be able to meet  it, but would obtain variances.  In the with alternate  disinfectants
 scenarios, 50 percent of plants would still require GAC to comply with  a THAA MCL of 10 ug/1.  About
 one-fifth of these plants would not meet the MCL, but would obtain variances. A THAA MCL of 10 ug/1
 would eliminate roughly three-quarters of the total quantifiable by-product cancer risk.

       For the segment of the water industry studied (representing 103  million persons; 43 percent of the
 total population served by community water supplies), the aggregate treatment cost of a THAA MCL of
 10 ug/1 is estimated to be $957 million (SWTR Scenario) to $974 million (ESWTR Scenario) per year in
 the without alternate disinfectant scenarios and $922 million (SWTR Scenario) to $963 million (ESWTR
Scenario) per year in the with alternate disinfectant scenarios.  These results  do  not indicate significant
cost savings from the use of alternate disinfectants because an MCL of this stringency would drive nearly
the same percentage of plants to GAC, as noted above.
                                               38

-------
        An alternative regulatory strategy might attempt to reduce by-product cancer incidence to the
 maximum extent possible without triggering irreversible and expensive technology shifts, such as switching
 to ozone and GAC. The results in Exhibits 33 and 34 indicate that standards equivalent to a THAA MCL
 of 30 ug/1 would force 20 to 21 percent of systems to GAC if alternate disinfectants were not permitted
 (without scenarios). If chloramines are permitted, the with scenario results show that only 4 to 11 percent
 of systems would have to resort to ozone or GAC to meet a THAA MCL  of 30 ug/1. Results for the with
 scenarios indicate 59 to 60 percent of plants could  achieve a THAA MCL of 30 ug/1 with either no
 treatment changes at all or with elimination of pre-chlorination. An additional 30 to 36 percent of plants
 could achieve this level with elimination of pre-chlorination, chloramines and optimization of coagulation.
 This proportion of chloramines usage is about equivalent to the current penetration of  this technology in
 systems subject to the interim TTHM standard.

        If alternate disinfectants are permitted (with scenario), the total treatment cost  of a 30 ug/1  MCL
 for the segment of the industry under study (surface water systems serving > 10,000 that  filter but do not
 soften) would be roughly 10 percent as much as for an MCL of 10 ug/1 ($71 to $120 million per year for
 the SWTR and  ESWTR, respectively vs. $1 billion per year). If chloramines are not used (without scenar-
 io), a THAA MCL of 30 ug/1 would cost about one-third as much as a THAA MCL of 10 ug/1 ($336 to
 $353 million per year vs. $1 billion per year). In the without scenarios, a THAA MCL of 30 ug/1 achieves
 50 percent as much reduction in cancer incidence as a THAA  MCL of 10 ug/1. In the with scenarios, a
 THAA MCL of 30 ug/1  achieves 38 percent as much  reduction in cancer  incidence as  a THAA MCL of
 10 ug/1.  The difference  results from the reliance on GAC as a last resort in the without scenarios.  Once
 use of GAC is  triggered, it over-shoots a target of 30 ug/1.

 6.2.4 Evaluation of Trade-Offs Within The Regulatory Impact Analysis Framework

        The framework  for regulatory decisionmaking is defined in the Safe Drinking Water Act.  The
 application of that statutory framework to the problem of disinfection by-products regulation, using DBF-
 RAM results, is described in a companion paper.55  A totally separate mandate, Executive Order 12291,
 specifies a cost-benefit framework for analyzing trade-offs during the rule development process. As stated
 at the outset of this paper, the goal of regulatory impact analysis is to develop and organize information
 on  benefits,  costs,  and economic  impacts so as to clarify  trade-offs  among alternative  regulatory
 options.*6 EPA guidance for developing regulatory impact analyses (RIAs) specifies that the comparison
 of regulatory options  is to be  achieved by quantifying and monetizing benefits and costs to the extent
possible.

        This section presents DBF-RAM results in the context of the  RIA requirement.  The discussion
 is presented in  three parts.  The first part (6.2.5.1) briefly  describes the conventional  approach that has
 been followed in assessing the benefits of SDWA regulations in previous RIAs.  The second part (6.2.5.2)
 discusses how the development of DBF regulations presents a completely different type  of problem due
 to the trade-offs between health risks, and describes the solution that has  been devised to cope with this
 analytical complication. The third part (6.2.5.3) presents DBF-RAM results in the context of this RIA
 framework.

 6.2.4.1  Economic Benefits Relevant to SDWA Regulations

        The benefits of SDWA regulations are usually evaluated in terms of two approaches: 1) damages
 avoided, and 2) willingness-to-pay.  It  is  possible to quantify the monetary value of direct economic
                                               39

-------
 damages, such as lost income and medical costs, that are associated with the incidence of adverse health
 effects.  Such damage estimates can be combined with exposure and risk models to yield aggregate
 national  estimates of the expected value of monetary damages that will result at alternative levels of
 regulation. The difference in the expected value of damages between one level of regulation and the next
 most stringent level (the damages avoided) may be compared to the difference in the associated treatment
 and monitoring costs. As long as the expected value of the damages avoided exceeds the difference in
 the treatment and monitoring costs,  the aggregate net benefits of regulation are positive.

        The concept of economic benefits is not merely the expected value of the avoided damages, but
 the wittingness-to-pay for a reduction in the risk of incurring those damages?  The willingness-to-pay
 reflects the intuitive notion that damages are, afterall, disagreeable.  Therefore, it usually exceeds the
 expected value of avoided damages.  It is presumed that people are willing to pay something extra for a
 margin of safety depending upon the degree of damages involved,  their assessment of the degree of risk
 (i.e., of the uncertainties affecting the estimate of the expected value of the damages), and their individual
 risk tolerance for the type  of damage involved.   It has been suggested that altruism  provides another
 motive for additional willingness-to-pay. Studies have shown that the safety of water supplies is perceived
 to  have a societal  value to consumers, reflecting concerns for sensitive populations  and civic pride.58
 Thus,  an extra measure of insurance is part of the benefit being  derived from SDWA regulations,
 supplemental to the expected value of the avoided health damages.  This willingness-to-pay concept of
 benefits is quite consistent with  the structure of the SDWA which stipulates that MCLs be established as
 close as feasible to the no adverse effect level (i.e., as close as feasible to the MCLG), plus a margin of
 safety.

        In the Surface Water Treatment Rule Regulatory Impact Analysis,59 for example, economic
 damages associated with contracting giardiasis were estimated to lie within a range of $1400 to $1800 per
 case (1986 dollars) based on valuation of medical costs and lost work.60  Using this estimate of the cost
 of illness together with estimates of the costs of coping with waterborne disease outbreaks, the expected
 value of  damages  avoided  was compared  to the projected  costs of installing  filtration.  The results
 indicated  that net benefits might be negative in the small system size categories, depending upon the
 assumptions made  regarding the risk of infection.61  However, viewing the  same results in light of the
 uncertainties not captured in the analysis, and in light of the presumed willingness to  pay for an extra
 margin of safety, it is possible to make a subjective judgement that net benefits are in fact positive. For
 example,  the above damage estimate does not take account of the fact that about 0.1 percent of cases of
 waterborne  disease may  result  in a fatality.62   In addition, there  are  significant economic impacts
 accompanying the epidemic occurrence of waterborne disease -- instances in which a substantial portion
 of the population is simultaneously infected.  Quantitative valuation of "willingness-to-pay" is inherently
 more difficult than  valuation of direct damages and is ultimately somewhat subjective.  (NB: neither the
 SDWA, nor EO 12291 require EPA  to demonstrate that net benefits are positive.)

        Similarly, regulatory impact analyses (RIAs) developed for SDWA regulations  governing
 carcinogens, have often  shown  negative net benefits in the small system size  categories.  Under the
 damages avoided approach, cancer incidence is valued at $400,000 per case, reflecting lost income and
 medical costs. Under the willingness-to-pay approach, cancer incidence is valued at $8,000,000 per case.
The willingness-to-pay figure is taken from economic research on the value of life which draws its conclu-
sions from data on wage rate differentials in high risk occupations and other evidence.63 The $8,000,000
per case figure is not  literally intended as an estimate of the value of life, but  rather of the willingness-to-
pay to reduce the risk of premature death.  Typically, RIAs  on SDWA regulations governing carcinogens
have used the  $8,000,000 figure.   Using this figure, the net benefits of  proposed  MCLs for many

                                                40

-------
contaminants are negative in the small system size categories and are sometimes negative in the larger
system size categories as well.  In the final analysis, determining the margin of safety for carcinogens is
a judgement call and the SDWA does not constrain that judgement to adhere to a cost-benefit criterion.

6.2.42  Benefits Concepts Relevant to Disinfection By-Products Regulation

        The regulation of disinfection by-products presents a distinctly different type of benefits valuation
problem, unlike that  posed  by any other drinking water contaminant or group  of contaminants.  The
valuation of benefits is complicated because there are multiple health risks involved and measures to
reduce one type of risk may  increase another.  The ideal solution would be one which minimizes the total
incidence of adverse health effects. Since the incidences of the different health effects are inversely
related to each other, a means of trading-off one benefit for another is needed.

        Exhibit  35 illustrates  a  transformation of  the  conventional  benefit/cost  framework into a
theoretically equivalent form  that is more suited  to  the multiple trade-offs involved.   It shows that
maximizing health damages  avoided per dollar of treatment expenditure is the same as minimizing health
damages incurred per dollar of treatment expenditure.  Mathematically, minimizing total social costs is
the "dual"  ~ the mirror image of maximizing  net benefits.  As described in Technical Note 2, both
formulations yield the same answer, but the dual is easier to work with since treatment costs and the costs
stemming from both types of health damages can be simply added together to compute the total cost.

        A means of capturing the extra willingness to pay for a margin of safety within such a "damages
avoided" context has been suggested by envisioning a category of "anxiety damages" (i.e.,  without  the
insurance benefit inherent in a safety margin, there would be comparable anxiety damages).64  Thus, total
willingness-to-pay equals the total  damages,  where total damages are conceived as the sum of direct
economic damages (lost work and medical expenses) plus anxiety damages.  Within this  framework,  the
health damages incurred are valued in this analysis on the basis of willingness-to-pay estimates: $3,000
per case of giardiasis and .$8,000,000 per case of cancer.

        The willingness-to-pay rationale implicit in this economic framework serves to clarify an important
aspect of the nature of potential trade-offs; they may consist of conflicts between competing safety margins
as much as trade-offs of actual health effects.

6.2.43  Analysis of Trade-offs In  Terms of Total Social Costs

        Viewed in the minimum total cost framework, the evaluation of disinfection by-product regulatory
alternatives entails  trade-offs between  five different  dimensions of costs.  As by-product controls  are
applied, they impose treatment and monitoring costs (dimension 1) while reducing by-product cancer risks
(dimension 2)  and potentially increasing endemic and  epidemic microbial risks  from source water
contaminants (dimension 3). As more stringent by-product controls are introduced, they may generate
additional cancer risks (e.g., bromate) and microbial risks (e.g., Cryptosporidium.  or pathogenic bacterial
growth  in the distribution system) as side effects (dimensions 4 and 5).

        In  order  to clarify the nature of the trade-offs within this minimum total cost framework, it is
necessary to model how the  five different dimensions change simultaneously; one or more types of costs
decreasing while others increase  as a  result  of the opposing relationships.  Currently,  the DBF-RAM
provides quantification covering only the first three dimensions (the core problem). Exhibits 36 through
39 present  results for the four different scenarios defined to analyze the core problem, showing a family

                                               41

-------
 of cost curves for each.  Plotted as a function of THAA MCLs are: by-product treatment costs (dimension
 1), costs of cancer cases incurred (dimension 2), costs of Giardia cases incurred (dimension 3), and the
 total cost. With respect to dimension 2, the current DBF-RAM does not include quantification of cancer
 risk associated with brominated haloacetic acids. With respect to dimension 3, the DBF-RAM quantifies
 increased Giardia outbreak risk, but the expected value  of damages associated with outbreaks have not
 been included in the total cost results to simplify the comparison  in this first iteration of the analysis.
 While the additional damages  associated with increased  outbreak risks would be significant, they would
 not change the shape of the resulting total cost curves, but would steepen the slope.  Costs of unknown
 cancer risks (dimension 4) and unknown microbial risks (dimension 5) introduced by use of alternate
 disinfectants are unqualified in the current DBF-RAM.

        Both  the cancer and  giardiasis incidence  estimates are based on maximum likelihood risk
 assessment procedures, as described  in Section 5.4 and Technical Note 1.  As noted in Section 5.4, the
 dose-response relationships presently being  utilized produce aggregate results for  both  cancer and
 waterbome disease  incidence  that  are much lower than  the  levels potentially  implied by  recent
 epidcmiological studies in both areas. Conceivably, both types of health risks could be much greater than
 indicated by the  current dose-response functions.

 6.2 A3.1  MCL Alternatives

        The total cost curves for the SWTR scenarios shown in Exhibits 36 and 38 have  an assymetric
 shape which suggests that the lowest  total cost will result from the least stringent THAA MCL (60 ug/1).
 This result  implies that the side-effect of increased  giardiasis risk outweighs the gains in cancer risk
 reduction. The trade-off is a reflection of the apparent flaw in the current SWTR with regard to the level
 of risk reduction assured at the first  customer, discussed at length  above.  Moreover,  if the potentially
 significant giardiasis outbreak risk were also reflected in the current set of total  cost curves, the picture
 would be even more assymetric.

        Exhibits 37 and 39 present total cost curves for the ESWTR scenarios. These are also assymetric,
 but have minimum points suggesting  least cost THAA MCLs lie in a range of 40 to 50  ug/1.  The reason
 there is a potential for  by-product  reduction in these two scenarios is that  an enhanced SWTR would
 eliminate the trade-off with respect to Giardia risk while  introducing very little additional treatment cost
 and contributing only a small offsetting increase in cancer  risk, according to the DBF-RAM.  Significantly,
 detailed results tabulated in Appendix A indicate that  the  increased outbreak risks induced by by-product
 controls would be eliminated in the ESWTR scenarios.

        Comparing results of with vs. without alternate disinfectant scenarios (Exhibits 36 vs. 38, and 37
vs.  39), indicates that the nature of the trade-offs is not  markedly different between the two scenarios.
As  discussed in section 6.2.3, above, the treatment costs are higher in the without alternate  disinfectant
scenarios, but  the difference is less significant at the most stringent MCLs because similar proportions of
plants would be driven to GAC in both scenarios.  It is at the most stringent MCLs, however, where the
trade-offs implied by the with  alternate  disinfectant scenarios would be different in ways that are not
shown in the present results due to  the inability to incorporate quantitative assessment of additional side
effects introduced by alternate disinfectants: carcinogenic by-products of alternate disinfectants  and
inadvertent microbial risks introduced by alternate disinfectants.

       The results summarized in  Exhibits 36 through  39  point to a conclusion that the potential for
trade-offs between cancer risk  (dimension 2) and microbial  risk (dimension 3) is significant without an

                                                42

-------
 enhanced SWTR in place.  The uncertainties associated with the risk assessment procedures assumed in
 DBF-RAM and the exposure assessment methodology of DBF-RAM are so great that there is no confident
 means  of defining an error bound around these results.  The individual microbial and  cancer  risk
 assessments are subject to a wide range of potential error. It is conceivable that either of the two could
 be in error in ways  which would cause the  total cost  curves  to exhibit very different shapes (e.g.,
 significant underestimation of cancer risk).  The exposure assessment produced by DBF-RAM, while
 producing roughly plausible results in terms  of mean values, could be in much greater error at the
 extremes where trade-offs may be more acute.  Consideration of the additional uncertainty presented by
 unquantified cancer and microbial risks introduced by use of alternate disinfectants (dimensions 4 and 5)
 further  complicates the picture.

 6.2.432  Technology-Based Alternatives

        The current set of DBF-RAM results indicates there is a significant potential for trade-offs between
 microbial and cancer risks if by-product MCLs are established in the absence of an enhanced SWTR.  In
 the alternative, there has been some mention of technology-based approaches in the hope that trade-offs
 may be  more avoidable with such approaches. The alternative to establishment of MCLs under the SDWA
 is the specification of a  treatment technique.  Normally, treatment techniques can be used only when
 measurement problems prevent the establishment of MCLs.  A rationalization might be made that, in this
 unique  case, the difficulty of relating MCLs to resulting health risks could be considered a measurement
 problem.

        One treatment technique approach would be to specify a membrane technology capable of reducing
 both microbial  and cancer risks simultaneously.  Because  membrane filtration removes microbial
 contaminants, it might provide a means of progressing without an enhanced SWTR.  The cost of installing
 nanofiltration  in all the large surface water systems that filter but do not soften has been very roughly
 estimated to be on the order of $5 billion per year ($150/household/yr. x [103 million persons/3 persons
 per household]).  By comparison, the results in Exhibit 36 indicate that the total social  cost of a  THAA
 MCL of 60 ug/1 in the SWTR With Alternate Disinfectants Scenario ~ a rough proxy for the status quo -
 - is about $1  billion per year, nearly  all in the form of cancer damages ($358 million/yr) and Giardia
 damages ($639  million/yr). In  addition to  the  unfavorable economics, it is  incorrect to conceive  of
 membrane technology as being free of trade-offs.  Membrane technologies would not remove bromide and
 may result in a shift to brominated by-products (in some systems) which may be more potent carcinogens.
 In addition the solid waste and water wastage impacts of these technologies would pose significant cost
 and environmental issues.

        Another technology-based alternative would be to adopt an approach similar to that used in the
 Lead and Copper Rule. The objective would be to require  optimized application of the simplest and most
 inexpensive by-product control strategies by all plants above a specified action level. Under this approach,
 systems above the action level would be required to demonstrate that they have eliminated prechlorination
 and optimized coagulation processes, relative to their unique raw water and treatment conditions.  DBF-
 RAM analysis of a similar option (81% not prechorinating and optimizing coagulation; see Exhibits 40
 and 41) shows that at the most stringent MCL of 10 ug/1, about one-third of  the known cancer risk
 (dimension 2)  would be removed at a fairly low cost cost ($122 million per year).  The apparent effective-
ness of  an option requiring elimination of prechlorination and optimization of coagulation as the only
available treatments reflects an inherent feature of the influent simulation — that most  of the aggregate
cancer risk is in the large number of plants that have average influent conditions, not in the small percent-
age of plants that have extreme influent conditions.  In this way, the technology-based approach  applies

                                               43

-------
the low cost treatments more extensively to more of they total known cancer risk, achieving significant
reduction without introducing the higher costs and unknown risks associated with alternate disinfectants.

       Such an approach would not be free of significant trade-offs, however.  One downside is that it
would provide uneven protection, leaving much cancer risk untouched in the small percentage of plants
that have extreme influent conditions.  Another major inescapable trade-off is that the elimination of
prcchlorination would generate offsetting increases in Giardia risks in about the same proportions in this
scenario as in the other DBF-RAM scenarios. In fact, results in Exhibits 40 and 41 indicate that it is these
two first steps in the least cost algorithm that are responsible for most of the increased risk of giardiasis.
The development of an enhanced SWTR would therefore also be  a requisite for implementation of this
strategy.

       An additional drawback is the fact that the optimization of coagulation could cause a shift to more
brominated by-products that are potentially more potent carcinogens. Furthermore, there are some plants
in which influent conditions (e.g., alkalinity) are particularly unsuited to optimization of coagulation, or
in which site conditions are particularly unsuited to the addition of contact time to compensate for the loss
of prechlorination. As with the membrane technologies, the cost and environmental consequences of the
solid waste and water wastage impacts of such a strategy would have to be fully evaluated.  All these
factors would have to be weighed by states in making individual optimization decisions.  A final obstacle,
therefore, would be the capacity of state primacy agencies to make plant-by-plant determinations regarding
the optimization of treatment.
7.0    Summary Observations

       The DBF-RAM  may be  viewed as simply an  explicit, systematic framework for testing the
hypothesis that a specified regulatory strategy will result in a net decrease in residual health risk.  In the
current analysis, the uncertainties and unknowns dominate.  The only clear conclusion seems to be that
an enhanced SWTR could eliminate increased microbial risk; and that  such enhancement might not
contribute significant additional treatment cost or cancer risk (in systems already chlorinating).

       The current DBF-RAM results are intended as starting points in this attempt to define and analyze
a manageable and meaningful core problem. The broader utility of the DBF-RAM framework lies in the
fact that modelling such a process requires explicit specification of the methodology and assumptions that
underly the trade-off functions.  Within  this explicit framework, it is possible to investigate  the many
remaining unknowns and the sensitivity of the results to uncertainties in the current place holder assump-
tions.

       The "what-if' phase of analysis that lies ahead has potential to reveal which unknowns and which
place holders have the greatest influence on the results.  The importance of missing information can be
evaluated by examining various additional place holder hypotheses regarding unknown processes, or by
simply accentuating the pure unknowns that lie outside the model, in the manner adopted in this paper.

       The DBF-RAM provides a frame of reference for evaluating, or anticipating the implications of
new scientific understanding or technology.  For example, the technology to practically  implement an
enhanced SWTR may not yet exist, but the potential of such a development has been usefully evaluated.
In this way, the DBF-RAM can serve as a framework for evaluating the value of research as well as the
relative importance of different research  objectives. Significantly, in the current DBF-RAM results, the

                                                44

-------
total social cost curve most closely approximating the status quo has an intercept that implies existing
health damages are on the order of one billion dollars per year.  This provides a benchmark against which
the value of research and of technology development can be assessed.

        A number of sensitivity tests of current place holder assumptions are presently underway in
response to comments offerred by the EPA Science Advisory Board.  The sensitivity of current results is
being evaluated with respect  to: influent TOC  assumptions, influent bromide  assumptions, seasonal
variation, over-design assumptions, prechlorination assumptions, chlorine dose and contact basin sizing
assumptions,  chlorine  decay  assumptions,  trihalomethane  formation  equations,   influent  Giardia
assumptions, and Giardia infection  modelling.  In addition to sensitivity tests, the modelling framework
is presently being expanded to encompass surface water plants which employ  lime  softening and to
examine differences in  the implied trade-offs when model assumptions are modified  to reflect the
circumstances faced by small water systems, serving fewer than 10,000 persons. In addition, a formation
equation to predict bromate occurrence and exposure is  being added.

        Ultimately, the utility of "what-if' analysis will diminish. The number of conceivable variations
of assumptions far exceeds the number of public water supplies.   At some point, it would make more
sense to apply the EPA Water Treatment Plant Model to each individual plant than it would to continue
to guess about them in aggregate.  Efficient selection of follow-up DBP-RAM analyses should still offer
further improvements in the understanding of the problem before that point is reached. Beyond that point,
however, there must be a recognition that no amount of analysis can overcome many of the uncertainties
inherent in this problem.  A modelling framework such as that of the DBP-RAM  is useful in gaining
understanding of the potential significance of uncertainties, unknowns, and variability, but it cannot supply
the measure of judgement that will eventually be required to devise  a solution in the  presence of such
factors.
                                              45

-------
References
 1.   USEPA,  Office of Policy Analysis, Guidelines  for Performing
     Regulatory Impact Analysis, December 1983,  OPA EPA-230-01-84-
     003.

 2.   Regli,  S., Rose, J.B.,  Haas, C.N., and Gerba,  C.P.,  Modeling
     the Risk from Giardia and Viruses in Drinking Water,  JAWWA,
     November 1991,  76-84.

 3.   USEPA,  Guidance Manual  for Compliance  With  the  Filtration
     Requirements  for Public  Water Systems  Using Surface  Water
     Sources,  March  31, 1989.

 4.   LeChevallier, M.W., Norton, W.D., and Lee, R.G.,   Occurrence
     of Giardia and Cryptosporidium  spp. in Surface Water Supplies,
     Applied  and Environmental Microbiology 57 (September  1991):
     2610-2616.

 5.   Ibid.

 6.   Amy G.L., Chadik, P.A. and Chowdhury,  Z.A., Developing  Models
     for Predicting  Trihalomethane  Formation Potential  and  Kinet-
     ics, JAWWA, 79(7).

 7.   USEPA, Regulatory Impact Analysis: Benefits and Costs  of Final
     Surface Water Treatment Rule, prepared by Wade  Miller Associ-
     ates, Inc., February  17,  1989.

 8.   Rose,  Joan B.,  Haas,  Charles N.,  and  Regli,  Stig, Risk
     Assessment and Control  of Waterborne Giardiasis,  American
     Journal of Public Health  81 (June 1991):  709-713.

9.   American Water Works Association, Government Affairs Office,
     Surface Water Treatment Rule Evaluation Project, Final Report,
     December  1987.

10.  Op. cit. note 8.

11.  Macler, B.A.  and Regli, S., Use of Microbial Risk  Assessment
     In Setting U.S. Drinking  Water Standards, USEPA, 1992.

12.  LeChevallier,  M.W., Norton, W.D., Lee, R.G.  and Rose,  J.B.,
     Detection  and Treatment  of Giardia and  Cryptosporidium In
     Water  Supplies, American  Water  Works Association  Research
     Foundation, January 1991.

13.  Bennett,  J.V.,  Holmberg,  S.D.,  Rogers,  M.F.,  and  Solomon,
     S.L., Infectious and  Parasitic Diseases  in Closing  the Gap;
                                46

-------
      The Burden of Unecessarv  Illness,  Edited by Amler, R.w. and
      Dull, H.B., Oxford University Press, New York, Oxford, 1987.

 14.  Payment,  P.,  Richardson,  L.,  Siemiatycki,  J.,  Dewar,  R.,
      Edwardes, M. and Franco,  E., A Randomized Trial to Evaluate
      the Risk  of Gastrointestinal Disease  due to Consumptionm of
      Drinking  Water  Meeting  Current Microbiological  Standards,
      American Journal of Public Health 81 (June 1991):  703-708.

 15.  Decision Research,  American Water Works Association Research
      Foundation National Trihalomethane Survey Report, prepared for
      Metropolitan Water District of Southern California, April 27,
      1987.

 16.  Cromwell,  J.E., III, Lee, R.G., Kawczynski, E., Development of
      the AWWA Water Industry Data Base:  A "Facts" Machine for the
      Future,  Annual Conference of the American Water Works Associa-
      tion,  Cincinnati, OH,  June 1990.
 17.


 18.


 19.



 20.




 21.
22

23


24



25
 Haas, C.N., Final Report of the 1990 AWWA Disinfection Survey,
 AWWA Disinfection  Committee,  May  20,  1991.

 Fair,  P.S.,  TSD  Plant  Data Base,  unpublished  data,  disk
 transmitted July 31,  1991.

 Krasner, S.W., McGuire, M.J.7 Jacangelo, J.G.,  Patania, N.L.,
 Reagan, K.M., Aieta, E.M., The Occurrence of Disinfection By-
 products In U.S. Drinking Water,  Journal AWWA, August  1989.

 Morris, R.D.,  Auder,  A.M., Angelillo, I.F., Chalmers, T.C.,
 Mosteller,  F.,  CMorination, Chlorination  By-Products,  and
 Cancer: A Meta Analysis, Am.  J. Public Health, 1992; 82-955-
 963.

 Letkiewicz,  F.J.,   Grubbs, W.,  Lustik,  M.,  Cromwell,  J.,
 Mosher, J., Zhang, X, and Regli,  S., Simulation of Raw Water
 and  Treatment  Parameters In Support of the Disinfection By-
 products Regulatory Impact Analysis, 1992, EPA-811-R-92-001.

 Op. cit. note  19.

 Hibler, C.P.,  Analysis of Municipal Water Samples for Cysts of
 Giardia, Advances  in Giardia  Research   (1988): 237-245.

 Grubbs,  W.D.,  Macler,   B.,   and  Regli,  S.,   Simulation of
 Microbial Occurrence,  Exposure and Health Risks After Drinking
 Water Treatment  Processes,  (1992), EPA-811-B-92-005.

 Gelderloos, A.B., Harrington,  G.W., Schaefer, J.K., and Regli,
 S., Simulation of  Compliance  Choices  to Meet Both Microbial
 and Disinfection By-Product Treatment Objectives, AWWA Water
Quality Technology Conference, Orlando, FL,  November 1991.

                           47

-------
26.  Malcolm Pirnie,  Inc., Water Treatment Plant Simulation Program
     — tfsers  Manual,  March  1992,  prepared for USEPA Office of
     Ground Water and Drinking Water., EPA-811-C-92-001.

27.  Gelderloos, A.B.,  Harrington,  G.W., Owen,  D.W.,  Regli, S.,
     Schaefer, J.K., Cromwell, J.E., and Zhang, X., Simulation of
     Compliance Choices for Regulatory Impact Analysis, 1992.
28.
29.
30.
31.
32.
33.
34.
35.
36.
37.

38.
39.
40.
41.
42.
43.
44.
45.
46.
Ib:
Ib:
Op
Re<
Grt
T\f*t
u&\
Op
Op
op
op
op
Coi
Dr-
L/X.
A
19!
Op
Op
Op
Op
Op
Op
Op
Op
Cr,
     Op. cit. note 26.

     Regli,  S.,   Cromwell,  J.E.,  Zhang,  X.,  Gelderloos,  A.B.,
     	,s,  W.,  Letkiewicz,  F.,  Macler,  B.A.,  FramevorJc  for
     Decision Makings EPA Perspective, 1992.

         cit. note 8.

         cit. note 2.

         cit. note 7.

         cit. note 24.

         cit. note 24.

     v,«~3onery,  P.J.,  and  Greenfield,  D.N.,  Pennsylvania Safe
     Drinking Water Program Filter Plant Performance Evaluations;
     A  Summary of Procedures and  Results,  January 1988-December
     1990, Pennsylvania Department of Environmental Resources.

         cit. note 19.

         cit. note 15.

         Cit. note 18.

         cit. note 19.

         cit. note 18.

         cit. note 17.

         cit. note 15.

         cit. note 20.

     Craun, G.F., Epidemiologic Studies of Organic  Micro-pollutants
     in Drinking Water,  in The Handbook of Environmental Chemistry,
                                48

-------
     Vol 5:  Water  Pollution,  O.  Hatzinger (ed)f Springer-Verlog,
     Berlin, 1991.

47.  Cantor, K.P., Hoover, R., Hartge P., Mason, T.J., Silverman,
     D.T.,  Altman, R.,  Austin,   D.F.,  Child,  M.A.,  Key,  C.R.,
     Marrett,  L.D.,  Myers,  M.H., Narayana,  A.S.,  Levin,  L.I.,
     Sullivan, J.W., Swanson,  G.M.,  Thomas,  O.B.,  West, D.W., J.
     National Cancer Institute, 1987; 79;1269.

48.  Op. cit. note 12.

49.  Op. cit. note 24.

50.  Op. cit. note 8.

51.  Op. cit. note 14.

52.  Glaze, W., The Dilemma of Competing Risks In Drinking Water,
     AWWA Annual Conference, Vancouver,  BC, 1992.

53.  LeChevallier, M.W.  and  Norton,  W.D., Relationship Between
     Treatment of Giardia, Cryptosporidium, Turbidity, and Particle
     Counts, American Water Works  Service Company, Belleville, IL,
     January 1992.

54.  Malcolm Pirnie, Inc.,  Technologies and Costs  for Control of
     Disinfection By-Products, July  15, 1992,  prepared for USEPA
     Office of Ground Water and Drinking Water.

55.  Op. cit. note 31

56.  Op. cit. note 1.

57.  Freeman,  M.A.,  The  Benefits of Environmental  Improvement,
     Resources for the Future, 1979,  Washington, D.C.

58.  Mitchell, R.  and Carson,  R.,  Valuing  Drinking  Water Risk
     Reductions Using  the Contingent Valuation Method: A Method-
     ological Study of Risks from THMs and Giardia, Resources for
     the Future,  report to USEPA,  Cooperative Agreement Grant# CR
     810 466 016, p 107.

59.  Op. cit. note 7.

60.  Harrington,  W., Krupnick, A., Spofford,  W.,  The Benefits of
     Preventing An Outbreak of Giardiasis Due to  Drinking Water
     Contamination, Resources for the Future, September  1985, Draft
     Final Report to the USEPA, Cooperative Agreement  Grant# CR 810
     466 010.
                                49

-------
61.  Cromwell,  J.E.,  III,  Costs and  Benefits of  Filtration of
     Surface Water Supplies, AWWA Annual Conference, Kansas City,
     MO., June 1987.
62.  Op. cit. note 13.
63.  Fisher,  A,  Chesnut,  L.G.,  Violette,  D.M.,   The Value of
     Reducing Risks of Death: A  Note On New Evidence,  Journal of
     Policy Analysis and Management, Vol. 8, No.l,  88-100 (1989)
64.  Op. cit. note 58, p9-6.
                               SO

-------
                                  Technical Note 1

                          Cancer Risk Assessment Procedures
       The cancer risk estimates for the trihalomethanes and  haioacetic acids included in the
DBF-RAM  are  based on  dose-response relationships developed  from  animal  bioassays for
carcinogenicity using the linearized multistage model for low dose extrapolations. In the DBF-
RAM, both  upper  limit and  maximum likelihood  estimate values for the  dose-response
relationships for these substances have been used, as discussed further below.

       The EPA carcinogen risk assessment guidelines provide a discussion and rationale for
computing and using  an upper limit value  to characterize  the carcinogenicity  dose-response
relationship for use in risk  assessments. While noting the need to examine and compare dose-
response results from various extrapolation models, these guidelines recommend  the use of the
linearized multistage model as the model of choice, emphasizing that this model provides a
plausible  upper  limit to the  risk that is  consistent with some  proposed  mechanisms of
carcinogenicity.

       It is noteworthy that the  EPA guidelines also state that such an upper limit estimate "does
not necessarily give a realistic prediction of the risk. The true value of the risk is  unknown, and
may be as low as zero. The range of risks, defined by the upper limit of the chosen model and
the lower limit  which may be as  low as zero, should be  explicitly stated.  An established
procedure does not yet exist for making "most likely" or "best" estimates."

       The linearized  multistage model produces values  for the "slope"  of a line relating the
"extra"  incidence of cancer over background rates (the y axis value)  with the dose of a
contaminant received (the x axis value).  The dose is usually expressed in terms  of mg/kg-day
(milligrams of the substance received per kilogram of body weight per day), and the  incidence
is  usually treated as a  unitless value.  Therefore, the units for this slope describing the risk are
(mg/kg-day)"1. (Note that the "unitless" incidence could also be expressed in units of "cases per
individual per lifetime.")

       The mathematical form  of the multistage model for determining the extra risk is:

                          A(d) = l-
where A(d) is the extra risk at dose d (see, e.g., Anderson et al., 1983). In this model, the "point
estimates" of the coefficients qt (and  therefore  the extra risk function A(d) at  any dose d) are
calculated by maximizing the likelihood function of the data.  The upper 95% confidence limits
on the extra risk are then determined from the upper 95% confidence limit on the parameter q^
which is denoted as "
-------
        As  noted above, the risk value is presented in units  of (mg/kg-day)'1.  It is often
 convenient to express the risk in terms of an exposure value that corresponds to specified risk
 levels, such as 104, 10'5, or 10"* (that is, exposure corresponding to a 1 in 10,000, 1 in 100,000
 or 1  in  1,000,000 chance  of a  carcinogenic response,  respectively).   For drinking water
 contaminants, these risk levels are usually expressed in terms of a drinking water concentration
 that an individual is assumed to experience for a lifetime. Most often for organics, the drinking
 water concentrations corresponding to these risk levels are expressed in units  of ug/I.   The
 arithmetic for converting the q2* and ql values to corresponding drinking water concentrations
 associated with various risk levels is detailed in the attached  Exhibit.  As noted there, these
 conversions involve the assumptions of a 70 kg individual consuming 2 liters of drinking water
 per day for a lifetime.

        Notwithstanding the EPA guidelines recommending the use of the upper limit for  risk
 assessments involving carcinogens, the DBF RAM uses, as previously noted, both the "maximum
 likelihood estimate" (
-------
to 600 ftg/L.  That is, an individual exposed to drinking water with approximately 600 /ig/L of
chloroform present for his/her lifetime, and  assuming the 70 kg body weight and 2 L/day
consumption, is estimated to have a one in ten-thousand chance of cancer (above background
cancer rates).. The chloroform drinking water concentrations corresponding to the 10"4, 10"5, and
10"*  risk levels for both the MLE and upper 95% confidence interval unit risks obtained by
similar calculations, including rounding, are shown below.
       Risk Level

         10-4

         10"5

         10-*
MLE

 10,000

  1,000

   100
Upper 95% CI

   600

    60

      6
(Units^

fig/L

//g/L
References:

Anderson, EL and  the Carcinogen  Assessment Group of the  U.S.  Environmental Protection
Agency. 1983.  Quantitative Approaches in Use to Assess Cancer Risk.  Risk Analysis, 3(4):
277-295.
USEPA, Guidelines for Carcinogenic Risk Assessment. Federal Register, 1986, 51(185): 33992-
34003.

-------
                            CO
                                           s   1?
                    To i?
                                                    O)
                     ro
                     o
                     CD


                     8
                     CJ
                                        *     *^-
                                        ^" c '»=   I
                                                                  CD
                                                                  CD
                                      CD O C- «   I

                                      iili   '
                                                    O)
                                          CO
                                           CA
                                                      ca
                                  OJ
                                                     O)
                   Ul CO
                   o> u.
                    o
                      en
                                        Si
                                        UJ  a>
                                                                                    tt>

                                                                              o <«

                                                                                E
                                                                             a
                                     I    S
                                                      O

                                                      s?
                                                         fi
                                                      o> S
                                                       So

                                                       Ifi
                                          etc
                                          ra

                                         UJ
                                                                                    til
                                                                ffl
                                                                                         XX

                                                                                         «= 3=
                                                                                          CO CO
                                                                      S S  w 2PH
                                                                     z n  « n co co
                                                                     o o  <» o 0,0,
                                                                                         in to
                                                                                               CO o* = ^
                                                        	• 	• pm 	» *W W

                                                        ??f 1"«
                                                        CQCQ jgCOcEOf


                                                        afcco ££>.>.
                                                        22 I 2

                                                        III III 3 III

                                                        i^ sd w =d
                                                                                         to
                                                                                                   CM in
                                                                                                   T-; in

                                                                                                   o o
                                                                                         (O T- i-
                                                                                                   in
                                                                                                   c\i in
                                                                                                o •r tri
                                                                                                   in
                                                                                         o o o o
                                                                                                O T-
                                                                                           T- ^ rr
                                                                                        «  CM  CM Cl  «- CM

                                                                                        O O O O O O
                                                                                        •   in
                                                                                                cn c
                                                                                        c  cv  cvi
                                                                                                   o c
                                                                                                   o o
                                                                                        CO CO CO
                                                                                                   CO O

-------
                                    Technical Note 2

   Equivalence of "Maximum Net Benefit" and "Minimum Total Social Cost" Objectives

 Introduction

       The issue of trade-offs among alternative regulatory options may be approached by either
 minimizing total social cost or maximizing the net benefit.  Total social cost includes 1) the cost
 of treatment and monitoring, and 2) the value of health damages incurred.  The net benefit includes
 1) value of direct health damages avoided, 2) "willingness-to-pay"  for a reduction in the risk of
 incurring health  damages, and  minus  3)  the cost of treatment  and monitoring.  The following
 discussion provides an intuitive method of showing the equivalence of two approaches without using
 a rigid mathematical proof. First, necessary and sufficient conditions  for maximizing the net benefit
 are derived. Second, the optimal degree of treatment maximizing the net benefit is shown to be the
 same as  the optimal degree of treatment minimizing  total social  cost.  Therefore, the equivalence
 of maximizing the net benefit and  minimizing total social cost objectives can be established.

 Maximizing Net Benefit

       The health effects associated with  microbial contaminants and disinfection by-products in
 drinking water are waterborne diseases and cancer, respectively.  Depending upon the degree of
 treatment applied, both types  of health damages may be assessed  by  estimating the "willingness to
 pay" and direct damages.

 1. Value of Health Damages Incurred

       The total value of health damages  incurred at any time can  be defined as the  sum of the
 value of waterborne diseases and cancer cases incurred  at that time. This  definition can also be
 expressed as
                                    Dt(6) = Gt(6)+ Ct(5)

where  t = any time instant, and t a 0,
       5 = degree of treatment, and 5 a 0,
       D,(6) = value of health damages  incurred at time t,
       G,(5) = value of waterborne disease incurred at time t, and
       C,(o) = value of cancer cases incurred at time t.

2. Value of Health Damages Avoided

       Similarly, the value of health damages avoided at any time is expressed as

                    = Dt=0(6=0) - Dt(6) = [Gt=0(5=0) - G,(6)] + [Ct=0(5=0) -
(1)
(2)
where  A,(o) = value of health damages avoided at time t,
       Dt=0(6=0) = value of health damages incurred at time 0 (baseline),
       Gt=0(S=0) = value of waterborne disease incurred at time 0 (baseline), and
       Ct=0(6=0) = value of cancer cases incurred at time 0 (baseline).

-------
 3. Cost of Treatment and Monitoring

        The cost of treatment and  monitoring under the regulation  may  be denoted by R,(8).
 Assuming that no treatment is installed before the regulation (i.e., a "clean slate"), the baseline cost
 of treatment and monitoring is zero, or
                                        Rt=0(6=0) = 0
                                                                                         (3)
 4. Net Benefit
       The net benefit  from the regulation is  the difference  between value of health damages
 avoided and the cost of treatment and monitoring.  It can be defined as
                                     Bt(8) = A,(8) - Rt(8)
                                                                                         (4)
where B,(8) = the net benefit at time t.

5. Maximization of Net Benefit

       The objective of the trade-off analysis is to solve for the optimal degree of treatment that
maximizes the net benefit. The optimal degree of treatment 8* is found when

                                                                                         (5)
                               !,(8') = Max Bt(8) = A,(8") - Rt(8*)
                                    8*0

Assuming that B,(8) has finite first order derivative and that second order derivative exists,  8* may
be found by applying the first order condition
                                        as
                                             a=8'
                                                     as
                                                          l«=8
                                                              . =0
                                                                                         (6)
and second order condition
                               6=6'
                                       as2
                                              8=8'
                                                      as2
                                                                                         (7)
Equation (6) is  the necessary condition for optimization.   The set of solutions  to this equation
includes all the  possible values of 8* that maximize the net benefit Bt(8), except when B,(8) is
maximized at the boundary 8 = 0.  Equation (7) is the sufficient condition for maximizing B,(8),
meaning that if 8" satisfies both equations (6) and (7) then 8* is the optimal degree of treatment to
maximize the net benefit, provided that B,(8) is not maximized at 6=0.

       Assuming that the degree of treatment 8 = 0 is not the optimal solution, equation (6) implies
that the marginal value of health damages avoided is equal  to the  marginal treatment expenditure
with respect to the degree of treatment when the  net benefit is maximized.

-------
Minimizing Total Social Cost

       Total social cost is the sum of the value of health damages incurred and the cost of treatment
and monitoring.  Let St(5) represent the total social cost, it is defined as

                                     St(5) = Dt(5) + R,(5)                                  (8)

Total social cost is minimized if the optimal degree of treatment 5" can be found for

                            St(6") = Min St(5) = Dt(5") + ^(5")                          (9)
                                    6*0

By applying the first order condition, the value of 6* may be solved from
           as
        aa
                                                      ad
                                                             ...  =0
To minimize S,(8), the optimal solution 6"  must also satisfy the second order condition
From equation (2), it is known that

                                  Dt(5) = Dt=0(6=0) - A.C5)

Substituting D,(5) from equation (12), equations (10) and (11) become
3S,(o)
  36  '
                                 as
                 35
                      l8=ft"
                                                                    !»=»
                                                                       .. = o
and
               3o2
ao2
                                                  3o2
                                                                                        (10)
                                                                                        (ii)
                                                        (12)
                                                                                        (13)
                                                                                        (14)
Recalling that Dt=0(6=0) is a constant describing the baseline condition, it has zero derivative.

-------
  The equations (13) and (14) can be rearranged as
                           as
                               18=8*
18=8*
  and
                                              «=8"
18-6
                                                               ..  =0
           .2  "«=»
                                                                ..  s 0
                                                                                         (15)
                                                                                         (16)
 Equivalence of the Two Optimization Processes

        Note  that the equations (15) and (16) are of the same  form as equations (6) and (7),
 respectively.  It follows that the sets of solutions to equations (5) and (9) are identical for 6 >  0.
 In other words, the two optimization processes are equivalent for any degree of treatment other than
 zero.  To complete the argument, it must be proven that if 5 = 0 is the optimal degree of treatment
 for maximizing net benefit B,(8), it also minimizes total social cost S,(5), and vice versa.

        If 5 = 0 is maximizing net benefit B,(S), then the net benefit associated with no treatment
 is greater than the net benefit associated with any positive degree of treatment, or
                                B,(6=0) = Max Bt(6) a B,(6>0)
                                          5*0

 By substituting for B,(6) from equation (4), the above equation is transformed into

                             ^(6=0) - R,(5=0) & A,(8>0) - R,(5>0)

 Replacing for A,(6) from equation (2), the inequality (18) becomes

                  [Dt=0(5=0) - D,(5=0)]- R,(8=0) * Dt=0(8=0) - Dt(5>0) - R,(6X))
                                           (17)
                                          (18)
                                          (19)
It is true that Dt=0(6=0) is equal to Dt(8=0) because health damages remain constant over time if no
treatment is installed.   Thus the above expression can be rearranged as
                            Dt(5=0) + Rt(8=0) * Dt(5>0) + Rt(5>0)

Recalling equation (8), the above inequality is shown to be

                                      S,(5=0) x St(5>0)
                                          (20)
                                          (21)
This inequality implies that 5=0 also minimizes total social cost S,(5).  Similarly, we can show that
if 6=0 minimizes total social cost S,(6), it also maximizes net benefit Bt(6).  Therefore, maximizing
the net benefit is equivalent to minimizing total social cost for any degree of treatment.

-------
EXHIBITS AND APPENDIX MATERIAL

-------
  Analysis of Potential Trade-offs In
Regulation of Disinfection By-Products
     John E. Cromwell, III & Xin Zhang
         Wade Miller Associates, Inc.

             Frank Letkiewicz
            Abt Associates, Inc.

         Stig Regli & Bruce Macler
    U.S. Environmental Protection Agency
    Exhibits & Appendix Material

-------
                                                  Exhibit  1
             Raw Water
               Quality
                                                  Treatment Practices
                                              Pre-Disinfection
                                              Coagulation
                                              Filtration —
                                              Softening —
                                              Disinfection
                                              By-Product Removal/Control-
                                              Corrosion Control	
                                              Post-Disinfection and Storage-
                                                                           Process
                                                                           Control
                                                                          Strategies
                                                  Distribution System
                                              Characteristics (Dwell Time)
       1 st Customer
                             Ava. 'Customer
                                                Last Customer
microbial
  risk
DBF
risk
microbiai
  risk
DBP
risk
    t
microbial
  risk
DBP
risk-

-------
                                    Exhibit  2
             Influent Giardia vs. Total Log Reduction (Winter) From LeChevailier
   10,000
    1,000
OJ
.2    100
"S
.2
a
I
      10
                                                                             • I
                                                                           i   I
                               SWTR Win.
                                                                                 10
                                      Total Log Reduction

-------
                                  Exhibit  3
•<£>  i
CO  !
Q
                    '"~>       /""">
                    •M*       ««—'
G      O
                                                                      "I
                                                                                            CD
                                                                                            CO
                                                                                            at
                                                E
                                                3

                                                                                                     X
                                                                                                     O
                                                                                            cn
O      O

-------
                                            Exhibit  4
                     Disinfection By-Products Regulatory Analysis
              Baseline Cancer Incidence —  Based on Occurrence Data
                                  (Surface Water Systems)
Annual Cases = Population Exposed (persons) x DBF Concentration (f/g/l) x Annual Risk Factor (cases/persons/year/fid


! Total Population (million persons) 3 '•"-
Average Concentrations fr/g/1)
, Chloroform
1 Bromodichloromethane
Dlbromochloromethane
Bromoform
TTHMs
Dichloroacetic Acid
Tichloroacetic Acid
THAAs ,

\ MLE Annual DW Risk Factors * i
| Chloroform ' ;
I Bromodichloromethane ;
! Dibromochloromethane
Bromoform
DichloronceticAcid
Tichloroacelic Acid
j Cancer Incidence based on MLE (cases/yr.)
Chloroform
Bromodichloromethane
Dibromochloromethane
Bromoform
TTHMs
! Dichloroncetic Acid
: Tichloroacetic Acid
THAAs
I Total
Upper 95% Cl Annual DW Risk Factors °
Chloroform
Bromodichloromethane
Dibromochloromethane
Bromoform
Dichloroacetic Acid
Tichloroacetic Acid
Cancer incidence Biased on 95% Cl (cases/yr.)
Chloroform
Bromodichloromethane
Dibromochloromethane
Bromoform
TTHMs:
Dichloroacetic Acid
Tichloroacetic Acid
THAAs
Total
High Case Estimate '
<1 0,000 people
17

77.2
24.8
10.4
1.4
114
27.7
16.6
'•'-' ;.:44


1.43E-10
3.33E-10
3.33E-10
1.78E-10
3.59E-08
8.68E-09

0.19
0.14
0.06
0.00
0.4
17.3
2.5
20


2.49E-09
> 10,000 people
' ;:V:145-

59.7
17.4
6.3
0.8
4.V' , v,"'; ,""84
22.1
17.00
*£. :•:;'" . ., v.'y.so


1.43E-10
3.33E-10
3.33E-10
1.78E-10
3.59E-08
8.68E-09

1.24
0.84
0.30
0.02
-2:4
115.0
21.4
138


2.49E-09
1.02E-08| 1.02E-08
1.02E-08I 1.02E-08
3.22E-09i 3.22E-09
1.13E-07
2.57E-08

3.3
4.4
1.8
0.1
10
54.3
7.4
62

1.13E-07
2.57E-08

21.6
25.7
9.3
0.4
57
361.0
63.4
! 424

Low Case Estimate ^
< 10,000 people
;i • ••>••: -mm
$.?„ .. ..'• /i.'S';":;''-




•':'':' . • ,••' ' ''^Tf'^:1:'-'''^' • '


sii;" ' • ' ".."''••• '*&<-'

:.::#;.:: . • .: ••'/..'














1';
















Si:-

> 10,000 people
'•:' ::'^-':.' .1
;::.;i;:. :;:..;. ,.-



	 q
; ;:ilS': .'!"••-.:'.





1.43E-
3.33E-
3.33E-
1 .79E-
3.59E-
8.68E-
••''•W-'-"
0
0
0
0
.. .':! , ' .' '(
a




2.49E-
1.02E-
1.02E-
3.22E-
1.13E-
2.57E-






10
2
1

1 Band on occurrcno* data from: dsirrftction by-products fiold studies data (EPA.OGWDW.TBO)
: B*Md on occuf r«fx» datt from Krasncn t/tjai.. 1989
1 Soure*: F«d»f«l Reporting D«ta System (FRDS)
4 maximum lik«Ihood •stimat* (MLE)
J upper 85% confidence Intervil

-------
                    Exhibit 5
CO  cc
CO
mmm

3
 O
Q.

            os  ^
              o
              CO
                                  c
                                  o
                                 oo
                           QOCO
                                            c
                                            o

                     §,S'5o,

                     feg.®?
                     j^f:
                                           V

-------
                     Exhibit 6
          CO
 CO     3
'co
          03
 CO      >*
          03
CL
00
Q
                  03
                  03
                 •ma
                  D)

                 08
00)
CT3      -^3
03 O      .g O
O C      ^ 
-------
Exhibit  7


90% 	

.a 70%
JQ UuJb [


3 «rvv /
O 20% /

0
SURFACE WATER WITHOUT FILTRATION/SOFTENING
CUMULATIVE PROBABILITY DISTRIBUTIONS
Actual and Model-predicted Influent TOC Concentration


ft LOGNORMAL CURVE

i 	 i"^L_ ' 	 i-'-TT.... 1
! .--•' ^** — r
h'lx^**"
j^ i
^ 1
/? • I
X'' ' i
' .•" I
..
i






























5 10 15 20 25
Influent TOC Concentration (mg/l)

30
100% T 	
& 90% •
(8 7fW .
.a /uvo i

l S ~w
0 n«n/
Actual and Model-predicted Bromide Concentration


	 FITTED LOGNORMAL CURVE i

I 	 —-^^=~- ,....,..
.-X"
: 1
7 • '
/ I
/ ° '











3 30% 1


0.0
;
!

!
.f




















0.5 1.0 1.5 2.0 . 2.5 3.0 3.5
Bromide Concentration (mg/l)
Actual and Model-predicted influent Giardia
100% -
2- 90% -
5 70% i
f\
| 40y^





	 — LeC

hevallier 46 cities





~ Currently Fitted Lo
1
gnormal Curve i


1
i





••^^^
^
JX?
/.:
S"
t




i
^f i
^x^1' :


rr-~~— K^ ; !
u-i***-**^
1
i











10 100 1,000 10,000 100,000
Influent Giardia Concentration in Logarithmic Seal* (cysts/100 L)
    Page

-------
Exhibit 8
-
S"
E
o
2 10 •
c
§
o
O
O
o 1-
4-*
0>
i

Simulated Influent TOC vs. Influent Giardia


i,^-^.— >^— — '
—
	 a
	 91
1 	
p —
-1-
1
i

i
i
o -I 	
1
_
s 	 	
1
100 i==
nt TOC Concentration (mg/1)
—A
_* O
1
0
0


a
-1— «-
•V ' 1 ! ' "" r— rr, 	
1 'ffl 	 	 Bf 	 •
Mil' yr^
i|®
i — •*
« 8
"*^j-
^
i : '
III f


8 s ®
fffi T la «
_.± 	
14 •»


1 i ' 1 	
ii— H«--
y l a la.
T^1tf:
— r^i

!

	 3 	 1-
	 . 	 h-

— i —


	

10 100 1,000
Influent Giardia Concentration (cysts/1 00 L)




^ ^^— p «-^

Simulated Influent TOC vs. Influent Bromide

i^— ^^— ^»— — •



.0001
_«^__^_
1


i
i i •: l !
€
£p 1 1 -

>
• | ' ' , <
19"
• ,,> I
i i
' j>> i i a-i
i . ' PL
9^ j ' * 3
S ' i ['
i I

' c, « *
p1 "T

. ' i
TT

^ ^'
/if 1 ^ 1 1












10,000 ;

i



4 i«*
[ «l

0.001 0.01 0.1 1
Influent Bromide Concentration (mg/1)
P 	 i


. 	
10

-------
Exhibit 9


0
ft\
u/

C
mmtm
*f\
CO
'o
0

0
O
C
CO
"5.


O








2f
£g
fll *""
£R — ^
ZM ^^ta
.E co
UJ















-^



















•—p*-




0
m CO
CO 0
coO
i- E
^





2o
/_ CM
•gl
"z





























<
0 (
W(
coc
o i
~" <












3>
/)
^
1^
1^
••
3
••
t













_^».



C
o
7^ "***
^ **~
CO
z


^_ X H
T3 O




2o
•§=
•-?
OT0"













^>












.^to.





W









C
O
«^» *•*

CO


•nO
\j v^
^3 ^f

-------
                   Exhibit  10
                                    U)

                                 =  5
                                   I

                                 =5 tn
                                               o
                                            —  o
                                                  I
                                                  0)
                                                  Q.
                                                  tn


                                                  o
                                                  «*^
                                                  o
                                                  0)
                                                  •a
                                                  a>
                                                  0>


                                                  CL
o
o
               - ajeu (ssaun;)

-------
                                               Exhibit  11
                                       Overview ofGlardla Modelling
     Hibler Data
      73 plants
  variation between
  plants: lognormai
                LeChevalier Data
                   46 plants

                variation between
                plants: lognormai
         t
  temporal variation
     within plants:
delta negative binomial
                distribution used
              for influent simulation

                        Assuming:

                        • 48% recovery
                        • 13% viability
                        • deletion of
                          values estimated at
                          limit of detection
                                                  100
                                           simulated influents
                                            Treatment Model
                                                  100
                                      plant effluents (1st customer)
                                                   t
                                           summary statistics
                                  f
                               mean
                                     90th%
 distribution used
  for simulation
  of within plant
temporal variation
                                 I
compute expected
value of Rose dose
response function
compute expected value
 and variance of Rose
dose response function
                              estimate
                             cf endemic
                              incidence
                                rate
                                   estimate
                                  of outbreak
                                     risk
Assuming:

• 5% failure rate
  (loss of 1 log)
• 25% secondary
  infection rate
• 1% per 30 days
  outbreak threshold
• viable cyst=
  infective in humans
• infection=jllness

-------
                                          Exhibit  12
                                Overview of National-Level Model
Process Characteristics
• SW
• Filtration
• w/out softening
   Simulated Raw Water
   Quality and Chemical
     Dosages for 100
         Plants
           TOC/UV 254
          Removal Model
          Alkalinfty/pH
             Model
       Chlorine/Chloramine
          Decay Model
        Treatment &
        Distribution
        Assumptions
        • SWTR/ESWTR
        • Lead
        • Taste
       Batch Mode
      Treatment
        Model
                                    THM Formation
                                        Model
       HAA Formation
          Model
                                      inactivation
                                        Model
                                      Predicted Plant Effluent
                                          and Distributed
                                           Water Quality
                                           for 100 plants
                                        Compliance Sorting
                                             Routine
                                       (least cost algorithm)
                  f
          Average Customer
              By-Product
            Concentrations
    Treatment Vectors
of Compliance Percentages
   for Alternative MCLs
Effluent (1st customer)
Giardia concentrations
 for alternative MCLs
                 I
                Cancer
                 Risk
                Model
        National
         Cost
         Models
                                       i
       Giardia
        Risk
       Model

-------
                           Exhibit  13

                 Compliance Sorting Routine
           MCL Option #1
              PLANT #1
   Read Predicted By-Product Concentrations

    For Status Quo Treatment Configuration
Below MCL
                         Above MCL
                       Read Predicted By-Product Concentrations

                     For Next Least Cost Treatment Configuration
                                   s This  ^^^   yes
                               Last Treatment ?
                                   vailabl
    Save Record of Effluent Data to Output File
                                 4,
                   Compliance Percentages
SWTR Scenario
W/AIt.Disinf.
No further treatment
Eliminate per-chlorination
Eliminate per-chlor & add ammonia
pre-chlor + ammonia + alum dose
pre-chlor + ammonia + alum + ozone
pre-chlor + ammonia + alum + ozone + GAC
100
61
21
18
0
0
0
TTHM
75
49 '
25
23 ,
3
0
0
MCLs (ug/1)
50
33
23
34
9
1
0
25
17
10
30
27
13
3

-------
                              Exhibit  14

                           The  Core Problem
       Influent
       Giardia
                    Influent
                     TOG
Influent
Bromide
                                Coagulation
                                    &
                                 Filtration
Giardia
removal
                                  Reduced
                                   TOG
   Log
Reduction
at the 1st
customer
                                                    Shift to
                                                   Brominated
                                                    Species
                    Reduced
                     Demand
        Reduced
       Precursors
              Giardia
            inactivatfon
                                             By-product
                                             formation
    C!2
Residual at
  the last
 customer

-------
                                 Exhibit 15
                     SWTR Scenario: With Alternate Disinfectants
                     TTHM MCL= 100 (ug/l)      #. of Plants = 39
10,000
 1,000
ra
1
OS
5
"£
.3
"2
  100
   10
                              Total Log Reduction Achieved
                                                                  I    !
1 • /
1 ! /
i ; ! /
! i i./
i /
I
!



. ,- i |
i ! |
j
                                                                            10

-------
                                   Exhibit  16
                       SWTR Scenario: With Alternate Disinfectants
                    TTHM MCL = 25 (ug/l)     #. of Plants Treating = 83
   10,000
   1,000
to
>>

3-
a

1
jj

(3



i
100
      10
                                          z
                                 Total Log Reduction Achieved
                                                           I     I     I
                                                                               10

-------
                                      Exhibit 17
                       SWTR Scenario: With Alternate Disinfectants
                             TTHM MCL = 100 ug/I to 75 ug/l
   100%
    90%
(0

Q>
S*
o
35
2
I
"3

o
                       10             100           1,000          10,000
                     Increase in Annual Giardia infections Per 1 Million People
100,000

-------
                                     Exhibit 18


                       SWTR Scenario: With Alternate Disinfectants
                             TTHM MCL = 100 ug/l to 25 ug/l
   100%
if)

3

o
   10%
    0% -T
                      10            100           1,000         10,000
                     Increase in Annual Giardia Infections Per 1 Million People
100,000

-------
                                      Exhibit  19
CO
   100%  —
   90%  —
   80%
   70%  —
   60%  —
   50%  —
               85%
O
09
S

-------
                                      Exhibit  20
   100%
   90%  	
   80%  	
   70%  	
CO

§ 60%  —


I
03
I

0)
   50%
   30%
               47%
SWTR Scenarios: With Alternate Disinfectants
        TTHM MCL = 100 ug/i to 25 ug/l
                                                             26%
I : --
20% — •
l
\ '
10% — •
i
1
I
n% — 	



•

7%

- 'I'l'T-l; :!:•*";;' ;'"1 ;";!"''.•
: , ..


•
17%













3%
, lji!!;!i&!':SM-?3!!!!:
               1-10           10-100          100-1,000        1,000-10,000      10,000-100,000

                       Increase in Annual Giardia Infections Per 1 Million People

-------
                                    Exhibit  21




                      SWTR Scenario: With Alternate Disinfectants



                         From TTHM MCL = 100 ug/l to 75 ug/l

                         Change in Risks Per 1 Million People
    10
I
—     i
0)


I   1
Q>
u

(0
O

13
o
£
u
0>
Q
                         T—T-TT
                              ! I
                                           i fel I!
                                           fO
                                                       I  :
                                                      TTT
                                                           i—r
                                                      >   ;  i
                     10            100           1,000

                            Increase in Annual Giardia Infections
                                                              10,000
100JOOO

-------
                                     Exhibit  22


                      SWTR Scenario: With Alternate Disinfectants

                         From TTHM MCL = 100 ug/l to 25 ug/l
                         Change in Risks Per 1 Million People
    10
I
I
2
I
o
0>
£>
U
0>
a
    0.1?
   0.01-
                                        i   i
I    !  t
                      i 9 | bi
Jfi.
                     10             100           1,000

                            Increase in Annual Giardia Infections
                                                       10,000
                                                                          I I
                   100,000

-------
                                    Exhibit 23
                       SWTF5 Scenario: With Alternate Disinfectants
                        THAA MCL = 60 (ug/l)     #. of Plants = 29
   10.000
    1.000
§

_o_
.2   100
ra
o>
                                                                -/+•
                                                           I   /  I
	 1 . w
• /
;•

i


!
;


: / i • ' i
I / . i I
i /
• y


i i
! i


i . /
i./
i /







i
i



                                 Total Log Reduction Achieved
                                                                                10

-------
                                    Exhibit  24
                       SWTR Scenario: With Alternate Disinfectants
                    THAA MCL = 10 (ug/l)    #. of Plants Treating = 80
   10.000
    1,000
i    100

ra

O


i
      10
F
*

                                Total Log Reduction Achieved
                                                                              to

-------
                                   Exhibit 25


                    SWTR Scenario: With Alternate Disinfectants
                           THAA MCL = 60 ug/l to 50 ug/l
100%
                    10            100           1,000          10.000
                  Increase in Annual Giardia Infections Per 1 Million People
100,000

-------
                                   Exhibit 26


                    SWTR Scenario:  With Alternate Disinfectants
                           THAA MCL = 60 ug/l to 10 ug/l
100%
                    10             100           1,000          10,000
                  Increase in Annual Giardia Infections Per 1 Million People
100,000

-------
                                      Exhibit 27
   100% —
   90% —
   80% —
   70% —
o  60% —
0)
OJ
m
   50% —
|  40%-
   30% —
   20% —
   10%
    0%
               95%
SWTR Scenarios: With Alternate Disinfectants
        THAA MCL = 60 ug/1 to 50 ug/l
                               1%
                     1%
                                                             3%
                                                                           0%
               1-10           10-100          100-1,000        1,000-10,000      10,000-100,000

                       Increase in Annual Giardia Infections Per 1 Million People

-------
                                      Exhibit 28
   100% —

   90%
SWTR Scenario: With Alternate Disinfectants
        THAA MCL = 60 ug/i to 10 ug/l
   80% --'
   70% —
CO        I

I  60% —
   50% —
0)
H  40%
                                                             29%
   20% —
   10% —

;









' • . :
x«'« i




,
"-i-"^ „<









_ ' %

*-r









.














12%














I

24%

^





'



.





— 1 — ::



iiyiil







I1-:- •

ijjj i
ij:j




1!








4%
J — ! — 1 	 1 — 1
               1-10           10-100          100-1,000       1,000-10,000     10,000-100,000

                       Increase in Annual Giardia Infections Per 1 Million People

-------
                                   Exhibit  29
                     SWTR Scenario: With Alternate Disinfectants



                          From THAA MCL = 60 ug/l to 50 ug/l

                         Change in Risks Per 1 Million People
    10
0>
u

0)

!°
'u
o

CO
U

ra

c
    0.1
0>
<0
to

£
u
0>
a
    o.or
-9-1-
                                                                         i  i
                     10            100            1,000


                            Increase in Annual Giardia infections
                                                     10,000
100,000

-------
               Exhibit 30
SWTR Scenario: With Alternate Disinfectants

       THAA MCL = 60 ug/l to 10 ug/l
    Change in Risks Per 1 Million People

10
1
0)
g
0 « -
n Annual Cancer Incid
j«
a, 0.1 «
CO
ra
O g
o '
0)
a g
(


r
!
I 1
1














i l
1 i
i
1
'
:
'




fl
1



1 >
! 1 |
M
III


i . t j































'
~rn 	 ; 	 : — rr-
| 1 i . 1 !
j j : ! i •» 1*
ill'® : ®
iiii
*

.

I
a




r-g-
Y)








-i-






!






















H ^— 1 W"--

81
— 31 —
. *
« « T !
a ® •
- 9
3 i
I
, • ' , . -fl 1 I I i




J.J.

(

J_

.*i i • a | -t1« < j
t • • l • ! i

. » 1 i • ' ' • ' ; !


! i I'" ! i 1
• : !
i
t
1
.OTr
1







!! 1 !
II!
li





i








!
J
;
' !
L. ...


i
i
10 100


















T1 	 9s"
91


a






















8 L M«
! i
9 i
aa.
e
^
9

i
1
«




1 1







I WJ



















___
4-lli
a u




















i ' 	










-=-
S gj
6



• • — g 	

»


9













i
!





! P i
i





i






1 1 . ! i |
i 1 i i j










1 • !
1









i i
1,000 10,000 100,(
                                                       000
      Increase in Annual Giardia Infections

-------
                                                   Exhibit 31
8?

t
•° S  _
 « a  to
 B w  «
 3 .«  «
z
                     *
                 las
                 a i— <
— o  o
2 «  <«
5 B  «
w «  u
                   B
                   a /-^


                lit
                |]
                   B
                   W
                <_ a
                   U
I'IS
                 2 -
                   ^ ^^

                "n £ ^


                £-1
        —
        y


        s
        'S
                   B
                   a

                   C/3
                           O «—
                                   oo f, r~- CM
                                      CN o\ o
                                              c
                                              «",
                                   — — <*•. o
                                 I  O OS O CM OS
                                ir; CM OO CM CM —
                                rr n- <*i ro ts -"
                                O\ — OO  — ' Os U-,
                                r~- o •—
                                — CM SO
                                CM CM — — f^. CM
                                   — CM C- OO CM
                                           rc O\
                            "3

                           U
                                 o o o
                                 SO V, ^T
                           c o
                           CM —
j*  tn    ^,
•t  O to <
 5  *3 «
 3  « «

.2 U >


5  8"S
 B  B "O ^
 a  « 'S ^.
   Sr j -i s>\
   w >
                                                               S2 u  M
                                                              ,9 e  «
                                                               S 5  S
                                                 s


                                              S I ?
                                              la Is C^.
                                              a a *c
                                              I- to MS
                                              a A

                                              B 3
                                                               U ^
                                                              A to
                                                                 W  «
                                                                 M

                                                                 O
                                                                 c.
                                                                 CS
                                                                 U
                                              2 £
                                              .o H
                                                 to
                                                 CS
                                                 B
                                                 a
                                                 u
                                                 C/5
                                                                 C/5
                                                                       r- o  •—
                                                                                1  00 TT — — .
                                                                          O O
                                                                          <•«-, V-.
                                                                                  r-  Os O
                               r-
                               oo
                               oc

                            CM" -o"
2
a
o

o
CM

O
                                                                                                     W
                                                                                                     o
                                                                    oc o CM
                                                                    <*•. CM CM
                                                         U

                                                         S

                                                         <
                  c o o o o o
                  \o in TT tf. CM i—
                                                                                    (S

                                                                                    •o
                                                                                    0)
(S

to
(8
                                                                                    O
                                                                                    U

-------
                                          Exhibit 32
8J






 I
2
•2 u ,u.
 a to n
 s « »
2 S »
a a 2

!3J

go<
a
                « o
                o w
                a n
   a

—  ° "T1
W  C !*••

"S  "3 2l

2  8S
|  Ht*
**""•*

I  § =
   a ^
             V-  «
             isi
             cs —
             £H|
                u
                u
               c/a
              1  CN VO OO CO CO
                           I  00 c*-. 0\ CO 00
                               •* *- 1 «/-.

                               OO U-. ff.
                          •"*• rr fi O) »— «->
                          oo oo
                                 oo
                                 vo
                                 ITS OC

                                 f. n-.
                      O
                          o o o o o o
5 ^
5 w «
y O i«
• 5^
^i U ^ rv
a a *o s&
O « *5 V-.
E 0 > <*
U CM <
U °
a
w ^^
2 u M "s
o S.S
1 0**
iJJs
« k. "S °^
c i
n
si?
a M C
u fr" t2<
|||
M ~ tq
O ^ ^^. ^j
CQ &Q ^^
^ t- '5 v^
g U .3 ^
a a § o;
Z M u
^^ f X W

M
ro
lit
r** *** ^^
&•
CJ



^ ,
53 ** ^
Hi
MM CQ 4A
ea « >-'
•» !•> »
£^ S
u


o
'C
•J
S
V]
|

C/3
a

1 CN VO CO C7\ CO



1 r^ ~o\ ff* r^ i^*
»-• C^l tM pj



1 O \O VO ON t^>
— -* «N Tf


1 O  O\ Vs,



1 0 — 00 04 VO
O w 10 OO

Csl O TJ- VO OO O
<*-, tS C5 SO >o rrj
1—1 "^ ^™





O\ O\ — • -H vo OO
oc so i/-, rr •** \o
CO OO \O. If. ** cf.
^ _r ," .^
v^i ^r t^





OO CO OO \O OO Tf
— — — tr; 00 t~







M
^^
J
^ o cs o o o o
•IM
^























S2
(8
0)
>.
o
CM
i_
0)
>
O
S
a
£
t»
£
£
=
o
T^
a
•o
0)
_N
"5
3
C
C
a
£
a
^2
"m
O
O
«

-------
                         Exhibit 33
                 Compliance Percentages
SWTR Scenario
W/Alt.Disin.
No further treatment
Eliminate pre-chlor
Eliminate pre-chlor & add NH3
pre-chlor + NH3 + alum dose
pre-chlor + NH3 + alum + Oz
pre-chlor + NH3 + alum + Oz + GAG
60
70
12
18
0
0
0
THAAMCLs
50 40 30
64
15
17
4
0
0
56
16
22
5
1
0
48
12
22
14
3
1
(ug/D
20
37
8
8
23
7
17
10
19
3
1
22
5
50
                Compliance Percentages
   Enhanced SWTR Scenario
         W/ Alt. Disin.
     THAA MCLs (ug/1)
60   50   40   30   20   10
No further treatment
Eliminate pre-chlor

Eliminate pre-chipr & add NH3

pre-chlor + NH3 + alum dose

pre-chlor + NH3 + alum + Oz
pre-chior -t- NH3 + alum + Oz + GAG
70   62   56   47   -35   17
11   16   16   12
17   18   16   12
10
     0
2,   3    10   18   18   15

0    1    2    10   16   15

000    1    17   50

-------
        Exhibit 34
Compliance Percentages
SWTR Scenario
W/out Alt Disin.
No further treatment
Eliminate pre-chlor
Eliminate pre-chlor & modify alum
Eliminate pre-chlor + alum + GAG
60
70
12
11
7
THAA MCLs
50 40 30
64
15
14
7
56
16
16
12
48
12
19
21
(ug/1)
20
37
8
20
35
10
19
3
21
57
Compliance Percentages
Enhanced SWTR Scenario
W/out Alt Disin.
No further treatment
Eliminate pre-chlor
Eliminate pre-chlor & modify alum
Eliminate pre-chlor + alum + GAC
60
70
11
12
7
THAA MCLs
50 40 30
62
16
15
7
56
16
15
13
47
12
21
20
(ug/1)
20
35
10
20
35
10
17
3
22
58

-------
                          Exhibit 35
Equivalence of "Maximum Net Benefit" and
  "Minimum Total Social Cost" Objectives
    $'s
Benefit: value of health
 damages avoided
                                        cost of treatment & monitoring
                                         Degree of
                                         Treatment
    $'s
   total social cost
                         minimum total
                          social cost
                                           cost of
                                          treatment &
                                          monitoring
                                                    cost of health
                                                      damages
                                                      incurred
                                          Degree of
                                          Treatment
 Maximizing Health Damages Avoided
 Per Dollar of Treatment Expenditure
                        Minimizing Health Damages Incurred
                        Per Dollar of Treatment Expenditure

-------
                                         Exhibit 36
                                                                           Draft: 13-May-\
    MODEL OUTPUT (surface w/o softening): SWTR W/ ALTERNATIVE DISINFECTION

    Treatment Code:
    1 — not requiring further treatment modification
    2 — eliminate pre—chlorination
    3 - eliminate pre-chlor + add ammonia
    4 — pre—chlor + ammonia + alum dose
    5 — pre—chlor -t- ammonia + alum + ozone
    6 — pre—chlor + ammonia + alum + ozone + GAC
    Population =
103,000,000
(persons)
THAA
MCL
fcg/0
60
50
40
30
20
10
Annual
Treatment Cost *
($M)
2
12
21
71
383
922
Cost of Cancer**
@ $8m/case
($M)
358
338
304
259
177
91
Cost of Giardia
@ $3,000/case
($M)
639
714
849
974
1,239
1,729
Total
Costs
($M)
1.0CX
1,0&
1,17^
1,3O
1,79!
2,74i
  Capital costs are annualized at 10% interest rate over 20 years.
** MLE of cancer incidence from HAAs.
                 3.5
            CQ
            &
            CO
            O
            O
                               SO       40       3O       20

                                  ALTERNATE THAA MCLs (M9/I)

-------
                                        Exhibit 37
                                                                           Draft: 13-May-
    MODEL OUTPUT (surface w/o softening): ENHANCED SWTR W/ ALTERNATIVE DISINFECTION

    Treatment Code:
    1 — not requiring further treatment modification
    2 — eliminate pre—chlorination
    3 — eliminate pre—chlor + add ammonia
    4 — pre—chlor + ammonia + alum dose
    5 — pre—chlor + ammonia + alum + ozone
    6 — pre—chlor + ammonia + alum + ozone + GAC
    Population =
103.000.000
(persons)
THAA
MCL
»0
60
50
40
30
20
10
Annual
Treatment Cost *
($M)
7
16
38
120
422
963
Cost of Cancer**
@ $8m/case
($M)
369
348
313
258
178
91
Cost of Giardia
@ $3,000/case
($M)
0.8
0.8
0.9
1.0
1.0
1.0
Total
Costs
($M)
377
364
351
379
601
1,055
 * Capital costs are annuaiized at 10% interest rate over 20 years.
** MLE of cancer incidence from HAAs.
                 1.5r
             m
             OT
             O
             o
                                               30

                                  ALTERNATE THAA MCLs
                                                        20
                      10

-------
r
                                                   Exhibit 38
                                                                                       Draft: 13-May-\
               MODEL OUTPUT (surface w/o softening): SWTR W/O ALTERNATIVE DISINFECTION

               Treatment Code:
               1 — not requiring further treatment modification
               2 — eliminate pre—chlorination
               3 — eliminate pre—chlor + modify alum dose
               4 — pre-chlor + alum dose + GAC
               Population =
                   103.000.000
                                            (persons)
==r=^=
THAA
MCL
WO
60
50
40
30
20
10
Annual Treatment
Cost *
($M)
,
104
Cost of Cancer**
@ $3m/case
l$M)
334
Cost of Giardia
@ $3,000/case
($M)
740
1181 3201 772
2021 270
831
3531 2061 972
5881 140
957 1 79
I
1,212
1,519
Total
Costs
($M)
1,1
1,2
1,3
1,5
1,9
2,5
           * Capital costs are annualized at 10% interest rate over 20 years.
          ** MLE of cancer incidence from HAAs.
                                3.5
CO
I

I
                         CO
                         8

3.0 r


2.5 r


2.0 r


1.5r


1.0 r


0.5 r


0.0 ^~
                                                                            TotalCost
                                                                                Cost of Giardia Cas«s i
                                                                                  Cost of Cancer Cases
                                     60
                50          40
                     ALTERNATE THAA
                                                                     30
                                                                               20
                                                                                          10

-------
                                                                            Drafc 13-May-
                                        Exhibit 39


     MODEL OUTPUT (surface w/o softening):  ENHANCED SWTR W/O ALTERNATIVE DISINFECT1OI

     Treatment Code:
     1 — not requiring further treatment modification
     2 — eliminate pre—chiorination
     3 — eliminate pre—chior + modify alum dose
     4 — pre—chlor + alum dose + GAC
     Population =
103,000.000
(persons)
THAA
MCL
(M9/I)
60
50
40
30
20
10
Ann.Trt.*
Cost
($M)
118
t 118
218
336
588
974
Cost of Cancer**
@ $8m/case
($M)
337
320
271
211
142
77
Cost of Giardia
@ $3,000/case
($M)
0.8
0.8
0.8
1.0
1.0
1.0
Total
Costs
($M)
455
438
490
548
731
1 ,052
 * Capital costs are annuaiized at 10% interest rate over 20 years.
** MLE of cancer incidence from HAAs.
                     1.5 r
               tn
               O
               co
               O
               O

                                                                        Treatment Cost
                                                                    Cost of Cancer Cases  I
                          60
                                               40         30
                                          ALTERNATE THAA MCLs fr/g/L)

-------
                                            Exhibit 40
                                                                  Draft: 10/12/92
Model Output (surface water systems w/o softening?: SWTR  w/ MODIFYING ALUM DOSE

Treatment Tier Code:
 1 - not requiring further treatment modification
 2 - eliminating prechlorination + modifying alum dose
Population •
103.000.000 (persons)
                   Number of Systems •
                322
THAA
MCL
(ug/D

60
50
40
30
20
10
Treatment
Code

2
2
2
2
2
2
% of Systems
Modifying
Alum Dose*

30
36
44
52
63
81
Cumulative
% of Systems
<= MCL*

93
93
68
79
65
43
Annual
Treatment
Cost <$M) "
CD
45
54
66
78
95
122
Cost <$M) of
Cancer "*
@$8M/Case
(2)
326
303
275
254
234
215
Cost ($M) of
Gtardiasts
@$3K/Case
(3)
796
853
917
1,086
1,237
1,261
Total
Social
Cost($M)
(1W2)-K3)
1,167
1,210
1,258
1,418
1,565
1,597
  * Includes 20% over-design factor.
 " Maximum Likelihood Estimates (MLE) of cancer incidence associated with HAAs.
"• Capital costs are annualized at 10% interest rate over 20 years.
                            Total Social Cost at Alternate THAA MCLs
                                                         .Total Social Cost
        200 +   Treatment Cost
          n •              •
                                                          Cost of Cancer Cases
                                                                     a	
                                                                    -+-
           60
    50
   40             30

Alternate THAA MCL (ug/Q
20
10
                                                                c:\017\dbp\lotus\sensffiv\old\ALMHAAXLS

-------
                                            Exhibit 41
                                                                                        Draff:
Model Output (surface water systems w/o softening): Enhanced SWTR w/ MODIFYING ALUM DOSE

Treatment Tier Code:
 1 - not requiring further treatment modification
 2 - eliminating prechlorination + modifying alum dose
Population =
103.000.000 (persons)
Number of Systems ••
fZZ
THAA
MCL
(ug/l)

60
50
40
30
20
10
Treatment
Code

2
2
2
2
2
2
% of Systems
Modifying
Alum Dose*

30
38
44
53
65
83
Cumulative
% of Systems
<= MCL*

93
93
87
80
65
42
Annual
Treatment
Cost <$M) **
(1)
45
57
66
80
98
125
Cost <$M) of
Cancer ***
@$8M/Case
(2)
332
300
280
258
235
216
Cost ($M) of
Giardiasis
@$3K/Case
(3)
0.8
0.8
0.9
1.0
1.0
1.0
Total
Social
Cost($M)
(l)-K2)+(3)
378
358
348
339
333
342
  " Includes 20% over-design factor.
 "* Maximum Likelihood Estimates (MLD of cancer incidence associated with HAAs.
*" Capital costs are annualized at 10% interest rate over 20 years.
    O 200 -
    a 150 ••
    § 100-•
    <  50
                           Total Social Cost at Alternate THAA MCLs
                                                         Total Social Cost
                                    Cost of Cancer Cases
                                      Treatment Cost
                                       40             30

                                    Alternate THAA MCL (ug/0
                                               20
                            10
                                                               c:\017\dbp\lotus\sensitiv\ old\ EALMHAA.XLS

-------
Appendix A

-------
4-)
•H
A
•H

X
1
?
   I
   l
   S
                               lOW 9NII33S SWHiSAS dO %
       I  -,
        13 + *
        E E E
             I
                                                                            KJ

" j
1
]
i
_
s
e
1

a,

P
1!
i!
Jsz
_*S
i*;
5*
Si
*"
iS£
1
sl
JC
fl
S
i
{• 8 S | |
I
1 i i ^ «
Jj » » » BJ
^
^ gg««gg isisSS illlil sliil-
8
« o o> o> o> o nqoEDcqo n<»r*.(qcpo cooui^r^-^r
^•^oooci ^*-oooo »-dddcici •^•ddddei
i
o> o o m o c oqqi>r^i>. to o -r q o a 
-------
•H




I
      I
f    i=

S    a"
I    €;
3.    s
                             1OW SNLL33W SWHLSAS dO %
              i: i
               l
Kill  !
                      si
                     '•3
                     K
            !?1
            a*
                                        oioneri  aiocb<\j
                                          "!      -^-
                          noock  CJOOB  oo»r-o
                               -
                                        ^ o o d  *-ooci
                          ea o ^ p*  o o to us  « a *- ^"   »•

                          r? r* in *•  cjuicviv*  nnoitD  nnKo
                          cS « « Si  cl w « J3  « « "- "  
-------
en
•H


 x
W

     •s

     i
                                                    oooo
                                                    in*r(oaj-*-




                                         HOA 9NIL33W SW3ISAS dO %
                -
                      >e
                   Jp  o


                   i  §

         S
liliii  i
i  11  111
— eg n «r us o


.
1
•s
a





«
1
jj
I



t
fi
1
j


4
CumulMlvo







,
|
£



|
£

3
^


§



^
3
£
i
s.
1
£
%oiSy*.
i
a
•o
*


Jt
1
i
a
s
£
%
|


|
1
"i
2
X
£

u
1
*
S

«
i

5
T

—
it
V
"a
c
£
§




>,>,>,>,
III!



a a 5 s

cootr^ooo co r»- oj CM o o> CTOIACMUIIA o •»— 10 01 •»- tv

SI5-SI5 Is§S^i igs^s? SSas-sg

o — o o o o raooooo CTCOEOOOC ncar«.tnv^

•"T tO (O l/> Ul lA *T 03  Oi o S' <73 Ul P* O> O S «- OJ ^T P- Ol O
-oonoo a^gj^-o ggjgr-^o J ; ^ g; £ „



888888 KSKKKH SSSSSS 3<3S$%i9.

                                                                                  3   S   8  8   S   <


                                                                                    tlfln* NOUMUN32NOO MV3TI

-------
AJ
•H
.o
•H


i
    §
    £
    B
    a
    3


    I

    E
    Ul
    -5
       l|!i«

       111!
      *• a • • i.
      5 i S. S.

       fill  i
      l»||e  a
      | cii» S.  ^
HOW SNUH3W SW3LSAS dO %

• 1
§1
1s
m
•
3
{
S

«
£
o

i,

E

•Italians of
ifAvaCu».j>
Jl
! £„
1*"
!3
i^l
*i
i
® >****.
^ S S I
II 2 2 1
I
j 8 :B 8
•g
S
w HgS |«gg i;iS
3
ItfQiTri (d ul c5 M «i ci *r «I
yi uionc^ lonoj** m CT o 0
81" "" ""'
1— — — w ^ — o e> ^cidd
-U!
i

ooocvi 03 q o c 03 at  w *r ^ tn^eiw mc3np3
i
j5538 SSS8 SSS8
i
/
|Ooi»- ^-e«»- nojoicQ
f N"' """' ^W'
! 52 52 52 5? tnmuito o o o o
ooao r*.f»is.N. uimuSOS



£

1


n — to in

isli

- ood


-------
•H

43

•H
                                                                   S
                                                                      ^
                                                                      _2_


                                                                      M
                                                                      <
                                                                   a
                                                                   CM
                              o
                                        9NI133N SIM31SAS JO %
     •3
     1






c
a
+
i!!
in
a a c
•*• + +
ill
i ii
fi s •i
T T r
ill
i i
» u» c




2
dirant modV
tan
add ammonia
ring farther Ire
pra-chlorinol
pro-chtot +
Ill
1 I 1
»-« «











i













!
703.000.000

.5
3
*

la



- Q
.=
S
JS „
E


S
i
3
£
8

S
*

*i
ii
i!
Iji
fi
*t
1

I
f
d



P S rt o o d d
i 1
1
— i
g s i i § 1 §
9 n < ri V of e'
i « a a s ? I
s


g ^_ _

eo^-^^^— qir-qqoq n -r- at a at at nqoooe eoqN. Ul l/> UJ 'T CO O C3 C\J O ^TOOOJOfO
ojcuojaicMcv cMMCMoicvicQ CM *-»-*- -^ *- oj^-*-*-^-»- cM^-^-^-^-ii w;:;;^:iitu
1


^8gSn ««§8B «««88 »""S "MM *«««•
/
i
i B «» nrrwo CMWO-UJO «n-«-tn

                                                                                       ;   \

                                                                                               \  \
                                                                                       !SSgSSS2"

                                                                                        inn) »iouytuNa3Nao wan


                                                                                                             ff
                                                                                                           f «
                                                                                      8  S  S  S   S

                                                                                        aim sasra toawo-KNv
     •g IS o "W
     3 S t —

 3   II1!


 % ill!
 s I !a§gjs

^*i Itfll
   5 g C .9 Q. Q.

-------
8


I
  vo

  4:

  •M
  •H
  A
  •rl

  X
  w
       «   *  8
       s   S  I
       I
       1
                       \
                        \
                                                    _  o
1
                              o
                                        o  o  o
                                 10W 9NU33W SW3J.SAS JO %
F   IISS  s
o   IK '§ »- t*
5   13.2 e
i  Jf jl
a  -H f i S. i
t  (5"i B " ^  >

§  ilii!  I


u i
1 1
1 1
£
J3
•fi

<
LJ

S
1
s
|
£

8?

23
1!
is
i
I «» i
• *:
i"
CO.
% •
o
*u
*!
M
1
JO
° § i § S
£| Jl -= d 0'
I
"1 3 i " s
1 s g' S a"
23
,^O--h» tr»-oo VOifi^ ^•^•co r» eo S T to r*. §
J
„ o w ^" f*. •T in v f«» o o co CM 02 CJ a> ^
^9 r^ r> <~ to !*• *• tn *- »— ^— ^r ^~ ^ CM
;Q
31

j|l
i|
ragggg ssss ss?s ssss


i §
«S d

I I
i S

O) o o fC ra CB N! tt
^••^•eoi- ^-cacM
CO^-O>O CO CM CM O

CO O 03 1 N CO a K Irt

CO a CM v* CO CM O O
CO i to CM ai COUJO^-
CO CM CM t£ 0 CM CM >i

CN|^^ CMr-^

SS55 SSSfe

(^•iniOB O CM CO Q
CO T 





o o o a o o o o
CM CM CM P* r~ t- ••-+*.
                                                                          S S S


                                                                      (!*« NOUVHlNa9HC» KV3M
                                                                    S  S  S  S  ^

                                                                      COO-OU-03JNIVKXMO >
                                         «»  T  PI   Ci   2


                                          (Tin) 83SV9 U33NV9 TIHV
                                                                           Vl
                                                                                s
                                                                                    3


                                                                                    S


                                                                                    si
                                                                                          Id
                                                                                          as
                                                                                          si
                                                                                           i

                                                                                  U
                                                                                  lip
                                                                                     *5
                                                             lit!
                                                               si*2"
                                                                                         [ilii

-------
§
t
Q
    r-
    •H
    X!
        a
        I
        £2
        a

        I
       .
        h«
        3
o   a  a>
                                         10W 9NLL33H SW31SAS dO %





c
u
4i


J§ E
+ ,
E E i
aj o "S
see
ill
|| j
S a S

•sss
! 1 1




i
1 1
hi
(tag farther Ire
pro-chlomaJ
pra -chtor +
— no« lequi
-elknlnala
- edmhtlB
«- « B












1
1
1

1.
1


1
1





*
I
103.000.000


*



i
i
£
3
3








9

I
g
1


11
|1

Ji
ICumuUIra
% of Bin.
a.
•8
*
iE
~
1
^
a
&

*•
«^
^
a



5t
£

3

«>
B

a
,.

1
|


=
'1
V
.'01
i
i
I


£ S S ' S S S
££££*£
s s £ Has.



S 3 S fls a CD

w oS « » o »


S §§SS? S?55?3 S5SSS8 S^SSSS S^SSS'S iliisz




^^TTTr ^^rTrT 9 T °. °. 99 .-)qq«o» n o *- ^ O O O OOOOCOO k.1 O CS «O Ul Ul U)O>CO*TCMO tf>CBCOOO(D
CMCUCUOJOlN CM CM OJ CM CM « CM CM CM »- ^- 5- CM •--—»- -^ — CM ^- -^ ^- »- S c« i= ?I " " **

a?§SK?i? ss?ss!8! a??sss s^ssss is§sss« ssssss

OV^QQS CMcotooiog ^OCM^OOQ >^ en ^~ os ci o m in c> t^ co o r-- o o ui o a
r^coraooo *o r«. ca es o S tnr^ooofi ^r u? p* 03 01 5 co^^oeoo ^- .§
                                                                               ly^l
                                                                               c B^ •«
                                                                               111*
   ill!
!  flls^
 •  £3-f5f
   s-s?ii
                                                                                                                     "Si
                                                                                                                     S?
                                                                               5l^i
                                                                               O

-------
I
    00

    4:
    •H


    I
         I
         u
         I

         i
         I
         S

         m
         a
         u
         g

         1
         o
         *
         §
!   s    I
   •o


   I,
    i:  si
              III!
        I
        5

                                §  g
                            co   r^  (o  to

                            "IOW 9NIX33W SW3XSAS dO %



M
S
1
•i
u

e
3
£
3
i


TJ
]

£
3





b
I
u

I
1

5



,
IE

1
Bv-PioAid

1
TJ
*
w
>>
W
•5
£
_I
X
1
5
S
e
i
1
i
IU
3
^
8

Ul
S

g
^
1

"d
V
*l
5
II
i



££.££££
£ ; £ £ £ £ £


8 8 S S 8 J
g««H 8*** SSSSJ «»« ^8* *«ti*
co CM o O co vj ^ w OI^OIT owcyo oo^*1^ cooto o
uj ^^ rt 55 irt ca CM  VCOO^
co co CM c>5 concMcu co cy CM CM n CM CM fO co CM CM ^. co CM CM ^
m*-oQ m »~ o> e» m p co aa tnoon tn CB ^r t» tocano
CM CM CM & CMCU*-^ ««*-*- CU^-—— CM*-— CM*--^
u> v **• 5 10 ^r ^" Sr m- ^ ^ n u> T- co o in ^r co tu in o CM 4«
o T- rt cs CMono 10 CM r-« CB r~- 01 o a tn m ID D f^- o CM y
r-DoS or-o« inr.0 -r in CD n •» «> - « •»

g^~'- gss- ss-2 »s-o,a c,^s-s ^nsis

QQQO O O O O O O O O O O O O O O O Q O O O~ O
© 5 S C m in in a •T^-'W* co co eo ft oroi 01 « — «- »- 5*
                                                                                     (I/Ml HOU.WUN33NO3 NVa<

-------
Appendix B

-------
2
I
 •H
 XI
 •r(



 I
 Z
 O


 5
 UJ
        Q

        Ul
        I

        I
        to

        •5
        S.

       I
           0     O
          g    -o



          1    lo

           S    t§
                                 e
                                 o
                 a •  e
fc     of«2o


O     c  S  SJS ,

jrf     EollS
g    i  c  •  • Q.

o     .1111
3    j- »•  CM  « •»•
                                I
                                a.
Can cat 1
Incidence
0 THAAs
S*
4
c
1 3
elihood Est
THAAs D(
je
-1 U.
E «
3
|
8 o
3" CM
— CO
•
u
_ CM
1 °
H
m U.
o 0
U "
3
{I
g* Custom*!)1
s DCAA TCAA
• <
1 1
2 u.
1 S
o
f u
I s
0
5 to
_ «
i Mean Concentrationi
TTHMs CF C
1 wo
3^3
E « v
S*
gi
ad
o a *i e r» o
» n n N *• ~
oooooo
oooooo
oooooo
O U T CO CM f
O » CO CM
a a co   o o o o
o o o o o o o
CO CM T- O IO » CO
f- «> IO » CM »-
CO CO CM CM V-
N e » w «o o
«r w n N ^ ^
oooooo
oooooo
» » » "T CO CO
oooooo
CO CO CO CO CM CM
oooooo
CO CO CM CM »- »-
OOOOOO
T- o a co r- in
i- r- o o o o

a a co e ^ CM

rC J^ S 2 co P
oooooo

o o a co co »
T- a « *- f» »
eo ^ « o » h.
•>• » co eo CM i-

oooooo
co n ^ eo CM v
*• «- CM
00 0
oo a
CM ^ C
•
U
3
THAA/TTK
•
•
i
3
•
u
e
e
•o
£
•
o
e
U
•
0
3
Alternate THAA
«
•
u
g
fl
0
£

„
O
at Alternate TTHU MCLi
•
o
c
e
TJ
•5 !!
e
o
O
CM CM IM ^ «- ^ ». w ,
ooooo tfkiAmi
ooooo ooo<
« » CO CM »- COIO^CC


§ K S S i
11 1 1
I\
1 \\\
^k\
^

o o us o u) o
(H1W) VVH WOUd 30N3Q!

^
\
\


n o in o in o
{31W) VVH WOUd BON 3d
V

(31H) VVH WOUd 30N3Q1C
• 100 «l «
" »• f »• 1
n M n u> i
s o o o i
O CM »- CD 1


Sin
«




in o
3NI U30NVO


-

N.

W O li
ONI U3ONVO
-
N, ;
Ml U30NVO
i *i n CM a
30000
» CO CM »-


8
8 !
_ o
si
8 E 1
1
,1
.
Q
fi
If
8 If
0 40 SO
= THAA MCL. (M8/1)
eludes 20% over -
L and those not m
- $£«
i • »
S is =|
< !•
°- n\
0 — .c
2 •«-
1!
0 $.
, a 1
40 SO CO 70 80 90 100
IRNATE TTHM MCL. &IQ/I)
f systems able to meet MCI
the last treatment tier of all
8 5 "* 1
o < e g i
o|
1 cummulative p«
2 mean concentn

-------
I
 •H


 X
   W

   5
   E
   IU
   5

   ee


   to
   I
   I
   a.



   O
         o
           !!
  o
  U
    I

    '•S
    o
    E




    ]|i:"
    ts 2 + S «^
    • 8 8 * o o
    5S2 E E E
    5 » « 2 E E
    * I i • • •

    C n <%_
      .5 S o
      "
   uj EO
a
o
        i i
          i i i
          a a a

          i i i
               a
               o
n


0
Cancer 1
Incidence
i) THAAs
MCL UCL
fTHAAs) (TTHM
U
•lihood Eati
THAAs DC
a
_l U.
E
3
• o
a" CM
~ CO
u
1 3
M
1 S
o *
a
(r
ige Customer)2
U DCAA TCAA
1 1
o
1 N
OB
*S 09
Mean Concwilrationi
TTHMs CF C
• w_
«l ift O
"3 »• a
E o
3*
» V- K CM CM -•
+ •+ n n N *-
o o o o o o
o o o o o o
o o o o o o
 r- « in *r T
o o a co i- «
T-
O O h* 4D ^ CM
O « O O CM »•
O O O O O O
a co o o « *•
O CO l>- O 10 «•
CM »- v- *•
o n a 10 a 10
» CO CM CM ^- T-
O O O O 0 O
o o o o o o

O IO O O 10 IO
O !*• O IO C9 CM
o e N
n 10 n
o o o
 »"-
"a £ 3
E » v
3-*
a z
CM N f a ce e »-
e> CM »- n n n n
n n io o o o o
o o o o o o o
co CM T- o 10  »
i- «- 0 O 0 0

co i*» o OU)
(31H) WH NOcU 30N3QIONI U3ONVO

n
\
\. -
\Q
3tnOU)OU)OU)<
O^TOe3O*M*-»
(3n«) WH WOHd 30N3QIONI U3ONVO
V :
^\
-
}tnoinou?ou)c
lVTnc3Oi«»-«'
(31W) WH WOUd 30N3QIONI H3QNVO
0 0 0 i- *•
M CM N CM »•
O IO IO IO IO
D 0 0 0 0
O » CO CM i-

i
8 |
0 °
|
Q H-
s 1
o
3
itf
if
8 11
30 40 SO
E THAA MCLi fag/I)
includes 20% over-
CL and those not m
o E a "" !
* 11
o -g c
•S if
1 »
xo so oo 70 ao eo 100
RNATE HHM MCL« (/ifl/l)
systems able to meet MCI
he last treatment tier of all
S!-' ° u
_i » . «
S« S
H '•=
* cummulative pt
2 mean conoentr

-------
r
CM
Ct

I
         •J

         d
           2  £

           1  i
           M  m
               >


               z

               UJ
              I
              co
              Q

              O
              z
              S.
          =   e
           S   S
          C  -0
          ""   C

           1  I
                          2   s
                    ifji!
               ~     « W W j*   Q
              A    -H I  I  «   °
                  •Q *« 1* i* v
                  o — • • o

                  0 §•• «*
              jj   1 o J J 2   «
              0   | c « . a   -3
              O   .1111    §•
              3   f-  • — i
1 w"
J5]
«ij
o ce n e ce o
v n r> N T- T-
§00000
OOOOOO

N  10 *
•-OOOOO

r-  e 10 n
a a n n o> r-
10 10 10 10 w w
OOOOOO
O - r- « •«•
o ca v v cci 10
CM T- »- •-
o » c» n a *
• ca 10
a a at a a a
o in o o n n
O 1^ tD K) O Cl
in » in
o o o

U
I 3
elihood Esti
THAA* DC
£
imum L
BF
i s
o
• 01
2 o
u

s S
3
t
• (Average Cuatomer)2
: THAA* DCAA TCAA
1 *

1 N
a
•5 CD
CM
Mean Concentrationi
TTHM* CF C
Ms
d!
3 X
10 «0 0 CM N 0 »
CM T- *• n n n N
m n 10 o o o o
o o o o o o o

r~ e «  » CM
O O> O « O CO
CM T- »- »-
OOOOOO
CM CM i- i- O CO
o o a  at at 01 eo *• i
a o> a a a ca
OOOOOO
IO H>1
OOOOO OOOOOO Ol


SU) O O V) Vt
t~ S in n 01 .
s N\ \
i ^V\\.\
^X\ '
^, .
S5SSS8S22
(31W) WH WOUd 3ON3QIONI U3ONVO

\-

-
AOtAOIOOIAOU
(TIN) WH nOUd 30N30IONI U3ONVO
V :
>=> :
(31H) WH WOUd 30N30IONI U3ONVO
U CM CM N C»



, I
3
o S
* 1
§ s
J_
• I
2
IO
J<

8 «l
20 30 40 50
ALTERNATE THAA MCL* (|I0/I)
enttier (include* 20% over-
eeting MCL and those not m
2 ||
i> ** JC
1*
£ 1
• S.
40 50 80 70 80 90 100
ERNATE TTHM MCU (pg/Q
system* able to meet MCI
he latt treatment tier of all
o g o ^ i
0 * • S
«!
' cummulative pi
2 mean concentr

-------
S!
 i
t

  I
w
IU
I
i
         (0
         o
         Ul
         c
        I

        1
        uJ
Treatment Code:
1 - not requiring further treatment modification
2 - eliminate pre-chlorination
3 - eliminate pre-chlor + modify alum dose
4 - pre-chlor •»• alum dose + GAC
S - pre-chlor -f ammonia -f ahim + ozone
6 - pre-chlor -t- ammonia -I- alum + ozone + GAC
Population = 103.000.000 (parsons)
Cancar 1
MCL MCL Incidence!
(THAAs) (TTHM.) THAA*
Cancar Incidence (Maximum Likelihood Estimate)
TTHMs CF C2B B2C BF THAA* DCAA TCAA
Mean Concentration* of By-Products (Averaga Customer)2
. TTH"-* 	 CF 	 C2B B2C BF THAAs DCAA TCAA
ICumulativa
MCL % of Sy*.
(TTHMs} f_MCL'
IO CM a CM CM i- CM CM to CM CM »- 10 <• O ^ CM «- MJ O » O ^ ^ O O O • ^ <- 0 0 0 0 O *•
 N » •» «r « »> CM — n n n n ft f n r> n n ft •*• n •> n N M *» N CM CM CM CM «-
oooooo oooooo oooooo oooooo oooooo oooooo
o u « n CM •- o 10  0 «> - o io «• o CM *- o «  •>
o r- «o « n CM
T-
Incidence (Maximum Likelihood Estimate)
C2B B2C BF THAA* DCAA TCAA
Cancar
TTHMs CF
Mean Concentration* of By-Products (Average Customer)2
TTHMs CF C2B B2C BF THAAs DCAA TCAA
• . '
s 2 a
|£v
r» r- o » m ~
o r» n co o o
n n co CM « r-
e n a CM CM i-
» » n « CM »•
oooooo
oooooo
* «r * «r « n
oooooo
•+ m m m N *-
oooooo
O CO CM CM *- O
oooooo
r- »- 0 CO «0 »
•^ — — b b b
CO CO 0 IO O T-
o o a co •> o
CM o co m o 0
CM CM ^ ^- ^
r- r- « r- in o
o 0 0 0 « OU)OtOOU>O
^vnncMCM — *-
(nn) WH WObd 3ON3QIONI U3ONVO "5
&0

(31W) WH WOUd 30N3QIONI fcHONVO
\ ;
O tnv^oocv
-------
Appendix C

-------

*•
I
   0
        W



        O


        2
        I


        0)

        c

        I
        <

        «
        CO
        -o

        C
        (0

        W
 CO

;o



_c


"5


 c
        2
        +*

        o
        o
        c
        o
        O















1
1
1
s
J
s
!
£
•e
C
3
|
£
i
X
'
jj
X
i
c
<5




























<
?
J





a


|




E
G
t
1


c
1
e
i
i
i
1


2
|
E
c
i
j;
"C
g
1

E

c
Jl
"



-
1



g

a
0
o
r>
01
c
a
a
i
c
S
E
£
[
i
i
£
c
a
TJ
C
e
i
£
B
C
.1
T
O
E
c
g
E
i
e
.1
1
E
c
8
E

i
0)

c
.!

1

e
5
E
i
as

c
.9
g
E
c
E
fi
u
a>
.g
E
c
i
E
C

j
^

as:*;:-- at?s;s- sfcs?-- 5^?2S-
in in •«• co eo ,- o o in m « co IOW^^«Y- i^i^ow^-w



N. to u> ^- co cj eo r«- ® in * ^- r- o in * co  f*» T->QCO o o o o o o o o o o c> o o o o o o

to co «o o to o to to in m ^- co (o in w in  to.co o co
- CM T- -r- T- f- CM i-111?- -r- '" CM'-r-^T-^- CM^*-^-^-





o co r** to u) *4* o co f*« o in ^ os co r^> r*** in ^f OB co t*- to m ^


if) <^ co  o ^ o> CD to CM CM in




st*^-010 ssss"1" as?^"10 gi^sss-

co to co o> Is- o co to to o> r*"- o co f*1* m o co o in r*» in o> jj* o>

COCOCOCM** COeScJCMCj" COe?CMCJ$2*- 5cOCOCMCI»-

o ^i- en ^ as ^*- o co en to o in o^tomoa^ r- 





                                                                                           I
                                                                                           e

-------
5

I
           o
      I
       £    i

      3  |

      ni
      5:1
•n     -=i   ra
J3     (0   ^

s     co   i
            C  CO
            Q)
            o
           o
                                • o
                          * 5 +
                              il
                             i
                             o o  a
                           I   I  I  I
             S«- p p    co a r- r^   coacow   coat^v    m at f I
             CM CM K    COT-T-T-.   CO .- 1- •r-.   OT-'T-».    3 »- 1- •

     a

      c    oocoiO:?   co m in to   co in ^ *:    *  «M

      i



           T- «*•««•«*    V- 'd- O I
           CO CM Ol M    tO CM CM i

     Q

      e    coaiaitt1   co co co >•   co eo r*. o   co co «> ie-    co co o 4"   co co <


   ?•§
   o  5
   P  E

      C|   r- O O O    r-OC90»   T- O CO «J   T-OI CO O    i- Ol h. >T   f- C» r-  CO

      §








      c    r- r- h- r»    r^ i-- ^ »   r- r- rf CM  . t~- 10 CM »-    r~ 10 CM e»   i- t CM  n-    ^co^oi-   •* co c.  ,
           CMCMCMVM-   CMv-T-v»   CMT-T-T"   CMr-T- '^B-    CM »- T-      CM r- »-




     £    CO CM CM CM    CO CM CM CM   CO CM CM tM   COMCMtV    CO CM CM N   CO CM C4  CM





   __ ,S    o' o' o' d    d d d d   o' d d d   d d d d    dodo   d d d  d

all
sr
                                             fill
                                              I  S
                                              SflP
                                                   3
                                                        co r-   r^ «
                                                        T- a> « ai   ^ CD co co    T- a co  i
                                                                in   COT-**-*-    coT-^reo    COT-I
                                                                H   r-iO'T*    i- 10 f  n.   T-m
                                                                                           T>'    CDCOr-O
                                                                                                  co ^- T--   o oo co <
                                                                                                                             SCO CM'!

                                                                                                                             ^- T- ',
                                                         I  CO CO M   CO CO I
                                                                                                CO CO CO •»-
                                                                             tot^eota    co^^-«^
                                                                             ^tocort    ^-rtCMW
           tncMooi    in c» in *   mcocMi
           ~  •      '    m co co co   wcoeoi
                                                         i  CM O 01
                                                         i  ^r *r n
                                                                                                       i M   m co CM ^    u>
                                                                                                                        to p -^r-
                                                                                                                        rt 
-------
Draft: 01 -Apr~02
Exhibit C-3
Concentrations of Individual THMs and HAAs at Alternate TTHM MCLs
- SWTR w/ Alternate Disinfection -
Treatment Coda: w/ Alternate Disinfection

1 - not requiring further treatment modification
2 - aiminate pra-chlorination
3 - eiminate pra-chlor + add ammonia
4 - pra-chlor + ammonia + alum do**
5 - pre-chtor + ammonia -t- alum dos* + ozone
0 - pre-chtor + ammonia + alum dos* + ozone + GAC


I Concentrations of By— Products (uoA.) (Averag* Customer)*
|
3
i
Bromoform
Dibfomochtofomethana
Bromodichlocomethane
Chloroform
§
e
c
E
I
e
i
E
i
o
.a
e
e
1
i
1
.a
•q
a
e
1
i
1
T
E
e
d|

•'& ':••'• ••$ 5*1;

CO CM C\! CM CM CM: CO CM CM CM CM 8 CO CM CM »- *- w CO CM *- v T- -^ C9 M v v v* :••». « *M v- f- *- *•
co a o o o o co co a o> at a co co o o <» » co eo o ai o a co co eo r- r*> tv co ca co r» r- «o


•'
1 • ' " • • %

CO CM CM CM CM tt COCUCJCMC4CM C9CMCMCMCM C*: CO N N CM CM rN: O CM CM CM CM CM CT CM CM *- v- r-
odddoGJ: d d d d d o d d d d d d d d d d do d d d d d d d d d d d d
d d d d d d d d d d d o d d d d d d d d d d d d d d d d d d d d d d d d
CM CO «O CO CO (Q CM C» (D O O ffl-: CJtf>mU> CMaiOv-^ v cy O W (A Ol '01- CM O CT W CO W
rv.o^r^r^r9: r- o co o 05 55 r- ca w CM cj cv r^cocMCMcMN- r- eo CM i- i- ^ r- « CM »- -r- T-

COCMCMCMCMCM O> CM -r- *- i- i"- COCM^^^-T- CJCM^-f- i- •--•- CO Ol v- •*- CO  I** f-- f^ 1^*-' ^~ O ^" 4D ID "O< •»" Ok (D IO *O W: ^" d IO ^ ^ >*4^'
cocor^i^r*f«« coT- co^^r^ ^!^" COV^«COCM eo T- ^ o» o to
Q^-^VT-W o o co co co eo o o r»- « m tf> ocomcomc* o co co o o O OCOCMCOGON

S;gS8§ SS8SSS ?588S8 ? S 8 8 8 S3 §38888 §g»»«>5
S5!§§§? ssssssj: sgoaaa sssaaais sfeaass ssscr^:*
SO O O O O' U) IO IO M IO If)' O O Q Q Q ^^ Q Q Q Q Q'Q "^ !Q !9 1^ IO IO IO IO U> U) IO IO
S S 5 S S h- P- r-. ?- ?- P. S S «5 * «5 85 SSSSion K 35 co rt rt co co rt CM CM N CM CM ci

* concentration at each treatment tier of all systems; those meeting MCL and those not meeting MCL

-------
1
            10


           O

           s
           I-    I


           il§
           C  '+=

           II
           <  ,S
           4-1   CO
          •o   o>
           C  5
           «  <
           OT   o


          S   c

          *_   (8


          *€

                LL1
          2
          •»•*

          0)
          o

          o

         O
 I
  I

 "So

:ts


ill
•  fc -8
                            i  •

                          s|fr

                          III.
                          r- CM n
                        r~ o o io    f~  II).:   i- h- « »    T- K H> - f ti








                        CO CM CM M    CO CM CM-CM'    CO CM *- »•••   CO CM T-'T-:   CO CM T- T».    CO CM *-•;-;•




                        co CB o 10    co O co Nc.    m a t- O    co o :   co co o *    eo eo on









                        T- o o o    T- o o :«»•    T- o co r»:   T- o eo •«;   i- o r- 10    »- o ^ i*»;














                        coeocois*-    coeor^V:    coN>cntM-'   cor* co:.^-   co to CM 01    co to ^'«3:









                        ffi M S °    K? S e°;">:    in o> o w':   m e» m,«»;   in o» v a    m « o (»
                        CM CM CM v*    CM CM T*-«ai'    CM v- v- i*-    CM i- s— TV.    CM ^ T-       CM v» ^ ;.•'•••••






                        CO CM CM CM:    CO CM CM CM    CO CM CM CM.   CO CM CM CM;    CO CM CM tf,    CO CM T-;{M




                 e

                 E      •» ^; co CO    »t co co•»:    * eo co n-   t co co «:    * co co *   * eo

                 n      O O O CJ    O O O:c£:    OOOCJ:   OOOcf:    ocicio'   OOO






                        t~ «o o «>:    r-. 

10 to o o o ci; o o o c! o o ci cj< o o o'cf. ci o c5 cl ci ci ci^ci- §r: r: IT- co e» t» t» eg a e *• ma 10 »• m en 10 N- eg a io:W: Ol CM :N-: CO T- ^ !». CO T- T- *• CO ^ T- • »•*• CO »- *- Sv-T--.;:; CMT-O - co co n- •+ m co w ^c co co « * co co w •<»• co co - v~' Q O; if O eo W, ^S- O h- N **• O I v- Q'CM- . •«• «$'«•>• SI « S Si SI S S CT ** j- eo o «>. r: co CD •<•' »- « in o r- co m «o COCJCMCM-' COCM^-^^— -COCUv-v. CO-C*i--r» COCMt-»-' COCM^--. O Cfe O CI-: Cft 01 CO;td: 0> CO f^ N' O> CO t^. (D- O) CO r*. IO: C* f* « ^' • '• ' , :'•' • ' '••'•;. CM O O O: CM O C9-CO CM O CO N:- CM O> CO Is.- CM CD h. IO: CM <* I O> Q-° ^ i— 5-? • ^ O'r f*- *D U> 'W; h» ^f CJ f» N- ^t1 OJ r^ ci B - rtcM • CM CO * r- CM CO *» SSS8 SSS8 8S88 d OJ c


-------
Exhibit C-5
dividuai THMs and HAAs at Alternate TTHM MCLs
iced SWTR w/ Alternate Disinfection -
icentrations of In
- Enhan
rtion
C t
O i
0 *
a
I
y

"j
•8
S
'S
s
!
£
modification
nonia
dose
dote + ozone
dose + ozone + GAC
I ill!
E e tl • • •
15 "S + + +
ill'1'1!

f ot g E E
C 1 1 • « •
c£H+++
IIJ55S
o-« ^ •= •= •=
X c c <* o o
" '1 •§ 1 1 1
ilim
1 1 1 1 1 1
t- :Q£
™;

CO CM CM CM CM CM
d d d d do
o' d d d d d
r- in E E 5 5,
^f CO CM CM CM CM

O Ol CO CO QQ-99-
CM o c» a c» o*
^SSSSS:
T-  CM w T- T- *-
w ? ^r 5 5 5
8S88§§

8»»tSS s?«;553t S?"?"?Z. 85™^^? S^SP""



00010000' coaoooo1 coooico N>-I<«: COCOOIN-KN- QOCOOKO .O'
i- O O O O O i- O O O C» O T- O  t~  O: co r~ oo co co o to h- to  to «t CM cj

COCMCMCMCMN- COCMCMCMCMOT COCMCMCMCMC4- COCMCMCMCMCM- COCMCMT-v-v-
•* CO CO CO CO CO •«• CO CO CO CO CO T CO CM CM CM «VI Tf CO CM CM CM CM ^- CO CM CM « CM
d d d o' o" d dddd do o d d o" do d d d d d d o' d d d d o'
d d d d d d d d d d d d d d d o' o' d d d d d d d d d d d d d
CM T- r^ -r- — ^ CM 1- O CO CO CO CM t- O CM CM CM CM ^ O CO O O CM ^ O in CO CO 1
^* ^ CO W W tO'' f** ^ ^ CM OJ W ^* ^ CT CM CM CM- r* *f O ^ ^ ^ ^« **f CO ^ T~ ^
•I- CO CM CM CM CM •«• CO CM CM CM CM T CO CM CM CM CM "T CO CM CM CM CM •» CO 1- •r- — T-

co a> to to co a o» co i-. r- r- f«. a> co t--  a to a CM co co r- i~- f» CMOCOKOO: CM cs t- co in » CM t» r- « •* »

r- i- ^ at tx ai ^- o o co ID o i- a co •» co r) i-aoi-oo r-a>coair^N

co o> o r*> ^ f^r' co o> co w m • 10- co ot co fli 01 < 9' co ot co ^ co f^« co CA co *t o 01
?S3c5c?k $^3gSg ?SSScSR $33S!S8 ?3SJ??S
N. ^ eo -a>^fooQ' r^ d « r*. «o » h- eo o ca o p r-*. N- a» o» to U)
to^rcoc?c5cq in o cb eo co rt IOCOCOCMOJM m c5 ?5 cj w S 10 to w t- -r- i-


•s
 *

I
 a
•s

-------
a
8

I
           O
           |
           S

           <



           (0
    'H     fl
           (0
X)
•H


I
       (0
       3
      •o
          C
          o
          o
          o
         o
                                         1
                                                         ••»    %£«'**">    5 £ ~ <" «> »   555


                                                > ca 10 « o *•    o c o « co i-    o e K> « co <-   r>r.r>



                                                .!>.«> ««T-    co r» «J n co —    co r- o » co T-   ce> «o o 10 n ^>
                                                    I Ol O CM « »    CM Ot 0 CO Ol O   CM Ot f. CM « IO    CM ta«cM    o o o at o n
                                                                                       OOCOOVCM    o o a a 10 '
                                                  a> a « e <«•  n CM o 10   ogar^T-o
                                                  a at o CM o i
                                                                                       8555'
                                                  O CO CO CO CO CM    COCOCOCOCMCM   CO CO CO CO CO CM    CO CO CO CO CM CM
                                              co co •»•**»    o co co co co «    n to » » •» »    co co co co co CM
                                              o" d d d d d    d d d d d d    d d d d d d    d d d d d d
                                              r*»roiwr«>r<»i^    i^ror^ctipK)   r^i^Ki
                                              odd odd    dodo* do*   ddo'<
is.   f. r. ci 19 o i
d   d d d d d
                                                      icocococo         CM CM T- T- <» I
                                                                                         SO
                                                                                         CO
                                                  coh»|^oio^   h-rooto^cM    eocorvi
                                                  ooono«   o ai ot r- « CM    o o ct a i
                                                                                                         89 r«» *- *
                                                                                                         S CM CM ?
                                                                                                          a to o K> » c«
                                                                                                         o o a « 10 CM
                                                 OCCOOh-CO    CD e « CM O CM
                                                                                           ,^^N.CO    oa a o CM o CM
                                                 <- a n T- r~. «    CM ai o »- o CM    — a» m T- r~ »    co o o <-  S S o   3 5 S r~ r~ m    c*o>eorr>   
-------
4

if
        u
        0>
        4—
        (0
        =
        (3  U
        «  O
        w *:
       < .£
       <  w

       x 5
*j    C  «-
•H    M  (0
^    *«•  r~
•H    ,«  ^

I   |I


     to
      (0 "J


     i cc
      £ CO
      H-   I
       o   '
       (0
     V
     o
     c
     o
    o


                 m*
                                           8J38S
                                                                          o a 10 •*   «> ID <«• •*:  e *• n *<
                                                     ,- »
d d d d:
KSSS
* 0 •>:«»:
0 01 — -W
::••:•:

noon- o CM CM w

d o o d d d d d.
00 h. <0 .K': CO 0 0 f» 8
o o o d- d d d d
SS88 SSS8
* n n n *• o o M-

j
                                                                                                         g
                                         2 SEE:  2551?.  2SS8;   25
                                                               n *> n :  o v n o   «> n >-
                                                                                                       g
                                                     58:?:  88 S£:
                                        RSS8   SSSS;  R8S8:  SSSS:  RSSR;   S2S
                                                                                 sss
                                        8888:  SSSS
                                                                                 ^ rg n «   ^ M a
                                                                               8888.  222S
                                                                                                        1

                                                                                                        1
                                                                                                        «

-------
?
I
   (0
   O
   o
   < o
   «'•§
     *
   « CO
   II
     co
    o
   O

1 — not r*a,ublnfl further kMtnwnt modification
2 - •Imk»t»p»«-chkxir»tkHi
3 — •Imlnate pia-ehtof + add ammonia
4 — pr«— chlor + ammonk + alum doc*
S - pca-chk* + ammonia + alum doM -f OZOM
8 - cxa-chk* + ammonta + alum dot* + OZOIM + QAC

|


«
if
|
>
3
d
S
«
I
I
I


858888 SSSSt* 825222 82?2a« 822<»-"> aSS*"-
eeoeev «o o o o » » o e e « • • o in e «> » w e « « •»<*•« o » 10 « « w
~- y> '*> 4"
''X ,4 '} -^

SSSSSSa 5S822? 555SS? 5X5=225 «SC:2?"; 5SJ:5^«
T ' •bv'' f ** f
>-f ' > ,V
: , X* "- "•-*
£ "§55358 S5S8S8 S5855S SJSSaS S58S8S S?SS8^
^ ' '"•=• \ , -?
; "" ^ f "' "*
c *-^--^— *pp9S5? »»«•!«.».- s«5:»»» S525S!? aSS™ 2'*

i '•& f * *
II i ';;.: ' ,» ' •
^ M* ^ • ff* m tn -^^' ^ ^ n C9 CO 4£ ^ C9 C9 n C9 49 ^ 49 CI N M (V *^ A CI ^ ^ M
• SSddod dddodS ddddoS ddddoct dddddet ddddod
,'.'.'•
5*" 'S'-*:: S- j
E ?;* -
r ~ ^ ^ ^ ^ K. CD r^ r- r- r- r. cooh-h-h-N. eoh.oeei« WKOOWO eooio^tift:
dddddd dddddd; dddddiS: ddddd* dddddet ooooo*
! ' '•'•'•; *• '»•*•>
5SS88S8 KS888S; SS8888 K88888 KS88S8 KSSSS5
e *o««on *«nn«W; •» « « CM ci:«* » n w « «:«!: » n »- »- - T* * '« »- *• " T-
| | ,: -:
E ,x«
e 0«,^^^^ wn^^^^: COM^^,.,. ncjooow n ^ a o a « o j- « K K K
i ' -s - • • ^ I
« ssaasra. ssaaaa saaaas sa8aa.a 551222^ Has"1""


fi 2SSSSS 25SSS? 25S8S* 2C88SIS 25S22? 25S22"
* '4 :-|
| sssaaai ssasss satsss 52222?: 322"-* 822»«:-
£ pggigSi CSS888: SSSSS8 SSSSSg SSSPKR SSCSSS;
8 *- " .;.:-.: *"":;'" , ^ , - "•
5 s;sssi §?sss8; ssjgciRife ssaassi; ?sa22e ssas^*
• •' ' '
_.^. ^. ** ^ *+ ** t* m«mfK91CMB. W) 0 O •> M M
• S ? 5 5 55 S"";?? SSfSSSS SSSS8R 8™R«S!8 loSST-^^
jsj ^ :^ ; -, "
g| ;, A

-------
-o
4
 

O
2

|

H  |
 o  c
 (0  O

II
      (0
  •H
  A
  •H
 4-   •o

 I§

 !l
 w m
 2

 02
 O

 O
o
                            !
                            I
                            835 2 £  S 2 1 1;  S 2 ** **  gjetiofl*  mono  eo a 10
                            Cf ^ ^»  CO *• «".^»:  CO T- ^ «*•  n T» ^   CO »• »••    n w ^



                                       :•:  !«• O O :«:  t» O
                                       ,. fa r« >.  T- c.  n ID <+:  co
                                                                  ^ a o ej
9
C CO CO CO C» CO CO to «• CO «0 CO »:
if "~~ " 	 -
E
! :g •**•«» v «• co •» *• «• co «•
' | M oooo oo do d d d efe
f I ci ci ei d ci o ci c> d d d ct
• SI ^ ^ r*. K h» r^ r^-K- K o o O'
1
!; S •*•»««» •*•«««»: •«• CO CO •»
II
It if
'1 E
E
§ g •«• CO CM CM : -T CO CM CM •»• CM CM ^
j
g| SSSS 5SS8 S S3 55 ft
1
f^
C CMT-^-O CM ^ O O CM O O 0
| E . >|
co to we* co 10 n »: co » o o
-
8S55: 825* 825*
CO CO CO 49 CO CO CO :•)•: CO CM CM CM
* CO CO * •«• CO CO «- r»co«» NOIOK
ooo d- cidooJ ci o ci ci
CMCftOCk: CMOIOICB CMOIOkCft.
f** O CO 9< r^ 4D CO C9' h« O 4D -9'
^«nw ^ o eo w ^ « c\i fci
^r CM *- T» ^* cj v o^ ^ •»- o-0-

ssaa sasjr sss- i
.
CM 0 O tS CM C» CO r^: CM Ol r> •«•
" " -, "
                                                                               g
                                                   i  SSSS:  SSSg:
                                             T- CM CO »  ^ CM CO V  T-CM««-  ^ CM CO »
                                                                              •8

                                                                              I
                                                                              •s

-------
   (ft

   0
   1.1
   v« *^
   0) Q
   ±2 0)
   < c
      Q
       O
      S
t  "*"  E
6  ?  o
4J   C 4-
5   « 5
'*   (ft *^
•*^  •»•  .r^

   PE
   "a §
   IS
   if
   •Si
    (ft in
    §  i
   1
   •»•*
    o
    o
    §
   u
                             ,_,_..,.—.  eaaaooe  g> a •> « « •  g a w £ <
                             ! w cd N N M  H T- i- <•-••-•«-  B ~ •-»-»• v  n *-~ i-
                                                        •  oa 10 ^ t» «•  ga 10 »- f* •*
                                                           «»-*•*•   ,  «i^*-*-   -v
                                                        .. ^         "-^     •    ••v-.
                                                        •  K IO e « « «•  r. IO .»»»«*  « » K r — :**:


                                        So o o •> »  o a a o »•«  » a a * «i e  w a o n o o
                  	   	        8855?  o» _--,-.,-  K,..-,-,-r  N-^^»-

             • I |f| nnonn-n-  nnnno:r*  nnnnnn  a o o« CTt«:

             ^E- _.».««« «.  vwnnn-n--  •»•»«««:•»  •»««€»«•»  ^tiojciiciw:  ^owoi^ni
                  ddddSS   dddddd  dddddret:  dddddrf  dddddxf  ddoooe';

             fc ~ i
              5- ^^^^^^.   »_i^>^»^i«.hi  h. r^ O O O O  t^h-OOOiW:  f".OOIOIO:«0-  KOIOt*^
                  dddddS   ddddSS  dddddS:  ddddo.eS;  dddddet  dddoo*
                E
              ,p ^pgggg   KfrgSSS  KS82S8  KS8888  SS8888  K!

                e *wcjcicj«i  *ndoioi«»  * n en 01 ft i»:  » n « ~ft-.m  »«CJ«^T'  *«^»-»-^.


                  jnojoiorw  *«~«««*  *Ji??^^  »«r^^?:  »5»°<»<»*  j-a-i-r*


                  ssaaaa  ss^aaa  saojojafc  58H5«:s; 535555  sat0-*""
                          •?                               ^
                   ooooo*  a««.«»*  a«r.««»o  a o iw o io:w. a e o a -r «r  a a 10 » n ft


                   01^.0000  oi»-oooo  «oaaa«  01 o o « o-«  M a t~ 
-------

-------

-------

-------

-------