United States
Environmental Protection
Agency
Office Of
The Administrator
(A-101F6)
EPA 101/F-91/047
February 1991
Nitrate Risk Management
Under Uncertainty
#90-2503
                              Printed on Recycled Paper

-------
                           DISCLAIMER

This report was furnished to the  U.S.  Environmental Protection
Agency by  the student identified  on  the  cover  page,  under a National
Network for Environmental  Management  Studies  fellowship.

The  contents are essentially  as received from the  author.  The
opinions, findings,  and conclusions expressed are  those of the author
and  not necessarily  those of the U.S. Environmental  Protection
Agency.  Mention,  if any, of company, process, or product names  is
not to be considered as an endorsement by  the  U.S.  Environmental
Protection  Agency.

-------
       NITRATE RISK MANAGEMENT  UNDER UNCERTAINTY
                           by


                      Yong W. Lee
            Department of Civil  Engineering
                  W 348 Nebraska Hall
                 University of Nebraska
             Lincoln, Nebraska,  68588-0531
This paper describes work performed under the National
 Network for  Environmental Management studies  (NNEMS)
          program (contract No. o-913265-oi-o)
         U.S. Environmental Protection Agencv
               library (PL.

-------

-------
                             ABSTRACT
    In many  areas throughout the U.S., groundwater  supplies are
contaminated by nitrates.  Nitrate contamination has been a subject
of concern because nitrate salt can induce infant methemoglobinemia
and cancer.    Nitrate  risk management  describes the  process of
evaluating  nitrate  control  strategies  and  selecting the  best
management scheme.   Available  nitrate  risk management strategies
(several potential nitrate control methods  and possible changes of
health risk level) can be investigated based on the acceptable risk
level and the reasonableness of the nitrate  control cost.  However,
the objectives of the risk reduction and cost are in conflict, and
each  phase of  the  risk  and  cost  analysis  is associated  with
uncertainty.   In this study,  a multicriterion  decision-making
methodology is  provided to assist decision makers  to determine,
under uncertain information,  which of the nitrate risk management
strategies "best" satisfies the reduction of both health risk and
cost.      A  numerical  example  is illustrated  to  show how  the
methodology can be applied to solve nitrate  risk management problem
under uncertainty.

-------

-------
INTRODUCTION
    The purpose of this study is to examine the problem of choosing
the "best" nitrate risk management scheme using a fuzzy composite
programming.   Nitrate contamination in groundwater  supplies has
been a  subject of concern  due  to the finding that  nitrates can
cause infant methemoglobinemia and human cancer.  Most nitrates in
groundwater  can  be  traced  to  the  excessive  application  of
commercial fertilizers.  However,  some nitrates are contributed by
the fixation of atmospheric nitrogen by plants, industrial wastes,
domestic wastewater, and animal wastes.   Generally, nitrate salts
reach groundwater by percolation through the soil.
    To  reduce  nitrate risk from  groundwater supplies,   several
strategies can be developed based  on the acceptable level of human
health risk, the  reasonableness of nitrate  control cost,  and the
technical feasibility  of nitrate control methods,  where  nitrate
control methods include nitrate  source control, development of new
water supply,  blending  two or more water  supplies,  and  direct
treatment of nitrates.   High  cost strategies may  provide  a high
degree of human health risk protection,  while low cost strategies
may not  provide adequate protection. In other words, the objectives
of the risk  reduction and cost are in conflict with each other.  In
addition, the objectives may be of varying degrees of importance.
Thus, the  ultimate  goal of the  nitrate risk  management  is  to
determine,   under  the different  importance  of objectives,  which
strategy "best" satisfies the  reduction  of both human health risk
and nitrate control cost (a risk versus  cost trade-off analysis).

-------
                                                                2
    In nitrate risk management process,  each phase of human health
risk analysis, and possibly nitrate control  cost and the technical
feasibility of each control method, is associated with uncertainty.
Sources  of  uncertainty include exposure assessment,  interspecies
conversion, dose-response assessment, and inaccuracy of data used.
The uncertainty in problem  variables (risk, cost and technical
feasibility) and its impact on results can be represented with an
application of fuzzy set theory.
    The  concept of  fuzzy set  theory  is  briefly introduced in
Appendix I.  The fuzzy  set approach, pioneered by Zadeh  (1965), has
been widely  applied to solve,  under uncertainty,  decision-making
problems (Watson et al., 1979; Zimmermann, 1985) and, specifically,
multicriterion decision-making problems  (Yager,  1977;  Bogardi and
Bardossy, I983b; Anandalingam and Westfall,  1988).
    In this study,  a fuzzy composite programming method is used to
formulate  a  methodology  for   solving   nitrate  risk  management
problem.  The fuzzy composite programming method has  been used as
a useful tool  to solve decision-making  problems where there are
conflicting objectives; values  of problem variables are uncertain;
and the objectives  are  of  varying degrees of importance (Bardossy,
1988;  Lee  et al.,  1990).   This is  a multi-level  multiobjactive
programming method using fuzzy sets.

HEALTH EFFECTS OF NITRATES
    Nitrate salts themselves  are probably harmless to humans.  The
human health hazard of  nitrates results from  the nitrites formed by

-------
the  bacterial  conversion of  ingested  nitrates.   These nitrites
induce methemoglobinemia and possibly cancer.

Infant Methemoglobinemia
     Nitrate ion has been known as the causative agent of methemo-
globinemia, which is more commonly called the "blue baby" syndrome.
This  disease  usually  occurs  to infants  and pregnant  women,  in
particular infants under 3 months of age are  more susceptible (WHO,
1978).   Sine the  pH of the infant's gastric juice is relatively
high  (from 5 to  7) , nitrate-reducing  bacteria  can live  in  the
infant's stomach rather than being limited to the intestines as in
older children and adults (Andersen,  1980).  As shown in Figure 1,
nitrate is reduced to nitrite by the nitrate-reducing bacteria in
the  stomach,  and the nitrite  is  absorbed in blood  and converts
hemoglobin (a protein in red blood corpuscles) into methemoglobin,
which can not carry oxygen to the body's tissues.
                             Nitrate
                                  nitrate-reducing bacteria
                             Nitrite
           Hemoglobin
methemoglobin
                       methemoglobin-reducing
                       bacteria
   Figure 1. Basic Reaction in the Development of Methemoglobinemia
            from Nitrate  (Winton et al.,  1971).

-------
    The American Public Health  Association  performed a survey of
infantile nethemoglobinemia cases (Walton, 1951).  Results of the
survey are summarized  in Table  1.   Table 1  also shows results of
the study conducted  by Winton et al.  (1971). Based  on these two
studies, the methemoglobinemia can occur to infants consuming water
which contains nitrates over 10 mg/1 (as N).

                Table  1. Nitrate Toxicity Studies

Nitrate-nitrogen
concentration (mg/1)
0-10
11 - 20

20 - 50
51 - 100
> 100
< 1.0

1.0 - 4.9
5.0 - 9.9
10.0 - 15.0

Number of humans
investigated

Of 214 data
available



63

23
20
5
Number with
methemo-
globinemia
0
5

—36
81
92
0

0
0
3(*)

Source

Walton
(1951)




Winton et
al. (1971)


(*) infants with methemoglobin  level  of  5.9  % higher than normal
level of  about 1.6 %.   A level greater  than 3 %  is  defined as
methemoglobinemia  (U.S. EPA, 1987).
Human Cancer Risk
    Nitrites, produced by the reduction of the nitrates, react in
the  stomach with  amines  and amides  to  form nitrosamines and
nitrosamides, which  induce  cancer (Mirvish, 1977).   There is no

-------
                                                                 5
conclusive evidence that nitrates are responsible for human cancer
yet.  However, some studies have shown that nitrate  ingestion can
be  the  cause of  human cancer  (Mirvish,  1977; Lu  et  al.,  1984;
Crespi and Ramazzotti,  1990).
    Since there are no data related directly to cancer incidence in
humans,  data obtained from animal experiments can  be applied  to
estimate  the  human   cancer  risk  caused  by nitrates.     This
interspecies conversion is performed under  the  assumption  that
nitrosamines are as carcinogenic in humans as  they are  in animals.
Although the interspecies conversion  is a controversial  issue,  it
may  be  the  "only"  method,   short  of  extensive epidemiological
studies,  for modelling the relationship  between nitrate  dose  to
humans and its corresponding  cancer response.
    Lee et al. (1990b)  performed a study to estimate human cancer
risk corresponding to a particular dose of nitrate.  In  this study,
animal data  obtained by Terracini et  al.  (1967) were used for the
interspecies conversion (from  animal to human), and  a  combined
probabilistic/fuzzy  set framework is used  to represent  the un-
certainty in values reguired for the interspecies conversion. Based
on  the  study, the relationship between  human nitrate  dose  (X,
g/day) and its corresponding  cancer response  (Y)  is  as follows:
       (1 +  expfZ,))'1 <; Y < (l + exp(Z2))-1                   (i)
where:
      Z,  = 3.331  + (1-p) 1.138  - 3.429(logcX) + (1-p)0.881|logftX|
      Z2  = 3.331  - (l-p)1.138  - 3.429(logeX) - (1-p) 0.8811 logeX| .
where the response Y is the probability of developing cancer,  and

-------
the value  p  (0  <  p £  1)  is used  to  define the  domain for  the
interval value of the response Y.  Thus,  the response Y is a fuzzy
number (fuzzy probability) that  belongs to  a given set (interval)
with a degree of membership  (p).   Figure  2 shows the  lower  and
upper bound lines for the response Y calculated by  using the value
p = 0.5 which is neither optimistic nor pessimistic.
      0.0
        0.0     0.8     1.6     2.4    3.2    4.0    4.8
                       HUMAN DOSE OF  NITRATE  (g/day)
5.6
6.4
   Figure 2. Response versus Human Dose of Nitrate  When p = 0.5.
NITRATE CONTROL METHODS
    Some studies have  identified methods  for  reducing or removing
nitrates  from water supplies  (Musterman and Bergo,  1980;  Sorg,
1980; Dahab,  1987).  In this section, the nitrate control methods
are reviewed  as follows:

-------
Nitrate source Control
    Contamination of  groundwater  by nitrate can be  decreased by
efficiently managing the agricultural, industrial and urban systems
which generate nitrates.  In other words,  nitrates can be reduced
by decreasing the use  of commercial  fertilizer in the agricultural
area and including a nitrification/denitrification process in the
wastewater  treatment  systems.    However,  if  these  methods  are
implemented,  a very  long period of time would  be required to
improve the quality of groundwater contaminated by nitrates because
of the large storage capacity for  nitrate  in the soil and the long
residence  time of  groundwater.    In  addition,  it  seems to be
economically and politically infeasible to  enforce stringent enough
fertilizer control actions which would reduce the contamination of
nitrate.

Development of New Water Supply
    The method of developing new water supplies can be considered
if nitrate-free water sources  exist in the area.   First, surface
water can be a source  of the new water supply. However, undeveloped
or partially developed surface water supplies can  be expensive
because of  the expenditure required to remove other contaminants
such as organic materials and suspended solids. Second, new wells
can be constructed to draw water from a nitrate-free  aquifer.  The
disadvantage of this method is that the quality of water drawn  from
the new wells  might be  changed with pumping and time.   The third
alternative  is to purchase water from a neighboring community.
However, in this case the community  would  have no control over its

-------
                                                                8
water  supply  and  would  become  dependent  on  the  neighboring
community.   Finally, bottled water could  be used for individual
systems or very small communities.

Blending Two or More Water Supplies
    Blending nitrate-contaminated water with uncontaminated water
of different source can  be a reasonable approach if a "good" source
of  water exists  in the  area.   To  implement  this  approach,  a
blending  facility and  new piping  systems  should  be constructed,
which may require a  large expenditure.

Treatment of Existing Water Supply
    Techniques for nitrate treatment include ion exchange, reverse
osmosis,  biological denitrification,  electrodialysis,  chemical
reduction, and distillation.   Currently, the first three mentioned
above can  be considered practical  and economical.   Those three
techniques are discussed below.

Ion exchange
    The nitrate treatment process by ion exchange is accomplished
by contacting nitrate ions in solution with the ion exchange resin
which in turn releases sodium ions  into the solution.  The nitrate
removal efficiency by this process varies with the quality of the
water.  For example,  high sulfate water can lower the efficiency of
the process because  ion exchange resins  generally prefer sulfate
ions over nitrate ions for exchange.  Thus, high sulfate water is
more costly to treat than low sulfate water.   As a result, if the
water is low in concentration of total dissolved  solids, the ion

-------
                                                                9
exchange technique for nitrate removal is practical and economical
(Sorg, 1980).   In addition, the cost of the brine disposal should
be considered  before the ion exchange  process is applied  for a
full-scale system. In some situations,  disposal of the regenerant
brine may account for a large fraction of the overall cost of the
process.  Potential  alternatives for the brine  disposal include
discharge to   the local  municipal wastewater treatment  plant,
application to the land,  transport  to  the  other  treatment plant,
and disposal in the ocean.
    In the  United States,  several  full-scale ion exchange plants
were  operated  at the Long  Island  in  New York (Gregg,  1973)  and
currently at  several locations  in  California (Lauch  and Guter,
1986; Guter and  Kartinen,  1989).  Each of the plant operates to
treat the part of the influent  water and blend the treated water
with  the  raw  water so  the  blended effluent water has a nitrate
concentration lower than the drinking water standard of 10 mg/1  (as
N).
Reverse osmosis
    Reverse osmosis  is a process  which is used  to  remove ionic
species (e.g., nitrates) from the water by forcing the water to be
transported across a  semipermeable membrane.  The nitrate removal
efficiency by this process  is dependent on operating pressure and
the type  of membrane.   This process involves high cost resulting
from  the  intensive energy requirement.  However,  the process is
fairly  effective for  other processes  under  certain situations.
That  is, the use  of the reverse osmosis process is warranted when

-------
                                                               10



the water  supply  is high in total dissolved solids,  and/or when



nitrate  and other  contaminant(s)  removal from  water supply  is



required (Sorg, 1980).  There are some full-scale reverse osmosis



plants in  the  United States such as one  in the  Greenfield, Iowa



(Laverentz, 1974)  and in the City of Cape Coral,  Florida (Ashton,



1986).  However, none of the plants  was  specifically designed to



remove nitrate.





Biological denitrification



    Biological denitrification process is conducted by contacting



facultive microorganisms with the water supply containing nitrates.



Under this  contact,  the microorganisms utilize the nitrates as a



terminal  electron  acceptor  in  place  of  the  molecular  oxygen,



reducing  the nitrates  to  the  nitrogen   gas which  is  harmless.



Although this process has a well established history in the realm



of wastewater treatment, full-scale application of the process in



the U.S. has not been introduced to  the  field  of water treatment



due to the potential contamination by organics  and microorganisms



of the treated water supply.   There  are  large-scale plants being



operated in Europe  (Rogalla et al., 1990).



    With laboratory-scale experiments, Dahab and Lee (1988) have



studied  the potential  of  using biological denitrification  for



nitrate removal in groundwater supplies.   The result of the study



showed that the concentration of nitrates as high as 100 mg/1 (as



N) can  reduced to  levels below  1.0  mg/1.  This high efficiency



(nearly  100 percent) can  not be  practically  achieved  by other



processes  available for  nitrate removal.   Furthermore,  wastes

-------
                                                               11
(excess biological growth) produced from the process is much easier
and less expensive to dispose of than  wastes  produced from other
nitrate  removal  processes   (Dahab,  1990).    Consequently,  the
biological denitrification process  can be used  as a pretreatment
step in  a water  purification system because the post treatment
steps are required to remove residual organics and microorganisms.

METHODOLOGY FOR NITRATE RISK MANAGEMENT
    In this section, a methodology is provided to assist decision
makers to determine  the  "best"  nitrate risk  management strategy,
that is,  the  one that best  satisfies  the objectives  of problem
variables  such as  risk, cost  and  technical  feasibility.    To
formulate  the  methodology,  fuzzy composite  programming  method
(Bardossy, 1988;  Lee et  al.,  1990) is  employed.   Fuzzy composite
programming organizes  a  problem  into  the following  sequential
format:    1)   define  management  strategies  2)   define  basic
indicators,   3) group basic  indicators into  progressively fewer,
more  general,  groups,   4)   evaluate  and  rank  the  management
strategies.

Selection of Basic Indicators
    To  represent  both potential  benefits and  problems  of  each
nitrate risk  management  strategy, basic  indicators  are selected.
These indicators are then used as the input variables to evaluate
each strategy.   This example is shown in Table 2.   Because the
selection of the basic indicators tends to  be  case specific, it is
difficult  to  generalize.   In  other  words,  the  kind  of  basic

-------
                                                               12
indicators varies with the nitrate control methods applied, and the
number of basic indicators depends on the  level of analysis desired
(preliminary or detailed).
          Table 2.  Basic indicators and  their Definition
Basic indicators
Methemoglobinemia
Human cancer
Adverse impact of
performance
Status of technology
Ease of operation
Cost
Definition
The level of methemoglobinemia risk
associated with nitrate contamination.
The level of human cancer risk related
to nitrate contamination.
Adverse environmental impact that occurs
by performing a process to remove
nitrates from water supplies.
Suitability of the method used to remove
nitrates, and assessment of whether or
not its technology is fully developed.
Evaluation of whether the employed
method or process is easy to operate.
The expenditure required to reduce
health risks and adverse environmental
impact below specified levels.
Composite Procedure
    The composite procedure involves a step by step regrouping of
a  set  of  various  basic  indicators  to form  a single  indicator
(Bogardi and Bardossy, 1983a).  Figure 3  shows  an example of the
indicator grouping.  The set of basic (first-level) indicators is
grouped into  a smaller subset  of second-level indicators.   For
example, the basic  indicators such  as risks of methemoglobinemia
and human cancer can be grouped  into human health risk, an element
of the subset of second-level indicators.   The status of technology

-------
                                                                 13
and the  ease of  operation are grouped  into the technical feasi-
bility, the other element of the subset of second-level indicators.
    Further, second-level indicators such as human health risk and
the adverse  impact of performance are grouped  into environmental
risk,  an  element  of  the  subset  of   third-level  indicators.
Similarly, the  final  indicator (system)  can be formed by composing
environmental risk,  technical feasibility and cost.
      Level 1
    Level 2
Level 3
Level 4
  Methemoglobinemia
    Human cancer
                        Human health
                           risk
   Adverse impact
   of performance
Adverse impact
of performance
                   Environmental
                      risk
 Status of technology
  Ease of operation
                         Technical
                        feasibility
                    Technical
                    feasibility
       Cost
    Cost
  Cost
              Figure 3. Example of Composite  Structure.
Trade-off Analysis
    The values of  the basic indicators for nitrate risk management
process  are  estimated  as  fuzzy numbers to  characterize their
uncertainty.  The  fuzzy numbers are values which belong to  a given
set (interval) with a degree of membership.   In order to evaluate
the various nitrate risk management strategies  under uncertainty,

-------
                                                                14


let Zj (x) be a fuzzy value  for  the i-th basic indicator, and  let


its membership function /jfZ,. (x)) be a trapezoid (Figure 4), where


x is  one element of the discrete  set  of nitrate risk management


strategies.  Note  that  if  the trapezoid is reduced to a vertical


line, it represents a so-called crisp number.   At  this time,  a


level-cut concept  (Dong and Shah,  1987) can be used to define  the


interval of each basic indicator at various levels of "confidence".


As shown in Figure  4, Z. h is the interval value of the i-th basic


indicator at level-cut h (i.e.,  a  < Z,  h < b) .
                                     i, n
                              most likely interval
           i.o H
             h -
           0.0
                                                  i   r
                                                 b
                          largest likely interval
          Figure 4. Fuzzy Estimate of the i-th Basic Indicator.
    Since the units of basic indicators such as risk and cost  are


different and thus difficult to compare directly,  the actual  value


of each  basic indicator  (Z, h(x))  should be transformed into  an
                            i ,n

index.  As shown in Figure 5, using the best value  (BESZ,-) of Z{  and


the worst  value (WORZ^)  of Z,  for the i-th basic indicator,  the

-------
                                                                   15
actual value Z, h(x) can  be transformed  into an index value Sf  WORZ,,  then
                            1      , Z,jh(x)  £ BESZ,
                     Z, h(x)  - WORZ,
                     BESZ, -  WORZi
-, WORZ, <  Z,fh(x)  < BESZ,.

 ,  Zj>h(x)  < WORZ,-
   2)  If BESZj  < WORZ,., then
                            1       ,  Zi/h(x)  < BESZ,.
                     Zj>h(x)  - WORZ,
                      BESZj  -  WORZj
 , WORZ, <  Zijh(x)  < BESZj

 ,  Zi/h(x)  >
                                                                 (2)
                                                                  (3)
a) BESZj >
1.0 -
d •
S|fh<*>
c •
0.0
b) BESZ,. <
1 f> J
d -
S|fh(*>
c •
0.0
WORZj
^^
^^
^^
WORZ,. a b BESZ, zi,h(x)
WORZ,
""\^
"^^
^^^^
BESZ, a b WORZ,
Figure 5. Transferring the actual value Z, h(x) into an index S.>h(X)

-------
                                                                16

    To assign the best and worst values (i.e., BESZ{ and WORZ.)  of

the i-th basic indicator,  we can use one of two options.   The first

option is  to  assign the best and worst values of the i-th  basic

indicator according to the overall best and worst values of the i-

th basic indicator, among the given strategies.  The second option

is to assign the best and worst values of  the  i-th basic  indicator

according to the opinion of expert.   Since the actual value Z.jh(x)

is an interval with lower bound a and upper bound b (i.e., Zf h(x)

e [a, b]), the index value S? h(x)  resulting from Zi h(x)  is  also an

interval  (i.e., S,. h(x) e [c,  d])  (Figure 5).

    Next.   index   values,   Lih(x),   for  second-level   composite
                             J »n

indicators  can  be calculated by using the index values of  basic

indicators, or:

                nj
    Ljrh (x) = C =wi.j (Si,h,j(x»PJ  1<1/V                         («>


where:
       n, = the number of element in the second-level group  j,

       S,. h - = the  index value  for the i-th  basic indicator in the
              second-level group j of basic indicators,

       w{ - = the weight reflecting the importance of each of basic
             indicators in group j; Z w, j = 1, and

       p. = the balancing factor  among indicators for group  j.


    Weights represent the relative importance between indicators in

a group.  The greater the importance  of an  indicator, the  greater

should be the weight assigned to it.  The balancing factors are

assigned  for each group  of indicators.     The balancing factor p

(p > 1)  reflects  the importance  of the maximal  deviations of the

-------
                                                                17

 indicators,  where maximal deviation means the  maximum difference

 between  an indicator value and the best value for that indicator.

 The processes  available for determining the  weights and balancing

 factors  are given in detail elsewhere  (Lee et al.,  1990).

      Using the index  values  for second-level indicators,  index

 values, Lkh(x), for third-level composite indicators can be defined

 as:
                 nk
    LM  <*>  =  t.^J-k                                (6)
             K™~i

where:

       Lk.h ~ the i°dex value of the third-level group  k  in the
             final group (system),

       wk = the weight representing the  importance among elements
            in  the final group, and

       p = the balancing factor for the  final group.

-------
                                                                18

   Note that the process of computing the index value is applicable

to fewer or more than the four levels used  in  this  study.


Ranking Management strategies

    As  the first  step  for  determining  the  ranking  of  various

nitrate risk management strategies,  let L(x) be the fuzzy number

representing the final composite indicator (system)  of  strategy x.

In other  words, the  index value,  L(x),  of  the final  composite

indicator is given as a  fuzzy number.  With the  help of two index

values  L^fx)  and  I^.gfx) (Equation 6),  the membership  function,

M(L(x)), of the fuzzy number L(x) can be  approximately calculated

from the piecewise linear function  (Figure  6):
                   L(X) - R^

                   	»  Rmin * L
-------
                                                                19

by applying  a ranking method  such as the  one  developed by  Chen

(1985).  This method determines the ranking of n fuzzy  numbers by

using a maximizing set and a minimizing set (Figure  7).
         0.0
                                                     L(x)
                                   Rmax ~" ^in

        Figure 6. Membership Function of Fuzzy Number L(x)
   membership
      value
    1.0 -
    0.0
                                max ~"  min
              Figure 7.  Ranking Method of Fuzzy Numbers.
                                           O  left utility value
                                               right utility value

-------
                                                                20

    The maximizing set M is a fuzzy subset with membership function

   given as:



             *"      TI"^'  Lmin - L - ^

                     0     ,  otherwise.

where  for  x =  1,  ...,  n,   Lmjn = min  [  min L^x)]   and Lmax = max

[max L^o (x)].

Then, the right utility value,  UR(x), for strategy x is defined as:

    UR(X)  = max{ min [MM(L), M(L(x))]}                         (9)

The minimizing set G is a fuzzy subset with membership function MG

given as:


                          '   min —   —  max
                     - Lmax                                    (10)

                     0     ,  otherwise.

The left utility value, UL(X),  for strategy x is defined as:

    UL(x)  = max{ min [MC(L) , /i(L(x))]}                         (11)

The total utility or ordering  value for strategy x  is:

    U(x) =  (UR(x)  + 1 - UL(x))/2                               (12)

The strategy that  has  the highest ordering value in  the  discrete

set of strategies is then  selected as the  "best" strategy.



ILLUSTRATIVE EXAMPLE

    A community with a nitrate water quality problem is chosen for

the illustrative example.   Currently,  the community  consumes one

million gallon  of  water per day  (MGD) ,  and the community's only

water source is groundwater that has nitrate  level  of 20  mg/1 (as

-------
                                                               21
N), sulfate level of  120  mg/1,  and  total  dissolved solids of 800
mg/1.
    Table 3 shows  five strategies available to  the community to
reduce  the  nitrate  level  in  the  community's  water  supply.
Strategies 1 and 2  treat all of the influent water (1.0 MGD) by
using biological denitrification  and  ion  exchange, respectively.
Strategies 3 and 4  treat 80 percent  of the influent water by using
biological denitrification and ion exchange, respectively, and the
treated water  (0.8 MGD)  is blended  with the  raw water (0.2 MGD).
These treatment rates  are  determined to achieve the target level of
nitrate in the  finished water (acceptable human health risk level) .
To calculate the treatment rates, it  is assumed that the nitrate
removal efficiency is 85 - 95  %  for biological denitrification and
60 - 80 % for  ion exchange.   Strategy  5 is  to blend the existing
water with groundwater of different  source.  The amount of ground-
water which can obtain  from the different source is 0.6 MGD, and
the water includes nitrates in the range  of  0 to 6 mg/1  (as N).

   Table 3.  Nitrate Risk Management Strategies for the Example
            Community
Strategy
1
2
3
4
5
Nitrate control method
Biological denitrification
Ion exchange
Biological denitrification
Ion exchange
Blending two water supplies
Nitrate in the finished
water, mg/1 (as N)
1-4
4-7
4-7
7-10
8-12

-------
                                                               22
  Table 4. Basic indicator values for five different strategies
strategies
1
2
3
4
5
basic
indicators
(1)
(2)
(3)
(4)
(5)
(6)
(1)
(2)
(3)
(4)
(5)
(6)
(1)
(2)
(3)
(4)
(5)
(6)
(1)
(2)
(3)
(4)
(5)
(6)
(1)
(2)
(3)
(4)
(5)
(6)
most likely interval
low
1.0
5.5xlO'10
11.4
0.4
0.2
950
0.9
l.lxlO'8
4200
0.6
0.5
570
0.9
l.lxlO'8
9.1
0.4
0.2
760
0.8
6.2xlO"8
3500
0.6
0.5
456
0.2
1.2xlO'7
0.0
0.9
0.8
523
high
1.0
9.2xlO'8
12.6
0.6
0.4
1050
1.0
l.OxlO'6
4640
0.8
0.7
630
1.0
l.OxlO'6
10.1
0.6
0.4
840
0.9
3.5xlO'6
3870
0.8
0.7
504
0.4
5.7xlO'6
0.0
1.0
0.9
578
largest likely interval
low
1.0
l.7xlO"11
9.6
0.3
0.1
800
0.8
3.4xlO"9
3540
0.5
0.4
480
0.8
3.4xlO'9
7.7
0.3
0.1
640
0.6
3.0xlO'8
2950
0.5
0.4
384
0.0
4.9xlO'8
0.0
0.8
0.7
440
high
1.0
3.7xlO'7
14.4
0.7
0.5
1200
1.0
2.0xlO"6
5300
0.9
0.8
720
1.0
2.0xlO"6
11.5
0.7
0.5
960
1.0
5.7xlO'6
4400
0.9
0.8
576
0.5
9.9xlO"6
0.0
1.0
1.0
660
* (1); methemogloblnemia,  (2);  cancer risk,  (3); adverse impact of
  performance, (4); status of technology,  (5);  ease of operation,
  (6); the total cost required  for capital and O & M (dollar/day).

-------
                                                               23
    Table 4 contains data of the basic indicators for each of the
strategies.  Data on cancer risk are calculated by using Equation
1,  and  cost data are obtained from the  literatures (Sorg, 1980;
Dahab, 1987; Guter and Kartinen,  1989).   The cost data assigned to
strategy  5  are  hypothetical.   Data  on the  adverse impact   of
performance correspond  to  the volume (m3/day)  of wastes produced
from  the process  used to  remove nitrate.    Data on  methemo-
globinemia, the status of technology and  the ease of operation are
assigned  by transferring qualitative definitions  into numerical
values. For example, the basic indicator "status of technology" can
be  characterized with qualitative  definitions such as bad, good,
excellent,  and numerical values  corresponding to the qualitative
definitions can be obtained using Figure 6.
   Numerical
    value
  1.0 -I
  0.5 -
  0.0
         bad
quite   good    very
good            good
excellent
qualitative
definitions
Figure 6. Transferring qualitative definitions to numerical values.

    As shown  in Table  4,  the value  of  each basic  indicator is
represented  by two  intervals  (i.e.,  most  likely  interval  and
largest likely interval) to reflect the uncertainty in each basic

-------
                                                               24



indicator.  These intervals are used  to  construct the membership



function (Figure 4)  for each basic indicator.  Table 5 displays the



best and worst values of  the six basic indicators, and the weights



and balancing factors assigned to each indicator are shown in Table
6.
                    Table 5. Best and Worst Values
Basic Indicators
Methemoglobinemia
Human cancer
Adverse impact
Status of technology
Ease of operation
Cost
Best value
1.0
l.7xlO'11
0
1.0
1.0
384
Worst value
0.0
9.9X10'6
5300
0.3
0.1
1200
           Table 6. Weights (w) and Balancing Factors (p)
Basic
indicators
Methemoglobinemia
Human cancer
Adverse impact
of performance
Status of
technology
Ease of
operation
Cost
w
0.6
0.4
1.0
0.5
0.5
1.0
P
1
1
2
1
Second- level
indicators
Human health
risk
Adverse impact
of performance
Technical
feasibility
Cost
w
0.9
0.1
1.0
1.0
P
2
1
1
Third-level
indicators
Environmental
impact
Technical
feasibility
Cost
w
0.4
0.3
0.3
P
2
Final
indicator
System

-------
                                                               25
    With  the help  of Tables  4,  5  and 6,  the fuzzy  composite
programming software developed by Lee et al.  (1990a)  was used to
determine which  of the  five  nitrate risk management strategies
"best" satisfies the objectives of basic indicator variables.  The
software  is  microcomputer-based  and  utilizes  the  methodology
described  in the  previous  section.    Using  the  software,  the
ordering value for  each of the five strategies can be calculated as
shown in Table 7.

      Table 7. Ranking of Nitrate Risk Management Strategies
Ranking
1
2
3
4
5
Strategies
Strategy 2
Strategy 4
Strategy 5
Strategy 3
Strategy 1
Ordering values
0.548
0.547
0.494
0.362
0.345
     The  microcomputer-based  software  provides  the  trade-off
information in numerical and graphical  form.  The graphical output
is displayed in the form of  a  box  so  it is easily interpreted by
decision makers.  Figure 7  shows an example for the boxes formed
between  environmental  risk  and  cost  for  each  nitrate  risk
management  strategy.    The  risk-cost  boxes  are produced  as  the
result  of  trade-offs  between  indicators  under  the  third-level
indicators, and the widths of the boxes represent the uncertainty
in trade-offs.  With the boxes formed between the risk and cost, we
can  identify  the status  of the  trade-off  analysis.   That  is,

-------
                                                                26


Strategy  1  provides the lowest  risk,  but it requires  very large


cost.  The  strategy 4 provides the  lowest cost,  but it does so at


a relatively high  risk.   Based  on the  widths  of  the risk-cost


boxes, Strategies  4 and  5 include relatively high uncertainty.


Using Equations  2 and  3,  the closer the  actual value of the i-th


basic indicator  is to  the best value, the closer  the index value


resulting from the actual value is to 1.0.  Thus, the "ideal point"


located at  (1,1)  in Figure 7  represents the lowest  cost and the


largest reduction in environmental  risk.
                                                              Ideal
                                                              Point
  O
 o
1 .U-
n q.
0.8-
n 7
0 6-
0.5-

0.4-
0.3-
0.2-

0.1 -
n n.
Strategy 4


Strategy 5












2




Strategy 3



W










Strategy 1


       0.0   0.1   0.2    0.3
 0.4   0.5   0.6

Environmental Risk
0.7    0.8   0.9    1.0
     Figure 7. Risk  versus Cost Boxes shown at Level-cut h=0.

-------
                                                               27



DISCUSSION AND CONCLUSIONS



    In this study, a multicriterion decision-making methodology was



developed  to  assist  decision  makers  to  solve  nitrate  risk



management problem under uncertainty. In the methodology, the basic



indicator values are transferred  into  fuzzy  numbers to represent



their uncertainty.  Using these fuzzy numbers, the uncertainty in



the basic indicator values can be incorporated into the trade-off



analysis.



    For the illustrative example,  strategy 2  is ranked as the best



of the five strategies  because  it has  the highest ordering value



(Table 7) .    Strategy  2 appears  to be  the  most cost  effective



alternative.  This can be demonstrated  by comparing with the "next



best" strategy for the cost.   Since  the ordering value of strategy



4  is close  to  the ordering value of  strategy  2,  it  may  be



reasonable to consider  a combined use  of  the strategies 2 and 4.



Combining the two strategies  means  changing  nitrate level in the



finished water because  both  of  them use ion exchange  process to



remove nitrate in the water supply (Table 3) .   Note that the final



result (ranking  of the strategies)  varies with the weights and



balancing factors assigned to each indicator and group.



    Results of this study lead to the following conclusions:



1. The uncertainty in the values of the basic indicators selected



   for nitrate risk management can be represented using the fuzzy



   set approach.



2. The fuzzy composite programming method can be a useful tool for



   solving  nitrate  risk  management  problem  where  there  are

-------
                                                               28



   conflicting objectives; values of the basic indicator variables



   are uncertain;  and the objectives  are  of varying  degrees of



   importance.



3. This study shows how the fuzzy composite programming method can



   be  employed to  formulate the  methodology  for nitrate  risk



   management.



4. Under the situation  of the illustrative  example,  ion exchange



   is  the  most effective nitrate  removal  method,  that  is,  the



   method shows the "best" compromise between environmental risk,



   technical feasibility, and nitrate control cost.







ACKNOWLEDGEMENT



    Research leading to this paper has been supported by the U.S.



Environmental Protection Agency  (contract No. U-913265-01-0).  The



author wishes to express  his sincere  appreciation  to Dr.  Bogardi



and Dr. Dahab, professors in the Department of Civil Engineering,



University of  Nebraska-Lincoln.   Their comments and suggestions



were very useful.

-------
                                                               29



APPENDIX I.- CONCEPT OF FUZZY SET



    Fuzziness  represents  situations  where membership in sets can



not be defined on a yes/no basis because the boundaries of sets are



vague  (Zadeh,  1965).   The central  concept of fuzzy set theory is



the membership function which represents numerically the  degree to



which an element belongs to a set.  In a classical set, a sharp or



unambiguous distinction exists between the members and non-members



of the set.  In other words, the value of the membership function



of each element in the  classical set is either 1  for members  (those



that certainly belong  into the set)   or  0 for non-members  (those



that certainly do not) .   However,  many  sets  such as the sets of



tall men,  beautiful  women, highly contaminated water or numbers



much greater than  1.0, do not exhibit this characteristics.  That



is, their boundaries are  fuzzy.



    Since the transition from member  to non-member appears gradual



rather than abrupt, the  fuzzy set  introduces vagueness  (with the



aim of  reducing  complexity)  by eliminating the sharp  boundary



dividing members  of  the  set  from  non-members  (Klir and Folger,



1988) .   Thus,  if an element  is a  member of a  fuzzy set to some



degree, the value of its membership function can be between 0 and



1. When the membership function of  an element can only have values



0 or 1, the fuzzy set theory reduces  to the classical set theory.

-------
                                                               30

APPENDIX II.- REFERENCES

  Anandalingam,  G.,  and  Westfall,  M.  (1988).  "Selection  of
    Hazardous  Waste Disposal  Alternative Using  Multi-attribute
    Utility   Theory  and   Fuzzy  Set   Analysis,"  Journal   of
    Environmental Systems, 18(1), 65-85.

  Andersen, D.R.  (1980).  "Nitrates in Water Supplies:  basis for
    concern  and  status  of  treatment  processes," Draft  paper,
    Department  of  Civil  Engineering,  University of  Nebraska,
    Lincoln, NE 68588.

  Ashton,  M.   (1986).  "Reverse Osmosis  can  be  Cost  Effective,"
    Public Works, 117(8), 55.

  Bardossy, A. (1988).  "Fuzzy Composite Programming," Proceedings,
    International  Conference on the  Foundations  of Utility and
    Risk, Budapest, Hungary.

  Bogardi, I., and Bardossy, A.  (1983a).  "Application  of MCDM to
    Geological  Exploration," In Essays  and  Surveys on Multiple
    Criterion Decision Making,  Ed.  by P. Hansen, Springer-verlag.

  Bogardi, I., and Bardossy, A.  (1983b).  "Regional Management of an
    Aquifer for Mining under Fuzzy Environmental Objectives," Water
    Resources Research, 19(6),  1394-1402.

  Chen, S.H. (1985) .  "Ranking Fuzzy Numbers with Maximizing Set and
    Minimizing Set," Fuzzy Sets and Systems,  17, 113-129.

  Crespi, M., and Ramazzotti, V. (1990). "Evidence that N-nitroso
    Compounds  contribute  to  the  Causation  of  Certain  Human
    Cancers," NATO workshop on Nitrate,  Lincoln, Nebraska.

  Dahab, M.F., and Lee,  Y.W.  (1988).  "Nitrate  Removal  from Water
    Supplies  using  Biological  Denitrification,  Journal  Water
    Pollution Control Federation, 60(9), 1670-1674.

  Dahab,   M.F.   (1987).   "Treatment  Alternatives  for  Nitrate
    Contaminated  Groundwater Supplies,"  Journal of Environmental
    Systems, 17,  65-75.

  Dahab, M.F.  (1990).  "Nitrate Treatment Methods: an  Overview,"
    NATO workshop on Nitrate, Lincoln, Nebraska.

  Dong,  W.  and Shah,  H.C.  (1987). "Vertex Method for Computing
    Functions of Fuzzy Variables,"  Fuzzy Sets and Systems, 24, 65-
    78.

  Gregg, J.C.  (1973). "Nitrate Removal at Water Treatment Plant,"
    Civil Engineering, 43(4), 45-47.

-------
                                                              31

Guter,  G.A.,  and  Kartinen,   E.O.   (1989).   "Alternatives   for
  Reducing Nitrate in Municipal Water Supplies," Presented at the
  Nitrate Drinking Water Symposium, Dallas, Texas.

Klir,  G.J.,  and Folger,  T.A.   (1988).  "Fuzzy Sets, Uncertainty,
  and  Information," Prentice Hall, Englewood Cliffs, NJ  07632.

Lauch,  R.P.,  and  Guter,  G.A. (1986).  "Ion  Exchange  for  the
  Removal of Nitrate from Well Water,"  Journal American Water
  Works Association,  78, 83-88.

Laverentz,  D.L.  (1974).  "Economic  Feasibility of  Desalting
  Systems for Municipal Water Supply in Iowa,  U.S. Department of
  the  Interior.

Lee,   Y.W.,  Bogardi,  I.,   and Bardossy,  A.   (1990a).   "Fuzzy
  Composite Programming Software and its Manual," Department of
  Civil Engineering,  University of Nebraska, Lincoln, NE 68588.

Lee, Y.W., Bogardi, I., and Bardossy, A.  (1990b). "Nitrate  Risk
  Analysis using Expert Knowledge-based Regression Method," Draft
  paper, Department of Civil Engineering,  University of Nebraska,
  Lincoln, NE  68588.

Lee,   Y.W.,   Bogardi,  I.,   and Stansbury,  J.   (1990).   "Fuzzy
  Decision-making  in  Dredged Material  Management,"  Accepted
  tentatively  by Journal  of   Environmental Engineering, ASCE,
  October.

Lu,  S.H.,  Bartsch,  H.,  and  Ohshima,  H.   (1984).  "In N-Nitroso
  Compounds:  Occurrence,  Biological  Effects  and Relevance  to
  Human Cancer," O'Neill,  J.K., von Borstel,  R.C.,  Long, J.C.,
  Miller, C.T. and Bartsch, H.  eds.  IARC Scientific Publications
  No. 57, International Agency  for Research on  Cancer, Lyon,  947-
  956.

Mirvish, S.S. (1977).  "N-nitrosocompounds, Nitrate, and Nitrite:
  possible  implications  for  the  causation  of  Human  Cancer,"
  Progress in Water Technology, 8,^195-207.

Mirvish,  S.S.   (1990).  "The  Significance  for Human Health  of
  Nitrate, Nitrite  and  N-Nitroso Compounds,"  NATO  workshop  on
  Nitrate, in Lincoln, Nebraska.

Musterman, J.,  and Bergo, D.A.  (1980).  "Nitrate Removal Processes
  and  their  Effectiveness,"  Presented at the Joint USEPA-IDEQ
  Public Meeting on Drinking Water Quality, Des Moines,  Iowa.

Rogalla,  F.,  de  Larminat, G.,  Coutelle,  J.,  and Godart,  H.
  (1990). "Experience with  Nitrate Removal Methods from Drinking
  Water," NATO workshop on Nitrate,  Lincoln, Nebraska.

-------
                                                             32

Sorg, T.J. (1980). "Compare Nitrate Removal Methods," Water and
  Waste Engineering, 17(12), 26-31.

Terracini, B.,  Mcgee,  P.N., and Barnes, J.M.  (1967).  "Hepatic
  Pathology  in  Rats on  Low Dietary Levels  of  Demethylnitros-
  amine," British Journal of Cancer, 21, 559-565.

U.S. Environmental Protection Agency, (1987). "Nitrate/Nitrite,"
  Health Advisory Office of Drinking Water.

Walton,  G.   (1951).  "Survey of  Literature Relating to  Infant
  Methemoglobinemia due to Nitrate-contaminated Water," American
  Journal of Public Health,  41, 986-995.

Watson,  S.R.,  Weiss,  J.J.,  and Donnell,  M.L.   (1979).  "Fuzzy
  Decision Analysis,"  IEEE Transactions  on Systems,  Man,  and
  Cybernetics, SMC-9(1), 1-9.

Winton, E.F., Tardiff, R.G., and McCabe, L.J. (1971). " Nitrate
  in Drinking Water," Journal American Water Works Association,
  63, 95-98.

WHO,  (1978).  "Task Group on Environmental  Health  Criteria for
  Nitrates,  Nitrites,   and N-Nitroso  Compounds,"  Nitrates,
  Nitrites, and N-Nitroso Compounds, World Health Organization,
  Geneva, 7-84.

Yager, R.R.  (1977).  "Multiple Objective  Decision-making using
  Fuzzy Sets," International Journal of Man-Machine Studies, 9,
  375-382.

Zadeh, L.A.  (1965).  "Fuzzy Sets,"  Information  and Control, 8,
  338-353.

Zimmermann, H.J. (1985). "Fuzzy Set Theory  and its Application,"
  Kluwer-Nijhoff Publishing, Hingham, MA 02043.

-------