Professional Documents
Culture Documents
Fuzzy Approach To The Environmental Impact Evaluation
Fuzzy Approach To The Environmental Impact Evaluation
www.elsevier.com/locate/ecolmodel
Received 21 September 1999; received in revised form 20 July 2000; accepted 17 August 2000
Abstract
Because of the imprecision of the ecological impacts and the frequent lack of quantitative information, fuzzy theory
provides a useful approach to the environmental impact evaluation. In this paper, the environmental parameters are
defined through fuzzy numbers. By considering the quality of the global component as derived from primitive
environmental components, for these primitive components the quality is generally derived from their fuzzy physical
parameters. Suitable operators are proposed in order to estimate the global environmental component quality as a
function of environmental primitive components quality that can decrease or increase the environmental quality. Also
in fuzzy form the magnitude of the plant impact factor that strikes the environmental components is esteemed. This
magnitude is a function of environmental parameters that can be fuzzy, as fuzzy can be the relation between these
parameters and the magnitude of the impact factor. Therefore, through the matrix method, the total environmental
impact (T.E.I.) in fuzzy terms is calculated, as in fuzzy form, the percentage of impact on every environmental
component is calculated. A criterion is proposed in order to compare fuzzy T.E.I. for various sceneries. This criterion
is based on a ranking method that considers the grade of prudence of the decision-maker and the acceptable risk.
Finally, a case study is reported as an explanation of the method. © 2001 Elsevier Science B.V. All rights reserved.
Keywords: Fuzzy set theory; Environmental impact; Ecological impact; Environmental decision making
0304-3800/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 3 0 4 - 3 8 0 0 ( 0 0 ) 0 0 3 8 0 - X
132 M. Enea, G. Salemi / Ecological Modelling 136 (2001) 131–147
represent the concepts of vagueness that natural environmental components are the elements
language uses for pointing out qualitative vari- struck by the impact factors typical of the plant
ables, such as warm, cold, lukewarm etc. considered. Some examples of environmental
A fuzzy set is defined by assigning to every components can be air, water, use of the territory,
object of the question a membership grade to the acoustic level and aesthetics.
whole that represents a concept (Zimmermann The quality is a characteristic of the environ-
and Zysno, 1991 Klir and Yuan, 1995). Such a mental components. Each component has an ini-
definition is very suggestive, because it allows the tial quality before the project begins and this
representation of some facts of a qualitative na- quality can be estimated subjectively through an
ture. In the case of the environmental quality, opportune numerical value according to a specific
representing it with a precise number (Hobbs et scale. This evaluation is often based on a relation
al., 1989 Salski, 1992) does not give credible re- between a parameter (e.g. physical or chemical)
sults. Such an observation induced us to try a and the quality level.
different approach in comparison to the ones The impact factors are those characteristics of
traditionally used in environmental impact evalu- the plant that can influence the environmental
ation (EIE). components. Common examples can be acoustic
Some authors employed the fuzzy logic in order issues, water pollution, ground occupation, trans-
to face environmental subjects. Salski (1992) in- port and production of solid waste. The impact
troduced the problem of uncertainty in ecological factor magnitude represents the environmental
research by fuzzy logic approach, applying ap- damage intensity caused by the impact factor
proximated reasoning to create a fuzzy knowl- considered. Such intensity can be expressed
edge-based model. Others have used it to map the through an opportune numerical value.
land by remotely sensed imagery (Foody, 1996 The assignment of these numbers (quality of
Metternicht, 1998); others developed fuzzy clus- the environmental component and impact factor
tering for the classification of chemicals in ecotox- magnitude) is the most subjective part of the
icological data analysis (Friederichs et al., 1996 entire analysis and is determined on the basis of
Melcher and Matthies, 1996) and of sites sus- an expert’s team judgment.
pected of being contaminated (Lhen and Temme, The matrix crisp EIE model begins with the
1996). individualization of the environmental compo-
Silvert (1997) describes an ecological impact nents of the site and of the plant impact factors,
analysis approach by fuzzy logic; he also proposes then proceeds with numerical estimate of the
fuzzy combination rules for partial memberships quality and of the magnitude impacts.
of impact and defuzzification rules. In the matrix method Q is the matrix of the
In this paper, a general environmental impact environmental components quality, where the qij
evaluation by fuzzy theory is proposed. This ap- element is equal to qi if the jth impact factor
proach can be adopted with many models of strikes the ith environmental component, other-
impact analysis. One of the simplest basic models wise zero if it does not strike; qi is the quality level
for the EIE is the matrix model (Caffrey and of the ith environmental component. G is the
David, 1975 Enea et al., 1994) whose input is the vector of the magnitude impacts, where the ele-
subjective evaluation of both the qualities of the ment gj represents how much the jth impact factor
environmental components and of the impact fac- strikes the environmental components. The
tors magnitude. To treat the fuzzy approach, it product matrix Q·G is the vector of the impacts
was opportune to start from such a model be- on the environmental components and the sum of
cause of its simplicity that leads to a better com- this element is the total environmental impact
parative valuation between the proposed and the (T.E.I.).
traditional model. Usually the values of the elements of the matrix
It is now opportune to make some remarks Q and the vector G are estimated very precisely,
about the nomenclature used in this paper. The although the expert assessment is vague, since
M. Enea, G. Salemi / Ecological Modelling 136 (2001) 131–147 133
If a fuzzy set on R possesses all the following To define the operators to compose the quali-
properties: ties of the primitive environmental components,
to be a normal fuzzy set; we note that usually the quality of the primitive
that hA is a closed interval for every h (0, 1]; environmental components is a function of ob-
that the support of A, 0 + A is limited; jects that add or subtract quality to the
it can be considered a fuzzy number. Special case environment.
of fuzzy numbers includes fuzzy intervals of real For instance, usually some flora and fauna
numbers. species (increasing types) add quality to the envi-
Let us consider a model with environmental ronment component, while gaseous or solid waste
components N and M as factors of impact and let (decreasing types) decreases it; we call the first
us assume that the jth environmental component increasing type and the second, decreasing types
has K primitive components. Since the quantifica- environmental components.
tion of the environmental component quality is So we can define primitive environmental com-
surely a vague concept, it seems appropriate to ponents as increasing or decreasing, that must
express such variables through a fuzzy set A:x follow different composition quality rules. To
[0, 1], where x points out a level of quality of the simplify this concept, let us consider the case of
environmental component that can be assumed to the presence in the same environment of two
vary between 1 and 10. The membership between different fauna species: one rare r and the other
0 and 1 quantifies how much the primitive envi- common c; if a= q(r) and b= q(c) are, respec-
tively, the quality level of these primitive environ-
ronmental component is at the level of quality x
mental components then obviously a b; then it
(Figs. 5–8 and 15). In such a way the quality of
is plausible in this case, in the calculation of the
the environmental component is qualified not by a
global environmental component quality, to give
precise value, but from a set of values, each one of
more importance to a rather than b.
them represents the quality of the component to a
In the case of two polluting gases, p1 and p2
certain degree of membership.
characterizing the air quality component, with
In the particular case, the fuzzy set that repre-
quality levels a and b, if the presence of p2 is more
sents the quality level can be reasonably chosen in harmful than p1, then obviously ab; then it is
such a way that it is a fuzzy number. For in- plausible in this case for the calculation of global
stance, the triangular or trapezoid memberships environmental component quality to give more
are fuzzy numbers that well lend in representing importance to b rather than a.
linguistic concepts. With the use of fuzzy num- Based on this observation, to define a specific
bers, fuzzy arithmetic operations can be defined. class of operators for each typology of primitive
As mentioned before, to determine the global environmental component, we list below some
component quality, we need to define a rule for properties relative to the operators for the compo-
the composition of the primitive component qual- sition of increasing and decreasing primitive envi-
ity. There are several numerical methods for the ronmental component qualities.
primitive quality aggregation; this paper points Let us suppose that the maximum value of the
out some functions for the quality aggregation, quality environmental component has convention-
based on the typology of the environmental ally been fixed equal to 10. The properties for the
component. quality composition f(a, b) of the increasing type
In the case that the primitive components qual- components would be:
ities are crisp, let q ipj be the quality of the ith 1. f(a, b)[0, 10]Öa, b[0, 10]
primitive component relative to the jth global 2. f(a, 0)= aÖa [0, 10] (boundary condition)
component. We can define a function q j =f({q pj}) 3. b 5c [ f(a, b)5 f(a, c)Öa, b, c [0, 10]
where q j is the resulting jth global component (monotonicity)
quality and {q pj}is a vector whose elements are 4. f(a, b)=f(b, a)Öa, b [0, 10] (commutativity)
q ipj. 5. f is a continuous function
M. Enea, G. Salemi / Ecological Modelling 136 (2001) 131–147 135
other quality component (also equal to a) can This function satisfies the following conditions:
also decrease the global quality (subidempotency) 1. f(a1, a2, a3,.....,an )[0, 10] Ö a1, a2, a3,....,an
because it is a decreasing type component.
[0, 10]
The function:
2. f(0, 0, 0,.....,0)=0; f(1, 1, 1,.....,1)= 1
f(a, b)=10 −min[10,((10 −a)w +(10 − b)w)1/w] (boundary condition)
3. f(a1, a2, a3,...,an )5 f(b1, b2, b3,..., bn ) if ai
w (0, ) (2)
5 bi Öi 05 i5 n
is a well-known class of functions (Yager, 1980)
(monotonic increasing in all its arguments)
as w varies, that has the properties (1 – 6) in terms
4. f(a1, a2, a3, . . . , an ) (is a continuous function)
of subidempotency, when w does not tend to ;
5. f(a1, a2, a3, . . . , an ) (is a symmetric function in
moreover it satisfies the associative property and
all its arguments)
is strictly monotonous. The function Eq. (2) in-
6. f(a1, a2, a3, . . . , an ) is an idemponent function
creases when w is rising.
that is f(a, a, a, . . . , a)=a
For w is:
The lower and the upper bound of this operator
f(a, b)= min(a, b) (2%) are min(ai ) and max(ai ) that are obtained for
h − and h .
When w0 we obtain the function fmin(a, b):
Another possible class of aggregation that cov-
fmin(a, b)= a when b =10 ers the entire range between the min(ai ) and the
max(ai ) operator is the ordered weighted averag-
fmin(a, b)= b when a =10
ing operations (OWA):
fmin(a, b)= 0 otherwise
f(a1, a2, a3, . . . , an )= % pi bi (4)
The function fmin(a, b) is the lower-bound of i
Eq. (1), while min(a, b) is the upper-bound with where bi for any i {1, 2, 3. . . n} is the ith largest
w, consequently: in a1, a2, . . . , an. That is the vector
b1, b2, b3, . . . , bn is a permutation of the vector
fmin(a, b)5 f(a, b)5min(a, b)
a1, a2, a3, . . . , an in which the elements are or-
The min(a, b) function has the properties (1 – 6) dered: bi ] bj if iB j for any pair
in terms of indemponent and (7). Such relation- i, j {1, 2, 3, . . . , n}. The vector p1, p2, p3, . . . , pn
ship in the specific case will express the concept is a weighting vector such that pi [0, 1] for all
that the quality of the global environmental com- i {1, 2, 3, . . . , n} and:
ponents is the minimum of the decreasing types
environmental primitive components, which %pi = 1 (5)
i
sometimes can be sufficiently considered as an
The lower and the upper bound of this operator
acceptable hypothesis.
are min(ai ) and max(ai ) that are obtained for the
In order to complete the set of numerical values
weighting vectors:
for f(a, b), it is necessary to consider when it is
opportune that f(a, b) has a numerical value be- p = 0, 0, 0, . . . , 1 and
tween min(a, b) and max(a, b); i.e. the resulting p = 1, 0, 0, . . . , 0
quality is a middle value between the quality of
Silvert (1997) proposed other classes of aggre-
primitive components. In this case, we can use
gation such as ‘the compensatory and function’
also the class of generalized means that covers the
by Zimmermann and Zysno (1991), as well as the
: ;
entire range between min and max operations.
‘symmetric sum’ by the same Silvert.
These are defined by the formula:
1/h
Now it is opportune to say something about the
%a hi extension principle (Zadeh, 1965). The member-
f(ai ) =
i ship rule that characterizes the elements of a
(3)
n general crisp set A¦U (universal set) can be
M. Enea, G. Salemi / Ecological Modelling 136 (2001) 131–147 137
established by the concept of the characteristic the extension principle or the a-cut
function A(x) taking only two values, 1 and 0, representation:
indicating whether or not x U is a member of A.
[ f(A, B)](z)= sup min(A(a), B(b))
Let us consider the following function of two real Ö(a,b) z = max(a,b)
(11)
numbers x and y:
h
( f(A, B))= {max(a, b) a, b(hA× hB)} (12)
z = f(x, y) (6)
Such an operation can be indicated syntheti-
If the real numbers x and y are extended to
cally as max(A, B)(z) not to be confused with the
crisp intervals A and B, respectively, then, the
standard fuzzy t-conorm max(A(x), B(x)).
usual function can map the crisp intervals A and
If we select f(a, b)=min(a, b) (decreasing type
B into the interval Z as follows:
environmental component) we obtain, employing
Z = {z z=f(x, y), x A, y B)} (7) the extension principle or the h-cut
representation:
Using the previous expression, the interval
function being the extension of Eq. (6) is denoted [ f(A, B)](z)= sup min(A(a), B(b)) (13)
Ö(a,b) z = min(a,b)
as h
( f(A, B))= {min(a, b) a, b( A× B)} h h
(14)
Z = f(A, B)
Such an operation can be indicated syntheti-
The characteristic function of the crisp set Z cally as min(A, B)(z) not to be confused with the
can be described as: standard fuzzy t-norm max(A(x), B(x)).
The f(A, B)(z) is a continuous fuzzy number
Z(z) = sup min[A(x), B(y)] (8)
{Öx,y z = f(x,y)} because the f(a, b) is continuous and
Eq. (8) is equivalent to Eq. (7), it represents the monotonous. For the class of Eq. (1) and Eq. (2)
basic concept of the extension principle, so when through the repeated application of the function
A(x) and B(y) are the membership of the fuzzy [ f(A, B)](z) we can calculate
j j j j
sets A and B (considered here as special cases of f(q 1p , q 2p , q 3p , . . . , q kp ) which can be indicated
characteristic function) Eq. (8) expresses a form of with q j:q [0, 1].
the extension principle. The extension principle is As stated before, the quality that is a character-
a principle for fuzzifying crisp functions: employ- istic of the environmental component is some-
ing this principle, standard arithmetic operations times a function of physical or chemical
on real numbers are extended to fuzzy numbers. parameters.
Since the values of the qualities of the primitive In order to define the quality of environmental
components are fuzzy numbers, it is necessary to components, it is worth considering the case in
apply the operators above defined Eqs. (1) – (4) which the quality is expressed as function of a
through the extension principle. For the increas- certain physical parameter f (Newman, 1989 Vis-
ing or decreasing types if A:a [0, 1] and B:b mara, 1995), that is a correspondence existing
[0, 1] are two fuzzy numbers, applying the between the value, which some physical quantity
operator f(a, b) we get: assumes in the site and the quality of the consid-
ered environmental component. Thus, we can
[ f(A, B)](z)= sup min(A(a), B(b)) (9)
Ö(a,b) z = f(a,b) define a function q( f ) that maps the values of the
This expression can also be written employing physical quantity to the quality of the component.
the h-cut representation, in terms of arithmetic For some cases, it is opportune to pass to a
operations on closed intervals of real numbers normalized scale by applying the following
(Klir, 1992): formula:
f(A, B)= { f(a, b) a, b hA × hB}
h
(10) fnorm = ( f− fmin)/( fmax − fmin)
If we select f(a, b) = max(a, b) (increasing type where f is the measure of initial conditions ac-
environmental component) we obtain, employing cording to the original scale; fnorm, the measure of
138 M. Enea, G. Salemi / Ecological Modelling 136 (2001) 131–147
the initial conditions according to the normal- pact magnitude with the physical parameter, or
ized scale; and fmin and fmax are the lower and the quantification of the surface occupied by the
upper bound of the interval in which the quality plant, since this quantification is often uncertain
component is different from zero. during the feasibility study of the project. Thus,
It is necessary then to choose and this is the besides the situation in which all the variables
most delicate phase of the crisp model, the func- and functions are crisp, the following cases can
tion q=f( f ) that best describes the environ- be found:
mental quality. a. fuzzy diagram, crisp occupied surface;
An example could be the case when, for the b. crisp diagram, fuzzy occupied surface;
water quality, the pH is the measure of a primi- c. diagram and occupied surface, both fuzzy.
tive environmental component quality. In this In the crisp model, if s indicates the surface
case if to a pH value between fmin and fmax is occupied by the plant, there is a function m=
associated a quality higher than zero and zero g(s) that returns the ‘m’ magnitude of the ac-
outside (obviously in absolutely pH is between 0 tion for a value of the occupied surface s.
an 14), a function that can represent the envi- Referring to case (a), this function m=g(s) can
ronmental quality is: depend upon a fuzzy parameter B:r [0, 1], as
in the following equation:
q= 1 −4( fnorm − 0, 5)2
q=0
if 0 5fnorm 51
otherwise m=10
s B
if s5 Sm
Sm (16)
If the value to be assigned to the physical m= 10 if s\ Sm
quantity is not of the crisp type, but is instead
For each s, m is a fuzzy number, that apply-
fuzzy, it is necessary to replace f in q( f ) with
ing the extension principle, is defined by:
the fuzzy value F:f [0, 1] of the physical quan-
tity and to use the operator: [g(s, B)](m)= sup B(r) (16%)
Ör m = g(s,r)
Q(q) = sup F( f ) (15) In the case (b) S:s [0, 1] indicates a fuzzy
Öf q = q( f )
number that expresses the surface occupied by
Of course the function q( f ) can also be a
the plant, if m=g(s) is the function relating the
fuzzy function; this case is similar to the case
impact magnitude with the physical parameter,
for the calculation of impact factor magnitude
also in this case, the magnitude will be a fuzzy
expounded in the next section.
number that when applying the extension princi-
ple, is expressed by:
In the crisp model, a certain value of magni- Finally in the case (c) the most complex, the
tude is assigned to every impact factor, generally magnitude will be a fuzzy number function of
through the individualization of the physical two fuzzy numbers S and B. Applying the ex-
measure of the impact factor on a scale of im- tension principle we have:
pact (Enea et al., 1994 Vismara, 1995). [g(S, B)](m)= sup min(B(r), S(s)) (18)
For instance, the elementary impact ‘ground Ö(r,s) m = g(r,s)
occupation’ can be represented by the diagram Sometimes the magnitude of an impact factor
as in Fig. 16; being the surface occupied by the is function, as before for environmental quality,
plant the input, it is possible to obtain the ‘m’ of several impact parameters, therefore also
magnitude of the action. in this case the composition rules (1–4) are
In such a case the uncertain elements can be applicable to determine the impact factor magni-
either the form of the diagram relating the im- tude.
M. Enea, G. Salemi / Ecological Modelling 136 (2001) 131–147 139
&
impact ii (yi ) on the environmental component i th 1990):
is contained in the T.E.I. this relationship is fuzzy.
By applying the extension principle we have SL(A]B)= [inf I(A, h)− inf I(B, h)] dh
U(A,B)
Ri :ri [0, 1], which represents in fuzzy form how
&
(24)
much the impact on the component i th is con-
tained in the global impact: SR(A]B)= [sup I(A, h)− sup I(B, h)] dh
S(A,B)
(25)
Ri (ri )
U(A, B)= {h 05h5 1, inf I(A, h)]inf I (B, h)}
= sup min[i1(y1), i2(y2), i3(y3)
S(A, B)={h 05h5 1, sup I(A, h)] sup I (B, h)}
N
Ö(y 1,y 2,y 3,...,yN) ri = yi/ % yk
k=1
Those areas are given in Fig. 1. U(A,B)= {h h 5 h5 1, inf I(A, h)] inf I(B, h)}
The degree to which A ] B is defined by the S(A, B)={h h5 h5 1, supI(A, h)] sup I(B, h)}
following relation:
(28)
C(A ]B)= max{[SL(A ] B) +SR(A ] B)]
With reference to fuzzy-set interpretation of
−[(SL(B] A)+SR(B ]A)],0} (27) possibility theory (Zadeh, 1978 Klir, 1992), h can
&
Fortemps (1996) defines the crisp relations be- be seen as an accepted grade of risk, because if dh
tween every pair (A,B): is the hbasic probability of nesting set hA then:
1. A ]B iff C(B] A) =0 h = dh
0
2. A\ B iff C(A]B) \ 0 (which implies C(B ]
A)=0) is the upper bound probability of the set: SA =
3. A $B iff C(A] B) = C(B ]A) =0 {x 0B A(x)5 h}.
The crisp preference ] structure is a complete, The Eq. (27%) with Eq. (28) definition for
reflexive and transitive relation, which implies U(A, B) and S(A, B), can be used to compare the
that \ and $ are transitive. The transitivity of T.E.I. for different plant typologies and different
this relation makes this ranking criterion suitable locations, taking into account the decision mak-
in order to compare the T.E.I. for many types of er’s grade of pessimism and of accepted risk.
scenery.
In the third case, if A and B are fuzzy T.E.I.
corresponding to two different sceneries the solu- 6. Example
tions are considered equivalent, that means they
belong to the same grade of preference. In the following an application of the proposed
The relation can also be modified to take into fuzzy approach is presented; such an example is
account the grade of prudence of the decision developed first in crisp with the purpose of under-
maker, by introducing a grade of prudence lining the differences between the two models.
i[0, 1], as in the crisp Hurwitz method used to The example concerns the T.E.I. for the work-
compare economic alternatives. The condition ing phase of an incinerator, in order not to weight
i= 0.5 will obviously identify the neutral deci- down the numerical part, we refer to five environ-
sion-maker. Considering this prudence coefficient, mental components and four factors of impact
the relation (27) can be modified as follows: only, this is because the example is finalized to
better clarify the methodology and not to ap-
C(A ] B)=
praise the case in matter. The values assigned to
max{[(1−i)SL(A]B) +iSR(A ] B)] the environmental component qualities and to the
impact factor magnitudes are merely indicative.
− [(1 −i)SL(B]A)+ iSR(B ] A)],0} (27%)
Using the classical crisp method (Fig. 2) the
In order to consider the acceptable risk level, matrix of assignment Q (5× 4) has five environ-
sets Eq. (26) can be calculated as: mental components and four impact factors.
The five environmental components are: through knowledge engineering techniques that
site noise; supply the estimates in order to value the T.E.I. in
road network; a more objective way. In this paper, such tech-
site aesthetics; niques are not studied, therefore the formalities to
site wholesomeness; quantify environmental parameters are put aside.
air quality. In the following figures, the membership of the
The four impact factors are: fuzzy number of the generic variable x is indicated
ground occupation; with v, for ease of representation; qi is the ith
noise emission; environmental component quality; mi is the ith
transport of solid waste and material; magnitude impact factor.
gaseous discharge.
Furthermore a vector G (4 × 1) is the vector of
the impacts magnitudes. As mentioned before the
values assigned to the elements of the matrix Q
and to the vector G, are merely indicative. From
the matrix product Q·G, we obtain the vector I
(5 × 1) of the partial impacts on the single envi-
ronmental components. The sum of the elements
of the vector I will supply the T.E.I.. The single
elements of the vector I divided by T.E.I. are the
Fig. 3. Site quality versus acoustic level (dB).
percentage incidence of the impact on the single
environmental components; these percentages are
indicated in Fig. 2.
In fuzzy procedure the elements of the matrix Q
are the global components quality in fuzzy form,
whose primitive are calculable through Eq. (15) if
a quality physical measure exists or through an
expert direct evaluation when there is not a phys-
ical parameter, associated to the considered envi-
ronmental component quality. In the second case,
a good evaluation of such a fuzzy set is possible Fig. 4. Fuzzy acoustic level of the site (dB).
142 M. Enea, G. Salemi / Ecological Modelling 136 (2001) 131–147
Fig. 8. Fuzzy wholesomeness quality of the site. Fig. 9. Air quality versus NO2 concentration (mg/m3).
M. Enea, G. Salemi / Ecological Modelling 136 (2001) 131–147 143
Fig. 13. Fuzzy air quality for NO2. Fig. 14. Fuzzy air quality for SO2.
144 M. Enea, G. Salemi / Ecological Modelling 136 (2001) 131–147
7. Conclusions
proach is pointed out. Fuzzy approach in fact alla valutazione del loro impatto sull’ambiente. Ing. Sani-
allows the decision maker to have a broader and tar.-Amb. 4, 35 – 45.
Foody, G.M., 1996. Fuzzy modelling of vegetation from re-
more realistic perception of the possible plant motely sensed imagery. Ecol. Model. 85, 3 – 12.
environmental impact; simultaneously although Friederichs, M., Fränzle, O., Salki, 1996. Fuzzy clustering of
allowing a better discretion in the choice among existing chemical according to their ecotossicological prop-
several alternatives, this approach highlights the erties. Ecol. Model. 85, 27 – 40.
accepted risk levels and the caution towards the Hobbs, B.F., StaKhiv, E.Z., Grayman, W.M., 1989. Impact
evaluation procedures: theory, practice and needs. ASCE
worst possible impacts. WR 115, 2 – 21.
The proposed approach, taking into account Kaufmann, A., Gupta, M.M., 1988. Fuzzy Mathematical
that both the evaluation of the environmental Models in Engineering and Management Science. North-
component quality and the impact factor magni- Holland, Amsterdam.
tude are imprecise, gives through suitable opera- Klir, G.J., 1992. Probabilistic versus possibilistic conceptual-
ization of uncertainty, in: AYY BM.
tors the real variability of the total plant impact Klir, G.J., Yuan, B., 1995. Fuzzy Sets and Fuzzy Logic —
on the environment assigning a membership value Theory and Application. Prentice Hall PTR, Upper Saddle
to every possible impact. This is more coherent River.
with the real world than the crisp method because Lhen, K., Temme, K.H., 1996. Fuzzy classification of sites
the plant impact on the environment is subjective suspected of being contaminated. Ecol. Model. 85, 51 – 58.
Melcher, D., Matthies, M., 1996. Application of fuzzy cluster-
and imprecise as subjective and imprecise is a ing to data dealing with phytotoxicity. Ecol. Model. 85,
fuzzy set. Moreover by means of this approach, it 41 – 50.
is possible to express either qualitative or quanti- Metternicht, G.I., 1998. Fuzzy classification of JERS-1 SAR
tative environmental factors. data: an evaluation of its performance for soil salinity
In this paper we have not proposed defuzzifica- mapping. Ecol. Model. 111, 61 – 74.
Nakamura, K., 1986. Preference relation on a set of fuzzy
tion rules because, in this case, it brings again to utilities. Fuzzy Sets Syst. 20, 147– 162.
crisp logic estimate, losing precious information Newman, P.J., 1989. Classification of Surface Water Quality.
about possible impacts. On the other hand, we Heineman Professional, New York.
have compared the T.E.I. for different sceneries, Roubens, M., 1990. Inequality constraints between fuzzy num-
by suggesting an approach that takes into account bers and their use in mathematical programming. In:
Slowinsky, R., Teghem, J. (Eds.), Stochastic Versus Fuzzy
the grade of prudence of the decision-maker and a Approaches to Multiobjective Mathematical Programming
level of acceptable risk. under Uncertainty. Kluwer, Dordrecht, pp. 321– 330.
Other researches could be made in order to Salski, A., 1992. Fuzzy knowledge-based models in ecological
estimate either the sort of function Eqs. (1) – (4), research. Ecol. Model. 63, 103– 112.
as well as w, h, p parameters; this estimate can be Silvert, W., 1997. Ecological impact classification with fuzzy
sets. Ecol. Model. 96, 1 – 10.
made by the expert’s opinions and through Vismara, R., 1995. Ecologia Applicata. Hoepli, Milan.
knowledge engineering techniques that are an ob- Yager, R.R., 1980. On a general class of fuzzy connectives.
ject of fuzzy theory research. Fuzzy Sets Sys. 4 (3), 235– 242.
Yager, R.R., 1984. Probabilities from fuzzy observation. In-
form. Sci. 32 (1), 103– 110.
Zadeh, L.A., 1965. Fuzzy sets. Inform. Cont. 8, 338– 353.
References Zadeh, L.A., 1978. Fuzzy set as a basis for a theory of
possibility. Fuzzy Sets Sys. 1 (1), 3 – 28.
Caffrey, P., David, M., 1975. Evaluation of environmental Zadeh, L.A., Kacprzyk, J., 1992. Fuzzy Logic for the Manage-
impact of landfills. J. Eng. Div. 2, 55–69. ment of Uncertainty. Wiley, New York.
Enea, M., Nicosia, S., Salemi, G., 1994. Rifiuti solidi urbani: Zimmermann, H.J., Zysno, P., 1991. Fuzzy Set Theory and Its
un confronto fra diversi sistemi di smaltimento preliminare Application, second ed. Kluver Academic, Boston.