You are on page 1of 9

A New Method for Site Suitability Analysis:

The Analytic Hierarchy Process


REZA BANAI-KASHANI erarchy process (AHP) with which error in judging the relative
Graduate Program in City and Regional Planning importance of factors in site suitability analysis can be both
Department of Geography and Ranning detected and corrected. The proposed approach is illus-
Memphis State University trated with an example to show how the AHP frames the site
Memphis, Tennessee 38152, USA evaluation problem and can aid in decision making involving
multiple criteria, factor diversity, and conditions of uncertainty.
ABSTRACT / A critical shortcoming of methods that are reliant The article concludes by suggestingthe potential application
upon the judgment of experts to determine site suitability is of the AHP in public choice decisions involving complex,
noted. The article introduces a new method, the analytic hi- controversial, and conflictual site selection processes.

Recendy, Anderson (1987) surveyed a class of example, LESA uses a USDA-recommended stan-
methods for land capability/suitability analysis. The dard of 100 points for land evaluation (LE) and a
seven methods ranged in degree of computational or maximum of 200 points for site assessment (SA)
analytical sophistication from the simple to the more (Steiner 1983).
elaborate. The methods include: pass/fall screening, This approach to site assessment is operational
graduated screening, weighted factors, penalty point when standards are known. However, in new
assignment, composite rating, weighted composite problems, for which no standards have been estab-
rating, and direct assignment. For detailed compar- lished, or when old standards are discredited (Lynch
. . . . . . . .~:c Anderson (1987), Chapin and Kaiser (1979,
,sum, 1984) and intangible criteria are used to assess alterna-
Chapter 9), and Steiner (1983, 1987) for a review of tives (Saaty 1987), the weighted-factors method is of
land evaluation and site assessment (LESA) as a suit- fimited use.
ability method used by US government agendes, as Lynch (1984) has warned against the possible
well as examples of applications of suitability proce- danger of generalization with the use of "performance
dures in the United States and in other countries. standards" and has stressed the use of "performance
A complex site selection process involves a measure dimensions." Lynch (1984, pp. 111-113) notes:
of trade-offs among the criterion factors. For example,
the pass/fail method of screening alternatives may not to assert that ideal densityis twelvefamilies to the acre, or that good
result in the selection of the best alternative where the cities are organized into residential neighborhoodsof 3000 persons
each, are statements too easily discredited. Situations and values
relationship among the criteria is not considered. differ. What we might hope to generalize about are performanced./-
Methods that account for the relationship among the mens/ons,that is, certainidentifmblecharacteristicsof the performance
suitability factors, such as nonlinear combination, [ofd~s] whichare due primarily to their spatial qualifiesand which
a r e measurable scales, along which different groups will prefer to
factor combination, and weighted composite rating
(Chapin and Kaiser 1979, Chapter 9, Anderson 1987), achievedifferent positions"[emphasisadded].
provide greater flexibility in a suitability analysis.
However, there is no formal process inherent in such The performance dimensions suggested by Lynch are:
methods that can aid the decision maker in deriving vitality, sense, fit, access, control, efficiency, and jus-
the relative importance of the criteria. tice. "Groups and persons will value different aspects
The weighted-factors method provides a procedure of them and assign different priorities to them. But,
where each suitability factor is assigned a score, which having measured them, a particular group in a real
is multiplied by the weight of that factor. The results situation would be able to judge the relative goodness
of the multplicafions are added, and thus a site com- of their place and would have the dues necessary to
posite score is determined. The composite score is improve or maintain that goodness" (Lynch 1984, p.
compared with a predetermined standard, which is 119). The performance characteristics are thus evalu-
used to select or reject a site. Land evaluation and site ated in relation to the goals and values and are
assessment (LESA) provides an application of this brought to bear upon the relevant features of settle-
method (USDA 1983, Wright and others 1983). For ment form.
Concomitantly, the dependence of the criteria on
KEY WORDS: Analytichierarchyprocess (NIP); Site suitability;Multi- (the characteristics of) the alternatives is called "struc-
criteriaevaluation;Judgmentconsistency tural dependence" by Saaty (1987), which is revealed

EnvironmentalManagementVol 13, No 6, pp 685-693 9 1969Spdnger-VerlagNewYorkInc


686 R. Banai-Kashani

in "relative" measurement. The approach allows for analysis can be both detected and corrected. A brief
the priorities of the criteria to change according to the description of the AHP method is given by developing
importance obtained from the ranking of the alterna- an example for site suitability analysis.
fives, i.e., to discover the relative weight of alternatives
through paired comparisons, rather than through a
Site Suitability Analysis by the AHP Method
standard (Saaty 1987, p. 165). Relative measurement is
thus contrasted with absolute measurement, which in- Suppose a decision maker is faced with a problem
volves known standards (Saaty 1987). of deciding among three alternative sites and has
A common feature of the suitability methods noted identified the criteria that guide the site selection pro-
above, as well as the checldist methods used in envi- cess. The decision maker may perceive the criterion
ronmental impact assessment (e.g., Leopold and factors as being varied in importance. Furthermore,
others 1971, Dee and others 1973), is their reliance cognizance is taken of the competing attributes of the
upon expert judgment. However, none incorporate a (site) alternatives, while deciding which alternative best
check on the consistency of value judgments in the meets the criteria. This problem involves a two-stage
process of deciding the relative importance of factors evaluation process: (1) determining a measure of the
or in evaluating preferences for alternatives in a site relative importance, or priority, of the criterion
suitability analysis. Yet, the problem of inconsistency factors, and (2) determining a measure of the relative
arises in decision making under conditions of limited weight of the alternatives. The structure of a hierarchy
information and uncertainty. Various sources of un- can be used to frame this site evaluation problem
certainty, such as about the planning environment, (Figure 1).
about value judgments, and about the decisions of This hierarchy is comprised of four levels. The first
other participants, contribute to errors in decision level simply states the problem objective: choosing an
making and forecasting by experts (Hall 1980, p. 250). optimal, or most suitable, site, subject to a set of cri-
[See Hall (1980) for a detailed account of the different teria. Three criteria are identified (as an example) at
sources of uncertainty.] the second level of the hierarchy: (site) slope, price,
Factors commonly considered in a suitability anal- and views. Three alternative sites, a, b, and c, are sped-
ysis, such as zoning, land value, and existing and fled at the third level of this hierarchy. At the fourth
planned uses (Anderson 1987), are evaluated in the level, the attributes (data) of the alternative sites are
face of the uncertainties of the economic, demo- specified. The objective is to determine which of the
graphic, and political environment. The expert plan- three sites should be selected vis-~t-vis t[ae relative im-
ning team encounters the uncertainties of the deci- portance given to the suitability criteria.
sions of other participants--the present and future The criteria include both quantitative (slope and
residents of a community, the developer, the land use price) and qualitative (views) factors. The first two
regulator/controller, and the polifician. Furthermore, factors lend themselves easily to quandfication,
variability is exhibited in the value judgments of the whereas the third factor (views) is restricted by the
decision makers, and decisions may be made in the type of measurement scale used. For example, a mea-
face of limited information. An alternative to the sure of landscape views can be given by the use of an
methods commonly used in suitability studies is called ordinal scale, such as good, fair, poor. The AHP offers
for which can aid the expert faced with decision an advantage of a ratio scale, which can be effectively
making under uncertainty. applied to the measurement of both qualitative and
In this article Saaty's (1980, 1982) analytic hierarchy quantitative factors in site suitability evaluation. [For a
process (AHP) is presented as a new method for site description of the different types of scales in the con-
suitability evaluation. The AHP has been shown to be text of site suitability studies, see Hopkins (1077).]
effective in evaluation problems involving multiple The AHP procedure involves performing compar-
and diverse criteria, measurement of trade-offs, and ison of pairs of factors within a set of reciprocal ma-
with limited data [see Saaty and Vargas (1987), and trices. The use of a matrix in site suitability analysis is
Zahedi (1986) for a survey of the method and its ap- not new. For example, in his important work Design
plications]. The AHP exhibits flexibility in dealing with Nature, McHarg (1969) used a matrix that formed
with both the qualitative (intangible) and quantitative the relationship between a set of factors to evaluate
(tangible) factors in a multicriteria evaluation problem. land use compatibility and/or suitaiaility. A combina-
Furthermore, the AHP provides a methodological tion of qualitative and quantitative factors was used--
framework within which the inconsistency in judging although the ordinal scale posed a limitation in arith-
the relative importance of factors in a site suitability metic operation. This limitation is surmounted if the
The Analytic Hierarchy Process 687

L1 : Objective Choosing Optimal Site

L2: Criteria slope price views


(s) (p) (v)

L3: A l t e r n a t i v e
Sites a b c
/J',, A',, a
a b c b c

L4: S i t e t [ I i I I I '{ "1


Attributes s s p v s p v
30% $p v
2 0 / excellent lo% $40/ good 5% $801 fair
8cre acre acre

F~um 1. A simple hierarchy for


Configuration of Sites site suitability analysis.

more versatile ratio scale is used. However, the recip- 1/3 and 1/6 in the second and third rows, first column
rocal matrix of pairwise comparisons can be solved to entries, respectively). For this matrix, all that remains
determine the priority of factors, which is expressed in is a comparison of the criterion of price with that of
relative (or percentile) weights. [It turns out that the views.
solution of a reciprocal matrix that has unit rank and Note that from the initial set of comparisons, which
in which all its elements are positive and its diagonal involved the factors slope, price, and views, the fol-
elements are one is a unique vector (upon normaliza- lowing scale relations were determined:
tion of its columns), which gives the relative weights of
s --- (3)P; s = (6)v
the elements along a ratio scale (Saaty 1980).] In com-
paring pairs of factors, the AHP scale of relative im- which implies
portance is used (see Table 1 for explanation).
P = (2)v
A matrix of three factors, slope, price, and views, is
formed (Table 2). The diagonal elements are assigned Thus, this value (2) is entered for the second row,
the value of unity (i.e., when a factor is compared with third column, of the criterion matrix. Now, to deter-
itself). mine the relative weight of the three criteria, the judg-
Anderson (1987) stressed that an essential property ments are synthesized. For this purpose: (1) add all the
of a site suitability method is to allow measurement of values in each column of the matrix; (2) divide each
the relative weight of the criterion, rather than to as- column entry by the column total; and (3) take an
sume the same weight for each criterion. The AHP average over the rows (or take geometric mean of each
can provide this flexibility. Suppose, in particular, in row). [The geometric mean (GM) of three factors, S, P,
comparing the criterion factor, slope with price, the and V, is determined by multiplying the factors and
former is given moderate importance, i.e., the value 3 then taking their cubic root: GM = (SPV)m.]
over the latter. (See the AHP scale description in Table For the three criteria, the relative importance ex-
1.) Further, suppose slope is given relatively strong pressed in percentages is derived (Table 3). Thus a
importance (value of 6) when compared to views. common scale 0 % - 1 0 0 % ) is obtained from the AHP
These values are placed in the first row of the matrix. procedure of assigning quantified judgments. The cri-
Since, by definition, the AHP comparison matrices are terion of slope emerges as the most important (with a
reciprocal, the remaining entries in the first column of relative weight of 67%), followed by the importance of
this matrix are automatically determined (shown by price (23%)and views (10%).
688 R. Banai-Kashani

Table 1. The AHP scale for paired comparisons." An essential contribution of the AHP as an alterna-
Intensity tive to the existing methods for site suitability studies
of importance Defmition and explanation lies in its internal procedure for detecting the inconsis-
tency of judgment, in comparing one factor, criterion,
Ib Equal importance--two activities
contn~tmte equa/ly to the objective. or activity with another. The A H P provides a proce-
dure for checking the consistency of expert judgment
3 Moderate importance--experience in the process o f pairwise comparison of factors. A
and judgment slighdy favor one
activity over another. problem o f inconsistency in assigning the relative
weights for the same criteria as specified before can be
Essential or strong importance-- demonstrated.
experience and judgment strongly
favor one activity over another. Suppose in comparing the criterion of price with
views the value 4 is assigned, rather than the value 2,
Demonstrated irnportance--an which preserves consistency in the matrix. (Recall that
activity is strongly favored and its
dominance is demonstrated in the comparison of price with views was earlier per-
practice. formed through the comparison of slope with price
and slope with views.) This comparison is shown in
Extreme importance--the evidence
favoring one activity over another Table 4 in brackets, resulting in a new, "inconsistent"
is of the highest possible order of matrix. After the same type o f calculation as shown
affLrmation. earlier, the result (the fourth column of the matrix)
2,4,6,8 Intermediate values between the two shows the relative importance o f the criteria. The rela-
adjacent judgements--when tive weight o f slope is decreased (from 67% to 63%),
compromise is needed. whereas the weight of price is increased (from 23% to
Redprocal of If an activity has one of the above 27%). T h e rankings of the criteria are not affected
above numbers numbers assigned to it when (despite some inconsistency) and still remain: slope,
compared with a second activity, followed by price, followed by views. Nevertheless, it is
then the second activity has the conceivable that in the general case the preference
reciprocal value when compared to order arrived at in an inconsistent matrix may turn out
the first.
to be different than in the case where consistency is
"Adaptedfrom Saaty0987). maintained. Thus, it is essential to determine a mea-
~'he scale l.l, 1.9..... 1.9,or even a freer one, can be used to com- sure of departure from consistency.
pare dements that are dose together, or are near equal in impor- To determine such a measure, the first column of
tance.
the inconsistent matrix (Table 4) is multiplied by 0.63,
the second column by 0.27, and the third by 0.10. The
result of this multiplication is shown in the first three
Table 2. Pairwise comparisons of the criteria. columns o f Table 5 in decimal form. Thereby, we seek
to contrast the inconsistency with the solution we
Criterion S P V
would have obtained if the judgments were random
Slope (S) 1 3 6 (Saaty 1982).
Price (P) gs 1 2 The row totals are given in column 4 and the crite-
Views (V) Vo 1/~ 1
rion weights are shown again in column 5. T h e row

Table 3. Determining the relative weight of criteria.


Step IIl
Step I Step II
Row Geometric
Criterion S P V S P V average % mean
Slope (S) 1 3 6 6/9 0/9 % 6/9 67 .67
Price (P) gs 1 2 2/9 V9 u % 23 .23
Views (V) V6 Y2 1 V9 V9 V9 V9 10 .10
Totals % % 9 1 1 1 1 100% 100%
The Analytic Hierarchy Process 689

Table 4. Inconsistent matrix. strates good consistency in judgment. If the value had
Criterion S P V Row avg exceeded 10%, the new value assigned in comparing
price with views (4) would need to be rethought. Why
Slope (S) 1 3 6 0.63 is 10% inconsistency tolerated? Saaty (1987, p. 172)
Price (P) ~ 1 [4] 0.27
Views (V) Ve [V,] 1 0.10 explains:
The reason is that inconsistencyitselfis important, for withoutit
totals are divided by the criterion weights, and their new knowledgewhichchangespreferenceorder cannot be admitted.
Assumingall knowledgeto be comi.qentcontradictsexperiencewhich
average value is taken, which is called'X,=,. requires continued adjustment in understanding. Thus the objective
Now, an index (C0 that provides a measure of de- of developinga wide-rangingconsistent frameworkdepends on ad-
parture from consistency can be computed: mitting someinconsistency.
Cl = ( k , ~ - n)/(n - 1) Thus, the AHP is conceptually predicated on the
where n denotes the size of the comparison .matrix behaviorally plausible paradigm of bounded ratio-
(Saaty 1980, 1982). For this example the value o f C l = nality and sets out to model decision making within the
(3.01 - 3)/2 = 0.005. The closer k,,~ is to n (or CI is framework of "satisficing e' [satisf=cing expresses the no-
to zero), the better the consistency of judgment; other- tion of being satisfied with "good enough" solutions,
wise, a problem of inconsistency is implied, explained rather than searching for optimal or the "best" solu-
further as follows. tions (Simon 1983, p. 85)].
Through this process we have, in effect, sought to Thus far, we have arrived at the relative impor-
assess whether the AHP ratio estimates (in paired com- tance of the three criteria, and we are assured of judg-
parisons of factors) are closer to being logically or con- ment consistency in the process. Now, however, we
sistently adjudicated or are closer to being randomly should take a further advantage of the AHP proce-
chosen (Saaty 1980). The random value o f Cl (denoted dure, which guides us through the remaining task of
by R/) for a matrix of size n = 3 is R / = 0.58. Random evaluating the three sites (a, b, c, in our example). We
consistency index (R/), which is derived from a sample address explicitly how each site fares when assessed in
of size 500 of a randomly generated reciprocal matrix the face of the three criteria. To do this, assume the
using the scale 1/9, 1/8. . . . . 1. . . . . 8, 9, is given by the following characteristics (data) for the three sites, sum-
size of the matrix (or the number of factors, n, in the marized in Table 6 (repeated from level 4 of the hier-
comparison matrix) (Saaty 1987, p. 171): archy in Figure l).
Site a has a steep slope, excellent views, and the
Size of matrix (n) lowest price (compared to b and c), whereas site c has
the lowest slope, high price, and fair views. Now, the
1 2 3 4 5 sites can be compared to assess how they meet the
three suitability criteria. The information of site char-
consistency acteristics is used in the paired comparisons. For the
index (R/) 0 0 0.58 0.90 1.12 criterion of slope, a matrix is formed (Table 7). In this
matrix, a preference is revealed for a site that has a
Size of matrix (n) less steep slope, since it is more economical for devel-
6 7 8 9 10 opment. The solution of this matrix shows in relative
weights (%) this revealed preference: site c (65%), site
Random b (25%), and site a (10%). Similarly, matrices for the
consistency criteria of price and views are formed (Table 7). [A
index (R/) 1.24 1.32 1.41 1.45 1.49 software package for the IBM PC called Expert Choice
(1988) is available to perform the AHP computations
The consistency ratio (CR) is found: of the sort presented here. In addition, a variety of
computer programs in Basic, Fortran, and APL is pro-
C R = CI/RI
vided in Saaty (1982) that facilitate the use of the AHP
= 0.005/0.58
interactively with either personal and/or mainframe
= 0.008, or 0.8%
computers.]
Thus, the judgment formed in the criteria matrix The values of consistency index (C0 and consis-
(Table 4) is acceptable if we use a limit suggested by tency ratio (CR), stated in the bottom of Table 7 indi-
Saaty (1982) that a C R value of less than 10% demon- cate that the comparisons were performed with good
690 R. Banai-Kashani

Table 5. Determining a measure of inconsistency in judgment with X~w,.


(4) (5) (6) (7)
Row Criterion
S P V total weight (4):(5) k~
Slope (S) 0.63 0.81 0.60 2.04 0.63 3.23 3.02
Price (P) 0.20 0.27 0.40 0.87 0.27 3.22
Views (V) 0.10 0.06 0.10 0.26 0.10 2.60

Table 6. Characteristics of alternative sites: data. with the rational planning paradigm, which specifies
Price (1) problem identification/goal formulation, (2) evalua-
Site Slope (%) (S/acre) Views tion of alternatives, (3) determination of choice, and
(4) implementation.] The interrelationship of factors
a 30 20 excellent
b io 40 good at each level of the hierarchy was analyzed with respect
c 5 80 fair to each factor at its preceding level. Thereby, the rela-
tionship among the objective, criteria, and alternatives
was accounted for. Multiplying the weights of factors
judgment consistency. However, a clear advantage of in the preceding level by the weight of factors in the
the AHP approach is that the three sites are now com- succeeding level resulted in the overall weights of
parable through a common relative scale (0%-100%), factors in the lowest level of the hierarchy, i.e., the
whether a criterion factor is qualitative or quantitative priorities of the alternative sites. This operation is
in nature. guided by a mathematical principle of aggregation in a
Finally, which site is the most suitable, given th e rel- hierarchy. Stated differently, the organizational ele-
ative importance arrived at earlier for the three suit- ments of the hierarchy (criteria) were brought to bear
ability criteria? In the criteria matrix (Table 3), the rel- upon the instrumental elements of the hierarchy (al-
ative importance of slope (67%), price (23%), and ternatives), and the composite weights of the elements
views (10%) was determined. In the Fmal step (Table were arrived at. Thus, the AHP provides an analytic
8), these values are multiplied by the relative weight of procedure (comparing elements) and a synthetic pro-
the alternative sites for each criterion (the last column cedure (compositing elements) in a unified evaluation
in the matrices given in Table 7, repeated in Table 8). framework.
The last column in Table 8, site priority, states the One can allow for greater complexity by increasing
composite scores to show how each site (a, b, c) meets the number of levels or factors in a hierarchy. For ex-
the criteria of slope, price, and views simultaneously. ample, the participants (e.g., developers, users, etc.)
Site c, which has a slope of 5%, and $80/acre price, and could be identified explicitly, thereby accounting for
fair views, emerges as the most suitable (with a relative the different priorities they attach to the suitability cri-
weight, or priority, of 47%). The AHP computations teria. [It is interesting to examine the scenario in which
carried out in Table 8 can also be illustrated graphi- the weightings of the criteria result in an outcome that
cally as shown in Figure 2. favors one party more than another. The AHP pre-
diction of such an outcome can inform the parties in-
volved and can provide a basis for negotiation so as to
Conclusion
arrive at collectively desirable decision outcomes.] Al-
A summary of the AHP as a method for site suit- ternatively, the land-use activities, such as residential,
ability evaluation is given here. First, the AHP pro- commercial, and industrial may be discerned as yet
vides a hierarchical structure that allows one to or- another set of factors in a level of the hierarchy (e.g.,
ganize the relationship of factors in a suitability anal- see Banai-Kashani 1987, 1988). This level by level das-
ysis. In the illustrative example presented, a simple sification and representation helps to capture the con-
hierarchy was used that had four levels (Figure 1). tribution of different types of factors in a complex site
Level 1 stated the objective of a site evaluation evaluation analysis. It also helps to avoid a potential
problem, subject to a set of suitability criteria. The cri- confusion that arises with an indiscriminant inclusion
teria were specified at the second level, and the alter- of very different types of factors in the analysis of their
native sites and their attributes were identified at the interrelationships, for example, the distinction among
third and fourth levels of this hierarchy, respectively. the criterion factors, activities; attributes of alterna-
[There is a certain conceptual resemblance of the AHP fives, and choices.
The Analytic Hierarchy Process 691

Table 7. Comparing sites for each criterion.


Relative Relative Relative
Slope a b c weight Price a b c weight Views a b c weight
a 1 Vs ]/6 0.10 a 1 2 4 0.56 a 1 3 7 0.67
b 3 1 Vs 0.25 b V= 1 3 0.32 b Vs 1 3 0.24
c 6 3 1 0.65 c V4 Vs 1 0.12 c V~ Vs ) 0.09
CI = 0.009 CR = 0.015 CI = 0.009 CR = 0.015 CI = 0.003 CR = 0.005

Table 8. Determining overall priorities,


Criterion Site
Sites Slope Price Views weight priority*

b 0.25 0.32 0.24 x P 0.23 = 0.264


c 0.65 0.12 0.09 V 0.10 0.474
9In detail: (0.67 x 0.10) + (0.23 x 0.56) + (0.10 x 0.67) = 0.262
(0.67 x 0.25) + (0.23 x 0.32) + (0.10 x 0.24) = 0.264
(0.67 x 0.65) + (0.23 x 0.12) + (0.10 x 0.09) = 0.474

/L
Slope Price Views

,1ILL
Composite Priorities

~r~-~.:~:$~~. Key ~ < 20%

20 - 40~

[ ~ > 40% Figure 2. Criterion-weighted, com-


posite priority mapping of the sites.

Second, there should be a limit on the n u m b e r o f sistency o f j u d g m e n t in paired comparison o f factors


factors that are included in each level o f a hierarchy in (Saaty 1977, 1980).
o r d e r to maintain good consistency in judging their Third, the A H P pairwise comparison approach
relative importance. Modern problem-solving tech- offers an account of the competing properties o f the
niques (e.g., Simon 1981, 1983) have stressed a limita- alternatives. After all, the three sites in the illustrative
tion in the natural cognitive ability o f the mind to pro- example had certain c o m m o n features: slope, price,
cess information. Psychological experiments have and views, which varied with relative intensities. T h e
shown that an individual cannot simultaneously com- A H P ratio scale provided a measure o f the intensities
pare m o r e than seven objects (plus or minus two) o f preferences for these features, which were revealed
without being confused (Miller 1956 in Saaty and in the comparative evaluation o f the site alternatives in
Vargas 1982, p. 22). Miller's observation has been con- accordance with the criteria. T h e suitability criteria
firmed further with the use o f A H P in testing the con- were determined by the relative importance given to
692 R. Banai-Kashani

slope, price, and views, which were derived through the AHP method in the realm of public choice theory
the paired comparisons of their interrelationship. (Buchanan and Tullock 1965) that remain to be inves-
The AHP approach has an interface with multicri- tigated.
teria evaluation methods that have recently emerged
in the planning literature (Saaty 1980, Nijkamp and Acknowledgments
Voogd 1983, Hughes 1986, Vargas 1987). However, a
The author gratefully acknowledges the editor of
key ingredient of the AHP approach for multicriteria
this journal for the opportunity to incorporate the
evaluation and decision making should be empha-
constructive comments of the referees in this version
sized. In the matrices of Table 7, alternatives (sites)
of the paper.
were compared for each criterion, with the criterion
itself considered as a given. However, the relative im- Literature Cited
portance of the criteria was derived through the AHP
paired comparison method (in the first stage of evalu- Anderson, L T. 1987. Seven methods for calculating land
ation) and was then brought to bear upon the weights capability/suitability. Planning advisory service (PAS) re-
port No. 402. American Planning Association, Chicago: 20
of the alternatives (in the second stage of evaluation).
PP.
Thus, the AHP provides an advantage compared to Banai-Kashani, A.R. 1987. A new paradigm for planning
existing muldcriteria methods, which do not aid the simulation. Pages 131-136 In E. Alexander (ed.), Pro-
decision maker in arriving at the relative weights of ceedings of the 1987 international congress on planning
the criteria and which treat the criteria as givens or and design theory. American Society of Mechanical Engi-
(pre)determined outside of the evaluation method. neers, New York.
An important advantage of the AHP method that Banai-Kashani, A.R. 1988. Toward a synthetic measure of
allows one to gauge the consistency of judgments in good serdemem form. Environment and Planning 15B:399-
412.
the evaluation process should be stressed. For ex-
Buchanan, J. M., and G. Tullock. 1965. The calculus of con-
ample, in evaluating three alternatives Al, A2, As, the sent. University of Michigan Press, Ann Arbor. 316 pp.
preferences may be ordered by: A 1 preferred to A~,
Chapin, F.S., and E.J. Kaiser. 1979. Urban land use plan-
and A 2 preferred to As. If consistency is to be held, it ning. University of Illinois Press, Chicago.
follows that logically A~ should be preferred to As. Dee, N., J. K. Baker, N. L. Drobry, K. M. Duke, I. Whitman,
However, perfect consistency does not always hold in and D. Fabringer. 1973. An environmental evaluation
practice. Limited information, uncertainty, and behav- system for water resources planning. Water Resources Re-
ioral as well as contextual variability are among the search 9(2):523-535.
contributing factors leading to a violation of the con- Expert Choice. 1988. Software Package. Decision Support
sistency principle. However, the recognition of incon- Software, McLean, Virginia.
sistency is important in itself. It calls for improving Hall, P. 1980. Great planning disasters. University of Cali-
judgments by learning, incorporating better informa- fornia Press, Berkeley, California. 308 pp.
tion, and a greater scrutiny of the behavioral and con- Hopkins, L.D. 1977. Methods of generating land suitability
textual environment of decision making. maps: A comparative evaluation. Jonnud of the American In-
s~*ute of Planner~ 43(4):386-400.
Furthermore, controversial site selection processes
Hughes, W. 1986. Deriving utilities using the analytic hier-
- - f o r example, the siting of nuclear power plants and archy process. Socio-Eanunnic Planning Science 20(6):393-
landfills for environmentally hazardous materials-- 395.
can fuel considerable public debate surrounding the Leopold, L.B., F.E. Clarke, B.B. Hanshaw, and J.R.
interpretation of the location criteria as well as the Balsiey. 1971. A procedure for evaluating environmental
choice of the optimal site. An account of consensus in impact. Geological survey circular 645. US Geological
decision-making, however, is indispensable, not only in Survey, Washington, DC.
the expressed opinions of the nonexpert (public Lynch, K. 1984. A theory of good city form. MIT Press,
groups), but also among the panel of experts them- Cambridge, Massachusettes.
selves (the planning staff). Toward this end, the AHP McHarg, I. 1969. Design with nature. Natural History Press,
consistency analysis can be fruitfully applied to pro- New York. 197 pp.
vide an account of consent, or dissent, in public deri- Miller, G. A. 1956. The magical number seven plus or minus
two: Some limits on our capacity for processing informa-
sion making and thereby formulating strategies for tion. Psycho/og/cdRet/av 63:81-97.
conflict resolution (e.g., Saaty and Alexander 1977,
Nijkamp, P., and H. Voogd. 1983. A survey of multicriteria
Saaty and Vargas 1987). Various sources of uncer- analysis for development planning. Pages 217-229 in L.
tainty, such as those identified at the outset of this ar- Chatterjee, P. Nijkamp (eds.), Urban and regional policy
ticle, suggest fertile ground for future applications of analysis in developing countries. Gower, Boston.
The Analytic Hierarchy Process 693

Saaty, IL W. 1987. The analytic hierarchy process--what k is Simon, H. 1981. The sciences of the artiticial. MIT Press,
and how it is used. MathematicalModeUing. 9(3-5):161- Cambridge, Massachusetts. 247 pp.
176. Simon, H. 1983. Reason in human affairs. Stanford Univer-
Saaty, R. W., and L. G. Vargas (eds.). 1987. The analytic hier- sity Press, Stanford, California. 115 pp.
archy process. M ~ Modelling 9(3-5).
Steiner, F. 1983. Resource suitability: Methods for analyses.
Saaty, T. L. 1977. A scaling method for priorities in hierar- Env/ronmenta/M~. 7(5):401-420.
ch~mal structures. Jounud of MoZhemat~a/ Psycho/og~
15(3):234-281. Steiner, F. 1987. Agricultural land evaluation and site assess-
ment in the United States: An introduction. Environmental
Saaty, T. L 1980. The analytic hierarchy process. McGraw- Managwnont 11(3):375-377.
Hill, New York. 287 pp.
Saaty, T.L. 1982. Decision-making for leaders. Lifetime USDA (Department of Agriculture). 1983. National agricul-
Learning, San Francisco. 291 pp. tural land evaluation and site assessment handbook. Soil
Conservation Service, Washington, DC.
Saaty, T.L. 1987. Rank generation, preservation, and re-
versal in the analytic hierarchy decision process. Decision Vargas, L. G. 1987. Priority theory and utility theory. Mathe-
Sc/oncea 18:157-177. mat/ca/Made//ng 9(3-5):381-385.
Saaty, T. L., and J. M. Alexander. 1977. The forward and Wright, L. E., W. ZitzJnann, K. Young, and R. Googim. 1983,
backward processes of conflict analysis. Belmvioral Science LESA--agricultural land evaluation and site assessment.
22:87-98. Journal of Soil and WaterConservation $8(2):82-86.
Saaty, T.L., and L. Vat'gas. 1982. The logic of priorities. Zahedi, F. 1986. The analytic hierarchy process--a survey of
Kluwer/Nijhoff, Boston. 299 pp. the method and its applications. Interfaces 16(4):96-108.

You might also like