Вы находитесь на странице: 1из 15

Cow.

Environ. and Urban Systems, Vol. 16, pp. IOl-115,1992

Printed in the USA. All tights reserved.

OlS6-9715/92 $6.00 + .oo


Copyright 0 1992 Pergamon Press Ltd.

A FUZZY GIS MODELING APPROACH


LAND EVALUATION

FOR URBAN

Daniel Z. Sui
Department of Geography, University of Georgia

ABSTRACT. Present analytical functions and conventional cartographic modeling techniques in


Geographic Information Systems (GIS) are based on Boolean logic, which implicitly assumes that objects
in a spadal database and their attributes can be uniquely defined. The inherent constraint of the classical
set theory does not allow for partial set membership conditions and imprecise information in GIS. The
inadequacy of the Boolean logic for the representation and manipulation of spatial data is a major
obstacle toward realistic GIS modeling. This paper demonstrates the usefulness of Zadeh's fuzzy set
theory in GIS modeling for urban land evaluation. The results indicate that incorporating fussy set theory
into GIS modeling can provide more &tails about the gradual transition of urban land value than the
traditional cartographic modeling approach. Fuzzy GIS modeling can also reduce the information loss by
obtaining membership grade for each individual land parcel. The membership function allows
identification of the extent to which a particular area belongs to a valuation class based on given criteria.

INTRODUCTION
The rapid development of Geographic Information Systems (GIS) has dramatically changed
the procedure for spatial data collecting, storing, analyzing and displaying. The 1980s saw a
great proliferation and adoption of GIS technology in the urban and regional planning community (Scholten & Stillwel, 1990). GIS applications in urban and regional planning rely overwhelmingly (though not exclusively) on the routine functions within a GIS. Although much of
the built-in analytical functionality reflects the perceived needs of the current market, it is
becoming increasingly clear that the standard GIS analytical operations are insufficient for the
diversified needs of GIS users (Burrough, 1990).
All the current GIS analytical functions used for cartographic modeling are based on the traditional Boolean logic, which is crisp, deterministic and precise in nature and gives no room
for the imprecisions in geographic information and human decision making processes. Each
Reprint requests should be sent to Daniel Z. Sui, Departmentof Geography,The University of Georgia, Athens, GA
30602.
Financial supportfrom National Science Foundationof Peoples Republic of Chii and Departmentof Geography,
The University of Georgia, is gratefully acknowledged. The authorwould like to express sincere thanks to Professor
Cheng Ji-Cheng. Peking University, and Professors C.P. Lo, James 0. Wheeler, Jean-Claude Thill, and Elizabeth
Hovinen. University of Georgia, for their constructivecomments on an earlier draft of this paper. Comments from two
anonymous reviewers are also gratefullyacknowledged. The authoris responsible for any remainingerrors.
101

102

D. Z. Sui

spatial entity is associated with a single thematic attribute. Sharp boundaries are imposed
between categories. A spatial entity can either belong to or not belong to a set, even though
most geographic phenomena change gradually in space and time. The conventional cartographic modeling technique, as discussed by Tomlin (1990), is based on the two-valued Boolean
logic and has been widely employed in urban and regional planning. However, the traditional
cartographic modeling technique has proved to be quite awkward in some GIS applications
where imprecision and vagueness prevail, because not all the entities in the spatial database can
be uniquely defined, either in the set of attributes or in their spatial delineation (Leung, 1988).
The deficiencies of the traditional Boolean logic for the design of spatial databases have
been recognized in recent years (Burrough, 1986; 1989). As an alternative to Boolean logic,
Zadehs fuzzy set theory has been proposed as the new logical foundation for GIS design
(Robinson, 1988; Leung, 1989). The potential applications of fuzzy logic in spatial data collection, representation, retrieval and display have been discussed in recent literature. Robinson
and Strahler (1984) have considered some of the more direct contributions to be made by fuzzy
set theory for the storage of geographic features as linguistic data. Robinson (1990) has also
developed a new automatic method for the interactive acquisition of fuzzy spatial relationships.
Built on the INGRES relational database, Kollias and Voliotis (1991) reported a prototype soil
information system with fuzzy retrieval capabilities based on fuzzy reasoning. A fuzzy information representation scheme and its implementation in conventional GIS software was developed at the University of Guelph (Wang, Hall, & Subaryono, 1990). Rowman (1990) reported
a formal fuzzy logic-based specification framework for geographic information, in which rules
of reasoning about time, space and accuracy have been compactly stated in a subset of secondorder calculus. This formalized representation can be easily modified to the particular needs for
specific applications. Some cartographers have also discussed how fuzzy logic can be incorporated into information display (Bouille, 1982; Muller, 1987). The applications of fuzzy mathematics in data collection have been well documented in remote sensing literature (Kent &
Mardia, 1988; Blonda, Pasquariello, Losito, Mori, Posa, & Ragno, 1991).
But so far few studies have indicated explicitly how to incorporate fuzzy set theory into GIS
modeling processes to handle the imprecisions in the spatial decision process, especially in an
urban context. Therefore, the objective of this paper is to explore the application of Zadehs
fuzzy set theory in GIS modeling for urban applications. This paper aims to contribute to the
literature by extending the strict Boolean logic-based cartographic modeling to a fuzzy domain.
Specifically, this research will test whether the results derived from the fuzzy modeling
approach are significantly different from those of the conventional method for urban land evaluation. After a conceptual discussion of how to incorporate fuzzy set theory into cartographic
modeling, a case study for urban land evaluation in Jining City, P.R. China, is presented based
on the fuzzy GIS modeling approach, which is followed by the results and discussions. A
future research agenda is laid out in the concluding remarks.
FUZZY SETS AND FUZZY GIS MODELING

To discuss how to incorporate fuzzy set theory into the GIS modeling process, it is necessary
to introduce some of the basic concepts for those not familiar with fuzzy sets. According to
Zadeh (1965). a fuzzy set is mathematically defined as follows:
If X=(x} denotes a space of objects, then the fuzzy subset A in X is a set of ordered pairs

A = KM(X) l=fupA(n)/n. . . . . . . . . . . . . .

(1)

where A is a fuzzy subset in universe of discourse U and &(x) : U+M is the membership
function which takes on values in an ordered set M (membership set). The integral sign is not

A Fuzzy G/S Modeling Approach for Urban Land Evaluation

103

Riemanns integral but is used to indicate that the operations apply to all of the items after the
sign. The slash f means with respect to. The value of CIA(X)is the grade of membership of
x in A. x E X means that x is contained in X. Usually l.tA(x) is a number in the range of 0 to 1,
with 1 representing full membership of the set and 0 nonmembership. The grade of membership of x in A reflects a kind of ordering that is not based on the probability but on admitted
possibility. The value of PA(x) of object x in A can be interpreted as the degree of compatibility of the predicate associated with set A and object x. In other words, PA(x) of x in A specifies
the extent to which x can be regarded as belonging to A. The closer the value of PA(x) to 1, the
more likely x is to belong to A.
Unlike the classic set, a fuzzy set does not have sharply defined boundaries. The notion of a
fuzzy set provides a convenient point of departure for the construction of a conceptual framework which parallels in many aspects the framework used in the case of ordinary sets, but is
more general than the latter. The fuzzy framework provides a natural way of dealing with problems in which the source of imprecision is the absence of sharply defined criteria of class
membership rather than the presence of random variables (Zimmermann, 1988). Figure 1 illustrates the difference between the ordinary Boolean sets and fuzzy sets.
All the Boolean algebra operations (like intersection, union, negation, etc.) can be easily
extended to fuzzy set operations (Kandel, 1986). Therefore, if we can find a scheme to represent the spatial data from a fuzzy set perspective, then all the Boolean logic-based operations in
GIS should also be extended to the basis of fuzzy logic, which will lead to fuzzy cartographic
modeling. The following description shows how the fuzzy set theory can be incorporated into
cartographic modeling in a raster-based GIS.
To represent geographic information in a fuzzy scheme, a raster data layer in a GIS can be
defined as a fuzzy set since most geographic features do not have sharp boundaries, and each
grid cell (pixel) as the set element. In this way, each grid cell in the raster file can be attached
with a group of membership grades to indicate the extent to which the cell belongs to certain
attributes. Usually there are multiple data layers involved in the GIS modeling process and
these layers are stored in GIS as different raster files (overlays). Mathematically, these raster
files in question can be expressed as a data layer set:

Nonmarks

.;

(a)

FIGURE 1. Boolean Set versus Fuzzy Set: (a) Boolean Set (b) Fuzzy Set.

104

D. Z. Sui

DL = {DLl, DL2, ..., DLi, ..., D&)


where DLi, i = 1,2, . ... n, represents each individual data layer considered in the modeling process.
Because the weight of each individual data layer is different in the modeling process, we
need to represent the weight as another set:
W = (WI, W2, ***,Wi, **.,Wn)
where Wi, i = 1,2, .... n, represents the weight of each individual data layer.
Finally there should be a set containing the decision space or valuation scale:
v = (V,, v2, .... Vk)..., V,}
where Vk, k = 1,2, ..., m, represents each individual decision space or valuation scale.
For instance, Vk may be the urban development senario set {Option 1, Option 2, .....) if we
want to model which development strategy is the best, or Vk may be the valuation scale (Most
Valuable, Very Valuable, ..... ] if we want to evaluate urban land values.
All of the grid cells in the i-th data layer can be represented as a subset on V, that is:
RDLi =

trijk )

where RDG is actually a fuzzy matrix:


rill
RDLi=

ri12

rilm

i2m

ri21

i22

e
.

e
.

Gk
.

*
.

it2

L: ritl

. . . . . . . . . (2)

itm
11

where rijk is the membership grade of j-th grid cell in the i-th data layer to k-th class.
For the cells with the same location in all the data layers considered (same location but different attributes), they can also be represented as a subset on V, that is:
cj = (rijk)

In matrix form, it is:

Cj=

lj2
2j2

a
.

*
.

*
*

nj2

.....
4

ljm 1
'2jm

(3)

ijk
:

rnjm

Similar to matrix (2), qjk is also the membership grade of j-th grid cell in the i-th data layer
to k-th class. Matrices (2) and (3) are the essential forms of the fuzzy representation of raster
format data. rijk can be calculated by specific membership functions in a particular application
context. Combined with the weight set, the cartographic modeling procedure, which is a process of overlaying multiple data layers in GIS, is actually an operation of set W and Cj, or in
algebraic ~~IIIIS,a fuzzy mapping from Cj to Bj through W:

A Fuzzy G/S Modeling Approach for Urban Land Evaluation

ljl

'2jl

rlj2
r2j2

ijk

105

(4)
.
Lnjl

'nj2

.
*

where Bj is the j-th grid cell in the result data layer and 0 represents the operator when performing cartographic modeling. So far, there are several fuzzy operation schemes available
(Dubois & Prade, 1988). The following algebraic multiplication operator is used in this paper:

Bj,glwi*

. . . . . . . . . . . . . . . . .

rgk

The algebraic multiplication operator is adopted as the fuzzy operator in this research,
because all of the evaluation criteria have been taken into consideration in a weighted manner,
which is empirically consistent with the cartographic modeling process in GIS.
A CASE STUDY IN JINING,

P.R. CHINA

Background
The development of a market-oriented economy in mainland China has led to an urgent need
for urban land evaluation by the local government because of the gradual transition of land
ownership from the public to the private sectors. Urban land evaluation is essentially the process to determine the relative value of a certain block of urban land for taxation and investment
purposes. The most commonly used conventional method by the Chinese government in urban
land evaluation is the Expert Weight Evaluation (EWE) method which manually overlays gridbased planning base maps. This manual method is not only labor-intensive, but also may distort the real values of certain land parcels because of its inability to model the imprecision and
vagueness in the evaluation process. Obviously, new techniques are needed to automate the
process and model the fuzzy nature of urban land values. The author believes that the fuzzy
GIS modeling technique outlined above can be used to achieve these objectives.
To test the applicability and feasibility of the fuzzy GIS modeling approach, Jining, a medium-sized city in Shandong Province, Peoples Republic of China, is selected as the study area
for urban land evaluation. Jining is located in the Southwest of Shangdong province (Figure 2).
Its proximity to the Grand Canal and a major railway route has enabled the city to become the
major commercial and manufacturing center in the region. Its population is 291,321, of which
191,518 live in the city proper. All the data employed in this research are provided by the
Jining City Planning Council.
Evaluation Criteria
Urban land evaluation is a very complex process, which usually requires simultaneous consideration of many socioeconomic, locational and physical factors. Based on theories from
land economics and National Urban Land Evaluation Standards (NLAB, 1989), seventeen criteria are used for evaluation purposes in this research. They fall into seven main categories:
infrastructure; shopping convenience; transportation accessibility; social welfare; population
distribution; environmental quality; and physical conditions (Table 1). The concentration of

106

D. Z. Sui

FIGURE 2. Location of Study Area.

these factors results in two types of differential rents which are reflected in the
value: (a) the first type (differential rent I) is caused by the varying locations, (b)
type (different rent II) is the result of varying amounts of capital investment at the
similar locations within the city (Yeates, 1990).
In accordance with the common practice of urban land evaluation in China, the
set (valuation scale) V is partitioned into a five-component
subset:

urban land
the second
essentially
evaluation

V = IV,, V2, v3, V4, Vs)


= (Most Valuable, Very Valuable, Valuable,
Less Valuable, Least Valuable}
This valuation scale for urban land evaluation
ments of intermediate-sized
cities in China.

has been widely

adopted by local govern-

Derivation of Membership Functions


Successful applications of fuzzy mathematics depend to a large extent on the appropriateness
of membership functions. Both inductive and deductive approaches are available to determine
membership functions (Dombi, 1990). Considering the complexity and the availability of the
relevant data and the computational load in a PC environment, two kinds of membership functions are employed in this paper:

107

A Fuzzy G/S Modeling Approach for Urban Land Evaluation


TABLE 1. Evaluation Criteria, Membershlp Functions and Weights

ClitUia

Infra-

structure

Shopping

Membership

Weight
I

Function

Dhnce to Paved Road (In)

Function I

0.0830

-toStreet
here&on

Function I

0.0272

Water Supply

FunctionII

0.0275

Sewcrcrge
Condition

Fuoction II

0.0019

Ruidential Energy Source

Fuctioa II

0.0078

Dktanceto City Shopping

Fuction I

0.3295

Diehnce to Community
Shopping Center (km)-

Function I

0.0599

Dietancs to Public Traneit

Function I

0.0589

Dietance to City Bur Stop (m)

Fuction I

0.1172

Dietarm to Long-Distance
Bur
Terminal (km)

Function I

0.0662

Distance to Reilway Station


(km)

Function I

0.0097

Dhtancr to School (km)

Function I

0.0330

(m)

0.1587

Center (Inn)
0.3894

Transportation
Accessibility

0.2520

SOCid

Welfare

Dietancr to Entateinment
Center (km)
Distance to Hospital (km)

Population
Distribution

Population Den&y
(l.OoOhnf)

Environment

Pollution haesement

Quality
Physical
Condition

Function I

I
I

Function I

0.0570

0.0090
I

1 0.0100

Function I

0.0781

Function II

I
Fouodation Strength @g/cm?

Function I

0.0318

0.0318

I
0.0330

0.0330

Multi-Section Linear Function


The National Standards for Urban Land Evaluation in China (NLAB, 1989) were adopted to es&
lish the membership functions in this research. For example, the membership functions for the distance to a city shopping center are the multi-section linear functions shown in Figure 3 and Table 2.
The intuition for choosing this multi-section linear form function is derived from the deficiencies of the Boolean logic-based evaluation method. Conventional evaluation techniques
such as the EWE method often employ simple thresholds to determine whether the urban land

D. Z. Sui

108

MO.3

VW

VdlJable

VakJable

Valuable

.5

1.0

1.5

Viii%

Least
Valuable

2.0

2.5

Distance to City Shopping Center (km)


FIGURE 3. Membership Function I: Multl-Section Llnear Function.

is most valuable, very valuable, valuable, less valuable or least valuable. In terms of GIS operations, the simple Map Algebra commands such as ASSIGN, CLASSIFY or RECLASSIFY are
used to implement the cartographic modeling procedure. For instance, in terms of distance to a
city shopping center, any location less than 1 kilometer to the closest city shopping center will
be rated as the most valuable, 1000 meters to 1500 meters will be very valuable land, 1500 to
2000 meters will be valuable land, 2OOClto 2500 meters less valuable land, and larger than
TABLE 2. Membershlp Function for Distance to City Shopplng Center

Most Valuable

Less Valuable

Least Valuable

109

A Fuzzy G/S Modeling Approach for Urban Land Evaluation

2500 meters least valuable land. If two locations are respectively 999 meters and 1001 meters
from the closest city shopping center, then these two locations will be evaluated to be VI (most
valuable) and V2 (very valuable) according to the threshold model, even though the difference
of the distance (and hence the urban land value) is not particularly significant. Sharp boundaries are imposed to the gradual transition of the urban land value. The limitations of this simplistic approach can be avoided by the multi-section linear membership functions. With the
multi-section linear membership function as given in Table 2, all the locations are evaluated by
calculating the mem~rs~p functions so that gradual change of the urban land value can be
modeled more realistically. Based on this membership function, the evaluation of our two
hypothetical sites becomes (V,/O.O02, V2/0.998, V$O, V4/0, Vs/O) and (VI/O, V2/0.998,
V3/0.002, V4/0, V@). Instead of being rated into different land categories, these two land
parcels should be treated as Very Valuable land based on the membership grade since, for
both locations, @V2= 0.998. The membership grade also tells us that the first location
rn~n~ly
tends to be Most Valuable (V~~.~2) but by no means down rated as Valuable
land (V3xf) and the second location marginally tends to be rated to Valuable land (V$O) but
by no means up rated as Most Valuable (VI/O). The gradual transitional nature of urban land
value has been vividly described by the multi-section linear membership function.
Characteristic Function
The ch~~te~stic action

(Figure 4) normally takes the follo~g

ma~ematic~ form:

Owhenx=Vi
u (x) =

i=l,2
{ 1 whenx+

,..,,

m.

. . . . ..(N

Vi

The c~~te~stic
function may be regarded as the special case of fuzzy sets. Actually ordinary sets always implicitly assume characteristic functions, or in other words, when U(x) contains only the two points 0 and 1, a fuzzy set becomes a nonfuzzy set (Kandel, 1986). In this
research, the characteristic function is used for categorical map data such as water supply,
sewage condition, residential energy source and environment quality. This, however, does not
indicate that these criteria can be clearly defined. The characteristic functions are used for these
criteria because the original data for these categorical maps were not available. It is too difficult to define membership functions solely based on these categorical maps.

Method of Determining Weights


Saatys Analytic Hierarchy Process (AI-P) is used in this research to determine the weight of
each individual criterion (Saaty, 1980). AI-P is a ma~ematic~ method to determine priority of
the criteria in the decision-making process. It is a three-part process which includes identifying
and organizing the criteria into a hierarchy; evaluating pairwise comparisons between the relevant elements at each level of the hierarchy; and the synthesis using the solution algorithm of
the results of the pairwise comparisons over all the levels. In order to deal with the inconsistencies of judgment, the graph theoretic eigenvector approach (Saaty, 1980) is used as the solution
algorithm. Forty experts from both academic i~ti~tio~
and urban planning agencies were
surveyed about the relative weights of the criteria considered in this research. The weight of
each individual criterion is shown in Table 1.
Compared to other expert survey-based methods of determining weights used in the urban
planning community such as the Delphi method (Thrall, Swanson & Nozzi, 1988), the advantage of the AHP method is that it can deal directly with the inconsistent judgments and also
yield a measure of the ~co~is~ncies of the respondents judgment. Saaty (1980) has shown
that &&n for all the reciprocal judgment matrices, where n is the cardinality of the square

D. Z. Sui

110

.
LL L
MOSt

ValiJable

VW

Valuable

Valuable

Less
Valllable

Lea.54
Valuable

Category
FIGURE 4. Membership Function II: Characteristic Function.

matrix obtained by expert survey, Lrnax is the maximum eigenvalue of the matrix, and that
&.,ax=nifandonlyifA
is a perfectly consistent matrix. Thus the consistency index
C.I. = (Xrnax- n)/(n-1)
which can be used as a measure of the inconsistency of a matrix. If C.I. equals zero, then A is a
perfectly consistent matrix. As A becomes more and more inconsistent, C.I. becomes larger
and larger. The C.I. obtained in this research is C.I. = 0.07, which indicates an overwhelming
consensus among the experts over the weight of each individual criterion. (Saaty recommends
C.I. I 0.1 will be considered consistent judgment.) AHP method has become so popular for
decision making in recent years because of its ability to gauge the inconsistencies of expert
judgments (Vargas, 1990).
Implementation on G/S
The multi-criteria fuzzy evaluation model for urban land evaluation was implemented on a
raster-based micro-GIS Peking University Remote Sensing Information System (PURSIS)
developed at the Institute of Remote Sensing, Peking University.
First, all of the 17 subfactor maps were converted into digital format using the data capture
subsystem of PURSIS. Seventeen 160 x 160 raster data files were obtained, each pixel representing a 5Om x 5Om square on the ground. Then a FORTRAN-77 module was written to perform the fuzzy operation on the digital file according to the algebraic multiplication operator
(equation 5). The module for implementing the multi-criteria fuzzy model was implemented
through two separate phases. In the first phase, the subcriteria within each of the seven major
criteria were overlaid based on the fuzzy operations given by equation (5). The fuzzy matrix
(equation 3) was calculated by either multi-section linear function or the characteristic function
(Table 1). In the second phase, fuzzy operations were performed to overlay the seven digital
files obtained in the first phase into the final evaluation map.
RESULTS AND DISCUSSIONS
To facilitate comparisons, the urban land in Jining City was also evaluated based on the
same criteria using conventional cartographic modeling. The result from the fuzzy modeling
approach was compared with that of the traditional method.

A Fuzzy G/S Modeling Approach for Urban land Evaluation

111

To fully appreciate the superiority of the fuzzy GIS modeling approach, let us first take a
look at some simple queries based on a single criterion using map algebra commands. The
users query is to display the Very Valuable based on the distance to the city shopping center.
Suppose we just query a small window area which contains only 16 pixels as shown in Table 3.
Using the Boolean logic-based method, Very Valuable land is defined by two distance
thresholds ranging from 1.0 km to 1.5 km. According to these thresholds, only three pixels
meet the requirement of Very Valuable iand. They are pixels P5, P6, and Pll ffable 4), even
though intuitively we think several other pixels such as P8, P4, and PlO are approximately
qualified to be Very Valuable. But these approximately qualified pixels are simply excluded
from the Boolean logic-based queries because a sharp boundary is imposed to the Very
Valuable land and other land.
If the same query, however, is made through the fuzzy membership functions as defined in
Table 2, the answers are completely different (Table 4). The multi-section linear membership
function allows the user to model the gradual transition of urban land value. The boundaries
between the Very Valuable land and the Most Valuable land, the Very Valuable land and
the Valuable land are now fuzzy. All the approximately qualified pixels whose membership function is not zero will be included. More importantly, the membership grade attached to
each pixel tells the extent to which it is qualified to be rated as Very Valuable. Table 4 shows
that among the 16 pixels queried 9 are qualified to be considered as Very Valuable land.
According to the membership grade in Table 3, we can recognize that P8, P4, PlO, P13, and P7

TABLE3. MembershInGradefor VervValuableLand

112

D. Z. Sui
TABLE 4. Query Results for Very Valuable Land: Boolean Logic versus Fuzzy Logic

Boolean Logic

Fuzzy Logic

mVery Valuable

Very Valuable

[lOOOm, 15OOm]
P5, P6, Pll

[Defined as in table 21
Pll, P9, P6, P4,

are more qualified to be Very Valuable land than P5, but these six pixels have been totally
excluded by the Boolean logic-based query. Such findings suggest that the information loss and
distortion via the Boolean logic-based operations are very significant. The case will become
even worse when multiple criteria are employed in the GIS modeling process.
Figures 5 and 6 show the final results obtained from the traditional cartographic modeling
and the fuzzy GIS modeling approach based on the same 17 evaluation criteria. For fuzzy
modeling, the map was made based on the maximum membership grade value of each cell.
By comparing these two sets of results, we can find that both the fuzzy set and Boolean operations give similar results with respect to the extremes. The general pattern of urban land value
tends to decrease from the urban center to the periphery. The most valuable land is situated in
the old section of the city where comprehensive infrastructure has been developed during the
long history of growth. The advantageous location and intensive capital investment in the traditional CBD may explain the concentration of the most valuable land around the center of
the city. The concentric distribution of urban land values is typical for monocentric cities
around the world.
The differences between the two sets of results are also obvious. The fuzzy set approach provides more detail about the gradual transition of the urban land values. The distribution of
Most Valuable land obtained from the fuzzy approach is more tuned to shopping convenience and transportation accessibility criteria. Three other Most Valuable land pieces stand
out on the map of the fuzzy cartographic modeling approach (Figure 6), which have been rated
as only Valuable land using the conventional approach (Figure 5). The boundary between
Very Valuable and Valuable land has also changed substantially on these two maps. In
sum, instead of showing a monotonous concentric pattern, the fuzzy GIS approach gives more
details about the changing urban land value in Jining than the Boolean logic-based traditional
approach. Such detailed information and the associated membership grades are important for
the investors and local governments and may result in different investment decisions, development strategies, and land tax policies. Therefore, the fuzzy set approach is more appropriate for
modeling the imprecision and vagueness in the cartographic modeling process than the
Boolean logic-based approach.
However, the application of fuzzy set theory in GIS modeling is not without its problems.
The most pressing problem for using fuzzy set theory in cartographic modeling is the derivation of the membership function used in the fuzzy operation. All the approaches currently used
to derive the membership function are ad hoc in nature. Recently Dombi (1990) gave a detailed
discussion of the membership function from both a theoretical and empirical perspective, but
little research has been done about the membership function which is most suitable for geographic information and spatial decision making. Additional study of the membership function
for the fuzzy aspect of geographic problems is urgently needed.

A Fuzzy GIS Modeling Approach for Urban Land Evaluation

773

km

Most Valuable
Very Valuable
Valuable
Less Valuable
teast Valuable
FlGURE5. Urban Land Value.Evaluation Map I: Conventional CartographicMtiellng ApproactL
CONCLUDING REMARKS
This paper has addressed a problem that is particularly common in GISmodelmg for urban
applications: namely neither the objects nor their attributes may be uniquely defined. The
black-and-white Boolean logic which is the basis for most cartographic modeling breaks down
where there is ambiguity, imprecision and class overlap. Zadehs (1965) fuzzy set theory, as a
sensible extension of the normal Boolean algebra, can cope with the problems of uncertainty in
terms of perceived possibilities.
The study on merging fuzzy sets theory with GIS modeling in this paper indicates the fuzzy
cartographic modeling approach can avoid the loss of information that often arises when the
strict Boolean logic is used on the basis of the crisp sets (YES/NO). Combining fuzzy set theo-

114

D.

Z. Sui

km

Most Valuable
Very Valuable
Valuable
Less Valuable
Least Valuable
FIGURE 6. Urban Land Value Evaluation Map II: Fuzzy GIS Modellng Approach.

ry with GIS modeling procedures not only endows the latter with the capability to deal with
imprecision and vagueness, but also promotes further applications of fuzzy sets in the spatial
decision making process. Fuzzy cartographic modeling is shown to have fewer limitations in
modeling reality than the conventional modeling approach. The inadequacy of traditional
Boolean logic for representation and manipulation of spatial data can be avoided by incorporating fuzzy set theory into the GIS modeling process, as exemplified in a case study for urban
land evaluation of Jining, RR. China.
Future research agendas should include the development of the most suitable membership
function for geographic information and spatial decision making processes. The author
believes that the development of appropriate membership functions will be instrumental for

A Fuzzy G/S Modeling Approach for Urban Land Evaluation

115

further explicit applications of fuzzy logic in geographic data collection, storage/retrieval, snalysis/manipulation and display/representation.
REFERENCES
Blonda. P. N., Pasquariello. P N., Lo&o. S., Marl, A., Posa, F., & Ragno, D. (1991). An experiment for the integration
of multitemporal remotely sensed images based on a fuzzy logic approach. International Journal of Remote
Sensing, 12,463-416.
Bouille. E (1982). Actual tools for carmgraphy today. Cartogrophico, 19.27-32.
Burrough, R A. (1986). Geographical information systemfor land resource assessment. Oxford, UK: Clarendon Press.
Burrough. P. A. (1989). Fuzzy mathematical methods for soil survey and land evaluation. Journal qf Soil Science, 40,
411-492.

Burrough, P A. (1990). Methods of spatial analysis in GIS, International Journal of Geographical Information
Systems, 4,221-223.

Dombi. J. (1990). Membership function as an evaluation. Fuzzy Sets and Systems, 35.1-21.
Duboii, D., & Prade, H. (1988). Possibility theory: An approach to computerized processing of uncertainty. New York:
Plenum Press.
Kandel, A. (1986). Fussy mathematical techniques with applications. New York Addison-Wesley.
Kent, J. T., & Mardia, K. V. (1988). Spatial classification using fuzzy membership models. IEEE Transactions on
Pattern Analysis and Machine Intelligence, 10,659-671.

Kollias, V. J., & Voliotis, A. (1991). Fuzzy reasoning in the development of geographical information systems FRSIS:
A prototype soil information system with fuzzy retrieval capabilities. International Journal of Geographical
Intormation Systems, 5,20%223.
Leung, Y. (1988). Spatial analysis andplonning under imprecision. Amsterdam, The Netherlands: North-Holland
Leung, Y. (1989). Fuzzy logic and knowledge-based GIS: A prospectus. Proceedings of the 12th Canadian
Symposium on Remote Sensing, 1,4X50.
Muller. J.-C. (1987). The concept of error in cartography. Cartographica, 24.1-15.
NLAB (National Land Administration Bureau, P.R. China). (1989). National urban land evaluation standards,
Peoples Republic of China (Zhong Hua Ren Min Gong He Guo Cheng Shi Tu Di Fen Deng Ding Ji Biao Zhun),

China: Beijing.
Robinson, V. B., & Strahler, A. H. (1984). Issues in designing geographic information systems under conditions of
inexactness. Proceedings of the 10th International Symposium on Machine Processing of Remotely Sensed Data,
179-188.
Robinson, V. B. (1988). Some implications of fuzzy set theory applied to geographic databases. Computers,
Environment and Urban Systems, 12,89-97.

Robinson, V. B. (1990). Interactive machine acquisition of a fuzzy spatial relation. Computer Md Geosciences, 16,
857-812.
Rowman, G.-C. (1990). Formal specification of geographic data processing requirements. IEEE Transactions on
Knowledge and Data Engineering, 2,3X)-380.

Saaty, T. (1980). The analytical hierarchical process: Planning, priority setting, resource allocation. New York
McGraw-Hill.
Scholten, G. I.. & Stillwell, J. H. (1990). Geographical information systems for urban and regional planning. London:
Kluwer Academic.
Thrall, G. I., Swanson, B.. & Nozzi, D. (1988). Green-space acquisition ranking program (GARP): A computer-assisted decision strategy. Computers, Environment and Urban Systems 12,161-184.
Tomlii, C. D. (1990). Geographic information systems and cartographic modeling. New Jersey: Prentice Hall.
Vargas, L. G. (1990). An overview of analytic hierarchical process (AHP) and its applications. European Journal of
Operations Research, 48.2-8
Wang, F.-J., Hall, G. B., & Subaryono. (1990). Fuzzy information representation and processing in conventionat GIS
software: Database design and application. International Journal of Geographical Information Systems, 4.261-283.
Yeates. M. (1990). The North American city. New York Harper and Row.
Zadch, L. A. (1965). Fuzzy sets. It$ormation and Control, 8.338-353.
Zimmennann. H.-J. (1988). Fussy set theory-An inferential mechanism in mathematical modelsfor decision support.
NATO AS1 Series, 727-741.

Вам также может понравиться