You are on page 1of 15

+ MODEL

ARTICLE IN PRESS
ENGEO-02587; No of Pages 15

Engineering Geology xx (2006) xxx – xxx


www.elsevier.com/locate/enggeo

Preparation of a geotechnical microzonation model using


Geographical Information Systems based on
Multicriteria Decision Analysis
Çağıl Kolat a,⁎, Vedat Doyuran a , Can Ayday b , M. Lütfi Süzen a
a
Department of Geological Engineering, Middle East Technical University, Ankara 06531, Turkey
b
Anadolu University, Research Institute of Satellite and Space Sciences, Eskişehir 26470, Turkey
Received 7 July 2005; received in revised form 8 July 2006; accepted 13 July 2006

Abstract

The purpose of this study is to develop a geotechnical microzonation model using Geographical Information Systems (GIS)
based on Multicriteria Decision Analysis (MCDA). As study area, the Eskişehir downtown area has been chosen. Eskişehir is one
of the most rapidly growing cities in central Turkey. The model inputs include slope, flood susceptibility, soil, depth to groundwater
table, swelling potential, and liquefaction potential. The weight and rank values are assigned to the layers and to the classes of each
layer respectively. The assignment of the weight/rank values and the analysis are realized by the application of two different
decision models, namely Simple Additive Weighting (SAW) and Analytic Hierarchy Process (AHP) methods. The geotechnical
microzonation maps prepared as outputs of these methods are found to be consistent with each other and confirmed by the experts
within the study area. The geotechnical microzonation map prepared using the AHP method is recommended as the final map of
the study.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Geotechnical microzonation model; Geographical Information Systems; Multicriteria Decision Analysis; Analytic Hierarchy Process;
Eskişehir

1. Introduction provides an effective solution for this requirement. Most


of these geotechnical site selection applications require
Rapidly growing cities with increasing population an enormous amount of data which must be geograph-
underline the requirement for new residential areas. En- ically related to each other. Conventionally, geo-
gineering geological evaluations should be performed in environmental evaluation and mapping were laborious
order to determine the most suitable residential areas. and time-consuming tasks because of the large amount of
Preparation of the geotechnical microzonation maps time and effort required for the manual handling and
processing of the spatial data (Dai et al., 2001). Con-
sequently, there exists a necessity of a system where all
⁎ Corresponding author. Tel.: +90 312 210 57 23; fax: +90 312 210
of these large quantities of data could be manipulated
57 50.
E-mail addresses: ckolat@metu.edu.tr (Ç. Kolat), with ease. Geographical Information Systems (GIS),
vedat@metu.edu.tr (V. Doyuran), cayday@anadolu.edu.tr (C. Ayday), being a computer-based system that enables acquisition,
suzen@metu.edu.tr (M. Lütfi Süzen). storage, retrieval, modeling, manipulation and analysis
0013-7952/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.enggeo.2006.07.005

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
2 Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx

of geographically related data (Aronoff, 1993; Worboys, the sense that results of the analysis (decisions) depend
1995), has provided a complimentary solution for this on their spatial arrangement. Accordingly, many real-
requirement. world spatial planning and management problems give
Site selection decision problems involve a set of rise to GIS based multicriteria decision making or spatial
geographically defined alternatives, from which a choice Multicriteria Decision Analysis (MCDA) (Malczewski,
of one or more alternatives is to be made on the basis of 1999).
multiple, conflicting and incommensurate evaluation GIS should be considered as a special-purpose digital
criteria. The alternatives are geographically defined in database in which a common spatial coordinate system

Fig. 1. Location map of the study area.

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx 3

is the primary means of storing and accessing data while been chosen (Fig. 1). Eskişehir is one of the most
processing the data to obtain information for decision rapidly growing cities in the Central Anatolian region
making. The ultimate aim of GIS is to provide support of Turkey. The study area is bounded by the
for decision making (Densham, 1991). This can be coordinates 4408166 N and 283628 E in the north-
achieved by integrating the MCDA and the analytical western edge and 4403718 N and 288712 E in the
capabilities of GIS (Diamond and Wright, 1988; Carver, southeastern edge in Universal Transverse Mercator
1991; Eastman et al., 1995; Jankowski, 1995; Keller, (UTM) projection (Zone 36 N, European Mean Datum
1996; Malczewski, 1999). 1950). The study area covers approximately 22.5 km2
The purpose of this study is to prepare a (∼ 5.1 km × 4.4 km). Since the study area is already
geotechnical microzonation model using MCDA tech- densely settled, the proposed microzonation model can
niques with GIS support and to propose a concise also be used to check the suitability of already settled
flowchart to be used in further similar studies. As study areas and also to determine if further precautions are
area, the sub-section of Eskişehir downtown area has needed for safer planning actions or modifications.

Fig. 2. Flowchart for the preparation of the microzonation map of the study area using GIS based MCDA.

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
4 Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx

2. Methodology From these three data sources six different predictor maps
were produced. These were slope, flood susceptibility, soil,
The steps followed throughout the study are pre- depth to groundwater table, swelling and liquefaction
sented in Fig. 2, where steps of implementing MCDA and potential layers. The next step was to assign weight and
GIS for the preparation of the geotechnical microzonation rank values to the layers and to the classes of each layer,
model are emphasized. The data sets used in this study can respectively. The assignment of the weight/rank values and
be grouped in three main data sources as topographical base the analysis were realized by the application of two
maps, lithological maps and geotechnical borehole data. different decision models, namely the Simple Additive

Fig. 3. a. The Digital Elevation Model (DEM) of the study area, b. Slope layer of the study area, c. Flood susceptibility layer of the study area.

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx 5

Weighting (SAW) and the Analytic Hierarchy Process discarded since they seem to produce significant errors in
(AHP) methods. representing the study area properly. The other four
This flowchart can be applied for further similar studies, methods were evaluated according to their accuracy. The
provided that the layers used for the analysis are determined accuracy assessment for DEM was performed by the
according to the needs of the study area. calculation of Root-Mean-Square Error (RMSE). For
computing the RMSE, sample points were taken accord-
3. Data evaluation ing to USGS mapping standards (USGS, 2004).
The true and interpolated topographic values of the
3.1. Topographical map sample points were examined, and compared by
calculating their RMSE; to be able to decide which
1:25.000 and 1:5.000 scale topographical maps were interpolation method should have been used. During the
first registered according to the Universal Transverse evaluation, it was observed that when the study area is
Mercator (UTM) projection system (Zone: 36 and divided into two sections (as flat area and gentle hills)
Datum: European 1950 — Mean) and digitized. These and two different methods are applied for each respective
digitized contours were used to produce the Digital section, the vertical accuracy of the produced DEM
Elevation Model (DEM) of the study area. increases. The RMSE values were calculated, compared
The cell size of the DEM was determined by the and evaluated separately for these two sections. As a
method proposed by Florinsky and Kuryakova (2000) result, the profile surface fitting algorithm was used for
where the adequate cell size for the DEM of the study the gentle hills section and the minimum curvature
area was determined as 10 m. Six different surface fitting algorithm for the flat area section. The integration of
methods were applied and evaluated according to their these two methods yielded an RMSE of 0.079 m in the
accuracy in order to decide the most appropriate method whole DEM (Fig. 3a), which is quite acceptable.
that represents the natural environment best. The surface
fitting methods applied were; minimum curvature, in- 3.1.1. Slope layer
verse distance, profiles, polynomial, triangulation and Slope is an important factor while considering the
kriging. Inverse distance and polynomial methods were ease of engineering construction and susceptibility to

Fig. 4. a. Lithology map of the study area (Ayday et al., 2001), b. Soil layer of the study area including the borehole locations.

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
6 Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx

Fig. 5. a. Depth to groundwater table layer of the study area including the borehole locations, b. Swelling potential layer of the study area including
the borehole locations, c. Liquefaction potential layer of the study area (modified from Ayday et al., 2001).

landsliding (Dai et al., 2001; Süzen and Doyuran, the ease of engineering constructions, since steep slopes
2004a). There is no slope stability problem encountered interfere with excavation processes.
in the study area (Koyuncu, 2001). Therefore, the slope The slope map was prepared in degrees using the DEM
layer will only contribute to the microzonation map in of the study area. Afterwards, the slope values were

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx 7

Table 1 (Lower Eocene); Conglomerate-sandstone, Tuff–Marl–


Assigned weight and rank values for the layers/classes of the study Clay and Limestone Members of Porsuk Formation
area
(Upper Miocene); Old Alluvium (Quaternary) and Young
Layers Weighting Classes Ranking Alluvium (Quaternary). In the study area, Mamuca and
Liquefaction 10 No liquefaction 5 Porsuk Formations were classified as bedrock.
potential layer potential The soil layer of the study area was based on the
Moderate liquefaction 2
lithology map and on the report prepared by Ayday et al.
potential
High liquefaction 1 (2001). The relevant data of 66 boreholes were reviewed
potential and evaluated. The liquid limit and plasticity index data
Flood susceptibility 8 Non-flood areas 5 were compiled for the depth range corresponding to the
layer Flood-prone areas 1 average foundation excavations and the data were
plotted into the plasticity chart in order to evaluate the
Soil layer 7 Bedrock 5
SC 4 soil behavior. The units classified as bedrock were not
MH/ML 3 taken into consideration. The soil layer of the study area
CL 2 was prepared according to the Unified Soil Classifica-
CH 1 tion System (USCS). Based on USCS, the main soil
Depth to groundwater 5 N10 m 5
classes include, SC (Clayey sands, sand–clay mixtures),
table layer 5–10 m 3
0–5 m 1 MH / ML (Inorganic silts, micaceous or diatomaceous
Swelling 4 Bedrock 5 fine sands or silts, elastic silts/Inorganic silts, very fine
potential layer Low expansion 4 sands, rock flour, silty of clayey fine sands), CH
Medium expansion 3 (Inorganic clays of high plasticity, fat clays) and CL
Slope layer 3 Most favorable 5
(Inorganic clays of low to medium plasticity, gravelly
slope class
Intermediate 4 clays, sandy clays, silty clays, lean clays). The results
slope class
Least favorable slope 1 Table 2
class Standardized rank values and normalized weight values
Layers Weighting Classes Ranking
subdivided into three main classes according to the (normalized) (standardized)
Guidelines for Urban Engineering Geological Investiga- Liquefaction 0.2703 No liquefaction 1
tions (South African Institute of Engineering Geologists). potential layer potential
Slopes between 2° and 6° were assigned as the most Moderate 0.4
liquefaction
favorable class, slopes less than 2° and slopes between 6°
potential
and 12° were assigned as the intermediate class, and slopes High liquefaction 0.2
greater than 12° were assigned as the least favorable class. potential
The slope layer of the study area is given in Fig. 3b. Flood susceptibility 0.2162 Non-flood areas 1
layer Flood-prone areas 0.2
3.1.2. Flood susceptibility layer
Soil layer 0.1892 Bedrock 1
The flood susceptibility of the study area was SC 0.8
examined in case of thunderstorms of long duration. MH/ML 0.6
The potential flood-prone areas (Fig. 3c) were defined CL 0.4
as the areas having slope values less than 2° within the CH 0.2
Depth to 0.1351 N10 m 1
Porsuk Stream floodplain. It should be noted that in the
groundwater 5–10 m 0.6
preparation of the flood susceptibility layer, it was table layer 0–5 m 0.2
assumed that the Porsuk Dam will not fail, even in case Swelling 0.1081 Bedrock 1
of an earthquake. potential layer Low expansion 0.8
Medium expansion 0.6
3.2. Lithology map and soil layer
Slope layer 0.0811 Most favorable 1
slope class
The lithology map of the study area was taken from Intermediate 0.8
Ayday et al. (2001). For the study area, six lithological slope class
units were identified (Fig. 4a). These units include Least favorable 0.2
slope class
Conglomerate-sandstone Member of Mamuca Formation

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
8 Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx

Fig. 6. The microzonation map prepared by using the SAW method and the histogram showing the boundaries of the three zones.

obtained were evaluated spatially and the boundaries of lyzed and evaluated in order to obtain the depth to
the soil classes were determined manually. The soil layer groundwater table, swelling potential and liquefaction
of the study area including the borehole locations is potential layers.
given in Fig. 4b.
3.3.1. Depth to groundwater table layer
3.3. Borehole data Groundwater is one of the main factors governing the
stability of foundation excavations as well as the ease and/
The borehole data used in this study was obtained or difficulty of the excavation works. In liquefaction as-
from Ayday et al. (2001). The borehole data were ana- sessment the position of the water table within non-

Fig. 7. Hierarchical structure used in the preparation of the microzonation map.

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx 9

Table 3
Comparison judgments from a fundamental scale of absolute numbers
4. Analysis
for assigning weight/rank values (Saaty, 2004)
4.1. Common steps of Multicriteria Decision Analysis
Weight/Rank Intensities
applications
1 Equal
3 Moderately dominant
5 Strongly dominant In order to obtain geotechnical microzonation model of
7 Very strongly dominant the study area two methods were applied: Simple Additive
9 Extremely dominant Weighting (SAW) and Analytic Hierarchy Process (AHP).
2, 4, 6, 8 Intermediate values In the analysis, weight values to the layers and rank
Reciprocals For inverse judgments
values to the classes of each layer were assigned. For
each class of the layer, rankings were given according to
cohesive sediments should also be known. The lithological their significance in foundation performance. After the
units exposed in the study area formed fair-to-poor aquifers rankings were assigned to the classes of each layer, the
due to the abundance of MH/ML type soils. Both the weights were assigned to layers according to their
bedrock and the alluvial deposits were hydraulically con- importance. The interaction between the layers was not
nected and serve as a single aquifer, which was unconfined. taken into consideration since the layers were assumed
The depth to water table layer was prepared by to be independent of each other.
considering the highest elevations of the static water levels. The weight and rank values of the layers and classes of
Thus, from the available geotechnical boreholes, six within each layer were standardized in order to obtain a common
the study area and six located at close vicinity, groundwater dimensionless unit. Afterwards, the output (microzona-
levels measured during May were compiled and depth to tion map) was created by multiplying the weight value
water table layer was prepared. Areas underlain by shallow assigned to each layer by the rank value given to the
(0–5 m) groundwater table were considered as least classes of that layer and finally by adding up the products.
favorable, between 5–10 m as favorable, and deeper than As a result, the microzonation map of Eskişehir
10 m as the most favorable (Fig. 5a). downtown area regarding the foundation suitability of
residential areas was grouped into different zones based
3.3.2. Swelling potential layer on the recommended subdivisions by The General
The swelling potential layer of the study area was Directorate of Disaster Affairs of Turkey (GDDA, 2000).
determined using 13 boreholes, for the first 2.5 m from the
surface, since below this depth, the soil moisture remains 4.2. Simple Additive Weighting (SAW)
constant. After compiling the Clay Content (%) and
Plasticity Index (%) from the borehole data within 0– In the SAW method, the weight and rank values are
2.5 m depth interval, these values were evaluated by using given totally based on expert opinion. In this method, all
the activity chart, and were found to be within the low and of the layers are concurrently considered in assigning
medium expansion classes. Therefore, the classes of the weight values, and all classes of each layer are also
swelling potential layer of the study area were determined concurrently considered while assigning rank values. As
as; low expansion and medium expansion (Fig. 5b). a result, six weight values were assigned to the six layers.

3.3.3. Liquefaction potential layer 4.2.1. Assigning weight and rank values
The liquefaction potential layer of the study area was In the assigning of weight and rank values, inverse
taken from Ayday et al. (2001), which was based on the weighting and ranking criteria was used. For weight
Seed and De Alba (1986) approach. In the calculations,
peak horizontal acceleration was taken as 0.4 g and the Table 4
grain size data for a depth of 5 m was used. The pairwise comparison table for assigning weight values
The liquefaction potential map was reclassified to
Liquefaction Swelling Soil Depth Slope Flood
obtain the liquefaction potential layer of the study area. to gwt
The classes of “bedrock” and “groundwater level deeper
Liquefaction 1 5 3 4 6 1
than 10 m” were assigned to the class of “no liquefaction”. Swelling 1/5 1 1/3 1/2 1 1/4
The liquefaction potential layer of the study area includes Soil 1/3 3 1 2 4 1
high liquefaction potential (Factor of Safety (FS) b 1.0), Depth to gwt 1/4 2 1/2 1 3 1/3
moderate liquefaction potential (1.0 b FS b 1.2) and no Slope 1/6 1 1/4 1/3 1 1/5
Flood 1 4 1 3 5 1
liquefaction potential (FS N 1.2) (Fig. 5c).

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
10 Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx

Table 5
The pairwise comparison tables for assigning rank values to the classes of each layer
Liquefaction potential layer No liquefaction Moderate High
potential liquefaction liquefaction
potential potential
No liquefaction potential 1 7 9
Moderate liquefaction potential 1/7 1 2
High liquefaction potential 1/9 1/2 1
Flood susceptibility layer Non-flood areas Flood-prone areas
Non-flood areas 1 1/5
Flood-prone areas 5 1
Soil layer Bedrock SC MH/ML CL CH
Bedrock 1 3 5 6 7
SC 1/3 1 3 5 6
MH/ML 1/5 1/3 1 3 4
CL 1/6 1/5 1/3 1 3
CH 1/7 1/6 1/4 1/3 1
Depth to groundwater table layer N10 m 5–10 m 0–5 m
N10 m 1 2 7
5–10 m 1/2 1 4
0–5 m 1/7 1/4 1
Swelling potential layer Bedrock Low expansion Medium
expansion
Bedrock 1 3 5
Low expansion 1/3 1 3
Medium expansion 1/5 1/3 1
Slope layer Most favorable Intermediate Least
slope class slope class favorable
slope class
Most favorable slope class 1 3 6
Intermediate slope class 1/3 1 5
Least favorable slope class 1/6 1/5 1

values, the assignment start from the least important microzonation map, the depth to groundwater table,
with the value of 1, the next least important is assigned swelling potential and slope layers have carried
the value 2, and the most important layer gets the value relatively low importance, since the possible problems
of 10. Similarly the rank values were quantified as the can be handled relatively more easily and practically.
least important class value being 1 and the most im- The weight assigning order of these three layers was
portant class value being 5. given according to the ease of the precautions needed, as
The assigned weight and rank values for the layers/ (in descending order): depth to groundwater table,
classes of the study area based on engineering judgment swelling potential and slope.
are given in Table 1. As can be observed from the table,
the most important layer was defined as the liquefaction 4.2.2. Standardization of rank and weight values
potential layer, followed by the flood susceptibility, soil, The simplest formula for standardizing the raw data
depth to groundwater table, swelling potential and slope is to divide each raw score by the maximum value for a
layers in decreasing order of importance. In the given criterion (Malczewski, 1999). Hence, the rank
determination of the weight values, the liquefaction values of the classes were standardized according to the
layer had the greatest value since the study area is relative distance between the origin and the maximum
located in the second degree earthquake zone (GDDA, rank value, using the following formula:
1996). The flood layer has taken the second important
role since there is a large floodplain in the study area and X Vij ¼ Xij =Xjmax
for the case of thunderstorms of long duration, there is a
possible risk of flood (In 1963, the overflow of Porsuk where X'ij is the standardized rank value for the ith class
Stream resulted in serious flood hazard). Soil layer was for the jth layer. Xij is the raw rank value, and Xjmax is
also important since the behavior of the soil under static the maximum rank value for the jth layer.
and dynamic loading conditions should be taken into On the other hand, the weight values were normal-
consideration during construction. In contributing to the ized by dividing each weight by the sum of the weights.

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx 11

Table 6 4.2.3. Result map of SAW


The computed weight and rank values For the preparation of the microzonation map of the
Layers Weighting Classes Ranking study area, the overlay operations of the layers were used.
Liquefaction 0.350722 No liquefaction 0.792757 The formula proposed by Malczewski (1999) for
potential layer potential obtaining the total scores was applied in this study.
Moderate 0.131221 Accordingly, each pixel of the output microzonation map
liquefaction
(Mi) was calculated by using the following summation:
potential
High liquefaction 0.076022 Mi ¼ Rj wj xij
potential
Flood susceptibility 0.256510 Non-flood areas 0.833333 where, xij = rank value of the ith class with respect to
layer Flood-prone areas 0.166667 the jth layer
Soil layer 0.184303 Bedrock 0.496900
wj = normalized weight of the jth layer.
SC 0.264927
MH/ML 0.129109 Thus the normalized weight value assigned for each
CL 0.069362 layer was multiplied by the standardized rank value
CH 0.039702 given to the classes of that layer. Finally the sum of the
Depth to groundwater 0.102535 N10 m 0.602629 products was calculated.
table layer 5–10 m 0.315029
The microzonation map of the study area was
0–5 m 0.082342
Swelling 0.057615 Bedrock 0.636986 categorized in three resultant classes as: Areas Suitable
potential layer Low Expansion 0.258285 for Settlement (SA), Provisional Settlement Areas
Medium expansion 0.104729 (PSA) and Detailed Geotechnical Investigation Re-
quired Areas (DGA) (Fig. 6). The boundary conditions
Slope layer 0.048315 Most favorable 0.634838
for the categories were evaluated according to the expert
slope class
Intermediate 0.287203 judgment taking into consideration the score distribu-
slope class tions by means of discrete histograms (Fig. 6).
Least favorable 0.077959
slope class 4.3. Analytic Hierarchy Process (AHP)

In the AHP method, pairwise comparisons form


Thus the sum of the normalized weight values was equal the backbone of the methodology. AHP provides us
to 1. The standardized rank values and the normalized with a way to derive from an observer's quantified
weight values are given in Table 2. judgments (i.e., from the relative values associated

Fig. 8. The microzonation map prepared by using the AHP method and the histogram showing the boundaries of the three zones.

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
12 Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx

with pairs of stimuli), a set of relative weights or comparison judgments presented in Table 3. The
priorities associated with individual stimuli (Özdemir, implementation of pairwise comparison tables for
2005). Hence, in assigning the weights of the layers, assigning rank values to the classes of each layer is
only two layers were considered at a time. In total 15 given in Table 5.
different weights were given, since in assigning each
weight, only two layers were considered at a time. In 4.3.3. Obtaining overall weight and rank values
other words, the total number of pairings where every In order to obtain the overall weight/rank values, the
layer was matched with all others is 15. Similarly, the eigenvector solution was used. This approach had been
rankings of the classes of each layer were also demonstrated mathematically by Saaty (1990). The
assigned considering only two classes at a time. solution for the eigenvector can be explained in the
The hierarchical structure used in the preparation of following steps:
the microzonation map is given in Fig. 7. In assigning
weights, pairwise comparisons of each layer with other i. A short computational way to obtain the weight/
layers were assessed, while in assigning rank values, rank values is to raise the pairwise matrix to pow-
pairwise comparisons of each class with other classes ers that are successively squared each time.
of the same layer were assessed. For the assignment of ii. The row sums are then calculated and normalized.
weight and rank values, the comparison judgments iii. The iteration is instructed to stop when the
scale from Saaty (2006) was used, as given in Table 3 difference between these sums in two consecutive
with their corresponding meanings. calculations is smaller than a prescribed value.

4.3.1. Assigning weight values Applying the above method, the overall weight
The pairwise comparison matrix for assigning values of the layers and the overall rank values of the
the weight values is given in Table 4. The logic classes of each layer were obtained. The pairwise
behind assigning values in the pairwise comparison comparison matrix was squared; row sums were
matrix can be explained through the liquefaction layer calculated and normalized, the matrix obtained was
as follows: squared; row sums were calculated and normalized
The liquefaction layer has equal importance with the again. This process was repeated until the difference
flood layer and moderate prevalence to soil layer, between the normalized values fell below the
whereas it has moderate–strong, strong, and strong– threshold of computing capabilities. When the differ-
very strong prevalence against depth to groundwater ence value appeared to be zero, then the normalized
table, swelling and slope layers, respectively. values were taken as weight/rank values for the
classes/layers. The computed weight and rank values
4.3.2. Assigning rank values are given in Table 6.
The AHP method can be used not only to assess
weights but also to assess the performance of 4.3.4. Consistency ratio
alternatives by pairwise comparison of the alternatives The consistency ratio was calculated in order to
(Janssen, 1992). In order to assign the rank values to determine whether the pairwise comparisons were
the classes of each layer, the pairwise comparison consistent or not. The consistency ratio (CR) is designed
matrix was prepared separately for the layers, using the in such a way that if CR b 0.10, the ratio indicates a
reasonable level of consistency in the pairwise compar-
Table 7
isons; if CR ≥ 0.10, the values of the ratio are indicative
Possible result numbers of the addition process with the corresponding
percentages of inconsistent judgments (Malczewski, 1999). The
consistency ratios for all of the pairwise comparisons
used in order to obtain the microzonation map were
calculated and found to be consistent.

4.3.5. Result map of AHP


The geotechnical microzonation map of the study
area using results of AHP method was prepared in a
manner consistent with that of the SAW method.
Thus, the normalized weight value assigned for each
layer was multiplied by the standardized rank value

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx 13

given to the classes of that layer. Finally the sum of maker to make better decisions under such circum-
the products was calculated. stances (Starr and Stein, 1976).
As in the SAW result, the AHP microzonation map In decision making context, a criterion would imply
of the study area was categorized again in three resul- some sort of standard by which one particular choice or
tant classes as: Areas Suitable for Settlement (SA), course of action could be judged to be more desirable
Provisional Settlement Areas (PSA) and Detailed Geo- than another. Actually in real life, every decision re-
technical Investigation Required Areas (DGA) (Fig. 8). quires the balancing of multiple factors so that in some
The boundary conditions for the categories were sense, everyone is well practiced in multicriteria de-
evaluated according to the expert judgment taking cision making. However, the human brain can only
into account the score distributions by means of simultaneously consider a limited amount of informa-
discrete histograms (Fig. 8). tion, so that all factors cannot be resolved in one's head
(Belton and Stewart, 2002).
4.4. Comparison of microzonation maps Usage of GIS based MCDA is essential in the
preparation of geotechnical microzonation maps due to
In order to compare the microzonation maps which the need for using a large amount of spatial data and
were prepared using SAW and AHP methods, the integrating the geographical data with the decision
common areas and noncommon areas were highlighted. maker's preferences.
During this process, firstly the classes of a categorical Two considerations are of critical importance for
map were transferred from the ordinal scale to the spatial Multicriteria Decision Analysis: (i) the GIS capa-
interval scale (Süzen, 2002). Numbers were assigned to bilities of data acquisition, storage, retrieval, manipula-
the classes of both microzonation maps, in such a way tion and analysis, and (ii) the MCDA capabilities for
that, the sum of these values from the two microzona- aggregating the geographical data and the decision
tion maps must be unique and represent different maker's preferences into uni-dimensional values of
conditions (Süzen and Doyuran, 2004b). In this study, alternative decisions (Carver, 1991; Jankowski, 1995).
two microzonation maps were added to find out the Accordingly, the possible sources of errors in our study
common and noncommon areas. The possible results of can be categorized as data related errors and errors
the addition process are shown in Table 7 with the resulting from the decision maker's preferences:
corresponding percentages obtained from the addition Considering the data related errors; the original
process. geological, hydrogeological, and geotechnical data
As a result of the comparison analysis of the study seem to be quite satisfactory for this study. However,
area, the two microzonation maps were found to give the number of boreholes (especially for the groundwater
similar outputs. In the comparison output map; only the level observations) could be increased to obtain more
classes of “11”, “22”, “33”, and “32” were found. The information for better geotechnical characterization of
absence of classes “13” and “31” shows that there the study area. For the preparation of the data layers, the
exists no conflict between the microzonation maps. continuous surfaces are formed from the interpolation of
Furthermore, classes “11”, “22”, and “33” stand for this raw point/line data. During this interpolation
exactly same classes both in SAW and in AHP to DGA, process, some errors may have occurred due to lack of
PSA and SA, respectively. Unfortunately class “32” information between the consecutive points/lines. We
means that the “PSA” class found in the microzonation tried to minimize these errors in the preparation of DEM
map prepared with SAW method is found as “SA” class of the study area by applying several surface fitting
in the microzonation map prepared with AHP method. methods and choosing the method with least RMSE
As a result, the ratio of the correct classification value. For the maps prepared using the borehole data, the
category (classes “11”, “22” and “33”) in the point data is interpolated manually as accurate as
comparison of two microzonation maps was found to possible.
be 98.39% of the total study area. In addition to the data related errors, there is un-
certainty involved in the specification of decision maker
5. Discussion preferences. In fact, the criterion map errors and de-
cision maker preference errors are interrelated. The
Multiple objectives are essential to many ‘real’ sys- information derived from criterion maps is an essential
tems. Frequently, these multiple objectives conflict with element for specifying the decision maker's preferences.
each other (as one objective is improved, the others may For reliable results, the decision maker is expected to be
deteriorate). Dimensional analysis can help the decision an expert to make preferences since the importance of

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
14 Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx

each criterion can be overestimated or underestimated 6. Conclusion


according to these preferences.
The subjectivity of the preferences comes mainly This study demonstrates the superiority of the usage of
from the assignment of weight and rank values. In the MCDA techniques with GIS for the preparation of the
scope of this study, the weight and rank values which are geotechnical microzonation map regarding the suitability
used both in the SAWand the AHP methods, are assigned of the residential areas. The important advantages of using
properly according to the engineering judgment. these techniques can be summarized as having relatively
The basic differences between the SAW and the AHP low cost, easy data manipulation, rapidly updating of data
methods lie in their objectiveness, easiness and evaluation and the possibility to produce various new scenarios.
opportunities. Although AHP is more complicated than In this study, the slope, flood susceptibility, soil,
SAW, AHP has more objective results. The basic strategy depth to groundwater table, swelling potential and
is to divide the decision problem into small, understand- liquefaction potential layers were prepared for the
able and manageable parts; analyze each part; and chosen study area in Eskişehir. The assignment of the
integrate the parts in a logical manner to produce a weight and rank values and the analysis were performed
meaningful solution (Malczewski, 1999). In the AHP by application of the SAW and the AHP methods. As a
method, this strategy is applied in assigning rank and result, the study area was categorized into three different
weight values since only two layers/classes are considered zones regarding the foundation suitability of residential
and compared at a time. This decreases the subjectivity of areas as: (1) Areas Suitable for Settlement; (2)
the study and brings an advantage to AHP method. Provisional Settlement Areas; (3) Areas requiring
Besides, the pairwise comparison for the determina- detailed geotechnical investigations. The maps prepared
tion of weights is more suitable than direct assignment using the SAW and the AHP methods were found to be
of the weights, because one can check the consistency of consistent with each other. The geotechnical micro-
the weights by calculating the consistency ratio in zonation map prepared using the AHP method is
pairwise comparison; however, in direct assignment of recommended as the final map of the study area.
weights, the weights are depending on the preference of According to this final map; the majority of the study
decision maker (Şener et al., 2006). area is found to be in the “Areas Suitable for Settlement”
On the other hand, the SAW method definitely has an (SA) and “Provisional Settlement Areas” (PSA) zones.
advantage in rapidity. In applying this method, the result The minority part of the study area is found to be in
can be realized quickly with the contribution of a DGA, in which detailed geotechnical investigations are
qualified expert. However, since SAW method uses required. For the already settled areas which are not
direct assignment of the weights/ranks, the qualification found in the first zone (SA), the necessity of the
of the expert needed is much more than needed in AHP. precautions must be considered. The methodology used
Furthermore, the AHP method provides the user to be for the chosen area in this study can also be applied to the
able to evaluate the situation in different aspects. In other locations and to other site selection procedures by
many decision problems, four kinds of concerns are reconstructing the necessary parameters appropriately.
considered: benefits, opportunities, costs, and risks
(BOCR); in which the first two are advantageous, and Acknowledgments
hence, are positive and the second two are disadvanta-
geous and are therefore negative (Saaty and Özdemir, This study was supported by METU Research
2003). They have shown that the negative priorities can Foundation Grant No: BAP-2004-07-02-00-57.
also be defined as relative numbers and used along with
positive priorities in AHP. Therefore, another important References
advantage of the AHP method is to lead the expert to be
able to evaluate the BOCR of the problem separately, Aronoff, S., 1993. Geographic Information Systems: A Management
Perspective. WDL Publications, Ottowa, Canada.
which is not possible when applying the SAW method. Ayday, C., Altan, M., Nefeslioğlu, H.A., Canigür, A., Yerel, S., Tün,
The microzonation maps obtained by using the SAW M., 2001. Preparation of Engineering Geological Map of Eskişehir
and the AHP methods are found to be consistent with Urban Area. Research Institute of Satellite and Space Sciences,
each other. The reason for this consistency lies in the Anadolu University (in Turkish).
proper assignments of weight and rank values by the Belton, V., Stewart, T.J., 2002. Multiple Criteria Decision Analysis — An
Integrated Approach. Kluwer Academic Publishers Group. 372 pp.
expert. The microzonation map prepared using the AHP Carver, S.J., 1991. Integrating multicriteria evaluation with Geograph-
method is chosen to be the final map of this study due to ical Information Systems. International Journal of Geographical
the clear advantages of this method. Information Systems 5 (3), 321–339.

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.
ARTICLE IN PRESS
Ç. Kolat et al. / Engineering Geology xx (2006) xxx–xxx 15

Dai, F.C., Lee, C.F., Zhang, X.H., 2001. GIS-based geo-environmental Malczewski, J., 1999. GIS and Multicriteria Decision Analysis. John
evaluation for urban land-use planning: a case study. Engineering Willey and Sons Inc. 392 pp.
Geology 61, 257–271. Özdemir, M.S., 2005. Validity and inconsistency in the Analytic
Densham, P.J., 1991. Spatial decision support systems. In: Maguire, D.J., Hierarchy Process. Applied Mathematics and Computation 161,
Goodchild, M.S., Rhind, D.W. (Eds.), Geographical Information 707–720.
Systems: Principles and Applications. Longman, pp. 403–412. Saaty, T.L., 1990. How to make a decision: the Analytic Hierarchy
Diamond, J.T., Wright, J.R., 1988. Design of an integrated spatial Process. European Journal of Operational Research 48, 9–26.
information system for multiobjective land use planning. Envi- Saaty, T.L., 2006. Rank from comparisons and from ratings in the
ronment and Planning. B 15 (2), 205–214. analytic hierarchy/network processes. European Journal of Oper-
Eastman, J.R., Jin, W., Kyem, P.A.K., Toledano, J., 1995. Raster ational Research 168 (2), 557–570.
procedures for multicriteria/multiobjective decisions. Photogram- Saaty, T.L., Özdemir, M., 2003. Negative priorities in the Analytic
metric Engineering and Remote Sensing 61 (5), 539–547. Hierarchy Process. Mathematical and Computer Modelling 37,
Florinsky, I.V., Kuryakova, G.A., 2000. Determination of grid size for 1063–1075.
digital terrain modelling in landscape investigations — exempli- Seed, H.B., De Alba, P., 1986. Use of SPT and CPT Tests for Evaluating
fied by soil moisture at a micro-scale. International Journal of the Liquefaction Resistance of Soils. Proceedings, Institu. ASCE.
Geographical Information Science 14 (8), 815–832. Şener, B., Süzen, M.L., Doyuran, V., 2006. Landfill site selection by
GDDA, 1996. Earthquake zoning map of Turkey. General Directorate using geographic information systems. Environmental Geology
of Disaster Affairs. Ministry of Reconstruction and Resettlement of 49, 376–388.
Turkey (in Turkish). Starr, M.K., Stein, I., 1976. The Practice of Management Science.
GDDA, 2000. Laws and Regulations: Regulations for the Construction Prentice-Hall Inc., Englewood Cliffs, New Jersey. 231 pp.
of Buildings in Hazard Areas, Ankara, pp. 244–332 (in Turkish). Süzen, M. L., 2002, Data Driven Landslide Hazard Assessment using
Guidelines for Urban Engineering Geological Investigations, South Geographical Information Systems and Remote Sensing, M.E.T.U.
African Institute of Engineering Geologists. Ph. D. Thesis, 196 p. http://www.metu.edu.tr/~suzen.
Jankowski, P., 1995. Integrating Geographical Information Systems Süzen, M.L., Doyuran, V., 2004a. Data driven bivariate landslide
and multiple criteria decision making methods. International susceptibility assessment using Geographical Information Sys-
Journal of Geographical Information Systems 9 (3), 251–273. tems: a method and application to Asarsuyu catchment, Turkey.
Janssen, R., 1992. Multiobjective Decision Support for Environmental Engineering Geology 71/3–4, 303–321.
Management. Kluwer Academic Publishers. 232 pp. Süzen, M.L., Doyuran, V., 2004b. A comparison of the GIS based
Keller, C. P., 1996, Decision Making using Multiple Criteria, NCGIA landslide susceptibility assessment methods: multivariate versus
Core Curriculum, Unit 57, Santa Barbara, CA: National Center for bivariate. Environmental Geology 45–5, 665–679.
Geographic Information and Analysis, URL: http://www.geog. USGS, 2004. Standards for Digital Elevation Models, U.S. Geological
ubc.ca/courses/klink/gis.nodes/ncgia/u57.html#UNIT57. Survey http://rmmcweb.cr.usgs.gov/public/nmpstds/acrodocs/
Koyuncu, P., 2001, Assessment of Geo-Engineering Properties of the dem/2DEM0198.PDF.
Geological Units in the City Center of Eskişehir and Preparation of Worboys, M.F., 1995. GIS — A Computing Perspective. Taylor and
Engineering Geological Map, M. Sc. Thesis, Hacettepe University Francis.
(in Turkish).

Please cite this article as: Çağıl Kolat et al., Preparation of a geotechnical microzonation model using Geographical Information Systems based
on Multicriteria Decision Analysis, Engineering Geology (2006), doi:10.1016/j.enggeo.2006.07.005.

You might also like