5th Annual International Workshop & Expo on Sumatra Tsunami Disaster & Recovery 2010
Landslide hazard assessment
using analytic hierarchy processing (AHP) and
geographic information system
in Kaligesing mountain area of Central Java
Province Indonesia
1
Syamsul Bachri and 2Rajendra P Shresta
1
Departement of Geography, Faculty of Social Science, State University of Malang
Jalan Semarang 5, Malang 65145, Indonesia
2
School of Environment, Resources and Development, Asian Institute of Technology
58 Moo 9, Km. 42, Paholyothin Highway, Klong Luang, Pathumthani 12120 Thailand
Email: 1syam_geoum@yahoo.com, 2rajendra@ait.ac.th
Abstract - Landslide is type of natural disaster which may meteorological disaster. Several studies showed that
cause huge losses of live and properties. Many landslides disasters are likely to occur due to their environmental,
triggering factors found in Kaligesing make this zone as climatic and geographical condition. In particular, landslide
landslide prone area in Indonesia. Zoning hazard area was a will occur in more intensive scale in countries which has
solution to assess landslide disaster since there is still a great
mountain geographic characteristic area.
danger of further landslides in the region and also it is strongly
linked with spatial issues. The combination of GIS and Landslide is a natural disaster which may cause huge
Analytic Hierarchy Process (AHP) are used to create landslide losses of lives and property. The occurrence of landslide is
hazard zone in this study. Factors, such as landform, land influenced by many factors, naturally occurring or by
utilization, slope steepness, soil texture and lithology are human activities. Natural landslide may occur by collapses.
considered for use in AHP through pair-wise matrix. The It is triggered by physical condition such as topography,
output of calculation was validated with present landslide climate, vegetation, land use and earthquake. It is also
location. Based on the judgment matrix and calculation, the commonly occurred due to over human exploitation, such as
result showed ë max = 5.406256, the feature vector of
， ， ，
normalization: F= 0.4042 0.2746 0.2018 0.0845 0.0349). ， forest logging, over grazing, etc. Moreover, landslide often
occurs in mountainous area which has low stability slope.
In this calculation, RI =1.12. According to relational formula,
Many landslide triggering factors are found in Indonesia
CI=0.101564. A consistency ratio (CR) was computed to verify
that the consistency of matrix. CR value of 0.090682, meant that makes Indonesia as the second highest number of
that the pair-wise matrix is consistent (threshold CR<0.10) and landslide in the world [3]. According to [4] there are 20
can be used for assigning the criteria weight. Spatial regions in Daerah Istimewa Yogyakarta (DIY) province -
distribution of the susceptibility classes of landslide in Central Java province and 29 regions in East Java province
Kaligesing showed that more than 40% of the study area was which are categorized as susceptible to landslide hazards.
categorized as landslide prone area with moderate All of the areas were witnessed mass movement and rock
susceptibility and 30.05% falling on high susceptibility class, movement. As it is mentioned by [5], most of the hilly areas
while the rest 20.78% was categorized as less susceptibility in Java are susceptible to landslide. The landslides in Java
class.
Island increase over time [6]. The occurrences of landslide
Key word: Landslide, AHP, GIS, Kaligesing, Java.
during 1990 and 2005 caused the death of 1,000 people. The
highest loss was incurred in 2005 that victimized 118
I. INTRODUCTION individuals. Furthermore, data showed by [7] the extreme
event was in January 2006 causing death of 142 people and
Currently around 70 percent the world’s population lives damaging 182 houses; and landslide in September 2000 in
are affected by natural disaster, such as earthquake, flood Purworejo caused death of 44 people and 20 people were
and landslide [1]. During 1991-2005, EMDAT reported injured as well as 77 houses were damaged [8]
total amount of economic lost due to natural disaster as Kaligesing, sub-district in Central Java Province,
379.15 US $ billion in developing countries [2]. Comparing Indonesia, is known as one of the landslide prone area [9].
with other part of continent, Asia is the world’s most Most part area of Kaligesing sub-district is upland area
disaster prone region. Geological disasters like landslide which varies in lithology, geomorphology and hydrology.
hold second position in number of occurrences after Hydro-
107
5th Annual International Workshop & Expo on Sumat
atra Tsunami Disaster & Recovery 2010
Due to these factors, this area is susc
usceptible to landslide constitute of andesit, dasit,
t, conglomerate, breksi andesit,
hazards. More than 10 occasion of land ndslide occurs in 2005 gravel and sand. In addition,, a mechanic weathering process
[10]. Generally, three cases of landslid
lide occur in each year
[11]. The degree of damage was differe erent at different areas
on houses and road network. The aver verage indemnification
was about 1,000,000-100,000,000 IDR R ffor each damage.
Management of land resources is very ve important due to
the susceptibility of this area. In order
er to curb and reduce
the impacts of landslide also to be more effective to
exploitation the land, landslide hazardd zoning
z can be one of
the solutions. Combination of GIS S and AHP are the
effective method for hazard assessmen ent, GIS has powerful
for spatial analysis while AHP has ce certain advantages in
multi-index integrated evaluation.
II. METHODOLOGY
GY Fig. 1. Locatio
tion of study area
A. Description of study area in this region is occurring fre
frequently due to such formation
The research area is located in Kal aligesing sub-districts condition as well as effectivee rainfall and steep slope.
Purworejo Regency, Central Java Province Indonesia Soil and slope map in this
his study area were presented in
(Fig.1). It lies between South Latitudede 7º 50’ 34” – 7º 51’ Fig. 2 and 3. Soil type in study area was formed and
45” and East Longitude 110º 07’ 46 46” – 110º 08’ 20”. controlled by geomorphic pprocess. Soil type’s based on
Kaligesing sub-district has 21 village ages, namely: Village USDA classification such ass vvertisol, inceptisol and enstisol
Somongari, Jati Rejo, Dono Rejo, Hulosobo,
Hu Kali Harjo, are found within Kaligesing ng sub district. Most of upper
Tlogoguwo, Kali Gono, Jelok, Ked edung Gubah, Purbo areas in Kaligesing developp by vertisol soil types which
Wono, Pandan Rejo, Ngaran, Tawangg Sari, S Gunung Wangi, have high content of expansivsive clays and active erosion and
Tlogo Rejo, Sudogoro, Tlogo Bu Bulu, Hardi Mulyo, mass movement processes.. This T area covers as a part of
Sumowono, Pucung Roto, and Ngardi di Rejo. The total area Kaligono village, Hulusob sobo, Ngaran, Sudorgo and
of the study area is approximately 7,472.89
7,4 ha with total Tlogoguwo. Furthermore, entisolse was found in the
population 35,895 citizens. Kaligesingg sub-district
s is located deposition and alluvium plain in which lies in part of Kaligono
between Purworejo City and Yogyakar karta Province. Due to and Kaliharjo village. Inceptis
tisols was placed in the hill foot
improvement of roads network, the dev evelopment process in slope which usually disturbeded by sedimentation and erosion
this area has increased particularly rly along the roads. process. In great group based
ed on FAO classification we also
Kaligesing sub district is an importan tant region for socio- found detailed soil types in this study area such as lithic
economic development particularly fruit production in hapludalfs, typic eutrudepts ts, typic hapludalfs, oxyaquic
Purworejo district Central Java province p Indonesia. eutrudepts, lithc udhorthentents, lithic udhorthents, typic
Unfortunately, Kaligesing is a landsli slide prone area. This hapludalfs, haplic udarent, nt, lithic calciustepts, haplic
study area represents characteristic feature
f of most of ustarents, vertic hapludalf alfs, typic eutrodepts. The
upland area in Central Java, Indonesia. domination of texture was occ ccupied by clay texture.
The climate of the study area is mostly sub-tropical
condition. Based on Oldeman classifica ication, Kaligesing sub
district is categorized as C3 class clim
limate. It has minimal
rainfall intensity, 200 mm in wet monthnth and 100 mm in dry
month. The average amount of rainfall ll in
i study area is about
2000 to 5000 mm in a year. Based B on Oldeman
classification, wet condition prevails in Kaligesing for 5–6
months and there is dry condition forr 2-42 months in a year.
So, 5-6 months are prone for occurrenc nce of landslides. This
condition represent of high perception dduring the year.
Based on geology map sheet Yogyak akarta scale 1:100,000,
the study area lies in three major ro rock formations: Old
Andesit Van bemmelen formation, Allu lluvium formation and
Jonggrangan formation. It was domina inated by Old andesit
Van bemmelen formation which forme med during Oligocene
up to first Miocene. This formatio tion is composed of
andesitic, andicitist lava flow and tuff.
f. The
T rock material are
Fig. 2. Soil map
ma of study area
108
5th Annual International Workshop & Expo on Sumat
atra Tsunami Disaster & Recovery 2010
processing (AHP) weighted d is tools that can be use in
processing data landslide relat
lated factors. The AHP system is
worked for ranking in a sett of o alternatives. First of steps in
AHP is create the AHP hierar rarchy. Then the second steep is
use Pair Wise to compare each ach of factors. Third procedure is
conducted all the priority intoin degree of susceptibility of
landslide.
Furthermore, analytical hie ierarchy model takes as an input
the pair wise comparisons and nd produces the relative weights
as output. The procedure consists
co of three major steps:
generation of the pair wise comparison
co matrix, computation
of the criterion weights, andnd estimation of the consistency
ratio. Following is the detailailed step of mapping landslide
hazard based on [15] analysi ysis hierarchy processing (AHP)
procedure:
Fig. 3. Slope map off study
s area 1. Generation of the pair wisese comparison matrix:
The method employs scal cale with values from 1 to 9 to
The elevation within Kaligesing subub district varies from rate the relative preference
ces of two criteria based on [16]
250 m to 800 m [12]. It is considered as hilly area. Based on (Table 2).
slope map which construct from conto ntour map, 7.42 Ha of 2. Computation of the criterioion weights:
area has falls on level slope (3-8%), of 10.83 Ha areas lies This step involves the follo
llowing operations:
on level slope 8-15 % slope class, 185.
85.65 Ha falls on level a. Sum of the values inn eeach column of the pair wise
slope (15-30%) and 368.31 Ha falls ls on level slope (30- comparison matrix.
45%). Similarly, 134.99 Ha falls on level
le slope (45-65%). b. Divide each element in the matrix by its column total
This study area was dominated by mounountain areas which lie (the resulting matrix is referred to as the normalized
in more than 15 % of slope level. pair wise comparison matrix)
m
c. Compute the averagee of o the elements in each row of
B. Methods of mapping landslide hazard ard the normalized matrix rix, that is, divide the sum of
There are many landslide ma mapping approach has normalized scores for or each row by the number of
been used in several study, including on-ground
on monitoring,
criteria. These average ges provide an estimate of the
remote sensing data, geomorphologic gic approach, factors
relative weights of thehe criteria being compared. The
overlay, statistic models, and geotech chnical process model
higher the weight is the more important the criteria.
[13]. Factors overlay method were thee ttechnique which this 3. Estimation of the consistenc
ency ratio:
study uses. Landslide location and land ndslide related factors This step is to determine ne whether the comparisons are
such as slope, soil texture, lithology, landform
la and land use consistent. It involves thee ffollowing operations:
were used for analyzing landslide ide susceptibility. A
a. Determine the weighte hted sum vector by multiplying
probability method was used for calculaulating the rating of the
the original pair wisese comparison matrix to sum of
relative importance of each factorr class to landslide
normalized scores forr each
e row matrix.
occurrence. In this study, the score off each factor can be
b. Determine the consist istency vector by dividing the
dispensed as the same or different valualue depends on expert weighted sum vector by the sum of normalized scores
judgment. The degree of impotance ce of value in each
for each row matrix.
parameter adapts from [14] (Table I).
c. Calculate lambda ( ), consistency index (CI), and
For calculating the weight of the relative
re importance of
Consistency Ratio (CR R)
each factor to landslide occurrencee is using analytical
hierarchy processes (AHP). The analysis hierarchy
a) = average value
ue of consistency vector
TABLE I λ−n
DEGREE OF IMPORTANCE
CE
b) = ,n is the number of criteria.
Intensity of Importance Nuumerical scale n −1
Equal importance 1 CI
Weak importance of one over another 3 c) CR= , RII is the random index, the
Essential or strong importance 5 RI
Demonstrated importance 7 consistency index
ex of a randomly generated pair
Absolute importance 9 wise comparisonn matrix. The RI depends on the
Intermediate values between two 2,, 4,
4 6, 8 number of elemeents being compared that taken
adjacent judgments
If activity I has one of the above Rec
eciprocal of above number from [16] (Table
le 3).
numbers assigned to it when compared
with activity j, then j has the reciprocal
value when compared with i
109
5th Annual International Workshop & Expo on Sumatra Tsunami Disaster & Recovery 2010
TABLE 2
PAIR WISE METHOD TABLE 4
RANKING FACTORS USED TO EACH PARAMETER IN HAZARD ASSESSMENT
A1 A2 A3 … An Criteria Sub criteria Factor (F)
A1 W1/W1 W1/W2 W1/W3 … W1/Wn Slope 3-15% 1
A2 W2/W1 W2/W2 W2/W3 … W2/Wn 15-30% 2
A3 W3/W1 W3/W2 W3/W3 … W3/Wn >30 % 3
. . . . . Soil texture Loamy sandy, sandy loam 1
. . . . . Sandy clay loam 2
. . . . . Loam, clay loam, silty clay, silty clay 3
loam
TABLE 3
RANDOM INDEX (RI)
Land use Rain fed paddy field, rice paddy Field, 1
n 1 2 3 4 5 6 7 8 grass land
Shrubs, perennial crop 2
RI 0.00 0.00 0.58 0.90 1.12 1.24 1.32 1.41
Mix perennial crop, settlement 3
n 9 10 11 12 13 14 15
Rock Andesit 1
RI 1.45 1.49 1.51 1.48 1.56 1.57 1.59
material Breksi andesit, dasit 2
Limestone, sediment breksi andesitic 3
Landform Alluvial plain 1
III. RESULTS AND DISCUSSIONS Colluviums-alluvial, foot slope, foot 2
slope of structural hills , foot slope of
denudation hills
Due to of several impacts of landslide, this study develop
Structural hills, denudation hills, karts 3
landslide hazard map for learning about potential local
Source: Adopted from Hadmoko, 2009
landslide hazard and taking step to reduce those hazard.
The basic assumption which used in this study is the geo-
morphological concept that said “the past and present are develop own rainfall station by researcher to get appropriate
the keys to the future [17]. This principle can simply explain data. Further, main idea for composing landslide hazard
through basic concept of landslide. Landslide is typically mapping is to reduce hazard its self. Thus, this research tried
occurred periodically lies on specific physical condition in to considered most factor that can be managed or control by
certain area like geologic, slope and soil condition which human being. According to degree of importance slope was
categorizes as indelible factors. Landslide that occurred in the most importance factor comparing with landform,
the past can be occurring in the future under similar lithology, land use and soil texture. Table 5, 6 and 7 from
condition. Thus, take similar factor as consideration analysis are revealed the processes of AHP justification.
composing landslide hazard susceptibility is an idea in this Based on the judgment matrix and to calculate, max =
5.406256, the feature vector of normalization: F=
， ， ， ，
research. Further, to achieve appropriate landslide hazard
map, this research also combining to other factor like land (0.4042 0.2746 0.2018 0.0845 0.0349). In this
use. Factor overlay method and analytic hierarchy calculation, RI = 1.12. According to relational formula,
processing (AHP) were used to develop landslide CI=0.101564. A consistency ratio (CR) was computed to
susceptibility map. Slope class, soil texture, land use, verify that the matrix is consistent. CR value is 0.090682,
lithology, and landform are factor that considered meaning that the pair-wise matrix is consistent (threshold
composing the map. Reference [18] conducted a study CR<0.10) and can be used for assigning the criteria weight.
based on a research report from [19] and found five classes
of landslide hazard. This research modified the finding and TABLE 5
ORIGINAL COMPARISON MATRIX
divided land slide hazard parameters of the respective study Soil
area into three classes based on the findings of [18]; high Slope Landform Lithology Land use Texture
susceptible, moderate susceptible and less susceptible. The Slope 1 2 3 5 7
selected factors and its score are revealed in Table 4. Landform ½ 1 2 5 7
AHP was used to determine degree of important for each
Lithology 1/3 ½ 1 5 7
factor related to the landslide occurrence. The relative
importance of criteria among one and another was assigned Land use 1/5 1/5 1/5 1 5
Soil
using pair wise method. The pair wise comparison for each Texture 1/7 1/7 1/7 1/5 1
Variable was justified in discussion with an expert and data
from local people perception as well as site visit. As result
of discussion with expert, rainfall factor that mentioned by
local people as factor causing landslide disaster in study
area cannot used to be considering factor due to rainfall data
coverage area. The coverage area is too small and the
pattern of rainfall is homogenous. In addition, in case of
Indonesia, the lowest level of rainfall station is placed
insub-district level. Hence, it should be improve with
110
5th Annual International Workshop & Expo on Sumat
atra Tsunami Disaster & Recovery 2010
TABLE 6
NORMALIZED COMPARISON MATRIX
M
Soil TAB
ABLE 8
SUSCEPTIBIL
ILITY CLASSES
Slope Landform Lithology
gy Land use Texture
No. Interval value Susc
sceptibility class Frequency of
Slope 0.459 0.520 0.472
72 0.30 0.259 landslide point
Landform 0.229 0.260 0.315
15 0.308 0.259 1 1-1.588 Les
ess susceptible 2
2 1.589-2.175 Mode
derate susceptible 8
Lithology 0.153 0.130 0.157
57 0.308 0.259 3 2.176-2.763 Hig
igh susceptible 16
Land use 0.091 0.052 0.031
31 0.061 0.185
Soil Texture 0.065 0.0371 0.022
22 0.0123 0.037
TABLE 7
RELATIVE WEIGHT OF CRITE
TERIA
Sum Weight
W
Slope 2.020 0.4042
Landform 1.373 0.2746
Lithology 1.008 0.2018
Land use 0.422 0.0845
Soil Texture 0.174 0.0349
According to the above factorss evaluation and its
weighted, the formula for landslide hazard
ha index is given
below:
5
LHI = W 1.F1 Fig. 4. Map of landslide hazard
1
In the formula: LHI— landslide hazard
rd index; W1-weight of TAB
ABLE 9
each index; F1—factor of each index.
x. Below calculation is DISTRIBUTION OF LAN
NDSLIDE HAZARD AREA
Landslide hazard Area (Ha) Percentage
representing the above formula: High susceptible 2245.60 30.05
Moderate susceptible 3672.92 49.15
Landslide hazard index (LHI) = 0.404
042*Slope + 0.2746*L Less susceptible 1552.86 20.78
+ 0.2018*LG + 0.0845*LU + 0.0349*S*ST Total 7472.89 100
30.05% falls on high susce ceptibility class, while the rest
Where landslide hazard index is the total susceptibility 20.78% is categorized as less le susceptibility class. Less
score, while the factors (F) are respec
pectively susceptibility susceptibility zone was characteristic
ch with less mass
score for Slope: score of slope; L: scocore of landform; LG: movement processes and d located on 3–8% slope
score of litology; LU: score of land use;
us and ST: score of obliqueness. This area lies on alluvial plain landform. While
soil texture. This equation is appliedied in map calculator the moderate landslide suscep
ceptibility areas was found to be
function of the Spatial Analysis extension
ext tool in the characterized by moderately ely steep to steep slope and
ArcView. Furthermore, in order to exc xcuse and validate the showed incidences of histo istorical mass movement. The
final landslide hazard map, final overlay
lay processing between regions of high landslide susc
usceptibility are the areas where
landslide hazard map and landslide distribution
d map was steep slope (>30%) were locatedl and mass movement
released. occurs frequently. Both old mass
m movement and new mass
After analyzing and calculated datada related landslide movement were observed too be b still active due to high rain
occurrences in this study area, we camame to the result. The intensity and topography condndition.
landslide hazard index value then was as classified into three
different classes to identify the susce
sceptibility level. The IV. CONC
NCLUSIONS
result of class of hazard based onn the maximum and
minimum value of the total score (Table (T 8). Following In summary, mapping landslide
la hazard through AHP
formula is applied to the value. analysis showed that most of the area Kaligesing is prone to
Spatial distribution of the suscep ceptibility classes of landslide hazards. In this study,
s landslide hazard were
landslide in Kaligesing is showed in Fig.
Fi 4. Table 9 shows divided into three category;; high
h susceptible, moderate and
more than 40 % of the study areaa was categorized as less susceptible. The proporti
rtion of the area mostly falls in
landslide prone area with moderate susceptibility
su class and moderate susceptible landsl dslide hazard with 49.15 %.
Effective rainfall, physiograph
aphic condition (slope, lithology,
and landform) and improperr land
l utilization were the causes
of landslide occurrences. Even
ven the technical already develop
111
5th Annual International Workshop & Expo on Sumatra Tsunami Disaster & Recovery 2010
landslide hazard zone, it is not enough to reduce the risk. Gadjah Mada University, Yogyakarta: Gadjah Mada
Collaboration management on landslide risk reduction University, 2005
between regions, departments concerned, universities, [11] Office of National Unity and Community Protection.
research centers, non-governmental organizations and local 2009,
peoples in landslide-prone play important role to better risk [12] Department of Forestry, 2007
management. [13] Zink, J.A., Lopez, J., Metternicht, G., Shrestha, D.P.,
Selem, L.V., ” Mapping and modeling mass movement
ACKNOWLEDGMENT and gullies in mountains areas using remote sensing
The authors thankful to DIKTI scholarship, who supported and GIS technique, Journal of Applied Geology, vol
this research activity through master scholarship, as well as 3(1), 2001
Asian Institute of Technology, State University of Malang [14] Saaty, T.L., Alexander, J.M, Thinking with Models:
and Gadjah Mada University for all services during research Mathematical Models in the Physical, Biological and
activity. Social Sciences, Pergamon Press, London, 1981
[15] Saaty, How to make decision: The Analytical
REFERENCES Hierarchy Process. European Journal of Operational
[1] UNDP, “Vulnerability and risk assessment,” New Research 48, 9-26, 1990
York, United Nation Development. 2nd ed, 2004 [16] Tripathi, N.K., Spatial analysis method in GIS
[2] ISRD, “Disaster impact reported. Asia, International (Lecturer notes, Course No AT76.9013, School of
Strategy for Disaster Reduction,” Retrieved April 1 Engineering and Technology). Bangkok: Asian
2010, from http://www.unisdr.org/disaster- Institute of Technology, 2009
statistics/impact-economic.htm, 2005 [17] Huabin, W., Gangjun, W., Weiya, X., Gonghui, W.,
[3] ILC (International Landslide Center), University of “GIS-based Landslide Hazard Assessment: an
Durham. Retrieved March 29 2009. from Overview,” Progress in Physical Geography 29, Vol
http://www.landslidecenter.org/database.htm, 2004 4, pp 548-567, 2005
[4] Korita, “49 Daerah di Pulau Jawa Rawan Longsor, [18] Hadmoko, D.S., Lavigne, F., Sartohadi, S., Hadi, P.,
Yogyakarta, Indonesia,” Retrieved July 8 2009, from Winaryo, “Landslide hazard and risk assessment and
http://www.ugm.ac.id/index.php?page=rilis&artikel=1 their application in risk management and land use
109, 2009 planning in eastern flank of Menoreh Mountains,
[5] Marfai, M.A and Widiyanto, W, ”Landslide hazard Yogyakarta Province, Indonesia,’’ Natural Hazard,
assessment and mitigation,” Proceeding research year DOI 10.1007/s11069-009-9490-0, 2009
book. Geography Faculty Gadjah Mada University. [19] PSBA-UGM, Research Center for Disaster, 2001
ISBN 979-8786-19-x, pp 36-42, 17 April 2002 (in
Indonesian)
[6] Hadmoko, D.S., “Toward GIS-based integrated
landslide hazard assessment: a critical overview,”
Indonesian Journal of Geography, vol. 34 (1), pp 55-
77, 2007
[7] DGHM, “Landslides database in Java Indonesia,”
Unpublished
[8] Marfai, M.A., King, L., Singh, P.L., Mardiatno, D.,
Sartohadi, J., Hadmoko, D.S., Dewi, A., “Natural
hazards in Central Java Province, Indonesia: an
overview,” Environmental Geology, DOI
10.1007/s00254-007-1169-9, 2007
[9] KR., “Longsor di kecamatan kaligesing Purworejo,”
Kedaulatan rakyat, Purworejo.
RetrievedMarch302009,from
http://www.kr.co.id/web/detail.php?sid=147425&actm
enu=35, 2007
[10] Bachri, S., “Kajian kemampuan lahan dan distribusi
titik longsor untuk penentuan prioritas penanganan
bencana longsor di kecamatan Kaligesing propinsi
Jawa Tengah Indonesia,” Bachelor thesis study,
112
syamsul bachri