investigating the effect of conservation …

JURNAL GEOGRAFI VOLUME 14 NO. 2

JURNAL GEOGRAFIMedia Pengembangan Ilmu dan

Profesi Kegeografianhttp://journal.unnes.ac.id/sju/index.php/ujet

INVESTIGATING THE EFFECT OF CONSERVATION TECHNIQUES ONTHE LAND DEGRADATION OF TROPICAL CATCHMENT PRONE TOLANDSLIDE

Nugroho, Christanto1; Junun, Sartohadi1; M Anggri, Setiawan1; M Pramono, Hadi1; Victor G,Jette2; Druba Phika, Shresth2;1. Departement of Environmental Geography, Faculty of Geography, Gadjah Mada University2. Faculty of Geo-Information Science and Earth Observation (ITC), Twente University, The

Netherlands

Info Artikel________________Keywords: ConservationTechnique, landdegradation, Modelliing,Tropical Catchment, Dieng

Abstract___________________________________________________________________Land degradation in Serayu watershed is a major concern in central Java andin Indonesia. As part of a broader effort to develop a land degradationassessment tool in tropical area, this study implemented a process-basedwatershed hydrology to assess the effect of conservation technique upon landdegradation by using PCRaster. STARWARS was used to assess the watershedhydrology in the area based on their land use/ land cover, soil, and slopeprofiles. The results from STARWARS were used as inputs for thePROBSTAB model to simulate the slope stability in the area. DEM scenariowere used, they are with terraces and without terraces.The models show that the landuse practice in the study area work like twoedges of sword. The promoting of bench terrace can be reducing the risk of soilerosion but in the other hands it increases in the risk of landslide. From theslope-stability modeling, we can see that the terrace increases the pore-waterpressure significantly which lead to the ideal conditions for the failures. Theextremely high intensity rainfall, in the other hands, may build a sharp increaseof pore-water pressure. The increasing probability of failure might cause thesoil erosion even worse. Therefore, in order to make the terrace practice iseffective to control the land degradation process; the terrace has to be wellmaintained.

AbstrakKata kunci: teknikkonservasi lahan,degradasi lahan,pemodelan, DAS Tropis,Dieng

Degradasi lahan di DAS Serayu sangat penting untuk dikaji. Sebagai bagiandari upaya pemahaman proses degradasi lahan di wilayah tropis, studi inibertujuan untuk mengetahui pengaruh teknik konservasi terhadap degradasilahan dengan menggunakan software PCRaster. Model STARWARSdigunakan untuk memodelkan kondisi hidrologi di wilayah penelitian. Hasildari model ini akan digunakan untuk input model PROBSTAB yang akanmemodelkan stabilitas lereng di wilayah penelitian. Skenario penggunaanDEM dalam model ini juga akan di gunakan, yakni DEM denganpemanfaatan lahan teras dan tanpa teras.Hasil pemodelan menunjukkan bahwa pengguaan lahan dengan teras sepertipedang bermata dua. Pemanfaatan teras dapat menurunkan risiko erosi,namun di sisi yang lain dapat meningkatkan risiko terjadinya tanah longsor.Dengan intensitas hujan yang sangat tinggi, risiko terjadinya longsor menjadisemakin tinggi. Apabila lereng mengalami longsor, maka erosi yang terjadiakan lebih parah daripada erosi pada lereng normal. Sehingga pengelolaanteras sangat penting dalam mengurangi tingkat degradasi lahan di wilayahDieng.

Alamat korespondensi: [email protected]


1. INTRODUCTIONJava, the fifth largest island in

Indonesia experiences severe landdegradation. Over 62% of the Indonesiapopulation lives in this island (~ 900people/km2). The history of landdegradation in the island is closely linked tothe rapid land use/land cover changes thatthe region experienced since 1700’s whenthe colonizers began to exploit the richtimber wealth (Lavigne and Gunnell, 2006).A major land degradation process in theregion is shallow landslides. Landslides inIndonesia are triggered by rainfall andconsequent increase in pore-water pressure(Marfai et al., 2008) or earthquakes.According to the database of theInternational Landslide Centre, Indonesia isthe second most landslide affected countryin the world after China (ILC, 2006).Between 2000 and 2007, Java experienced57 landslides causing 249 human fatalities(DGHM, 2007). The fatalities and damagesdue to landslides are only to increase giventhe current trend of migration to theuplands, consequent rapid land use/landcover changes and continuing deforestation(Hadmoko et al., 2009).

Java’s one of the most landslide proneareas is the Dieng Mountains which isfrequently reported in the press due to thealmost annual occurrence of the monsoonrelated hazards. A recent example of thedisastrous effect of landslides in the region isthe 9 November 2010 landslide in Sitiengwhich caused human fatalities and damagedhouses. Several researchers have attemptedto assess the landslide susceptibility of theregion (eg., (DGHM, 2007; Hadmoko andLavigne, 2007; Marhaento, 2006) usingstatistical and heuristic methods, accordingto which the highest landslide susceptibilityin the region is to areas with steep slopes(>25) along the escarpments. However,susceptibility only implies the spatialprobability of landslide occurrence (vanWesten et al., 2005).

Recently, Hadmoko et al., (2009)illustrated a method to derive both spatialand temporal probabilities (rainfallthresholds) for Indonesia with detailedrainfall and corresponding landslideinventory data. Applicability of this methodis limited in most landslide prone regions ofthe country owing to the lack of suchhistorical data. Given these limitations oflandslide hazard assessment using statisticaland heuristic methods a pragmaticalternative for landslide hazardquantification (spatio-temporal probabilities)in the region is the application of physically-based shallow landslide initiation models.Initiation of shallow landslides like debrisflows is a consequence of perchedgroundwater tables or successive wettingfronts. Thus a model that attempts tocapture the process must seamlesslyintegrate the generation of perchedgroundwater tables as a result of rainfall andthe consequent failure. Takara et al., (2008)conducted sediment-runoff modeling using aphysically-based dynamic model in theregion and illustrated the use of such modelsfor the early warning of debris flows.However, their model does not explicitlyinclude pore-water pressure fluxes (the maintrigger of shallow landslides in the region)thereby limiting its applicability for bothlandslide hazard quantification andlandslide early warning.

Thus there is an apparent lacuna of amodel that can quantify the spatial andtemporal probability of landslides in theregion. Such a model necessarily has tofunction in situations lacking historicallandslide and rainfall data and should becapable of simulating the main trigger, pore-water pressure fluxes, as a consequence ofrainfall input. Studies elsewhere have shownthat physically-based dynamic models thatcouple slope hydrology and slope stabilityare suitable to evaluate shallow landslideinitiation conditions in regions lackinghistorical data (eg., (Simoni et al., 2008; van


Beek and van Asch, 2004)). A set of modelsthat has been applied and proven to beuseful in regions lacking historical data(Kuriakose et al., 2009b) are theSTARWARS (Storage and Redistribution ofWater on Agricultural and RevegetatedSlopes) coupled with PROBSTAB(Probability of Stability) (van Beek, 2002).In this research these models were appliedto Serayu sub-catchment in the DiengMountains and validated against anavailable landslide inventory.

In the other hands, Dieng is one of themost productive areas in Indonesia, in termof agriculture.In order to support theagricultural activity, terracing was promotedas the most common land managementpractice in this area. In the sense of farming,there are several advantages of this landmanagement. By building terraces, thefarmer will be able to grow crops in such asteep land, generate more lands and reducethe erosion.

Due to its rough terrain and highintensity of rainfall, constructing terrace inDieng might lead to other land degradationproblem. The probability of infiltration willbe increased as the slope reduces locally bythe bench terrace. The objective of this studyis to model the effect of terraces on the land

degradation processes and assess howeffective is the terrace practice in the studyarea.1.1 Study Area

The study area is the Kali Putihcatchment of Sitieng Sub-District,Wonosobo, Indonesia covering an area of 1km2 (Figure 1). The sub-catchment islocated in the upper part of Serayuwatershed. The highest and lowest points inthe study area are 2375 and 1675 m a.m.s.l,respectively.

There is significant evidence from theregion that shallow landslides initiate in thewet season immediately after the clearing ofnatural forests (Lavigne and Gunnell, 2006).The current climate change scenario forJava proposes an increased influence of ElNino effects. Though the total rainfall willreduce owing to increased periods ofmeteorological drought, the intensity ofrainfall is expected to increase (Hulme andSheard, 1999). This increase will invariablyreflect in the number of landslides occurringin the region. Through this research weattempt to understand the effect of landmanagement practice and its relation to thehydrological and mechanical response ofslopes to the current rainfall pattern of theregion.

Figure 1. Study Area, Tieng, Wonosobo


2. METHODS2.1 The Models

The models used for the studycomprise a distributed dynamic hydrologicalmodel (STARWARS) that is coupled to astability model (PROBSTAB); both weredeveloped by van Beek (2002). The dynamicspatial outputs of the hydrological model arethe inputs for the slope stability model. Itwill run on daily basis. An added advantageof the models is its open architectureallowing parameterizations appropriate forthe study area. Both the models areembedded in PCRaster®, a GIS with anadvanced Environmental ModellingLanguage (www.pcraster.nl). The modelschema is illustrated in Figure 2.

STARWARSSTARWARS is an acronym for STorageAnd Redistribution of Water onAgricultural and Re-vegetated Slopes. Themodel simulates the spatial and temporaldynamics of moisture content and perchedwater levels in response to gross rainfall (P)and evapotranspiration.

PROBSTABPROBSTAB is an acronym for PROBabilityof STABility. The slope stability model isbased on the infinite slope form of theMohr-Coulomb failure law as expressed bythe ratio of stabilizing forces (shear strength)to destabilizing forces (shear stress) on afailure plane parallel to the ground surface.The model is modified to account for root-induced cohesion and surcharge that werenot originally considered by van Beek . Theequation used is:= ( ) {( . . ) } . ∅[( . ) ].where, Cs and Cr are the soil and rootinduced cohesion respectively (kPa), s unitweight of soil, w unit of weight of water, vsurcharge due to weight of vegetation (kPa),Zs soil thickness (m), Zw water height (m), slope angle and angle of internal friction.The unit weight of soil is the sum of the unitweight of saturated soil and unsaturated soilrespectively.

Figure 2. Methodology Flow-chart


2.2 The parametersA detailed list of inputs necessary for

the models is provided in van Beek (2002)and Kuriakose et al., (2009b). Individualparameters were prepared as detailed below.

Climatic parametersDaily rainfall data was available for the areafrom 1989 to 2009 (daily basis) and 2010(detailed data). From the analysis of theaverage monthly rainfall it is evident the dryseason in the region starts in May and endsin September and the wet season starts inOctober and extends up to April. Potentialevapotranspiration (RPET) was derivedfrom Penman’s equation (Penman, 1948)and interception and bulk throughfall ratiowas computed from Quickbird imageoutside the model environment as suggestedby Kuriakose et al., (2006; 2009b). All non-intercepted rainfall was assumed to bereaching the land surface while any canopystorage was made available forevapotranspiration. RPET was adjusted tocrop-specific PET using crop factors of therespective land use types. All RPETexceeding the available canopy storage in a

given day was assumed to be the fraction ofPET that is to occur from the soil (FPET).Within STARWARS, evapotranspirationoccurs at the actual rates (which is FPETadjusted to crop-specific FPET) when thesoil is close to saturation; else FPET wasscaled down linearly depending on the soilmoisture content in a given time step.

The DEMAn available semidetail-dem was combinedwith Trimble DGPS and semi-automaticelevation measurements to generate a digitalelevation model (DEM) with a terraceimpresion (Figure 3). In order to update theDEM by inputting the terrace on the DEM(figure 4), digital terrain modeling wasapplied. One of the inputs is terrace map.The terrace map was derived from high-ressatellite image of Quickbird recorded on2009. Digital terrain modeling was produceusing semi-automatic method using ILWIS3.8. The result then was validated withseveral data: 1. Elevation point from DGPS,2. Transect line from field work, includingnumber of terraces on each transect anddifferent height of the terraces.

Figure 3. Dem generation


Figure 4. DEM with terrace impression

Soil depthSoil depth was one of the importantparameters to run the slope stability model.This information was obtained by fieldinvestigation as suggested by Kuriakose etal., (2009a). About 108 locations weremeasured. Soil depth was measured fromtop soil to the bedrock at the escarmnet,terrace cutting, river valley or using borhole. These measurements were interpolatedusing ILWIS software to produce a spatiallydistributed soil depth map. Four methodswere compared namely, Krigging ,Moving

Average, Trend Surface, Moving Surface(figure 5). The accuracy of the results wastested based on the coefficient ofdetermination (R2) (Nash and Sutcliffe,1970) between observed and predicted soildepths at a selected set of validationlocations. 20% from the total sample wererandomly selected for this validationprocesses. Total 22 points were selected.From the result, Figure 5, we understandthat kriging result was significantly betterthan the moving average results and thusthis map was used for the modeling.

Figure 5. Soil depth modeling. a. Krigging b.Moving Average c. Trend Surface d. MovingSurface

a. b. c. d.


Soil propertiesSeveral soil properties were necessary forrunning the model. Undisturbed anddisturbed soil samples were collected fromthe field in order to derive the soilmechanical and hydrological properties. Astratified random sampling was conducted.Measured variables include saturatedhydraulic conductivity, infiltration rate andbulk density. Cohesion and angle of internalfriction were derived from undisturbedsamples tested in the laboratory usingtriaxial (UU) tests.

3. RESULTS AND DISCUSSIONThe STARWARS was run on a

daily basis from 1st January – 8thNovember 2010 to get the initial conditionsfor modeling the 9th November 2010 slopehydrology. The results of the 8th November2010 was reported and used as the initialcondition for the simulation of 9thNovember 2010. The slopes were mostsaturated on the 13.10 pm with the overallhighest mean water level reported being 1.8m. There was no data available for assessingwhether this date was in reality the mostsaturated. However, from the averagemonthly rainfall it is known that Decemberis the wettest month in the region and hencethis model prediction is plausible. So also, itis known that it was on 13.00 pm that theSitieng landslides occurred in the region andthus the assumption that landslides occur asa consequence of high pore water pressureseems valid.

A sensitivity analysis was performedto understand the contribution of individualparameters to the overall performance of themodel. This in part compensates for the lackof appropriate calibration and validation ofSTARWARS. The sensitivity analysis wasperformed keeping the same initialconditions as that used for the actualsimulation of November 2010 and themeteorological data of the month. Severalsimulations were carried out, each time

varying only one of the parameters by acertain percentage. The model output basedon which sensitivity was assessed was totalsoil water storage. The parameters towardswhich sensitivity was tested were ksat, n, αand soil depth. The parameter change thatwas imposed was 25%, 50% and 100% ofthe standard deviation to the meanparameter value and the correspondingresults were also quantified in terms ofnormalized change in soil storage. Figure 6-a shows the sensitivity of STARWARStowards individual parameters. It is evidentthat soil water storage predictions bySTARWARS are highly sensitive to soildepth followed by porosity. This clearlyreflects reality and thus has a physicalmeaning; with increase in porosity and soildepth there is more pore space available forwater to be stored.

(a)

(b)

Figure 6. Sensitivity of STARWARS model(a) and Sensitivity analysis of PROBSTABmodel (b)


A sensitivity analysis was performedto assess the normalized change in overallfactor of safety towards slope, angle ofinternal friction, bulk density and cohesion.Figure 6-b shows the result of this sensitivityanalysis. It is evident that there is a linearrelationship between slope angle and factorof safety. With decreasing slope angle thefactor of safety increases rapidly and viceversa. In descending order, the otherparameters to which factor of safety aresensitive is angle of internal friction,cohesion and soil depth. In contrast withslope angle, the slope stability increases asthe friction angle increases. This was alsothe case with cohesion. The safety factorincreases as cohesion increases implyingthat cohesive soils are less prone to slopestability problems.

3.1 Model validityFigure 7 shows overall stability in

the November 2010 overlaid by thelandslide locations. The result shows that all

the landslides occurred in unstable area. Theprobability of occurrence of landslides canbe calculated using the prior probabilityconcept (Equation 3) developed by Bonham-Carter et al., (1994).

Pp =( )( ) Equation 3

where Pp is the spatial probability oflandslides, Npix(slide) is the number of pixelsoccupied by landslides and Npix(total) is thetotal number of pixels occupying the map.Based on the area occupied by the actuallandslides (pixels) of the November 2010and the total number of pixels in the studyarea (pixels), the actual probability oflandslides at that time was 0.09. Thepredicted overall probability of landslides inthe area was 0.127 given 8851 pixels havingFS<1. With reference to the location oflandslides the model was >70% accurate,while in terms of the area predicted as failedthe model was only 10% accurate.

Figure 7. Overall factor of safety a). No terrace b).With terraces

4. CONCLUSIONSThe models shows that the landuse

practice in the study area work like twoedges of sword. We understand from severalpaper that promoting of bench terrace canbe reducing the risk of soil erosion but in theother hands it increases the risk of landslide.From the slope-stability modeling, we can

see that the terrace increase the pore-waterpressure significantly which lead to the idealconditions for the failures. The extremelyhigh intensity rainfall, in the other hands,may build a sharp increase of pore-waterpressure. The increasing probability offailure might cause the soil erosion evenworse. Therefore, in order to make the


terrace practice is effective to control theland degradation process; the terrace has tobe well maintained. The farmers in the studyarea always maintain the terrace. Theyimmediately re-construct the terrace afterthe landslide event take place. In this way,terrace practice will effectively control theland degradation processes.

Even when geomorphic processesare well understood, their descriptions mustoften be simplified in order to operate withinlandscape models. Model predictions arelikely sensitive to both the processesincluded and how they are simplified(Lancaster and Grant, 1999). It will not belong since now that with such well-definedmodels, it will be possible to answerquestions that might seem very simple suchas “Where will the land degradation occur;why do they occur and when will theyoccur”?

5. REFERENCES

Ang, A.H.-S. and Tang, W.H., 1984.Probability Concepts in EngineeringPlanning and Design: Decision, Risk,and Reliability (V II). John Wiley &Sons, New York, 562 pp.

Bonham-Carter, G.F., 1994. Geographicinformation systems for geoscientists:modelling with GIS. Computer Methodsin Geosciences, Vol 13. PergamonPress, Oxford, 398 pp.

Borga, M., Dalla Fontana, G., Da Ros, D.and Marchi, L., 1998. Shallow landslidehazard asessment using a physicallybased model and digital elevation data.Environmental Geology, 35(2-3): 81-88.

DGHM, 2007. Landslide database in Java,Indonesia. Directorate of Geology andHazards Mitigation, Govt. of Indonesia,Jakarta (Unpublished).

Geertz, C., 1963. Agricultural involution:the process of ecological change inIndonesia. University of Chicago Press,Chicago.

Greenway, D.R., 1987. Chapter 6:Vegetation and Slope Stability. In: M.G.Anderson and K.S. Richards (Editors),Slope Stability. John Wiley and SonsLtd., West Sussex, England.

Hadmoko, D.S. and Lavigne, F., 2007.Landslides in Menoreh Mountains,Central Java, Indonesia: dynamics,triggering, spatio-temporal analysis andassociated hazards, InternationalConference of International Associationof Geomorphologists on EnvironmentalChange in the Tropic, 22-25 June, KotaKinabalu, Sabah, Malaysia.

Hadmoko, D.S., Lavigne, F., Sartohadi, J.,Samodra, G. and Christanto, N., 2009.GIS application for comprehensivespatial landslide analysis in Kayangancatchment, Menoreh Mountains, Java,Indonesia. In: J.P. Malet, A. Remaitreand T.A. Bogaard (Editors), LandslideProcesses: From GeomorphologicMapping to Dynamic Modelling - Atribute to Prof. Dr. Theo van Asch.European Centre on GeomorphologicalHazards (CERG), Strasbourg, France.

Hulme, M. and Sheard, N., 1999. ClimateChange Scenarios for Indonesia.Climate Change Research Unit, WorldWildlife Fund, Norwich, UK.http://www.cru.uea.ac.uk/~mikeh/research/indonesia.pdf, Accessed on: 19July 2009.

ILC, 2006. The International LandslideCentre landslide fatality database.International Landslide Centre,University of Durham. 01 February.http://www.landslidecentre.org/database.htm, Accessed on: 09 August 2009.

Kuriakose, S.L., Devkota, S., Rossiter, D.G.and Jetten, V.G., 2009a. Prediction ofsoil depth using environmental variablesin an anthropogenic landscape, a casestudy in the Western Ghats of Kerala,India. Catena, 79(1): 27-38.

Kuriakose, S.L., Lele, N.V. and Joshi, P.K.,2006. Fractional Vegetation Cover


Estimation from MODIS NDVI Data –A comparison of methods andapplication in Dynamic NumericalModelling for Debris Flow (Urul Pottal)Initiation Prediction. In: S. Subbaih(Editor), International Conference onNatural Hazards and Disasters - Localto Global Perspectives, November 25th

to 27th. IGS Conference Publication(CDRom): Sri KrishnadevarayaUniversity and The Indian GeographicalSociety. ISSN No. 0973-5062,Anantapur, India.

Kuriakose, S.L., van Beek, L.P.H. and vanWesten, C.J., 2009b. Parameterizing aphysically based shallow landslidemodel in a data poor region. EarthSurface Processes and Landforms, 34(6):867-881.

Lancaster, S.T. and Grant, G.E., 1999.Modeling the interaction of Landslides,Debris Flows and the Channel Network.Corvallis Forestry ResearchCommunity, Corvallis, Oregon, USA.

Lavigne, F. and Gunnell, Y., 2006. Landcover change and abrupt environmentalimpacts on Javan volcanoes, Indonesia:a long term perspective on recent events.Regional Environmental Change, 6: 86-100.

Marfai, M.A. et al., 2008. Natural hazardsin Central Java Province, Indonesia: anoverview. Environmental Geology, 56:335-351.

Marhaento, 2006. Landslide hazard analysisusing heuristic-statistic method incombination with multi temporallandslide data. M.Sc Thesis, GadjahMada University, Yogyakarta.

Nash, J.E. and Sutcliffe, J.V., 1970. Riverflow forecasting through conceptual

models part I -- A discussion ofprinciples. Journal of Hydrology, 10(3):282-290.

Penman, H.L., 1948. Natural evaporationfrom open water, bare soil and grass.Proceedings of the Royal Society ofLondon, A(194): S. 120-145.

Simoni, S., Zanotti, F., Bertoldi, G. andRigon, R., 2008. Modelling theprobability of occurrence of shallowlandslides and channelized debris flowsusing GEOtop-FS. HydrologicalProcesses, 22(4): 532-545.

Takara, K., Apip and Bagiawan, A., 2008.Study on Early Warning System forDebris Flow and Landslide in theCitarum River Basin, Indonesia. In: K.Sassa, P. Canuti, L. Picarelli and Y.Yueping (Editors), Chapter 20:Monitoring, Prediction and EarlyWarning; Landslides - Disaster RiskReduction. Springer, Tokyo, pp. 369-374.

van Beek, L.P.H., 2002. Assessment of theinfluence of changes in Landuse andClimate on Landslide Activity in aMediterranean Environment. PhDThesis, University of Utrecht, Utrecht,The Netherlands, 363 pp.

van Beek, L.P.H. and van Asch, T.W.J.,2004. Regional assessment of the effectsof land-use change and landslide hazardby means of physically based modeling.Natural Hazards, 30(3): 289-304.

van Westen, C.J., van Asch, T.W.J. andSoeters, R., 2005. Landslide hazard andrisk zonation: why is it still so difficult?Bulletin of Engineering Geology andEnvironment, 65(2): 167-184.

investigating the effect of conservation …

Documents