You are on page 1of 13

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/271384570

The determination of the Landslide occurence
probability by spatial analysis of the Land
Morphometric...

Article in Carpathian journal of earth and environmental sciences · May 2014

CITATIONS READS

7 188

5 authors, including:

Dănuț Petrea Bilasco Stefan
Babeş-Bolyai University Academia Română Filiala Cluj Napoca Colect…
18 PUBLICATIONS 45 CITATIONS 55 PUBLICATIONS 91 CITATIONS

SEE PROFILE SEE PROFILE

Sanda Rosca Iuliu Vescan
Babeş-Bolyai University Babeş-Bolyai University
45 PUBLICATIONS 63 CITATIONS 17 PUBLICATIONS 52 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Unmanned aerial vehicle (UAV), Ground-penetrating radar (GPR) and G.I.S. spatial analysis for
landslides monitoring in Romania View project

All content following this page was uploaded by Sanda Rosca on 26 January 2015.

The user has requested enhancement of the downloaded file.

ro. e-mail: sbilasco@yahoo. together with be exposed to landslides. No. dpetrea@geografie. based on phenomenon. Transylvania Plateau. hydro-climatic conditions.Carpathian Journal of Earth and Environmental Sciences. probability. 91 . Faculty of Geography. Even the methodology adopted by Romanian Governamental Decision no. 400015. The implementation of the model in a relief unit characterized by a high morphological and morphometric variability. rosca_sanda@yahoo. Republicii Street. erosional processes and human activities generate The identification of the areas of susceptibility useless of agricultural lands inducing major negative associated with the landslide occurrence probability. Ștefan BILAȘCO2. 2007). 9. 2. Vol. may lead to a true estimation of the probability that a certain territory It is well known that landslides. Romania. The value of the Relative Operating Characteristics curve (0. May 2014. Cluj-Napoca. Iuliu VESCAN1 & Ioan FODOREAN1 1 "Babeş-Bolyai" University. the goal was to reduce the subjectivity level that occurs.ubbcluj. 9. GIS spatial analysis. p.ubbcluj. INTRODUCTION complex spatial analysis models. Cluj-Napoca Subsidiary Geography Section. Romania. Thiery. We believe that the identifying the preparing factors and the probability quantitative assessment of each morphometric factor of the occurrence and evolution of the studied and the integrated analysis of all factors.com. hazard coefficients 1. fioan@geografie.ubbcuj.ro. 1998).ro 2 Romanian Academy. In the investigated territory. triggering and further Statistical spatial analysis models are based on evolution of the landslides.com Abstract: The diagnosis and the prediction of landslides are essential problems in territorial planning from the efficient land-use perspective. area (Surdeanu. gives a high level of generality to the model. For the moment there are two natural hazards with the highest occurrence research directions: the heuristic one and the frequency and they have the widest manifestation quantitative one (Van Westen. Specialized literature emphasizes 91 . 2004. The main goal of this study is to identify the probability of the damaged lands occurrence (affected by landslides) by introducing and using some variables and coefficients out of the record which will be integrated in a complex model of GIS spatial analysis. declivity and the fact that the landslides are still active is a result of not suitable land-use in the transition period from centralized agriculture (in communism) to a traditional agriculture. implications on the economic development in the based on GIS models is a frequent subject in the affected areas. landslides represent the specialized literature. landslides are the most frequent process. in most of the methods. 400006 Cluj-Napoca. The predictability model presented in this study is The premise of the study is that the territorial part of the second category and uses the statistical morphometry represents the main factor which spatial analysis. In this context there were elaborated an amount of investigation methods. Sanda ROȘCA1. Key words: landslides. contributes to the occurrence. GIS data bases. These are the results of the favorable litology. In this manner. which means a considerable fragmentation of the properties. 447/ 2003 is quite incomplete. vescan@geografie.848) used to validate the model indicates the high predictability of the model and suggests its possibility to be used for other theories with similar morphological characteristics. In Romania. in the process of marking the variables. but there isn`t any consensus yet.102 THE DETERMINATION OF THE LANDSLIDE OCCURRENCE PROBABILITY BY GIS SPATIAL ANALYSIS OF THE LAND MORPHOMETRIC CHARACTERISTICS (CASE STUDY: THE TRANSYLVANIAN PLATEAU) Dănuț PETREA1.

not only from a tectonic. as Bălteanu (2009). the bedrock and external modeling. Lee & Pradham. Hârjoabă (1968). this methodology allows units can be identified: depressions. density GIS spatial analysis models for the of fragmentation. slope aspect. possibilities for the storage and management of large amounts of databases. one appreciating that this is based on elements which are stresses the multi-varied statistical analysis. Rădoane et al. (2011). international specialized literature. in the hydrogeological) imposes an amount of subjectivity transition area of the Transylvanian Plateau to the which derives not only from the different formation neighboring mountain area and we analyzed them of the applicants but also from the deficient types. Manea & Surdeanu (2012). derived from DEM. Saha view as a reflex of the varied coniditions imposed by (2002). which is based on the continuity (lithology. Tarnavelor Plateau has the largest area of all Considering the necessity to respect the the three units. Minucsér (2013). on the territory of the Transylvanian The study of landslides is an old Plateau one identifies a high frequency and variety preoccupation of Romanian researchers: Tufescu of types and forms of landslides. Onac et al. (2000). 1995. Filip (2008). Dhakal et al. The point of view. like a different structure with common and scales of the maps and available data for characteristics. For increasing the accuracy of the results and based on identifying the relations between the for quantifying the predictability degree of the preparing factors as independent variables and the model we used a number of nine morphometric existent landslides as dependent variables. 30%. which gives a statistical index to independently from the interpretation that is given each factor implied in the model. Arghiuş et al. Martiniuc (1961). (2013) for Romania. seismicity). Transylvanian Plain affected by landslides. and more recent: Micu & an important agricultural region of Romania. which it is formed. the Government of Romania and Tarnavelor Plateau. due to its different geological structures on 7th of May 2003. The large as not so efficient by many specialists because the spatial diversity of these morphometric units mark of some variables (lithological. variables (hypsometry. stream power index. The economic and (1966). Armaș The Transylvanian Plateau is a complex (2011).. some conclusive results. climatic and imposed their inclusion in this study. (1995).. low plateaus. STUDY AREA made by Bălteanu & Micu (2009). from morphometric and geomorphological point of Nagrajan et al. applied difficult to quantify and do not have a spatial by Carrara et al.. but also Plain 18% and Somesan Plateau 12% of the entire 92 . (Fig. slope. Bilaşco et al. First of all. depth of fragmentation. all about 70% of its area is agricultural land. being based on the by the researcher to some data that can often be presence of landslides in the analyzed territory.. different regions of the country. profile probability. Bălteanu (1983). followed by the Transylvanian demands of the present methodology. research was 2. In these (2007). Therefore three morphographic In some situations. territorial impact of these phenomena is higher due Băcăuanu (1980). starting from their presence or existing model with some unitary morphological and absence in each uniform unit from a morphometric morphometric variables. wetness determination of the landslides occurrence index. each with different issued the Decree 447/2003 which settles the morphometric characteristics. using the bi-varied statistical approach. Sarkar & Kanguno (2004). to the fact that the Transylvanian Plateau represents (1995).. statistical analysis. main reason for the study is the fact that they are Another type of approach is the bi-varied rigorously and objectively determined. using classical identification. but also approaches have been made by Chung et al.many types of statistical approaches. methodology for the elaboration and for the content The border area with the neighbouring of natural risk maps for landslides with application mountains presents a very high morphometric guidelines published in Monitorul Oficial 305 on the variety.. we analysis of the contribution of every factor in consider that there is a possibility to validate the landslides occurrence. 1). even though it is regarded high transition hills near the mountains. (2000). In the physical geographical unit.. similar structural and lithological point of view. curvature) which are considered to be essential in were developed at the same time with the increase of spatial analysis and planning studies. This incomplete or inadequate from the point of view of type of approach is most frequently used and is scale. hydrogeology. In this purpose. plan curvature. 1999. (2008). Transylvanian Plateau can be divided into three Because of the bigger extension of the areas major units: Somesan Plateau.. conditions. inventory and mapping From regionally point of view the methodology. Surdeanu (1998).

Each element of the morphometric database is included in the spatial analysis model as a parameter for the identification of the landslide occurrence probability. wetness index. plan curvature. slope aspect.hypsometry factor Fdd. and the influence of the SPI and WPI factors in the The Digital Elevation Model. The intervals choosed for each coefficient were selected on the basis of the studies mentioned in references via expert knowledge-based approach. (Table 1). In the spatial analysis model.slope factor Fa. The probability values of the parameters were obtained based on Decree 447/2003. density of fragmentation. To implement the the validation of the model to quantify its degree of equation by means of the above-mentioned factors.Probability coefficient Fs. RESULTS equation. modeled databases morphometric point of view. drainage density. we use the determinist quantitative analysis based on mathematic indexes integrated into a formula that takes into consideration all the parameters identified spatially before: KM.aspect (slope aspect) factor Fh.density of fragmentation factor Fda. The the probability of occurrence and the development of hypsometric factor has a double role: first as a the landslides are divided in two major categories: database shaped on the basis of contour lines and the quantitative or statistical models and semi. Geographical position of the study area Fspi. was based on the creation and analysis of the 93 . predictability. which gives high accuracy results.stream power index factor Fcpl. which analysis sub-models to identify the spatial extension are derived from the Digital Elevation Model (DEM). taking into account the territorial extension for each characteristic interval. represents the support for the parameters that compose the spatial analysis 4.profile curvature factor All these factors are analyzed as GIS databases to be The GIS spatial analysis models that analyze integrated in the spatial analysis model.plan curvature factor 3. and second as a derived database and active The achievement of a GIS spatial analysis component of the spatial analysis equation which will model supposes the following of several steps: the finalize the final model for the identification of the creation of the database. drainage network which is the base for deriving other quantitative or deterministic models.MATERIAL AND METHODS Fcpr. Using the TopoGrid Function of the ArcGis software. The execution of the spatial analysis model we obtained a 20 m resolution DEM.study area. starting from the primary (40%) compared to the regional units that compose morphometric database (contour lines obtained by the major unit. landslide occurrence probability. profile curvature. factors. probability coefficient) and derived databases (slope angle. It is also the most complex from a vectorization of 1:25000 maps). Spatial analysis is based only on the it was necessary to set up two additional GIS spatial morphometric characteristics of the territory. seen as a final result of the modeling. stream power index. hipsometry and depth of fragmentation). The contact area of the Transylvanian The spatial analysis model is developed using Plateau with the mountains has a large extension more database structures.wetness index factor Figure 1. modeled database.depth of fragmentation factor Fwi. (the Digital Elevation Model. the spatial modeling proper.

8 Very high >400 m .8 Very high 30. W slopes (sunny and semi-sunny) are >0. . because the river network is almost absent.1-0.high values characterise the confluence areas and the surfaces with a lack of vegetation with the presence of linear erosion.the surfaces with 400-700 m.51-0. SW.51-0.5 15.3 Average 1-1. landslides.1 Low 165-400 m probability the altitude extended surfaces unltil 400 0. S. Fwi Average.5 501-700 m High .31-0.51-0. characteristic to >0. >0. >0.31-0.51-0. S shadow and half shadow slopes with low Fa sussceptibility for landslides.1 Low 0-100 m fragmentation . characteristic intervals and the probability class where Table 1.8 High 2-2. terraces 0. NE plain and extended watersheds are.8 High SE.8 high 18-20 the water from the soil.8 Very high 701 – 900 m mountain contact areas.1-0.31-0.51-0.low values emphasise low saturation in water index Wetness Index < 0.1 Low 0-14 surfaces identified with watershed area and low 0.plane surfaces.8 High >35° very high susceptibility for landslides and gravity >0. E and NW exposure are high 0.there are considered to be surfaces with low < 0.SE.3 Average N.1-0. -high values of the parameter indicate surfaces 0.31-0. 0. Average- Fda 0. not affected by slope Hipsometry processes and which are influenced by river bed Average. < 0. identified mainly. in general. we identified the class out of the total area of the analysed territory. < 0. spacific to valley corridors.8 Very high SW.3 Average 14-16 hills.1-25° Fs high . < 0. .1-400 overlayed on high slope sufaces).5 m/kmp average section of the slope.Low slope surfaces (<5°) characteristic for major 0. For each parameter. specific to terraces 0.5 E. Computed statistical values for the GIS model variables Para- Code Value P Intervals Explanation meter .low values of drainage density. in general. not Aspect (slope aspect) Average. terraces plain and extended slopes are.moderately inclined surfaces (5.databases and the thematic maps specific for every they belong according to the weight of the area of that parameter. Fh 0.low values of the depth emphasizes a low potential 0. W highly sussceptible to landslides because of the high amount of precipitation and high sunny degree which is favourable to frost and defrost processes. 0. presents an accentuated morphodynamic potential. not affected by Slope angle Average.31-0. processes.1-0.1-35° processes.5 200.31-0.5 m/kmp area induce low probability to occur landslides Drainage density Average.1-15°) are preferable for mass movements 25. in the 94 .1-0. NE.1-15° river beds of the main rivers.slopes with N.5-2 m/kmp . NW .3 Average 401-500 m meters which are. affected by landslides. impose a high susceptibility due to regresive erosion of the 1 and 2 river flows.1 Low PLAIN .3 Average 100.highly inclined surfaces (>25°) presents high and 0.8 Very high >2. in general.3 Average 5.More than 900 m altitude relief. .8 High 300.1-200 m Depth of for landslides.5 16-18 characterized by a high saturation index regardind high 0.51-0.1 Low 0-5° .1-300 m .1 Low 0-1 m/kmp .1-30°.5 m/kmp .average values of drainage density.5 1.1-0. specifical to high Fdd 0.8 High 901-1800 m fronts are partially affected by slope processes.high values of the depth induce a high potential for high landslides to occure due to high altitude gauge (arear 0. identified in slope 0. 0. < 0.

the values closed to 0 are for the bent linear 0.1 Low >0 .1-0.8 Very high >0.the values closed to 0 are given to the linear 0.1 high surfaces. . Plan Curvature Average. generationg a low probability. 4. < 0.2 average to high probability. the probability of landslide occurrence is coefficient depending on the area.2-0. two-equation system: the equation calculating the 95 .2 network which acts to the level of the slope by depth and regresive giving a high and very high probability. assessment of every probability interval for each o z – the value of the probability parameter was not possible. Fspi 0. The proposed analytical procedure considers vp = a + z (4) where: that the probability for a landslide to occur on a o vp – the probability value. slope area. we will give an example of the The calculation of the probability value calculation of the probability value for the intervals according to the proposed procedure is based on a of the slope angle factor (Table 2). this fact determines a high probability and very high that the landslides to >0.31-0.1 Low < .3 Average -0. probability interval.5 flow and high level of infiltration of the water into the soil.high values of the SPI characterise the areas where high 0. Fcpr .51-0. of the probability interval. the probability value.3 Average 0-1 power of the drainage and. The analysis and creation of the value of the probability coefficient specific for the databases extension area of the probability interval (2 and 3) and the equation calculating the value of the interval The methodology for the determination of the probability coefficient (4).negative values represents concave surfaces 0.8 High 1.0. generating a high susceptibility for landslides.1 high flow and low infiltration level. the minimal value probability interval out of the total studied area.1 . aspects that induce a very high probability to occur >0.1-1.8 Very high >0. Because the area of o y – the interval area of spatial extension.8 High 0. Therefore.4 occur.09 – 0 overlayed over the accumulation zones with a low probability.low values for the SPI emphasise a low erosional Stream Power 0.31-0. in general to the low declivity slopes with a Fcpl 0. In order to determine coefficient depending on the area.21-1.2 high surfaces with a average to high probability. high probability interval is very high on the entire o z – the value of the probability study area. the qualitative as percentage. Profile Curvature Average.2 .8 Very high >20 river beds.negative values characterise convexe slopes with 0. a low Index Average probability for the occurence of the landslides.8 High 0. probability value proposed in the Government (x * y) / 100 = z (2) Decree 447/2003 is based on the quantitative and x = vpmax .5 0. at the contact between slopes and major >0.1 Low <0 .vpmin (3) where: qualitative analysis of the landslides for every o x – the value of the probability interval. this paper.5 landslides. of course.1. implementation is very large. we proposed an analytical o vpmax. if the spatial extension of the interval.8 Very high >1. < 0.3 Average 0-0.51-0. vpmin – the maximal value of the procedure which takes into account the area of every probability interval. territory is directly proportional to its spatial o a – the basic value of the probability extension. .31-0. .1-0.51-0.5 1. a high dynamics due to their shape.1-0.1-0. high and that interval receives a high value in the To illustrate the analytical method proposed in value interval.1-0.5 0-0.high values characterise concave slopes with high 0. < 0.maximum positive values indicates convexe surfaces for forming the elementar hydrographical >0.4 the erosional power is high.

00002 characterized by low probability are very small. a grid type For high probability raster database has been obtained.19 * 7) / 100 = 0.035 = 0. Hypsometry raster database. High 0.80 11. reveals the large Very High 701 – 900 m 0.002 = 0.48 V The database corresponding to the slope factor has been created taking into account the grid type 4.16 8475. the probability values for every altitudinal intervals have a reduced territorial extension. Transylvanian Plateau reveals the considerable Table 2. Medium 15.2.51 81. of 7%. we computed the Transylvanian Plateau.1° – 35° 0. high and very high probability facts. Higher altitudes Value (km2) 0-2° are specific for Târnave Plateau and the Low 0.2. Values and probability classes (slope) extension of low altitude areas.51 interval.79 spatial extension of low probability (35% of all areas) followed by the average probability (57%).1° .002 high weight (27.1° -30° Probability Area High 0.1 – 15 0. (0.01 = 0.38 The quantitative analysis of the probability Medium High 501-700 m 0.3.31 + 0. insolation. 2008) and temperature.98 NE 4. Hypsometric values and probability classes 25.03 9732.51 + 0.03 stability of the slopes in the analysed territory.77 Medium 401-500 m 0. derived from DEM.21 by means of climatic parameters that are unequally distributed on the surface: solar radiation.80 covering only 0. Someșan Plateau.71 2.80 + 0.03 8898.01 The analysis of the probability indices has been 0.11 development of geomorphological slope processes 0. Taking into account the above-mentioned The average-high.47 High NV necessary to be quantified.35 6333. Values and probability classes (slope aspect) assessment of the landslide occurrence probability. showing the very (0. This indicates a good (0.29 * 1) / 100 = 0.41%. For low probability 1% and 0.1° – 25° Table 4.15 6938.32 performed according to the probability values presented in table 3. performed High 901-1800 m 0.80 604. The analysis of the probability factors N Medium 0. followed by average-high.68% respectively.1 Slope angle Medium E The slope angle is a parameter absolutely 0. One remarks that the areas (0.57 6008.5° Transylvanian Subcarpathian Hills (the transition Medium 5. as well as Probability Intervals in the contact basins and corridors.2 Slope aspect For average probability The slope exposure to the sun influences the (0.49 values specific for the slope angle factor. 0 + 0.66 Probability Intervals >35° Value (km2) Very High 30. Very High 0.53 at the level of the entire study area. high and very high probability with weights of 24.01) / 100 = 0.53%) of the average probability 0.1 * 35) / 100 = 0. 0. Probability Area Probability Intervals Value (km2) Low plan 0.. both quantitatively.21 14421. and the The hypsometric analysis of the indications from table 2. and qualitatively.2.11 = 0.000002 = 0.84 5967. probability value for each interval of the morphometric factors taken into account for the Table 3.10 + 0.21 sector to the mountain unit).03 4. 23.35 6153.20 * 57) / 100 = 0.2.51 58.38 Low 165-400 m 0.03 S as a factor that generates landforms emerged as a SV consequence of those processes.00 4. interval have been chosen (Table 4) according to 96 . as a SE factor that enables slope processes.54% of the total area of the Based on these equations.84% For very high probability and 23.1° – 20° 0. By reclassification. largely distributed in Probability Area the Transylvanian Plain.32 1792.20 * 0.00 High 20.01% respectively.05 137. For average-high probability precipitation (Dragotă et al.

with 22% and 11% These two factors can be associated because respectively. 51.4. favor a higher degree of water (km2) Value infiltration. followed by the average power index (SPI) and average-high intervals. due to the Probabi- nature of their mechanisms of development and Area Probability Intervals lity manifestation.40 numerical probability value in the created database.28 does not exist in the study area.14 5647. one remarks the very high extension of the low 4. lowlands means a high fragmentation of the slope.400 m 0. Concerning the weight of each probability Very High >501 m 0. - (Table 5). The low the slope.01% in the very 0. The depth of fragmentation They are identified mainly in Someșan The landform vertical fragmentation. indicating a relative stability of the they are part of the same category of factors that slopes in the study area.0. option is based on the fact that the development and further evolution of landslides take place on very 4.72% of the Probability Intervals lity (km2) area is included in the average probability class. One may notice that the high probability interval has not been given a 1. This Western Carpathians.05 13655. A high value of the drainage network density characteristic of the large watershed areas. which is and river floodplains.which the database representing the hypsometric The high and very high probability intervals probability factor has been finalized. calculated at the basis or on favourable for the emergence of landslides.5 m/km2 48. Density of fragmentation small areas. as a Plateau and the Transylvanian Plain.31 48. 4.52 high probability category.53 1826. These three probability categories are located in the Having in mind the purpose of the analysis. take into consideration the topography of the terrain 97 .5 This is due to the fact that this difference in height High 0. the high area of study we opted for the hectare size quadrat as Târnave Plateau and the contact area with the calculation units for the depth of fragmentation. The average. difference in height on a certain calculation area. for the Probabi- Area depth of land fragmentation factor. Value 0 .23% and 2.33 2627.5. high. At high values.40%.6.5 m/km2 0.83 interval on the total area of the analysed territory. in their turn.5/4. high and very high probability covers small highlights territorial discrepancies regarding the areas.04 present a low landslide occurrence probability High 401 .2.56 fragmentation are associated to stable and relatively 101 – 200 m Medium 0. Of course. The minimal values of the density of Low 0 . in this Transylvanian Subcarpathian Hills. on the (density of fragmentation) extension areas of each interval.100 m 0.04 12123. m/km2 Very High >2. it determines a high instability of the values of the depth of fragmentation are usually slope. Therefore.81 1449.49 average-high probability and only 0.20 13038.2. Wetness index (WI) and stream probability interval (54%). The areas with cover small areas of 7% and 6% respectively.2.5-2 Medium High 0. out of low and average landslide occurrence probability the total area of study. have the highest weight (38.63% respectively). factors. It is part of the category of preparing indicates unstable ground.500 m . usually with high slopes.09% in the low probability class.5 m/km2 0. Values and probability classes methodology proposed in this study. The calculation of the probability coefficient value (Table 6) is based. 0. so the analysis of the depth of The landform density of fragmentation fragmentation according to the classical represents an important indicator in the calculation methodology (by km2) does not meet the modeling and determination of the landslide occurrence objectives.19% in the Low 0. they define area favourable for the development of geomorphological with low landslide occurrence probability.2. m/km2 2 – 2. probability factor for the emergence of landslides. The high value of the difference in height probability.68% and 33. 0.13%.10 Table 6. with weights of 25. according to the Table 5. They Medium High 301 .5 – 1 m/km2 Medium 1 –1.80 1. slope processes. Values and probability classes (depth of fragmentation) These processes.99 stable slopes and landforms as regards the effects of 201 –300 m linear erosion exerted by the drainage network.

20% for low slope pronounced concavity and therefore there is a and average-high probability and 49% for average very high landslide occurrence probability. The concave slopes with divergent flow and a low to other probability intervals have very low weights.01 35.80 2168. two indicators may be between 12. As for the territorial extension of the probability Table 7.7% Value (km2) Low 0 .02 [SPI] Very High >1. The low curvature.19 12115.09 .51 0.21 – 1.64 flat areas of watersheds and plateaus. where there is a high landslide occurrence probability (Table 8). The zero value is associated to linear slopes. average landslide occurrence probability.2. The areas The plan curvature represents the perpendicular with the largest extension (73.0. High 1.20 0. with an Medium -0.81 occurrence probability.14 0. 7% for high probability.40 13004. [WI] Medium High 1.80 0. terrace Very High >0.2 0.29 and are directly proportional obtained: the plan curvature and the profile to the landslide occurrence probability. due to the incipient development of a probability.61 and 51.flow accumulation The areas with a prominent slope and exposure 400 – area in m2 of the DEM cell change are identified with the help of the slope .53 2574.0 0.30 class. followed by areas positive values highlight the convex slopes.16 0.1 0.80 0. Values and probability classes (WI) intervals according to the profile curvature. 0.001)).76 weight of the very high probability interval is almost Very High >20 0.2 0.1 0. with presenting a low probability. Low <0 0. stream within a certain catchment area (SPI). According to the The calculated values of the wetness index are method of curvature analysis.19% for the high probability class while the High 18 . with low direct impact on the landslide trigger mechanisms and (almost null) landslide occurrence probability.8 and 1. the GIS Probability Area analysis shows the very large extension of the average Probability Intervals and average-high probability. convex or concave).31 0.44 risers and steep banks.. Positive values highlight the probability.58 null as compared to the entire area of the study unit (Table 9). therefore having a negative values define convex slopes.64 (1991) for SPI. Values and probability classes (SPI) Probability Area performed according to the equations derived by Probability Intervals Value (km2) Beven & Kirkby (1979) for WI and Moore et al. taking into consideration the The high values are characteristic for areas with shape of the slope (linear.45%. Values close 98 . Positive values are High 0. Values and probability classes (profile curvature) ranging between -2. The average probability class. The other three intervals Medium 14 . implemented in ArcGIS as: Medium 0-1 0.001)*(“slope”/100+0. with a weight of convergent flow and a high landslide occurrence 26.10 Ln((“accumulation”*400)/Tan(“slope”)).54%) are part of the direction on the maximum slope orientation.07 characteristic for moderate and steep slopes. Table 9.003 where: Ln.4 0. The analysis of the SPI calculated values.32 Ln((“accumulation”+0.60% respectively.4 0.02 6667. values correspond to lands with low degree of water The profile curvature provides information saturation which involves a low landslide probability. The spatial modeling of these two factors has been Table 8. The relatively high water saturation. about the areas with moderate and high flow on the topographical surface.24 18535.66 present lower weights.1 – 1. Tan – mathematical identifiers 4. Plan and profile curvature accumulation . which The extension of areas with different degrees present an average or average-high landslide of probability varies between 4% for very high occurrence probability.and highlight the degree of water accumulation on covering extremely small areas at the contact certain areas (WI) or the power of water erosion in a between the plateau and the mountains.34 5030.7. highlights the areas Probability Intervals Probability Area with potentially high erosion exerted by water Value (km2) streams.02 5128.45.19 10377.slope in radians topographical surface curvature.00 incipient drainage network and low landslide Medium High 0 – 0. 0. a high probability of landslide emergence. The negative values are characteristic for permanent drainage network in those sectors. The negative values of the SPI indicate the Low < . with weights of 48.2 0.14% for the low probability Medium High 16 .51 49.43 and 25.1 – 0.18 0. probability (Table 7).

intermediate models have other three intervals: 9.17% for the low probability been conceived and later integrated in the final interval. plan curvature. mathematical equations transposed into GIS spatial Regarding the territorial extension of the analysis functions.2 – 0.98 most important one was to create. In their turn. Spatial analysis using the Raster Calculator function shaped as: The spatial modeling of the landslide SquareRoot (("Fs" * "Fa")/7 * "Fh" * "Fdd" occurrence probability demands the use of several * "Fda" * "Fwi" * "Fspi" * "Fcpl" * "Fcpr") spatial analysis techniques.2 0.25 included in ArcGIS programme. the complex model based on the High probability equation and on the entry modeled and High 0.69 spatial analysis.0. Very High >0.91%) modeling techniques.79 The model has been executed using the equation (1) implemented in a GIS environment 4.to 0 are associated to linear flow slopes which present and the processing of thematic databases by means of an average to high landslide probability.15% for the very high probability (Table 10).70% for the average probability and structure of the probability model (Fig.12 2947.1 0. spatial analysis and represent defining elements of the probability techniques have been used by database reclassification.07%) and high (46. The second and Medium 0 . by means of of average-high (32. the main entry elements in the spatial analysis equations meant to Table 10: Values and probability classes (plan curvature) determine the landslide occurrence probability.5 0.1 – 0. these techniques suppose the use of specialized software Figure 2: GIS spatial analysis model of the landslide occurrence probability 99 .37 8083. The main purpose is to generate probability intervals. calculation equation implemented in the process of The probability coefficients derived from each spatial analysis and identification of areas with factor are represented by raster format databases. 2). Probability Area The first one was to calculate the coefficients Probability Intervals Value (km2) representing WI and SPI according to the functions Low <0 0. GIS spatial analysis and probability intervals and the small extension of the reclassification of databases. All derive and represent the probability coefficients the factors analysed above constitute raster databases defining the analyzed factors. by means of Medium 0. In order to 0.01 2310.80 38. different degrees of landslide occurrence probability.5 0. one remarks the very large extension Starting from the spatial databases.28 derived databases.3. in their turn.64 11824. stored in different database structures. 11. defined according to the terrain new attributes. These represent.

In the Niraj drainage basin. covering a total area of 5510 ha (Fig. Model validation In order to establish the model degree of predictability. Arieș and Olt (in fewer cases). we opted for its validation by means of ROC curve determined at the level of two distinctive areas. a morphometrically complex landform unit. It is important to Figure 4: ROC curve for the high landslide occurrence stress the fact that. validated by the average category of landslide occurrence existence of landslides (true positive). Lands corresponding to it are mainly located along the large valleys of Someș and Mureș (mostly). and Someşan Plateau which is. covering very small areas at the contact with the mountains. topographical maps and field inventory. 4). Morphometry based landslide probability map represent areas in which the model has calculated a of the study area very high probability but where there are no landslides (false positive). most of the territory belonging the OY correspond to areas in which the model has to the Transylvanian Plateau is included in the determined a high probability. The result of the model is materialized in a has been finished. The probability. considered to be representative for the Transylvanian Plateau: Niraj drainage basin. The ROC method allows the evaluation of the predictability rate of the methods applied comparing the obtained probability map with the map of the existing landslides. The values represented on On the whole. a number of 322 landslides have been identified. 4. in its turn. The ROC curve is calculated for the 10 percent threshold which represents the percentage of the analyzed area for which the false positive and the true positive values are determined.30% of the total area of the analysed unit. (Fig. 126 landslides have been detected and mapped covering a total area of 984 ha. The average-high and high probability categories are almost unsizeable in the graphical representation and in the quantitative spatial analysis. The low probability category represents 40. no areas with very high landslide raster database which highlights the landslide occurrence probability included in the category > 0.761). Transylvanian Plateau. Lands comprised in this probability class (0. once the spatial analysis model probability 100 .63% of the total area of the plateau. followed by the average class (0.806) category cover larger areas of Târnave Plateau. The landslide inventory was based on SPOT imagery.977). They have a weight of only 0. 3). Someșan Plateau and the Transylvanian Subcarpathians. on larger areas. 3).8 occurrence probability for every pixel of the have been identified at the level of the entire analysed area (Fig. and the very high class (0. remarkable because its high landform diversity makes it a true “synthesis” of the whole Transylvanian Plateau. which defines an area with a weight of best validation was obtained for the high probability 59.4.07%. as well as in the Transylvanian Plain. The values of the OX Figure 3. In Someşan Plateau.

it susceptibility assessment in the Codrului Hills excludes the subjectivity of the analyst in assessing (North-Western part of Romania). the GIS-assisted spatial Bălteanu. 43-69. O. 1983. considered to be representative for the morphometric variety of the The results of this study allow us to state that Transylvanian Plateau.. More than that. and integration of factors as databases in a GIS 235-241. results in the investigation of other territories with The fundament of the model lays on our firm similar features (in the sense that there is a lack of conviction that morphometry represents a very landform morphometric homogeneity). București. In (extremely important) may be interpreted in a conclusion. București. which manner. A physically based. the researcher’s subjectivity in the procedures of Băcăuanu. model is that.. morphometry.. Editura Academiei Române.I. M... Landslide investigation: analysis of the terrain morphometric features ensures from morphodynamic mapping to hazard a high resolution and a systematic. Society (in Romanian). spatial analysis model. 24. As shown above.. Unlike these methods. the model may be adopted of stability or instability. and their integration as key factors in a It can be concluded that the applied model has methodological approach meant to determine the a very good correlation with the probability classes landslide occurrence probability is fully justified and active landslides and the causative factors scientifically and very useful because of its practical selected are relevant for the models applied. sensitive and true “geoindicator” of the terrain states More importantly. has a double function. Man.. implications. The model is materialized in hectare size raster Bilaşco. D. Cs. has been tested by its validation using the ROC (2). based on statistical and Bălteanu. The Moldavian Plateau-Nature. Application in Buzau Subcarpathians deterministic analysis. The results show a good predictability of the The scientific relevance of the model model and prove its usefulness for the practical proposed in this study starts from the premise that research. 101 . to cartographic or statistical deficiencies.. & Micu. Journal of Earth and Environmental Sciences. Strasbourg. compared to other models (like the one proposed by means of law). & Keller. it has a higher REFERENCES degree of objectivity because all the parameters may be analysed according to the same criteria. Case study: Cornu Village/Romania. playing a part proposed in this approach may be applied with good both in the preparation and triggering the landslides. 1980. 5. 2011. as a working alternative for the methodologies that These states are in fact prefigured by means of are inherently exposed. Editura Științifică și Enciclopedică. 8(3). Editura CERG. A case-study in the Romanian unitary interpretation of the landform defining Subcarpathians: Muscel Catchment. D. Process from Geomorphologic Mapping to The proposed model is based on the analysis Dynamic Modelling. the Someş plateau. & Kirkby MJ. 156 p. the deficient character of the available data or Zeitschrift fur Geomorphologie 55/2. 2009.L. as a defining factor of the terrain All these facts allow us to state that the model dynamic state. most methods (including the 137-144. assessment regarding the importance of that factor. 2011. contributes to the improvement of the methodology We consider that a major advantage of this regarding the prediction of landslide processes. Hydro. An analytic multicriteria account factors which frequently cannot be properly hierarchical approach to assess landslide calibrated due to the inadequate scale of the existing vulnerability. Romanian Journal of The accuracy of the final results of the model Geography. variable contributing area model of basin present different landslide occurrence probabilities. Landslide higher degree of resolution. V. in a more or less obvious the two above-mentioned functions. Case-study: models. E.. Șt..Gh. A. V. Munteanu. Corina. it is our opinion that this model relatively constant (homogeneous) metrics according constitutes an original approach which significantly to the spatial attributes of the landforms. This methodology allows the (in Romanian). database structures that may prove to be extremely I. Filip S... 1979. hydrology. CONCLUSION curve on very large areas. Bull. 209-229. 91-111. Arghiuş. Statistical model using gis for the useful in the development of other spatial analysis assessment of landslide susceptibility. 2013. which proves the viability the analysis of the land morphometric characteristics and veracity of this method. Sci. Editura Academiei Române. determination of the spatial extension of areas which Beven K. & Mihăiescu. Ozunu. one derived from Decree 447/2003) take into Armaș Iuliana. în Landslide characteristics. Roşian. Carpathian the importance of certain factorial characteristics. objective and assessment. Horvath. at a Arghiuș. maps. R.

& Arora. 49. PhD Thesys. 39-56. 1961. Roy. Ş.J. Cardinali. 70(5). Publisher.. 5. 31–47. Kluwer geomorphological.. Mukherjee. 2. accelerated erosion (in Romanian).. Use of GIS Technology in the Prediction susceptibility mapping based on terrain and and Monitoring of Landslide Hazard. A. Ehrlich M. based quantitative method. Franţa. Guzzetti. of Remote relevance of pluvial regime for landslides genesys Sensing. 58. Env. in Hazards. 1998. 2014 Accepted for publication at: 20. 2013. Using Dhakal. nr.K.S. XL. A. 354-357.. teză de doctorat.P. 23 (2). Supplementary Issue *** 2003. nr. A. 2004. Engineering and Remote Sensing.. Nagrajan. Romanian). Landslides Hazard cartographie morphodynamique. Rădoane. Presa Universitară Clujeană. nr. 1999. V. Photogrammetric Hârjoabă. Dordrecht.V. 02. Landslide the cube matrix method for landslide susceptibility hazard mapping and its evaluation using GIS: An assesment. 59-64. 2012. M. Grayson.. 617–625. Geo-information tools for Martiniuc.. Brazil. Maria. Editura Academiei. A. Studii şi investigation of sampling schemes for a grid-cell Cercetări GGG. Landslide nr. 02. Y. 128 p.. 2007. and biological applications. Tufescu. T. 25. V. I. Landslide hazard P..P.. Hydrological Processes. 1995. Rio assessment in the Curvature Carpathians and de Janeiro. Jumar. Surdeanu.. Jour. 2014 102 View publication stats ...J. Academic Publisher. 242-257. M.. 55. Kluwer Academic Geografie.. 20. 1966. S. B. Gupta. malaysia using frequency ratio and Thiery. & Van Westen.. C. 2003. A. May. M. Bull. 2014 Published online at: 25. Strei Valley. F. 2013 Revised at: 10. R. Revista de Geomorfologie. 1995. & Pradhan. 175. Eds. & Ladson. 981–989.D. A. editori Universităţii „Ştefan cel Mare” Suceava. C. 2000. Rădoane. Natural climatic factors for tropical monsoon regions. Secţiunea Carrara. & Moore..B.. (in Szeghed. The (Ganga) valley. Lacerda W. Spatial analysis of geohayard on the rules related to the drafting and content of the Fruška Gora mountain.. Cluj- regarding Baia Mare Depression and Mountains (in Napoca. I. & Ichim. Kheire. universitatea din Caen. developments. Zeitschrift für Fontoura S. Todică. P. 2008. R. & Kanungo. Netherlands. Geomorphical Information susceptibility of Măhăceni Tableland.. 2. nr. 1968. C. A. Guzzetty. S. 1991. 2007. (Alpes de sud.R. at Selangor. Guzzetty. Natural relief modelling and Cluj-Napoca. & Carrara. Seria Geografie. M. 2004. Volume 53. București. Photogrammetric Eng.. Van Westen. I. An integrated Supcarpathians). Onac. Systems in Assessing Natural Hazards. Received at: 08. The Geography of the degraded lands Filip. D. 107-133. Susceptibilite du basin de Barcelonnette logistic regression models. Himalayas. Landslides. Case study: Neamţ County. V. based landslide hazard zonation in the Bhagirathi Dragotă Carmen. S.. 2002. & Reichenbach. & Surdeanu. Government Order no.. & Micu Dana.. Amada. Geomorphologie. S. C-J. 135. Landslide hazard symposium on landslides. Saha. France) aux mouvements de versant: Manea. Micu. Analele Systems in Assessing Natural Hazards.. evaluation & 35.. 2013. 4. Int.. F. 2000. 02. V. Fabbri. analise spatiale et Assessment in the Upper and Middle Sectors of the modelisation probabiliste. Romania. Geol..G. Netherlands. Dordrecht. & Aniya.. F. A. June 28 -July 2 2004. B. Case study Muscel Basin (Buzău Sarkar. Reichenbach. M. remote sensing and GIS. 357–369. R. 2008. Romania. 274 p. Editura Lee. Sayao A.. M. 3-30.. stabilisation: proceedings of the 9th international Micu. & Bălteanu. Landslides and ways of prevention landslide risk assessment: an overview of recent and stabilization (in Romanian). 11. 250 p. Editura Presa Universitară Clujeană. 7. Tutovei Hills Landform (in Romanian). concerning the approval of the methdologycal Minucsér. Landslide hazard mapping Academiei RSR. Carrara. 238 p. V. Eng.R... M. 2009. 4. & Surdeanu. 447 since April.. 2008. Subcarpathians. Romanian) published in Monitorul Oficial.. and evolution. D. 1995. Geomorphical Information Digital terrain modelling: areview of hydrological.. Anul XVII. Cardinali. Chung. A. M. 111-118. 10. N. 403-426.K. Natura. GIS- & Remote Sensing. 33-41. A study of environmental geomorphplogy (in Romanian). University of natural hazards at floods and landslides maps. 66(8). The Multivariate regression analysis for landslide application of GIS for mapping landslide hazard zonation. 14. no.. Present Environment approach for landslide susceptibility mapping using and Sustainable Development. București.