Landslide susceptibility mapping of SE Serbia using GIS

1 Geological survey of Serbia, Rovinjska 12, Belgrade. E-mail: s.djokanovic@yahoo.com Abstract. Landslides represent a great problem in Serbia. According to current estimates 30–35 % of Serbia is affected by landslides. In this paper a landslide susceptibility analysis is done for SE Serbia. Study area covers 1507 km2. Relief is hilly or mountainous and characterized by high altitude differences. Analysis is done by geographic information system (GIS) and evaluation by analytic hierarchy process (AHP). For susceptibility assessment are used four factors: lithology, slope angle, distance to rivers and distance to faults. The most landslides are formed on slope steepness less than 30°. There is four classes of susceptibility in study area. Zone of very high susceptibility make 63.9 % of the study area. Zone of high susceptibility covers 15.7 % of the study area. The moderate class occupies 37.4% and zone classified as having low susceptibility accounts for 10 % of study area. Final landslide susceptibility map of SE Serbia is satisfactory.


Introduction
Landslides represent a great problem in the countries all over the world. Landslides are common in japan (YoSHImATSu & AbE, 2006), China (DAI & LEE, 2001a), korea (LEE & mIn, 2001), malaysia (LEE & PRADHAm, 2007), Iran (PouRGHASEmI et al., 2012;mAn -SouRI DAnESHvAR, 2014), Tukey (AkGun & buLuT 2007), uSA (WACHAL & HuDAk, 2000), Italy (GuZZETY et al., 2008;PELLICAnI et al., 2014), Austria (ZIEHER et al. 2016) and many others. Landslides endanger human lives, facilities, roads, forests and agricultural land. In the framework of the European Soil Thematic Strategy European union recognized landslides as a soil threat requiring specific strategies for priority area identification, spatial hazard assessment and management (GunTHER et al., 2013). The Strategy considers landslides as one of eight soil threats in Europe. So it is very important to identify areas where landslides can occur in the future (GunTHER et al., 2013).
In order to reduce damage we need to define which areas are susceptible to landslides. Sites that are prone to landslides should therefore be identified in advance to avoid such damage (DAI et al., 2001b). Susceptibility maps show were landslides may occur (CHACon et al., 2006). Landslide susceptibility maps contain information on the type of landslides that might occur and on their spatial likelihood of occurrence (CoRomInAS et al., 2014). For landslide susceptibility analyses detection of the location of landslides is very important (LEE et al., 2001). Lack of data prevents the quantitative determination of the probability of slope failure (CoRomInAS et al., 2014).The susceptibility term is a function of potential landslide occurrence and landslide related factors (LEE et al., 2001). FELL et al. (2008) give definition of landslide susceptibility as a quantitative or qualitative assessment of the classification, volume, area and spatial distribution of landslides which exist or potentially may occur in an area. According to these authors the aim of susceptibility mapping is to include the maximum number of landslides in the highest susceptibility classes.
In this paper is considered the landslide susceptibility in southeast Serbia. For landslides susceptibility is used the regional scale map 1:100 000. Factors considered for analysis are lithology, slope angle, distance to rivers and distance to faults. Landslides susceptibility analysis is done using geographic information systems (GIS) and AHP. For AHP evaluation is used the extension from marinoni. Landslides data used in this paper are taken from basic engineering geological map (sheet bela Palanka) done by Geological survey of Serbia (GZS). Landslides data are collected by field study.

Materials and methods
GIS and AHP are used to map and evaluate landslide susceptibility in SE Serbia. AHP is multicriteria method whose author is SAATY (1980). AHP enables a comparison of factors and determine the weight of each factors using a matrix in which all elements are compared with each other. The comparison values strongly depend on expert judgement and experience. Description of compared factors is shown in Table 1. To reduce the influence of subjectivity and possibility of inconsistencies,

SonjA ĐokAnoVić
Geol. an. Balk. poluos., 2019, 80 (2), 103-116 SAATY (1980) defined the consistency ratio (CR) as follows: CR = CI/RI CI-consistency index, RI-average resulting consistency index, depends on matrix order. If CR < 0.1 the judgments are seemed trust worthy. A CR ³ 0.1 requires revision of the jud gments in the matrix, identifies reasons for inconsistencies and repeats the process of comparing.
Four factors are used in analysis lithology, slope angle, distance to rivers and distance to faults. Each factor is then assigned a number from Saaty scale to gain a weight. Weights of factor mean their relative importance to slope instability in the study area. Then all factors are reclassified into four classes. Classified map is then overlained and we create final map of susceptibility in the study area. Susceptibility map overlay with landslide inventory map in the verification process.

Study area
Study area is located in southeast Serbia and covers an area of 1507 km 2 (Fig. 1). Geomorphologic features are conditioned by lithology and tectonic structure. Lithology influenced the morphometric characteristic (slope angle) and tectonic setting to the existence of larger morphostructures. Relief is characterized by high altitude differences (over In the base of the mountains rivers formed valleys with elevation 200-500 m. The major rivers are južna morava, nišava, Lužnica, koritnička and kutinska. nišava river composite valley is cut in rocks with distinguish resistance to erosion. The highest width is in bela Palanka basin and the smallest width is in Sićevo Gorge. The gorge parts of the canyon have very steep walls up to 500 m high (mEnkovIć, 2011). All rivers drain study area into the južna morava. Climate of study area is moderate continental with element of mountains climate in the highest part. Summers are hot with little rainfall and winters are cold with lots of snow. Autumns are wormer then springs.

Geology
The oldest rocks in the study area are crystalline schists. These are upper Proterozoic, Proterozoic-Cambrian and Cambrian age (vujISIć et al., 1980). These rocks are mostly albite-chlorite-muscovite schists with lenses of albite, actinolite and chloriteepidote schists and muscovite quartzite. Quartz conglomerate and quartzite are Cambrian age. other Paleozoic rocks are Silurian, Devonian, Carboniferous and Permian age. The upper Silurian is preserved in small area in the core of Suva Planina (vujISIć et al., 1980). Schist and meta sandstone of modra Stena are supposed to be Silurian age. Devonian (lower, middle and upper) consists of flysch. Carboniferous rocks unconformably overlain Devonian and pass upward into Permian red sandstone. The study area mostly consist of mesozoic formations (Triassic, jurassic and Cretaceous). Triassic rocks are transgressive over Permian sandstone while jurassic rocks unconformably overlain Permian sandstone or Triassic rocks (vujISIć et al., 1980). The jurassic is composed of calcareous rocks, clastic rocks and flysch. Lower cretaceous is the most widely spreaded mesozoic formations. The upper cretaceous is predominantly developed in nE part of study area. It consists of sedimentary, calcareous and sedimentary-volcanogenic formation. The Tertiary rocks are Paleogene and neogene age. The Paleogene is preserved in nE of study area. These rocks lie transgressively over maastrichtian formations (vujISIć et al., 1980). The koritnik-babušnica basin is represented by oligocene rocks. These lie trans gressively over Cretaceous formations. The neogene rocks are deposited in basins which are formed under the influence of longitudinal dislocations (Zaplanje, bela Palanka, babušnica, Pirot, niš, and Leskovac basin). Tertiary volcanic rocks consist of andesite and andesite tuff and dacite. The Quaternary rocks are developed mainly in nišava, južna morava, koritnička, kutinska and Lužnica valley. Study area mostly belongs to Carpato-bal kanides and the smaller part belongs to Serbian-macedonian massif.

Digital elevation model
The digital elevation model (DEm) was created by digitizing of contours on the topographic map at scale 1:100 000. Equidistant between contours is 100 m. DEm was created in Arc map 10.1 by interpolation. DEm was done with 100 × 100 mand standard deviation 2.5 m (Fig. 2). The slope angle map was derived from DEm.

Landslides susceptibility factors
The relationship between susceptibility to landslides and the factors is complicated and depends heavily on the specific conditions (mA et al., 2013). Data for landslide susceptibility analysis are obtained from geological, engineering-geological and topographical maps. Geological and topogra-phical maps need to be digitize first. All maps are digitized in AutoCAD map 2014 and then export to GIS software Arcmap 10.1. AHP extension from marinoni is used In this paper for landslide susceptibility is used four factors: slope, lithology, distance to faults and distance to rivers. For each factor is done map with different classes..

Slope
Slope angle is important factor for stability. Slope angle is an essential component of slope stability analysis (DAI et al., 2001b). The steeper the slope, the greater the landslide probability. but, not always and not necessary. In lacustrine rocks landslides are formed in slope with angle <15°. Different authors use different values for this factor so there are no unique values for slope angle. TEmESGEn (2001) use intervals of 10 degrees and distinguishes 7 classes. mARjAnovIć et al. (2013) use <5°, 5-10°, 10-15° and >15°. Slope map is obtained from DEm with resolution 100 × 100 m. For study area four classes of slope angle are classified (Fig. 6) less than 5°( class 1), 5-15° (class 2), 15-30° (class 3) and more than 30° (class 4).

Distance to rivers
Rivers and streams are important factor for soil stability (PouRGHASEmI et al., 2012). Landslides may occur as a result of undercutting toe of the slope due to erosion. It happens often in Serbia. Streams are also important because they can cause gully erosion. DAI & LEE (2001a) suggest that the buffer zone should be 50 m and the maximum distance is over 300 m. TEmESGEn (2001)  3. sandstone, conglomerate, shale, siltstone pyroclastic; 4. gravel, sand and clay (deluvium, deluvium-proluvium and proluvium) and lacustrine clay, marl, sand and gravel; 5. schists. For study area is much better buffer zone of 100 m. According to distance from rivers and streams four classes are classified (Fig. 7) less than 100 m (class 4), 100-200 (class 3), 200-300 (class 2) and more than 300 m (class 1).

Distance to faults
Faults are important factor for slope instability. Faults are important because the rock in this zone is cracked and weakened. Fault represents predisposed direction where landslides can occur. Fault information is also used frequently as one of the factors in a statistical assessment (vAn WESTEn et al., 2008). Faults data are taken from engineeringgeological map. Determined and presumed faults are taken into account while photogeological are not considered. In study area there is no active faults. vAn

Landslide data
The knowledge of the landslides in a particular area is expressed by a landslide inventory map which shows the locations and outlines of landslides (CHACIon et al., 2006). Landslide inventory is an inventory of location, classification, volume, activity and date of occurrence of landsliding (FELL et al., 2008). Landslide map is obtained from the latest engineering-geological map. In the engineering-geological map landslides are classified as active or dormant. Landslide map of the study area is shown in Fig. 9. Landslides mapping is made by using topographical maps at scale 1:25 000. Landslides are then modified for scale 1:100 000.

Results and discussion
For landslides susceptibility analysis seven maps are done (DEm, landslides, lithology, slope, rivers and faults). maps are created by AutoCAD map 2014 and exported to the Arcmap 10.1. The landslide map shows 1297 landslides, 139 active and 1158 dormant with average density of 0.9 landslides per km 2 . The total landslide area is 114 km 2 which makes 7.5% of the study area.
The lithological map shows that class 5 (units defined as the most sensitive) covers 34 %, class 4 makes 3.3%, class 3 makes 24%, class 2 occupies 32% and class 1 (the least sensitive) makes 6.7% of the study area. The largest part of study area almost equally consists of lithological classes 5 and 4, class 2 is in the middle, while classes 1 and 4 are represented less than 10%.
In order to compare maps the pairwise comparison matrix is created ( Table 2). As a result of comparing the weights of each map are gained. obtained CR=0.025 means that judgement is consistent. The greatest significance is given to lithology (56.929) and slope (26.427). The least significance is given to distance to rivers (10.552) and distance to faults (6.092).
After pairwise comparison the landslide susceptibility map of SE Serbia is created. This map is then reclassified and four classes of susceptibility are created (Fig. 10) very high (class 1), high (class 2), moderate (class 3) and low (class 4). Zone of the very high susceptibility represents 36.9% of the study area. Zone of the high susceptibility covers 15.7% of the study area. In this zone the chance for landslide development is high. The lithology of zone is diverse and slopes are of variable steepness. The moderate class occupies 37.4% of the study area and zone classified as having low susceptibility accounts for 10% of study area. Fig. 9. Landslide map of study area. Landslide susceptibility map of SE Serbia is generally satisfactory. Steep slopes are favorable to the development of landslides but in study area, these slope are made of strong limestone with no landslide. because of that in this case slope angle has small influence to the development of landslides.

SonjA ĐokAnoVić
verification of susceptibility map is done by overlapping with existing landslides (Fig. 11). Total area of existing landslides is 114 km 2 . About 8.37% of landslides are in class of very high susceptibility, 11.73% of landslides are in high susceptibility, 5.43% of landslides are in moderate class and 5.2% are in low susceptibility class. From here we can conclude that the landslide susceptibility map for study area is satisfying.

Conclusion
Landslide susceptibility map for SE Serbia shown in this paper is a result of selected factors relevant for susceptibility. This is the first time that landslide susceptibility mapping is done for this area of Serbia. The quality of final map depends on the quality and quantity of data we have. Geological map is not suitable for determining lithological composition because it shows stratigraphic approach. For this reason, in this paper, the engineering-geological map was used. The final susceptibility map of study area is satisfying.
GIS is a very powerful tool which allows susceptibility analysis easily and quickly to be performed. Problem is reported due to analysis as a result of used AHP extension. AHP extension from marinoni does not provide possibility of forming a complex tree of comparison with several factors and subfactors. There is also no possibility to compare and evaluate classes within the same factor.