The methodological approach for the structural and landslide recognition was based on an integration of the available topographic, geological and geomorphological information and on data arising from air-photo interpretation (black and white aerial photos from September 1990 at 1 : 33 000 scale), orthophotos dating to July 2008 (1 : 10 000 scale) and Google Earth satellite images dated, respectively, May 2010, April 2011, July 2014, August 2015 and June 2016. In addition, field investigations were carried out from November 2017 to December 2018. The recognized tectonic lineaments and landslides were mapped at a 1 : 10 000 scale. The azimuthal analysis of the collected tectonic data, useful for the recognition of each fault system, was carried out using the GeoRose software.
The inventory of mass movements enclose location, area, perimeter and typology. The latter was classified following the scheme proposed by Cruden and Varnes (1996). In the study area were mapped mainly recent landslides, using several geomorphological criteria based on the morphological freshness of landslides (Keaton and DeGraff, 1996), such as: presence of bare scarps or with poor vegetation cover, depletion and deposition areas well developed, irregular slope profiles, ground cracks, slope ruptures. Finally, the collected data were digitized and stored in a GIS geo-database that allowed the construction of the landslide and fault density maps. The maps were created by means kernel density algorithm with a spatial resolution of 10 m and search area of 1 km2, using Esri ArcGIS 10.1 software. The landslide density (LD) represents the percentage of landslide area occurring in each km2; while the fault density (FD) represents the total length of tectonic lineaments affecting each km2 (Table 1).
Parameters Formula Unit Landslide density (LD) LD=(La/A)×100 (%) Fault density (FD) FD=Fl/A (km-1) Local relief (LR) LR=hmax–hmin (m) Drainage density (DD) DD=Sl/A (km-1) Hypsometric integral (HI) HI=(hmean–hmin)/(hmax–hmin) (-) La. Total area of landslides within unit area; A. unit area of 1 km2; Fl. total length of faults (km); Sl. total length of streams (km); hmax. altitude max (m); hmin. altitude min (m); hmean. altitude mean (m).
Table 1. Mathematical formulas of the parameters used in this study
The morphometric analysis, performed with GIS techniques, was used to calculate some relief features as slope gradient, plan curvature and local relief, which are considered the main parameters influencing landslides occurrence (Scotti et al., 2014; Conforti et al., 2012; Molin et al., 2012; Lee and Min, 2001; Pike and Wilson, 1971). The spatial distribution of these geomorphic attributes was elaborated by a DEM, with a spatial resolution of 5 m, using the Spatial Analyst tools of the ArcGIS software. The used DEM was downloaded from the website of the Centro Cartografico della Regione Calabria (http://geoportale.regione.calabria.it/opendata). The slope gradient and plan curvature maps were automatically derived from DEM. In particular, the plan curvature, which represents the shape of topography, is theoretically defined as the rate of change of slope gradient or aspect in a particular direction (Wilson and Gallant, 2000). The curvature value can be evaluated calculating the value of the radius of curvature of the specific direction and was obtained directly from the derivatives of a topographic surface (Wilson and Gallant, 2000). Plan curvature is described as the curvature of a contour line formed by intersecting a horizontal plane with the topographic surface. The influence of plan curvature on landslides is the control of the surface and subsurface hydrological regime of the slope. Positive (> 0) values of plan curvatures define convexity in the downslope direction, negative (< 0) values of plan curvatures characterize concavity of slope curvature in the downslope direction. Values of plan curvatures around zero indicate that the surface is flat. These three values allowed to classify the plan curvature map in three classes: convex, concave and flat, respectively.
The map of local relief highlights the maximum difference in height for unit area (Table 1). This parameter provides an evaluation of tectonic uplift and fluvial erosive action (Aringoli et al., 2014; Molin et al., 2012; Della Seta et al., 2004). The map, computed from DEM, was produced by subtracting the maximum value of elevation from the minimum one within a moving window sized 1 km2 (Aringoli et al., 2014; Scotti et al., 2014; Molin et al., 2012; Agliardi et al., 2009).
The drainage network was automatically extracted from the DEM, with the support of the Hydrology tool of ArcGIS and following the methodology described by Peckham and Jordan (2007). Subsequently, the drainage lines were hierarchically ordered according to the Strahler (1964) hierarchic scheme. Because the drainage network pattern and its orientation are highly influenced by structural features of the area (Ietto and Perri, 2015; Molin et al., 2012; Capolongo et al., 2005; Della Seta et al., 2004), a statistical azimuthal analysis of stream orientations was carried out using the GeoRose software. This step allowed the assessment of tectonic characteristics on the geometries of the drainage channels. Furthermore, a map of the drainage density was constructed with the aim to observe the texture of the stream network and the degree of dissection of the drainage basin (Molin et al., 2012; Della Seta et al., 2004). The map of the drainage density, defined by the total length of stream channels per unit area (1 km2 in this study, Table 1), was obtained using the line density analysis tool in ArcGIS software (Moglen et al., 1998).
The spatial relationships between tectonic lineaments, morphometric parameters (slope gradient, plan curvature, local relief and drainage density) and landslide occurrence were evaluated by comparing the landslide inventory map with fault density map and each morphometric parameter showed in its specific map. The comparison process was carried out by the zonal statistic function in the ArcGIS spatial analysis toolbox. Therefore, spatial correlations between landslide density map and each selected morphometric parameter were determined using the Pearson correlation coefficient. The correlation was considered statistically significant at p < 0.01. The statistical analyses were done using the software SPSS 19.0.
2.1. Data Collection
2.2. Morphometric Features Analysis
2.2.1. Relief features
2.2.2. Drainage network features
2.3. Relationships between Tectonic Lineaments, Morphometric Parameters and Landslides
The tectonic data were obtained integrating the existing geological information (Cucci and Tertulliani, 2006; Tortorici et al., 2003, 1995; Casmez, 1971) with the interpretation of aerial photos and field survey. The spatial pattern of the tectonic features is showed in Fig. 2a, in which the study area appears to be affected by several fault lineaments. The structural survey mapped 231 tectonic lineaments in the Mesima Basin area, mainly oriented in NE-SW and WNW-ESE directions, as showed by the rose diagram in Fig. 2b. A minor frequency of tectonic lineaments with NNE-SSW trending was identified and mapped as well. Finally, a number of minor faults were recognized with N-S and E-W directions (Fig. 2b). All these fault systems, are linked to the Quaternary tectonic extensional phase that characterize the whole Calabria region (Cucci and Tertulliani, 2006; Tortorici et al., 2003, 1995).
The system of NE-SW trending normal faults is part of a large tectonic structure occurring in the Southern Calabria along the boundary between the uplifted Serre and Aspromonte mountain ranges and the Late Pliocene–Pleistocene basins of Mesima and Gioia Tauro (Pirrotta et al., 2016). This tensive structure played a key control on the morphostructure of the Monte Poro Plateau and of the Capo Vaticano Promontory (Ietto and Bernasconi, 2005). Furthermore, the NE-SW fault system has given rise to catastrophic historical earthquakes with epicentres within the study area, including the ones that occurred in 1 783 with magnitudes greater than 6 (Cucci and Tertulliani, 2006; Jacques et al., 2001; Boschi et al., 1995). Several authors (Bozzano et al., 2011; Keefer, 2002; Chiodo et al., 1999; Cotecchia et al., 1986) argued for a relationship between the high seismicity of the Mesima Basin area and the widespread instability phenomena triggered during the 1 783 seismic sequence.
In the southern sector of the Mesima Basin, the boundary between the Paleozoic crystalline rocks and the Plio-Pleistocene sediments is marked by WNW-ESE tectonic lineaments (Fig. 2). This fault system mainly dips toward SW.
The NE-SW and WNW-ESE extensive tectonic systems affect the morphology of the study area, which is arranged at steps on right and on left sides of the Mesima Graben. Both fault systems produce well-developed ridges and fault escarpments, with triangular and/or trapezoidal facets and juxtapose the Neogene–Quaternary sediments with the underlying Paleozoic crystalline basement. Furthermore, the extensive tectonic systems transversely cut and dislocate several orders of terraced surfaces towards the Mesima Valley (Tortorici et al., 2003).
Starting from the Lower–Middle Pleistocene the extensional processes were accompanied by a strong regional uplift (Antonioli et al., 2009, 2006; Cucci and Tertulliani, 2006). Direct effects of the tectonic uplift produced high local relief with strong erosive energy, narrow and fault-aligned valleys with steep slopes and deepening of the drainage networks (Ietto et al., 2015).
The map of fault density, constructed on the basis of the cumulative length of faults per unit area, is showed in Fig. 3a. The fault density values range from 0 to 2.7 km-1, with an average value of 0.55 km-1. Higher values of fault density (> 1.5 km/km2) were recorded along the horst structure of the Serre Massif and the Capo Vaticano Promontory, both crossed by the NE-SW fault system (Fig. 2). Lower values (< 0.25 km-1) were mainly observed along the alluvial and coastal plain, in the southernmost sector of the Mesima Basin (Fig. 3a).
The maps of the morphometric parameters concerning the slope gradient, the slope curvature and the local relief are given in the Fig. 3. The analysis of the data shows that the Mesima Basin is characterized by a significant geomorphic heterogeneity linked to the lithological and structural characteristics of the area. Mountain and hill landscapes affect about the 80% of the territory, in which the slope gradients range from 0 to 71.2 degrees, with an average value equal to 15 degrees (Fig. 3b).
Landsurfaces falling between 0 and 5 slope gradient value account for the 24% of the total area. Such landsurfaces mainly occur in the lower part of the basin and, subordinately, in the summit part of the flat or gently-sloping highlands. Slopes falling within the 5–10 slope gradient class are also relatively diffused (18% of the total area) and are scattered in the local landscape. More than 27% of the study area displays slope gradients ranging from 10 to 20 degrees; whereas about 31% of the basin is affected by steep slopes with values of slope gradient more than 20° (Fig. 3b). The steep slopes mainly occur along the Serre Massif mountain ridge, dominated by a granitic complex (Graessner et al., 2000 and references therein), which borders the left flank of the basin. Subordinately, slope gradient with values > 30° occurs in the central sector of the basin (Fig. 3b). Usually, they form steep erosional scarps carved in the sand and sandstone deposits.
The local relief map (Fig. 3c) provides a useful tool to show the spatial variation of the fluvial cuts in the study area. The obtained values of local relief range from 0 to 382.9 m, with an average value equal to 146 m. The highest values of local relief are achieved in the left flank of the basin (Fig. 3c), in which lies the area of maximum altitudes (Fig. 1). The lowest values were mainly observed on the floodplain and, secondly, on the gently rolling upland surfaces of the Serre Massif and Capo Vaticano Promontory (Fig. 3c). The spatial distribution of the local relief data, however, shows that high values (> 200 m) are recorded in the areas characterized by hard rocks, by steep slopes (> 20° in average) and by deep-narrow valleys. In addition, comparison analysis between the map of local relief and fault distribution highlights that the highest values of local relief are achieved in the fault-dominated landscapes.
Concerning the morphology of the Mesima Basin landscape, the concave slopes dominate on the 35% of the entire area, whereas 21% of it is affected by slopes with a convex shape (Fig. 3d). Finally, flat curvature characterizes the remaining 44% of the basin area. Concave and planar topographies were mainly found on gentle slopes. Conversely, steep slopes are affected by a convex morphology. Valleys floor are mainly dominated by a concave shape and, subordinately, by a flat morphology. In the areas constituted by clayey lithologies, the slopes morphology is very articulated, where concave-convex shapes are dominant.
The upstream part of the Mesima catchment is a fluvial network made up of deep and narrow incisions, with river gradients up to 9%, excavated in bare rock. The gradient percentages decrease notably only towards the mouth, where the fluvial axes cross the Neogene and Quaternary sedimentary cover that characterize the hilly sub-basins and coastal plain. In these areas the watercourses are wide and the valleys have a floor slightly concave. The geometrical analysis of the Mesima River Basin shows a drainage network well developed with a maximum hierarchy order (Strahler, 1964) equal to 8. This value testifies that the basin has a quite complex configuration (Fig. 4a) because of a strong influence of the topography, lithology and tectonic systems. The hypsometric curve of the entire Mesima Basin, computed on the basis of Strahler's methodology (Strahler, 1964, 1952), shows a concave-convex shape (Fig. 4b). According to several authors (Scotti et al., 2014; Huang and Niemann, 2006; Hurtrez et al., 1999; Willgoose and Hancock, 1998; Pike and Wilson, 1971), this curve profile highlights a strong influence of the tectonic uplift on the geomorphic evolution, causing intense denudation processes on the slopes and a deep dissection of the drainage network (Kusre, 2013; Ciccacci et al., 1992). The hypsometric integral (HI) value of the curve (Table 1), is equal to 0.32 and shows a Monadnock stage (HI < 0.35 in Strahler, 1952).
Figure 4. (a) Drainage network map; (b) hypsometric curve of the Mesima River; (c) main stream parameters.
In the study area the analysis of other quantitative geomorphic parameters proves that the total number of the fluvial axes was 6 949, the total length of the hydrographic network was about 2 656 km (Fig. 4c) and the stream frequency amount to about 8.6 km-1. According to Sreedevi et al. (2009), these values are indicative of a high rate of dissection. Overall, the drainage pattern is sub-dendritic, though in the upstream areas of the catchment an angular and locally sub-parallel pattern is very common in the first- and second-order sub-basins. This kind of pattern arise from faults and fractures affecting the high-competence rocks (Fig. 2a). A trellis-like pattern affects the hilly sub-basins, which are carved in clayey deposits. Instead, a pinnate pattern is common in the lower part of the basin that is dominated by the presence of a number of short and low-order tributaries. Further, the drainage network is less developed in the basin areas dominated by conglomerate deposits.
The drainage density map of the Mesima River Basin displays significant variability with values ranging from 0.8 up to 8.2 km-1 (Fig. 5). The average value of drainage density is 3.3 km-1 and more than 60% of the study area shows values between 3 and 5 km-1. Drainage density values strictly depend by rock erodibility and by slope steepness (Sreedevi et al., 2009; Ciccacci et al., 1992). In the study area, high values (> 5 km-1) of drainage density were collected in the areas constituted by crystalline rocks and clayey deposits, where high values of local relief occur. Contrary, low drainage density values (< 2 km-1) were observed in the alluvial plain and in the areas dominated by conglomerate and sand deposits, which are characterized by high permeability and low values of local relief (Fig. 3d). In according to Radaideh et al. (2016), the spatial distribution of drainage density is also influenced by tectonic lineaments distribution. The comparison between drainage density and fault density (Table 2) showed a direct and meaningful relationship (r=0.48). The analysis also showed that slope gradient distribution and drainage density have a direct and significant relationship (r=0.71) (Table 2).
LD (%) FD (km-1) SG (°) LR (m) DD (km-1) LD (%) 1.00 - - - - FD (km-1) 0.58 1.00 - - - SG (°) 0.87 0.56 1.00 - - LR (m) 0.75 0.39 0.89 1.00 - DD (km-1) 0.66 0.48 0.71 0.62 1.00 Significant at p < 0.01.
Table 2. Pearson's correlation among landslide density (LD), fault density (FD), slope gradient (SG), local relief (LR), and drainage density (DD) analysed in the study area
The fluvial organization is strongly influenced by the structural-geological characters of the study area. Several authors (Ietto et al., 2015; Capolongo et al., 2005; Mayer et al., 2003; Tortorici et al., 2003) asserted that the fluvial axes disposed perpendicular to the axis of the mountain ranges and parallel to fault alignments indicate an evolution of the drainage network controlled primarily by local tectonics and subsequently by regional Quaternary uplift. In the Mesima Basin many geomorphological anomalies, indicating fault activity and tectonic disturbance, were observed (Fig. 4a). Indeed, many fluvial channels run parallel to the direction of the tectonic lineaments and of the main geological structures. Deflected branches, alignment of deflections and variance among river branches were observed in the drainage network as well. Finally, fluvial axes lying along the fault lines are entrenched and show meander-shape. For these reasons, the influence of tectonics on drainage network was investigated through the azimuthal comparison between streams and faults. The rose diagrams in Fig. 6 show the drainage preferential trends from first to eighth hierarchic orders, as following: NE-SW trend is present on all the orders; WNW-ESE trend mainly affected the streams from third to fifth order; NNE-SSW trend is present mostly on the main streams (6th, 7th, 8th order) and, secondly, in the first and second order; N-S trend mostly influences the first and the second order; finally, the E-W fault direction mainly controls the streams belonging to the third and fourth order (Fig. 6).
Figure 6. Rose diagrams of the drainage trend corresponding to the recognized hierarchic orders (from 1st to 8th order).
According to several authors (Pérez-Penã et al., 2010; Ribolini and Spagnolo, 2008; Malik and Mohanty, 2007, and many others), the obtained results show that the azimuthal distribution of higher order of fluvial streams is strongly controlled by local tectonic systems, conversely, a weak control of the fault lineaments was observed on lower order of drainage network. This characteristic suggests that other factors may play a more important role on the azimuthal arrangement. Therefore, lithologies and denudation processes, like landslides and water erosion, can be relevant on the geomorphological evolution of the drainage network. Consequently, in accordance with several authors (Santangelo et al., 2013; Rapisarda, 2009), the role of denudation processes in the low degree of hierarchical organization of the drainage network is highlighted.
The geomorphological analysis provided information about the spatial distribution, geometry and typology of the landslides (Fig. 7a). The results point out that the study area is affected by widespread instability phenomena, which contribute significantly to control the present-day landscape evolution of the Mesima Basin. Many landslides were observed along the deeply carved fluvial valleys or close to the main tectonic lineaments. The latter, cause pre-existing weakness zones in the rock masses encouraging the groundwater drainage and the deep weathering processes, either held responsible of instability processes (Conforti and Ietto, 2019; Borrelli and Gullà, 2017; Liu et al., 2007; Parise et al., 1997; Sorriso-Valvo et al., 1996; Dramis and Sorriso-Valvo, 1994).
The geomorphological survey allowed to map (Fig. 7a) a total of 1 991 landslides (Table 3). The cumulative area affected by all mapped landslides is 63.3 km2, which constitute the 7.9% of the whole study area, with a landslide frequency roughly equal to 2.5 landslide/km2. The size of landslide body ranges from 25.1 m2 to 757 312.8 m2, with an average value of 27 798.8 m2 (Fig. 8a). In particular, Fig. 8b shows that landslide sizes greater than 10 000 m2 are the 50% of the total number of instability phenomena and only 5% exceeds 100 000 m2. The analysis of cumulative area-frequency distribution of the landslides shows that the frequency decreases for the instability phenomena with size less than 200 m2 (Fig. 8b). This result suggest that the smaller landslides are underestimated in our landslide inventory system, because of the low image resolution used for mapping mass movements (Malamud et al., 2004; Stark and Hovius, 2001).
Landslide type N (-) Landslide size (km2) Total landslide area (%) Min Max Mean Fall 37 25.1 65.5×103 17.5 ×103 0.6 1.0 Slide 1 351 63.7 631.1×103 32.8 ×103 44.3 70.1 Flow 341 56.5 228.0×103 20.3 ×103 6.9 10.9 Complex 262 107.3 397.1×103 43.4 ×103 11.4 18.0 Total 1 991 25.1 631.1×103 27.8 ×103 63.2 100
Table 3. Distribution of type, number and area of landslides for the landslide inventory of the Mesima River Basin
Figure 8. (a) Box plot of the landslide area distribution: the lower and upper limits of each box are the 25th and 75th percentiles, the whiskers show the data range, the vertical line in the box represent the median value; (b)frequency-area distribution of the landslide inventory.
The recognized landslides include shallow (depth < 2 m) and deep-seated (depth > 2 m) failures. The deep-seated failures involve the cover materials and/or the bedrock; whereas most of the shallow landslides are triggered in the cover materials of soil or surficial weathered bedrock. Some shallow landslides are also represented by a partial reactivation of pre-existing deep landslides.
According to the Cruden and Varnes (1996) classification scheme, the types of landslide were classified into slide, flow, and complex (Table 3 and Fig. 9). Slide-type landslides, mainly constituted by rotational slides, are the most frequent kind of instability because they represent more of the 66.4% of the mapped landslides (Table 3). Flow-type landslides, constitute about 12.5% of the checked unstable phenomena; whereas 18.3% is the percentage reaches for the complex movements. The complex landslides, consisting of multiple types of landslide, are mainly formed by slide that evolve in flow-type. The fall-type of landslide is very rare and often un-mappable because of its relatively small dimensions. This kind of instability represents only the 2.8% of the mapped landslides and mainly occurs along steep scarps cut in granitic or gneissic rocks and on sandstone deposits, which overlie the highly erodible clayey deposits.
Figure 9. Examples of landslide typologies affecting study area: (a) rock-fall; (b) rock-slide; (c) earth-flow; (d) complex landslide (slide-earth flows).
The map of landslide density portrays the sectors of the study area characterized by a different concentration of instability phenomena (Fig. 7b). The landslide density values ranging from 0.1% to 71%, with mean value of 10.5%. The map shows that high values (> 20%) of the landslide density are achieved on the foothills areas of the Serre Massif, on the mountain slopes of the Aspromonte Massif and along the slopes surrounding the Capo Vaticano Promontory (Fig. 7b). Very high values of landslide density (> 40%) are concentrated along strike-slip faults, where the relative tectonic uplift and the local relief values reach high rates (Ietto and Bernasconi, 2016; Ietto and Perri, 2015; Catalano et al., 2008; Tortorici et al., 2003). On the contrary, the alluvial and coastal plain and the summit part of the flat or gently-sloping highlands are characterized by very low values (< 5%) of landslide density (Fig. 7b). The landslides distribution suggests that the occurrence is not random, but lithology, tectonic setting and morphometric features, play an important role in determining the landslide disposition (Conforti and Ietto, 2019; Roda-Boluda et al., 2018; Borrelli et al., 2014). Indeed, the comparison between the landslide and lithology maps shows that the highest landslide concentration occurs in the silty clay deposits and in the granitic rocks highly tectonized and weathered. In these areas, it was recognized and mapped more than 62% of the total number of landslides (Fig. 10a). Instead, the 17% of mass movements affect slopes carved in the sand and sandstone deposits (Fig. 10a).
Figure 10. Areal distribution of landslides compared to (a) lithology (in which 1. alluvial deposit, 2. eluvial/colluvial deposit, 3. silty clays, 4. conglomerates and sands, 5. sand and sandstones, 6. evaporitic limestones, 7. gneiss, 8. granite); (b) plan curvature.
The landslide index, expressed as the ratio in percentage between the landslide area and the area of each lithology class in the Mesima Basin, shows values equal to 12.9%, 12.4%, and 10.9% for gneissic rocks, silty clay deposits and eluvial/colluvial deposits respectively (Fig. 10a). Lowest values of landslide index were obtained for conglomeratic and sandy deposits (2.2%) and alluvial materials (0.6%). Finally, field observations demonstrated that eluvial/colluvial deposits, which often cover the igneous and metamorphic basement rock, are mainly involved by shallow landslides (such as earth flow), in agreement with several researches in similar contexts (Conforti and Ietto, 2019; Ietto et al., 2017; Calcaterra and Parise, 2005).
The fault density map and the maps regarding the investigated morphometric parameters (slope gradient, plan curvature, local relief and drainage density) were compared with the landslide inventory map (Figs. 10b, 11). The comparisons suggest that the landslide distribution is strongly influenced by local variations of the observed geo-morphometric features. In particular, the comparison between landslide inventory and fault density maps shows that the landslide frequency increases with the growth of fault density (Fig. 11a). The finding shows that more than 50% of the mapped landslides fall in the areas with fault density > 1 km-1. Indeed, several landslide scarps are aligned with the major fault lineaments. Therefore, many landslides are closely related to the local tectonic setting, in agreement with several authors (Conforti and Ietto, 2019; Borrelli and Gullà, 2017; Conforti et al., 2015; Calcaterra and Parise, 2010; Parise et al., 1997; Sorriso-Valvo et al., 1996; Dramis and Sorriso-Valvo, 1994) that claim the role of the faults as responsible for intense fracturing and deformation of the rocks, steep slopes and fluvial undercutting of the slope. In addition, the comparison between landslide and fault density maps shows a positive significant correlation (r=0.58, Table 2). This result testifies that the landslide occurrence increases with the increasing of the fault density. Indeed, in the basin areas constituted by weathered granitic rocks the value of fault density is greater and a higher landslide density occurs (Fig. 7b).
Figure 11. Relative and cumulative frequency distribution of landslides compared to: (a) fault density, (b) slope gradient, (c) local relief and (d) drainage density. Box-plots showing distribution of landslides with respect to fault density and morphometric parameters. The lower and upper limits of each box are the 25th and 75th percentiles, the whiskers show the data range, the horizontal line in each box represents the median and the black point represents the mean values.
Slope gradient carries out a significant influence on landslide occurrence: all things being equal, steeper slopes are more inclined to failure than gentler slopes. Figure 11b shows the relationship between the slope gradient and the distribution of mass movements. The comparison suggests that the landslides frequency reaches high values on the steep topographies. Indeed, more than 85% of landslides trigger on slopes with gradients more than 20°, which characterize only the 31% of the study area. These results infer that slope gradient plays a key role on the control of landslide occurrence (Fig. 11b). Similar results were reported by Conforti and Ietto (2018), Khazai and Sitar (2003) and Wasowski et al. (2002). In addition, the same comparison between slope gradient and landslide occurrence shows that the last one increases up to a slope gradient value of 45°. If this value is exceeded, landslide occurrence generally decreases. However, the frequency of the steeper slope gradient (over 45°) is underrepresented. Fewer instability phenomena occur with the increasing of steepness (Fig. 11b), because the more weathered layers are thinner or absent along the steep slopes (Ietto et al., 2017). A strong spatial correlation was found between landslide density and slope gradient (r=0.87, Table 2). Indeed, the slope angle is the main parameter of the slope stability, because an increase of shear stress occurs with a progressive rise of the slope gradient (Saha et al., 2002).
The comparison between the local relief map and the landslide inventory map shows that more than 80% of the landslides occur in the areas characterized by high values of local relief (> 160 m) (Fig. 11c). The highest landslide concentration was found along slopes with values of local relief between 190 and 240 m. Relatively few landslides (less than 4%) were identified in the areas with moderate or low local relief values (< 100 m) (Fig. 11c), because of the gentle morphologies of the area. Further, the same comparison of maps shows a significant positive Pearson correlation (Table 2).
The analysis of the plan curvature shows a fairly homogeneous distribution of landslides on concave and on flat slopes, with a slight predominance of the first ones (Fig. 10b). The landslides occurrence in the convex slopes is represented by only the 7.3% of the total landslides mapped. The Fig. 10b shows that the landslide index is equal to 7.5% for concave slopes, 6.1% for flat slopes and 2.1% for areas with convex morphology. On concave slopes the dominant types of landslide are mainly complex- and flow-type; whereas on convex slopes the slide-type mass movements are prevalent. In addition, on convex morphologies, characterized by steep slopes, rare phenomena of rock-fall occur as well.
The comparison between landslide occurrence and drainage density is shown in Fig. 11d. The latter, points out that landslide phenomena increase with the growth of the drainage density up to a value of 5 km-1. If this value is exceeded, the landslide occurrence gradually diminishes. The highest value of the landslide frequency is achieved within areas with a drainage density of 4 km-1. With regards to cumulative landslide (Fig. 11d), the analysis highlight that about 65% of landslides occur in areas with drainage density values between 3 and 6 km-1. Finally, the correlation analysis between drainage density and landslide density shows a good relationship (r=0.66) (Table 2). This sharp relationship could be linked to the geomorphologic modification of the area mainly caused by gully erosion and fluvial undercutting, as well as by water saturation of the shallower horizon. These phenomena are able to influence the triggering landslides (Ietto et al., 2017; Conforti et al., 2016; Vennari et al., 2014; Sanchez et al., 2010).
Previous researches regarding relationship between the morphometric features and landslides are very rare in Calabria. For example, Carrara et al. (1977) argued about the comparison of some morphometric attributes of landslides occurring in the Crati and Ferro basins (North Calabria); they declared a sharp relationship between the slope instability and the lithologies outcropping in either basin areas. Other researches carried out in Daunia region (Apulian Apennines, Central Italy) by Mancini et al. (2010) found out that land cover, lithology and slope exposure are strongly related to the occurrences of slope failure. Similar results were obtained by several authors in worldwide researches, such as: in Japan by Oguchi (1997), in France by Ribolini and Spagnolo (2008), in Spain by Pérez-Peña et al. (2010), in Romania by Petrea et al. (2014) and in India by Malik and Mohanty (2007). All these researchers affirmed that the analysis of morphometric features constitutes a key factor for determining the probability of landslide occurrence in an area.