Inventory of Vietri-Maiori landslides induced by the storm of October 1954 (southern Italy)

On 25 October 1954, a storm hit the area surrounding the villages of Vietri sul Mare and Maiori of the Amal ﬁ Coast (southern Italy) causing more than 300 deaths and severe damage to infrastructures and agriculture. This event has been among the most catastrophic historically documented in Campania Region. On this basis, and considering the lack of an existing complete characterization of the event in terms of triggered slope processes, we used multiple sets of stereoscopic aerial photos and a LiDAR-derived high-resolution topography to produce an event landslide inventory map. Our map provides an overview of the landslide distribution and extent in the area that mainly su ﬀ ered the e ﬀ ect of the storm and is the basis for a morphometric characterization of landslide source areas that we present in this paper as simpli ﬁ ed statistical analysis. In addition, we compared the rainfall distribution with the spatial density of source areas.


Introduction
On October 25th, 1954, a storm hit the area surrounding Vietri sul Mare and Maiori villages of the Amalfi Coast (Campania Region, southern Italy, Figure 1(a)) inducing floods and diffuse landslides that caused more than 300 deaths and severe damage to infrastructures, buildings and local agriculture (e.g. Esposito, Porfido, & Violante, 2003;Lazzari, 1954;Penta, Lupino, Camozza, & Esu, 1954;Tessitore, Di Martire, Martino, & Calcaterra, 2011;Violante, 2009). This event is among the most catastrophic historicallydocumented in Campania Region, causing a total economic loss evaluated in about L45 billions (corresponding to an actual value of about €700 millions), and is among those of highest magnitude in Italy (i.e. rainfall intensity;De Luca, Furcolo, Rossi, Villani, & Vitolo, 2010). It was characterized by a total rainfall higher than 500 mm fallen in approximately 16 h (registered by the Salerno rain gage); the 95% of the total fell in about 7 h (between 19:00 and 2:00 of 25-26 October 1954;S.I.M.N., 1954). It is interesting to note that the storm occurred after a long dry period; this condition possibly influenced its potential as landslide triggers (Fiorillo & Wilson, 2004). This event has a number of historical analogues in Campania Region triggering floods and landslides. Vennari, Santangelo, Parise, and Santo (2016) recognized approximately 500 rainstorms inducing floods between 1540 and the beginning of 2015. A further example is the storm that, in October 2015, hit the central sector of Benevento Province (southern Italy) that caused 2 deaths and diffuse damages (e.g. Guerriero et al., 2018;Magliulo & Cusano, 2016); this event triggered also a number of destructive landslides in association with floods (e.g. Santo, Santangelo, Forte, & De Falco, 2017).
The occurrence of these landslides is controlled by the geologic characteristics and morphology of mountainous systems also influenced by human modifications (e.g. Vitale & Ciarcia, 2018). These reliefs, consisting of complex associations of monoclines that forms high ridges developed in the Apennine (NW-SE) or anti-Apennine direction (NE-SW), are predominantly formed by limestone sequences (Figure 1(b)). Such sequences are mantled by layered air-fall pyroclastic soils (Matano, De Chiara, Ferlisi, & Cascini, 2016), derived by the volcanic activity of Vesuvius and Phlegraean Fields, piled up onto a karstified surface of bedrock (Figure 1(b)). Lazzari (1954) considers bedrock surface asperities generated by karstic processes as fundamental for the keeping of pyroclastic cover on slopes. These pyroclastic covers are generally composed of ashy and pumiceous layers, often overlying buried soils, both with variable physical and mechanical properties (Esposito & Guadagno, 1998). This stratigraphy (and its hydrologic/mechanic significance) as well as the high inclination of hillslopes make them prone toward landsliding. In this geological context, vocated to catastrophic events, the evaluation of landslide susceptibility and hazard is of basic importance for land planning and landslide risk mitigation measurement and strategy adoption Revellino, Guadagno, & Hungr, 2008). Since both susceptibility and hazard evaluation are based on the landslide history of the zone, it is crucial to have a complete knowledge of landslides occurred in the area in terms of location, extent, distribution, typology and possibly age. Most of these data are easily derivable from geomorphological (e.g. Conforti, Pascale, & Sdao, 2015;Conforti, Pascale, Pepe, Sdao, & Sole, 2013;Guerrero et al., 2012;Lazzari & Gioia, 2016;Lazzari, Gioia, & Anzidei, 2018;Murillo-García & Alcántara-Ayala, 2017) and event landslide inventories that assumes dramatic importance especially for deriving reliable information about landslide age and conditions of triggering (e.g. Ardizzone et al., 2012;Fiorillo, Diodato, Meo, & Pagnozzi, 2018).
On this basis and in order to fully document the effects of the October 25th, 1954, storm and, overall, to provide a contribution toward a better knowledge of landslide susceptibility and hazard in the eastern sector of the Amalfi Coast, we prepared an event inventory map of landslides triggered by the storm. In our knowledge, our map completely documents for the first time landslides induced by the storm within the area that mainly suffered its effects. This also as consequence of the short-living landslide trace signature along slopes (few years). Landslide mapping was completed through visual interpretation of historical aerial-photo into a GIS environment and was supported by recently available LiDAR data. On the basis of these data, we made a statistical analysis of morphometric parameters characterizing the identified landslide source areas. Results from this analysis can be considered as a basis of an improved regional landslide susceptibility evaluation. To supplement our analysis, we used available reconstructed rainfall spatial distribution characterizing the storm to make a comparison with spatial density of recognized landslide sources.

Data and mapping methods
The 1:10,000 inventory map of landslides triggered by the October 25th, 1954, storm in the Vietri sul Mare-Maiori area, was made using both 800 dpi negativescans (∼0.4 m ground sample distance) and hardcopy of black and white stereoscopic aerial photos (scale 1:33,000) taken in June 1955 and April 1956 (8 and 18 months after landslides event, respectively). Additional data are recently acquired digital LiDAR topography (1 m single sided pixel dimension, 2012), which were used to derive a 4 m (single sided pixel dimension) shaded relief base for the map.
On the basis of data availability and its processing, the study area can be divided into four different zones where different mapping techniques were used for landslide identification (Figure 1(a)). In zone 1 and 2, landslide mapping was completed into GIS environment through visual interpretation of digital ortho-rectified aerial images (e.g. Keaton & Degraff, 1996). Images taken in 1955 were used to derive the ortophoto for zone 1 while those taken in 1956 were used for zone 2. In zone 3 and zone 4, it was not possible to complete image digital ortho-rectification due to the absence of a complete image coverage immediately outside from the study areas (i.e. polygon 3 and 4 of Figure 1), thus landslide mapping was manually completed through visual interpretation of stereoscopic images using a stereoscope (e.g. Santangelo et al., 2015). Manually mapped landslides were visually transferred to the base map into GIS environment. Landslide transfer was completed considering the topography of the area and the position of landslides mapped at the boundary of zone 1 and 2. It is important to notice that zone 1 and 4 cover most of the Regina Major river basin, which was the area mostly affected by landslides. Zone 2 includes the Bonea basin, which was highly affected as well. Landslide mapping was completed using post-event imagery in the area that mainly suffered the effect of the storm, as reported by historic documents. We assumed that, overall, the mapped landslides were caused by the event of October 1954. This assumption could be reasonable considering the magnitude of the event and typical short living signature of landslide involving the pyroclastic soils mantling slopes of this region.
Landslide description was completed using terminology and classification derived from Hungr et al. (2001). Especially, in our study area landslides are debris avalanches and debris flows. In both cases, landslide initiation is connected with the occurrence of soil slides that materialize the source area (Hungr et al., 2014). In the map, landslides were depicted by polygons without a specific identification of source, transport and deposit areas (Figure 2). This because of their highly variable size that makes not possible to depict these areas for many of them at the map scale. In addition, it was not possible to identify individual landslides where they tend to coalesce, or objectively distinguish debris avalanche deposit from debris flow deposit next channels. This drawback is mainly connected to the difficulty in interpreting the signature of individual landslides from historic aerial imagery and to its local absence due to successive landslides and/or avalanche to flow transition (i.e. cascade effect). This represents a limitation of our dataset (e.g. Marc & Hovius, 2015). To partially overcome this issue, Figure 3 provides an example of possible interpretation of landslide processes within the area covered by Figure 2. In this area, debris avalanches occurring along open slopes and soil-slide induced debris flow occurring along channels are marked with different colors. In this zone, as well as in the study area, landslide source areas have a triangular or trapezoidal shape and can be identified as the upper zone of each single landslide characterized by a length along the landslide axis (i.e. axial length) of few to several meters. This simplification is needed since it is not recognizable from aerial image interpretation.  Once identified, we consistently selected a 6 m buffered circular area around source area apex (i.e. the upper point of the zone, Figure 2), which was used as reference for deriving, from the available LiDAR dem, elevation, acclivity, planar curvature and profile curvature of slopes. In this way, we associated with each apex (i.e. source area) the distribution mode of each parameter considered for the analysis (Figure 2). These data were used as basis to make a simplified statistical analysis consisting in the computation of univariate distribution of morphometric parameters that, being indicative of topographic condition promoting landslide initiation, might guide regional landslide susceptibility estimation. These landslide properties were derived together with landscape properties, in order to allow a simplified comparison. It is clear that for basin or slope scale analysis a further evaluation of the morphometric and stratigraphic singularities (i.e. natural and man-made discontinuities) recognized as responsible for landslide initiation is needed (e.g. . To supplement our analysis, we used the spatial distribution of rainfall reconstructed by the Italian Hydrographic Service (S. I.M.N., 1954) after the event to make a visual-comparative analysis with the spatial density of landslides source areas. The rainfall map was prepared on the basis of data registered by the rain gauge network and local estimates in the Regina Major basin (Figure 1 and inset map of the Main Map). Especially, quantitative data consisted in cumulative rainfall during the event. Landslide source-areas density was calculated as the number of source area per square kilometer. Such calculation was automatically completed into GIS environment.

The landslide inventory map
The Main Map reports the location, extent and distribution of landslides mapped in the study area through GIS-aided and manual visual-interpretation of processed and stereoscopic aerial photos. Inset graphs show major morphometric features of both landscape and landslide source areas in terms of elevation, acclivity, planar and profile curvatures of slopes. The inset map shows the spatial distribution of total rainfall and offers a comparative overview with the spatial density of landslide initiation zones. Overall, we identified more than 1500 landslides of variable dimension (few examples are reported in Figure 4). Most of them occur in the central sector of the Regina Major basin and within the Bonea basin. All of these landslides are typically soil slides evolving mainly into debris avalanches and subordinately into debris flows where they reach creeks. Debris avalanche axial length variates from few to hundreds of meters while debris flows travel for hundreds to thousands of meters before to create fan-shaped deposits. Of particular relevance is the debris flow deposit that reached the beach of Vietri sul Mare, which was responsible for its enlargement and created underflow deposit at the mouth of the Bonea stream (Lazzari, 1954;Violante, 2009).
The morphometric analysis of landslide source areas, represented in the inset graphs of the Main Map, indicates that such areas preferentially occurs along slopes characterized by very low planar and profile curvature and inclination around 40°, but variable between 20°and 60°. The mode of distribution of landslide slope angle is slightly higher than that of the landscape. These morphometric values are consistent with the recognized triggering mechanism characterized by an initial slide evolving mainly into avalanche and subordinately into flow. The elevation is highly variable but characterized by a mode of the fit around 500 m. Additionally, it is observable a secondary peak in the elevation bar diagram around 200 m. Our interpretation is that most of the landslides were triggered by the storm at high elevation (i.e. distribution fit mode) while a number of them occurred at lower elevations next the slope base (secondary peak in the bar graph) due to channel erosion induced by debris flowing.
The inset map of the Main Map shows the distribution of cumulated rainfall in the south-western sector of the Lattari Mountain range where the study area is located. The spatial distribution of rainfall of the Italian Hydrographic Service indicates a rainfall concentration on the Salerno-Vietri sul Mare-Maiori area with a peak higher than 500 mm, as registered by meteorological stations, and an east-west elongation. Within the study area, cumulative rainfall variates between 500 mm of the north-eastern sector to approximately 100 mm of the south-western edge. Most of the area is characterized by rainfall higher than 300 mm and more than 400 mm were registered across the Regina Major and the Bonea basins. This distribution seems to match the spatial distribution of landslide source areas but it does not fully match their spatial density which is maximum across the eastern sector of the central Regina Major basin. This mismatch might be related to the limited number and distribution of the meteorological stations that registered the event across the study area, consequently used for its reconstruction, in the assumption that morphometric parameter distribution is approximately steady across the area.

Conclusions
The Main Map provides an overview of the distribution of landslides triggered by the storm of October 25th, 1954, in the area surrounding the villages of Vietri sul Mare and Maiori of the Amalfi Coast. For what we know, our map represents the first inventory completely covering the area that mainly suffered the effects of the storm. Overall, we identified more than 1500 landslides that commonly triggered as top-soil slides and subsequently evolved into debris avalanches and flows. Our supplementary analysis indicates that most of them occurred along slopes characterized by extremely low planar and profile curvature (i.e. planar slope), which is a common condition for debris avalanche development. Landslide initiation occurs mainly for slope inclination around 40°. Such data identify the general morphometric characteristics that promote landslide initiation on natural slopes of these geological environments, characterized by steep limestone slopes mantled by pyroclastic deposits, and can be considered as a reference in regional landslide susceptibility evaluation perspective. In this way, the significance of our inventory and collateral analysis is also related to the recent rapid-growing of urban areas (last decades) that is increasing the overall risk connected with the occurrence of rainfall-induced landslides and floods in this area. Additionally, our comparison between density of source areas and rainfall distribution seems to underline a specific spatial cause-effect relation, even if the high intensity of the rainfall, conditioned by the local orographic effect, does not allow a more detailed match between rainfall distribution and landslide source areas.

Software
The map was made using Golden Software Map Viewer 8. Dem analysis and rainfall distribution reconstruction was completed using Quantum GIS, Open Source Geographic Information System, licensed under the GNU General Public License. Quantum GIS is an official project of the Open Source Geospatial Foundation (OSGeo, public domain software).