Landslides DOI 10.1007/s10346-012-0348-2 Received: 16 August 2011 Accepted: 17 July 2012 © Springer-Verlag 2012 P. De Vita I E. Napolitano I J. Godt I R. Baum Deterministic estimation of hydrological thresholds for shallow landslide initiation and slope stability models: case study from the Somma-Vesuvius area of southern Italy Abstract Rainfall-induced debris flows involving ash-fall pyroclastic deposits that cover steep mountain slopes surrounding the Somma- Vesuvius volcano are natural events and a source of risk for urban settlements located at footslopes in the area. This paper describes experimental methods and modelling results of shallow landslides that occurred on 5–6 May 1998 in selected areas of the Sarno Mountain Range. Stratigraphical surveys carried out in initiation areas show that ash-fall pyroclastic deposits are discontinuously distributed along slopes, with total thicknesses that vary from a maximum value on slopes inclined less than 30° to near zero thickness on slopes inclined greater than 50°. This distribution of cover thickness influences the stratigraphical setting and leads to downward thinning and the pinching out of pyroclastic horizons. Three engineering geological settings were identified, in which most of the initial landslides that triggered debris flows occurred in May 1998 can be classified as (1) knickpoints, characterised by a downward progressive thinning of the pyroclastic mantle; (2) rocky scarps that abruptly interrupt the pyroclastic mantle; and (3) road cuts in the pyroclastic mantle that occur in a critical range of slope angle. Detailed topographic and stratigraphical surveys coupled with field and laboratory tests were conducted to define geometric, hydraulic and mechanical features of pyroclastic soil horizons in the source areas and to carry out hydrological numerical modelling of hillslopes under different rainfall conditions. The slope stability for three representative cases was calculated considering the real sliding surface of the initial landslides and the pore pressures during the infiltration process. The hydrological modelling of hillslopes demonstrated localised increase of pore pressure, up to saturation, where pyroclastic horizons with higher hydraulic conductivity pinch out and the thickness of pyroclastic mantle reduces or is interrupted. These results lead to the identification of a comprehensive hydrogeomorphological model of susceptibility to initial landslides that links morphological, stratigraphical and hydrological conditions. The calculation of intensities and durations of rainfall necessary for slope instability allowed the identification of deterministic hydrological thresholds that account for uncertainty in properties and observed rainfall intensities. Keywords Ash-fall pyroclastic deposit . Debris flows initiation . Somma-Vesuvius Introduction Rainfall-induced shallow landslides present a significant hazard to community and infrastructure in the Campania region of southern Italy. Here, the mountain ranges that surround the Somma-Vesuvius volcano complex are mantled by ash-fall and pyroclastic deposits that are susceptible to landslides that mobilise into debris flows. For example, on 5–6 May 1998, 159 people were killed by debris flows resulting from heavy rainfall in the Sarno Mountain Range. Debris flows involving pyroclastic deposits represent a relevant societal risk in other areas of the world owing to frequent occurrence of landslides and growing human activity in zones that surround volcanoes (e.g. Capra et al. 2003; Crosta et al. 2005; Paudel et al. 2007; Devoli et al. 2009). Although this type of landslide is generally recognised as a very rapid flow-like mass movement, initiation mechanisms still remain a fundamental question for susceptibility assessment. Different triggering mechanisms have been proposed considering several factors including the critical function of natural morphological features of the slopes as well as artificial modifications. Flow-like mass movements are typically triggered by landslides (Iverson et al. 1997; Jakob and Hungr 2005) that involve small volumes of soil and entrain additional material as they move downhill (Celico and Guadagno 1998; Fiorillo et al. 2001; Guadagno et al. 2005). Often three different phases are recognised. The first is a soil slip (Campbell 1974) or debris slide (Cruden and Varnes 1996) that generally involves a few cubic metres of soils, usually with a gravel content >20 % (debris). The intermediate is debris avalanche (Hungr et al. 2001), in which mobilised material, falling on the downslope nearly saturated soils, provokes liquefaction (Hutchinson and Bhandari 1971; Sassa 1985); this process leads to the increase of the amount of soil involved in the movement and the typical triangular downslope widening shape of landslide scars on open slopes. The third is debris flow that occurs when the flow is channelised by the stream network (Hungr et al. 2001; Revellino et al. 2004). Depending on morphology of the slope and on continuity of pyroclastic cover, the succession of different evolution- ary stages can be different: the first phase (slide) can directly develop into a channelled flow (debris flow) or into an avalanche only (debris avalanche). In some cases, initial slides do not evolve through each of the subsequent phases, stopping their movement along the slope. Among factors that directly control susceptibility and location of the initial triggering areas, morphological discontinuities of slopes, including man-made cuts and fills, have been recognised as predisposing factors (Celico and Guadagno 1998; Guadagno et al. 2005). Types of morphological factors that enhance susceptibility for initial landslides are, for example, natural scarps (both above and below), artificial road cuts and road-fill prisms. Other similar conceptual models have been formulated (e.g. Cascini et al. 2008) considering the role of zero-order basins in relation to bedrock topography in defining the shape and distribution of source areas (Tzukamoto 1973; Dietrich et al. 1986). Several conceptual hydrogeological models have been proposed in order to explain hillslope hydrological processes and to comprehend the cause–effect relationship between heavy rainfall occurrence and initial instability in pyroclastic soils. Some of the models assume the formation of an occasional water table (Celico and Guadagno 1998) or Landslides Original Paper