Central Umbria, IT

1 Assisi-and-Collazzone Umbria-ItalyThe study area is located in Umbria, central Italy, where climate is Mediterranean and rainfall occurs mostly from October to December and from March to May. In the region crop out sedimentary rocks pertaining to the Umbria-Marche stratigraphic sequence, Lias to Eocene in age, overlaid by lake and fluvial deposits, lower Pliocene to Quaternary in age, and by fluvial deposits of Recent age. The structural setting is complex, and results from the superposition of two tectonic phases associated to the formation of the Central Apennines mountain chain. The area is seismically active, and earthquakes with local magnitude ranging between ML 4.6 and ML 5.8 have occurred recently. The last destructive earthquake sequence in the area occurred between May 1997 and April 1998, and was followed by a new sequence in December 2009. Due to the geological and climatic settings, landslides are abundant in the area, and contribute to shape the landscape. More than 80% of the area is covered by vegetation, including 41% forest, 10% grassland, and 42% cultivated areas. The research will focus mainly in two study areas: (1) the Collazzone basin and (2) the town of Assisi.The Collazzone area is located 20 km S of Perugia, and extends for 79 km2 with elevations in the range from 145 m along the Tiber River flood plain to 634 m at Monte di Grutti. In the area, landscape is hilly and lithology and the attitude of bedding planes control the morphology of the slopes. Valleys oriented N–S are shorter, asymmetrical, and parallel to the main direction of the bedding plains, whereas valleys oriented E–W are longer, symmetrical, and mostly perpendicular to the direction of the bedding planes. In the area crop out sedimentary rocks, including: (i) recent fluvial deposits, chiefly along the main valley bottoms, (ii) continental gravel, sand and clay, Plio-Pleistocene in age, (iii) travertine deposits, Pleistocene in age, (iv) layered sandstone and marl in various percentages, Miocene in age, and (v) thinly layered limestone, Lias to Oligocene in age. Mass movements are abundant and frequent in the area, and include shallow soil slides and flows, deep-seated slides and flows, and compound failures. Shallow landslides occur primarily on cultivated or abandoned areas.

Assisi is built along the NW sector of the Monte Subasio, a distinct physiographical feature in central Umbria, and is bounded to the SW by the Valle Umbra plain. Sedimentary rocks crop out in the area, where layered and massive limestone, marl and clay pertaining to the Umbria-Marche stratigraphic sequence, Lias to Eocene in age, are overlaid by lake deposits, lower Pliocene to Quaternary in age, and by fluvial deposits of Recent age. Ivancich is a neighborhood in the Assisi municipality located SE of the mediaeval part or the town. Built mainly in the period 1960 – 1970, the neighborhood is a residential area of one- to three-storied private homes, and hosts the Assisi hospital and a Franciscan convent. In the area, an active deep-seated landslide is present. Geomorphological, geotechnical and topographical investigations have revealed that the Ivancich landslide is an old (ancient) translational slide that involves the debris deposit that covers the bedrock, represented by a politic-sandstone unit. More recent slides have developed inside the old landslide deposit. The recent slope failures have caused damage to roads, and private and public buildings, including the Assisi hospital.

Tested LAMPRE products

1 LIM-Product icon 2 ELIM-Product icon OFF 3 LSMM-Product icon OFF 4 LStats-Product icon OFF 5 3DSDM-Product icon6 LRIM-Product icon OFF

Physiographic settings

Hilly and mountainous landscape with large open valleys and intra mountain basins drained by the Tiber River and its tributaries.

Relevant phenomena

Shallow and deep-seated, rapid to slow moving landslides, in urban and suburban areas.


Umbria was selected as test site for a new procedure for the landslide susceptibility assessment at regional scale. The skill performance of the model in terms of predictions and uncertainties, have been evaluated. The output is a new susceptibility map.

LAMPRE products testing

1 LIM-Product iconWorking in the Collazzone area (Umbria, Italy), LAMPRE has experimented with new methods and techniques for preparing and updating existing LIMs, by exploiting very-high resolution (VHR) stereoscopic satellite images.

LIMs are usually prepared using primarily the visual interpretation of stereoscopic aerial photographs (Figure A) aided by field surveys. This is a time-consuming operation that makes the production of LIMs over large areas difficult. Modern VHR stereoscopic satellite images are comparable in quality and resolution to the traditional, medium-scale aerial photographs, and can be used by expert geomorphologists to detect and map landslides, and to prepare LIMs.LIM figA LIM figB LIM figC

LAMPRE has advanced the capability to exploit stereoscopic satellite imagery to prepare and to update LIMs. For the Collazzone area (Italy), LAMPRE used pairs of very-high resolution (VHR) stereoscopic satellite images taken by the WorldView-2 satellite on 14 April 2014 to recognize and map slow moving landslides that had left faint but recognizable signs on the slopes. This was done by creating an oriented stereoscopic model of the area covered by the images using the Leica Photogrammetry Suite (LPS) ERDAS IMAGE® software and the Rational Polynomial Coefficients provided with the satellite imagery. This oriented stereoscopic model was used in Stereo Analyst ArcGIS® software to obtain a 3D view of the satellite imagery. Planar StereoMirror® System technology was then used to visualize and navigate this 3D scene (Figure B). In the dynamic 3D environment a trained geomorphologist identified and mapped numerous landslides in the study area (Figure C). The 3D information on the individual landslides obtained visually by the interpreter was stored in a GIS database, and used to update an existing multi-temporal LIM for the Collazzone area (Figure D).LIM figD

The production of a LIM, and of seasonal or multi-temporal LIMs, requires the ability of the interpreter to recognize landslides, or portions of the landslides that have left subtle but discernable signs (e.g., topographic or land cover changes) visible in the satellite images. With modern stereoscopic satellite imagery and 3D digital visualization technology, trained investigators can detect and map landslides, reducing the production time of LIMs.

3 LSMM-Product icon OFFLAMPRE has prepared LSMMs for areas ranging from a few to several thousand square kilometres, including the Briga catchment (Messina, Italy), and the Umbria region (Italy). To prepare the LSMMs, LAMPRE adopts a statistical approach to landslide zonation.

Landslides cover 10% of the hills and the mountains of Umbria (Italy). Exploiting an existing landslide inventory map, and morphological, geological and land-cover information in a Geographical Information System (GIS), LAMPRE has prepared a landslide susceptibility model and the associated zonation map shown in Figure A. In the map, the different colours represent areas expected to be prone to (red, orange) or free of (light and dark green) landslides. Figure B shows the uncertainty associated with the landslide susceptibility zonation.LSMM figA LSMM figB

It is important to evaluate the quality of the susceptibility zonation by comparing it to observed landslide inventories. The map in Figure C shows the geographical location of the areas identified correctly and incorrectly as landslides using the susceptibility model. The four-fold plot shows the proportion of the territory that was classified correctly and incorrectly as landslide prone or landslide free. In the plot, correct predictions are: True Positives (TP) where landslides were predicted in the model and observed in reality, and True Negatives (TN) where landslides were neither predicted nor observed. Also shown is the proportion of the region classified incorrectly. False positives (FP) show the proportion of the area predicted as landslide prone where no landslides were observed, and False Negatives (FN) show areas where no landslides were predicted in the model but were observed in reality. Figure D shows a Receiver Operating Characteristic (ROC) plot used to evaluate quantitatively the performance of the landslide susceptibility zonation.LSMM figC LSMM figD

In the ROC plot, the larger the red area under the curve, the better the susceptibility model and the associated terrain zonation. The combination of the susceptibility map (Figure A), the model uncertainty map (Figure B), and measures of the quality of the susceptibility model (Figures C and D), are useful for land planning and management. This combination helps to evaluate the potential impact of landslides, to do landslide early warning, and to construct scenarios of landslide abundance and activity in a changing climate.

4 LStats-Product icon OFFLAMPRE has determined the statistics of landslide areas for a number of existing inventories, including a geomorphological inventory in Umbria and event inventories in Italy and Taiwan, and for a new inventory of sub-marine landslides offshore Israel.The areal extent of individual landslides in a region follows known statistical distributions. However, determining these statistical distributions is not simple, and LAMPRE has developed new software to facilitate this task.LStats FigA LStats FigBFigure A portrays a portion of a typical landslide inventory map, with many small landslides, several landslides of intermediate size, and only a few large landslides.Figure B shows a histogram of the distribution of the area of the landslides in the map. In the chart, each vertical bar portrays the number of the landslides in a category of areas, shown for convenience in logarithmic coordinates.LStats FigC LStats FigD LStats FigEFigure C portrays a Kernel Density Estimation (KDE) of the landslide areas. The small vertical bars along the x-axis show the number and size of the individual landslides.Figure D portrays a more advanced analysis that shows, in log-log coordinates, the probability density function (pdf) of the landslide area. The associated uncertainty around the pdf is shaded in grey. Determining the statistical distribution of landslide areas helps to investigate the variations of landslides in time.Figure E portrays the probability density functions of landslide areas for landslides of different ages, in the same region. The statistics of landslide size (area, volume) is a component of landslide hazard, and mandatory information for the preparation of landslide hazard models and the associated maps and scenarios.
5 3DSDM-Product iconLAMPRE has prepared 3DSDMs for the El Portalet (Huesca, Spain) and for the Ivancich (Assisi, Umbria, Italy) landslides. In both areas, the slow-moving, deep-seated active landslides affect structures and infrastructures, producing significant damage.

To prepare the 3DSDMs, LAMPRE exploited Finite Element Modelling (FEM) to integrate in-situ monitoring data, surface and sub-surface geological, geotechnical and groundwater information, and surface deformation measurements obtained through advanced Differential SAR Interferometry (DInSAR).3DSDM FigDLAMPRE has developed a 3DSDM of the Ivancich (Assisi, Italy) active landslide (Figure D) using space-borne DInSAR measurements and geological, geotechnical and groundwater subsurface information. Inclinometer data were used to define the geometry of the sliding surface of the landslide. LAMPRE has obtained the best-fitting surface deformation field for the landslide through the minimization of the error between the numerical results and the measured landslide displacement rates (Figure D).

LRIM FigB6 LRIM-Product icon OFFLAMPRE has prepared LRIM for the Collazzone area (Umbria, Italy), and for two regions outside of the LAMPRE test sites, including Northridge (California, USA) and Su-Hua (Taiwan).

The LAMPRE LRIM process is shown in Figure A, where the inputs are an existing landslide susceptibility model and map (LSMM), a digital elevation model (DEM), and a road network map. The user defines the number of landslides to be dropped during the triggered landslide event simulation, and then, landslides have a lower or higher probability of being dropped based on each slope unit’s landslide susceptibility. Landslide areas (how many large vs. small) and shapes are randomly selected from already established statistical distributions. A synthetic landslide inventory is thus produced, and this is repeated hundreds of times. The location, shape and area of landslides in each synthetic inventory will be different.LRIM FigA

Figure B shows an example applied to the Collazzone area (Umbria, Italy), with one of the hundreds of synthetic landslide inventory maps created, along with the resultant road blockages (red triangles) over the 153 km of road network. The exact spatial location of both model landslides and road blockages shown are for one given simulation, and will vary for other simulations. The enlargement shows a detail of the LRIM for a portion of the Collazzone area.

Hundreds of these “synthetic” inventories were created and overlaid with the road network maps for the two areas, and the number, size and road network impacts calculated (last part of the process shown in Figure A), including “direct” road blockages (red triangles). By repeating the process many times, and for different scenario inputs, LAMPRE is able to define a range of possible outcomes in terms of road network disruption. This aids LRIM users to consider different scenarios of impact and plan accordingly.

Read 2800 times
More in this category: « Messina Province, IT

Contact us

Fausto Guzzetti 
Consiglio Nazionale delle Ricerche - Istituto di Ricerca per la Protezione Idrogeologica

  • Tel. +39 075 501 4402


GA n°: 312384
Project type: Small collaborative project
Start date: 01/03/2013
Duration: 24 months
Total budget: 2,488 mln. €
EC funding: 1,964 mln. €
Total effort in person-month: 284
Other info: Visit CORDIS



The research leading to these results has received funding from the European Union Seventh Framework Programme (FP7/2007-2013) under grant agreement nº 312384. LAMPRE is managed by the Research Executive Agency (REA)

LAMPRE is on Twitter

LAMPREproject FInal Conference at the Natural Sciences museum in Bruxelles showing LAMPRE products and results #FP7 #Dissemination pic.twitter.com/xddv3voJwW

Wednesday, 25 February 2015 via Twitter Web Client