The main objective of my PhD research is to develop an object-oriented, hierarchical and multi-scale geomorphological approach to studying “landslide systems” meaning sets of landslides of different type evolving on the long-term with mutual interaction (sensu Guida et al. 1988, 1995; Coico et al. 2013; Valiante et al. 2016). The proposed approaches aim: 1) to improve the existing or new inventories, defining an object capable of storing both spatial and temporal relations between landslides in a single dataset, avoiding physical data fragmentation and logic inconsistency; 2) to build a robust conceptual model for the practical management of complex arrangements of landslides and their evolution. This work also aims to contribute to the overall theme of landslide hazard assessment and mitigation, focusing on those cases where complex spatio-temporal arrangements of landslides interacts with engineering structures or infrastructures, for better understanding and quantify the interactions at various spatio-temporal scales between engineering works and natural processes. The research has been conducted following three main strategies: 1) a “Top-Down approach” based on morphometric analyses on Digital Elevation Models (DEMs) to find whether a portion of landscape shows a set of “topographic signatures” ascribable to landslide systems; 2) a “Bottom-Up approach” based on the reconstruction of the landslide system through field activities starting from any of the landslides composing the system itself; 3) comparison of the above strategies using a training-target approach on selected case studies significant for different Italian landscapes. The “Top-Down” approach is based on the application of morphometric techniques using Digital Elevation Models, such as Topographic Position Index (TPI) (Weiss 2001; Paron and Vargas 2007; De Reu et al. 2013), useful for the semi-quantitative delineation of main landforms, and Slope – Area Plots (Montgomery and Foufoula-Georgiou 1993; Booth et al. 2013; Tseng et al. 2015), exploited for the estimation of the erosional processes type acting on the slopes, and extended also to gravity-driven processes. Basically, the graphical plot of the topographic steepness as function of the drainage area can be subdivided in four main plot regions or curve segments, each one representing a dominant geomorphic process: I) hillslopes; II) hillslope-to-valley transition; III) debris flow dominated channels or landslides driven channels; IV) alluvial channels. The “Bottom-Up” approach follows the GmIS_UniSA method proposed in Dramis et al. 2011. In the first steps data collected from field activities were stored referring to a symbol-based representation (SGN 1994; APAT 2007; ISPRA 2018) similarly to what has been done by many authors (Gustavsson et al. 2006; Devoto et al. 2012; Miccadei et al. 2012; Del Monte et al. 2016 among the others), in the next steps the original data is extended from the symbol-based to a full-coverage representation. The latter is then reclassified using the proposed object-oriented data model. Such object-oriented data model is based on the assumption that any entity can be represented by exactly one object regardless of its complexity or inner structure (Egenhofer and Frank 1992). Complexity is then handled through the classification process: a real-world feature and its behaviour is described and encapsulated in a class definition, then any operation of simplification or generalization can be performed defining a set of sub-classes and super-classes. Any feature described by a class definition is an object (an instance of that specific class); simplifying, a class is the description of a feature and its behaviour while an object is the feature itself (Atkinson et al. 1990; Chaudhri 1993; Kösters et al. 1997). The described classification process results in a set of classes linked by parent-child relationship (generalized and specialized classes) and sibling relations (classes sharing a common super-class) in a hierarchical structure. Hierarchies are usually exploited to model, and therefore better handling complexity of natural systems; in this perspective a hierarchy is defined as a multi-level or layered system where each level can be decomposed in a number of interrelated subsystems until a non-decomposable elementary subsystem is defined (Simon 1962; Odum and Barrett 2005; Wu 2013). Depending on the objectives of a particular study or analysis, the hierarchical level closer to the study object is called focal level or level 0 which sets the starting point for decompositions (levels -x) or generalizations (levels +x) (Wu 1999). Applying the previous concepts, the basic landslide inventory is built by means of usual techniques, such as field survey, remote sensing, desk studies, etc., then an object-oriented hierarchical model is applied resulting in a hierarchical classification of landslides. The focal level is set at the input inventory containing individual landslides as one object differentiated by type of movement. The proposed model assumes that a “functional interaction” (i.e. dynamic interaction) exists if the condition of spatial and temporal overlap between landslides is verified. This assumption can be evaluated through the integration of two topological models. The Dimensional Extended nine-Intersection Model (DE-9IM) (Egenhofer and Herring 1990a, b; Egenhofer and Franzosa 1991; Clementini et al. 1994) and the Region Connection Calculus (RCC8) (Randell et al. 1992; Cohn et al. 1997). Starting from the focal level, 2 levels of generalization are defined based on the topological relation between landslides: i) if two or more landslides of the same type have a 2-dimensional relation between their interior portion, they can be simplified in a landslide complex object having the same type of movement as the input features; ii) if two or more landslide complexes have a 2-dimensional relation between their interior portion or with the interior of another landslide which is not part of the input complex, they can be simplified in a landslide system object. A level of decomposition has been also implemented describing landslide components. Once derived a landslide system, it is useful to define its Reference Hillslope, meaning the minimal portion of territory in which it is likely to evolve. To address this task Surface Networks can be a valid technique in order to objectively define the minimal portion of the topographic surface in which a gravitational process can develop and evolve. The extraction of Surface Networks from DEMs (Pfaltz 1976; Wolf 1991; Schneider 2003; Rana 2004) is based on the detection of the characteristic features of a surface called critical elements, such as critical points (local minima, local maxima and local saddles) and critical lines (ridgelines, connecting peak and passes, and courselines, connecting pits and saddles). This data structure has been exploited to decrease complexity of topography representing just its “mathematical skeleton” (Guilbert et al. 2016). In order to test these methodologies, three italian case studies have been selected choosing sites with different geological settings and thus landsliding style. The choice of the study areas has been made also picking landslide recently reactivated with a great impact on anthropic activities. The selected case studies are: • Corniolo - Poggio Baldi (FC) along the Bidente River valley; • Roscigno (SA) on the south-western slopes of Mt. Pruno; • North-eastern slope of the Rocca di Sciara relief in the valley of the Northern Imera river, close to Scillato village (PA). The Corniolo – Poggio Baldi case study has been selected for the last reactivation of the Poggio Baldi landslide in March 2010. The movement developed as a rock-wedge slide evolving in a flow-like movement that produced the damming of the Bidente river and the formation of the Corniolo Lake, which is partially still present today. The main geological settings of the area are made of a sandstone-marly flysch with a dip-slope attitude. The case of Roscigno refers the history of the abandoned “Old Roscigno” rural village. This ghost town has been transferred from about sixty years due to landsliding activity and is nowadays part of the Cilento UNESCO - Global Geopark. The village was built on the south-western slope of Mt. Pruno, mainly composed of terrigenous deposits such as calcarenitic-marly flysch, tectonically overlapping a clayey-marly flysch. The main movement affecting the slope is a deep-seated rock slope deformation, on top of which several shallow landslides developed, such as rotational slides and mud flows. The Rocca di Sciara case study has been chosen for the last reactivation of the lower portion of the slope in April 2015. The event caused severe damages to the road network, also involving the Palermo – Catania highway leading to the failure of the Imera viaduct. The geological settings of the slope are made of a dip-slope bedding heterogenous sequence of limestone megabreccias and thick-bedded calcarenites, thin-bedded or laminated calcilutites and clayey flysch. During these three years of research, several survey activities have been performed in order to reconstruct the geological and geomorphological setting of the case studies. All these activities were supported by the object-oriented perspective defined before, allowing objects definition and description directly on the field. Both the “Top-Down” and the “Bottom-Up” strategies have been applied to the case studies. As for the first strategy, the contributing area reclassification shows mainly the hypothetical landslide-related channel as linear features, while the TPI reclassification highlights concave morphologies that can be related to landslides components, such as detachment areas, trenches, counterslopes, and so on. Both these methods can be useful techniques to assess potential landslides affected areas for a better planning of further activities such as field surveys, which are the starting point of the second strategy. Following the data collection, by direct surveys, desk studies or remote sensing, all the information has been rearranged within the object-oriented logic perspective; then, the hierarchical model allows to derive higher rank units, such as landslide complexes and landslide systems. Based on these derived objects, through the integration of Surface Networks it is possible to define the so-called “Reference Hillslope” for each landscape object. Every landslide is characterized not only by its attributes but also by it spatial and temporal relations with the other movements. Coupling this object-oriented hierarchical approach with a temporal characterization of landslide features in the form of “events”, semantically defined, it is possible to build an object-oriented and event-based database capable of storing both spatial and temporal relations between landslides. The Top-Down approach showed some limitation in the recognition of deep-seated movements, while the Bottom-Up approach allowed the automatic reconstruction of the landslide hierarchies starting from the landslide inventory. A landslide system built with an accurate spatio-temporal inventorying of landslides can be a tool for the fast retrieval of useful information such as “how many events affected a slope and how they developed, their magnitude and frequency and how they interacted”. All these data regarding the past and present activity of a slope are the assumption for understanding its most likely evolution, thus, to contribute to the formulation of reactivation scenarios. Moreover, the definition of the “reference hillslope” allow to objectively define the area/volume to be investigated starting from a reference object – a landslide, a landslide complex or a landslide system - , both for the planning of remediation and monitoring activities and as a starting point to search whether a landslide interacts with other geomorphic processes or anthropogenic activities.

Integration of object-oriented modelling and geomorphometric methodologies for the analysis of landslide systems / Valiante, Mario. - (2020 Feb 07).

Integration of object-oriented modelling and geomorphometric methodologies for the analysis of landslide systems

VALIANTE, MARIO
07/02/2020

Abstract

The main objective of my PhD research is to develop an object-oriented, hierarchical and multi-scale geomorphological approach to studying “landslide systems” meaning sets of landslides of different type evolving on the long-term with mutual interaction (sensu Guida et al. 1988, 1995; Coico et al. 2013; Valiante et al. 2016). The proposed approaches aim: 1) to improve the existing or new inventories, defining an object capable of storing both spatial and temporal relations between landslides in a single dataset, avoiding physical data fragmentation and logic inconsistency; 2) to build a robust conceptual model for the practical management of complex arrangements of landslides and their evolution. This work also aims to contribute to the overall theme of landslide hazard assessment and mitigation, focusing on those cases where complex spatio-temporal arrangements of landslides interacts with engineering structures or infrastructures, for better understanding and quantify the interactions at various spatio-temporal scales between engineering works and natural processes. The research has been conducted following three main strategies: 1) a “Top-Down approach” based on morphometric analyses on Digital Elevation Models (DEMs) to find whether a portion of landscape shows a set of “topographic signatures” ascribable to landslide systems; 2) a “Bottom-Up approach” based on the reconstruction of the landslide system through field activities starting from any of the landslides composing the system itself; 3) comparison of the above strategies using a training-target approach on selected case studies significant for different Italian landscapes. The “Top-Down” approach is based on the application of morphometric techniques using Digital Elevation Models, such as Topographic Position Index (TPI) (Weiss 2001; Paron and Vargas 2007; De Reu et al. 2013), useful for the semi-quantitative delineation of main landforms, and Slope – Area Plots (Montgomery and Foufoula-Georgiou 1993; Booth et al. 2013; Tseng et al. 2015), exploited for the estimation of the erosional processes type acting on the slopes, and extended also to gravity-driven processes. Basically, the graphical plot of the topographic steepness as function of the drainage area can be subdivided in four main plot regions or curve segments, each one representing a dominant geomorphic process: I) hillslopes; II) hillslope-to-valley transition; III) debris flow dominated channels or landslides driven channels; IV) alluvial channels. The “Bottom-Up” approach follows the GmIS_UniSA method proposed in Dramis et al. 2011. In the first steps data collected from field activities were stored referring to a symbol-based representation (SGN 1994; APAT 2007; ISPRA 2018) similarly to what has been done by many authors (Gustavsson et al. 2006; Devoto et al. 2012; Miccadei et al. 2012; Del Monte et al. 2016 among the others), in the next steps the original data is extended from the symbol-based to a full-coverage representation. The latter is then reclassified using the proposed object-oriented data model. Such object-oriented data model is based on the assumption that any entity can be represented by exactly one object regardless of its complexity or inner structure (Egenhofer and Frank 1992). Complexity is then handled through the classification process: a real-world feature and its behaviour is described and encapsulated in a class definition, then any operation of simplification or generalization can be performed defining a set of sub-classes and super-classes. Any feature described by a class definition is an object (an instance of that specific class); simplifying, a class is the description of a feature and its behaviour while an object is the feature itself (Atkinson et al. 1990; Chaudhri 1993; Kösters et al. 1997). The described classification process results in a set of classes linked by parent-child relationship (generalized and specialized classes) and sibling relations (classes sharing a common super-class) in a hierarchical structure. Hierarchies are usually exploited to model, and therefore better handling complexity of natural systems; in this perspective a hierarchy is defined as a multi-level or layered system where each level can be decomposed in a number of interrelated subsystems until a non-decomposable elementary subsystem is defined (Simon 1962; Odum and Barrett 2005; Wu 2013). Depending on the objectives of a particular study or analysis, the hierarchical level closer to the study object is called focal level or level 0 which sets the starting point for decompositions (levels -x) or generalizations (levels +x) (Wu 1999). Applying the previous concepts, the basic landslide inventory is built by means of usual techniques, such as field survey, remote sensing, desk studies, etc., then an object-oriented hierarchical model is applied resulting in a hierarchical classification of landslides. The focal level is set at the input inventory containing individual landslides as one object differentiated by type of movement. The proposed model assumes that a “functional interaction” (i.e. dynamic interaction) exists if the condition of spatial and temporal overlap between landslides is verified. This assumption can be evaluated through the integration of two topological models. The Dimensional Extended nine-Intersection Model (DE-9IM) (Egenhofer and Herring 1990a, b; Egenhofer and Franzosa 1991; Clementini et al. 1994) and the Region Connection Calculus (RCC8) (Randell et al. 1992; Cohn et al. 1997). Starting from the focal level, 2 levels of generalization are defined based on the topological relation between landslides: i) if two or more landslides of the same type have a 2-dimensional relation between their interior portion, they can be simplified in a landslide complex object having the same type of movement as the input features; ii) if two or more landslide complexes have a 2-dimensional relation between their interior portion or with the interior of another landslide which is not part of the input complex, they can be simplified in a landslide system object. A level of decomposition has been also implemented describing landslide components. Once derived a landslide system, it is useful to define its Reference Hillslope, meaning the minimal portion of territory in which it is likely to evolve. To address this task Surface Networks can be a valid technique in order to objectively define the minimal portion of the topographic surface in which a gravitational process can develop and evolve. The extraction of Surface Networks from DEMs (Pfaltz 1976; Wolf 1991; Schneider 2003; Rana 2004) is based on the detection of the characteristic features of a surface called critical elements, such as critical points (local minima, local maxima and local saddles) and critical lines (ridgelines, connecting peak and passes, and courselines, connecting pits and saddles). This data structure has been exploited to decrease complexity of topography representing just its “mathematical skeleton” (Guilbert et al. 2016). In order to test these methodologies, three italian case studies have been selected choosing sites with different geological settings and thus landsliding style. The choice of the study areas has been made also picking landslide recently reactivated with a great impact on anthropic activities. The selected case studies are: • Corniolo - Poggio Baldi (FC) along the Bidente River valley; • Roscigno (SA) on the south-western slopes of Mt. Pruno; • North-eastern slope of the Rocca di Sciara relief in the valley of the Northern Imera river, close to Scillato village (PA). The Corniolo – Poggio Baldi case study has been selected for the last reactivation of the Poggio Baldi landslide in March 2010. The movement developed as a rock-wedge slide evolving in a flow-like movement that produced the damming of the Bidente river and the formation of the Corniolo Lake, which is partially still present today. The main geological settings of the area are made of a sandstone-marly flysch with a dip-slope attitude. The case of Roscigno refers the history of the abandoned “Old Roscigno” rural village. This ghost town has been transferred from about sixty years due to landsliding activity and is nowadays part of the Cilento UNESCO - Global Geopark. The village was built on the south-western slope of Mt. Pruno, mainly composed of terrigenous deposits such as calcarenitic-marly flysch, tectonically overlapping a clayey-marly flysch. The main movement affecting the slope is a deep-seated rock slope deformation, on top of which several shallow landslides developed, such as rotational slides and mud flows. The Rocca di Sciara case study has been chosen for the last reactivation of the lower portion of the slope in April 2015. The event caused severe damages to the road network, also involving the Palermo – Catania highway leading to the failure of the Imera viaduct. The geological settings of the slope are made of a dip-slope bedding heterogenous sequence of limestone megabreccias and thick-bedded calcarenites, thin-bedded or laminated calcilutites and clayey flysch. During these three years of research, several survey activities have been performed in order to reconstruct the geological and geomorphological setting of the case studies. All these activities were supported by the object-oriented perspective defined before, allowing objects definition and description directly on the field. Both the “Top-Down” and the “Bottom-Up” strategies have been applied to the case studies. As for the first strategy, the contributing area reclassification shows mainly the hypothetical landslide-related channel as linear features, while the TPI reclassification highlights concave morphologies that can be related to landslides components, such as detachment areas, trenches, counterslopes, and so on. Both these methods can be useful techniques to assess potential landslides affected areas for a better planning of further activities such as field surveys, which are the starting point of the second strategy. Following the data collection, by direct surveys, desk studies or remote sensing, all the information has been rearranged within the object-oriented logic perspective; then, the hierarchical model allows to derive higher rank units, such as landslide complexes and landslide systems. Based on these derived objects, through the integration of Surface Networks it is possible to define the so-called “Reference Hillslope” for each landscape object. Every landslide is characterized not only by its attributes but also by it spatial and temporal relations with the other movements. Coupling this object-oriented hierarchical approach with a temporal characterization of landslide features in the form of “events”, semantically defined, it is possible to build an object-oriented and event-based database capable of storing both spatial and temporal relations between landslides. The Top-Down approach showed some limitation in the recognition of deep-seated movements, while the Bottom-Up approach allowed the automatic reconstruction of the landslide hierarchies starting from the landslide inventory. A landslide system built with an accurate spatio-temporal inventorying of landslides can be a tool for the fast retrieval of useful information such as “how many events affected a slope and how they developed, their magnitude and frequency and how they interacted”. All these data regarding the past and present activity of a slope are the assumption for understanding its most likely evolution, thus, to contribute to the formulation of reactivation scenarios. Moreover, the definition of the “reference hillslope” allow to objectively define the area/volume to be investigated starting from a reference object – a landslide, a landslide complex or a landslide system - , both for the planning of remediation and monitoring activities and as a starting point to search whether a landslide interacts with other geomorphic processes or anthropogenic activities.
7-feb-2020
File allegati a questo prodotto
File Dimensione Formato  
Tesi_dottorato_Valiante.pdf

accesso aperto

Tipologia: Tesi di dottorato
Licenza: Creative commons
Dimensione 14.49 MB
Formato Adobe PDF
14.49 MB Adobe PDF

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11573/1361628
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus ND
  • ???jsp.display-item.citation.isi??? ND
social impact