L'utilizzo dell'interferometria radar satellitare nella caratterizzazione dei fenomeni franosi a differenti scale d'indagine

Slope processes are a many-sided issue, related to many interconnected factors, which need different approaches to be fully understood. Different investigation methods needs different observation scales depending on their limits and data availability. Classical field tools (such as GPS stations, piezometers and inclinometers) and geological surveys represent an effective approach to landslide investigations also allowing a sub-surface knowledge; but data collection start only from the beginning of survey. Air-borne and space-borne remote sensing techniques allow both, small and large scale analysis. Air-borne data are available from the 50 but their acquisitions are discontinuous in time; space-borne SAR (Synthetic Aperture RADAR) data are available from 1992 (ERS-1 mission) until now, providing twenty years of information about land and object displacements but their application in landslide analysis need an in depth study. This research aimed to investigate the potential for satellite A-DInSAR (Advanced Differential Interferometric Synthetic Aperture RADAR) techniques in landslide identification and characterization, in addition to airborne or terrestrial investigations and geological field surveys. I worked mainly on applicability and interpretation of A-DInSAR techniques at different scale: regional scale (Agno Valley), large slope scale (Rovegliana), and single slope scale (Prezzo landslides). The satellite remote sensing techniques called A-DInSAR includes Small Baseline Subset (SBAS) (Berardino et al., 2002) and Persistent Scatterers (PSInSAR) (Ferretti et al., 2001) algorithms. The main advantages of these techniques are: i) availability of 20-years of data and possibility of reconstruct displacements dataset; ii) application on small and large observation scale. The main limits of advanced DInSAR techniques are temporal decorrelation and geometrical distortions. Temporal decorrelation is due to changes of electro-magnetic response of objects with time caused by atmospheric phenomena or anthropic changes or vegetation growth. Satellite look angle of 23.3° (for used ERS and ENVISAT images) and right side-looking acquisitions mode, are the responsible for geometric distortions effects: layover and shadow. These effects, together with aspect and inclination of slopes, need to be taken into account before starting an investigation of mountainous area, because slope instability processes could be located in area affected by layover or shadow effects. Another issue to consider in landslide analysis with A-DInSAR techniques is the measurement of displacement along the radar-target line of sight that provides a 1D displacement velocity. The 3D velocity problem (W-E, N-S, vertical directions) can be solved using both, ascending and descending images. Aiborne remote sensing methods, LIDAR (LIght Detection and Ranging) and optical images photo-interpretation permit to trace the outline of slope instabilities, to classify their characteristics (state of activity, frequency, etc.) and to identify geological and tectonic settings of landslide prone areas. But, their discontinuous acquisitions lead to low precision in displacement measurements. Terrestrial remote sensing techniques (laser scanning, Ground-Based-InSAR, photogrammetry, GPS station) permit large scale investigations, seeking out details of the instability phenomena; they represent a useful complement to conventional field mapping and rock mass discontinuity characterization. These methods are a useful approach because allow i) investigating vertical rock face, which is no detected by satellite InSAR or airborne techniques and ii) the construction of detailed 3D model and 3D displacement data of the phenomena. In contrast they need long and hard-working data collection. In the end, geological-geomorphological and geotechnical field surveys allow to verify and complete the data collected using techniques described before, sometime discovering new important details such as geological settings predisposing to landslide. Regional scale analysis illustrates a feasibility study on the limits of A -DInSAR applicability in landslide phenomena analysis. Layover and shadow maps (LS map) of Agno basin were combined with morphological characteristics (slope and aspect) and land use data. The Agno Valley is located in the NW sector of Vicenza Province. The valley in characterized by gentle relief (only NW part present high relief over 1500 m and slope over 40°), the mean altitude is 600-700 m. The main morphometric features are easterly aspect and slope values between 10° and 30°. Rotational/translational slides and slow flows are the common landslide types. The two LS maps, one for ascending and one for descending acquisition mode were created. These maps show the areas affected by geometric distortions, which are devoid of RADAR information and are unusable for instability processes investigations (17.8% of total Agno Valley area for descending track and 11% for ascending one). Combining ascending and descending LS maps, only 1.3% of Agno Valley is simultaneously affected by LS distortions; therefore these areas cannot be investigated through interferometric techniques (using ERS-1/2 and ENVISAT images with incidence angle of 23.3°). Focusing on relationship between aspect/slope factors and visible area, it has been analysed slope and aspect index (normalized invisible areas divided by normalized visible areas for each class) for both ascending and descending acquisition mode. When the index is greater than one, the number of invisible pixels is greater than the visible ones. The results show that slope values greater than 30° are the main morphometric limit for the application of InSAR techniques, whereas the aspect class that hindered the use of InSAR methods on instability characterisation is the East one for descending orbit and West ones for descending track. Comparing land use data and LS maps, the results show that 44% of visible area fall into woody area, the 11% is part of urban zone and 22% fall within grass land class. Therefore a big percentage of visible pixels fall into "problematicà" land use classes for interferometric processing: woody and grass land areas have low density of PS and SBAS data, due to the sparse presence of scatters. Regarding Italian Landslide Inventory, the 18.2% of mapped landslides are hit by layover effect in descending track, whereas the 10.2% is “not visible” in ascending images and only 1.4% of known instable phenomena is totally “invisible” to both descending and ascending track. These results show a good inclinations of Agno Valley to be investigated by interferometric way. For the large slope analysis, time-series displacement datasets derived from SBAS and PS interferometric processing of ERS and ENVISAT images (ground pixel resolution of 25x25m) were analysed to evaluate the real contribute of these two innovative techniques and to determine the state of activity of landslide phenomena affecting Rovegliana area (North sector of Agno Valley). Several time-consuming interferometric processing were conducted to find out the best interferometric-procedure parameters to process mountainous and vegetated areas such as the test areas. Rovegliana slope is mainly affected by rotational, translational and shallow movements that involve mainly the quaternary deposits and fractured rocks and no displacement data are available. Due to quite large extension of total area affected by landslides (4 Km2), and to high density of vegetation, field detection methods (eg. GPS, laser scanning) and the aerial ones, are expensive and time-consuming. PS and SBAS data, obtained from intererometric processing, were very helpful in this morphologic context. In fact, the presence of scattering houses on the entire slope, facilitated the unwrapping step of interferometric processing, although the presence of vegetated areas. PS data allowed the identification of new landslides, not previously recognized in the Landslide National Inventory. PS data also helped following geological-geomorphological investigations and historical aerial optical images analysis finalised to the delimitation of these new instabilities. Moreover, satellite DInSAR data helped to defined the state of activity and, in some cases, the return time of landslides. The large slope scale case of Rovegliana demonstrated that interferometric techniques can help in the identification and characterisation of the landslide processes in those areas where: i) vegetation is strongly present, ii) the extension of instable area is too large for classical methods (GPS or laser scan) iii) a displacement dataset is absent or deficient, and iv) geologic or morphologic evidence are lacking. At the single slope analysis, Prezzo landslides (Giudicarie Valley, Trento) were analysed via PS and SBAS techniques, LIDAR data and field surveys. The interferometric processing were computed using ERS and ENVISAT ascending images because the LS map of descending images highlights layover effect on Prezzo area. Thus, the capability of interferometric techniques to gain 3D displacement data is null. But the ascending LOS is quasi-parallel to landslide movement, so the assumption that LOS velocity data detect the real displacement can be done and it was supported by the final results. Moreover, ERS ascending dataset has many corrupted and unprocessable images, and available images (19) were no sufficient to compute PS processing. Both, PS (2004-2010) and SBAS (1995-2010) data distributions, show a clear identification of the unstable area on Prezzo village, where the houses represent good scatterers, but outside the village, where trees and grass land prevail, data density is very low. In the case of Prezzo landslide GPS and interferometric data were compared to demonstrate the good capability of PS and SBAS to detect the displacement velocity and to identify different displacement zones. The high resolution of LIDAR data improved the geological, geomorphological and structural field surveys allowing a complete analysis of the landslide process. The parallel use of different remote sensing methods supported by field data, demonstrate that A-DInSAR data can improve a reliable identification and delimitation of landslide areas and the history of its displacements. Satellite remote sensing techniques allowed to overcome the usual drawbacks of conventional field detection and classical monitoring methods (e.g. GPS, laser scanning, optical photo, LIDAR), especially when used over small and medium areas (up to 5-6 km2). Furthermore, the twenty years of interferometric data allow the definition of long displacements dataset permitting to evaluate temporal evolution of instabilities, also when a monitor system is lacking or incomplete. But in some cases, for example high velocity phenomena or steep slope location, terrestrial remote sensing detection is needed to outpace the intrinsic limits of satellite InSAR and air-borne techniques. Therefore a simultaneous use of satellite, airborne, terrestrial and field data can lead to a complete landslide identification and characterizations reducing the costs and the spending-time of the analysis

Nel presente lavoro sono riportati i risultati sulle potenzialità di applicazione e sull'interpretazione dei dati interferometrici, ottenuti attraverso le tecniche avanzate DInSAR (Differential Synthetic Aperture RADAR Interferometry), ai fini dell’analisi della franositàa diverse scale d’indagine. Le immagini SAR dei satelliti ERS 1, ERS 2 ed ENVISAT, sono state processate tramite le tecniche Persistent Scatterers (PS) e Small Baseline Subset (SBAS) utilizzando il software SARscape. Alla scala del bacino È stato eseguito uno studio di fattibilitàper individuare i fattori che limitano la possibilità di ottenere informazioni dal trattamento interferometrico dei dati SAR. L’area analizzata È la Val d’Agno, una valle situata nella parte nord-occidentale della provincia di Vicenza, caratterizzata da pendenze medie comprese tra 20° e 30° e da versanti principalmente orientati verso i quadranti orientali. Le aree affette da distorsioni geometriche (layover e shadow) sono state combinate con i fattori morfometrici (pendenza ed esposizione) e con le carte di uso del suolo, per capire quali fattori limitano l’applicazione delle tecniche interferometriche. I risultati mostrano che le aree del bacino in cui non È possibile ottenere informazioni sono molto limitate e l’acquisizione di dati SAR È ostacolata principalmente da elevati valori di pendenza. Confrontando le frane presenti nel catalogo IFFI (Inventario dei Fenomeni Franosi d’Italia), con le aree in layover, si ottengono risultati favorevoli all’utilizzo delle tecniche DInSAR nell’area della Val d’Agno. Alla scala del versante sono stati analizzati i risultati ottenuti dall'analisi PS ed SBAS delle immagini sia ascendenti che discendenti dei satelliti ERS e d ENVISAT al fine di valutare l’evoluzione temporale e lo stato di attività dei fenomeni franosi nell'area di Rovegliana, situata nella parte nord-orientale della Val d’Agno. Le tecniche interferometriche hanno individuato molti target in movimento, anche se l’area è molto boscata e ciò ha permesso di indentificare nuovi fenomeni franosi, oltre a quelli presenti nel catalogo IFFI. Le informazioni derivanti dai dati interferometrici, congiuntamente all'analisi geologica-geomorfologica e alla fotointerpretazione delle immagini aeree ottiche, hanno agevolato la delimitazione delle instabilità. Le serie degli spostamenti derivanti dai processamenti PS ed SBAS dello stack di immagini SAR ha reso possibile la definizione dei tempi di ritorno di alcune frane. Soprattutto i dati SBAS sono stati utili a questo scopo poichè presentano una minore oscillazione dei valori di spostamento rispetto ai dati PS e quindi permettono l’individuazione di periodi di stasi e di accelerazione del movimento franoso. Alla scala locale, l’area studio di Prezzo (valle delle Giudicaria, Trento), si È dimostrata ideale per l’integrazione delle metodiche DInSAR con i dati LIDAR, la fotointerpretazione ed i rilievi geologici. La frana di Prezzo È un movimento traslazionale studiato dal Servizio Geologico della Provincia di Trento fin dal 1999 con GPS e mire ottiche. Si sono potuti confrontare i dati di spostamento PS con i dati del monitoraggio terrestre scoprendo una quasi perfetta correlazione tra le velocità dei due metodi e validando cosÌ i dati interferometrici. La frana di Prezzo È soggetta a fenomeni di layover in modalità discendente, perciò i dati di velocità calcolati con l’analisi PS si riferiscono a spostamenti lungo la LOS ascendente. La corrispondenza dei valori È dovuta al fatto che il movimento di frana È quasi parallelo alla direzione della LOS e perciò quasi tutto il movimento reale È misurato dall’analisi interferometrica. Questo a sottolineare che la buona riuscita di un’analisi interferometrica differenziale dipende da molti fattori tra cui la relazione tra movimento, angolo di vista del satellite e direzione LOS. Inoltre, gli scatteratori individuati dall’analisi PS, identificano molto bene l’area instabile caratterizzata da velocità comprese tra 20 e 5 mm/anno, da quella stabile (dato sempre confermato dalla rete GPS e da quella ottica). Nella comprensione della dinamica delle aree instabili la frequente mancanza di dati relativi all’evoluzione temporale ed anche spaziale degli spostamenti rappresenta un limite per la loro caratterizzazione. L’analisi interferometrica delle immagini RADAR satellitari offre il grande vantaggio di fornire vent’anni (dal 1992 ad oggi) di dati relativi agli spostamenti delle aree instabili. I limiti imposti dalla stessa tecnica (decorrelazioni temporali e spaziali, misure lungo LOS), dalla morfologia del territorio (aspect e slope) e dalla tipologia di movimento (velocità), ostacolano la sua applicazione, che deve essere valutata di volta in volta, in funzione dei limiti sopra brevemente elencati. Per questo motivo, il dato interferometrico satellitare non sostituisce le altre metodologie di telerilevamento o le indagini in sito, ma opportunamente integrato ad esse, aumenta le potenzialità e le possibilità di una corretta analisi dei processi d’instabilità dei versanti