Διερευνώνται οι συνθήκες ευστάθειας σε συγκεκριμένα πρανή επί της Ε.Ο. Πατρών – Τριπόλεως αμέσως μετά τον οικισμό Δίβρη. Για τον σκοπό αυτό, έγινε λεπτομερής μηχανική περιγραφή των ασυνεχειών, ταξινομήσεις βραχομάζας με τα συστήματα RMR και GSI καθώς επίσης και συστηματική δειγματοληψία βραχωδών ασβεστολιθικών τεμαχών μεγάλων διαστάσεων. Τα δείγματα αυτά μεταφέρθηκαν στο Εργαστήριο Τεχνικής Γεωλογίας, για διάτρηση και κοπή, ώστε να διαμορφωθούν κυλινδρικά δοκίμια κατάλληλα για την εκτέλεση εργαστηριακών δοκιμών τριαξονικής φόρτισης, μοναξονικής θλίψης, σκληρότητα Schmidt, σημειακή φόρτιση, διάτμηση ασυνεχειών κ.λπ. Με δεδομένα, τα αποτελέσματα της εργαστηριακής έρευνας αλλά και τις επιτόπου παρατηρήσεις και καταγραφές, εκτιμήθηκαν τα μοντέλα δυνητικών ολισθήσεων στα ασβεστολιθικά πρανή, ενώ με την εφαρμογή μεθόδων οριακής ισορροπίας μελετήθηκε η ευστάθειά τους. Προς την κατεύθυνση αυτή χρησιμοποιήθηκαν τα λογισμικά WEDGE και ROCSTA του Εργαστηρίου Τεχνικής Γεωλογίας καθώς και το Swedge και RocPlane της Rocscience, για τις περιπτώσεις σφηνοειδών και επιπέδων ολισθήσεων αντίστοιχα. Τέλος, έγινε σύγκριση των αποτελεσμάτων προσομοίωσης των παραπάνω λογισμικών και εξήχθησαν ενδιαφέροντα συμπεράσματα σχετικά με την εφαρμογή τους και τις δυνατότητες ρεαλιστικού σχεδιασμού των μέτρων προστασίας. / In this postgraduate work an effort is presented to apply computational methods of Laboratory of Engineering Geology in order to analyze Limestone slope stability in high-risk areas of Patras - Tripolis national road and to design reliable, safe, preventive and protective measures. The geological conditions of the area including the geotectonic evolution of geological Olonou - Pindos zone are described. The discontinuities were examined in detail, while the rock mass quality, was estimated using classification system such as RMR, SMR and GSI systems. The topographic map of the region of interest was outlining with a brief description of georeferencing and GIS software. It is necessary to note separate chapters of the work, were focused on stability and analysis using several software, Rocksta, RocPlane of Rocscience, Wedge, Swedge of Rocscience, Dips. Finally, a comparative overview of these programs and drawing conclusions is included.

H παρούσα εργασία αφορά τη διερεύνηση ευστάθειας των πρανών του ορύγματος στη Χ.Θ. 18+535 –18+832 της Ιόνιας Οδού. Στην περιοχή αυτή προβλέπεται να εκσκαφτεί μονόπλευρο όρυγμα επί της κύριας οδού (αριστερό πρανές), μέγιστου κατακόρυφου ύψους 30m. Στην δεξιά παρειά θα κατασκευαστεί τοίχος από οπλισμένο έδαφος για την αντιστήριξη του αυτοκινητοδρόμου και της βοηθητικής οδού. Στην περιοχή εκτελέστηκαν έξι γεωτρήσεις (BE1-139, BW6-610, BC4-435, BC4-436, BW4-437, BC4-438, BW4-439) στην παρούσα φάση μελέτης και μία γεώτρηση (ΓΓ4) και δύο φρέατα (Φ3 και Φ4) σε προγενέστερο στάδιο έρευνας. Σύμφωνα με τα αποτελέσματα των γεωτεχνικών ερευνών η περιοχή αποτελείται από φλύσχη και πιο συγκεκριμένα από εναλλαγές ψαμμίτη και κροκαλοπαγούς. Οι γεωτρήσεις που εκτελέστηκαν στην περιοχή του ορύγματος έδειξαν επιφανειακά και μέχρι το βάθος του 1.00μ το μανδύα αποσάθρωσης του υποκείμενου κροκαλοπαγούς αποτελούμενου από κροκάλες και αργιλοϊλυώδεις χάλικες με άμμο, χαμηλής πλαστικότητας έως μη πλαστικούς, πολύ πυκνής απόθεσης. Έπειτα από το βάθος αυτό και μέχρι το πέρας των γεωτρήσεων το βραχώδες υπόβαθρο αποτελείται από ψαμμίτες και κροκαλοπαγή σε εναλλαγές. Ο προσανατολισμός των συστημάτων διακλάσεων και κυρίως της στρώσης κρίνεται δυσμενής λόγω δημιουργίας δυνητικών επιπέδων ολίσθησης και σφηνών. Επιπλέον λαμβάνεται υπόψη η επιδεκτικότητα στην αποσάθρωση που παρουσιάζει ο φλύσχης ως υλικό, όταν εκτεθεί σε περιβαλλοντικούς παράγοντες με την πάροδο του χρόνο, σε συνδυασμό με απότομες κλίσεις πρανών. Ο σχεδιασμός των πρανών έγινε με στόχο την επίτευξη αποδεκτών συντελεστών ασφαλείας έναντι επίπεδης ολίσθησης, ολίσθησης σφηνοειδούς τύπου και ανατροπών στις περιοχές όπου τα πρανή διαμορφώνονται σε κερματισμένη βραχόμαζα, αλλά και κυκλικής ολίσθησης όπου τα πρανή θα εκσκαφτούν εντός εδαφοποιημένων υλικών. Επιπλέον αποσκοπεί στην εκσκαφή των πρανών σε ασφαλείς κλίσεις και τον περιορισμό της έκτασης των εκσκαφών για περιβαλλοντικούς κυρίως λόγους. Για την εκσκαφή των ορυγμάτων με κλίση 3:1 προκύπτει πρόβλημα δυνητικών ολισθήσεων και θα πρέπει να κατασκευαστούν μέτρα αντιστήριξης καθώς και μέτρα αποστράγγισης. Τα απαιτούμενα μέτρα αντιστήριξης και αποστράγγισης και προστασίας από διάβρωση και αποσάθρωση αναλύονται και παρουσιάζονται στην παρούσα μελέτη. / -

Les phénomènes d’instabilité générés par les séismes dans les pentes représentent un risque naturel majeur à l’échelle mondiale. Ce risque a, de plus, tendance à croître du fait d’une urbanisation croissante dans des zones à fort aléa, comme en témoignent de nombreux évènements catastrophique à travers le passé. Pour garantir la sécurité des personnes et des biens face à ce type d’aléa, il faut améliorer les prédictions des déplacements générés par les séismes dans les pentes.Les premiers travaux scientifiques visant à corréler les caractéristiques des instabilités dans les pentes aux paramètres sismiques sont de nature empirique et remontent aux années 1980. Ces méthodes ne permettent pas d’expliquer tous les mouvements de terrain observés à travers le monde, notamment lorsque les effets de site modifient sensiblement la distribution des mouvements dans les pentes et génèrent des déplacements plus grands ou plus petits que ceux prédits par les lois empiriques qui négligent les effets de site.La première partie de cette thèse présente une nouvelle base de données mondiale de 277 glissements de terrain d’origine sismique ou non. L’analyse statistique conduite sur les données de cette base a montré que la forme géométrique moyenne des glissements de terrain reste stable lorsque le volume des instabilités augmente.La deuxième partie de la thèse est dédiée à l’analyse de la stabilité du glissement de terrain de Diezma (Espagne) au moyen de trois méthodes : la Méthode de NEWMARK (1965) basée sur le principe de l’équilibre limite en 2D qui suppose des conditions elasto-parfaitement-plastiques ; l’analyse par différences finies dans le domaine temporel avec le code FLAC en 2D et en 3D qui suppose des conditions visco-elasto-parfaitement-plastiques ; et l’analyse par recombinaison modale dans le domaine fréquentiel avec le code CESAR-LCPC en 2D et en 3D qui suppose des conditions visco-élastiques.Situé dans une région sismiquement active, le glissement de terrain de Diezma est susceptible d’être affecté par des séismes. Une analyse des déplacements induits dans ce versant par onze signaux sismiques différents a été menée dans l’objectif d’établir des corrélations entre les paramètres macro-sismiques des scenarios sismiques appliqués et les déformations calculées par les trois méthodes.Les résultats montrent des différences marquées à la fois qualitatives et quantitatives. Les conclusions principales sont : (a) Les déplacements obtenus par les trois méthodes ont des ordres de grandeur différents et peuvent donc conduire à une surestimation ou à une sous-estimation des déplacements ; (b) Les deux méthodes numériques montrent que les déplacements sont fortement conditionnés par la géométrie des modèles ; seule la méthode par différences finies fait apparaître une réponse en termes de déplacements qui dépend du scénario sismique considéré; (c) Les résultats des simulations2D et 3D ne sont pas comparables ; des analyses complémentaires doivent encore être menées pour guider l’utilisateur dans le choix de la méthode la plus appropriée; (d) La Méthode de NEWMARK (1965), dont l’utilisation reste très répandue de nos jours, est tout à fait appropriée à l’étude de cas simples mais elle peut se révéler inexacte lorsque la structure géologique / topographique du versant conduit à un fort effet de site car ce dernier n’est pas pris en compte par cette méthode / Seismically induced slope deformation is a worldwide common phenomenon that poses an increasing and considerable threat to fast expanding urbanization, and a great number of catastrophic events throughout the past attest thereof. For this reason, displacement predictions allowing for proper slope surveillance became a major concern.Early attempts to relate slope failures to seismic parameters are of empirical nature and date back to the 1980s. Although having proven stable, these relations are frequently disturbed by site effects causing outliers in terms of smaller or greater displacements than expected.The first part of this thesis presents a newly build chronological database of 277 globally distributed seismically and non-seismically induced landslides. A comprehensive statistical analysis was conducted on the data of this database with the main result that – statistically seen – the average geometrical shapes of landslides differ only proportionally.The second part of the thesis is dedicated to a comparative slope stability analysis of the Diezma Landslide (Spain) by means of three methods: the limit-equilibrium based NEWMARK-Method (1965) in 2D under elasto-perfectly-plastic conditions; finite-difference analysis in the time domain with the code FLAC in 2D and 3D under visco-elasto-perfectly-plastic conditions; and modal recombination analysis in the frequency domain with the finite-element code CESAR-LCPC in 2D and 3D under visco-elastic conditions.Located in a seismically active region, the Diezma Landslide is likely to be affected by earthquakes and was therefore considered as a suitable model case. A broad analysis of expected displacements was conducted using eleven strong-motion seismic scenarios. Moreover, the thesis searches for potential relations between macro-seismic parameters of the applied earthquake scenarios and the predicted deformations obtained from the three methods.It appeared that – due to the functionality of the methods – results differ quantitatively as well as qualitatively, and so does their suitability. Major findings are: (a) Results from the three methods are of different orders of magnitude and, thus, can easily lead to over- or under-estimations of displacements; (b) Both numerical methods reveal a strong influence of the model-geometry on the predicted displacements, whereas a scenario-dependent slope behavior manifested itself only within the finite difference analysis; (c) The switch from 2D to 3D does not necessarily result in a similar performance in each dimension and results must be critically judged before further use; (d) The NEWMARK-Method (1965) has proven itself once more to be appropriate for first slope assessments but not for sophisticated evaluations of ground response to seismic shaking

Escorregamentos de taludes no Sudeste do Brasil são causados principalmente, pelo efeito da água proveniente das chuvas. Nos últimos anos, vem se incrementando o número de desastres naturais, ao passo são registradas mudanças climáticas, que podem exercer influência na ocorrência de chuvas extremas. Muitas encostas permanecem grande parte do ano com o solo em estado não saturado, porém variações nas condições ambientais podem ocasionar mudanças bruscas da sucção, reduzindo ou até mesmo eliminando-a e gerando pressões neutras positivas. A dissertação apresenta, um estudo sobre a influência das chuvas, consideradas extremas, no processo de infiltração e de este nos eventos de escorregamentos, considerados catastróficos pela sua dimensão, e que causaram prejuízos ambientais, econômicos e sociais no Brasil. Para o estudo foram selecionados dois eventos de escorregamentos translacionais rasos relativamente típicos, considerados catastróficos, um deles aconteceu na região da Serra de Cubatão em janeiro do ano 1985, e o outro na Região Serrana do Rio de Janeiro em janeiro de 2011. Estes dois eventos apresentaram características de precipitações e mecanismos de escorregamentos próprios, que permitem ilustrar os diferentes mecanismos atuantes em cada caso. / Landslide in southeastern Brazil, are mainly caused by the effect of water from rainfall in infiltration process, in recent years has been increasing the number of natural disasters, while climate change that may exercising influence on the occurrence of extreme rainfall are recorded . Many slopes remain a large part of the year with unsaturated soil condition; however, changes in environmental conditions can cause sudden changes of suction, reducing or even deleting it and generate positive pore pressures. This dissertation presents a study about the influence of rainfall, considered extreme in the infiltration process and this one in the events regarded by their size of catastrophic landslides, which caused environmental, economic and social losses in Brazil. For the study were selected two events of shallow translational landslides relatively typical, considered catastrophic, one of them occurred in the Serra de Cubatão region on January 1985 and the other in the mountainous region of Rio de Janeiro on January 2011. Both events exhibit characteristics of rainfall and sliding mechanisms themselves, allowing illustrate the different mechanisms active in each case.

This thesis contains two main parts. The first part presents a database compiling 137 landslide-triggering earthquakes (LTEs) worldwide, with magnitudes greater than the minimum observed threshold for causing landslides (M4.5), for the period of 1998 -2009. Our data sources include a comprehensive review of the existing literature on earthquake-triggered landslides (ETLs), and also a USGS-based earthquake catalog (PAGER-CAT) that contains information on earthquake-triggered secondary events. Only 14 earthquakes out of the 137 seismic events induced significant numbers of landslides (>250). We compared the number of ETLs with the total number of earthquakes with M ≥ 4.5 (n=68,734) during the same period of time. The results show that only 0.2 % of ETLs and only 4.5% of earthquakes of M > 6 resulted in landslide. In addition, we compiled a database of 37 large-scale landslides, involving initial failure volumes of greater than 20 Mm3 that occurred worldwide between 1900 to 2010. The database contains large-scale earthquake-triggered (n ETLs=18) and non-earthquake-triggered landslides (n NETLs=20), i.e., ca. 50% of large-scale landslides were induced by seismic activity. Surprisingly, the volume-temporal frequency curves of ETLs and NETLs show almost identical slopes and intercepts. Thus, for a given volume, the annual frequency of ETLs is almost identical to that of NETLs in the 110 year period. In contrast to previous studies, this thesis found that the volume of the largest landslide triggered by a given landslide-triggering earthquake is not a function of earthquake magnitude. Peak ground motions (PGA, PGV, and PSA) were calculated for the 18 large-scale ETLs at the site of each occurrence and the resulting values show a correlation with the volume of landslides below the threshold of ca. 80 Mm3. Above this threshold, the relationship between peak ground motions and ETL volume shows complex and nonlinear behavior. The results suggest that 1) other special conditions are required for significant earthquake-triggered landslides to occur, and 2) that very large earthquake-triggered landslides (volume greater than 80 Mm3) result from complex progressive failure mechanisms initiated by seismic shaking (i.e., above this threshold volume, landslide volume is independent of PGA, PGV, and PSA). A detailed analysis of the two 1985 Nahanni earthquakes and the North Nahanni rockslide triggered by the first main shock is carried out in the second part of the study. The North Nahanni rockslide, Northwest Territories, Canada was triggered by the earthquake of M=6.6 on October 5th, 1985. The slide occurred in a Palaeozoic carbonate sequence along a thrust fault, which partly follows bedding and partly cuts across bedding. The sliding surface within the limestone consisted of two planes; the lower plane dipped at 20° while the upper plane dipped at 35°. Slope stability analysis is performed using discontinuum numerical modeling. Static slope stability analyses indicate that the sliding rock was marginally safe for the sliding surface friction angles of 24o or higher. Dynamic analyses of the co-seismic movements are conducted by applying a series of sinusoidal waves to the base of the model. The amplitudes of the October earthquake's seismic waves are estimated using strong motion data available from the second main shock. The results, from the dynamic analysis indicate that the slope becomes unstable for given seismic inputs at a specific range of friction angles (24o to 30o) for the sliding surface and the deformation behavior of the North Nahanni rock masses is dependent on the frequency of the seismic signals. Because the static slope stability analysis showed that the slope was close to instability prior to the seismic shaking, we suggest that the 1985 Nahanni earthquake operated as a trigger event that accelerated the occurrence of the slide. This finding supports our earlier results of the global scale study, which showed that the triggering event does not change the general trend of the frequency-volume distribution of landslides; however, it can accelerate the occurrence of slope failure.

Abstract A small opencast coal mine has been developed over previous underground workings in the Malvern Hills, inland Canterbury, New Zealand. The coal measure strata dip at ~45° to the southeast, and consist of finely laminated mudrocks with multiple coal seams of varying thickness. Production is in the range 10,000 to 15,000 tonnes per annum from two principal seams with an aggregate thickness of ~4.5m. The open pit has been designed with footwall batters parallel to bedding, vertical bench separation of 15m, and the highwall formed to a nominal 4V:1H. Preliminary examination of the open pit mine site in 2003 indicated that footwall failures involved de-lamination due to drying out on exposure, and buckling and/or shearing along bedding surfaces. During mine development it became apparent that the batters formed easily where thin (less than 0.3m thick) coal seams were present in the sequence. In the 2004 campaign the pit floor was lowered, with a new batter and bench formed to expose the 3m thick Main Seam coal. The day after completion of this batter, a large buckle failure occurred involving the entire length of the pit (85m along strike), and a 2m thick intact slab with a total volume of ~3700m³ translated down dip 6.2m on the base of a thin coal seam to form a pronounced buckle at the toe. Even though footwall batters are cut to the angle of dip, which is entirely realistic geotechnically, the de-coupling and buckling that occurred compromised the safety and economics of the whole operation. Buckling failure in moderately dipping soft rock sequences has been identified in footwall slopes of coal mining operations. Models used in the literature to simulate similar footwall failures include: the Euler solution using column and beam buckling theory to calculate the kinematic feasibility of a slab-buckle, conceptual modelling using a base friction table, and numerical modelling using distinct element analysis. Back analysis of the Malvern Hills failure was necessary to investigate the controls on the footwall stability, and for future mine design. Engineering geological description of the pit and slab materials was done, and an engineering geological model created. Samples of the slab material and failure surface were collected by coring and trenching, with testing of these materials to establish the required parameters for use in the Euler solution. Back analysis using three different forms of the Euler solution provided unrealistic results that overestimated the overall length of a stable slope by more than 10 times. An engineering geology reassessment was undertaken, and a number of inadequacies in the Euler solution methodology were identified particularly in relation to pore pressure and elasticity considerations. Given that the Malvern Hills toe-buckle slab failure displays both elastic and plastic deformation components in the soft mudrocks, and the slab itself cannot be considered as homogenous, reservations must exist about conventional predictive analytical techniques for pit slope failures of this type. No further large scale slab-buckle failures have developed at the mine site, in part because of the slow rate of coal extraction, but precautionary drainage of the footwall slopes has been undertaken to improve overall batter stability. The location of the slab-buckle failure on a critically positioned pre-sheared thin coal seam with full hydrostatic head is considered the most probable cause, rather than inherent instability of the generic bench and batter arrangement adopted. The adoption of a precedent based engineering geology approach to future mine design is considered the most appropriate solution in the circumstances.

This thesis contains two main parts. The first part presents a database compiling 137 landslide-triggering earthquakes (LTEs) worldwide, with magnitudes greater than the minimum observed threshold for causing landslides (M4.5), for the period of 1998 -2009. Our data sources include a comprehensive review of the existing literature on earthquake-triggered landslides (ETLs), and also a USGS-based earthquake catalog (PAGER-CAT) that contains information on earthquake-triggered secondary events. Only 14 earthquakes out of the 137 seismic events induced significant numbers of landslides (>250). We compared the number of ETLs with the total number of earthquakes with M ≥ 4.5 (n=68,734) during the same period of time. The results show that only 0.2 % of ETLs and only 4.5% of earthquakes of M > 6 resulted in landslide. In addition, we compiled a database of 37 large-scale landslides, involving initial failure volumes of greater than 20 Mm3 that occurred worldwide between 1900 to 2010. The database contains large-scale earthquake-triggered (n ETLs=18) and non-earthquake-triggered landslides (n NETLs=20), i.e., ca. 50% of large-scale landslides were induced by seismic activity. Surprisingly, the volume-temporal frequency curves of ETLs and NETLs show almost identical slopes and intercepts. Thus, for a given volume, the annual frequency of ETLs is almost identical to that of NETLs in the 110 year period. In contrast to previous studies, this thesis found that the volume of the largest landslide triggered by a given landslide-triggering earthquake is not a function of earthquake magnitude. Peak ground motions (PGA, PGV, and PSA) were calculated for the 18 large-scale ETLs at the site of each occurrence and the resulting values show a correlation with the volume of landslides below the threshold of ca. 80 Mm3. Above this threshold, the relationship between peak ground motions and ETL volume shows complex and nonlinear behavior. The results suggest that 1) other special conditions are required for significant earthquake-triggered landslides to occur, and 2) that very large earthquake-triggered landslides (volume greater than 80 Mm3) result from complex progressive failure mechanisms initiated by seismic shaking (i.e., above this threshold volume, landslide volume is independent of PGA, PGV, and PSA). A detailed analysis of the two 1985 Nahanni earthquakes and the North Nahanni rockslide triggered by the first main shock is carried out in the second part of the study. The North Nahanni rockslide, Northwest Territories, Canada was triggered by the earthquake of M=6.6 on October 5th, 1985. The slide occurred in a Palaeozoic carbonate sequence along a thrust fault, which partly follows bedding and partly cuts across bedding. The sliding surface within the limestone consisted of two planes; the lower plane dipped at 20° while the upper plane dipped at 35°. Slope stability analysis is performed using discontinuum numerical modeling. Static slope stability analyses indicate that the sliding rock was marginally safe for the sliding surface friction angles of 24o or higher. Dynamic analyses of the co-seismic movements are conducted by applying a series of sinusoidal waves to the base of the model. The amplitudes of the October earthquake's seismic waves are estimated using strong motion data available from the second main shock. The results, from the dynamic analysis indicate that the slope becomes unstable for given seismic inputs at a specific range of friction angles (24o to 30o) for the sliding surface and the deformation behavior of the North Nahanni rock masses is dependent on the frequency of the seismic signals. Because the static slope stability analysis showed that the slope was close to instability prior to the seismic shaking, we suggest that the 1985 Nahanni earthquake operated as a trigger event that accelerated the occurrence of the slide. This finding supports our earlier results of the global scale study, which showed that the triggering event does not change the general trend of the frequency-volume distribution of landslides; however, it can accelerate the occurrence of slope failure.

Due to rapidly growing population of Ankara city (Turkey) and traffic load, it is required to widen some of the existing highways. One of them is Ankara-EskiSehir (E-90) highway that connects highly populated areas to the city center. During widening, several cut slopes were formed along the highway route. However, some instability problems such as small-sized rock falls and rock detachments have occurred along a cut slope in highly jointed, folded and sheared limestone. They caused local degradation of the cut slope. The cut slope has a slope angle varying from 71&deg / to 84&deg / and contains several shear zones. In this study, the relationships between the existing detachment zones and various parameters (e.g. block size, point load strength index, weathering, shear zone, daylight zone) considered to be important for slope instability were investigated using GIS-based statistical landslide susceptibility analyses in order to predict the further aerial extension of the detachment zones with time. During the overlay analyses, statistical index and weighting factor methods were used by means of TNT-MIPS software. The outcomes of the analyses using both methods are compared and evaluated together with the field observations to check the reliability of the methods and to assess the detachment zones that may develop in the future. Additionally, limit equilibrium analyses were also carried out for the determination of the possible large scale mass failures. The overlay analyses indicate some risky zones where detachments are likely to occur in the future. On the other hand, the limit equilibrium analysis of the rock mass using Bishop simplified method shows that except one section no mass failure is expected in the cut slope. Suitable remediation measures which include the use of wire mesh, shotcrete, toe support, and concrete barrier blocks or catch/barrier fences are recommended for these zones.

It is well recognised that the inherent soil variability and the effect of vegetation, in particular the effect of tree root reinforcement, have a significant effect on the stability of a natural slope. However, in practice, these factors are not commonly considered in routine slope stability analysis. This is due mainly to the fact that the effects of soil variability and vegetation are complex and difficult to quantify. Furthermore, the available slope stability analysis computer programs used in practice, which adopt conventional limit equilibrium methods, are unable to consider these factors. To predict the stability of a natural slope more accurately, especially the marginally stable one, the effects of soil variability and vegetation needs to be taken into account. The research presented in this thesis focuses on investigating and quantifying the effects of soil variability and vegetation on the stability of natural slopes. The random finite element method (RFEM), developed by Griffiths and Fenton (2004), is adopted to model the effect of soil variability on slope stability. The soil variability is quantified by the parameters called the coefficient of variation (COV) and scale of fluctuation (SOF), while the safety of a slope is assessed using probability of failure. In this research, extensive parametric studies are conducted, using the RFEM, to investigate the influence of COV and SOF on the probability of failure of a cohesive slope (i.e. undrained clay slope) with different geometries. Probabilistic stability charts are then developed using the results obtained from the parametric studies. These charts can be used for a preliminary assessment of the probability of failure of a spatially random cohesive slope. In addition, the effect of soil variability on c'–ϕ' slopes is also studied. The available RFEM computer program (i.e. rslope2d) is limited to analysing slopes with single-layered soil profile. Therefore, in this research, this computer program is modified to analyse slopes with two-layered soil profiles. The modified program is then used to investigate the effect of soil variability on two-layered spatially random cohesive slopes. It has been demonstrated that the spatial variability of soil variability has a significant effect on the reliability of both single and two-layered soil slopes. Artificial neural networks (ANNs), which are a powerful data-mapping tool for determining the relationship between a set of input and output variables, are used in an attempt to predict the probability of failure of a spatially random cohesive slope. The aim is to provide an alternative tool to the RFEM and the developed probabilistic stability charts because the RFEM analyses are computationally intensive and time consuming. The results obtained from the parametric studies of a spatially random cohesive slope are used as the database for the ANN model development. It has been demonstrated that the ANN models developed in this research are capable of predicting the probability of failure of a spatially random cohesive slope with high accuracy. The developed ANN models are then transformed into relatively simple formulae for direct application in practice. The effect of root reinforcement caused by vegetation is modelled as additional cohesion to the soils, known as root cohesion, cr. The areas affected by tree roots (i.e. root zone) are incorporated in the finite element slope stability model. The extent of the root zone is defined by the depth of root zone, hr. Parametric studies are conducted and the results are used to develop a set of stability charts that can be used to assess the contribution of root reinforcement on slope stability. Furthermore, ANN models and formulae are also developed based on the results obtained from the parametric studies. It has been demonstrated that the factor of safety of a slope increase linearly with the values cr and hr, and the contribution of root reinforcement to a marginally stable slope is significant. In addition, probabilistic slope stability analysis considering both the variability of the soils and root cohesion are conducted using the modified RFEM computer program. It has been demonstrated that the spatial variability of root cohesion has a significant effect on the probability of slope failure. / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1349971 / Thesis (Ph.D.) - University of Adelaide, School of Civil, Environmental and Mining Engineering, 2009

It is well recognised that the inherent soil variability and the effect of vegetation, in particular the effect of tree root reinforcement, have a significant effect on the stability of a natural slope. However, in practice, these factors are not commonly considered in routine slope stability analysis. This is due mainly to the fact that the effects of soil variability and vegetation are complex and difficult to quantify. Furthermore, the available slope stability analysis computer programs used in practice, which adopt conventional limit equilibrium methods, are unable to consider these factors. To predict the stability of a natural slope more accurately, especially the marginally stable one, the effects of soil variability and vegetation needs to be taken into account. The research presented in this thesis focuses on investigating and quantifying the effects of soil variability and vegetation on the stability of natural slopes. The random finite element method (RFEM), developed by Griffiths and Fenton (2004), is adopted to model the effect of soil variability on slope stability. The soil variability is quantified by the parameters called the coefficient of variation (COV) and scale of fluctuation (SOF), while the safety of a slope is assessed using probability of failure. In this research, extensive parametric studies are conducted, using the RFEM, to investigate the influence of COV and SOF on the probability of failure of a cohesive slope (i.e. undrained clay slope) with different geometries. Probabilistic stability charts are then developed using the results obtained from the parametric studies. These charts can be used for a preliminary assessment of the probability of failure of a spatially random cohesive slope. In addition, the effect of soil variability on c'–ϕ' slopes is also studied. The available RFEM computer program (i.e. rslope2d) is limited to analysing slopes with single-layered soil profile. Therefore, in this research, this computer program is modified to analyse slopes with two-layered soil profiles. The modified program is then used to investigate the effect of soil variability on two-layered spatially random cohesive slopes. It has been demonstrated that the spatial variability of soil variability has a significant effect on the reliability of both single and two-layered soil slopes. Artificial neural networks (ANNs), which are a powerful data-mapping tool for determining the relationship between a set of input and output variables, are used in an attempt to predict the probability of failure of a spatially random cohesive slope. The aim is to provide an alternative tool to the RFEM and the developed probabilistic stability charts because the RFEM analyses are computationally intensive and time consuming. The results obtained from the parametric studies of a spatially random cohesive slope are used as the database for the ANN model development. It has been demonstrated that the ANN models developed in this research are capable of predicting the probability of failure of a spatially random cohesive slope with high accuracy. The developed ANN models are then transformed into relatively simple formulae for direct application in practice. The effect of root reinforcement caused by vegetation is modelled as additional cohesion to the soils, known as root cohesion, cr. The areas affected by tree roots (i.e. root zone) are incorporated in the finite element slope stability model. The extent of the root zone is defined by the depth of root zone, hr. Parametric studies are conducted and the results are used to develop a set of stability charts that can be used to assess the contribution of root reinforcement on slope stability. Furthermore, ANN models and formulae are also developed based on the results obtained from the parametric studies. It has been demonstrated that the factor of safety of a slope increase linearly with the values cr and hr, and the contribution of root reinforcement to a marginally stable slope is significant. In addition, probabilistic slope stability analysis considering both the variability of the soils and root cohesion are conducted using the modified RFEM computer program. It has been demonstrated that the spatial variability of root cohesion has a significant effect on the probability of slope failure. / http://proxy.library.adelaide.edu.au/login?url= http://library.adelaide.edu.au/cgi-bin/Pwebrecon.cgi?BBID=1349971 / Thesis (Ph.D.) - University of Adelaide, School of Civil, Environmental and Mining Engineering, 2009