Ocean-island flank collapses are amongst the most dangerous of all landslide related hazards in the world, as they have the potential to trigger ocean-wide tsunamis that can cause damage and loss of life to communities thousands of kilometres from their source of origin. The implications for landslide-induced tsunami originating from high volcanic islands in the Pacific are serious; and consequent hazards to life, infrastructure, and emergency management need to be constantly reviewed, monitored, and investigated. Ta’u, the easternmost inhabited island in the Samoa Islands volcanic chain, exhibits a series of down-faulted benches on its southern flank; believed to be the remnant of catastrophic collapse involving ~30km³. An historical map of Ta’u, charted during the first United States exploring expedition into the Pacific Ocean (Charles Wilkes Expedition), suggests that the event was recent; having occurred less than 170 years ago. A collapse event of this magnitude would have generated a locally devastating tsunami, with possible impacts experienced at the regional level. However, there exists no written or oral record of such an event. It appears that half the island, involving an estimated 30km³, disappeared off the map less than 170 years ago without anybody noticing it. A number of key questions thus emerged. Did this event actually happen within the last 170 years, and if so, how and why could it have gone unnoticed? Is the event much older than the impression obtained from the literature? More importantly, what is the likelihood of a future collapse and subsequent tsunami, and what would the hazard impacts be at the local and regional levels? These questions formed the research basis for this thesis. Specific aims were developed to address the issues identified, and a range of inter-disciplinary scientific techniques using innovative methods and new datasets were implemented to achieve them. The results demonstrate that the collapse most likely occurred more than 170 years ago, raising serious debate on the accuracy of observations made during the Charles Wilkes Expedition. The results also show that the eruptive-hazard at the site exists. Given that the nature and frequency of active volcanism in the area is uncertain, the risk of a future collapse and subsequent tsunami in the medium-term is considered high. The inter-disciplinary approach to landslide-tsunami hazard investigation on an oceanic island presented in this thesis, can be developed and applied by disaster managers to similar hazard investigations on other oceanic islands. Ultimately, the increase in knowledge-base can be used as a tool for developing safer and more resilient coastal communities.

Ocean-island flank collapses are amongst the most dangerous of all landslide related hazards in the world, as they have the potential to trigger ocean-wide tsunamis that can cause damage and loss of life to communities thousands of kilometres from their source of origin. The implications for landslide-induced tsunami originating from high volcanic islands in the Pacific are serious; and consequent hazards to life, infrastructure, and emergency management need to be constantly reviewed, monitored, and investigated. Ta’u, the easternmost inhabited island in the Samoa Islands volcanic chain, exhibits a series of down-faulted benches on its southern flank; believed to be the remnant of catastrophic collapse involving ~30km³. An historical map of Ta’u, charted during the first United States exploring expedition into the Pacific Ocean (Charles Wilkes Expedition), suggests that the event was recent; having occurred less than 170 years ago. A collapse event of this magnitude would have generated a locally devastating tsunami, with possible impacts experienced at the regional level. However, there exists no written or oral record of such an event. It appears that half the island, involving an estimated 30km³, disappeared off the map less than 170 years ago without anybody noticing it. A number of key questions thus emerged. Did this event actually happen within the last 170 years, and if so, how and why could it have gone unnoticed? Is the event much older than the impression obtained from the literature? More importantly, what is the likelihood of a future collapse and subsequent tsunami, and what would the hazard impacts be at the local and regional levels? These questions formed the research basis for this thesis. Specific aims were developed to address the issues identified, and a range of inter-disciplinary scientific techniques using innovative methods and new datasets were implemented to achieve them. The results demonstrate that the collapse most likely occurred more than 170 years ago, raising serious debate on the accuracy of observations made during the Charles Wilkes Expedition. The results also show that the eruptive-hazard at the site exists. Given that the nature and frequency of active volcanism in the area is uncertain, the risk of a future collapse and subsequent tsunami in the medium-term is considered high. The inter-disciplinary approach to landslide-tsunami hazard investigation on an oceanic island presented in this thesis, can be developed and applied by disaster managers to similar hazard investigations on other oceanic islands. Ultimately, the increase in knowledge-base can be used as a tool for developing safer and more resilient coastal communities.

På Kanarieön La Palma finns sprickzonen och vulkanen Cumbre Vieja. Under ett vulkanutbrott år 1949 bildades ett förkastningssystem längs vulkanens västra sida som skulle kunna vara ett förstadium till en framtida kollaps av vulkanen. Denna kollaps skulle kunna leda till bildandet av en tsunami när kollapsmaterial rasar ner i havet likt ett jordskred. Tsunamin kan komma att spridas över stora delar av Atlanten och i olika grad påverka de omkringliggande kontinenterna. Det finns olika teorier om hur stort jordskredet kommer att bli och om det kommer att ske successivt eller kollapsa som en enda enhet. I detta arbete studeras fyra kollapsscenarier med volymer av 20 km3, 40 km3, 80 km3 och 450 km3 närmare. Utifrån dessa volymer skapas riskanalyskartor i GIS över ön Teneriffa och New Yorks storstadsområde som visar hur långt över land tsunamin når vid de olika kollapsscenarierna. Dessa kartor jämförs sedan med markanvändningskartor över Teneriffa och New Yorks storstadsområde. Trots att tsunamivågen kan bli mycket hög vid Teneriffas kust kommer stora delar av ön att undkomma tsunamin, vilket bl.a. beror på öns höga höjd över havet. Då de flesta byggnader är belägna vid kusten innebär det att många människor ändå riskerar att drabbas. New Yorks storstadsområde är istället lågt beläget, vilket bidrar till att många kustnära områden kommer att drabbas trots att våghöjden har avtagit väsentligt. Då detta område är mycket tätbefolkat kan en tsunami därför leda till en stor påverkan. / On the Canary Island La Palma there is a volcanic ridge called Cumbre Vieja. During an eruption in 1949 a fault system was formed along the western flank of the volcano that can be an early stage of a future flank collapse of Cumbre Vieja. During this collapse a large volume of rock material will fall into the ocean like a landslide, which could lead to the formation of a tsunami. The tsunami could spread over large parts of the Atlantic Ocean and to varying extent affect the surrounding continents. There are different theories of how big the landslide will be and if it will collapse gradually or as a coherent block. In this study four collapse scenarios with different volumes, 20 km3, 40 km3, 80 km3 and 450 km3, are studied more closely. Based on these volumes, vulnerability maps are created in GIS covering the island Tenerife and the New York metropolitan area. The maps show the land reach of the tsunami for each of the collapse scenarios. These maps are thereafter compared with land use maps over Tenerife and the New York metropolitan area. Despite the high amplitude of the tsunami wave at the coast of Tenerife, large parts of the island will remain unaffected by the tsunami. This is due to, among other things, the high altitude of the island. However, most buildings are located along the coast, which means that many people are at risk. The New York metropolitan area has, on the other hand, low altitude which is one explanation why many coastal areas will be affected despite the significant decrease in tsunami wave height. A tsunami would have a major impact due to these areas being densely populated.

L’archipel des Açores dans l’océan Atlantique est édifiées sur un épais plateau océanique, à proximité de la jonction triple entre les plaques Nord-américaine (Na), Nubienne (Nu) et Eurasienne (Eu). La formation du plateau et l’origine du volcanisme ont été le plus souvent attribués à la présence d’une instabilité mantellique. Cependant, la répartition et la morphologie des édifices volcaniques semblent avoir été grandement influencés par la déformation régionale liée à la migration de la frontière de plaque (Eu/Nu). En effet, la frontière serait passée d’une faille transformante aujourd’hui inactive, la zone de fracture est des Açores (EAFZ), à un rift ultra lent actif appelé le Rift de Terceira (TR).Lors de ce travail, nous utilisons le volcanisme comme marqueur de la déformation régionale. Nous nous intéressons particulièrement aux îles de S. Miguel et Graciosa, qui sont localisées à l’intérieur du TR, et à Santa Maria, une île volcanique éteinte qui se situe entre la EAFZ et le TR. De par leur position, ces trois îles constituent donc des cibles particulièrement appropriées afin d’étudier l’architecture et l’évolution de la frontière de plaque Eu/Nu durant les dernier Millions d’années. A partir de nouvelles données géomorphologiques, stratigraphiques, géochronologiques et tectoniques, couplées aux données bathymétriques et géophysiques disponibles, nous reconstruisons les étapes successives de construction et de démantèlement de ces îles puis discutons de leur signification géodynamique. Ces données sont ensuite complétées par des expériences de mécanique des fluides afin d’investiguer les liens possibles entre un panache mantellique, la migration de la frontière de plaque sur plusieurs échelles d’espace et de temps.Les résultats montrent que les édifices localisés dans le TR se construisent via des pulses volcaniques courts (<100 kyr) et relativement synchrones, séparés par des épisodes d’effondrements catastrophiques. Nous proposons qu’une telle évolution reflète des épisodes brefs et intenses de déformation régionale le long de la frontière de plaque active. La distribution des marqueurs tectoniques ainsi que leurs orientations N110 et N150 dans la partie Est de S. Miguel, nous conduit à proposer que l’extension oblique du TR est principalement accommodée par les failles bordières majeures du rift. Nous identifions une nouvelle tendance tectonique orientée N50° qui pourrait représenter des failles transformantes accommodant les variations d’obliquité du TR. L’activité de île de Santa Maria est ici datée entre 5.7 et 2.8 Ma. S. Maria a été façonnée par plusieurs effondrements sectoriels catastrophiques, le plus probablement déclenchés par les mouvements tectoniques régionaux. Nous identifions également une nouvelle structure de type graben reliant les îles de S. Maria et S. Jorge plus loin au NW. La forme de ce graben est semblable au TR et est située entre l’ancienne et la nouvelle frontière Eu/Nu. Nous interprétons ce graben comme un ancien rift transitionnel et donc comme une ancienne frontière de plaque Eu/Nu. A partir de nos données géochronologiques, nous proposons que la partie Est de ce rift transitionnel aurait migré vers la partie Est du TR entre 2.8 et 1.7 Ma.La migration de la frontière Eu/Nu a été interprétées par Vogt and Jung (2004) comme résultant de sauts successifs vers le NE de l’axe du Rift afin de maintenir sa position au dessus d’un point chaud fixe. Nos expériences de mécanique des fluides suggèrent que l’archipel des Açores, comme celui des Canaries, du Cap Vert, de Madère ainsi que les volcans sous marins de Great Meteor sont la signature en surface d’un groupe d’instabilités mantellique prenant naissance et remontant à partir du sommet d’un dôme thermochimique situé dans le manteau inférieur. De plus, Ces panaches secondaires pourraient être suffisamment faibles pour adapter leurs mouvements aux équilibres de forces pré-existants, notamment la structure et la morphologie de la lithosphère. / The Azores archipelago in the Atlantic comprises nine volcanic islands which developed on a thick oceanic plateau close to the Triple Junction between the North American (Na), the Nubian (Nu), and the Eurasian (Eu) lithospheric plates. The formation of the plateau and the origin of the volcanism remain controversial, but have been generally attributed to a plume-like mantle instability. However, the distribution of the volcanic edifices east of the Mid-Atlantic Ridge (MAR) appears greatly influenced by regional deformation associated with the northward migration of the Eu/Nu plate boundary from an extinct old transform fault, the East Azores Fracture Zone (EAFZ), up to the presently active ultra-slow Terceira Rift (TR). In this thesis, we use the volcanism as a marker for regional deformation. We especially focus on S. Miguel and Graciosa, which are located within the TR, and on S. Maria, an old volcanically extinct island located between the EAFZ and the TR. These three islands thus constitute particularly suitable targets to track the architecture and the evolution of the Eu/Nu plate boundary during the last few Myr. From new geomorphological, stratigraphic, geochronologic, structural/tectonic data, and existing bathymetric and geophysical data, we reconstruct the successive stages of growth and destruction of the islands, and discuss their geodynamic meaning. These data are then complemented by fluid dynamic modelling using laboratory experiments to examine the possible links between mantle instability, plate boundary migration and the development of the volcanism on various spatial and temporal scales.The new results on the islands show that the edifices located within the TR grew through short (<100 kyr) and partly synchronous volcanic pulses, separated by catastrophic sector collapses. We propose that such evolution reflects brief and intense episodes of regional deformation along the still active Eu/Nu plate boundary. The distribution of tectonic markers and the recognition of N110 and N150 tectonic structures in eastern S. Miguel leads us to propose that oblique extension in the TR is mainly accommodated by the master faults of the rift, and that the TR is presently not the locus of appreciable sea-floor spreading. Furthermore, we identify a new N050 trend, which may represent transform faults accommodating the variation in obliquity of the TR. The activity of S. Maria is here dated between 5.7 and 2.8 Ma. Like the recent islands, S. Maria experienced catastrophic flank collapses, most probably triggered by regional tectonics. We identify a new graben structure linking Santa Maria to the island of S. Jorge further NW. The shape of this graben is similar to the TR and it is located between the EAFZ and the current plate boundary. We interpret this graben as a former transient rift, and therefore an old Eu/Nu plate boundary. From the new data, we propose that the eastern part of the transient rift migrated to the eastern part of the TR between 2.8 Ma and 1.7 Ma.The overall migration of the Eu/Nu plate boundary to the north and the creation of the Azores plateau has been interpreted by Vogt and Jung (2004) as resulting from successive NE jumps of the rift axis to maintain its position over a fixed ‘hotspot’. Our fluid mechanics experiments suggest that the Azores, as Canary, Cape Verde, Madeira Islands and Great Meteor seamounts might be the surface signature of a cluster of mantle instabilities rising from the top of a large thermochemical dome located in the lower mantle. However, such secondary plumes present a strong time-dependence 5-40 Myr time scale. Moreover, they could be sufficiently weak to adapt their motions to the pre-existing force balances and morphology of the lithosphere. We therefore present a scenario of the Azores area evolution combining a triple junction and decompression melting buoyant material (i.e. such in volatiles and/or temperature) under a thickening lithosphere.

Les effondrements massifs de flancs sont des évènements destructeurs qui affectent tous les édifices volcaniques. Ces effondrements peuvent impliquer quelques dizaines, voire centaines de km3 de roche. Les volcans boucliers océaniques, bien que possédant de faibles pentes, sont également affectés par ces épisodes destructeurs, entraînant la formation de tsunamis. Dans le contexte de bouclier océanique, ces déstabilisations n’ont jamais été observées. Actuellement, seules des reliques sont présentes sur les îles volcaniques, avec la présence de cicatrices d’effondrements, de dépôts détritiques en mer ou de dépôts de tsunami sur les îles voisines. Les relations entre le magmatisme des volcans boucliers et les effondrements de flanc sont peu contraintes. Afin de mieux comprendre ces relations, deux effondrements de flanc de volcans boucliers océaniques ont été étudiés : l’effondrement de Monte Amarelo sur l’île de Fogo, dans l’archipel du Cap Vert, et l’effondrement de Güímar, situé sur la rift-zone Nord-Est de Tenerife, dans l’archipel des Canaries. Les deux archipels résultent de l’activité d’un panache mantellique sous la plaque africaine.Les produits volcaniques pré et post-effondrement de ces deux secteurs ont été étudiés d’un point de vue géochimique (majeurs, traces, isotopes Sr-Nd-Pb), pétrologique et géochronologique (K-Ar, Ar-Ar), de manière à identifier à la fois les sources et les processus magmatiques mis en jeu lors de leur formation. L’évolution temporelle des sources, ainsi que des processus magmatiques, a été reconstruite afin d’identifier d’éventuels liens avec l’effondrement de flanc étudié.Les résultats montrent que l’évolution du magmatisme de l’île de Fogo amène à la formation de zones superficielles de stockage, de complexes intrusifs et d’éruptions explosives conduisant à de nouvelles instabilités de l’édifice. Suite à l’effondrement, ces zones de stockage sont déstabilisées en quelques milliers d’années. Les processus magmatiques lithosphériques (assimilation, fusion partielle) sont également perturbés mais sur une période plus longue (plusieurs dizaines de milliers d’années). L’effondrement de Güímar ne montre aucun lien avec le magmatisme de l’île de Tenerife.La différence principale entre ces deux contextes est la localisation de la zone effondrée par rapport au système magmatique. En effet, l’effondrement de Güímar est situé en périphérie du système magmatique et ne montre aucun lien avec ce dernier ; à l’inverse l’effondrement du Monte Amarelo, situé à l’aplomb du système magmatique, se répercute rapidement sur ce système à faible profondeur, mais également à des profondeurs lithosphériques avec un délai plus long. / Massive flank collapses are destructive events that affect all volcanic edifices. They can take off huge volumes of rock, from tens to hundreds km3. Oceanic shield volcanoes are also affected by such events even if they have shallow slopes, thus, tsunamis could also be generated in this context. However, a shield volcano flank collapse has never been observed. Nowadays, only relics are visible, such as collapse scars, detritic deposit offshore or tsunami deposits on nearby islands. The relationships between collapse and magmatic history of oceanic shield volcanoes are poorly constrained. Two flank collapses are studied in this thesis, with the aim to better understand these relationships: the Monte Amarelo collapse, on Fogo Island (Cape Verde), and the Güímar collapse, located on North-East rift-zone of Tenerife (Canary Islands). Those archipelagos are the result of hot spot activity below the African plate. Geochemical (major and trace elements, and Sr-Nd-Pb isotopes), petrological and geochronological (K-Ar and Ar-Ar) analyses were carried out on volcanic samples as to identify the source and magmatic processes at stake during magma genesis. The temporal evolution of source and magmatic processes is reconstructed in order to track possible links with the flank collapse. The magmatic system of Fogo Island evolves through time, favouring the formation of superficial storage zones, intrusive complex and explosive eruptions prior to the collapse, which participate to the instability of the edifice. Following the Monte Amarelo collapse, shallow storage zones are destabilized within a few thousand years. Lithospheric magmatic processes (assimilation, partial melting) are also affected but on a longer timescale (tens of thousands years). The Güímar collapse shows no links with the magmatic evolution of Tenerife Island. The main difference between the two collapses is the location of the collapse area with respect to the plumbing system. Güímar collapse is located at the periphery of the plumbing system and show no link with the magmatic history. Contrariwise, the Monte Amarelo collapse is located directly above the plumbing system and influence rapidly the superficial plumbing system, and the deep plumbing system in the long term.