La propagation sismique dans les milieux poreux est classiquement associée à des phénomènes de dispersion et d’atténuation des ondes sous l’effet des mouvements fluides. Dans certaines conditions, celle-ci peut également être associée à une conversion d’énergie sismique en énergie électromagnétique dite « sismoélectrique ». La théorie des phénomènes sismoélectriques combinant la théorie de l’électrocinétique à la poroélasticité de Biot, repose en grande partie sur les développements de Pride (1994). Sur la base de ces développements théoriques, Pride et Haartsen (1996) relient le champ électrique cosismique à l’accélération sismique qui le génère par une fonction de transfert. Nous proposons une étude quantitative des couplages sismoélectriques en vue de valider la théorie de Pride et sa généralisation en milieu non saturé. Dans ce but, nous avons développé une expérience en laboratoire sur un sable de quartz non-consolidé, menée dans la gamme du kilohertz sur la base d’un dispositif d’acquisition électrique modulable. Deux méthodes de traitement des signaux sont proposées, l’une temporelle, l’autre spectrale, permettant d’obtenir une analyse complète des vitesses de phase, atténuations et fonctions de transfert. Les expériences réalisées se sont focalisées sur l’étude du rôle de la conductivité du fluide et de la saturation en eau dans le phénomène sismoélectrique cosismique. Une étude time-lapse a ainsi pu être réalisée dans des situations de changements de salinité et de teneur en eau. Dans tous les contextes, l’étude quantitative des rapports d’amplitudes des champs sismoélectriques et sismiques E/ü montre une bonne corrélation avec les prédictions théoriques. Par ailleurs, l’étude des variations de saturation dans une gamme allant de la saturation résiduelle en eau (Sw = 0.3) à la saturation totale, montre que: i) les atténuations et fonctions de transfert ont des comportements reliés à la distribution des fluides qui influencent fortement les propriétés mécaniques du milieu ; ii) une inversion de polarité du champ sismoélectrique peut être observée dans le cas très particulier des milieux non consolidés. / Seismic propagation within porous media is usually associated with wave attenuation and dispersion phenomena related to fluid flow. Under certain circumstances, it may also be correlated to a conversion of seismic into electromagnetic energy known as “seismoelectric”. The understanding of seismoelectric phenomena, combining the theory of electrokinetic to Biot’s poroelasticity, relies mainly on the formulation by Pride (1994). On basis of these theoretical developments, Pride and Haartsen (1996) defined a transfer function expressing the link between the coseismic seismoelectric field and the seismic acceleration at its origin. We propose a quantitative analysis of coseismic seismoelectric couplings with the purpose of validating Pride’s theory and generalizing it to partially saturated media. With this aim in view we developed a laboratory experiment involving an adjustable device for electric acquisitions, conducted within the kilohertz range on unconsolidated quartz sand. Experimental data were subsequently processed in both time and frequency domains, enabling a full analysis that embraces phase velocities, attenuations and transfer functions. The conducted experiments focused on the impact of fluid conductivity and water saturation with regard to the coseismic seismoelectric phenomenon. Time-lapse monitoring were accordingly run under varying salinity or water content. In all scenarios, the quantitative analysis of the electric-to-seismic amplitude ratio E/ü appeared in good agreement with theoretical projections. Moreover, investigations of saturation variations, ranging from the residual water saturation (Sw = 0.3) to full saturation, showed that: i) the behavior of attenuations and transfer functions are directly related to fluid distribution, that greatly impacts the mechanical properties of the medium; ii) in the very peculiar case of unconsolidated media, polarity inversion of the coseismic seismoelectric field may be experienced.

The aim of this research was to study the relationship between rock slope stability and glacial processes. An in-depth analysis of our current understanding of how glaciated rock slopes develop instability and movement during deglaciation is presented; this shows that understanding is incomplete without an appreciation of the variable mechanical behaviour of glacier ice. In this thesis, I argue that: (1) The ductile behaviour of ice at low strain rates allows movement of rock slopes buttressed by ice. Field evidence and simple force models are used to explore rate of movement of ice-contact slopes and the conditions under which they evolve. The results indicate that large rockslides can move and deform glacial ice at rates of 10-2 to 102 m-yr. This implies that ice-contact slope movement may be important for slope evolution and the erosion and entrainment processes of glaciers; and (2) the elastic strength of glacier ice at the high strain rates associated with seismic shaking enables ice to modify the response of the surrounding rock to seismic shaking. To explore this, numerical analyses of the interaction between glacial erosion, glacier mass, topography, and earthquake shaking intensity are undertaken. Shaking of mountains of variable shape and with different levels of ice inundation is simulated using FLAC 6.0. The results suggest that complete inundation by ice can significantly reduce shaking intensity. This, in combination with glacial steepening of slopes, may make recently deglaciated slopes more prone to coseismic failure. In the final chapter of the thesis, I present a conceptual model of the evolution of slope stability during stages of glaciation and deglaciation. The model incorporates the ideas presented in the thesis. I then offer recommendations for how our understanding of these processes can be further advanced.

The Alpine Fault is a major plate boundary structure, which accommodates up to 50-80% of the total plate boundary motion across the South Island of New Zealand. The fault has not ruptured historically although limited off-fault shaking records and on-fault dating suggest large to great (~ Mw 8) earthquakes (every ~100-480 years; most recently in 1717), making it potentially one of the largest onshore sources of seismic hazard in New Zealand. The central section of the Alpine Fault, which bounds the highest elevations in the Southern Alps, is one of the most poorly characterised sections along the fault. On-fault earthquake timing in addition to the amount of dextral slip during major earthquakes was unknown along a 200-km-long section of the central Alpine Fault, while the amount of co-seismic hanging wall uplift was poorly known, prior to the present work. In this thesis I address these knowledge gaps through a combination of light detection and ranging (lidar), field, and stratigraphic mapping along with sample dating to constrain earthquake timing, style of faulting, and hanging wall rock uplift rates. Using lidar data coupled with field mapping I delineated the main trace of the Alpine Fault at Gaunt Creek as a north-striking fault scarp that was excavated and logged; this is part of a 2-km-wide restraining bend dominated by low-angle thrust faulting and without the clear strike-slip displacements that are present nearby (<5 km distant along strike in both directions). Where exposed in this scarp, the fault-zone is characterized by a distinct 5-50 cm thick clay fault-gouge layer juxtaposing hanging wall bedrock (mylonites and cataclasites) over unconsolidated late-Holocene footwall colluvium. An unfaulted peat at the base of the scarp is buried by post-most recent event (MRE) alluvium and yields a radiocarbon age of A.D. 1710–1930, consistent with sparse on-fault data, validating earlier off-fault records that suggest a 1717 MRE with a moment magnitude of Mw 8.1 ± 0.1, based on the 380-km-long surface rupture. Lidar and field mapping also enabled the identification and measurement of short (<30 m), previously unrecognized dextral offsets along the central section of the Alpine Fault. Single-event displacements of 7.5 ± 1 m for the 1717 earthquake and cumulative displacements of 12.9 ± 2 m and 22 ± 2.7 m for earlier ruptures can be binned into 7.1 ± 2.1 m increments of repeated dextral (uniform) slip along the central Alpine Fault. A comparison of these offsets with the local paleoseismic record and known plate kinematics suggests that the central Alpine Fault earthquakes in the past 1.1 ka may have: (i) bimodal character, with major surface ruptures (!Mw 7.9) every 270 ± 70 years (e.g. the 1717 event) and with moderate to large earthquakes (!Mw 7) occurring between these ruptures (e.g. the 1600 event); or (ii) that some shaking data may record earthquakes on other faults. If (i) is true, the uniform slip model (USM) perhaps best represents central Alpine Fault earthquake recurrence, and argues against the applicability of the characteristic earthquake model (CEM) there. Alternatively, if (ii) is true, perhaps the fault is “characteristic” and some shaking records proximal to plate boundary faults do not necessarily reflect plate-boundary surface ruptures. Paleoseismic and slip data suggest that (i) is the most plausible interpretation, which has implications for the understanding of major plate-boundary faults worldwide. Field mapping, geological characterisation, geophysical mapping, and optically stimulated luminescence (OSL) dating of on-fault hanging wall sediments were used to better constrain the geometry and kinematics of Holocene deformation along the rangefront of the Southern Alps at the Alpine Fault near the Whataroa River. The fault here is dextral-reverse, although primarily strike-slip with clear fault traces cutting across older surfaces of varying elevations. Deformational bulges are observed along these traces that are likely thrust-bounded. A terrace of Whataroa River sediments was found on the hanging wall of the Alpine Fault approximately ~ 55-75 m (when considering uncertainties) above the floodplain of the Whataroa River. OSL ages for a hanging wall sediments of 10.9 ± 1.0 ka for the aforementioned terrace, 2.8 ± 0.3 ka for Whataroa River terrace deposits in a deformational bulge, and 11.1 ± 1.2 ka for a rangefront derived fan indicate Holocene aggradation along the rangefront and hanging wall uplift rates of 6.0 ± 1.1 mm/yr. The sub-horizontal, laterally continuous, and planar-bedded Whataroa-sourced terrace deposits suggest that the adjacent bounding faults are steeply-dipping faults without geometries in the shallow subsurface that would tend to cause sedimentary bed rotation and tilting. Using data from the approximately 100-m deep pilot DFDP boreholes together with lidar and field mapping, I present a review of the Quaternary geology, geomorphology, and structure of the fault at Gaunt Creek, and estimate new minimum Late-Pleistocene hanging wall rock uplift rates of 5.7 ± 1.0 mm/yr to 6.3 ± 1.1 mm/yr (without considering local erosion) that suggest that the Southern Alps are in a dynamic steady state here. GPS-derived “interseismic” vertical uplift rates are < 1 mm/yr at the Alpine Fault, so the majority of rock uplift at the rangefront happens during episodic major earthquakes, confirming with on-fault data that slip occurs coseismically. Notably the uplift rates from both Mint and Gaunt Creek are consistent between the two sites although the primary style of faulting at the surface is different between the two sites, suggesting consistent coseisimc uplift of the Southern Alps rangefront along the Alpine Fault in major earthquakes. This thesis collected new on-fault datasets that confirm earlier inferences of plate-boundary fault behaviour. This study of the high-uplift central section of the Alpine Fault provides the first on-fault evidence for the MRE (i.e. 1717) and repeated of dextral slip during the MRE and previous events as well as new hanging wall uplift data which suggests that the majority of rangefront uplift occurs in earthquakes along the Alpine Fault. Because the fault has not ruptured for ~300 years, it poses a significant seismic hazard to southern New Zealand.

京都大学 / 0048 / 新制・課程博士 / 博士(理学) / 甲第21950号 / 理博第4528号 / 新制||理||1650(附属図書館) / 京都大学大学院理学研究科地球惑星科学専攻 / (主査)教授 釜井 俊孝, 教授 千木良 雅弘 / 学位規則第4条第1項該当 / Doctor of Science / Kyoto University / DGAM

Local site conditions are known to influence ground motion during earthquake events and increase the severity of damage. Data from earthquakes are useful to study the response but they are available only from active regions. Ubiquitous ambient vibrations on the other hand offer a more practical approach to quantify site responses. This thesis explores the use of various methods for obtaining site responses. The primary area of study is the Kachchh rift basin, NW India, a Mesozoic rift that features significant lateral variations in surface geology and has experienced ground responses during 1819 and 2001 earthquakes. The Mw 7.6, 2001 event was followed by hundreds of aftershocks, which were recorded by temporary networks. In this study we have used earthquake signals as well as ambient vibrations to understand site response in various parts of the basin. In addition we have collected data from a few sites from the Indo-Gangetic plains and Kathmandu valley, both affected by large earthquakes, 1934 the M ~ 8 (Bihar) and 2015, Mw 7.8 (Nepal). Velocity and acceleration records from a network of eight stations in the Kachchh Rift were used to evaluate site responses using Standard Spectral Ratio (SSR) and Horizontal to Vertical spectral ratio (HVSR-E) methods. Ambient vibrations were analyzed following Nakamura’s H/V method (HVSR-AV), for data collected from 110 sites that represent different field conditions within the Kachchh Rift. Fundamental resonance frequency (f0) varied between 0.12 – 2.30 Hz, while the amplification factor (A0) was in the range of 2.0 – 9.1. We found that higher A0 and liquefaction index (Kg) values were mostly associated with higher liquefaction potential. Using a close network of stations, we studied the role of site response in damage to the Bhuj city that suffered maximum damage in 2001; our results suggest that site response was not a significant factor. Studies based on passive data were complemented by Multi-channel Analysis of Surface Waves (MASW) to map shear wave velocities of the various subsurface units up to depths of 10m (Vs10) and 30m (Vs30). Our results imply average Vs could be a good proxy to characterize site amplifications where sediment thicknesses are shallow. Power law relationship between f0 and thickness (h) suggest a strong positive correlation (r = 0.89) adding credence to HVSR-AV method, making it a cost-effective alternative to MASW to infer site conditions. Further, to understand the influence of topography on site effects, we analyzed data from hills, valleys and their edges, both from the Kachchh rift and Kathmandu valley. Sites on the edges of valleys showed multiple, fuzzy peaks in the low frequency range (< 1 Hz) and broad peaks attributable to sites prone to higher damage. Spectrograms generated through Huang-Hilbert Transforms (HHT) suggested focusing of energy in narrow frequency bands on the edges, while valleys tend to scatter energy over wide frequencies. Although our current results are based on limited observations, we recognize spectral analysis as a powerful tool to quantify site effects in regions with significant topography. It is known that coseismic liquefaction could lead to nonlinear behavior wherein the near-surface soil layer loses its shear strength, causing a reduction of its fundamental resonance frequency. We used data from selected sites of coseismic liquefaction to highlight the significance of nonlinear effects in site response. Earthquake signals and ambient vibrations from Umedpur, a region that experienced intense liquefaction during 2001 were used in this analysis. Here we followed an empirical decomposition method based on HHT and signals were decomposed as many intrinsic mode functions (IMFs) that showed characteristic peaks for events of various values of PGAs. Thus, the first IMF for events with relatively higher PGAs (0.03g) showed distinct peaks for the S wave coda part, which were not noted for those with lower PGA (0.01g). These observations in a region of coseismic liquefaction are useful in developing models for quantifying nonlinear behavior. In conclusion, site response studies using different types of data and processing techniques in regions affected by recent earthquakes brings out the scope and limitations of each of these sets of data and techniques. This study suggests that ambient vibrations provide reasonable estimates of site response and can be reliably used in regions where earthquake data are not available.