Multiscale Ambient Noise Tomography for volcano monitoring and geothermal exploration in volcanic areas

El desarrollo de un mŽtodo no lineal multiescala para la Tomograf’a de Ruido Ambiental es, sin duda, el aspecto m‡s innovador de esta tesis. Este mŽtodo es altamente eficiente, y sus fortalezas radican en la utilizaci—n de fuentes pasivas naturales, lo que permite la exploraci—n del subsuelo sin necesidad de detonaciones costosas y reduce los gastos log’sticos asociados al transporte de camiones Vibroseis o martillos neum‡ticos. Desarrollamos un mŽtodo de inversi—n no lineal multiescala que muestra mejoras con respecto a los mŽtodos de inversi—n lineales y no lineales convencionales. Aplicamos este mŽtodo para estudiar el volc‡n Misti (Perœ) y las islas de Gran Canaria y La Palma. La selecci—n de las ‡reas de estudio se hizo con el objetivo de la vigilancia volc‡nica, y en el caso de La Palma y Gran Canaria, el objetivo se extiende a la exploraci—n geotŽrmica. Las Islas Canarias representan la regi—n espa–ola con mayor potencial para la utilizaci—n y posible explotaci—n de recursos geotŽrmicos debido a su actividad volc‡nica actual o extinta. Esta tesis se centra en la generaci—n de modelos 3D de recursos geotŽrmicos de alta entalp’a en las islas de Gran Canaria y La Palma. Estos modelos pueden allanar el camino para el establecimiento de las primeras plantas geotŽrmicas de alta entalp’a en nuestro pa’s. Esto podr’a actuar como motor para el incipiente sector geotŽrmico, estimulando el desarrollo geotŽrmico e impulsando sectores relacionados como centros de investigaci—n enfocados en el conocimiento del subsuelo, empresas de perforaci—n y firmas de ingenier’a para el desarrollo de proyectos. El objetivo es identificar ‡reas con anomal’as de velocidad indicativas de estructuras geol—gicas asociadas con posibles fuentes geotŽrmicas. Un ejemplo ser’a una anomal’a negativa de velocidad causada por un sistema hidrotermal.

9.1 A nonlinear multiscale inversion for ambient noise tomography Our proposition introduces an innovative nonlinear multiscale inversion method for ANT. The outcomes derived from various setups of synthetic models exhibit substan- tial variations, contingent upon the chosen inversion and regularization strategies. To address this, we systematically conducted extensive testing across various models. In most instances, the applied inversion methods demonstrated the capability to effec- tively detect and accurately locate the intended target anomalies. However, the anomaly shape and velocity value were not satisfactorily retrieved for various combinations of in- version methods and regularisation strategies. On average, the nonlinear multiscale in- version yields superior results to linear and nonlinear inversions. It consistently enables accurate determination of both the shape and velocity value for positive and negative anomalies. These findings hold true for both idealized and real network configurations. Furthermore, we observed that the nonlinear multiscale inversion shows considerable improvement compared to the nonlinear inversion when the ray density is low. Regarding the regularization methods employed in this study, it s clear that TSVD produces the least desirable outcomes across all types of inversions. Its impact on the inversion process is adverse, resulting in the appearance of artifacts such as ghost anomalies in the final tomographic images. Conversely, TIKH regularization demon- strates promising results, which are evident in the statistical analysis for all types of inversions. Specifically, the choice of ?? = 0.5 for TIKH regularization consistently yields superior performance compared to other parameter combinations. We demonstrated that the nonlinear multiscale inversion performs better with only a slight increase in computing time compared to traditional nonlinear inversion. There- fore, the non-linear multiscale ambient noise tomography can be applied satisfactorily to image complex geological areas such as volcanoes or industrial applications such as the exploration of geothermal resources. 9.2 Ambient Noise Tomography of Misti volcano (Peru) Our ANT of Misti provides the first 3-D S-wave velocity model of this andesitic stratovol- cano, for which few previous geophysical studies exist [142, 57, 107, 104, 108, 38, 4]. We have processed a dataset from 23 temporal seismic stations with broadband and short-period sensors. We applied a nonlinear multiscale inversion to obtain the 2D group velocity maps and a transdimensional approach for the inversion in depth. The obtained 3-D S-wave velocity model shows the presence of three high-velocity and two low-velocity anomalies. The high-velocity zones could be related to a dense intrusive complex of dikes associated with the formation of Misti volcano. Two high- velocity anomalies, located at the western-northwestern and southeastern parts of the crater, could be related to the Falla Chili, which could have generated a weakness zone through which the magmatic materials have risen. On the other hand, the low- velocity zone imaged on the volcano s western flank could be related to the volcano s hydrothermal system for the top part and to an ancient caldera wall for the bottom part, where fractured and weakened materials could explain the velocity reduction in this zone. Conversely, the low-velocity anomaly located under the crater of Misti volcano could be related to fracture zones produced by the occurrence of repeated eruption episodes during the recent geological history of the volcano, through which gases could ascend. Finally, we underline that more multidisciplinary studies would be required to cor- roborate further our findings and, in general, to understand better the internal structure of Misti volcano. We believe that our findings directly affect the quality of the volcano monitoring system and, therefore, helping mitigate the volcanic risk associated with this volcano. 9.3 Ambient Noise Tomography of Gran Canaria island We obtained the first 3-D S-wave velocity model of Gran Canaria island using ANT. We used data from 28 temporary and two permanent broadband seismic stations. We applied a conventional methodology to extract dispersion curves for all the pairs of stations. Subsequently, we applied a non-linear multiscale inversion to obtain 2- D group velocity maps for different periods according to [29]. Finally, we applied a Bayesian transdimensional approach to perform the depth inversion. The 3-D S-wave velocity model shows the presence of very well-differentiated velocity anomalies: two with high-velocity (H1 and H2) and three with low-velocity (L1, L2 and L3). The high- velocity anomalies could be related to intrusive volcanic bodies, related to the early phases of the island formation and aligned along the main structural trend of the is- land. One of these intrusions (H1) separates the island in two parts corresponding, respectively, to the Paleocanaria, the oldest part of the island and the more recent Neocanaria. Conversely, the low-velocity zones could be associated with porous and highly fractured materials produced during the more recent volcanism on the island. Further, we observe a spatial correlation between the conductive anomalies observed in the resistivity model from [93] and strong lateral gradients in the S-wave velocity model. These conductive anomalies could represent evidences of hydrothermal circu- lation and thermal anomalies, associated with lateral contacts between different geo- logical units and/or faults. These interesting geological structures could be of interest in the context of geothermal exploration of the island. This study has corroborated the existence of such structures, which were previously inferred only by resistivity models. 9.4 Ambient Noise Tomography of La Palma island A 3-D S-wave velocity model of La Palma island was obtained by unifying results from a new ANT model obtained in this study and a LET model obtained by D Auria et al. [45]. We applied ANT on data recorded by 38 broadband seismic stations to extract the sta- tion pairs dispersion curves. Subsequently, we obtained the 2-D group velocity maps through a non-linear multiscale inversion taking the topography into account[29]. Fi- nally, we derived some S-wave 1-D profiles using a Bayesian tridimensional inversion. The final 3-D S-wave velocity model shows the presence of two high-velocity zones (H1 and H2) and four low-velocity zones (L1, L2, L3 and L4). The high-velocity anomalies H1 and H2 are interpreted as a plutonic intrusion re- lated to the island s ancient volcanism and more recent solidified intrusive dyke com- plexes, respectively. From the point of view of geothermal exploration, the most inter- esting imaged features are the low-velocity anomalies L1, L2, and L3. The low-velocity zones L1 and L2 are interpreted as hydrothermal alteration zones associated with an active or fossil hydrothermal system in the Cumbre Vieja volcanic complex. Velocity variations estimated before the Tajogaite eruption seem to favour the hypothesis of an active hydrothermal reservoir. The anomaly L3 is interpreted as associated with frac- tured rocks favouring the ascent of hot fluids toward the surface in the island s southern part. This hypothesis could also explain the geochemical and geophysical anomalies observed in previous studies[141, 131, 132]. Finally, the low-velocity anomaly L4 could be related to landslide deposits produced during destructive episodes of the island s geological history. 9.5 General conclusion and perspectives As previously mentioned, the nonlinear multiscale ambient noise tomography method developed in this Ph.D. study offers several advantages over conventional linear and nonlinear inversion techniques. This method has allowed us to gain deeper insights into the geological structures of Misti volcano, Gran Canaria, and La Palma Island. Our findings align well with previous geophysical studies[57, 93, 35, 34, 51, 45, 30]. For Misti volcano, we are considering the application of additional geophysical tech- niques such as magnetotelluric and gravimetry to further characterize the stratovol- cano s structure. Additionally, a local earthquake tomography could be conducted to increase the depth of our S-wave velocity model, providing more insights into the deep crustal structure. Comprehensive geophysical exploration surveys at the island-wide scale are being considered to better understand Gran Canaria Island s geothermal potential. These surveys may include spontaneous potential measurements and detailed investigations in areas showing high geothermal potential, such as seismic tomography, magnetotel- luric surveys, electrical tomography, gravimetry, and more. We emphasize the impor- tance of conducting multiparametric geochemical surveys across the entire island. In the case of La Palma Island, more detailed geophysical and geochemical explo- ration surveys are deemed necessary, especially in the southern part of the island where S-wave low-velocity anomalies are shallower. These efforts would further enhance the characterization of La Palma s geothermal potential. Furthermore, applying clustering and machine learning techniques for quantita- tively comparing resistivity, density, and S-wave velocity models is being considered. This approach aims to combine models obtained from various techniques to improve the interpretation of the geological context in different study areas [24, 130, 13, 52, 67]. Future research focuses on extracting attenuation information from ambient noise cross-correlations for noise-based monitoring and imaging purposes. Time-lapse mon- itoring of seismic velocity can be crucial for understanding volcanic system dynamics and potentially forecasting eruptions. This technique relies on seismic ambient noise interferometry to detect subtle changes in the medium, particularly relative velocity variations (dv/v) by analyzing phase changes in ambient noise cross-correlation coda waveforms. In addition to velocity variations, the study of seismic attenuation changes extracted from the coda of ambient noise cross-correlations holds promise for volcano monitoring. We observed that obtaining a 3-D attenuation model from ambient noise cross- correlations is feasible for ambient noise tomography. Our ongoing work is focused on developing a 3-D attenuation model for La Palma Island. The methodology employed for ambient noise attenuation tomography includes several essential steps: (I) Measure- ment of ???? using the lapse-time dependence method, which involves assessing ???? as a function of coda window length (????) for different onset times of the coda (????). (II) Performing a 2-D spatial attenuation inversion using linear inversion techniques and sensitivity kernels, considering diffusion as the scattering regime. (III) Conducting a 1- D depth inversion utilizing linear inversion techniques and sensitivity kernels derived from the Rayleigh wave fundamental mode, extracted from a previous S-wave veloc- ity model. Preliminary results indicate a strong correlation between high-attenuation zones and low-velocity zones, high-conductivity zones, and low-density zones. These future research avenues are promising for advancing our understanding of geological environments such as volcanoes and geothermal fields where fluids are in- volved.