Geothermal energy is a promising renewable resource, and its development in volcanic regions has unique challenges because of the interaction between volcanic processes, magmatic intrusions, and geothermal reservoir dynamics. In this thesis, the evolution of geothermal reservoirs in Chile is explored through the analysis of surface deformation and seismic data. The focus is on two case studies: Cerro Pabellón, which is the first installed geothermal power plant in South America, and Cordón Caulle, a volcanic complex that underwent a significant eruption in 2011-2012 and is a potential geothermal resource. At Cerro Pabellón, analyzing the InSAR time series, from the Sentinel-1A/B satellite mission between 2014 and 2021, reveals an increase in subsidence rates near the geothermal area -- up to -7.2 mm/yr along the radar line of sight -- coinciding with the power plant initial operation date (Lobos et al., 2023). Analytical models suggest that the source of deformation is located at a depth between 370 and 660 m, a location likely compatible with the geothermal reservoir depth. The deformation source probably resulted from fluid extraction in combination with the decrease of pressure in the geothermal reservoir. Apart from the observed surface deformation, between March 2017 and August 2020 there was some seismicity detected in the Pabelloncito graben close to the geothermal plant; however due to limited coverage of the seismic network with the closest stations > 30 km away, it is not possible to link this seismicity directly to the power plant. The rate of earthquakes does not increase with geothermal production, and the events are not preferentially located near the power plant. Seismicity does appear to be triggered by distant earthquakes. At Cordón Caulle, we combine a satellite-derived time series of topographic change from 2011-2019 with field observations to document the multi-source mechanisms of elevation change during and following the 2011-2012 eruption, which includes the emplacement and subsequent subsidence of a laccolith, lava flows, and tephra deposits. We identified three distinct periods of topographic change: (i) rapid surface uplift to 410 m/y during the laccolith and lava flow emplacement phase (Jun-Aug 2011), (ii) accelerated subsidence rates to 12 m/y (Aug 2011-Feb 2013), and (iii) a period of slower subsidence at a rate of 0.3 m/y (Feb 2013-Nov 2019) (Lobos et al., 2022). We map the spatial distribution of the ratio of subsidence in periods ii-iii to uplift in period i to reveal the influence of various volcanic and geomorphological processes (cooling, contraction, erosion) in different locations. Furthermore, we created a seismic catalog before and during the 2011 Cordón Caulle eruption using machine learning techniques for phase picking and event detection, which provides evidence for earthquakes associated with magma migration and eruption. For detection and then localization of seismic events, we apply deep-learning phase-pickers (GPD, U-GPD) and three-phase associators (GaMMA, PhasePApy, PyOcto), to produce earthquake catalogs to constrain magmatic and eruption dynamics. We find two distinctive periods of seismicity corresponding to two distinctive phases of eruptive behavior: a phase of increased seismic activity related to the eruption onset and a gradually declining phase toward the end. Spatial changes in the event distribution showed a high degree of clustering in the vicinity of the 2011 vent, indicating that volcanic unrest and magma migration were still in progress. Comparisons of the three-phase associators indicate their complementary strengths, with PyOcto detecting the most events, GaMMA identifying deeper events with well-constrained locations, and PhasePApy capturing events with lower station coverage.