Abstract: Cite this article as Löer, K., T. Toledo, G. Norini, X. Zhang, A. Curtis, and E. H. Saenger (2020). Imaging the Deep Structures of Los Humeros Geothermal Field, Mexico, Using Three-Component Seismic Noise Beamforming, Seismol. Res. Lett. 91, 3269–3277, doi: 10.1785/ 0220200022. Supplemental Material We present a 1D shear-velocity model for Los Humeros geothermal field (Mexico) obtained from three-component beamforming of ambient seismic noise, imaging for the first time the bottom of the sedimentary basement ∼5 km below the volcanic caldera, as well as the brittle-ductile transition at ∼ 10 km depth. Rayleigh-wave dispersion curves are extracted from ambient seismic noise measurements and inverted using a Markov chain Monte Carlo scheme. The resulting probability density function provides the shear-velocity distribution down to 15 km depth, hence, much deeper than other techniques applied in the area. In the upper 4 km, our model conforms to a profile from local seismicity analysis and matches geological structure inferred from well logs, which validates the methodology. Complementing information fromwell logs and outcrops at the near surface, discontinuities in the seismic profile can be linked to geological transitions allowing us to infer structural information of the deeper subsurface. By constraining the extent of rocks with brittle behavior and permeability conditions at greater depths, our results are of paramount importance for the future exploitation of the reservoir and provide a basis for the geological and thermodynamic modeling of active superhot geothermal systems, in general. Introduction Los Humeros volcanic complex (LHVC; Fig. 1), located in the eastern part of the Trans-Mexican volcanic belt (TMVB), hosts a conventional geothermal field (Ferrari et al., 2012; GutiérrezNegrín, 2019). On-going hydrothermal activity makes the LHVC a favorable area for geothermal exploitation, and a geothermal power plant has been operating since the 1990s. The LHVC has been identified as an important natural laboratory for the development of general models of superhot geothermal systems (SHGSs) in volcanic calderas (e.g., Jolie et al., 2018). Although extensive geological field studies and well log analyses have provided many constraints on the near-surface geology of the caldera complex and conventional geothermal reservoir, conditions at depths greater than 2–3 km are largely unknown and currently being studied intensively (Jolie et al., 2018). It is assumed that superhot fluids could exist in the carbonate rock basement underlying the caldera (Jolie et al., 2018). These rocks might exhibit secondary permeability related to the damage zone of active resurgence faults and inherited pervasive basement structures (Lorenzo-Pulido, 2008; Rocha-López et al., 2010; Norini et al., 2015, 2019; Jolie et al., 2018). The maximum depth of these brittle structures is defined by the brittle-ductile (BD) transition zone, which thus plays an important role in geothermal exploration because upper crustal faults and fractures behave as hydraulic channels for the circulation of geothermal fluids (e.g., Ranalli and Rybach, 2005). In SHGSs that exhibit a positive thermal anomaly, the depth of the BD transition may differ from areas with a normal thermal gradient, as rocks become progressively more ductile with increasing temperature. Thus, a positive 1. Department of Civil and Environmental Engineering, Bochum University of Applied Sciences, Bochum, Germany; 2. Now at Department of Geology and Geophysics, University of Aberdeen, Aberdeen, United Kingdom; 3. German Research Centre for Geosciences GFZ, Section 4.8 Geoenergy, Section 2.2 Geophysical Deep Sounding, Potsdam, Germany; 4. Istituto di Geologia Ambientale e Geoingegneria, Consiglio Nazionale delle Ricerche, Area della Ricerca CNR—ARM3, Milan, Italy; 5. School of Geosciences, Grant Institute, University of Edinburgh, Edinburgh, United Kingdom; 6. Institut für Geophysik, ETH Zürich, Zürich, Switzerland; 7. Institut für Geologie, Mineralogie und Geophysik, Ruhr-Universität Bochum, Bochum, Germany; 8. Fraunhofer-Einrichtung für Energieinfrastruktur und Geothermie IEG, Bochum, Germany *Corresponding author: katrin.loeer@hs-bochum.de; katrin.loer@abdn.ac.uk © Seismological Society of America Volume 91 • Number 6 • November 2020 • www.srl-online.org Seismological Research Letters 3269 Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/91/6/3269/5176463/srl-2020022.1.pdf by University of Edinburgh user on 08 November 2020 thermal anomaly could potentially limit the volume of rocks in which secondary permeability may exist. We use three-component (3C) beamforming to extract structural information from ambient seismic noise. 3C beamforming is an array technique, which, like standard beamforming, not only estimates the dominant propagation direction and wavenumber of a recorded wavefield, but in addition determines the polarization of the wavefield by comparing phase shifts across different components (Riahi et al., 2013). As a result, different wave types can be distinguished and their propagation parameters analyzed separately. This allows us, for example, to estimate wavefield composition and surface-wave anisotropy, which is, however, beyond the scope of this study. Here, we consider fundamental mode Rayleigh waves only and extract dispersion curves from frequency–wavenumber (f-k) histograms; these are inverted for a shear-velocity depth profile using a reversible-jump Markov chain Monte Carlo (rj-McMC) algorithm. Although this algorithm is computationally expensive, it has the advantage of providing uncertainties for the velocity profile by finding the distribution of models that are consistent with data. 3C beamforming does not require impulsive (man-made or natural) seismic sources and is thus cheap, flexible, and applicable also in aseismic areas. Whereas cross-correlation-based ambient noise methods typically rely on month-long recordings, from beamforming, we extract stable dispersion curves from only 1 day of seismic noise data. Depending on the array geometry and seismic noise spectrum, the depth sensitivity of 3C beamforming can exceed that of other seismic methods by several kilometers, as we will show in this study. The analysis of four reflection seismic lines recorded across the LHVC, for example, provided 2D velocity maps and seismic sections down to 6 km at the most (Jousset, Ágústsson, et al., 2019). Ambient noise cross-correlation methods applied in the same area, but using a larger array, produce 3D tomographic Figure 1. (a) Simplified geological map of the Los Humeros volcanic complex (LHVC) and surrounding basement, on a shaded relief. The trace of the A-A′ geological cross section of panel (b) is shown. Triangles denote seismic station locations of the dense broadband (DB) network, circles denote geothermal wells. In the upper-right inset, the location of the LHVC within the TransMexican volcanic belt (TMVB) is indicated. (b) A–A′ schematic geological cross section showing the subsurface geometry of the main structures and stratigraphic units. Trace of the geological cross section is shown in panel (a). Modified from Norini et al. (2019). ENE, east-northeast; LH, Los Humeros caldera ring fault; LHh, inferred flexure plane of the Los Humeros trap-door caldera; LP, Los Potreros caldera ring fault; TF: thrust fault; RF, resurgence fault (red lines); WSW, west-southwest. The color version of this figure is available only in the electronic edition. 3270 Seismological Research Letters www.srl-online.org • Volume 91 • Number 6 • November 2020 Downloaded from http://pubs.geoscienceworld.org/ssa/srl/article-pdf/91/6/3269/5176463/srl-2020022.1.pdf by University of Edinburgh user on 08 November 2020 images down to a maximum of 10 km depth (Granados Chavarria et al., 2020; Martins et al., 2020). In a similar manner, a recent local earthquake tomography study provides information only of the upper 3–4 km (Toledo et al., 2020). We show that 3C beamforming provides information to greater than 10 km depth. In the following, we describe geology and available datasets, introduce both 3C beamforming and the rj-McMC inversion algorithm, and summarize our findings in Los Humeros and their implications for SHGSs, in general. Geology of LHVC The LHVC basement is composed of Mesozoic sedimentary rocks involved in the Late Cretaceous–Eocene compressive orogenic phase that generated the Mexican fold and thrust belt (sedimentary basement unit in Fig. 1) (Fitz-Díaz et al., 2017; references therein). The sedimentary basement rests above the Precambrian–Paleozoic crystalline basement of the Teziutlan Massif unit, made of greenschists, granodiorites, and granites (e.g., Suter, 1987; Suter et al., 1997; Ortuño-Arzate et al., 2003; Ángeles-Moreno, 2012; Fitz-Díaz et al., 2017) (Fig. 1a,b). Since the Eocene, the area underwent a limited extensional tectonic phase, associated with northeast-striking normal faults and the emplacement of Eocene–Miocene granite and granodiorite magmatic intrusions (Fig. 1a). The TMVB volcanic activity occurred from 10.5 to 1.55 Ma with the emplacement of fractured andesites, basaltic lava flows, and few volcaniclastic levels (old volcanic succession unit in Fig. 1) (e.g., Yanez and Garcia, 1982; Ferriz and Mahood, 1984; López-Hernández, 1995; Cedillo-Rodríguez, 1997; Carrasco-Núñez, Hernandez, et al., 2017; Carrasco-Núñez et al., 2018). Volcanic activity resumed ∼700 ka ago with the emplacement of the Pleistocene– Holocene LHVC (LHVC unit in Fig. 1) (e.g., Carrasco-Núñez, Hernandez, et al., 2017; Carrasco-Núñez et al., 2018). This volcanic complex represents a basaltic andesite–rhyolite system of two nested calderas, namely the outer Los Humeros caldera and the inner Los Potreros caldera (Carrasco-Núñez, Hernandez, et al., 2017; Calcagno et al., 2018) (Fig. 1a). The LHVC caldera stage occurred between ∼165 and ∼69 ka and consisted of two major ca