The largest earthquakes worldwide occur in ?subduction zones? where plates under the oceans descend beneath continents and into Earth?s deep interior. These earthquakes, which occur on faults that separate the descending oceanic plate from the overlying continental plate, can have devastating human and economic impacts. Regions of the world at risk from these earthquakes include the U.S. Pacific Northwest and Alaska, Central and South America, Japan, and Southeast Asia. In this project, Meltzer, Beck, and their team will analyze data recorded by seismometers that they deployed in northern Ecuador between 2020 and 2022, in coordination with European scientists. Their analyses will result in detailed 3D images of the subduction zone beneath Ecuador, where a magnitude 7.8 earthquake occurred in 2016. Using machine learning techniques, the researchers will also comb the seismometer data for signs of small earthquakes, whose locations can outline the geometry of the main subduction zone fault and other local faults. The research team will use the images and earthquake data to infer which parts of the main subduction zone fault could give rise to large earthquakes. These findings and related activities by the research team and Ecuadoran colleagues will help Ecuador anticipate and plan for potentially devastating subduction zone earthquakes. Increasingly high resolution seismic studies reveal the complexity and diversity of structure and physical properties within and between subduction zones. Downdip and along-strike heterogeneity influences plate coupling and slip behavior on the plate interface, in the downgoing plate, and in the upper plate, ultimately culminating in large magnitude earthquakes with significant human and economic impacts. Between October 2020 and November 2022, the team collected onshore-offshore active and passive seismic data recorded by a dense array of 3-component short period nodes and broadband sensors in coastal Ecuador. These data were acquired as a complement to the HIPER project (High Resolution Imaging of the Pedernales Earthquake Rupture Zone) supported by French and German funding agencies. The data were collected in and adjacent to the 2016 Mw 7.8 Pedernales rupture zone to significantly densify the onshore portion of HIPER for active-source recording, and to widen the aperture of passive recording to improve lateral coverage and imaging at depth across the forearc. The combined arrays present an exceptional opportunity for multiscale 3D imaging to examine the relationship between physical properties, structure, fluids, and variations in slip behavior from seismic to aseismic. Analysis of these data will employ machine learning to significantly increase source detection, and a combination of traveltime and ambient noise tomography and high frequency receiver functions as input to a multi-step series of inversions to constrain earthquake locations and seismic structure along and above the seismogenic zone and across the forearc. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.