The Raton Basin is known as an area of injection induced seismicity for the past two decades, but the reactivated fault zone structures and spatiotemporal response of seismicity to evolving injection have been poorly constrained in the past due to scarce public monitoring. Application of a machine-learning phase picker to four years of continuous data from a local array enables the detection and location of ~38,000 earthquakes. The events between 2016-2020 are ~2.5-6 km below sea level and range from ML<-1 to 4.2. Most earthquakes occur within previously identified ~N-S zones of seismicity, however our new catalog illuminates these zones are composed of many short faults with variable orientations. The two most active zones, the Vermejo Park and Tercio, are potentially linked by small intermediate faults. In total, we find ~60 short (<3 km) basement faults with strikes from WNW to slightly east of N. Faulting mechanisms are predominantly normal but some variability, including reverse dip-slip and oblique-slip, is observed. The Trinidad fault zone that hosted the 2011 Mw 5.3 earthquake is quiescent during 2016-2020, likely in response to decreased wastewater injection after 2012 and the shut-in of two nearby wells in 2015. Unlike some induced seismicity regions with higher injection rates, Raton Basin frequency-magnitude and spatiotemporal statistics are not distinguishable from tectonic seismicity. The similarity suggests that induced earthquakes in the Raton Basin are dominantly releasing tectonic stress.

Accurate earthquake location and magnitude estimation play critical roles in seismology. Recent deep learning frameworks have produced encouraging results on various seismological tasks (e.g., earthquake detection, phase picking, seismic classification, and earthquake early warning). Most existing machine learning earthquake location methods utilize waveform information from a single station. However, multiple stations contain more complete information for earthquake source characterization. Inspired by recent successes in applying graph neural networks in graph-structured data, we develop a Spatio-Temporal Graph Convolutional Neural Network (STGCN) for estimating earthquake locations and magnitudes. Our graph neural network leverages geographical and waveform information from multiple stations to construct graphs automatically and dynamically by an adaptive feature integration process. Given input waveforms collected from multiple stations, the neural network constructs different graphs and fuses spatial-temporal consistency effectively from various stations based on graphs’ edges. Using a recent graph neural network and a fully convolutional neural network as baselines, we apply STGCN to earthquakes cataloged by Southern California Seismic Network from 2000 to 2019 and induced earthquakes collected in Oklahoma from 2014 to 2015. STGCN yields more accurate earthquake locations than those obtained by the baseline models and performs comparably in terms of depth and magnitude prediction, though the ability to predict depth and magnitude remains weak for all tested models. Our work demonstrates the potential of using graph neural networks and multiple stations for better automatic estimation of earthquake epicenters.

Seismicity in the Raton Basin over the past two decades suggests reactivation of basement faults due to wastewater injection. In the summer of 2018, 96 short-period three-component nodal instruments were installed in a highly active region of the basin for a month. A machine-learning based phase picker-(PhaseNet) was adopted and identified millions of picks, which were associated with events using an automated algorithm – REAL (Rapid Earthquake Association and Location). After hypocenter relocation with hypoDD, the earthquake catalog contains 9259 M -2.2 – 3 earthquakes focused at depths of 4-6km. Magnitude of completeness (Mc) varies from -1 at night to -0.5 in daytime, likely reflecting noise variation modulated by wind. The clustered hypocenters with variable depths and focal mechanisms suggest a complex network of basement faults. Frequency-magnitude statistics and the spatiotemporal evolution of seismicity are comparable to tectonic systems.