We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
Machine Learning Reveals Additional Hydraulic Fracture‐Induced Seismicity in the Eagle Ford Shale.
- Authors
Fasola, Shannon L.; Brudzinski, Michael R.
- Abstract
A recent study found ∼90% of earthquakes in the Eagle Ford shale in Texas from 2014 to 2018 were spatiotemporally related to hydraulic fracturing (HF) wells. Since then, earthquakes have remained prevalent, including a new region of seismicity in Live Oak county. We sought to perform a deeper exploration of how HF has contributed to recent seismicity using template matching and repeating signal detection (RSD), which employs machine learning to search for repeating earthquakes with the potential to find something different than the template catalog. RSD identified new bursts of seismicity with a shorter S‐P time (∼2 s) than the previous catalog (>4 s), which temporally correlated with HF near station N4 735B. The short S‐P events have smaller magnitudes (ML < 2.0), consistent with the idea that HF‐induced seismicity in the Eagle Ford is likely more pervasive than previously reported, but detection is limited by the density of stations. RSD and template matching identified 1,600 earthquakes correlated with HF from 2019 to February 2020. We confirmed newly detected HF‐induced seismicity in Live Oak county did not occur until January 2019. Despite similar cumulative volume prior to and after 2019, the onset of detectable seismicity did not occur until HF injection exceeded 2 million barrels per month over this area, supporting the notion that injection flux is a stronger influence on the seismicity occurrence than cumulative volume. When considering the full catalog, the likelihood of seismicity also correlated with proximity to mapped faults, with 13% of wells <2 km from faults having seismicity compared to only 3% of wells >2 km. Plain Language Summary: We compared times and locations of hydraulic fracturing (HF) wells in the Eagle Ford shale in Texas with a catalog of earthquakes we enhanced through two methods of seismogram similarity detection. One method utilizes a known earthquake recording to search for similar recordings (template matching), while the other method (repeating signal detection) searches a broader range of detected signals to identify earthquake swarms. The latter method found a region of small‐magnitude earthquakes that was not previously detected. This result suggests that earthquakes induced by HF are more pervasive than previously reported, but detection is limited by how many seismometers are recording. Both methods identified ∼1,600 earthquakes correlated with HF from 2019 to early 2020. A new region of seismicity related to HF in Live Oak county illustrated that the total amount of fluids injected per month in this area (injection flux) has a stronger influence than the total volume injected, providing new insight into the physics of induced seismicity and strategies for mitigation. We found a higher likelihood of seismicity the closer HF wells were to mapped faults, supporting the notion that proximity of injection to susceptible faults is a critical factor. Key Points: Injection flux has a stronger influence on HF‐induced seismicity than total volumeMachine learning (RSD) helped to identify a new region of seismicity related to hydraulic fracturingHF‐induced seismicity more pervasive than previously reported but limited by density of stations
- Subjects
TEXAS; EAGLE Ford Shale; INDUCED seismicity; MACHINE learning; SHALE; EARTHQUAKE swarms; HYDRAULIC fracturing; EARTHQUAKE magnitude; SEISMOGRAMS
- Publication
Journal of Geophysical Research. Solid Earth, 2023, Vol 128, Issue 2, p1
- ISSN
2169-9313
- Publication type
Article
- DOI
10.1029/2022JB025436