Large pieces of equipment

Microseismic Event Relocation Using Local, High-Resolution Velocity Models and the Impacts on Locations and Magnitudes

01. Background

This study aims to contrast processing workflows, event locations, and magnitudes between TexNet and a private seismic monitoring array (operated by ESG Solutions), highlighting the significance of local arrays and high-resolution velocity models for seismicity monitoring. Seismic data from a private regional array in Howard County, TX is utilized to compare results from seismicity located with a global velocity model and TexNet’s location algorithm, with those generated using a proprietary workflow and detailed velocity models which were constructed from in-situ, local subsurface data provided by operators. This area has seen an increase in seismicity and felt events in recent years, leading to the implementation of seismic response plans - correctly classifying seismicity in these response areas is crucial to avoid incorrectly assigning blame and for the prevention of large events in the future. Accurate velocity models are a necessary tool for determining event depths and magnitudes with respect to these response plans.

Fig. 1 - IASP91 velocity model, cited by TexNet as the velocity model used forevent processing outside of the Delaware Basin. This model extends toover 6000km below the surface and contains little near-surface detail.

Fig. 2Similar P-wave velocity depth slices from the 3D PSDM Tomographic Velocity Model and Log & VSP Data Velocity Model (in depth)that were used to calculate event locations in this project. Both show shallow velocity variations and westward dipping features.

02. Methodology

To best understand the driving factors influencing differences in event locations and magnitudes between TexNet and ESG’s private array, we compared the following datasets:

  • (A) - 3D Pre-Stack Depth Migration (PSDM) Tomographic Velocity Model & ESG’s processing workflow.
  • (B) - Log & VSP Data Velocity Model (in depth) & ESG’s processing workflow.
  • (C) - IASP91 velocity model & ESG’s processing workflow.
  • (D) - TexNet’s processing results (taken from the TexNet online catalog) using theIASP91 velocity model.


A subset of 11 high-magnitude events recorded by both TexNet and ESG’s private array has been used for comparison. Event epicentral locations, depths, and magnitudes were compared to evaluate the impact and reliability of each velocity model and processing combination.

Fig. 3 - The traces above show the three velocities models within the samedepth range. While basement velocities are similar, shallow velocitiesdiffer significantly. The Log & VSP model has the slowest near-surfacevelocities, while the IASP91 model has fast near-surface velocities.

03. Analysis

This comparison used 11 high-magnitude events that occurred between June 6th - October 28th, 2022. The 3D simplex location method, a robust algorithm for determining hypocentral calculations, was employed with the same stations and arrival time picks between the 3D PSDM, Log & VSP, and IASP91 models. These results are compared to locations reported by TexNet for the same events. All local magnitudes were calculated using the equation described by Kavoura et al (2020). The events reported with ESG's processing used only TexNet stations to calculate ML in order to compare values more evenly - including closer private stations can skew local magnitudes, which are heavily influenced by a station's distance from an event. Moment magnitudes were calculated using a net moment approach to fit event frequency spectra. A Mw estimation for TexNet's catalog was derived by constraining events to their reported locations and adjusting spectral amplitudes accordingly.

Fig. 4 - Relocated event depths vs northing are shown above, with a histogram highlighting mean depths on the right. The Log & VSP model depths are consistently shallower while the TexNet depths are the deepest on average. This can be attributed to the near-surface velocities in each model and the station coverage around these events.

Fig. 5 - The map above shows broadband seismometer locations (colored by provider) and event locations (colored by relocation method) within the Howard Co seismic monitoring array. The higher magnitude events analyzed in this project are circled to show the spread in locations per event, with the middle clusters containing several different events.

04. Results

The main results from this processing comparison are listed below:

  • Event epicentral locations show significant spread based on proximity to the private array and station coverage - events with full station coverage exhibited less horizontal dispersion.
  • Velocity models with slower near-surface velocities produced shallower event depths. The Log & VSP model had a mean depth of 10,175ft, the 3D PSDM model mean depth was 19,900ft, while TexNet's mean depth was 24,815ft due to fast shallow velocities.
  • Moment and local magnitudes are higher when events locate deeper due to an increased source-sensor distance and less dense station coverage - TexNet's Mw estimations are 0.4Mw and 0.3ML above those of the Log & VSP model events.



Both the Log & VSP and 3D PSDM models produced 33% fewer 3.0ML events than the TexNet catalog, highlighting the importance of local and shallow geological data when constructing velocity models used in seismic response decision-making and hazard assessment.

Fig. 6 - The histograms below highlight differences in moment and local magnitudes for the 11 reprocessed events. Shallow event depths lead to lower magnitudes on both scales, while deeper TexNet locations provide higher magnitude estimations. This difference is most notable with respect to moment magnitudes, where there is an average difference of 0.4Mw between the Log & VSP model locations and TexNet locations. Local magnitude means differ by 0.3ML between the same models.

05. Extended Catalog Analysis

Along with more accurate velocity models and event locations, local monitoring arrays allow for improved event detection in areas relevant to operators. ESG's monitoring array in Howard Co detected nearly four times as many Mw 1.0+ events from January - June 2023 compared to TexNet. A velocity model comparison was conducted with this extended ESG catalog to determine if similar results would be seen on a more robust dataset. TexNet's results were excluded because most of these smaller events were not reported. The expanded catalog underscores that slower near-surface velocities produce shallower event depths and generally lower magnitude estimations. The Log & VSP model again provided the shallowest events and lowest MLs, with nearly the same Mw estimations as the 3D PSDM model. Consistency in magnitude calculations with the local array highlights the importance of nearby stations for reducing bias and magnitude overestimations.

Fig. 7

Fig. 8 - Extended ESG catalog event depths are shown above, with histograms highlighting mean depths on the right. The Log & VSP model produced the shallowest depths on average and the IASP91 model producing the deepest events. These depths highlight that velocity models created with accurate near-surface velocities can locate events shallower than generalized velocity models.
Fig. 9 - Moment and local magnitude histograms for the extended ESG catalog. While these results are mostly consistent between models, slow near-surface velocities still tend to produce lower magnitudes on both scales. Greater consistency among these results highlights the importance of using nearby stations to reduce magnitude overestimations.

06. Conclusion & Significance

  • Variations in near-surface velocities yield notable differences in event depths and magnitudes. High-resolution velocity models accommodate near-surface velocities based on local geological information, while global models can be oversimplified. Differences in event depths due to variations between velocity models can lead to magnitude discrepancies. Events located deep due to generalized velocities may be assigned an overestimated magnitude, leading to incorrect decision-making with respect to active operations and reporting.
  • Event locations and magnitude estimations can be improved using dense local arrays which reduce uncertainty in hypocentral locations and provide better spatial coverage with closer proximity recordings.
  • It is recommended that both regulatory bodies and operators consider integrating local monitoring arrays and high-resolution velocity models into their seismic monitoring practices to acquire the most accurate event locations and magnitudes.
Fig. 1 - IASP91 velocity model, cited by TexNet as the velocity model used forevent processing outside of the Delaware Basin. This model extends toover 6000km below the surface and contains little near-surface detail.
Fig. 2 - Similar P-wave velocity depth slices from the 3D PSDM Tomographic Velocity Model and Log & VSP Data Velocity Model (in depth)that were used to calculate event locations in this project. Both show shallow velocity variations and westward dipping features.
Fig. 3 - The traces above show the three velocities models within the samedepth range. While basement velocities are similar, shallow velocitiesdiffer significantly. The Log & VSP model has the slowest near-surfacevelocities, while the IASP91 model has fast near-surface velocities.
Fig. 4 - Relocated event depths vs northing are shown above, with a histogram highlighting mean depths on the right. The Log & VSP model depths are consistently shallower while the TexNet depths are the deepest on average. This can be attributed to the near-surface velocities in each model and the station coverage around these events.
Fig. 5 - The map above shows broadband seismometer locations (colored by provider) and event locations (colored by relocation method) within the Howard Co seismic monitoring array. The higher magnitude events analyzed in this project are circled to show the spread in locations per event, with the middle clusters containing several different events.
Fig. 6 - The histograms below highlight differences in moment and local magnitudes for the 11 reprocessed events. Shallow event depths lead to lower magnitudes on both scales, while deeper TexNet locations provide higher magnitude estimations. This difference is most notable with respect to moment magnitudes, where there is an average difference of 0.4Mw between the Log & VSP model locations and TexNet locations. Local magnitude means differ by 0.3ML between the same models.
Fig. 7
Fig. 8 - Extended ESG catalog event depths are shown above, with histograms highlighting mean depths on the right. The Log & VSP model produced the shallowest depths on average and the IASP91 model producing the deepest events. These depths highlight that velocity models created with accurate near-surface velocities can locate events shallower than generalized velocity models
Fig. 9 - Moment and local magnitude histograms for the extended ESG catalog. While these results are mostly consistent between models, slow near-surface velocities still tend to produce lower magnitudes on both scales. Greater consistency among these results highlights the importance of using nearby stations to reduce magnitude overestimations.