Oklahoma's induced seismicity strongly linked to wastewater injection depth

See allHide authors and affiliations

Science  16 Mar 2018:
Vol. 359, Issue 6381, pp. 1251-1255
DOI: 10.1126/science.aap7911
  • Fig. 1 Wastewater injection and seismicity in Oklahoma.

    (A). Map showing the surface expression of key geologic provinces of Oklahoma (41) and class II UIC wells (35) (related to oil and gas production). Black polygon defines the “Oklahoma induced seismicity zone” delineated by the USGS (5). (B) Map showing total injected volume (bbl) for 2011 to 2015 (30-km Gaussian kernel density). Symbols denote locations of earthquakes from 2011 to 2016, including the 5 November 2011 Prague (7, 10, 11) and 3 September 2016 Pawnee (18, 21, 28, 44) events. (C) Map showing total seismic moment release from 2011 to 2016 (20-km Gaussian kernel density), with mapped faults in the sedimentary cover including the Nemaha uplift (arrow) also shown (42) [this does not include all large-scale basement structures (17)]. Yearly injection volumes and seismic moment release are shown in fig. S2.

  • Fig. 2 BN structure, unconditional rank correlation coefficients, and cumulative distribution functions (CDFs) for Test 2.

    (A) The saturated network and (B) the unsaturated network. Arcs denote influence between nodes. (C) Unconditional rank correlation coefficients for all nodes with log10 annual moment release (empirical data and BNs). This correlation is a measure of the strength of influence of each observable on the node of interest (moment release). The BN uses joint normal copulae to represent the empirical data—the closer the values in the BN rank correlation matrix are to the empirical rank correlation matrix, the better the model approximates the data (matrix determinants can also be compared). Full correlation matrices are provided in table S5. (D and E) CDFs of log10 annual moment release for the BN forecast (mean and median estimate), linear regression model, and empirical data for (D) saturated BN and (E) unsaturated BN. This function shows how closely the model output distribution matches the observations. Results are shown for Test 2–learning using 90% of the observations, testing with the remaining 10% (see fig. S11).

  • Fig. 3 Probability distributions for well depth and distance to basement using BN inputs for January 2011 to December 2015 (wells ≥1 km deep and injection records >10,000 bbl/month).

    Probability density for (A) well depth below surface; (B) depth relative to basement and (C) distance to basement; identifying (i) cases with low or no seismic activity (log10 moment in 1 year ≤13; shaded blue) and (ii) cases with relatively high seismic activity (log10 moment in 1 year ≥15; shaded pink). The distributions in (B) and (C) show that wells drilled closer to the basement are associated with higher seismic moment release. Distributions for injection volume are shown in fig. S7.

  • Fig. 4 Simulated impact of raising the injection well level or capping monthly injection volume.

    (A) For a subset of records, well depth was reduced by 1 km and used to generate revised estimates for annual moment release using the BN (colored points) and linear regression models (black points). The plot shows change in annual moment versus original well depth and predicts the greatest reduction in moment release for wells initially closest to the basement interface. The joint effect of cumulative volume is also seen—wells with higher cumulative injections are expected to experience proportionally greater moment reduction. (B) Predicted change in mean annual moment resulting from a regulatory limit of 15,000 bbl/day (450,000 bbl/month) versus monthly injection with no cap. This plot shows cases of high injection only (where the limit is applied). Points are colored by total reduction in annual volume to illustrate the importance of maintaining injection limits over time. Inset: CDFs for moment release (BN mean estimates) show the overall effect on all high-volume wells.

Supplementary Materials

  • Oklahoma's induced seismicity strongly linked to wastewater injection depth

    Thea Hincks, Willy Aspinall, Roger Cooke, Thomas Gernon

    Materials/Methods, Supplementary Text, Tables, Figures, and/or References

    Download Supplement
    • Materials and Methods
    • Figs. S1 to S18
    • Tables S1 to S7
    • References
    Data File S1

Stay Connected to Science

Navigate This Article