Characterization of Youngstown Induced Seismicity
Robert L. WalkerArman Khodabakhshnejad
Mork Family Department of Chemical Engineering and Materials ScienceInduced Seismicity Consortium
Outline
Background
Problem Statement
Correlation between Fluid Injection and Seismicity
Prior Work
Our Strategy
Conclusion
Event Background
(Credit USGS ,Youngstown 7½-minute quadrangle)
(Credit Ohio Department of Natural Resources, 2012)
Potential Examples of Induced Seismicity
(From Seeber et. al, 2004)
(From Nicholson et. Al, 1988)
1986?
In Northeast Ohio….
1987 & 2001?
In the Continental US….
(Credit National Research Council, 2012)
Why is This Important?
(From W. Y. Kim, 2008)
(Credit Organization of
American States)
(Credit University of
Arizona)
So You Want to Cause an Earthquake….
The time tested approach: Literature has suggested that Seismic
Tremors can also be generated from….• Mining• Other Earthquakes• Geothermal Energy Generation• Dams• Hydrocarbon Reservoir Depletion• Reservoir Changes [i.e. Lake]• Nuclear Tests• Skyscraper Construction [Taipei 101]• Hydraulic Fracturing• Wastewater Disposal• Rainfall and Snowmelt (one study, at least)
Mitigation: Spotlight on Ohio
• It is thought that larger (M3-4) events are preceded by micro-earthquakes
• Key to anticipating [and hopefully avoiding] larger events is a sensitive, robust seismometer network • Ohio has ~180
Class II injection wells, now classified as either “shallow” or “deep,” respective to a 7000 foot reference depth.
• Effective Oct. 2012, all “deep” wells are required to submit a seismic monitoring plan, or if possible create their own seismic network.
The Problem
(Credit National Research Council)
Unfortunately, seismometer networks may not be up to the task
ObjectiveDevelopment of a model that can detect an
increased likelihood of Induced Seismic events
(From M.D. Zoback, 2012)
This model will take inspiration from previous correlations, as well as the proposed “Traffic Light Model” of M. D. Zoback
Current Thinking
Water/fluid pressurein fault = p
• For an to occur, the stress must exceed the critical shear stress on the fault: τcritical = c + (σn - p)
• Function of Hydraulic Stress only
Majer, 2011
Working Hypothesis
Mechanisms of Seismicity
Mechanical based seismicity
Lubrication based seismicity
50 100 150 200 250 300 350 4000
0.5
1
1.5
2
2.5
33.5
4
4.5
Day
Mom
ent
Mag
nitu
de
Youngstown Seismicity Over Time
(Credit Ohio Department of Natural Resources, 2012)
Water/fluid pressurein fault affects coefficient of friction µτ critical = c + µ(σn - p)
Correlations Between Fluid Injection and Seismicity
(Majer, 2011, Relation after McGarr, 1976)
(Ake et. al, 2005)
Probabilistic Model
Model Development Flowchart
B Value Analysis
Nearby Seismic Events
∑ Energy Transfer
Corrected Pressure
Data
Data Analysis has yielded attributes related to seismic
activity
Adaptive NeuroFuzzy Inference System will
be employed to predict seismic events
Physically derived model will test
predictive model
Generated Catalog
from Gathered
Data Past Seismic Records
0 100 200 300 400012345
Injection data
Pressure
Injection Attributes
Artificial Intelligence
Qualitative & Quantitative
Models
Flow Rate
Pressure Fluctuation
s
Energy
Analogy
ANFIS
Lubrication Model
b – Value analysis from
generated catalog
PREDICTION
Northstar #1 Data
0 50 100 150 200 250 300 3500
0.2
0.4
0.6
0.8
1
1.2
Normalized Cumulative Injection Normalized Total Energy Release
Day
Normalize
d
QC/Initial Testing: Surface Pressure Estimation
QC/Initial Testing: Energy Transfer Check
After removing the high energy outlier….
𝐸𝑡𝑟𝑎𝑛𝑠𝑓𝑒𝑟𝑒𝑑=𝑃𝑟𝑒𝑠𝑠𝑢𝑟𝑒×𝑉𝑜𝑙𝑢𝑚𝑒
Looks awful lonely, doesn’t it?
Hm. Guess not. Suppose there’s more to it.
Conclusion
- Based upon initial analysis of available data, a pattern of seismicity can be identified.
- Cumulative fluid injection, injection pressure, and past seismic events can serve as fundamental components of a predictive model.
- We believe that the “dual mechanism” source of seismicity may be able to explain certain patterns of seismicity and perhaps large seismic events.
- If so, this pattern could as a basis for a predictive model.