Examination of Possibly Induced Seismicity from Hydraulic Fracturing in the Eola Field, Garvin County, Oklahoma

[editor’s note: the full report from the Oklahoma Geological Survey is available as a downloadable PDF at the link below]

http://www.ogs.ou.edu/pubsscanned/openfile/OF1_2011.pdf

Summary

On January 18, 2011, The Oklahoma Geological Survey (OGS) received a phone call from a resident living south of Elmore City, in Garvin County, Oklahoma, that reported feeling several earthquakes throughout the night. The reporting local resident had also offered that there was an active hydraulic fracturing project occurring nearby. Upon examination there were nearly 50 earthquakes, which occurred during that time. After analyzing the data there were 43 earthquakes large enough to be located, which from the character of the seismic recordings indicate that they are both shallow and unique. The earthquakes range in magnitude from 1.0 to 2.8 Md and the majority of earthquakes occurred within about 24 hours of the first earthquake. Careful attention and significant effort was put into obtaining the most accurate locations possible and gaining a reasonable estimate in the error in locations. The nearest seismic station is 35 km away from where the earthquakes occurred. Formal errors in location are on the order 100-500 m horizontally and about twice that for depth. Examination of different velocity models would suggest that the uncertainties in earthquake locations should be about twice the formal uncertainties. The majority of earthquakes appear to have occurred within about 3.5 km of the well located in the Eola Field of southern Garvin County. The Eola Field has many structures, which may provide conduits for fluid flow at depth. The well is Picket Unit B well 4-18, and about seven hours after the first and deepest hydraulic fracturing stage started the earthquakes began occurring. It was possible to model 95% of the earthquakes in this sequence using a simple pore pressure diffusion model with a permeability of about 250 mD (milliDarcies). While this permeability may be high it is less than those reported for highly fractured rock. The strong correlation in time and space as well as a reasonable fit to a physical model suggest that there is a possibility these earthquakes were induced by hydraulic-fracturing. However, the uncertainties in the data make it impossible to say with a high degree of certainty whether or not these earthquakes were triggered by natural means or by the nearby hydraulic-fracturing operation.

 

Advertisements

Fracking Hell: Oklahoma, Earthquakes, Injection Wells, and Data Accessibility

Attempting to mash the earthquake and underground data into a cohesive user-interface has proved to be, to put it mildly, daunting.  It was much easier to find sources of earthquake data than it was to find any source of well data with any fields relevant to my needs.

The earthquake data was relative easy to come by, for example I found the following sources:

I downloaded the entire earthquake dataset from the Advanced National Seismic System (ANSS) beginning in 1898 through the present day and imported the data into an Apache Lucene index.  In short order I had a searchable earthquake index lacking but a few location-centric fields:

  • The country in which the earthquake occurred.
  • The state in which the earthquake occurred.
  • The county in which the earthquake occurred.

In order to associate the above needed fields with the earthquake data, I downloaded two ESRI-formatted shapefiles from the National Atlas:

And one shapefile from Mapping Hacks:

I then wrote a Java program that would read each earthquake record and link it to its associated country, state, and county available from the respective shapefile of each.  To do this I used a Java library at GeoTools-8.0-M3-bin.zip from GeoTools.org.

Well data was much more difficult to come by, especially with any fields relevant to my needs, for example:

  • The type of well, for example “oil”, “gas”, “inj” (for injection) was available as data, just not available as a field upon which one could query.  In other words, I could not query for just underground injection wells (“inj”).
  • The date each well became active, let alone its filing date, was not available via the web interface.
  • The location of each well, in latitude and longitude, was not available either.
  • Given the lack of the above information, I didn’t even concern myself with the lack of well depth information.

As an exercise in personal fortitude, I downloaded the wells for each county in the State of Oklahoma from the Oklahoma Corporation Commission’s Well Data System into one Excel spreadsheet per county.  I then wrote a Java program that read the well data within each county’s Excel spreadsheet and posted it to an Apache Lucene index.  I then zipped the Apache Lucene index and pushed it a web site so that it could be queried and viewed using Apache Solr’s VelocityResponseWriter browser interface.  The results of this effort can be viewed and queried here.

So, in concluding this post, I find the earthquake data adequate for my present needs but the well data lacking any useful date or location information to allow me to associate the earthquakes to the wells by either location or time.  As I am a persistent researcher, my next post will detail my further attempts at locating and downloading well data.

Commentary on Oklahoma Geological Survey’s “Position Statement on Triggered or Induced Seismicity”

The following is my commentary, [bracketed in green text], concerning a faithful reproduction of the Oklahoma Geological Survey’s “Position Statement on Triggered or Induced Seismicity” statement. The only differences in the reproduction are to be found in the formatting of the bullet lists as the original, visible here, had different bullet imagery and some “[sic]” notes within the statement where some grammatical errors were present in the original PDF.

Oklahoma Geological Survey
The University of Oklahoma
Mewbourne College of Earth & Energy
G. Randy Keller, Director and State Geologist

Position Statement on Triggered or Inducedi Seismicity

    • We take allegations of triggered seismicity seriously and are actively assessing the possibility that this may be occurring in Oklahoma. [Editor's note: It is a good thing that OGS is taking these recent swarms of earthquakes seriously.]
  • It is well understood that earthquakes can be triggered by fluid injection at depth. [Editor's note: It is also a good thing that OGS acknowledges that swarms of earthquakes can occur due to injection activities.]
    • It is also well documented from the historical and geologic record that Oklahoma frequently has naturally occurring earthquakes. [Editor's note: It is not the frequency of earthquakes that is at question, rather it is rising quantity of earthquakes of significant historical magnitude that has caught the attention of concerned citizens. One more significant point to be made is "that Oklahoma frequently has naturally occurring earthquakes"? Does this phrase imply that the author has a ready method, without an extensive scientific study, to easily identify which earthquakes are "natural" and which are not? If so, as the author suggests in the aforementioned "frequently..." phrase, couldn't the non-natural earthquakes be identified as those not readily identified as "natural" utilizing the implied easy process?]
    • It is true that the past few years have seen a significant increase in earthquake activity within Oklahoma.  While we are studying the possibility that some of this activity could be related to oil and gas operations, it is unlikely that all of the earthquakes can be attributed to human activities. [Editor's note: The "unlikely that all of the earthquakes can be attributed to human activities" goes without saying as that is not what is at issue here. Rather it is the recent swarms of earthquakes that appear to be co-located with some injection wells that are of concern, especially the recent rise of earthquakes of significant historical magnitude.]
      • It is however possible that there have been incidents of triggered seismicity within Oklahoma over the past several years. [Editor's note: Alright, I'll concede that the author of this statement might be signaling some open-mindedness with this sentence.]
    • If cases of triggered seismicity are identified, our goal is to adequately and thoroughly document and research these cases, so that proper steps can be taken to mitigate the likelihood of future triggered seismicity.[Editor's note: "If cases of triggered seismicity are identified" is a loaded phrase with "identified" being the operative word. I've read many a scientific study that concluded with phrases of "waffle words" indicating that more study was needed before a definitive answer might be possible. For an applicable analogy, please read up on the history of research into the cancer causality of cigarette smoking.]
      • We will also attempt to work collaboratively with operators of these wells to gain as great of an understanding of the triggering process as possible. [Editor's note: "attempt to work collaboratively with operators of these wells" is definitely the operative phrase in this sentence. Based upon my efforts to obtain oil well information from both state agencies and oil industry sources for "TheFrackingWall" blog, I can save the OGS a lot of grief by certifying that Hell hath no ice machines yet. Simply put, "collaboration" is not going to occur on a proactive basis on the part of state agencies or private sector actors. Case and point: Tried to obtain any latitudes and longitudes for oil and gas wells from the Oklahoma Corporation Commission's Well Browser tool lately? Yes, I can personally certify that only piping hot java is served in Hell these days.]
      • When earthquakes appear to be triggered, there are methods to, with the cooperation of operators, assess whether seismicity truly is triggered or not.[Editor's note: Once again, it is delusional for anyone to think that there will be "cooperation of operators". The best one can hope for is "coercion of operators". Once again, has the author of this statement attempted to use the well browsing tool at http://www.fracfocus.org? As a software developer of thirty years experience, I can certify that the well browser at FracFocus.org is not friendly to analysts of any sort. If I'd have to speculate, I'd say that FracFocus.org was implemented in the manner that it was for the very reason that it would be vexing for analysts of any type. How does the author of this statement, after seeing the frustrating web site that is FracFocus.org, think that s/he will receive any significant "cooperation of operators" in the effort to "assess whether seismicity truly is triggered"? The author is more likely to enjoy the "pleasures" of delays and obfuscations than that of cooperation.]
    • We consider a rush to judgment about earthquakes being triggered to be harmful to state, public, and industry interests. We are taking a measured and scientific approach to addressing issues so that any conclusion that earthquakes are linked to oil and gas activities can be scientifically defensible. [Editor's note: Oddly enough, this is where anecdotal evidence is more helpful. A time-series analysis of earthquake swarms coupled with the start of injection operations of suspect wells will likely return more common-sense evidence of a correlation than a massive in-depth study more likely to conclude with the aforementioned "waffle words". In addition, it might be more productive to conduct a "toggle-switch" experiment with a suspect injection well. "Toggle" the suspect injection well "off" and monitor the affected area to see if the earthquakes subside over time. Then "toggle" the suspect injection well "on" at its former injection pressure and monitor for any increase in earthquake swarms. Wash, rinse, repeat. This sort of approach is done all the time in software development and it is highly effective in finding hidden bugs and undesirable behavior.]
  • We are currently working on addressing cases in Oklahoma where triggered seismicity has been suggested.  This requires careful examination of a large amount of complex data and does not happen quickly.[Editor's note: Once again, wouldn't a time-series analysis reduce the effort required to come to a common-sense conclusion about triggered seismicity? Additionally, wouldn't a "time-series tripwire" system be a reasonable proactive undertaking? That is, a system utilizing past time-series swarm patterns in the vicinity of suspect wells might be able to proactively raise "red flags" about other injection wells.]
  • We have also completed an evaluation of one such case documented in OGS Open File (OF1-2011) that is available from the OGS website.  This case involved hydraulic fracturing of a well in Garvin County in South-Central Oklahoma.[Editor's note: One study, noted without a hyperlink, that ended with "waffle phrases" in the first paragraph of the conclusions is not an omen of good portent as to the results of any future studies. As a matter of fact, it is the first paragraph of the "Conclusions" of the mentioned study that provides the proof needed by skeptics that fluid injection will ever be "identified" as the cause of the earthquake swarms in question.]
  • We are actively working with and advising the Oklahoma Corporation Commission concerning issues of possible triggered seismicity.[Editor's note: It would be to all parties interest if the OCC modified its well browser site to return latitudes and longitudes of wells as well as a comprehensive download covering all of the wells in the database.]
  • It is important to note that earthquake processes in the stable interior of continents occur on the order of hundreds to ten’s [sic] of thousands of years while meaningful earthquake monitoring in Oklahoma and the central United States as a whole has only been possible for about 40 years.[Editor's note: Once again, it's not about the scale of hundreds or thousands of years, rather it's about the swarms of earthquakes occurring in the vicinity of suspect injection wells.]
    • Earthquakes often cluster in space and time in any tectonic setting, making any short-term trends in seismicity difficult to interpret.[Editor's note: Wax-on, wax-off, Daniel-san. If in doubt, "toggle" the suspect injection well "off" and "on". Just make sure that the suspect well injects at the actual pressure they were injecting during the earthquake swarm, not the "nominal" pressure with which they claim they were injecting. To paraphrase President Ronald Reagan: "Trust, yet verify."]
    • Many processes are thought to occur within the Earth that could potentially generate the seismicity patterns observed in recent years.[Editor's note: Once again, "toggle" the suspect well "off" and "on". In addition, over time, injection wells should start showing somewhat distinctive patterns of earthquakes patterns when associated with their injection logs.]
    • Most earthquakes in Oklahoma do not occur on well-known faults, but there are many undocumented faults throughout Oklahoma due to its complex geologic history.  Thus, we are working on a new fault map for the state.[Editor's note: Noting that one is working on a more comprehensive fault map for the state is not a helpful statement. It implies that one will need "cooperation of operators" to obtain the needed seismic data to do so. It is reasonably questionable as to whether cooperation is in the operators economic self-interest. As a general rule, these sorts of studies may render useful results many years down the line but by that time the damage is already done.]

________________________

iInduced seismicity is the more colloquial term, but triggered seismicity is the more accurate term for earthquakes inadvertently cause [sic] by anthropogenic activities.  This indicates that the stress released in the earthquake was accumulated through natural processes, but the mechanism that caused the stress release was due [to] the affects [sic] of human activities.[Editor's note: The use of the word "inadvertently" gives the impression that triggered seismicity is unintentional or accidental, or more succinctly a by-product of innocence. Use of such words could also signal the likely outcome of any studies conducted by the overseen agency.]

100 E. Boyd, Room N-131, Norman, Oklahoma 73019-0628
Phone (405) 325-3031  Fax (405) 325-7069 www.ogs.ou.edu
A STATE AGENCY SERVING THE PEOPLE OF OKLAHOMA

Preliminary Analysis of FracFocus.org Data

Abstract

This post covers the oil/gas well and chemical data extracted from the FracFocus.org website by the staff of SkyTruth.org.  The FracFocus.org website is gradually becoming the default repository for general hydro-fracturing well information as well as the chemicals used on each well during the fracturing process.  The “findings” I present will be sample  Structured Query Language (SQL) queries of a substantial subset of the extracted FracFocus.org data (800,000+ records).

FracFocus.org

When I first saw FracFocus.org’s website, I thought I had found an answer to my oil and gas well location information difficulties, as well of chemicals used in their operations.  Then I tried to use it and encountered another “fracking wall.”  As objective as I can be now and in my professional opinion based upon my experience cited in my biography below, FracFocus.org’s website at that time was intentionally constructed to make information extraction difficult.

I found a Frequently Asked Questions webpage upon which I found the following question:

“Are the records from FracFocus available in a digital format such as Excel?”

It was answered as follows:

“No. FracFocus was originally designed to serve records one at a time in Adobe pdf format in order to ensure accurate, unaltered and uncompromised data. Consequently the chemical information gathered does not currently reside in a database or spreadsheet format.”

Another question on the same page was as follows:

“Why can’t the system show me the information on more than one well at a time?”

It was was answered as follows:

“The purpose of the FracFocus records presentation system was to provide those who may live near a well that has been fractured with information concerning the materials used to fracture the well.  All information other than the information used in the search form is available only in an Adobe pdf format.  As such, information such as Ingredients, Trade Names and CAS numbers is not available for search or data aggregation purposes.”

It was this answer that surprised me the most.  Once again, I was stymied.  At the time this was only a “night job” self-education project, so I moved on to looking for well information from more easily-obtained sources.

SkyTruth.org Extracts Data from FracFocus.org

Several months later, while surfing the web, I noted that SkyTruth.org had extracted data from the FracFocus.org website.  I went to the article and attempted to obtain the data from the links they had posted on the http://frack.skytruth.org/fracking-chemical-database/frack-chemical-data-download web page.  I found it to be a little bit difficult to use for analytical reasons, so I applied some of my programming skills and pulled all of the extracted data into two data sets: one of “reports” that contained information about each well, and another of “chemicals” citing which chemicals were used in each reported well.  I then joined these two data sets into a “blended” data set and pushed all three of the resulting data sets to the “Download” section of an open-source repository at http://code.google.com/p/fracking-analysis/.

Database Preparation

As I wished to do some Structured Query Language (SQL) analysis of the extracted data and take a swipe at creating a high-speed, easily-searchable website using Apache Solr 4 as the back-end search engine, I imported the data into two tables within a MySQL database on an Amazon Web Services (AWS) Relational Database Service (RDS) instance.  I used an AWS RDS instance so that I could expand or contract its capacity and capability at will and as needed.  Having been pointed to a source of Chemical Abstract Number  information coupled with recognized and suspected toxicities, I consolidated this data into a table suitable for import into the MySQL database as well.

Once these three tables, as well as a “view” that joined the chemicals to their respective well reports, had been created in the MySQL database, I backed it up and posted it to the open-source repository at http://code.google.com/p/fracking-analysis/downloads/detail?name=frackanalysis_db_backup.zip&can=2&q=.

The tables presently contained within the database are as follows:

  • “chemicals” – a table of approximately 800,000 records citing information about each chemical documented as having been used on the reported well.
  • “reports” – a table of approximately 27,000 records citing information about each well reported as of September 21, 2012.
  • “uvw-chemicals-reports” – a view joining the each record in the “chemicals” table with its respective “reports” table.
  • “cas_chemicals” – a table of approximately 8,000 rows citing information about each chemical with its noted toxicity (e.g. cancer, respiratory, blood) and its category (e.g. recognized or suspected).

The above database and the tables within it are the basis of the queries cited below.

Simple Analytical Queries

Let’s start off with some simple analytical queries of the database.  These will be prefaced with an explanation of their purpose in the format of a natural language query.

“What is the maximum number of chemicals entered for a reported well?”

SELECT MAX(`Count`) Max_Count FROM (SELECT COUNT(*) `Count`    FROM frackanalysis.`chemicals` GROUP BY pdf_seqid) subqry1;

Which gives one an answer of 310.

“What is the minimum number of chemicals entered for a reported well?”

SELECT MIN(`Count`) Min_Count FROM (SELECT COUNT(*) `Count`    FROM frackanalysis.`chemicals` GROUP BY pdf_seqid) subqry1;

Which gives one an answer of 1.

“What is the average number of chemicals entered for a reported well?”

SELECT AVG(`Count`) Avg_Count FROM (SELECT COUNT(*) `Count`    FROM frackanalysis.`chemicals` GROUP BY pdf_seqid) subqry1;

Which gives one an answer of 29.0704.

More Complex Analytical Queries

Let’s continue with some more complex analytical queries of the database.  These will be prefaced with an explanation of their purpose in the format of a natural language query.

“What is the maximum number of chemicals entered for a reported well that are recognized carcinogens?”

SELECT MAX(`Count`) Max_Count FROM (SELECT COUNT(*) `Count` FROM frackanalysis.`chemicals` c INNER JOIN frackanalysis.`cas_chemicals` cc ON c.`cas_number` = cc.`CAS_EDF_ID` WHERE c.`cas_type` = ‘valid’ AND cc.`Toxicity` = ‘cancer’ AND cc.`Category` = ‘recognized’ GROUP BY pdf_seqid) subqry1;

Which gives one an answer of 6.

“What is the average number of chemicals entered for a reported well that are recognized carcinogens?”

SELECT AVG(`Count`) Avg_Count FROM (SELECT COUNT(*) `Count` FROM frackanalysis.`chemicals` c INNER JOIN frackanalysis.`cas_chemicals` cc ON c.`cas_number` = cc.`CAS_EDF_ID` WHERE c.`cas_type` = ‘valid’ AND cc.`Toxicity` = ‘cancer’ AND cc.`Category` = ‘recognized’ GROUP BY pdf_seqid) subqry1;

Which gives one an answer of 1.2445.

A More Interesting Analytical Query

Let’s continue with some more interesting analytical queries of the database.  These will be prefaced with an explanation of their purpose in the format of a natural language query.

“What are the most reported chemicals that are recognized carcinogens in descending order?”

SELECT * FROM (SELECT cc.`Chemical_Name`, COUNT(*) `Count` FROM   frackanalysis.`chemicals` c INNER JOIN frackanalysis.`cas_chemicals` cc ON c.`cas_number` = cc.`CAS_EDF_ID` WHERE c.`cas_type` = ‘valid’ AND cc.`Toxicity` = ‘cancer’ AND cc.`Category` = ‘recognized’ GROUP BY cc.`Chemical_Name`) subqry1 ORDER BY `Count` DESC;

Which gives one the resultset below:

NAPHTHALENE 6691

BENZYL CHLORIDE   1636

FORMALDEHYDE      1563

EPICHLOROHYDRIN   378

ATTAPULGITE 334

NITRILOTRIACETIC ACID, TRISODIUM SALT MONOHYDRATE     259

ACRYLAMIDE  233

BIS(2-CHLOROETHYL) ETHER      204

ETHYLBENZENE      134

THIOUREA    96

ACETALDEHYDE      29

1,4-DIOXANE 11

ETHYLENE OXIDE    11

ANTIMONY TRIOXIDE 5

NITRILOTRIACETIC ACID   1

propylene glycol butyl ether  1

Summary

The queries above are just a few of the many that could be made against the cited database and its tables.  A more thorough set of queries seem to be in order by professionals more versed in health studies.  Now that a database suitable for analysis is available, such may yet be possible.    As the records in the reports table contain latitude and longitude values, queries relevant to location seem to be in order as well.  A list of recommendations concerning further development of this database appears below.

Recommendations

  • A table of CAS chemicals along with their toxicity on a relative scale be found and imported into the database.  Perhaps information from the Material Safety Data Sheets (MSDS) can be used to obtain the relative toxicity as the MSDS ranks chemicals on a scale from 0 to 4.
  • A table of cities is obtained from a source like the National Atlas Data Download and the reported wells plotted in relation to said cities.
  • A table of wells drilled be imported into the database so that a query of wells drilled versus wells reported to the FracFocus.org site can be made to assess the level of reporting compliance in jurisdictions in which reporting is mandatory.
  • A table of schools is imported into the database so that proximity-oriented queries can be made against both wells and chemicals.
  • A table of earthquakes is imported into the database so that proximity-oriented and time-sequenced queries can be made against reported wells.
  • An open-source repository is created for storing and retrieval of both the data and analytical queries.
  • An open-source, high-speed search engine web site is created as an aid to the data analysis community as well as a demonstration site for presentations to policy makers as to what is possible with a relative minimum of effort.

Biography

David Darling is a U.S. Army veteran (tactical nuclear weapons) with a B.S. in Zoology and over thirty years of professional experience in Information Technology as a software developer, corporate-level database developer/administrator and search engine developer/administrator. He presently authors software in Java, Java Server Faces, PrimeFaces, coupled with the Apache open-source technologies of Velocity, Solr, and Hadoop.  Additional authorship in databases is done in SQL Server, MySQL, PostgreSQL/PostGIS, and Oracle. He especially enjoys any applications dealing with mapping, spatial data, and Geographical Information Services (GIS). He has extensive experience in petrochemical, law enforcement, local/county/regional governments, voter registration, taxation and property tax appraisal, telecommunications, real-time tracking (spatial/status), and health-care information.