Project Description
Below is a write-up for an analysis performed on the relationship between the volume of injected fluids and occurrence of
earthquakes in Oklahoma. Assocated documentation and code can be found on the GitHub repository:
Oklahoma Earthquakes GitHib repository.
The project was a personal side project not associated with the class. The goal was to intersect my geologic background and
recently learned analyical methods learned. Below the formal write-up is an informal discussion where I discuss my thoughts
on the project and lessons learned.
Earthquakes in Oklahoma: Exploring the Relationship Between Waste Water Injection and Seismic Events
Introduction
Oklahoma has experienced a marked increase in seismicity since 2009 which has been strongly linked to the increase in saltwater disposal (SWD) volumes injected into the Arbuckle Group [1, 2, 3]. This SWD is a direct result of hydralic fracturing (informally known as fracking); the hydraulic fracturing requires an enormous amount of water that must be disposed of after the hydrocarbon recovery process - in 2015, the year with the highest injection volume, nearly 330,000,000 m³ of wastewater were injected into the Arbuckle Group in Oklahoma, with more volume injected into various other formations. This project attempts to model the realationship between the increase of injection volumes and the increase in seismic events within Oklahoma.
Background
Geologic Setting
The Arbuckle Group is a relatively poorly understood geological formation of approximately Late Cambrian - Ordovician age (approximately 480-490 million years old). The Arbuckle Group is primarily composed of nearly pure carbonates, which, through karst dissolution processes, have produced significant amounts of porosity making it a suitable candidate for wastewater storage [4]. However, the Arbuckle Group has hydraulic communication with the crystalline basement below containing pre-existing faults under stress. It is hypothesized that injection of wastewater into the Arbuckle Group causes pressure on the pre-existing faults, thus reducing the frictional resistance, inducing seismicity [1].
Wastewater Injection Background
Waste water injection wells are used to inject fluid into porous geologic formations. The waste water is typically a brine mixed with proprietary chemicals used in the hydraulic fracturing process. The wells included in this study are Class II wells, used for fluids associated with oil and natural gas production.
Data Acquisition
Acquisition of data for this project was relatively stright forward: injection data was downloaded as XLSX files from the Oklahoam Coproration Commission's (OCC) website [5], and the earthquake data was retreived via API from the United States Geological Survey (USGS) [6]. Additionally Oklahoma Geological Survey (OGS) has data on seismic events; however, their download interface is less suited for large scale downloads and do not include earthquake data post 2018.
Methods
This section details pre-processing of earthquake and injection data prior to linear regression modeling,
along with further processing descisions resulting from exploratory data analysis.
Wastewater Injection Data
Wastewater injection data from the the OCC website were provided in three major categories:
1) Daily injection volumes from SWD wells located in the Area of Interest (AOI) completed in the Arbuckle
Group. Each years' data is contained in one XLSX file, and the years 2012-2021 are represented,
and updated daily. However, daily injection rates for these wells were not required until 27 March
2016, thus, data prior to that date are considered incomplete. As shown in Figures 1 and 2, the
dropoff in injection volumes during 2020 are possibly die to incomplete data.
2) Monthly injection volumes for all SWD wells located in Oklahoma. Each years' data is contained
in one XLSX file; there is monthly data for 2011-2020. It's noted, though, that data post 2019 is
considered incomplete and is subject to being updated.
3) Yearly injection volumes for all SWD Wells located in Oklahoma from 2006-2010.
During intital viewing of the data, it was noticed that there was a discrepency between volumes
from the daily and monthly injection data files: after approximately 2015, the daily rates summed to
a monthly volume were high as 3,179,746 m³ (~840,000,000 gallons; 20,000,000 barrels). After analysis
of the data, it was determined that there were different reported volumes for the same wells in
both data sets. Ostensibly, the monthly data should have the complete data, as, per the OCC website,
the monthly injection volumes sorted by year "...contains annual reporting of volumes for all injection
wells submitted to the Underground Injection Control (UIC) Department..." [5], while the daily data should only
be a subset of the total wells located in the AOI. Figure 1 (below) displays the discrepency.
Figure 1: Comparison of Wastewater Injection Volume Sources
Figure 1 displays the differences in volumes between the monthly, daily, and combined data sources. As stated above, daily
volumes (red; note that daily volumes were grouped and summed by month) were not required to be submitted
to the OCC until 27 March 2016; however, the large increase in volume corresponds
to March 2015 - it is possible that there is a typo on the OCC website. Due to the lack of data prior to
March 2015, the daily data could not be used on its own for the analysis. Monthly data is available as far
back as 2011, and makes up the bulk of the injection numbers used for analysis. Monthly data is available for
every well in Oklahoma - as such, the monthly data were filted by well identification number to match wells
included in the daily data, as those wells are located in the AOI. However, after filtering the monthly data
and grouping the daily data by month, the monthly injection volumes did not match. The blue line in Figure 1
displays the filtered monthly injection volumes.
The daily and monthly data were then combined, with duplicate well identification numbers with lower total monthly
volumes dropped; the green line in Figure 1 displays the combined data. After combination, it was decided that
the monthly data would be used for analysis without combining with the daily data since it is impossible to know whether the daily or
monthly datasets contain the values more closely aligned with true injection volumes, thus, there was no adequate
basis for determing which of the values to drop. However, the volumes and trends identified here within are
analagous to other independent and peer reviewed studies [1, 7].
It should be noted that, due to the limitations and goals of this study, the well injection volume data
here within should not be considered completely representative of the actual well injection volumes: a more complete
study should involve communication with the OCC to resolve possible discrepencies in their data.
Seismic Event Data
Data were downloaded from the USGS website via an API call; data cleaning involved removing earthquake entries with negative magnitude and those with missing data (e.g., no magnitude, geographic coordainates). The earthquake data were then filtered for only events which occur wthin 20 km of an injection well. The spatial relationship between the seismic event location and injection locations is noted to be an important attribute: 20 km was chosen to be a reasonable distance after comparison with other studies. Lastly, only earthquakes of magnitude 3 or higher were included in the analysis, per convention with literature [1, 3, 7, 8]. The interactive map below displays the earthquakes per year.