Hydraulic Fracturing and Earthquakes

by Robert A 

Less than a decade after the record high oil prices of 2004-2008 a new method of producing oil and natural gas known as hydraulic fracturing has radically changed the American and global energy landscape almost doubling American oil and gas production, contributing to the economic recovery as well as to a 50 percent drop in global oil prices.

Hydraulic fracturing, commonly known as fracking, uses a combination of horizontal drilling, explosives and a mixture of chemicals, sand and water to create fissures in underground geological formations thus releasing trapped oil and gas. While its economic impact has been widespread, there have been concerns that hydraulic fracturing can cause earthquakes in areas that have had little history of seismic activity and increase the frequency of earthquakes in areas where minor activity is considered normal.

Hydraulic fracturing for oil and natural gas extraction continues to expand in the U.S. (see figure 1) and is predicted to grow in countries like China, Argentina and Europe. It is therefore essential to fully understand the effects of fracking on seismicity in order to avoid negative human, infrastructure, regulatory and financial repercussions.

Figure 1. Hydraulic Fracturing Across the United States

Source: Post Carbon Institute

Research, Goals and Methods

A series of studies on the subject, including one from the U.S. Geological Survey, concluded that while hydraulic fracturing can cause micro tremors the actual cause of the earthquakes appeared to be the injection of fracking waste water into underground wells. Fracking produces large quantities of toxic waste water some of which is injected into deep, underground disposal wells thus putting pressure on geological formations and potentially causing earthquakes.

This study aims to complement previous research on the potential seismic effects of fracking by analyzing detailed public data from 2005 to 2011 of 67 counties in Pennsylvania with extensive fracking but with only minor presence of injection wells. The limited number of injection wells help to isolate the real effect of fracking and avoid colinearity issues and interaction effects when both fracking and disposal wells are present in great numbers.

The data is analyzed using OLS regression models that control for the effects of waste water injection wells as well as historical earthquake data to avoid attributing effects to fracking that might in fact be due to historical earthquake trends. The results  are then compared to data from Ohio which, has similar geology but features numerous fracking and injection wells.

Analysis

As observed in the Table 1 the summary statistics show a marked increase in the average number of earthquakes for the 67 Pennsylvania counties between 2005 and 2011 when hydraulic fracturing was introduced as a method of extracting oil and natural gas. The incidences of earthquakes between 2005 and 2011 were double the average for a similar 7 year period spanning from 1991-1997 and over 7 times higher than in the 1998-2004 and 1984-1990 periods.

Table 1. Summary Statistics of Earthquakes Across PA Counties

Variables	Mean	Std. Dev.	Min.	Max.
Drilled wells between 2005-2011	72.66	176.61	0.0	949.0
Active Disposal Wells	0.13	0.34	0.0	1.0
# of Earthquakes between 2005-2011	0.45	1.75	0.0	13.0
# of Earthquakes between 1998-2004	0.06	0.24	0.0	1.0
# of Earthquakes between 1991-1997	0.22	1.07	0.0	7.0
# of Earthquakes between 1984-1990	0.07	0.40	0.0	3.0
Observations	67

A similar earthquake pattern could be observed across counties in Ohio (table 2) which also saw a rapid increase in fracking wells in the 2005 to 2011 time period.

Table 2. Summary Statistics of Earthquakes Across OH Counties

Time Period	Number of Earthquakes
2005-2011	34
1998-2004	12
1991-1997	6
1984-1990	12

Source: USGS

Table 3. Earthquake Occurrence in Pennsylvania Counties by Fracking

Wells between 2005-2011

	(Logged Variables)	(Categorical Variables)
log(Total Drilled Wells)	-0.038
	(-1.750)
Active Disposal Well	0.401**	0.383***
	(2.860)	(5.451)
# of Earthquakes 1998-2004	0.545*
	(2.007)
# of Earthquakes 1991-1997	0.173**
	(2.764)
# of Earthquakes 1984-1990	-0.229
	(-1.676)
Presence of Wells 2005-2011		-0.101*
		(-2.007)
Earthquake occurrence 1998-2004		0.480***
		(4.195)
Earthquake occurrence 1997-1997		0.672***
		(5.853)
Earthquake occurrence 1984-1990		0.089
		(0.744)
Constant	0.133	0.066
	(1.940)	(1.674)
R-Squared	0.37	0.70
Observations	67	67

Note: OLS estimates with t statistics in parentheses. * p < 0.05, ** p < 0.01, *** p < 0.001

Source: USGS and Stateimpact.org

Table 3 shows the results of two regression models of earthquakes by fracking wells.

Model 1

Due to a skewed distribution of the data (see Graph 1 in appendix), the dependent and effect variable values in Model 1 were transformed in order to more accurately estimate the causal relationship between earthquakes and fracking. The relationship between the total number of earthquakes and fracking wells per county was tested against the potential confounding effects of injection wells and earthquake data from prior periods. Model 1 seemed to suggest a negative effect resulting in a 0.038% average decrease in the likelihood of earthquakes for every additional 1% rise in the number of fracking wells. However statistical analysis indicated that this effect could likely be due to random correlations and is not statistically significant. On the other hand, while limited in number, the injection wells present in the PA counties appeared to have a significant effect on the increase in earthquake incidence, resulting on average in a 0.401 point increase per injection well. Earthquakes from the 1991-1997 period also appeared to have a positive effect on earthquake incidences in the 2005-2011 period.

Model 2

This model measures the probability earthquake occurrence given the presence or absence of fracking wells, injection wells and prior earthquakes in the 7 year periods. Again counties that have or are nearby injection wells appear to have a 0.383 higher probability of experiencing earthquakes. Fracking wells also appear to have a significant effect on earthquake occurrence however, contrary to what might be believed, their presence supposedly lowers the probability of earthquakes by 0.101.

Earthquake incidences in prior periods also suggest a higher probability of future earthquakes. These results are not surprising as geological formations are in slow but constant movement and tend to naturally generate more or less regular seismic activity. Thus, areas tend to experience more or less consistent earthquakes with similar magnitudes over decades and centuries.  

It might very well be that fracking activity in Pennsylvania simply coincided with a natural cycle of increased seismic activity. This conclusion however, does not explain the considerable effect of injection well presence.

It should be noted that while the regression results, especially in model 2, do show on average statistically significant effects of injection wells and prior earthquakes on the incidence of future earthquakes, the results cannot be taken at face value due to the relatively small number of observations.

Conclusions

The above presented regression results do not seem to support the notion that fracking wells increase the chances of earthquakes. They do however, suggest that prior earthquakes and injection wells have an effect on increased earthquake incidences.

The limited number of injection wells in Pennsylvania prevents us from making major conclusions on the effect of injection wells. Next door Ohio however, had 177 Class II disposal wells in 2011, with a large number of them used for fracking waste water disposal. The overlay of injection wells and earthquake incidences in Ohio over the 2005-2011 period does show a certain level of correlation between the two (figure 4).    

Figure 4. Overlay of Injection Wells and Earthquake Incidences between 2005 and 2011

The results of this study appear to be in line with those conducted by the USGS and independent researchers. While limited data prevents us from making any major conclusions, the results do not show any support for concerns that fracking wells cause earthquakes or increase the incidence of earthquakes.

Appendix

As shown in graph 1, the data suffers from a potential 0 problem with a considerable number of PA counties not having fracking wells or earthquakes. Moreover the skewed data distribution requires the logging of both the independent and dependent variables in order to properly conduct a linear regressions.