Fracking, Mountaintop Mining, and More…My Summer at SkyTruth

 Hi, my name is Jerrilyn Goldberg.  Over the course of  two months last summer I worked as an intern at SkyTruth. In September I started my junior year at Carleton College in Northfield, Minnesota, majoring in environmental studies and physics. Over the course of my internship I contributed to SkyTruth’s Mountaintop Removal (MTR) research by creating a mask to block out rivers, roads, and urban areas that could be confused with mining activity by our analytical model. I also helped classify many of the ~1.1 million control points that allow us assess the accuracy of our MTR results.

To analyze the accuracy of the MTR results we obtained through our Earth Engine analysis, we dropped 5,000 randomly distributed points at each of 10 sample areas for each year between 1984 and 2016. These points were manually classified as being `mine` (if it overlapped a user IDed mine location) or `non-mine` (if it overlapped anything other than a mine). A subset of those manually classified points were then used to assess the accuracy of the output from our Earth Engine analysis

In addition to the MTR project, I created a story map illustrating the development of Marcellus Shale gas drilling and hydraulic fracturing (fracking) in Pennsylvania, and discussing the environmental and public health consequences fracking is having on some rural Pennsylvania communities. Check it out here. Through my research for the story map, I learned about the hydraulic fracturing process. I also learned about many of the political and social complexities surrounding the fracking industry in Pennsylvania, including conflicts between economic and community interests. Our goal with this story map is to present an accessible and accurate narrative about the fracking industry in Pennsylvania, which begins with understanding what’s actually going on now.

Click the image above to visit Jerrilyn’s interactive story map.

I started by learning about SkyTruth’s FrackFinder Pennsylvania data and methodology from the 2013 project. I read through our GitHub repository and figured out why the FrackFinder team chose their methodology and what the results represented. (While I was familiar with the general concept of the project, I did not know much about the specifics beforehand.) With this in mind, I set out to update the dataset with well pads built after 2013.


I quickly realized that this task presented many questions such as, which of the many state oil and gas datasets actually contained the information I sought. I selected the Spud Data, which contains all of the individual locations where operators have reported a drilling start-date for a permitted well. I filtered to include only unconventional horizontal wells drilling for natural gas and excluded those reported as ‘not drilled.’ To account for some missing drilling locations which I noticed while reviewing the latest Google base map imagery, I also download the Well Inventory Dataset which includes all permitted oil and gas wells along with their status. From here I filtered out all the spuds and wells not listed as drilled in 2014, 2015, or 2016 and joined the files. After joining the layers, I formed a well pad dataset by creating a 150 meter buffer around the wells, dissolving overlapping areas, then locating the centers of each buffer. This step effectively says ‘create a 150 m radius circle around each point, but when these overlap, clump them into one circle, then find the center of that new circle.’ Finally, I found all the buffers that overlapped with FrackFinder drilling locations from 2013 and earlier, and eliminated all of those centroids.

Hydraulic fracturing well locations in Pennsylvania by year through 2015.

A quick note about the imagery: USDA collects high resolution aerial imagery as part of the National Agriculture Imagery Program (NAIP), which at the time of my project was last collected for Pennsylvania in 2015. While I worked hard to eliminate inaccurate points, I was unable to verify all of these with the existing NAIP imagery. That said, I found that the other points accurately represented the general well pad locations and thus chose to include the points for the first half of 2016, even though I obviously couldn’t verify the existence of those recent drilling locations on the mid-summer 2015 NAIP imagery.


At the same time I found The Nature Conservancy’s (TNC’s) 2010 Energy Impact Analysis, which looked at the predicted development of wind, shale gas, and wood fuel usage in Pennsylvania. Part of TNC’s study identified three construction scenarios for how many wells and well pads could be built in Pennsylvania by 2030. With an assumption that 60,000 new wells would be drilled between 2010 and 2030, the study predicted between 6000 and 15000 new well pads would be built to host those wells. Each scenario featured a different distance between pads and a different number of wells per pad (because that number stays constant at 60,000 new wells). I found some data from TNC’s study hidden on an old SkyTruth backup with help from Christian and David. With the FrackFinder data, my update, and the ‘informed scenarios’ in hand, I started trying to figure out an appropriate way to synthesize the three datasets, to identify which TNC drilling scenario best fits what is actually happening..


One roadblock in conducting a thorough analysis and comparison was that TNC’s research makes a quantitative prediction about the possible volume of infrastructure development instead of a more tangible spatial prediction. The study distributes the predicted numbers of new well pads across the counties of Pennsylvania, which overlay the region of Marcellus Shale with ideal conditions for hydraulic fracturing for natural gas. All of the included counties now contain at least one well pad. I did notice that since 2010, about 1/3 of the well pads estimated by the low impact scenario (6000 well pads) have already been constructed. If the rate of development between 2010 and 2016 remains constant, Pennsylvania will surpass TNC’s low impact scenario.

An example of The Nature Conservancy’s “low” impact scenario for fracking well construction across a section of Pennsylvania.

The Nature Conservancy’s medium impact scenario for future fracking well construction across a section of Pennsylvania.

The Nature Conservancy’s high impact scenario for future fracking well construction over a section of Pennsylvania.


Fracking Pennsylvania” uses maps and other media to create a narrative of hydraulic fracturing and its consequences. While originally intended for the community members we work with in southern Pennsylvania, I hope this story map becomes a useful tool for many different communities grappling with fracking.


While I have my time in the Watchdog spotlight, I want to publicly thank everyone here for welcoming me into the awesome world of SkyTruth. I’m so grateful for the learning opportunities I had last summer and for all of the support I received. Special thanks to Christian for introducing me to SkyTruth and to John for helping me improve my Story Map even though he is definitely one of the busiest people in the office. I look forward to sharing my experience through the Carleton Internship Ambassador program this year.  

FrackFinding Success in Three States

Since the launch of FrackFinder, we’ve found great success in our efforts in Pennsylvania, Ohio, and West Virginia enlisting the public to help us analyze aerial imagery across the Marcellus and Utica shale gas-drilling regions. The results have been unique datasets that are being used, or can be used, by researchers to study the impact fracking has on public health and the environment. What we’ve learned is helping us refine our tools and methods for future rounds of FrackFinder. Here we’ll give a rundown of the results of our efforts and what we’ve done with them, as well as links to the data we’ve made available free for public use.

Pennsylvania Fracking Sites Map

Our motivation behind the FrackFinder project was to fill gaps in publicly available information related to where fracking operations in the Marcellus Shale were taking place. Seeing an opportunity to make this info available to the public, but lacking state data, we began mapping fracking sites ourselves. The locations of drilling sites, also known as “well pads,” were hard to come by, but state permits for drilling individual oil and gas wells were easily accessible. Unfortunately drilling permits aren’t very useful on their own. The permits are just approvals to drill: they don’t say if the site is active, when drilling and fracking began or ended, or if development of the drill site ever happened at all. Luckily, each permit provides the exact location where the operator is authorized to drill their well. By pairing the location information from the permits with available high-resolution aerial survey photography from multiple years, it is possible for us to learn where active well pads are and narrow down when they were built to within a span of a couple of years.

Of course, analyzing multiple years’ worth of imagery for thousands of permit locations is a monumental task.  To get the job done, we looked to crowdsourcing to speed up the process. Crowdsourcing also gives us the opportunity to reach the public, get people interested in citizen science, and provide them the opportunity to see the impact of fracking for themselves. It’s important for people to understand the large footprint fracking has compared to historical oil and gas drilling in the region, and seeing just how close many well pads are to farms and homes can change some people’s perspective on the issue.

Timelapse image showing how close drilling is to homes, and how big modern fracking operations are.

Our first phase of FrackFinder took place in Pennsylvania.  For this project we had 3,000 locations to examine on three different years of imagery, and we asked 10 volunteers to look at every site: a grand total of  90,000 image analysis tasks. Participants were presented with an image of a location corresponding to a drilling permit and were asked to determine if the site was active or inactive on the basis of visible infrastructure.  All the tasks were knocked out in three weeks, thanks in part  to a Washington Post article mentioning the project published around the time of our FrackFinder launch. In the quality assurance phase, we found that if seven of the ten participants for a given task agreed there was active drilling then our experienced in-house analysts agreed with the crowd, so we established 70% crowd consensus as an acceptable threshold to confirm if there was indeed drilling at a location.  This first project went so well that we quickly supplemented it with another year of imagery.  The final map we produced shows the location of active well pads in imagery from 2005, 2008, 2010, and 2013, and we intend to update it with 2015 imagery in the near future.

Marcellus Shale fracking sites in Pennsylvania in 2005, 2008, 2010, and 2013. Click on this image to link to the full interactive map.

Pennsylvania Impoundments Map

Not long after publishing the data on well pad locations from the first phase, we were approached by researchers from Johns Hopkins University who were interested in our data. They wanted to study the public health impacts of living near a modern fracking site, and the state couldn’t provide anything comparable to what we had at the time. They were specifically interested in how volatile chemicals coming off drilling-related fluid impoundments would affect people living nearby. While we had locations for the wells from our first FrackFinder project, we didn’t have information on the size, location and timing of the impoundments that may contain drilling and fracking fluids.

Hydraulic fracturing-related fluid impoundments in Pennsylvania. Click on the image to link to the full interactive map.

Using the same imagery we had prepared for the first round of FrackFinder, we launched another round of crowd-assisted image analysis using the same methods to determine the presence of impoundments. After the public identified water bodies that were likely related to drilling, our analysts verified that they were impoundments and delivered the data to the researchers. The Pennsylvania FrackFinder project was the first time we used crowdsourcing to create a high-quality data set for use in actual research.  And it has paid off in improving the public’s understanding of the health risks posed by living near modern drilling and fracking activity. The Johns Hopkins researchers have published the following peer-reviewed studies based in part on our work:

Ohio Well Pads Map

Ohio was the first state outside of Pennsylvania to have its own FrackFinder spinoff. Instead of launching a public crowdsourcing project we enlisted the help of students at Walsh University in Ohio who were interested in studying the impact of fracking on the environment and looking to get experience with GIS image analysis. We asked students to delineate all terrain that was modified to accommodate the drilling activity, including forest clearcutting around actual fracking infrastructure. This not only provided an educational opportunity for the students, but it allowed us to build and experiment with tools we plan on using in the future to let the public delineate fracking sites and create complex polygons, rather than simply confirming the presence or absence of a well pad at a specific point. This work hasn’t been used for research yet, but it still produced a high-quality data set that is available to anyone who would wish to use it in the future to quantify the ecological footprint of fracking-related land use, and explore the habitat and ecosystem impacts of modern drilling and fracking.

Utica Shale fracking well pads in Ohio. Click on the image to link to the full interactive map so you can zoom in and see the outlines of fracking sites delineated by students at Walsh University.

West Virginia Well Pad and Impoundment Map

Due to time constraints, we conducted the first round of West Virginia FrackFinder internally, and now have a multiyear map and dataset showing the locations of Marcellus and Utica Shale drilling sites statewide. We plan on launching a new public FrackFinder round this summer using the same area delineation technique that was demonstrated in Ohio. In West Virginia, we delineated the footprints of well pads and fluid impoundments, but not the broader area of clearcutting and landscape modification surrounding the drilling sites as was done in Ohio. When we launch our next public FrackFinder round we will ask the public to delineate this “impact halo” around well pads to help determine the ecological footprint of fracking in the state.

Marcellus and Utica Shale fracking sites in West Virginia in 2007, 2009, 2011, and 2014. Click on the image to link to the full interactive map.


Fracking-related fluid impoundments in West Virginia for the same years as the map above. Click to go to the full interactive map.

The data we produce for West Virginia is being used by researchers at UC Berkeley and at Downstream Strategies. They will perform a geospatial proximity analysis to see how fracking activity near sensitive populations in schools, hospitals, homes, and rehabilitation centers, paired with different chemicals used in fracking, affects public health. The results of their research will be detailed in a comprehensive white paper that will be published with policy makers in mind.

Confirmed: EPA Findings Edited to Downplay Fracking Impacts

Documents obtained by journalists at Marketplace and APM Reports revealed that federal officials made eleventh-hour edits to the Environmental Protection Agency’s (EPA) long-awaited Draft Assessment on the Potential Impacts to Drinking Water Resources from Hydraulic Fracturing Activities. The unsubstantiated edits downplayed the risks of hydraulic fracturing leading to a flurry of headlines claiming the EPA found little risk in fracking.

In fact, the more nuanced language of the report found evidence of contamination events and threats to groundwater, but ultimately the EPA lacked the data to conclude if fracking was having “widespread, systemic impact…” on drinking water. We wrote about these contradictions between the EPA press release and the actual report in June 2015 post entitled:

Word Games Continue: Just What Evidence Did EPA Not Find?

Earlier in 2016 the EPA Science Advisory Board also criticized the edited conclusions and called on the Agency to substantiate their claims or consider revising the report.

Words matter. Science matters. Don’t take headlines and executive summaries for granted, especially as we head into a political transition already swamped with climate deniers and a who’s who of the fossil fuel industry. Become as informed as you can from primary sources, and also support watchdogs and journalists who have proven effective at accurately reporting on what is happening in the world.

Read the full story from Marketplace and APM Reports:

EPA’s late changes to fracking study downplay risk of drinking water pollution

Leaving a MARC: Cutting a Swath though Pennsylvania

Construction work on the MARC 1 pipeline right-of-way in the Endless Mountains of Pennsylvania. Photo by J. Henry Fair, flight by LightHawk.
Fracking is not the only part of oil and gas drilling that has an impact on the landscape and the environment. Case in point: the newly-built MARC 1 pipeline runs for 39 miles through Bradford, Sullivan, and Lycoming counties in northeast Pennsylvania, carrying natural gas produced by horizontal drilling and hydraulic fracturing (fracking) in the Marcellus Shale. Along the route this pipeline crosses 71 roads, 19 named streams and rivers, many small unnamed creeks, and cuts through a densely forested swath of the beautiful Endless Mountains.
The MARC 1 pipeline right-of-way crossing a stream in northeast Pennsylvania. 
Photo by J. Henry Fair, flight by LightHawk.

Construction of the pipeline began in the fall of 2012, and we were interested in illustrating construction-related impacts. Finding info on pipeline routes, however, is no simple task. The first map which turned up was a scanned pdf created by Central New York Oil and Gas Company (CNYOG); a deeper dig for a more accurate map turned up the Department of Transportation’s National Pipeline Mapping System (NPMS), but unfortunately NPMS data is not available for download. So we decided to create our own map of the pipeline – informed by the CNYOG map, and validated against the NPMS data:


The pipeline map was created by tracing the route on aerial and satellite imagery available in Google Earth. Imagery was collected during the pipeline’s construction which helped us do a pre- and post-construction comparison. Road and stream data from the US Census Bureau’s Tiger/Line was used to calculate the number of roads and streams which were intersected by the pipeline. Here is a side-by-side look at a selected site along the pipeline route before and during construction:

MARC 1 pipeline crossing field and forest near Sugar Run, PA. Compare imagery from 2011 and 2012.
Directions to this location.

MARC 1 also traverses Pennsylvania State Game Land for 1.5 miles, with the right-of-way occupying 21 acres of this prime habitat and hunting / recreation area:


We used the USGS’s 2011 National Land Cover Dataset to assess the area and types of land use impacted by the construction of the pipeline. Overall, construction of the Marc-1 pipeline right-of-way impacted over 400 acres of land, 318 of which were forested (see the exact breakdown of land cover types at right and raw data here).

Now the MARC 1 pipeline is a done deal and some of the impacts will eventually fade into the background, but the corridor through forest and woody wetlands will remain. From air emissions and habitat fragmentation to property rights issues, we need to be careful not to overlook the environmental impact of pipeline building, especially as developers focus their efforts on expanding pipeline capacity to keep up with oversupply of natural gas.

If you want proof of that, look no further than the MARC 2 pipeline. Yes, developers were already proposing a 30-mile MARC 2 pipeline less than two years after the MARC 1 pipeline was completed. Stay alert…


MARC 1 crossing field and forest near the Susquehanna River close to Sugar Run, PA. 
Compare imagery from 2011 and 2012. Directions to this location.


MARC 1 traversing rural Bradford County, PA near Foster Branch, a tributary of the Susquehanna River. 
Compare imagery from 2011 and 2012. Directions to this location.

Freaky Fracking – Mapping How Wellpads are Carving Up Ohio

We don’t usually do seasonally-themed maps, but this map tracing the footprint of wellpads in Ohio’s Marcellus and Utica Shales just happens to work best with the colors associated with All Hallows Eve. Over the past year our FrackFinders and partners at Walsh University have helped us map shale drilling in eastern Ohio. Below you can see the total area of area of wellpads displayed using scaled “bubbles” which show the size of the wellpads relative to each other. This top-level view helps you see where drilling is the densest, and locate the largest and smallest wellpads. 

If you zoom in closer, you will see the actual outline of individual wellpads to scale. In this map there are 320 sites, all traced out by students at Walsh University participating in our FrackFinder collaborative image analysis projects. The median area of these wellpads is 13,787 square meters, or 3.4 acres. If you recall, we recently used this number to help visualize similar drilling in western Pennsylvania

The largest pad was 17 acres and the smallest pad was 0.6, and all told we found 1,100 acres of Ohio fields and forest converted to gravel wellpads. The total impact of drilling extends beyond just the wellpads we mapped in this phase of the project, so in future we will be working to repeat this approach in other states and looking at total landscape impacts. This tutorial video we created for the project will show you exactly what we’ve mapped here. 
Understanding the public and environmental impacts of drilling is complicated, especially since these industrial operations are scattered all across the landscape; some sites are remote while others are right next to homes and farms. But the data you help create in these projects enable SkyTruth and our partners to correlate this data about when and where drilling occurred with public health and environmental data. This research is starting to bear fruit as our partners at Johns Hopkins recently released a study showing that living in the most active quarter of Pennsylvania’s Marcellus gasfield was associated with a 40% increase in the likelihood of pregnant mothers giving birth prematurely. Scary indeed.

Inside a Hotspot: A Timelapse of Shale Drilling in Pennsylvania

Earlier this month we published a map of active Marcellus shale wellpads in Pennsylvania as observed on aerial survey imagery from 2005, 2008, 2010, and 2013 by our FrackFinder citizen scientists. Now we thought we’d take a closer look at one of those hotspots of drilling activity, specifically an area in Washington County, PA near Cross Creek County Park and the town of Hickory. 

For this visualization, we created a 3.4 acre buffer around each active wellpad, a number we derived from our related work mapping the footprint of wellpads in Eastern Ohio. We have not yet measured the cumulative footprint of drilling activity in Pennsylvania, so we used the median area for wellpads in Ohio’s Marcellus and Utica shale play. 

However, the impact of drilling is not just restricted to the gravel parking lot around a wellhead, it extends to service roads, pipelines, waste impoundments, gas separators, compressor stations, etc. So to visualize that impact, we have also included a snapshot of the aerial survey imagery for the same area from each of the respective years. 

These visuals are cumulative, meaning that not every wellpad was visibly active at the time of the aerial survey. However, given the predicted lifespan of shale wells we can expect that almost all of these sites could be expanded and re-fracked several times over the coming decades. If you want to take a closer look you can download high-resolution stills from our album over at Flickr or explore the interactive map of all observed, active wellpads in Pennsylvania. 

This kind of dense drilling activity in close proximity to homes and towns is cause for serious concern with recent findings by our partners at Johns Hopkins who found that “expectant mothers living in the most active area of fracking drilling and production activity were 40 percent more likely to give birth prematurely (before 37 weeks of gestation).” Our goal with these maps and mapping projects is that the resulting data will be used to better understand the public health and environmental impacts of resource extraction activities like fracking.

Word Games Continue: Just What Evidence Did EPA Not Find?

UPDATED – Dec. 2, 2016: Documents obtained by journalists at Marketplace and APM Reports found that officials made eleventh-hour edits to downplay the risks of fracking. Earlier this year the EPA Science Advisory Board also criticized these top-level conclusions and called on the EPA substantiate their claims or consider revising the report. Also, replaced broken link to EPA press release. 

Yesterday, the U.S. Environmental Protection Agency (EPA) published a series of draft reports on their findings from five years of research and literature review on the question of whether or not fracking contaminates groundwater. But if you just read the headlines you might have been confused about
what the EPA had actually concluded. As Forbes pointed out, the headlines were a bit contradictory. 


But the bigger news is that even EPA was inconsistent about the findings of their own report. The press release from EPA states that their assessment (emphasis added): 

“…shows hydraulic fracturing activities have not led to widespread, systemic impacts to drinking water resources and identifies important vulnerabilities to drinking water resources.”

However, the Executive Summary of the report puts things differently (emphasis added):
We did not find evidence that these mechanisms have led to widespread, systemic impacts on drinking water resources in the United States. 

These two statements may look similar, but there is a big difference between saying that you did not find find any evidence of a crime and definitely claiming that you have proven the suspect’s innocence. But try telling the House Natural Resources Committee that fracking has never been proven NOT to cause contamination, and members of Congress will laugh aloud and joke about pigs not flying to Mars. Seriously (check out 1:12:10).

But buried on page 22 of the 28-page executive summary, the EPA goes on to say (again, emphasis added):

This assessment used available data and literature to examine the potential impacts of hydraulic fracturing from oil and gas on drinking water resources nationally. As part of this effort, we identified data limitations and uncertainties associated with current information on hydraulic fracturing and its potential to affect drinking water resources. In particular, data limitations preclude a determination of the frequency of impacts with any certainty.

So in short, the EPA didn’t find proof of wide-spread contamination from fracking, but they lack the data to say with any certainly whether that means anything at all. At least they acknowledged that they found “specific instances where one or more mechanisms led to impacts on drinking water resources, including contamination of drinking water wells.” Which is actually big news coming from an agency which had previously stopped short of such a conclusion. 

Unfortunately, this contradiction between headlines from the EPA PR office and the finely-nuanced findings of the EPA scientists just underscores a point we made by in 2013. Word games are still misleading the American public about frackingand “…[w]hile cases of contamination caused by fracking remain obscured by lack of information and tricky linguistics, we know that a growing number of citizens are reporting harm and environmental contamination in unconventional oil and gas fields, and especially from wells that have been fracked.”