Sunday, May 15, 2011

Does Fracking Contaminate The Water Table? : Deconstructing Penn State Study

The Wild Eyed Liberals at the ArkTimes took a break from bloviating outrage at Secretary of State Mark Martin for things like failing to keep records on a state vehicle that he is not required to keep records on (seriously, they griped about that). They took time out to bloviate outrage at the private businesses that provide them with the electricity they voluntarily use to warm their homes and publish their libels. In this case I am talking about natural gas drilling, specifically with the relatively new technology of horizontal fracturing.

They normally don't need any rational facts to take ahold of, for example their various attacks on Martin, but in this case they had a study from Penn State University to back up their contention that the process is responsible for methane contamination of the water table. Now the Arktimes folks can be discounted a raving lunatics, but Penn State deserves a fair hearing.

I am here to report on the study. The bottom line is that it is fatally flawed because they used sampling methods that were so contradictory to good study design that I am confident many of the 8th graders in my old Earth Science classes could have spotted the flaws. More about that in a bit.

The study found not a trace of fracking chemicals in the water, but it did find natural gas. They went to pains to show that much of it was not surface methane, like swamp gas, but rather came from underground sources. One problem with that is that there is underground gas that is in shallow formations, then there is the deep-formation gas that the drillers in the study area are actually fracking in. That gas is thousands of feet, maybe even 10,000 feet, below the water table, and the wells use triple pipe incasement far below the bottom of the water table.

There are some places in the earth where underground but still shallow deposit natural gas seeps out all by itself into the water table. I believe some areas around Hutting for example, had tapwater you could set on fire even back in the 50s, long before horizontal drilling and fracking were available. The study went to a great deal of trouble to show that the gas found in local water was mostly non-biological (not swamp gas) but did an awful job of discerning whether or not areas that had a lot of such gas in the water table did so because of drilling or simply because they were one of the many areas on earth where methane naturally seeps into the groundwater regardless of what is going on 8,000 feet of rock below that.

(CONTINUED ON THE FLY)

6 Comments:

Blogger Mark Moore (Moderator) said...

Here is how they said they TRIED to discern the difference:

They took water sample in three main areas. Sometimes they took water samples close to an active well, within 1KM, and other times they took water samples in the same region that were more than 1KM from any active well.

The result? There was ON AVERAGE 17 times more methane from samples taken close to the wells as there was in samples taken farther from 1KM from any active well. Case closed? Not so fast! Averages are only as good as the data which make them up. Let's see how they sampled...

One of the three regions was the Catskills. They took 5 non-active samples in this area and got an average of 1.9 milligrams of methane in the sample with a standard deviation of 6.3! That is far too few samples with a deviation that large.

You could get that result with four readings near zero and one that is above the average methane contamination for an active site. In other words, a 1 in 4 chance that an inactive site will have just as much methane as an active site. But even that stat overstates the methane around active sites as you will see.....

8:08 PM, May 15, 2011  
Blogger Mark Moore (Moderator) said...

They sampled 13 active sites in the Catskills, with an average of 26.8 mg of methane and a standard deviation of over 30!

Again their standard deviation is larger than the average, which indicates to me they found an outlier such as a well or a gas line spewing a leak and therefore not indicative of normal operations. Leaks will eventually be noticed and repaired. When the SD is that large compared to the average, it signals one very hot spot. Most of the other sites could have been close to the levels found at the non-active sites.

8:15 PM, May 15, 2011  
Blogger Mark Moore (Moderator) said...

This is further supported by their sampling at the next site, the Genesee. Here is a place where very little methane naturally flows out of the earth. They took 6 readings at non-active sites and found an average of only 1.5 mg of methane (with a standard deviation of 3.0, still high compared to the average, but closer than the proportion at the Catskills.)

Now you might think that including so many non-active site measures for this region, that they would want to take some non-active site measures too. That is, they would if the goal was a good study, rather than a study that showed what they wanted in advance for the study to show. They only took ONE (1) active site sample in this region. It was lower than the average from the non-active sites in the region at 0.3 mgs!

If they had taken seven more readings like that, it would have brought the averages back into line. That is, the active sites and the non-active sites would have had much closer to the same readings. But they did not do that. They went to a site where the gas was naturally low and took a bunch of non-active site readings which had the effect of driving the non-active site methane average down, but took only one reading from an active site so that the active site average remained high.

But it gets worse...

8:22 PM, May 15, 2011  
Blogger Mark Moore (Moderator) said...

The table I am drawing my data from is on page two of the study btw, at the link I provided on the front page.

The last site was the Lockhaven site. Here we have a place where the methane readings were getting high enough to be a problem. The took seven samples from active sites which had a whopping average of 50.4 Mg of gas in them, though the standard deviation was 36 mg.

The seven readings from these sides drove up the "active site" numbers. So how many readings from inactive sites did they take at this location? NONE. Inexcusable.

The researchers give every appearance of picking their sample sites to skew the averages. If a site has high methane, they take lots of active site samples but much fewer or even no inactive site samples. If a site has low methane then they take many inactive site samples and one or few active site samples.

To me that brings into question their whole site selection process. Perhaps they heard anecdotal evidence of a gas hot spot and selected that as the place to do the most active site tests. That says nothing about the water quality associated with natural gas drilling as a whole. It's like judging a factory that turns out a million candy bars by hearing about one lot of 100 that has been recalled and using that lot for testing to determine the safety of the bars as a whole.

8:32 PM, May 15, 2011  
Blogger Mark Moore (Moderator) said...

I am amazed that such sloppy work passed peer review. My opinion is that this study reflects poorly on both the Penn State researchers and the reviewers who failed to call them on these middle-school science fair level mistakes. That's the kindest conclusion I can come to. The other possibility is that they were deliberately cherry-picking their active vs. inactive site selection with the purpose of skewing the results. That's not a competence issue, it's an integrity issue. Since I can't know their hearts, I will simply evaluate their processes: FAIL.

8:36 PM, May 15, 2011  
Anonymous Cal Beisner said...

Mark, you wrote in one of your comments: "Sometimes they took water samples close to an active well, within 1KM, and other times they took water samples in the same region that were more than 1KM from any active well." Of course, that skews the results, too, if the drilling companies are more likely to drill close to natural sources than far from them. I.e., the samples weren't random.

10:39 AM, July 11, 2011  

Post a Comment

Links to this post:

Create a Link

<< Home