/* Added by TWP, 10/12/2012 */ /* End of addition */

One of the live oaks that bless my home

Friday, March 11, 2016

Is U.S. Shale Oil & Gas Production Peaking? Part I: Gas Production

Part I of this post shows my calculations of ultimate gas recovery from the Barnett, Fayetteville, Haynesville and Marcellus shales.  They might deliver 6-7 years of natural gas consumption in the U.S. in 2015, or might deliver only 3 years worth of U.S. gas consumption.  In Part II, I will show my calculations of ultimate recovery of oil and gas in the Eagle Ford and Bakken shales that ultimately might deliver 6-12 months of additional gas consumption.  I will also discuss the physical reasons for the negative impact oil production from these two shales has had on global oil prices.

As Asjylyn Loder and others at Bloomberg have noted, another 19 billion dollars of debt of shale oil and gas producers is going into default as of the second week of March, 2016:
Since the start of 2015, 48 oil and gas producers have gone bankrupt owing more than $17 billion, according to law firm Haynes and Boone. Fitch Ratings Ltd. predicts $70 billion of energy, metal and mining defaults this year, and notes that $77 billion of energy bonds are bid below 50 cents, according to a note Thursday (3/09/2016)
Asjylyn Loder, just like Mason Inman, is one of the few thoughtful reporters informed on the subjects of shale oil and gas production, and their economics.

In December 2014, Mason wrote a controversial article published by Nature, one of the most influential scientific journals:
"Natural gas: The fracking fallacy: The United States is banking on decades of abundant natural gas to power its economic resurgence. That may be wishful thinking."
In his Nature article, Mason included this quote from me:
The results are “bad news”, says Tad Patzek, head of the University of Texas at Austin's department of petroleum and geosystems engineering, and a member of the team that is conducting the in-depth analyses. With companies trying to extract shale gas as fast as possible and export significant quantities, he argues, “we're setting ourselves up for a major fiasco”.
This particular quote reverberated in the media around the world and caused a storm of ad hominem attacks on me.  Here is the letter signed by Mr. Howard Gruenspecht, Deputy Administrator of the U.S. Energy Information Administration.  In that letter I was called a relatively minor player in the definitive UT BEG Shale Study and, much worse, President of ASPO, the Association for the Study of Peak Oil.

Ouch, that hurts! Of course there were other letters and more howling.

Never mind that the BEG study used this model of shale gas production in all of their calculations.  Our model was published in the Proceedings of the U.S. National Academy of Sciences (NAS) and was awarded by NAS the 2013 Cozzarelli Prize for best paper in engineering.  And never mind that in early 2011, I made an accurate prediction of gas production from the Barnett shale, based on the model originally proposed by Dr. M. King Hubbert, who was the original pre-ASPO researcher.  Of course, early on, in 1956, Hubbert was called a lunatic and idiot. But I wrote on this subject many times, see here and here and several other posts, and do not want to repeat myself.

Since in this post I am talking about oil and gas, I will express their production in the units of energy, instead of barrels of stock tank oil and standard cubic feet of gas.  To obtain numerical answers between 0 and 30, I will use exajoules (EJ). 1 EJ = ten to the power eighteen of joules is an astronomical amount of energy.  1EJ when digested as food, feeds 320 million Americans for one year.  Americans use about 100 EJ per year of primary energy, mostly as fossil fuels, most of them oil and gas. Also, 1EJ equals the high heating value of 1 (0.8 to be exact) trillion standard cubic feet of natural gas (1 Tcf).  Currently, U.S. consumes about 28 Tcf (35 EJ) of natural gas per year.

The plots below have consistent x- and y-scales for all mudrock (shale) plays I consider: the Barnett, Fayetteville, Haynesville, Eagle Ford and Bakken.  I do not have much faith in the Marcellus shale calculation; it is a relatively young development and production data reporting in Pennsylvania is incomplete and awful.  I am very grateful to my friend, Dr. Art Berman, for providing me with the spreadsheets that contain data dumps from the Drillinginfo well production database.

Disclaimer:  The plots below are only for the active producing wells up to date.  To the extent that robust drilling will continue unabated in each play, the three shale gas plays described below might produce another 50 percent more energy.  If the rate of drilling rapidly diminishes, as is the case today, the ultimate production from these three shales may be only 10-20% above the projections in this blog.

Let's do a quick summation: According to my calculations, the Barnett shale will ultimately deliver 9 months of U.S. gas consumption, Fayetteville 4 months, Haynesville 6 months, and Marcellus 14 months.  Suppose now that the Marcellus shale will produce as much as these three plays together, or 19 months of U.S. gas consumption (40% more than my calculation).  Also suppose that the ultimate recovery from all four shales will exceed my current predictions by 50%.  This brings us to (19+19)*1.5 = 5 years of U.S gas consumption (I always round my recovery estimates).  I will add the estimated ultimate gas production from the Eagle Ford (5 months of U.S. consumption) and Bakken (2 months of U.S. consumption) in Part II of this blog, and they will add another 1.5 years of U.S. gas consumption.  So much for the 100 or 200 years of gas supply from shales "predicted" by the robust technologists.

Warning:  When I said in December 2014, that my country is setting herself up for a major fiasco, I was not idly joking nor was I seeking cheap publicity. I merely tried to encourage those who care to listen to refocus their thinking.  Is anyone listening a year and a half later?
Click on the image to see it in full resolution. Production rate of natural gas from the Barnett shale in EJ/year.  1EJ = 0.8 Tscf.  Data source: Drillinginfo and Texas Railroad Commission, accessed January 2016.
Click on the image to see it in full resolution. Cumulative production of natural gas from the Barnett shale in EJ.  1EJ = 0.8 Tscf.  The estimated total gas production from the Barnett is equal to about 9 months of current gas consumption in the U.S.  Data source: Drilling Info and Texas Railroad Commission, accessed January 2016.
Production rate of natural gas from the Fayetteville shale in EJ/year.  1EJ = 0.8 Tscf.  Note that  because of infill drilling and pad drilling two Hubbert curves match the reported production. Data source: Drillinginfo and Texas Railroad Commission, accessed January 2016.
Cumulative production of natural gas from the Fayetteville shale in EJ.  1EJ = 0.8 Tscf.  The estimated total gas production from the Barnett is equal to about 4 months of current gas consumption in the U.S.  Data source: Drillinginfo and Texas Railroad Commission, accessed January 2016.
Production rate of natural gas from the Haynesville shale in EJ/year.  1EJ = 0.8 Tscf.  Likely, because of depletion of the very high initial pressure and new wells that got better with time, two Hubbert curves match the reported production. Data source: Drillinginfo and Texas Railroad Commission, accessed January 2016.
Cumulative production of natural gas from the Haynesville shale in EJ.  1EJ = 0.8 Tscf.  The estimated total gas production from the Barnett is equal to about 6 months of current gas consumption in the U.S.  Data source: Drillinginfo and Texas Railroad Commission, accessed January 2016
Production rate of natural gas from the Marcellus shale in EJ/year.  1EJ = 0.8 Tscf.   The production data from Pennsylvania are listed every 6 months and are of poor quality. Data source: Drillinginfo and other sources, accessed January 2016.
Cumulative production of natural gas from the Marcellus shale in EJ/year.  1EJ = 0.8 Tscf.    The estimated total production from  the Marcellus is equal to about 14 months of U.S. gas consumption. The production data from Pennsylvania are listed every 6 months and are of poor quality. Data source: Drillinginfo and other sources, accessed January 2016.

19 comments:

  1. Such a great information on Save on electricity. Also helpful to search and full details on Natural Gas quote. Thanks for it...

    ReplyDelete
  2. Mr. Patzek

    It is difficult for me to believe that this is a serious presentation.
    I have heard over the years numerous references to you and your work as extremely biased, but that in itself never dissuades me from observing and evaluating presented data.

    This piece above leaves me aghast.

    You don't trust the production figures from the Marcellus?
    Sir, I can assure you that the state tax people, royalty owners, industry participants who share in the proceeds as non-operators ALL watch production figures quite keenly.

    In case you are unaware, the Pennsylvania production figures are compiled and published monthly starting January, 2015.

    Furthermore, it is universally recognized that production output in shale plays other than the Appalachian Basin area continue to be adversely impacted both by the huge resource potential as well as favorable economics from the Marcellus.

    I do not know your rationale for excluding the Utica play in your work, but it is considered by many to be comparable to the Marcellus. Current production and studies support thus premise.

    ReplyDelete
    Replies
    1. Dear Mr. Coffeeguyzz,

      Have you actually read any of my scientific papers and established independently that they are biased? Or are you repeating bar gossip while hiding behind a pseudonym?

      As to your factual questions, please also read my reply to the comment below.

      I do not trust MY analysis of the Marcellus sufficiently to draw firm conclusions about the ultimate recovery from this play, because it still may be too early to draw such conclusions. Therefore, I allow the Marcellus to produce 4.2 more gas than current cumulative production, which may be significantly too much. I gave all other plays a factor of 3 for their ultimate recoveries relative to current cumulative production. This is a high multiplier and puts me in the biased too optimistic camp, but it doesn’t matter, given my main argument.

      My main argument that production potential of all shale plays in the U.S. has been vastly exaggerated for political and propaganda reasons is unchanged and now supported by sufficient data. While the overall resource is giant, the recovered fraction will remain small because of the generally poor quality of this resource. For the record, let me restate the obvious: Some operators in the small sweet spots in all plays will make a lot of money; most others will lose money and go bankrupt. In the old fashioned reservoir engineering practiced by people of my age, these sweet spots are called reservoirs.

      Delete
  3. 1. Missing discussion of the Utica (growing fast and young, academic geology based report discusses huge resource in place).

    2. Misses the impact of price (and gas on gas competition) driving some basins down. In particular the Haynesville.

    3. The ability to ad hoc add extra Gaussian curves allows you a lot of parameter flexibility to fit any curve. Consider that a single bell curve has adjustable parameters for tightness, mean location in time, and area under the curve. When you have two of them, that is a fair amount of adjustable parameters. And for the H in particular, the improvements in drilling and cost efficiency are incremental and evolutionary (not an all new practice or discovery justifying a new curve).

    4. For curves that have not turned yet (Marcellus), there is a lot of flexibility to fit different curves. You really can't make a good judgment on eventual area under the curve based on the run to date.

    5. The Marcellus is takeaway restricted. This will tend to retard how fast the curve goes up, but preserve more resource to be used later. Will cause deviations from the bell curve shape. And for that matter the speed of adding takeaway capacity will drive the eventual production curve shape. A model just based on fitting Gaussians lacks any input to reflect this.

    ReplyDelete
  4. @ Anonymous: I wish you were not. However, please let me try to answer your reservations:

    1. Missing discussion of the Utica (growing fast and young, academic geology based report discusses huge resource in place).

    I can’t work on every single play in the U.S., because it is an after-hours hobby. I have looked at one small part of the Utica as a consultant and it was lousy. Too deep, too lean in hydrocarbons, and wells were too expensive together with the price of land. I bet the prices have changed in the last two years.

    2. Misses the impact of price (and gas on gas competition) driving some basins down. In particular the Haynesville.

    This is a short blog, not a book. For an analysis of the Haynesville, please look at the Bureau of Economic Geology reports from the Sloan project. I am a coauthor of several of them. You can also look up my publications on the web. Briefly, however, Haynesville is a Marcellus that isn’t. For the same productivity, a Marcellus can be 40-50% cheaper. Therefore, the Haynesville can only come back when Marcellus peters out, or natural gas price is $7-12/mcf. There is a lot of complex economics and many parameters in the BEG papers. You make like them more, but for the last 5 years Barnett wells liked my simplistic analysis better.

    3. The ability to ad hoc add extra Gaussian curves allows you a lot of parameter flexibility to fit any curve. Consider that a single bell curve has adjustable parameters for tightness, mean location in time, and area under the curve. When you have two of them that is a fair amount of adjustable parameters. And for the H in particular, the improvements in drilling and cost efficiency are incremental and evolutionary (not an all new practice or discovery justifying a new curve).

    Now that’s a remarkably wicked non-argument. The Barnett, Marcellus, Eagle Ford and Bakken can be matched with a single Hubbert curve. That’s just three parameters: year of the peak, height at the peak in Bcf/day, and growth rate in %/year. These three parameters model 16,000 wells in the Barnett, 14,000 in the Eagle Ford, 10,000 wells in the Bakken, and ~10,000 wells in the Marcellus (please try to find the exact number of active wells in the Marcellus and let me know if you beat Pennsylvania’s confusing non-reporting). You also use 3 parameters to calculate the always wrong in the long run hyperbolic decline rate of a single well in each of these plays. This would make 48,000 parameters for the Barnett shale. Plus 32,000 ad hoc parameters for the exponential decline rates that follow. Do 3 parameters strike you now as excessive?

    The Fayetteville (5,400 wells) has two cycles because of the infill wells and pad drilling. Haynesville also has two cycles (3,300 wells) for other reasons: the initial very high downhole pressure and well-choking, surface facilities, proppant embedment, etc. That’s 6 parameters, not 16,200 + 10,800 parameters for the Fayetteville if done well-by-well. Ad hoc? A lot of parameter flexibility? I don’t think so. The reason for the robustness of this approach is the omnipresent Central Limit Theorem. I wrote about it in my blog several times and co-authored and important paper on coal, where we thoroughly explained our arguments.

    Remarks 4 and 5. Even though the Marcellus production data fit a Hubbert curve remarkably well, I do not believe my Marcellus prediction and give it 2x1.4x1.5 = 4.2 times more cumulative production than thus far. We’ll see by how much I have overestimated the ultimate Marcellus production, but it makes no difference to my main point. By the way, do you still remember what my main point is?

    ReplyDelete
  5. Thanks for the detailed response. A lot of my reservations remain. I'm not sure how useful it is for me to continue to peck at them (maybe even annoys you) but I do appreciate your time in the response. Ahh screw it...here's a couple things:

    A. "With four parameters I can fit an elephant, and with five I can make him wiggle his trunk." https://en.wikiquote.org/wiki/John_von_Neumann

    B. If you had fitted the H a year or two ago, it would have been a single curve only. In fact, I have had arguments with peakers several times who had done exactly that--I pointed out how the H was getting a "tail" that wasn't dropping as fast as it should. Now that you have two curves it somewhat validates me. However, could you have known to add that second curve a year or two ago? Seems very after the fact.

    ReplyDelete
    Replies
    1. Your point about fitting data is quite fitting, no pun intended. You should then realize that a modern reservoir simulation run may use 7 billion parameters to initialize it. One should perhaps ask if problem at hand is overfitted or not. If this problem is overfitted, then predictive value of the simulation is low.

      The fat tail point is also good. In unpublished work I have used the Generalized Extreme Value (GEV) distributions that arise in fitting whatever statistics one may want to perform on the thousands of wells in a given shale at a fixed time. The GEV distributions are however inappropriate to model time evolution of total rate of production from an oil/gas province or similar. That’s at least what I currently think, but still must prove mathematically. For total rates as a function of time, Gaussians are appropriate.

      The Hubbert curves are really Gauss distributions that arise from the Central Limit Theorem for the means of discrete, uncorrelated random variables with arbitrary distributions and finite variances. The Gaussian only conveys information about what happened or is a continuation of the same trend. This condition obtains if the same large set of wells, already accounted for in the fit, continues into the future. If there are essentially different sets of wells accessing that which has not been accessible with the initial set of wells, another Gaussian arises with time.

      One of the frequent accusations tossed at King Hubbert, one of the two best American-born scientists to date, is that he did not predict the peak level of global petroleum production, even though he was right about the timing 50 years earlier. Hubbert underpredicted the level of global peak production by a factor 2.5. He could not possibly foreseen the Prudhoe Bay, Cantarell, much of Burgan, Iraq, offshore GOM, North Sea, Nigeria, Angola, Brazil, some of Siberia, and so on. His prediction was not an invisible hand of God, but a darn good 50-year prediction by a genius. Ask yourself how many economic models of the entire planet would hold for 60 years within a factor of two or three, when they also have three parameters.

      In summary, prediction is very difficult especially about the future, as Niels Bohr used to say. Predictions from a Gaussian are as robust as predictions can be, but are not magic. I have been quite successful in predicting the future, but I used intuition and educated guesses in addition to science. Please look at my early 2011 prediction of the future gas production in the Barnett, posted a month ago or so. There is as much intuition in it as science. I am pretty sure that my predictions of the Barnett and Fayetteville are almost right. The EF prediction is too low. The Bakken one is low but not by as much. The Haynesville prediction is likely too low if a future development happens years from now. The Marcellus is tricky. Its best areas have been pretty much drilled up, while other areas are usually poor to very poor, but we don’t know. I very much doubt that Marcellus will ever produce 4 times as much as to date. Only time will show the manifold errors of my ways.

      Delete
    2. Feynmann? Franklin? Edison? (Who is the other one?)

      I read the 1956 paper on FF and nuclear power. It is very accessible--I mean that in a good way. You say the critics cite H being off on the peak amount and the defenders cite him being on with the peak timing. But I have a different problem: it is the area under the curve. That is what his whole method is based on. I can forgive him being off with timing or height or even with having multiple peaks. But being off on the amount is the much bigger issue. His prediction of US gas is way off low too.

      To really get into H's work, would need to look at the underlying studies particularly Pratt. While the 56 paper is a nice exercise, a nice explication, what would be more meaningful would be detailed critique and analysis of the underlying studies. And bottom line...he was WRONG about calculating all the produced oil from all the sedimentary basins of the earth (his phrase).

      He even does speak about the phenomenon of multiple peaks. And he cites the pretty example of Indiana production and how seismic technology enabled new peaks. But he doubts there will be new technologies other than gradual improvement of EOR. Yet he missed hz drilling, DW, geosteering, etc.



      Delete
  6. Mr.Patzek

    Re the Utica ... "too deep, too lean in hydrocarbons".
    Well, you may not be able - inclination wise nor tim wise - to peruse the 200 page study on the Utica recently released by the West Virginia University folks (spoiler alert, near 800 Tcf recoverable) but anyone following this field should at least be aware of the Deep Utica well from EQT, the Scotts Run.

    This well, online with just over 8 months production data, has produced over 7 1/2 Bcf.
    It was just ranked the #1 US shale gas well in output by Rystad.
    It has a vertical depth of 13,400'.
    It has a lateral length of just over 3,200'.

    Too deep, too lean indeed!

    ReplyDelete
    Replies
    1. All right, so there are sweet spots in the Utica. My experience there is very limited and I'll try to improve it, time permitting.

      Still Utica has to compete with the Marcellus. This competition is very difficult given Utica's depth and a harder overburden, if I recall correctly.

      Delete
    2. Mr. Patzek

      100 mile distance separates southernmost Antero's Rymer 4HD from Consol's Gaut 4IH. The Gaut's 9,900 psi flowing casing pressure was exceeded by Rice's John Brigg's 10,200 psi FCP.
      The Briggs is expected to flow 12 MMcfd for 700 days.

      Ohio's website, oilandgas.ohiodnr.gov, shows three dozen Dry gas Utica wells produced over one billion cf in the fourth quarter with another 35 producing over 800 million cubic feet.
      Th Utica is coming on very strong/very quickly.

      Delete
  7. I think your main point is that the shales will not deliver the century of Obamerican bliss? Only like a decade or less. Is that a fair caricature of your main point? ;)

    Nevertheless, somehow despite all the doom and gloom about the shale gas, we have somehow managed to grow NG production from under 20 TCF/year to within a TCF of 30 TCF/year within the last decade. [And yes, those are wet numbers but dry shows the same story and I bet you don't like me to put the NGLs with the crude...if I put them on their own it is good too...stunning increase in propane production).

    https://www.eia.gov/dnav/ng/hist/n9050us2A.htm

    Plus we even have LOWER prices now. Even with the higher volume consumed:

    https://www.eia.gov/dnav/ng/hist/rngwhhdA.htm





    ReplyDelete
    Replies
    1. Yes, you got my main point all right.

      Your other arguments are all valid, but they have a time expiry on them. The question to ask is how will we be able to make America run for another century? Or at least a few decades?

      To each of your points, I can provide a qualification and/or a counterpoint, but those would require me to write too much. Perhaps one day we could discuss these points over a beer?

      Delete
    2. Ted,

      I appreciate the work you have done here. Jean Laherrere has emailed me the same kind of oil production profiles on the Bakken and U.S. oil production.

      I write about how energy will impact the precious metals, mining and overall economy on my blog the SRSroccoReport.com.

      I get my share of emails from folks who tell me that the Bakken holds 22 billion barrels of technically recoverable oil and so on and so forth. Thus, no reason to worry. Party on dude.

      For some reason, either these individuals can't see the forest for the trees, or denial of the real consequences of rapidly falling U.S. oil & gas production is too great.

      So, I gather we will just have to wait around until 2020 to see how things turned out.

      Unfortunately, it will be much worse than we probably realize.

      best regards,

      Steve

      Delete
  8. It would be an honor to buy you a beer. Large, cold, free, expensive. You got it! You deserve it for putting up with me.

    P.s. I was going to add Teller as a great American scientist. Wasn't sure if you knew of his fundamental discoveries (e.g. surface area measurements) that have application to catalysis and geology. Or just thought of him as a bomb maker only. But...I got a little worry and looked at his bio and saw he was Hungarian.

    ReplyDelete
  9. Do you like Fourier series? Bessel functions? Or just bell curves?

    I mean even aside from politics, am trying to find a place where we can find a commonality of interests. I know on politics we are probably far apart. I'm an old Cold Warrior. Actually had a Pole dissident give me training in the military on Soviet foreign policy. Can't remember all the nuances of that time, but the parts where he talked about the black market behind the curtain had the class laughing out loud. Sort of a relief of tension compared to what we had to do in spying on our enemy.

    P.s. I did have a Russian tell me that I needed to take the Pole's words with a grain of salt when describing the Russians. Still at the time, I felt like I was getting an inside look.

    ReplyDelete
    Replies
    1. I smiled reading your comment. I like all sets of orthogonal, better yet, ortho-normal, complete sets of base functions.

      The base functions in this particular case of resource production - Gaussians - resolve statistical trends in the time evolution of large sets of wells, fields, mines, forests, etc.

      As you will see from my forthcoming paper, remarkably few Gaussians are needed to do the job, and they predict the trajectory of rate of decline of older sets of producers remarkably well.

      When it comes to predicting the future, Gaussians are insufficient and all else is as well. But they do account for a lion share of the possible future production if large enough sets of producers are captured.

      My wife and I were very much involved with Solidarity and are no friends of communism. The Soviet-style communism impacted our lives quite negatively. Several relatives in both of our families were murdered by the Soviets. But this would be the subject of an entirely different story.

      Delete
  10. Such a great information on Save on electricity. Also helpful to search and full details on Natural Gas quote. Thanks for it...

    ReplyDelete
  11. It would be great to read more concerning that blog..!Thank you for posting this blog.
    gclub
    โกลเด้นสล็อต
    gclub

    ReplyDelete