Bakken Projections

The Dynamic Context Server features an interactive Bakken Oil model showing the Red Queen effect. The model uses historical oil well count and cumulative production to estimate average well output over time and then project that a number of months into the future.

The North Dakota Mineral Resources Department releases monthly data which we use to fit against. The start of the model is set to when the oil production statistics began and continues to the recent month 378:

DC at the peak oil climate and sustainability blog recommended a projected growth of 150 new wells a month.

This leads to a maximum asymptotic value of 1.4 million barrels a day.

The Red Queen dilemma describes the effect of requiring ever increasing rates of increase to make up for quickly declining wells. Depending on the ultimate reserve potential of the Bakken formation, the production may never reach close to 1.4 million barrels per day before the number of new wells starts to decline.

Related pages:

Note: The early data from NoDak is contaminated with production from conventional non-fracked wells. In terms of a fit for recent data this does not make much of a difference, but as DC notes, it will not give you insight into when the fracking actually commenced and what the fracking baseline was, which may actually be about 25x lower in pre-boom times than that shown. In other words, the true Bakken output would have started much closer to zero in 1980 and the actual boom-era starting point is around 2002 or month 275 according to DC’s analysis.


33 thoughts on “Bakken Projections

  1. Note that 150 wells added per month was suggested in part to match the figure from Rune Likvern. Other numbers such as 200 wells per month would lead to higher asymptotic values such as 1.8 MMb/d. It is unlikely that the economics and eventual decrease in new well EUR will enable these levels of production, 175 new wells added per month is about as high as it will go, and if there is no decrease in new well EUR we could see 1.6 MMb/d, but we will quickly run out of space for new wells at this rate and if 48,000 producing wells is the maximum as suggested by the NDIC, we will see a rapid dropoff in output when new wells can no longer be added.



  2. WHT,

    Is it possible to add decreasing new well EUR in your Bakken Model, possibly by convolving with a percentage decrease of 0.8 % per month starting in Jan 2015 (maybe use a 0 % decrease from month 1 to month 396). At some point EUR will decrease, we don’t know when or by how much, though your server is so cool you could allow a range of inputs from 10 % annually to 30 % annually and also allow the starting month of the EUR decrease to vary from month 379 to 500. Just a thought, I could try to adjust your code and then email it to you to see if I am on the right track. I am still working on my EF data.



  3. DC, To do it exactly as you describe is not so easy, since I am taking advantage of a stationary extraction process to make the convolution calculation a two-liner in the language that I use in the server.

    Monthly_Wells derivative Number_Wells/Time,
    Total convolve Single_Well*Monthly_Wells

    On the other hand, we could potentially just scale back the number of new wells that come online. A decreasing EUR is equivalent to a scaled-back “efffective” number of new wells that come online in this case.

    This is basically applying some mathematical symmetry to the analysis to make it into a simpler problem.


    • I suppose that could work. I was thinking several convolutions could be performed as in your world shock model. Couldn’t the decreasing EUR be somewhat analogous to an extraction rate in your shock model. Early on EUR is 100 %, later it falls to some lower level. It seems that in the shock model, the extraction rate could be at 20 % for a period and then fall gradually by 10 % annually to 2 %. I was thinking decreasing EUR per well could be handled in the same way. In year 32 it is 100 %, it falls to 90 % in year 33, etc.. I know what my response would be, “you write it”. I may play with your shock model to see if I can make it work using ADA. If I can do that, I will send you the code, which will just be a modification of your original code in the Oil Conundrum.
      I am not up to speed on the code used in the DCS, so it is really beyond me at this point to create something.

      The effective well idea is interesting, but I think it would be confusing to many people. I get it though, essentially if the EUR is say 75 % of the present new well, we would simply add 150*.75 or 112.5 new wells rather than 150, a simple and elegant solution (and much less work). I guess you could just add an input of annual EUR decrease to the red queen model and handle it this way and it would be no different than my more complicated solution(which I cannot implement).

      I will run a quick test in my spreadsheet to verify that the two approaches are equivalent (as you said the symmetry suggests that this should be true.)


  4. Hi WHT,

    In your Red Queen 2 Model I came across the following code:

    cumulative_diffusional_dispersion(Amp, D, Theta, Year, Result) :-

    Tau is (1-exp(-Theta*Year))/Theta,

    Result is Amp/(1+sqrt(D/(Tau+0.0001))).

    I am a little confused by this. I assume this is the OU diffusion model.

    Previously I thought you had given me the following formulation for the OU diffusion model:

    Cumul_output = Amp/(1+1/(sqrt(D*Tau))), where
    Tau = (1-exp(-Theta*year)))/Theta

    This may just be a lack of understanding the code, on my part. Should your code have been:
    Tau is (1-exp(-Theta*Year))/Theta,
    Result is Amp/(1+1/(sqrt(D/(Tau+0.0001)))). ?

    I am also unclear about why the 0.0001 is needed, or is this just to prevent infinite values at t=0?

    Also, I confirmed that your suggested effective well idea to account for a decrease in the EUR of new wells, works just fine. If EUR decreases at an increasing rate from a 0 % annual rate to a 9 % annual rate over an 18 month period starting in Jan 2014 we come fairly close to the USGS mean estimate for TRR from 1953 to 2073, if we assume about 48000 for a maximum in producing wells and about 2000 producing wells added per year. Maximum output in this scenario is about 1.1 MMb/d in mid 2017. Note that economics is ignored in this case, though under reasonable assumptions this same output rate will be reached, but fewer than 48000 wells will be added and lower cumulative output is reached (less than 8.4 Gb) under any reasonable economic assumptions.



  5. DC,
    You are correct, the D*tau combination is what it should be. I inverted D to be 1/D and adjusted the constants that way.
    Result is Amp/(1+sqrt(1/(D*(Tau+0.0001))))
    This doesn’t make any difference in the result but it does show sloppiness on my part.

    The 0.0001 factor is so that the time doesn’t blow up at zero just as you expected. The Result is cumulative so that it should be close to 0 initially and then go to Amp at long Tau.

    As to the second part : I am glad that the symmetry scaling approach works.

    If you think of a convolution kind of as f(t)*g(t) where g(t) = k*h(t)
    f(t)*(k*h(t)) = (k*f(t)) * h(t)
    by associativity, where we switched the constant multiplier to the other function, and all is square. That’s what was running through my head when you posed the request, and why I thought to take the lazy way out of it.


  6. Ok, thanks. When I tried to match your code in excel I ended up with something that ended up looking like a hyperbola, as I am not as accomplished as you in mathematics I got confused. I may try to implement something in excel that can be easily adjusted like on your DCS, unfortunately I am not sure I will be able to write the entroplet code properly.




    A new post is up at my blog. A new interactive spreadsheet is available with more flexibility, like Paul’s Red Queen 2 model the wells added/month and # of future months can be changed. In addition the annual maximum rate of decrease in new well EUR can be changed, when that decrease begins, and how long it takes to ramp up to its maximum rate of decrease. Enjoy.



    • I just wanted to be clear that the scenario with TRR=16 Gb is very unrealistic, I would expect no higher than 1.5 MMb/d from the Bakken/ Three Forks and more reasonable scenarios have a TRR between 8 and 9 Gb and a peak of 1.1 to 1.3 MMb/d in the 2018 to 2020 time frame.



  8. I did use the img tag, there seems to be a problem with word press.

    Supposedly the default settings in word press do not allow images to be posted in comments, you would probably need to create a dummy account to try this yourself. If you try the plug in let me know and I will try to post an image.


    I show my code above in quotes.



  9. Ok. Thanks.

    I keep making incremental changes to my interactive spreadsheet, but what I should really do is try to understand your context server better so I can contribute. My latest incarnation is to allow the user to ramp the number of wells added per month from a starting value. For example the user inputs 150 wells per month as a starting value (in Sept 2013), and then can specify that this rate increases by 2 wells per month for 50 months so that wells added per month ramps linearly from 150 to 250 over 50 months and then remains at that level until the total months are reached ( a maximum of 336 in my spreadsheet).

    Eventually I would like to introduce economics into the model, but I am finding it a challenge to automate that process. So far the spreadsheet only gives an expected TRR and indicates breakeven when breakeven is reached (see the breakeven chart). I am not sure if you have checked it out, I will put my latest on google drive with the link below if you are interested.

    I checked out your new model more closely, nice job, as always.

    Hard ramp is pretty clear ( no EUR decrease), diminishing returns I would think means that new well EUR decreases. By how much as an annual percent or monthly percent? I assume the intermediate also has diminishing EUR but maybe by half of the diminishing returns?

    I am trying a chart below, it is a jpg:



  10. DC,
    The intermediate ramp is interesting.

    The hard ramp is clear as you say, it rises at the constant rate of X wells/month until it hits the limit.

    The diminishing returns ramp is essentially an exponentially damped response that hits the limit asymptotically.

    The intermediate is what I would call a critically damped response which follows the hard ramp more closely and then transitions into the limit more tightly. But it is smooth and so doesn’t show the sharp discontinuity of the hard ramp.

    This matches the Goldman Sachs chart that you linked to and that what your chart follows. It does look like they are seeing a 60,000 well limit, which seems like a large number and that may not be sustainable,


    • I do not think the Goldman Sachs scenario is very plausible, my aim was really to show that unreasonable assumptions need to be made to approach a 2 MMb/d peak in 2023, my guess is that the high end will be near the USGS F5 estimate of 11 to 12 Gb and that scenarios up to about 9 Gb for TRR are more reasonable.
      So I take it you are using your effective well idea to simulate a decrease in new well EUR? To my eye the diminishing returns model looks more realistic. Nice work! Is it possible to allow the user to input the month that the transition begins and the number of months until the maximum rate of decrease in delta wells added occurs. I will investigate a little further to see the shape of your wells added curve.



      • The diminishing return models the difficulty of constructing a new well when the pickings get thin.

        The difference between 60,000 wells and 30,000 is huge for sustaining anything close to 2 million barrels a day.


  11. Your diminishing returns looks a lot like the effective well idea that you suggested. In fact for a well EUR decrease starting in Jan 2014 and reaching its max monthly rate 18 months later, 60000 wells is like 26500 effective wells if the max annual EUR decrease is 12 % (which results in 8.5 Gb for my model) and real wells are added at 200 per year. The models don’t quite match up, but we are using different well profiles, so we wouldn’t expect a match, your peak is lower and I am not sure of the rate that your well additions decelerate, for my model the maximum monthly rate of decrease is about 1 % per month (1.0596%), this rate being reduced in a linear fashion over 18 months


    • 60,000 wells with a typical well spacing of 320 acres — 2 wells/square mile — covers an area of 173 miles by 173 miles.

      This is the green square on a map of North Dakota. Is this even a remotely possible scenario? Incredible what we will do to extract oil.


  12. Thx, WebHT & DC for hashing this all out. You do a great service for all of us paying attention to reality, and not hype. No, we will not do 60,000 wells, but we will bust a lot of sod & break a lot of hearts in the desperate attempt to squeeze the Bakken dry.


  13. According to recent presentations at the Geological Society of America meeting in Denver:

    Though it’s true that oil production in fracked wells declines very quickly, any production peak in any of these tight oil fields would be market-driven, not driven by geology, said EIA analyst John Staub.

    “To fully drill-out the Bakken, for example, (total wells would be) in the 40,000 range,” he said. “They’re drilling 1,800 wells a year now. There’s a long ways away to fully drilling that out. David’s (Hughes) come in and talked with us. Yes, the wells decline quickly, but there are a lot of opportunities to drill wells.”

    With 40,000 wells, peak within 5 years and max 1.7 million barrels/day.


  14. Pingback: Simple models of forced warming | context/Earth

  15. Pingback: Dealing with the Dynamics of Diffusional Sequestration | context/Earth

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s