Wednesday, September 25, 2013

Update to North Dakota Bakken / Three Forks Scenarios

The link above is to a Rigzone article discussing the April 2013 USGS Bakken/Three Forks estimate, which was referenced by Robert Rapier at

The USGS slide presentation can be found at

and click on the slide presentation pdf link near the bottom of that page.

 Near the end of that slide presentation it breaks out the 7.4 BBO for all of the US Bakken/Three Forks into Montana and North Dakota sections of the play.  North Dakota has a mean estimate of 5.8 BBO of undiscovered technologically recoverable resources (TRR) in the Bakken/ Three Forks.

I had been ignoring the "undiscovered" and was incorrectly interpreting the estimate as the full TRR, until I read the following in the Rigzone article:

"While the estimates for Bakken oil resources would appear to be the same for 2008 and 2013, the size of the Bakken estimate increased as well, a USGS spokesperson told Rigzone in an email statement."

"It's deceptive, because although our current estimate of 3.65 billion barrels of oil for the Bakken is numerically the same as the 2008 assessment for the Bakken, you have to remember that oil companies have been producing millions of barrels of oil since the 2008 assessment, gradually transforming the undiscovered resources to the proven reserves then production barrels," the spokesperson commented. "Because our assessments do not include proven reserves or produced barrels of oil, the 3.65 billion does represent an increase."

 About 2.2 BBO of proven reserves have been added in North Dakota since 2005 (from 0.4 BBO in Dec 2005 to 2.6 BBO in Dec 2011 according to the EIA).  If it is assumed that all of these reserves were due to the Bakken/ Three Forks play, then a mean estimate for the total TRR for North Dakota might be as large as 8 BBO and the F5 estimate (or 3P estimate) for total TRR might be as large as 11 BBO.  The F95 (or 1P) estimate of total TRR for the Bakken/Three Forks is about 6 BBO.

Based on this, many of my recent Bakken scenarios might underestimate future Bakken output so I have revised them accordingly.

Figure 1

The range of the TRR from the low to high scenario is quite close to the USGS F95 and F5 estimates of 6 to 11 BBO.  These scenarios all increase the new wells added to an annual rate of 2200 wells per year by mid 2016 (185 wells/month starting in June 2016). Previous scenarios added wells at a rate of 1800 wells per year until Dec 2014 and then increased the wells added to a maximum of 2000 wells per year in mid 2016.  Note that the NDIC (North Dakota Industrial Commission) expects the future wells added per year to be between 1800 and 3000 wells/year.  I have also presented scenarios at (Shale Oil Boom thread) and where the wells added per year remain at the present level of 1800 new wells per year.
I decided to use three different well profiles to give my low, medium, and high scenarios, note that these well profiles are different from my previous post.
Med TRR- qi=10570, b=0.81, di=0.0883, EUR(10)=259 kb, EUR(20)=311 kb, EUR(30)=339 kb
High TRR- qi=11750, b=1.26, di=0.145, EUR(10)=283 kb, EUR(20)=371 kb, EUR(30)=430 kb
Low TRR- qi=10500, b=0.45, di=0.069, EUR(10)=235 kb, EUR(20)=256 kb, EUR(30)=264 kb
These well profiles are Arps hyperbolics with equations of the form (in Excel notation):
Monthly output=qi/(POWER((1+b*di*t),(1/b))), and t is months from first output and output is in barrels per month.  I use t at mid month so t=0.5, 1.5,….., 359.5.
Figure 3
Well productivity (EUR) for all three TRR scenarios is given in figure 3 above.
 When we consider economically recoverable reserves (ERR) we need to make assumptions about real oil prices, average well cost, transport costs, royalty and tax payments, operating expenses (OPEX), and other costs.  We also need to assume an annual discount rate used to calculate the net present value (NPV) of future income streams from sales of output from a given well.
I assume transport costs of $9/ barrel, OPEX of $4/barrel, royalty and taxes of 20 % of well head revenue (refinery gate revenue-transport costs), other costs of $3/barrel, and an annual discount rate of 12.5 %.  Real Oil prices follow the EIA’s AEO 2013 reference case adjusted to May 2013 $, and well costs are assumed to fall from $9 million per well in Jan 2013 to $7 million in Feb 2016 (about 8% per year) and then remain at that level.

Figure 4

As profits fall below $1 million per well in 2019, it is assumed that fewer new wells are added per day and this reduces the rate of decrease in well EUR.  The following chart shows the progression of new well EUR over time when economics are introduced as well as the changes in the annual rate of decrease of new well EUR and how this corresponds to the number of new wells added per day.

Figure 5

Note that the Maximum rate that wells are added over the mid 2016 to mid 2018 period is 2200 new wells per year (about 6 new wells per day), by 2025 the rate that new wells are added has fallen by a factor of 6.  After 2040 prices are assumed to remain flat and it is no longer profitable to add new wells, total producing wells reach a maximum of about 26000 wells in late 2040.

Below is a chart of the ERR for all three scenarios with the same underlying economic assumptions used for the Low TRR and high TRR cases.
Figure 6

Clearly there are many assumptions which have been made to create these scenarios.  All may be proven wrong by future events.  Well costs may not decrease as fast as I have assumed (or they may decrease faster), well productivity decrease is assumed to begin in Jan 2014 and gradually increase to an annual rate of decrease of 16 % ( my guess is that the range is likely to be between 12 % and 24 %), this guess could also be either low or high.  Real oil prices could follow the EIA's AEO 2013 low or high price case rather than the reference case assumed here.  In the future I may explore a few of these possibilities, clearly high prices are more likely under the low scenario and low prices are more likely in the case of the high scenario. 
A final note on the decrease in EUR of the average new well.  The 16 % annual rate was not chosen randomly, the recent USGS mean estimate for the North Dakota portion of the Bakken/Three Forks was chosen as the target for my medium TRR scenario. It was assumed that this decrease in EUR would gradually accelerate beginning in Jan 2014 and the shape of the % EUR curve in figure 3 above was thought to be reasonable and results in a TRR very close to my target. 


  1. Dennis - While (I think) I totally grok the concept of diminishing returns that you, WebHT, AB & others point out about the Bakken, my understanding of your analytic math is rudimentary. So I did a simple check. Tell me if this makes sense: I multiply the mid-range total output of a well from figure 2 (rounded to 350,000) by the rounded total wells drilled figure of 45,000, and I arrive at 15.75 BBO. So that tells me that if anything, your figures are optimistic relative to the USGS high estimate of 11 BBO. Does this make sense? Thanks for your great work on this!

    1. Hi Clifman,

      You are almost correct, but the Medium scenario matches pretty closely with the mean USGS estimate and is not an optimistic estimate. Keep in mind that the “average well” does not remain the same over time. The math (except maybe the hyperbolic function) is mostly just arithmetic (I do not have the math chops of WHT). Your 16 BBO estimate is pretty much spot on, but it would need to assume that EUR never decreases and the entire Bakken/Three Forks play has uniform well productivity. In other words, either there are no sweet spots or technology advances so rapidly that it counteracts the decreasing EUR so that the average EUR of new wells remains constant over time. I do not believe that either of these optimistic scenarios will be the case, the EUR of new wells will eventually decrease as the sweet spots get drilled up.

      How does the TRR get to be 8 BBO rather than 16 BBO for the Medium scenario? Take a close look at figure 3, especially the blue curve. If I had made my model a little less “realistic” and the blue curve were a straight line between Jan 2014(100 %) and Dec 2033 (0 %) then we would be at 50 % of the Jan 2013 EUR in 2023 and our “average well” over the entire 2014 to 2033 period would have an EUR of 170 kb.

      When we multiply 170 kb by 48000 wells we get about 8 BBO, so my medium scenario matches up pretty closely with the USGS mean estimate and is not as optimistic as you imagine.

      Note that the high scenario does not work as well when we do the same kind of simple estimate and the TRR approaches 10 BBO rather than the 11 BBO shown in figure 1. This is because I take the TRR out to Dec 2073 so these “high” scenario wells have been producing at least 40 years (2033 to 2073) rather than 30 years.

      For the high scenario the EUR(40)(40 year EUR) is 475 kb and our estimate becomes 11.4 BBO, using the straight line assumption for the decrease in EUR.

      Hope this helps.

      Dennis Coyne

  2. I am thinking the average won't go over 300K
    Good work

    Have you tried the server yet?

    1. WHT,

      I agree if you mean the EUR(30) for the average well in the ND Bakken/ Three Forks. I have not tried the server yet, What did you use for your well profile? Is it a hyperbolic or OU diffusion? I commented on your blog recently, basically I was commenting on the OU diffusion model an the difficulty with fitting to short term data (24 to 36 months), it seems a hyperbolic fit to the 24 months of data we have and then fitting an OU diffusion model to the hyperbolic works best, then a hybrid hyperbolic/ OU diffusion model can be used where a hyperbolic is used for the first 24-30 months with an OU diffusion model thereafter. I forgot to include the oil produced in my TRR estimate, when this is added the mean USGS ND Bakken/Three Forks TRR becomes about 8.4 BBO.


    2. Sorry for the delay in responding. (I converted your images in the comments to display over at, btw)

      I did use an OU profile for the latest fit and the mean is 280,000 barrels cumulative. I also did the painstaking job of pulling each of the monthly records so that I would get a true cumulative count instead of the "active" count that the NoDak crew puts on their site.

      I think the strong flattening of the OU curves is vital to getting a good fit because it simulates the wells that have been shut in and gone inactive after some time. OTOH, if we only consider the active wells, then the hyperbolic curve seems to work better. The reason for this is the weeding out of the inactive curves does the cumulative flattening.

      I will look into first 24 months of data some more.