Thursday, March 10, 2011

Bakken Shale Oil

A look at shale oil:
I have also looked a little more at the Eagle-Ford and Niobrara plays. My conclusion is that all of the producers are seriously exaggerating what they have, but what they have is still valuable.

In the analysis below I have concluded that the Bakken could probably be brought to 3 million barrels/day by 2020 and maybe the sum of the other shale plays could reach a similar level. (Note added 14/8/11 - As pointed out below, there probably is not enough space to drill to ever get to the necessary number of wells, so my conclusion is probably wrong, by a large amount).The USA uses about 20 Mb/d (for me M=million, but in the oil industry it is thousand and MM is million), and imports about 13 Mb/d. I expect imports to decline by 1/2 before 2020. Most of the decline could be offset by rapid development of domestic production, which is now viable due to price and technology. To get there, by my estimates, the industry would have to add about 140 horizontal rigs/year, on average, for the next 10 years. Given the MENA unrest, and the sustained high price for oil, that development now has a high likelihood of happening.

These horizontal wells, on average, seem to be 1-2 miles deep, and near 2 miles horizontal. That's 3 to 4 miles of drill pipe per well. 60,000 wells in the next 10 years could mean 200,000 miles of pipe. I don't know how much steel is in a pipe, but let's guess that 1 car equals 200 feet of pipe. That comes to about 5 million cars over 10 years, or about 4% of USA annual car production. I am surprised that it is that small. Even if I'm off by a factor of 2, we could supply all of the pipe just by building 5% fewer cars at 5% less steel per car. Given the expected increase in gasoline prices that is almost certain to happen anyway.

However the big limitation is environmental. Getting to 3Mb/d in the Bakken means drilling 30,000 producing wells by end 2020, and drilling near 5000 wells in the single year of 2020, vs about 500 in 2009. Where does the water come from, and how is the contaminated water treated for multiple frac'ing of 30,000 wells, or 5000 wells per year? Technically it's feasible, practically I'm not so sure.

Some urls:

My analysis:

Looking at Continental Resources’ (CLR) and Brigham’s reports we find one excellent consistency and a couple of major inconsistencies.
Brigham’s average production of 1800 b/d for the first week, 1100 b/d for the first 30 days, and 830 b/d for the first 60 days falls on a smooth curve that includes CLR’s 450 b/d for the first 90 days. This curve implies a production rate of 90 b/d at the end of the first year, only 5% of the first week’s average production rate.
If we take a production rate at the end of the first year of 100 b/d, declining to zero at the end of the 10th year, we have 165 kb for those 10 years. Add in the 75 kb from the first year and we have 240 kb EUR. This is a long way from the 300 to 400 kb EUR projected by CLR in their EUR vs frac stages chart. Their median EUR projection seems to be about 360 kb which is still high vs the decline curve, but much lower than the 518 kb they have modeled. Very strange.
I have tried curves using the 30, 60, and 90 day average rates as end of period rates, or as average for each of the first, second, and third 30 day periods instead. Thus I have a range of EURs from 240 to 410 kb, with a middle value about 340 kb. However the low end is the one consistent with both companies’ real wording. The safest EUR number to use is probably 250 kb/well.
In their curve of IP vs EUR Continental shows a first 30 day IP of 600 b/d corresponding to an EUR of 500 kb, and a 30 day average IP of 1000 b/d as an outlier vs 1100 as an average for Brigham. However a 30 day IP of 600 subtends a 90 day average production a long way below 450b/d, and therefore a much lower EUR. CLR’s EUR estimates look totally unrealistic, and inconsistent with their other data.
There is a lot of inconsistency in the Continental data. Their 30 day production rate curve vs frac stages must be for a lower producing set of wells than the 90 day curve.  Both curves suggest that anything over 18 stages is well into diminishing returns. Near 15 stages looks optimum.
From Nov ’09 to Aug ’10 CLR has gone from 4 to 18 rigs, and from about 25 to about 33 “days on well”, about 10 wells/rig/yr at end 2010.  Elsewhere they report 800 “gross” wells for 562 “net” wells. This would suggest 30% non-producers, , ie 7 producing wells per rig per year.   With 164 rigs  in Q4 2010, one would expect an average of at least 120 for the year, providing less than 700 new wells. A judgment call would suggest 6 producing wells per rig per year.
CLR then shows about 2300 wells producing in the Bakken at mid 2010, of which 2075 are horizontal multi frac wells. 2300 wells produce about 350 kb/d at end 2010, or 150 b/d/well. It looks like there were 400 wells added in 2007, 550 in 2008, 500 in 2009, and > 600 in 2010. An average production in the first year of 500 b/d/well, declining to 160 in the second year, 80 in the third year and 65 in the 4th year, etc.,  would give us near 150 b/d/well at end 2010, which is consistent, and is also consistent with an EUR of about 250 kb/well. So we now have some numbers to project future growth in output.
CLR claims there were 164 active rigs working in the Bakken in late 2010. If we add 60 rigs per year to 2020, and maintain 7 wells/rig/yr, and if my calculation is correct, we would be producing 3 million barrels/day in 2020. If present production declines by 4%/yr through 2020, we would have a net gain in output of 0.5 million barrels/d to offset declining imports, 10% of the expected decline if we are lucky.

See, (8/14/11) and especially note this contribution from Rockman
FOR ALL Another meaty story by HO...mucho thanks. I've never worked this trend so have nothing to contribute in particular about it. But I want to point out to those who don't appreciate why this is an "unconventional" play. HO's general decline curve says it all. This is not a field in the conventional sense. It is a "play". Compare to the Mother of conventional fields: Ghawar. It is still producing a significant volume after many decades. But more important: many of the original wells are still producing. They have periodically drilled new wells to enhance recovery but the main reservoirs are still prodcing. HO’s decline curve clearly shows the Bakken is a very different animal. Original Ghawar wells came on at 5,000+ bopd and held there for decades. For the most part a Bakken well’s production becomes relatively insignificant after several years compared to thir imtial rate. There is some history of success re-frac’ng such wells but seldom as valuable as the initial completion.
No doubt the Bakken, Eagle Ford and similar unconventional plays will add significantly to our reserve base. But at a cost that requires higher oil prices. But it’s still important to remember there is an end to this road also. These are not the first major fractured oil reservoirs to be developed with frac’d horizontal wells. The Austin Chalk (a “limestone” shale) had it big day in Texas. Same big initial rates, same decline rate profile, same boost to oil production, same hype. And for the most part the play is dead even at current high oil prices. And for a simple reason: almost no more drill sites left to poke. In fact the current boom in Eagle Ford drilling is a result of the decline in AC drilling: it’s one of the few oil plays left in Texas. And in time it will be fully developed just as the Bakken will. And once most of the finite number of locations is drill the oil rates will drop off a cliff.
These plays can help give us some breathing room. But they don’t change the game significantly IMHO. Would be nice if we use this crutch to give us a little more time to make adjustments to PO. But I have my doubts.

Saturday, February 26, 2011

Shale Gas

Projecting the Near Future of USA Natural Gas Availability    Feb 11, 2011
1)       A look at conventional production

Let’s start with Jean Laherrere’s curve from 2006, which suggests that conventional NG production was about to “fall off a cliff”, which as we all know, didn’t happen. Was Jean simply wrong, or is there something more at play? A second look at discoveries vs conventional (marketed minus unconventional) shows that the area under the discovery curve is quite larger than under the conventional curve, so that there was still more to produce before the fall off as of 2006.  However, conventional production in the USA declined 4% from end 2002 to end 2006, so if the decline were to be arrested, it clearly would have required an increase in drilling. Web Hubble Telescope, in his magnum opus entitled The Oil Conundrum, has noted that oil producers will increase annual production as a % of reserves in order to keep production on a plateau when faced with a decline. It seems likely that NG producers would do the same. In fact vertical rigs for drilling conventional plays increased rapidly from 653 working rigs at mid 2002 to 1173 rigs at mid 2007.

Some of the results of that effort can be seen in Texas with a large increase in producing wells after 2002, simply keeping production flat. Vertical operating rigs dropped off sharply after the economic crisis in 2008 down to 275-330 rigs from Q2 2009 to end 2010 (a 74% decline), and at least some of these rigs were used for shale gas development, suggesting that there is little more economic potential for increasing conventional production, or even keeping it on a plateau. We may be at the edge of the cliff.
Indeed, scaling the discovery/production curve with graph paper and eyeball, (or shifting the discovery curve 27 years instead of 23 years) it looks like all of the area under the discovery curve has been produced by the end of 2010. In fact, conventional production has decreased by about 2 Tcf from 2006 through 2010, if the EIA numbers are right.  There is a strong possibility that EIA has overestimated shale gas production for 2009/10 (see below), in which case conventional production has remained on a plateau, increasing the probability that the production rate can no longer be maintained. If conventional production does now drop off parallel with the discovery curve, it will drop by >10 Tcf/yr by 2015, or by > 2 Tcf/yr per year going forward.
 Now the big question is, can increases in shale gas production offset such a rapid decline in conventional production, or is the famous NG glut about to turn into scarcity? To address that question we have to get into an analysis of shale gas production potential.

2) What about shale gas?
There is considerable variability among sources and plays. Berman provides a curve that implies an average EUR of about 1.25 Bcf/well for the Barnett, but says he calculates an EUR of .84 Bcf. The USGS assessment of 14 “units”, ignoring the 4 lowest, would average closer to Berman’s lower EUR, but their curve of decline implies an EUR of about 1.0 Bcf. However they provide a forecast with a 50% probability of 2.1 Bcf, which seems very high, perhaps based on the Woodford (see below).
At 1.0 Bcf EUR, and using Berman’s decline rates the first year production would average 520 kcf/d, for the Barnett. (Bermann uses M for 1000 and MM for 1 million, but as an electronic engineer, I will us k for 1000 and M for 1 million). Pennsylvania has reported production of 180 Bcf from 632 operating wells between end June 2009 and end June 2010, giving 285 mcf/well/yr or 780kcf/well/day for the Marcellus shale, which is at least in the same ballpark. The USGS data gives a first year average well production of 930 kcf/d for the Fayatteville, and 1100 kcf/d for the Woodford. The big outlier is the Haynesville, which is twice as deep as the average for the above, is under correspondingly more pressure, and has denser gas pack. First year production averages about 4400 Kcf/d/well, roughly 5 times the average of the shallower plays.
Berman concludes that the EUR is proportional to the first years production rate. Using the above numbers, and an EUR of 1.0 Bcf for the Barnett, we get EURs for the other plays as 1.5 Bcf for the Marcellus, 1.8 for the Fayatteville and 2.1 for the Woodford. Production falls off to about 50% of the first year in the second year, 67% of the second year in the third year and 88 % of the third year in the fourth year. The first 5 years produce about 70% of EUR. Using these results we get an unweighted average EUR of about 1.6 Bcf per well across these 4 plays, better than Berman’s estimate, but much worse than operators’ claims. The Haynesville might have an EUR of near 5 Bcf/well.
I have estimated rigs used for drilling for gas from Baker-Hughes data, with realistic SWAGs of % of total horizontal rigs and total gas rigs. On average there were about 440 horizontal NG rigs operating in 2008, with a peak at mid-year and then a decline to about 420 at year end. Rigs dropped off sharply in 2009 to a bottom near 350, and then picked back up again to 480 by year end, giving a time weighted average of near 400. Growth continued through 2010, to average about 565, with a probable start for 2011 of 670. It is assumed that the 2300 uncompleted wells at end 2010 were drilled using vertical rigs, in order to avoid losing leases.
Pennsylvania numbers give 7 operating wells/rig/year. Other data suggest 9 or 10 wells/rig/yr. USGS has unproductive wells at 10 percent of total, in early years declining to about 3% in 2009. I have done a piece-wise linear estimate (see below) of rigs, wells/year, and production/well using 10 wells/rig/yr as a base case with the above first year production figures mix weighted to the higher producers after 2006, with declining production year to year as described above, and added a generous increment for vertical wells, and it simply isn’t possible to get to the EIA numbers for 2009/10. Other years are pretty close. 2011 completions of the 2300 carryover wells have been assumed at 14/rig/yr for the horizontal rigs.
I have to conclude that the EIA numbers are wrong, and that shale gas production has been like 2005-0.18Tcf, 2006-0.4 Tcf, 2007-1.0 Tcf, 2008-2.1 Tcf, 2009-2.3 Tcf, 2010-3.5 Tcf. The EIA has 2006/7/8/9 as 0.3/1.0/2.3/3.4 Tcf. We are close for 2006/7/8, but then my estimate drops to 68% of  the EIA number for 2009. It should be noted that the EIA had 2009 total consumption at the same level as 2008, when in fact it should have dropped by at least 5% due to the economy.
Estimates for wells/rig/yr and first year average production rates  are probably optimistic and no allowance has been made for  rigs moved to drilling for NGL. Wells/rig may drop if a lot of new rigs are deployed. First month average productivity may not be maintained as drilling moves away from the best of the sweet spots. The 2011 estimate could easily be 20% high.
If conventional gas drops off by 2 Tcf in 2011, shale gas might optimistically replace 70%. We are likely to finish the 2010/11 withdrawal season with about 1.5 Tcf in storage, and if summer of 2011 is not as hot as 2010 we could reduce summer consumption by 300 Bcf, or maybe 200 Bcf after allowing for growing demand from the recovering economy. If we can draw down ending withdrawal season storage by 400-500 Bcf without causing any problems we can just offset the 2 Tcf decline in conventional production, and squeak through without disruption. If the drop off is just a little more than 2Tcf, we might still squeak through if we can increase imports. Are the variables too optimistic? Can cash strapped producers increase working rigs fast enough?
Prices will go back up, and drilling will pick up again. With no more shut in wells to bring on in 2012, no more room to draw down storage, little room for increased imports, if we have another 2 Tcf decline in conventional production, we will be in severe difficulty by mid 2012. Shale gas cannot make up the difference.

Piece-wise Linear Shale Gas Production Estimate - Feb 11 2011
·         Base case- 100 rigs @ 10 wells/rig/yr, all produced on first day of year. First year average well productivity = 100 kcf/d.
o    Annual production = 1000X365X100 = 36.5 x10e6 kcf/yr
·         Real first year case – each rig drills 1 well in each of 10 equally spaced periods through the year. The first well produces at the the first year average rate for 365 days. In between wells produce at consecutively higher rate for shorter time. The last well produces at the first month peak rate (140% 0f first year average) for 36 days. Total first year production rate multiplier “m” is 62% of base case.
·         Successive years for a single well produce at reduced rates relative to first year average as  2nd year – 50%, 3rd year – 33%, 4th year – 29%, 5th year – 26%, 6th year – 24%, 7th year – 23%. Graphing these average rates one can determine start and end rates for each year. The first and last period wells’ start and end average rates can then be determined and the averages averaged to get the decline rate for the ensemble of wells drilled through the year. Relative to the base case the ensemble successive year production rate multipliers “m” are 2nd year – 75%, 3rd year 42%, 4th year 31%, 5th year – 27%, 6th year – 25%.
·         For each year we can now identify multipliers n and p, where n100 is the average number of rigs drilling 10 wells each, and p100 is the average production rate for the year in kcf/d/well.
·         Using the constant K = 365x10e9 we can then calculate any year’s production for any beginning year as mxnxpxK for that year.
·         We can then sum the productions from each prior year and the current year to get total production for the current year.
·         Further adjustments must be made for fewer wells/rig/year in early years, and for production in any given year  from vertical wells. A learning curve wells/rig was assumed for 2005/6/7 as 5, 7, and 9 wells, with successive years at 10. Vertical well production is a reasonable fudge factor that gets close to EIA totals for 2007/8.
·         2009/10 production has been adjusted down by 1000 and 1300 producing wells respectively to account for the 2300 non-producing wells at the end of 2010, and 2011 has been adjusted up by the full 2300.
·         “p” has been determined using the above mentioned average first year production for rates for the 5 plays where figures are available. A mix has been SWAG’d between Barnett and Other.  Other is the simple average of The Marcellus, Fayateville and Woodford rates for 2007/8/9, with 5% Haynesville added for 2010/11.2005/6 are assumed to be Barnett only. The mix weighted average rate is then calculated for each year.
·         “n” for 2005 through 2007 has been taken from a study I did in 2009, and 2008/9/10/11 has been SWAG’d from several sources but based mainly on Baker-Hughes rig counts.
UPDATE – 2/14/11
·         After thought, added 2/14/11. - The 2300 well carryover could come from using vertical rigs to drill wells in 2010 to avoid losing leases. In this case all of the horizontal wells drilled in 2009/10 could be produced in 2009/10, which would raise production in 2009 through 2012.
·         Also a more detailed analysis of Baker-Hughes data has led to a revision of the 2010 rig count from 500 to 565.
·          Assuming completions of the 2300 wells could be done in 2011 by horizontal rigs at the rate of 14 wells/rig/yr that would raise the average wells/rig to 11 for 2011.
·         The end result table has been updated accordingly
End result:                            
                                                                                2005       2006       2007       2008       2009       2010       2011       2012
n    (100 producing rigs)                                     1.6          2.4           3.1           4.4           4.0            5.65        6.5          7.5
p (production rate 100 kcf/d)                             5.2          5.2          5.8          6.3            7.6            8.8          10.0        10.0
Wells drilled                                                        1600       2400       3100       4400       4000       5000       6500       7500
New wells produced                                           1600       2400       3100       4400       4000       5000       8700       7500
Vertical well multiplier                                        1.5          1.4          1.2          1.5          1.3          1.4          1.1          1.1
Wells/rig adjustment                                           0.5          0.7          0.9          1.0          1.0          1.0          1.1          1.0
Shale gas produced (Tcf/yr)                               .18          .42          1.0          2.1          2.3          3.5          5.0          5.4
EIA                                                                                           0.3          1.0          2.3          3.4          4.4

Addition 2/26/11
Since completing the above, I came across the following chart (slide 7) from an Oct 2010 EIA presentation here . The chart shows shale gas production in Bcf/yr, by play, from 2000-2010.  From end 2006 to end 2010 production rate increases 0.9Tcf/yr/yr average with 2010 up 1.1 Tcf/yr.
  Consider the Barnett production. Working rigs have swung between 86 and 76 since late 2009, with a probable average near 81. Note that with no increase in rigs, drilling has just barely offset declines during 2010, and without an increase soon, production will begin to decline. With Haynesville rig count down, production rate 2011 will likely increase <1.2 Tcf/yr including holdover completions..
Scaling this chart we can easily determine year end production rates, and average annual production by basin and total.
Total results are:                                                  2008       2009       2010       2011 Est.
Year end production rate (Tcf/yr)                    2.2          3.2          4.4          5.6
Production for the year      (Tcf)                        2.0          2.8          3.9          5.3
My original estimate            (Tcf)                        2.1          2.3          3.5          5.0
EIA production for the year (Tcf)                      2.3          3.2          4.4
Comments :      1) I underestimated Haynesville production for 2009/10
                           2) EIA used year end production rate instead of production for 2008/09/10  
Having underestimated Haynesville shale gas production, I decided to do some more digging. First was a thought experiment to see if the EIA numbers were realistic. Starting with first year production of 4.4 Mcf/day and the decline rates used above, and a guess of 7 wells/rig/yr, it only required about 10 rigs in 2008, 40 in 2009, and 70 in 2010 to provide the EIA production – all OK. Then I dug up an actual average rig count of 170 in 2010, down to 150 in Jan 2011. At 5 wells/rig/yr and a 1st year average of about 2.7 Mcf/d  that works out. However see the following  figure. Very advanced multi-stage frac’ing and we get  9  Mcf/d in the first month, declining =>80% in the first year, so about 4.0 Mcf/d average for the first year, pretty close to the original figure. Of course that seems to be based on about 5 wells operating for 9 months and 52  for 1 month. There are probably near 1000 wells operating now, so the averages have probably declined. I will go with the first year 2.7Mcf/d average, and a lot more rigs than expected.

With the very large first year production, producers get a quick payback, and then, in the above curve. estimated decline flattens at a level about 3-4 times higher than the shallower plays. Haynesville should be the center of attention, so why a declining rig count?.  With no increase in working rigs it would take 6 years at the 2010 rig count to increase annual production by 1 Tcf, and would take almost 20 years for cumulative annual declines to offset new first year production, assuming the above estimate is correct.  By doubling rigs in 2011 to 340, we could increase  production vs 2010 by 0.8 Tcf/yr in 1 year, and 1.0 Tcf/yr in the second year.
However, note that  this estimate was based on only 9 months data, and only a few wells.  What if production simply falls off rapidly (as might be inferred from the "Normalized Production History")? A given well's production could go to near zero in a very short time, maybe less than 7 years. In such a case, at constant rig count, production growth stops in year 8. Doubling the rig count in 2011 would increase projected production by 1 Tcf in the 3rd year. With only 3 years experience on the earliest wells, it is not possible to make a high confidence level projection yet.
Given the time required to build rigs, train crews, negotiate leases etc., it is likely that it would take 3 to 4 years to double the 2010 rig count. Conventional production could fall by 6 to 8 Tcf/yr in that time. 2012 still looks highly problematic. 

References Texas NG production info to 2005 - basis for modeling conventional prod’n? Lots of info, quality hard to assess. several shale gas articles Interesting analysis, but I have no hot-links to the figures mentioned. Interesting data source and  Some helpful USGS analysis, quite conservative for a change.