Lessons from the Barnett Shale suggest caution in other shale plays

By: Arthur Berman

Published: Mar 29, 2010

The Potential Gas Committee announced in late June that shale gas has increased the US resource base to more than 1,800 Tcf. Shale players are now more confident than ever that they are in the right business. The miracle of low risk and high reward must be working. The executives of these companies and the investment analysts that promote their stock have proclaimed a new miracle in which high capital costs combined with low gas prices somehow result in profit.

Since little is known about the commercial potential of new shale plays like the Marcellus and Haynesville, I decided to see what could be learned from the robust production history of the Barnett Shale. What I found surprised me. Most reserve predictions based on hyperbolic production decline methods were too optimistic when compared with production performance. There is little correlation between initial production rates (IP) and ultimately recoverable reserves (EUR). Average well life is much shorter than predicted, and the volume of the commercially recoverable resource has been greatly over-estimated. Core areas of the play do not have appreciably higher average EURs than the play overall, and the EUR from horizontal wells is not significantly greater than from vertical wells. Finally, average well performance has decreased consistently since 2003 for horizontal wells.

The Barnett Shale was the first shale gas play to be commercially developed and is the standard of comparison for this play type. There are now almost 12,000 producing Barnett wells, of which two-thirds are horizontal and one-third are vertical. Cumulative gas production is 5.64 Tcf, of which 3.62 Tcf comes from horizontal wells and 2.02 Tcf from vertical wells.

In 2007, I projected EUR for almost 2,000 horizontal wells in the Barnett Shale (World Oil, November 2007). At that time, these were the only horizontal wells with enough production history to evaluate. Now, with two additional years of production, I revised the decline curves for the same control set of 1,977 horizontal wells. The overall EUR decreased 30% from my previous estimate, and the average per-well EUR fell from 1.24 Bcf to 0.84 Bcf. The reason is clear: most wells do not maintain the hyperbolic decline projection indicated from their first months or years of production. Production rates commonly exhibit abrupt, catastrophic departures from hyperbolic decline as early as 12-18 months into the production cycle but, more commonly, in the fourth or fifth years for the control group. Pressure is drawn down and hydraulically produced fractures close.

Workovers and additional fracture stimulations may boost rates back to previous levels, but rarely restore a well to its initial decline trajectory. More often, a steep hyperbolic or exponential terminal decline follows attempts to remedy a well’s deteriorating performance. This observation casts doubt on the validity of the common practice of “group curve fitting” used to predict EUR from early production in newer shale plays.

I decided to examine the validity of the other common technique used in new play evaluation: prediction of EUR from IP. I projected decline curves for all the horizontal wells in the Barnett, and the resulting cross-plot of IP versus EUR provided a broad range of EURs that might be associated with a particular IP. For example, the well with the highest EUR in the Barnett Shale (8.8 Bcf) has a good correlation with IP (7.94 MMcfd). The next-best well has a EUR of 8.6 Bcf and a poor correlation to an IP of 4.28 MMcfd, while the fourth-best well has a EUR of 7.1 Bcf and an even poorer correlation to the IP of 1.9 MMcfd. In the end, I would prefer a high IP to a low one but, since approximately half of the EUR is produced in the first year, a well’s early decline rate is more important than IP in predicting reserves.

Operators often state that shale plays have about a 30 to 40-year production life, but I found that the average commercial life for horizontal wells is about 7.5 years, although the mode is four years. There are many wells that should have 8-12 years of production but few that will extend beyond 15 years. About 75 percent of predicted EUR in horizontal Barnett wells has been produced by Year 5. In the control group, the first wells were drilled in 2003, and already 15% have reached their economic limit five to six years into their production life cycle.

The average EUR for all horizontal Barnett wells is 0.81 Bcf (the mode is 0.5 Bcf/well). This is about one-third of the 2.5 Bcf/well average predicted by many operators. My decline projections indicate that only about 300 horizontal wells in the play (4% of total) will reach or exceed a 2.5 Bcf threshold. This seems consistent with the average to-date cumulative production of 0.46 Bcf/well.

The variance between reserves that I calculate and those claimed by operators in the Barnett Shale is because of differences in approach. Most operators project at least 40 years of production for their wells. I project to an economic limit of 2,000 Mcf per month because this is the threshold below which cost exceeds revenue based on $3.50/Mcf netback gas price, a 25% royalty, and average operating costs from operator 10-K SEC filings. Operators use a terminal decline rate of 4-8% even though there is no evidence for these rates from any shale well. I use 15% because this is corresponds to the decline rate for Barnett Shale wells with the longest production history. Operators typically use a “type curve” model to compute reserves that is a mathematically approximation of average decline rates. I projected each of the 1,977 wells in my control group individually using standard hyperbolic rate vs. time graphic methods, and used these results as a model to project reserves for wells outside the control group.

Since I averaged EURs over an area that comprises almost 15,000 square miles (9.5 million acres), I focused on the core areas or “sweet spots” centered in Tarrant and Johnson counties. While these areas include many of the best wells in the play, the average EUR is not much better than the overall play average. The average EUR for wells in the Tarrant County core area is 0.95 Bcf, and 0.84 Bcf in the Johnson County core area.

The US Geological Survey estimates technically recoverable Barnett gas resources of 26 Tcf, and many operators believe that this is too low. My calculations suggest that the Barnett EUR, based on 11,817 horizontal and vertical wells, will be about 8.8 Tcf. An additional 23,000 wells are required to reach 26 Tcf, at a cost of more than $75 billion for leasing, drilling, and completion alone.

Another surprise is that horizontal wells do not have significantly higher EUR (0.81 Bcf) compared to vertical wells (0.62 Bcf). Horizontal completions only result in a 31% improvement in reserves for about 2.5 times the cost. Put another way, the nominal unit cost (leasing, drilling and completion costs only) of gas from a horizontal well is approximately $4.30/Mcf compared to $2.05/Mcf from a vertical well.

A final surprise is that well performance in the Barnett Shale has not improved over time because of new completion technologies or better knowledge about drill site selection and reservoir distribution, as many people assume. For horizontal wells, the peak average EUR of 1.14 Bcf/well occurred in 2003. Average EUR has declined every year since down to 0.59 Bcf/well in 2008.

When I state concerns about the economics of shale plays in the current low-price gas environment (World Oil, January 2009), operators dismiss them because they are hedged at higher gas prices. If every operator in the Barnett Shale was hedged at a netback gas price of $8/Mcf, only 31% of horizontal wells would break even or make money. At $6/Mcf, only 15% of wells would reach this commercial threshold.

I am disturbed that public companies and investment analysts make fantastic claims about the rates and reserves for new shale plays without calibrating them to the only play that has significant production history. Almost every assumption used by the industry to support predictions about the Haynesville or Marcellus Shale plays is questionable based on well performance in the Barnett Shale. While it is true that every play is different and the Barnett Shale does not perfectly predict what will happen in other plays, it seems reasonable to temper and calibrate our uncertainty with what is known. There are many lessons we can learn from the Barnett Shale, and they all suggest a cautious approach to developing new shale plays.

Data used above were provided courtesy of IHS Inc. However, the analysis and opinions expressed here are solely those of the author and do not represent those of IHS or any other organization.