Unconventional Resources in US: Potential & Lessons Learned

Size: px

Start display at page:

Download "Unconventional Resources in US: Potential & Lessons Learned"

Justina Hawkins
5 years ago
Views:

1 Unconventional Resources in US: Potential & Lessons Learned Looking at Barnett Shale from top of Barnett Pass, British Columbia, Photo by John McCall Tad Patzek, Petroleum & Geosystems Engineering, UT Austin 22 nd International Conference Oil-Gas AGH 2, Cracow, June 9

2 Acknowledgement This research has been supported in part by a grant from the Sloan Foundation to the Bureau of Economic Geology at the University of Texas in Austin and Rice University in Houston, Texas. p./73

3 Summary of Conclusions Gas production from the Barnett shale follows closely a multi-hubbert curve model (Patzek, 27, 28, 29), (Patzek & Croft, 2) The post-28 (right-most) Hubbert curve is very steep and its area (cumulative gas produced) is small; not a good sign With over 4 thousand Barnett wells and up 6 years of production in some wells, there seems to be enough data to draw quantitative conclusions about current and future production of shale gas There is a problem, however, almost no one knows how to draw the conclusions; only rough sketching comes to mind p.2/73

4 Multi-scale shales Classical depositional models will not work: We are cursed with 6 orders of magnitude of grain sizes and irregular grain shapes Source: André Kempe et al., PNAS, 99(4), 97, 22 p.3/73

5 Gas Rate from Barnett Shale EJ/Year Source: Texas Railroad Commission. EJ Tcf. p.4/73

6 Cum Gas from Barnett Shale EJ Source: Texas Railroad Commission. EJ Tcf. p.5/73

7 A High Gas Rate Scenario 3 Data High EJ/year Source: Texas Railroad Commission. A 26 Tcf scenario. p.6/73

8 High Cum Gas from Barnett Shale 3 Data High 25 2 EJ Source: Texas Railroad Commission. EJ Tcf. p.7/73

9 Disclaimer... I have used a constant gas price of $4./Mcf Therefore my calculations should be used only for planning purposes and do not reflect the historical profits realized over the last 6-7 years from historical wells Work to do: Evaluate wells the average annual gas price starting from when a well was drilled Everything I will say, is equally applicable to condensate- and oil-producing shales p.8/73

10 More Conclusions... Barnett shale is very heterogeneous areally and gas production is highly dependent on time Cumulative gas production by well in a given county is approximately lognormally distributed for fixed production times (,2, or 3 years in this presentation) One can therefore calculate the expected values of cumulative gas production by well after, 2, or 3 years, the median values, and the most probable (mode) values According to the results, no mean, median, or most probable well in the Barnett shale is profitable for three years, if its drilling cost is $3 million, and the gas price is close to $4/Mcf p.9/73

11 More Conclusions... The shale heterogeneity and different completion and production strategies result in the standard deviations from the expected production values (means) that are of the same order as the means Therefore, it is difficult to predict a priori the probability of a desired level of production in a given geographic area One can run Monte Carlo simulations and calculate the probabilities of exceeding a certain cumulative production after, 2, or 3 years (only the first-year results are shown here) The Monte Carlo simulations reveal that high-productivity wells are possible, but highly unlikely p./73

12 More Conclusions... According to the Monte Carlo simulations, chances of drilling and completing a Barnett shale well that will pay for itself in year are about: /8 in Tarrant county / in Johnson county in all other Barnett shale counties A novel and potentially ground-breaking approach to estimating gas shale production is sketched; it still needs further research and development p./73

13 Map of Barnett Shale in Texas Clay Montague Cooke Jack Wise Denton Northing, 5 ft Palo Eastland Erath Parker Hood Somervell Tarrant Johnson Dallas Ellis Bosque Hill 35.2 Comanche Hamilton Easting, 5 ft Source: HPDIP, from Dr. Peter Valko, Feb 3, 2 p.2/73

14 Barnett wells in Tarrant county Northing, 5 ft Easting, 5 ft p.3/73

15 First year gas cums in Tarrant county Zoom on the highest well density p.4/73

16 Barnett wells in Johnson county Northing, 5 ft Easting, 5 ft p.5/73

17 First year gas cums in Johnson county Zoom on the highest well density p.6/73

18 Barnett wells in Denton county Northing, 5 ft Easting, 5 ft p.7/73

19 First year gas cums in Denton county Zoom on the highest well density p.8/73

20 Barnett wells in Wise county Northing, 5 ft Easting, 5 ft p.9/73

21 First year gas cums in Wise county Zoom on the highest well density p.2/73

22 The Lognormal Distribution In a log-normal distribution of a random variable X > that has outcomes {x}, µ is the mean and σ standard deviation, respectively, of ln x (the variable s logarithm is normally distributed). In cartesian coordinates, µ and σ are the location and scale parameters, respectively [ ] pdf(x;µ,σ) = f(x;µ,σ) = x exp (lnx µ)2 2πσ 2 2σ 2 where pdf = f is the probability density function The distribution mode (pdf s peak value) is Mode = exp(µ σ 2 ) p.2/73

23 The Lognormal Distribution, cntd. Median is defined as Median(µ, σ) = m f(x;µ,σ) dx = m f(x;µ,σ) dx = 2 The lognormal distribution s median is equal to Median(µ) = exp(µ) The most important mean or expected value of the lognormal distribution is ] E(X) = m = xf(x;µ,σ) dx = exp [µ+ σ2 2 p.22/73

24 The Lognormal Distribution, cntd. The variance of the lognormal distribution is V(X) = x 2 f(x;µ,σ) dx = [ exp ( σ 2) ] exp ( 2µ+σ 2) The standard deviation is s = V Note that µ and σ describe the normal distribution of lnx, while m and s the lognormal distribution of x p.23/73

25 Generalized Extreme Value (GEV) The novel GEV approach is based on the observation that gas production from a shale is an extreme event that corresponds to a certain level of changing the shale structure and connecting to it The GEV fits of county-wide Barnett shale data are almost perfect, much better than the lognormal fits (here I only show the nine largest counties after 3 years) With better fits, I can obtain tighter bounds on the uncertainties of production volumes, i.e., narrower 95% confidence intervals p.24/73

26 Tarrant county after. yrs Experimental pdf GEV Lognormal.4 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.657, µ =.377, σ =.275 p.25/73

27 MLE of GEV pdf in Tarrant county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after. yrs p.26/73

28 Tarrant county after. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.657, µ =.377, σ =.275 p.27/73

29 Tarrant county after 2. yrs.4.2 Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.883, µ =.4639, σ =.335 p.28/73

30 MLE of GEV pdf in Tarrant county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 2. yrs p.29/73

31 Tarrant county after 2. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.883, µ =.4639, σ =.335 p.3/73

32 Tarrant county after 3. yrs.4.2 Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.793, µ =.548, σ =.3453 p.3/73

33 MLE of GEV pdf in Tarrant county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 3. yrs p.32/73

34 Tarrant county after 3. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.793, µ =.548, σ =.3453 p.33/73

35 Johnson county after. yrs Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.298, µ =.386, σ =.834 p.34/73

36 MLE of GEV pdf in Johnson county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after. yrs p.35/73

37 Johnson county after. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.298, µ =.386, σ =.834 p.36/73

38 Johnson county after 2. yrs.6.4 Experimental pdf GEV Lognormal.2 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.568, µ =.4429, σ =.2599 p.37/73

39 MLE of GEV pdf in Johnson county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 2. yrs p.38/73

40 Johnson county after 2. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.568, µ =.4429, σ =.2599 p.39/73

41 Johnson county after 3. yrs.4.2 Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.82, µ =.5449, σ =.3228 p.4/73

42 MLE of GEV pdf in Johnson county.4.35 σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 3. yrs p.4/73

43 Johnson county after 3. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.82, µ =.5449, σ =.3228 p.42/73

44 Denton county after. yrs Experimental pdf GEV Lognormal 3.5 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.2427, µ =.429, σ =.938 p.43/73

45 MLE of GEV pdf in Denton county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after. yrs p.44/73

46 Denton county after. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.2427, µ =.429, σ =.938 p.45/73

47 Denton county after 2. yrs Experimental pdf GEV Lognormal 2.5 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.2789, µ =.249, σ =.327 p.46/73

48 MLE of GEV pdf in Denton county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 2. yrs p.47/73

49 Denton county after 2. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.2789, µ =.249, σ =.327 p.48/73

50 Denton county after 3. yrs Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.2758, µ =.2427, σ =.53 p.49/73

51 MLE of GEV pdf in Denton county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 3. yrs p.5/73

52 Denton county after 3. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.2758, µ =.2427, σ =.53 p.5/73

53 Wise county after. yrs Experimental pdf GEV Lognormal 3.5 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.3, µ =.48, σ =.26 p.52/73

54 MLE of GEV pdf in Wise county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after. yrs p.53/73

55 Wise county after. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.3, µ =.48, σ =.26 p.54/73

56 Wise county after 2. yrs Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.595, µ =.232, σ =.56 p.55/73

57 MLE of GEV pdf in Wise county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 2. yrs p.56/73

58 Wise county after 2. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.595, µ =.232, σ =.56 p.57/73

59 Wise county after 3. yrs Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.78, µ =.2828, σ =.866 p.58/73

60 MLE of GEV pdf in Wise county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 3. yrs p.59/73

61 Wise county after 3. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.78, µ =.2828, σ =.866 p.6/73

62 Parker county after. yrs Experimental pdf GEV Lognormal 3.5 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.278, µ =.337, σ =. p.6/73

63 MLE of GEV pdf in Parker county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after. yrs p.62/73

64 Parker county after. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.278, µ =.337, σ =. p.63/73

65 Parker county after 2. yrs Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.54, µ =.254, σ =.554 p.64/73

66 MLE of GEV pdf in Parker county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 2. yrs p.65/73

67 Parker county after 2. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.54, µ =.254, σ =.554 p.66/73

68 Parker county after 3. yrs Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =., µ =.2797, σ =.27 p.67/73

69 MLE of GEV pdf in Parker county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 3. yrs p.68/73

70 Parker county after 3. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =., µ =.2797, σ =.27 p.69/73

71 Hood county after.97 yrs 6 5 Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =-.225, µ =.599, σ =.878 p.7/73

72 MLE of GEV pdf in Hood county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after.97 yrs p.7/73

73 Hood county after.97 yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =-.225, µ =.599, σ =.878 p.72/73

74 Hood county after.94 yrs Experimental pdf GEV Lognormal 2.5 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =-.795, µ =.265, σ =.276 p.73/73

75 MLE of GEV pdf in Hood county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after.94 yrs p.74/73

76 Hood county after.94 yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =-.795, µ =.265, σ =.276 p.75/73

77 Hood county after 2.92 yrs Experimental pdf GEV Lognormal 2.5 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =-.2524, µ =.358, σ =.544 p.76/73

78 MLE of GEV pdf in Hood county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 2.92 yrs p.77/73

79 Hood county after 2.92 yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =-.2524, µ =.358, σ =.544 p.78/73

80 Hill county after.83 yrs 6 5 Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.48, µ =.625, σ =.888 p.79/73

81 MLE of GEV pdf in Hill county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after.83 yrs p.8/73

82 Hill county after.83 yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.48, µ =.625, σ =.888 p.8/73

83 Hill county after.67 yrs Experimental pdf GEV Lognormal 3 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =., µ =.254, σ =.292 p.82/73

84 MLE of GEV pdf in Hill county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after.67 yrs p.83/73

85 Hill county after.67 yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =., µ =.254, σ =.292 p.84/73

86 Hill county after 2.5 yrs Experimental pdf GEV Lognormal 3.5 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.22, µ =.2925, σ =.356 p.85/73

87 MLE of GEV pdf in Hill county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 2.5 yrs p.86/73

88 Hill county after 2.5 yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.22, µ =.2925, σ =.356 p.87/73

89 Erath county after.78 yrs 2 Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =-.57, µ =.739, σ =.48 p.88/73

90 MLE of GEV pdf in Erath county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after.78 yrs p.89/73

91 Erath county after.78 yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =-.57, µ =.739, σ =.48 p.9/73

92 Erath county after.56 yrs 9 8 Experimental pdf GEV Lognormal 7 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =-.523, µ =.92, σ =.72 p.9/73

93 MLE of GEV pdf in Erath county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after.56 yrs p.92/73

94 Erath county after.56 yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =-.523, µ =.92, σ =.72 p.93/73

95 Erath county after 2.33 yrs 7 6 Experimental pdf GEV Lognormal 5 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =-.34, µ =.46, σ =.86 p.94/73

96 MLE of GEV pdf in Erath county.2.. σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 2.33 yrs p.95/73

97 Erath county after 2.33 yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =-.34, µ =.46, σ =.86 p.96/73

98 Jack county after. yrs 4 2 Experimental pdf GEV Lognormal Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.49, µ =.669, σ =.54 p.97/73

99 MLE of GEV pdf in Jack county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after. yrs p.98/73

100 Jack county after. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.49, µ =.669, σ =.54 p.99/73

101 Jack county after 2. yrs 9 8 Experimental pdf GEV Lognormal 7 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.6, µ =.24, σ =.834 p./73

102 MLE of GEV pdf in Jack county...9 σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 2. yrs p./73

103 Jack county after 2. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.6, µ =.24, σ =.834 p.2/73

104 Jack county after 3. yrs Experimental pdf GEV Lognormal 3.5 Probability Density Block Extremum, Bcf/well year GEV pdf: ξ =.97, µ =.483, σ =.52 p.3/73

105 MLE of GEV pdf in Jack county σ µ MLE = Maximum Likelihood Estimate, 95% CI for µ and σ after 3. yrs p.4/73

106 Jack county after 3. yrs Cumulative Probability Fitted Generalized Extreme Value CDF Empirical CDF R 95% CI Block Extremum, Bcf/well year GEV cdf: ξ =.97, µ =.483, σ =.52 p.5/73

107 Cumulative gas in Tarrant county Cumulative gas production, Bcf GEV µ ± σ Fit Rescaled time on production p.6/73

108 Cumulative gas in Johnson county Cumulative gas production, Bcf GEV µ ± σ Fit Rescaled time on production p.7/73

109 Cumulative gas in Denton county Cumulative gas production, Bcf GEV µ ± σ Fit Rescaled time on production p.8/73

110 Cumulative gas in Wise county Cumulative gas production, Bcf GEV µ ± σ Fit Rescaled time on production p.9/73

111 Cumulative gas in Parker county Cumulative gas production, Bcf GEV µ ± σ Fit Rescaled time on production p./73

112 Cumulative gas in Hood county Cumulative gas production, Bcf GEV µ ± σ Fit Rescaled time on production p./73

113 Cumulative gas in Hill county Cumulative gas production, Bcf GEV µ ± σ Fit Rescaled time on production p.2/73

114 Cumulative gas in Erath county Cumulative gas production, Bcf GEV µ ± σ Fit Rescaled time on production p.3/73

115 Cumulative gas in Jack county Cumulative gas production, Bcf GEV µ ± σ Fit Rescaled time on production p.4/73

116 Revenue in Tarrant Co. after. yrs.5.45 Random sampling of pdf Sample based GEV pdf.4.35 Probability Density Revenue, Million k=.92, µ =.286, σ =.85 p.5/73

117 Revenue in Tarrant Co. after. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.92, µ =.286, σ =.85 p.6/73

118 Revenue in Johnson Co. after. yrs.7.6 Random sampling of pdf Sample based GEV pdf.5 Probability Density Revenue, Million k=.39, µ =.2542, σ =.7227 p.7/73

119 Revenue in Johnson Co. after. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.39, µ =.2542, σ =.7227 p.8/73

120 Revenue in Denton Co. after. yrs.4.2 Random sampling of pdf Sample based GEV pdf Probability Density Revenue, Million k=.2499, µ =.568, σ =.375 p.9/73

121 Revenue in Denton Co. after. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.2499, µ =.568, σ =.375 p.2/73

122 Revenue in Wise Co. after. yrs.9 Random sampling of pdf Sample based GEV pdf.8.7 Probability Density Revenue, Million k=.573, µ =.5956, σ =.3989 p.2/73

123 Revenue in Wise Co. after. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.573, µ =.5956, σ =.3989 p.22/73

124 Revenue in Parker Co. after. yrs.9 Random sampling of pdf Sample based GEV pdf.8.7 Probability Density Revenue, Million k=.77, µ =.555, σ =.3897 p.23/73

125 Revenue in Parker Co. after. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.77, µ =.555, σ =.3897 p.24/73

126 Revenue in Hood Co. after.97 yrs.4.2 Random sampling of pdf Sample based GEV pdf Probability Density Revenue, Million k=-.994, µ =.6375, σ =.3349 p.25/73

127 Revenue in Hood Co. after.97 yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=-.994, µ =.6375, σ =.3349 p.26/73

128 Revenue in Hill Co. after.83 yrs.4.2 Random sampling of pdf Sample based GEV pdf Probability Density Revenue, Million k=.8, µ =.656, σ =.357 p.27/73

129 Revenue in Hill Co. after.83 yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.8, µ =.656, σ =.357 p.28/73

130 Revenue in Erath Co. after.78 yrs 2.5 Random sampling of pdf Sample based GEV pdf 2 Probability Density Revenue, Million k=.22, µ =.299, σ =.82 p.29/73

131 Revenue in Erath Co. after.78 yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.22, µ =.299, σ =.82 p.3/73

132 Revenue in Jack Co. after. yrs 2.8 Random sampling of pdf Sample based GEV pdf.6.4 Probability Density Revenue, Million k=.248, µ =.2754, σ =.23 p.3/73

133 Revenue in Jack Co. after. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.248, µ =.2754, σ =.23 p.32/73

134 Revenue in Tarrant Co. after 2. yrs.35.3 Random sampling of pdf Sample based GEV pdf.25 Probability Density Revenue, Million k=.58, µ =.8553, σ =.774 p.33/73

135 Revenue in Tarrant Co. after 2. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.58, µ =.8553, σ =.774 p.34/73

136 Revenue in Johnson Co. after 2. yrs.4.35 Random sampling of pdf Sample based GEV pdf.3 Probability Density Revenue, Million k=.88, µ =.7689, σ =.2 p.35/73

137 Revenue in Johnson Co. after 2. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.88, µ =.7689, σ =.2 p.36/73

138 Revenue in Denton Co. after 2. yrs.8.7 Random sampling of pdf Sample based GEV pdf.6 Probability Density Revenue, Million k=.2784, µ =.855, σ =.5246 p.37/73

139 Revenue in Denton Co. after 2. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.2784, µ =.855, σ =.5246 p.38/73

140 Revenue in Wise Co. after 2. yrs.7.6 Random sampling of pdf Sample based GEV pdf.5 Probability Density Revenue, Million k=.863, µ =.929, σ =.657 p.39/73

141 Revenue in Wise Co. after 2. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.863, µ =.929, σ =.657 p.4/73

142 Revenue in Parker Co. after 2. yrs.7.6 Random sampling of pdf Sample based GEV pdf.5 Probability Density Revenue, Million k=.427, µ =.87, σ =.5899 p.4/73

143 Revenue in Parker Co. after 2. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.427, µ =.87, σ =.5899 p.42/73

144 Revenue in Hood Co. after.94 yrs.8.7 Random sampling of pdf Sample based GEV pdf.6 Probability Density Revenue, Million k=-.696, µ =.469, σ =.53 p.43/73

145 Revenue in Hood Co. after.94 yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=-.696, µ =.469, σ =.53 p.44/73

146 Revenue in Hill Co. after.67 yrs.8.7 Random sampling of pdf Sample based GEV pdf.6 Probability Density Revenue, Million k=.9, µ =.76, σ =.55 p.45/73

147 Revenue in Hill Co. after.67 yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.9, µ =.76, σ =.55 p.46/73

148 Revenue in Erath Co. after.56 yrs.4.2 Random sampling of pdf Sample based GEV pdf Probability Density Revenue, Million k=-.95, µ =.479, σ =.278 p.47/73

149 Revenue in Erath Co. after.56 yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=-.95, µ =.479, σ =.278 p.48/73

150 Revenue in Jack Co. after 2. yrs.4.2 Random sampling of pdf Sample based GEV pdf Probability Density Revenue, Million k=.546, µ =.452, σ =.32 p.49/73

151 Revenue in Jack Co. after 2. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.546, µ =.452, σ =.32 p.5/73

152 Revenue in Tarrant Co. after 3. yrs.35.3 Random sampling of pdf Sample based GEV pdf.25 Probability Density Revenue, Million k=.249, µ =2.26, σ =.3452 p.5/73

153 Revenue in Tarrant Co. after 3. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.249, µ =2.26, σ =.3452 p.52/73

154 Revenue in Johnson Co. after 3. yrs.35.3 Random sampling of pdf Sample based GEV pdf.25 Probability Density Revenue, Million k=.33, µ =2.835, σ =.2562 p.53/73

155 Revenue in Johnson Co. after 3. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.33, µ =2.835, σ =.2562 p.54/73

156 Revenue in Denton Co. after 3. yrs.7.6 Random sampling of pdf Sample based GEV pdf.5 Probability Density Revenue, Million k=.2826, µ =.9674, σ =.599 p.55/73

157 Revenue in Denton Co. after 3. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.2826, µ =.9674, σ =.599 p.56/73

158 Revenue in Wise Co. after 3. yrs.7.6 Random sampling of pdf Sample based GEV pdf.5 Probability Density Revenue, Million k=.97, µ =.22, σ =.728 p.57/73

159 Revenue in Wise Co. after 3. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.97, µ =.22, σ =.728 p.58/73

160 Revenue in Parker Co. after 3. yrs.5.45 Random sampling of pdf Sample based GEV pdf.4.35 Probability Density Revenue, Million k=.326, µ =.376, σ =.7622 p.59/73

161 Revenue in Parker Co. after 3. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.326, µ =.376, σ =.7622 p.6/73

162 Revenue in Hood Co. after 2.92 yrs.7.6 Random sampling of pdf Sample based GEV pdf.5 Probability Density Revenue, Million k=-.2374, µ =.454, σ =.687 p.6/73

163 Revenue in Hood Co. after 2.92 yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=-.2374, µ =.454, σ =.687 p.62/73

164 Revenue in Hill Co. after 2.5 yrs.7.6 Random sampling of pdf Sample based GEV pdf.5 Probability Density Revenue, Million k=.52, µ =.767, σ =.5456 p.63/73

165 Revenue in Hill Co. after 2.5 yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.52, µ =.767, σ =.5456 p.64/73

166 Revenue in Erath Co. after 2.33 yrs.4.2 Random sampling of pdf Sample based GEV pdf Probability Density Revenue, Million k=.2, µ =.5765, σ =.3345 p.65/73

167 Revenue in Erath Co. after 2.33 yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.2, µ =.5765, σ =.3345 p.66/73

168 Revenue in Jack Co. after 3. yrs.9 Random sampling of pdf Sample based GEV pdf.8.7 Probability Density Revenue, Million k=.63, µ =.5887, σ =.3998 p.67/73

169 Revenue in Jack Co. after 3. yrs Cumulative Probability Fitted CDF Sample based CDF Revenue, Million k=.63, µ =.5887, σ =.3998 p.68/73

170 GEV statistics of cum gas: Means Tarrant Johnson Wise Parker Hood Denton Hill Jack Erath After period After period 2 After period Expected values of revenue, $ million per well Source: HPDIP, from Dr. Peter Valko, Feb 3, 2 p.69/73

171 GEV statistics of cum gas: SDs Tarrant Johnson Wise Parker Hood Denton Hill Jack Erath After period After period 2 After period Standard deviations of gas revenue, $ million per well Source: HPDIP, from Dr. Peter Valko, Feb 3, 2 p.7/73

172 Remedial Technology Suppose that you had a technology that would allow you to stop drilling and completing a Barnett well if its 3-year production would be below Bcf You would spend only $ million on that well You would charge this sunk cost to the project and drill another, better well for $3 millions Such strategy could limit your negative cash flow quite considerably at $4/Mcf p.7/73

173 Profit by county after 3 yrs Johnson Discard unprofitable wells Blind drilling Tarrant Parker Denton Wise Hill Hood Erath Profit from drilling $4./Mcf Source: HPDIP, from Dr. Peter Valko, Feb 3, 2 p.72/73

174 Conclusions Economic gas production from gas shales is difficult but possible, even at $4-5 per standard cubic feet of gas We still understand too little about the mechanisms of gas/condensate flow through shales, and fracture generation and connectivity Significant fundamental and applied research is needed to predict and improve the long-time performance of hydrofractured horizonal wells in shales Full water reuse and recycling will be necessary to allow for drilling unconventional gas wells in many parts of the U.S. p.73/73

Statistics & Flood Frequency Chapter 3. Dr. Philip B. Bedient

Statistics & Flood Frequency Chapter 3 Dr. Philip B. Bedient Predicting FLOODS Flood Frequency Analysis n Statistical Methods to evaluate probability exceeding a particular outcome - P (X >20,000 cfs)