moment tensor inversion of induced seismicity under different station ... - mit erl moment... ·...

Moment tensor inversion of induced seismicity under different station configurations in oil/gas fields Chen Gu* and M. Nafi Toköz, Massachusetts Institute of Technology Summary Induced seismicity occurs both in conventional oil/gas fields due to production and water injection and in unconventional oil/gas fields due to hydraulic fracturing. Source mechanisms of these induced earthquakes are of great importance for understanding their causes and the physics of the seismic processes in reservoirs. Previous research on the analysis of induced seismic events assumed a double-couple (DC) source mechanism. However, recent studies have shown a non-negligible percentage of a non-double-couple (non-DC) component of source moment tensor in hydraulic fracturing events (Rutledge et al. 2003, Šílený et al., 2009; Warpinski and Du, 2010; Song and Toksöz, 2011). In this study, we determined the full moment tensor of a large number of earthquakes from an oil/gas field. The seismicity of the field and source mechanisms of events using DC assumption have been studied extensively (Sarkar, 2008; Li et al., 2011, 2012). Song and Toksöz (2011) developed a full waveform based complete moment tensor inversion method to investigate induced events. In this study, we inverted and compared the source mechanisms of induced events with and without DC constraints. We investigated the accuracy of moment tensors on number of stations, azimuthal distribution, and signal to noise ratio. We conducted tests with synthetic data to validate the method before the application to the real data. Our results show that the double-couple component of the MT is determined most accurately. The accuracy of isotropic and CLVD components depend on data quality and station coverage. Introduction Induced micro-earthquakes widely happen in conventional and unconventional oil/gas fields. Induced seismicity study is of great importance in monitoring and understanding the processes of hydraulic fracturing, fluid injection and oil/gas extraction. Previous research on the analysis of induced seismic events in conventional oil/gas fields assumed a double-couple (DC) source mechanism. However, recent studies have shown interests in a non-double-couple (non-DC) component of source moment tensor in hydraulic fracturing events (Šílený et al., 2009; Warpinski and Du, 2010; Song and Toksöz, 2011). In this study, we determined the full moment tensor of the induced seismicity from a conventional oil/gas field. The

seismicity of the field and source mechanisms of events using DC assumption have been studied extensively (Sarkar, 2008; Li et al., 2011, 2012). We used the approach developed by Song and Toksöz (2011) for full moment tensor inversion. We applied this full moment tensor inversion method to the selected induced events. We compared the full moment tensor results with the DC moment tensor results. Our results show a varying percentage of non-DC components in this conventional oil/gas field. Double-couple component of the MT is determined most accurately. The accuracy of isotropic and CLVD components depends on data quality and station coverage. Methodology and Data The data used are from an oil and gas field in Oman, studied extensively by Sarkar (2008) and Li et al. (2011). The events were located using NonLinLoc algorithm which utilizes a probabilistic, non-linear, global-search earthquake location algorithm (Lomax et al., 2000). The locations were validated using synthetic seismograms (Li et al., 2011). We determined the moment tensor of a sub-set of the induced earthquakes using both DC and full moment tensor models. We first construct a Green’s function library, and calculate the synthetic seismograms for a point moment tensor source using the discrete wavenumber integration method (Bouchon, 1981). The observed seismogram 𝑣! of the ith component at the nth geophone of location 𝑥!! is modeled by

𝑣! 𝑥!!, 𝑥!, 𝑡 = 𝑚!"𝐺!",! 𝑥!!, 𝑥!, 𝑡 ∗ 𝑠 𝑡!

!!!

!

!!!

, (1)

where 𝐺!",! 𝑥!!, 𝑥!, 𝑡 is the spatial derivative of the Green’s function at the nth geophone of the location 𝑥!! due to a point moment tensor source of the location 𝑥!, and 𝑠 𝑡 is the source time function. The best solution of DC moment tensors and source mechanisms was determined with the best waveform fitting using the criteria described by Li et al. (2011, 2012), and Song and Toksöz, (2011, 2013).

Source mechanism and DC moment tensor inversion

The seismicity data are from both surface and downhole monitoring networks (Figure 1). For the surface monitoring network, five surface stations, instrumented with SM-6B geophones, have been set up since 1999. For the downhole network, thirty-three SM-7m seismic sensors were instrumented in five boreholes ranging from depth 750m to 1250m in the year 2002. The data used in the studies consist of 800 events located by the surface network and 2000 events from the downhole network. Unfortunately there is little overlap between the two sets of events. This field is dominated by two fault systems – the northeast-southwest trending main faults and northwest-southeast trending auxiliary system. Synthetic tests To validate our method, we first applied our full waveform inversion method to the synthetic data. The configuration of the seismic source and stations was shown in Figure 1. The source mechanism was set to be purely double-couple, with strike of 305°, dip of 20°, and rake of 85°. A layer velocity model was used in the synthetic tests (Figure 1, lower right). A 5% Gaussian noise was added to the synthetic data. The full moment tensor inversion shows good recovery of the original source mechanisms. Recovered source parameters are: Strike = 306°, Dip = 21°, Rake = 87°, DC% = 96%, CLVD% = 3%, ISO = 1%.

Figure 1: Map view and side view of the stations and located events for both near-surface network and downhole network (Sarkar 2008). The red dots denote the location of the detected events, and the green triangles show the location of the stations. The black lines are the identified faults. Top: The green triangles (VA11, VA21, VA31, VA41, and VA51) are the five near-surface stations. These stations are located in shallow boreholes 150 m below the surface. Bottom: The green triangles are the thirty-three downhole stations. These downhole stations are located in five vertical wells (YA, YB, YC YD, and VE).

Figure 2: The configuration of source and stations of our synthetic data. Left: The cube on the left shows the locations of the source and the stations. The blue star denotes the source, the green triangles denote the stations. We use the red lines to mark stations with the same x, y coordinates. Right: The beach ball on the top right shows the mechanism of the synthetic source. The black triangles denote the T, P, N axes, and the green triangles denote the projection of the 5 stations on the source focal plane. The velocity model is shown at the bottom right.

0

2

4

6

8

10

12

VA11VA21

VA31 VA41

VA51

Sout

h to

Nor

th (k

m)

0 2 4 6 8 10 12

0

1

2

3

4West to East (km)

Dep

th (k

m)

0

2

4

6

8

10

12

YAYBYC

YDYE

Sout

h to

Nor

th (k

m)

0 2 4 6 8 10 12

0

1

2

3

4West to East (km)

Dep

th (k

m)

05

10

0

5

10

0

1

2

3

4

T1

West to East (km)

T4

M4T5

T2

T3

South to North (km)

Dept

h (k

m)

0 1000 2000 3000 4000 5000 6000 7000

0

500

1000

1500

2000

2500

3000

3500

4000Velocity (m/s)

Dept

h (m

)

VpVs

x = 8.18 km, y = 6.22 km, z = 1.14 km Strike = 305°, Dip = 20°, Rake = 85°DC%: 100, CLVD%: 0, ISO%: 0 T

N PT1

T2 T3

T4

T5


We also tested the effects of station configurations on the inversion results and compared with the original source parameters with the different station configurations. The synthetic tests show that the coverage of stations around the focal sphere, not the number of stations, determined the source mechanism recovery quality.

Results for field data We selected 10 events from the surface monitoring network and 10 events from the downhole monitoring network. We listed the source mechanism and DC moment tensor results in Tables 1 and 2 with the order of station coverage and average signal-to-noise ratio (SNR AVE.) for these selected events. For the seismic events with good signal-to-noise ratio and station coverage (e.g. 20010035, 20000129, yb0210230130), we found a good variance reduction, and a

DC dominant source mechanism, which is what we expected in a conventional oil/gas field. Most of the full moment tensor and DC moment tensor results give consistent focal plane directions, which are parallel to the two directions of the main fault system in this oil/gas field. Table 1: Full moment tensor and DC moment tensor for 10 events monitored by surface station network. Events are listed in order of qualities (S/N ratio).

Table 2: Full moment tensor and DC moment tensor for 10 events monitored by downhole station network. Events are listed in order of qualities (S/N ratio).

Conclusion Based on the synthetic simulation and the study of induced seismic events from an oil/gas field, we can state that reliable moment tensor inversion for small events requires seismic records with good signal-to-noise ratio, good azimuthal distribution and focal sphere coverage of stations, and a detailed velocity model. Accurate velocity models improve the solutions.

!!

Surface Events

Source Mechanism DC Moment Tensor SNR AVE.

STA CONFIG Beach

Ball Strike

(°) Dip (°)

Rake (°)

DC %

CLVD %

ISO %

VR %

Beach Ball

Strike (°)

Dip (°)

Rake (°)

VR %

20010035

135 232

63 83

351 208 74 5 21 57

135 231

63 81

350 207 56 93.0

20000129

81 198

88 5

95 27 98 1 2 44

80 192

88 5

95 22 44 66.2

20010057

122 228

71 75

160 23 63 17 20 26

127 226

72 63

152 20 22 65.5

20000061

41 310

37 86

4 127 66 4 30 42

68 330

87 16

172 75 39 38.4

20000003

118 213

67 67

157 23 63 6 31 52

118 217

68 68

156 24 47 23.6

20000126

265 336

82 80

353 185 36 25 39 39

252 342

88 79

349 182 31 21.6

20010143

62 192

7286

160 26 19 36 44 3

15 206

34 56

81 96 3 13.7

20010014

80 144

53 73

197 34 41 57 3 33

56 144

50 88

182 40 33 8.0

20010069

91 352

74 17

73 167 46 18 36 32

78 246

85 5

91 78 29 4.7

20000107

47 136

88 39

101 359 10 45 45 37

46 150

87 13

103 346 32 4.6

!

!!

Downhole Events

Source Mechanism DC Moment Tensor SNR AVE.

STA CONFIG Beach

Ball Strike

(°) Dip (°)

Rake (°)

DC%

CLVD%

ISO%

VR %

Beach Ball

Strike (°)

Dip (°)

Rake (°)

VR %

yb0212060053

320 249

90 83

7 182 52 33 14 46

352 134

21 73

126 77 33 101.0

yb0210230130

282 195

65 89

359 155 92 6 1 20

284 194

64 89

359 154 20 54.4

yb0211060116

343 234

89 63

29 171 66 33 0 15

342 173

80 10

88 101 14 48.4

yb0212060017

58 152

85 88

357 185 88 8 4 11

61 152

84 88

358 186 11 46.8

yb0210210072

140 275

54 89

148 46 30 52 18 14

183 272

60 89

1 150 14 168.9

yb0210210113

192 269

40 90

190 51 75 17 8 14

178 270

33 89

178 57 14 113.8

yb0210190234

223 302

63 80

356 205 51 22 28 19

290 322

70 23

258 300 18 51.0

yb0212140059

51 81

87 90

184 5 7 64 29 16

172 80

70 86

4 160 9 50.1

yb0302020013

267 61

89 85

347 193 6 64 30 9

340 81

58 73

340 214 4 79.9

yb0212080046

25 132

89 78

167 5 27 15 58 19

12 233

44 54

238 297 17 19.3

!!

Figure 3: Moment tensor inversion results using data from different station configurations. Different colors denote different number of stations used for inversion. The red, blue and black beach balls show the results using five, four, and three stations, respectively.

T1

T2 T3

T4

T5

T1

T2 T3

T4T1

T2 T3T5

T1

T2

T4

T5

T1

T3

T4

T5T2 T3

T4

T5

T1

T2 T3

T1

T2

T4

T1

T3

T4

T2 T3

T4 T1

T2T5

T1

T3

T4

T2T3

T5

T2

T4

T5

T1 T4

T5 T3

T4

T5

Strike: 304º DC: 100%Dip: 20º CLVD: 0%Rake: 84º ISO: 0%

















Double-couple component of the MT is determined most accurately. The accuracy of isotropic and CLVD components depends on data quality and station coverage. High frequency data from downhole sensors resolve detailed source properties, but the source mechanism determination requires detailed modeling. Acknowledgement The authors thank Petroleum Development Oman (PDO) for providing field data. This research was also supported by Aramco, Halliburton/Pinnacle, and Kuwait-MIT Center.

moment tensor inversion of induced seismicity under different station ... - mit erl moment... ·...

Documents