Rapid 3D inversion of gravity and gravity gradient data to test geologic hypotheses

Transcript

1. Rapid 3D inversion of gravity and gravity gradient data to test geologic hypotheses Leonardo Uieda Valéria C. F. Barbosa Observatório Nacional Rio de Janeiro, Brazil GGHS2012 S7-044
2. Forward modeling
3. Observed Predicted Model
4. ✔ Control ✔ Prior information ✔ Speed ✗ Tedious ✗ Gravity + gradients ✗ 3D
5. Geophysical inversion
6. Inversion Observed data 3D model
7. Prior information Regularization: • Damping • Smoothness Inversion Observed data 3D model
8. ✔ Automatic fit ✔ Gravity + gradients ✔ 3D ✗ Control ✗ Prior information ✗ Speed
9. Source http://xkcd.com/747 Something in the middle
10. Planting anomalous densities Uieda and Barbosa (2012), Geophysics
11. Observed data Mesh
12. Zero density contrast
13. Seeds (user specified) Predicted data
14. Neighbors
15. The best New predicted data
16. The best New predicted data
25. Fit!
26. Seeds = Skeleton Source http://www.sciencebuzz.org
27. Inversion Body Source http://dinocrisis.wikia.com
28. ✔ Control ✔ Prior information ✔ Speed ✔ Automatic fit ✔ Gravity + Gradients ✔ 3D (only need skeleton) Seeds
29. Example applications
30. Redenção granite
31. Outcrop
32. Oliveira et al. (2008)
33. Seed Density contrast = -0.09 g.cm-3
35. Quadrilátero Ferrífero
36. Complex geology and topography
37. Complex geology and topography Iron ore deposits
38. Seeds
40. Hypothesis testing Registro do Araguaia intrusion
41. Registro do araguaia • Not outcropping • Alkaline intrusion • Density contrast ≈0.3 g.cm-3
42. After Dutra et al. (2012)
43. After Dutra et al. (2012)
44. 3D gravity inversion (Dutra et al., 2012) Bottom = 18 km
45. Hypothesis 1
46. 18 km
49. Hypothesis 1
50. Hypothesis 1 X
51. Hypothesis 2
52. 18 km 5 km
55. Hypothesis 2
56. Hypothesis 2 X
57. Hypothesis 3
58. 5 km
61. Outcropping
62. Hypothesis 3
63. Hypothesis 3 X
64. Hypothesis 4
65. 5 km
66. Density 0.5 g.cm-3
68. Not outcropping
69. Not outcropping
70. Hypothesis 4
71. Hypothesis 4 V
72. Conclusion • 3D forward modeling = difficult • Traditional 3D inversion = not flexible • Planting densities = fast + flexible