3D Tomography using Efficient Wavefront Picking of Traveltimes Abdullah AlTheyab and G. T. Schuster King Abdullah University of Science and Technology (KAUST) 1
Outline Introduction Areal Picking 3D Tomography using Areal Picks Conclusion 2
Introduction For conventional acquisition geometry, receiver lines are sparse. Picking is done on time- offset sections. first-arrivals x y t 3
Field Data Example 4 3D OBS data parameters: – 234 OBS stations – 129 source-lines – 50m inline spacing – 400m OBS spacing – 40-50m water depth Source boat sail lines Receiver stations
Human picking time 30,186 sections to pick, each with 360 receivers. Estimated picking time: – 2 section/minute → 251hrs – 8 hr/day: 31 days 5 12 km 0 3 Time [sec] CRG Shingling Low SNR
Quality Control and Cycle-skipping 12 km 0 3 Time [sec] Shingling Traveltime [sec] Traveltime Map 14 2 Distance [km] 1 3 Shot 73Shot 74Shot 75 6
Memory Footprint The size of the data is 80 GB (at 4ms sampling, after windowing). Interactive picking software require: – Large memory, – Swapping to hard drives. Memory access pattern for QC is complex. 7
Conventional Picking Approach Disadvantages: 1.Large human piking time (31 days) 2.Laborious to QC and correct picks 3.Large memory footprint (80 GB) 8
Outline Introduction Areal Picking 3D Tomography using Areal Picks Conclusion 9
Areal picking For conventional acquisition geometry, receiver lines are sparse. Picking is done on time- offset sections. first-arrivals x y t 10
x y t Areal picking For dense-receiver acquisition geometry We propose picking on time- slices (Areal Picking). 11
Areal picking For dense-receiver acquisition geometry We propose picking on time- slices (Areal Picking). y t x 12
Areal picking For dense-receiver acquisition geometry We propose picking on time- slices (Areal Picking). y t x 13
Areal picking For dense-receiver acquisition geometry We propose picking on time- slices (Areal Picking). y t x 14
Areal picking: Interpolation We implemented a program that does real-time interpolation. 15 Cartesian picks Polar interpolation Continuous Polygon Picks are interpolated in polar-coordinates.
Field Data Example 2 4 y[km] 14 x [km] 19 4 Time 0.8 sec 16
Field Data Example 2 4 y[km] 14 x [km] 19 4 Time 0.8 sec 17
Field Data Example y[km] 14 x [km] 19 4 Time 2.4 sec 18
Field Data Example y[km] 14 x [km] 19 4 Time 2.4 sec 19
Field Data Example: Human picking time ms time-slice spacing for 5 Hz FWI. 234 shots x 15 slices/shot= 3,510 slices (vs. 30,186 sections) to pick. Estimated picking-time: slices/minute: 30 hrs hr/day: 4 days (vs. 31 days)
Field Data Example: Quality Control Polygon must not cross. y[km] 14 x [km] 19 4 Time 2.4 sec 21
Field Data Example: Quality Control Min Apparent velocity Max 22 Detect mispicks. Apparent Velocity Map Explore regional trend
Field Data Example: Memory footprint 80 GB → Slicing for 5Hz FWI → 2 GB Slices are spaced at ½ of the shortest period. 23
Outline Introduction Areal Picking 3D Tomography using Areal Picks Conclusion 24
Polygon resampling y[km] 14 x [km] 19 4 Picked Traveltime Map Regularized Traveltime Tomography 0 3 Traveltime [sec] 25
Tomography using Areal Picks y[km] 14 x [km] 19 4 Picked Traveltime Residual Regularized Traveltime Tomography Residuals [sec] 26
Tomography using Areal Picks Cycle skipping Count Traveltime Error [sec] Traveltime Error Histogram 27
Final Traveltime Tomogram depth slice x [km] y [km] 10 inline xline 0 z [km] y [km] Velocity [m/s] 018 Structural cross-section
Field Data Example: Waveform Comparison km 0 3 Time [sec] Observed
Field Data Example: Waveform Comparison km 0 3 Time [sec] Calculated
Field Data Example: Waveform Comparison km 0 3 Time [sec] Observed
Field Data Example: Waveform Comparison km 0 3 Time [sec] Calculated
Field Data Example: Waveform Comparison km 0 3 Time [sec] Observed
Field Data Example: Waveform comparison km 0 3 Time [sec] Calculated
Field Data Example: Waveform Comparison km 0 3 Time [sec] Observed
Field Data Example: Waveform Comparison km 0 3 Time [sec] Calculated
Outline Introduction Areal Picking 3D Tomography using Areal Picks Conclusion 37
Conclusions Areal picking allows for building 3D tomograms in reasonable time. Advantages of areal picking: – About 70-90% reduction in human picking time (31 vs. 4 days) – Easier QC and correct mispicks – Much lower memory footprint (80 GB vs. 2 GB) 38
Thank you Acknowledgments: Pemex for providing the data. Sponsors of CSIM Saudi Aramco for supporting the FWI project. Research Computing at KAUST. 39