Loading...
Please wait, while we are loading the content...
Similar Documents
Interpretation of First‐Arrival Travel Times with Wavepath Eikonal Traveltime Inversion and Wavefront Refraction Method
| Content Provider | Semantic Scholar |
|---|---|
| Author | Rohdewald, Siegfried R. |
| Copyright Year | 2011 |
| Abstract | We describe blind interpretation of the synthetic traveltime dataset, made available by Colin Zelt for the SAGEEP 2011 “Seismic Refraction Shootout” session. A 1D initial model (with horizontal layering, parallel to smoothed topography) is obtained automatically from the traveltimes, without requiring the user to assign first breaks to assumed refractors. This initial model is then iteratively refined with WET (Wavepath Eikonal Traveltime inversion: forward model synthetic traveltimes with Eikonal solver, back-project misfit along wavepaths aka Fresnel volumes, in a SIRT-like algorithm). Alternatively, traveltimes are interpreted with WR (Wavefront Refraction, layer-based ray inversion) method. For WR, first breaks need to be assigned to refractors interactively. WR has problems with imaging faults, pinchouts, outcrops and other velocity anomalies, which violate the WR assumption of laterally continuous layers. And the assignment of first breaks to hypothetical and mathematically idealized refractors is subjective and non-unique. But WR is still useful to detect lateral change of velocity, independent of WET. Artefacts of the 1D initial model (horizontal layering in basement) are progressively removed, with increasing number of WET iterations. Fit of WET model to WR interpretation (fault in basement) improves with increasing iteration count, even after the RMS error stops decreasing. This demonstrates that using solely the RMS error as a criterion for determining the optimum number of WET iterations is unreliable, and may stop WET prematurely. We propose the following criteria, to determine the optimum number of WET iterations : I. explain traveltimes with smooth minimum-structure model, II. minimum correlation of final model with layering of initial model, III. reasonable fit with WR interpretation, and IV. small RMS error. Also, we match areas of low ray coverage to apparent lowvelocity zones in the sedimentary overburden and depressions in the bedrock surface. Introduction Wavepath Eikonal Traveltime Inversion WET inversion (Schuster, 1993) uses the Fresnel volume approach (Watanabe, 1999) to model propagation of first-break energy, in a physically meaningful way. While ray-tracing methods assume that the frequency of the source signal is infinite and model wave propagation along “thin rays”, WET partially models finite-frequency effects such as diffraction and scattering, using wavepaths aka Fresnel volumes or “fat rays” (Husen, 2001). For each source and receiver, WET forward models traveltimes to all grid nodes with an Eikonal solver (Lecomte, 2000) and back-projects traveltime residuals along wavepaths, in a SIRT-like algorithm. WET naturally smoothes the tomogram, but requires careful first break picking. Bad picks can result in artefacts (“engraving” of wavepaths in tomogram), especially in low-coverage situations (low ratio of shots to receivers). As shown with interpretation of synthetic data (Sheehan, 2005; Jansen, 2010), refraction tomography and WET (with a 1D initial model) work well in many situations where conventional layer-based refraction methods fail. Refraction tomography blurs sharp velocity contrasts, and images them with gradients. |
| Starting Page | 31 |
| Ending Page | 38 |
| Page Count | 8 |
| File Format | PDF HTM / HTML |
| DOI | 10.4133/1.3614086 |
| Alternate Webpage(s) | http://rayfract.com/pub/SAGEEP11.pdf |
| Alternate Webpage(s) | https://doi.org/10.4133/1.3614086 |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
