Transmission seismic imaging of the subsurface using simulated evolution techniques

Abstract

Transmission seismic imaging, popularly known as tomography, provides a means of estimation of velocity distribution in the subsurface by inverting P or S-wave data, from either time of arrival or full waveform. The accuracy and resolution of the tomographic reconstruction depends both on the data acquisition geometry and the inversion algorithm used. In many practical applications such as cross-borehole tomography, a limited range of viewing angles and ray density coverage results in nonunique solutions. The number of raypaths and sampling density differ enormously for various acquisition geometries used to record seismic arrivals. Individually, crosshole or vertical seismic profiling (VSP) data suffer from incompleteness. But a combined crosshole-VSP data interpretation can overcome this limitation, as explored in the present paper. Since finding the optimal solution, even from poor initial models, is extremely important in subsurface imaging, global optimisation using simulated evolution for the traveltime inversion is proposed in this paper. Genetic algorithm and evolutionary programming, the two main avenues of simulated evolution, are implemented for the tomographic applications. The potential of these methods is demonstrated on a number of numerical examples of which the results of an ?L?-shaped anomalous zone in a low velocity medium and two voids embedded in a high velocity host rock are presented here. The velocity tomograms of the first model present a somewhat distorted image evolved from VSP and crosshole data using the genetic algorithm approach. A combined crosshole-VSP traveltime inversion by both genetic algorithm and evolutionary programming gave solutions close to the actual models, the resolution by genetic algorithm being better than that of evolutionary programming. Similar results are also observed for the other model from crosshole-VSP combined interpretation. Shear-wave seismic tomographic surveys were conducted at North Searsole colliery, Raniganj Coalfield, India to detect abandoned galleries in coal seams for the purpose of stabilisation. The proposed simulated evolution tomographic imaging schemes have been applied for the traveltime inversion of VSP, crosshole and crosshole-VSP combined data to estimate S-wave velocity distribution in the subsurface region between the boreholes NS-002 and NS-003. It is found that S-wave velocity variation ranges between 0.1 ? 1.6 km/s by both genetic algorithm and evolutionary programming schemes when crosshole-VSP data are interpreted together, and between 0.4 ? 1.6 km/s by VSP and 0.5 ? 1.6 km/s by crosshole tomography. Thus individual tomographic imaging (crosshole/VSP) underestimated S-wave velocity variation in the region. The low shear-wave velocity zones estimated as 0.12 km/s by genetic algorithm and 0.32 km/s by evolutionary programming for crosshole-VSP data could be detected at a depth of 30 ? 36 m in the Kenda Bottom coal seam. As validated by the existing mine plan, these low velocity zones are the probable galleries in the coal seam. Traces of the Kenda Top and the Kenda Bottom coal seams could also be delineated by these schemes. Although the extreme heterogeneity associated with these coal mine workings poses a challenge to the accurate image reconstruction, the simulated evolution algorithms when implemented for the combined crosshole-VSP data inversion yield better subsurface resolution.