Traveltime inversion for a 3-D near surface velocity model
MetadataShow full item record
The near surface environment is often the source of the most severe lateral velocity variations present in the seismic section. Near surface lateral velocity variations distort the traveltimes of deeper events and are the most serious limitation in achieving accurate structural maps. This work discusses the development of a near surface velocity model for a shallow marine data set. The near surface model consists of three components. The first is a model of the laterally variable seafloor depth and topography. Below the seafloor, the model consists of the compressional wave velocity as a function of depth which reaches a maximum depth of approximately 500 meters. The presence of vertical and lateral velocity gradients is recognized. Embedded within this slowly varying background velocity field are a number of local lens-like velocity anomalies. The lens anomalies represent the major lateral velocity variations present in the near surface. Autocorrelograms of the deeper pre-stack data are used to obtain the seafloor model. The period of the first water layer reverberation is used to estimate the water depth. These data are enhanced by a deconvolution algorithm which improves the agreement at the line intersections. Measured first arrival times from the pre-stack data are used to develop the subseafloor velocity model. A multichannel filter algorithm is devised to estimate the traveltime deviations produced by the lens anomalies and the common shot statics. These traveltime deviations are the higher spatial frequency components of the first arrival times and are produced by the higher spatial frequency components of the velocity model. The output from the algorithm consists of a sixteen layer traveltime (velocity) perturbation model. The estimates of the lens anomaly and shot static produced traveltime deviations are subtracted from the first arrival times to isolate the slowly varying background components. These data are then inverted using the Generalized Linear Inversion and Tausum algorithms to obtain the laterally varying background velocity model.