Using a local Monte Carlo strategy to assess 1-D velocity models from wide-angle seismic travel-time data and application to the Rockall Trough

We present a statistically based search strategy to explore velocity–depth model space derived from the inversion of seismic refraction and wide-angle data. The method is based on the Metropolis algorithm and computes the likelihood of any given model of fitting the observed data. By iteratively perturbing the model, the model space is sampled and the resulting probability density function provides a quantified measure of the velocity resolution as a function of depth. Unlike manual analysis, where a single layer is perturbed to test its sensitivity ignoring the effect on the deeper layers, this method computes the fit for the whole model at each iteration and only selects the models that achieve a specified global fit. We show results of the algorithm from the 1-D inversion of Expanding Spread Profiles from the central part of the Rockall Trough to the west of Britain. As expected, the method highlights the areas of the model that are both well and poorly constrained and shows the degradation of resolution with increasing depth such that for high quality data the accumulated velocity errors at basement depths is +/− 5%. This error increases as the data quality decreases and the estimated pick error increases.