A Self-Adjustable Input Genetic Algorithm for the Near-Surface Problem in Geophysics

The method of exploration seismics aims at creating an image of the earth's subsurface structures by active acoustic reflection measurements. However, seismic images acquired from land data are often severely degraded due to complex propagation effects near the surface of the earth. Although some methods have been proposed to address the near-surface problem, it remains largely unsolved. We propose a solution that involves an estimation of the true wave propagation effects through the near-surface area in order to compensate for them without explicitly estimating a velocity-depth model. The estimated one-way propagation operators describe wave propagation between the surface level (i.e., the acquisition level) and a laterally consistent datum reflector level. They are parameterized by one-way travel times along a predefined lateral grid. Based on this solution we present a self-adjustable input genetic algorithm (SAIGA) to estimate these travel time functions. SAIGA is an advanced and scalable genetic algorithm that can overcome the hurdle of excessive calculation time due to large 3-D data volumes, as it optimizes the parameters on a representative subset that is randomly selected and periodically updated from the full input dataset. Finally, we apply SAIGA to a 3-D field dataset containing 2.4 million traces yielding good results within a reasonable calculation time.