A statistical learning approach for seismic QC efficiency

Abstract

Processing large marine seismic surveys can be challenging with regards to quality control (QC) efficiency because of the large amount of data that must be validated for each processing step (it is common to reach more than 10 6 shots for one survey). Statistical and machine learning related techniques are natural tools to tackle such a challenge.Using data from a 10,000 km² survey in the Espirito Santo basin, we implemented a statistical approach to detect problematic shots and a supervised learning approach based on logistic regression to detect the presence of swell noise with an accuracy of 100% in our validation set.