Paleokarst caves recognition from seismic response simulation to CNN detection

Abstract

Paleokarst systems, found in carbonate rock formations worldwide, have potential for creating vast reservoirs and facilitating hydrocarbon migration. Thus, studying these systems is essential for the exploration and development of carbonate reservoirs. Our proposed approach is to use a convolutional neural network (CNN) based method to automatically and precisely identify cave features within 3D seismic data. We present an efficient method to produce ample amounts of 3D training data, which is comprised of synthetic seismic data and labels for cave features contained in the seismic data, as a solution to bypass the labeling task for training the CNN. This workflow uses point-spread functions (PSFs) to simulate cave response in the seismic data and allows us to easily generate realistic and diverse synthetic training datasets with different geological structures and cave features. By training the CNN with these synthetic datasets, it can effectively learn to detect cave features in field seismic volumes. We have evaluated the effectiveness of our method using multiple examples and found that it performs more accurately than previous methods, including seismic attributes and other CNN-based paleokarst characterization methods.