How to transform the seismic time series to a graph?

Learning the spatial structure from seismic time series is significant to interpret the network of depositional reservoirs. As a network analysis tool, graph theory has been successfully applied in geoscience. However, it is difficult to directly use a popular transformation to map the seismic time series with peaks/troughs to a graph, which will be the basic for the structured deep learning. Thus, inspired by the family of visibility graph algorithms, we present a seismic top-bottom visibility graph that better represents the peak and trough of seismic record and the reflectivity sequence. Results on Ricker wavelets, a reflectivity sequence from the depositional cycle, a synthetic seismic record and real seismic traces from tight reservoirs, demonstrate that the proposed top-bottom visibility approach is potential to efficiently capture the graphical structures of reservoirs.