Using vehicle-induced noise as a local seismic source through ML and interferometry

In our research, we successfully demonstrated that vehicles moving along the seismic line can serve as moving seismic sources, providing valuable data for subsurface characterization. To achieve this, we implemented a machine learning algorithm capable of accurately identifying each time window that contains truck transit. This algorithm efficiently detected the presence of vehicles by analyzing the seismic recordings, allowing us to precisely track their positions along the seismic line. To generate virtual shot gathers, we employed seismic interferometry. Using the closest receiver to the source as a virtual source, we effectively simulated the presence of an impulsive source at the location of each identified vehicle. This enabled us to reconstruct the wavefield response as if it were generated by an active seismic source. By utilizing the phase shift method, we calculated dispersion images that captured the variation of surface wave velocities across the surveyed area. These dispersion images provide crucial information about the subsurface properties and allow us to infer the structural characteristics beneath the seismic line. Comparing the estimated DCs derived from the computed dispersion images of the vehicle-induced noise data with those obtained from active seismic sources, we established a consistent pattern. This finding indicated that the information obtained from the vehicle-induced noise data was reliable and comparable to that obtained through traditional active seismic surveys. This consistency demonstrated the feasibility and effectiveness of utilizing vehicle-induced noise as a valuable seismic source for subsurface characterization. Overall, our research successfully demonstrated that vehicles moving along the seismic line can be utilized as moving seismic sources, providing valuable insights into subsurface structures. Through the implementation of a machine learning algorithm for vehicle detection, the application of seismic interferometry to generate virtual shot gathers, and the computation of dispersion images, we established the consistency between vehicle-induced noise data and active seismic sources. These findings open up new possibilities for cost-effective and environmentally friendly subsurface characterization techniques.