Due to advances in recording capabilities, seismic data for land oil-industry surveys is now often recorded continuously. A large portion of these data is usually discarded since it is not directly used for conventional imaging purposes. However, there is a growing consensus that these "surplus" recordings contain a variety of useful information about the subsurface, especially for academic seismologists but also for oil explorers. This paper presents an example of extraction of 3D information about the deep crust from such recordings. In this case the deep information was retrieved from a post-harvest but pre-correlation dataset by means of extended correlation. Although signal penetration in this particular area is limited by an extremely thick (>20,000 ft) overlying sedimentary cover, the exercise demonstrates the tremendous potential of systematically mining the rapidly expanding database of continuously recorded oil exploration data, and the clear need to preserve rather than discard the unconventional portions of those records.
This paper reports a pilot study of a large N (number of recording channels), vibroseis dataset acquired in southeast New Mexico in 2014 and provided to Cornell's Department of Earth and Atmospheric by Fairfield Nodal. The dataset consists of uncorrelated records harvested for a conventional reflection survey, cut into 21 second shot gathers for conventional correlation with a 16 second, 4-76 Hz upsweep. Although ideally our analysis would have been carried out on data harvested specifically to allow full correlation to the crustal depths of interest, this shorter conventional harvest was more quickly available at this point, and serves to illustrate that even these conventional harvests can be productively used. The additional correlated data lengths were obtained by the technique of extended correlation (Okaya, 1986) in which bandwidth is sacrificed for greater correlated record length.
Geology and seismic survey
The seismic data presented here were acquired by Fairfield Nodal Acquisition and Processing Company in South East New Mexico (Figure 1) and was provided to Cornell University specifically to demonstrate extraction of deep reflectivity. The harvested length of data is 21 second, which when correlated with the 16 second upsweep of 4-76 Hz, yields a conventional correlated record length of 5 seconds. This travel time corresponds to approximately 15 km depth for relevant velocities.