Search

Here we analyse ambient noise (AN) data generated during drilling of exploration boreholes and recorded using a dense array deployed over one of the numerous shallow iron-ore mineralization targets in the Pilbara region (Western Australia). Drilling and drilling-related operations were reoccurring in a sequence as described by the drillers' field notes, which created the rare opportunity to analyse AN data in time segments when only one type of technical process was predominantly active. Consequently, most of the recorded AN sources did not overlap in time and space. We extract the recordings in 15-min-long segments matching the time-span of single field-note entry and identify individually acting AN sources associated with specific field operations. The temporal variations of noise spectrograms and AN cross-correlations show dependency on the sequence of a few consecutive field operations and specific frequency–amplitude patterns associated with single field operations. These changes are directly reflected by the events visible in the retrieved virtual-source gathers (VSG), implying significant changes in noise temporal and spatial stationarity. Some VSGs represent the mixed contributions of surface and air waves. To remove the contributions of these arrivals to the reflection imaging, we visually inspect all data and select only field operations acting as stationary-phase sources specifically for the reflection retrieval. This was done for different receiver configurations inside PilbArray, and as a result, we obtain a collection of VSGs containing coherent body-wave reflections. Database of visually inspected VSGs is used to develop and benchmark a semi-automatic curvelet-based method for accurate parametrization of the reflection events retrieved from passive data and to compare the imaging quality of the different field operations. Common-midpoint stacks from manually and automatically selected VSGs show reflectivity consistent with the one obtained from the active-source data and related to the structure hosting shallow iron mineralization. Our results demonstrate the capacity of AN seismic interferometry to retrieve body-wave reflections and image shallow mineralization. They also provide an intermediate step toward automating the passive reflection imaging with similar data sets. [ABSTRACT FROM AUTHOR]

Seismic interferometry (SI) enables the retrieval of virtual sources at the location of receivers. In the case of passive SI, no active sources are used for the retrieval of the reflection response of the subsurface, but ambient-noise recordings only. The resulting retrieved response is determined by the illumination characteristics of the recorded ambient noise.Characteristics like geometrical distribution and signature of the noise sources, together with the complexity of the medium and the length of the noise records, determine the quality of the retrieved virtual-shot events. To retrieve body wave reflections, one needs to correlate body-wave noise. A source of such noise might be regional seismicity. In regions with notable human presence, the dominant noise sources are generally located at or close to the surface. In the latter case, the noise will be dominated by surface waves and consequently also the retrieved virtual common-source panels will contain dominant retrieved surface waves, drowning out possible retrieved reflections. In order to retrieve reflection events, suppression of the surface waves becomes the most important pre-processing goal.Because of the reasons mentioned above, we propose a fast method to evaluate the illumination characteristics of ambient noise using the correlation results from ambient-noise records. The method is based on the analysis of the so-called source function of the retrieved virtual-shot panel, and evaluates the apparent slowness of arrivals in the correlation results that pass through the position of the virtual source and at zero time. The results of the diagnosis are used to suppress the retrieval of surface waves and therefore to improve the quality of the retrieved reflection response. We explain the approach using modelled data from transient and continuous noise sources and an example from a passive field data set recorded at Annerveen, Northern Netherlands. [ABSTRACT FROM AUTHOR]