The coherency of ambient seismic noise recorded during land surveys and the resulting implications for the effectiveness of geophone arrays

Abstract

The use of analog sensor arrays is often assumed to provide signal-to-ambient-noise improvements proportional to the square root of the number of sensors being summed. We determined via numerical modeling and field experiments that the improvements sought were significantly hindered once the ambient noise exhibited coherence over the array being summed. As a first step, a numerical model was developed to explore the optimal sensor spacing based on the average correlation coefficient between sensors. Field experiments were then carried out to measure ambient noise using closely spaced geophones at several sites in Perth, Western Australia. We show that the measured noise at six test sites was strongly coherent over distances of up to 10 m, with the level of coherency being inversely proportional to frequency. The resulting optimum geophone spacing under our field conditions was determined to be 7.5 m. These results offered further encouragement to reassess our use of analog arrays and to consider recording the output from individual sensors. In addition to the benefits from postacquisition noise suppression, this approach will enable increased trace density without degradation of the signal caused by strong coherent noise sources over acquisition spreads.

Methods for reducing unwanted noise (and increasing signal) in passive seismic surveys

“…Wind noise is related both to the impact of the wind on the geophone case and to waves resulting from the movement of vegetation. (Dean et al 2015) with the latter considered to be the most significant. To reduce the impact of wind noise on the geophone casing a bucket is often placed over the instrument to shield it from the direct effects of the wind.…”

Geometry-independent realistic noise models for synthetic data generation

“…For example, Pearce and Barley (1977) assume that noise is stationary -an unrealistic assumption. Whereas another commonly used method of distributed surface sources generally fails to capture the complexities of recorded noise (Dean et al, 2015). The Isolated-CoVAriance (ICOVA) method of Birnie et al (2016) overcame these assumptions to generate noise models that accurately imitate recorded noise, nevertheless these models are restricted to the array geometry in which the sample noise has been recorded.…”

Analysis and models of pre-injection surface seismic array noise recorded at the Aquistore carbon storage site

“…However, this approach only serves to produce a distorted signal by weighting the sampled recorded noise by a signal having Gaussian distribution and so is not meaningful. A more deterministic noise modelling method is that of distributed surface sources where source properties, such as direction, amplitude and source time functions, are randomly distributed [e.g., Sylvette et al, 2006;Lunedei and Albarello, 2015;Dean et al, 2015]. While this modelling method provides significant improvements on the WGN modelling assumption, it is a theoretical modelling method independent of recorded noise and therefore has limitations to the extent to which it can model the complex properties of recorded noise.…”

“…While this modelling method provides significant improvements on the WGN modelling assumption, it is a theoretical modelling method independent of recorded noise and therefore has limitations to the extent to which it can model the complex properties of recorded noise. This is discussed by Dean et al [2015] who state "Although the modeled data have the same characteristics as the field measurements, it is unlikely that models can be built with the geologic, geographic, and meteorological detail required to create accurate models". A recent advancement in representing realistic noise in synthetic datasets is to directly incorporate a sample of recorded noise into the synthetic dataset (referred to as a 'cut-and-paste' job).…”