Ambient noise cross-correlations applied to reservoir scale OBS-recordings Cornelis Weemstra 12 , Alexander Goertz 1 , Brad Artman 1 and Lapo Boschi 2 ID: S33A-2069 1 Spectraseis AG, Z¨ urich, Switzerland 2 ETH, Z¨ urich, Switzerland Abstract Surface waves extracted from the ambient seismic wavefield via interferometry can be used for velocity inversion. In order to invert shear wave velocities at the reservoir scale, Bussat & Kugler (2009) adapted this approach to Scholte waves at frequencies up to 1 Hz. These waves were extracted from comparatively short OBS recordings, ranging from several hours to a few days. We extended the analysis of the pressure components. This passive seismic data set was acquired in April/May 2007. It is recorded over a ˜220 km2 survey area at an average depth of 360 m, offshore Norway. Data was recorded at 117 seabed locations by 16 ocean-bottom seismometers (OBS), equipped with a broadband seismometer and a differential pressure gauge (DPG). The instruments have a flat response to particle velocity between 240 s and 50 Hz and data was acquired with a sampling rate of 125 Hz. The stations were systematically relocated after 1 to 2 days of recording except for two stations which were recording continuously. The main energy in the data below 5 Hz stems from swell noise, ocean microseisms and Scholte waves traveling along the seabed. After filtering the multicomponent data between 0.01 and 10.0 Hz, we analyzed them with respect to azimuthal variation of the incident wavefield in order to find directions of possible dominant sources that may violate the assumption of a diffuse wavefield and impact the retrieval of interstation Green’s functions. Future work will be to analyze the influence of different temporal normalization techniques on the Green’s function gathers. The ultimate goal of our effort is to understand the effect of temporal normalization on the decay of amplitude with distance, i.e. the relation between temporal normalization and attenuation of the obtained Green’s functions. Synthetic data will be used to test this effect. We seek processing techniques that preserve the attenuation characteristics of the subsurface best. For future research this will provide us with the opportunity to invert for the quality factor using a seismic data set of this scale. Data Characteristics Survey characteristics ◮ 4-component data acquired through end of April/begin of May 2007 (a period of roughly 12 days). ◮ Only differential pressure gauge (DPG) data used. ◮ 16 broadband ocean-bottom seismometers (BBOBS). ◮ Survey area of ∼ 220km 2 . ◮ Different lines deployed successively. ◮ Nominal station interval along the lines of 500 m. ◮ Average water depth of 360 m. Figure 1: The configuration of the Astero survey. The stations are shown by the color-filled circles. The colors represent the start of recording of the stations. To the right you can see the period of recording of each station. Figure 2: The configuration of the Astero survey. The stations are shown by the color-filled circles. The colors represent the duration of recording of the stations. To the left you can see the start of recording of each station. Two stations (P01 and P02) were recording continuously over the whole period of 12 days. Processing sequence 1. Traces are cut in time-windows of 60 seconds. 2. The time-windows are detrended. 3. One-bit normalization of the data. 4. Cosine taper of 2.5% of the trace length. 5. Fourier transformation of the traces. 6. Multiplication of the spectra of the synchronous time-windows, i.e. actual cross-correlation. 7. Inverse Fourier transformation. Signal to Noise Ratio The Signal to noise ratio (SNR) of the Green’s functions as stated in the figures or given in the caption of the figures is an empirical one. The Cross correlations are first normalized such that it’s maximum value has an amplitude of one. The SNR is defined as the maximum amplitude in the velocity wedge of 350 m/s to 750 m/s divided by the standard deviation on the noise windows. The noise windows are defined as the windows corresponding to higher and lower velocities besides a transitional margin. This margin is outside the velocity wedge and it’s width depends on the frequency range of the Cross correlations. Longer periodic cross correlations have longer margins between ’signal window’ and ’noise window’. Acknowledgements First of all, we would like to thank Statoil for providing us with this dataset. We would also like to thank the parties that were involved in the recording of the data: Bergen Oilfield Services and the Scripps Institution of Oceanography. References Bussat, S. & Kugler, S., 2009. Recording noise - Estimating shear-wave velocities: Feasibility of offshore ambient-noise surface-wave tomography (ANSWT) on a reservoir scale: 79th Annual International Meeting, SEG, Expanded Abstracts, 1627 - 1631. Green’s Function Gathers Figure 3: Green’s function gathers (GFG’s) for three frequency bands. Only cross-correlations based on synchronous recordings of more than 4 hours, interstation distances of more than 3.2 km and SNR higher than 4 are displayed. Violation of the Equipartitioned Wavefield Assumption Figure 4: The difference between the amplitude of the causal and anti-causal Green’s function plotted on the radial axis as function of back-azimuth. Only cross-correlations based on synchronous recordings of more than 4 hours, interstation distances of more than 3.2 km and a SNR higher than 4 are taken into account. Attenuation Figure 5: Amplitude decay of the Green’s function with distance along the lines. Attenuation with distance along the lines ◮ The amplitudes of the Green’s Functions for 5 periods of synchronous recordings are examined. ◮ Again, only cross-correlations based on synchronous recordings of more than 4 hours, interstation distances of more than 3.2 km and a SNR higher than 4 are taken into account. ◮ Positive interstation distances correspond to energy traveling in the direction of increasing station number along the concerning line. ◮ The cross-correlations of the different lines correspond to synchronous recordings of differing length: 1. Dark blue: Line A; 9 stations; 30 hours of synchronous recording. 2. Light blue: Line A; 12 stations; 12 hours of synchronous recordings. 3. Green: Line D; 8 stations; 11 hours of synchronous recordings. 4. Red: Line B; 9 stations; 7 hours of synchronous recordings. 5. Yellow: Line B; 9 stations; 7 hours of synchronous recordings. ◮ The aim is to get a first insight in the attenuation of the amplitudes with distance. ◮ Because of the synchroneity and 2 dimensional station distribution, the variation in time and direction of source activity is taken out of the equation. ◮ The graphs show 1 √ r fitted to the highest amplitude of the Green’s function measured on each line. The curve is fitted to this amplitude at an interstation distance of 3.2 km. Model setup for synthetic data Figure 6: Setup and velocity and Q model that will be used to test the dependence of the Green’s function amplitude on attenuation. ◮ 2D finite difference forward propagator with a model with a length of 20 km and a depth of 5 km. ◮ 100 sources randomly placed along the surface. ◮ Water column of 300 m depth and at the bottom 400 receivers with an interstation distance of 50 m. Conclusions ◮ Only 4 hours of data can be sufficient to obtain a Green’s function by cross correlating pressure gauge data in this specific environment ◮ The dominant source direction changes with frequency and time. ◮ The Green’s function’s amplitudes decay with distance. ◮ The amplitudes seem to decay a bit more than the theoretically expected geometrical attenuation (Q?). ◮ A lot more research needs to be done to understand the dependence of the Green’s function on attenuation of the medium. Institute of Geophysics, ETH Z¨ urich, Switzerland Mail: cornelis.weemstra@tomo.ig.erdw.ethz.ch WWW: http://www.seg.ethz.ch View publication stats View publication stats