Empirical 3D depth/time dependent coherent noise generation for use in statistical models of seismic data

• To address the complications of measuring the time to frequency spectrum for 3D TL random noise, we use a modified STFT.
• By matching the real data amplitude with the phase of modeled noise, a 3D time-lapse modeling noise is produced.
• The resulting noise realization matches the coherent characteristics of the measured noise, as desired.
• The method is fast and easy to implement for the generation of many realizations of noise for statistical approaches.
• The use of the presented method for a statistical determination of plume detectability is a part of the future work.

Seismic monitoring feasibility analysis performed for a particular site usually involves the prediction of the time-lapse seismic signal and an assessment of the detectability of such a signal. In order to do this, level and other characteristics of time-lapse noise should also be known. In general, the observed time-lapse noise is spatially correlated, band-limited and exhibits lateral and temporal variations of its characteristics. If one wants to perform a statistical analysis of the detectability of the time-lapse signal one would need to have multiple realizations of such noise. To obtain such realizations, we propose a method for the modification of a single measured time-lapse noise volume by changing the phase spectra of the noise in sliding windows to random phases to preserve the spatial and temporal variability of the noise. To demonstrate the performance of this approach, we apply the method to CO2CRC Otway Project 4D seismic data.