Enhanced Reservoir Definition - A Novel Vectorized Disorder Seismic Attribute for Spatial Incoherency Mapping

Summary Although most of the regions in seismic data are usually stratified, coherent and continuous, there are areas in seismic data that look chaotic, i.e., high randomness zones. The randomness of seismic data is mainly due to three reasons: energy of random reflections or refractions from certain subsurface geological structures, e.g. fraction zone, gas chimney, karst collapse, and terminated unconformity;suboptimum acquisition and consequential low data quality, e.g. all kinds of noise;imperfect data processing/imaging procedure, e.g. inaccurate migration velocity, imaging artifacts, and operator aliasing. Thus, a reliable method to estimate the spatial distribution of the randomness level in seismic data is essential and indispensable for reservoir characterization. Estimation of randomness spatial distribution with high resolution will assist reservoir characterization in multiple ways, including: 1. identify certain geological features such as: fracture zones, gas chimneys, karst collapse, terminated unconformity or other random reflection/refraction related features; 2. indicate seismic data quality and random noise level, e.g. recognizing boundary problem between merged surveys; 3. provide covariance matrix for seismic inversion or uncertainty index for interpretation and reservoir simulation. Thus, random- ness level measurement is an important task in seismic image processing/imaging, which helps both to improve seismic processing/imaging quality control (QC) and to better expose some geology structures, such as subsurface anomalies, fracture zones and gas chimneys. There have been numerous early efforts at estimating the randomness level in seismic data, e.g. cross-correlation- based algorithms, structure-tensor-based algorithms, grey level co-occurrence-matrix-based algorithms, and convolutional filtering-based algorithms. Among them the Disorder algorithm is a convolutional filtering-based algorithm designed using an optimization approach. By design, the Disorder is insensitive to faults, channel edges, unconformities, and other linear discontinuities of geological features. However, the Disorder algorithm suffers from two major is- sues: 1. An-isotropy and 2. Non-smooth-able. Thus, we proposed novel Vectorized Disorder algorithm, which in comparison to Disorder algorithm provides desirable (much more isotropic and smoother) randomness distribution estimation.