Constraining the Inversion’s Low Frequency Model via Seismic Attributes Analysis
Building the low frequency model for seismic inversion plays a key role in the inversion process as it initially as it initially establishes the spatial distribution of reservoir parameters the spatial distribution of the reservoir parameters which could produce misleading inversion results from the seismic data. Commonly used, the model-based inversion algorithm depends heavily on the accuracy of the input low frequency model as it merges it with the measured seismic. Using a standard workflow, we interpolate the well log acoustic impedance spatially across our control points and then assign distance-based weight to each. Such a workflow would not honor an anomalous signal, such as a change in geological depositional environment, between those wells.
The low frequency model can be derived using different techniques that honor the seismic amplitude away from the well, especially in the absence of sufficient control wells during exploration stage. In our work flow, two models are generated: the first one, “plain vanilla”, which uses simple interpolation between well logs, and the second one which is constrained using seismic attributes to guide the interpolation away from the wells in order to assist in qualifying the interpretation of the subsurface.
This dual-interpolation approach was tested on both synthetic and 3D field seismic data, using control wells on field data. We aimed to resolve reservoir parameters and to reduce risk in the exploration stage, in which the interpreter seeks to identify sand bodies in varying alluvial environments. We will show that in these exploration settings, a more representative low frequency model enhances our inversion product by resolving tuning and wavelet effects.
In conclusion, we determined that using the seismic amplitude to constrain the low frequency model building not only helped to improve the inversion process at well locations, but it also yielded lower error residual and, hence, produced a better representation of the subsurface sand distribution.