Bayesian Linear Inversion
LTrace Bayesian Linear Inversion provides a deterministic seismic inversion based on the linearized Bayesian methodology for estimating elastic properties from seismic data. This plugin provides both acoustic and elastic inversion. The input data are seismic data, wavelet and low-frequency model of the properties. The results are comparable to a constrained sparse spike inversion but are obtained 6x to 10x faster, since no optimization is required and the solution is analytically computed. The software also implements an approximation on the horizontal continuity model along an input horizon, allowing the user to impose lateral correlation on the results with a low overhead.
Features:
- Deterministic acoustic and elastic inversion for estimating Velocities or Impedances and Density.
- Runs 6x to 10x faster than a Constrained Sparse Spike inversion.
- Horizontal correlation along an input horizon.
- Quality Control (QC) for comparing the inversion results with well log data and iterative parameters adjustment.
- Accelerated by CUDA on NVIDIA® GPUs.
Check out our plugin at the:
Acoustic inversion
Absolute Acoustic Impedance remains a cornerstone of modern seismic interpretation projects. It enables interpreting layer properties without distortions caused by tuning effects. LTrace Acoustic Bayesian Linear Inversion provides a deterministic acoustic seismic inversion using the linearized Bayesian methodology for estimating acoustic impedance from seismic data. The input data are the full stack seismic data, the wavelet and the low frequency model of acoustic inversion.
Elastic inversion
LTrace Bayesian Linear Inversion provides a deterministic elastic seismic inversion using the linearized Bayesian methodology for estimating velocities or impedances and density. We use angle stacked seismic data – such as near, mid and far – along with low frequency models to quickly perform the inversion.
For the elastic inversion, the software uses the latest Graphics Processing Unit (GPU) technologies to accelerate the inversion processing. If the user has an NVIDIA® GPU, the inversion will be automatically processed using it. Using GPUs provides a performance gain over CPUs of 4 to 7 times in our benchmarks, depending on the model of the GPU and the vertical gate selected at the inversion.
Theoretical concepts
Using the convolutional seismic modeling and under the Gaussian assumption for the seismic errors and the prior distribution of the elastic properties, the Bayesian posterior distribution is analytically calculated:
The methodology is based on (Buland, A. & Omre, H., 2003) with some particularities. The prior distribution of elastic properties includes the property correlations, low frequency models and 3 dimensional spatial correlation models. We developed the software as a plugin for OpendTect, an open source platform used throughout the industry. We have employed the latest GPU-based libraries for extreme parallelism and acceleration, therefore, if the user has an NVIDIA® GPU, the process will run on it using only 10 to 20% of the time compared to running the inversion on CPU.
OpendTect BLI Webinar
Video tutorials
