On 20 September 2023, Alvaro Chaveste and Rocky Roden presented a new Machine Learning (ML) methodology for Reservoir Characterization, enabling lithofacies prediction between wells in weeks and at a fraction of the cost of sophisticated seismic inversion methods. The new ML methodology computes Self Organized Maps (SOM), an unsupervised form of ML, and statistically correlates the ML output to lithofacies from petrophysical logs. The methodology, developed using synthetic seismic data and applied to conventional and unconventional (shale) plays, defines the lithofacies of interest at high resolution. Click the movie to the right to see the methodology in action!
Supporting resources on lithofacies prediction using Paradise
Comparison to seismic inversion for lithofacies prediction
Inversion-based reservoir characterization computes rock properties (i.e., P- and S-wave Impedance and density) through seismic inversion. The desired reservoir properties (porosity, lithology, and fluids) are obtained through petrophysical inversion that relates rock properties to reservoir properties through rock physics models.
Seismic inversion computes the broadband reflectivity from seismic and complements it with a low-frequency model obtained, usually, from seismic velocities. Meeting these requirements implies that data should match physical models and have accurate seismic velocities. Great challenges are encountered by processors and interpreters who often fail to fully meet these conditions, which results in uncertainties in the Rock Properties computed from seismic.
Ideally, reservoir properties would be computed from rock properties through petrophysical inversion; nevertheless, the forward rock physics models that relate rock properties to reservoir properties are non-linear and mathematically complex. A common practice to circumvent these inconveniences has been to select, in cross-plots of rock properties, areas (volumes in 3D cross-plots) that correspond to target reservoir properties.
The alternative ML approach to reservoir characterization by Geophysical Insights has these technical advantages over seismic inversion for lithofacies prediction, saving time and money:
- Process is not deterministic – does not require data to match physical models
- No requirement to determine the wavelet and low-frequency trend in the data, which are non-trivial
- Employs multiple seismic attributes in ML clustering that identify natural patterns in the data relating to lithology
- Works on thin and thick beds as ML clustering is done at the seismic sample interval, below the wavelets resolution limit
- Neurons generated by SOM have some of the characteristics of inverted data; they map intervals, not just stratigraphic interfaces
Contact us to schedule a 15-minute consultation about applying ML-based Lithofacies Prediction to your area of interest. More on the process is outlined in the images below.
Amplitude Trace and Seismic Attributes Computed from Amplitude
Multidimensional vectors input into SOM are obtained by taking the same time/depth sample for all attributes. Each sample is classified by a neuron, which provides interpretable data below the seismic wavelet resolution.
“SOMs find facies patterns first, which can then be labeled by the geologist using log data or geologic analogs. The flexible nature of SOMs makes them a useful unsupervised approach for seismic facies analysis, particularly where there may be limited well control.”
– Dr. Heather Bedle, assistant professor at the University of Oklahoma and principal investigator for the Attribute-Assisted Seismic Processing & Interpretation (AASPI) Consortium
SOM Topology with Seismic Amplitudes and Lithofacies
The SOM topology map at left corresponds to the SOM neurons in the middle track. The left-most well track shows the seismic trace at the well location. The right-most trak is the lithofacies computed from well logs. Formation thicknesses observed in the lithofacies are smaller than a seismic trough (reference the shallow clays).
Four Lithofacies Estimated from Petrophysical Evaluation of Five Niobrara Wells
The lithofacies from the well shown in the right-most track of Figure 2 is the well-log representation of the lithofacies shown here as PDF plots. The economic importance of the play is evidenced by the appearance or hydrocarbons in all lithofacies.
SOM-to-Lithofacies Correlation in Niobrara Formation
The SOM-to-Lithofacies correlation ties the neurons from seismic data to the lithofacies from well logs. In the lower right, the ‘winning’ lithofacies topology shows the lithofacies corresponding to the neurons computed from seismic. The lithofacies from the well are inserted for reference at the well’s location. Multi-client data presented with permission from Geophysical Pursuit, Inc. & Fairfield Geotechnologies.
Schedule a 15-minute
consultation on your
area of interest
- Discuss your area of interest, including lithofacies in the region
- Understand the extent of the well logs and petrophysical data available
- Assess how machine learning will help accelerate lithofacies prediction