PREDICT: PeRmEability DIstributions of Clay-smeared faulTs [Github]

Computes upscaled fault permeability distributions using a parameter-based, probabilistic description of clay and sand smears. Written in MATLAB.


  • Generates multiple realizations of the fault core materials in shallow (depth < ~3km), poorly-lithified siliciclastic sequences, using a high-resolution grid (fine grid). The computation of fault permeability is performed in a given throw window, and the code can be run in 2D or 3D.

  • Outputs the directional components of the fault permeability tensor (dip-normal, kxx; strike-parallel, kyy; and dip-parallel, kzz) in an upscaling (coarse) grid defined by the user (3D version). For example, this flexibility is useful to assign fault permeability in subsequent flow simulation models.

  • The fault permeability is obtained via flow-based upscaling of the fine grid permeability. PREDICT uses MRST to perform permeability upscaling (finite volume method)

  • Incorporates uncertainty in the geological variables that control the fault material distribution and their properties. Therefore, the output permeability is given as a set of probability distributions.

  • Mostly vectorized where possible and can be run in parallel using MATLAB's parfor, so it is fast and scales well with the number of computing cores available.

convergence_nSim_3D.png

Convergence of output probability distributions vs number of realizations performed, for five different example sequences (A to E). N = 1000 realizations are enough.

times.png

Computing times for an example sequence vs cores available

example_output.png

Example of output probability distributions of fault permeability for each directional component, and their correlation.