Stochastic simulation of cave systems with ODSim

in: Proc. 28th Gocad Meeting, Nancy

Abstract

Cave systems are a challenge in both groundwater and petroleum modelling as their high spatial heterogeneity considerably in uences ow in many carbonate reservoirs. The complexity of the dissolution process involved in karst development make the building of accurate 3D description of karst networks geometries dicult. For this reason, Henrion et al. [2007] develop an alternative approach based on stochastic methods to generate 3D realistic cave systems. In the continuation, this paper focus on the extraction of channel ow network and the condidtioning of karst realizations to eld data and secondary informations such as water level. The rst step of the method consists in building the structural model which includes a network of discontinuities such as fractures and bedding planes. The second step reproduces the selective nature of karst processes (enlarging the fractures presenting the most favourable conditions for dissolution): the structural model is discretized into a graph of connectivities [Vitel, 2007] allowing the use of a graph search algorithm to nd least-cost paths (preferential ow paths) between given initial nodes (sinks) and goal nodes (sources). In Henrion et al. [2007], fractures connected to extracted path were the skeleton of further karst development occurring on the whole fracture. Here, to be more consistent with chenalized ow encountered on fracture and bedding planes, karst is directly simulated around the extracted ow paths ("dissolution paths"). At this step data conditioning may be introduced. Eventually, to simulate the extension of conduits around "dissolution paths", the 3D distance to these paths which can be seen as a map of the potential of karstication is computed. A distance cuto is then generated by sequential Gaussian simulation and is used to perturb the initial distance function, creating realistic karst geometry. Data conditioning may also be included at this last step.

Download / Links

    BibTeX Reference

    @INPROCEEDINGS{223_pellerin,
        author = { Pellerin, Jeanne and Henrion, Vincent and Caumon, Guillaume },
         title = { Stochastic simulation of cave systems with ODSim },
     booktitle = { Proc. 28th Gocad Meeting, Nancy },
          year = { 2008 },
      abstract = { Cave systems are a challenge in both groundwater and petroleum modelling as their high spatial
    heterogeneity considerably in
    uences 
    ow in many carbonate reservoirs. The complexity of the
    dissolution process involved in karst development make the building of accurate 3D description of
    karst networks geometries dicult. For this reason, Henrion et al. [2007] develop an alternative
    approach based on stochastic methods to generate 3D realistic cave systems. In the continuation,
    this paper focus on the extraction of channel 
    ow network and the condidtioning of karst realizations
    to eld data and secondary informations such as water level. The rst step of the method consists
    in building the structural model which includes a network of discontinuities such as fractures and
    bedding planes. The second step reproduces the selective nature of karst processes (enlarging
    the fractures presenting the most favourable conditions for dissolution): the structural model is
    discretized into a graph of connectivities [Vitel, 2007] allowing the use of a graph search algorithm
    to nd least-cost paths (preferential 
    ow paths) between given initial nodes (sinks) and goal nodes
    (sources). In Henrion et al. [2007], fractures connected to extracted path were the skeleton of further
    karst development occurring on the whole fracture. Here, to be more consistent with chenalized
    
    ow encountered on fracture and bedding planes, karst is directly simulated around the extracted
    
    ow paths ("dissolution paths"). At this step data conditioning may be introduced. Eventually,
    to simulate the extension of conduits around "dissolution paths", the 3D distance to these paths
    which can be seen as a map of the potential of karstication is computed. A distance cuto is then
    generated by sequential Gaussian simulation and is used to perturb the initial distance function,
    creating realistic karst geometry. Data conditioning may also be included at this last step. }
    }