Stochastic simulation of channel migration conditioned to subsurface data

in: 10th International Geostatistical Congress, Valencia, Spain

Abstract

Integrating geological concepts in quantitative subsurface models of channelized systems while honoring all available data is a major research area. Advances in seismic imaging permits the direct observation of architectural elements which were preserved in the sedimentary record and reflect the most recent state of the sedimentary system. Sedimentological concepts are then useful to interpret subscale structures which are poorly imaged owing to limited seismic bandwidth. In this work, we propose to anchor a channel object model on these evidences and to simulate its migration backward in time. The objective is then to reconstruct possible geobodies and the associated heterogeneities. In principle, such a backward modeling is not possible because depositional processes are irreversible. However, we do not use explicit physical or energetic considerations in the backward modeling but rather geometric methods. Also, we compensate the ill-posedness of the retro-migration process by introducing stochastic components to reproduce both gradual and drastic changes of system evolution. More precisely, we suggest to identify abandoned meander and point bar facies in order to integrate them into a possible channel object. These remnants of the paleo-locations of channel paths can also be simulated to overcome blurred parts of the seismic image or unexplored areas. Abandoned meanders correspond to discrete modifications of channel paths while point bars document the continuous evolution of the channel loops. Indeed, abandoned meanders are relicts of paleo-channels which were once continuous. However, gradual retro-migration of this channel is able to produce point bars geometry but does not preserve the sinuosity of the channel. Indeed, it fails in reproducing abrupt changes of channel trajectory which occurs during meander cut-off. Together, these evidences of the migration can be dated relatively one to another in order to classify their integration into the backward migration process. Conceptual geometric aspects of the river migration are also reproduced such as its three-dimensional components (i.e. lateral, vertical or downstream evolution) inside the floodplain. As a first proof of concept, the method has been applied on satellite images of the Tyung river (Russia) which presents an interesting channel belt more or less preserved according to the areas. The first results are quite encouraging and permit to approach the underground architecture of a reservoir. In a further step, it could be used in building realistic heterogeneity models for reservoir fluid flow simulation.

Download / Links

    BibTeX Reference

    @INPROCEEDINGS{Parquer2016b,
        author = { Parquer, Marion and Collon, Pauline and Caumon, Guillaume },
         title = { Stochastic simulation of channel migration conditioned to subsurface data },
         month = { "sep" },
     booktitle = { 10th International Geostatistical Congress },
          year = { 2016 },
      location = { Valencia, Spain },
      abstract = { Integrating geological concepts in quantitative subsurface models of channelized systems while honoring all available data is a major research area. Advances in seismic imaging permits the direct observation of architectural elements which were preserved in the sedimentary record and reflect the most recent state of the sedimentary system. Sedimentological concepts are then useful to interpret subscale structures which are poorly imaged owing to limited seismic bandwidth.
    In this work, we propose to anchor a channel object model on these evidences and to simulate its migration backward in time. The objective is then to reconstruct possible geobodies and the associated heterogeneities. In principle, such a backward modeling is not possible because depositional processes are irreversible. However, we do not use explicit physical or energetic considerations in the backward modeling but rather geometric methods. Also, we compensate the ill-posedness of the retro-migration process by introducing stochastic components to reproduce both gradual and drastic changes of system evolution.
    More precisely, we suggest to identify abandoned meander and point bar facies in order to integrate them into a possible channel object. These remnants of the paleo-locations of channel paths can also be simulated to overcome blurred parts of the seismic image or unexplored areas. Abandoned meanders correspond to discrete modifications of channel paths while point bars document the continuous evolution of the channel loops. Indeed, abandoned meanders are relicts of paleo-channels which were once continuous. However, gradual retro-migration of this channel is able to produce point bars geometry but does not preserve the sinuosity of the channel. Indeed, it fails in reproducing abrupt changes of channel trajectory which occurs during meander cut-off. Together, these evidences of the migration can be dated relatively one to another in order to classify their integration into the backward migration process. Conceptual geometric aspects of the river migration are also reproduced such as its three-dimensional components (i.e. lateral, vertical or downstream evolution) inside the floodplain.
    As a first proof of concept, the method has been applied on satellite images of the Tyung river (Russia) which presents an interesting channel belt more or less preserved according to the areas. The first results are quite encouraging and permit to approach the underground architecture of a reservoir. In a further step, it could be used in building realistic heterogeneity models for reservoir fluid flow simulation. }
    }