The study of fluvial deposits has been a relevant topic for petroleum and hydrogeology during the last decades. Reservoirs of fluvial origin are playing a major role in newly proven reserves of hydrocarbons. An important characteristic of many fluvial reservoirs is the presence of sinuous sand-filled channels within a background of floodplain shale. The complex geometry and the heterogeneity of these deposits makes their characterization challenging.
We propose a new approach to model channels geometry using continuous variable multiple-point geostatistical method (Direct Sampling, DS). The idea is to simulate the direction taken by the channel along the river axis, discretized as a 1D process. The simulations are based on training images consisting of orientations observed on present-day rivers. The simulated meanders show a high degree of realism, including meanders presenting multiple internal bends. Oxbow lakes do occur in the simulation, which are thereafter modeled as separate sedimentary bodies. Since the simulations are 1D, the computational cost is minimal.
Conditioning to observed channels locations is accomplished by the addition of a trend to the simulated channels. However, this conditioning method can produce artifacts when a large number of conditioning data are present, resulting in a loss of realism. In such cases, we propose conditioning through an inverse procedure that is guaranteed to honor the channel orientations observed in the training image.
Once the geometry of the river skeleton is defined, sedimentary deposits are generated iteratively by following sedimentary rules describing the processes of meander migration as observed in active systems.
The methodology is illustrated with a case study based on a highly sinuous meandering river located in the Bolivian Amazon basin.