Abstract: |
Braided river systems are characteristic of high energy environments which result in highly heterogeneous deposits. In Alpine regions such as Switzerland, they constitute an important part of the alluvial aquifers which are taped for drinking water supply. To model the heterogeneity of such deposits, several 3D sedimentary model set up exist. Among them, descriptive methods translating geological or geophysical site-specific data into conceptual facies models or structure imitating methods including a wider variety of stochastic models exist but do not include strong geological conceptual knowledge and their degree of realism is rather weak. Another way of modeling those environments is to use process imitating methods, but the conditioning to field data is often very difficult. In this work, we propose a new method combining high resolution analog data with the use of the Direct Sampling method embedded in a pseudo-genetic algorithm. On the one hand, we use high resolution Digital Elevation Models (DEM) obtained from Airborne photography and LIDAR acquired at successive time steps on analog sites. These data provide both high resolution and large scale information. On the other hand, the Direct Sampling algorithm is a recent multiple points statistics method allowing to work directly with continuous variables. Here the training images represent topography variations due to erosion and deposition processes on a time interval. Successive topography surfaces are built by iterative simulations conditional to the previous time steps. The over-lapping and crossing surfaces produce volumes that are populated by different sedimentary textures and structures, according to the sedimentary raw material, volume shapes and analog outcrops. |
