Building a training image with digital outcrop models

Current standard geostatistical approaches to characterizing subsurface heterogeneity may not capture realistic facies geometries and fluid flow paths. Multiple-point statistics (MPS) has shown promise in portraying complex geometries realistically; however, realizations are limited by the reliability of the model of heterogeneity upon which MPS relies, that is the Training Image (TI). Attempting to increase realism captured in TIs, a quantitative outcrop analog based approach utilizing terrestrial lidar and high-resolution, calibrated digital photography is combined with lithofacies analysis to produce TIs. Terrestrial lidar scans and high-resolution digital imagery were acquired of a Westwater Canyon Member, Morrison Formation outcrop in Ojito Wilderness, New Mexico, USA. The resulting point cloud was used to develop a cm scale mesh. Digital images of the outcrop were processed through a combination of photogrammetric techniques and manual digitizing to delineate different facies and sedimentary structures. The classified images were projected onto the high-resolution mesh creating a physically plausible Digital Outcrop Model (DOM), portions of which were used to build MPS TIs. The resulting MPS realization appears to capture realistic geometries of the deposit and empirically honors facies distributions.