Quantifying Natural Delta Variability Using a Multiple-Point Geostatistics Prior Uncertainty Model: Bridging the Gap Between Quantitative Surface Dynamics and Machine Learning

Abstract

Deltas can take on a range of forms as a result of their intrinsic (autogenic) variability. Here we attempt to quantify uncertainty associated with autogenic pattern variability from overhead photographs of an experimental delta evolving under fixed boundary conditions, using two modern multi-point geostatistical methods. Specifically, we evaluate whether and to what degree the autogenic variability in a laboratory experiment can be represented and reproduced by a multiple-point geostatistical prior uncertainty model. We used (1) distance-based clustering of overhead snapshots of the experiment and (2) a rate of change quantification by means of a computer vision algorithm, to identify a set of training images from which a set of geostatistical model realizations can be generated. We show quantitatively that, with either training image selection method, we can statistically reproduce the natural variability of the delta formed in the experiment. Furthermore, the patterns represented in the set of training images can be defined as the ‘eigen-patterns’ of the natural system. The eigen-pattern in the training image sets display patterns consistent with previous physical interpretations of the fundamental modes of this type of delta system: a highly channelized, incisional mode; a poorly channelized, depositional mode; and an intermediate mode. We consider it very encouraging that the eigen-patterns, identified geometrically without reference to process dynamics, nonetheless correspond to fundamental dynamic modes of the system. Additional multi-point geostatistical investigations of experiments with different but steady boundary conditions may offer insight into how much the construction of the prior model depends on the dynamics of the system under consideration. Refining existing metrics such as the ratio of overbank to channel deposits may improve the linkage between sub-surface architecture, system dynamics and prior model construction, to improve uncertainty quantification in subsurface reservoirs using Bayesian approaches.

AAPG Datapages/Search and Discovery Article #90291 ©2017 AAPG Annual Convention and Exhibition, Houston, Texas, April 2-5, 2017