Calibration workflow for 3D Gravity Flow modeling
- Start date: 2019
- End date: 2020
- Funded by: Eni S.p.A.
- Topic: Stochastic approaches leading to uncertainty quantification of models for sediment gravity flows.
We focus here on the use of 3D Gravity Flow as a computational suite to model sediment gravity flows. Lithified accumulations of these deposits may give rise to significant hydrocarbon reservoirs and numerical tools such as 3D Gravity Flow can assist to assess location, overall shape, and internal characteristics of these sediment bodies. 3D Gravity Flow mimics these sediment events through a system of coupled partial differential equations that are solved by way of a finite volume scheme. Typical outputs of interest include granulometric volumes and total thickness of deposits throughout the domain considered. The amount of model parameters required by 3D Gravity Flow is case-dependent and increases with the complexity of the system considered, at the expenses of computational cost. The latter might become so important to hamper model calibration. System characterization is further complicated by the observation that model parameters are typically affected by uncertainty and are not known a priori. Other sources of uncertainty underpinning our predictive ability include conceptual model uncertainty, i.e., the incomplete knowledge of all processes driving the system dynamics and their mathematical rendering at a given scale of interest, the topography of the bottom of the underwater system, as well as the proper resolution of the numerical grid employed to solve the selected system equations. The uncertainty stemming from all of these elements propagates to model outputs (i.e., modeling goals). We propose a strategy for the construction of an automatic calibration tool and a quantification of the way uncertainty propagates from model inputs to modeling goals. The general framework rests on the sequential use of (a) global sensitivity analysis (GSA), (b) model reduction, and (c) stochastic inverse modeling techniques.