RT Journal Article ID 07c42d2178285bf1 A1 Santos, Susana A1 Gaspar, Ana Teresa F. S. A1 Schiozer, Denis J. T1 COMPARISON OF RISK ANALYSIS METHODOLOGIES IN A GEOSTATISTICAL CONTEXT: MONTE CARLO WITH JOINT PROXY MODELS AND DISCRETIZED LATIN HYPERCUBE JF International Journal for Uncertainty Quantification JO IJUQ YR 2018 FD 2018-02-15 VO 8 IS 1 SP 23 OP 41 K1 petroleum field development K1 risk analysis K1 geostatistics K1 reservoir simulation K1 Latin Hypercube K1 Monte Carlo K1 joint proxy models AB During the development of petroleum fields, uncertainty quantification is essential to base decisions. Several methods are presented in the literature, but its choice must agree with the complexity of the case study to ensure reliable results at minimum computational costs. In this study, we compared two risk analysis methodologies applied to a complex reservoir model comprising a large set of geostatistical realizations: (1) a generation of scenarios using the discretized Latin hypercube sampling technique combined with geostatistical realizations (DLHG) and (2) a generation of scenarios using the Monte Carlo sampling technique combined with joint proxy models, entitled the joint modeling method (JMM). For a reference response, we assessed risk using the pure Monte Carlo sampling combined with flow simulation using a very high sampling number. We compared the methodologies, looking at the (1) accuracy of the results, (2) computational cost, (3) difficulty in the application, and (4) limitations of the methods. Our results showed that both methods are reliable but revealed limitations in the JMM. Due to the way the JMM captures the effect of a geostatistical uncertainty, the number of required flow simulation runs increased exponentially and became unfeasible to consider more than 10 realizations. The DLHG method showed advantages in such a context, namely, because it generated precise results from less than half of the flow simulation runs, the risk curves were computed directly from the flow simulation results (i.e., a proxy model was not needed), and incorporated hundreds of geostatistical realizations. In addition, this method is fast, straightforward, and easy to implement. PB Begell House LK https://www.dl.begellhouse.com/journals/52034eb04b657aea,215dae1860d04f34,07c42d2178285bf1.html