Incorporating uncertainty in static or dynamic modeling

When modeling a reservoir, a large range of interpretations can result from a single data set. With uncertainty analysis, you explore the parameter ranges rather than focusing on specific parameter values. This way you can determine the degree of uncertainty in your model outcomes that are due to uncertainty in the input parameters. JewelSuite Subsurface Modeling allows you to capture uncertainty at every step of the workflow and to propagate that uncertainty through the workflows so that its impact on the business decisions can be quantified. Uncertainty can be represented at different levels, from probability density functions (PDFs) for individual parameters to a set of alternative subsurface model concepts. The uncertainty representations are embedded in the workflows, more specifically on the workflow input forms where the model parameters are defined. This promotes auditability of the uncertainty ranges and consistency with the supporting data.

In the application, uncertainty can be incorporated in static (volumetric) calculations as well as dynamic simulations, however, for dynamic simulation the set of uncertain parameters to propagate to the run is more limited. In case of volumetric uncertainty calculations, uncertainty is embedded at two levels:

1. Uncertainty in the form of alternative high-level scenarios, the so-called 'subsurface concepts' (not applicable to dynamic simulation)
2. Uncertainty as parametric uncertainties embedded in model parameters throughout the workflows

Setting up concepts with the concept strip

Some uncertainties can only be incorporated via alternatives to a single interpretation. This range of alternative interpretations of a single data set is called conceptual uncertainty. One concept is a 'string' of consecutive modeling steps, where each modeling step is represented by one domain model (e.g. Fault Model 1, 3D Structural Model 1, 3D Grid 1, etc.). By creating various concepts and combining them into one set of input data (the so-called subsurface model) for volumetric calculations, you can as such incorporate conceptual uncertainty.

A typical example of a conceptual uncertainty is a reservoir where structural uncertainty exists around the presence of faults (for example related to the fault pattern in regard to seismic dimensions). Either the reservoir is densely faulted, or there are no faults at all. By creating two concepts (one with a structural model which is faulted, and one with a structural model that has no faults at all) and combining both concepts into one probabilistic calculation, you can as such capture the associated uncertainty in the outcome of your volumetrics calculation. Note that incorporating conceptual uncertainty via multiple concepts is only possible for volumetric uncertainty calculations, not for dynamic simulation. For dynamic uncertainty modeling, your input always consists of one single concept.

In JewelSuite, you create concepts (either volumetric concepts or simulation concepts) with the concept strip. The final set-up of the uncertainty run is done with the study strip, where the volumetric or simulation concept(s) are used as input.

Incorporating parametric uncertainties

Uncertainties which are related to input parameters which you specify on the workflow forms (e.g. the fluid level of a fluid model) and which are defined by Probability Density Functions (PDFs) are the so-called parametric uncertainties. For each parametric uncertainty you can indicate if it is an independent or correlated uncertainty (e.g. you can set porosity and permeability to be correlated uncertainty parameters). When set to correlated, you must specify a parent-child relationship between the correlated uncertainty parameters. Once defined, the computational model can be simulated by sampling a single value for each parameter distribution per realization. In the application, the parameters for which you can define parametric uncertainty can be recognized by a symbol () next to the parameter's entry field on the workflow form. Clicking on the symbol will open an Uncertainty Parameter dialog where you can specify the PDF for the input parameter. This PDF is sampled during the uncertainty calculation with the STUDY strip (for dynamic simulation incorporating parametric uncertainties, sampling is done by the CMOST engine of CMG). For an overview of all parameters in the application that can be made uncertain, see Overview of all uncertainty parameters in the application.

Probabilistic calculations with the study strip

With the STUDY strip you specify in more detail how the uncertainty calculation must be performed. The input consists of the concepts that you created with the concept strip (see previous paragraph). For the volumetric uncertainty calculations, setting up the study means specifying the sampling technique, choosing your uncertainty parameters, defining parent-child relations, etc. For dynamic simulation, setting up the study is more limited, as the actual uncertainty run is performed by the CMG simulator.