Uncertainty Inherent in Geologic Data, Interpretations and Conceptual Models Used in Groundwater Flow Modeling
Abstract
When using MODFLOW-2000, parameter estimation and the calculation of flow model uncertainty are generally based on a static hydrogeologic framework. Although the properties of hydrogeologic units may be estimated and uncertainty ascertained during model calibration, the spatial location of the boundaries of the units are not usually modified. As such, the hydrogeologic framework is essentially viewed as absolute, and as a result having no uncertainty. In reality there are numerous types of uncertainties inherent in geologic data that could be acknowledged and accounted for. Framework models often include mixtures of quantitative geologic data, interpretations based on quantitative data, and qualitative, tacit geologic knowledge. Interpolated surfaces are created from these "data" and then "stacked" in geologic or stratigraphic modeling software to create volumetric units. In many cases, more than one geologic conceptual model may honor all available data. Because the task of building a digital 3D framework model is labor-intensive and technically difficult, it is usually not possible to construct 3D frameworks to test all possible conceptual models and decisions regarding the preferred conceptual model are often made prior to the creation of a digital model. Once a model is completed in the digital realm, the technical capabilities of the software and visual appeal can effectively mask a lack of data or flaws in the conceptual model. What is needed is a way to explicitly identify uncertainty inherent in framework models so that alternatives can be explored more efficiently, and so that the geologic and mathematical sources of error are more transparent. Geologic error and uncertainty inherent in any digital 3D framework model might be subdivided into five general categories: (1) interpolation, the construction of surfaces between data points; (2) inference, the presence of information that constrains the location or property of a feature without being actual data from that feature; (3) extrapolation, methods for extending calculated surfaces beyond the range of the data; (4) interpretation and geologic inference that are based on data, such as geologic cross sections; and (5) paradigms, which includes elements that are constructed based on a conceptual model of the system. These five categories are not strict statistical definitions but are concepts that describe types of uncertainty within the framework model, promote understanding of the logic that went into building the model, and allow effective scrutiny of the model and possible alternatives. These subdivisions help focus attention on where the framework model has the greatest likelihood of being accurate and where alternatives are possible and therefore, where and what methods should be used to quantify error in a particular aspect of the model. By quantifying error in the framework model, model developers and users are better equipped to evaluate the use of particular aspects of the model in groundwater flow modeling.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2002
- Bibcode:
- 2002AGUFM.H71A0775S
- Keywords:
-
- 1829 Groundwater hydrology;
- 1899 General or miscellaneous