Prioritizing Major Predictive Uncertainty Sources in Coupled Hydrosystem Models

Coupled hydrosystem models (CHMs) can provide useful insights to hydrogeological systems. However, inherent assumptions and simplifications in CHMs result in model prediction uncertainty. Typically, numeric CHMs are used in a deterministic way without rigorous quantification of prediction uncertainty. In order for CHMs to be useful and preserve their credibility, it is important for them to have quantifiable measures of uncertainty that stakeholders, who may not have a background in CHMs, can understand (Pappenberger and Beven 2006).

Numeric CHMs have three main inherent uncertainty sources as described by (Gupta et al. 2005):

• Data: errors in observations used to build, calibrate and operate the model; and the scarcity of available data;

• Parameter: there may be many sets of parameter values that behave equally well during calibration, yet lead to different predictions;

• Conceptual: unknowns with respect to the specific location of aquitards and facies boundaries, the incorporation and the representation of physical processes and their connections, the geometry and limits of the model domain. 

This project has three main objectives: 1) to determine the dominant predictive uncertainty sources in CHMs; 2) to determine the conditions in which a specific uncertainty source dominates; and 3) to determine which uncertainty sources are irreducible. A methodology to create the CHMs and quantify uncertainty is developed.