Model uncertainty--parameter uncertainty versus conceptual models

Area of Science:

• Hydrogeology
• Environmental Modeling
• Water Resource Management

Background:

• Model structure uncertainty is a primary driver of predictive uncertainty in simulations.
• Traditional uncertainty studies often focus solely on parameter uncertainty, neglecting structural variations.
• Groundwater modeling requires robust methods to address diverse sources of uncertainty.

Purpose of the Study:

• To investigate the extent to which parameter uncertainty analysis can account for model structure errors in groundwater models.
• To compare predictive uncertainties arising from different conceptual groundwater models.
• To assess the influence of calibration data types on the significance of conceptual model uncertainties.

Main Methods:

• Development of three distinct groundwater models based on varying hydrogeological interpretations.
• Inverse calibration of each model using groundwater heads and streamflow data.
• Monte Carlo simulations for parameter uncertainty analysis within each conceptual model.
• Comparative analysis of predictive uncertainty intervals across the different models.

Main Results:

• Significant differences in predictive uncertainty intervals were observed among the three conceptual models.
• Discrepancies in predictions were most pronounced for data types not utilized during model calibration.
• Conceptual model uncertainties increase in importance for predictive simulations involving extrapolated data.

Conclusions:

• Parameter uncertainty analysis alone is insufficient to encompass model structure errors in groundwater modeling.
• The choice of conceptual model and its underlying hydrogeological interpretation critically influences predictive uncertainty.
• Future research should prioritize methods that explicitly address and quantify model structure uncertainty, especially when predictions extend beyond calibration data.