Detection of hidden model errors by combining single and multi-criteria calibration

Area of Science:

• Environmental modeling
• Hydro-biogeochemistry
• Landscape-scale processes

Background:

• Environmental models use mathematical equations to simulate landscape processes, with observations for validation.
• Model performance and uncertainties are typically assessed using single or multi-criteria methods.
• Existing assessment methods can lead to errors, such as overestimated uncertainty or ignored model deficiencies.

Purpose of the Study:

• To combine single and multi-criteria model assessment approaches for a comprehensive evaluation.
• To identify and rectify common mistakes in environmental model assessment.
• To guide model users and developers in detecting and correcting hidden errors.

Main Methods:

• Utilized a coupled hydro-biogeochemistry landscape-scale model for simulations.
• Simulated 14 target values including discharge, stream nitrate, soil moisture, soil temperature, and trace gas emissions (N2O, CO2).
• Employed a combined single and multi-criteria assessment approach for model validation.

Main Results:

• Identified typical mistakes in single-criterion (ignored wrong processes) and multi-criteria (overestimated uncertainty) assessments.
• Revealed five types of posterior probability distributions for model parameters, each linked to specific error types.
• Successfully identified mismatched parameters, obsolete parameters, flawed structures, and wrong process representations.

Conclusions:

• Combining single and multi-criteria assessment is crucial for uncovering hidden errors in environmental models.
• The presented method aids in diagnosing model deficiencies, including parameter and structural issues.
• Recommends integrating diverse observations (physical, chemical, biological, ecological) for thorough model assessment.