A benchmark methodology for managing uncertainties in urban runoff quality models

Area of Science:

• Environmental Engineering
• Computational Hydrology
• Statistical Modeling

Background:

• Urban runoff quality models are crucial for environmental management.
• Comparing and validating these models presents significant challenges.
• Existing methods may not fully capture parameter uncertainties and interactions.

Purpose of the Study:

• To present a novel benchmarking methodology for urban runoff quality models.
• To utilize Bayesian theory and Markov Chain Monte Carlo (MCMC) methods for model comparison.
• To quantitatively assess parameter uncertainties and sensitivities in urban runoff models.

Main Methods:

• Developed a benchmarking methodology based on Bayesian inference.
• Employed the Metropolis algorithm, a Markov Chain Monte Carlo (MCMC) sampling technique.
• Applied the methodology to four configurations of pollutant accumulation/erosion models across four street subcatchments.

Main Results:

• The Metropolis algorithm provided reliable inferences of model parameters.
• Quantitative assessments of parameter uncertainties and their interaction structures were achieved.
• The methodology effectively demonstrated sensitivity of model output distributions to parameters.

Conclusions:

• The proposed Bayesian benchmarking methodology is effective and efficient for comparing urban runoff quality models.
• Reliable parameter inference aids in improving the mathematical conceptualization of model equations.
• This approach enhances the understanding and application of urban runoff quality models in environmental management.