Bayesian-based calibration for water quality model parameters.

Area of Science:

• Environmental Science
• Water Quality Modeling
• Computational Hydrology

Background:

• Traditional water quality model calibration methods often struggle with local optima and parameter equifinality, limiting accuracy and efficiency.
• Accurate parameter calibration is crucial for reliable water quality predictions and effective environmental management.

Purpose of the Study:

• To propose a novel Bayesian-based method for water quality model parameter calibration.
• To improve calibration efficiency and accuracy, avoiding local optima and parameter equifinality.
• To enhance the convergence speed of the calibration process.

Main Methods:

• The parameter calibration problem was reformulated as a posterior probability function sampling problem.
• Bayesian inference and the Markov Chain Monte Carlo (MCMC) algorithm were employed for sampling.
• Optimized initial sampling values were set to improve MCMC convergence speed.

Main Results:

• The proposed Bayesian method achieved a mean relative error (MRE) of less than 10% for parameter calibration.
• Specific MREs for D x and D y were 5.3% and 8.3%, respectively.
• Calibration efficiency was improved by optimizing initial MCMC values, Markov chain length, and proposal distribution.

Conclusions:

• The Bayesian-based method offers an efficient and accurate approach for water quality model parameter calibration.
• Optimized MCMC initial values significantly enhance calibration efficiency.
• The method demonstrates wide applicability across various water quality models.