Estimation of parameter uncertainty for an activated sludge model using Bayesian inference: a comparison with the...

Area of Science:

• Environmental Engineering
• Biotechnology
• Mathematical Modeling

Background:

• Activated sludge models (ASMs) commonly use frequentist inference for parameter assessment and uncertainty estimation via covariance matrices.
• Alternative Bayesian inference methods, treating parameters as probability distributions, offer potential advantages for ASM analysis.

Purpose of the Study:

• To apply and compare Bayesian and frequentist inference methods for uncertainty assessment in an ASM-type model.
• To evaluate the performance of both methods for a model incorporating intracellular storage and biomass growth.

Main Methods:

• Practical identifiability was assessed using respirometric profiles (oxygen uptake rate) and probabilistic global sensitivity analysis.
• Parameter uncertainty was estimated using both Bayesian and frequentist inferential procedures.
• A comparative analysis of the strengths and weaknesses of each approach was conducted.

Main Results:

• Bayesian inference was found to be a more general methodology, encompassing frequentist approaches under specific conditions.
• The study demonstrated the practical identifiability of the ASM model parameters using the chosen methods.
• Comparative results highlighted the distinct outcomes and insights provided by each inferential procedure.

Conclusions:

• Bayesian inference is a more versatile and comprehensive approach for addressing inferential challenges in activated sludge modeling.
• The study encourages the adoption of Bayesian inference for robust uncertainty quantification in ASMs.
• Understanding the relationship between Bayesian and frequentist methods provides a foundation for advanced modeling techniques.