Assessing wastewater micropollutant loads with approximate Bayesian computations.

Area of Science:

• Environmental science
• Stochastic modeling
• Statistical inference

Background:

• Wastewater production is inherently stochastic, necessitating complex models for predicting chemical and pharmaceutical loads.
• Formal statistical inference methods are often lacking for these models due to intractable likelihood functions.

Purpose of the Study:

• To investigate the application of Approximate Bayesian Computation (ABC) methods for parameter inference in stochastic environmental models.
• To assess the performance of ABC algorithms in water quality modeling and environmental systems analysis.
• To infer the number of wastewater pulses using high-resolution sewer data.

Main Methods:

• Utilized Approximate Bayesian Computation (ABC) for Bayesian inference with computationally intractable likelihoods.
• Applied three different ABC algorithms to analyze benzotriazole and total nitrogen loads in sewer systems.
• Investigated the sensitivity of results to substance characteristics and catchment properties.

Main Results:

• All tested ABC algorithms performed well in inferring the number of wastewater pulses.
• Uncertainty in the inferred number of wastewater pulses ranged from 6% to 28%.
• Results demonstrated higher sensitivity to substance characteristics than to catchment properties.

Conclusions:

• ABC methods provide a powerful and applicable approach for parameter updating in stochastic environmental models.
• These methods can enhance the predictive capabilities of water quality models using monitoring data.
• Further research and careful tuning are recommended for optimal application of ABC in environmental analysis.