Automatic model calibration of combined hydrologic, hydraulic and stormwater quality models using approximate...

Area of Science:

• Environmental Engineering
• Hydrology
• Water Resource Management

Background:

• Automatic model calibration is crucial in water engineering but often hindered by complex likelihood function evaluations.
• Existing methods face challenges in accurately modeling urban hydrologic, hydraulic, and stormwater quality processes.

Purpose of the Study:

• To present an innovative Approximate Bayesian Computation (ABC) framework for automatic model calibration.
• To address the limitations of traditional calibration techniques, particularly the evaluation of likelihood functions.
• To apply ABC to a combined urban hydrologic, hydraulic, and stormwater quality model.

Main Methods:

• Developed a novel calibration framework utilizing the Approximate Bayesian Computation (ABC) technique.
• Applied the ABC framework to a complex, multi-input-output urban water model.
• Validated the model's performance against observed data from three catchments and commercial software (MIKE URBAN).

Main Results:

• The ABC-based model successfully simulated runoff hydrographs and total suspended solid (TSS) pollutographs within 95% confidence intervals.
• Model validation demonstrated strong performance, meeting statistical criteria like coefficient of determination (CD), root mean square error (RMSE), and maximum error (ME).
• The calibrated model showed good agreement when compared with MIKE URBAN, a leading commercial modeling software.

Conclusions:

• The developed ABC framework offers an effective solution for automatic calibration and uncertainty estimation in complex water models.
• This approach overcomes the computational burden associated with likelihood function evaluation in traditional methods.
• The framework is highly applicable to advanced hydrologic, hydraulic, and stormwater quality models with multiple inputs and outputs.