Bayesian Model Averaging Approach for Urban Drainage Water Quality Modelling

EasyChair Preprint 11219

Abstract

The uncertainty in urban drainage water quality modelling is highly relevant in any practical application. Several models are available in the literature for such tasks, and one of the most challenging choices is selecting the most appropriate approach for the specific application. The Bayesian Model Averaging approach attempts to support the modeller in such choices by providing a method to identify and select the best-performing models and average their output response to reduce the related uncertainty. This method dates back to studies conducted by Bates and Granger (1969), whose analysis used model-averaging techniques in economic forecasting and determined that a pooled forecast of competing models outperformed any single model’s prediction. Techniques such as equal weight, Granger-Ramanathan averaging, and Bates-Granger averaging linearly combine the deterministic model outputs into another single-point deterministic forecast. This research applies the Bayesian Model Averaging to an actual catchment and is compared with several single water quality models. The analysis showed that the Bayesian Model Averaging approach outperformed all single-model applications.

Keyphrases: Bayesian approach, Water quality modelling, modelling average technique, uncertainty analysis, urban drainage

@booklet{EasyChair:11219,
  author    = {Gabriele Freni and Mariacrocetta Sambito and Stefania Piazza},
  title     = {Bayesian Model Averaging Approach for Urban Drainage Water Quality Modelling},
  howpublished = {EasyChair Preprint 11219},
  year      = {EasyChair, 2023}}