Predicting Dissolved Oxygen Concentration in Urban Watersheds: A Comparison of Fuzzy Number Based and Bayesian Data-Driven Approaches
AbstractDissolved Oxygen (DO) concentration is significantly and adversely impacted by urbanisation. However, the processes that govern DO concentration in riverine environments are complex, difficult to understand and to model. In the Bow River in Calgary, AB, Canada there is a need to predict DO concentration as part of the overall goal to improve river health for downstream users. In this research three data-driven models are used to predict DO concentration in the Bow River are compared. Fuzzy linear regression performed better than both simple linear regression and Bayesian linear regression. The fuzzy number based model was better able to capture the daily variability in DO in the Bow River and substantially improved low DO predictions.
Environmental Modeling, Risk Assessment and Decision Making (EMR)