Uncertainty in Modeling Pollutant Build-up and Wash-off Processes at the Urban Catchment Scale

Zhaokai Dong, Pradeep Goel, and Clare Robinson

ABSTRACT

Build-up/wash-off models are widely used to predict stormwater pollutant loads across various land uses, establishing a “baseline” for evaluating potential stormwater management scenarios. At the urban catchment scale, these models are typically calibrated using data collected at the outlet. The resulting parameter values are then interpreted as representative of build-up and wash-off processes across multiple distinct land uses. However, the “lumped” calibration approach – fitting model output to outlet data by adjusting parameter values – may not yield a unique or representative set of parameter values for distinct land uses. This calibration uncertainty may lead to biased pollutant load predictions across land uses, even when the model output matches well the outlet data. In this study, we used the Generalized Likelihood Uncertainty Evaluation (GLUE) approach to investigate parameter variability and prediction uncertainty for a stormwater quality model simulating total suspended solids, total phosphorus, and soluble reactive phosphorus loads in a 240-ha urban catchment in London, ON, Canada. Our results suggest that while the model effectively replicated observed pollutant loads at the outlet, the “optimal” parameter values varied across events. Furthermore, parameter likelihood distributions showed seasonal patterns, indicating the influence of seasonal variability on parameterization. Additional monitoring data did not reduce model uncertainty, suggesting that uncertainty is driven primarily by limitations in the model structure and the “lumped” calibration approach rather than by data availability. These findings imply that, for event-based load simulations, the model represents just one possible fit between inputs and outputs rather than providing precise predictions based on a representative set of calibrated parameters. Transitioning from a “lumped” to a “distributed” land-use-based calibration approach may improve parameter reliability but would require more extensive data collection across multiple distinct land uses.

 Click here to download a static PDF version of the presentation.


Permanent link: