Research papersHighway runoff quality models for the protection of environmentally sensitive areas

- Compiled 940 monitored highway runoff events from 14 sites located in 5 countries.
- Developed novel highway runoff quality models using artificial neural networks.
- Inclusion of a seasonal term within the ANNs improved the prediction accuracy.
- ANNs trained to directly predict heavy metals were more accurate using surrogate.
- Developed a novel framework for the design of roadside ditch treatment systems.

This paper presents novel highway runoff quality models using artificial neural networks (ANN) which take into account site-specific highway traffic and seasonal storm event meteorological factors to predict the event mean concentration (EMC) statistics and mean daily unit area load (MDUAL) statistics of common highway pollutants for the design of roadside ditch treatment systems (RDTS) to protect sensitive receiving environs. A dataset of 940 monitored highway runoff events from fourteen sites located in five countries (Canada, USA, Australia, New Zealand, and China) was compiled and used to develop ANN models for the prediction of highway runoff suspended solids (TSS) seasonal EMC statistical distribution parameters, as well as the MDUAL statistics for four different heavy metal species (Cu, Zn, Cr and Pb). TSS EMCs are needed to estimate the minimum required removal efficiency of the RDTS needed in order to improve highway runoff quality to meet applicable standards and MDUALs are needed to calculate the minimum required capacity of the RDTS to ensure performance longevity.