Stochastic simulation of partially-mixed, event-driven treatment wetlands

Abstract

Simulated rain events and runoff produce a more realistic influent to a stormwater treatment wetland than do an averaged inlet flow and concentration. Network flow models provide a more realistic prediction of a treatment wetland’s internal flows than does a plug flow or complete mixing estimation. Therefore, a combination of simulated rain events and a partial mixing model was chosen to produce realistic and descriptive predictions of treatment wetland function. The internal flow and mixing of the wetland was modeled using the network model ZDM. First-order pollutant removal was added to the hydraulic model, and the combination was calibrated to experimental nitrate data from Des Plaines River Wetland Demonstration Project wetland EWC1. Rainfall probability distributions were used in Monte Carlo simulations, and coupled with simple pollutant deposition and runoff from a watershed, to provide varying concentrations and flows into a wetland. The model was then combined with the variable rain and inlet flows produced by the Monte Carlo simulation, to provide predictions of the expected performance of a treatment wetland for nitrate reduction. The results of the simulation contain more detail than a time averaged model, which predicts only flowrate and concentration. The simulations produced such averages, but they also produced cumulative frequency distributions of anticipated exit concentrations. Using the ZDM model, an equation was developed to estimate the average conversion provided by a treatment wetland.