Denitrifying bioreactors that target filtration of nitrate from farm drainage water are gaining recognition as a tool for reducing nitrate pollution from agricultural areas. While the hydrologic regime and nitrate concentrations constitute two fundamental environmental variables that determine the size of a denitrifying bioreactor, the issue of over- or under-treatment of water that might otherwise promote undesirable pollution swapping phenomena and construction costs also need to be factored into the overall design process. Conventional methods for optimizing the design of denitrifying bioreactors generally rely on deterministic models, even though many of the design parameters are not known with confidence.
In this work we demonstrate how stochastic optimization can be used to enhance the long term performance of a groundwater fed, woodchip denitrifying bioreactor. We apply an alternative design philosophy and demonstrate how the bioreactor design process can be improved through application of stochastic methods. The design aspect of an ‘in-stream’ bioreactor planned for installation on a farm in New Zealand is structured as a multi-objective performance optimization problem that is solved in a stochastic framework, using freely available open source tools. Uncertainty considerations regarding values of physical parameters that govern bioreactor performance are incorporated into the assessment, from which a Pareto set of optimal designs was obtained. A 75 m long bioreactor of 1.5 m height was selected as the preferred choice from the optimal set of design solutions. Assuming a 10-year operational life, it is predicted that the cost of nitrate removal by the planned denitrifying bioreactor will be NZ$9.70/kg-N (approx. AU$9.20/kg-N).