Determining effective placement strategies for denitrifying bioreactors in New Zealand river systems
Haddon Smith1, Thomas Graham-Murdoch1, Richard Clarke1, Naresh Singhal2, Alys Clarke3, Vinod Suresh3, Andrew Mason1, and Stephen Waite 1
1. Department of Engineering Science, University of Auckland
2. Department of Civil and Environmental Engineering, University of Auckland
3. Auckland Bioengineering Institute, University of Auckland
The state of New Zealand’s waterways is a subject of growing concern, due to decreasing trends in a range of water quality measures. In particular, the nitrogen species nitrate shows increasing concentrations at monitored sites within agricultural regions. This has prompted research into how nitrate can be removed from waterways, with the use of denitrifying bioreactors being one promising method. Aside from the design of such bioreactors, the problem of determining the most effective spatial placement within river networks is one research avenue that can help maximize the impact of bioreactor installations. The aim of this project was to use optimization techniques alongside water quality modelling theory to find effective placements solutions. A number of methods were developed so the advantages and disadvantage of different approaches could be evaluated. A mixed integer program (MIP), solved using commercial solver Gurobi, was shown to be effective. A greedy sequential heuristic algorithm was also implemented with successful results. A dynamic programming method was used, allowing the use of non-linear objective functions, however it was shown to be limited in its ability to scale to large networks. The placement strategies resulting from optimisation are shown to give additional benefit to water quality compared to random or intuitively generated solutions, and give a number of insights into how bioreactors should be placed. Firstly, placements in high concentration rivers are more effective that placements in lower concentrations areas. Secondly, placing units high up in river networks is shown to be advantageous to overall river health. Thirdly, the performance of the greedy sequential heuristic indicates that there is a low level of dependence between the effectiveness placement locations. Fourthly, a small number of large installations is shown to be more effective than a large number of small installations in some cases. Following on from this, the robustness of solutions to seasonal variation and variations in bioreactor extraction rates was studied. Furthermore, a number of insights were gained into what characteristics designers of bioreactors should try to achieve, such as the extraction rates necessary for different levels of impact on water quality. Lastly, the value of using optimization in choosing the placement of treatment facilities in rivers systems is demonstrated, through the use of the MIP model to find placement strategies designed to meet water quality improvement aims published by the Waikato River Authority.
This presentation is eligible for the ORSNZ Young Practitioners Prize.