Understanding Complex Flow Pathways within Lab-Scale Denitrifying Bioreactors with a Conservative Tracer

Highlights Storm size and media age did not significantly impact removal rate based on simulated storm events in lab bioreactors. Post-hoc cluster analysis of bromide tracer time series identified distributed and preferential flow patterns. Distributed flow cluster had significantly higher removal rate and removal efficiency than preferential cluster. Abstract. Denitrifying bioreactors are designed to reduce excess nitrate (NO 3 - ) pollution from agricultural fields. During storm-induced flow events, flow rate and hydraulic retention time (HRT) can fluctuate widely, which may disturb denitrification within bioreactors in ways not captured by current steady-flow models of NO 3 - removal rate (RR). This study investigated RR and removal efficiency (RE) during storm flows of variable size and duration to close the gap between existing steady-flow models and real-world flow event conditions. Three simulated storm flow events were run through six lab bioreactors, and RR and RE were calculated during and following each event. Our results show that bioreactors were able to attenuate flow such that event size and duration were not significant explanatory variables of RR. A k-means cluster analysis on characteristics of the outflow bromide tracer time series for each bioreactor during each event identified two major flow patterns. The cluster exhibiting a more distributed bromide load through time had a significantly higher average RR than the cluster with a bromide load time series more characteristic of preferential flow (24.8 and 22.0 g N m -3 d -1 , respectively, p = 0.01). Bioreactors did not consistently display a given flow pattern but often changed between events. This suggests that flow patterns within lab bioreactors for each event are a more significant driver of RR than the external factor of the inflow hydrograph. Keywords: Denitrifying bioreactor, k-Means clustering, Stormwater, Tracer.