Design approach tradeoffs and annual performance prediction for denitrifying bioreactors treating subsurface drainage

Long-term empirical assessments of nutrient treatment technologies for agricultural outflows underpin the ability to accurately design, model, and implement such practices for water quality improvement. The conservation practice standard for denitrifying 'woodchip' bioreactors in the USA has existed for a decade but the performance of this standard has not been comprehensively evaluated. Annual nitrate-nitrogen (N) removal was monitored at ten full-size bioreactors in Illinois, USA that included bioreactors designed according to the practice standard or designed to increase treatment loading (n = 22 and 14 site-years, respectively). Three modeling approaches (correlation; a priori; Bayesian model averaging) were used to deepen understanding of correlative and predictive factors for a range of performance metrics. Across the dataset, which is the largest compilation of full-size denitrifying bioreactor performance to date, the bioreactors provided a median edge-of-field N removal efficiency of 17 % and N load reduction of 49 kg N per bioreactor per year. As annual N loading from the drainage areas increased, the edge-of-field N removal efficiency decreased (Spearman correlation coefficient ρ: 0.67) whereas annual N load removed by the bioreactors increased (ρ: +0.63). A bioreactor design paradigm focused on maximizing the annual N load removed would require an intentional willingness to define performance beyond the efficiency-based metrics that are currently used in the practice standard. Overall, these ten bioreactors provided N removal across all years which reiterates this practice's dependability and demonstrates possibilities for further design optimization.