A Bayesian modeling framework to predict stormwater pollutant reduction in bioretention media

Abstract

Bioretention is a green stormwater infrastructure practice that has been shown to reduce some of the harmful effects of urban stormwater on downstream receiving waters. Typically, media used for bioretention consist of high percentages of sand to realize high hydraulic conductivities and smaller percentages of soil or organic materials to improve the retention of stormwater pollutants. The objectives of this study were to evaluate four bioretention media mixes that differed in sand, compost, and water treatment residuals content and to predict pollutant removal based on realistic stormwater concentrations using Bayesian statistics. Pollutants of interest were chemical oxygen demand, ammonia, nitrite, nitrate, total Kjeldahl nitrogen, as well as total and dissolved forms of phosphorus, copper, lead, and zinc. Water quality data were collected for eleven storms, and cumulative rainfall after construction was used as a measure for the media age. The data were evaluated for the probability to observe pollutant reduction and for the probability to observe high-performance reduction (HPR) with a minimum pollutant reduction of 60%. The chemical oxygen demand had a probability for reduction of P = 0.09 and HPR of P = 0.00, resulting in pollutant export in the majority of storms. HPR reduction goals were met for total lead, zinc, and copper, and reduction probabilities were generally P > 0.5 for all total and dissolved metals. Age and media composition affected phosphorus and nitrogen removal, with increased pollutant retention observed as the media aged. Lower compost percentages resulted in less nutrient export. This probabilistic model of realistic stormwater concentrations in combination with bioretention pollutant removal experiments results in a holistic view of credible pollutant outflow concentrations that can be used by stormwater managers to alleviate the effects of stormwater on surface water quality.