Biofilter Performance Research Articles

Start-Up and Performance of a Full Scale Passive System In-Cluding Biofilters for the Treatment of Fe, as and Mn in a Neutral Mine Drainage

Passive mine drainage treatment plants are the scene of many chemical and biological reactions. Here, the establishment of iron (Fe), arsenic (As), and manganese (Mn) removal was monitored immediately after the commissioning of the Lopérec (Brittany, France) passive water treatment plant, composed of aeration cascades and settling ponds followed by pozzolana biofilters. Iron and As were almost completely removed immediately after commissioning, while Mn removal took more than 28 days to reach its maximum performance. Investigations were performed during two periods presenting strong variations in feeding flow-rates: from 2.8 m3.h−1 to 8.6 m3.h−1 and from 13.2 m3.h−1 to 31.3 m3.h−1. Design flow rate was reached during the first week of the second period. Dissolved Fe and As were not affected by the decrease in residence time while Mn was only slightly affected. Microbial communities in biofilter presented similarities with those of the pond sludge, and genera including Mn-oxidizing species were detected. Proportion of bacteria carrying the aioA gene encoding for As(III)-oxidase enzyme increased in communities during the second period. Results suggest Mn removal is mainly associated with bio-oxidation whereas removal of Fe and As could be mainly attributed to chemical oxidation and precipitation of Fe, possibly helped by As(III) bio-oxidation.

Quorum sensing improved the low-temperature performance of biofilters treating gaseous toluene

Biofilters usually have poor VOC removal performance at temperatures lower than 20 °C. In this study, two quorum sensing (QS) enhancement methods, which are addition of exogenous N-acyl-homoserine lactones (AHLs) and inoculation of AHL-producing bacteria, were applied in biofilters treating gaseous toluene at a low temperature of 12 °C. Their effects on biofilter performance and biofilm characteristics were investigated. The results showed that adding exogenous AHLs and AHL-producing bacteria in biofilters raised the average toluene elimination capacity by 39% and 26% respectively, and raised the average mineralization efficiency by 25% and 47% respectively in first 24 days. In addition, the two QS enhancement methods could increase the attached biomass by 48% and 87% respectively and made the biofilm distribute more uniform by increasing its extracellular polymeric substances content and microbial adhesive strength. The two QS enhancement methods resulted in more mesopores in biofilm, lower O/C and (O+N)/C of organic elements in biofilm, and increased the solubility of toluene in liquid phase, which all benefit VOCs mass transfer in biofilters. These results demonstrate that QS enhancement methods have the potential to optimize the biofilm and thus improve the performance of biofilters treating VOCs at a low temperature. This work provides us a new choice to improve industrial-scale biofilters treating VOCs at high latitude regions or in winter.

Evaluation of bioscrubber and biofilter technologies treating wastewater foul air by a new approach of using odor character, odor intensity, and chemical analyses

The treatment of raw foul air that could escape to the atmosphere from the head space of the incoming wastewater into a Southern California Water Resource Recovery Facility was evaluated by using a 1/20th scale pilot unit consisting of five different biological media technologies, operating side by side, under different operating conditions. The removal of six different odor characters from eight chemical odorants present in the foul air were assessed. These were rotten egg (Hydrogen Sulfide), rotten vegetables (Methyl Mercaptan), canned corn (Dimethyl Sulfide), rotten garlic (Dimethyl Disulfide), earthy/musty (2-Methyl Isoborneol and 2-Isopropyl 3-Methyl Pyrazine) and fecal (Skatole and Indole). This is the first time a study evaluates specific odors by simultaneously employing sensory analyses using the Odor Profile Method, which defines the different odor characters and intensities, together with chemical analyses of the compounds causing these odors, known as odorants.The paper discusses the efficiencies in removing odor characters as well as odorants by two different bioscrubbers (reticulated polyurethane cube foam and polypropylene mesh with layered polyester foam) and three different biofilters (engineered media, seashells, and lava rock).The results show that the two bioscrubbers, even with greater empty bed gas retention times, did not provide significant improvement in odor intensity and odorant removal. However, the biofilters showed that larger empty bed gas retention times provided significant improvements in diminishing the odor intensities and better odorant removal.The biofilter with lava rock media at 45 s empty bed gas retention time provided the best treatment among the technologies tested, achieving the following odorant reductions: 99.8% for hydrogen sulfide, 98.4% for methyl mercaptan, 57.0% for dimethyl sulfide, and 52.7 for dimethyl disulfide. This biofilter also achieved the following odor intensity reductions: 47% for rotten vegetable odors, 50% for earthy/musty odors, and 100% for fecal odors. The odor panel detected odors by the Odor Profile Method that were below the detection limit of the corresponding chemical analytical method for specific chemical compounds causing these odors.Differences were observed between the performances of bioscrubbers and biofilters, based on odorant removal compared to those based on sensorial analyses, indicating that both analyses are required to understand more fully the odor dynamics. Furthermore, a total odor removal of 99.2% was observed by the dilution to threshold olfactometer method even though nearly half of the rotten vegetable and earthy/musty odors remained based upon the Odor Profile Method. This shows the olfactometer method did not correctly define the degree of odor nuisance in the foul air in this study.Bioscrubbers have in general a better economic return when used at low EBGRTs and as preliminary (first stage) treatment systems. Biofilters are more effective when used at high EBGRTs and can be used as stand-alone or polishing systems.

Application of co-immobilized microbial biochar beads in hybrid biofilter towards effective treatment of chrome tanning wastewater

This paper aims to scrutinize the comparative performance of hybrid biofilter via column mode to boost the reduction rate of Cr (VI) and organic pollutants in various pre-treated tannery wastewater. The pre-treated water through aluminium formate, aluminium sulphate, and ferric chloride were passed through two divergent biofilter medium discretely with (i) immobilized sodium alginate microbial beads (Im/S.AM.B) and (ii) co-immobilized microbial biochar beads (Co-im/B.M.B). Biosorption efficiency of the hybrid biofilter was examined with respect to pH, initial adsorbate concentration, time, flow rate, breakthrough curve, and regeneration studies. The results exhibited that aluminium formate coagulant pre-treated wastewater when passed through Im/S.AM.B under optimized conditions, the reduction rate of Cr (VI), COD and sulphate was 92.5%, 89.5% and 75%, respectively, whereas Co-im/B.M.B surpassed it by 6.7%, 6.5% and 19% respectively, with a contact time of 90 min, flow rate 2 mL/min, depth 30 cm and pH 6. Further, the raw and treated sample of Co-im/B.M.B was studied for its morphology and surface chemistry via FE-SEM, EDX, FTIR and XRD analysis. The results revealed that Cr (VI) reduction is predominantly through precipitation, electrostatic attraction, and metal complexation. It was observed that electrostatic attraction of cationic charged biochar surface to Cr (VI) ions results in the reduction of Cr (VI) to Cr (III) and complexation between Cr (III). Further, the functional groups present on the microbial biomass (DPAML065) surface were also found responsible for Cr (VI) reduction. Thus, Co-im/B.M.B in hybrid biofilter presents an economical approach in reducing Cr (VI) from tannery wastewater.

A mixed agro-waste based biofilter for the removal of methyl ethyl ketone: Kinetics and modeling

Biofiltration has grown worldwide, as an inexpensive and consistent air pollution control technology for the elimination of numerous organic compounds. In this work, the biofiltration of methyl ethyl ketone (MEK) has been studied. The agro-wastes namely pressmud and cornstack are mixed and used as packing materials for the removal of MEK. The performance of the biofilter has been explored over 200 days. Experiments are carried out in four periods. In every period, the biofiltration of pollutants is performed for various inlet concentrations (IC) of MEK (0.2–1.2 g m−3) and gas flow rates (GFR) (0.03 m3 h−1 to 0.12 m3 h−1). The biofilter performances are investigated by their removal rates, elimination capacity, temperature, pressure drop, carbon dioxide production, and variation of bed height at the time of operations. For the IC of 0.2 ± 10% g m−3 and GFR of 0.03 m3 h−1, the removal efficiency (RE) of MEK is 97% using the mixture of cornstack (80%) and pressmud (20%). The MEK removal efficiency decreases, when the cornstack ratio increases. The maximum RE is obtained at the IC of 1.2 g m−3 and GFR of 0.12 m3 h−1 for all the tested conditions. The pressure drop (PD) is also measured for all the operating conditions. It is found that the PD rises with an increase in height. The temperature is monitored throughout the biofiltration process. It is observed that an increase in RE leads to a rise in temperature in the biofilter. SEM images confirm the growth of microbes on the packing material. The kinetics of removal of MEK is investigated by using Ottengraf–van den Oever model.

Acute dose-response exposure of a peracetic acid-based disinfectant to Atlantic salmon parr reared in recirculating aquaculture systems

There is a high regard for peracetic acid (PAA)-based disinfectants in recirculating aquaculture systems (RAS) because of the low risk of bioaccumulation, fast degradation with neutral residuals and minimal impact on biofilter performance. However, the no-observed-effect concentration in Atlantic salmon parr is unknown. The present study evaluated the effect of an acute PAA exposure on Atlantic salmon parr health and welfare by evaluating survival, swimming behaviour, appetite and histopathological alterations in the gills and skin. Nine experimental RAS units were employed, where each unit was dedicated for one PAA concentration (0.0, 0.05, 0.1, 0.2, 0.4, 0.8, 1.6, 3.2 and 6.4 mg/L). Fish were exposed to the target PAA concentration in a static system for 1 h and the exposure protocol was repeated after a 52 h recovery. The fish survival was 100%, 80% and 0%, respectively ≤1.6, 3.2 and 6.4 mg/L. Fish swimming behaviour was normal in PAA ≤ 1.6 mg/L, whereas it become erratic with air gasping for PAA ≥ 3.2 mg/L. The fish appetite did not change among the different PAA treatment groups. Skin and gill histopathological alterations were pronounced in PAA ≥ 3.2 mg/L, characterized by a poorer skin condition and necrotic gill lamella. The skin acidic mucous cells density was 55% lower in the 6.4 mg/L group than the 0 mg/L group. The sub-lethal water pH values observed in the 6.4 mg/L group after PAA administration may have played a confounding and compounding factor to the PAA toxicity response in this group. In conclusion, the current study identified the no-observed-effect concentration for PAA to be below 1.6 mg/L for Atlantic salmon parr and provided insights into its use as a water prophylactic strategy in RAS. Toxicity of PAA-based disinfectants is influenced by its acidified nature, which can interfere with the water pH of low alkalinity aquaculture systems. Further studies should evaluate the health and welfare consequences of a long-term PAA exposure in Atlantic salmon parr.

The role of ammonia oxidizing microorganisms in biofiltration for the removal of trace organic compounds in secondary wastewater effluent

Although nitrifying microorganisms play an important role in TOrC biotransformation, ammonia-rich environments appear to hinder important cometabolic processes, whereas low-dose monochloramine has minimal impact on biofilter performance.

The regenerative role of biofilm in the removal of pesticides from stormwater in biochar-amended biofilters

We studied the impact of a microbial biofilm on biochar-amended biofilter performance and lifetime, and the potential for leaching of pesticide transformation products.

Enhancing recovery performance of the toluene-removing biofilter after the short/long interference-shutdown period

In this study, the feasibility of three methods on enhancing the recovery performance of biofilter after the interference and starvation periods was evaluated. Results show that despite the pressure drop risk, supplementation of 7.5% (w/v) Polyethylene glycol-600 (PEG-600) resulted in quick recovery on removal efficiency in both short- and long-term interference shutdown experiments. Tinidazole Tablets (2 mg/L), a Bacteroidetes-specific antibiotic, are more suitable to apply as a one-time shot to improve recovery of biofilter as the second dose of Tinidazole Tablets was no longer effective presumably caused by the increased drug resistance. It is worth noting that the maximum elimination capacity of 134 g/(m3·h) was observed with Pseudomonas putida (P. putida) BRJC1032 addition. The biodegradation kinetic, biological characteristics and microbial community evolution in biofilters were systematically analyzed for finding the suitable methods to enhance recovery performance, which is of great value for the further industrial application of the biofilter technology.

Laboratory Media Test Comparisons to Long-Term Performance of Biofilter and Media Filter Treatment-Train Stormwater Controls

AbstractStormwater biofilters and bioretention controls have been extensively studied and are commonly encouraged as effective stormwater controls. Most of these controls incorporate media to enhan...

Two quorum sensing enhancement methods optimized the biofilm of biofilters treating gaseous chlorobenzene

In this study, effects of two quorum sensing (QS) enhancement methods on the performance and biofilm of biofilters treating chlorobenzene were investigated. Three biofilters were set up with BF1 as a control, BF2 added exogenous N-acyl-homoserine lactones (AHLs) and BF3 inoculated AHLs-producing bacterium identified as Acinetobacter. The average chlorobenzene elimination capacities were 73 and 77 g/m3/h for BF2 and BF3 respectively, which were significantly higher than 50 g/m3/h for BF1. The wet biomass of BF2 and BF3 with QS enhancement eventually increased to 60 and 39 kg/m3 respectively, and it was 29 kg/m3 for BF1. Analysis on biofilms in three biofilters showed that distribution uniformity, extracellular polymeric substances production, adhesive strengths, viability, and metabolic capacity of biofilms were all prompted by the two QS enhancement methods. Comparisons between the two QS enhancement methods showed that adding exogenous AHLs had more significant enhancing effect on biofilm due to its higher AHLs level in start-up period, while AHLs-producing bacteria had an advantage in enhancing bacterial community diversity. These results demonstrate that QS enhancement methods have the potential to optimize the biofilm and thus improve the performance of biofilters treating recalcitrant VOCs.

Efficiencies of indigenous South African plant biofilters for urban stormwater runoff water quality improvement with a focus on nutrients and metals

Abstract In South Africa, urban activities contribute high levels of pollution to rivers and groundwater via stormwater runoff. In reducing urban stormwater loads of engineered plant biofiltration, an effective and self-sustaining component of green infrastructure is a treatment option. The country's extensive natural biodiversity offers untapped potential of indigenous species' use in plant biofilters. This paper presents the findings of a plant biofilter column experiment, which investigated the performance of nine indigenous plant species under varied urban stormwater pollutant load strengths. Average significant loads of dissolved Cd (>98%), Cu (>84%), Pb (>99%) and Zn (>95%), as well as NH3-N (>93%), were removed by the plant biofilters, whereas the removal of -N (−37 to 79%) and -P (−81 to 63%) was more variable. Biofilters equipped with indigenous plant species were on average at least 11% more efficient than unvegetated soil in the removal of urban nutrient and metal pollutants. Over time, planted biofilters improved nutrient and metal removal efficiencies. The results support the inclusion of indigenous plants in biofilters within urban stormwater green infrastructure initiatives. Further research to inform plant biofilter design practicalities and assess plant biofilter performance in the field is warranted.

Evaluation of enzyme activity for monitoring biofiltration performance in drinking water treatment

Many water providers monitor adenosine triphosphate (ATP) as an indicator of biological acclimation of their biofilters; however, strong correlations between ATP concentration and filter performance (e.g., organic matter or disinfection by-product precursor removal) are not typically observed. As an alternative, this study evaluated the use of enzyme activity for monitoring biological processes within filters. Recent studies have proposed that enzyme activity may be used as an indicator of biofilter function as it provides a means to quantify biodegradation which may allow for a more accurate measure of degradation potential and to gain a better understanding of biofilter performance. Sampling was completed from full- and pilot-scale biofilters to assess impacts associated with pre-treatments, varying sources waters, as well as pre-treatment and operating conditions. Enzyme activity (carboxylic esterase, phosphatase, ß-glucosidase, α-glucosidase, ß-xylosidase, chitinase, and cellulase) and ATP were measured from the top 5 cm of biofilter media representative of typical full-scale sampling; water quality parameters included dissolved organic carbon (DOC) and disinfection by-products (DBPs): trihalomethane (THM) formation potential (FP), and haloacetic acid FP (HAA FP). Results confirmed that ATP was not a reliable monitoring tool for DOC and DBP FP reduction in biofilters. A strong relationship was observed between esterase activity and DOC reduction; chitinase activity significantly correlated to THM FP reduction for filters treating three different source waters and HAA FP reduction achieved by filters treating the same source water with a range of pre-treatment and backwash conditions. This study showed that enzyme activity may be appropriate for monitoring biological processes within drinking water filters and may act as a surrogate for the removal of organic compounds.

Biological responses to P‐limitation in indigenous bacteria isolated from drinking water

AbstractBiofilm formation in direct biofiltration causes operational issues such as accelerated head loss accumulation rate resulting in short filter runtime. Applying coagulant in low P level source water may impose bacteria to the P‐limited condition. It is hypothesized that the nutrient limitation, specifically P, as an essential macronutrient for bacteria causes stress and triggers excess extracellular polymeric substances (EPS). Two of the most biofilm‐producing bacteria were isolated from a full‐scale biofilter, provided the opportunity to study the indigenous biofilter bacteria. The isolated bacteria were identified using full‐length 16S rRNA and characterized. The biological behavior of the species was studied under different P‐limited conditions in a nutrient‐limited medium simulating freshwater nutrient availability. Carbohydrate and Protein‐EPS were increased when decreasing the available P in the medium, suggesting that lack of P can trigger higher EPS.Article Impact StatementThis study demonstrated that the lower P levels are directly related to excessive extracellular polymeric substances production and consequently the biofilter performance.

How well do stormwater green infrastructure respond to changing climatic conditions?

Urban green infrastructure (GI) such as biofiltration systems (also known as bioretention, biofilters, rain gardens) are being increasingly implemented in different parts of the world to enhance urban greening whilst controlling stormwater pollution. Treatment effectiveness has been shown to be reliant upon the surrounding climate, including rainfall patterns (e.g. length of wet and dry weather) and temperature. Plant species play a key role in treatment, yet, the response of different plant species to pollutant removal under different climatic and environmental conditions is still not fully known. This information is pertinent to ensure sustained biofilter performance under a range of conditions. This study investigates the change in biofilter nutrient and E. coli removal performance in the presence of four plant species (Canna indica, Carex appressa, Ginkgo biloba and Miscanthus sinensis) with distinct above and below ground characteristics, when exposed to wet weather, dry weather and cold weather conditions. The extent of the decline in nutrient treatment performance following extended drying and under cold conditions was found to be species-specific. Miscanthus sinensis was the best performer and was also found to be cold-resistant. With the exception of Miscanthus sinensis, all plant species were more impacted by the cold temperature (decrease from 70% to 45% TN removal on average; <1 to 30% decrease across species) compared to 4-weeks drying (decrease from 60% to 54% TN removal on average; <1–16% decrease across species) in systems equipped with a submerged zone. In contrast, the 4-weeks drying (E. coli decrease from 2.60 to 1.08 log removal on average) had a more pronounced effect on E. coli removal performance compared to the cold conditions (E. coli decrease from 1.70 to 1.38 log removal on average). This study shows that plant species relative contribution to pollutant removal is pollutant-specific and varies under different climatic conditions. A mix of plant species may ensure higher level of system resilience under a variable climate. The study also highlights the need to understand more about the microbial ecology of these plant systems to optimise both nutrient and pathogen removal for sustained long term performance.

Denitrification performance and mechanism of biofilter constructed with sulfur autotrophic denitrification composite filler in engineering application

Sulfur autotrophic denitrification (SAD) is a promising technology due to its low cost and low sludge production. Based on previous studies on SAD materials as well as the denitrification mechanism of SAD technology, this study constructed two biofilters with a sulfur autotrophic denitrification composite filler (SADCF) to investigate the application potential of SAD technology. The feasibility of a SADCF-based biofilter was demonstrated, with a maximum nitrate volume load of 0.75 kg N/(m3·d) and low accumulation of nitrite and ammonium. In addition, an improved backwashing method (air–water backwashing) was obtained by comparing two different backwashing methods. Furthermore, some iron reducing bacteria (0.4% Geothrix) along with a rapid proliferation of the main sulfur-oxidizing bacteria (23.0% Thiobacillus and 27.7% Ferritrophicum) were found under real-world operating conditions. Overall, the results of this study provide a case reference for the operation of SADCF-based biofilters and the application of SAD technology in engineering.

Metagenomic insights into the explanation of biofilter performance distinction induced by dissolved oxygen increment

In this study, a set of novel double-layer-packed sequencing batch biofilm reactors (SBBRs) were constructed to treat secondary effluent with a low carbon-to-nitrogen ratio (C/N = 3.0). All the reactors were operated under identical conditions except dissolved oxygen (DO) in the aerobic section, which was controlled at seven different levels (0.5, 1.0, 1.5, 2.0, 2.5, 3.5 and 4.5 mg/L). It was found that the aerobic section contributed most to performance at an optimal DO of 2.0 mg/L. The change in DO accounted for the varied bacterial communities, which resulted in different performances. Both bacterial community structure and potential functions displayed close associations with performance implying nutrients were mainly removed by simultaneous nitrification-endogenous denitrification and phosphorus removal (SNEDPR) process. Moreover, the proliferation of nonfunctional nitrifiers/denitrifiers led to the inhibition of denitrification resulting in a lower removal of total dissolved nitrogen and phosphorus.

Performance evaluation of a full-scale open bed biofilter through on-site measurements and CFD analyses

Composting process is applied to stabilize organic solid waste. But this method releases odorous organic compounds as by-product. Several methods are employed to cope with this nuisance problem. In this study, a full-scale biofilter unit was evaluated which is used in the treatment of composting waste gas. The performance of the open bed biofilter was evaluated by measurements volatile organic compounds (VOCs) and total organic carbon (TOC) from the inlet sampling ports and the outlet of the biofilter. Different sampling points were selected at the outlet of the biofilter in order to better represent removal efficiencies (RE) and determine the variability over the biofilter surface. The average TOC removal rate was 67 ± 11 %. Fifty-six different VOCs were sampled and then quantified by GC–MS. The average inlet concentrations for the target VOCs fed to the biofilter unit in November, January, April, May, and June were 57.9, 11.51, 4.54, 5.09, and 10.26 mg m−3, respectively and the removal efficiencies for these inlet concentrations were 45 %, 62 %, 32 %, 32 %, and 40 %, respectively. The deterioration of the packing material, caused reduced removal efficiencies by the time. The VOC removal rates were lower than TOC removal rates due to the selected VOC species. Further analyses were conducted with computational fluid dynamics (CFD). The effectiveness of the waste gas distribution was evaluated. Flow distribution within the biofilter was not uniform. This was also supported by the measurement results. Regions of the biofilter volume that are close to the inlet pipe had smaller retention times, which resulted in reduced treatment efficiencies.

Water quality and hydraulic performance of biochar amended biofilters for management of agricultural runoff

This research evaluated the effect of biochar amendment rate on nitrogen species and organic carbon removals and hydraulic performance in biofilter columns treating dairy farm runoff. Initial studies compared the performance of sand columns amended with two types of biochar with different specific surface area (SA) and cation exchange capacity (CEC) with an un-amended sand column. The results showed that biochar enhanced N-species removal due to its unique physicochemical properties. In subsequent tests, two biofilter columns with different biochar fractions (20% and 50% by volume) were operated at varying hydraulic loading rates and antecedent dry conditions. Total nitrogen, ammonia, organic nitrogen and organic carbon removals were significantly higher in the column with the higher biochar fraction. The high CEC of biochar increased ammonium retention during the application period, allowing for nitrification during the antecedent dry periods (ADPs) when aerobic conditions developed in the media pores. High biochar SA also resulted in greater retention of DON and DOC by adsorption. A variable saturation flow model of biochar amended biofiltration was developed using HYDRUS-1D software. The model was calibrated using data from conservative tracer and moisture content studies. Model results showed that the high microporous structure of the biochar increases the time needed to reach full saturation, lowers the saturated conductivity and increases the hydraulic retention time in the medium. This calibrated model can be used to design field scale biofilter systems for managing agricultural runoff.

Machine learning approaches for predicting the performance of stormwater biofilters in heavy metal removal and risk mitigation

The increasing amount of data on biofilter treatment performance over the past decade has made it possible to use data-driven approaches to explore the relationships between biofilter performance and a range of input variables. The knowledge gap lies in lack of models to predict the biofilter performance considering both design and operational variables, especially for heavy metals. In this study, we tested three machine learning (ML) approaches, namely multilinear regression (MLR), artificial neural network (NN), and random forest (RF), to predict biofilter outflow concentrations of heavy metals (Cd, Cr, Cu, Fe, Ni, Pb and Zn) using a range of design and operational factors as input variables. The results show that RF performed relatively better than other two models, with median Nash-Sutcliffe Efficiency (NSE) values of 0.995, 0.317, 0.762, 0.636, 0.726, 0.896 and 0.656 for Cd, Cr, Cu, Fe, Ni, Pb and Zn, respectively during model training. However, all the models were less accurate during model validation, with the better performance found for Cd (average NSE=0.964), Zn (0.530) and Ni (0.393) and poorer performance observed for Cu (0.219), Pb (0.058), Fe (-0.054) and Cr (-0.062). Infiltration rate (IR) and inflow concentration (Cin) were sensitive to all pollutants’ removal in biofilters. The ratio of system size to catchment size was also found to be important for Zn, Ni and Cd, while ponding depth was an important variable for Cd. Based on thousands of hypothetical design and operational scenarios (generated using raw data), the best ML models were used to predict the biofilter outflow concentrations and estimate the risk quotient (RQ) values with regards to reuse of treated stormwater for various purposes. Results suggest that biofilters were able to reduce health risks associated with heavy metals in stormwater and therefore produce reliable water fit for reuses such as irrigation, swimming, and toilet flushing. Modelling results showed that biofiltration did not meet the requirements for drinking when Cd contamination exists. Explorative analysis also demonstrated how the key operational and design variables can be optimised to further reduce the health risks that can be fit for drinking purposes (i.e., RQ value <1).