The influence of vegetation in stormwater biofilters on infiltration and nitrogen removal: preliminary findings T. Pham*, E. G. Payne*, T. D. Fletcher**, P. L. Cook *, A. Deletic * and B. E. Hatt *, * Department of Civil Engineering, Monash University, Wellington Rd, Clayton, Victoria, 3800, Australia (E-mail: tracey.pham@monash.edu.au; Emily.payne@monash.edu.au) ** Dept. of Resource Mgt & Geography, The University of Melbourne, Burnley Campus, 500 Yarra Boulevard, Burnley, Victoria, ABSTRACT This study provides new information on the influence of vegetation type on the infiltration rate and removal of nutrients in biofilters. Compared to previous studies, a wider range of plant species (11 species native to Western Australia, and 11 native to Victoria) and plant types (lawn grasses, grasses, sedges, rushes and shrubs) were examined in a laboratory-column study. The preliminary results indicate that the presence and type of vegetation influence infiltration rate and nitrogen removal in biofilters. In the early stages of the experiment, vegetation reduced the infiltration rate of columns as compared to soil-only controls, but over time this trend began to change, with. Importantly, whilst all plant species were able to reduce effluent concentrations for Total Nitrogen (TN) and nitrogen oxides (NOx), soil-only control columns leached NOx. Columns with submerged zones universally resulted in greater reductions in concentration for TN and NOx than traditional non-submerged columns (although the differences were often minor). Whilst there were some differences in removal performance between plant species, these differences were also relatively small, with TN outflow concentration averaging 0.44 mg/L and 0.26 mg/L for non-submerged and submerged columns respectively, an 80% and 88% reduction from the inflow concentration of 2.27mg/L. These preliminary results demonstrate that as long as the media is carefully selected, designers can choose from a relatively wide range of plant species and still achieve effective nutrient removal. KEYWORDS biofiltration, bioretention, nitrogen, phosphorus, vegetation selection INTRODUCTION Changes in land-use due to urbanisation mean that urban waterways experience increased peak flows and increased runoff volumes at higher frequencies as compared to their pre-developed conditions (Kibler, 1982). Further, urban stormwater pollutants originating from anthropogenic activities deteriorate the water quality of the receiving waterway (Klein, 1979). Biofiltration systems, which consist of a vegetated basin overlying a soil-based filter media, aim to mitigate the impact of urban stormwater, both in terms of hydrology and water quality. Past research has demonstrated that biofilters are able to attenuate flows (Hunt et al., 2008, Barber et al., 2003), as well as reduce suspended solids, heavy metal and phosphorus concentrations in stormwater (Hatt et al., 2007, Bratieres et al., 2008). Many studies have explored ways to enhance biofilter performance by modifying the traditional biofilter design. For example, introduction of a submerged zone containing a carbon source can greatly enhance nitrate removal by promoting denitrification, thereby improving overall nitrogen removal rates (Zinger et al., 2007a, Kim et al., 2003). Yet, nitrogen removal in biofilters continues to be inconsistent (Bratieres et al., 2008). One aspect of biofilter design which has not been thoroughly investigated is vegetation selection. While many studies have demonstrated that the presence of vegetation is crucial (Henderson et al., 2007, Lucas and Greenway, 2008, Bratieres et al., 2008), only a few studies have considered the actual mechanisms by which plants contribute to biofilter performance. To reduce clogging in biofilters, some authors suggest that plant species with thick roots may be preferred as they leave larger macropores for water flow in the soil when the root decays (Le Coustumer et al., 2009, Read et al., 2010). Although Le Coustumer et al. (2009) speculated that vegetation species which have a high proportion of fine roots should be avoided; they may worsen the effect of clogging by reducing pore space, Read et. al (2008, 2010) showed that a dense fine root system is important for maximising nutrient removal. Other studies also suggest that variation in plant species may account for variations in treatment performance. Bratieres et al. (2008) concluded that Carex appressa and Melaleuca ericifolia were significantly better for nitrogen removal because the roots of these species are better able to exploit the surrounding soil than other plants studied. Large amounts of microscopic root hairs on Carex roots, and arbuscular mycorrhizal fungi growing on the roots of Melaleuca led to increased root surface-area of these species. The study by Read et al. (2010) showed that variation in growth and morphology of plant species influenced pollutant removal. They recommended that plants chosen for biofilters should have high growth rates, high root density, high root mass and deep roots (FAWB, 2009). Unfortunately, the study conducted by Read et al. (2010) encompassed a relatively short time frame so that vegetation effects over the long term were not tested. Furthermore, only plants indigenous to south-eastern Australia were tested. This paper presents the preliminary findings of a larger column study whose broader aims are to understand the role of vegetation in nitrogen removal by biofilters. The paper aims to identify any early outstanding plant types or species which are demonstrating an ability to 1) reduce clogging by maintaining the infiltration rate of biofilters, and 2) minimise nitrogen concentrations in outflows. It specifically tests species from both south-eastern and south-western Australia. METHODS Experimental Setup Twenty-two Australian plant species (11 indigenous to Western Australia (WA), and 11 indigenous to Victoria (VIC)) were used for this study. Plant types included lawn grasses, grasses, sedges, rushes and shrubs in order to represent the range of plant types and morphology commonly found in biofilters (see Table 1). Ten plants from each species (except C. appressa which had a total of 15 plant representatives) were sourced from nurseries local to the plant species and planted into 300mm high x 150mm diameter black planter bags filled with loamy sand; the media typically used in biofilters. Plants were grown outside and allowed four months to acclimatise to their new growth conditions. During this time, they were watered with 400mL of tap water at a frequency sufficient to maintain them in a vigorous growing condition (typically every 2-3 days). Following the four-month acclimatisation period, each plant was planted individually into 150 mm diameter laboratory-scale test columns (that is, one plant per column). Of these ten replicate columns, five contained a 300mm deep Submerged-Zone (SZ) which was created by a raised outlet pipe, and five were non-submerged (NS) (i.e., the outlet was located at the base of the column) (See Figure 1). A mixture of pine chips and sugar-cane mulch was added as a carbon source into the submerged sections (transition and gravel layer) of SZ columns to assist with denitrification as specified by FAWB (2009). The biofilter columns were 800mm in height and contained four main sections (see Figure 1). The upper 200mm consisted of a ponding zone constructed from perspex to allow sunlight to reach the plant and allow the accurate measurement of the ponding depth in the column. The rest of the column was made of a PVC pipe and designed in a similar manner to that of field-scale biofilters: 300mm filter media, 200 mm transition layer, and 100 mm drainage layer (FAWB, 2009). The inside of the PVC pipes were sand-papered to reduce any preferential flow down the sides of the column. Additionally, ten control columns (unvegetated) were constructed; five to be subject to WA climate conditions and five to be subject to Victorian conditions (see details of dosing below). Table 1: Plant species chosen for this study. (L) - Lawn Grass, (G) - Grasses, (S) - Sedges, (R) -Rushes, (SH) - Shrubs VIC Plants WA Plants Soft Leaf Buffalo (LG) Poa labillardieri (G) Poa sieberiana (G) Carex appressa (S) Veletene (LG) Sporobolus virginicus (G) Austrodanthonia caepitosa (G) Poa poiformis (G) Cyperus gymnocaulos (R) Juncus kraussii (R) Gahnia trifida (S) Carex tereticaulis (S) Melaleuca incana (SH) Astartea scorpia (SH) Hypocalymma augustifolium (SH) 200 mm Ponding zone (clear perspex) 300 Loamy sand filter media (Ks = 180 mm/hr) 200 Sand transition layer (SZ with carbon addition) 100 Gahnia sieberiana (S) Juncus pallidus (R) Dianella revoluta (S) Dianella tasmanica (S) Allocasurina littoralis (SH) Leptospermum continentale (SH) Hakea laurina (SH) Gravel drainage layer SZ with carbon addition) SZ outlet Normal outlet Figure 1: Column configuration. A “Normal outlet” situated at the base of the column is used for Non-Submerged (NS) columns, while “SZ outlet” is used to create a Submerged-Zone (SZ). Stormwater Dosing A semi-natural stormwater was used for dosing the columns as this offered the best compromise between providing real stormwater quality and consistent concentrations for experimental control. The semi-natural stormwater was manufactured by mixing a concentrated “slurry” with dechlorinated tap water and chemicals to obtain target sediment and pollutant concentrations which were consistent with Australian urban stormwater quality as specified by Duncan (1999) and Taylor et al. (2005). The concentrated “slurry” was formed by passing sediment collected from a nearby stormwater pond through a 300 µm sieve. The columns were dosed twice weekly with the seminatural stormwater from 17th of December 2010 to simulate a ‘wet period’. Based on climate patterns local to the plant species and assuming a biofilter sized to 2.5% of its catchment area, dosing volumes of 3.7L and 4.2L were calculated for Victorian and WA columns, respectively. Infiltration measurements Infiltration measurements were conducted every three months to determine the influence of vegetation on the evolution of hydraulic conductivity over time. The infiltration rate was calculated by pre-soaking the columns (approximately 24 hours before commencement of the experiment) and measuring the drop in water level every two minutes for at least one hour. For extremely slow draining columns, measurements were taken every two minutes for the first hour, and then randomly throughout the day with the time noted so that the infiltration rate could be calculated. Infiltration measurements were also taken the following day for columns that were still draining. It is important to note that the measurements do not necessarily represent a saturated hydraulic conductivity, but the infiltration rate, measured using pre-wetted conditions typical of a biofilter in field practice. Ensuring saturation of the columns (which could have taken longer than 48 hours) was not considered desirable, given the likely artificial impacts on treatment performance. This paper presents results from the infiltration measurements conducted on the 21st and 22nd of June 2011 for NS columns and 5th and 6th of July 2011 for SZ columns, but data on the temporal evolution of infiltration rate over the experimental period to date (December 2010 to December 2011) are also presented. Stormwater Sampling To determine the influence of vegetation on pollutant removal, inflow and outflow water quality samples were collected on a monthly basis. A sub-sample was collected from the inflow tank, while the whole outflow volume of the columns were collected into 5L bottles then sub-sampled. All water quality samples were analysed by a NATA-certified laboratory for Total Nitrogen (TN), Total Dissolved Nitrogen (TDN), ammonia (NH3), and oxidised nitrogen (NOx). Additionally, Dissolved Organic Nitrogen (DON) was calculated by subtracting the sum of NH3 and NOx from TDN and Particulate Organic Nitrogen (PON) was calculated as the difference between TN and TDN. This paper only examines the results of the June 2011 dosing run.. Data analysis To determine whether vegetation significantly influenced the data collected (i.e infiltration rates and effluent N concentrations), an independant t-test was carried out between soil-only and vegetated columns for both NS and SZ designs separately. Levene’s test was used to assess the equality of variances of the data sets. In instances where homogeneity of variance was not assumed, Welch’s ttest was employed. One-way ANOVA was used to further investigate the influence of plant types (Lawn Grass, Grass, Sedge, Rush, and Shrubs) and plant species. To compare vegetated types and species against control columns, a post-hoc multi-comparison test was employed (Tukey or GamesHowell test depending on the homogeneity of data). Additionally, Pearson’s correlation determined any associations between infiltration rate and nitrogen species. SPSS v.19® was used for all analysis. A critical value of p= 0.05 was considered for hypothesis tests. Finally, Cube-root transformation was applied to infiltration data in order to meet normality assumptions. RESULTS AND DISCUSSION Infiltration Rate The presence of vegetation appears to slow the infiltration rate when compared to soil-only columns. The mean infiltration rates for soil-only columns were 70 mm/hr and 68 mm/hr for NS and SZ designs respectively, while the mean infiltration rates for vegetated columns were 49 mm/hr and 35 mm/hr (for NS and SZ designs respectively). Particularly slow draining plant species for NS and SZ columns (respectively) included J. palllidus (µ= 36 mm/hr and 11 mm/hr), M. incana (µ= 16 mm/hr and 10 mm/hr) and Velvetene lawn grass (µ= 4 mm/hr and 13mm/hr). It is hypothesised that the root volume of these species may be a factor causing the drainage rates to be slower than other columns; this will be investigated further in a final harvest which will examine vegetation characteristics. Vegetation type has a significant effect on infiltration rates (Figure 2). Lawn grass (p <0.05), rushes (p = 0.001) and shrubs (p<0.05) were significantly slower draining than soil-only control columns for NS column designs. For SZ columns, all plant types observed significantly slower infiltration rates than soil-only (p<0.05). Lawn grasses had the lowest mean infiltration rate of 18 mm/hr and 19 mm/hr for NS columns and SZ columns respectively, compared to 70 mm/hr and 68 mm/hr for soil-only NS and SZ columns, respectively. The lawn grass covers the entire surface area of the column thus its root may reduce flow pathways more so than other plant types. Figure 2: Mean, maximum and minimum bar graph comparing infiltration rates of various plant types: Lawn grass (n= 10), Grass (n=25), Sedge (n=30), Rush (n=15), Shrub (n=30) and Soil-Only controls (n=10) for each NS and SZ column design. It thus appears that in the early stages of the experiment, vegetation has a negative influence on infiltration rate, but this may be an artefact of the small column size and may also be a temporary effect. These results were taken in June 2011 early in the study and during winter when vegetation is mostly dormant. During more active seasons, root growth disturbs the soil in the vicinity of the roots, thereby causing movement in the soil substrate. The movement of soils may enlarge existing macropores or create new ones for water flow (Angers and Caron, 1998). Several studies have observed initially low infiltration rates of soils under plants which progressively increase with time (Lewis et al., 2008, Meek et al., 1992). The hydraulic conductivity of vegetated field biofilters studied by Lewis et al. (2008) initially decreased to 15mm/hr in winter, however improved over time. Figure 3 suggests that similar behaviour is occurring with the columns; it is thus possible that the vegetated systems will end up with the highest infiltration rates, at least for some vegetation types. Ongoing experiments will help to answer this question. Figure 3: Temporal evolution of infiltration rate over the duration of the experiment to date, showing the increase in variability of infiltration over time for soil-only (n=10) and vegetated systems (n=10). Stormwater treatment Vegetated columns significantly reduced the outflow concentrations of TN, TDN and NOx from their inflow concentrations (µ = 2.27 TN mg/L, 1.89 TDN mg/L, 0.97 NOx mg/L). Furthermore, the presence of vegetation significantly reduced the outflow concentrations of TN, TDN and NOx compared to soil-only controls regardless of plant type (see Figure 4). The mean TN outflow concentrations of soil-only columns were 1.71 mg/L and 1.35 mg/L (NS and SZ columns respectively). On the other hand, vegetated columns had a mean TN outflow concentration of 0.44 mg/L and 0.26 mg/L (NS and SZ columns respectively). SZ columns further enhanced pollutant removal of TN, TDN and NOx. Figure 4: Mean, maximum and minimum bar graph comparing inflow TN concentrations (n=15) to various plant types: Lawn grass (n = 10), Grass (n=25), Sedge (n=30), Rush (n=15), Shrub (n=30) and Soil-Only controls (n=10) for each NS and SZ column design Overall, NH3 concentrations in the effluent were extremely low compared to the mean inflow concentration of 0.39 mg/L. The mean concentrations of NH3 for vegetated columns were 0.003 mg/L and 0.019 mg/L for NS and SZ respectively, which was not significantly different from the mean concentration of soil-only columns; 0.004 mg/L and 0.021 mg/L for NS and SZ respectively. Zinger et al (2007) attributes the addition of a carbon source, which promotes ammonification and possibly dissimilatory nitrate reduction to ammonium (DNRA), to the increased outflow of NH3 in the SZ columns compared to NS columns. Given that the outflow concentrations of NH3, DON and PON were low for all columns (soil-only and vegetated) compared to inflow concentrations, the effect of vegetation is less important than has been shown in previous studies (e.g. Bratieres et al., 2008). Interestingly, soil-only columns for both NS and SZ designs leached NOx (µ = 1.58 mg/L and 1.14 mg/L respectively) compared to the inflow NOx concentration of 0.97 mg/L (See Figure 5). Greater NOx removal from vegetated as compared to unvegetated systems has also been observed in other studies (Henderson et al., 2007, Denman et al., 2007). These results suggest nitrification is taking place, but that there is a lack of denitrification occurring to complete the nitrogen removal process. On the other hand, plant root exudates from vegetated columns may provide a carbon supply to support heterotrophic denitrifying bacteria, leading to effective NOx transformation to gaseous end products. Plants may also be responsible for the direct uptake of some of the NOx, along with facilitating microbial uptake, particularly around the rhizosphere. Given that TDN (which consists of DON, NH3, and NOx) makes up 80% of TN in typical urban stormwater (Taylor et al., 2005), any removal of dissolved nitrogen offered by vegetation is crucial to overall TN removal, hence amplifying the importance of vegetation in biofilters. Figure 5: Mean, maximum and minimum bar graph comparing inflow NOx concentrations (n=15) to various plant types: Lawn grass (n = 10), Grass (n=25), Sedge (n=30), Rush (n=15), Shrub (n=30) and Soil-Only controls (n=10) for each NS and SZ column design Infiltration rate and treatment performance As expected, the treatment ability of the columns is related to their infiltration rate. However, this was only noticed for vegetated columns; there was no correlation between infiltration rate and the concentration of any nitrogen species for soil-only control columns (n= 20). This further highlights the effect of vegetation on both infiltration rate and treatment performance. Outflow concentrations of TN (p<0.001), TDN (p<0.001), NOx (p<0.001) and PON (p<0.05) were all positively correlated with infiltration rate for vegetated columns. In other words, slower draining columns provided a longer detention time for treatment processes to occur, and resulted in low pollutant outflow concentrations. NH3 effluent concentrations weakly correlated with the infiltration rates for columns with shrubs (n= 60, R= 0.026) and sedges (n= 60, R= -0.263) only. It is difficult to tell whether improved treatment performance by columns with a SZ is the primary result of anoxic conditions within the saturated zone, which promote denitrification, or the increased retention time of effluent within these columns, providing greater opportunity for biotic uptake or transformation. Further studies are required to better understand nitrogen pathways within biofilters, and this could include the use of isotopes to identify key processes and storages. IMPLICATIONS FOR BIOFILTER DESIGN The presence of vegetation in biofilters influences both infiltration rate and treatment performance. At least in the early stages of this study, vegetation has reduced the infiltration rate, but in turn this decreased outflow concentrations of nitrogen, a pollutant which is typically difficult to consistently remove. In part, this may be because increased detention time simply gives more time for biological transformation and uptake processes to take place. Designers should thus consider making larger biofilters with slower-draining media, because these systems are likely to have improved overall performance. Ideally, vegetation selection for biofilters should be a balance of infiltration rates and treatment performance. Although biofilters planted with lawn grasses provide easy maintenance, the preliminary results of this study demonstrate that the nature of their design (that is, lawn grasses covering the entire surface area of the biofilter) may encounter some drainage issues. Nevertheless, the results are preliminary and lawn grasses should not be ruled out entirely, especially if their infiltration rates improve in the future. Vegetation is essential for nitrogen removal, and from this study, it appears that provided the media is selected carefully so that it does not contain excessive nutrients (FAWB, 2009), many plant species can be effective in ensuring high levels of nitrogen removal. Consequently, there is a relatively large range of plant species to choose from which will still achieve effective nutrient removal. Furthermore, vegetation selection should consider the ability to survive and contribute to biofilter performance under stressful conditions such as drought. Future work will investigate the effect of dry periods on vegetation, and overall biofilter performance. ACKNOWLEDGEMENT This project was funded by an ARC Linkage Project (LP0990153) with partner funding from Melbourne Water and the WA Department of Water. We would like to thank J. Read, R. Williamson, F. Winston, T. Hines and K. Browne for their contribution to this project. REFERENCES ANGERS, D. A. & CARON, J. 1998. Plant-induced changes in soil structure: Processes and feedbacks. Biogeochemistry, 42, 55-72. BARBER, M. E., KING, S. G., YONGE, D. R. & HATHHORN, W. E. 2003. Ecology ditch: A best management practice for storm water runoff mitigation. Journal of Hydrologic Engineering, 8, 111-122. BRATIERES, K., FLETCHER, T. D., DELETIC, A. & ZINGER, Y. 2008. Nutrient and sediment removal by stormwater biofilters: A large-scale design optimisation study. Water Research, 42, 3930-3940. DENMAN, L., MAY, P. B. & BREEN, P. F. 2007. An investigation of the potential to use street trees and their root zone soils to remove nitrogen from urban stormwater. Australian Journal Water Resource 10, 303-311. DUNCAN, H. P. 1999. Urban Stormwater Quality: A Statistical Overview. Melbourne, Australia: Cooperative Research Centre for Catchment Hydrology. FAWB 2009. Stormwater Biofiltration System: Adoption Guidelines. HATT, B. E., DELETIC, A. & FLETCHER, T. D. 2007. Stormwater reuse: Designing biofiltration systems for reliable treatment. Water Science and Technology. HENDERSON, C., GREENWAY, M. & PHILLIPS, I. 2007. Removal of dissolved nitrogen, phosphorous and carbon from stormwater by biofiltration mesocosms. Water Science and Technology, 55, 183-191. HUNT, W. F., SMITH, J. T., JADLOCKI, S. J., HATHAWAY, J. M. & EUBANKS, P. R. 2008. Pollutant removal and peak flow mitigation by a bioretention cell in Urban Charlotte, N.C. Journal of Environmental Engineering, 134, 403-408. KIBLER, D. 1982. Urban Stormwater Hydrology, USA, American Geophysical Union. KIM, H., SEAGREN, E. A. & DAVIS, A. P. 2003. Engineered bioretention for removal of nitrate from stormwater runoff. Water Environment Research, 75, 355-367. KLEIN, R. 1979. Urbanization and stream quality impairment Water Resources Bulletin, 15, 948-963. LE COUSTUMER, S., FLETCHER, T. D., DELETIC, A., BARRAUD, S. & LEWIS, J. F. 2009. Hydraulic performance of biofilter systems for stormwater management: Influences of design and operation. Journal of Hydrology, 376, 16-23. LEWIS, J. F., HATT, B. E., DELETIC, A. & FLETCHER, T. D. 2008. The impact of vegetation on the hydraulic conductivity of stormwater biofiltration systems. 11th International Conference on Urban Drainage. Edinburgh, Scotland, UK. LUCAS, W. C. & GREENWAY, M. 2008. Nutrient retention in vegetated and nonvegetated bioretention mesocosms. Journal of Irrigation and Drainage Engineering, 134, 613-623. MEEK, B. D., RECHEL, E. R., CARTER, L. M., DETAR, W. R. & URIE, A. L. 1992. Infiltration rate of a sandy loam soil: effects of traffic, tillage, and plant roots. Soil Science Society of America Journal, 56, 908-913. READ, J., FLETCHER, T. D., WEVILL, T. & DELETIC, A. 2010. Plant traits that enhance pollutant removal from stormwater in biofiltration systems. International Journal of Phytoremediation, 12, 34-53. READ, J., WEVILL, T., FLETCHER, T. & DELETIC, A. 2008. Variation among plant species in pollutant removal from stormwater in biofiltration systems. Water Research, 42, 893-902. TAYLOR, G. D., FLETCHER, T. D., WONG, T. H. F., BREEN, P. F. & DUNCAN, H. P. 2005. Nitrogen composition in urban runoff - Implications for stormwater management. Water Research, 39, 1982-1989. ZINGER, Y., FLETCHER, T. D., DELETIC, A., BLECKEN, G.-T. & VIKLANDER, M. 2007a. Optimisation of the nitrogen retention capacity of stormwater biofiltration systems. 6th International Conference on sustainable techniques and strategies in urban water management. Lyon, France: NOVATECH 2007.