Final report for GNE16-124
Project Information
Our research attempts to isolate the effects of low-phosphorus compost and vegetation on the performance of bioretention cells in an agricultural landscape. In June 2016, a stormwater bioretention system was installed at the UVM Miller Research Complex, a dairy teaching and research facility comprising of 2.7 acres impervious area, and with uses associated with a working agricultural landscape, including movement of silage, manure, etc. Runoff from the surrounding landscape is captured in grass-lined swales and moves through a settling forebay before it is split among three bioretention cells, each with a unique soil or vegetation treatment. One cell has neither vegetation nor compost, one cell is planted with vegetation (Panicum virgatum) and no compost, and one cell is planted with vegetation and includes a layer of low-phosphorus compost. Low phosphorus compost is defined as being entirely composed of leaves and plant material, and is less than 0.2% total phosphorus by dry weight. The cells are experimentally paired, allowing for a factorial comparison of the compost and vegetation’s affect on performance. Preliminary results have suggested that, after one growing season, the presence of compost and vegetation has no statistically significant effect on nutrient and sediment concentration reduction. The presence of compost does, however, significantly affect above-ground biomass of planted species and the amount of labile phosphorus in the shallow horizon of the of the bioretention soil media. Thus, the use of low-phosphorus compost in bioretention cells may improve plant growth and establishment without contributing significant excess nutrients to effluent, but may load additional phosphorus to the media. We suggest continued monitoring of these cells, as an affect of vegetation may become apparent as the plants continue to grow and the additional phosphorus contributed by the compost to the soil media may decrease the ability for long term sorption and treatment.
Introduction:
The purpose of this project is to test the efficiency of bioretention cells used to treat stormwater from an agricultural production area (i.e., barns, grain storage, parking lots, etc). Bioretention is an increasingly used stormwater best management practice that promotes the infiltration and treatment of stormwater through natural processes such as sedimentation, filtration, sorption, plant uptake and storage, and microbial decomposition. Bioretention cells have been shown to be effective at reducing elevated hydrological conditions and pollutant loads in stormwater (Davis et al. 2001).
Activities related to farming such as wetland draining, soil compaction, and clearing of native vegetation have reduced the ability of landscapes retain and naturally treat stormwater, leading to greater volumes and velocities of runoff entering downstream water bodies (Ruhl 2000). Constituent within this runoff can be water quality pollutants including nutrients, sediment, heavy metals, and pathogenic organisms, which can negatively affect aquatic habitat and pose a risk for human health. In addition to fields and pastures, stormwater managers must also consider the production area of farming operations. The centralized production area is unique in its characterization as a stormwater landscape; it can have a high impervious area similar to that of an urban or suburban watershed but can contribute substantial stormwater pollutants similar to production fields and agriculture. Bioretention may provide a means by which diffuse stormwater runoff in an agricultural landscape or farm production area can be retained and treated.
A benefit of bioretention is the flexibility that designers have to meet water quality goals by altering aspects of the cell’s major design components such as media composition and planting plan. The use of compost in bioretention is often recommended for plant establishment; however, studies have observed nutrient leaching from organic matter in bioretention cells in the initial years after installation (Hunt et al. 2006). Stormwater from agricultural areas often already has high levels of nutrients, and bioretention cells accepting this water may not require compost to meet planting goals. A net export of nutrients would be counter to the goals of stormwater management, especially in nutrient sensitive watersheds. There is, therefore, a need for better understanding of compost’s nutrient contribution to bioretention cells and impact on stormwater treatment.
The type and presence of vegetation within bioretention cells can also have a significant impact on stormwater treatment. A species that is used in bioretention systems that also has agricultural uses is Switchgrass (Panicum virgatum). While graminoids with extensive root systems have generally been shown to be effective in stormwater treatment (Read et al. 2008), there are no field studies that specifically quantify the effectiveness of Switchgrass in bioretention.
Through a factorial design, our research will isolate the effects of compost and vegetation (i.e., Switchgrass) on the ability of bioretention cells to treat pollutants from stormwater runoff in a farm production area. Stormwater pollutants of interest in this study are nutrients (nitrogen and phosphorus) and total suspended solids (TSS), chosen for their traditionally high load in agricultural areas (Ruhl 2000). Additionally, we will determine if these design components have a significant impact on bioretention cells’ change in nutrient concentration with soil depth, providing us with a greater understanding on their propensity to leach or export nutrients. This study will allow us to quantify how compost or vegetation influences stormwater treatment and provide recommendations for their future use in bioretention projects.
Our study will have a direct economic benefit to farmers considering using bioretention to treat stormwater from their property. Bioretention cells may take up a relatively large area of land depending on the watershed they are treating, and farmers would benefit from minimizing the cost of vegetation or compost application. By testing a monoculture of Switchgrass, we will be able to estimate the effectiveness of a single plant and provide suggestions on bioretention planting density and pollutant removal.
The compost used in our study is a low-phosphorus leaf mix that may have a lower chance of leaching nutrients than other bioretention projects. By isolating its effect on pollutant removal and nutrient leaching, we will be able to comment on whether compost in general is required for agricultural bioretention cells, and if a low nutrient mix can meet performance goals.
- Quantify the effects of both compost and vegetation on the removal of nutrients and TSS by bioretention cells treating stormwater from a farm production area.
- Measure the change in vertical concentration of nutrients within the bioretention soil media and determine to what extent it is affected by compost and vegetation.
Cooperators
Research
Study Site
Construction of three bioretention cells was completed in fall 2016 at the University of Vermont Paul R. Miller Research Complex (MRC), a dairy and equestrian teaching and research facility located in South Burlington, VT (44° 27' 33.411" N, 73° 11' 21.9696"). South Burlington has average high and low temperatures of 12.9 and 2.33 °C, respectively, average annual rainfall of 93.4 cm, and precipitation 151 days of the year (U.S. Climate Data 1981-2010). The bioretention cells treat runoff from buildings, rooftops, grassy lawns, paved and dirt parking and driving lanes, and some areas where dairy cows and farm equipment cross paths between paddocks and the barns. The entire drainage area is 12,974 m2 with four dominant land cover types: pavement, rooftop, grass, and dirt/gravel roads. The complex can be considered as a unique stormwater runoff landscape with uses resembling both that of a typical suburban/ institutional area and an agricultural production facility. The landscape is within the Potash Brook Watershed (HUC-8, 01100002), a tributary to Lake Champlain.
The bioretention cells’ surfaces are trapezoidal and have an approximate total area of 249.3 m2, 3 x 83.1 m2 cells, together representing ~1.9% of the total drainage area. The vertical profile of the cells from top to bottom includes a 7.6-cm layer of pea-stone, a 76-cm layer of sand-based bioretention media, a second 15.2-cm layer of peastone, and a 30.5-cm layer of roughly 2.2-cm diameter gravel. The sand-based bioretention media is 90% silica sand, 8% fine gravel, and 2% silt and clay, with an average porosity of 0.25. Each bioretention cell has a maximum ponding depth of 30.5 cm, resulting in a total of 53 m3 of storage capacity per cell (excluding the volume of the media). The cells are unlined; surrounding soils range from dense clay to dense sandy loam, with intermittent pockets of loose gravel and urban fill. A perforated 10.2-cm diameter PVC underdrain pipe runs along the longitudinal center of each cell within the gravel layer, conveying water that does not seep into the subgrade, into three separate outflow structures, one per bioretention cell.
Compost and Vegetation Treatments of Bioretention Cells
Each bioretention cell has a unique treatment that allowed for experimental comparison. Cell 1 (C+V+) was installed with a layer of low-P compost and planted with Switchgrass (Panicum virgatum), Cell 2 (C-V+) had no compost and was planted with Switchgrass, and Cell 3 (C-V-) had neither compost nor vegetation. Low-P compost was defined as being entirely composed of leaves and plant material and excluding manures and foodscraps, with less than 0.2% total phosphorus by dry weight. In Cell 1, a 7.6-cm layer of compost was added between the sand-based bioretention media layer and upper peastone layer. In the two planted cells, Cell 1 and Cell 2, 300 Switchgrass plants were installed from 10.2-cm pots at an approximate density of one plant per 0.25 m2. Plants were rooted in the low-P compost in Cell 1 and sand-based bioretention media in Cell 2, with the base of the plants in both Cells 1 and 2 surrounded by peastone (as a mulch alternative). Plants were installed between June 1 and June 8, 2016 and for the first six weeks after installation were watered with an oscillating sprinkler for approximately two hours, three to five times per week. All cells, including the non-vegetated Cell 3, were watered equally to maintain consistency.
Stormwater Flow
During storm events, runoff is collected and channeled by two grass lined swales into a 0.75-m deep by 15-m diameter circular sediment forebay, meant for settling large incoming solids. Stormwater exits the forebay via a 10.2-cm diameter PVC upturned elbow pipe. The bioretention inflow consists of a three-way splitting structure designed to direct approximately equal volumes into each of the bioretention cells via three separate 15.2-cm diameter upturned elbow pipes that were leveled and placed at the same elevation using a laser transit level. Stormwater from the splitting structure enters the bioretention cells through a 10.2-cm diameter PVC inlet pipe, and spreads across the surface of each cell before percolating through the media. Water that is not lost via seepage into the surrounding soil is collected by the 10.2-cm-diameter perforated PVC underdrain and channeled into a 76-cm diameter outflow sampling structure, one per bioretention cell. When the outflow sampling structure fills, stormwater will overflow into another 15.2-cm diameter upturned elbow pipes that discharges to a grassy swale and eventually into Potash Brook. In extreme rainfall events (T ≥ 25 Years, 8.8-cm/24 hours), excess stormwater will bypass the system via emergency spillways in the bioretention cells and forebay that conveys runoff directly to the discharge swale.
Runoff Sampling
Runoff samples were collected during storm events using a flow-based sampling protocol (Leecaster et al. 2002). For the flow calculation, the upturned 15.2-cm diameter PVC elbow pipes in the inflow splitting structure and three outflow sampling structures were treated as rectangular sharp-crested weirs without end contractions. Before storm events, the sampling structures were filled with tap water to overflow the weirs and calibrate a pressure transducer level sensor (Teledyne ISCO 720 Module, Lincoln, NE) to zero (0.00 m). During storm events, the water height above the weirs was measured by a pressure transducer and converted to flow via the equation:
[1] Q = 3.33LH3/2
where Q is the flow (m3/s), L is the length of the weir (m), and H is the measured height of the water above the weir crest (m) (U.S. Dept. of the Interior Bureau of Reclamation 2001).
Height was recorded and flow was calculated every minute and converted to a measure of volume (V, m3) via the equation:
[2] V = ∫ Q(t) ∂t
The trapezoidal method of numeric integration was used to estimate the area under the hydrograph (i.e. volume) based on discrete sampling points along the curve. Trapezoidal integration as a general function of estimating area under a curve (i.e. volume of the hydrograph) is illustrated by:
[3] A = Σ (t2-t1) * [(f)t2-(f)t1]/2
where A is the area under the curve, t1 and t2 are time points (i.e. minutes), and (f)t1 and (f)t2 are the flow rates at t1 and t2, respectively.
A 900-mL sample of runoff was taken for every set amount of volume that was calculated to have passed through the sampling structure with a maximum total of 24 bottles that could be filled by the auto samplers (Teledyne ISCO ISCO 6712, Lincoln, NE). The volume between samples was pre-determined in order to best capture the entire duration of a predicted storm, and varied from one storm event to the next. Inflow volume between samples was determined by using the forecasted rainfall depth in the Curve Number Equation (Akan 1993), and dividing by 24. Outflow volume between samples for each cell was estimated as one third the inflow volume. A total of thirteen storms were sampled between June 22nd to November 3rd, 2016 (Appendix A).
Water Quality Analysis
Nutrient Concentration
Stormwater samples were analyzed for concentrations of Total Nitrogen (TN), combined Nitrate and Nitrite (NOx-N), Ammonium (NH4-N), Total Phosphorus (TP), and Soluble Reactive Phosphate (SRP). Concentrations were measured for every sample bottle taken during a storm event. Soluble nutrient species (NOx-N, NH4-N, and SRP) were prepared for analysis by filtration through a 0.45μm pore nylon mesh filter. Total nutrient species (TN and TP) were prepared for analysis via persulfate digestion of an unfiltered sample, which oxidized all forms of nitrogen and phosphorus into NOx-N and SRP, respectively. All preparations for nutrient analysis were done within 48 hours of the sampled storm event.
Nutrient concentrations were determined via flow injection analysis and automated colorimetry (Lachat Instruments QuickChem8000 AE, Hach Inc., Loveland, CO). In this procedure, the concentration of solute is directly proportional to its color and the absorbance read at 520 nm for NOx-N (magenta), 660 nm for NH4-N (emerald green), and 880 nm for SRP (blue) (APHA 2010 – 4500 P-B). Each analysis was calibrated with 12 standards of NOx-N, NH4-N, and SRP in deionized water ranging in concentration from 0.005-10.0 mg/L along with two Quality Control Checks in a similar range. The instrument was recalibrated or samples were reanalyzed with new standards if Quality Controls deviated by greater than 10% of their expected value. If preliminary results showed a wide range in concentration values, results were calibrated along two different curves. When concentrations were less than 0.1 mg/L for either nitrogen or phosphorus from the full calibration curve, a low calibration curve was used instead, ranging from 0.005-0.2 mg/L.
TSS Concentration
Total suspended solids (TSS) concentration was measured by taking a 400-mL subsample from each bottle (APHA 2010 – 2540D). Deionized water was first passed over a Whatman 47-mm standard glass fiber filter and dried at 100 oC overnight. Filters were then weighed and had a subsample applied to them over a vacuum filter from a vigorously shaken bottle. The subsample of 400 mL was used for analysis, unless clogging of the filter was observed, in which case a smaller sample of 100 mL was used. Once the entire subsample has passed through the filter, it was dried again overnight at the same temperature and its final weight recorded. Final TSS concentrations were calculated as the difference between filter weights divided by the subsample volume.
Mass Removal
Mass was calculated using stormwater volume and pollutant concentrations of twelve storms for nitrogen (TN, NOx-N, NH4-N) and TSS, and of eleven storms for phosphorus (TP, SRP). Several storm events were missing either flow or water quality data due to instrument error, and were therefore left out of analysis (Appendix A). Mass of nutrients and TSS that passed through the sampling structures during a storm event were calculated via the equation:
[4] M = Σ(VC)
where M is mass, V is volume of stormwater passing through the sampling structure during a sampling interval, and C is the concentration of the stormwater pollutant (i.e. nutrients, TSS) during the same interval. Concentration in the sample bottle was multiplied by the preceding volume, and, in the last sample taken, by the final volume that did not result in a bottle being filled. If the event had no volume measured from its outflow structure, mass was assumed to be either fully retained by the bioretention cell or to have seeped into the surrounding soil. All mass was reported in kilograms (kg).
The mass of stormwater pollutants removed by the bioretention cell per storm (excluding seepage) was estimated via the equation:
[5] MR = MI – (MO + MS)
where MR is the mass removed by the cell, MI is the mass into the cell via the inflow structure, MO is the mass out of the cell via the outflow structure, and MS is the mass out of the cell via seepage to the surrounding soil media. MS was calculated as the EMC of the outflow event for an event multiplied by the estimated volume of seepage from the cell. The estimated volume of seepage per cell for a storm event was calculated via the equation:
[6] VS = (VP + VI) - VO
where VS is the volume of seepage from a cell, VP is the volume of precipitation fallen on a cell during a storm event (i.e. cm x m2), VI is the volume of stormwater entering the cell (i.e. 1/3 of total inflow volume for an event), and VO is the volume of stormwater exiting through the outflow structure. For the purpose of this estimation, I assumed the media was at field capacity and there was no storage during a storm event. Equal influent volume between the cells could not be assumed before the installation of the baffle in the inflow structure, and therefore the mass retention of each bioretention cell was only calculated for two storms after this date (October 22 and 28). Evapotranspiration was assumed to be negligible in this model due to high frequency of rainfall between these dates and relatively cooler ambient temperatures.
Bioretention Media Water Extractable Phosphorus (WEP)
Samples of the sand-based bioretention media were taken using a 2.54-cm soil core on June 8 and again on November 7, 2016 using methods similar to that of Muerdter et el. (2015). The dates of sampling represent the first day of installation of the bioretention cells and one week after the final water quality sampling date, respectively. Three equidistant points (2.74 m) along the center transect of each bioretention cell were marked with flags for sampling locations. At each of these locations, two separate cores were taken 10 cm perpendicular to center transect line for each sampling date; to the left in June and the right in November. Peastone mulch, and compost in Cell 1, was cleared away from the sampling locations so that cores were taken from the top of the sand-based bioretention media in all three cells. Media was extracted to a depth of 40 cm and divided into five separate segments of 0-5 cm, 5-10 cm, 10-15 cm, 25-30 cm, and 35-40 cm (lowest sampling depths chosen to represent the full extent of the soil core). Similar depth segments of the two cores samples taken at each location per time point were composited and evenly mixed. This resulted in a total of 15 media samples per cell per sampling time (5 depths x 3 sampling points). Media samples were allowed to air dry for one week in paper envelopes before weighing and analysis.
A measure of Water Extractable Phosphorus (WEP) concentration was obtained for both sample dates. Three grams of soil were combined with 30 mL of deionized water and shaken for 1 hour. Samples were then centrifuged for 10 minutes at 5000 RPM, and the supernatant liquid was extracted using a polypropylene syringe and filtered through a 0.45μm pore nylon mesh filter. This solution was then analyzed for SRP using automated molybdate blue colorimetry. WEP concentration was reported in mg P/kg Media (dry weight).
Vegetation Harvest and Sampling
A total count of surviving Switchgrass was done on October 22, 2016; survival was defined in this experiment as a plant being alive at the end of the growing season, and does not include a count after overwintering. A plant was considered to have survived if it could be positively identified at the time of counting. Surviving plants were divided by the total originally planted (i.e. 300), to obtain a measure of plant survivorship in each cell.
On November 7, 2016, a representative subsample of Switchgrass above-ground biomass was harvested from each of the two planted cells. One m2 PVC quadrats were placed along a systematic grid within the two cells in nine locations, and all Switchgrass plant material within these areas was harvested at 2.54 cm above the peastone layer. The harvested contents of each quadrat were placed in separate paper bags and dried at 100 oC for 24 hours. Total biomass (sans paper bag) was weighed immediately after drying. The average biomass per harvested quadrat was assumed to be representative of the entire cell and was multiplied by the total area to obtain an estimate of total biomass.
Statistical Analysis
Nutrient and TSS concentration reduction was compared across the three treatments using Analysis of Covariance (ANCOVA), in which the average outflow concentrations of individual storm events from separate treatments were the tested dependent variables and the inflow concentration was the independent covariate. In this model, a main effect of inflow concentration upon treatment outflow concentration was tested, and if no significant effect was found, a test of interaction between the treatments was performed. In the event that influent concentration was to have a significant effect on treatment effluent concentration, a one-way ANOVA with Least Squares Difference analysis was used to test for a significant difference across treatments.
Individual treatment reduction of stormwater pollutant concentration was analyzed by comparing influent concentrations of storm events with the individual treatment effluent concentrations in a paired samples t-test.
A four-way Analysis of Variance (ANOVA) with two-way interaction was used to compare WEP concentration in the bioretention media, testing for a significant effect and interaction on treatment, time, media depth, and distance from cell inlet.
Switchgrass biomass was compared across the planted bioretention cells, Cell 1 (C+V+) and Cell 2 (C-V+) using independent samples t-test.
A threshold of p<0.05 was used as an indicator of significance in all tests. Values between 0.05 and 0.1 were considered “marginally significant”. Statistical models were run on IBM SPSS, Version 23.
Results
Nutrient and TSS Removal Performance
Concentration Reduction
Nitrogen – The average influent concentration of TN per storm measured from twelve events was 4.00 (± 1.87) mg/L, ranging from a low of 1.47 mg/L to a high of 14.2 mg/L. The average effluent TN concentration of Cell 1 (C+V+) was 3.31 (± 1.12) mg/L, a 17.1% reduction from seven events; Cell 2 (C-V+) was 2.46 (± 0.68) mg/L, 38.5% reduction from eight events; Cell 3 (C-V-) was 2.65 (±1.04) mg/L, a 33.8% reduction from four events. Only Cell 2 significantly reduced TN concentration via a paired samples t-test with influent concentrations (p = 0.019). Influent TN concentration did not significantly affect the difference in treatment performance (p = 0.984), nor was there a statistically significant difference in effluent TN concentration across treatments (p = 0.984).
The average influent concentration of NH4-N per storm measured from thirteen events was 0.369 (± 0.212) mg/L, ranging from a low of 0.027 mg/L to a high of 1.52 mg/L. The average effluent NH4-N concentration of Cell1 was 0.060± (0.023) mg/L, an 83.7% reduction from seven events; Cell 2 was 0.023 ± (0.012) mg/L, a 93.8% reduction from nine events; Cell 3 was 0.020 ± (0.008) mg/L, a 94.6% reduction from five events. Influent concentration of NH4-N had a significant effect on the difference in treatment performance (p = 0.016), such that higher influent concentrations were correlated with higher effluent concentrations in Cell 1. Effluent from Cell 1 and was significantly greater than that of Cell 2 and 3 (p < 0.001). Reductions of concentration from the influent were statistically significant via paired samples t-test for Cell 2 (p = 0.024) and Cell 3 (p = 0.008), and marginally significant for Cell 1 (p = 0.054)
The average influent concentration of NOx-N per storm measured from thirteen events was 0.230 (± 0.188) mg/L, ranging from a low of 0.027 mg/L to a high of 1.34 mg/L. The average effluent NOx-N concentration of Cell1 was 2.23 (± 0.41) mg/L, a 970% increase from seven events; Cell 2 was 1.70 (± 0.63) mg/L, a 739% increase from nine events; Cell 3 was 1.75 (± 0.76) mg/L, a 761% increase from five events. All treatments significantly increased concentration compared to influent in a paired samples t-test (p <0.001). Influent NOx-N concentration did not significantly affect the difference in treatment performance (p = 0.465), nor was there a statistically significant difference of NOx-N effluent concentration between treatments (p = 0.294).
Phosphorus - The average influent concentration of TP per storm measured from eleven events was 1.50 (± 0.362) mg/L, ranging from a low of 0.558 mg/L to a high of 3.08 mg/L. The average effluent TP concentration of Cell 1 was 0.112 (± 0.042) mg/L, a 92.5% reduction from six outflow events; Cell 2 was 0.088 (± 0.024) mg/L, a 94.1% reduction from seven events ; Cell 3 was 0.066 (± 0.037) mg/L, a 95.6% reduction from three events. Both Cells 1 and 2 were found to significantly reduce influent concentrations via paired samples t-test (p = 0.001 and p < 0.001, respectively), and Cell 3 was found to have a marginally significant effect (p = 0.054). Influent TP concentration did not significantly affect the difference in treatment performance (p = 0.197). Treatments’ effluent concentrations were not found to be significantly different from each other (p = 0.194); however, effluent concentration from Cell 1 was marginally significantly greater than Cell 3 (p = 0.080).
The average influent SRP concentration per storm measured from twelve events was 0.887 (± 0.196) mg/L, ranging from a low of 0.258 mg/L to a high of 3.08 mg/L. The average effluent SRP concentration of Cell 1 was 0.052 (± 0.019) mg/L, a 94.1% reduction from six outflow events; Cell 2 was 0.035 (± 0.007) mg/L, a 96.1% reduction from eight outflow events; Cell 3 was 0.046 (± 0.013) mg/L a 94.8% reduction from four outflow events. All treatments were found to significantly reduce influent SRP concentration via paired samples t-test (p < 0.001). Influent SRP concentration did not significantly affect the difference in treatment performance (p = 0.747), nor was there a statistically significant difference in effluent concentration between treatments (p = 0.226), however, effluent concentration from Cell 1 was marginally significantly greater than Cell 2 (p = 0.100).
TSS – The average influent TSS concentration per storm measured from twelve events was 155.7 (± 197.0) mg/L, ranging from a low of 9.2 mg/L to a high of 1137.8 mg/L. The average effluent TSS concentration of Cell 1 was 4.1 (± 4.2) mg/L, a 97.4% reduction from eight events; Cell 2 was 4.4 (± 6.3) mg/L, a 97.2% reduction from eight outflow events; Cell 3 was 1.5 (± 1.9) mg/L, a 99.0% reduction from four events. All treatments had a marginally significant reduction of TSS concentration via paired sample t-test (Cell 1 - p = 0.077; Cell 2 - p = 0.057; Cell 3 - p = 0.051). Influent TSS concentration did not significantly affect the difference in treatment performance (p = 0.835), nor was there a statistically significant difference in effluent concentration across treatments (p = 0.812).
Mass Removal
Table A reports the estimated mass of stormwater pollutants retained by the bioretention cell media during the two sampled storms that occurred after modifications to the inflow splitting structure to equalize volume of all three bioretention cells’ inflow. The storms occurred on October 22 and 28 and rainfall depth recorded was 2.78cm and 1.02 cm, respectively. Rainfall was sparse in the weeks prior to the October 22, with inflow being last observed 33 days prior.
Table A – Average stormwater pollutant mass retention of MRC Bioretention Cells for October 22nd and 28th Storms. These events were sampled after the installation of the inflow baffle, and can therefore be assumed to have equal volume directed to all three cells.
|
Cell 1 (C+V+) |
Cell 2 (C-V+) |
Cell 3 (C-V-) |
||||
|
MR % |
MR (kg) |
MR % |
MR (kg) |
MR % |
MR (kg) |
|
|
TSS |
85.73% |
1.286 |
96.47% |
1.233 |
83.21% |
1.406 |
|
TN |
-51.89% |
-0.030 |
2.17% |
0.004 |
-12.70% |
-0.019 |
|
NOx-N |
-1440.11% |
-0.050 |
-1161.24% |
-0.039 |
-1919.71% |
-0.065 |
|
NH4-N |
67.95% |
0.003 |
86.37% |
0.003 |
80.09% |
0.003 |
|
TP |
80.44% |
0.022 |
92.45% |
0.023 |
79.68% |
0.020 |
|
SRP |
91.37% |
0.016 |
94.70% |
0.016 |
96.46% |
0.017 |
This limited data set suggests a possible effect of treatment on bioretention TSS and nutrient mass retention. The general trend was Cell 2 had a greater retention of nutrient and sediments than Cell 1 and Cell 3. This pattern is similar for TN, NH4-N, TP, and TSS. The difference between treatments is shown most notably by a low, but positive retention of TN by Cell 2 (2.17%), but a net export by Cell 1 (-51.89%) and Cell 3 (-12.70%). NOx-N was an exception to the pattern with mass exported at notably higher levels in Cell 3 (-1919%) than Cell 1 (-1440%), but both still greater than Cell 2 (-1161%).
Media Water Extractable Phosphorus (WEP) Concentration
The average WEP concentration across all depths in Cell 1 decreased between June and November, from 1.53 (± 1.22) mg P/kg Media to1.15 (± 0.34) mg P/kg Media, due primarily to the significant lowering in the shallow layer. The four deeper layers of Cell 1 have an average increase in concentration over time. Both Cell 2 and3 showed a net increase in average WEP concentration over time from 0.317 (± 0.046) mg P/kg Media to 0.700 (± 0.248) mg P/kg Media and from 0.368 (± 0.106) mg P/kg Media to 0.706 (± 0.279) mg P/kg Media, respectively.
There was a statistically significant effect of treatment on WEP concentration (p = 0.001), with Cell 1 significantly greater than Cell 2 (p = 0.003) and Cell 3 (p = 0.004). There was also an effect of depth (p < 0.001) and interaction of depth and treatment (p = 0.016) on WEP concentration. These results suggest greater WEP concentrations in the shallow depths of the bioretention media, and that Cell 1 had higher concentrations in its shallow depth than either Cell 2 or Cell 3. There was no statistically significant effect of time or distance from inlet on WEP concentration.
Vegetation Survivorship and Biomass
The survivorship of the planted species, Switchgrass (Panicum virgatum), was similar between the two vegetated bioretention cells with a total count of 252 for Cell 1 (C+V+) and 238 for Cell 2 (C-V+), on October 22nd (136 days since planting). Out of the initial 300 per cell planted on June 8, this is a relative survivorship of 84% and 79.3%, respectively.
The above-ground biomass accumulation of the cells was significantly different via independent samples t-test, with Cell 1 yielding an average of 0.346 (± 0.103) kg/m2 and Cell 2 yielding an average of 0.037 (± 0.013) kg/m2 (p < 0.001). Factoring this measure of biomass by the total planted area of the cells (i.e. 83.1 m2) yields total above-ground biomass measurements of 28.75 kg for Cell 1 (C+V+) 3.07 kg for Cell 2 (C-V+), 936% more biomass in the cell with low-P compost.
Discussion
Nutrients and TSS Removal Performance
Low-P compost effects on water quality: Cell 1 (C+) vs. Cell 2 (C-)
Overall, the presence of compost had no effect on TSS concentration reduction and only a marginal effect on nutrient concentration reduction. This effect can be seen by Cell 1 (C+V+) having effluent TP concentrations marginally higher than Cell 2 (C-V+) though both bioretention treatments significantly reduced concentrations from the influent. This may suggest a potential for the low-P compost to leach some phosphorus, but not to an extent that it severely depresses concentration reduction potential of bioretention cells treating this type of mixed institutional and agricultural runoff. Compost and native soils with high P fractions have been shown to decrease concentration reduction in bioretention cells, and at times even lead to a net increase of P (Hunt et al. 2006, Bratieres et al. 2008, Paus et al. 2014, Chahal et al. 2016). This study contributes to the available literature which reports this range in potential for excess phosphorus leaching from bioretention, and underscores the importance of soil media specification and design.
Also, related to the presence of compost, the effluent NH4-N concentration of Cell 1 was found to be significantly affected by influent concentration and significantly greater than Cell 2’s effluent concentrations. This suggests that NH4-N removal performance was poorer in the presence of compost and especially during storms where NH4-N influent loading was high. One possible cause of this could have been a relation between storm intensity and observed effluent. It is possible that heavy, intense rainstorms that carried higher concentration of influent NH4-N to the cells from the watershed also caused greater leaching of NH4-N from the compost in Cell 1. Li et al. (2006) noted a significant correlation between maximum rainfall intensity and the concentration of stormwater pollutants. Also, simulated high intensity or high volume rain events have been shown to increase effluent pollutant concentration of bioretention mesocosms (Yang et al. 2013). However, the sample size of storms for this study was small, and available rainfall data did not explicitly point to any patterns related to storm intensity and NH4-N concentrations. Future study of these cells should take the intensity of rainfall into consideration to further explore this hypothesis.
Of the pair, only Cell 2 was found to significantly reduce TN concentration from influent levels, though effluent concentration was not significantly different between treatments. The lower reduction of TN and NOx-N concentration may be indicative of leaching of compost by the Cell 1, a phenomenon seen in several previous studies that have used organic soil amendments in bioretention (Davis et al. 2001, Hsieh and Davis 2005, Mullane et al. 2015, Hurley et al. 2017). Here I make note that the treatment of a vegetated bioretention cell without added compost was the only configuration tested to significantly reduce both Total Nitrogen and Total Phosphorus. This finding is noteworthy, and raises questions about whether compost is necessary or advisable to achieve nutrient-related water quality goals for stormwater.
An analysis of the mass retention by the different treatments from two storms in which equal flow between the cells could be assumed also suggested that the presence of low-P compost has a marginal negative effect on bioretention pollutant mass removal. Mullane et al. (2015) noted a similar contribution of nutrients from compost to the effluent of bioretention mesocosms, but observed a decreasing effect over time as nutrients washed out of the system. The estimations of mass retention by the MRC bioretention cells are from a very limited data set taken in the first year of operation; it is possible that the nutrient leaching may decrease over time.
Panicum virgatum effects on water quality: Cell 2 (V+) vs. Cell 3 (V-)
The presence of vegetation had no effect on TSS concentration reduction and only a small effect on nutrient concentration reduction. The only detectable difference between treatments, was that TN influent concentration was significantly reduced by Cell 2 but not Cell 3. This has been seen in previous studies, with a positive effect of vegetative uptake of nitrogen (Bratieres et al. 2008, Lucas and Greenway 2008), but little to no effect on phosphorus concentration (Read et al. 2008). I should note, however, that bioretention vegetation’s capacity for nutrient uptake has been reported to change over the course of vegetation establishment (Houdeshel et al. 2015). This bioretention system was still in its first year of installation at the time of this study, and it is possible that a greater difference between treatments will be seen in subsequent years.
Media Water Extractable Phosphorus (WEP) Concentration
As expected, the WEP concentration of all treatments significantly decreased with depth, suggesting phosphorus sorption by the media in a top-down fashion. Over time, however, there was an increase in the concentration of the shallowest 15 cm of all treatments, with the exception of the first 5 cm of Cell 1. This was similar to results found by Muerdter et al. (2015), who noted a loading in shallow depths, but no significant increase in Mehlich-3 phosphorus concentration beyond media background levels past a depth of 12 cm in a seven year old bioretention cell. Together these studies underpin the importance of the first 10-15 cm of bioretention media for phosphorus removal, a sentiment echoed by other bioretention studies focused on other stormwater pollutants such as sediment (Hsieh and Davis 2005), heavy metals (Sun and Davis 2007), and pathogenic organisms (Rusciano and Obropta 2009).
Compost had an immediate and significant effect on media WEP concentration. The highest concentrations of WEP measured in the shallow layers of Cell 1 in June, suggest the low-P compost leached loose, labile forms of phosphorus immediately after placement. The effect persisted into November; however, there was a convergence of average concentration across treatments, possibly due to of some initially leached phosphorus migrating to the lower media or discharging with the effluent. A contribution of phosphorus to bioretention media by compost is expected, and in many ways the goal of organic amendment, but long term loading onto media that has low sorption capacity could result in rapid saturation and decreased P removal. This finding provides a greater understanding of the mechanism by which leaching occurs when a layer of compost is added atop of the bioretention media, as opposed to mixed throughout.
Sands are often used as bioretention media due to their high rates of hydraulic conductivity and storage capacity (Fassman-Beck et al. 2015); however, their P sorption potential is typically low due to their relatively few Al and Fe complexion sites (Xu et al. 2006). This lower capacity could result in a decreased treatment lifespan of a bioretention cell receiving high loads of P, as complexion sites become saturated and begin desorption (Del Bubba et al. 2003). My study has shown that P loading onto sand-based bioretention media can be immediately apparent in the first year of installation, and exacerbated by the presence of a low-P compost layer. Also, use of bioretention cells to treat agricultural production facilities is a relatively novel idea; P loading in this landscape is generally greater than the urban sites where bioretention has been most extensively studied. Future research should consider the long term implications of sand-based bioretention media accepting this level of P loading.
Panicum virgatum Survivorship and Biomass
Overall, Switchgrass proved to be a vigorous and hardy species, well suited for bioretention. When planted in compost, the Switchgrass grew rapidly and fully covered the area of the bioretention cell within the first few months. Even without compost, the plants were still able to survive an abnormally dry growing season and significantly reduce the concentration of nutrients in runoff. This fact leads us to recommend Switchgrass and other native C4 grasses capable of tolerating low nutrient environments in bioretention projects that abstain from compost amendment. Additionally, these types of plants may be well suited for bioretention in agricultural landscapes, where they can be readily harvested and utilized for other purposes such as biofuel or livestock bedding. One challenge of using Switchgrass was that it was a warm-season grass that took several months to establish after planted in spring and required frequent watering through early summer. Bioretention projects considering Switchgrass should be aware of its seasonality and plant accordingly to minimize extra management.
Low-P compost was found to have a significant effect on the total biomass growth of planted Switchgrass, but interestingly, little to no effect on the total survivorship. This suggests that low-P compost may aid in the initial growth of vegetation in bioretention cells, but Switchgrass is able to survive and uptake a similar concentration of phosphorus when planted directly in sand-based media. Since there were detectably greater phosphorus concentrations Cell1’s (C+V+) effluent and media, it can be assumed that the difference in biomass does not account for a total greater uptake of phosphorus. That is, the larger plants in this cell were not taking up enough phosphorus to overcome what was added by the compost in the first year. However, as plants mature and as nutrient leaching diminishes with time, vegetative uptake may be able to overcome the effect and result in a net phosphorus removal.
The role of these plants in bioretention phosphorus cycling should continue to be studied. Switchgrass biomass production has been shown to increase with age (Frank et al. 2004). By the time of the first season’s harvest, the Switchgrass in Cell 1were at or close to their maximum size (1-1.5 m), but the Switchgrass in Cell 2 were significantly smaller, some near seedling size. The Switchgrass in Cell 2 (C-V+) are expected to continue to grow and eventually approach the same biomass as those planted in Cell 1. Plant uptake of phosphorus has been noted as a primary mechanism for removal in previous bioretention studies (Lucas and Greenway 2008); future bioretention studies should consider quantifying the change in nutrient water and media concentration as plants mature.
Several questions remain about the role of Switchgrass on bioretention performance that were not answered in this study. The Switchgrass used in this study were an open-pollinated variety with intended use for ecological restoration; it is possible that varieties bred for other purposes could have a different effect on bioretention performance. For example, varieties bred for biomass accumulation may have a more pronounced affect on pollutant uptake (Reed et al. 1999) or varieties used for slope stabilization may influence soil media structure (Simon and Collison 2002). Also, different placement of Switchgrass in bioretention cells could be explored; Switchgrass in my studied were evenly spaced in rows, but grouping individuals could allow for natural benefits of intraspecific symbiosis such as shared mycorrhizal communities(Hart et al. 2003). Finally, a monoculture was studied in this experiment for the purpose of isolating the effects of a single species. Floristic diversity increases the productivity and chemical cycling of an ecosystem (Tilman et al. 1997); future studies could consider comparing how monocultures in bioretention compare in performance to communities with several species adapted to cohabitating with Switchgrass.
Uncertainty and Future Research on Bioretention Hydrology
Several issues limited my ability to study the cells’ hydrology. One issue was the abnormally dry weather conditions of our sampling season; Vermont was in a Stage 2 drought during my sampling (NOAA National Centers for Environmental Information 2016). Runoff from low volume storms (< 13.5 m3) can be fully retained by the forebay in this system; in an effort to increase flow through the cells such that bioretention performance could be better evaluated, including low-volume events, a shallow trench (8.8-cm deep, and approximately 1-m wide) was carved across the forebay from the inflow swales to the forebay outlet structure, which is the ihanflow to the bioretention cells. This trench channeled stormwater directly to the bioretention cells’ inlet, minimizing forebay residence time and allowing the sampling of more low-volume storms. This feature is temporary, and the forebay will be restored at a later date. A comprehensive water budget of the cells should be completed once the system is restored to its intended designed state and normal weather conditions can be assumed.
Another issue was that after the first half of the sampling season, it was observed that higher intensity storms delivered a greater volume to the center cell, Cell 2. As a modification, on October 15, a fiberglass baffle was installed in the splitting structure to dissipate flow, reduce turbulence, and more evenly distribute the influent volume across the three bioretention cells for all storm intensities. Equal inflow volume splitting to the three cells could not be assumed up to this point, and therefore a mass balance and measure of media retention could only be obtained for two storms (October 22 and 28). A more detailed mass and water balance model can be produced for these cells as more storm events are sampled with the assumed equal inflow splitting.
Finally, the effects of the experimental treatments on cell hydrology were not explored. The additional layer of compost in Cell 1 is expected to provide extra storage and the positive influence on growth may increase evapotranspiration rates compared to Cell 2. Similarly, the evapotranspiration of the Switchgrass present in Cell 2 could influence moisture content and storage compared to Cell 3. These may have a significant effect on the retention and pollutant removal of the bioretention cells overall, and should be considered in future years of sampling.
Conclusions
This study provides a better understanding of the benefits and drawbacks of using low-P compost in bioretention cells. On the one hand, the bioretention cell with the low-P compost had vigorous establishment of Switchgrass in the first year with only slightly higher effluent stormwater nutrient concentrations than the bioretention cell with no compost. However, on the other hand, the presence of this compost may shorten the treatment lifespan of a bioretention cell, as it was shown to significantly increase the concentration of labile phosphorus within media that may have low sorption potential. Also, while it was important for growth, the compost had no effect on the number of plants that survived, suggesting that at least some types of vegetation can successfully establish without added compost, low-P or otherwise.
Therefore, my recommendation of low-P compost for use in bioretention projects is dependent upon the situation. When fast plant establishment is required, I recommend its use over compost derived from manure or other enriched feedstocks which have been shown to have a greater leaching potential than leaf-based composts in other bioretention experiments (Hurley et al. 2017). However, I acknowledge that the presence of the low-P compost still had some deleterious effects upon stormwater nutrient treatment and encourage exploration of no-compost options whenever possible, especially in nutrient-sensitive watersheds.
At the time of this study, there was little to no effect of the presence of vegetation on either nutrient treatment or media phosphorus concentration in the Miller Research Complex Bioretention Cells. I note that my comparison took place after only five months, and it is possible that an effect could become apparent as the vegetation grows. I encourage further research into species that can survive low nutrient bioretention media. Finally, continued testing of different Switchgrass varieties could increase choice among practitioners and encourage nutrient sensitive bioretention designs.
Bioretention cells, regardless of their soil and vegetation selection, were highly effective at the removal of particulate nutrients and sediment from stormwater in a mixed-use agricultural setting. Farmers who choose this technology to treat runoff from their production facilities can expect high removal rates of sediment and phosphorus; however the removal of nitrogen in the form of nitrate will continue to pose a challenge. My research suggests that low-nutrient alternative composts and the use of a hardy, farm plants like Switchgrass in bioretention can improve nutrient removal performance in mixed-use agricultural landscapes. I encourage bioretention as one form of stormwater treatment technology that should be considered by farmers as water quality concerns shift into the agricultural sector and operations are required to manage their nutrient-rich stormwater on site.
Education & Outreach Activities and Participation Summary
Participation Summary:
2016 Student Green Stormwater Infrastructure Symposium Poster at the University of Vermont
2016 International Low Impact Development Conference Poster in Portland, Maine
Tour of Miller Research Center Bioretention Cells for undergraduate landscape design class and exhibition for regulators (EPA, DEC, local stormwater professionals).
Publication in Science of the Total Environment (Pending)
Project Outcomes
My and my advisor's attitude towards sustainable agriculture remain optimistic. Our research suggested that it is possible to have a significant positive affect on water quality in mixed-use agricultural landscapes using low-impact landscape design. In rural areas like Vermont, this will be important in meeting larger watershed goals of water quality. It is our intention to further promote bioretention as one sustainable option to treating high nutrient concentration runoff originating in mixed-use agricultural landscapes. Research on these filters will continue, exploring more design options and quantifying treatment potential of soil and vegetation pallets.