1. Introduction
Rainfall and snowmelt on urban or agricultural landscapes typically produces enough water to generate flow over the surface of the landscape, which is called stormwater runoff. In agricultural watersheds, there are often buried perforated pipes called drain tiles that collect soil moisture and discharge it downstream, which allow the soils to be farmed. The flow from these drain tiles, called tile drainage, has important characteristics such as volume and rate as well as water quality characteristics such as temperature, pH, and pollutant concentrations. One such pollutant of concern is phosphorus, which can be either particulate (≥0.45 µm in size) or soluble (dissolved and <0.45 µm). In stormwater runoff, tile drainage, and surface water bodies, soluble phosphorus is most often in the form of phosphate (
PO43−) [
1]. Though many terms such as dissolved phosphorus, soluble reactive phosphorus, and ortho-phosphate have been used, this article will use the term “phosphate” to describe the soluble reactive phase of phosphorus.
In agricultural watersheds, phosphate sources include natural organic matter, crop biodegradation, natural and synthetic fertilizers, and livestock waste, among others. Phosphate is more bioavailable than particulate phosphorus [
2], and thus typically limits biological growth in temperate non-marine surface water ecosystems [
3,
4,
5]. When in excess, however, phosphate often generates nuisance algae blooms and eutrophic conditions in these ecosystems [
6].
Sedimentation (i.e., particle settling) and filtration (i.e., sieving) are two mechanisms used by typical urban stormwater control measures (SCMs) such as wet ponds, dry ponds, and sand filters to capture particulate forms of phosphorus. Phosphate, however, is often not captured by most SCMs because a chemical or biological process is necessary to do so. For example, sand filters capture approximately 80% of total suspended solids [
7] by filtration. Particulate phosphorus, which is about 55% of total phosphorus [
7], will also be captured but phosphate remains untreated.
Many phosphate-controlling strategies have been proposed and utilized for agricultural runoff, including optimizing fertilizer P use-efficiency, refining animal feed rations, using feed additives to increase P absorption by the animal, treating manure to lower its soluble P content, moving manure from surplus to deficit areas, managing soil P levels, finding alternative uses for manure other than land application [
8,
9]. Other strategies include conservation practices for critical areas of P export from a watershed, such as reduced tillage, buffer strips, terracing, and cover crops [
8,
9].
Constructed wetlands have been used to capture phosphorus from agricultural drainage with removal rates of 2% to 44% for total phosphorus [
10,
11,
12,
13,
14]. One study of 17 intensively studied constructed wetlands reported total phosphorus capture of 1 to 88% and dissolved reactive phosphorus (DRP) capture of −19 to 89% [
15]. The large capture range for both total phosphorus and DRP was found to be due to variations in soluble fraction, runoff volume, wetland age, surface area to contributing area ratio, and other site-specific factors. Particulate capture in these constructed wetlands were reported to be predictable based on flow and settling characteristics, but capture of DRP was less predictable, likely due to uncontrolled transformations within the wetlands. These studies suggest the need for a treatment system that consistently captures both particulate phosphorus and phosphate.
SCMs must be improved to capture soluble pollutants such as phosphate [
16]. Previous studies have found that adding metals such as steel wool or elemental iron to sand filter media resulted in significant removal of phosphate [
17,
18,
19] from synthetic and natural stormwater. As stormwater passes through the sand mixed with iron filter media, the elemental iron rusts to form iron oxides, which binds with phosphate via surface adsorption. These iron enhanced sand filters (IESFs) are substantially different than constructed wetlands. Constructed (and natural) wetland systems rely on shallow water depth (typically < 30 cm) and long residence time to capture pollutants through settling of particles and bio-chemical interactions with vegetation, the soil, and organisms within the water and soil. IESFs, however, use the physical process of filtration to capture particles and a chemical sorption reaction with iron to capture phosphate [
17,
18]. Thus, IESFs typically have a shorter residence time and deeper water depth (typically 30–180 cm) because the water is treated quickly by the IESF media.
With this knowledge, Wright Soil and Water Conservation District (SWCD) installed an IESF in 2012 to limit the total phosphorus and phosphate load from an agricultural subwatershed. A drain tile was intercepted and re-routed away from an existing ditch system and into the IESF. The performance of this IESF was assessed by monitoring natural rainfall-induced tile drainage events for two rainy seasons. The main objectives of this study were to: (1) assess the performance of a three- to four-year-old IESF with regards to the capture of total phosphorus and phosphate from agricultural tile drainage; (2) investigate maintenance requirements; and (3) compare measured data to previously published performance of IESFs.
3. Results and Discussion
The performance of the IESF for capturing total phosphorus and phosphate from agricultural tile drainage was assessed by monitoring natural rainfall-induced tile drainage events during 2015 and 2016. Routine and non-routine maintenance performed since installation of the IESF resulted in a reduction between influent and effluent total phosphorus and phosphate loads on the receiving water body for all rainfall-induced events measured during this study. The performance results of the IESF based on monitoring is discussed in the following sections, though all data and results, which includes some information based on grab samples, are reported in [
27].
3.1. Rainfall Event Performance and Continuous Monitoring
3.1.1. Rainfall and Tile Drainage Volume
Rainfall depth varied from 0.05 to 7.32 cm per event with an average of 2.23 ± 0.70 (α = 0.05,
n = 30) cm. As is typical for rainfall, the average (2.23 cm) was greater than the median (1.82 cm) for the rainfall depth, which was also true for the rainfall duration, average rainfall intensity and tile drainage volume because several large rainfall events skew the average towards larger values and many small events skew the median towards smaller values. Annual precipitation measured at a municipal airport (Buffalo, MN; KCFE) approximately 8.7 km northwest of the IESF was 53.7 cm for the 2015 water year (1 October 2014–30 September 2015) and was 82.4 cm for the 2016 water year (1 October 2015–30 September 2016). The total measured rainfall depth at the IESF site in 2015 was 29.97 cm and was 37.1 cm in 2016, which are approximately 50% of the average annual precipitation because it excludes precipitation that occurred during non-sampled rainfall events and several snowmelt and rainfall events that occurred outside the monitoring periods. Multiplying the total measured rainfall (66.95 cm) by the contributing watershed (~7.45 ha) results in a total rainfall volume of approximately 49,878 m
3. Tile drainage per event varied from 28.2 to 487 m
3 with an average of 178.5 ± 48.4 (α = 0.05) m
3 and a total tile drainage volume of 5891 m
3. Dividing the tile drainage (5891 m
3) by the rainfall volume (49,878 m
3) results in a tile drainage “runoff coefficient” of approximately 0.12. To determine if this agricultural watershed was representative of the region, the rainfall and tile drainage volume were compared [
27] to other agricultural sites monitored by Discovery Farms Minnesota (DFM). The monitored rainfall depth at the IESF site was 10–25% less than the average rainfall measured at the DFM sites because equipment failure at the IESF site prevented measurement of some rainfall events during the monitoring period of June through November in both 2015 and 2016. The tile drainage “runoff coefficient” for the IESF (0.12) was similar to the average “runoff coefficient” measured at the nine DFM sites (0.125) [
27]. This suggests that tile drainage characteristics of the IESF site are consistent with other agricultural sites in a similar climate for the time period between 2011 and 2016.
3.1.2. Rainfall-Induced Flow vs. Baseflow
Agricultural watersheds produce tile drainage whenever there is excess soil moisture. When monitoring equipment was operational, tile drainage was almost continuous. It is unclear whether this was typical, or if rainfall and runoff characteristics for 2015 and 2016 were unusual. It is also unclear if or to what extent the upstream wetland affected the tile drainage conditions during the monitoring study.
The tile drainage measured from 26 June to 20 November 2015 was 3167 m3. Of this volume, approximately 2489 m3 was rainfall-induced tile drainage in which samples were collected, approximately 221 m3 was non-sampled rainfall-induced tile drainage, and approximately 457 m3 was baseflow. This corresponds to approximately 86% rainfall-induced tile drainage (7% non-sampled) and 14% baseflow. The tile drainage measured from 14 May to 7 November in 2016 was 6388 m3. Of this volume, approximately 3653 m3 was sampled rainfall-induced tile drainage; 2282 m3 was non-sampled rainfall-induced tile drainage; and approximately 454 m3 was baseflow. This corresponds to approximately 93% rainfall-induced tile drainage (36% non-sampled) and 7% baseflow. Overall for 2015 and 2016, 90% of the tile drainage was rainfall-induced (26% non-sampled) and 10% was baseflow. Approximately twice as much volume was measured in 2016 compared to 2015, though the baseflow amount was nearly identical (457 m3 in 2015 vs. 454 m3 in 2016). This was expected because the increase in tile drainage was in response to additional rainfall.
3.1.3. Total Phosphorus
Summary statistics for total phosphorus are shown in
Figure 3 and reported in detail in [
27]. Analysis of volume-weighted composite samples collected as part of this project revealed that the influent total phosphorus Event Mean Concentration (EMC) varied from 138 to 1516 μg/L with a volume-weighted average EMC of 370 ± 168 (α = 0.05,
n = 20) μg/L. Capture of total phosphorus by the IESF was the combination of particulate phosphorus captured on or within the IESF by filtration and capture of phosphate within the IESF media through sorption with iron. Effluent total phosphorus EMC varied from 56 to 343 μg/L with a volume-weighted average EMC of 125 ± 30 (α = 0.05) μg/L. The total phosphorus load was reduced by 66.3% from 1274 g to 429 g. The arithmetic average (sum divided by
n) influent and effluent load was 63.7 ± 22.1 g and 21.5 ± 9.6 g, respectively. Because the influent volume is assumed to equal the effluent volume for every rainfall-induced tile drainage event, the total load reduction percent was mathematically equivalent to the average load reduction (66.3%). The 95% confidence interval of the average as given by (2) was calculated based on the standard deviation of percent load reduction for 20 events. Thus, the total phosphorus load decreased by 66.3% ± 6.7% (α = 0.05,
n = 20).
Rainfall-induced tile drainage, total phosphorus EMC and load for each event in 2016 are shown in
Figure 4. All events exhibited positive total phosphorus capture (i.e., effluent EMC < influent EMC). Two events had influent EMCs greater than 1000 μg/L, but 13 of the 20 events had influent EMCs between 200 and 800 μg/L. By contrast, only one event had an effluent EMC greater than 200 μg/L and 9 of the 20 events had effluent EMCs less than 100 μg/L for total phosphorus. Thus, the total phosphorus EMC was reduced for all events.
The total phosphorus load reduction for each event varied from 42% to 95% with an overall total phosphorus load reduction of 66.3% ± 6.7% (α = 0.05,
n = 20). A substantial portion of the influent total phosphorus load (323.6 g, 25%) was contributed by two events: 19 August 2016 (178.3 g) and 15 September 2016 (145.3 g), as shown in
Figure 4. These two events contributed approximately 17.8% of the rainfall-induced tile drainage in 2016. Three other events each contributed between 100 and 130 g of total phosphorus load, which accounts for approximately 344.6 g (27%) of the overall influent total phosphorus load and 26% of the tile drainage. Thus, five of the 20 events (25%) contributed approximately 52.5% of the influent total phosphorus load and 43.7% of the tile drainage. This shows that a relatively small number of events with large tile drainage can contribute most of the total phosphorus load and tile drainage in a region where thunderstorms are common.
Total phosphorus was not measured in samples from 2015, but the total phosphorus influent load for all 2016 events was 1274 g. If this load is attributed to the entire 7.45-ha contributing watershed, the total phosphorus load per land area is approximately 171 g per ha (g/ha). This load only represents tile drainage between June and November 2016, and excludes any load that may have been contributed by non-sampled events. By comparison, ten Discovery Farms Minnesota (DFM) sites reported total phosphorus loads for the entire 2016 water year ranging from 11.2 to 157 g/ha for tile drainage [
28]. The tile drainage at the IESF site appears to contain more total phosphorus load per area in six months than the annual load at DFM sites.
3.1.4. Phosphate
Summary statistics for phosphate are shown in
Figure 5 and reported in detail in [
27]. The influent total phosphorus EMC varied from 18 to 358 μg/L with a volume-weighted average EMC of 162 ± 33 (α = 0.05) μg/L. All events exhibited positive capture (i.e., effluent EMC < influent EMC) of phosphate, likely due to sorption with iron oxide surfaces in the IESF [
17,
18]. As a result, the effluent EMC varied from 8 to 127 μg/L with a volume-weighted average EMC of 57 ± 13 (α = 0.05) μg/L. The phosphate load was reduced by 63.9% from 956 g to 345 g. The arithmetic average (sum divided by
n) influent and effluent load was 30.8 ± 13.9 g and 11.1 ± 5.1 g, respectively, and the 95% confidence interval of the average as given by (2) was calculated for 31 events. Thus, the phosphate load decreased by 63.9% ± 7.7% (α = 0.05,
n = 31).
Rainfall-induced tile drainage, phosphate EMC and load for each event in 2015 and 2016 are shown in
Figure 6. While the influent phosphate EMC appears to increase from 2015 to 2016, the volume-weighted average EMCs were 158 ± 37 μg/L and 165 ± 54 μg/L, respectively. The volume-weighted effluent phosphate EMCs were also similar in 2015 and 2016; 56 ± 18 μg/L and 60 ± 19 μg/L, respectively. Thus, there was minimal, if any, change in EMC from 2015 to 2016.
The influent phosphate load varied from 2.8 to 157 g per event with an average load of 31 ± 14 (α = 0.05,
n = 31) grams per event and a total influent load of 956.1 g (
Figure 5). The effluent load varied from 0.4 to 61.7 g per event with a volume-weighted average of 15.6 (α = 0.05) grams per event and a total effluent load of 345.1 g. The load reduction for each event varied from 9% to 87% with an overall phosphate load reduction of 63.9% ± 7.7% (α = 0.05,
n = 31).
The influent phosphate load for all events in 2015 and 2016 was 956.1 g. A substantial portion of that load (271.9 g, 28%) was contributed by two events: 19 August 2016 (157 g) and 16 July 2015 (114.8 g), as shown in
Figure 6. These two events contributed 16.5% of the tile drainage. Seven other events each contributed between 50 and 80 g of phosphate load in 2015 and 2016, which accounts for approximately 461 g (48%) of the overall influent phosphate load and 36.7% of the tile drainage. Thus, nine of the 31 events (29%) contributed approximately 76.7% of the influent phosphate load and 53.1% of the tile drainage. This further shows that a relatively small number of events with large tile drainage can contribute most of the phosphate load and tile drainage in an area where thunderstorms are common.
3.2. Soluble Fraction
A pollutant can either exist in soluble phase (e.g., aqueous) or in particulate phase (e.g., attached to a mineral or incorporated into organic matter). The total concentration is the sum of the soluble concentration and particulate concentration, where the soluble portion is defined as smaller than 0.45 μm [
24]. The soluble fraction can be calculated by dividing the soluble concentration by the total concentration. Calculating the soluble fraction can provide insight into the relative performance of the different treatment mechanisms. For sand filters, a portion of the particulates are captured by filtration on the surface and within the media. The IESF also captures a portion of the soluble phase of phosphorus (phosphate) through adsorption with iron.
The soluble fraction of phosphorus (%) for the rainfall event samples collected throughout this project are presented in
Figure 7. The influent soluble fraction varied from approximately 5% to 90%, with an average of 43.2% ± 12.6% (α = 0.05,
n = 18), which is similar to values measured in urban stormwater [
17,
18]. The effluent soluble fraction varied from approximately 15% to 75% with an average of 43.5% ± 10.0% (α = 0.05,
n = 18). For comparison, nine Discovery Farms Minnesota (DFM) sites reported an average of 60% soluble fraction for tile drainage samples collected between 2011 and 2016. The soluble fraction reported by DFM is expected because tile drainage has already been filtered by the soil. For unknown reasons, the 2015–2016 influent to the IESF had an unusually high percentage of particulate phosphorus.
Performance data previously discussed in
Section 3.1 show that phosphate and total phosphorus were captured by the IESF at approximately the same rate and the average soluble fraction of influent and effluent were similar (i.e., 43% and 44%, respectively). Thus, particulate phosphorus and soluble phosphate are captured at roughly the same level of performance. Relative to sand filters in urban settings, the particulate phosphorus capture performance was low, likely due to the smaller size of particulates in the agricultural tile drainage.
3.3. Hydraulic and Phosphate Loading Rate
The IESF in this study was constructed in October 2012. Rainfall and flow rate data were not collected at the IESF until June 2015. Rainfall, however, is measured at a municipal airport (KCFE, Buffalo, MT, USA) approximately 8.7 km northwest of the IESF. The rainfall data measured at KCFE was correlated (R
2 = 0.869) to the rainfall data measured at the IESF during this project [
27], as
where
RainIESF = precipitation measured at the IESF site,
RainKCFE = precipitation measured at the Buffalo Municipal Airport (KCFE). Using Equation (3), the precipitation at the IESF can be predicted from historical rainfall data at KCFE for the periods when precipitation was not measured at the IESF. Approximately 253.6 cm of precipitation fell at KCFE between 30 October 2012 and 18 October 2016, which can be extrapolated to total precipitation depth of approximately 301.2 cm at the IESF using Equation (3).
Multiplying the contributing area (7.45 ha) by the rainfall depth (301 cm) yields a predicted rainfall volume of approximately 224,500 m3. Given an estimated tile drainage “runoff coefficient” of 0.12, the estimated tile drainage was 27,000 m3, which corresponds to the total volume of water treated by the IESF since it was constructed. Dividing the volume treated (27,000 m3) by the surface area of the IESF (92.9 m2) yields a depth treated of approximately 290 m.
Overall, the performance of the IESF in this study (63.9% phosphate and 66.3% total phosphorus load reduction) is comparable to other agricultural treatment practices and studies of IESFs. The IESF captures more phosphorus than constructed wetlands, for both total phosphorus (2–44%, [
10,
11,
12,
13,
14]) and phosphate (9%, [
12]), though the data falls within the upper range reported for a study of 17 intensively studied constructed wetlands (TP = 1–88%; DRP = −19 to 89%, [
15]). Laboratory experiments of IESFs in previous studies found an average of 88% phosphate capture with a total treated depth of 200 m [
18]. The amount of tile drainage water treated by the IESF in this study from October 2012 to October 2016 was approximately 290 m of treated depth, which exceeds previously investigated treated depths (200 m) [
17,
18]. An IESF trench after less than one year of operation exhibited an average of 60% phosphate load reduction for 7.2% iron by weight, and 78.8% phosphate load reduction for 10.7% iron by weight [
29]. The measured performance is also expected to be less in field studies compared to laboratory studies due to ion competition with phosphate of natural stormwater and tile drainage compared to laboratory synthetic stormwater.
3.4. Maintenance
Regular, routine maintenance began within one or two years of construction and consisted of Wright SWCD staff visiting the site once or twice per month to (1) remove vegetation, iron ochre, and algae from the IESF, and (2) scrape and level the surface as needed. These activities occurred during the months of May through September of each year and required one or two individuals less than approximately one hour each to complete per site visit. In addition, non-routine maintenance was needed in May 2016 to remove a substantial accumulation of vegetation, iron ochre, and algae from the surface and required two individuals for approximately 2 h each.
Iron ochre is a waste product from bacteria that oxidize dissolved minerals such as iron and was visible on the IESF as a rust colored sludge when wet and as a rust colored thin crust/cake when dry. It is likely that the dissolved iron in the water from the tile drainage was sufficient to support bacteria that produce iron ochre. The accumulation of iron ochre and subsequent biofouling reduced the hydraulic capacity of the IESF in locations near the inlet to near zero, resulting in small pools of water between rainfall-induced tile drainage events. Algae sometimes grew within standing water on the IESF surface, and was also removed during routine maintenance. The combination of iron ochre, algae, and biofouling caused “creeping failure” on the surface of the IESF, moving slowly from the inlet towards the outlet. If vegetation, iron ochre, and algae were not removed during routine maintenance, accumulation would begin to clog the entire IESF surface and prevent treatment. These maintenance activities are recommended as part of the management of other, similar IESF systems.
5. Conclusions
An iron enhanced sand filter (IESF) was installed in Wright County, MT, USA to reduce soluble phosphate and total phosphorus loads from agricultural tile drainage. For this study, monitoring equipment was installed to collect volume-weighted composite samples during rainfall-induced tile drainage events of 2015 and 2016. Approximately 90% of the measured tile drainage corresponded to rainfall-induced events and approximately 10% corresponded to baseflow. During the study period, 33 rainfall events were monitored and IESF capture performance was determined for phosphate and total phosphorus.
The rainfall depth of events measured from approximately June through November in 2015 and 2016 varied from 0.05 to 7.32 cm per event with an average of 2.23 ± 0.70 (α = 0.05, n = 30) cm, a total of 66.95 cm, and an effective tile drainage “runoff coefficient” of 0.12. The total phosphorus load reduction varied from 42% to 95% with a volume-weighted average reduction of 66.3% ± 6.7% (α = 0.05, n = 20) in 2016. The phosphate load reduction varied from 9% to 87% with a volume-weighted average reduction of 63.9% ± 7.7% (α = 0.05, n = 31) in 2015 and 2016. The total phosphorus load reduction (66.3%) was similar to the phosphate load reduction (63.9%) because the IESF removed both particulate phosphorus and soluble phosphate. In addition, the influent soluble fraction varied from approximately 5% to 90% with an average of 43.2% ± 12.6% (α = 0.05, n = 18) and effluent soluble fraction varied from approximately 15% to 75% with an average of 43.5% ± 10.0% (α = 0.05, n = 18). The average influent and effluent soluble fractions are similar, thus suggesting that particulate phosphorus and soluble phosphate are captured at approximately the same level of performance for agricultural tile drainage.
Maintenance of the IESF consisted of: (1) removal of vegetation, iron ochre and algae; and (2) leveling and scraping of the IESF surface. This occurred approximately once or twice per month during the growing season (May to September) of each year and each occurrence required typically less than one hour per person for one or two people. Routine and non-routine maintenance performed since installation of the IESF resulted in reduction of total phosphorus and phosphate loads on the receiving water body for all rainfall-induced tile drainage events measured during this study. If vegetation, iron ochre, and algae were not removed during routine maintenance, accumulation would clog the IESF surface and prevent treatment. This level of maintenance was satisfactory to ensure proper flow rate and contact between the water and the IESF media and is expected to continue throughout the life of the IESF. These maintenance activities are also recommended as part of the management of other, similar IESF systems.