Introduction
Roadways are used by vehicles and livestock within farm boundaries (termed lanes in New Zealand). On dairy farms, roadways are used by dairy cattle twice daily to go to the milking shed, during which time they deposit dung on the roadway. Runoff from these roadways is acknowledged as a significant source of contaminant transfer to nearby streams (Hively et al., 2006). However, apart from a few studies they remain an understudied part of farm scale losses. Monaghan & Smith (2012) found that the concentration of faecal contaminants in runoff decreased with increasing distance away from the milking parlour. Furthermore, McDowell et al. (2007) found that in one case, runoff from a roadway accounted for 80% of the phosphorus (P) load in a nearby stream. However, the amount of dung-P available for loss is highly variable depending on deposit weight, P content and number (Vadas et al., 2015).
Our objective was to determine the concentration of contaminants in roadway runoff and determine if the load and concentration of contaminants (P, Escherichia coli [E. coli] and suspended sediment [SS]) could be decreased with a filter material as has been used to reduce P and SS loads in runoff from grassland and cropland (Ballantine & Tanner, 2010; Buda et al., 2012; Karczmarczyk et al., 2016). We modified the filter material (steel melter slag) to increase P sorption given the sensitivity of downstream waterbodies in the study area. To place the P results of the present study into a broader context an uncertainty analysis was conducted using ranges of values gathered from the literature pertaining to deposit weight, P content and number of deposits per hectare of roadway.
Materials and methods
Site description and field setup
The Mangakino stream feeds into the P-limited Lake Rerewhakaaitu (Abell et al., 2010, 2011), one of the Rotorua lakes. Land use within the catchment is representative of the surrounding area with a dominance of pastoral agriculture (70% dairy, 7% sheep and beef, 15% forestry blocks and 8% other mixed land uses). In July 2006, 200 L of a 1% solution of aluminium chlorohydrate (ACH; Orica Chemicals, Newmarket, Auckland, NZ) was sprayed onto 2 t of steel melter slag; hereafter called ACH-altered slag. We chose to use ACH-altered slag because of its high P sorption affinity and low cost and toxicity compared to other readily available products (McDowell et al., 2008). The ACH-altered slag was placed in a 50-m long ditch that had been dug alongside one half of the roadway. On the other side, a similar ditch was dug but no ACH-altered slag installed (Figure 1). The roadway cut across the stream, resulting in runoff discharging into the stream. At these points, a perforated polyvinyl chloride (PVC) pipe was connected to tipping-buckets to collect ACH-altered slag and control runoff (Figure 1). The tipping buckets were calibrated to catch 1% of sample after each tip. Samples from the tipping buckets were collected in response to rainfall events (>10 mm) for 12 mo. A subsample of each water sample was filtered (<0.45 μm cellulose acetate syringe filter) and measured for dissolved reactive P (DRP) and, after persulfate digestion, total dissolved P (TDP). An unfiltered subsample was also measured for total P (TP) after persulfate digestion (Eisenreich et al., 1975). All P determinations were made colorimetrically (Watanabe & Olsen, 1965). Dissolved organic P (DORP) was obtained by the difference of TDP−DRP, and particulate P by the difference of TP−TDP. SS was determined on the remaining sample volume by filtration through a GF/A glass fibre filter paper and weighing the oven-dried (105°C) residue. The faecal indicator bacteria – E. coli was enumerated using the Colilert® media and the Quanti-Tray® system (IDEXX Laboratories, Westbrook, ME, USA). Data from runoff events during the year were checked for normality and log-transformed if necessary, before being analysed via a two-tailed t-test of paired data. Data for concentrations were multiplied by the volume of runoff to calculate loads on a per event basis. These were summed across all events and expressed on a per hectare basis (viz. yield) assuming a catchment area of 200 m2.
Uncertainty analysis
An uncertainty analysis was conducted specifically for P using an Excel-based Monte Carlo model. Inputs into the model were data ranges as in Table 1 for dung wet weight (kg), number of deposits on roadway per ha, dung total P concentration (g/kg dry weight). For the Monte Carlo analysis of total P deposited on a 1 ha section of roadway, we assumed all input variables had a uniform distribution. The model was run 1,000 times, whereby the model selected a value for each parameter within the range specified in Table 1. This enabled a probability distribution of outcomes to be achieved. The results of the present study were then compared to this full range of possible outcomes and loads were compared using the actual runoff data.
Weight, frequency and P concentration of dairy cattle dung used in the uncertainty analysis
| Parameter | References | |
|---|---|---|
| Dung wet weight (kg) range = 1.5–2.7 | • | 1.5–2.6 kg (Haynes & Williams, 1993) |
| • | 2.0 kg (Krol et al., 2016; Bacher et al., 2018) | |
| • | 2.5 kg (Bell et al., 2015) | |
| No of deposits on roadway/ha cow = 1–1.5 | • | Dairy cattle defecate 10.5 times/cow per day (Oudshoorn et al., 2008). Therefore, a low percentage of this could be on roadways |
| • | Davies-Colley et al. (2004) investigated a 200 m roadway, 245 cows and showed that five deposits were made during each milking event | |
| Dung P concentration (g/kg dry weight) = 4–8 g P/kg | • | Estimated for lactating dairy cattle (McDowell, 2006; Vadas et al., 2015) |
Results
Field trial
Data for the concentration of P fractions, SS and E. coli in runoff from the ACH-altered slag and control section of the roadway are given in Table 2, along with the probability of a significant difference (P < 0.05). In all cases, concentrations were greater in the control than in the ACH-altered slag sections for DRP, PP, TP and SS. Concentrations of DRP, TP and SS were well in excess of recommended limits for freshwater eutrophication in disturbed (i.e. agricultural) lowland streams and for aquatic ecosystems and E. coli concentrations deemed fit for contact recreation (Australian and New Zealand Governments, 2018). However, this should only be taken as an indication of the potential to enrich the receiving stream and runoff will likely be diluted during stormflow. Contextually, concentrations of at least TP were like those noted for septic tank discharge (1–14 mg/L) (Withers et al., 2011).
Mean ± s.e. (range) of the volume of runoff and concentrations of P fractions, sediment and E. coli, and the probability of a significant difference between the ACH-altered slag and control roadways
| Parameter | ACH-altered slag | Control | Significant difference (probability for t-test) |
|---|---|---|---|
| Runoff (L)1 | 367 ± 69 (53–1,760) | 588 ± 123 (53–1,751) | 0.574 |
| DRP (mg/L) | 0.060 ± 0.013 (0.004–1.560) | 0.301 ± 0.053 (0.004–2.312) | 0.001 |
| DOP (mg/L) | 0.149 ± 0.049 (0.001–0.906) | 0.135 ± 0.022 (0.001–0.571) | 0.798 |
| PP (mg/L) | 0.604 ± 0.048 (0.240–6.860) | 2.046 ± 0.556 (0.010–6.640) | 0.010 |
| TP (mg/L) | 0.813 ± 0.136 (0.176–7.623) | 2.482 ± 0.590 (0.166–8.853) | 0.006 |
| SS (mg/L) | 395 ± 93 (240–13,680) | 2,720 ± 1,103 (26–4,950) | 0.049 |
| E. coli (cfu 100/mL) | 3,629 ± 1,406 (100–24,190) | 5,755 ± 1,901 (50–14,500) | 0.0882 |
1Number of runoff events = 18.
2Data required log transformation for comparison of means.
The loads across all events along with the percent decrease due to treatment with ACH-altered slag are given in Table 3. Loads from the ACH-altered slag treatment were generally lower than from the control treatment except for DOP (−7%), a P species that commonly exhibits poor sorption characteristics (Andersen et al., 2016). In contrast, 88% of DRP and 96% of PP species were mitigated by the ACH-altered slag inferring that efficient sorption and filtration occurred. The efficiency of removal decreased with increasing event size, becoming ineffective for events with about 1,000 L of runoff, which equated to about 40 mm of rainfall (Figure 2).
Loads of runoff, P fractions, sediment and E. coli in the ACH-altered slag and control roadways and the percentage mitigation (i.e. the fraction of load from ACH-altered slag vs. control roadways)
| Parameter | Mean load for ACH-altered slag | Mean load for control | Percent decrease |
|---|---|---|---|
| Runoff (L/ha) | 3,133,000 | 3,243,750 | 3 |
| DRP (kg/ha) | 0.3 | 2.5 | 88 |
| DOP (kg/ha) | 0.5 | 0.5 | −7 |
| PP (kg/ha) | 0.7 | 17.0 | 96 |
| TP (kg/ha) | 1.5 | 20.0 | 93 |
| SS (mg/ha) | 0.4 | 21.6 | 98 |
| E. coli (cfu) | 4.44E+10 | 1.93E+11 | 77 |
Uncertainty analysis
Using the parameter ranges in Table 1, a probability distribution based on 1,000 runs of the model is shown in Figure 3. This shows that for approximately 50% of outcomes, 2.5 g TP/ha was available to be lost in runoff. When this is converted into a load using the runoff data presented in Table 3 the minimum TP load is 7.7 kg/ha and the maximum is 462.2 kg/ha.
Discussion
It should be noted that this trial did not have spatial replication nor directly measure manure loads. As such our findings should be taken as a proof of concept, informing the design of a more comprehensive study.
Considering that deposition of dung on the roadway occurs daily, concentrations of P lost are about one-fifth of those expected in runoff from a fresh dung patch as documented by McDowell (2006). However, this does depend on the rate of drying, diet and above all climatic conditions.
Installing ACH-altered slag alongside a roadway could be viewed as like an active filter bed used as pre or post treatment for wetlands (Ballantine & Tanner, 2010). These beds are influenced by flow rates and the concentration of inflowing water. In studying active filters with steel melter slag as a retention material, Shilton et al. (2005) noted that overall P removal from inflowing waste (piggery effluent or domestic effluent) water in two steel melter slag filters – one at Ashurst and another at Waiuku (New Zealand) – was 72 and 77%, respectively, but was most efficient in summer and autumn during low flows. Our data indicated a similar effect in that DRP load from the ACH-altered slag treatment was much lower than that of the control section at lower flows, but decreased at higher flows becoming ineffective when >1,000 L of runoff or 40 mm of rainfall occurred resulting in the filter being overtopped (Figure 2).
The cost per kg of P retained by unaltered steel melter slag when used in P-socks in a streambed to remove P from baseflow was estimated to be 30 USD (McDowell et al., 2007), which precluded its use when compared to dosing streams with alum (at about 10 USD/kg P precipitated) (Pilgrim & Brezonik, 2005), but was more cost-effective than dosing the roadway directly with alum (McDowell & Nash, 2012; Smith & McDowell, 2016). However, the difference in cost between ACH-altered slag and unaltered slag is negligible as it is a waste material and the 1% solution of ACH is cheap. The resulting cost effectiveness would be about 3 USD/kg P retained. However, this would vary according to site characteristics (e.g. slope) and frequency of road use by cattle. It is also likely that the P-sorption capacity of the ACH-altered slag would decrease with time as sorption sites became occupied. Although the P sorption capacity of a similar product was not exceeded when included as part of a backfill for an artificial drainage network, this was exposed to much less sediment load (McDowell et al., 2008). We have no data to confirm the longevity of our material beyond the 12-mo length of the trial. It should also be noted that other filter media may be more suited to a specific geographical location. Decision support tools are now available to help match pollutant and filter medium type (Ezzati et al., 2019).
In terms of the uncertainty analysis results the present study falls within the lower range of potential losses, that is, 20 kg/ha. This range represents all possible sites where deposit weight, TP content and number of deposits range within a hectare of roadway. Preliminary data suggests that the use of ACH-altered slag was effective at mitigating P losses in runoff from a roadway (93% of TP) supplying P loss into the Mangakino stream. We therefore recommend the use of a material like ACH-altered slag to remove P from runoff from roadways entering streams. However, the uncertainty analysis did suggest that the deposition of dung-P on the roadway could be higher, which may reduce the efficiency of the ACH-altered to retain P. A reduced efficiency would need to be factored into the design of any filter bed.