2. Riparian Attenuation of Faecal Microbes

2.1 Introduction

There is a widespread perception that fencing to exclude livestock from stream channels and a proportion of riparian land is likely to be the most effective measure in reducing faecal contamination by grazing cattle. Not only does this prevent the deposition of faecal material directly to streams and near channel contributing areas, but also the dense vegetation associated with riparian buffer strips (RBS) reduces the momentum of overland flow, promoting entrapment of faecal material (and other particulates). However, although many studies of sediment and nutrient entrapment in RBS have been conducted, understanding of microbial processes within them is poor, limiting evaluation of their effectiveness. To improve understanding, MAF Policy funded an initial field study of riparian attenuation of faecal microbes during the financial year 2001/02 (Collins et al. 2002). The key findings arising from this research were:

When liquid farm dairy wastes were injected over 2-3 minutes onto saturated pasture soils, irrigated to generate surface runoff, there was a rapid ‘first flush’ of high Campylobacter and E. coli concentrations downslope for at least 5 m.

When effluent injection stopped (but surface runoff continued), concentrations typically decreased by 2-4 orders of magnitude over 60 minutes.

When surface runoff stopped (c. 60 minutes after injection), outflow microbial concentrations were consistently higher than background concentrations, indicating that not all the microbes applied had been flushed out of the plots.

It was not possible to quantify Campylobacter retention because of the imprecision associated with the 3-tube MPN analysis. High background concentrations precluded quantification of E. coli retention.

Grass length had no apparent impact on the retention of Campylobacter or E.coli.

Plot length (in a downslope direction) affected: outflow rate, and timing and magnitude of peak microbial concentration.

Surface runoff 5-12 days after the initial experiments did not remobilise significant numbers of the retained microbes. It is not clear whether this is because of die-off, flushing in sub-surface drainage, retention onto soil particles, or a combination of some or all of these processes.

These findings raised implications for the design (e.g. the impact of plot length) and effectiveness (e.g. the lack of re-mobilisation of microbes in subsequent events) of RBS. However, key questions with respect to riparian attenuation of microbes remained, particularly with respect to the impact of soil type, slope angle, type of faecal material, flow rate, and bypass in subsurface flow, upon RBS effectiveness. Given the potential importance of RBS as a mitigation measure, further plot studies have been conducted within Objective 2 of the PTRRP. There were three key objectives of this latest research:

To determine the impact of flow rate upon microbial attenuation.
To determine the relative importance of subsurface flow (under a RBS) as a transmission pathway.
To compare microbe attenuation under differing types of applied faecal material (liquid dairy effluent versus relatively dry pats).

2.2 Methodology

The study objectives were achieved through a series of plot experiments that involved the flushing of dairy farm effluent and cow pats into sloping grass strips, by surface runoff generated using a ‘sprinkler’. The cow pat experiments also used a rainfall simulator to generate rainsplash erosion of the pats, as would occur under natural rainfall. Both surface and subsurface outflow at the lower end of a plot was sampled for microbial analysis during each experiment. This information, coupled with a known input of microbes to each plot, enabled the attenuation of Campylobacter and E. coli to be quantified.

2.3 Experimental Design

Experimental plots were established for the initial RBS study (Collins et al. 2002) on the Ruakura campus farm, Hamilton. Four of these plots were used in this latest study. The plots lie on a sloping paddock (10-15° ) underlain by a Hamilton clay loam soil, and have not been grazed since October 2001. The plots are 2 m wide and 5 m long.

During the experiments two of the plots had short grass (7-10 cm) to simulate recently grazed pasture, while the other two had long grass (30–37 cm) to simulate a grass buffer strip. Each plot was bound along its sides by sheet metal inserted approximately 5 cm into the soil, to minimise lateral flow of water out the plot. At the lower end of each plot a trough was dug. Sheet metal was pushed horizontally into the exposed soil face, along the width of the plot, approximately 5 cm below the soil surface. Foam sealant was applied from the sheet metal upwards to just below the ground surface, across the width of the plot. This allowed collection of surface flow. The metal sheet overhung open plastic piping into which flowed surface runoff generated on the plot. This piping sloped laterally so that runoff flowed out of the lower end. At a depth of about 30 cm (corresponding to a clay-rich layer that impeded vertical movement of water) a second sheet metal strip was pushed horizontally into the exposed soil face, along the width of the plot. The metal strip overhung plastic piping and this design collected subsurface runoff draining out of the exposed soil face between depths of 5 and 30 cm.

Water was applied to the top of each plot using a ‘sprinkler’ which comprised a hosepipe, to which was attached a closed plastic tube that lay across the top of a plot. Ten holes were drilled in the plastic tube to ensure an even application of water to the plot (Photograph 2.1). A water meter was attached to the hosepipe to determine input rates. The water supply was subject to changes in pressure.

During the cow pat experiments, water was also applied to each plot using rainfall simulators. These each provided very heavy ‘rainfall’ at a fixed rate of 50 mm per hour over an area of 1 m2. The rain fell from a height of 1 m and, therefore, had less erosive power than natural raindrops of the same intensity and size. Effluent and cow pats were obtained from a local dairy farm, and stored overnight at 10° C. A strain of C. jejuni, isolated from dairy land drain sediments, was added to both effluent and pats.

2.3.1 Soil Drainage

The Hamilton clay loam underlying the plots has developed within about 30 cm of younger volcanic tephra that lies upon older, strongly weathered volcanic tephra. Water movement through the older clayey tephra is slower than that through the younger tephric material. As a consequence, the upper part of the soil profile becomes waterlogged during heavy or prolonged rainfall. Water also moves through the soil profile by the process of bypass flow. Bypass flow occurs when water (and entrained contaminants) flow through soil cracks rather than through small soil pores, and as a consequence there is limited beneficial interaction with the soil. McLeod et al. (2003) have analysed the Hamilton clay loam and rated it as having high bypass flow for microbes. This means that surface-applied microbes in effluent are likely to be rapidly transmitted to depth in the soil.

Experimental Procedure

Expt.	Num	E/P	S/R	Purpose	Date	Q_In	Q_Out
1	4	E	S	Objectives A, B, C	21/08/02	10.2 – 10.6	2.6 - 4.6
2	4	E	S	Objective A	16/10/02	4.0 – 6.6	0.4 – 1.3
3	2	P	S, R	Objective C	22/01/03	11.7 – 14.7	1.5 – 3.0

Three sets of experiments (Table 2.1) were conducted to address the objectives of the study. During each set, experiments were conducted on 2-4 plots.

Table 2. 1 Overview of the three sets of experiments. Num indicates the number of individual plots used in each experiment. ‘E’ indicates that effluent was applied, ‘P’ indicates that pats were applied. ‘S’ indicates that the sprinkler was used, ‘R’ indicates that the rainfall simulators were used. Q_In and Q_Out (surface plus subsurface) give the range of input and output flows (L/min) from the plots under each set of experiments.

2.4.1 Experiment 1

Experiment 1 involved the application of effluent to four plots. The inflow rate of water was 10-11 L/minute, sufficient to saturate the soil and generate overland flow. The outflow rate was steady and similar to that seen during heavy rainfall. Once the plots had attained a hydrological steady state by pre-watering, the hosed water was temporarily stopped and 20 L of dairy effluent was applied to each plot using watering cans fitted with a distributor to ensure an even application (Photograph 2.2). Effluent was applied evenly across the plots 0-30 cm above the top end of the plot over a period of c. 2 minutes, the time taken for the sprinklers to apply 20 L of water. This practise minimised the changes in outflow from a plot whilst the hosed input of water was temporarily stopped.

Once effluent application was completed, the water supply was turned back on. The plastic pipe (attached to the hosepipe) was placed above the band of effluent enabling surface runoff to wash down through the effluent. Water was applied for 40 minutes. Surface and subsurface outflow from the lower end of each plot was measured at designated intervals using a graduated plastic container held beneath the collection pipe. One-litre samples of outflow were also collected at designated intervals for microbial analysis. Samples of background outflow (prior to the addition of effluent), and of the effluent, were also taken for microbial analysis. All samples were placed in a chilly bin and transported to the laboratory within 1 hour of collection. Analysis was completed within 6 hours of collection.

2.4.2 Experiment 2

Experiment 2 involved the application of effluent to four plots using the same procedure used in experiment 1, but applying a slow flow rate (4.0 – 6.6 L/min).

2.4.3 Experiment 3

In Experiment 3, cow pats rather than liquid dairy effluent were applied to two of the four plots. As with experiments 1 and 2, the surface of the plots was saturated prior to the application of faecal material. Four pats, each weighing 1.25 kg were placed along the top of each plot, and surface runoff generated from a sprinkler upslope flushed pat material down into the plots.

Additionally, two rainfall simulators were placed over the pats with one simulator encompassing a pair of pats (Photograph 2.3). Both the sprinkler and simulators applied water to the plots for 40 minutes. The cow pat experiments were conducted using a fast flow rate (11.7 – 14.7 L/min) to provide a direct comparison with experiment 1 whereby similar flow rates were used to flush liquid dairy effluent through the plots.

2.4.4 Microbial Analysis

(a) E. coli Analysis

E. coli were analysed using a commercial MPN technique involving Colilert and QuantitrayTM (IDEXX, USA). Trays were incubated at 35° C for 24h and E. coli identified under UV light (366nm). The concentration of E. coli was determined from MPN probability tables supplied by the manufacturer.

(b) Campylobacter Analysis

Samples were analysed for C. jejuni using a 5-tube (i.e. 5 tubes per dilution) two-stage MPN technique (MIRINZ Manual 2001). A ten-fold dilution series was inoculated into Campylobacter medium. The concentration of confirmed Campylobacter in the original sample was determined by reference to five-tube MPN probability tables.

Photo 2. 1 Application of water using the sprinkler.

Photo 2. 2 Application of effluent.

Photo 2. 3 Experimental design for the cow pat experiments.

2.5 Results

2.5.1 Objective A: To Determine the Impact of Flow Rate Upon Microbial Attenuation

Relevant Experiments: A comparison of attenuation between experiments 1 and 2.

Table 2.2 shows that flow rate has a clear impact upon the recovery (the percentage of the applied microbe recovered in the outflow, i.e. the inverse of attenuation) of microbes over a 40-minute period. At high flow recovery varies between 16% and 62% for E. coli, and between 13% and 89% for Campylobacter. In contrast, recovery at slow flow ranges between <1% and 5% for both microbes.

Also of note is the impact of flow rate upon surface runoff microbial concentrations over time (as illustrated by E. coli data from one plot, Figure 2.1). Under fast flow there was an initial first flush of high Campylobacter and E. coli concentrations, with the peak occurring within 10 minutes of effluent application. Following the peak, concentrations typically decreased by a 1-2 orders of magnitude for the remainder of the experiment. This response was also observed in the first plot study (Collins et al. 2002). In contrast, under the slow flow rate, microbe concentrations did not peak until at least 30 minutes after effluent application and, thereafter, did not decline for the remainder of the experiment.

Table 2.2: Microbial recovery (the percentage of the applied microbe recovered in the outflow) under fast and slow flow rates. Q_In and Q_Out are the mean input and output flow rates respectively (L/min) over the duration of each plot experiment. Outflow includes subsurface flow. EC_Rec and C_Rec are the recovery percentages for E. coli and Campylobacter respectively.

			FAST	FLOW			SLOW	FLOW
Plot	Q_In	Q_Out	EC_Rec	C_Rec	Q_In	Q_Out	EC_Rec	C_Rec
1	10.2	3.4	16	89	6.6	0.6	< 1	2
2	10.6	4.8	25	74	5.2	0.5	< 1	1
3	10.5	4.5	62	46	4.6	1.3	4	5
4	10.3	3.7	42	13	4.0	0.7	< 1	< 1

Table 2. 2 Microbial recovery (the percentage of the applied microbe recovered in the outflow) under fast and slow flow rates. Q_In and Q_Out are the mean input and output flow rates respectively (L/min) over the duration of each plot experiment. Outflow includes subsurface flow. EC_Rec and C_Rec are the recovery percentages for E. coli and Campylobacter respectively.

2.5.2 Objective B: To Determine the Relative Importance of Subsurface Flow as a Transmission Pathway

Relevant Experiments: Experiment 1

Table 2.3 indicates that the contribution of subsurface flow to the total volume of outflow was variable (4% to 24%) between plots under the fast flow rate. Some subsurface runoff was observed to flow beneath the collecting pipe on plots 1 and 3. Consequently the volume of subsurface runoff and its microbial load are underestimated for these two plots.

Whilst fast flow microbial concentrations were similar in surface and subsurface flow, the greater volume of surface flow generally yielded a much larger microbial load: subsurface flow contributed between < 1% and 13% of the total number (surface and subsurface combined) of E. coli recovered in the outflow. For Campylobacter this figure ranged between < 1% and 24% (Table 2.3).

Plot	%_SubQ	%_SubECQ	%_SubCCQ
1	24	13	24
2	4	< 1	< 1
3	7	2	2
4	5	3	10

Table 2. 3 The percentage contribution of subsurface flow (%_SubQ) and its microbial load (E. coli = %_SubEC; Campylobacter = %_SubC) to total (surface plus subsurface) outflow, and total outflow microbial load, respectively.

2.5.3 Objective C: To Compare Microbe Attenuation Under Differing Types of Applied Faecal Material (Liquid Dairy Effluent Versus Relatively Dry Pats)

Relevant Experiments: A comparison of Experiments 1 and 3

A comparison of Tables 2.2 and 2.4 indicates that recovery of E. coli under fast flow rates is markedly lower from cow pats (2-4%) than effluent (16-62%). This is also true of Campylobacter whereby recovery from cow pats is no greater than 5%, but from effluent ranges between 13% and 89%.

Outflow microbial concentrations from pats (Figure 2.2) were typically at least an order of magnitude lower than those from effluent (despite a similar number of microbes applied to the plots for each type of faecal material) and, in contrast to effluent concentrations, continued to increase slowly over the duration of an experiment. This response suggests that as a pat becomes steadily more saturated over time so the rate of break-up increases.

Plot	QS_In	QR_In	Q_Total_In	Q_Out	EC_Rec	C_Rec
1	12.8	1.9	14.7	3.5	2	< 1
2	10.0	1.7	11.7	3	4	5

Table 2. 4 Flow rates and microbial recovery for the cow pat experiments. QS_In and QR_In are the input flow rates (L/min) from the sprinkler and rainfall simulator respectively. Q_Total_In provides the sum of these, and Q_out is the outflow rate (L/min). EC_Rec and C_Rec are the recovery percentages for E. coli and Campylobacter respectively.

2.6 Discussion

The relationship between flow rate and microbial entrapment raises important implications for buffer strip design, particularly if appreciable faecal contamination is only delivered by surface runoff during large events. It is important to note, however, that these experiments were conducted once the surface of the plots was saturated. Drier antecedent conditions would probably have given rise to differing entrapment rates.

Vertical movement of soil water and microbes may be slow via the soil matrix or rapid within bypass flow caused, for example, by worm-holes or soil cracks. Bypass flow within the Hamilton clay loam (underlying the plots) enabled a substantial proportion of microbes to travel rapidly to depth, although this varied considerably between plots. Microbes transported by this process are not therefore attenuated by surface vegetation or filtering by the soil matrix. Under more poorly drained soils (typical of natural riparian zones) this bypass flow may be more marked, and could limit the effectiveness of buffer strips, especially for microbes. Consequently, the widely held notion that RBS will be most effective upon poorly drained soil may only hold true where bypass flow is not excessive.

Appreciable release of microbes from cowpats appears to require at least 10 minutes of substantial rainfall and surface runoff. This contrasts to the readily mobile microbes within liquid effluent, giving rise to the contrasting outflow concentrations between pats and effluent. Under very prolonged rainfall this contrast is likely to diminish as pats progressively break up.

Contact for Enquiries

MAF Information Services
Pastoral House
25 The Terrace
PO Box 2526
Wellington, NEW ZEALAND

Fax: +64 4 894 0721
Contact this person