Management Strategies to Mitigate Faecal Contamination Inferred from Analysis of Data from the Waikato Region - Relationships between environmental factors and faecal

4.1 Intra-Site Variability

4.1.1 Flow and Turbidity

At 12 (of the 73) sites, instantaneous flow was recorded at the time samples were collected for E. coli analysis, enabling the relationship between these two variables to be examined. The resultant scatter plots (Figure 8) generally exhibit a positive linear relationship between log-transformed E. coli concentration and flow, although 2 sites, (4 and 65) exhibit an inverse relationship. The data at most sites are strongly skewed as most samples were collected at low to medium flow with just 2 or 3 samples collected at high flow. This skewness is also evident within the turbidity data (Figures 9 and 10). The increase in E.coli concentration with increasing flow generally observed at these sites is in accordance with findings from a number of other studies (for example, Nagels et al. 2001). This behaviour can be attributed to the wash-in of faecal material from grazed paddocks, and the entrainment of bacteria, previously deposited, from within bed sediments. As flow rises, turbidity increases and sunlight inactivation of bacteria suspended in the water column is decreased. In contrast, however, faecal bacteria discharged from point sources will tend to decrease in concentration with increasing flow, due to dilution. Watershed 4 (upstream of Okauia on the Waihou river) is characterised by significant point sources, and this may be the reason for the inverse relationship observed. Watershed 65 does not have any significant discharge of point sources. It should be noted, however, that sampling at both these sites encompassed a relatively limited range of flow, and it may be that over a greater flow range, a positive correlation between faecal bacteria and flow may be apparent. Intermittent cattle access to streams may contribute, at some sites, to the relatively high variance in faecal contamination observed at low flows (Figure 8). Cattle within a stream have been shown to increase turbidity and concentrations of faecal contamination by 2-3 orders of magnitude (Davies-Colley et al. 2002). Sampling undertaken downstream of an access point clearly, therefore, has the potential to record high levels of faecal contamination even at relatively low flow.

The relationships illustrated in Figure 8 indicate that some of the variance within a site can be attributed to variations in flow at the time of sampling. As a consequence, in order to aid interpretation and prediction of faecal contamination, attempts were made to minimise the flow-induced variance within the data. Two approaches were used. The first excluded from the analysis those E. coli samples that had been collected at greater than mean flow. This required that a flow record be available for each site. At 12 sites flow is recorded directly. For the remaining 61 sites a surrogate flow record was derived from a nearby stream or river. These surrogate records were obtained from FoSRT, Mighty River Power, Emco, Contact Energy, and Carter Holt Harvey. This approach typically led to 2 or 3 E.coli values being excluded from each site, resulting in a new set of (lower) median values for the region. Although this modified dataset was characterised by a lower variance, use of it resulted in a poorer relationship between median E. coli and all of the environmental variables. The use of surrogate flow data that was often derived from a river tens of kilometres from the sampling site was probably a key reason for this. Such a distance would have given rise to differences in the state of flow (at least on occasions) between the surrogate and actual sites.

The second approach used the turbidity dataset to discriminate those values statistically defined as an outlier or extreme outlier. It was assumed that high turbidity samples were collected at high flow and the corresponding E.coli concentration was, therefore, excluded from the faecal dataset. Evidence that this assumption was valid is provided by the flow-turbidity and E. coli-turbidity relationships derived for those 12 sites where instantaneous flow was recorded (Figures 9 and 10). Both sets of relationships are characterised by considerable scatter. However, they do illustrate a correlation between high flow, and high turbidity and E. coli concentrations. This suggests that turbidity may be a useful surrogate variable for E. coli, and that processes of entrainment of fine sediment generally may be similar to those mobilising bacteria. As with the mean flow criteria of the first approach, however, this second method led to no improvement in prediction of median E. coli values from the range of environmental factors. The inability of either method of accounting for flow-induced variance to improve explanation of faecal contamination, probably relates, in part, to the generally small reduction in median E.coli values that resulted.

4.1.2 Seasonality

To assess the impact of season upon faecal contamination, E. coli concentrations at 12 sites were divided into winter (collected between April and September) and summer (collected between October and March) samples. No seasonal differences were found at these sites possibly because these were masked by the impact of flow variations. At five sites (15, 60, 61, 65, and 67) daily solar radiation data was available from one of 3 automatic weather stations located within 20km. At four of these sites a weak inverse relationship (R2 ranged between 0.006 and 0.14) was found between E. coli concentration and mean daily solar radiation averaged over the day, and preceding day, of sampling. These relationships are consistent with die-off due to solar radiation.

4.2 Inter-Site Variability

4.2.1 Livestock

It has been well established that grazing animals are an important causal factor in the faecal contamination of streams (section 1). This is reflected within the Waikato dataset through a comparison of median E. coli concentrations in streams draining pastoral catchments (>90% pasture) with those draining forested (>90% forest) ones (Figure 7). Furthermore, a correlation (R = 0.48, Rs = 0.59) exists between median E. coli concentration and the percentage of pastoral land within a catchment, across the region. This relationship strengthens (R = 0.54, Rs = 0.58) when stock density (stock units per km2) rather than pastoral land is used as the independent variable. A further slight strengthening of this relationship (R = 0.58, Rs = 0.60) occurs when just cattle stock density is used as the predictor (Figure 11a), probably because cattle are attracted to water, depositing faecal material directly to streams. It is likely that the presence of livestock would be a stronger prediction of faecal contamination if cattle access to streams, or conversely, their exclusion from them, were known for each watershed. The acquisition of such data was not feasible at a regional scale in the present study.

Figure 11a-k. Relationships between environmental factors and median E. coli

4.2.2 Forest and Non-pastoral Vegetation

Catchments characterised by forest or non-pastoral vegetation exhibit relatively low median E. coli concentrations (Figure 7), and a weak inverse relationship exists (R = -0.39) between the percentage of land under indigenous vegetation and median E. coli, across the region. This linear relationship strengthens (R = -0.52) when the percentage of all non-pastoral vegetation is used as the independent variable (Figure 11b). The Spearman Rank correlation coefficient (Rs) is –0.63, indicating a relatively strong degree of monotonicity between non-pastoral vegetation and median E. coli. Figure 11b suggests this relationship may be non-linear. The exclusion of livestock from such land is clearly the principle reason for this correlation. Non-zero E. coli concentrations are observed, however, even in fully forested catchments (e.g. site 28), and this background level of contamination can probably be attributed to birds and mammals such as pigs, deer, possums and rats. This contamination from wild animals may be accentuated by the scarcity of ground vegetation under the shade of riparian trees, which might otherwise attenuate faecal material entrained in overland flow (R. Davies-Colley pers. comm.). Furthermore, shading by riparian vegetation will reduce, relative to pastoral catchments, the rate of sunlight die-off on land and in the water.

4.2.3 Soil Drainage

A relatively strong correlation (R = 0.69, Rs = 0.61) exists (Figure 11c) between median E. coli and the percentage of a catchment characterised by poorly drained soil (soil drainage classes 1 and 2). This is consistent with the hydrological characteristics of such soils whereby a relatively large volume of surface runoff is generated, rapidly transporting entrained faecal material to surface waters. Where such soils underlie grazed pastoral land, stock trampling may further impede the infiltration of water (Nguyen et al. 1998). The rapid hydrological response of such soils may also act to increase peak flows in receiving waters. Higher flow velocities associated with such a response would in turn lead to a greater relative entrainment of faecal material within such streams.

The analysis of soil drainage properties in this study did not account for the presence of subsurface drains, often installed under otherwise poorly drained dairying land. Although these drains reduce ponding and overland flow of surface water, they may also act to provide a flowpath by which subsurface faecal contamination can be rapidly transported to the channel network. Such an inference is supported by the recovery of E. coli concentrations in excess of 10,000 cfu/100mL from subsurface drains in Northland (L. Nguyen pers. commun.). Consequently, the correlation between median E. coli and poorly drained soil may, in part, be explained by the presence of subsurface drains.

An inverse relationship (R = -0.42, Figure 11d) is implied between median E. coli and the percentage of well-drained soils (soil drainage class 5) within a catchment, whilst the Spearman rank correlation coefficient (Rs = -0.57) indicates a relatively strong degree of monotonocity between the two variables. Reference to the matrix of correlation between independent variables (Table 3) suggests that this is partly an indirect relationship whereby the presence of well drained soil simply reflects the absence of poorly drained soil. It may, however, also indicate that well drained soils have a direct impact through minimising the generation of overland flow and enabling rainwater to infiltrate down through the soil horizons. This would lead to reduced contamination as the soil matrix is generally effective at filtering soil water through the attachment of faecal material to soil particles. It is important to note, however, that if a soil is exceptionally permeable due to macropores, then the filtration process will be minimal and subsurface flow may be of low bacterial quality.

4.2.4 Wetlands and Inland Water

A weak inverse relationship exists between the percentage of land within a catchment classified as wetland, and median E. coli. This relationship strengthens (R = -0.32, Rs = -0.47) if those catchments without any wetlands are excluded from the analysis (Figure 11e). With the exception of one site, all those catchments with wetlands are also characterised by > 40% pasture. These findings suggest that the presence of a wetland may act to trap faecal material, reducing levels of faecal contamination. It is likely that trapping efficiency will be dependent upon the rate of water moving through a wetland, decreasing with increasing flow.

The percentage of inland water (streams, rivers, ponds and lakes) within a catchment essentially exhibits no correlation with median E. coli. Although length of streambank may correlate positively with faecal contamination through influencing the degree to which cattle have access to waterways, such a relationship may be offset by the presence of lakes and ponds that may act to trap faecal material. However, no such conclusions can be drawn from this dataset.

4.2.5 Rainfall

An inverse relationship (R = -0.33, Rs = -0.28) exists between mean annual rainfall and median E. coli. This is probably because those areas within the region that receive high rainfall are also characterised by relatively low pastoral land use, as reflected by the negative correlation coefficient between these two variables (Table 3).

4.2.6 Slope and Catchment Area

A weak inverse relationship is apparent between dominant slope angle and median E. coli. This strengthens slightly when the percentage of steep slopes (Figure 11f) within a catchment is used as the independent variable (R = -0.49, Rs = -0.32). Steep slopes might be expected to exhibit a positive relationship with faecal contamination through increasing overland flow to streams. However, little pastoral land within the region lies on steep slopes, as indicated by the negative correlation coefficient between these two variables (Table 3).

It may be speculated that larger catchments, and therefore, a greater stream and river length may increase the time available for die-off and deposition of faecal material. In addition, the larger the catchment, the more likely it is to encompass flat land characterised by lower flow velocities that favour deposition. Catchment area, however, appears to have no influence upon median E. coli levels. This lack of a correlation may relate to the spatial distribution of contaminant sources within catchments. A point source discharge of faecal material located close to the catchment outlet, for example, will mask the impact of the processes outlined above.

4.2.7 Urban Area

The presence of urban areas (Figure 11g) appears to correlate weakly (R = 0.39, Rs = 0.47) with median E. coli concentrations. Urban runoff is known to have relatively high levels of faecal contamination, attributed, for example, to bird droppings and dog faeces. Furthermore, point source discharges from treated waste water (and some industries) within urban areas are likely to provide an appreciable input of faecal contamination direct to a stream or river.

4.2.8 Point Source Discharges

No strong correlation is apparent between median E. coli and the 3 categories of point sources across the region. The volume of non-dairy point sources (per day per km2) apparently provides the strongest linear relationship (R = 0.39, Rs = 0.34), Figure 11h, primarily since this factor explains much of the variance caused by site 64, which has the joint highest median E.coli (1300 cfu/100mL) across the region. The stream network in watershed 64 directly receives 63 m3/day/km2 of non-dairy effluent, a far higher volume than any other watershed. Excluding site 64 from the bivariate relationship results in effectively no correlation between median E. coli and non-dairy point sources.

The direct discharge of dairy effluent to surface water is low, with a mean discharge of 1 m3/day/km2 across the region. This reflects a policy in recent years to reduce the direct discharge of effluent to surface water, and is the likely reason for the apparent lack of a strong correlation (R = 0.32, Rs = 0.51) with median E. coli concentrations (Figure 11i). The Spearman Rank correlation is stronger probably because it is better able to account for the skewed nature of the dairy point source data.

The lack of a strong correlation (R = 0.22, Rs = 0.36) between point source discharges from utilities and median E. coli may reflect the inclusion of discharges within this category that are not faecally-contaminated. In addition, recent improvements in the treatment of sewage and industrial waste-water (Waikato State of the Environment Report 1998, Vant 2001) may limit any relationship.

4.2.9 Effluent Discharges to Land

No strong correlation is apparent between the 3 sources of effluent discharge to land, and median E. coli across the region. Non-dairy effluent (Figure 11j) provides the strongest relationship (R = 0.46, Rs = 0.51). Expressing this data in terms of the number of occurrences, rather than the volume of discharge, may have masked a stronger correlation with the faecal contamination of streams. A tentative conclusion may be drawn, however, that effluent discharge to land does not markedly impact upon bacterial water quality.

4.2.10 Ponds

Since the presence of a pond generally reflects consent to discharge dairy effluent direct to a stream, a strong correlation exists between the two (Table 3). As with dairy point source discharges, the presence of dairy ponds shows no strong correlation (R = 0.33, Rs = 0.58) with median E. coli concentration. If climatic conditions are appropriate and effluent is contained within a pond for sufficient time, then significant microbial die-off may occur prior to discharge. This may act to weaken any correlation between ponds and discharge of dairy waste with faecal contamination of streams.

4.2.11 Turbidity

Median turbidity is a relatively strong predictor (R = 0.65, Rs = 0.71) of median E. coli, across the region (Figure 11k). A degree of correlation is to be expected given that both stream bed-sediments and the microbes settled within them are subject to entrainment as flow velocity increases. In addition, the processes by which overland flow detaches and transports soil particles on the hillslope are also applicable to faecal material. Median turbidity has a relatively strong correlation (R = 0.79) with the percentage of poorly drained soil within a catchment, and some correlation (R = 0.44) is apparent with the density of cattle (Table 3).