Influence of seasonal and inter-annual hydro-meteorological variability on surface water fecal coliform concentration under varying land-use composition

w a t e r r e s e a r c h 4 8 ( 2 0 1 4 ) 1 7 0 e1 7 8

Available online at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/watres

Influence of seasonal and inter-annual hydrometeorological variability on surface water fecal coliform concentration under varying land-use composition Jacques St Laurent*, Asit Mazumder Water and Aquatic Sciences Research Program, Department of Biology, University of Victoria, PO Box 3020 Station CSC, Victoria, British Columbia, Canada V8W 3N5

article info

abstract

Article history:

Quantifying the influence of hydro-meteorological variability on surface source water fecal

Received 6 June 2013

contamination is critical to the maintenance of safe drinking water. Historically, this has

Received in revised form

not been possible due to the scarcity of data on fecal indicator bacteria (FIB). We examined

24 August 2013

the relationship between hydro-meteorological variability and the most commonly

Accepted 13 September 2013

measured FIB, fecal coliform (FC), concentration for 43 surface water sites within the

Available online 25 September 2013

hydro-climatologically complex region of British Columbia. The strength of relationship was highly variable among sites, but tended to be stronger in catchments with nival

Keywords:

(snowmelt-dominated) hydro-meteorological regimes and greater land-use impacts. We

Fecal coliform

observed positive relationships between inter-annual FC concentration and hydro-

Drinking source water

meteorological variability for around 50% of the 19 sites examined. These sites are likely

Climate variability

to experience increased fecal contamination due to the projected intensification of the

Land-use

hydrological cycle. Seasonal FC concentration variability appeared to be driven by snow-

Waterborne disease

melt and rainfall-induced runoff for around 30% of the 43 sites examined. Earlier snowmelt in nival catchments may advance the timing of peak contamination, and the projected decrease in annual snow-to-precipitation ratio is likely to increase fecal contamination levels during summer, fall, and winter among these sites. Safeguarding drinking water quality in the face of such impacts will require increased monitoring of FIB and waterborne pathogens, especially during periods of high hydro-meteorological variability. This data can then be used to develop predictive models, inform source water protection measures, and improve drinking water treatment. ª 2013 Elsevier Ltd. All rights reserved.

1.

Introduction

Fecal contamination of drinking source water results in waterborne disease outbreaks and millions of cases of gastroenteritis throughout the world (Bartram and Cairncross,

2010). Examining the processes that drive variability in source water contamination is therefore critical to improving the safety of drinking water and maintaining public health. The identification and quantification of drivers of contamination allows us to target source water protection efforts and

* Corresponding author. Tel.: þ1 250 472 4789; fax: þ1 250 472 4766. E-mail addresses: [email protected] (J. St Laurent), [email protected] (A. Mazumder). 0043-1354/$ e see front matter ª 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.watres.2013.09.031

w a t e r r e s e a r c h 4 8 ( 2 0 1 4 ) 1 7 0 e1 7 8

increases our capacity to anticipate variability in source water contamination, which can be used to improve the specificity of drinking water treatment processes (Charron et al., 2004; Delpla et al., 2009). Climate factors are primary among drivers of source water contamination due to their influence on the transport of contaminants by precipitation-induced runoff. Runoff generated by rainfall and snowmelt (encapsulated by the term hydro-meteorological) is associated with increased levels and variability of fecal contamination in downstream surface water (Kistemann et al., 2002; Cha et al., 2010). However, the relationship between hydro-meteorological variability and fecal contamination levels tends to vary significantly within and among watersheds (Wilkes et al., 2009). Investigating the strength, influence, and variability of this relationship among surface source water sites is increasingly important given the significant changes to temperature and precipitation patterns observed at local, regional, and global scales (Bates et al., 2008). British Columbia (BC) is projected to experience higher annual mean temperatures, accompanied by increased total precipitation and changes to the fraction of annual precipitation stored as snow (Schnorbus et al., 2011). The corresponding increase in runoff and change to snowmelt volume will likely alter fecal contamination levels and variability in surface source water. Variability in FC concentration tends to be positively related to precipitation (Cha et al., 2010), therefore increases in total annual precipitation will likely elevate contamination levels, although this relationship has yet to be examined over an inter-annual period. Precipitation is increasing at a rate of around 7% per degree Celsius increase in temperature (Wentz et al., 2007), which will result in a 6e12% increase in mean annual precipitation across BC by 2050 (Schnorbus et al., 2011). The corresponding increase in runoff is likely to elevate the transport potential of fecal contaminants and resultant levels of fecal contamination in downstream surface source water. Seasonal variability in FC concentration can be influenced by variability in surface runoff (Dorner et al., 2007), therefore current periods of peak fecal contamination may be altered by changes to existing hydro-meteorological regimes. Hydrometeorological regime is determined by the timing, volume, and extent of snowpack accumulation and associated snowmelt, which vary in relation to cold-season temperatures (Marsh and Woo, 1981). Projected increases in cold-season temperatures are expected to transition nival-regimes towards hybridregimes, and hybrid-regimes towards pluvial-regimes (Dery et al., 2009); (descriptions of hydro-meteorological regimes are given in the supplementary material (SM)). The impact of precipitation-induced runoff on surface water fecal contamination is influenced by land-use and management factors that alter watershed characteristics, such as ground permeability and the presence of riparian vegetation (Perdek et al., 2003; Tate et al., 2004). These factors alter the capacity of surface runoff to transport fecal contaminants into surface water and therefore likely contribute to the high variability observed among relationships between hydro-meteorological conditions and fecal contamination variability (Kay et al., 2008). The distribution of fecal contaminant sources within a catchment may determine the relative influence of climate

171

variability on source water fecal contamination levels among catchments. Point-source contamination tends to be directly discharged into receiving water, generating contamination variability that is often unrelated to runoff variability. Conversely, diffuse contamination requires transport by means of runoff into surface water (Kloot, 2006). Therefore, variability in surface water fecal contamination may be more strongly associated with runoff in catchments where contamination is associated with diffuse fecal sources. Although previous studies have provided strong evidence for an association of higher fecal contamination with greater runoff, there is little evidence available to determine the influence of hydro-meteorological variability on seasonal and inter-annual fecal contamination levels (Wilkes et al., 2009; Dorner et al., 2007; Sigua et al., 2010). A better understanding of these interactions is critical for assessing the potential for changes in climate to influence surface source water contamination levels and the risk this contamination will present to public health (Patz et al., 2008). In this study, we examined climate forcing of surface water fecal contamination across a range of hydro-meteorological regimes and land-use scenarios. We measured the extent to which seasonal fecal contamination variability was determined by snowmelt and rainfall variability, and examined how the strength of this relationship varied in relation to hydro-meteorological regime. We also quantified the relationship between inter-annual hydro-meteorological variability and FC concentration to see how fecal contamination levels responded to long-term changes in snowmelt and rainfall. Site characteristics associated with climate forcing of seasonal and inter-annual fecal contamination were identified in order to categorize those catchments most vulnerable to changes in climate altering surface water fecal contamination levels and variability.

2.

Methods

2.1.

Study region

The province of BC in the west of Canada was selected as a study region due to its complex hydro-climatology, largely resulting from its exposure to the Pacific Ocean in the west, successive mountain ranges throughout the interior, and the continental expanse to its east. Coastal BC has a temperate, mild and wet, oceanic climate due to the Kuroshio Current that transports warm tropical water into the northeast Pacific Ocean. Precipitation decreases towards the interior, due to the rain-shadow cast on the leeward side of successive mountain ranges, and temperature ranges increase in the absence of the ocean to moderate seasonal fluctuations in insolation (Shabbar et al., 1997). These conditions generate a range of hydro-meteorological regimes, which tend to be nival and hybrid in the interior and at higher elevations (>500 m) and pluvial near the coast and at lower elevations (<650 m). Sample sites were generally located in the south of BC where population density is greatest, between 48.50 and 56.10 latitude N and 114.90 and 124.40 longitude E, and had an elevation range of 170 me1700 m. A map of the study region and sample site locations, and sample site coordinates and

172

w a t e r r e s e a r c h 4 8 ( 2 0 1 4 ) 1 7 0 e1 7 8

characteristics, are presented in the SM (Fig. S1 and Table S1). Mean river discharge and rainfall varied significantly among sites, ranging from 0.25 m3 sec1e2580.90 m3 sec1 and 0.42 mme29.82 mm, respectively. Seasonal and annual summary data for river discharge and rainfall values for each site are presented in the SM (Tables S2 and S3, respectively).

Mazumder (2012)). Variability in mean FC concentration among sites was associated with land-use composition, by means of diffuse contaminant transport into surface water, among a sub-set of sites called “group 1” (n ¼ 20) whereas the remaining sites that had no such association are a sub-set called “group 2” (n ¼ 23). Site group is labelled in Table S1.

2.2.

2.3.

Data collection

Fecal contamination was measured using the fecal indicator bacteria (FIB) group called fecal coliforms (FC). FC concentration, measured in colony-forming units (CFU) 100 ml1, is generally considered a robust indicator of fecal contamination and the likelihood of waterborne pathogen presence (Jokinen et al., 2012). Measurements of FC concentration (n ¼ 7798) for 43 surface water sites within BC were obtained from the Environment Canada Water Quality Monitoring Program (Environment Canada, 2003). All FC data was produced using standardized sampling and enumeration methods provided by the BC Ministry of Environment (BC MoE, 2009). We used an inclusion criterion of a minimum of two FC measurements each month for the duration of each data set, which varied from 4 to 7 years within the period from January 2000 to December 2006 (summary data for site FC concentrations are given in Tables S1, S2, and S3). Mean temperature and river discharge on the day of sampling and three-day cumulative rainfall prior to sampling were obtained from the Environment Canada meteorological and hydrometric stations with greatest proximity to each surface water sampling location. Sample site catchment characteristics and land-use composition were obtained from the government of BC and quantified using ArcGIS 9.3 software (full details of land-use analyses are provided in St Laurent and

Statistics

Multivariate linear regression (MLR) was used to quantify the extent of both seasonal and inter-annual FC concentration variability explained by river discharge and rainfall variability. The maximal MLR model (general mathematical representation presented in the SM) was fitted and then simplified by removing non-significant variables until the minimum adequate model was obtained. Seasonal variability was examined using mean values for each month in order to increase the accuracy of the annual trend for each parameter and obtain a standardized and appropriate sample number on which to perform MLR. Interannual variability was examined using a subset of sites (listed in Table S3), selected on the basis of having seven years of consecutive data. Trends for each variable were extracted by smoothing mean monthly time series data using weighted (span of 0.75) quadratic least squares regression. Trend values were differenced before MLR was performed. A regression tree was used to identify climate and/or landuse variable threshold values associated with climate forcing of fecal contamination (mathematical representation of deviance, minimized to identify variable threshold values, is presented in the SM). One-way ANOVA was used to compare r-square values among sites of different land-use group and hydro-meteorological regime.

Table 1 e Summary statistics of MLR models used to examine seasonal FC concentration variability as function of river discharge and rainfall variability. Site name

Site no.

H-m regimea

River discharge b

Cheakamus River below sewage plant Cheakamus River on lake road Columbia River at Birchbank Columbia River at Nicholson Elk river at Hway 93 Elk River at Sparwood Fraser River at Hansard Kettle River at Carson Kootenay River at Creston Kootenay River at Fenwick Nicola River Peace River Salmon river at Salmon Arm Salmon River at Silver Creek Similkameen River at US boarder a

3 4 7 8 11 12 14 18 21 22 28 31 36 37 40

Hybrid Hybrid Hybrid Nival Nival Nival Nival Nival Nival Nival Nival Nival Nival Nival Nival

Rainfall c

t-value

p-value

e e 3.320 e 7.940 2.558 e e 2.844 4.593 2.594 e e 2.728 2.446

e e 0.008 e <0.001 0.031 e e 0.017 0.001 0.027 e e 0.023 0.035

b

t-value 5.535 4.022 e 24.831 3.072 8.245 2.980 3.491 e e e 4.213 2.156 3.158 e

Full model c

p-value <0.001 0.002 e <0.001 0.013 <0.001 0.015 0.006 e e e 0.002 0.057 0.012 e

Adj r2d

p-valuec

0.729 0.580 0.477 0.982 0.918 0.959 0.441 0.504 0.392 0.646 0.343 0.604 0.249 0.554 0.312

<0.001 0.002 0.008 <0.001 <0.001 <0.001 0.015 0.006 0.017 <0.001 0.027 0.002 0.056 0.011 0.034

Hydro-meteorological regime. The test statistic, which is calculated by dividing the coefficient by its standard error. A larger t-value, therefore, indicates a greater probability of the coefficient being different from 0. c Derived by comparing the t-value with values in the Student’s t distribution. This gives the probability of obtaining a t-value at least as large as the one calculated under the assumption that there is no relationship between FC concentration and river discharge and/or rainfall. d The fraction of variation in FC concentration accounted for by river discharge and/or rainfall variability. b

w a t e r r e s e a r c h 4 8 ( 2 0 1 4 ) 1 7 0 e1 7 8

All data handling, graphics, and statistical analysis were performed with the statistical software R (R Development Core Team, 2009). An Alpha (a) value of 0.05 was used as a significance threshold for all model parameters and statistical tests.

3.

Results

3.1.

Seasonal FC concentration variability

Surface water FC concentration tended to vary seasonally, with 75% of sites having significantly different mean FC concentrations among seasons (Table S2). However, evidence for hydro-meteorological variability driving seasonal FC concentration variability was only observed among half of these sites (Table 1). The proportion of sites that showed evidence of climate forcing varied among hydro-meteorological regimes, as did the season in which peak FC concentration occurred. The majority of the 27 sites with nival hydrometeorological regimes experienced peak FC concentration during summer (59%; 16 sites), a minority during spring (15%; four sites), and a single anomalous site in winter (4%). Seasonal FC concentration variability was significantly related to hydro-meteorological variability for around half of these sites (56%, or nine, and 50%, or two, of the sites with peak fecal contamination in summer and spring, respectively). Both

rainfall and snowmelt appeared to drive peak FC concentration in summer (Fig. 1b), whereas peak FC concentration in spring was driven principally by snowmelt (Fig. 1c). FC concentration did not vary significantly among seasons for 22%, or six, of the sites (p-value for ANOVA among seasons are presented in Table S2). Mean FC concentration was not significantly different among seasons for 71%, or five, of the sites with hybrid hydrometeorological regimes. However, mean FC concentration was significantly greater during summer in the Columbia River (site 7) and during winter in the Cheakamus River (site 3). Peak FC concentration was associated with snowmelt in summer and with rainfall in winter (Fig. 1a). Sites with pluvial hydro-meteorological regimes presented no evidence of seasonal FC concentration variability being driven by hydro-meteorological variability. Highest mean FC concentration occurred with greatest frequency during summer (44%, or four, of the sites), followed by fall (22%, or two, sites), and winter (11%, or one, site). FC concentration did not vary significantly among seasons for 22%, or two, of the sites. The proportions of FC variability explained among sites with nival and hybrid hydro-meteorological regimes were very similar (r2 ¼ 0.60  0.07 and 0.58  0.07 for hybrid and nival sites, respectively). However, the range of values was larger and the confidence of the mean r2 value greater among nival sites due to a larger n (r2 range for nival sites ¼ 0.25 to 0.98 (n ¼ 12) and r2 range for hybrid sites ¼ 0.48 to 0.73 (n ¼ 3)); (Table 1). Evidence for climate forcing of seasonal FC concentration variability was associated with relatively lower mean rainfall (<2.55 mm) and relatively higher temperature variance (standard deviation (SD) of >8.05  C). River discharge and catchment area variability among sites did not appear to influence climate forcing. There was no evidence to suggest that climate forcing of seasonal FC concentration variability was greater among group 1 sites, in which fecal contamination levels varied in relation to land-use composition, than group 2 sites, in which fecal contamination levels did not vary in relation to land-use composition.

3.2.

Fig. 1 e Seasonal FC concentration variability in relation to hydro-meteorological variability, which drove high fecal contamination during winter in the Cheakamus River (site 3, panel a), during summer in the Salmon River (site 37, panel b), and during spring in the Nicola River (site 28, panel c).

173

Inter-annual FC concentration variability

Inter-annual surface water FC concentration variability was related to hydro-meteorological variability for 47% (9 of 19) of the sites examined (Fig. 2). The proportion of inter-annual FC concentration variability explained by rainfall and snowmelt variability was very similar (r2 ¼ 0.33  0.12 and 0.32  0.14, respectively), whereas a greater proportion of FC concentration variability was explained when both rainfall and snowmelt were related to FC concentration variability (r2 ¼ 0.50  0.06); (Table 2). Evidence for climate forcing of inter-annual FC concentration variability was associated with relatively higher mean rainfall (<2.95 mm) and relatively higher temperature variance (SD of >8.81  C). River discharge and catchment area variability among sites did not appear to influence climate forcing. Again, there was no evidence to suggest that climate forcing was greater among group 1 than group 2 sites. The coefficient values obtained from the MLR models (MLR summary data in the SM) showed river discharge to have a

174

w a t e r r e s e a r c h 4 8 ( 2 0 1 4 ) 1 7 0 e1 7 8

Fig. 2 e Inter-annual variability of FC concentration for sites where it was significantly related to river discharge (RD) and rainfall (R); (note that y-axis scales differ among graphs).

greater influence on inter-annual FC concentration variability than rainfall among sites for which both river discharge and rainfall were related to FC concentration variability. The relative impact of river discharge and rainfall on FC concentration variability (from highest to lowest: sites 31, 40, and 39, respectively) was consistent among sites (Fig. 3).

4.

Discussion

We quantified climate forcing of seasonal and inter-annual surface water fecal contamination variability and examined how and why this varies among sites. Spring snowmelt and subsequent increases in rainfall volume in nival catchments appeared to drive peak fecal contamination levels during summer, especially among those with greater land-use impacts. Conversely, rainfall did not appear to be a significant driver of seasonal fecal contamination variability in pluvial catchments. Peak fecal contamination during summer in these catchments likely resulted from greater production and release of fecal contaminants. Our observations are broadly consistent with those of previous studies (McDonald et al.,

2008; Dorner et al., 2007), but also extend our understanding of climate forcing of fecal contamination by demonstrating how it tends to increase in relation to a more nival hydrometeorological regime and greater land-use impacts within the surrounding catchment. Furthermore, inter-annual variability in snowmelt and rainfall appeared to drive variability in fecal contamination levels for half of the sites examined. Variability in snowmelt tended to drive greater changes in fecal contamination than rainfall. These results demonstrate that the positive relationship observed between precipitation events and fecal contamination levels (Kistemann et al., 2002; Cha et al., 2010) extends to inter-annual periods, which are particularly relevant to changes in climate. An observational study, such as this, is limited by not being able to isolate individual climate and land-use factors in order to unequivocally demonstrate the cause of such variability in climate forcing. However, we inferred plausible explanations for differences in the relative influence of hydro-meteorological factors on seasonal and inter-annual fecal contamination variability among sites by examining the differences in their hydrometeorological and land-use characteristics.

175

w a t e r r e s e a r c h 4 8 ( 2 0 1 4 ) 1 7 0 e1 7 8

Table 2 e Summary statistics of MLR models used to examine inter-annual FC concentration variability as function of river discharge and rainfall variability. Site name

Site no.

River discharge t-value

Cowichan River Fraser River at Hope Kettle River at Carson Kootenay River at Creston Kootenay River at Fenwick Peace River Salmon river at Salmon Arm Similkameen River at Princeton Similkameen River at US boarder

4.1.

10 15 18 21 22 31 36 39 40

5.025 4.990 11.072 e e 2.947 3.641 4.229 10.225

Seasonal fecal contamination variability

The relationships between seasonal hydro-meteorological variability and fecal contamination were highly variable among sites and hydro-meteorological regimes, although they tended to be stronger than those measured in previous studies (Cha et al., 2010; Sigua et al., 2010; Wilkes et al., 2009). Given that site specific factors around a given surface water location may strongly affect the influence of hydro-meteorological regime on seasonal fecal contamination variability, conclusions drawn from these relationships are highly generalized and certainly will not be valid in all cases. The majority of sites in catchments with nival hydrometeorological regimes experienced greatest fecal contamination during spring and summer. This was due to increasing contaminant transport by snowmelt and rainfall-induced runoff, likely augmented by a greater abundance of fecal waste due to increased wildlife and livestock activity in the local catchment. However, the relative influence of snowmelt and rainfall on fecal contamination variability resulted in peak contamination occurring during different seasons among these sites.

Fig. 3 e Relative response of fecal contamination levels to inter-annual hydro-meteorological variability among sites, demonstrated by the percentage change in annual mean FC concentration due to a 1% increase in annual mean river discharge or rainfall volume. Bars and whiskers indicate 95% confidence intervals. Note that site 15 was excluded from graph as the response of FC concentration to a 1% increase in river discharge was anomalously high (4.97 ± 1.98%).

p-value <0.001 <0.001 <0.001 e e 0.004 <0.001 <0.001 <0.001

Rainfall

Full model

t-value

p-value

Adj r2

p-value

e e e 4.599 9.447 5.854 e 3.300 3.911

e e e <0.001 <0.001 <0.001 e <0.001 <0.001

0.220 0.232 0.600 0.199 0.558 0.487 0.130 0.401 0.603

<0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001

The differences in seasonal fecal contamination variability between Nicola River (site 28, Fig. 1 graph c) and Salmon River at Silver Creek (site 37, Fig. 1 graph b), despite their very similar climate characteristics, suggest that catchment characteristics and waste management strongly moderated hydro-meteorological impacts on the seasonal variability of surface water fecal contamination. Only 0.54% of the catchment upstream from the Nicola River sampling location is used for agriculture, in comparison to 14% of the Salmon River at Silver Creek catchment. The extended period of high contamination in spring and the secondary peak during fall in the Salmon River at Silver Creek would have coincided with periods of manure spreading in the upstream catchment. Therefore, the greater influence of rainfall within this catchment can be attributed to a higher abundance of fecal waste, which also resulted in a higher mean FC concentration (72 CFU 100 ml1 and 11 CFU 100 ml1 for Salmon River at Silver Creek and Nicola River, respectively). By contrast, the annual variability of fecal contamination in the Nicola River more strongly resembled the “first flush” effect (Hathaway and Hunt, 2011), where fecal contaminants accumulated in the catchment were “flushed” into the river with the onset of snowmelt, resulting in a sharper peak and decline in contamination. The subsequent low contamination levels were likely due to the depleted abundance of fecal sources, which reduced the relative impact of subsequent rainfall events. The only evidence for winter rainfall driving peak fecal contamination was observed in the Cheakamus River (Site 3, Fig. 1 graph c), which had a hybrid meteorological regime. Fecal contamination was low during peak river discharge in July, which likely originated from glacier-melt higher upstream and thus had little relation to runoff in the local catchment. In January, however, rainfall-induced runoff appeared to increase the transport of fecal contaminants into the river, driving a peak in both the flow volume and FC concentration. We expected to observe a similar relationship of high winter rainfall driving increases in fecal contamination in surface water for pluvial catchments. By contrast, peak fecal contamination occurred with greatest frequency during the summer. This suggests that land-use impacts may be a more significant determinant of seasonal fecal contamination variability among these sites than hydro-meteorological variability. Contributions of fecal waste from farms, wildlife,

176

w a t e r r e s e a r c h 4 8 ( 2 0 1 4 ) 1 7 0 e1 7 8

and human recreation activities, such as bathing and camping, tend to be greatest during the summer (Meays et al., 2006a; Sigua et al., 2010). Therefore, summer rainfall events would have generated highly contaminated runoff, which would have received relatively low dilution due to low river flow volumes (McDonald et al., 2008). The accumulation of fecal contamination in surface water may have been enhanced during summer by increased FC survival and in-situ growth, which is associated with increased temperatures, organic matter, and dissolved solids (Tiefenthaler et al., 2009). The climate characteristics, of low rainfall and high temperature variance, associated with evidence for climate forcing of seasonal fecal contamination variability identify surface water sites in nival catchments as the most vulnerable to changes in climate altering peak contamination variability. This indicates that source water protection measures, including monitoring and modelling efforts, should be focused towards surface source water sites within catchments with nival hydro-meteorological regimes.

4.2.

Inter-annual fecal contamination variability

The positive relationship between surface water fecal contamination and hydro-meteorological variability over an inter-annual time period has only previously been measured in coastal marine water (Lipp et al., 2001; Chigbu et al., 2004). In this study, however, we were able to observe variability in the relative influence of river discharge and rainfall on FC concentrations among surface water sites. This variability is important to examine as it can help identify surface source water sites that may be more vulnerable to changes in climate altering mean fecal contamination levels. The Fraser River at Hope (site 15; Fig. 2 graph b) MLR coefficient estimate for river discharge (b ¼ 59.08) was anomalously high due to the large range in FC concentration (0e199 CFU 100 ml1) that occurred over a relatively small range in river discharge (3.74e5.12 m3/sec). This likely resulted from the mobilisation of point-source contamination, as there is little to no agricultural activity in the surrounding catchment. Point source contamination can generate high fecal contamination variability (St Laurent and Mazumder, 2012), which may account for why relatively small changes in river discharge drove relatively large changes in FC concentration. The Peace and Cowichan Rivers (sites 31 and 10; Fig. 2 graphs f and a, respectively) also demonstrated relatively large changes in fecal contamination in response to river discharge variability. In the case of the Peace River this was likely due to relatively small changes to the high flow volume (mean ¼ 1498.18 m3/sec) being associated with relatively large changes in FC concentration, which would have resulted from the substantial increase in contaminant transport associated with greater snowmelt within this watershed. In the case of the Cowichan River, which had the greatest extent of agricultural cover (45%) among the sites, activities within the catchment would have provided an abundance of diffuse sources of contamination to be mobilised into the river during periods of high runoff. The Salmon River at Salmon Arm (site 36; Fig. 2 graph h) showed the lowest response in fecal contamination to changes in river discharge. This was due to relatively low FC concentration variability (range ¼ 58.8e112.19 CFU 100 ml1)

among consistently high fecal contamination levels (mean ¼ 87.27 CFU 100 ml1); therefore, changes in river discharge did not drive high variability in fecal contamination levels. Inter-annual rainfall variability had the greatest impact on fecal contamination levels in the Kootenay River at Creston (site 21; Fig. 2 graph d), likely due to the surrounding catchment receiving the greatest mean rainfall (4.09 mm) and having the highest proportion of agricultural cover (19%) among these sites. These conditions would have resulted in high volumes of rainfall-induced runoff encountering an abundance of diffuse fecal sources, which would have greatly elevated fecal contamination during periods of high rainfall. Conversely, the Peace river (site 31; Fig. 2 graph g) demonstrated the lowest response in fecal contamination levels to changes in inter-annual rainfall variability, likely due to having a combination of the lowest annual mean rainfall (1.75 mm), a very low range of FC concentrations (2.33e5.53 CFU 100 ml1), and little agricultural cover (3%) in its local catchment. The climate characteristics of sites that showed evidence for climate forcing of inter-annual fecal contamination variability did not strongly differentiate them from those with no evidence for climate forcing. This may have arisen due to the sample of catchments having predominantly nival, and therefore very similar, hydro-meteorological regimes (n ¼ 15, 1, and 3 for nival, hybid, and pluvial, respectively). Therefore, our observations are limited to indicating that climate forcing of fecal contamination is more likely to occur in nival influenced catchments with greater total rainfall. We hypothesized that surface water fecal contamination originating from diffuse sources would be more strongly related to hydro-meteorological variability than contamination from point sources. Our observations indicate that interactions between hydro-meteorological variability and different sources of fecal contamination are too complex to support this hypothesis. The examples of the Fraser River at Hope and the Cowichan River demonstrate that climate factors can strongly influence surface water fecal contamination by mobilising both point and diffuse sources. As climate forcing was not greater among group 1 than group 2 sites, runoff mediated transport did not appear to be a more significant diver of diffuse source than point source contamination. On the one hand, this might have resulted from the impediment of diffuse fecal contaminant transport, due to factors such as entrapment in riparian vegetation, retention ponds, and artificial dams (Ferguson et al., 2003). On the other hand, runoff variability may have mobilized point sources of contamination by causing combined sewer overflows and septic leakage (Arnone and Walling, 2007). These observations suggest that climate factors are important drivers of both diffuse and point source contamination, although their subsequent influence on surface water contamination levels is strongly dependent on site specific land-use factors.

4.3.

Implications and recommendations

Quantifying and anticipating drivers of surface source water fecal contamination is necessary for improving the safety of

w a t e r r e s e a r c h 4 8 ( 2 0 1 4 ) 1 7 0 e1 7 8

drinking water and public health. This has been demonstrated by the positive association between precipitation and waterborne disease outbreaks in North America (Curriero et al., 2001), and by the highly seasonal variability of waterborne disease outbreaks in Canada, which were twofold greater during spring and summer than during fall and winter between 1975 and 2001 (Thomas et al., 2006). The relationships observed in this study demonstrate the capacity of precipitation variability and seasonal changes in hydro-meteorological conditions to drive drinking source water contamination, which can subsequently lead to waterborne disease. Minimising the risk of waterborne disease requires consideration of how projected changes in climate may influence surface source water quality. We observed climate forcing of seasonal and inter-annual FC concentration variability for only 30% and 50% of the sites, respectively. Changing hydro-meteorological conditions, however, are likely to have significant impacts on fecal contamination levels and variability among these sites. Our analysis suggests that increasing temperatures and total rainfall in nival catchments will necessitate surface source water quality protection by reducing the release and transport of fecal contaminants during summer, fall, and winter. In pluvial catchments, however, land-use factors appeared to determine seasonal fecal contamination variability, therefore, source water protection efforts need to be targeted towards the summer period despite this being the time of lowest rainfall and runoff. The reduction and control of point and diffuse sources of contamination is also likely to reduce the negative impact of increases in total precipitation on surface source water quality. Lack of data for FIB and waterborne pathogens is a major limiting factor when investigating hydro-meteorological impacts on fecal contamination. Sampling variability among surface water sites tends to be very high, with the majority only being sampled on a few occasions each year. This is particularly inadequate with regard to bacterial populations, which are highly variable in surface water due to the many factors that influence their transport, fate, and survival in the environment (Meays et al., 2006b). Furthermore, monitoring of fecal contamination tends to be limited to a single FIB despite the likelihood that other FIB and waterborne pathogens may vary in their response to hydro-meteorological drivers. Anticipating FIB and waterborne pathogen variability in surface source water requires the development of process models, the accuracy of which is contingent on extensive reliable data regarding bacterial inputs, temperature, hydrology, soil and vegetation properties, and the condition of the riparian zone (Zhu et al., 2011). Model calibration and accuracy can be increased through the provision of high quality, high frequency, and long term data sets. There is a need, therefore, to increase both the frequency and range of FIB and waterborne pathogens monitored in surface source water in order to measure and anticipate the impacts of changing hydrometeorological conditions on fecal contamination levels. This can be facilitated by the development and use of quantitative polymerase chain reaction methods to enumerate FIB and waterborne pathogens in source water samples. Our analysis suggests that increased monitoring of FIB and waterborne pathogens should be prioritized towards surface

177

water sites located in watersheds with nival hydrometeorological regimes and greater land-use impacts. High frequency and long-term data for such sites will improve our capacity to identify and target source water protection measures towards those that are most vulnerable to climate change impacts on fecal contamination levels and anticipate the need to adjust drinking water treatment disinfection processes so that they safely neutralise microbial pathogens in raw water while also minimising disinfection by product formation in finished water.

Acknowledgements This work was funded by the National Science and Engineering Research Council of Canada (NSERC) through the Res’eau Waternet Research Network Grant, Public Health Agency of Canada Grant, and NSERC-Industrial Research Chair Grant to A. Mazumder. We would like to thank the British Columbia Ministry of Environment for access to data, Klaas Broersma for helpful comments and edits, and two anonymous reviewers for their insightful criticism.

Appendix A. Supplementary data Supplementary data related to this article can be found at http://dx.doi.org/10.1016/j.watres.2013.09.031.

references

Arnone, R., Walling, J., 2007. Waterborne pathogens in urban watersheds. J. Water Health 05 (1), 149e162. Bartram, J., Cairncross, S., 2010. Hygiene, sanitation, and water: forgotten foundations of health. Pub. Library Sci. Med. 7 (11), 1e9. Bates, B., Kundzewicz, Z., Wu, S., Palutikof, J. (Eds.), 2008. Climate Change and Water. Technical Paper of the Intergovernmental Panel on Climate Change. IPCC Secretariat, Geneva, p. 210. British Columbia Ministry of Environment (BC MoE), 2009. Water and Air Monitoring and Reporting. Sampling Methods and Quality Assurance. BC MoE, Victoria, BC. http://www.env.gov. bc.ca/epd/wamr/labsys/lab_meth_manual.html (accessed 25.07.11.). Cha, S., Lee, S., Park, Y., Cho, K., Lee, S., Kim, J., 2010. Spatial and temporal variability of fecal indicator bacteria in an urban stream under different meteorological regimes. Water Sci. Technol. 61 (12), 3102e3108. Charron, D., Thomas, K., Waltner-Toews, D., Aramini, J., Edge, T., Kent, R., Maarouf, A., Wilson, J., 2004. Vulnerability of waterborne diseases to climate change in Canada: a review. J. Toxicol. Environ. Health (Part A) 67, 1667e1677. Chigbu, P., Gordon, S., Strange, T., 2004. Influence of inter-annual variations in climatic factors on fecal coliform levels in Mississippi Sound. Water Res. 38, 4341e4352. Curriero, F., Pat, J., Rose, J., Subhash, L., 2001. The association between extreme precipitation and waterborne disease outbreaks in the United States, 1948e1994. Am. J. Public Health 91 (8), 1194e1199. Delpla, I., Jung, A., Baures, E., Clement, M., Thomas, O., 2009. Impacts of climate change on surface water quality in relation to drinking water production. Environ Int. 35, 1225e1233.

178

w a t e r r e s e a r c h 4 8 ( 2 0 1 4 ) 1 7 0 e1 7 8

Dery, S.J., Stahl, K., Moore, R.D., Whitfield, P.H., Menounos, B., Burford, J.E., 2009. Detection of runoff timing changes in pluvial, nival, and glacial rivers of western Canada. Water Resour. Res. 45, W04426. http://dx.doi.org/10.1029/ 2008WR006975. Dorner, S., Anderson, W., Gaulin, T., Candon, H., Slawson, R., Payment, P., Huck, P., 2007. Pathogen and indicator variability in a heavily impacted Watershed. J. Water Health 05 (2), 241e257. Environment Canada (EC), 2003. Environment Canada Water Quality Monitoring Program, Online Water Quality Database. EC, Ottawa, ON. http://waterquality.ec.gc.ca/waterqualityweb/ searchtext.aspx (accessed 25.07.11.). Ferguson, C., Husman, A., Altavilla, N., Ashbolt, N., Deere, D., 2003. Fate and transport of surface water pathogens in watersheds. Crit. Rev. Environ. Sci. Technol. 33, 299e361. Hathaway, J., Hunt, W., 2011. Evaluation of first flush for indicator bacteria and total suspended solids in urban stormwater runoff. Water, Air Soil Pollut. 217, 135e147. Jokinen, C., Edge, T., Koning, W., Laing, C., Lapen, D., Miller, J., Mutschall, S., Scott, A., Taboada, E., Thomas, J., Topp, E., Wilkes, G., Gannon, V., 2012. Spatial and temporal drivers of zoonotic pathogen contamination of an agricultural watershed. J. Environ. Qual. 41 (1), 242e252. Kay, D., Crowther, J., Stapleton, C., Wyer, M., Fewtrell, L., Anthony, S., Bradford, M., Edwards, A., Francis, C.A., Hopkins, M., Kay, C., McDonald, A., Watkins, J., Wilkinson, J., 2008. Faecal indicator organism concentrations and catchment export coefficients in the UK. Water Res. 42, 2649e2661. Kistemann, T., Claben, T., Koch, C., Dangendorf, F., Fischer, R., Gebel, J., Vacata, V., Exner, M., 2002. Microbial load of drinking water reservoir tributaries during extreme rainfall and runoff. Appl. Environ. Microbiol. 68 (5), 2199e2197. Kloot, R., 2006. Locating E. coli contamination in a South Carolina watershed. J. Environ. Manage. 83, 402e408. Lipp, E., Schmidt, N., Luther, M., Rose, J., 2001. Determining the effects of El Nin˜o-Southern Oscillation events on coastal water quality. Estuaries 24 (4), 491e497. Marsh, P., Woo, M., 1981. Snowmelt, glacier melt, and high arctic streamflow regimes. Canadian J. Earth Sci. 18, 1380e1384. McDonald, A., Chapman, P., Fukasawa, K., 2008. Microbial status of natural waters in a protected wilderness area. J. Environ. Manage. 87, 600e608. Meays, C., Broersma, K., Nordin, R., Mazumder, A., Samadpour, M., 2006a. Spatial and annual variability in concentrations and sources of E. coli in multiple watersheds. Environ. Sci. Technol. 40, 5289e5296. Meays, C., Broersma, K., Nordin, R., Mazumder, A., Samadpour, M., 2006b. Diurnal variability in concentrations and sources of E. coli in three streams. Can. J. Microbiol. 52, 1130e1135.

Patz, J., Vavrus, S., Uejio, C., McLellan, S., 2008. Climate change and waterborne disease risk in the Great Lakes region of the U.S. Am. J. Prev. Med. 35 (5), 451e458. Perdek, J., Arnone, R., Stinson, M., 2003. Managing Urban Watershed Pathogen Contamination. U.S. Environmental Protection Agency, Cincinnati, Ohio, pp. 3e24. EPA/600/R-03/ 111. R Development Core Team, 2009. R: a Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. ISBN 3-900051-07-0, http://www. R-project.org. Schnorbus, M., Bennett, K., Werner, A., Berland, A., 2011. Hydrologic Impacts of Climate Change in the Peace, Campbell, and Columbia Watersheds, British Columbia, Canada. Pacific Climate Impacts Consortium: University of Victoria, Victoria, BC, p. 157. Shabbar, A., Bonsal, B., Khandekar, M., 1997. Canadian precipitation patterns associated with the Southern Oscillation. J. Clim. 10, 3016e3027. Sigua, G., Palhares, J., Kich, J., Mulinari, M., Mattei, R., Klein, J., Muller, S., Plieske, G., 2010. Microbiological quality assessment of watershed associated with animal-based agriculture in Santa Catarina, Brazil. Water Air Soil Pollut. 210, 307e316. St Laurent, J., Mazumder, A., 2012. The influence of land-use composition on fecal contamination of riverine source water in southern British Columbia. Water Resour. Res. 48, W00M03. http://dx.doi.org/10.1029/2012WR012455. Tate, K., Pereira, M., Atwill, E., 2004. Efficacy of vegetated buffer strips for retaining Cryptosporidium parvum. J. Environ. Qual. 33, 2243e2225. Thomas, K., Charron, D., Waltner-Toews, D., Schuster, C., Maarouf, A., Holt, J., 2006. A role of high impact weather events in waterborne disease outbreaks in Canada, 1975 e 2001. Int. J. Environ. Health Res. 16 (3), 167e180. Tiefenthaler, L., Stein, E., Lyon, G., 2009. Fecal indicator bacteria (FIB) levels during dry weather from Southern California reference streams. Environ. Monitoring Assess. 155, 477e492. Wentz, F., Ricciardulli, L., Hilburn, K., Mears, C., 2007. How much rain will global warming bring? Science 317, 233e235. Wilkes, G., Edge, T., Gannon, V., Jokinen, C., Lyautey, E., Medeiros, D., Neumann, N., Ruecker, N., Topp, E., Lapen, D., 2009. Seasonal relationships among indicator bacteria, pathogenic bacteria, Cryptosporidium oocysts, Giardia cysts, and hydrological indices for surface waters within an agricultural landscape. Water Res. 43, 2209e2223. Zhu, Z., Broersma, K., Mazumder, A., 2011. Model assessment of cattle and climate impacts on stream fecal coliform pollution in the Salmon River watershed, British Columbia, Canada. Water Air Soil Pollut. 215, 155e176.