Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Influence of climate and land use on nutrient and bacterial dynamics in surface waters of the Lower Fraser… Ross, James Donald 2006

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Notice for Google Chrome users:
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.

Item Metadata

Download

Media
831-ubc_2007-267868.pdf [ 25.3MB ]
Metadata
JSON: 831-1.0100420.json
JSON-LD: 831-1.0100420-ld.json
RDF/XML (Pretty): 831-1.0100420-rdf.xml
RDF/JSON: 831-1.0100420-rdf.json
Turtle: 831-1.0100420-turtle.txt
N-Triples: 831-1.0100420-rdf-ntriples.txt
Original Record: 831-1.0100420-source.json
Full Text
831-1.0100420-fulltext.txt
Citation
831-1.0100420.ris

Full Text

Influence of cl imate and land use on nutrient and bacterial dynamics in surface waters of the Lower Fraser Val ley, British Co lumb ia by James Donald Ross B.Sc. Queen's University, 1995 M.Sc. Queen's University, 1998 A thesis submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in The Faculty of Graduate Studies (Resource Management and Environmental Studies) The University of British Columbia December 2006 © James Donald Ross, 2006 ABSTRACT It is understood that intensive agricultural activities can adversely impact surface-water quality resulting in r isks to ecosys tem and human health. What is less clear are the l inks between agricultural land use (type and intensity), environmental condit ions and surface-water quality at varying spatial and temporal sca les . There are also chal lenges with detecting agricultural influence on surface waters in a timely and accurate manner. This is of concern in the Lower Fraser Val ley as this region has exper ienced significant agricultural intensification and population growth in recent years. This study examined inf luences of agricultural land use , cl imate and hydrology on water quality in three watersheds to identify land-use pract ices and environmental condit ions producing the greatest risk of contaminat ion. Th is was accompl ished through an intensive surface-water sampl ing program to a s s e s s nutrient and bacterial dynamics in the Hatzic, Elk Creek and Sa lmon watersheds, combined with hydrometric and meteorological monitoring from 2002-2005. Spect roscop ic techniques (absorption and f luorescence) were a lso evaluated as tools to detect and quantify agricultural. influence. Consistent correlations between agricultural land use and contamination (nutrient and bacterial concentrat ions) were observed ac ross all watersheds. Seasona l trends were consistent, with nutrient concentrat ions peaking during winter months (illustrating strong hydrological control over mobil isation and transport) and bacterial concentrat ions peaking during summer months (illustrating the supply-constrained nature of bacterial stores). Contaminant concentrat ions correlated with measures of agricultural intensity. L ivestock operat ions represented the highest-risk land use for contamination, with even smal l operat ions producing observable impacts on water quality. Temporal ly , the greatest risk of bacterial contamination was assoc ia ted with storm events preceded by per iods of dry weather during summer months. Absorpt ion and f luorescence were effective measures of agricultural inf luence as they quantify and character ize agriculturally-derived d isso lved organic matter. Advan tages of these techniques include rapid sample process ing, minimal requirements for sample treatment and vo lume. Further, they provide qualitative information regarding water quality, water source and land use that is not avai lable from nutrient or bacterial ana lyses alone. T h e s e techniques do not accurately detect contaminants in areas with minimal agricultural inf luence and therefore are limited as direct indicators of bacterial or nutrient concentrat ions. T A B L E OF CONTENTS ABSTRACT ii T A B L E OF CONTENTS iii LIST OF TABLES vi LIST OF FIGURES viii LIST OF ABBREVIATIONS AND COMMON TERMS xii ACKNOWLEDGEMENTS xiii 1. Introduction 1 1.1. Background 1 1.2. Drinking water and public health in British Co lumb ia 2 1.3. Resea rch context 3 1.4. Thes is organization 5 2. Literature review 7 2.1. Introduction 7 2.2. Agriculture in the Lower F raser Va l ley 7 2.2.1. Overv iew 7 2.2.2. Land-use change: 1986-2001 9 2.3. Susceptibi l i ty of water resources to contamination in the L F V 12 2.3.1. Surficial geology and aquifer vulnerability 12 2.3.2. Cl imate 13 2.4. Nutrients 15 2.4.1. Nutrient sources in agricultural watersheds 15 2.4.2. Transport p rocesses 17 2.4.3. Ecologica l and human health impacts 24 2.5. Pathogens 26 2.5.1. Pathogens of concern 26 2.5.2. Pathogen sources in agricultural watersheds 32 2.5.3. Transport p rocesses 37 2.5.4. Ecologica l and human health impacts 43 2.6. Water-quality monitoring 48 2.6.1. Detection methods 49 2.6.2. Parameters 51 2.6.3. Monitoring f rameworks • 54 2.7. Spect roscopy and water quality 59 2.7.1. Theoret ical background 59 2.7.2. Absorbance spect roscopy 61 2.7.3. F luorescence spect roscopy 63 2.7.4. Advantages and limitations of spectroscopic techniques 66 2.7.5. Appl icat ions of spectroscopy in environmental research 68 2.8. Current gaps and opportunities 78 3. Site descriptions and methods 80 3.1. Introduction 80 3.2. Site select ion and overview 80 3.2.1. The Hatzic watershed 83 3.2.2. The Elk Creek watershed 85 3.2.3. The Sa lmon River watershed 86 3.3. Field methods 88 3.3.1. Meteorological monitoring 88 3.3.2. Hydrometric monitoring 89 iii 3.3.3. Water quality monitoring 92 3.4. Laboratory methods 95 3.4.1. Nutrient analys is 95 3.4.2. Microbiological analys is 96 3.4.3. Spectrophotometr ic analys is 96 3.5. Data analysis and representation 97 4. Influence of land use, climate and hydrological conditions on nutrient and bacterial cycling in an agricultural watershed 99 4.1 . Introduction 99 4.2. Methods 99 4.3. Resul ts 100 4.3 .1 . Hydrology and cl imate of the Hatzic watershed 100 4.3.2. Spat ial trends in surface water quality 108 4.3.3. Tempora l trends in sur face water quality 116 4.3.4. Storm-event dynamics 121 4.4. Conc lus ions 1 2 6 5. The effect of agricultural intensity on surface-water quality in the Lower Fraser Valley 129 5.1. Introduction 129 5.2. Methods 129 5.3. Agricultural intensity 129 5.3.1. Quantifying agricultural intensity 129 5.3.2. Hatzic, Elk Creek and Sa lmon watersheds 130 5.4. Resul ts and d iscuss ion 133 5.4.1. Compar i son of land-use types 133 5.4.2. Cumulat ive downst ream impacts 144 5.5. Conc lus ions 151 6. Absorbance spectroscopy as a tool to detect agricultural influence on water quality 154 6.1. Introduction 154 6.2. Methods 155 6.2.1. Samp le collection 155 6.2.2. Spect roscop ic analys is 156 6.2.3. Spectral s lope 157 6.3. Resul ts and d iscuss ion .....158 6.3.1. Absorbance spec t ra (200-800 nm) 158 6.3.2. Far U V (190-220 nm) 161 6.3.3. U V - C (220-290 nm) 164 6.3.4. Vis ible (440 nm) 169 6.3.5. Spectra l s lope 169 6.3.6. Storm event dynamics - 172 6.3.7. Absorbance spect roscopy as a qualitative tool 176 6.3.8. Absorbance spect roscopy as a quantitative tool 182 6.4. Conclusions 186 7. Fluorescence spectroscopy as a tool to detect agricultural influence on water quality 188 7.1. Introduction 188 7.2. Methods 189 7.3. Resul ts and d iscuss ion 190 7.3.1. E E M features 190 7.3.2. Site-to-site variability 191 7.3.3. Land use and f luorescence intensity 197 7.3.4. Tempora l trends 200 iv 7.3.5. Relat ionships with other indicators 203 7.4. Conc lus ions. . . ! 208 8. Integrated Discussion and Conclusions 211 8.1. Introduction 211 8.2. Integrated d iscuss ion 211 8.2.1. The link between agricultural land use type and intensity and nutrient and bacterial contributions to surface waters in agricultural watersheds 212 8.2.2. The influence of meteorological condit ions on nutrient and bacterial dynamics in surface waters in agricultural watersheds 215 8.2.3. The potential of spect roscopic techniques to detect agricultural influence 216 8.3. Limitations and opportunities for future work 219 References 224 Appendix A - Real-time, field-based spectroscopic sensors 244 V LIST OF TABLES Table 2-1 - Typical moisture content and nutrient content of var ious manures on a wet-weight bas is (modified from B C Ministry of Agriculture, Food and Forests , 2004) 16 Table 2-2 - Preva lence of common waterborne pathogens in l ivestock populat ions and geometr ic mean concentrat ions of organ isms (per gram of feces) . All numbers are for fresh manure except those in () which represent samples from stored manure. Modif ied from Hutchison et a l . (2005a), except where otherwise noted. N D = no data 34 Table 2-3 - Waterborne d isease outbreaks in B C from 1980-2004. Modif ied from (BC Auditor Genera l , 1999; Peck , 2004) 45 Table 2-4 - C o m m o n detection techniques for var ious water-quality parameters 50 Table 2-5 - The multi-barrier approach for drinking water sys tems in the context of the "source-to-tap" framework (modified from O'Connor , 2002) 56 Table 3-1 - Compar i son of the Hatzic, Elk Creek and Sa lmon watersheds in terms of s ize , land use and stocking densit ies (modified from Schre ier et a l . , 2003). " X " denotes no data avai lable 83 Table 3-2 - Land-use categor ies for the three watersheds based on contributing a rea 93 Table 4-1 - Daily rainfall at four weather stations in the Lower Fraser Val ley for the major storm event in October, 2003 (Environment C a n a d a , 2005). Numbers in brackets represent station elevation 106 Table 4-2 - Descript ive statistics for ammonia , nitrate and orthophosphate (2002-2005) . Note that concentrat ions below detection limits were ass igned a value of 0.5 x detect ion limit 109 Table 4-3 - Descript ive statistics for total and fecal coliform by station (2003-2005) 111 Table 4-4 - Descript ive statistics for total and fecal coliform data by land use (2003-2005) 114 Table 4-5 - Descript ive statistics for total and fecal coliform by land use and s e a s o n 120 Tab le 5-1 - An imal unit convers ion coefficients (Beaul ieu et a l . , 2001) 130 Table 5-2 - Land-use distributions and nutrient surp luses for the Hatzic, Elk Creek and Sa lmon watersheds (note that nutrient surp luses are calculated based on the E A ' s within which these watersheds are located) 131 Table 5-3 - Est imates of l ivestock populations in the Hatz ic watershed 132 Table 5-4 - An imal stocking densit ies for the Hatzic, Elk Creek and Sa lmon watersheds (see text for sources) 133 Table 5-5 - Water-qual i ty data by land use and watershed (Salmon data include groundwater) 134 Table 6-1 - Samp le collection dates (*) and rainfall (mm) for the previous 5 days for the Hatzic, Elk Creek and Sa lmon watersheds 156 Table 6-2 - Absorbance wavelengths used as proxies for common water-quality parameters 157 Table 6-3 - Absorbance at 280 nm by land use for the Hatzic, Elk Creek and Sa lmon watersheds 167 Table 6-4 - Absorbance at 440 nm by land use for the Hatzic, Elk Creek and Sa lmon watersheds 169 Table 6-5 - M e a n S values by land use and watershed for the three days of sample collection 170 Table 6-6 - M e a n D O C concentration by land use and watershed for samp les col lected on August 31, 2005 .• 183 Table 6-7 - Correlat ions between nutrient and bacterial indicators and absorbance at specif ic wavelengths 184 Table 7-1 - Wavelengths for f luorescence peaks a s s e s s e d in this study 190 Table 7-2 - M e a n f luorescence intensities by land use for the Hatzic, Elk Creek and Sa lmon watersheds. 198 vi Tab le 7-3 - Spea rman rank correlat ions between f luorescence intensity and nutrient and bacter ia concentrat ions for all samp les from the three watersheds (numbers in brackets represent sample size) 204 Table 7-4 - Spea rman rank correlations between f luorescence parameters and water-quality indicators, by watershed (numbers in brackets represent sample size) 205 Tab le 7-5 - Spea rman rank correlat ions between f luorescence intensities and bacterial and nutrient concentrat ions for agricultural sites in the Hatzic, Elk Creek and Sa lmon watersheds (numbers in brackets represent sample size) 206 vii LIST OF FIGURES Figure 2-1 - Map of the Lower Fraser Val ley with census enumerat ion areas. Note that R ichmond, Langley, Matsqui and Chi l l iwack include severa l enumerat ion areas (East and Wes t R ichmond, North, Central and South Langley, North, South and Wes t Matsqui and East and Was t Chi l l iwack). ..8 Figure 2-2- Average monthly rainfall at Abbotsford airport for the period 1971-2000 (Environment C a n a d a , 2005) 14 Figure 2-3 - a) simplif ied nitrogen cycle showing critical p rocesses , and b) key chemical reactions in the nitrogen cycle (note: the equat ions for ammonif icat ion and immobil ization are omitted for clarity as they are multi-step p rocesses) . Modif ied from Pierzynski et al . (1994) 18 Figure 2-4 - Simpli f ied phosphorus cyc le , illustrating sources and pathways by which phosphorus is lost from the soil profile. Modif ied from Sutton (1997) 22 Figure 2-5 - Percentage of drinking water in British Co lumb ia der ived from different sources as of 2001 (BC Provincial Health Officer, 2001) 47 Figure 2-6 - A framework for source protection as part of the multi-barrier approach (modified from Federal-Provincial-Terr i tor ial Commit tee on Drinking Water and C C M E Water Quality Task Group , 2004) : , 58 Figure 2-7 - UV-v is ib le-near infrared spectrum with wavelengths in nm 59 Figure 2-8 - Jab lonsk i d iagram illustrating the p rocesses of absorpt ion, internal convers ion and f luorescence. T imes assoc ia ted with each process are also provided (modified from Lakowicz, 1999). 60 Figure 2-9 - Schemat ic representation of an absorbance spectrophotometer 62 Figure 2-10 - a) Idealised absorbance spectrum and emiss ion spectrum (explained in Sect ion 2.7.3) illustrating the change in absorbance with wavelength and b) typical absorbance spectrum for a surface-water sample , showing decreas ing absorbance with increasing wavelength 63 Figure 2-11 - Emiss ion spectra for river water (at three excitation wavelengths) illustrating severa l features: 1) Rayle igh-Tyndal l scattering, 2) R a m a n scattering and 3) variat ions in f luorescence intensity based on excitation and emiss ion wavelengths (see text for explanation) 64 Figure 2-12 - Contour plot of an exci tat ion-emission matrix for river water showing the Rayle igh-Tyndal l line, R a m a n line and f luorescence peaks 66 Figure 2-13- Average composi t ion of d isso lved organic matter in river water with 5 mg-L" 1 d isso lved organic carbon. Modif ied from Thurman (1985) 70 Figure 2-14 - Abso rbance spectrum for E C - 4 col lected on December 20, 2004 (dashed) and second-derivative of the s a m e spectrum (solid). Note second-der ivat ive peak at 224 nm 77 Figure 3-1 - Lower Fraser Val ley with locations of three study watersheds 81 Figure 3-2 - Mean monthly precipitation and mean daily temperatures (minimum, maximum and average) at Chi l l iwack (near the Elk Creek watershed) from 1971 - 2000 (Environment C a n a d a , 2005) 82 Figure 3-3 - M a p of the Hatz ic Val ley watershed showing topography, major s t reams, sampl ing sites and hydrometric and cl imate stations 84 Figure 3-4 - Map of the Elk Creek watershed showing topography, major s t reams, sampl ing sites and hydrometric and cl imate stations 86 Figure 3-5 - Map of the Sa lmon River watershed showing contours, major s t reams, sampl ing sites and hydrometric and cl imate stations. Note also the location of the Hopington aquifer 87 Figure 3-6 - a) meteorological station located in the Hatzic watershed, b) data logger, power supply and field laptop used to download data 89 VIII Figure 3-7 - Photographs illustrating: a) cable housing, Y-joint and removable cap providing a c c e s s to sensors for c leaning and calibration (lower hydrometric station); and b) data logger mounted in wooden cabinet on tree at upper hydrometric station 91 Figure 4-1 - 24-hour rainfall, water level and speci f ic conductance for the upper and lower hydrometric stations in the Hatzic watershed. Black horizontal l ines denote missing data as a result of logger failure. Three major storm events are also noted (see text for details) 101 Figure 4-2 - Water temperature for the period of record for the upper and lower hydrometric stations. ..103 Figure 4-3 - Rainfal l , water level and turbidity data illustrating: a) c lear correlation between the three var iables at the lower hydrometric station during the moderate wet season of 2002-2003, and b) an example of fouling of the turbidity sensor at the upper station after heavy rains 105 Figure 4-4 - a) Lower hydrometric station from the west, illustrating low-flow condit ions typical of a dry period during winter, and b) looking south at the s a m e hydrometric station from the bank during the October, 2003 event (note data logger and submerged dock railing for reference) 107 Figure 4-5 - Boxplots of a) total and b) fecal coliform concentrat ions demonstrat ing increasing bacterial contributions with downstream distance from forested headwaters. Note: outliers were not removed as they represent verified results 112 Figure 4-6 - Log10 Feca l coliform vs . Log10 E. coliior March 23 and May 1, 2003 showing strong posit ive correlation between concentrat ions of the two bacter ia in stream samples . Symbo ls also illustrate concentrat ions by land-use category 113 Figure 4-7 - M a p of the Hatz ic watershed illustrating locations of horse and dairy and beef cattle operat ions and stream sampl ing sites with the highest observed fecal coliform concentrat ions (red dots) 115 Figure 4-8 - Nutrient concentrat ions vs. water level (upper hydrometric station) for the period of record. Note higher concentrat ions in first wet s e a s o n . R e d bars denote mean concentrat ions for day of sampl ing 117 Figure 4-9 - a) rainfall, s tage and chloride and nutrient concentrat ions, and b) rainfall, s tage, turbidity and coliform concentrat ions measured every three hours from March 18-21, 2005 at the lower hydrometric station (turbidity measured at upper station) 123 Figure 5-1 - a) Median fecal and total coliform concentrat ions, and b) mean nutrient concentrat ions at forested sites in the Hatzic and Elk Creek watersheds. Error bars represent 9 5 % conf idence intervals 135 Figure 5-2 - Med ian total and fecal coliform concentrat ions for agricultural si tes in the Hatzic, Elk Creek and Sa lmon watersheds. Total coliform and fecal coliform values were higher in the Elk Creek watershed than in the Hatz ic watershed (P < 0.001 and P = 0.018, respectively) and Sa lmon watershed (P = 0.030 and P = 0.048, respectively). Error bars represent 9 5 % conf idence intervals. 140 Figure 5-3 - M e a n N H 4 + , N 0 3 " and P 0 4 3 " concentrat ions at agricultural sites in the Hatzic, Elk Creek and Sa lmon watersheds (error bars represent 9 5 % conf idence intervals) 141 Figure 5-4 - Downstream trends in N 0 3 " in the mainstem river of a) the Hatzic and b) the Elk Creek watersheds and for tributaries to the mainstem in c) the Hatz ic and d) the Elk Creek watersheds. Percentages represent the proportion of total contributing a rea under agriculture 145 Figure 5-5 - Downstream trends in N 0 3 " concentrat ions in the Sa lmon watershed, for a) the Sa lmon river and b) Cogh lan Creek above and below its conf luence with the Sa lmon river. Note the significant influence of the Hopington aquifer on surface-water nutrient levels 147 Figure 5-6 - Downstream trends in log fecal coliform concentrat ions in the mainstem river of a) the Hatzic and b) the Elk Creek watersheds and for tributaries to the mainstem in c) the Hatz ic and d) the Elk Creek watersheds 149 Figure 5-7 - Downstream trends in fecal coliform concentrat ions in the Sa lmon watershed for a) the Sa lmon River and b) Cogh lan Creek above and below its conf luence with the Sa lmon River 150 ix Figure 6-1 - a) plot of a typical absorbance spectrum (solid line) and the In-transformed spectrum for the s a m e sample (dashed line); b) least-squares fit to In-transformed absorbance data between 290-450 nm (r2=0.99) 158 Figure 6-2- a) absorbance spect ra for samples col lected in the Hatzic watershed on August 31 , 2004, plotted by land use ; b) absorbance spect ra for samples col lected in the Sa lmon watershed at: 1) S A -1G (dashed, right axis), a deep well with significant N 0 3 " contamination (-11 mg-L" 1) as illustrated by the peak at ~ 220 nm and 2) S A - 3 G (left axis), a municipal well with N O V concentrat ions below detection limits. The sharp drop at 350 nm is an analytical artifact. Note the different sca les for the two Y - a x e s 160 Figure 6-3 - Abso rbance at 220 nm vs. land use for the Hatzic, Sa lmon and Elk Creek watersheds combined 161 Figure 6-4 - a) linear regression of absorbance at 220 nm vs. nitrate concentrat ions measured using the Lachat Q u i k C h e m method, with r2 = 0.91 (r 2 = 0.85 with three highest nitrate va lues removed); b) linear regression of second-derivat ive absorbance at 224 nm vs. nitrate, with r 2 = 0.99 (r 2 = 0.99 with three highest nitrate va lues removed) 163 Figure 6-5 - Absorbance spectra for all sites in the Hatz ic watershed over the three days of sampl ing. Note the consistent presence of a shoulder at 280 nm at forested and mixed sites on August 31 , 2004. Group ing of spect ra by land use is still observed but the distinction is not as obv ious as for the Elk Creek due to the lower intensity of land use in this catchment 165 Figure 6-6 - Abso rbance spect ra for all sites in the Elk Creek watershed over the three days of sampl ing. Note the consistent presence of a shoulder at 280 nm at forested and mixed sites on August 31, 2004. Note a lso the consistent distinction between land-use types and the consistent ly high absorbance va lues for the sites under the greatest degree of agricultural inf luence ( E C - 4 , E C - 5 and EC-6) 166 Figure 6-7 - Abso rbance at 280 nm by station for the: a) Hatzic, b) Elk and c) Sa lmon watersheds 168 Figure 6-8 - Abso rbance at 440 nm vs . spectral s lope for the Hatzic, Elk Creek and Sa lmon watersheds. Note clustering of samp les by land use 171 Figure 6-9 - Water level, 1-hour rainfall and spectral propert ies of samples col lected at the lower hydrometric station in the Hatzic S lough during a storm event, March 18-21, 2004. Note increase in A 4 4 0 during second storm peak 174 Figure 6-10 - Spectra l s lope vs. A 4 4 0 for the storm event in the Hatzic watershed. Chlor ide concentrat ions are represented by bubble s ize. Numbers refer to the consecut ively numbered samp les which were col lected every 3 hours, from 09:00 on March 18 to 09:00 on March 21 , 2005 176 Figure 6-11 - T ime ser ies of absorbance spect ra for storm event in Hatzic watershed. Samp les were col lected every 3 hours starting at 09:00 on March 18 (Y-axis lines align with discrete spect ra for each sample and correlate with the data points on Figure 6) 180 Figure 6-12 - In-transformed spectra for the storm event captured in the Hatz ic watershed. Note the clear dec rease in spectral s lope assoc ia ted with the second peak in river stage, suggest ing a shift to increased high-molecular-weight compounds in transported D O M 181 Figure 6-13 - Absorbance va lues vs. a) nitrate, b) ammonium and c) fecal coliform concentrat ions for all three watersheds. Note the strong, consistent correlation between 2nd derivative absorbance at 224 nm compared to correlations for N H 4 + and fecal col i form, which vary by watershed 185 Figure 7-1 - E E M illustrating typical agricultural f luorescence patterns as well as regions used for extraction of f luorescence peaks ( T r y p t o p h a n and Tyr=tyrosine) 191 Figure 7-2 - E E M ' s for: a) HV-2 , b) H V - 5 , c) HV-9 , d) HV-14 , e) HV-18 and f) HV-20 in the Hatzic watershed (Aug. 13, 2004) illustrating variability in f luorescence patterns by land use. Note the increase in humic-l ike and protein-like f luorescence intensity at agricultural si tes and the lack of protein f luorescence at the urban site 193 Figure 7-3 - E E M ' s for: a) E C - 9 , b) E C - 1 4 , c) E C - 1 , d) E C - 4 , e) S A - 1 9 and f) S A - 5 in the Elk Creek and Sa lmon watersheds (August 13, 2004). The impact of point-source agricultural contaminat ion on f luorescence intensity can be seen at E C - 4 (note the difference in scale) . Simi lar f luorescence x patterns were observed at S A - 1 9 and S A - 5 ; however, the influence of dilution by groundwater can be observed at S A - 5 194 Figure 7-4 - Mean f luorescence intensity for: a) protein-like f luorescence and b) humic-l ike and fulvic-like material at each station over the three days of sampl ing ( G W = groundwater) 196 Figure 7-5 - Percentage of contributing a rea under agriculture vs. a) fulvic-like f luorescence, b) T 2 8 o and c) Tyr22o- Note the increased protein-like f luorescence assoc ia ted with HV-18 199 Figure 7-6 - F luorescence intensities for samp les col lected during the Hatzic watershed storm event from March 18-21, 2005 ; 202 Figure 7-7 - T 2 8 o vs . log fecal coliform stratified by land use across the three watersheds. R 2 va lues for agricultural, forested, mixed and urban sites were as fol lows: R 2 = 0.775, 0.004, 0.085 and 0.228, respectively 208 xi LIST OF ABBREVIATIONS AND COMMON TERMS Abbreviation Explanation A n Absorbance at wavelength n A L R Agricultural land reserve C D O M Chromophor ic d isso lved organic matter C F U Colony-forming units D O C Disso lved organic carbon D O M Disso lved organic matter E M M A End-member mixing analys is E x / E m Exci tat ion-emission wavelength pair E A Enumerat ion a rea E E M Exci tat ion-emission matrix F V R D Fraser Val ley Regiona l District G V R D Greater Vancouver Reg iona l District H A C C P Hazard analysis critical control points IR Infrared L F V Lower Fraser Val ley M P N Most probable number N P S Non-point source P A H Polycycl ic aromatic hydrocarbons P C A Principle components analys is rs Spea rman rank correlation coefficient S Spectral s lope S T P S e w a g e treatment plant S U V A Speci f ic ultraviolet absorbance T n Tryptophan f luorescence intensity at wavelength n Tyr n Tyrosine f luorescence intensity at excitation wavelength n UV Ultraviolet UV-V i s Ultraviolet-visible ACKNOWLEDGEMENTS I would like to acknowledge the support, gu idance and contributions of the many people who helped make this project possib le. Firstly, I would like to thank my Superv isor , Dr. Hans Schreier, whose pass ion for sc ience, research, field work and "making a difference" were an inspiration from the start. I would also like to express my gratitude for his cont inued support throughout this process, both f inancial and otherwise. I would also like to acknowledge the contributions of my committee members . Dr. K e n Hall , thank you for a lways making the t ime to d i scuss the finer points of water chemistry and the chal lenges of sample collection and analys is . To Dr. Dan Moore, thank you for initially broadening my perspect ives regarding headwater hydrology and providing thoughtful input regarding the design and results of this study. And finally, thanks to Dr. Judy Isaac-Renton who helped me to span the gap between environmental sc ience and public health and provided immeasurable support for this project as an advisor on all i ssues whether related to medicine or not. I would also like to acknowledge the National Sc ience and Engineer ing Resea rch Counc i l for f inancial support in the form of a Post -Graduate Fel lowship. I am greatly indebted to several residents in the Lower Fraser Val ley, without whom none of the field work for this project could have been completed. This is particularly true in the Hatzic watershed where several famil ies not only put up with monitoring stations on their property for 2-3 years, but a lso spent hours in the field helping to install, maintain (and protect!) our various p ieces of equipment. To the Schroots family, thank you for your watchful eye over our hydrometric station on your property, and for our many enlightening d iscuss ions on the history of the Hatzic watershed. To the W e b b family, my eternal thanks for your help in establ ishing and maintaining the hydrometric station on your property and for your help in keeping it dry during the f loods! I only hope my rainfall records are as accurate as yours. A lso , thank you to the K o k a s k a family for al lowing us to locate a cl imate station on your property, and more importantly, for helping us get it up and down the mountain! I would most likely still be putting it together had it not been for your help. I would also like to thank the F V R D , and in particular, G raham Daneluz, for their early support of this project. To Bruce A d a m s and all of the folks at Jou le Microsys tems, I am grateful for your pers istence, patience and tireless efforts in attempting to implement your technology in a field setting, despite its intended use in much c leaner and safer environments. For the tireless analysis of water samp les and for xiii support in the laboratory I am indebted to many people. Thank you to Caro l Dyck and Karen Fergus for guidance on nutrient analysis and for a lways process ing so many samp les on such short notice. To S e l e n a Shay , thank you for your input and gu idance regarding microbiological techniques. A lso , to Dr. Fred Rose l l , my s incere thanks for your invaluable gu idance regarding spectroscopic analys is , and your humour, which helped to pass those many hours in the lab. Thanks also to G i n a Bestbier and Jenni fer Macdona ld for your input and support on all things related to G I S . To those friends who helped in the field and accepted payment in the form of donuts and sandwiches , thank you. Thanks a lso to the many fr iends and family who provided such great encouragement over the years, your support helped more than you'll ever know. To my parents, you have a lways supported me and encouraged me to pursue my pass ions. It is for this reason that this document even exists. Thank you for your never-ending support and for bel ieving that this would someday be done! Finally, to Heather, what can I say? The last three years have been an incredible journey. Y o u have been an untiring source of support, adv ice, much-needed humour and most importantly, encouragement , when it was needed most. I can't thank you enough, and I look forward to the lifetime of opportunities ahead of us. xiv Chapter 1 1. Introduction 1.1. Background C a n a d a ' s agricultural sector has exper ienced significant growth in recent decades due to technological advances that have al lowed increased crop and l ivestock production on a relatively stable agricultural land base . This sector has also grown substantial ly in British Co lumb ia (BC), generat ing approximately $2.4 billion in revenue in 2004 (BC Ministry of Agriculture and Lands , 2006). S ince 1971, crop production in B C has increased, while total a rea under field crops and vegetables has remained constant or has dec reased . Similarly, l ivestock numbers have increased during this t ime. Between 1971-2001, cattle, poultry and swine numbers in the province increased by 4 3 % , 1 4 0 % and 1 1 1 % respectively, again, with little change in total a rea under l ivestock. Such increases in agricultural intensity across C a n a d a have generated concerns regarding the impact of farming activities on both ecosys tem and human health (Coote and Gregor ich, 2000; C h a m b e r s et al . , 2001 ; Environment C a n a d a , 2001). An increase in crop yields per unit a rea requires more intensive appl icat ions of mineral fertil izers and agr ichemicals, resulting in the accumulat ion of nutrients and pest ic ides in the soil profile. This material is then avai lable for transport to groundwater or nearby surface-water sys tems. Similarly, the growing productivity of l ivestock operations results in local ized surp luses of animal manure which can act as significant sources of nutrients and pathogens 1 . In both c a s e s , the assimilat ive capaci ty of the land base dec reases as intensity increases. The Lower Fraser Val ley (LFV) in southwestern British Co lumb ia is one of the most productive agricultural areas in C a n a d a , and is the most intensively farmed region in B C . In recent years, the intensity of both l ivestock and crop agriculture in this region has increased substantial ly, with the most significant change observed in poultry numbers, which increased by more than 8 0 % between 1991-2001 (Schreier et a l . , 2003). This has resulted in significant nutrient surp luses in watersheds in the region and impairment of surface water and groundwater resources as a result of nutrient and bacterial loading (Berka et al. , 2001 ; Schre ier et al . , 2003 ; Hii, 2004). The Fraser Val ley is a lso experiencing significant urban development and population growth. By 2031 , the population of the Fraser Val ley Regional District 1 In this document, the term "pathogen" is used to refer to organisms (bacteria, protozoa, v i ruses, etc.) that cause d isease in their hosts (humans or animals). Where appropriate, a distinction is made between human and animal pathogens. 1 Chapter 1 is projected to increase by 8 2 % (from 254,229 in 2003 to 462,666), with the greatest growth forecasted to take place in Abbotsford, Miss ion and Chi l l iwack (82%, 9 2 % and 84%, respectively) (Urban Futures, 2005). A s a result of this growth and agricultural intensification, drinking water suppl ies in the L F V are under increasing pressure. Currently, there are approximately 494 public drinking water sys tems within the boundar ies of the Fraser Health Authority, which spans from Delta to Boston Bar (Zubel, 2005). At the end of 2004, the Health Authority c lassi f ied 116 of these sys tems as representing a potentially "high risk" to human health, and 40 of the sys tems in this region were under boil water advisor ies. Nineteen of these advisor ies were issued for sys tems with surface-water sources (Zubel, 2005). 1.2. Drinking water and public health in British Columbia W h e n compared to other regions ac ross C a n a d a , B C has historically had the highest incidence of enteric (waterborne and foodborne) i l lness in C a n a d a ( B C Auditor Genera l , 1999). In fact, between 1980-2004, there were 29 recorded outbreaks of waterborne d i sease in the province ( B C Provincial Health Officer, 2001). The majority of these (93%) were assoc ia ted with rural drinking-water sys tems deriving water from st reams, lakes or reservoirs. S u c h surface water sources supply approximately 7 6 % of the province's drinking water (BC Provincial Health Officer, 2001). The human health risks assoc ia ted with surface-water sources are general ly cons idered higher than for groundwater due to the susceptibil i ty of the former to influence by local land-use activities and meteorological condit ions. In particular, the impact of agricultural activities on surface-water quality in B C has been noted ( B C Auditor Genera l , 1999; Coote and Gregor ich , 2000; Hooda et al . , 2000; B C Provincial Health Officer, 2001). G iven B C ' s dependence on surface sources for drinking water, there is a clear need for effective protection of surface-water sources in B C . In 2003, there were roughly 340 boil-water advisor ies in p lace at one t ime across the province, representing more than 1 0 % of the province's water sys tems (Chr is tensen, 2003). A report by the province's Auditor Genera l noted the importance of source-watershed protection, and conc luded that the Prov ince was not adequately protecting drinking-water sources from human activities ( B C Auditor Genera l , 1999). The need for improved source protection was also noted by the Provincial Health Officer (BC Provincial Health Officer, 2001) and by an independent drinking water review panel (Drinking Water Rev iew Pane l , 2003). A s a result, new legislation and regulations (the Drinking Water Protection Act and Regulation) were enacted in 2003. This legislation 2 Chapter 1 recognizes the need for source-watershed protection and requires water suppl iers to conduct "source-to-tap" assessmen ts to identify contamination risks and to prepare mitigation plans to address these risks. The legislation also requires routine monitoring commensurate to the risk of contamination based on these assessmen ts . 1.3. Research context In the L F V , agricultural activities represent a significant threat to water quality. Severa l studies have a s s e s s e d the influence of agricultural land use on surface-water quality (Schreier et al . , 1999; Be rka et a l . , 2001 ; Schre ier et al . , 2003; Schre ier et al . , 2004; Smith, 2004) and groundwater quality (L iebscher et al . , 1992; W a s s e n a a r , 1995; Zebarth et a l . , 1998; Hii, 2004) in this region. However, few have addressed the combined ecosys tem and human health risks assoc ia ted with agricultural intensification. In other words, there have been few studies to a s s e s s nutrient and bacterial contributions to surface waters over a range of spatial and temporal sca les in this region. Further, little research has been conducted to a s s e s s the influence of agricultural intensity on the timing and degree of contaminant contributions to surface waters. A s noted by Environment C a n a d a (2001) in an assessmen t of threats to surface-water quality, there are several gaps in our current understanding of surface-water contamination by nutrients and waterborne pathogens. In terms of nutrients, these included: 1) availability of monitoring data and 2) the need for data regarding trends and p rocesses at different spatial sca les (i.e., from plot to region). In terms of waterborne pathogens, gaps included: 1) a need to identify basel ine trends in surface-water pathogen concentrat ions to understand the impact of agricultural activities, 2) identification of contamination "hot spots" in agricultural watersheds to support targeted monitoring and 3) improved understanding of environmental inf luences on pathogen dynamics . Simi lar gaps with respect to pathogen loading were identified by the B C Provincial Health Officer in an assessmen t of drinking-water quality across the province (BC Provincial Health Officer, 2001). Finally, Environment C a n a d a (2001) noted several knowledge needs with respect to agricultural contamination of surface waters and groundwaters. Speci f ic to agricultural watersheds, these included: 1) a better understanding of the influence of different agricultural activities on contaminant production and mobil ization, 2) a better understanding of the links between hydrological condit ions and agricultural contaminant dynamics over varying spatial and temporal sca les , and 3) a more effective approach to water-quality risk assessmen t for point-source and non-point-source (NPS) contaminant loading, and for 3 Chapter 1 cumulat ive effects. Similar gaps in terms of contamination p rocesses and monitoring data have a lso been noted by the B C Auditor Genera l (1999), Smith and M c R a e (2000), C h a m b e r s et al . (2001), the B C Provincial Health Officer (2001) and Krewski et al . (2002). This thesis attempts to address many of the gaps descr ibed above by taking a multidisciplinary approach to the issue of surface-water contamination in agricultural watersheds. Recogn iz ing the link between issues of environmental contamination and public health, a col laborative project was initiated with the Institute for Resou rces , Environment and Sustainabil i ty ( IRES) and the B C Centre for D isease Control (BC C D C ) . The purpose of this col laboration was to improve understanding of the link between environmental var iables, land-use activities and potential threats to human and ecosys tem health. Another objective of the thesis was to a s s e s s novel monitoring techniques for the detection of agricultural inf luence to support rapid and accurate detection of agricultural contamination. In order to develop a regional picture of the inf luences of land-use on surface-water quality, research was conducted in three agricultural watersheds in the L F V . The mult i-watershed approach was chosen in order to minimize the effects of si te-specif ic var iables on the trends and correlat ions observed and to identify consistent patterns and mechan isms of agricultural influence across several si tes. Specif ical ly, this thesis a ims to accompl ish the following object ives: 1) Develop an improved understanding of the link between agricultural land-use pract ices and nutrient and bacterial dynamics in sur face waters in agricultural watersheds by: a) Capturing basel ine data for bacterial concentrat ions in surface waters in forested regions of agricultural watersheds in the Lower Fraser Val ley b) A s s e s s i n g bacterial contributions from agricultural activities through a compar ison with basel ine data c) Quantifying the link between nutrient and bacterial dynamics in surface waters d) A s s e s s i n g the link between agricultural intensity and the degree of surface-water impairment e) Gain ing an improved understanding of the role of agricultural activities in cumulat ive impacts to surface-water quality 2) A s s e s s the influence of meteorological condit ions on nutrient and bacterial dynamics in surface waters in agricultural watersheds, by: a) Assess ing the link between meteorological and hydrometric parameters and surface-water quality at different temporal sca les b) Identifying periods of high risk in terms of surface-water impairment by assess ing seasona l trends in nutrient and bacterial cycl ing in surface waters 4 Chapter 1 c) A s s e s s i n g the consis tency of these trends across severa l watersheds to determine if the results are sca lab le to the regional level 3) A s s e s s spect roscopic techniques for the detection of agricultural inf luence on sur face water quality in order to support a more proactive, r isk-based approach to surface-water quality management by: a) A s s e s s i n g the utility of absorbance and f luorescence spect roscopy for detecting agricultural contaminants (nutrients and bacteria) b) A s s e s s i n g the potential for these techniques to provide additional information regarding contaminant sources and mechan isms of mobil ization and transport to surface waters Another objective of this project was to test f ie ld-based spectroscopic sensors as a mechan ism for real-time monitoring of agricultural inf luence. However, as descr ibed in Appendix A , technological and logistical difficulties prevented this evaluat ion. 1 .4. Thesis organization The overall aim of this thesis to identify the land-use activities and environmental condit ions which lead to the greatest risk of nutrient and bacterial contributions to surface waters, and to determine if these results are sca lab le to the L F V and other regions by assess ing trends across multiple watersheds. Further, this thesis provides a preliminary assessmen t of spect roscopic techniques which have significant potential for rapid and sensit ive monitoring of surface-water contaminat ion, particularly in agricultural environments. The remainder of this thesis is divided into 7 chapters. Chapter 2 provides the context for this work by summar iz ing the relevant literature. This includes a review of historical agricultural land-use trends in B C , a d iscuss ion of agricultural contaminants and their sources and mechan isms of transport, an overview of drinking-water quality i ssues in the province and a review of existing water-quality monitoring tools and technologies. This is fol lowed by a descript ion of laboratory and field methods in Chapter 3. Chapters 4 and 5 descr ibe the results of an intensive sampl ing program conducted in the Hatzic, Elk Creek and Sa lmon watersheds from 2002-2005. Chapter 4 focuses on the impact of moderate-intensity agriculture on surface-water quality in the Hatzic watershed, and descr ibes the influence of meteorological and hydrological condit ions on contaminant mobil ization and transport. Chapter 5 compares surface-water impairment across three watersheds with varying levels of agricultural intensity 5 Chapter 1 to a s s e s s the implications of increasing agricultural production on a stable land base in terms of water quality. Chapters 6 and 7 outline the results of the evaluation of absorbance and f luorescence spectroscopic techniques, respectively, for the detection of agricultural influence on surface waters. Chapter 6 descr ibes the potential of absorbance spect roscopy as both a qualitative and quantitative tool that has several advantages over traditional monitoring technologies, including small sample s ize, minimal sample preparation and rapid analys is . Chapter 7 provides a similar assessmen t of f luorescence techniques. Finally, Chapter 8 provides an integrated d iscuss ion regarding the conclus ions and implications of this work and areas for future research. 6 Chapter 2 2. Literature review 2.1. Introduction Farming activities in agriculturally dominated watersheds have the potential to introduce a variety of contaminants into local surface waters. These include nutrients from fertilizers (chemical or manure), animal and human pathogens from livestock wastes , metals, pest ic ides, herbicides and endocr ine (hormone) disrupting subs tances . The potential for sur face water impairment is a function of the type and intensity of agricultural activity, and is directly inf luenced by local geophys ica l , hydrological and meteorological condit ions. The purpose of this chapter is to review the current literature regarding surface water contamination in agricultural watersheds and the potential use of spect roscopic techniques to detect and quantify such contaminat ion. This chapter begins with an overview of trends in agricultural activity in the Lower Fraser Val ley and descr ibes the vulnerability of water resources in the region to agricultural contaminants. The second section descr ibes agricultural contaminants in greater detail and traces their fate from sources on the land surface, through mobil ization and transport p rocesses to their del ivery to, and cycl ing within, surface waters. Current monitoring methodologies and techniques for these contaminants are a lso reviewed, and their strengths and w e a k n e s s e s are d i scussed . The chapter conc ludes with a review of spect roscopic techniques as tools for water-quality assessmen ts . 2.2. Agriculture in the Lower Fraser Valley 2.2.1. Overview The Lower Fraser Val ley (LFV, see Figure 2-1) contains the most productive and intensively farmed agricultural land in British Co lumbia . To understand the value of this region to the province's economy, and the dynamics of land use change in recent years, it is instructive to review statistical data col lected through the Agricultural C e n s u s of C a n a d a , conducted every 5 years (2001 being the most recent). The following review is based on data extracted from the Agricultural C e n s u s of C a n a d a for the L F V from 1986-2001 by Schre ier et al., (2003) for '13 municipalit ies (Langley, R ichmond, Surrey, Chi l l iwack, Delta, Matsqui , Agass i z , Abbotsford, Miss ion, Map le R idge, Pitt Meadows , N icomen and Burnaby). These fall within the greater Fraser Val ley which covers a total of 1.7 million ha, with 7 Chapter 2 approximately 129,210 ha zoned as Agricultural Land Reserve (ALR) . T N ts Hope Burnabv Pitt Meadows Maple J *— Ridge Agassiz Vancouver Nicomen Mission West ' | D e l t a Delta i Richtporid "' X / ' East S. Surrey North Surrey Figure 2-1 - Map of the Lower Fraser Valley with census enumeration areas. Note that Richmond, Langley, Matsqui and Chilliwack include several enumeration areas (East and West Richmond, North, Central and South Langley, North, South and West Matsqui and East and Wast Chilliwack). In 2001 , the agricultural sector in British Co lumb ia accounted for over 30,000 j obs 2 and generated $2 .22 3 billion in gross farm receipts ( G F R ' s ; a measure of total revenue generated by agricultural holdings). Of this total, $1.42 billion (approximately 64%) was generated in the L F V , a region that contains only 3 .5% of the total farmed area in B C . Farm land occup ies 83,308 ha within this region, with approximately 39,700 ha in the Greater Vancouver Reg iona l District ( G V R D ) and 48,700 ha in the Fraser Val ley Reg iona l District (FVRD) . This is less than the a rea of the A L R , as much of the A L R is held for conservat ion, rural residential or recreational uses . A c r o s s the region, 6 2 % of farm land is under crops, while the remainder falls under other u s e s 4 (17%), unimproved pasture (14%) and improved pasture (7%). Compared to the rest of the province, the L F V is unique in a number of ways. Ave rage farm s ize is much smaller, at 16 ha compared to the provincial average of 128 ha [and a national average of 274 ha 2 Including 1,900 jobs in aquaculture. 3 For this analysis census data are provided for illustrative purposes and are rounded for clarity 4 Land under farm buildings, barnyards, lanes, home gardens, g reenhouses , woodlots, sugarbush, bogs, marshes, etc. 8 Chapter 2 (Statistics C a n a d a , 2001)]. The relatively high productivity and agricultural intensity of the region is illustrated by the G F R ' s per hectare, which at nearly $16,200 are almost 20 t imes higher than the provincial average. Per- farm G F R ' s average $259,293 and are more than twice the provincial average ($109,540). Finally, profitability in terms of return on operating expenses was approximately 5 0 % greater for L F V farms (15.5% return) when compared to the provincial average (10.2%). Whi le it is important to note that these average values mask significant variability among municipalit ies in the L F V , the above data illustrate the high productivity, intensity and profitability of agricultural land in the L F V relative to the rest of the province. 2.2.2. Land-use change: 1986-2001 Between 1974-1975, the Agricultural Land Rese rve (ALR) was establ ished in British Co lumb ia to preserve agricultural land for farming purposes and prevent encroachment by urban land uses (Agricultural Land Commiss ion , 2004). A s a result, the total amount of agriculturally zoned land has remained relatively constant at approximately 4.7 million ha (this has changed slightly as a result of applications for rezoning). However, the type and intensity of farming pract ices has evolved significantly in recent years. Signif icant changes in the economic and competit ive landscape in the' agriculture sector have forced the industry to adapt in order to remain competit ive and viable. Free trade agreements have opened new markets for Canad ian products, but have a lso led to increased competit ion from international suppl iers, resulting in significant downward pressure on prices in local markets, particularly for commodity products. This downward pressure, combined with increasing costs of production (labour, property, farm inputs, etc.), has resulted in the amalgamat ion and intensification of farming operat ions to ensure maximum yield per unit a rea (Artemis Agri-Strategy Group, 2001a). A similar trend in decl ining farm numbers and increasing productivity was observed in all Canad ian provinces between 1996-2001 (Statistics C a n a d a , 2001). In the past 20 years, there have been significant changes in agricultural production in the L F V . The total number of farms in the region increased steadily from 1986 to 1996, with a rise in hobby farms (horses and ponies), field crops, g reenhouses and tree fruit and berry operat ions. Between 1996 and 2001 ; however, there was a nearly 1 3 % drop in the number of farms, accounted for mainly by lower 9 Chapter 2 numbers of horse and pony, cattle and swine operat ions while the number of field crop, poultry, dairy and berry operat ions remained relatively constant. A s the total amount of agricultural land has remained relatively stable, this indicates a trend towards amalgamat ion of farming operat ions in the region, primarily driven by a need to increase economies of sca le to remain competit ive in local and international markets (Artemis Agri-Strategy Group, 2001b). There has also been a trend towards higher stocking densit ies (animals per farm) ac ross the L F V . Between 1991-2001, the number of animals per farm in the L F V increased by approximately 2 5 % , 50%, 7 0 % and 7 5 % for cattle, pigs, dairy cows and ch ickens, respectively (Schreier et al . , 2003). This is partially expla ined by decreas ing farm numbers, but is a lso attributable to growth in the total number of animals over this time (particularly ch ickens, as descr ibed below). G r o s s farm receipts have increased from 1986-2001 by 146% (~$577 million to ~$1.42 billion), with total farm expenses keeping pace at 1 4 5 % (~$500 million to ~$1.2 billion). Due to consol idat ion and increasing intensification, the average profit per farm in the L F V has increased markedly from $21,143 in 1986 to over $36,610 in 2001 (adjusted for inflation to 2001 dollars). There has been a commensurate increase in returns per hectare, with va lues growing from $1,341/ha in 1986 to $2,260/ha in 2001 (2001 dollars). Average return per farm and return per hectare in the L F V are 1.6 and 11.9 t imes the provincial averages, respectively. The two most significant changes to the type of agricultural pract ices in the L F V in the past 20 years have been : 1) a major increase in total poultry numbers, and 2) a significant expansion in the total land under greenhouse operat ions. Whi le the numbers of cattle, sheep and pigs in the region have general ly dec reased , and horses and dairy cows have shown moderate increases (10-20%), the number of ch ickens in the region has grown by 124% (from 6.8 million to 15.3 million). Poultry farm numbers during this time have remained relatively constant, resulting in average current stocking densit ies of 11,130 birds per farm vs. 4,870 per farm in 1986. Greenhouse operat ions have also expanded/signif icant ly due to growth in year-round demand for greenhouse vegetables and f lowers both domest ical ly and in the United States. The total a rea under g lass g reenhouses has increased by almost 3 0 0 % from 935 000 m 2 in 1986 to 3.6 million m 2 in 2001. The most significant growth took place between 1996 and 2001 , and has been concentrated in Delta, Langley and Matsqui (which account for almost two thirds of total a rea under g reenhouses in the LFV) . 10 Chapter 2 The changes in agricultural intensity and type descr ibed above have the potential to influence surface water and groundwater quality in the L F V (particularly due to the high nutrient concentrat ions of poultry manure, descr ibed below). Groundwater contamination is of particular concern due to the high susceptibil i ty of local aquifers to non-point source contamination and the rel iance by the local population on groundwater for drinking and residential uses (Liebscher et al . , 1992; B C Provincial Health Officer, 2001 ; Hii, 2004). However, attention has more recently turned to surface water contamination for several reasons. Firstly, in many communit ies, local surface water sources are used as drinking-water suppl ies. A lso , the surface water s t reams in the region provide approximately 6 5 % of the habitat in the L F V for Fraser River C o h o sa lmon, and 8 5 % for chum sa lmon, generat ing concerns that water quality impairment arising from agricultural land uses may impact the health of these fish stocks (Fraser Bas in Counc i l , 2001). Finally, in many locations there is a strong coupl ing between surface water and groundwater sources , meaning that contaminants in sur face waters can affect groundwater sys tems and vice versa ( B e r k a e t a l . , 2001). The changes in land use descr ibed above have resulted in the development of significant nutrient surp luses in the L F V . Schreier et a l . (2003) used nutrient budgets and land-management data from Statist ics C a n a d a ' s C e n s u s of Agriculture to a s s e s s trends in nutrient surp luses in the L F V over time (1991-2001). A compar ison of agricultural inputs to calculated losses and removals illustrated that >65% of census enumerat ion areas in the L F V have nitrogen (N) surpluses of >100 kg-ha" 1-yr" 1, and a similar proportion had phosphorus (P) surp luses of >50 kg-ha" 1-yr" 1. Zebarth et al . (1998) constructed N budgets for typical poultry operat ions and raspberry f ields, which are commonly the receiving environments for poultry wastes. It was noted that not only are the rates of waste production increasing (due to growing numbers of animals), but the inputs of nutrients to the region as a whole are a lso increasing. The latter issue is a result of the marked growth in the poultry industry. A s chicken feed is not grown locally, significant vo lumes are imported annually. B e c a u s e ch icken manure is managed locally, this results in substantial net increase in nutrient stores in the L F V . This is in contrast to cattle feed, a large proportion of which is grown locally. The impact of growing nutrient surp luses is exacerbated by changing cropping patterns. In regions of the L F V where chicken operat ions are common, there has also been a trend towards increasing smal l fruit (raspberries, strawberries, etc.) production. These crops general ly require well-drained soi ls, are planted in wide rows, and have lower nutrient requirements than corn or g rasses (Hii, 11 Chapter 2 2004). A s a result, the exposure of local groundwaters and surface waters to contamination by fertilizer and manure from these operations has increased. 2.3. Susceptibility of water resources to contamination in the LFV This sect ion descr ibes the regional geophys ica l and meteorological character ist ics of the L F V , as they play an important role in surface water and groundwater contaminat ion. 2.3.1. Surficial geology and aquifer vulnerability The L F V is underlain by a complex ser ies of Quaternary deposi ts reaching depths of up to 300 m. Relatively impermeable tills, deposi ted during periods of glacial advance , are interspersed with more permeable glaciofluvial deposi ts from interglacial per iods. Low-permeabi l i ty marine and glaciomarine sediments are observed at depth and throughout the Quaternary sequence in low-lying areas near the coast, as a result of isostatic depress ion during glacial advances (Dakin, 1993). Many aquifers in the region are found within thick sandy or gravelly glaciofluvial deposi ts that extend to the surface, and are thus suscept ib le to influence by agricultural activities. The B C Ministry of Water, Land and Air Protection deve loped a classif icat ion system that ranks aquifers based upon their susceptibil i ty to contamination and the level of local development and demand (Kreye and W e i , 1994). Of the 6 h igh-demand aquifers in the region (those with > 20 domest ic wel ls per square kilometre), 4 are classi f ied as "highly vulnerable." Recent studies have documented the impairment of groundwater resources in this region. W a s s e n a a r (1995) used an analysis of N and oxygen (O) isotopes to a s s e s s the sources , mechan isms and timing of groundwater contamination in the Abbotsford aquifer. In this study, nitrate (N0 3~) va lues in domest ic wells exceeded 150 mg-L" 1 ; significantly higher than the 45 mg-L" 1 N 0 3 " drinking water quality guidel ine recommended by Health C a n a d a . In the s a m e study, an analysis of tritium (a radioactive isotope of hydrogen) indicated that groundwater in the upper aquifer was less than 10 years old, indicating a rapid recharge rate. Using the distinctive isotopic signatures of poultry manure, chemica l fertilizers and septic wastes , it was determined that poultry manure was the primary source of N 0 3 " - N to the aquifer. An analysis of oxygen isotopes ( 5 1 8 0 ) also indicated that nitrification took place soon after the application of manure in the summer months, but that the subsequent transport of N 0 3 " to the aquifer was accompl ished during heavier 12 Chapter 2 autumn rains (based on the 5 1 s O signatures of summer vs. fall rains). This supports the hypothesis that there is a direct and rapid link between land management activities and groundwater quality in the region. Using similar techniques, Mitchell et a l . (2003) attributed groundwater contamination in deep wells in northern Wash ington State to poultry operat ions in the L F V , while contamination of shal lower wells was assoc ia ted with local agricultural activit ies. In another assessmen t of N 0 3 " contamination in the Abbotsford aquifer, Zebarth et al . , (1998) monitored 35 p iezometers and observed that N 0 3 " concentrat ions at 15 of these sites never dropped below the Health C a n a d a guidel ine of 10 mg-L" 1 N 0 3 " - N , while only 12 sites had more than half of the samples under the guidel ine. Hii et a l . (2004) conducted a review of groundwater contamination in several L F V aquifers that serve approximately 120,000 residents (including Chi l l iwack, Abbotsford and Matsqui). They noted total N (nitrate + nitrite) levels of over 30 mg-L" 1 , a long with guideline exceedences for several metals and organic contaminants. 2.3.2. Climate Regiona l precipitation patterns play a significant role in the dynamics of water quality, particularly for surface waters. Precipitation in the L F V is highly seasona l (Figure 2-2), with the majority falling during the late autumn and winter months. It is during this time that contaminants on the land surface are suscept ib le to mobil ization and transport to surface st reams and aquifers. 13 Chapter 2 300 Figure 2-2- Average monthly rainfall at Abbotsford airport for the period 1971-2000 (Environment Canada, 2005) Many studies have noted the role of this seasona l cl imatic regime on water quality in the L F V (Hall et a l . , 1991; Schre ier et al . , 1999; e.g., Berka et al. , 2001 ; Smith, 2004; Macdona ld , 2005). Hall et al. (1991), in a review of water quality data throughout the Fraser River Bas in , observed a dilution effect during winter months for major cat ions. However, Berka et al . , (2001) and Smith (2004) noted that concentrat ions of N 0 3 " and orthophosphate (P0 4 3 ~) and bacter ia (total and fecal coliform) during wet season sampl ing in the S u m a s River watershed were consistently higher than dry s e a s o n va lues. A similar trend was observed in the Elk Creek watershed by Schre ier et al . (2004) for nutrients in surface waters, with peaks in concentration general ly occurring in October or November , the time during which the first heavy rains of the season mobil ize surface nutrient sources . Heavy rains throughout the winter months resulted in cont inued loading to surface waters; however, a seasona l hysteresis was evident as nutrient sources were depleted over time (local regulations prohibit application of manure to agricultural f ields once heavy rains begin in autumn). S h a w and Tuominen (1998) also descr ibed seasona l trends in surface water fecal coliform levels. During summer months, bacterial concentrat ions showed a marked dec rease at several sites a long the Fraser River from Miss ion to the Fraser River Estuary. This seasona l depress ion in coliform numbers was 14 Chapter 2 attributed to lower rainfall and improved chlorination of effluent from sewage treatment plants. A s descr ibed below, the influence of rainfall on water quality depends largely on site-specif ic factors and p rocesses (contaminant type, sources, transport pathways, groundwater and surface water interactions, etc.), and it is therefore difficult to descr ibe this inf luence in general terms in the L F V . 2.4. Nutrients This section provides an overview of nutrient sources in agricultural operat ions and descr ibes the p rocesses governing their mobil ization and transport to surface waters. This is fol lowed by a d iscuss ion of the ecological and human health risks assoc ia ted with nutrients. The focus of this sect ion is on N and P, due to their relative abundance (arising from human inputs), and their impact on biological productivity and human and ecosys tem health. 2.4.1. Nutrient sources in agricultural watersheds Nutrients are added to agricultural fields in the form of chemica l fertilizer, manure, sewage s ludge and industrial waste products in order to replace elements essent ia l for plant growth and ensure maximum yield for a given crop. These e lements include N, P, potassium (K), ca lc ium (Ca), magnes ium (Mg) and sulphur (S) and some metals, including iron (Fe), manganese (Mn), z inc (Zn), boron (B), chlorine (CI), cobalt (Co) and copper (Cu). In agricultural areas, manure is a lso appl ied to fields as a means of d isposal in regions with high-intensity l ivestock operations. In the L F V , a moderate market exists for sel l ing these wastes as fertilizer to nearby regions that are in overall nutrient deficit. S ince the early 1990's, the Susta inable Poultry Farming Group has been transporting poultry manure to nutrient-poor regions, initially to Merritt, and more recently to Delta (Schreier et a l . , 2003). However, this is not feasible for all operators as the cost effect iveness of this option depends on manure type (nutrient concentration per unit weight), operation s ize (and the potential for economies of scale) , availability of process ing facilities and transport distance (Wohl, 1996). In British Co lumbia , manure is commonly used as a component of the annual fertilization process. For field crops, manure can account for up to 7 5 % of a crop's total N requirements (BC Ministry of Agriculture Food and Fisher ies, 2004). The nutrient and moisture contents of different types of manure are outl ined in Tab le 2-1. Whi le actual va lues vary depending on feeding strategies, storage techniques and the age of the manure, the data illustrate two important characterist ics of these materials: 1) their 15 Chapter 2 high moisture content (which makes transport out of the region financially impractical) and 2) the substantial ly higher nutrient content of poultry manure relative to other l ivestock. This is particularly relevant in the L F V where poultry production has undergone significant expans ion in recent years, as descr ibed above. Table 2-1 - Typical moisture content and nutrient content of various manures on a wet-weight basis (modified from BC Ministry of Agriculture, Food and Forests, 2004) Manure Moisture Content Total N Content Total P 20 5 Total K20 (%) (kg/t) Content (kg/t) Content (kg/t) Beef (solid) 68 4.2 4.8 8.2 Dairy (solid) 77 3.9 3.4 9.0 Dairy (liquid) 91 2.9 2.1 4.5 Swine (covered) 93 6.3 3.3 3.9 Poultry (broiler) 25 31.6 22.8 12.2 Poultry (layer) 35 22.8 29.2 11.2 Manure management pract ices vary by region in B C ; however, provincial statistics provide a good indication of how they may contribute to surface water and groundwater contaminat ion, particularly when considered in the context of other Canad ian ' provinces. In B C , most l ivestock operat ions (79.7%) apply excess manure to local f ields during the spring, just prior to the growing season , to ensure uptake rates are maximized (Beaul ieu, 2004). However, nearly 2 5 % of l ivestock farms apply manure during the late fall and winter, when rates of uptake are lowest and total precipitation and precipitation intensities are highest (Beaul ieu, 2004). A review of manure storage pract ices across C a n a d a (Statistics C a n a d a , 2003) revealed that B C had the highest percentage of farms (43.2%) with no manure storage facilities (liquid or solid). This was 1 9 % higher than the national average. The timing of manure incorporation into soi ls a lso inf luences the degree to which nutrients (and pathogens) will be avai lable for transport to nearby sur face waters. In B C , 6 8 . 7 % of farms (representing 55 .4% of all manure produced in the province) de layed incorporation of manure into the soil for longer than 7 days , compared to the national average of 52 .4% (Beaul ieu, 2004). A s demonstrated by Muel ler et al. (1984), early incorporation can result in more than a 20-fold dec rease in the amount of nutrients exported from agricultural f ields. 16 Chapter 2 Commerc ia l fertilizers are a lso used to augment field nutrient levels over and above manure inputs. In 2001 , 61 .6% of B C field crop producers appl ied commerc ia l fertil izers, compared to the national average of 7 4 . 5 % (Korol, 2003). Th is lower value is likely expla ined by the fact that much of the province's crop production is located in the L F V where the significant nutrient surplus from livestock operat ions reduces the need for commerc ia l nutrient inputs. Aerial deposit ion is a lso an important source of nutrient accumulat ion. Whi le losses to the atmosphere do occur, atmospher ic interactions result in a net increase of N in soi ls. Brisbin (1995) deve loped a nutrient ba lance model for the L F V and identified two components of atmospher ic nitrogen input: 1) background deposit ion and 2) return flow, which was a s s u m e d to be 3 0 % of the volati l ized nitrogen (not including denitrification) resulting from manure management pract ices. In this model , atmospher ic exchange of P and potassium (K) were considered negligible. Nitrogen returns as a result of atmospher ic deposit ion in the L F V are considerable, as noted by Schre ier et al . (1999), who est imated average wet and dry N H 3 deposit ion in the region of up to 9.5 kg N-ha" 1 ,yr" 1 . 2.4.2. Transport processes The mobil ization and transport of nutrients from agricultural f ields to surface waters and groundwaters are complex p rocesses . This is because nutrients can occur in many chemica l forms (e.g., nitrogen occurs as nitrate (N0 3~), nitrite (N0 2 " ) , ammon ia (NH 3 ) and ammonium (NH 4 + ) among many others) and in different physical states (dissolved, adsorbed or in gaseous form). E a c h chemica l form and physical state is governed by different mobil ization and transport p rocesses . Further, several transport pathways and mechan isms exist between agricultural fields and surface waters, with each controlled by site-specif ic geophysica l factors and responding differently to environmental condit ions. It should be noted that there is a significant body of literature in the field of landscape ecology that addresses the mechan isms through which land use inf luences surface-water quality (a comprehens ive review is provided by Baker , 2005b). This sect ion reviews many of these p rocesses , with emphas is on the roles of land-management activities, hydrological var iables and cl imate. 2.4.2.1. Nitrogen Before descr ibing the details of N transport, it is necessary to briefly review the N cycle (Figure 2-3). From a functional perspect ive, the N cyc le involves the transformation of unreactive N gas (N 2) to biologically access ib le forms (through fixation, mineralization and nitrification) and then back to N 2 17 Chapter 2 (Chambers et a l . , 2001). The literature on this topic is extensive, and for detai ls regarding the entire N cycle, the reader is referred to Pierzynski et al . (1994) and C h a m b e r s et al. (2001). a) Denitrification Nitrification II N 0 2 |jj Nitrogen fixation (N gas to ammonia) N 2 + 8H* + 8e" + 1 6 A T P - » 2 N H 3 + H 2 + 16ADP + 16Pi (i) Soil org. N Ammonifi cation Nitrification - Step 1 (ammonium to nitrite) NH„* + 1.50 2 -> N0 2 - + H 2 0 + 2 H + Nitrification - Step 2 (nitrite to nitrate) NO 2 - + 0 . 5 O 2 - > N O 3 -00 (iii) •/Nitrification I Figure 2-3 - a) simplified nitrogen cycle showing critical processes, and b) key chemical reactions in the nitrogen cycle (note: the equations for ammonification and immobilization are omitted for clarity as they are multi-step processes). Modified from Pierzynski et al. (1994). Nitrogen is present in the environment in both organic and inorganic forms. Organic N is def ined as N that is bound to carbon (shown in Figure 2-3 as b iomass N and soil organic N). It is der ived from biological p rocesses and can be found in both soluble and particulate forms. Part iculate forms are compr ised of living and dead organisms, while soluble organic N is derived from human and animal waste, or the breakdown of the particulate form. In its organic form, N must go through the process of mineralization (transformation to N H 4 + or N0 3 " ) before it can be assimi lated by plants. Inorganic N (somet imes referred to as mineral N) is der ived from microbial p rocesses that convert N 2 and organic N into several different forms which can be assimi lated by plants. From a water-quality perspect ive, the inorganic forms of N of greatest interest are N 0 3 \ N0 2 ~, N H 3 and N H 4 + . Figure 2-3 illustrates the key chemical reactions involved in the convers ion of N between these forms. In particular, N H 4 + and N 0 3 " are of primary concern in surface waters and groundwaters due to their solubility and stability in solution. Nitrite is often converted to N 0 3 " , and N H 3 is converted to ionized N H 4 + . Ammon ium predominates at pH levels below approximately 8.75, while N H 3 is more common under more bas ic condit ions (Pierzynski et a l . , 1994). The application of N to agricultural f ields in the form of chemica l and organic fertilizers initially leads to a surplus in the upper soil profile once the nutrient requirements of crops are exceeded . 18 Chapter 2 Ammon ium and N 0 3 " are then avai lable for transport either laterally v ia over land flow or throughflow, v ia artificial subsurface drainage networks, or downward through leaching into local aquifers. The mobility of these two spec ies of N is governed by their specif ic chemica l properties, local geophysica l and meteorological parameters and by the type of land management activities employed. Due to it's positive charge, N H 4 + has a higher propensity for retention in the upper soil profile through cation exchange mechan isms than N 0 3 " (Hooda et a l . , 2000). A s a result, a large portion of N is lost through leaching and transported via subsurface flow as N 0 3 " . Nitrogen that is present in excess of crop requirements is subject to biological denitrification (conversion to N 2 gas). The degree to which this occurs is a function of the organic carbon content in the soi l , as this acts as an energy source for denitrifying bacter ia (Chambers et al . , 2001). A s N 0 3 " migrates downward through the soil profile, ongoing biological dissimilat ion is poss ib le ; however, such p rocesses are often restricted at depth due to a lack of organic carbon to act as an energy source. A s a result, N O V accumulat ion in groundwater aquifers is common where a cont inuous source is avai lable. The permeabil i ty of the soil profile and underlying material determine the sub-sur face mobility of N 0 3 " . A s would be expected, and has been demonstrated previously (e.g., Bergstrom and J o h a n s s o n , 1991; e.g., Berka et a l . , 2001 ; C h a e et al . , 2004), wel l-drained soi ls favour N mobility, while fine-textured soils promote lateral flow to nearby surface waters. Bergstrom and J o h a n s s o n (1991), in a control led lysimeter study, noted a three-fold increase in N losses through leaching for sandy soi ls , as compared to clay-rich soi ls. It should be noted; however, that sandy soils rich in organic matter showed minimal losses s to leaching due to denitrification. Cl imate and hydrology strongly influence N movement in watersheds. A review of the literature reveals the complexity of the relationships between precipitation, watershed hydrology and N transport. Van Herpe and Troch (2000) outl ined the role of seasona l hydrological condit ions, precipitation and land use on stream water N 0 3 " concentrat ions for several sub-catchments in the Zwa lm watershed in Belg ium. In general , they noted a positive correlation between d ischarge and N 0 3 " concentrat ion; however, this relationship was much stronger during the winter months (in summer , N 0 3 " peaks were more delayed). This was attributed to differing seasona l hydrological states, with the dry summer months resulting in hydrologically d isconnected groundwater stores, in contrast to the wet winter months during which these sources were strongly coup led, leading to faster hydrological response. A s a result, s t ream flow during summer storms was initially compr ised of relatively N 0 3 " -poor over land flow. Subsur face sources that 19 Chapter 2 were richer in N 0 3 " responded more slowly, leading to a later peak in N 0 3 " concentrat ions when compared to winter storms. It should also be noted that N avai lable for transport during summer months is general ly lower due to greater biological uptake assoc ia ted with the growing s e a s o n . Similarly, in catchments with wel l-drained soi ls, a negative relationship between precipitation and N concentrat ions may be observed (e.g., Schre ier et a l . , 1999; K e m p and Dodds, 2001). In such situations, s t reams that are strongly l inked to contaminated groundwater sources will show much higher N 0 3 " concentrat ions during basef low condit ions, with dilution by over land flow taking place during storm events (Smith, 2004). The role of land-management activities in N leaching is a lso significant. In early studies of N dynamics on experimental plots, Rolston and Broadbent (1977), Rolston et al . (1978) and Rolston et al . (1979) [as reviewed by Sharp ley et al . (1998)], deve loped winter and summer N budgets for experimental plots under three condit ions: 1) manure application, 2) rye-grass cropping and 3) a control treatment (no crop or manure applied). During the summer , they demonstrated that manure addit ions resulted in the denitrification of 7 9 % of total N (due to high biological productivity arising from ample organic carbon and warm temperatures), while plots under crops lost 6 6 % to leaching. Uncropped control plots exper ienced the greatest loss to leaching (87%) due to a lack of uptake and denitrif ication. Winter data showed a marked difference, with manured fields losing 7 7 % of total N (due to reduced microbial activity under cooler, wetter condit ions). L o s s e s to leaching in the winter months were lowest for the cropped plot (39%), as the coo l -season grass that was planted thrived in the lower-temperatures, resulting in greater metabol ism of avai lable N. Under winter condit ions, leaching in the uncropped plot was the dominant p rocess , with a loss of 9 9 % of total N. These early studies demonstrate the fundamental dynamics of N on agricultural f ields and the influence of precipitation, temperature, organic matter content and active crops on the fate of avai lable N. It is important to note; however, that f ie ld-based studies do not a lways show consistent results, due primarily to the strong influence of si te-specif ic condit ions on these p rocesses (Chambers et a l . , 2001). The spec ies of N observed in surface waters vary depending upon the location of the sampl ing site and site-specif ic condit ions. In a comprehens ive study of nutrient f luxes in the UK , Russe l l et al. (1998) a s s e s s e d the speciat ion of N at the outlet of four watersheds with varying degrees of cropped land and pasture over a complete water year. It was noted that only 8 % of total exported N was in particulate form (the majority of which was particulate organic N). Total d isso lved N formed the largest component of 20 Chapter 2 total N loads (76-82%) with N O V and N 0 2 " dominat ing. Disso lved organic N compr ised 13 -16% of annual N exports, while N H / - N made up only 0.3-1.2% of the total. This is in contrast to a study by Heathwaite and Johnes (1996) which noted that over 9 0 % of surface runoff from pasture was in the form of N H 4 + . The observed differences are likely due to the sca les represented by each study. S a m p l e s from the former study represent integrated catchment-scale N loading, and therefore include N inputs from several pathways, including over land flow, throughflow and contributions from groundwater. The latter study observed high N H 4 + concentrat ions in over land flow draining directly from heavily grazed lands. A lso , the timing of sampl ing relative to grazing or manure or fertilizer application can influence the relative proportion of N appear ing as N H 4 + vs . N 0 3 ' . Th is was illustrated by Pierson et al . (2001) and Smith et a l . (2003) for ch icken and cattle wastes, respectively. Both studies showed relatively high N H 4 + concentrat ions immediately after application of NH 4 + - r i ch manure, as there had been little time for convers ion to N 0 3 " v ia nitrification. Kurz et al. (2005) noted similar trends after the application of N fertilizer to three study sites in Ireland. Yet another factor, subsur face drainage networks, may also strongly inf luence N 0 3 " mobility within a catchment. Randal l and Mul la (2001) provide an overview of several studies assess ing links between precipitation and N 0 3 " content in tile dra inage. B e c a u s e an artificial drainage network essential ly acts as a direct link to nutrient stores within the soil profile, response to rainfall events is general ly rapid, with concentrat ion being a function of drainage vo lume and N availability. The above examp les illustrate a variety of p rocesses involved in nutrient mobility and transport in agricultural watersheds. Further, they highlight the need to consider si te-specif ic var iables when identifying and quantifying nutrient sources and pathways in such watersheds and when making inferences regarding the mechan ics of nutrient transport based on water-quality data. 2.4.2.2. Phosphorus Figure 2-4 illustrates the soil P cyc le, including primary P sources, interactions between inorganic and organic P, and routes by which P is lost from the soil profile. Al though concentrat ions of P in surface waters and groundwaters tend to be lower than those of N, P is commonly cons idered a greater threat to water quality because it is often the limiting nutrient for aquatic biological activity in fresh waters (Wetzel , 1983; Bechmann et a l . , 2005). Like N, P is found in both particulate and d isso lved phases . The particulate form includes plant and animal t issue, P precipitates and P adsorbed to sediments, while the 21 Chapter 2 dissolved form includes inorganic P (or P 0 4 3 " ) and organical ly bound phosphates . Operat ional ly, most studies define particulate P as that which does not pass through a 0.45 um filter. Figure 2-4 - Simplified phosphorus cycle, illustrating sources and pathways by which phosphorus is lost from the soil profile. Modified from Sutton (1997). The p rocesses governing P transport from agricultural soi ls to surface waters and groundwater differ from those assoc ia ted with N. In contrast to N, P is strongly adsorbed in most soi ls, and thus has a strong tendency toward transport in particulate form. This suggests that groundwater and surface water contamination with soluble P is minimal; however, the amount of soluble P in a soil is inversely related to P sorption capacity and strength, both of which are variable. Sorpt ion capaci ty and strength are a function of soil texture (finer grained material has a higher sorption capacity), soil chemistry (P is strongly bound to ca lc ium, a luminum, iron and manganese) and current P saturation (which is strongly affected by the history of P application at a given site). T h e s e characterist ics can change over time as a result of land management pract ices. For example , Labosk i and L a m b (2004) observed a significant decrease in soil sorption capacity after manure appl icat ions in five of seven soil types in experimental plots. Further, a strong negative correlation between sorption strength and P saturation was observed, indicating that surplus P appl icat ions can increase soluble P availability by s imul taneously reducing soil capaci ty and binding strength. There are differing opinions in the literature regarding the dominant form of P transport from agricultural fields to surface waters. Hooda et al . (2000), in a review of research on water quality concerns related to agriculture, noted that P transport is dominated by the particulate fraction. Similarly, Hart et al. (2004) reviewed several studies of P runoff from agricultural land in N e w Zea land where this trend was observed, with particulate P account ing for 6 2 - 9 1 % of total P loads. However, in the s a m e 22 Chapter 2 review, it was rioted that numerous studies descr ibe the opposite trend, with the d isso lved form (primarily organical ly bound P) dominating P export. This difference was attributed to variations in site-specif ic factors that affect the availability and mobil ization of different forms of P (described above), the sca le of the monitoring programs (i.e., plot, sub-catchment or catchment scale) and the timing of water sampl ing relative to P addit ions and rainfall events (Hart et a l . , 2004; Haygarth et a l . , 2005). The spatial sca le of sampl ing is an important var iable as the ratio of soluble to particulate P observed in a water sample is a function of the original ratio (as it is transported from the site), and the t ime between mobil ization and sampl ing, as P adsorption can take place in transit. A s an example, C o o k e (1988, in Hart et a l . (2004)) noted that soluble P dominated in surface runoff (comprising 6 3 % of total P) from a 16 ha catchment in New Zea land , but more than 8 5 % of P exported from the catchment via s t reams was in particulate form. In-stream processes play a major role in the ba lance between soluble and particulate P phases in surface waters, and under wel l -mixed condit ions an equil ibrium between the two forms may exist during basef low (McDowel l et al . , 2003). The equil ibrium P concentrat ion ( E P C ) of sediments in suspens ion determines the mobility of P between the two fractions (with a high E P C relative to stream P concentrat ions leading to desorpt ion, and vice versa). McDowel l et al . (2003) noted a significant positive correlation between E P C and stream soluble P concentrat ions during basef low condit ions, but not during storm flow as this equil ibrium cannot be maintained under condit ions of rapid particulate and soluble P loading. Interestingly, in a study of the influence of stream sediments on in-stream P cycling in a watershed inf luenced by sewage treatment plant (STP) d ischarges, Jarv ie et al . (2005) noted that bed sediments acted alternately as a net sink or source of P, depending on the soluble P concentrat ion of stream waters. During per iods of S T P effluent d ischarge, the net flux of soluble P was from the water co lumn to bed sediments. During periods of low effluent d ischarge, the opposite was true due to the reversal in the diffusion gradient at the sediment boundary layer. The availability of soluble P in a saturated soil profile may also result in downward mobility of P with the potential to reach subsur face drainage networks and groundwater sources . The presence of. artificial or natural subsur face drainage networks facilitates the delivery of soluble P to surface waters and groundwater due to shorter contact time between soluble P and soil particles (Sharpley et al . , 1995; Hooda et a l . , 2000; Lazzarotto et al . , 2005). S i m s et al. (1998) conducted a comprehens ive review of research into P leaching in agricultural drainage sys tems, and noted that under certain condit ions, this 23 Chapter 2 source, general ly thought to be negligible, can actually result in significant P loss relative to' over land transport. T h e s e studies indicate the influence of antecedent soil condit ions, spatial sca le and transport pathways on P availability and transport. It is clear that the risk of P loading to surface waters and groundwaters is exacerbated by intensive agricultural activity due to the dec reased sorption capaci ty of agricultural soi ls and the presence of sub-sur face drainage networks which directly link P sources to receiving waters. The resulting impacts to ecosys tem health can be significant, as descr ibed below. 2.4.3. Ecological and human health impacts 2.4.3.1. Ecological impacts The ecological impacts assoc ia ted with nutrient enrichment and eutrophication of sur face waters have been reported extensively and thus will not be covered in detail here. The primary issues of concern are the promotion of excess algal and plant growth leading to dec reased d isso lved oxygen levels (resulting from decomposit ion), dec reased photic zone depth in both rivers and lakes, impairment of spawning beds and reduction in aquatic spec ies diversity (Chambers et al . , 2001). Apart from ecological impacts, such impairment of water quality a lso limits the use of aquatic resources for drinking water, commerc ia l and sport f ishing and recreational purposes (Haygarth et a l . , 2005). For a comprehens ive review of the mechan isms and ecological impacts related to eutrophication in aquatic environments, the reader is referred to Wetze l (1983). For a review of this issue from a Canad ian perspect ive, Coote and Gregor ich (2000) and Chambers et al . (2001) are recommended. 2.4.3.2. Human health impacts Both N and P are essent ia l e lements for living organ isms; however, in certain forms and in high concentrat ions both can be harmful to human health. Phosphorus , in forms commonly found in the environment, is not directly toxic (Chambers et al . , 2001). However, excess ive P in aquatic sys tems can lead to b looms of toxin-producing algae. For example, Microcystis, a blue-green algae, is known to produce hepatotoxins (which affect the liver), while two genera {Anabaena and Aphanizomenon) are known sources of neurotoxins which can c a u s e significant neurological impairment or death in humans and animals (Chambers et al. , 2001). In marine environments, Pfiesteria piscicida has been l inked to large-scale fish kills (Burkholder et al . , 1992). Whi le b looms of these organ isms are not controlled 24 Chapter 2 exclusively by P concentrat ions, P is often the limiting element for growth, and therefore strongly inf luences algal production (Wetzel, 1983). Nitrogen also contributes indirectly to human and animal health effects by promoting algal production, but in contrast to P, there are several forms of N that are thought to be directly toxic to living organisms. The exact role of these forms of N in caus ing human i l lness is unclear. Historically, high N 0 3 " concentrat ions in drinking water have been assoc ia ted with methemoglobinemia, or "blue-baby syndrome" (Coote and Gregor ich , 2000; Knobe loch et al . , 2000; C h a m b e r s et al . , 2001 ; Environment C a n a d a , 2001). Infants under 3 months of age have traditionally been cons idered at greatest risk for this condit ion which is traditionally thought to occur when N 0 3 " is converted to N 0 2 " by bacter ia in the s tomach. This convers ion leads to oxidation of ferrous iron to ferric iron in hemoglobin and results in reduced oxygen carrying capaci ty in the b lood. However, in a recent review, Fewtrell (2004) quest ioned the role of nitrates as a causat ive factor s ince a c lear exposure- response relationship has not been identified. He went on to state that although nitrates may play a role in the onset of methemoglobinemia (along with other factors), attributing this condit ion to drinking-water N 0 3 " levels is inappropriate given the current level of ev idence. A similar conclus ion was reached by Avery (1999), who stated that severa l factors, such as diarrhea and gastrointestinal and urinary tract infections, can directly cause methemoglobinemia in infants without exposure to nitrates in drinking water. A number of studies have investigated the possib le causa l link between high drinking-water N 0 3 " levels and var ious forms of cancer . This link is suspected because once ingested, N 0 3 " is reduced to N 0 2 " , which can interact with compounds in the s tomach to produce N-ni trosoamines and N-ni trosoamides, both of which are among the strongest known carc inogens (Cantor, 1997). This hypothesis is supported by experimental studies, but epidemiological ev idence linking drinking-water N 0 3 " to cancer is mixed. Severa l authors report a link between nitrates and different types of cancer, including, for example, urothelial cancer (Volkmer et a l . , 2005), non-Hodgkin 's l ymphoma (Ward et al . , 1996), brain cancer (Ward et al . , 2000), bladder cancer (Weyer et al . , 2001) and gastric cancer (Xu et al . , 1992). In contrast, there are several studies which report no relationship between the inc idence of different forms of cancer and drinking-water nitrates (e.g., De Roos et al . , 2003 ; Mens inga et a l . , 2003; Ward et al . , 2003; C o s s et al . , 2004). Nitrates in drinking water have been implicated in other condit ions, including hyperthyroidism (Seffner, 1995) and insul in-dependent d iabetes (Kostraba et al . , 1992). In addit ion, the U S Centers for 25 Chapter 2 Disease Control and Prevent ion tentatively l inked high N 0 3 " concentrat ions to spontaneous abortions in LaGrange County, Indiana (Centers for D i sease Control and Prevent ion, 1996). Nitrates may also have a negative impact on animal health. A recent review of endocr ine disruption in invertebrates suggests that nitrates may play a key role in this p rocess in wildlife populat ions (Guillette and Edwards , 2005). The inconsistent f indings regarding the role of nitrates (and their derivatives) in human and animal i l lness make interpretation of cause-effect relationships difficult. Exper imental data suggest that a link to d i sease does exist; however, the retrospective and/or ecological (i.e., populat ion-based) design of many studies limits the ability to control for confounding factors and clearly establ ish that link (Cantor, 1997). The only way to reliably establ ish such a link would be through controlled dose- response exper iments in an imals in a laboratory setting. 2.5. Pathogens This sect ion provides an overview of waterborne human and animal pathogens of concern in North Amer ican agricultural watersheds. It first reviews var ious types of common pathogens (bacteria, protozoa and viruses), and then descr ibes their sources , assoc ia ted mobil ization and transport p rocesses and conc ludes with a d iscuss ion of the potential impacts of these organ isms on human and ecosys tem health. 2.5.1. Pathogens of concern While a wide variety of bacteria, protozoa and viruses are shed from humans and domest ic and wild animal populat ions, relatively few are known to cause waterborne d i sease outbreaks (Rosen , 2000). The organisms reviewed below are currently regarded as those posing a significant risk to human health in agricultural environments. Background on each type of organism is provided and is fol lowed by descript ions of specif ic pathogens and their assoc ia ted health impacts. 2.5.1.1. Bacteria Bacter ia are predominantly s ingle-cel led organ isms of various shapes , including spherical (coccus), rod-shaped (bacil lus), comma-shaped (vibrio), spiral (spirillum) or corkscrew-shaped (spirochete), and are general ly 0.5 to 5.0 um in s ize (Rosen , 2000). They are found in almost every environment on earth; however, those found in the intestinal sys tems of an imals and humans (enterobacteria) are of primary concern. Whi le there are numerous waterborne bacterial pathogens, those 26 Chapter 2 of greatest interest in agricultural watersheds with l ivestock are enterohemorrhagic Escherichia coli, Campylobacter, Salmonella and Yersinia (Ferguson et a l . , 2003). E. coli 0157:H7 Escherichia coli (E. coli) is a member of the coliform group of bacter ia and is commonly found in the intestines of animals and humans. There are over 100 strains of E. coli (most of which are non-pathogenic and serve useful functions in the gastrointestinal tract), and it is often used as an indicator of fecal contamination in water. E. coli 0157:H7 is an enterohaemorrhagic strain of E. coli (intestinal bleeding is commonly assoc ia ted with infection) and is highly infectious, with as few as 10 cel ls caus ing human i l lness (Rosen , 2000). The cel ls produce toxins (Verotoxin, Shiga- l ike toxin) that may impair kidney function (hemolytic uremic syndrome, or H U S ) , cause kidney failure and/or break down intestinal lining. These toxins pose the greatest risk to the young, elderly and immunocompromised (Szewzyk et al . , 2000). A s a result, the incidence of H U S assoc ia ted with E. coli outbreaks varies depending on the population involved (e.g., community outbreak vs. a nursing home outbreak), and can range from 0 -15% (Food and Drug Administrat ion, 2005). The mortality rate for H U S has been est imated at 3 -5% (Boyce et al . , 1995), but can be as high as 5 0 % among the elderly when observed with fever and neurologic symptoms, a condit ion known as thrombocytopenic purpura (Food and Drug Administrat ion, 2005). E. coli 0157:H7 was one of the primary pathogens involved in the Walker ton, Ontario outbreak in May, 2000, during which over 2,300 people fell seriously ill, and 7 people died as a result of contamination of a municipal water system with agricultural runoff (Hrudey et a l . , 2003). Campylobacter Severa l spec ies of Campylobacter can c a u s e i l lness in humans; however, Campylobacter jejuni is the primary cause of all d iagnosed c a s e s (Rosen, 2000). The most common sources of Campylobacter in agricultural watersheds include wild bird and poultry populat ions, cattle, pigs, dogs and cats. The organism can be spread v ia surface water and groundwater (Szewzyk et al . , 2000), and can be foodborne or spread person-to-person. A s few as 400-500 organisms may be sufficient to produce i l lness in humans, with onset of i l lness (campylobacteriosis) general ly occurr ing 2-5 days after exposure (Food and Drug Administrat ion, 2005). Compl icat ions from campylobacter ios is are rare and the fatality rate is low (0.1%, or 1 in 1,000 cases ) , but kidney failure, H U S , diarrhea (often bloody), reactive arthritis or 27 Chapter 2 infections in any major organ are possib le in suscept ib le populat ions (the immunocompromised or patients with a ser ious pre-existing condition) (Food and Drug Administrat ion, 2005). Livestock operations are reservoirs for Campylobacter and play a significant role in their d ispersal in the environment (Stanley and J o n e s , 2003). Campylobacter jejuni is thought to have played a role in the Walker ton outbreak descr ibed above (Hrudey et al . , 2003). A lso , in European countr ies, it is considered a significant threat to drinking water quality, having been responsible for the majority of outbreaks in private water sys tems in Eng land and W a l e s in the early-mid 1990's (Szewzyk et al . , 2000). Salmonella Salmonella spec ies are a well recognized c a u s e of enteric i l lness in humans, with up to 4 million infections occurr ing annual ly in the United States alone (Rosen , 2000). Salmonella enteriditis and Salmonella typhi (responsible for typhoid fever) are strains commonly observed to infect humans. The infective dose for Salmonella is relatively smal l , at 15-20 bacterial cel ls (Food and Drug Administrat ion, 2005). The majority of Salmonella-re\ated c a s e s are the result of foodborne t ransmission from poultry, beef, pig and dairy products. However, the potential exists for waterborne t ransmission to humans from livestock sources . Hooda et a l . (2000) cite severa l studies descr ib ing the longevity of Salmonella as ranging from days to several years after infected animal wastes have been spread on agricultural f ields and incorporated into soi ls. Despite this potential, there are few documented outbreaks of waterborne i l lness attributed to Salmonella spp. (Angulo et al . , 1997; Taylor et a l . , 2000) and there are no known c a s e s l inked directly to agriculture. Yersinia Yersinia has been implicated in both foodborne and waterborne outbreaks of enteric i l lness; however, few strains are pathogenic to humans (Szewzyk et al . , 2000). Yersinia enterocolitica is the primary c a u s e of human infections, which number approximately 3,000-20,000 per year in the United States (Rosen , 2000). 2.5.1.2. Protozoa Pro tozoa are single-cel led organ isms (a subtype of parasite) with more complex structures and life cyc les than bacter ia (Rosen , 2000). The two protozoan parasi tes of primary concern with regard to agricultural watersheds are Giardia and Cryptosporidium, with s ize ranges of 8-15 um, and 4-6 um, respectively (Szewzyk et al . , 2000). T h e s e organisms reproduce only within the gastrointestinal tract of 28 Chapter 2 humans and animals (not in the environment) and are more resistant to environmental stressors and traditional water treatment techniques than most other pathogenic microorganisms (Health C a n a d a , 2004b). The lifecycle of these parasi tes is particularly well suited to dispers ion and subsequent infection v ia contaminated drinking water. Infection occurs when the parasi tes either attach themselves to the intestinal lining (Giardia) or invade host cel ls (Cryptosporidium) where they thrive and reproduce. Prior to being shed by the host organ ism, they e n c a s e themselves in a protective cyst wall (Giardia) or oocyst wall (Cryptosporidium) that is highly resistant to environmental degradat ion and standard chemica l disinfectants (Szewzyk et al . , 2000). O n c e ingested, and if environmental condit ions are favourable, the organism excysts and infects the new host. Giardia Giardia lamblia (also referred to as Giardia intestinalis or Giardia duedenalis) is a protozoan parasite found in the intestinal tracts of many different hosts, including humans, dogs, cats, bears, muskrats, cattle and pigs, and is capable of c ross-spec ies t ransmission (Olson et a l . , 1999; R o s e n , 2000). W h e n the feeding stage (trophozoite) of the organism detaches from the intestinal wall, it forms a cyst, divides within the protective cas ing and is shed to the environment with the feces (Health C a n a d a , 2004b). O n c e ingested, excystat ion is tr iggered by s tomach acids and enzymes . Two trophozoites emerge and infect the intestinal tract v ia asexua l reproduction (Health C a n a d a , 2004b). Giardia is the most commonly reported intestinal parasite worldwide. In C a n a d a , the prevalence (percent of population infected) in humans is approximately 1-5%, although this number may be much larger as many c a s e s are thought to be asymptomat ic (Health C a n a d a , 2004b). Of the 29 documented waterborne d isease outbreaks in B C between 1980-2002, 13 were attributed to Giardia (Chr is tensen, 2003). Preva lence ranges from 10-100% in cattle and from 1-20% in pigs (Olson et al . , 1999). Giardia lamblia cysts can survive for extended per iods in surface water and drinking water sys tems. Bingham et al. (1979) demonstrated that cysts could survive for up to 77 days in tap water at 8 °C, but that longevity dec reased with increasing temperatures. The infectious dose for G. lamblia is very low; in mice 1-10 organisms generate infection (giardiasis) (Stachan and Kunstyr, 1983). Onset of i l lness is general ly within one week of cyst ingestion, and may last 1 to 2 weeks ; however, chronic c a s e s lasting years have been reported (Food and Drug 29 Chapter 2 Administrat ion, 2005). Symptoms of the acute phase of the i l lness may include nausea , d iarrhea, malaise and somet imes low-grade fever or chi l ls; however, this usual ly resolves spontaneously in healthy individuals (Health C a n a d a , 2004b). Whi le it is bel ieved that giardiasis is a zoonotic d i sease (transmitted from animals to humans), little ev idence from controlled exper iments is avai lable to definitively link l ivestock sources to human i l lness v ia waterborne t ransmission (Health C a n a d a , 2004b). However, given the very low infectious dose , high prevalence in l ivestock animals, longevity of cysts in the environment and number of known outbreaks of giardiasis in rural water sys tems in B C (and elsewhere), it is considered a significant human health risk when detected in water suppl ies. Cryptosporidium Cryptosporidium infects a wide range of hosts, including humans, cattle, goats, sheep , pigs, horses, dogs, cats, mice, voles and raccoons (Rosen , 2000). It has a relatively complex l i fecycle (see Smith and R o s e (1998) for a complete overview), but is similar to Giardia in that it infects the intestine. However, rather than affixing to the intestinal wal l , Cryptosporidium invades cel ls that line the intestine (enterocytes). It then reproduces intracel lular^, and is transmitted in an environmental ly resistant cas ing (oocyst). Approximately 8 0 % of oocysts produced in the intestine are "thick-walled" and are general ly excreted from the body. These oocysts are much more resistant to environmental degradat ion than Giardia cysts (Olson et al . , 1999). The remaining 2 0 % of oocysts are "thin-walled" and readily rupture in the intestine, reinvade new enterocytes and continue to reproduce leading to autoinfection, thus prolonging the infection (Rosen , 2000). A n infected person or animal can shed oocysts for several days over the course of the i l lness, with concentrat ions of up to 10 million oocysts per gram of feces . A s with Giardia, the infectious dose for Cryptosporidium \s very low. The infectious dose , as est imated by Health C a n a d a (2004b), is 39 oocysts in humans and 10 in an imals ; however, one organism is thought to be enough to produce i l lness in a healthy individual (Food and Drug Administrat ion, 2005). There is currently no known treatment for cryptosporidiosis. The infection is self-limiting in healthy individuals (lasting 2-4 days) but can be fatal for those with compromised immune sys tems (Food and Drug Administrat ion, 2005). Cryptosporidium parvum is thought to be the spec ies of primary concern in terms of human i l lness, and infections in cattle populat ions have been well documented (Casemore et al. , 1997). Other 30 Chapter 2 zoonot ic sources have been identified, including deer (Ong et al . , 2002), suggest ing that multiple exposure paths exist in rural watersheds. In B C , Cryptosporidium was the causat ive agent in three of the 29 waterborne d i sease outbreaks documented from 1980-2002. Two of these (one in Cranbrook and one in Ke lowna, in 1996) were among the largest outbreaks during that time, and in two of the three outbreaks cattle were identified as the source (BC Provincial Health Officer, 2001). In the spring of 2001 , an outbreak of cryptosporidiosis in North Battleford, Saska tchewan , resulted in 1,907 conf irmed c a s e s (5,800-7,100 estimated). The community der ives its drinking water from the North Saska tchewan River, and while the source of the pathogen was not conf i rmed, it is thought that upstream sewage treatment plants may have p layed a role (Laing, 2002). Cryptosporidium was also the cause of the largest documented waterborne d isease outbreak in North Amer ica , which took place in Mi lwaukee in 1993, and affected over 400,000 people (Mackenz ie et a l . , 1994). B a s e d on an analysis of death certif icates for two years after the outbreak, cryptosporidiosis was listed as an underlying or contributing cause for 54 deaths in the Mi lwaukee area, compared to four deaths for the two years prior to the outbreak (Hoxie et a l . , 1997). Other parasites It should be noted that other protozoan parasi tes are capab le of caus ing human i l lness. In B C , for example, the world's largest outbreak of waterborne toxoplasmosis (caused by the parasite Toxoplasma gondii) occurred in late 1994 and early 1995, infecting 100 people in the municipality of Victoria (Bowie et al . , 1997). This outbreak was assoc ia ted with peaks in local rainfall and turbidity in the municipal drinking water reservoir, and it is hypothesized that the parasite was der ived from one or more infected wild or domest ic cat(s). 2.5.1.3. Viruses Vi ruses are one of the the smal lest known pathogens (from 20-300 nm), consist ing of a nucleic acid core (of R N A or DNA) encapsula ted in a protein shell (Health C a n a d a , 2004a). All v i ruses are inactive unless they are within living cel ls. V i ruses reproduce by instructing a host cell to produce multiple copies of its genet ic material (as well as the protective protein shell) using the cel l 's own processes , until it ruptures and the viruses are re leased and continue infecting other cel ls. Enteric v i ruses are those that reproduce only in the host 's gastrointestinal tract (Rosen , 2000). 31 Chapter 2 Like protozoan parasi tes, v i ruses are shed by infected hosts in high concentrat ions, reaching up to 1 billion v iruses per gram of feces (Health C a n a d a , 2004a). They are also resistant to environmental degradat ion in water, where they may remain infectious for months (Szewzyk et al . , 2000), particularly at low temperatures and/or when adsorbed to sediments (Health C a n a d a , 2004a). Pathogenic v i ruses are regularly detected in surface waters. For example, Ehlers et al . (2005) detected viable enteroviruses in 2 8 . 5 % of river water samples , and 2 6 . 7 % of spr ing/dam surface water samples in se lected sites ac ross South Afr ica. Interestingly, this study also detected viruses in treated drinking water samp les , all of which met current Wor ld Health Organizat ion (WHO) drinking water guidel ines for heterotrophic plate counts and fecal coliform counts. The most common source of v i ruses is human sewage derived from sewage treatment plants (e.g., Lodder and Husman , 2005), sewage leaks or septic sys tems (Rosen , 2000). In agricultural watersheds, v i ruses may also be der ived from sewage s ludge spread on agricultural f ields (Health C a n a d a , 2004a ; Carter, 2005). Whi le interspecies t ransmission of infectious v i ruses is documented, there have been no documented c a s e s that would suggest a link between human i l lness and l ivestock sources of waterborne v i ruses (Rosen , 2000). However, Cl iver and Fayer (2004) suggested that the potential for waterborne viral z o o n o s e s does exist, particularly given the number of v i ruses produced in infected animals, their consistent detection in surface waters, and the significant potential for genetic mutation (leading to a change in host specificity) during the process of viral synthesis. They descr ibed severa l potential scenar ios in which c ross -spec ies , waterborne t ransmission could take p lace, and note that there are several examples of interspecies viral infection (e.g., avian inf luenza, bovine spongiform encephalopathy) where waterborne t ransmission has not been documented, but cannot be ruled out. 2.5.2. Pathogen sources in agricultural watersheds In agricultural watersheds, there are three primary sources of pathogenic organ isms: 1) wildlife, 2) l ivestock (direct deposit ion into surface waters, transport from manure storage facilities or from agricultural fields receiving manure) and 3) humans (septic sys tems, sewage treatment plants or sewage sludge). The following descr ibes pathogen levels in the wastes generated by each of these sources . 2.5.2.1. Wildlife Wild animals are known reservoirs of pathogenic bacteria, protozoa and v i ruses (Daszak et a l . , 2000). In many c a s e s , these pathogens are host-specif ic (they infect a limited range of host spec ies) , and 3 2 Chapter 2 are not pathogenic in humans. Wildlife sources represent a risk to human health when : 1) they carry and/or are infected by microbial strains that are a lso pathogenic to humans or 2) wildlife-specific pathogens mutate so as to be able to infect humans. Cryptosporidium is a good example of a pathogen that is highly infectious to human and animal hosts. In total, 15 spec ies of Cryptosporidium are known to infect vertebrate hosts, seven of which are infectious to humans (Fayer, 2004). These spec ies , C. baileyi, C. canis, C. felis, C. meleagridis, C. muris, C. hominis and C. parvum, were originally thought to be speci f ic to ch ickens, dogs , cats, turkeys, mice, humans and mice, respectively. However, each spec ies has been found to infect humans, and C. hominis has also been shown to infect marine and land mammals (Fayer, 2004). Cryptosporidium parvum and C. hominis are considered the most infectious strains to healthy humans (Sturdee et a l . , 1999), and they have been identified in 155 spec ies of mammals , including severa l spec ies of sheep, deer, bear, fel ines and rodents. The other spec ies are commonly found in immunocompromised individuals (Fayer, 2004). Recent ly, O n g et a l . (2002) identified a cervine (deer) C. parvum genotype in humans in B C , illustrating the potential for the emergence of new wildlife sources of cryptosporidiosis. Strains of G. lamblia are also infective in multiple spec ies . Severa l different strains of this organism have been identified, and while they demonstrate similar morphological character ist ics, they are genetical ly unique and therefore are infectious to different hosts. Giardia strains found in humans fall under two genetic groupings, known as Assemb lage A and Assemb lage B. A s s e m b l a g e A organisms (thought to be the more infectious of the two assemblages) are commonly found in humans, cats, dogs, deer and beavers , while Assemb lage B parasites have been observed in humans, chinchi l las, beavers and rats (Trout et al. , 2003 ; Thompson , 2004). Other wildlife spec ies are known reservoirs for Giardia] however, research is needed to genetical ly character ize these Giardia strains to understand the role of wildlife in waterborne outbreaks of giardiasis in humans (Thompson, 2004). Importantly, while there are several wildlife reservoirs for Giardia strains that are pathogenic to humans, there is little direct ev idence of a link between waterborne outbreaks of giardiasis and wildlife sources (Thompson, 2004). Relatively little work has been done to determine the role of wildlife in the t ransmission of waterborne bacterial pathogens to humans. Severa l studies have noted the presence of pathogenic bacteria in wildlife spec ies , however their role in human i l lness is not well understood. Wi ld deer have been shown to carry £. co / /0157 :H7 (Sargeant et al . , 1999); however, prevalence is general ly low (< 3%) in populat ions studied (Sargeant et a l . , 1999; F ischer et a l . , 2001 ; Renter et al . , 2001). The s a m e 33 Chapter 2 bacterium has a lso been observed in rats (Cizek et al . , 1999) and var ious bird spec ies in c lose proximity to farms (Nielsen et al . , 2004) indicating that there is likely two-way t ransmiss ion between farm and wild animals in agricultural environments. 2.5.2.2. Livestock Pathogens der ived from livestock enter surface waters v ia direct contact (i.e., l ivestock access ing streams or standing water bodies) or v ia transport from manured agricultural f ields or manure storage piles. Table 2-2 provides a summary of pathogen levels in l ivestock manure from recent studies, and descr ibes the prevalence of infection (% of animals infected) where avai lable. Table 2-2 - Prevalence of common waterborne pathogens in livestock populations and geometric mean concentrations of organisms (per gram of feces). All numbers are for fresh manure except those in () which represent samples from stored manure. Modified from Hutchison et al. (2005a), except where otherwise noted. ND = no data. Microorganism Animal Host Cattle Pigs Poultry Sheep Prevalence C o n e . Prevalence C o n e . Prevalence C o n e . Prevalence C o n e . (%) (cfu/g) (%) (cfu/g) (%) (cfu/g) (%) (cfu/g) E. coli 0157:H7 13.2 1200 11.9 3900 ND ND 20.8 780 (9.1) (260) (15.5) (1300) (22.2) (250) Salmonella spp. 7.7 2100 7.9 600 17.9 220 8.3 710 (10) (2500) (5.2) (610) (11.5) (4000) (11.1) (5800) Campylobacter spp. 12.8 320 13.5 310 19.4 260 20.8 390 (9.8) (530) (10.3) (1600) (7.7) (590) (11.1) (100) C. parvum 5.4 19 13.5 58 260 ND 29.2 10 (2.8) (10) (5.2) (33) (590) (0) (0) G. lamblia 100' 2230' ND ND ND 1.52 450 2 1 Ra ls ton et a l . (2003) in a 27 week study of G. lamblia and C. parvum p reva lence in newborn ca l ves 2 G i a n g a s p e r o et a l . (2005) Table 2-2 illustrates several important points regarding l ivestock as sources of waterborne pathogens. Firstly, pathogens were detected in considerable numbers in both stored and fresh manure for all animal types. Whi le pathogen prevalence varies significantly from study to study, and over t ime (see below), it is clear that l ivestock are reservoirs for a range of zoonot ic pathogens. Further, in most c a s e s , storage of animal wastes for extended periods of time did not result in a substantial drop in the number of organisms detected, indicating that both fresh and stored wastes are important pathogen sources that must be properly managed. 34 Chapter 2 Animal age also plays a role in prevalence. Hutchison et a l . (2005a) demonstrated significantly higher levels of E. coli 0157:H7 and Campylobacter spp. in wastes from young ca lves , lambs and piglets (< 3 months of age), and observed a higher prevalence of these pathogens in wastes derived from animal populat ions containing young stock. Similar trends have been observed e lsewhere (e.g., Garber et al . , 1995; Ralston et a l . , 2003), and in a review of several studies, O lson et a l . (2004) noted that the average age for peak shedding of (oo)cysts (cysts and oocysts) from cattle is approximately five weeks and one-two weeks , respectively. This indicates that the s e a s o n during which l ivestock animals are borne is a period of increased risk of surface water contamination from these sources . The survival of pathogenic organisms in manure and soil is a function of severa l factors, including initial pathogen concentrat ions, length of storage period, type of storage and environmental var iables (temperature, moisture, pH , U V exposure, nutrient availability). Pathogenic bacter ia, for example, show increased longevity in soi ls with high soil moisture content (Rosen , 2000; Duffy, 2003; G u a n and Holley, 2003). Mubiru et a l . (2000), in a control led laboratory study, observed pathogenic E. coli survival in two soil types, and noted that culturable bacter ia were still present after 8 weeks , with higher concentrat ions in f iner-grained, moister soi ls. Temperature also plays a significant role in pathogen survival, although the impact varies by pathogen type. In genera l , as temperatures increase (above 4 °C), longevity dec reases . However, f reezing temperatures also increase pathogen mortality. In a 12-week study, O lson et al . (1999) .assessed (oo)cyst survival in water, cattle manure and soil at -4, 4 and 25 °C. After one week in soi l , water and feces , Giardia cysts were no longer viable at -4 °C and 25 °C. Cys ts kept at 4 °C were infective for 11 weeks in water, seven weeks in soil and one week in feces . Cryptosporidium oocysts were much more resistant and survived the entire 12 weeks in water and soi l , when kept at -4 and 4 °C. Higher temperatures also appear to adversely impact the survival rates of v i ruses and bacter ia. Himathongkham et al . (1999) examined survival rates of E. coli 0157:H7 and Salmonella typhimurium in manure and manure slurry at 4, 20 and 37 °C and noted that bacter ia survived more than seven t imes longer at 4 °C than at 37 °C. In a review of several studies on pathogen survival, G u a n et al . (2003) noted that most pathogens can survive for at least 30 days in cold (4-6 °C) soi l , and that Cryptosporidium shows the greatest resi l ience to freezing temperatures, while E. coli and Salmonella are the most resilient in warmer soi ls (20-30 °C). 35 Chapter 2 Survival rates in manure tend to .decrease after manure is appl ied to agricultural f ields. Bolton et al. (1999) a s s e s s e d the survival of E. coli 0157:H7 in samples of inoculated cattle manure under different environmental condit ions (stored in plastic at 10 °C, stored in plastic outside and spread on grazing land), and observed culturable bacter ia after 99 days in both containers and after 50 days in the surrounding soils of the grazing land. Hutchison et al . (2005b) appl ied l ivestock wastes that had been inoculated with E. coli 0157:H7, C. jejuni, Salmonella and C. parvum to grass pasture and monitored their concentrat ions over time. Pathogen concentrat ions showed a marked decl ine within 24 hours of appl icat ion. The average time required for a 1-Log reduction was 1.94 days for bacteria and ranged from 8-31 days for oocysts (with little or no decrease in viability over that time). 2.5.2.3. Humans Pathogens assoc ia ted with human sewage reach surface waters v ia sewage treatment plant (STP) effluent to st reams, application of sewage s ludge to agricultural lands and through leaching from septic sys tems, it is suggested that human-infective strains of certain pathogens (such as G. lamblia) were first introduced to current wildlife hosts through contact with water contaminated by human fecal matter (Thompson, 2004). Consequent ly , these hosts now serve to amplify and spread these pathogens in natural sys tems. S e w a g e treatment plant effluent represents a significant point source of enteric pathogens to surface waters. Payment et a l . (2001) a s s e s s e d the removal of indicator bacter ia and specif ic pathogenic v i ruses and protozoa by a primary S T P (that does not employ chlorination) d ischarging to the St. Lawrence river at Montreal. Remova l rates of C. parvum, G. lamblia, E. coli and v i ruses were 27%, 76%, 1 2 % and 0%, respectively, resulting in significant inputs of pathogens to the river at this point, and a potential health impact on recreational users and biota downstream. Char les et al . (2003a) conducted a review of pathogen loads from S T P ' s discharging effluent into drinking water catchments in Sydney, Austral ia, and noted that 7 6 % of effluent samp les contained Cryptosporidium oocysts and 17% were positive for enteric v i ruses. S e w a g e treatment plants that chlorinate effluent prior to d ischarge have also been observed to re lease significant quantities of pathogens to the environment. Whi le bacterial reduction does take place, removal and/or inactivation of protozoan parasi tes is often incomplete, due to the resistance provided by the protective cas ings of (oo)cysts (e.g., Br iancesco and Bonadonna , 2005). 36 Chapter 2 The use of biosol ids (the isolated portion of sewage sludge remaining after treatment) as an amendment to agricultural soi ls can also contribute significantly to the concentrat ions of human pathogens, heavy metals and , potentially, endocr ine disrupting substances in soil (Coote and Gregor ich, 2000). Whi le treatment of biosol ids is required prior to application to agricultural fields in C a n a d a (Chambers et al . , 2001), total inactivation of microbes is often not ach ieved (Rose et al . , 1996). A s noted in a review by G e r b a and Smith (2005), the pathogenic organ isms found in sewage s ludge are similar to those in animal manure, but a lso include human v i ruses. A s descr ibed above, and as noted by G e r b a and Smith (2005), these organ isms can persist in the soil for months to years and can act as a reservoir for contamination of crops and nearby surface waters. Househo ld septic sys tems, common in agricultural watersheds, can be hydrologically l inked to surface water and groundwater networks and contribute to loading of waterborne pathogens. The degree to which such contaminat ion takes place is a function of the septic system maintenance record, its proximity to surface waters and groundwaters and the texture of surrounding soi ls, with f iner-grained materials reducing the potential for pathogen transport (Char les et a l . , 2003b). Harwood et al . (2000) used the antibiotic resistance patterns of fecal coliform and fecal streptococci to identify sources of bacterial contaminat ion of surface waters in Flor ida, and identified septic sys tems as a primary source of contaminat ion. Groundwater contamination with sept ic tank effluent is a lso common (Scandura and Sobsey , 1997; DeBorde et al . , 1998; Bopp et al . , 2003 ; Char les et a l . , 2003a ; Char les et al . , 2003b), and has been implicated in waterborne d i sease outbreaks (e.g., Bopp et a l . , 2003). 2.5.3. Transport processes In a review of transport mechan isms for waterborne pathogens, Ferguson et al . (2003) noted that transport to surface waters is governed by three groups of p rocesses : 1) those that inf luence adsorption to and desorption from particulate matter, 2) hydrological/meteorological p rocesses and 3) mechanical and biological p rocesses that inf luence transport pathways. Simply put, these define the availability of free organisms for transport, the existence of a transport force and medium and the availability of a pathway for transport, respectively. A similar f ramework is used in the following d iscuss ion . 2.5.3.1. Adsorption and desorption processes Adsorpt ion and desorpt ion p rocesses are control led by the physical characterist ics of the microbial pathogens involved (hydrophobic and hydrophilic properties, s ize and morphology), the nature 37 Chapter 2 of the soil solution and by the soil texture, chemica l composi t ion, electrostatic potential and bulk characterist ics (Ferguson et al . , 2003). Hydrophobic organisms preferentially adsorb to soil particles, a process that is enhanced as the ionic strength of the soil solution increases (Jewett et al . , 1995). Further, for v i ruses, variations in surface charge assoc ia ted with protein coats on different v i ruses (or strains of the s a m e virus) influence adsorpt ive behaviour (Sobsey et al . , 1995). In a soil co lumn experiment, Sobsey et al . (1995) also noted the roles of temperature, soil texture and soil solution on virus transport. Due to the availability of fewer adsorpt ion sites, coarser-gra ined, sandy soi ls resulted in dec reased retention of v i ruses relative to sandy loam and c layey soi ls. The addition of organic-r ich water to the soil co lumn reduced virus retention in all soil types due to competit ion for binding sites by organic compounds . Finally, higher temperatures (25 °C compared to 5 °C) resulted in fewer v i ruses being re leased from soil co lumns for all soil types due to a dec rease in virus survival and increased adsorpt ion rates. Similar inf luences on bacterial adsorpt ion have also been observed. He ise and Gust (1999) util ized a control led flow experiment over a sandy substrate to compare the distribution of nourished bacter ia, starved bacter ia and inert microspheres to determine the relative inf luence of biological and physical parameters on cell movement. The distribution of inert particles and starved bacter ia correlated with flow patterns. In contrast, nour ished bacter ia were more equally d ispersed across the sediment-water interface, indicating that they exert some control over their transport (such as active movement and adsorption) that is independent of flow. These results suggest that prediction of bacterial transport using particulate (soil) transport models , while useful in providing order-of-magnitude est imates, may not accurately reflect p rocesses at the cellular or soil particle level (Heise and Gust , 1999; Tyrrel and Quinton, 2003). Further, mitigation strategies a imed at limiting soil (and therefore microbial) transport may not have the desired effect (Jamieson et al . , 2004). It should be noted; however, that the impact force assoc ia ted with simulated rainfall has been observed to detach and transport bacterial cel ls individually (Muirhead et a l . , 2005), as descr ibed in more detail below. Oocys ts and cysts also have an affinity for adsorpt ion. Whi le few studies have experimental ly a s s e s s e d (oo)cyst attachment (Ferguson et al . , 2003), two have provided s o m e insight into the p rocesses involved. In a controlled experiment, Dai and Boll (2003) used flow cytometry and confocal microscopy to investigate the degree to which (oo)cysts attach to soil particles relative to positively and negatively charged beads . They quantified the negative surface charges on soil part icles, negatively charged beads and (oo)cysts, and noted that neither cysts nor oocysts attached to beads or soil particles of like charge. 38 Chapter 2 Attachment to positively charged beads ; however, was observed. This is in contrast to studies which demonstrated attachment of (oo)cysts to organic matter in sewage and to sed iments (Medema et al. , 1998; Searcy et a l . , 2005, respectively). The d iscrepancy observed for organic matter is attributed to differing charge character ist ics and adsorpt ion behaviour for soil vs . biological aggregates found in sewage effluent (Dai and Bol l , 2003). The d iscrepancy in the latter study (Searcy et a l . , 2005), is credited to the use of higher suspended sediment concentrat ions as part of the experimental protocol. 2.5.3.2. Hydrological and meteorological processes Water is the primary transport medium linking pathogens to surface waters and thus, it is not surprising that many studies have observed correlations between rainfall events and waterborne d isease outbreaks. Curr iero et al . (2001) compared the inc idence of waterborne d i sease outbreaks in the United States to precipitation levels between 1948 and 1994. A total of 6 8 % of all outbreaks were preceded (in the s a m e month) by rainfall events above the 8 0 t h percenti le for their region. Kovatz et al . (2005), in an international study of the seasonal i ty of laboratory-confirmed c a s e s of Campylobacter infection, a lso suggested a link between cl imate and incidence of infection. Further, extreme precipitation has been implicated as a causat ive factor in several outbreaks [e.g., Mi lwaukee, W iscons in (Mackenz ie , 1994), Walkerton, Ontario (Hrudey et al . , 2003) and numerous outbreaks in Finland (Miettinen et a l . , 2001)]. These observat ions are supported by plot and watershed-sca le studies linking rainfall to increased pathogen concentrat ions in surface waters. Schi jven et a l . (2004) identified the importance of rainfall impact force in (oo)cysts mobil ization by applying the s a m e volume of precipitation in mist and drop form to different types of manure in a laboratory environment. R e l e a s e of oocysts and cysts was four and nine t imes higher, respectively, as a result of drip appl icat ion. This was attributed to increased re lease eff iciencies resulting from the impact forces of water droplets. In addit ion, cumulat ive release values were greater for oocysts than cysts, suggest ing that re lease eff iciencies are higher for smal ler parasit ic pathogens. .Muirhead et al . (2005) a s s e s s e d the effects of rainfall on E. coli mobil ization from cowpats after grazing using simulated rainfall. Transported cel ls were partitioned into at tached (to particles) and unattached fractions. They observed that E. coli concentrat ions in surface runoff were strongly correlated to those in cowpats, and that cowpats served as a viable bacteria source for more than 30 days . Further, the majority of transported cel ls were not only unattached to soil or organic particles, but were a lso 39 • Chapter 2 transported as single cel ls, rather than in cell c lumps. This is in contrast to studies which have noted cell c lumping in surface waters (e.g., Kiorboe et al . , 2002), and is thought to represent the initial transport phase of these cells (associated with mobil ization by rainfall impact forces) prior to aggregat ing in surface waters. T h e s e results suggest that bacterial transport model ing, which has often equated bacterial transport with particulate (sediment) transport, needs to account for different bacterial mobil ization and transport thresholds due to dif ferences in mass and density when compared to sediment. Severa l researchers have examined the relationship between rainfall and stream water quality in an effort to quantify the water-quality risk assoc ia ted with storm events (Hansen and Ongerth, 1991; Atherholt et a l . , 1998; K is temann et al . , 2002). K is temann et a l . (2002) investigated the influence of rainfall on pathogen concentrat ions in tributaries of three G e r m a n drinking water reservoirs, two of which had significant pathogen sources (agriculture or sewage treatment plants), while the third was almost entirely forested. Significantly higher bacterial concentrat ions were observed in surface waters of all three watersheds during rainfall events. Parasi te concentrat ions were only observed to increase significantly in the non-forested watersheds where significant parasite sources were known to exist. A similar trend was documented by Atherholt et al . (1998). In an attempt to correlate (oo)cyst concentrat ions with more easi ly measured water quality parameters, they noted significant correlat ions between rainfall and parasite numbers in the Delaware River, and attributed the observed trend to increased over land transport as well as resuspens ion of stream-bottom sediment. Whi le pathogen concentrat ions general ly show posit ive correlations with rainfall, microbe loading per unit of precipitation appears to follow a seasona l pattern that is control led by source availability and rainfall f requency, thus complicat ing the development of a quantitative relationship between these two var iables. Hunter and McDona ld (1991) observed significantly higher fecal coliform concentrat ions in surface waters during summer months, and assoc ia ted these values with higher concentrat ions in over land flow from agricultural f ields at this t ime. The dec rease in concentrat ions observed during the winter months was attributed to seasona l changes in the land store of enteric bacter ia resulting from reduced application of manure and continual f lushing by winter rains. Similar seasona l trends have been observed e lsewhere (e.g., Hunter et a l . , 1999; e.g., Rodgers et al . , 2003). In-stream processes are also important in pathogen mobil ization as s t reambed sediments can serve as a significant store of v i ruses, bacter ia and protozoa. Pathogen concentrat ion in sediments may be up to 1000 t imes those in the water co lumn (Buckley et a l . , 1998). This has been attributed to 40 Chapter 2 continual sedimentat ion of organ isms adsorbed to sediments, stream a c c e s s by l ivestock (Nagels et al . , 2002) and prolonged survival of bacter ia when assoc ia ted with sediments (Sherer et a l . , 1992). To illustrate the importance of in-stream bacterial stores, Muirhead et al . (2004) created artificial f loods downstream of a water reservoir during dry condit ions. By eliminating bacterial transfer from the land surface via over land flow, they were able to a s s e s s bacterial loading from the s t reambed alone. For three success ive f lushing events of equal magnitude, turbidity and E. coli concentrat ions showed characterist ic hysteretic responses, with peaks at the onset of maximum flow, fol lowed by rapid decl ines for both parameters. Bacter ial concentrat ions during the first storm event increased by more than two orders of magnitude. Subsequent f loods over the next two days resulted in similar, but success ive ly smal ler peaks for both parameters, suggest ing source exhaust ion over t ime. A similar study (Nagels et al . , 2002) documented the impact of a natural and an artificial f lood on the s a m e river. The natural event resulted in an increase in stream E. coli concentrat ions to a maximum of approximately 40,000 M P N / 1 0 0 mL (most probable number per 100 mL), an increase of over two orders of magnitude from background levels. The artificial f lood produced a peak concentration of over 12,000 M P N / 1 0 0 mL, demonstrat ing the signi f icance of the in-stream store to event-related bacterial concentrat ions in surface waters. Interestingly, the peak bacterial concentrat ion for the natural storm was observed prior to peak flow, while the artificial f lood generated maximum concentrat ions after peak flow was reached, due to the relative s teepness of the artificial f lood front. Relatively little work has been done to a s s e s s the role of s t ream-bed sediments on (oo)cyst and virus concentrat ions in overlying waters. Work in the marine environment indicates that concentrat ions of v i ruses in sediment can exceed those in overlying seawater (Label le et al . , 1980), that bacterial concentrat ions in suspended floes can exceed those in the sediment and water column (Schendel et al . , 2004) and that sediments can increase survival time of v i ruses significantly (Labelle and Ge rba , 1980). Further, the work of M e d e m a et al . (1998) suggests (oo)cysts can settle out of suspens ion when significant amounts of organic matter are present. These studies indicate that stream sediments must be cons idered a potentially significant source for all pathogen types. 2.5.3.3. Mechanical and biological processes Over land transport is often cons idered the primary pathway for pathogen movement (Jamieson et al . , 2004). However, subsur face flow must be considered when assess ing pathogen transport and 41 Chapter 2 mitigation strategies, particularly where field drainage sys tems are present as they can represent a direct, subsur face link to surface waters. Ogden et a l . (2001) a s s e s s e d bacterial leaching to drainage sys tems after slurry appl icat ions in a plot-scale experiment. Between 0.2 and 1 0 % leaching of E. coli bacter ia to subsur face drainage sys tems was observed as result of rainfall, with the highest concentrat ions observed within a week of slurry appl icat ion. Hunter et al . (1992) conducted a study to a s s e s s the relative contribution of over land flow, matrix throughflow and macropore flow to surface water bacterial cycl ing in an upland watershed in northern Eng land. Each pathway was observed to transport fecal bacter ia to surface waters, with macropore and over land flow contributing most to stream bacterial loading. For all pathways, loading was positively correlated with stream gauge height, illustrating the role of rainfall in surface and subsur face transport. It should be noted that rainfall is not required to initiate downward transport of pathogens. The application of liquid manure can lead to an almost instantaneous response in bacterial water quality in sub-sur face soi ls and drains (Dean and Foran, 1992). The presence of macropores in a soil profile al lows transport of larger vo lumes of water and particulates down the soil profile than would be possib le through the soil matrix, thus encouraging downward mobility of pathogens (Jamieson et al . , 2002). Phys ica l soil character ist ics, biological activity and land management pract ices are the primary drivers of macropore development in a soil profile. Soi l texture has been observed to influence the development of macropores and the degree to which non-matrix (macropore) flow takes p lace. Flury et al . (1994) a s s e s s e d the development of preferential flow pathways in soi ls of different texture and noted that structured, clay-rich soi ls are highly prone to macropore development when compared to poorly-structured, sandy soi ls. Natsch et al. (1996) observed the influence of root growth and soil biological activity on downward movement of bacter ia in two inoculated soi ls (one planted with wheat and regularly p loughed, the other a grass land that had not been tilled in several years). Downward migration in the grass land plot occurred to greater depths due to extensive earthworm activity, a wel l -developed, deep root network and the lack of an impermeable plough pan that had deve loped in the wheat field as a result of regular tillage. Regular tilling of the soil not only results in a plough pan, it a lso continually destroys macropore networks, and greatly dec reases downward mobility of pathogens. Davies et al . (2004) observed a similar trend for oocyst transport and noted that over land flow was five t imes higher for unvegetated plots than vegetated plots. This resulted in significantly higher numbers of transported oocysts in over land flow. 42 Chapter 2 2.5.4. Ecological and human health impacts 2.5.4.1. Ecological impacts Emerging infectious d i seases often ar ise from wild, domest ic or l ivestock animal populat ions and are general ly cons idered in light of their potential impacts to human health (Daszak et al . , 2000). Indeed, in a review of the risk factors assoc ia ted with emerging infectious d iseases , Taylor et al . (2001) illustrated that 7 5 % are c a u s e d by zoonot ic pathogens. Recent work suggests ; however, that emerging pathogens (those that have a sudden increase in range, have moved from one spec ies to another, increased in severity, undergone a change in pathogenesis or are the result of recently evolved pathogens) can also significantly impact wildlife populat ions (Daszak et al . , 2000; Environment C a n a d a , 2001 ; D a s z a k et al . , 2004). Daszak et al. (2000) referred to the t ransmission of emerging pathogens amongst populat ions in the s a m e geographic region (often between l ivestock and local wildlife) as spil l-over, and noted that spil l-back, the re-transmission of a similar or more virulent form back to the original populat ion, can have devastat ing effects on l ivestock. One of the most dramatic examples of this is the wel l -documented global decl ine in amphibian populat ions observed in natural environments over the past severa l d e c a d e s (Houlahan et a l . , 2000). Numerous potential c a u s e s were proposed, including habitat loss, increased U V radiation, pollution and cl imate change. Berger et al . (1998) proposed that the observed decl ines in Central Amer i ca and Austral ia were actually the result of a fungal pathogen (phylum Chytridiomycota) that attacked epidermal cel ls of host organisms. S ince then, this pathogen, which is thought to have emerged from humans, has been implicated in dec l ines of several other amphibian populations globally (Daszak et al . , 2004). Whi le water plays a role in t ransmission of wildlife pathogens (e.g., avian botul ism, commonly observed in and around shal low lakes), limited conclusive ev idence exists for waterborne t ransmission of emerging infectious d i seases from livestock or humans to wildlife populat ions. G iven the potential for emergence in wildlife populat ions as a result of genetic mutation and geographic expans ion, and the subsequent threat posed to humans and l ivestock by spi l l -back, it has been recommended that a more proactive approach to wildlife infection be adopted (Daszak et al . , 2000). 43 Chapter 2 2.5.4.2. Human health impacts The i l lnesses caused by some common waterborne pathogens were descr ibed previously (Sect ion 2.5.1). This sect ion descr ibes health impacts at the population level by reviewing the incidence of waterborne outbreaks in B C , across C a n a d a and internationally. Waterborne disease in British Columbia Recent ly, reviews were conducted of waterborne d i sease outbreaks in B C s ince 1980 (BC Provincial Health Officer, 2001 ; Chr is tensen, 2003) and of drinking water infrastructure and drinking water sources in the province (BC Auditor Genera l , 1999). Table 2-3 provides a summary of the 29 outbreaks to take place over this time period. This list illustrates s o m e important points about waterborne i l lness in B C in terms of incidence, common pathogens, pathogen sources and mechan isms of infection. 44 Table 2-3 - Waterborne disease outbreaks in BC from 1980-2004. Modified from (BC Auditor General, 1999; Peck, 2004) Year Location Pathogen Number of cases Lab-confirmed / Epidemiological estimate Water Source Suspected Pathogen Source 1980 Nakusp Campylobacter 1 2 / 8 0 0 Surface Wildlife 1981 100 Mile House Giardia 69 Surface Beaver 1982 Kimberly Giardia Surface Wildlife 1984 Chilliwack Salmonella 82 Surface and groundwater Human (broken watermain) 1985 Creston Giardia 72 Surface Beaver 1986 Penticton Giardia 362 Surface and groundwater Beaver 1986 Penticton Giardia 109 /3 ,125 Surface and groundwater Beaver 1987 Black Mountain Giardia 60 Surface Wildlife/cattle 1987 Kamloops Campylobacter Surface Wildlife 1988 Near Lytton Salmonella Surface (spring fed) Wildlife 1990 Kitimat Giardia 28 Surface Beaver 1990 Creston Giardia 130 Surface Wildlife 1990 Fernie Giardia 50 Surface (spring fed) Wildlife 1990 West Trail/Rossland Giardia >40 Surface Wildlife 1990 Matsqui Unknown Surface 1991 Barriere Giardia 25 Surface (now groundwater) Wildlife 1991 Granisle* Unknown Surface 1991 Fort Fraser* Unknown Surface 1992 Kaslo Campylobacter 10 Surface Wildlife 1993 Fernie Campylobacter 35 Surface Cattle 1995 Victoria Toxoplasmosis 110/3 ,000 Surface Cats/cougar 1995 Revelstoke Giardia 62 Surface Beaver/wildNfe 1995 Revelstoke Campylobacter 71 Surface Beaver/wildlife 1996 Cranbrook Cryptosporidium 29 /2 ,097 Surface Calves 1996 Kelowna Cryptosporidium 177/ 10,000 Surface Human 1996 Valemount Giardia 10 Surface Wildlife 1997 Princeton Unidentified virus 146 Groundwater Human (sewage break) 1998 Chilliwack Cryptosporidium 19 Surface and groundwater Cattle 1998 Camp Malibu Campylobacter . 26 Surface Wildlife * Suspected outbreak Chapter 2 Table 2-3 emphas izes several key points regarding waterborne i l lness in the province: 1) waterborne i l lness has impacted thousands of people in British Co lumb ia over the past 25 years, 2) the majority of outbreaks (97%) took place in sys tems with surface-water or mixed surfacewater and groundwater sources , 3) 5 9 % of outbreaks were due to protozoan parasi tes, and 4) most outbreaks were assoc ia ted with smal l water sys tems. A s a result of the 29 outbreaks, there were 1,734 laboratory-conf irmed i l lnesses. The true impact; however, was likely much greater as the number of unreported c a s e s in such outbreaks can range from 10 to 1000 t imes those that are reported (Environment C a n a d a , 2001). Of the total number of people affected by an outbreak, only a smal l percentage will see a phys ic ian, and only a percentage of those visits will involve a full microbiological workup (Krewski et al . , 2002; M e d e m a et al . , 2003). A s a result, it is likely that entire outbreaks go unreported, and that the number of conf irmed c a s e s in observed outbreaks significantly underest imates the true impact on public health (Medema et a l . , 2003). From 1986-1998, B C had the highest inc idence of enteric d i sease rates in C a n a d a , with rates ranging from approximately 140 to 210 c a s e s per 100,000 individuals. This is 30-3 0 0 % higher than incidence rates in other Canad ian provinces (BC Provincial Health Officer, 2001). Of the 29 water sys tems involved in the B C outbreaks, 28 der ived their water from surface water sources (or, in some c a s e s , a combinat ion of surface and groundwater)., Approximately 7 6 % of the province's drinking water is derived from surface-water sources , as illustrated in Figure 2-5 (a disproportionately high percentage compared to the rest of Canada) . There are currently over 3,300 water sys tems in British Co lumbia , 96 of which are large municipal sys tems that serve 9 0 % of the province's population. The remaining 1 0 % are served by smal ler private and public sys tems which are much more difficult to monitor, inspect and regulate. In 2003 there were as many as 340 boil-water advisor ies in p lace at one time, indicating that there was sufficient concern regarding water quality to require users of these smal ler sys tems to disinfect their water prior to consumpt ion. 46 Chapter 2 Surface water (public) 69% Figure 2-5 - Percentage of drinking water in British Columbia derived from different sources as of 2001 (BC Provincial Health Officer, 2001). Protozoan parasi tes, particularly Giardia, were responsible for the majority of B C outbreaks over the past 25 years. In most c a s e s , it is bel ieved that contamination was der ived from wildlife sources ; however, severa l c a s e s of l ivestock contamination were a lso observed. Th is is not surprising g iven many of these smal l water sys tems are in rural communit ies. Further, most of the sys tems are suppl ied with surface water, which is much more likely than groundwater to become infected by protozoan pathogens. Most of the 29 outbreaks took place in small water sys tems. Many of these sys tems (30%) did not utilize any form of water treatment or disinfection (no filtration or chemica l addit ives). In several others, chemica l disinfection was accompl ished with chlorine or ch loramines which were ineffective against protozoan parasi tes (Chr istensen, 2003). National and international burden of waterborne disease In C a n a d a , waterborne d i sease outbreaks have been recorded by Health C a n a d a s ince 1974, and from that t ime until 1996 over 200 outbreaks were identified. T h e s e outbreaks resulted in 8,000 conf irmed c a s e s of i l lness (Todd and C h a p m a n , 1974-1996 in Environment C a n a d a , 2001); however, for the reasons descr ibed above this is likely a significant underest imate. Further, this number does not include the impact of the Walker ton, Ontario, and North Battleford, Saska t chewan , outbreaks. In the United States, severa l researchers have attempted to est imate the inc idence of waterborne d i sease in humans. R o s e et a l . (1999) cited an estimate (from the U S C D C ) of up to 900,000 c a s e s of 47 Chapter 2 i l lness and up to 900 deaths resulting from waterborne infections each year. Morris and Levine (1995 in M e d e m a et al . , 2003), in an assessmen t of the waterborne d isease burden in the U S , est imated that 560,000 people may suffer from moderate to severe symptoms, and that 7.1 million people suffer from mild to moderate infection as a result of waterborne d isease . Cons ider ing the many similarit ies between drinking water sys tems in C a n a d a and the U S , it is not unreasonable to est imate that the incidence of waterborne d i sease in C a n a d a could be 1 0 % of that in the U S , on the basis of relative population s ize (Environment C a n a d a , 2001). This suggests that such i l lness affects between 50,000 and 100,000 people annual ly in C a n a d a . Global ly, assess ing the burden of waterborne d isease is more compl icated. The W H O est imates that 1.1 billion people worldwide do not have a c c e s s to c lean water (World Health Organizat ion, 2000). Whi le the impact of this in terms of d isease burden is not known, an extrapolation of data for the United States and C a n a d a , two highly industrial ized countr ies, indicates that it is severe . Pruss et al . (2002) estimate the impact of d isease arising from water, sanitation and hygiene issues to be approximately 4 % of all deaths and 5 .7% of the global d i sease burden. Whi le the above are est imates of the impact of waterborne i l lness, they indicate the magnitude of the waterborne d isease burden at the local , national and international level. P russ et al . (2002) suggest that much of the d i sease burden attributed to water, sanitation and hygiene is preventable. A s descr ibed below, considerable effort is now focused on the development of innovative monitoring technologies and r isk-management f rameworks in order to address this issue. 2.6. Water-quality monitoring Given the strong dependence in B C on surface water as a drinking water source, and the significant potential for contamination of these waters with chemica l or biological agents, it is clear that water quality monitoring is critical to ensuring the safety of drinking water suppl ies. This sect ion descr ibes the common techniques and tools used for detecting and quantifying the contaminants descr ibed above, and includes a review of their effect iveness, advantages and d isadvantages. This is fol lowed by a d iscuss ion of recent ly-developed f rameworks that have been recommended to ensure drinking water quality. The purpose of this sect ion is to illustrate the strengths and w e a k n e s s e s of current monitoring methodologies, and is not intended as a comprehens ive review and descript ion of each avai lable method. For such details, the reader is referred to the Amer ican Publ ic Health Assoc ia t ion 's Standard Methods for 48 Chapter 2 the Examination of Water and Wastewater (Amer ican Publ ic Health Assoc ia t ion , 1999), hereafter referred to as "Standard Methods." 2.6.1. Detection methods Table 2-4 summar izes the common methods used to detect different contaminant types (nutrients, bacter ia, protozoa and viruses) in water. Issues of importance for a given monitoring technique include t imel iness, detection limit and accuracy. T h e s e are addressed in greater detail below. 49 Table 2-4 - Common detection techniques for various water-quality parameters. P h y s i c a l / C h e m i c a l Var iab les Variable/ Organism Method Name Conductivity Temperature pH Dissolved oxygen Chloride Nutr ients Phosphorus Nitrate Nitrite Ammonium Pro tozoa Giardia Bacter ia T ° t a l Coliform Fecal Coliform NA NA NA NA Standard Methods -Method 4130 Standard Methods - Method 4130 Standard Methods - Method 4130 Standard Methods - Method 4130 Standard Methods -Method 4130 E P A 1623 Cryptosporidium E P A 1623 Vi ruses Enteric Viruses Standard Methods - Method 9222 Standard Methods - Method 9222 Standard Methods - Method 9510 Method Description Water volume required Field/ Laboratory Turnaround Time Detection Comments Limit Field sensor Field sensor Field sensor Field sensor Flow injection analysis Flow injection analysis Flow injection analysis Flow injection analysis Measured in situ Measured in situ Measured in situ Measured in situ < 1 ml < 1 ml < 1 ml Flow injection analysis < 1 ml 1 ml Filtration, immunomagnetic 10-100 L separation, immunofluorescence assay Filtration, immunomagnetic 10-100 L separation, immunofluorescence assay Membrane filter technique 100 L for members of the coliform group Membrane filter technique 100 L for members of the coliform group Virus concentration and 2-200 L enumeration Field/laboratory Field/laboratory Field/laboratory Field/laboratory Laboratory Laboratory Laboratory Laboratory Laboratory Laboratory Laboratory Laboratory Laboratory Laboratory Instant Instant Instant Instant 4-6 hours 4-6 hours 4-6 hours 4-6 hours 4-6 hours 36 hours 36 hours 48 hours 48 hours 48-72 hours NA NA NA NA Variable .02 mg-L"1 .05 mg-L"1 .1 mg-L"1 6 mg-L"1 Variable Variable Variable Variable Variable Reagents, specialized equipment and trained technicians required Reagents, specialized equipment and trained technicians required Reagents, specialized equipment and trained technicians required Reagents, specialized equipment and trained technicians required Reagents, specialized equipment and trained technicians required Difficult to estimate % viable cysts Recovery is not 100% Difficult to estimate % viable cysts Recovery is not 100% o Chapter 2 2.6.2. Parameters 2.6.2.1. Physical and chemical variables Temperature, d isso lved oxygen, specif ic conductance or conductivity (a measure of total d isso lved sol ids), and pH are commonly measured in the field using hand-held sensors . A s s e s s i n g these parameters in the field is advantageous as they are quickly measured and can be used to gain preliminary insight into contaminant sources and flow pathways. Water from different stores in a watershed (groundwater, surface runoff) often has characterist ic va lues for these parameters due to the different physical and chemica l characterist ics and p rocesses assoc ia ted with each . For example, groundwater, due to prolonged contact t ime with bedrock or surficial materials, often carr ies a greater d issolved load than surface runoff, and therefore has a higher conductivity (and often a consistently different temperature). Further, surface-runoff sources can be d iscerned based on these variables with agricultural si tes often having higher conductivity than forested sources (Dow and Zampel la , 2000). W h e n monitoring for agricultural contaminat ion, these parameters are valuable proxies for water impairment, but they do not provide a direct measure of contamination or potential risk. 2.6.2.2. Nutrients Nutrient concentrat ions are often measured using colourimetric techniques that require filtered samp les , spec ia l ized laboratory equipment, reagents and trained operators. The time required for analysis is general ly several hours, when sample and reagent preparation are included. Monitoring of surface-water nutrient concentrat ions is a means to detect and quantify the influence of agricultural land use on surface water quality. Natural, or background, levels of N 0 3 " , P and N H 4 + tend to be around 0.1 mg-L" 1 , .005-.02 mg-L" 1 and 0.1-3.0 mg-L" 1 , respectively in mesotrophic sys tems. Thus , va lues above these ranges are often an indication of agricultural contamination (Chapman and K imstach , 1992). As ide from the direct impact of nutrient loading on water quality, nutrients can a lso serve as a surrogate for potential pathogen loading from agricultural operat ions. 2.6.2.3. Bacteria Compared to v i ruses and protozoa, detection and quantification of coliform bacteria in water is relatively easy . Broadly def ined, col i forms are a group of rod-shaped, gram-negat ive bacter ia that include the genera Escherichia, Enterobacteria, Klebsiella, Citrobacter and many others. This group, a lso referred 51 Chapter 2 to as "total col i forms", is often used as an indicator of fecal contamination because many of its members are found in the intestinal tract of warm-blooded organisms. However, many bacter ia in the total coliform group are non-fecal (thriving in the environment), and, therefore, have limited use as indicators of contaminat ion, due to the potential for false posit ives (i.e., they may be present when fecal contamination has not occurred). Instead, the non-fecal bacter ia are used to a s s e s s bacterial growth in drinking-water treatment and distribution sys tems. The fecal (or thermotolerant) col i forms are better indicators of fecal contamination because they are almost a lways der ived from the gut of warm-blooded organisms. Whi le these col i forms may also ar ise from non-fecal sources (e.g., industrial wastes) they are a more specif ic indicator than the total coliform group and are the most commonly used indicator of bacterial pollution (Rosen , 2000; Jam ieson et al . , 2004). Escherichia coli, a member of the fecal coliform group, are specif ic to the intestinal tract and are the best indicators of fecal contamination among col i forms. Whi le some strains are pathogenic (e.g., E. co / /0157 :H7 , descr ibed in Sect ion 2.5.1.1), the majority are not. The most common technique for enumerat ion of coliform bacteria from water samp les is Method 9222 (membrane filter technique for members of the coliform group) outlined in Standard Methods (American Publ ic Health Assoc ia t ion , 1999). The technique involves filtering between 100-1000 mL of water onto a membrane filter, placing the filter into a nutrient-rich growth medium and incubating it at 35 °C or 44.5 °C for 22-24 hours. The two temperatures are used to differentiate between total col i forms and E. coli (35 °C) and fecal col i forms (44.5 °C), and for each temperature a different culture medium is used. After incubation, colonies that have grown on the filter are identified and enumerated and the concentration is given in colony forming units (cfu) per 100 mL. Recent work has focused on the validity of these bacter ia as indicator organisms. Whi le fecal col i forms and E. coli are accepted indicators of fecal contamination, they do not a lways correlate well with viral and protozoan concentrat ions in water (Krewski et al . , 2002). This is primarily due to: 1) dif ferences in susceptibi l i ty to environmental s t ressors, 2) differing methods of reproduction (neither viruses nor protozoa can reproduce outside of a host), 3) different mobil ization and transport patterns and, 4) in treated water, differing levels of resistance to traditional methods of disinfection (Heise and Gust , 1999; Krewski et al . , 2002). Whi le many studies report a significant correlation between indicator bacteria and viral or protozoan pathogens in untreated source waters (e.g., Lecheval l ier et a l . , 1991), several report poor or no correlation with these pathogens (Grabow, 1996; Ehlers et al . , 2005) or with waterborne i l lness (Craun et a l . , 1997). 52 Chapter 2 2.6.2.4. Protozoa Detection methods for Giardia and Cryptosporidium are significantly more expensive, complex and t ime consuming than for bacteria, as protozoa cannot be cultured. The most common method for enumerat ion of these organ isms involves three s teps: 1) filtration, 2) elution and separat ion and 3) enumerat ion. B e c a u s e (oo)cysts tend to be present in relatively low concentrat ions, and because recovery of (oo)cysts is not a lways complete, this method requires the filtration of large vo lumes of water (commonly 10-100 L). All material on the filter is then eluted and centr i fuged. The supernatant fluid is then removed and magnet ic beads are used to separate the (oo)cysts from the remaining detritus. The magnetic beads selectively attach to the (oo)cysts as they are bound with Giardia- and Cryptosporidium-speci f ic antibodies (a process referred to as immunomagnet ic separat ion). The (oo)cysts are then extracted using a magnet, stained with f luorescently labeled ant ibodies and enumerated using f luorescence and differential interference contrast microscopy (US E P A , 2001). This analysis requires approximately 1.5 days (S. Shay , personal communicat ion). Al though this method is currently the standard, it has severa l constraints which limit its w idespread use. T h e s e include the significant costs assoc ia ted with the equipment and materials involved, as well as the person hours required to process each sample (the enumerat ion step can take several hours). Further, recovery of (oo)cysts is often incomplete, meaning that false negat ives are possib le. Molecular techniques such as polymerase chain reaction ( P C R ) are a lso used to detect (oo)cysts and identify live and dead oocysts (Mahbubani et a l . , 1991). Molecular techniques also al low for the identification of specif ic protozoan genotypes which ass is ts in identifying contaminant sources (Ong et a l . , 2002). 2.6.2.5. Viruses Regular screening for v i ruses in surface waters is not currently done, primarily because existing methods are evolving rapidly (American Publ ic Health Assoc ia t ion , 1999). Further, many outbreaks attributed to waterborne enteric v i ruses have been assoc ia ted with significant contamination events involving human sewage , and as a result, can be detected using traditional indicator bacteria. However, because v i ruses are more resistant to environmental degradation and disinfection than bacter ia (Health C a n a d a , 2004a) and because they are increasingly observed in sur face waters (Ehlers et a l . , 2005) and 53 Chapter 2 treated drinking water suppl ies that have met current standards for bacterial water quality (Vivier et a l . , 2004), methods for detection and enumerat ion are continually being refined. Standard methods for the enumerat ion of v i ruses in water involve col lection of a sample of appropriate volume (2-200 L, depending on source), the concentrat ion of v i ruses from that sample and the culturing of these v i ruses (in primate cells or living organ isms, such as mice) to enable identification and enumerat ion. The details of this procedure are beyond the scope of this sect ion, and the reader is referred to Amer ican Publ ic Health Assoc ia t ion (1999) for a complete descript ion. There are several limitations and/or chal lenges assoc ia ted with the detection of v i ruses in water samp les , including: 1) the smal l s ize of virus part icles, 2) low virus concentrat ions in water, and significant variation in concentrat ions, 3) the inherent instability of v i ruses, 4) interference from d isso lved and suspended materials and 5) limitations in current estimation and quantification techniques. Currently, the p rocesses descr ibed in Standard Methods require several days to complete. Therefore, they have limited use as a preventative monitoring tool. However, recent progress in detection techniques has been made using a combinat ion of viral culture and P C R . This has al lowed more rapid identification and enumerat ion of waterborne viruses (Abbaszadegan et al . , 1999; Vivier et al . , 2004). These techniques have yet to be incorporated into Standard Methods. It should be noted that bacter iophages (viruses that infect bacteria) have also been proposed as indicators of fecal pollution. They are useful indicators as they often survive longer that bacter ia, are specif ic to their hosts and, in groundwater, can travel farther than larger pathogens (Abbaszadegan et a l . , 1999; Health C a n a d a , 2004a). To date, eff icacy studies of these organ isms as indicators have yielded inconsistent results due to the many var iables that affect bacter iophage survival in water, as well as a lack of standardizat ion of detection and enumerat ion techniques (Ashbolt et a l . , 2001 ; Health C a n a d a , 2004a). A s a result, they have not gained widespread acceptance as effective indicators. 2.6.3. Monitoring frameworks The approach used to monitor surface water quality in a given watershed is determined by the objectives of the monitoring program. If the objective is to gain insight into speci f ic physical , chemica l or biological p rocesses , a focused program may be implemented to test a def ined research hypothesis. Such programs are often des igned to capture spatial and/or temporal variations in water-quality 54 Chapter 2 parameters at the sca les of interest, and over a specif ied time period, and are therefore unique to each project. Where drinking water is derived from surface-water sources , the objective of a monitoring program is first to a s s e s s the suitability of the source. O n c e the source is chosen , the objective is to detect potential contamination events and support necessary interventions to minimize the risks of human health impacts. Whi le such programs may a lso illustrate cause-effect relationships between activities of human and/or animal populat ions and water quality, the primary goal is the protection of human health. A s a result, such monitoring programs are often more robust and more rigorous than those des igned solely for research purposes. This sect ion provides a review of two methodologies des igned to monitor drinking-water quality that are based on the principles of risk management . The Multi-barrier Approach and the Hazard Analys is and Crit ical Control Point framework, are two methodologies that have been appl ied to drinking-water sys tems in response to several waterborne outbreaks (Krewski et a l . , 2002; Hrudey et al . , 2003). These frameworks demonstrate the importance of source watershed protection and provide the context for the following sect ion (2.7 Spect roscopy and water quality), which descr ibes spectrophotometry as a tool for water quality monitoring. 2.6.3.1. The Multi-barrier Approach The premise behind the multi-barrier approach is that redundancy minimizes the risk assoc ia ted with the failure of any one point in a drinking water sys tem. In other words, no process will have a 0 % risk of failure, and therefore, multiple p rocesses with low failure rates, working in ser ial , offer the greatest chance of protection against a failure in any one process (Anonymous, 2002; Krewski et a l . , 2002; O 'Connor , 2002). In the context of drinking water, this approach is appl ied from "source-to-tap," meaning that multiple barriers are implemented between the source watershed and the end user, as illustrated in Table 2-5. This framework requires that each barrier have unique modes of failure (i.e., a failure in one will not result in a failure in any of the others), and that no barrier be relied upon at the expense of the others, as this could result in a failure of the overall sys tem. 55 Chapter 2 Table 2-5 - The multi-barrier approach for drinking water systems in the context of the "source-to-tap" framework (modified from O'Connor, 2002). Component of DW Risk Barrier Tools System Source watershed Contaminat ion of drinking water Source protection Watershed management plan • Water quality monitoring programs Treatment system • Fai lure of Chemica l and physical • Filtration treatment treatment . Chemica l /UV/o ther • Disinfectant by- disinfection products • Monitoring Distribution system • Regrowth Ongoing monitoring • Chlor ine residual • Infiltration and treatment Line pressure (to prevent ingress) • Monitoring End user • Illness from Publ ic health • Communi ty health contaminated survei l lance, outbreak survei l lance water detection and • Emergency • Outbreak response response plans The multi-barrier approach has been adopted in principal as a framework for drinking water quality protection at the Federa l level in both C a n a d a and the United States, and forms the foundation for guidel ines and legislation related to drinking water sys tems. In both countr ies, gu idance documents have been deve loped to assist municipal and provincial/state governments in the implementation of the multi-barrier framework (US E P A , 2003; Federal-Provincial-Terr i tor ial Commit tee on Drinking Water and C C M E Water Quality Task Group, 2004). A review of the multi-barrier approach would not be complete without mentioning recent advances in the application of more advanced risk management methodologies to drinking-water protection. The Hazard A s s e s s m e n t and Crit ical Control Point ( H A C C P ) framework is a commonly employed risk assessmen t and management tool. This approach, initially deve loped by N A S A and the Pi l lsbury C o m p a n y to ensure the safety of food consumed by astronauts, is based on the principle of controll ing key "risk points" within a system to limit the potential for contaminat ion, and it has been successfu l ly adapted to water sys tems within the multi-barrier framework. The process begins with a full drinking-water system assessmen t in order to identify key hazards or r isks for contamination. Critical control points within this system are then identified so that controls can be implemented to eliminate or reduce the identified risks. E a c h control point is c losely monitored to ensure that quality parameters are maintained below pre-Chapter 2 defined thresholds. Further, p lans for corrective action are def ined in advance to ensure proactive response if thresholds are exceeded , and continual improvement is ensured through a ser ies of performance measurement p rocesses and reviews (Federal-Provincial-Terr i tor ial Commit tee on Drinking Water and C C M E Water Quality T a s k Group, 2004). The H A C C P approach is similar to other "quality management sys tems" such as ISO 9000 and ISO 14000 in that it is des igned to minimize risk by identifying all potential hazards, implementing p rocesses to control them and ensur ing continual improvement through ongoing review and revision. The integration of these strategies into national water-quality management strategies is already taking place in several countr ies. Austral ia 's Drinking Water Guide l ines are often cited as a good example of a comprehens ive , national water quality management system (Anonymous, 2002). Of particular re levance to this study is the emphas is p laced by these f rameworks on source protection, the first in a ser ies of protective barriers from source to tap. Both f rameworks stress a thorough source-watershed review, an assessmen t of potential risks (based on potential contaminants and water resource vulnerability) and continual monitoring in order to minimize the risk of contamination and to support a proactive approach to water quality management . This is represented graphical ly in Figure 2-6, which illustrates the components of source protection as outl ined in Canad ian guidance documents . 57 Chapter 2 Figure 2-6 - A framework for source protection as part of the multi-barrier approach (modified from Federal-Provincial-Territorial Committee on Drinking Water and CCME Water Quality Task Group, 2004) Water-quality monitoring in source watersheds is integral to the multi-barrier approach. A comprehens ive monitoring program provides the information necessary to: 1) identify and quantify contamination risk, 2) determine the vulnerability of source waters, 3) quantify spatial and temporal trends in contaminant loading in order to target remediation and risk mitigation activities and 4) provide information to stakeholders (end users, water sys tem managers , etc.) for dec is ion-making purposes (Federal-Provincial-Terri torial Commit tee on Drinking Water and C C M E Water Quali ty Task Group, 2004). A s outlined above, monitoring for all potential hazards assoc ia ted with agricultural activities is not feasible due to the costs , time required and limitations of many tests in terms of contaminant detection. It is therefore necessary to utilize a few key indicators that provide a realistic approximation of water quality risk. In agricultural watersheds, the greatest potential risks are likely to be assoc ia ted with farming operat ions (unless large wildlife populat ions are a lso present). Thus , monitoring for agricultural inf luence on surface water quality is essent ia l . The following sect ion descr ibes how f luorescence and absorbance properties of surface waters may be used to accompl ish this. 58 Chapter 2 2.7. Spectroscopy and water quality In recent years there has been significant focus on the potential for spectroscopic properties of water to act as water-quality indicators and to assist with source and flow-path identification. This sect ion provides an overview of the theory and principles of spect roscopy as related to water analys is . It then reviews the recent literature concerning the use of this technology to a s s e s s water quality and makes inferences regarding contaminant loading and transport p rocesses in marine and freshwater environments. 2.7.1. Theoretical background Spect roscopy is the study of the absorption or emiss ion of electromagnetic energy by atoms or molecules in order to quantitatively or qualitatively study e lements and compounds. It is based on the principle that e lements or compounds absorb and emit energy at characterist ic wavelengths when irradiated with energy at a known wavelength (or wavelengths). This review focuses on the use of absorbance and emiss ion (f luorescence) spectroscopy as a tool for assess ing water quality using light energy in the ultraviolet (UV), visible (Vis) and infrared (IR) range (between 200 and 900 nm, see Figure 2-7). Infrared V is ib le Ultraviolet > > UV UV UV Infrared Red Orange Yellow Green Blue Violet Longwave Midrange Shortwave (UV-C) Extreme/Fa o o o i- cn t~~ co i n m o i n o m CO Figure 2-7 - UV-visible-near infrared spectrum with wavelengths in nm. W h e n a substance is irradiated with light energy, the incoming energy is converted into internal energy through absorpt ion. This energy is then diss ipated in the form of f luorescence. T h e s e p rocesses are il lustrated schemat ica l ly us ing a Jab lonsk i d iagram (Figure 2-8). Absorpt ion results in the excitation of a molecule to a higher-energy state. The absorbed energy is translated into rotational, vibrational modes or it can elevate the molecule to an excited state, depending on the wavelength of incoming light. 59 Chapter 2 B e c a u s e vibrational and rotational energy levels are relatively c lose together (vibrational levels fall between the rotational levels shown in Figure 2-8), long-wave radiation (i.e., in the infrared range) often results in excitation to these levels. Higher-energy visible or U V radiation results in a transition to a higher electronic state (S1 , S 2 , etc.), when its wavelength, multiplied by P lanck 's constant, equals the energy difference between the ground state and the excited state (Sharma and Schu lman , 1999). W h e n this occurs , energy is almost immediately d iss ipated through a process cal led internal convers ion or vibrational relaxation (a radiat ionless transition to lower energy levels), until the molecule reaches the lowest vibrational level of the first exci ted state. The molecule then returns to any of the vibrational levels of the ground state. This p rocess results in the emiss ion of light energy in the form of f luorescence (Lakowicz, 1999). 4 Vibrational | 3 levels j 0 of 2nd excited i lL state (S2) r, Vibrational levels J _ ot 1st excited A D state (S1) A b s o r p t i o n ^ ? 10"15 s econds Fluorescence 10-9-10-7 Ground }• State Figure 2-8 - Jablonski diagram illustrating the processes of absorption, internal conversion and fluorescence. Times associated with each process are also provided (modified from Lakowicz, 1999). Due to energy loss during this process (through heat, col l isions or vibration), energy is emitted at a longer wavelength (lower energy level) than that of the excitation energy. The difference between excitation and emiss ion wavelengths is termed the Stokes shift. Whi le the entire process takes only a 60 Chapter 2 fraction of a second , the t ime required for each step (absorption, internal convers ion and f luorescence) varies by orders of magnitude as illustrated in Figure 2-8. 2.7.2. Absorbance spectroscopy Absorbance spect roscopy is the qualitative and quantitative study of absorpt ion patterns of compounds , general ly in solution. Subs tances in solution are identified qualitatively according to unique absorption patterns (i.e., the wavelengths at which absorpt ion occurs) . The concentration of a solute can also be determined using absorbance spect roscopy as the amount of energy absorbed by a substance in solution is directly related to its concentrat ion, as determined by the Beer-Lamber t Law, descr ibed below (Lakowicz, 1999). Abso rbance is a d imension less variable that is calculated as : A = log l 0 ( J P 0 /P) where: P 0 = radiant power of incident radiation at a given wavelength P = radiant power of outgoing light The Beer-Lambert law descr ibes how absorbance var ies as a function of the concentration of a substance in solut ion: A-£bc where: £ = molar absorptivity of the substance (wavelength dependent) b = pathlength of the sample container (absorbance increases with pathlength) c = concentration of the subs tance in solution It should be noted that there is some confusion in the literature regarding the nomenclature related to absorpt ion spectroscopy, particularly the terms "absorbance" and "absorption coefficient" (Hu et al . , 2002). Throughout this document, the terms absorbance and absorpt ion refer to the d imension less parameter def ined by the equat ion: A = log 1 0 (Po/P) . The instrument used in U V - V i s absorbance spect roscopy (also referred to as UV-V is -N IR spectroscopy, for near-infra-red) is an absorbance spectrophotometer, illustrated schemat ical ly in Figure 61 Chapter 2 2-9. A variety of lamps are employed as light sources for absorbance spectroscopy, including tungsten, deuterium and xenon. An important characterist ic of the lamp is the spectral range of the emitted light. Both tungsten and deuterium lamps are often used ; however, their spectral ranges are 340-1000 nm and 180-400 nm, respectively, meaning that two light sources are required to span the full UV- IR spectrum. Xenon lamps emit light across the full spectral range (180-1100 nm) and are therefore more versati le and more commonly used (Boeker and van Grondel le , 2001). The wavelength of light reaching the sample is controlled by a monochromater, or filter, which selects the specif ic wavelength for analys is , or adjusts to scan across a range of wavelengths. The sample , commonly in liquid form, is held in a cuvette, ideally made of quartz to allow transmission of U V radiation (plastic cuvettes are not UV-transparent) . Light Excitation Sample Detector Source Monochromater cuvette Figure 2-9 - Schematic representation of an absorbance spectrophotometer. An absorbance spectrum is a plot of absorbance as a function of wavelength (Figure 2-10). The shape of the absorbance spectrum is used to qualitatively identify one or severa l compounds in solut ion, based on the location of absorbance peaks . W h e n several absorbing compounds are present in solution, their absorbance peaks may overlap, thus making it difficult to resolve individual max ima and to detect individual compounds . This issue is addressed using derivative spect roscopy, whereby the derivative (first, second , nth) of the absorbance values is plotted against wavelength (descr ibed in Sect ion 2.7.5.3). Note also that the emiss ion spectrum is commonly a mirror image of the absorbance spectrum as the s a m e energy transitions are favoured for both absorbance and emiss ion (Lakowicz, 1999). However, this is not observed if there are compounds in solution that do not f luoresce (Coble, 1996). 62 Chapter 2 Figure 2-10 - a) Idealised absorbance spectrum and emission spectrum (explained in Section 2.7.3) illustrating the change in absorbance with wavelength and b) typical absorbance spectrum for a surface-water sample, showing decreasing absorbance with increasing wavelength. 2.7.3. Fluorescence spectroscopy Fluorescence or emiss ion spect roscopy deals with the release of energy from a molecule or atom that has been irradiated with a high-energy light source. A s with absorbance, compounds in solution have characterist ic f luorescence signatures which provide both qualitative and quantitative information. The location of f luorescence peaks are often given as exci tat ion-emission pairs, with the first number representing the excitation wavelength, and the second the emiss ion wavelength at which the peak appears . The intensity of a f luorescence peak is measured in f luorescence units which are d imension less, and calculated a s : F = 0o(l-e*c) where: F = f luorescence intensity (j) = quantum efficiency (number of photons emitted as a percentage of those absorbed, with a maximum value of one) / 0 = radiant power (amount of energy assoc ia ted with incoming radiation) £ = molar absorptivity of the substance (wavelength dependent) b = pathlength of the sample container c = concentrat ion of the subs tance in solution 63 Chapter 2 A f luorescence spectrophotometer is schematical ly simi lar to an absorbance spectrophotometer. Similar light sources are used (with xenon lamps being most common) , a monochromater is employed to control the wavelength of radiation directed at the sample and the samp le is conta ined in a quartz cuvette. A second monochromater is used between the sample and the detector to detect f luorescence at specif ic wavelengths. Severa l methods are employed to analyse and represent f luorescence data. Excitation spectra are generated by measur ing the intensity of emiss ion at one wavelength, while varying the excitation wavelength. Emiss ion spectra, on the other hand, are obtained by measur ing the variation in emiss ion or f luorescence intensity ac ross a range of wavelengths, while maintaining a f ixed excitation wavelength (Figure 2-11). The type of spect ra recorded depends on the purpose of the analys is . 800 —] 200 250 300 350 400 450 500 550 600 Emission wavelength (nm) Figure 2-11 - Emission spectra for river water (at three excitation wavelengths) illustrating several features: 1) Rayleigh-Tyndall scattering, 2) Raman scattering and 3) variations in fluorescence intensity based on excitation and emission wavelengths (see text for explanation). Figure 2-11 illustrates several important features of emiss ion spect ra that are relevant when analys ing f luorophores in water. W h e n incident light reaches the water samp le in the cuvette, it is either absorbed or transmitted, as descr ibed above. However, it can also be scat tered by molecules and 64 Chapter 2 particles in suspens ion , resulting in detection of a strong energy signal at the s a m e wavelength as the incoming radiation. Rayle igh and Tyndal l scattering are terms used to descr ibe scattering that ar ises from molecules and part icles, respectively. A second type of scatter, R a m a n scattering, is similar but involves some loss of energy to the molecules in solution. This occurs when a smal l portion of the energy of incoming photons is lost to rotational or vibrational p rocesses within a molecule (but it does not result in a transition of the molecule to a higher excited state). These photons, having lost some energy, are then scattered and detected at a slightly longer wavelength (Lakowicz, 1999). B e c a u s e R a m a n scattering involves only a smal l proportion of incoming radiation, it is weaker than Rayle igh Tyndal l scattering, and usually weaker than the f luorescence arising from the sample (both types of scatter are also illustrated in Figure 2-12). T h e s e forms of scatter are relevant as they can interfere with the analysis of f luorophores in solution, particularly if f luorescent peaks are found near the excitation wavelength. This is addressed either by subtracting the emiss ion spectra of a de- ionized water "blank" or through the application of algorithms to remove scatter peaks and subsequent interpolation of the remaining data (Zepp et a l . , 2004). To fully character ize a water sample using f luorescence spectroscopy, emiss ion spect ra must be obtained over a range of excitation wavelengths. A useful way to represent and analyse these data is to create a matrix of resulting f luorescence intensities (referred to as an exci tat ion-emission matrix, or E E M ) and plot a surface of these va lues, in either two or three d imensions (Figure 2-12). This type of plot clearly illustrates the Rayle igh-Tyndal l line (occurring along the line where excitation equals emission), R a m a n scattering and three f luorescence peaks commonly observed in surface waters (associated with humic material, fulvic material and tyrosine, expla ined in greater detail below). It is a lso very useful for an initial, qualitative assessmen t of compounds in a water sample , as they will have peaks in known regions on the E E M plot (i.e., known exci tat ion-emission pairs). A quantitative assessmen t of concentrat ion can be obtained by measur ing f luorescence intensity in these peak regions. Col lect ion of f luorescence information in this format at the time of analysis is a lso advantageous as it al lows for the data to be ana lyzed in many different ways after col lect ion. This includes E E M ' s , excitation spectra, emiss ion spectra or synchronous-scan spectra (obtained when a sample is irradiated ac ross a range of excitation wavelengths, and f luorescence is detected at Ex+n, where Ex=excitation wavelength and n is a constant integer) (Coble, 1996). 65 Chapter 2 440 420 400 350 400 450 500 Emission Wavelength (nm) 55C Figure 2-12 - Contour plot of an excitation-emission matrix for river water showing the Rayleigh-Tyndall line, Raman line and fluorescence peaks. 2.7.4. Advantages and limitations of spectroscopic techniques The use of spectroscopic techniques for the assessment of water quality has become increasingly common in recent years. This is due to recent technological improvements that have improved the capabil i t ies and performance of f luorescence and absorbance spectroscopy equipment. Spect roscopic techniques offer several advantages over traditional water-quality monitoring approaches, including 1) rapid analysis, 2) low sample vo lumes, 3) high sensitivity, 4) minimal sample preparation (filtering), 5) no requirements for reagents, 6) non-destructive analysis and 7) the potential for real-time monitoring using automated sensors (Reynolds and Ahmad , 1997; A h m a d and Reynolds, 1999; Baker and Curry, 2004). Absorbance scans require very little time after water samples have been filtered, with one scan from 200-900 nm lasting approximately one minute (depending on the scan rate). Absorbance spectrophotometers can also be equipped with mult i-sample trays allowing automated throughput of several samples in ser ies. F luorescence scans require minutes to several hours, depending on the range 66 • Chapter 2 of excitation and emiss ion wavelengths used, and on the desi red signal to noise ratio, or S N R (a lower S N R requires longer data collection t imes at each exci tat ion-emission pair). Very little sample is required for absorbance and f luorescence scans . Most standard cuvettes hold 3-4 mL of sample , with only 2 mL required for standard analys is . Smal ler cuvettes can also be used, requiring only 80-90 u.L of sample . A s a result, the majority of the original sample is avai lable for other ana lyses . Both f luorescence and absorbance scans are highly sensit ive. Depending on the f luorophore, f luorescence scans detect in the range of parts per million (ppm) or parts per billion (ppb). Measures of absorbance can also be used to detect in these ranges; however, f luorescence tends to be more sensit ive due to the relatively lower background s ignal . In other words, when measur ing absorbance, the detector is compar ing the difference between two relatively strong signals of the s a m e wavelength, while a f luorescence system is detecting a relatively weak emiss ion signal against very low background levels (Lakowicz, 1999). Spect roscop ic techniques require minimal sample treatment. Filtering is necessary to remove particulates that could affect light t ransmiss ion; however, no other treatment is necessary. A s many optically active components of interest are d isso lved (see below), filtering does not adversely affect the analys is . Neither f luorescence nor absorbance techniques require the addit ion of reagents, further simplifying the analysis p rocess . Further, both types of scan are non-destructive, meaning that the sample is avai lable for further ana lyses if required. Whi le both techniques offer significant advantages over traditional approaches to water-quality assessment , they do have limitations. These limitations must be accounted for in order to ensure proper interpretation of results. Firstly, f luorescence is sensit ive to severa l environmental var iables, including pH and temperature. C h a n g e s in pH can result in a shift in the wavelength of f luorescence peaks or the elimination or creation of f luorescence due to chemical alterations assoc ia ted with the presence of an acid or base (Sharma and Schu lman , 1999). Temperature also strongly inf luences f luorescence as it directly controls sample viscosity. A higher-viscosity sample results in more molecular col l is ions, thus increasing energy loss in forms other than f luorescence emiss ion (i.e., heat). General ly , higher temperatures result in dec reased f luorescence. Many f luorescence spectrophotomers are equipped with temperature-stabi l izing cuvette holders to address this problem. 67 Chapter 2 Quench ing is a process by which quantum efficiency of a fluorophore is reduced, through one of several mechan isms. Col l is ional quenching involves loss of excitation energy from an excited molecule as heat (rather than f luorescence) as a result of contact with other molecules. Static quenching occurs when a f luorophore forms a non-f luorescent compound as a result of interactions with another substance in solution. Resonance energy transfer is another form of quenching whereby an exci ted molecule transfers energy v ia electronic coupl ing to another molecule in a ground state, without a coll ision actually taking place (Lakowicz, 1999). Inner-filtering is another type of quenching arising from the presence of other absorbing compounds in solution (or high concentrat ions of the compound of interest). This can result in attenuation of incoming or emitted radiation, thus altering the f luorescence intensity measured at the detector. The absorbance of excitation energy and emitted radiation are referred to as primary inner-filtering and secondary inner-filtering, respectively. These p rocesses result in a non-l inear relationship between concentrat ion and f luorescence (i.e., very high f luorophore concentrat ions could actually produce very low f luorescence intensity) and thus, must be corrected for when significant absorbance is measured in a given sample . O h n o (2002) suggests a threshold of 0.3 absorbance units at 254 nm as the level beyond which correction for inner-filtering is required. The primary limitation assoc ia ted with absorbance spect roscopy is the potential for overlapping or adjacent absorbance peaks assoc ia ted with several compounds in solut ion. This overlap results in compound peaks that are broader and higher than those for individual compounds , and reduces the specificity of the scan unless mathematical techniques are appl ied to the data. 2.7.5. Applications of spectroscopy in environmental research Fluorescence and absorbance spect roscopy have been increasingly appl ied to environmental research in marine, freshwater and wastewater sys tems. This sect ion provides an overview of the appl icat ions of f luorescence and absorbance spect roscopy in environmental research. Of particular re levance to this study are the unique absorbance and f luorescence characterist ics of d issolved organic matter (DOM). This sect ion begins with an overview of these characterist ics. This is fol lowed by a review of the relevant literature. A s much of the recent work involving these techniques has been conducted in marine and freshwater environments, this review is divided broadly under these two headings. 68 Chapter 2 2.7.5.1. Spectroscopic properties of D O M Dissolved organic matter is def ined as organic material that p a s s e s through a 0.45 um filter (McDonald et al . , 2004), and is produced from the degradat ion of terrestrial and aquatic plant material. It is re leased through the chemica l , biological or physical breakdown of these materials and can be produced either autochthonously (i.e., in a water body through microbial metabolism) or al lochthonously (i.e., on the land surface) and transported to surface waters during storm events. D O M is difficult to character ize chemical ly because it represents a broad range of compounds with varying molecular weights. A s a result, traditional analytical techniques are complex and require large sample vo lumes (Leenheer and Croue , 2003). A significant proportion (30-60%) of D O M in surface waters is compr ised of humic substances (Figure 2-13), def ined as non-volati le, coloured, polyelectrolytic ac ids that range in molecular weight from 500 to 5000 (Thurman, 1985). The primary constituents of humic matter are humic and fulvic ac ids which are chemical ly similar; however, fulvic ac ids have lower molecular weights (500-2000) and are soluble in water at any p H . Humic ac ids are soluble in water only under alkal ine condit ions and are larger (molecular weights greater than 2000). Due to their solubility, fulvic ac ids are general ly more mobile than humic acids and tend to be found commonly in surface waters, while humic acids occur most commonly in the sol id phase in soi ls (Tipping, 2002). A s illustrated in Figure 2-13, D O M also contains amino acids and proteins, which have unique f luorescence properties, and are thought to produce protein-like f luorescence commonly observed in natural wa te rs . . 69 Chapter 2 Hydrophilic acids / / Carbohydrates Humic materials Carboxylic acids Amino acids Hydrocarbons Figure 2-13- Average composition of dissolved organic matter in river water with 5 mg-L" dissolved organic carbon. Modified from Thurman (1985). Absorbance is an effective indicator of chromophor ic (light absorbing) D O M concentrat ion in natural waters (Green, 1992; Del Casti l lo et al . , 1999; Kowa lczuk et a l . , 2003 ; McDona ld et al . , 2004). Chromophor ic D O M ( C D O M ) refers to that portion of the D O M pool that absorbs light in the U V - A , U V - B and visible wavelength ranges (Blough and Del Vecch io , 2002). Al though detection of individual compounds in solution using absorbance can be difficult due to broad and over lapping absorbance peaks , the s lope of the absorbance curve (referred to as spectral s lope, or S) can be used to character ize the composi t ion and source of C D O M in natural water samp les (Carder et a l . , 1989; Twardowsk i and Donaghay, 2001; Blough and Del Vecch io , 2002; Kowalczuk et a l . , 2003 ; Twardowsk i et a l . , 2004). This s lope is often used as a proxy for the proportion of C D O M compr ised of fulvic vs . humic ac ids as fulvic ac ids exhibit higher absorptivit ies at shorter wavelengths ( -280 nm) than humic ac ids which absorb in the blue range of the spectrum (-440 nm) (Thurman, 1985). A s descr ibed by Carder (1989) and Blough and Del Vecch io (2002), s lopes of -0 .02 or greater reflect dominance of low-molecular-weight fulvic material. Lower s lopes (-0.010) represent C D O M dominated by humic ac ids . v Analys is of f luorescence properties of C D O M are a lso c o m m o n due to greater sensitivity of f luorescence techniques (Blough and Del Vecch io , 2002). Two distinct c l asses of f luorophores are commonly found in C D O M in natural waters: humic-l ike and protein-like materials (Leenheer and Croue , 2003). The f luorophores are referred to as "humic-l ike" and "protein-l ike" b e c a u s e f luorescence for the former occurs at the s a m e excitat ion/emission pairs as f luorescence for isolated humic and fulvic 70 Chapter 2 materials. F luorescence for the latter occurs at the s a m e excitat ion/emission wavelengths as the known aromatic amino acids tyrosine and tryptophan. The optically active components of D O M , therefore, are a potentially useful indicator of agricultural inf luence as they are present in all natural waters, but their composi t ion and concentrat ion vary depending on source, dominant vegetation type, soil type, land use and microbial activity (Kirk, 1994; Leenheer and Croue , 2003). 2.7.5.2. Spectroscopy in marine research O n e of the first appl icat ions of modern, high-sensitivity spect roscopic techniques in marine research was conducted by Cob le et al . (1990), who used f luorescence spectroscopy to analyse C D O M extracted from seawater samp les col lected at depths ranging from the surface to 375 m. They noted three f luorescence max ima in E E M ' s from these samp les , referring to them as regions A ( E x / E m 5 = 260/435), B (Ex /Em = 285/335-345) and C (Ex /Em = 345/445-450). T h e s e max ima were attributed to f luorophores in C D O M assoc ia ted with different sources (river, marine, and soil) and their relative intensities were observed to vary with depth, suggest ing that this technique could be used to dist inguish C D O M sources , transport pathways and marine mixing p rocesses . In a continuation of this work, Cob le (1996) a s s e s s e d f luorescence from seawater and riverine samples , and illustrated a progressive change in the location of humic-l ike f luorescence peaks as freshwater became mixed with estuarine waters. This transition was attributed to a change in source of humic material from terrestrial to marine, particularly in eutrophic waters. Del Cast i l lo et al . (1999) used absorbance and f luorescence spect roscopy to determine the nature and extent of the contribution of the Or inoco River p lume to high concentrat ions of d isso lved organic carbon (DOC) and C D O M in the Car ibbean S e a . Both techniques were used to character ize freshwater organic matter, which compr ised a significant proportion of that observed in marine samples . The identification and quantification of C D O M was of particular importance due to its inf luence on the location of the photic zone , and this was accompl ished using absorbance at 300 hm (A 3 0 0 ) as an index of C D O M concentrat ions. F luorescence was also used to identify changes in C D O M composi t ion with increased mixing between the marine and riverine endmembers by assess ing the location of f luorescence peaks for different humic materials. Const i tuents of terrestrial (riverine) C D O M changed very little below salinities of approximately 30, but shifts in emiss ion max ima were observed above this threshold (higher 5 E x / E m = Exci tat ion/Emission wavelength at which f luorescence is observed, and is given in nm. 71 Chapter 2 salinities indicate dominance of marine water), suggest ing significant mixing and a change in the dominant source of C D O M from terrestrial to aquatic. Similar results were observed at or near this threshold by G u e g u e n et al . (2005) in the Western Arctic O c e a n and by Jaffe et a l . (2004) in a study assess ing C D O M sources and mixing p rocesses in an estuary in southwestern Florida. In the latter study, a f luorescence index of E m . 4 5 0 / E m . 500 at an excitation of 370 nm (first used by Mcknight et al . (2001)) was also successfu l ly used to differentiate terrestrial vs . marine C D O M , with higher values being indicative of marine production. A lso , because terrestrially-derived C D O M underwent conservat ive mixing in the estuary, it was used to determine the degree to which marine and fresh water mixing occurred in geomorphological ly compartmental ized sub-regions. C h e n and Gardner (2004) used a similar approach to a s s e s s C D O M cycl ing and physical mixing p rocesses in the Gulf of Mex ico using a submers ib le pump linked directly to a shipboard laboratory. This al lowed for analysis of spectroscopic var iables along multiple transects and profiles to quantify changes in C D O M concentrat ion and source both spatially and with depth. Their analys is indicated that coastal C D O M was dominated by terrestrial material from the Miss iss ippi and Atchafa laya Rivers , that end-members from a given river can vary over time due to seasona l effects and that p lumes from each river could be identified and t raced based on their unique spectroscopic properties (the Atchafa laya River has significant wet land inf luence compared to the Mississ ippi) . This technique has a lso been useful for identifying significant marine C D O M sources at depth that are assoc ia ted with bacterial breakdown of organic detritus at the pycnocl ine (Chen et al . , 2004). G iven the extensive information contained within E E M plots (potentially several thousand data points), and the numerous optically active compounds found within C D O M , multivariate statistical procedures are popular for E E M interpretation and analys is . Person and Wedborg (2001) appl ied principal component analys is (PCA) to E E M ' s to identify structural di f ferences related to varying C D O M sources in the Baltic S e a . By first normalizing all s cans to the s a m e f luorescence intensity at one excitat ion/emission wavelength pair, it was possib le to remove concentrat ion effects from the data sets and isolate structural dif ferences. Plotting the resulting P C A scores against the first and second principal components resulted in clustering of sample sites according to dominant humic source (terrestrial vs. aquatic). This technique al lowed the authors to make inferences regarding mass mixing p rocesses between the Baltic S e a (dominated by terrestrial humic materials from large riverine sources) and the Atlantic O c e a n (with relatively little terrestrial humic input) due to their characterist ic E E M structures. A 72 Chapter 2 similar approach was used by B o e h m e et al . (2004) on a large data set of over 600 samp les to identify C D O M end-members and mixing patterns in the Gulf of Mex ico. It is worth noting that several authors have suggested that a three-dimensional multivariate model cal led Paral lel Factor Analys is ( P A R A F A C ) is more appropriate for analysis of E E M ' s , as P C A is essential ly a two-dimensional technique with limitations in terms of data treatment and identification of underlying spect ra of f luorophores in solution (Jiji et al. , 1999; S tedmon et a l . , 2003 ; Fulton et al . , 2004). This approach al lows the extraction of peaks assoc ia ted with individual C D O M components that might otherwise be masked by overlapping peak f luorescence wavelengths. 2.7.5.3. Freshwater systems Recent ly, f luorescence and absorbance properties of organic matter in freshwater environments have been investigated as tools to a s s e s s : 1) paleoenvironmental trends, 2) hydrological f lowpaths and mixing p rocesses and 3) contaminant loading and source tracking. Baker et al . (1993) identified annual , luminescent bands in spe leothems (stalactites and stalagmites), and noted that luminescence, from humic and fulvic ac ids derived from overlying soi ls, could provide an accurate proxy of paleoprecipitat ion. This approach has been appl ied e lsewhere with similar results (Baker et al . , 1999; Proctor et al. , 2002). Wol fe et a l . (2002) appl ied the f luorescence index of McKnight et a l . (2001) to humic materials extracted from sediment cores from lakes in the Co lorado Front Range (core sub-samples were subjected to two wet chemical extraction techniques in order to isolate humic materials). Using this technique, it was possib le to trace changes in C D O M sourcing (terrestrial vs. aquatic) over t ime and thereby reconstruct a pattern of eutrophication resulting from anthropogenic nutrient loading in the latter half of the 2 0 t h century. These observat ions were val idated by compar ing this index to other accepted measures of eutrophication (diatom assemb lages and C : N ratios), thus indicating the potential of this index as a proxy for paleoenvironmental change. A s demonstrated in the marine studies descr ibed above, C D O M has the potential to serve as a conservat ive tracer, which, in freshwater sys tems is useful for assess ing f lowpaths and the contributions of different water sources to stormflow. Newson et al . (2001) identified distinct source areas in a watershed in northern Eng land and a s s e s s e d their contributions to catchment flow based on the luminescence properties of unique organic components of different soi ls. Baker and Spencer (2004) used absorbance and f luorescence to character ize changes in C D O M along the entire profile of the River Tyne 73 ^ Chapter 2 in a large ( -3000 km 2 ) catchment. Disso lved organic matter contributions from three main tributaries were identifiable and quantif iable based on unique organic matter signatures assoc ia ted with a peat-dominated source area, a relatively undeveloped upland catchment and a deve loped catchment with treated sewage d ischarges. Due to the colour assoc ia ted with peaty soi ls, absorpt ion at 340 nm and D O C concentrat ions were used to identify water from the first source area. Protein-l ike f luorescence, which was low in undeveloped catchments, showed significant increases assoc ia ted with sewage treatment d ischarge, and served as a useful signature of water from the third source area. Further, these characterist ics could be used to dist inguish between C D O M sources in estuarine samp les at the outlet of the watershed, indicating the value of this approach in source apportionment and studies of large-scale C D O M cycl ing. Ka tsuyama and Ohte (2002) used f luorescence intensity (in conjunction with d isso lved organic carbon and S i 0 2 ) as a tracer to quantify the relative input of saturated throughflow, non-saturated throughflow and rainfall to stormflow in a forested headwater catchment in Japan . Us ing end-member mixing analysis ( E M M A ) (Hooper et a l . , 1990), they demonstrated that non-saturated throughflow dominated stormflow and that there were minimal interactions between this zone and the saturated zone during storm events. Contaminant detection and source identification are areas where spect roscopy offers significant advantages over other techniques for reasons ment ioned above. Recent research has focused on refining these techniques to aid in detection of several types of contaminants, including hydrocarbons, pest ic ides, bacteria, industrial outflows, landfill leachate and agricultural effluent. Jiji et al . (1999), in a laboratory experiment, demonstrated that f luorescence spect roscopy combined with parallel factor analysis could be used to detect and quantify pest ic ides and polycycl ic aromatic hydrocarbons (PAH's ) in solution with detection limits of 1.1-13 ppb and 0.2 ppb, respectively. G i a n a et al . (2003) used f luorescence spect roscopy to identify bacteria in suspens ion and correctly c lassi f ied organ isms to one of three spec ies by applying P C A to E E M data. Other studies have a lso demonstrated that f luorescence techniques can be used to detect the influence of landfill leachate (Baker and Curry, 2004; Baker , 2005a) and industrial effluent (Baker, 2002a) on surface water quality. Resea rch in the a rea of sewage contaminat ion indicates that absorbance and f luorescence spectroscopy can be appl ied to process control in sewage treatment plants and environmental monitoring. Reynolds and A h m a d (1997) descr ibed the correlation of absorbance at 254 nm to b iochemical oxygen demand (BOD) in treated and untreated sewage , and several studies have utilized this measure to identify sewage treatment plant d ischarge to surface waters (Baker, 2001; Baker and Spencer , 2004). 74 Chapter 2 Similarly, A h m a d and Reyno lds (1999) noted that f luorescence at E x / E m = 248/350 is strongly correlated to B O D , and proposed this as a measure for on-line p rocess control in S T P ' s . Ga lapa te et al . (1998) used synchronous-scan f luorescence spect roscopy to identify a characterist ic f luorescence peak for sewage between 512 nm and 531 nm in the laboratory, and used these peaks to detect sewage effluent in Kurose River (Japan) downstream from a S T P (no peaks were observed in samp les taken upstream of the S T P ) . A f luorescence ratio (tryptophan-like to fulvic-like f luorescence intensity) was used by Baker (2001) to identify sewage contamination in rivers inf luenced by S T P d ischarge, and by Baker by et a l . (2003) to identify sewage and grey-water contamination in surface waters of an urbanized catchment in northeastern Eng land. Of re levance to this thesis is the use of spect roscopy in the identification of agricultural inf luence on water quality. F luorescence and absorbance have proven to be useful indicators of agricultural influence on surface waters and groundwaters. Baker (2002b) first examined f luorescence properties of farm wastes by analyz ing samples of si lage liquor (effluent from stored, wet g rasses that have fermented), pig and cattle slurry (liquid manure) and sheep barn wastes. E a c h waste type had high tryptophan concentrat ions and was discernible by its tryptophan:fulvic-like f luorescence intensity ratio, with s i lage effluent having a value >20 and the slurries and sheep barn wastes having lower va lues (-2-5 and -0 .5 -4 .0 , respectively). A s these values are higher than those observed in natural river waters (generally <1.0) this illustrates the potential of this index as an indicator of agricultural contamination. Recent work has been conducted to a s s e s s the link between protein-like f luorescence observed in natural waters and bacterial concentrat ions (Elliott et a l . , 2006b; Elliott et a l . , 2006a). T h e s e studies illustrated that both tyrosine-l ike and tryptophan-like f luorescence are found in isolated bacterial cultures. Thus , bacterial concentrat ions contribute at least partially to protein-like f luorescence observed in natural waters. Recent studies have a lso been a imed at d iscerning C D O M der ived from forested vs. agricultural sub-catchments in agricultural watersheds. S tedmon et al . (2003), col lected surface water samples from several sub-catchments in the Horsens watershed in Denmark to determine if distinctive f luorescence patterns could be detected for forested, agricultural and estuarine water. Using P A R A F A C analys is , they extracted five components that correlated with f luorescence peaks for terrestrial humic-l ike material and tryptophan-like material. The extraction of five components for these two broad categor ies of C D O M suggests that each is compr ised of multiple organic components that are controlled by different 75 Chapter 2 processes . Us ing these components , S tedmon et al . (2003) deve loped a fingerprint for source areas based on their relative abundance. A follow up to this study (Stedmon and Markager, 2005) involved a more extensive data set (1,200 samples) , al lowed further refinement of the P A R A F A C model and resulted in the extraction of eight components (four terrestrial, two anthropogenic/agricultural and two protein-like). Aga in , C D O M sources could be identified by the relative intensities of each component . Forested sites tended to be character ized by the highest f luorescence intensities, and relative peaks in emiss ion for humic-l ike materials. Agricultural sites showed lower f luorescence intensities (likely due to U V and microbial degradat ion of C D O M ) , but relatively higher intensities for fulvic-like and tryptophan-like materials. Co-variabi l i ty between the components was ana lysed by plotting each against the other, and five groupings were observed in the plotted data (sites dominated by forest, >50% agriculture, >75% agriculture, estuarine si tes, and lakes). T h e s e plots a lso illustrated which C D O M components were likely controlled by similar physical , chemica l and biological p rocesses , as they showed strong linear correlations. Finally, recent work by Ohno et al . (2006) represents one of the first rigorous assessmen ts of the properties of C D O M isolated from plant residues and agricultural amendments (beef, dairy, poultry and pig manures), and provides further support for the .observed strong correlation between spectroscopic properties and agricultural inf luence. Th is assessment , which compared plant- and manure-der ived C D O M , noted that C D O M derived from beef and poultry manures was character ised by significantly higher molecular weights than that der ived from plant residue, it was also observed that C D O M derived from beef and poultry manure had the highest molar absorptivit ies (at 280 nm) when compared to plant-derived C D O M . These studies indicate that C D O M source tracking in agricultural watershed is possib le, and that agricultural influence on surface waters can be detected. Absorbance spectroscopy has significant potential as a tool to detect agricultural influence given its effect iveness in quickly and accurately determining N 0 3 " concentrat ions (the N 0 3 " ion absorbs strongly in the 210-220 nm range). One of the chal lenges assoc ia ted with this technique is that readings can be confounded by interferences from other compounds in solution (metals, d isso lved organic matter, etc.) that a lso absorb in this wavelength range. This results in broad absorbance bands as a result of overlapping peaks (Crumpton et a l . , 1992; Twardowski et a l . , 2004). This issue can be addressed either by process ing water samples to remove interfering compounds or by correcting the data for the interference. Renn ie et a l . (1979) accompl ished the former for raw water, treated water and wastewater 76 Chapter 2 samples by adding sodium hydroxide (to increase pH), and pass ing them through a carbon filter to remove C D O M and d issolved iron and manganese . Measurements of absorbance at 210 nm for these p rocessed samp les resulted in N 0 3 " concentrat ions that did not differ significantly from establ ished methods. The latter method, involving a mathematical correction for absorbing subs tances , has traditionally been used in organic-r ich waters. Abso rbance at another wavelength where N 0 3 " absorbance is negligible (several have been used) is measured as a proxy for C D O M concentrat ion. This value is then subtracted from absorbance at 210 nm to determine N 0 3 " concentration (e.g., Thompson and Blankley, 1984). A third method for address ing interferences from other compounds involves the calculation of derivatives of the original absorbance spectra. The first derivative def ines, the rates of change at each wavelength along the original spect rum, and in so doing minimizes broad featureless peaks . The second derivative calculates rates of change of the first-derivative curve, thus emphas iz ing sharper peaks in the original absorbance spect rum. A s illustrated in Figure 2-14, the s e c o n d derivative of the original absorbance spectrum produces a peak at - 2 2 4 nm, allowing the extraction of the N 0 3 " s ignal (Cahil l , 1979; Suzuk i and Kuroda, 1987). c J3 si < 0.008 r— 0.004 -0.004 1 — -0.008 v u c CO .Q O CO J3 CO •D 400 500 600 Wavelength 700 800 Figure 2-14 - Absorbance spectrum for EC-4 collected on December 20, 2004 (dashed) and second-derivative of the same spectrum (solid). Note second-derivative peak at 224 nm. 77 Chapter 2 Severa l authors have used this technique to quantify N 0 3 " concentrat ions in water samples . Crumpton et al . (1992) demonstrated that this approach could be used to determine N 0 3 " content in surface-water samples without treatment (i.e., without extraction or digestion). Ferree and Shannon (2001) found that absorbance spect roscopy could accurately a s s e s s N 0 3 " concentrat ions in wastewater using N 0 3 " sp iked samples and compar ison with ion chromatography. Kar lsson et al . (1995) observed similar results when analyz ing unfiltered wastewater samp les for N 0 3 " concentrat ions ranging from .5-13.7 mg-L" 1 , illustrating the potential for real-time process control in S T P ' s . 2.8. Current gaps and opportunities Due to increased awareness regarding the risks assoc ia ted with waterborne d isease , and several highly-publicized waterborne outbreaks, substantial resources have been directed at improving our understanding of i ssues related to water contaminat ion, water-quality monitoring and risk management . Signif icant progress has been made in each of these areas; however, several gaps still exist. Environment C a n a d a (2001) descr ibed several "knowledge needs" related to waterborne pathogens, including a need to improve the t imel iness and effect iveness of pathogen detection techniques, and to identify contamination "hotspots" in source watersheds through improved monitoring techniques. The s a m e report a lso descr ibed the need for improved monitoring of nutrient contributions to surface waters and groundwaters. The B C Provincial Health Off icer 's annual report on drinking-water quality made several recommendat ions to improve the province's ability to manage water related health risks. These recommendat ions included rigorous testing and analys is of new surface-water sources , improved monitoring of existing sys tems and improved contaminant source identification. Further, Krewski et al. (2002) s t ressed the need for water-quality monitoring programs to detect contamination events in a timely fashion so interventions can be implemented to minimize public health risk. G iven the observed increase in agricultural intensity in the L F V and a growing demand for drinking water arising from regional population growth, an improved understanding of the links between agricultural land use and surface-water quality is critical. This thesis attempts to address the above gaps by identifying the links between cl imate, hydrology, land use and water quality in watersheds with differing types and intensities of agriculture. The primary objective is to identify the land use activities and meteorological/hydrological condit ions leading to the greatest risk of water contamination in order to 78 Chapter 2 support a more proactive approach to water-quality risk management . Further, this thesis attempts to address the current gaps related to monitoring technologies (t imeliness, sensitivity, etc.) by assess ing the utility of spect roscopic tools for monitoring agricultural influence on water quality. 79 Chapter 3 3. Site descriptions and methods 3.1. Introduction This chapter descr ibes the research approach used in this thesis, the field methods employed for collection of meteorological , hydrometric and water-quality data, and laboratory methods employed for microbiological, chemica l and spectroscopic analys is of water samp les . 3.2. Site selection and overview R e s e a r c h for this thesis was conducted in three catchments in the L F V , the Hatzic, Elk Creek and Sa lmon watersheds (Figure 3-1). The criteria for evaluating and selecting these watersheds were des igned to ensure the field sites would meet the objectives of the research, and were as fol lows: 1) land use in the watersheds must be dominated by agricultural activities, 2) both crop and l ivestock agriculture must be present, 3) the watersheds must represent differing intensities of agricultural activity and 4) they must be c lose enough to Vancouver to enable frequent sampl ing and equipment maintenance. The three watersheds met these criteria, with the Elk Creek watershed representing the higher-intensity agricultural site. The Sa lmon watershed was added later in the study to augment the evaluation of spectroscopic techniques. The watershed is dominated by agriculture, but unlike the Hatz ic and Elk Creek watersheds, it offered the opportunity to evaluate f luorescence and absorbance character ist ics of nitrate-rich groundwaters. 80 Chapter 3 123 W W 122'3p'0"W 122WW I 1 T—" 123WW 122"30'0"W 122*0'C"W Figure 3-1 - Lower Fraser Valley with locations of three study watersheds. The climate of the L F V is highly seasona l , with the majority of annual rainfall occurr ing between October and April (Figure 3-2). During the wet season , the region is dominated by low-pressure, marit ime sys tems del ivered by prevail ing westerly winds (during these months the Wester l ies shift southwards to an average latitude of 45 0 N). Upon landfall, these relatively warm, moisture-r ich air m a s s e s are driven to higher elevat ions by coasta l mountains, resulting in the observed peak in precipitation during this t ime. Occas iona l ly , colder, northern air m a s s e s meet these maritime sys tems and produce snowfal l . However , on average, snow accounts for only 4 % of annual precipitation. Pro longed precipitation during these months results in elevat ion of the water table, which frequently leads to dra inage problems and flooding on agricultural land (Bertrand et al . , 1991). During summer months, the Wester l ies shift north to an average latitude of 55 0 N, resulting in warmer, drier condit ions ac ross the region. A precipitation gradient is observed ac ross the L F V , with mean values increasing from southwest to northeast. A similar trend occurs for mean temperatures as the moderating maritime influence dec reases inland (Bertrand et a l . , 1991). 81 Chapter 3 300 250 Mean Precipitation (mm) Mean daily temp, (max.) Mean daily temp, (min.) Mean daily temp, (avg.) S 200 H c o 150 m 100 50 30 20 10 -10 -20 Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec. Month Figure 3-2 - Mean monthly precipitation and mean daily temperatures (minimum, maximum and average) at Chilliwack (near the Elk Creek watershed) from 1971 - 2000 (Environment Canada, 2005). Table 3-1 provides a compar ison of the Hatzic, Elk Creek and S a l m o n watersheds in terms of area, land use, l ivestock densit ies and nutrient surp luses. It should be noted that the animal numbers are for the C e n s u s of Agriculture Enumerat ion Areas (EA's) within which the watersheds fall. In each case , there is agricultural land in the E A that falls outside the bounds of the watershed. However , these numbers provide an indication of total animal numbers and density. Agricultural intensity, and the potential for impairment of surface waters and groundwaters is also il lustrated using nutrient surplus calculat ions from Schre ier et a l . (2003). 82 Chapter 3 Table 3-1 - Comparison of the Hatzic, Elk Creek and Salmon watersheds in terms of size, land use and stocking densities (modified from Schreier et al., 2003). "X" denotes no data available. Variable Hatzic Elk Creek Sa lmon Genera l A r e a (km*) 84 33 80 Forest (%) 68 71 5 Agriculture (%) 14 21 45 Urban/residential (%) 11 2 36 Other 7 6 14 L ivestock 1 Chickens (#) X 1,156,486 942,708 Catt le (#) 7,841 23,766 5,221 Pigs (#) X 22,835 4,409 S h e e p (#) 568 1,351 1,209 Horses (#) 210 499 1,739 Stocking density ( A U E / h a ) 2 1.25 11.04 3.9 Nutrient Nitrogen 13 132 18 surp luses (kg/cropped ha) Phosphorus 16 65 30 1 - L i v e s t o c k d a t a b a s e d o n C e n s u s of A g r i c u l t u r e e n u m e r a t i o n a r e a s , w h i c h i n c l u d e a g r i c u l t u r a l l a n d o u t s i d e o f w a t e r s h e d b o u n d a r i e s . T h e s e v a l u e s a r e p r o v i d e d to i l l u s t r a t e a n i m a l d e n s i t y o n l y . 2 - A U E = A n i m a l U n i t E q u i v a l e n t , a v a l u e b a s e d o n m a n u r e p r o d u c t i o n p o t e n t i a l (1 b e e f c o w = 1 A U E ) . T h e a r e a v a l u e u s e d i n t h e c a l c u l a t i o n w a s d e r i v e d b y s u b t r a c t i n g a g r i c u l t u r a l a r e a u n d e r c r o p s f r o m t o t a l a g r i c u l t u r a l a r e a f o r t h e e n u m e r a t i o n a r e a . 3.2.1. The Hatzic watershed The Hatzic watershed (Figure 3-3) is located east of the city of M iss ion , on the north s ide of the Fraser River. It is dominated by low-mid intensity agriculture consist ing primarily of l ivestock farms (mainly dairy and beef cattle) and fruit and greenhouse operat ions. The watershed covers approximately 83 k m 2 and is bordered to the east, west, and north by steep mountain s lopes . Low-lying areas are predominantly agricultural with some residential developments and numerous recreational properties (on Hatzic Lake). The watershed is drained by Patt ison Creek and Lagace Creek , both of which flow into Hatzic S lough on the val ley floor. Severa l other agricultural s loughs also flow into Hatz ic S lough before it reaches the lake. 83 Chapter 3 1 2 2 ° 1 8 ' 0 " W 1 2 2 ° 1 2 ' 0 " W I A Climate station 1k Hydrometric stations O Sampling sites o | _ C M 0 0.5 1 1.5 2 ] Kilometers Contour Interval = 20 m 2 2 1 8 ' 0 " W 1 2 2 ° 1 2 ' 0 " W Figure 3-3 - Map of the Hatzic Valley watershed showing topography, major streams, sampling sites and hydrometric and climate stations. The watershed drains into the Fraser River through four f loodboxes and a pump station located near the outlet of Hatzic Lake. During heavy rains and the spring freshet, these f loodboxes remain c losed and drainage is controlled by the pump station. The pump capacity is often exceeded and several major 84 Chapter 3 f loods in low-lying agricultural and residential areas have occurred over the past 50 years (Associated Engineer ing, 1992). The problem is exacerbated in upstream (northern) areas of the watershed where channel avuls ions resulting from stream aggradation are common . This often leads to flooding of agricultural land. Severa l large, active landsl ides in the headwaters have been est imated to supply 5,000 - 10,000 m 3 of sediment per year to these streams (Associated Engineer ing, 1992), and as a result, a long-term channel dredging initiative has been undertaken to mitigate f lood risk. S u c h flooding is of concern in agricultural watersheds as it directly links s t reams with manure and fertilizer sources on agricultural f ields. 3.2.2. The Elk Creek watershed The Elk Creek watershed is located near the city of Chi l l iwack on the south s ide of the Fraser River. The area of the watershed is approximately 33 k m 2 , and the southeastern portion is dominated by the steep s lopes of Elk Mountain (1,429 m) and Mt. Thurston (1,626 m). Much of the cropland in the watershed is under corn, and l ivestock farming is dominated by cattle and ch icken operat ions. Ch icken numbers in the East Chi l l iwack E A (of which the Elk Creek watershed is a part), have increased drastically in the past 20 years from approximately 402,000 in 1986 to over 1.1 million in 2001 (Schreier et al . , 2003). Expans ion of urban and recreational land uses is a lso taking place, primarily on the upland s lopes located in the southern half of the watershed. From 1905 to 1998, the Elk Creek served as a primary water supply to the City of Chi l l iwack, but was decommiss ioned after Giardia and Cryptosporidium were detected in surface water samples (water treatment to that time was limited to chlorination). 85 Chapter 3 121°48'0"W I • Cl imate station * Hydrometr ic stat ions o Sampl ing si tes Contour Interval = 20 m Figure 3-4 - Map of the Elk Creek watershed showing topography, major streams, sampling sites and hydrometric and climate stations. 3.2.3. The Salmon River watershed The Sa lmon River watershed is located northeast of Langley, along the southern shores of the Fraser River and is approximately 80 k m 2 in area. It is topographically less rugged than the Elk Creek and Hatzic watersheds, with a maximum elevation of approximately 140 m. A s a result, much of the watershed is developed and there is very little contiguous forest present. There are more than 550 farms 86 Chapter 3 in the area, with a larger proportion of hobby farms than in the other two watersheds (predominantly horse farms on marginal soi ls). Crop agriculture is dominated by cranberry operations, and there are significant cattle and chicken populations in the watershed. More than 13,000 people live in the area, and, as of 1997, the population was served by over 4,000 septic systems (Schreier et al . , 1999). 122°36'0"W 122<30'0"W 122°24'0"W , I 1 I , 1 1 1 122°36'0"W 122°30'0"W 122°24TJ"W Figure 3-5 - Map of the Salmon River watershed showing contours, major streams, sampling sites and hydrometric and climate stations. Note also the location of the Hopington aquifer. The watershed is underlain by the Hoppington aquifer which contributes significantly to streamflow during summer months. This cold water source, combined with gravel-bed st reams, produces excellent spawning habitat for sa lmon. However, this aquifer is unconfined and nitrate contamination from septic sys tems and agricultural operations has been documented, with concentrations regularly exceeding the Health C a n a d a guideline of 10 mg/L (Schreier et a l . , 1999). 87 Chapter 3 3.3. Field methods 3.3.1. Meteorological monitoring The purpose of the meteorological monitoring program was to gain an understanding of the inf luence of precipitation and temperature on surface-water quality at a range of sca les (i.e., from storm-event to seasona l ) . Of particular importance to this thesis was the col lection of high-resolution rainfall data in order to a s s e s s the impact of rainfall events on contaminant mobil ization and transport. In the Hatzic val ley, meteorological data were col lected between November , 2002, and August, 2005 , at a station establ ished in the northeastern region of the watershed, on a smal l promontory where Patt ison Creek reaches the val ley floor (Figure 3-3). This site was located on a private woodlot (with permiss ion of the landowner), and was chosen for two reasons. Firstly, it is near the headwaters of Patt ison Creek , which is the largest tributary feeding into the mainstem river of the watershed. Second ly , it is located approximately 280 m above the val ley floor, thereby providing a more realistic assessmen t of rainfall in the higher-elevation headwaters of the watershed. Scaffolding was used to elevate all sensors approximately 4 m in order to avoid interception losses from nearby trees and to prevent snow burial during winter months (as this site was located on private property, falling of the trees to completely el iminate interference was not possible) . The station included components from Forest Technology Sys tems (FTS) , namely a R G - T model t ipping-bucket raingauge (resolution: 0.254 m m ; accuracy: ± 2 % at 50 mm/hr), a THS-1 air temperature thermistor (resolution: 0.1 °C; range: -51-+60 °C; accuracy: 0.2 °C) and a P G - 4 precipitation co lumn (capacity: 1410 mm), all of which were connected to an F W S - 1 2 S data logger. Data were col lected for each sensor at 15-minute intervals and downloaded to a laptop on a regular bas is (every 1 -3 months). 88 Chapter 3 Figure 3-6 - a) meteorological station located in the Hatzic watershed, b) data logger, power supply and field laptop used to download data. For the Elk Creek watershed, rainfall data were obtained from the City of Chi l l iwack for a tipping-bucket raingauge located on Marble Hill (Figure 3-4). 3.3.2. Hydrometric monitoring Hydrometric data were col lected in the Hatzic and Elk Creek watersheds to investigate the relationships between rainfall and stream hydrology and the influence that both variables have on surface-water quality. In the Hatzic watershed, two hydrometric stations were establ ished on the mainstem river in July and December, 2002 and remained operational until August, 2005. Each station was equipped with 4 sensors : 1) an Analite 195 turbidity probe (equipped with an automated, anti-fouling wiper), with a range of 1-1000 N T U (nepholemetric turbidity units), 2) a Unidata four electrode, temperature-compensated conductivity sensor , with a range of 0-200,000 u.s-cm" 1, 3) a Unidata hydrostatic water depth probe (Model 6508), with a range of 0 - 2 m and 4) a temperature sensor (part of the hydrostatic probe). All sensors were connected to a Unidata "Display Starlogger" data logger (model 6004-2) which was downloaded on a regular basis (every 1-3 months). Sensors were c leaned during every visit and calibrated at least once every 6 months. The criteria for select ing both sites were as follows: 1) that they al low a compar ison of "headwater" and "downstream" hydrologic condit ions (i.e., a compar ison of hydrometric variables under 89 Chapter 3 forested, un-inf luenced condit ions vs . agricultural condit ions), 2) that they be located in morphological ly appropriate reaches of the stream (straight banks, single channel , minimal turbulence), 3) that the upper station be located far enough downst ream from the headwaters to limit potential damage by debris transported from 2 s lope failures in the upper catchment and 4) that they both be located on private property in order to minimize the potential for vanda l ism. B a s e d on these criteria, the sites were establ ished on upper Lagace Creek and the Lower Hatzic S lough (Figure 3-3). Each site was similarly constructed (Figure 3-7). The data loggers were housed in sea lab le , hard plastic c a s e s , which were in turn stored inside locked wooden cabinets. Leading from the cabinet was a 3.8 cm A B S pipe that housed the sensor cab les . This pipe fed into a sect ion of 7.6 cm pipe using a "reducing Y-joint", with one 3.8 cm branch and two 7.6 cm branches; one that sc rewed onto the main pipe, and the other with a removable cap . The sensors were mounted on a rod and rested at the bottom of the 7.6 cm pipe, which was perforated (to allow free movement of water around sensors) and capped . The rod could be pulled out when the cap at the top of the main pipe was removed allowing the sensors to be c leaned , cal ibrated and re-inserted into the pipe to the exact s a m e depth. To ensure that the pressure t ransducers remained at the s a m e depth over the course of the study, the 7.6 cm pipe was anchored to fixed points in the stream or on the shore. 90 Chapter 3 Figure 3-7 - Photographs illustrating: a) cable housing, Y-joint and removable cap providing access to sensors for cleaning and calibration (lower hydrometric station); and b) data logger mounted in wooden cabinet on tree at upper hydrometric station. Discharge was measured at the upper hydrometric station using a Swoffer Model 2100 Current Velocity Meter, and a top-set wading rod following standard procedures (Ministry of Environment Lands and Parks, 1998). Velocity measurements were obtained at 0.6 of the depth at 50 cm intervals across the channel . Stage was col lected from the automated sensor prior to, and just after, velocity measurements , and averaged if there was any change over the measurement time. Unfortunately, over the course of the study, several severe weather events resulted in transport of significant quantities of gravel to the region upstream of this station. This material was subsequently excavated on three separate occas ions . A s a result, the morphology of the channel changed significantly during the three years the station was in operation and it was not possible to establ ish a reliable rating curve for this site. In the Elk Creek watershed, hydrometric data for six stations were obtained from the City of Chil l iwack from July, 2002 to May, 2005. Each station consisted of an Isco 4150 A rea Velocity Meter (AVM) and an Isco data logger that recorded flow at 5-minute intervals. Due to maintenance issues, complete records are not available for any station; however, a compar ison with rainfall data al lows an assessment of streamflow response to storm events. Due to limitations associated with these hydrometric 91 Chapter 3 data, the influence of storm events on water quality will be a s s e s s e d primarily using rainfall data from the Marble Hill ra ingauge. 3.3.3. Water quality monitoring The water-quality monitoring program in each watershed was des igned to: 1) determine basel ine values for nutrient and bacterial concentrat ions from forested sub-catchments, 2) a s s e s s the influence of agricultural activities on surface water quality and 3) to determine spatial and temporal trends in contamination in order to identify t ime periods and land use activities assoc ia ted with the greatest risk in terms of water contaminat ion. To accompl ish this, water samples were obtained from stations representing var ious land use types (forested, agricultural and urban) and different hydrological condit ions (tributaries, mainstem river and sloughs) and samp led under a range of meteorological and hydrological condit ions (dry s e a s o n vs. wet s e a s o n and during storm events). The locations of sampl ing stations in each of the three watersheds are illustrated above (Figure 3-3, Figure 3-4 and Figure 3-5). The dominant land uses assoc ia ted with each station in the three watersheds are provided in Table 3-2. Land-use categor ies were ass igned to each station based on activities in their respective contributing area. Sub-ca tchments with less than 1% disturbance were categor ized a s "forested". T h o s e with between 1-10% agricultural activity were ass igned to the "mixed-use" category, while sites with greater than 1 0 % agricultural land use were ass igned to the "agricultural" category. Si tes with no agricultural inf luence but greater than 1% of total contributing a rea under residential development were categor ized as "urban". 92 Chapter 3 Table 3-2 - Land-use categories for the three watersheds based on contributing area. atershed Station Land use Area Forest Agriculture Urban/Rural Other (%) category (ha) (%) (%) Residential (%) Hatzic HV-1 Agricultural 961.5 80.9 15.8 3.3 0.0 HV-2 Forested 152.5 99.0 0.0 1.0 0.0 HV-4 Forested 150.1 100.0 0.0 0.0 0.0 HV-5 Mixed 227.9 97.1 1.1 1.8 0.0 HV-6 Forested 465.9 99.8 0.0 0.2 0.0 HV-8 Mixed 93.5 86.3 3.7 9.4 0.0 HV-9 Agricultural 45.0 69.7 29.4 0.8 0.0 HV-10 Mixed 206.8 92.3 1.8 5.9 0.0 HV-11 Mixed 1566.9 91.4 1.0 7.6 0.0 HV-12 Forested 111.8 100.0 0 0.0 0.0 HV-13 Mixed 340.9 98.1 1.2 0.7 0.0 HV-14 Agricultural 3444.5 80.5 10.3 9.2 0.0 HV-15 Agricultural 770.0 85.7 10.9 3.4 0.0 HV-16 Agricultural 984.0 76.0 20.0 4.0 0.0 HV-17 Agricultural 4416.9 81.2 10.9 8.0 0.0 HV-18 Agricultural 457.7 73.5 12.5 14.0 0.0 HV-19 Agricultural 4718.5 80.4 11.7 7.8 0.0 HV-20 Urban 826.0 67.0 0.0 37 0.0 Elk Creek EC-1 Agricultural 3128.8 72.9 19.7 1.6 5.9 EC-2 Agricultural 3087.4 73.8 18.6 1.6 5.9 EC-3 Mixed 2017.1 82.3 8.1 0.6 9.0 EC-4 Agricultural 68.4 0.0 100.0 0.0 0.0 EC-5 Agricultural 120.4 0.0 100.0 0.0 0.0 EC-6 Agricultural 44.8 0.0 100.0 0.0 0.0 EC-7 Mixed 566.6 66.7 6.3 1.6 25.4 EC-8 Forested 1313.1 96.5 0.7 0.0 2.8 EC-9 Forested 1247.3 100.0 0.0 0.0. 0.0 EC-10 Mixed 250.0 95.4 0.0 4.6 0.0 EC-11 Urban . 300.5 92.2 0.2 7.5 0.0 EC-12 Urban 326.6 86.5 3.1 10.3 0.0 EC-13 Mixed 331.9 98.7 1.3 0.0 0.0 EC-14 1 Mixed 282.5 100.0 0.0 0.0 0.0 EC-15 Agricultural 769.9 80.5 14.8 4.7 0.0 Salmon SA-1 Agricultural 6790.1 6.8 45.8 29.4 18.0 SA-2 Agricultural 5363.0 5.8 42.2 32.8 19.2 SA-3 Agricultural 1016.6 5.1 46.0 30.7 18.2 SA-4 Agricultural 2236.4 5.1 36.7 33.1 25.1 SA-5 Agricultural 1417.3 5.3 48.2 34.4 12.1 SA-6 Agricultural 4153.3 5.7 41.9 32.8 19.5 SA-7 Agricultural 1855.3 4.4 40.6 27.2 27.8 SA-9 Agricultural 1320.4 5.7 42.4 17.2 34.8 SA-14 Agricultural 577.7 12.4 62.4 17.4 7.8 SA-17 2 Agricultural 475.3 5.3 38.6 12.5 43.5 SA-19 Agricultural 581.9 2.6 56.2 32.1 9.1 1- EC-14 was assigned to the category of "Mixed" as it was determined after site selection that significant logging activities had occurred upstream. 2 - SA-17 was defined as "Agricultural" as the majority of "other" was cleared, but undeveloped Crown land. 93 Chapter 3 In the Hatzic watershed, 20 sampl ing sites were initially se lected. Two of these, HV-3 and HV-7 were el iminated from the sampl ing program as both were ephemera l , only flowing during the wettest months of the winter s e a s o n . In the Elk Creek watershed, 15 stations were se lected to represent a range of forested and agricultural condit ions, while in the Sa lmon watershed, 11 sites were chosen to represent agricultural condit ions, and to include st reams under the influence of N 0 3 " -rich groundwater. Due to a limited range in elevation across the Sa lmon watershed, delineation of contributing areas is approximate. Further, unlike the Hatzic and Elk Creek catchments, there is significant rural residential (low density, hobby farms or similar) land use in the watershed. A s agriculture was the dominant land use in each of the subcatchments, and residential development was relatively low-density, each of the subcatchments was categor ized as "agricultural." Sampl ing was t imed to capture seasona l condit ions throughout the water year and was most frequent during the wet s e a s o n (October-April) as this was the most hydrologically active and variable time of the year. Sampl ing during summer months was also conducted to capture low-flow condit ions and to determine the influence of storm events on water quality after prolonged dry periods. All samples for chemica l analysis were col lected in ac id -washed 250 ml or 500 ml low-density polyethylene (LDPE) bottles. Samp les were stored on ice in a cooler during sampl ing and refrigerated within 4-6 hours of col lect ion. All samples were ana lysed within 48 hours, and usually within 24 hours. Whi le analysis within 24 hours is recommended, analysis after 48 hours with refrigeration has been shown to produce comparab le results (Kotlash and C h e s s m a n , 1998). At each site, depth integrated samples were col lected by hand by opening the bottle near the stream bed and drawing it towards the surface as it filled with water. Samp les for microbiological analysis were col lected in sterile, factory-sealed bottles, which were also stored on ice in a cooler during sampl ing. All microbiological samples were returned to the B C Centre for D isease Control (BC C D C ) for analysis within 4-6 hours of sampl ing where they were subsequent ly ana lysed within 24 hours. 94 Chapter 3 3.4. Laboratory methods 3.4.1. Nutrient analysis Ana lyses for nutrients (N0 3~, N H 3 and P 0 4 3 " ) were conducted at the University of British Co lumb ia Soi l Chemist ry Laboratory using a Lachat Instruments Q u i k C h e m FIA+ 8000. The methods used to detect and quantify these compounds were Q u i k C h e m 12-107-04-1-B, Q u i k C h e m 10-107-06-2-A and Q u i k C h e m 10-115-01-1-A, respectively (Lachat Instruments). For N 0 3 " analys is , method 12-107-04-1-B involves reducing N 0 3 " to N 0 2 " by pass ing the sample through a co lumn containing copper coa ted cadmium. Nitrate concentrat ion was then determined by diazot izing with sulphani lamide dihydrochloride and measur ing absorpt ion of the resulting magenta dye at 520 nm. Va lues were expressed as N 0 3 - N with a minimum detection limit is 0.025 mg-L" 1 . For N H 4 + analysis, method 10-107-06-2-A involved heating samples with sal icylate and hypochlorite in an alkaline phosphate buffer, resulting in the production of an emera ld green dye that is in proportion to N H 4 + concentrat ion. The colour was then intensif ied by the addition of sodium nitroprusside. Resul ts were expressed as N H 4 + - N . The detection limit for this method is 0.1 mg-L" 1 . Ana lys is for P 0 4 3 r involved digestion in the presence of sulphuric acid and persulphate to hydrolize polyphosphates and organic P to P 0 4 3 " . The P 0 4 3 " ion reacts with ammonium molybdate and antimony potass ium tartrate under acidic condit ions to produce ascorb ic ac id which absorbs at 880 nm. Absorpt ion at this wavelength is measured to determine concentrat ion of P 0 4 3 " - P , with a detection limit of 0.02 mg-L" 1 . Ana lys is for chloride (CI) was conducted using Q u i k C h e m method 10-117-07-1-A. This method involves measur ing the absorption of ferric thiocyanate at 480 nm. This compound is produced as a result of the liberation of thiocyanate from mercuric thiocyanate through the formation of soluble mercuric chloride. Absorpt ion at 480 nm is proportional to CI" concentrat ion. The detection limit for this method is 6 mg-L" 1 . D isso lved organic carbon (DOC) was ana lyzed in the U B C Civi l Environmental Engineer ing Laboratory using a Sh imadzu (TOC-500) Total Organic Carbon Analyzer . Concentrat ions of D O C were determined by subtracting d issolved inorganic carbon from total d isso lved carbon, and were expressed as m g - L 1 . Prior to analysis, all samp les were filtered using Whatman #41 filter paper. Quali ty control was ensured through the analysis of standards after every 10 samples . 95 Chapter 3 3.4.2. Microbiological analysis All microbiological ana lyses for total and fecal col i forms were carr ied out in the Environmental Microbiology Laboratory at the B C C D C following Method 9222 of Standard Methods (American Publ ic Health Assoc ia t ion, 1999). Va lues were reported as colony forming units per 100 ml (cfu/100 ml). W h e n excess ive growth on the filter prevented enumerat ion, the result was reported as "overgrowth" or " O G " . 3.4.3. Spectrophotometric analysis Prior to spectrophotometr ic analysis, all samples were filtered through 2.5 um pre-ashed Whatman 42 filters to remove particulate matter that could interact with light during absorbance or f luorescence scans . Absorbance spect ra were col lected using a Cary 4000 UV-V i s absorbance spectrophotometer. Samp les were p laced in quartz cuvettes with a 1 -cm pathlength, and absorbance was measured between 200-800 nm at 1 nm intervals, using a scan rate of 600 nm/min. For each scan , basel ine correction was appl ied, whereby the absorbance spectrum for a Mil l i-Q de- ionized water sample was subtracted from that of the field sample in order to measure absorbance only for d isso lved materials in the sample . Post -process ing of absorbance scans (to obtain second derivative curves) was accompl ished using the "Maths" function within the Cary S c a n software (version 3.0(182)). Duplicate scans were col lected once for every 20 samp les . F luorescence readings were acquired using a Var ian C a r y Ec l ipse f luorescence spectrophotometer. Emiss ion scans were acquired for several excitation wavelengths (between 220-450 nm at 5 nm increments). Emiss ion was measured between 230-600 nm at 2 nm increments for each excitation wavelength (producing a total of 47 emiss ion spectra for each sample) . Excitation and emiss ion slits (which control the resolution of the emiss ion spectrum) were set to 5 nm. Corrected spectra (provided by Var ian, Inc.) were used to account for variat ions in emiss ion that arise from the instrument itself (primarily a result of the changing intensity of the excitation source with wavelength). The excitation filter, used to select the excitation wavelength of interest, was set to automatic. The emiss ion filter was a lso set to automatic in order to eliminate residual excitation light. Prior to each sample run, a blank of de- ionized water was scanned at E x / E m : 350/395 (the R a m a n peak for water) for two minutes to a s s e s s system stability (Baker, 2001). De- ionized blanks were scanned once every 20 samples . Emiss ion at 395 nm averaged 18.3 + 0.60 over the course of the study 96 Chapter 3 with no drift observed. All E E M scans were normal ized to a R a m a n peak of 20.0 to eliminate concentration effects and allow compar isons of D O M components across si tes. Duplicate scans were also col lected during each sample run. Duplicate scans were subtracted from the original and the percent difference was calculated for all 47 emiss ion spectra for each duplicate water sample (8742 data points). The difference between original and dupl icate scans was consistently less than 5 % for more than 9 9 . 5 % of all data points. Severa l authors note the need to apply corrections to f luorescence data to account for the inner-filtering effect assoc ia ted with high D O M concentrat ions in the sample (Mobed et al . , 1996; Lakowicz , 1999; Ohno , 2002). Correct ion for inner-filtering was not appl ied to samp les in this study for two reasons. Firstly, absorbance at 254 nm in most samples (93%) was below 0.3, indicating that D O M concentrat ions were not sufficient to produce a significant inner-filtering effect (Ohno, 2002). Second ly , a primary objective of this study was to a s s e s s f luorescence spect roscopy as a rapid and s imple technique for water-quality determination with minimal data-processing requirements. It was therefore dec ided to conduct the assessmen t using the raw f luorescence data. 3.5. Data analysis and representation A significant chal lenge in the analys is of environmental data involves the use of inferential statistics on data that are not independent either in t ime or in space . A critical assumpt ion of statistical tests is that all data points, or replicates, are independent. In environmental monitoring, due to repeated sampl ing at individual s i tes over time, or sampl ing of sites that are not geographical ly d ispersed, the assumpt ion of independence is not often satisf ied. The use of inferential statistics to test for dif ferences in such data was termed pseudorepl icat ion by Hurlbert (1984). Us ing inferential tests for samples col lected from the same site over t ime (and counting them as replicates) is termed temporal pseudorepl icat ion, while spatial pseudorepl icat ion results from analysis of data col lected from locations that are not sufficiently geographical ly distant. The result of such ana lyses is to artificially inflate the degrees of freedom, thereby increasing the potential for a Type 1 error (detecting a significant difference when one does not exist). To address this issue, when conduct ing compar isons of water-quality data between land uses or seasons , all data for each variable were averaged for each site to produce one value per site, per variable. In contrast to pseudorepl icat ion, this results in a substantial dec rease in sample s ize (in s o m e 97 Chapter 3 c a s e s up to an order of magnitude). This reduces the statistical power of the analysis (and the ability to detect real differences), but was preferred over the interpretation of invalid results that would ar ise by using the entire data set inappropriately. For Chapters 6 and 7, where correlat ions between spectroscopic var iables and water quality parameters were a s s e s s e d , ana lyses were conducted on pooled data and on data from individual sampl ing dates to see if there was any influence of autocorrelation. Resul ts for both were consistent, and so pooled data are presented. All statistical ana lyses were conducted using S P S S (13.0 and 14.0). Prior to analys is , all water-quality var iables were a s s e s s e d for normality through visual inspection of histograms and Q - Q plots and by using the Kolmogorov-Smirnov test (with Lill iefors signif icance correction). None of the var iables met the criteria for normal distribution. Therefore, a conservat ive approach was used for compar ison tests. The non-parametric Mann-Whi tney test was used for inter-group compar isons (i.e., those compar ing water-quality var iables across land uses or seasons) and the Bonferroni correction was appl ied to multiple group compar isons to account for the increased probability of Type 1 errors. Spea rman rank order correlations were used to test for associat ions between var iables. Boxplots produced in S P S S were used in this thesis as a tool for data analys is and visual representation. In these plots, each box d isp lays the median value (horizontal line within the box) and the 2 5 t h and 7 5 t h percenti les (the lower and upper bounds of the box). The minimum and maximum values that are not outliers are represented by the upper and lower horizontal l ines outside the box. Outl iers (more than 1.5 box lengths away from the box edge) are represented by circ les and extreme va lues (more than three box lengths from the edge of the box) are represented by aster isks. All other 2-d imensional graphs were produced using Grapher 6.0 (Golden Software), and all contour and surface plots were produced using S igmaPlo t 9.0 and 10.0 (Systat). All maps were produced using A r c M a p , within A rcG IS Desktop 9.0 (ESRI) . 98 Chapter 4 4. Influence of land use, climate and hydrological conditions on nutrient and bacterial cycling in an agricultural watershed. 4.1. Introduction This chapter descr ibes the results of a multi-year (May, 2002 - August , 2005) assessmen t of surface-water quality in the Hatzic watershed. The objectives of this study were to: 1) determine "basel ine" nutrient and bacterial levels in surface waters draining forested, undeveloped sub-catchments, 2) quantify the impact of agricultural land use on surface-water quality, including cumulat ive impacts assoc ia ted with increased agricultural inf luence and 3) a s s e s s spatial and temporal trends in nutrient and bacterial cycl ing in order to identify high-risk condit ions for contamination. To date, few multi-year studies have addressed the combined influence of cl imate and agricultural land use on bacterial and nutrient cycl ing in surface waters in the L F V . This study a s s e s s e s basel ine condit ions in undeveloped subcatchments, and the mechan isms of nutrient and bacterial transfer from agricultural f ields to surface waters at a range of temporal and spatial sca les . The chapter begins with an overview of the hydrological and meteorological condit ions observed during the study. This is fol lowed by a d iscuss ion of spatial trends in surface-water quality in terms of nutrients and bacter ia (with particular emphas is on the role of differing land uses) . Tempora l trends from storm-event to annual sca les are then descr ibed, and are fol lowed by conc lus ions. 4.2. Methods Methods used for the collection and analys is of data descr ibed in this chapter are descr ibed in detail in Chapter 3. Data presented in this chapter include var iables measured at two hydrometric stations (temperature, specif ic conductance, water level and turbidity), one meteorological station (air temperature and rainfall) in the Hatzic watershed. This chapter a lso descr ibes results from chemical and microbiological ana lyses conducted on grab samp les (specific conductance, N 0 3 _ , N H 4 + and P 0 4 3 _ concentrat ions and fecal col i form, total coliform and E. coli concentrat ions). Note that, for simplicity, the term "conductivity" is used to refer to specif ic conductance. 99 Chapter 4 4.3. Results 4.3.1. Hydrology and climate of the Hatzic watershed 4.3.1.1. Hydrometric parameters Figure 4-1 provides an overview of hydrological and cl imatological condit ions in the Hatzic watershed during the study. The strong seasonal i ty of rainfall and streamflow is evident in this graph. B a s e d on historical data (1959-2000) for the Hatzic watershed, wet s e a s o n rainfall (October - April) averages 220 mm-month" 1 , compared to 89 mm-month" 1 during the dry season (Environment C a n a d a , 2005). St ream response to rainfall events was rapid at the upper station with peak flow arriving within hours of peak rainfall. Onset of max imum flow at the downstream station was general ly de layed by several hours and the peak was broader and receded more slowly due to the integration of flow from approximately half of the watershed (44 ha) and the influence of Hatzic Lake, located approximately 1 km downst ream. A broad peak in water level was visible during each summer season at the lower station. T h e s e peaks are due to the deliberate elevation of lake water levels (flow is controlled at the f lood boxes) for recreational uses during this t ime. 100 Figure 4-1 - 24-hour rainfall, water level and specific conductance for the upper and lower hydrometric stations in the Hatzic watershed. Black horizontal lines denote missing data as a result of logger failure. Three major storm events are also noted (see text for details). Chapter 4 Speci f ic conductance showed a marked difference between the upper and lower stations, with mean values of 51.9 uS;cm"1 and 80.1 uS-cm"1, respectively. This difference is attributed to the higher total contributing a rea under agriculture for the lower monitoring station (883 ha vs. 28 ha for the upper site), and is supported by the fact that mean conductivity in surface-water samples from agricultural si tes was significantly higher than in samples from forested sites (93.8 uS-cm"1 vs . 29.1 uS-cm"1, respectively, P < 0.001). This link between agricultural land use and elevated surface-water conductivity va lues has been observed in previous studies (e.g., Dow and Zampe l la , 2000; C h e n et a l . , 2006) and is attributed to higher inputs of ions such as N 0 3 " and CI" der ived from agricultural activities (tillage and appl icat ions of organic and chemica l fertil izers). These data indicate that the two stations were situated appropriately to represent forested and agricultural inf luence on hydrometrics parameters. It should also be noted that an earl ier study of groundwater quality in the watershed (Magwood, 2004) revealed elevated conductivity levels in groundwater samp les , with va lues ranging from 17 - 3,290 uS-cm"1. This influence on surface-water conductivity can be seen at both hydrometric stations during summer months (Figure 4-1), as values showed a continual increase in the absence of contributions from rainfall. During the wet s e a s o n s , conductivity at both stations showed a rapid downward response to storm events. Pre-storm levels were quickly re-establ ished as the relative contribution of rainfall to streamflow dec reased on the falling limb of the storm hydrograph. Water temperature var ied similarly for both stations over the period of record (Figure 4-2). Minimum va lues of 0.4 °C and 1.4 °C were observed during January , 2004 for the upper and lower stations, respectively. Max imum values at the upper and lower station were 23.3 °C and 19.3 °C, respectively (in July, 2004 for the upper station and July, 2003 for the lower station). The shal lower c ross-sect ional profile at the upper station resulted in the lower minima and higher max ima that were observed at this site, as well as the greater diurnal fluctuations observed. 102 ecu C (V *» rb k <D (D 3 •o <o c o 3" (D TJ fl> 5' Q. <0 o o Q. o (D C TJ TJ (D O € (D 3" *< O. —i o 3 o w ST 5' 3 CO Temperature (°C) Temperature (°C) ro o o ro o ro Nov-02 -Dec-02 -Jan-03 -Feb-03 -Mar-03 -Apr-03 -May-03 -Jun-03 -Jul-03 -Aug-03 -Sep-03 -Oct-03 -Nov-03 -Dec-03 -Jan-04 -Feb-04 -Mar-04 -Apr-04 -May-04 -Jun-04 -Jul-04 • Aug-04 -Sep-04 Oct-04 Nov-04 —I Dec-04 Jan-05 —| Feb-05 Mar-05 Apr-05 May-05 Jun-05 Jul-05 Aug-05 x O CO S o c T J CD Chapter 4 The water level data for both upper and lower stations contain a unique daily cyc le in water depth that is easi ly observed during per iods of low flow. To a s s e s s this relationship without the influence of precipitation, a set of 4400 data points was extracted from the time ser ies (July 22 - September 5, 2003) during a low-flow period with no recorded rainfall. These data suggest that there is a relationship between stage and water temperature; however, the nature of this relationship is different between the two stations. At the upper station, there was a strong and significant negative correlation (r s = -0.833, P < 0.001) between diurnal water temperature and water depth. A similar pattern was observed by Bond et a l . (2002), and was attributed to daily evapotranspirat ion losses in forested catchments. This may explain the results of the present study, as the contributing a rea to this station was 9 1 % forested. The observed trend may also have been due to an inadequately vented air tube. The venting tube is des igned to allow the sensor to compensate for changes in atmospher ic pressure, and in these stations, ran from the sensor to the data logger box. Inadequate venting of the data logger box may have c a u s e d the sensor to calibrate for pressure changes assoc ia ted with warm, expanding air in the case , resulting in the observed inverse relationship. The relationship at the lower site was a lso significant (rs = 0.440) but was posit ive and slightly weaker. The daily peaks at this station contained secondary max ima, suggest ing that tidal changes in the Fraser River (which can be in excess of 1.0 m) may be influencing Hatzic Lake and Hatzic S lough levels (it is unknown what role, if any, the f lood boxes at the outlet of the watershed may play in mitigating this effect). These patterns are observed through the hydrometric record; however, because the amplitude of the cyc le is general ly smal l (4-6 cm) in compar ison to storm-induced depth increases, it does not obscure the rainfall-water level relationship. Turbidity was monitored at both hydrometric stations over the course of the study. During the first wet season (2002-2003), rainfall amounts were moderate and turbidity probes recorded elevated concentrat ions of suspended particulates assoc ia ted with storm events (Figure 4-3). During the second and third wet seasons (2003-2004 and 2004-2005), several extreme storm events (described below) del ivered large quantities of clastic material to the mainstem river (from active landsl ide scars located in the headwaters). Both sensors were overwhelmed during these events (turbidity exceeded the upper limit of both probes), likely due to sediment accumulat ion in the sensor housing. Despite continual maintenance, both sensors repeatedly recorded maximum values due to fouling over the latter two wet seasons . A s a result, useful turbidity data are only avai lable on a periodic bas is . 104 s I 20 — J J 40 -S l 60 80 0.8 S 0.6 0.4 0.2 Stage Turbidity Rainfall 11 I, iM v,iV 'i i> I ' 400 300 — too 1/1/03 2/1/03 a) 3/1/03 Date (m/d/yy) 4/1/03 5/1/03 0 | f 4 0 F 80 !1 120 -2 1.2 200 ft £ 0.8 0.4 10/1/03 b) Stage Turbidity 1 0 0 0 Rainfall 800 600 ?. 3 400 200 11/1/03 12/1/03 1/1/04 Date (m/d/yy) Figure 4-3 - Rainfall, water level and turbidity data illustrating: a) clear correlation between the three variables at the lower hydrometric station during the moderate wet season of 2002-2003, and b) an example of fouling of the turbidity sensor at the upper station after heavy rains. o Chapter 4 4.3.1.2. Meteorological parameters Three wet s e a s o n s were monitored during this study (a portion of 2002-2003 and all of 2003-2004 and 2004-2005). During the latter two seasons , the three largest storm events caused extensive flooding in the Hatzic watershed (Figure 4-1). The first, in October 2003, resulted from the arrival of several moist, Paci f ic sub-tropical air m a s s e s that stal led in a low pressure system over the G V R D and L F V for several days (October 15 - 22). Record 24- and 48-hour rainfall totals were recorded at several meteorological stations on Vancouver Island and the main land. The return period for the storm was calculated at 100 years in many locations (Northwest Hydraul ic Consul tants and Scott Resources Serv ices , 2005). A n analysis of rainfall data (Table 4-1) from several stations in the L F V indicated that rainfall was particularly heavy in the Hatzic watershed (although higher rainfall va lues are due in part to the higher elevation of the Hatzic cl imate station). Peak rainfall intensity reached 19.3 mm-hr - 1 , and remained high for several days , resulting in a maximum 48-hour total of 257.3 mm (October 16-17). Table 4-1 - Daily rainfall at four weather stations in the Lower Fraser Valley for the major storm event in October, 2003 (Environment Canada, 2005). Numbers in brackets represent station elevation. Station Oct. 15 Oct. 16 Oct. 17 Oct. 18 Oct. 19 Oct. 20 Oct. 21 Oct. 22 Total Abbotsford 12.4 93.8 73.6 T 11.8 50.6 25.0 23.0 290.2 Airport (57.9 m) Chilliwack 22.6 100.3 28.2 8.9 20.5 73.8 0 23.0 277.3 (11.0 m) Mission, 32.2 123.2 62.0 15.4 17.0 66.8 1.8 27.8 346.2 West Abbey (221 m) Hatzic 11.4 136.4 120.9 1.3 32.5 64.3 24.9 35.1 426.7 Climate Station (270 m) Signif icant f looding was observed throughout the watershed (Figure 4-4), with the majority of lowland agricultural fields being completely submerged. Water levels at the lower hydrometric station increased from 0.27 m to a peak of 2.0 m (the upper limit of the hydrostatic depth probe), and Hatzic S lough breached its banks in several locations. Conduc tance at the lower hydrometric station initially decl ined, but as water levels peaked , it subsequent ly increased to a maximum of 213 uS-cnrf 1 (the highest value observed during the study) and remained high during the recess ion of the storm peak. A s a prolonged peak was not observed at the upper hydrometric station at the s a m e time, this is attributed to 106 Chapter 4 an increase in d issolved solids derived from nearby agricultural fields that had been submerged for more than 11 days. Figure 4-4 - a) Lower hydrometric station from the west, illustrating low-flow conditions typical of a dry period during winter, and b) looking south at the same hydrometric station from the bank during the October, 2003 event (note data logger and submerged dock railing for reference). The second largest event took place from November 21-27, 2004. Rainfall for the first three days was moderate (24-hour maxima from 5.3-16.5 mm), but was followed by 137.2 mm on November 24 t h , with peak rainfall intensities of 10.2 mm-hr ' 1 . Aga in , excess ive flooding occurred in low-lying areas, but was shorter in duration (less than 2 days) when compared to the 2003 event and total rainfall vo lumes were much lower (186.2 mm). The third notable event (January 16 -23 , 2005) produced total rainfall of 340 mm, a 48-hour maximum of 209 mm, and again resulted in significant flooding throughout the watershed, with agricultural fields submerged for nearly 5 days. It is notable that, while agricultural fields were submerged for several days during these latter f loods, neither event produced a peak in conductivity at the lower hydrometric station as was observed in October, 2003. This is attributed to differing antecedent condit ions and the longer duration of the 2003 flood. The 2003 event was preceded by several months of dry weather al lowing accumulat ion of animal wastes in storage areas and on pasture fields, as well as the accumulat ion of organic matter and agr ichemicals on crop fields. Further, 107 Chapter 4 conductivity only exceeded pre-storm levels during the 2003 event after agricultural fields had been submerged for nearly five days, and peak values were not observed until nearly 11 days after f looding began. 4.3.2. Spatial trends in surface water quality Spatial trends in nutrient and bacterial concentrat ions were a s s e s s e d in order to determine the impacts of agricultural land-use pract ices on surface water quality and to identify significant point sources of agricultural contaminat ion. 4.3.2.1. Nutrients 4.3.2.1.1. Site-to-site variability Nutrient concentrat ions in surface waters throughout the Hatzic watershed were low (Table 4-2), reflecting the relatively low intensity of agricultural activities. The percentage of samples with concentrat ions below detection limits for NOV, N H 4 + and P 0 4 3 " were 4 .3%, 3 9 % and 44%, respectively. These samples were ass igned a concentrat ion of 5 0 % of the detection limit. M e a n N 0 3 " va lues were highest at stations HV-13 and HV-20 (1.0 mg-L - 1 and 0.67 mg-L ' 1 , respectively), and lowest at stations HV-6 and HV-1 (0.23 mg-L ' 1 for both). In contrast, the highest mean N H 4 + concentrat ions were observed at station HV-18 at 0.34 mg-L" 1 , with stations H V - 1 , HV-8 , HV-10 and HV-16 having similar mean concentrat ions of 0.24, 0.21, 0.19 and 0.22 mg-L" 1 , respectively. Mean N H 4 + was lowest at HV-20 , at 0.09 mg-L" 1 . Max imum mean P 0 4 3 " concentrat ions were also observed at HV-18 (0.10 mg-L" 1), and the second highest mean value was observed at HV-9 (0.08 mg-L" 1). The two lowest average P 0 4 3 " concentrat ions were observed in forested subcatchments at HV-2 and HV-12 (0.04 mg-L" 1 for both). 4.3.2.1.2. Land-use influence A compar ison of results stratified by land-use using the Kruskal-Wal l is test revealed significant dif ferences for N 0 3 " between land-use categor ies (P = 0.03, n = 18). A s descr ibed in Chapter 3, all statistical compar isons were conducted using mean values from each station in order to avoid pseudorepl icat ion. This significantly reduced the sample s ize for each test as only one value was used for each station. W h e n post-hoc analys is was conducted using Mann-Whi tney tests (with Bonferroni correction to account for multiple compar isons) , the resulting signi f icance values were above P = 0.05. A s a result, the following descr ibes qualitative dif ferences for these parameters by land use. 108 Chapter 4 A s illustrated in Table 4-2, maximum mean N O V concentrat ions were observed at the urban site (HV-20). It is important to note that, because the urban site was not repl icated, it is not possib le to draw conclus ions regarding the impacts of urban land use in general . M e a n N H 4 + and P 0 4 3 _ concentrat ions were highest at agricultural sites. Interestingly, mean N O V concentrat ions were higher at mixed sites than at agricultural si tes. At each mixed site, agricultural operat ions were located just upstream of the sampl ing location. The difference observed likely reflects direct N 0 3 " inputs from these farms. Finally, N 0 3 " concentrat ions at agricultural sites were higher than those measured in forested subcatchments . Table 4-2 - Descriptive statistics for ammonia, nitrate and orthophosphate (2002-2005). Note that concentrations below detection limits were assigned a value of 0.5 x detection limit. Nutrient Land U s e N Min. Max . Mean Median S D 1 N H 4 + Comb ined 317 0.05 1.22 0.16 0.11 0.16 Forested 77 0.05 0.48 0.14 0.05 0.12 Agricultural 137 0.05 1.22 0.19 0.12 0.18 Mixed use 87 0.05 0.68 0.16 0.11 0.15 Urban 16 0.05 0.31 0.09 0.05 0.08 N O 3 - Comb ined 317 0.01 2.51 0.44 0.37 0.32 Forested 77 0.01 1.25 0.30 0.26 0.18 Agricultural 137 0.01 1.76 0.41 0.34 0.31 Mixed use 87 0.20 2.51 0.57 0.46 0.39 Urban 16 0.39 1.17 0.67 0.59 0.27 P 0 4 3 ' Comb ined 291 0.01 0.40 0.06 0.03 0.08 Forested 71 0.01 0.28 0.04 0.01 0.06 Agricultural 126 . 0.01 0.40 0.07 0.03 0.08 Mixed use 80 0.01 0.35 0.06 0.03 0.08 Urban 14 0.01 0.34 0.06 0.02 0.09 - S tandard Deviat ion Agricultural si tes general ly produced elevated nutrient levels; however, the three nutrients did not vary similarly. An analysis of Spearman rank correlations between these var iables for agricultural sites revealed significant correlat ions between N H 4 + and P 0 4 3 " (rs = 0.255, P = 0.004, n = 126). A similar correlation was observed for mixed sites (rs = 0.235, P = 0.036, n = 80). In forested subcatchments, mean nutrient concentrat ions were consistently lower than for all other land uses . Peak values were typically observed during the wet season , reflecting the mobil ization of natural degradat ion byproducts in forest litter. 109 Chapter 4 4.3.2.2. Bacteria 4.3.2.2.1. Site-to-site variability Bacterial concentrat ions were highly variable throughout the watershed (Table 4-3). The ranges for total and fecal col i forms were 10 - 29,400 cfu-100 ml" 1 and 0 - 10,400 cfu-100 ml" 1 , respectively. W h e n all samp les were considered (N = 192), the two var iables were strongly correlated (rs = 0.736). Feca l coliform bacter ia were detected in all but 7 samp les , 6 of which were col lected from forested sites. Max imum values for fecal coliforms were recorded at HV-18 , a s low-moving agricultural s lough that f lows through several cattle and horse farms. Large numbers of waterfowl (>100 ducks and geese) were also frequently observed upstream of HV-18 during winter months. The next highest max imum values were observed at HV-5 , HV-17 , HV-13 and HV-19 . 110 Chapter 4 Table 4-3 - Descriptive statistics for total and fecal coliform by station (2003-2005). Station N Min. Max. M e a n Median S D 1 Total Col i form HV-1 12 110 5,300 1,152 475 1,620 HV-2 11 23 1,250 269 150 362 HV-4 10 24 2,900 478 167 877 HV-5 11 100 14,700 2,923 800 4,872 HV-6 12 12 320 100 76 95 HV-8 10 100 4,200 1,618 1,090 1,578 HV-9 10 100 7,800 2,513 1,700 2,513 HV-10 10 130 4,200 1,593 1,150 1,492 HV-11 11 30 2,100 641 360 713 HV-12 10 44 5,000 821 305 1,501 HV-13 10 300 6,300 3,017 2,500 2,543 HV-14 10 390 6,020 2,304 1,900 1,674 HV-15 10 780 4,800 2,018 1,550 1,332 HV-16 11 350 8,500 2,505 2,300 2,159 HV-17 11 340 6,700 2,381 1,670 2,099 HV-18 10 110 29,400 5,525 2,435 8,745 HV-19 11 460 4,400 2,095 2,510 1,383 HV-20 10 470 7,400 3,203 2,950 2,473 Feca l Col i form HV-1 12 0 410 70 24 114 HV-2 12 0 40 10 9 12 HV-4 10 0 710 76 10 223 HV-5 11 2 7,400 1,027 50 2,218 HV-6 12 0 20 6 3 6 HV-8 10 1 240 91 63 88 HV-9 10 7 370 100 72 107 HV-10 10 2 710 152 39 252 HV-11 11 2 76 23 10 25 HV-12 10 0 180 27 10 55 HV-13 11 4 2,400 335 30 733 HV-14 10 10 910 128 49 276 HV-15 10 10 170 50 30 51 HV-16 11 10 140 56 51 48 HV-17 11 1 2,900 390 76 858 HV-18 12 2 10,400 1,395 395 2,900 HV-19 11 10 2,120 311 108 611 HV-20 10 1 1,410 333 111 481 1 - Standard deviation A consistent increase in bacterial concentrat ions was observed along the mainstem from the forested headwaters (HV-6) to the final sampl ing site above Hatzic Lake (HV-19), reflecting the cumulat ive influence of moderate-intensity agricultural land use on surface-water quality (Figure 4-5). This co inc ides with an increase in the percentage of total land use accounted for by agriculture in the contributing a rea of each station, and also reflects contributions from tributaries with high bacterial concentrat ions derived from livestock operat ions (described below). m Chapter 4 o o HV-19 b) "i 1 r HV-11 HV-14 HV-17 HV-19 Site Figure 4-5 - Boxplots of a) total and b) fecal coliform concentrations demonstrating increasing bacterial contributions with downstream distance from forested headwaters. Note: outliers were not removed as they represent verified results. O n two occas ions (March 12 and May 1, 2003) samples were ana lysed for E. coli as part of an experiment at the B C Centre for D isease Control to a s s e s s a new £. coli culturing technique. This also provided an opportunity to determine the relationship between fecal coliform bacter ia, which are known to have non-fecal origins, and E. coli which is der ived solely from the intestinal tracts of animals and humans. O n March 12, the highest E. coli concentrat ions were observed at HV-1 and HV-19 (exact counts could not be obtained as excess ive E. coli colony growth prohibited enumerat ion, returning a result of "too numerous to count"). The next five highest values (ranging from 100-160 cfu-100 ml"1) were observed at H V - 5 , HV-8 , HV-9 , HV-15 , and HV-18 , all of which are agricultural si tes, except for HV-5 (mixed use). O n May 1, the highest count was observed at HV-20 (1080 cfu-100 ml"1) on Draper Creek . This is substantially higher than the Health C a n a d a guidel ine for recreational water quality (200 cfu-100 ml" 1), and represents a potential health risk as Draper Creek flows into Hatzic Lake near the beach at Nei lson Park, a popular summer recreation site. The next highest value observed was at HV-19 (75 cfu-100 ml" 1). All other sites had concentrat ions below 30 cfu-100 ml" 1 . For the two days on which both fecal coliform and E. coli data were col lected, a significant positive correlation between the variables was observed (r s = 0.728, P < 0.001, n = 35). This is illustrated graphical ly in Figure 4-6, as is the influence of land use, demonstrated by the grouping of samp les according to land-use category. This analysis suggests the fecal coliform data col lected during this study 112 Chapter 4 provide a valid indication of true fecal contaminat ion. Whi le it is recognized that the smal l sample s ize (n = 35) and relatively low bacterial counts on these two days prevent the extrapolation of this relationship to the entire data set, this is a useful means of confirming the value of fecal coliform bacteria as indicators of fecal contamination in this instance. • ^ A g r i c u l t u r a l : ' • ' ' . . # F o r e s t e d . ; . • ' . : x " : • A M i x e d ' '. " • X U r b a n . .:' r:-:::':;':r?r::: t • * • » —1 I ' " ' l I "" L 0.00 1.00 2 .00 3.00 E. Coli (LogW) Figure 4-6 - Log10 Fecal coliform vs. Log 10 E. coliior March 23 and May 1, 2003 showing strong positive correlation between concentrations of the two bacteria in stream samples. Symbols also illustrate concentrations by land-use category. 4.3.2.2.2. Land-use influence The highest median total and fecal coliform concentrat ions in the watershed were observed at the urban site (HV-20). Med ian values were used as an estimate of centrality due to the large standard deviat ions observed with coliform data (Table 4-3). The second and third highest median values were observed at agricultural and mixed use sites, respectively, and the lowest concentrat ions were consistently found at forested sites (Table 4-4). Med ian total and fecal coliform results from forested sites were significantly lower than for agricultural sites when all data were averaged for each station (P = 0.002, n = 13 for both). There were no significant dif ferences between agricultural and urban sites for either type of bacteria, or between mixed use and urban sites for fecal coli form. 3 . 0 0 -(Log10) CO p 2 . 0 0 " ifon Col Fecal 1 . 00 " o.ooH 113 Chapter 4 Table 4-4 - Descriptive statistics for total and fecal coliform data by land use (2003-2005). Bacter ia Land Use N Min. Max. M e a n Median S D Total coliform Comb ined 192 10 29,400 1,901 885 2,956 Forested 45 10 5,000 383 114 852 Agricultural 85 100 29,400 2,520 1,780 3,524 Mixed use 52 30 14,700 1,951 850 2,751 Urban 10 470 7,400 3,203 2,950 2,473 Feca l coliform Comb ined 196 0 10,400 260 29 986 Forested 46 0 710 27 9 106 Agricultural 87 0 10,400 330 64 1,189 Mixed use 53 2 7,400 333 40 1,096 Urban 10 2 1,410 333 111 481 Of particular interest is the substantial impact of cattle and horse operat ions on surface water bacterial concentrat ions throughout the watershed. Of the five highest fecal coliform values recorded during the study, three were at agricultural sites that either drained cattle and horse operat ions directly (HV-17 and HV-18) or were near the outlet of the mainstem river below severa l cattle operat ions (HV-19). The remaining two max ima were observed at mixed-use sites (HV-5 and HV-13) where the only inf luence in each forested subcatchment was one cattle operation (Figure 4-7). Whi le it is not possib le to relate animal densit ies to stream coliform levels at this sca le (agricultural census data are only provided at an aggregate level for the C e n s u s Enumerat ion Area), it is clear that individual l ivestock operat ions act as point sources for indicator bacteria, and therefore, potentially pathogenic microorganisms. These data a lso indicate that "total a rea under agriculture" is not an effective indicator of water-quality risk. In studies on the influence of urban land use on water quality, several authors have noted a threshold of approximately 1 0 % impervious a rea to generate a detectable change in water quality (e.g., Arnold and Gibbons , 1996; e.g., Hall et al. , 1999). For mixed sites in this study, it is apparent that the threshold is much lower, and that one farm can have a significant impact on the microbiological quality of surface waters. 114 Chapter 4 122°18'0"W 122°12'0"W _ l ' — i_1 1 122*18'0"W 122-12'0'W Figure 4-7 - Map of the Hatzic watershed illustrating locations of horse and dairy and beef cattle operations and stream sampling sites with the highest observed fecal coliform concentrations (black arrows). Bacter ia l concen t ra t ions in forested regions of the wa te rshed were signi f icant ly lower than in agricultural subca t chmen ts ; however , in contrast to nutrient leve ls , they w e r e obse rved to e x c e e d C a n a d i a n gu ide l ines for both recreat ional water quality and dr inking water quality (Table 4-3). Recrea t iona l gu ide l ines require that m e a n fecal col i form or E. coli concent ra t ions (from at least 5 s a m p l e s taken over less than 30 days ) not e x c e e d 200 cfu-100 ml" 1 (Federa l -P rov inc ia l Work ing G r o u p on Recrea t iona l W a t e r Qual i ty, 1992) and drinking water gu ide l ines require that no fecal col i forms be detected (Federa l -Prov inc ia l -Terr i tor ia l Commi t tee on Dr inking Water , 2006) . W h i l e consecu t i ve s a m p l e s were not co l lected within the 30 days required, the recreat ional limit of 2 0 0 c fu -100 ml" 1 w a s e x c e e d e d at forested s i tes on one o c c a s i o n at H V - 4 (780 cfu-100 ml" 1), and near ly e x c e e d e d at H V - 1 2 (180 cfu-100 ml ' 1 ) . Further, feca l co l i fo rms were detected in 8 7 % s a m p l e s f rom forested s i tes. M a x i m u m 115 Chapter 4 concentrat ions were observed after prolonged dry periods during summer months suggest ing the accumulat ion of fecal materials from wildlife sources during this time. Further, E. coli were detected in all samples from forested sites (concentrations ranged from 1-37 cfu-100 ml' 1). T h e s e results suggest the presence of active wildlife sources of fecal material in the watershed, which are adversely affecting water draining forested subcatchments . Whi le bacterial concentrat ions tend to be low, significant sp ikes were observed indicating the potential for non-bacterial pathogens in these waters (i.e., parasi tes and viruses). A s several residents obtain drinking water from surface sources draining forested subcatchments in the watershed, regular testing for pathogens should be conducted to ensure adequate treatment. 4.3.3. Temporal trends in surface water quality Tempora l trends at different sca les (storm, seasona l , annual) were a s s e s s e d to gain insight regarding the time periods of greatest contributions of agricultural runoff to surface waters. 4.3.3.1. Nutrients Figure 4-8 is a plot of nutrient levels for the entire period of record, along with water levels at the upper hydrometric station (this station is used as it was not inf luenced by the artificial elevation of summer water levels assoc ia ted with recreational use of Hatzic Lake). It is immediately apparent when looking at the aggregate record that N 0 3 " and P 0 4 3 ' levels were highest during the wet s e a s o n of 2002-2003. A compar ison of data for the three wet seasons on record revealed that mean concentrat ions for N 0 3 " were higher in the first wet season than in the second or third (0.64 mg-L" 1 vs . 0.46 mg-L" 1 and 0.42 mg-L" 1). A similar trend was observed for P 0 4 3 " (0.11 mg-L" 1 vs 0.02 mg-L" 1 and 0.01 mg-L" 1). Mean N H 4 + concentrat ions were higher in the first wet season than in the third (0.20 mg-L" 1 vs . 0.13 mg-L" 1) but lower than in the second (0.30 mg-L" 1) . 116 1.6 1.2 0.8 0.4 0 CO co Date (mmm-yy) Figure 4-8 - Nutrient concentrations vs. water level (upper hydrometric station) for the period of record. Note higher concentrations in first wet season. Red bars denote mean concentrations for day of sampling. • Chapter 4 The observed difference is attributed to the extraordinary rainfall events observed in the watershed in the second and third wet s e a s o n s (the two largest events were each greater than one-third of the total rainfall observed from December to Apri l in the first season) . All wet -season sampl ing for the latter two seasons was conducted after these events (both occurred in the autumn). It is likely that these events resulted in bas in-sca le f lushing of nutrient stores, with all subsequent events of smal ler magnitude producing lower nutrient concentrat ions in surface waters than would be expected under more typical rainfall condit ions. A distinct seasona l trend is a lso visible in Figure 4-8. M e a n wet -season N O V concentrat ions were higher than those observed during the dry s e a s o n for all three years of study. Although N addit ions to agricultural soi ls (in the form of fertilizer and manure) are highest during spring and summer months, increased uptake on the land surface and in s t reams (due to plant and microbial metabol ism) reduces the amount of N avai lable for transport. Further, higher f lows during the wet s e a s o n result in greater mobil ization of surface and subsur face sources than occurs in drier, summer months. A weak positive correlation is in fact observed between water depth at the upper hydrometric station and stream N O V concentrat ions when all stations are cons idered simultaneously. Stratifying the correlation by land use reveals dif ferences in this relationship at the site level. Agricultural si tes show a significant positive correlation between N 0 3 " and water depth (rs = 0.622, P < 0.001, n = 98). This is expected as agricultural si tes have greater N 0 3 " stores which are progressively a c c e s s e d as water levels increase during a storm event (e.g., Moreau et a l . , 1998; e.g., Moog and Whit ing, 2002). This correlation increases to rs = 0.727 (P < 0.001, n = 36) when only mainstem stations are cons idered as the cumulat ive upstream agricultural inf luence is greatest at these sites. At forested sites, the relationship between depth and N 0 3 " is negative and weaker (r s = -0.320, P = 0.020, n = 53). There are two factors that likely contributed to this observed trend. Firstly, N 0 3 " stores in forested subcatchments are limited relative to agricultural sites and rapidly depleted as flow increases. A s a result, N 0 3 " concentrat ions at t imes of increased d ischarge are consistently lower. Second ly , these forested st reams drain s teep s lopes and respond to rainfall more quickly than the upper hydrometric station. A s a result, water depth at that station may be a lagging indicator of depth at these sites. Mixed-use sites fall between forested and agricultural stations, with a weak positive correlation (rs = 0.276, n = 62) between water depth and nitrate. This reflects the influence of a smal ler agricultural land base (in terms of percentage area) when compared to sites c lassi f ied as "agricultural." 118 Chapter 4 Ammon ia concentrat ions a lso varied seasonal ly , with higher concentrat ions during, the wet season . Unl ike NOV, a correlation between water levels and N H 4 + was not observed at any site, suggest ing that different p rocesses control the mobil ization and transport of these two N spec ies (cf. K e m p and Dodds, 2001). No consistent seasona l trend was observed for P 0 4 3 \ This is likely due to the low P 0 4 3 ' addit ions assoc ia ted with moderate agricultural intensity. A lso , because P is commonly a limiting nutrient (Wetzel , 1983) any excess P is likely either adsorbed in the soil profile or metabol ized by plants or aquatic biota in s t reams draining agricultural f ields. 4.3.3.2. Bacter ia, A seasona l difference was observed (Table 4-5) for bacteria, with median fecal coliform concentrat ions being significantly higher during summer months when all sites and all seasons are cons idered (P = 0.030, n = 38). Med ian concentrat ions were also higher for total col i forms during summer months, although this difference was not significant. This pattern is in contrast to seasona l trends for nutrients, which had consistently higher concentrat ions in samples col lected during winter months. This difference in both seasona l and inter-annual trends is due to the unique p rocesses controll ing the availability of bacterial vs . chemica l contaminants for mobil ization and transport. With continual addition of animal wastes and fertil izers to agricultural soi ls, both N and P accumulate in the soil profile and underlying groundwaters over t ime (as descr ibed in Chapter 2). In the absence of rainfall, plant and microbial metabol ism (and, to a lesser extent, atmospher ic losses) are the primary p rocesses acting to reduce these stores. However, as demonstrated by Schre ier et a l . (2003), the application of nutrients in excess of the amounts lost to these p rocesses is common . The resulting surplus nutrients are therefore avai lable for transport during storm events (thus the positive correlation between stream flow and N 0 3 ' concentrat ions descr ibed above) . 119 Chapter 4 Table 4-5 - Descriptive statistics for total and fecal coliform by land use and season. Bacteria Land Use Season N Min. Max. Mean Median SD Total coliform Agricultural S u m m e r 31 100 8,500 3,007 2,580 2,253 Winter 54 110 29,400 2,241 1,185 4,073 Forested S u m m e r 17 10 5,000 778 260 1,302 Winter 28 22 700 143 105 143 Mixed S u m m e r 20 30 14,700 3,142 1,795 3,764 Winter 32 97 5,600 1,208 480 1,513 Urban Summer 4 1,700 7,400 5,240 5,930 2,531 Winter 6 470 3,100 1,845 1,955 1,279 Feca l coliform Agricultural Summer 31 14 2,900 322 120 629 Winter 56 0 10,400 334 45 1,412 Forested S u m m e r 17 1 710 64 10 172 Winter 29 0 10 5 2 5 Mixed Summer 20 1 7,400 735 73 1,718 Winter 33 2 900 90 20 196 Urban S u m m e r 4 230 1,410 748 675 556 Winter 6 1 118 56 48 46 Bacterial concentrat ions in agricultural soi ls, on the other hand, are controlled to a great extent by the dynamics of bacterial decay, which prevent the accumulat ion of significant stores over time (Beaudeau et al . , 2001 ; Jam ieson et al . , 2004). Whi le manure application rates are highest during spring and autumn, the combinat ion of bacterial inactivation, deplet ion by early wet season rains and prohibitions on manure spreading in late autumn and winter months results in consistently lower bacterial concentrat ions in surface waters during the wet season . It should be noted that lower bacterial concentrat ions do not necessar i ly indicate a lower risk of contamination of surface waters with waterborne pathogens, as more resilient organ isms (viruses and parasites) are not subject to in-situ decay at the s a m e rates. The highest median concentrat ions for both fecal and total col i forms in the watershed were observed on September 16, 2003. This is notable as sampl ing on this day coinc ided with a relatively smal l rainfall event (16.5 mm on September 16, and a total of 31.7 mm from the 14 , h -16 t h ) , which was preceded by two months of dry weather (total rainfall over the previous 60 days was 33.3 mm). Max imum fecal coliform concentrat ions for 10 sites were recorded on this day, and four samples had concentrat ions > 2,000 cfu-100 ml" 1 . Two of these were at mixed sites located downstream of cattle operat ions (HV-5 and HV-13 , with va lues of 2,000 and 2,400 cfu-100 ml" 1 , respectively). The two others were at HV-17 and HV-19, the lowermost stations on Hatzic S lough (2,900 and 2,120 cfu-100 ml" 1 , respectively). These are Chapter 4 substantial ly higher than the median values for these stations over the course of the study, which ranged from 30-108 cfu-100 ml" 1 . Two factors contributed to the higher concentrat ions observed on this day. Firstly, dry antecedent condit ions al lowed the accumulat ion of bacterial stores on the land surface and in stream sediments, as has been noted in other studies (e.g., Wi lk inson et a l . , 1995). Whi le data regarding the exact timing of manure appl icat ions to agricultural fields during this time are not avai lable, manure spreading during summer months was observed on several occas ions during the study. Resident l ivestock populations were also continually grazing during this per iod. Both activities contributed to the development of surface bacterial stores in the absence of any appreciable rainfall in the previous 60 days. Further, background levels of total and fecal coliform bacteria (at forested sites) were also elevated, suggest ing accumulat ion of bacteria from active wildlife fecal sources (or potentially non-fecal sources) during this warm, dry period. Second ly , samp les were col lected during the rising limb of the storm hydrograph. A s surface and in-stream bacterial sources are a c c e s s e d during the rising limb, bacterial mobil ization and transport is greatest, and tends to increase until the onset of the recess ion limb (Nagels et al . , 2002; Rodgers et al . , 2003 ; Muirhead et al . , 2004). It is therefore likely that samp les col lected during the rising l imb on other dates would have produced similarly elevated va lues. The importance of antecedent condit ions is demonstrated when compar ing this event to another in March , 2005. During s torm-based sampl ing from March 17-21, 2005 (described below) peak fecal coliform concentrat ions were again observed during the rising limb of the hydrograph; however, they did not surpass those observed during the September 16 t h event, which was less than one-third the s ize of the March , 2005 event. Al though the latter event was larger (75.3 mm), and had similar peak intensities (approximately 5 mm-hr" 1), it took place near the end of the 2004-2005 wet s e a s o n , and it was preceded by several storms of greater magnitude, which likely depleted surface and in-stream fecal bacterial stores. 4.3.4. Storm-event dynamics Storm-event sampl ing was conducted at the lower hydrometric station on one occas ion to a s s e s s event-scale relationships between rainfall, streamflow and water quality. Samp les were col lected every 3 hours beginning at 09:00 on March 18, 2005 and ending at 07:15 on March 21 using an I S C O 3700 automated water sampler p laced at the lower hydrometric station. Total rainfall during this t ime was 75.4 121 Chapter 4 mm; with three distinct events totalling 23.6 mm, 15.5 mm and 35.3 mm between March 19-21 (approximately 1 mm fell on the first day of sampling). B e c a u s e these events took place in rapid success ion , each resulted in success ive ly higher water levels, thus providing the opportunity to a s s e s s the response of water-quality var iables to a s tepped hydrograph with three sequential ly higher flow regimes. Figure 4-9 illustrates the response of hydrometric and water-quality var iables to these precipitation events. The initial increase in water level was observed approximately 5.5 hours after the first event began on March 19. Conductivi ty and temperature (not shown) dec reased at the onset of the first event as a result of dilution, with similar responses during the subsequent two events. Chlor ide a lso showed a marked dec rease with rainfall, as descr ibed below. Turbidity va lues at the lower station were not included in this analys is due to sensor fouling. However, data from the upper station are provided for reference. Whi le they should be interpreted with caut ion, they provide an indication of suspended sediment levels in the mainstem river over the course of the three events. 122 Chapter 4 c 2 — ' ~ Slage 2 E * - * * Chloride g • • • Nitrate-N o ,1 •c ^ Si Ammonia-N X X X Orthophosphale-P TP - 1 ! 1 I 1 0.6 — j 2 0.5 to • • * * • • * • • • • • -4 / 1 / 0.3 r 6) £. a 75 0.2 2 0.16 ^ E a o E 0.04 E m CM LO to o « LO in LO LO O §3 C3 C3 (J> o> CM LO in in LO «- o co to a) CO CO CO CO aj — T-T - ^ o i o i o i o i ^ CO CO CO C*} co co ro co « % % cj m x w — i ~ i - CM in in m tfi Q O tO tO o CVJ C\| C J P4 CO CO CO CO O C O C O OT in in in in CO CO CO CO 25000 r— 1600 [— 400 \ - 1200 r 300 800 I 400 100 Figure 4-9 - a) rainfall, stage and chloride and nutrient concentrations, and b) rainfall, stage, turbidity and coliform concentrations measured every three hours from March 18-21, 2005 at the lower hydrometric station (turbidity measured at upper station). 123 Chapter 4 4.3.4.1. Nutrients Of the three nutrient spec ies studied, only N 0 3 " showed an appreciable and consistent response to rainfall and elevated streamflow. Of all var iables measured during these events, streamflow (expressed as stage) and N 0 3 " were most strongly correlated (rs = 0.875, n = 24). For each event, peak N 0 3 " concentrat ions were observed early on the rising limb, and subsequent ly dec reased at peak flow, suggest ing either source exhaust ion or dilution with N 0 3 " poor rainwater. Interestingly, as flow receded and water levels dec reased , N 0 3 " increased slightly prior to the next event. This could represent the contribution of subsur face sources to stream N loads. Earl ier sampl ing of shal low wel ls (<10 m) in the watershed revealed N 0 3 ' levels as high as 12 mg-L ' 1 , with 12 wel ls having concentrat ions greater than 3 mg-L ' 1 (Magwood, 2004), suggest ing that groundwater could contribute significantly to stream N 0 3 " . Further, leaching of soil N 0 3 " v ia throughflow may a lso contribute to the observed increase. Chlor ide va lues and conductivity can be used as a proxy for the relative contribution of pre-event water to total flow (Laudon and S laymaker , 1997; Peters and Ratcliffe, 1998). Over the three rainfall events, CI" was significantly positively correlated with conductivity (rs = 0.708, n = 24). Before the onset of the first event, CI" concentrat ions were relatively consistent, with an average of 9.3 mg-L" 1 . During peak flows, concentrat ions reached local min ima (between 6.5-8.1 mg-L" 1) . O n the falling limb of the hydrograph, CI" concentrat ions were rapidly restored to pre-event levels until dilution by the next event began. During the third storm event (the largest of the three), the greatest dilution took place and CI" values dropped to a minimum of 6.5 mg-L" 1 . Nitrate va lues were negatively correlated with both CI" and conductivity (rs = -0.617 and -0.800, respectively; n = 24), as increases in both var iables represent the greater relative contribution of N0 3 " -poo r rainfall to streamflow. It is likely that N 0 3 " loads to the stream at this t ime represent a combinat ion of contributions from event water (transporting N 0 3 " from the land surface and soil profile in over land flow and throughflow, respectively), and pre-event water; however, the relative contribution of each source is not known. The continual increase in N 0 3 " concentrat ions with higher water levels suggests ample N stores in the watershed, both in the soil profile and in d isso lved form in shal low groundwater sources. It was not possib le to investigate hysteresis effects for these events as the autosampler reached capaci ty as rainfall ceased during the final event. 124 Chapter 4 Storm-event and seasona l dynamics of N G Y concentrat ions in surface waters suggest that, despite relatively low nutrient surp luses (Schreier et al . , 2003) N accumulat ion in agricultural soils and groundwater does take place during summer months, and that mobil ization of these stores is a function of rainfall and streamflow. Consistent increases in N O V concentrat ions with river stage, while smal l in absolute terms, suggest that these stores are geographical ly extensive and suscept ib le to transport as they become hydrologically connected to the drainage network. T h e s e event-scale observat ions indicate that throughflow and groundwater do contribute to total stream N O V during these events. W h e n combined with the results descr ibed previously (Sect ion 4.3.3.1), these data suggest that the wet seasons , during which N stores are hydrologically l inked to surface waters, are periods of peak N input. 4.3.4.2. Bacteria Total and fecal col i forms were detected in all samples , including pre-event samp les , suggest ing the presence of active in-stream stores. Both types of bacteria responded similarly across all three storm events, although concentrat ions differed by more than an order of magnitude. M a x i m a for total and fecal col i forms reached 25,000 and 1,400 cfu-100 ml" 1 , respectively. During the first event, peaks in total and fecal coliform concentrat ions occurred 30 minutes after peak flow. With the pass ing of the f lood peak, concentrat ions for both types of bacteria dec reased , but remained above pre-event levels. Concentrat ions for the first sample col lected on the next rising limb remained low as depth had barely su rpassed the previous peak (by 1 cm) and new bacterial stores had not been a c c e s s e d . The large increase observed in the next sample (during the second event) indicates that streamflow had increased sufficiently to mobil ize new bacterial sources . Total col i forms and fecal coliforms reached local max ima just prior to the peak in flow and near the end of the storm peak, respectively. The third and largest event produced the highest fecal coliform concentrat ions and the third highest total coliform counts observed over the three days. Both peaked on the rising limb, approximately four hours before maximum flow, and dec reased steadily over the next six hours until sampl ing conc luded. It is worth noting that the min ima observed at the conclus ion of the third event were nearly an order of magnitude higher than pre-event levels. This indicates that, while sources a c c e s s e d by increased flow were being depleted, they were still actively contributing to significantly e levated bacterial concentrat ions even on the receding limb. 125 : ; Chapter 4 A compar ison with CI ' data indicates that elevated bacterial concentrat ions were often assoc ia ted with CI" minima. Chlor ide concentrat ions were negatively correlated with total coliform (rs = 0.505, P = 0.012, n = 24) and fecal coliform (rs = 0.678, P < 0.001, n = 24). During these minima, relative contributions to streamflow by rainfall were highest, suggest ing that over land transport was a dominant transfer mechan ism for bacter ia from the land sur face. However, the fact that this relationship is not stronger suggests that a portion of the bacterial load is derived from in-stream sources. A s pre-storm CI" concentrat ions are re-establ ished in the time between rainfall events, bacterial concentrat ions appear to stabil ize or dec rease , but remain above pre-event levels. This reflects the ongoing contribution of in-stream sources as overland flow dec reases . Turbidity data from the upper hydrometric station provide an indication of the timing of sediment transport to the mainstem over the course of the three events. A s there was a slight delay in the response of the lower station to rainfall (as descr ibed in Sect ion 4.3.1.1) the turbidity data are likely two-three hours ahead of the data descr ibed above. A s illustrated in Figure 4-9, when this time difference is accounted for, peaks in turbidity coincide with the rising limb of each event and with peaks for total and fecal col i forms. This would be expected as these turbidity peaks represent sediment contributions from overland flow and resuspens ion of in-stream sources, both of which would contribute to e levated bacterial concentrat ions. 4.4. Conclusions The impacts of land use on surface water quality are of growing concern, particularly in the L F V due to increasing intensification of agricultural activities, population growth and a dependence upon local water suppl ies, many of which utilize surface sources . Whi le agricultural land use in the Hatzic watershed is not as intense as e lsewhere in the L F V , its inf luence on water quality is apparent. B a s e d on this assessmen t of surface-water quality over 2.5 years, the following conc lus ions can be drawn: 1) Moderate agricultural intensity results in minor nutrient contributions to surface waters The influence of agricultural activities on surface waters was observed throughout the Hatzic watershed with consistently higher nutrient concentrat ions observed for agricultural and mixed sites. However, nutrient delivery to surface waters was relatively limited and comparable to contributions from the urban site. This reflects the moderate intensity of farming activities in the watershed [N concentrat ions were consistently below Canad ian Environmental Quality Guide l ines for the protection of aquatic life (13-126 Chapter 4 16 mg-L' 1 ) and P concentrat ions were below detection limits in 4 4 % of samples] . These values place surface waters in the ol igotrophic-mesotrophic range (Wetzel, 1983), and are considerably lower than surface water concentrat ions in other agricultural ly-dominated watersheds in the L F V (Schreier et al . , 1999; Be rka et al . , 2001 ; Schreier et al . , 2004). 2) Moderate agricultural intensity produces significant bacterial contributions to surface waters In terms of risk, results from the present study suggest that fecal contamination from livestock and wildlife represent the greatest water-quality hazard in the Hatzic watershed. Feca l col i forms were detected in 9 8 % of samples col lected (N = 87) in agricultural subcatchments , indicating the presence of robust sources of fecal contamination. Of particular interest was the fact that individual l ivestock operat ions acted as significant point sources of fecal contamination at mixed sites. S o m e groups have found strong positive correlations between different indices of agricultural intensity and water-quality var iables (e.g., Be rka et a l . , 2001), while others have found weak correlation, or no relationship (e.g., Hunter et a l . , 1999). Consistent increases in bacterial concentrat ions were observed in a downstream direction (a result of increasing contributing a rea under agriculture). However, the data for mixed sites illustrate that this is a poor predictor of water-quality risk as individual agricultural operat ions can act as significant point sources of contaminants. 3) Mid-summer or early wet season rainfall events represent high-risk periods for bacterial contamination of surface waters Antecedent condit ions were observed to play a strong role in determining the risk of bacterial contamination of surface waters. Thus , the highest bacterial concentrat ions were observed during a relatively smal l storm event, after a prolonged dry period during which bacterial accumulat ion in surface soils of both forested and agricultural sites took place. A s a result, fecal coliform concentrat ions in sites draining both types of subcatchments exceeded Health C a n a d a recreational water quality guidel ines at 11 out of 18 sampl ing sites, in many c a s e s by more than an order of magnitude. The summer months appear to be t imes of peak bacterial concentrat ions in surface waters in forested and agricultural catchments, particularly after prolonged periods of dry weather during which fecal matter accumulates on the land sur face. This would suggest that, in B C , mid-summer rainfall events 127 Chapter 4 and the onset of wet s e a s o n rains represent t imes of significant risk in terms of pathogen contributions to sur face waters. 4) Active wildlife sources of fecal contamination are present in undeveloped subcatchments Background nutrient concentrat ions in forested subcatchments were consistently low and appeared to represent little risk in terms of human or ecosys tem health. However, bacterial concentrat ions (including E. coli and total and fecal coliforms) were relatively high in some undeveloped subcatchments . This suggests that fecal inputs to surface waters from wildlife sources was taking p lace, and could represent a risk to human health where these streams are used as sources of drinking water. Severa l residents obtain drinking water from surface sources draining forested subcatchments. W h e r e this occurs , regular testing for pathogens is recommended in order to a s s e s s the risk assoc ia ted with these water sources. The observat ion of fecal bacteria in undeveloped subcatchments has implications for the numerous drinking-water sys tems in B C that rely on surface-water sources . The consis tency in patterns of bacterial concentrat ions in these subcatchments suggest they are representative of those in undisturbed catchments used as drinking-water sources e lsewhere in the province (e.g., G V R D , Capi tal Regional District, etc.). A s stated above, the period of greatest risk for contamination of these water sources appears to be during mid-summer or early wet -season rainfall events. 5) Extreme rainfall events impact catchment-scale nutrient stores and nutrient contributions to surface waters Inter-annual variation in nutrient concentrat ions was observed and revealed the role of extreme events in depleting catchment-scale nutrient stores. This trend is particularly evident in the Hatzic watershed due to relatively moderate nutrient addit ions which result in s lower recovery of these stores over time. Bacterial concentrat ions did not exhibit a similar response, reflecting the more transient nature of bacterial stores (they do not persist from season to season) , and the relative speed with which they are replenished on an annual bas is . 128 Chapter 5 5. The effect of agricultural intensity on surface-water quality in the Lower Fraser Valley 5.1. Introduction In Chapter 4 it was demonstrated that moderate agricultural intensity can result in measurable impairment of surface waters, particularly in terms of bacterial concentrat ions. Land-use intensity in the L F V varies significantly across watersheds; however, little work has been done to evaluate the link between agricultural intensity and surface-water quality in this region. This chapter a ims to do so by assess ing surface-water quality across three watersheds of varying agricultural intensity (the Hatzic, Elk Creek and Sa lmon watersheds) . In particular, the objectives of this chapter are to: 1) determine if spatial and temporal trends in background levels of nutrient and bacterial concentrat ions are similar in undeveloped catchments across watersheds, 2) determine if the effect of land-use intensity on nutrient and bacterial concentrat ions can be detected through a compar ison of spatial and temporal trends across the three watersheds and 3) a s s e s s if agricultural intensity inf luences watershed-scale patterns of nutrient and bacterial mobil izat ion, transport and inputs to surface waters. 5.2. Methods Methods used for the collection of field samples and for laboratory analys is are descr ibed in detail in Chapter 3. Samp le collection in the three watersheds was t imed to occur on the s a m e days to ensure similar environmental condit ions across all sites. Statistical methods used in this chapter for analysis of correlations and compar isons of means are also outl ined in Chapter 3. 5.3. Agricultural intensity 5.3.1. Quantifying agricultural intensity In order to a s s e s s agricultural land-use intensity for the Hatzic, Elk Creek and Sa lmon watersheds it is necessary to quantify the intensity of crop and livestock operat ions, as both contribute to the impairment of water quality through manure and fertilizer appl icat ions. A n effective way to determine agricultural intensity is through the use of nutrient budgets to calculate surp luses or deficits based on total 129 Chapter 5 nutrient inputs (fertilizer, manure, atmospher ic deposit ion) and outputs (crop growth, atmospher ic losses) for a given agricultural a rea . Any surplus represents the potential nutrient loss to surface waters and groundwater (Oborn et a l . , 2003). Schre ier et a l . (2003) calculated nutrient surp luses by E A in the L F V based on data from the Canad ian C e n s u s of Agriculture (1986-2001) conducted by Statist ics C a n a d a . These data are used here to determine the potential for nutrient enrichment of surface waters based upon the availability of surplus N and P. Animal density is a lso an effective index of the intensity of l ivestock operat ions. To measure density in a reas with multiple l ivestock types, a method is required to convert all an imals to equivalent units. This can be accompl ished using feed requirements or waste production characterist ics and results in a value of "animal unit equivalents" or A U E ' s (Beaul ieu et a l . , 2001). A s waste production is the primary concern in this study, the latter method will be used. Convers ion factors for A U E ' s have been deve loped by several provincial agricultural ministries, but in a recent assessmen t of l ivestock density across C a n a d a , Beaul ieu et al . (2001) developed a common set of coefficients to be appl ied nationally (Table 5-1). O n c e l ivestock numbers are converted to A U E ' s , this value is divided by the total area under agriculture to determine a measure of density ( A U E - h a 1 ) (Beaul ieu et a l . , 2001). Table 5-1 - Animal unit conversion coefficients (Beaulieu et al., 2001) Livestock Type An imal Type A U E Convers ion Factor 1 Poultry Broilers 200 Layers 120 Pullets 300 Catt le Dairy cow 0.75 Dairy bull 0.75 Beef cow 1.0 Heifer 1.2 Steers 1.3 Ca lves 3.8 Pigs Boars /sows 5 S h e e p R a m s 7 E w e s 5 Others Horses 0.75 Goa ts 7 - number of animals that equal 1 A U E . 5.3.2. Hatzic, Elk Creek and Salmon watersheds 5.3.2.1. Nutrient Surpluses B a s e d on an assessmen t of manure production, fertilizer appl icat ions, uptake by crops and atmospher ic interactions, Schreier et al. (2003) calculated nutrient budgets for the N icomen, East Chapter 5 Chi l l iwack and North Langley E A ' s (Table 5-2). These E A ' s contain the Hatzic, Elk Creek and S a lm on watersheds, respectively. Surp luses or deficits were expressed in kg-ha" 1 for 2001 . For the N icomen E A , smal l N and P surp luses of 13 kg-ha" 1 and 16 kg-ha" 1 , respectively, were observed. A s agricultural intensity in the Hatzic watershed is lower than that throughout the rest of the N icomen E A (based on field observat ions, and as illustrated by nutrient concentrat ions in surface waters descr ibed in Chapter 4), this is likely an overest imate of surplus nutrients in the Hatzic watershed. Table 5-2 - Land-use distributions and nutrient surpluses for the Hatzic, Elk Creek and Salmon watersheds (note that nutrient surpluses are calculated based on the EA's within which these watersheds are located) Hatzic Elk Creek Salmon Genera l A r e a (ha) 8400 3300 8000 Forest (ha / %) 5 7 1 2 / 6 8 2343 / 71 4 0 0 / 5 Agriculture (ha / %) 1 1 4 0 / 1 4 693 / 21 3600 / 45 Urban/residential (ha / %) 924 / 1 1 6 6 / 2 2880 / 36 O t h e r ( h a / % ) 5 8 8 / 7 1 9 8 / 6 1 1 2 0 / 1 4 Nutrient Nitrogen (kg/cropped ha) 13 132 18 1 surp luses Phosphorus (kg/cropped 16 65 30 ha) 1 - note that N surp luses within the Sa lmon watershed were found to be significantly higher than for the E A as a whole, as descr ibed by Schre ier et al . (1999). In the North Langley E A , N and P surp luses of 18 kg-ha" 1 and 30 kg-ha" 1 , respectively, were observed. This estimate masks significantly higher N appl icat ions in s o m e parts of the Sa lmon watershed. A detai led analysis of nutrient surp luses in the Sa lmon watershed (Schreier et al. , 1999) revealed that local ized N 0 3 " surp luses were in excess of 100 kg-ha" 1 . Schre ier et a l . (1999) a lso est imated surplus N applications over the highly-vulnerable Hopington aquifer, which underl ies the middle sect ion of the watershed. Us ing a model that accounted for N derived from commerc ia l agriculture (fertilizers), hobby farms (manure), septic sys tems and atmospheric inputs, N surp luses ranging from 9-268 kg-ha" 1 were calculated, with an average of 68 kg-ha" 1 in the area over the aquifer. Nitrogen and P surp luses of 159 kg-ha" 1 and 65 kg-ha" 1 , respectively, were calculated for East Chi l l iwack. Us ing C e n s u s of Agriculture data, Macdona ld (2000) used residual, or surplus N as an indicator of agricultural production intensity across C a n a d a . Reg ions were ass igned to one of four c l asses based on total surplus N as fol lows: C l a s s 1 = < 20 kg-ha" 1 , C l a s s 2 = 21-40 kg-ha" 1 , C l a s s 3 = 41-60 131 Chapter 5 kg-ha" 1 and C l a s s 4 = > 60 kg-ha" 1 . Much of the L F V was ass igned to C l a s s 4, which was descr ibed as having significant potential for N losses to surface water and groundwater. 5.3.2.2. Animal Densities Animal densit ies in the Hatzic watershed were low, and were est imated based on the number of dairy, beef, sheep and horse operations identified during land-use surveys conducted by the Fraser Val ley Regiona l District, in conjunction with the B C Ministry of Agriculture (Graham Daneluz, personal communicat ion). B e c a u s e census data on animal numbers are only avai lable for the N icomen E A , assumpt ions were made regarding the number of animals per farm in the watershed (Table 5-3), as agricultural intensity is not consistent across the E A . These values were based on a review of census data and on animal numbers observed during field work and farm visits. B a s e d on these visits and observat ions of pasture fields over several years of fieldwork, the mid-range est imate is taken to be the most accurate, providing a stocking density for the watershed of 1.25 A U E - h a " 1 . Table 5-3 - Estimates of livestock populations in the Hatzic watershed Dairy Beef Horse1 Sheep* AUE AUE-ha"1 # of farms 11 15 33 3 N A N A Animals / farm 20 20 2 20 692 0.61 (low estimate) A n i m a l s / f a r m 40 40 5 40 1424 1.25 Mid-range estimate A n i m a l s / f a r m 60 60 8 60 2158 1.89 High estimate 1 - Low values are used for horses as the total number of horses in the entire E A was only 210 (Schreier et al . , 2003) 2 - An average convers ion factor of 6.0 was used for sheep (rams = 7 and ewes = 5) Stocking densit ies in the Elk Creek watershed were significantly higher than in the Hatzic watershed, based on calculat ions for the E A as a whole (agricultural intensity is more consistent in the East Chi l l iwack E A than in the N icomen EA) . Va lues for cattle, poultry, horses, goats and sheep were used to calculate a total of 35,445 A U E ' s in the E A , which has a total a rea under non-crop agriculture of 3,204 ha (out of a total of 9,867 ha of agricultural land). This results in a stocking density of 11.04 A U E - h a - 1 for land under l ivestock. Whi le the animal numbers used for this calculat ion apply to the E A as a whole, observat ions of several large poultry and cattle operat ions (at least four large poultry farms and 15 large cattle operations) within the watershed boundary during field work suggest that 11.04 A U E - h a " 1 is a reasonable estimate for the Elk Creek catchment. 132 Chapter 5 For the Sa lmon watershed, l ivestock populat ions are dominated by ch ickens, with far fewer cattle and pigs than in the East Chi l l iwack E A . The total number of A U E ' s in the E A is 15,098, leading to a stocking density of 3.9 A U E - h a " 1 for 3,878 ha of non-crop agricultural land (out of a total of 6,507 ha of agricultural land). The above data indicate that agricultural intensity, based on animal numbers, is lowest in the Hatzic watershed and highest in the Elk Creek watershed, with the Sa lmon watershed falling in between (Table 5-4). Stocking densit ies in the Sa lmon watershed are far below those in the Elk Creek catchment, based on calculat ions for their respective E A ' s . However, it should be noted that the value of 3.9 A U E - h a " 1 in the Sa lmon watershed is above 2.8 A U E - h a " 1 , identified by Breaden and Lovejoy (1990) as the level beyond which contamination of water resources becomes critical. Table 5-4 - Animal stocking densities for the Hatzic, Elk Creek and Salmon watersheds (see text for sources). Watershed AUE's Area under AUE-ha" (area livestock (ha) under livestock) Hatzic 1,424 1,130 1.25 Elk Creek 35,445 3,204 11.04 Sa lmon 15,098 3,878 3.9 It must be acknowledged that the above calculat ions for nutrient surp luses and stocking densit ies are subject to error given the nature of the avai lable data and assumpt ions related to nutrient model ing and animal unit equivalency. However, they do illustrate a substantial gradient in agricultural intensity between the three watersheds that is sufficient to support an analys is of its inf luence on water quality 5 . 4 . Results and discussion 5.4 .1. Comparison of land-use types Surface-water quality was compared across similar land-use types in the three watersheds to a s s e s s the influence of land-use intensity on nutrient and bacterial concentrat ions (Table 5-5). This analysis also provides an opportunity to compare background levels of nutrients and bacter ia at forested sites across watersheds. Note that there were no forested or mixed sites in the Sa lmon watershed, and therefore only data from the Hatzic and Elk Creek watersheds are compared for these land-use categor ies. 133 Chapter 5 Table 5-5 - Water-quality data by land use and watershed (Salmon data include groundwater). Watershed Land-use Parameter N Min. Mean Median Max. Standard (mg-L" 1 / cfu-100 ml"1) Deviation Hatzic Forested N 0 3 " 77 0.03 0.3 0.26 1.25 0.18 N H 4 + 77 0.05 0.15 0.1 0.48 0.12 P 0 4 3 " 71 0.01 0.04 0.04 0.28 0.06 Feca l coliform 46 0 27 9 710 106 Total coliform 45 10 383 114 5,000 852 Mixed N C V 87 0.20 0.57 0.46 2.51 0.39 N H 4 + 87 0.05 0.17 0.11 0.68 0.14 P 0 4 3 " 80 0.01 0.06 0.03 0.35 0.08 Feca l coliform 53 1 333 40 7,400 1096 Total coliform 52 30 1,951 850 14,700 2751 Agricultural N C V 137 0.03 0.41 0.34 1.76 0.31 N H 4 + 137 0.05 0.20 0.12 1.22 0.18 P 0 4 3 " 126 0.01 0.07 0.04 0.40 0.08 Feca l coliform 87 0 330 64 10,400 1189 Total coliform 85 100 2,520 1,780 29,400 3524 Urban N C V 16 0.03 0.10 0.08 0.31 0.08 N H 4 + 16 0.39 0.67 0.59 1.17 0.27 P 0 4 3 " 14 0.01 0.06 0.03 0.34 0.09 Feca l coliform 10 1 333 111 1,410 481 Total coliform 10 470 3,200 2,950 7,400 2473 Elk C reek Forested N O V 15 0.25 0.40 0.42 0.53 0.09 N H 4 + 15 0.05 0.15 0.13 0.30 0.09 P 0 4 3 " 15 0.01 0.02 0.01 0.12 0.04 Feca l coliform 15 2 26 10 100 28 Total coliform 15 18 331 170 1,630 482 Mixed N C V 38 0.29 0.90 0.89 1.657 0.38 N H 4 + 38 0.05 0.17 0.14 0.38 0.11 P 0 4 3 " 38 0.01 0.03 0.01 0.18 0.05 Feca l coliform 38 2 291 45 2,200 582 Total coliform 38 50 3,068 1,200 31,000 5746 Agricultural N C V 44 0.03 1.163 0.82 4.25 0.93 N H 4 + 43 0.05 0.46 0.32 3.58 0.62 P 0 4 3 " 43 0.01 0.05 0.02 0.28 0.07 Feca l coliform 42 18 1,717 270 20,500 4475 Total coliform 42 58 8,754 3,930 52,200 12277 Urban N 0 3 " 15 0.71 1.013 1.12 1.28 0.22 N H 4 + 15 0.05 0.13 0.12 0.27 0.08 P 0 4 3 " 15 0.01 0.04 0.01 0.24 0.07 Feca l coliform 15 6 1,137 240 6,100 1857 Total coliform 15 130 5,091 2,900 19,100 5975 Sa lmon Agricultural N C V 34 0.11 3.50 2.69 12.00 2.92 N H 4 + 34 0.05 0.08 0.07 0.17 0.03 P 0 4 3 " 34 0.01 0.02 0.02 0.10 0.02 Feca l coliform 33 0 231 61 3,700 672 Total coliform 32 0 1,936 860 7,700 2280 134 Chapter 5 5.4.1.1. Forested sites Median fecal coliform concentrations at forested sites were 9 cfu-100 ml" 1 and 10 cfu-100 m l 1 for the Hatzic and Elk Creek catchments, respectively (see Table 5-5 and Figure 5-1). Median total coliform concentrations were higher in forested sites of the Elk Creek watershed (170 cfu-100 ml" 1 vs . 114 cfu-100 ml" 1 in the Hatzic watershed), but this difference was not statistically significant. Total and fecal coliform concentrations were positively correlated in the Hatzic watershed (rs = 0.799, n = 39) and Elk Creek watershed (rs = 0.732, n = 15) suggest ing that concentrations of both are control led by similar p rocesses . I Fecal Coliform I Total Coliform Elk Creek a) Watershed b) Watershed Figure 5-1 - a) Median fecal and total coliform concentrations, and b) mean nutrient concentrations at forested sites in the Hatzic and Elk Creek watersheds. Error bars represent 95% confidence intervals. Across all forested sites, fecal coliform concentrations (when averaged for each site over the period of record) were significantly higher during summer months (P = 0.038, n = 14). No significant difference was observed for total coli form. During winter months, total and fecal coliform concentrations in forested sub-catchments of both watersheds were low. Feca l coliform concentrat ions at all sites were at, or below, 10 cfu-100 ml" 1 on all days (except for one sample at E C - 8 col lected on December 13, 2004, when a concentration of 25 cfu-100 ml" 1 was observed). During summer months, variability across sites was greater and the maximum values for all forested sites were observed. The two highest concentrations recorded at forested sites (710 cfu-100 ml" 1 and 180 cfu-100 ml" 1 at HV-4 and HV-12, respectively) were observed on September 16, 2003. These samples were col lected during a minor 135 Chapter 5 rainfall event that took place after 60 days with minimal precipitation (33.3 mm), and represent the highest fecal coliform concentrat ions observed at forested sites during the present study. E levated coliform concentrat ions during summer months are expected, as several factors contribute to increased availability of fecal bacter ia during this t ime. Wildlife activity (waterfowl, rodents, etc.) in and around waterways would be expected to increase during summer months, thus leading to greater potential for direct fecal loading to surface waters. Further, during summer months, warmer air temperatures are more conducive to the survival of thermotolerant bacter ia such as fecal col i forms in the environment. Growth of indicator bacter ia colonies in forest soils has been demonstrated in sub-tropical watersheds (Hardina and Fuj ioka, 1991; Byappanahal l i and Fuj ioka, 1998; So lo-Gabr ie le et a l . , 2000) and temperate environments (Byappanahal l i et a l . , 2003), suggest ing that elevated levels in summer months may represent a combinat ion of increased fecal loading from wildlife and increased in situ reproduction. These data suggest that persistent sources of fecal bacteria exist in forested catchments of both watersheds. Whi le concentrat ions were general ly low, smal l rainfall events produced significantly elevated bacterial concentrat ions in undisturbed subcatchments , particularly after prolonged dry periods (e.g., on September 16, 2003). S u c h condit ions should be considered as high risk for drinking-water sys tems utilizing surface-water sources, even in a reas not inf luenced by agricultural or urban land uses . At forested si tes, N H 4 + concentrat ions also showed a significant seasona l trend, with concentrat ions being higher in winter (P = 0.002, n = 14). No other seasona l dif ferences were observed. Further, within forested si tes, there were no significant dif ferences for N 0 3 " , N H 4 + or P 0 4 3 between watersheds. However, mean concentrat ions for N 0 3 " were slightly higher in Elk Creek sites and, P 0 4 3 " levels were elevated slightly in the Hatzic watershed (Figure 5-1 b). 5.4.1.2. Mixed sites Mixed sites in the Hatzic and Elk Creek catchments contained agricultural, urban and recreational land uses in subcatchments that were otherwise dominated by forest. The influence of these land uses was evident in both watersheds (Table 5-5). M e a n N O V concentrat ions at mixed-use sites were 0.57 mg-L" 1 and 0.90 mg-L" 1 in the Hatzic and Elk Creek catchments, respectively, and were significantly higher than at forested sites in the Hatzic watershed (P = 0.016, n = 10). Med ian total coliform concentrat ions at. mixed sites were 850 cfu-100 ml" 1 and 1200 cfu-100 ml" 1 for Hatzic and Elk Creek , 136 Chapter 5 respectively. In the Hatzic watershed, the median fecal coliform concentration was 40 cfu-100 ml" , compared to 45 cfu-100 ml" 1 in the Elk Creek catchment. A compar ison of mixed sites between the two watersheds revealed significant dif ferences only for P 0 4 3 " (P = 0.008, n = 10). M e a n N 0 3 " concentrat ions were higher in Elk Creek , due primarily to elevated values at E C - 1 3 and E C - 1 4 (1.23 mg-L" 1 and 1.35 mg-L" 1 , respectively), but this difference was not significant when all data for each site were averaged. At E C - 1 3 , this was due to agricultural inf luence from a dairy farm located directly adjacent to the sampl ing site. Site E C - 1 4 is not subject to agricultural influence, but recent timber harvesting activities in the headwaters of this subcatchment likely resulted in increased mobil ization of soil nutrients (cf. Briggs et a l . , 2000), leading to higher-than-expected va lues at this site. M e a n va lues for P 0 4 3 " at E C - 1 4 were a lso higher than would be expected from a predominantly forested subcatchment (fifth highest mean concentration in the watershed), and were likely also due to post-harvest losses. For all var iables other than P 0 4 3 " , no significant difference was observed between the two watersheds at mixed sites. A s the land-use activities upstream of mixed-use sites vary significantly from site to site and from watershed to watershed (cattle operat ions, urban development, golf course, etc.) this was not necessar i ly expected. A review of standard deviat ions indicated substantial variability in these outcomes within this land-use category, but there was no significant difference when all mixed sites were grouped. Seasona l ly , N H 4 + and N 0 3 " concentrat ions were significantly higher at mixed-use sites in winter when all mixed sites were considered (P = 0.011 and P = 0.043, respectively, n = 20). Similarly, total and fecal coliform concentrat ions were significantly higher in summer (P = 0.019 and 0.005, respectively, n = 20). No statistically significant dif ferences were observed for other var iables between seasons . 5.4.1.3. Agricultural sites Hatzic The maximum bacterial concentrat ions observed at agricultural si tes in the Hatzic watershed were observed at HV-18 . The maximum total coliform concentration observed was 29,400 cfu-100 ml" 1 while the peak fecal coliform value was 10,400 cfu-100 ml" 1 . Feca l and total coliform concentrat ions were not consistently high, as illustrated by far lower median values of 64 cfu-100 ml" 1 and 1,780 cfu-100 ml" 1 , 137 Chapter 5 respectively. Max imum nutrient concentrat ions for N 0 3 " , N H 4 + and P 0 4 3 " were 1.76 mg-L" 1 , 1.22 mg-L" 1 and 0.40 mg-L" 1 and were observed at HV-16 , HV-18 and HV-9 , respectively. B a s e d on a seasona l compar ison of concentrat ions N 0 3 " and N H 4 + concentrat ions were highest during winter months (P < 0.001 and P = 0.007, respectively, n = 16). Total and fecal coliform concentrat ions were highest in summer months (P = 0.028 and P = 0.038, respectively, n = 16). Elk Creek In the Elk Creek catchment, peak bacterial concentrat ions were regularly observed in two regions of the watershed: 1) at E C - 4 , E C - 5 and E C - 6 , the three sites with contributing areas containing 1 0 0 % agriculture, located in the low-lying, northeastern portion of the watershed, and 2) at the urban sites E C -11 and E C - 1 2 in the southwest corner of the watershed (descr ibed below). Max imum total and fecal coliform concentrat ions were observed at E C - 4 (52,200 cfu-100 ml" 1 and 20,500 cfu-100 ml" 1 , respectively). Median concentrat ions at this site were more than an order of magnitude higher than at any other site during this study. These values can be attributed to a high-density dairy cattle farm located directly east of the sampl ing site, adjacent to the ditch where E C - 4 samples were col lected. Numerous cattle (50-100) were consistently observed grazing at this site, and bacterial concentrat ions suggest that fecal sources are c losely l inked to surface waters at this location. The highest observed nutrient concentrat ions in the Elk Creek watershed were a lso recorded at E C - 4 , E C - 5 and E C - 6 . The peak N 0 3 " value (4.25 mg-L" 1) and the peak P 0 4 3 ' va lue (0.28 mg-L" 1) were both observed at E C - 6 . The max imum N H 4 + concentration (3.58 mg-L" 1 , almost three t imes higher than the next highest value) was recorded at E C - 4 providing further ev idence of a direct link between nearby high-density cattle operat ions and water quality in this region. A s in the Hatzic watershed, N 0 3 " concentrat ions were significantly higher in summer months (P = 0.002, n = 12). Total coliform concentrat ions were highest in summer months (P = 0.041, n = 12). Salmon In the Sa lmon watershed, a maximum fecal coliform concentrat ion of 3,700 cfu-100 ml" 1 was observed at S A - 1 , near the outlet of the watershed. This reflects the cumulative impact of upstream agricultural contributions. The highest N 0 3 " value recorded at any surface-water station during this study (4.44 mg-L" 1) was observed at S A - 5 ; however, e levated values were a lso observed at S A - 1 4 and S A - 1 9 , (3.6 and 2.79 mg-L" 1 , respectively). These elevated values are attributed to 50 large commerc ia l farms and > 100 horse farms in the central portion of the watershed [as note by Schre ier et al . (1999)], which 138 Chapter 5 are directly adjacent to surface waters at, or upstream of, these sites. T h e s e operat ions are also underlain by the unconf ined Hopington aquifer, which is heavily contaminated with nitrates and contributes significantly to N 0 3 " contributions to surface waters at S A - 5 and S A - 7 . The highest N H 4 + values in the watershed (0.17 mg-L" 1) were observed at S A - 1 9 and are also attributed to the relatively intensive agriculture in this region of the watershed. A maximum P 0 4 3 " concentration of 0.10 mg-L" 1 was found at S A - 1 7 . Simi lar to the Hatzic and Elk Creek watersheds, fecal and total coliform concentrat ions were highest in summer months (P < 0.001 for both, n = 24). No significant seasona l difference was observed for N 0 3 " , N H 4 + or P 0 4 3 " . This was due to nutrient contributions from the Hopington aquifer during the drier summer months (described in greater detail below). Inter-watershed comparison A compar ison of bacterial concentrat ions across the three watersheds revealed that land-use intensity exerts strong control over sur face water quality. B a s e d on Mann-Whi tney tests (with signif icance values adjusted using Bonferroni correction), fecal coliform concentrat ions were significantly higher at agricultural si tes in the Elk Creek watershed than in the Hatzic (P = 0.024, n = 14) and Sa lmon (P = 0.009, n = 18) watersheds (Figure 5-2). No significant difference was observed between the Sa lmon and Hatzic watersheds for either variable (likely due to the high maximum values at HV-18) . 139 Chapter 5 S a l m o n Figure 5-2 - Median total and fecal coliform concentrations for agricultural sites in the Hatzic, Elk Creek and Salmon watersheds. Total coliform and fecal coliform values were higher in the Elk Creek watershed than in the Hatzic watershed (P < 0.001 and P = 0.018, respectively) and Salmon watershed (P = 0.030 and P = 0.048, respectively). Error bars represent 95% confidence intervals. The trend for nutrients was not as consistent (Figure 5-3). Ammon ium concentrations were significantly higher in the Elk Creek watershed than in the Sa lmon watershed (P < 0.001, n = 18). Mean N H 4 + concentrat ions in the Elk Creek watershed were also higher than in agricultural sites in the Hatzic watershed, but this difference was not significant. The elevated values in the Elk Creek catchment are attributable to substantial N H 4 + inputs at E C - 4 arising from runoff from intensive cattle operations located nearby. Elevated N H 4 + concentrat ions in surface waters are thought to result from direct application of manure from cattle (Eghbal l et al. , 1997), ch ickens (Pierson et al . , 2001) and hogs (Gangbazo et a l . , 1995; G a n g b a z o et al . , 1999), and levels above 0.02 m g - L 1 may be toxic to fish and aquatic biota (Cooper, 1993). In this instance, direct contributions of fecal material into the slough from cattle were responsible for the high concentrations observed. This is supported by the fact that the highest fecal coliform concentrations observed during the study were also observed at this site. It is worth noting that the next highest N H 4 + concentrat ions were observed at HV-18 , a lso a site under significant influence from cattle operations. 140 Chapter 5 3.00 H c 200 o 1 o c 01 u C o 01 5 LOO H 0.00 f~) Ammonium • Nitrate | < ' 11 f I' • | • I M . i. | 111 it.-T Hatzic Elk Creek Watershed I Salmon Figure 5-3 - Mean NH4', N03" and P043" concentrations at agricultural sites in the Hatzic, Elk Creek and Salmon watersheds (error bars represent 95% confidence intervals). Significantly higher N 0 3 " concentrations were observed in the Sa lmon watershed than in the Elk Creek (P = 0.006, n = 18) or Hatzic systems (P < 0.001, n = 20). G iven the nutrient surpluses descr ibed in Sect ion 5.3.2, this difference is not surprising. Elevated N 0 3 " va lues observed in the Sa lmon watershed in both the dry and wet seasons are attributed to two sources. Firstly, due to significant local ized N surpluses, on the land surface and in the upper soil profile, N 0 3 " contributions to surface waters are significant during the wet season , as descr ibed for the Hatzic watershed in Chapter 4. The second source contributing to high surface-water N 0 3 " concentrat ions is the Hopington aquifer. This unconfined aquifer underlies the central portion of the watershed and contributes significantly to streamflow during baseflow conditions in late summer . Long-term N surpluses in the watershed have led to significant N 0 3 accumulat ion in the Hopington aquifer. Schre ier et al . (1999), in a study of groundwater N 0 3 " levels in the watershed, noted that N 0 3 " levels in 1 3 % of wel ls sampled (n=70) exceeded the 10 mg-L" 1 limit contained within the Canad ian Drinking Water Guidel ines. During the present study, samples col lected from a deep (48 m) groundwater well (SA-1G) drawn from the Hopington aquifer, had mean N 0 3 " concentrations of 11.95 mg-L" 1 . During basef low condit ions, 141 Chapter 5 contributions from N 0 3 - r i c h groundwater result in elevated surface-water N O V concentrat ions during the dry s e a s o n . A s a result, the significant seasona l dif ferences observed in other watersheds were not present in this catchment (it should be noted that seasona l dif ferences were observed at S A - 1 7 where influence from the aquifer is minimal). Groundwater contributions are likely also significant during the wet season , when they are combined with contributions from the land surface and upper soil profile. High groundwater N 0 3 ' levels may a lso be inf luenced by the presence of over 4000 sept ic sys tems in the watershed (Schreier et a l . , 1999). Sept ic sys tems only retain an average of 2 0 - 5 5 % of total N from sewage (Coote and Gregor ich, 2000), and therefore likely add to the elevated surface water concentrat ions observed, particularly during the wet s e a s o n when the local water table is e levated. T h e s e data indicate that significant soil N 0 3 ' stores have accumula ted within the watershed due to long-term surplus N appl icat ions and significant growth in the number of septic sys tems in recent years (Schreier et a l . , 1999). The data also indicate that there is considerable interaction between the surface water and groundwater sys tems. This has significant implications for aquatic ecosys tem health due to the high potential for eutrophic condit ions in these waters, particularly when N 0 3 " loading occurs during summer months. There is a lso a potential threat to human health assoc ia ted with elevated N 0 3 ' levels, as descr ibed in Chapter 2. Orthophosphate concentrat ions were significantly higher in the Hatz ic watershed than in the Sa lmon (P < 0.001, n = 20) and Elk Creek (P = 0.003, n = 18) catchments; however, va lues were low in all three watersheds. There are two factors that most likely contribute to the low values across watersheds. Firstly, P 0 4 3 ' is strongly adsorbed to soil particles, thus limiting its mobility in d isso lved form. Second ly , P is often a limiting nutrient in aquatic and terrestrial sys tems and excess P is rapidly assimi lated through metabol ic activity (Wetzel , 1983). These data suggest that the intensity of l ivestock operat ions in the three watersheds was not sufficient to generate the surp luses needed to produce significantly higher P 0 4 3 ' levels in surface waters. It should be noted; however, that the highest mean P 0 4 3 ' levels in the Hatzic and Elk Creek watersheds were assoc ia ted with intensive cattle operat ions. This trend that has been observed in several other watersheds in C a n a d a and the United States (e.g., Schepe rs and Francis , 1982; Coote and Gregor ich, 2000). It is attributed to dec reased P adsorpt ion capacity in soi ls after prolonged surp luses assoc ia ted with manure application and direct contributions from grazing animals. 142 Chapter 5 5.4.1.4. Urban sites At a watershed sca le , percentage area under urban influence is minimal in the Hatzic and Elk catchments (11% and 2%, respectively). For individual sampl ing sites, the max imum percentages of total contributing a rea under urban land uses were 1 1 . 1 % and 3 7 % for sites EC-11 and HV-20 , respectively. Despite the relatively smal l percentage of total a rea under urban development, urban sites in both watersheds had concentrat ions of nutrients and bacteria that exceeded those observed at several agricultural sampl ing stations. In the Elk Creek watershed the greatest urban influence in terms of contributing a rea was at site E C - 1 1 , with 1 1 . 1 % under residential development. S a m p l e s col lected from EC-11 had the 4 t h and 6 t h highest median fecal and total coliform concentrat ions observed in the watershed, respectively. This inf luence extended downstream to E C - 1 2 where the 7 t h and 5 t h highest median values for these var iables were observed. M e a n concentrat ions for N H 4 + at urban sites were significantly lower than for agricultural sites, but no significant dif ferences were observed between the two groups for N 0 3 " - N or P 0 4 3 " - P . At 3 7 % , the percentage of contributing area under residential development for HV-20 in the Hatzic watershed was the highest observed in both watersheds. No significant difference in concentrat ions for any of the variables was observed between the two watersheds or between seasons (this is likely a result of the low sample s ize when all va lues are averaged at the site level). Impairment of surface-water quality as a result of urban development has been documented in several studies and contaminant levels often show strong agreement with the total impervious a rea within a catchment (Arnold and G ibbons , 1996; Hall et a l . , 1999; Mall in et al . , 2001). Arnold and G ibbons (1996) suggested that 1 0 % impervious a rea was the threshold above which deleterious effects in receiving waters could be detected. A similar threshold was observed by Mall in et al . (2001) in two watersheds in North Caro l ina where population and percentage impervious a rea were strongly correlated to fecal coliform concentrat ions in tidal c reeks. Increased fecal bacterial concentrat ions in s t reams influenced by urban development has been attributed to several sources including storm sewer overflows and direct deposit ion of feces by house pets, rodents and other smal l mammals (e.g., raccoons) on paved sur faces and lawns (Young and Thacks ton , 1999; Mal l in et a l . , 2001). This material is then easi ly mobi l ized during storm events and rapidly conveyed to surface waters. In the Elk Creek watershed, samples from EC-11 and E C - 1 2 exceeded the Canad ian guidel ine for recreational water quality (200 cfu-100 ml"1) for four out 143 Chapter 5 of seven samp les , while at HV-20 four out of 10 samples exceeded this value. The guideline requires that the mean of five samples col lected within 30 days not exceed this number, and while insufficient samples were col lected to determine if the guidel ines were officially exceeded , the trend descr ibed above suggests that this is likely. 5.4.2. Cumulative downstream impacts 5.4.2.1. Nutrients A compar ison of cumulat ive downstream impacts on water quality revealed differing patterns ac ross the three watersheds for nutrients. Downstream trends are descr ibed below with reference to N O V concentrat ions as they showed the greatest spatial variation in response to land use. In both the Hatzic and Elk Creek watersheds, a discernible longitudinal trend was observed for N 0 3 ' on the mainstem river. Concentrat ions increased as the percentage of total contributing a rea under agriculture increased from 0 % in the forested headwaters to approximately 1 2 % and 2 0 % near the outlet of the Hatzic and Elk Creek watersheds, respectively (Figure 5-4a and Figure 5-4b). A long the mainstem river, N 0 3 ' concentrat ions were significantly correlated with contributing area under agriculture in the Hatzic watershed (rs = 0.427, P < 0.001, n = 85) and Elk Creek watershed (rs = 0.369, P = 0.023, n = 38). 144 Downstream 0% 1% 9.3% 11% 11.7% r HV-6 HV-11 < — i 1 i hv-14 hv-17 hv-19 Site o 1.00-b) 0%.: Downstream 0.7% 8.1% 18.6% 19.7% ST 1 L E J ' EC-3 Site Downstream 1.2% 10.9% 9.3% 11% 11.7% hv-13 hv-15 HV-14 Site Mainstem i M r j LJ IT" hv-17 hv-19 d) Downstream 100% 100% 18.6% 19.7% Mainstem Site Figure 5-4 - Downstream trends in N03" in the mainstem river of a) the Hatzic and b) the Elk Creek watersheds and for tributaries to the mainstem in c) the Hatzic and d) the Elk Creek watersheds. Percentages represent the proportion of total contributing area under agriculture. Chapter 5 W h e n tributaries to the mainstem were cons idered, this s a m e cumulative trend was often not observed due to significant contributions from agricultural operat ions. In these c a s e s , the relationship between N 0 3 ' and contributing area under agriculture did not hold, as demonstrated at HV-13 where 1.2% agricultural land use in an otherwise forested subcatchment produced median N 0 3 ' levels that exceeded those observed in areas under greater agricultural inf luence (Figure 5-4c). This influence decreased downstream as the tributary reached the mainstem. A similar trend was observed in the Elk Creek watershed (Figure 5-4d). Contributions were derived from sites under 100% agricultural inf luence ( E C - 4 , E C - 5 and E C - 6 ) , that were located along agricultural s loughs found entirely within the agricultural ly-dominated northeast region of the watershed. In both watersheds, N 0 3 ' concentrat ions dec reased downstream of point sources . This was likely a result of both dilution of N O V concentrat ions by mainstem flow, and of in-stream metabol ism of excess N 0 3 ' (Mulhol land and Hill, 1997). These data reflect the influence of sca le on nutrient concentrat ions and longitudinal trends in surface waters. A s noted by Buck et al . (2004), at the watershed sca le , total upstream area under agriculture is often a strong predictor of surface-water nutrient concentrat ions. At smal ler sca les ; however, local land-use activities have a greater inf luence upon water quality as the impact of dilution by upland and groundwater f lows is lower (see also W o o d et al . , 2005). Different trends were observed along the mainstem in the Sa lmon watershed for N 0 3 " concentrat ions, due to the strong interaction between surface water and groundwater (Figure 5-5). Nitrate loading to surface waters increases substantial ly downstream of S A - 9 due to intensive agriculture in the central portion of the watershed. This trend is similar to that observed by Schre ier et al . (1999), and illustrates the significant interaction between groundwater and surface water sys tems in this watershed, particularly during basef low condit ions in summer months. A s a result, the strongly seasona l patterns seen in the Hatzic and Elk Creek catchments were not observed in the Sa lmon watershed. The influence of N 0 3 " contributions from the Hopington aquifer extends downstream beyond the boundary of the aquifer itself, and represents the indirect influence of intensive agricultural pract ices located above this unconfined aquifer in the central portion of the watershed. 146 Chapter 5 5.00-, D o w n s t r e a m 4.00-01 3.00-2.00-1.0CH 0.00-I I I SA-17 SA-9 G r o u n d w a t e r i n f l u e n c e SA-7 SA-4 SA-6 Site SA-3 SA-2 SA-1 ! G r o u n d w a t e r j i n f l u e n c e i.oo-n I SA-19 SA-5 SA-6 SA-3 SA-2 SA-1 b) Site Figure 5-5 - Downstream trends in NOV concentrations in the Salmon watershed, for a) the Salmon river and b) Coghlan Creek above and below its confluence with the Salmon river. Note the significant influence of the Hopington aquifer on surface-water nutrient levels. 147 Chapter 5 5.4.2.2. Bacteria In the Hatzic and Elk Creek watersheds, bacterial concentrat ions showed similar downstream trends to those observed for nutrients. A long the mainstem, fecal coliform concentrat ions showed a significant posit ive correlation with contributing a rea under agriculture for the Hatzic watershed (rs = 0.676, P < 0.001, n = 54) and for the Elk Creek watershed (rs = 0.654, P < 0.001, n = 37). A s with nutrients, trends observed in tributaries were not as consistent. In both watersheds, e levated bacterial concentrat ions were general ly observed in proximity to l ivestock operat ions. In the Hatz ic watershed, median fecal coliform levels cont inued to increase downst ream; however, concentrat ions near l ivestock operat ions (e.g., site HV-13) were highly variable, and often exceeded those observed downst ream (Figure 5-6c). In the Elk Creek catchment, due to the significant bacterial inputs assoc ia ted with sites E C -4, E C - 5 and E C - 6 a decreas ing downstream trend was observed (e.g., Figure 5-6d). 148 a) HV-6 HV-11 Site 3> 3.00-E b) D o w n s t r e a m 0% 0.7% 8.1% 18.6% 19.7% EC-9 EC-8 EC-3 Site X i EC-1 D o w n s t r e a m 1.2% 10.9% 9.3% 11% 11.7% J HV-13 HV-15 HV-14 Site HV-17 HV-19 d) Site Figure 5-6 - Downstream trends in log fecal coliform concentrations in the mainstem river of a) the Hatzic and b) the Elk Creek watersheds and _,. for tributaries to the mainstem in c) the Hatzic and d) the Elk Creek watersheds. Chapter 5 The influence of groundwater contributions was again evident in the Sa lmon watershed. Unlike the Elk Creek and Hatzic watersheds, a consistent downstream increase in bacterial concentrat ions was not observed (Figure 5-7a and Figure 5-7b). This is attributable to dilution of bacterial concentrat ions by groundwater. This influence of groundwater contributions was lower at S A - 1 , likely as a result of inputs from Davidson creek which had consistently higher bacterial concentrat ions than sites over, or downstream of, the aquifer (as measured at SA-14) . a) 3 . 5 0 -E o 3 . 00 -o 2.50-S 2 . 0 0 -o 1.50-H Downstream Groundwater influence I I I I I I 1 I SA-17 SA-9 SA-7 SA-4 SA-6 SA-3 SA-2 SA-1 Site 3.50-E o 3.00-o O 2.50-o o £ 2.00-o O 1.5(H b) Downstream Groundwater influence — I 1 SA-19 SA-5 SA-6 SA-3 Site SA-2 SA-1 Figure 5-7 - Downstream trends in fecal coliform concentrations in the Salmon watershed for a) the Salmon River and b) Coghlan Creek above and below its confluence with the Salmon River. 150 Chapter 5 Differing patterns and intensities in agricultural activity across watersheds did not result in differing downstream trends in agricultural contaminants along mainstem river sys tems in the Hatzic and Elk Creek watersheds. In both watersheds, intensive l ivestock operat ions located in upstream locations contributed significantly to local bacterial and nutrient concentrat ions in tributaries. Upon entering the mainstem river in both watersheds, these contributions were diluted and values from sites along the mainstem represented a more integrated signal of land-use inf luence. The unique cumulat ive downstream impacts in the Sa lmon watershed were due to the strong interactions between groundwater and surface water in this catchment. The unconf ined Hopington aquifer provides a mechan ism for increased N retention within the watershed A s a result of strong l inkages between these two sys tems, lateral and longitudinal nutrient f luxes were different in both spatial and temporal extent when compared to the Hatzic and Elk Creek watersheds. The Hopington aquifer provides a direct link between the intensive agricultural activity in the central portion of the watershed and surface waters throughout the lower half of the catchment, resulting in spatial trends that are not necessar i ly attributable to upstream land-use type or intensity. Temporal ly, the accumulat ion of substantial N stores within the aquifer results in year-round N inputs to the surface-water network. The risk of eutrophication is therefore far greater within the Sa lmon watershed, as N loading during productive summer months is relatively high. 5.5. Conclusions The varying intensity of agricultural activities in watersheds throughout the L F V (described in Chapter 2) necessi tates an understanding of the implications of land-use intensity for surface-water quality and human health. This study, which represents the first multi-year assessmen t of water quality ac ross these three watersheds, addressed this quest ion through an analys is of nutrient and bacterial dynamics as affected by agricultural intensity, and supports the following conc lus ions. 1) Bacterial concentrations in forested subcatchments are consistent across watersheds in the LFV in terms of intensity and timing Background concentrat ions of bacter ia in forested subcatchments of the Elk Creek watershed were not significantly different from that observed in the Hatzic watershed. The consis tency from year-to-151 Chapter 5 year and between these two watersheds suggests that the values and trends in bacterial concentrat ions observed provide a reasonable indication of basel ine levels for these parameters in this region. In both watersheds, bacterial concentrat ions in forested subcatchments were low, but peaked during summer months, reflecting increased wildlife activity and in situ survival and reproduction at this time (as compared to winter months). A s noted in Chapter 4, smal l rainfall events during summer months can increase in-stream bacterial concentrat ions significantly, and therefore represent a t ime of increased water-quality risk. 2) A positive correlation exists between agricultural intensity and surface-water impairment from fecal sources Agricultural intensity (in terms of nutrient surp luses and animal densit ies) was highest in the Elk Creek watershed, with the contributing areas for three sites in this catchment under 100% agricultural land use. This intensity was reflected in indicators of fecal loading as both fecal coliform and N H 4 + concentrat ions were higher at these sites than at any other in the three watersheds. A similar trend was observed within each watershed, with the sites under the greatest agricultural inf luence showing consistently elevated contaminant levels. The strong influence of land-use intensity is a lso visible in the cumulat ive downstream impacts observed in each watershed. This trend suggests that agricultural intensity represents a useful indicator for identifying peak nutrient and bacterial inputs to surface waters. It also indicates that further intensification of agricultural operat ions on the fixed land base within the A L R in this region will lead to increased risk of fecal contamination of surface waters and assoc ia ted risks to human and ecosys tem health. 3) Seasonal trends in surface-water nutrient and bacterial concentrations are similar across land use types Seasona l patterns of nutrient and bacterial concentrat ions were similar across all land use types for the three watersheds. Nutrient concentrat ions were consistently higher in winter (reflecting the hydrological controls on mobil ization and transport), whereas bacterial concentrat ions were higher in summer (reflecting the seasonal i ty of fecal sources and in situ reproduction). In the Sa lmon watershed, seasona l dif ferences for N 0 3 " concentrat ions were not statistically significant, due to the substantial inputs from the Hopington aquifer. The presence of a robust subsur face 152 Chapter 5 N 0 3 ' store within the watershed extends the influence of intensive agriculture in the central portion of the watershed both spatially and temporally. Spatial ly, the high nutrient surplus focused over the aquifer is distributed beyond its borders as a result of strong interactions with the surface-water sys tem. Temporal ly, unlike the Hatzic and Elk Creek watersheds, N 0 3 " contributions to surface waters occur year-round, resulting in a greatly increased risk for eutrophication during summer months when in-stream productivity is highest. 4) Cumulative, downstream impacts are controlled by both land use and hydrological dynamics and are scale dependent A s demonstrated in each watershed, cumulat ive trends in nutrient and bacterial concentrat ions were inf luenced by total a rea under agricultural land use. In the mainstem of the three watersheds, bacterial concentrat ions consistently increased in the downstream direction, and va lues were positively correlated with total upstream area under agriculture. The s a m e pattern was observed in the Hatzic and Elk Creek mainstems for nutrient concentrat ions. In the Sa lmon watershed, the unconf ined Hopington aquifer serves as a long-term store for surplus N arising from intensive agricultural activity in the central portion of the catchment, and extends the influence of these activities downstream through substantial inputs to the Sa lmon River. The influence of sca le in nutrient and bacterial f luxes was evident, particularly in agricultural tributaries where nutrient and bacterial concentrat ions often dec reased in the downstream direction. At the catchment sca le , total area under agriculture was positively correlated with nutrient and bacterial concentrat ions. In smaller-order s t reams, this relationship was not observed due to significant inputs assoc iated with animal a c c e s s and/or manure appl icat ions. This highlights a significant limitation with the use of "area under agriculture" as an indicator of surface-water contaminat ion at all sca les , and the need to target monitoring programs and management opt ions at both the plot and catchment sca le in order to address surface water impairment in agricultural catchments. 153 Chapter 6 6. Absorbance spectroscopy as a tool to detect agricultural influence on water quality 6.1. Introduction A s demonstrated previously, agricultural land-use activities can act as significant point and non-point sources of contaminants such as nutrients and waterborne pathogens. The detection of such influence on surface-water quality is difficult. Whi le concentrat ions of these and other agriculturally-derived contaminants in surface waters are often correlated, these correlat ions are not consistent. A s a result, multiple detection methods are required in order to identify agricultural inf luence. Detection methods for these contaminants are often t ime-consuming and expens ive and can require large water vo lumes, spec ia l ized equipment and training. W h e n monitoring the quality of surface water used for drinking or recreation in agricultural watersheds, the ability to character ize agricultural influence with a rapid, sensit ive technique would allow a more proactive approach to the protection of human and ecosys tem health. Whi le one detection method may not provide sufficient information regarding the presence or concentration of all potential agricultural contaminants, rapid qualitative and/or quantitative detection of agricultural influence qualitatively and/or quantitatively could provide the necessary information to initiate precaut ions or corrective act ions while more detai led analyses are conducted. A useful indicator of agricultural influence on surface waters must either directly reflect the presence of contaminants, or represent a unique attribute of agricultural vs . non-agricultural runoff. A s descr ibed in Chapter 2 (Section 2.7.5.3), absorbance spect roscopy is a potentially useful tool to detect such influence for several reasons. Firstly, absorbance in the U V has been shown to vary proportionately with N 0 3 ' concentrat ions in wastewater sys tems (Ferree and Shannon , 2001) and in surface waters (Crumpton et al . , 1992). Second ly , absorbance in the UV-v is ib le range has been used to character ize chromophoric D O M ( C D O M ) in freshwater and marine environments and to identify source areas based on geophys ica l characterist ics and land use. Finally, absorbance spect roscopy offers several advantages over traditional detection techniques in terms of sample vo lume, preparation t ime and analysis time. The purpose of this chapter is to a s s e s s the utility of absorbance spect roscopy as a tool to rapidly and accurately detect agricultural influence in bulk surface-water and groundwater samples . This 154 Chapter 6 represents the first mult i-watershed assessmen t of absorbance parameters (absorbance, second-derivative absorbance and spectral slope) as qualitative and quantitative indicators of nutrient and bacterial concentrat ions at different spatial and temporal sca les . To accompl ish this, water samp les were col lected from three watersheds (Hatzic watershed, Elk Creek watershed and Sa lmon watershed), each of which is dominated by agricultural activities of varying type and intensity, as descr ibed in Chapter 3. The objectives of this assessmen t were to: 1) determine if UV-V i s absorbance spectroscopy could be used as a qualitative and/or quantitative technique for the detection of agricultural contamination of surface waters, and 2) determine the degree to which absorbance spect roscopy could be used to discriminate between water sources and transport pathways to better understand flow routing and contamination dynamics in agricultural watersheds. 6.2. Methods 6.2.1. Sample collection Samp les were col lected and ana lysed following the techniques descr ibed in Chapter 3. O n three separate dates, samp les were col lected from 11-19 surface-water and groundwater monitoring sites (see Chapter 3 for site locations) in each watershed under dry and wet condit ions (Table 6-1). A ser ies of samples was also col lected over the course of a rainfall event in the Hatz ic watershed to determine if the absorbance properties of water samples reflected event-scale agricultural inf luence on surface waters. 155 Chapter 6 Table 6-1 - Sample collection dates (*) and rainfall (mm) for the previous 5 days for the Hatzic, Elk Creek and Salmon watersheds. Date Hatzic Elk Creek Salmon August 27, 2004 5.84 1.4 0 August 28, 2004 14.48 4.8 2.6 August 29, 2004 0 0 0.2 August 30, 2004 0 0 0 August 31 , 2004* 0 0 0 5-Day Total 20.32 6.2 2.8 December 9, 2004 12.19 4.4 11.6 December 10, 2004 82.8 77.4 56.7 December 11, 2004 0 3.4 1.8 December 12, 2004 0 0 0 December 13, 2004* 9.14 7.4 5.2 5-Day Total 104.13 92.6 75.3 February 11, 2005 0 0 0 February 12, 2005 14.48 13.8 8.4 February 13, 2005 1.52 6 10.8 February 14, 2005 4.32 20.2 3.8 February 15, 2005* 0.25 0 0 5-Day Total 20.57 40 23 6.2.2. Spectroscopic analysis Water samples were ana lyzed following the techniques descr ibed in Chapter 3, with complete absorbance spectra col lected from 200-800 nm. The shape and features of these spect ra were first a s s e s s e d qualitatively to determine if samples col lected from different sources d isplayed characterist ic patterns. Quantitative ana lyses were then conducted using absorbance data for specif ic wavelengths in order to a s s e s s spatial and temporal patterns and to evaluate correlat ions with nutrient and bacterial concentrat ions in water samples . The wavelengths used in this study are listed in Table 6-2, along with common water-quality parameters for which they are used as proxies. 156 Chapter 6 Table 6-2 - Absorbance wavelengths used as proxies for common water-quality parameters. Wavelength Parameter Reference A220 N 0 3 " (Dress et al . , 1998) 2 n d derivative @ 224 nm N C V (Cahil l , 1979; Crumpton et a l . , 1992; Ferree and Shannon , 2001) A254 B O D (Reynolds and A h m a d , 1997) A280 Aromaticity; B O D (Chin et al . , 1994; Brookman, 1997) A300 C D O M concentration (Green, 1992; Del Cast i l lo et al . , 1999) A350 C D O M concentration (Kowa lczuke t a l . , 2003) A440 Colour/concentrat ion of humic substances, D O C (Cuthbert and Delgiorgio, 1992; Yacob i et a l . , 2003 ; McDona ld et a l . , 2004) Spectra l s lope D O M composi t ion, humic vs. fulvic content (Carder et a l . , 1989; Blough and Del Vecch io , 2002) * A 2 2 0 = absorbance at 220 nm 6.2.3. Spectral slope Spectral s lope (S), def ined in Chapter 2 (Section 2.7.5.1), was calculated as the s lope of a least-squares regression line through log-transformed absorbance data (Green and Blough, 1994; Yacob i et al . , 2003; Zanard i -Lamardo et al . , 2004) as shown in Figure 6-1. Abso rbance spect ra were log-transformed to generate a near-l inear plot to which a least-squares line could be fit in order to obtain an approximation of S. It has been noted that the use of a linear fit (as opposed to a non-l inear fit to the raw spectrum) can enhance the relative weighting of low absorbance values at longer wavelengths where absorbance readings can be near detection limits, thus overestimating S (Twardowski et al . , 2004). This can be particularly pronounced in transformed spect ra for samples with low C D O M concentrat ions as they display significant variation from linearity at longer wavelengths. To account for this, the calculat ion of spectral s lope was limited to 290-450 nm, as this wavelength range showed minimal departures from linearity. A s outl ined by Blough and Del Vecch io (2002), this is a commonly -used technique that produces S values with a strong linear correlation to those calculated through non-l inear fitting techniques. Due to the characterist ic shape of absorbance spectra, s lope va lues for least-squares fit l ines are negative. In the literature, S is consistently reported as a positive integer, and therefore, following this convent ion, all S va lues here are reported as posit ive numbers. 157 Chapter 6 Figure 6-1 - a) plot of a typical absorbance spectrum (solid line) and the In-transformed spectrum for the same sample (dashed line); b) least-squares fit to In-transformed absorbance data between 290-450 nm (r2=0.99). 6.3. Results and discussion This section descr ibes the results of absorbance analysis by first providing an overview of the shapes and features of complete absorbance spectra, and then by descr ibing quantitative characterist ics for specif ic absorbance wavelength ranges. Quantitative analysis was focused on three regions of the spectra, as each provided information regarding different aspects of water quality. These regions were: 1) the far-UV (190-200 nm), 2) the U V - C range (220-290 nm) and 3) the visible range, with emphas is on absorbance at 440 nm. Event-scale absorbance trends for the March, 2005 storm event in the Hatzic watershed are then descr ibed. This is followed by a d iscuss ion regarding the value of absorbance as a qualitative and quantitative tool in the analysis of water quality. 6.3.1. Absorbance spectra (200-800 nm) Absorbance spectra for all samples displayed a similar pattern of near-exponential decl ine with increasing wavelength. A visual inspection of spectra from differing land uses and from surface water and groundwater sites revealed consistent structural differences reflecting the varying composi t ion of C D O M in these samples (Figure 6-2). Agricultural sites displayed higher absorbance va lues across the entire wavelength range and relatively featureless spectra, both of which reflect relatively high concentrat ions and more heterogeneous C D O M compared to surface waters without significant agricultural influence. 158 Chapter 6 Absorbance values were lower for samp les col lected from forested and mixed-use sites in the Hatzic and Elk Creek watersheds. Many spect ra from forested and mixed sites showed minimal increases in absorbance with decreas ing wavelengths until reaching a shoulder at approximately 280 nm. Absorbance subsequent ly increased sharply between 2 8 0 - 2 0 0 nm. This shoulder at 280 nm was most pronounced in those samples col lected in August. 159 200 300 400 500 600 700 800 200 300 400 500 600 700 800 Wavelength (nm) b) Wavelength (nm) Figure 6-2- a) absorbance spectra for samples collected in the Hatzic watershed on August 31, 2004, plotted by land use; b) absorbance spectra for samples collected in the Salmon watershed at: 1) SA-1G (dashed, right axis), a deep well with significant N03 contamination (-11 mg-L"') as illustrated by the peak at ~ 220 nm and 2) SA-3G (left axis), a municipal well with N03" concentrations below detection limits. The sharp drop at 350 nm is an analytical artifact. Note the different scales for the two Y-axes. • Chapter 6 Spect ra for samp les col lected from groundwater sources ( S A - 1 G and S A - 3 G ) showed the lowest absorbance va lues across most of the wavelength range, reflecting low C D O M concentrat ions common for groundwaters (Figure 6-2b). Nitrate contamination in these samples was easi ly detected through visual inspect ion. Samp les high in N 0 3 " produced peaks in the absorbance spect ra at - 2 2 0 nm (for reasons descr ibed in Sect ion 6.3.2), as illustrated for station S A - 1 G in Figure 6-2b. 6.3.2. Far UV (190-220 nm) A s descr ibed in Chapter 3, absorbance at 220 nm (A 2 2 0 ) is a good indicator of N 0 3 " concentrat ion due to the strong absorbance of the N 0 3 " ion in the 210-220 nm range. Figure 6-3 illustrates the range in A 2 2 o va lues by land use for all three watersheds combined. A s expected, values were highest where peak N 0 3 ' concentrat ions were observed across all three watersheds (agricultural sites, and, particularly, groundwater site S A - 1 G in the Sa lmon watershed). 3 . 0 0 H 2 . 0 0 H o CM in XI < 1.00-0 . 0 0 H Urban Agricultural Land U s e Figure 6-3 - Absorbance at 220 nm vs. land use for the Hatzic, Salmon and Elk Creek watersheds combined. Linear regression was used to determine how well A 2 2 0 predicted actual N 0 3 " concentrat ions measured using a Lachat Instruments Q u i k C h e m FIA+ 8000. A s illustrated in Figure 6-4a, A 2 2 0 explains 9 1 % of the var iance in measured N 0 3 " va lues (n = 134). For many agricultural sites, absorbance va lues 161 Chapter 6 overest imate N O V concentrat ions. This is most prevalent in s low-moving agricultural s loughs (e.g., H V - 1 , HV-15 , HV-18 , E C - 4 , E C - 5 , and EC-6 ) where C D O M concentrat ions are high due to autochthonous production, most notably for samples col lected during the summer (August). An abundance of C D O M in these samples resulted in increased absorbance in the range of 220 nm, leading to the overestimation observed. Conduct ing the s a m e analysis using second-der ivat ive absorbance at 224 nm (Figure 6-4b), produced a stronger relationship, with absorbance explaining 9 9 . 8 % of the var iance in N 0 3 " concentrat ions (n = 124). Note that this relationship remains strong with the three highest N 0 3 ' va lues removed (r2 = 0.996, n = 121). Further, the relationship is stable across a range of N 0 3 " concentrat ions (from detection limit - 12.0 mg-L - 1 ) . T h e s e results are in accordance with other studies that have compared the results of second-der ivat ive spect roscopy to traditional ion chromatography. Crumpton et al . (1992) found no significant difference between this method and the automated cadmium reduction method for the analysis of surface waters. Similarly, Ferree and Shannon (2001), in an analysis of wastewater samples , observed excellent correlation between the second derivative technique and ion chromatography. This method represents a significant t ime sav ings over the traditional Q u i k C h e m method as each sample can be ana lysed in less than 60 seconds , and requires only that the sample be filtered to remove particulate matter in order to avoid light scattering during the absorpt ion scan . Another advantage, as illustrated in Figure 6-2, is that absorbance spectra a lso provide greater qualitative information than nutrient analysis a lone, as it is possib le to visually differentiate water source (e.g., surface-water vs . groundwater source). 162 Figure 6-4 - a) linear regression of absorbance at 220 nm vs. nitrate concentrations measured using the Lachat QuikChem method, with r2 = 0.91 (r2 = 0.85 with three highest nitrate values removed); b) linear regression of second-derivative absorbance at 224 nm vs. nitrate, with r2 = 0.99 (r2 = 0.99 with three highest nitrate values removed). Chapter 6 6.3.3. UV-C (220-290 nm) The U V - C range of the absorbance spectrum is an area of interest as it e n c o m p a s s e s the range in which 7r-7r* electron transitions occur for phenol ic subs tances , aniline derivat ives, benzo ic ac ids and polycycl ic aromatic hydrocarbons (Chin et a l . , 1994; Peuravuor i and Pihlaja, 1997; Khorassan i et al . , 1998; Duarte et al . , 2003). This transition represents the elevation of an electron from a pi bonding orbital to a pi antibonding orbital, and is the most commonly observed (Lakowicz, 1999). These subs tances are common precursors to (or components of) humic subs tances , particularly those derived from terrestrial sources . A s a result they have the potential to provide insight regarding C D O M source and transportation (Chin e ta l . , 1994). A s descr ibed in Sect ion 6.3.1, many absorbance spect ra col lected in August, 2004, showed a pronounced shoulder at approximately 280 nm. Figure 6-5 and Figure 6-6 show spectra plotted by land use for each day of sampl ing for the Hatzic and Elk Creek watersheds (the Sa lmon watershed is not included as it contained no completely forested subcatchments) . The shoulder at 280 nm was observed primarily in samples col lected in forested and mixed-use subcatchments. Spect ra col lected in agricultural subcatchments showed less definition in this area. This likely reflects higher total organic matter concentrat ions and a greater variety of organic matter sources (al lochthonous as well as autochthonous sources such as algae and rooted macrophytes that were commonly observed in agricultural s loughs), rather than a lack of absorbance at 280 nm. The one except ion to this trend was at HV-18 , where a peak at 280 nm was observed for samples col lected in December and February. 164 August 31 , 2004 December 20, 2004 February 16, 2005 200 300 400 500 600 700 800 200 300 400 500 600 700 800 200 300 400 500 600 700 800 Wavelength (nm) Wavelength (nm) Wavelength (nm) Figure 6-5 - Absorbance spectra for all sites in the Hatzic watershed over the three days of sampling. Note the consistent presence of a shoulder at 280 nm at forested and mixed sites on August 31, 2004. Grouping of spectra by land use is still observed but the distinction is not as obvious as for the Elk Creek due to the lower intensity of land use in this catchment. August 31, 2004 December 20, 2004 February 16, 2005 200 300 400 500 600 700 800 200 300 400 500 600 700 S00 200 300 400 500 600 700 800 Wavelength (nm) Wavelength (nm) Wavelength (nm) Figure 6-6 - Absorbance spectra for all sites in the Elk Creek watershed over the three days of sampling. Note the consistent presence of a shoulder at 280 nm at forested and mixed sites on August 31, 2004. Note also the consistent distinction between land-use types and the consistently high absorbance values for the sites under the greatest degree of agricultural influence (EC-4, EC-5 and EC-6). rji Chapter 6 Absorbance at 280 nm was positively correlated with total upstream area under agriculture (rs = 0.669, P < 0.001, n = 92). Va lues were higher at agricultural sites than at forested and mixed sites in all watersheds (Table 6-3); however, these values were not significant when averaged for each site and signif icance levels were adjusted using Bonferroni correction for multiple compar isons . For agricultural sites, mean A 2 8 o va lues appear to cor respond well with agricultural intensity (Table 6-3). Whi le dif ferences between watersheds were not statistically significant using the above approach, this trend appears to reflect dif ferences in land-use intensity descr ibed in Chapter 5. Table 6-3 - Absorbance at 280 nm by land use for the Hatzic, Elk Creek and Salmon watersheds. Watershed Land use N Minimum Mean Maximum SD 1 Hatzic Forested 12 .021 .050 .084 .020 Mixed 15 ' .025 .041 .074 .013 Agricultural 24 .042 .121 .348 .075 Urban 3 .105. .133 .182 .043 Elk Creek Forested 5 .036 .042 .048 .004 Mixed 13 .023 .047 .096 .025 Agricultural 15 .068 .172 .335 .080 Urban 5 .031 .041 .047 .007 Salmon Agricultural 35 .011 .141 .451 .078 1 - Standard Deviation In the Hatzic watershed, peak A 2 8 o va lues at each station were observed during the August sampl ing (Figure 6-7a), with the only except ions being HV-2 , HV-4 , HV-8 and H V - 1 1 , all of which are forested except HV-8 (mixed). Samp les col lected in the s lowest-moving agricultural s loughs (HV-1 , HV-15, HV-16 and HV-18) and in the urban subcatchment (HV-20) showed the highest A 2 8 0 va lues over the three days of sampl ing. Data col lected from the Elk C reek watershed did not show a similarly consistent pattern. Samp les col lected at s o m e stations showed max ima in August, while others peaked in December (Figure 6-7b). It was noted; however, that samples col lected in c lose proximity and along the s a m e watercourse d isplayed similar patterns in terms of timing for max imum values. In the Sa lmon watershed, peak A 2 8 0 va lues were observed at all stations for samp les col lected in December , except for S A - 1 7 (Figure 6-7c). Peak values for A 2 8 0 were not observed at any station in the three catchments for samples col lected in February, likely reflecting a dilution effect after several months of winter precipitation. 167 Chapter 6 0.40 H 0.30-o CO » 0.20 < Date • 31-AUG-20D4 • 13-DEC-2004 • 15-FEB-20D5 ^ A* J* > «i* «»» i # >.NN . .«? . N W > V a) * * ^ -f ^ ^ -T 4* * f 4* ^ ^ Site 0.00 -—p EC1 EC2 EC3 EC4 ECS EC6 EC7 EC8 EC9 EC10 EC11 EC12 b) S i te 0.50 0.40 » 0.30 (N I 0.20 0.10 0.00 c ) SA-1 SA-2 SA-3 SA-4 SA-5 SA-6 SA-7 SA-9 SA-14 SA-17 SA-19 Site Figure 6-7 - Absorbance at 280 nm by station for the: a) Hatzic, b) Elk and c) Salmon watersheds. 168 Chapter 6 6.3.4. Visible (440 nm) A relationship between land use and mean A 4 4 0 va lues was observed (Table 6-4). W h e n all sites were cons idered, forested and mixed sites consistently had significantly lower A^o va lues than agricultural sites (P < 0.001 for both, n = 32 and n = 36, for forested and mixed site compar isons, respectively). In the Hatz ic watershed, mean A 4 4 0 absorbance at the one urban site (HV-20) was lower than that observed at agricultural sites. In the Elk Creek watershed, mean va lues at urban sites were lower than at agricultural sites (Table 6-4), but not significantly so . Ac ross all watersheds, A 4 4 0 was also positively correlated with total upstream area under agriculture (r s = 0.635, P < 0.001, n = 92). Between watersheds, when consider ing mean va lues for all sites, no significant dif ferences between mean absorbance were observed between the Hatzic and Elk Creek catchments. Table 6-4 - Absorbance at 440 nm by land use for the Hatzic, Elk Creek and Salmon watersheds. Watershed Land use N Minimum Mean Maximum SD 1 H a t z i c F o r e s t e d 12 .000 .006 .020 .005 M i x e d 15 .000 .005 .013 .004 A g r i c u l t u r a l 24 .006 .016 .044 .010 U r b a n 3 .014 .015 .017 .002 E l k C r e e k F o r e s t e d 5 .001 .004 .007 .003 M i x e d 13 .000 .006 .010 .003 A g r i c u l t u r a l 15 .004 .019 .038 .009 U r b a n 5 .004 .007 .008 .002 S a l m o n A g r i c u l t u r a l 31 .004 .019 .067 .013 1 - Standard Deviation 6.3.5. Spectral slope Table 6-5 contains mean spectral s lope (S) va lues according to land use and watershed for the three days of sampl ing. The greatest variability was observed in the Hatzic and Elk Creek watersheds, with max imum values observed at forested sites in August. A compar ison of S values between the two watersheds by land use category revealed no significant dif ferences. Mean S values in the Sa lmon watershed were relatively consistent over the three days of sampl ing, reflecting the comparat ively homogenous agricultural land-use in the catchment. 169 Chapter 6 Table 6-5 - Mean S values by land use and watershed for the three days of sample collection. Watershed Land use Aug. 31,2004 Dec. 13, 2004 Feb. 15, 2005 Hatzic Forested 0.021 0.011 0.014 Mixed 0.019 0.010 0.014 Agricultural 0.014 0.012 0.013 Urban 0.016 0.013 0.012 Elk C reek Forested 0.022 0.014 0.010 Mixed 0.017 0.012 0.012 Agricultural 0.016 0.014 0.014 Urban 0.013 0.010 0.010 Sa lmon Agricultural 0.014 0.013 0.015 Plotting S against A 4 4 0 provides valuable insight into the composi t ion of C D O M in surface water samples as it provides an indication of C D O M concentrat ion ( A 4 4 0 ) and the relative concentrat ion of high-molecular-weight humic substances vs. lower-weight fulvic materials (S). A s illustrated in Figure 6-8 when plotted against these two var iables, stations tend to separate into one of two groups. The first are those sites that show minimal variation in C D O M concentrat ion but noticeable variability in C D O M composi t ion as illustrated by a range of va lues for S over the three days of sampl ing. These sites are predominantly found in forested and mixed-use sub-catchments. The second grouping contains those stations with C D O M concentrat ions that varied over the three days of sampl ing, but have consistently low S values. For all three watersheds, this group was consistently compr ised of agricultural s i tes. 170 0.035 H 0.030-4 0.005 H Hatzic Watershed Elk Creek i 1 r 0 000 0.010 0.020 0.040 Absorbance @ 440 nm (a.u.) T 1 r 0.050 0.060 0.070 O Agricultural A Forested • Mixed + Urban 1 1 1 0.000 0.010 0.020 ~T~ 0.030 0.040 0 050 0.060 0.070 Absorbance @ 440 nm (a.u.) Salmon ~i 1 r 0.000 0.010 0.020 0.030 0.040 0.050 0.060 Absorbance @ 440 nm (a.u.) 0.070 Figure 6-8 - Absorbance at 440 nm vs. spectral slope for the Hatzic, Elk Creek and Salmon watersheds. Note clustering of samples by land use. Chapter 6 Forested and mixed-use sites in the Hatz ic and Elk Creek catchments showed the greatest variation in spectral s lope, with the highest S values observed during the August sampl ing period. Samp les col lected from these sites during or immediately after larger rainfall events in December and February showed much lower S values that were comparable to, or lower than, those observed in agricultural sites. For forested sites in the Hatzic watershed (the only watershed where cont inuous water-level data were col lected), S showed a strong negative correlation with stage (r s = -0.887, P < 0.001, n = 12). This trend was also observed for mixed-use sites (r s = -0.756, P = 0.001, n = 15) and agricultural sites (rs = -0.767, P < 0.001, n = 24); however, no correlation existed for urban site HV-20 . A s noted by Blough and Del Vecch io (2002) S is larger for fulvic ac ids and is inversely correlated with molecular weight. Lower S values therefore suggest a greater proportion of the C D O M load is compr ised of fulvic ac ids. A s fulvic ac ids are more soluble than humic acids, this would be expected in undisturbed subcatchments during dry weather. During winter months when increased rainfall results in greater al lochthonous C D O M inputs (Thurman, 1985), it is likely that lower S va lues reflect an increase in the proportion of C D O M compr ised of less-soluble humic acids (due to higher rainfall and decomposi t ion of autumn leaf litter). Agricultural sites showed minimal variation in S, with va lues ranging from 0.010 - 0.020. The low S va lues at these sites reflect the relatively greater proportion of higher molecular weight humic materials assoc ia ted with agricultural amendments , as noted by Ohno et al . (2006). T h e s e sites did, however, display a broader range of A 4 4 0 va lues, reflecting variations in C D O M concentrat ion. Max imum A 4 4 0 va lues were consistently observed in agricultural s loughs in the Elk Creek and Hatzic watersheds. Higher C D O M concentrat ions and lower spectral s lopes at these sites likely reflect the presence of higher-molecular-weight organic matter as a result of both greater autochthonous (in situ) production and greater contributions from the adjacent land surface during rainfall events (Ohno et a l . , 2006). Urban sites in both the Hatz ic and Elk Creek catchments d isplayed similar properties. 6.3.6. Storm event dynamics In order to a s s e s s the change in C D O M concentration and composit ion over the course of a storm event, samples were col lected every three hours over the 36-hour event in the Hatz ic watershed from March 18-21, 2005 (Figure 6-9). The storm was compr ised of three rainfall events with rainfall of 23.6 mm, 15.5 mm and 35.3 mm. Absorbance at 220 nm reflects N 0 3 " concentration and therefore 172 Chapter 6 showed a strong posit ive relationship with river stage. Va lues for A 2 5 4 , A 2 8 0 , A 3 0 0 , A 3 4 0 and A 4 4 0 showed a large increase during the second rainfall event, which had the lowest total rainfall of the three events. For both the first and second events these values peaked just prior to the water-level max imum. 173 = 2 2 E 6 -0.8 0.7 — \ 0.6 — \ o ra 0.5 w 14 h 12 \ 0.4 -0.3 -0.2 Stage *• A Spectral Slope • • Absorption @ 220 nm ^ • Absorption @ 440 nm • • Chloride (mg-L 1) • • o> CM i n co i— OO CO CO CM LO o CO CO CO CO CO CO in o CO CJ> CM <Ji CD CT) CI CO CO CO CO o p o uri co ^ i— T— CM LO LO LO p o p CO LO 10 0.5 o I 0.08 0.4 ® - 0.3 <S 0.2 0.06 0.04 @ 0.02 0.012 0.01 - 0.008 a o co 75 o CD - 0.006 M - 0.004 0.002 CM LO LO LO 00 LO T- O CO CD CM LO LO LO LO cn LO O) Gi <Ji CM CO CO CO CO CO Date/Time CO CO CO CO CO CO CO CM CM CM CM CO CO CO CO Figure 6-9 - Water level, 1-hour rainfall and spectral properties of samples collected at the lower hydrometric station in the Hatzic Slough during a storm event, March 18-21, 2004. Note increase in A 4 4 0 during second storm peak Chapter 6 Although the third event was the largest in terms of total rainfall and the subsequent response of stream stage, it did not generate an equal ly large response in absorbance va lues. Absorbance at 220 nm increased with stream stage during this event. A similar trend was observed for absorbance in other U V wavelengths ( A 2 5 4 , A 2 8 o and A 3 0 0 ) , with max imum va lues as high as , or higher than those observed during the second event. Absorbance values at longer wavelengths in the visible range ( A 4 4 0 , A t 6 5 and A 6 6 5 ) did not show the s a m e positive correlation with stream stage ( A 4 6 5 and A 6 6 5 not shown). These values peaked during the second event, and while they did respond to the third event, va lues did not approach those observed during the second event. T h e s e data suggest that the second event resulted in the transport of easi ly access ib le organic matter that was rich in humic material, possib ly reflecting mobil ization of near-stream stores in agricultural regions of the watershed. The lack of a response in absorbance at longer wavelengths during the third event indicates that eas i ly -access ib le C D O M stores had been depleted by this t ime. It a lso likely reflects an increase in the relative contribution of rainfall to streamflow during this time. This is supported by the fact that both conductivity and chloride values reached absolute min ima during this event as well . Over the three events, absorbance parameters correlated strongly with bacterial and nutrient concentrat ions. Absorbance at 280 nm showed the strongest correlation with fecal coliform concentrat ions (r s = 0.732, P < 0.001) and A 2 2 0 showed a similar relationship with N 0 3 " (rs = 0.832, P < 0.001). Whi le correlations with fecal col i forms were strong, maximum absorbance values did not correspond to peak bacterial concentrat ions during the third event, highlighting different mobil ization and transport mechan isms for C D O M and bacter ia. These trends in absorbance in the U V and visible ranges were reflected in the spectral s lope values during the storm. Min ima in S were consistently observed on the rising limb of each rainfall event, and the lowest va lues were observed during the second storm peak. Va lues reached a minimum just prior to the increase in stream stage during this event, and remained low for 9 hours while absorbance at longer wavelengths ( A 4 4 0 and A 6 6 5 ) reached max imum values. After the second event, S briefly reached pre-storm levels, and then fluctuated moderately during the third event but did not reach the minimum values observed in the first and second events. A s illustrated in Figure 6-10, each event was character ized by a slightly different organic-matter signature. The first event was short l ived, with only one peak in A 4 4 0 and a concurrent dec rease in S. The second event, captured by samples 16, 17 and 18, reflects the contributions of humic-rich organic matter, 175 Chapter 6 and was character ized by high C D O M concentrat ions and the lowest spectral s lopes observed in any of the watersheds over the three days of sampl ing. Chlor ide va lues were depressed during this t ime; however, to a lesser extent than observed during the third event. This suggests that streamflow during the second event, while diluted by rainfall, was compr ised of pre-event water. Data for the third event suggest an exhaust ion of stores of easi ly transported organic matter, a reduction in the relative content of humic ac ids and a greater proportion of streamflow compr ised of rainfall, as indicated by lower CI" concentrat ions (Figure 6-10). 0.01 0.02 0.03 0.04 0.05 0.06 A440 Figure 6-10 - Spectral slope vs. A440 for the storm event in the Hatzic watershed. Chloride concentrations are represented by bubble size. Numbers refer to the consecutively numbered samples which were collected every 3 hours, from 09:00 on March 18 to 09:00 on March 21, 2005. 6.3.7. Absorbance spectroscopy as a qualitative tool Severa l authors have demonstrated the value of absorbance spect ra in the qualitative assessmen t of water in marine and freshwater environments (Khorassani et a l . , 1998; Vail lant et a l . , 176 Chapter 6 2002; Langergraber et a l . , 2004). S u c h analysis includes assessmen t of the location and intensity of individual absorbance peaks and shoulders and of spectral s lope. In the present study, qualitative variations in spectra observed between land-use types and water sources (i.e., groundwater vs . surface water) supported rapid determination of: 1) relative N 0 3 " concentrat ions, 2) relative C D O M concentrat ions and 3) relative C D O M composi t ion. T h e s e three characterist ics provide an indication of water source (i.e., surface water vs . groundwater), land-use type (forested vs. agricultural) and land-use intensity. For example, in the Sa lmon watershed, where samp les were col lected from surface water and groundwater sites, water source could easi ly be dist inguished based on a visual compar ison of absorbance values between 245-800 nm, a range that e n c o m p a s s e s the wavelengths for absorbance by humic subs tances . Va lues for groundwater samples from S A - 1 G and S A - 3 G were at, or near, zero throughout this range, while absorbance for all surface-water si tes was consistently higher, particularly between 200-400 nm (Figure 6-2b). These data suggest minimal C D O M content in these deep groundwater samples (the depths of the wells at S A - 1 G and S A - 3 G are 48 m and 74 m, respectively). This is consistent with the general ly low concentrat ions of organic matter in deep groundwaters. Organic materials initially derived from the litter layer are rarely transported to significant depth as they are either consumed by heterotrophic microbes or adsorbed onto the sur face of soil particles during downward transport (Thurman, 1985). Relat ive N 0 3 " concentrat ions for both groundwater sources were a lso easi ly d iscerned through visual inspection in the 220-225 nm range (Figure 6-2b). A s illustrated in Figure 6-5 and Figure 6-6 for each of the three sampl ing dates, an inspection of spect ra consistently a l lowed visual differentiation by land-use type. Further, s i tes observed to be under the greatest degree of agricultural influence in both watersheds (as indicated by consistently higher average nutrient and bacterial concentrat ions throughout the present study), consistent ly d isplayed the highest absorbance values across the entire wavelength range. For the Elk Creek catchment, there was an obvious visual distinction between agricultural and mixed sites for samples col lected on December 20, 2004 and February 16, 2005. For samples col lected on August 31 , 2004, there was some overlap of spectra for mixed sites and agricultural sites (not including E C - 4 , E C - 5 and E C - 6 , which could be easi ly differentiated for each day of sampling). This was likely due to relatively dry condit ions resulting in minimal transport of agriculturally-derived C D O M to surface waters. This was not the c a s e for E C - 4 , E C - 5 and E C - 6 as each of these sites is located in a s low-moving agricultural s lough with limited riparian vegetation and located in c lose proximity to cattle operations. T h e s e condit ions favour in-stream 177 Chapter 6 biological productivity and the resulting spect ra likely reflect the combined influence of agriculturally-derived C D O M and autochthonous C D O M production. Similar trends were observed in the Hatzic watershed; however, the distinction between spect ra representing different land uses was not as clear. This reflects the general trend of lower-intensity agriculture in this watershed (see Chapter 5). A s observed in the Elk Creek catchment, the site with the highest nutrient and bacterial concentrat ions (HV-18) consistently d isp layed the highest absorbance values across the entire wavelength range. This is a lso a s low-moving agricultural s lough draining a region with active cattle operat ions bordering the st ream. It appears , therefore, that qualitative assessmen t of entire spect ra is a useful comparat ive tool to a s s e s s relative agricultural land-use intensity at different sites. Visual inspection of spect ra can also provide insight regarding C D O M composi t ion. A s descr ibed above (Section 6.3.2), spectra for samples col lected from forested and mixed sites in August showed a distinctive shoulder at - 2 8 0 nm. Thurman (1985) noted that humic ac ids, due to their darker colour, absorb in the visible range, while fulvic ac ids absorb at wavelengths near 280 nm (thus the value of spectral s lope as an indicator of humic vs . fulvic acid content). This shoulder appears to reflect the presence of lower molecular weight fulvic acids and a relative lack of humic materials which tend to obscure the peak at 280 nm at agricultural ly-dominated sites. This would be expected in relatively "clear," headwater s t reams in forested sites as fulvic ac ids are more easi ly d isso lved from surrounding soi ls (Tipping, 2002). This peak is not observed under wet, winter condit ions indicating either a f lushing of fulvic ac ids from these sites, or an increase in the relative abundance of terrestrially-derived humic acids transported during rainfall events (or, more likely, a combinat ion of the two). Spect ra with low absorbance values and a visible absorbance peak at 280 nm therefore appear to indicate dominance of C D O M by fulvic ac ids under dry-weather condit ions. Qualitative analysis is a lso useful to a s s e s s relative changes in water quality at one site over t ime. This is illustrated in Figure 6-11 and Figure 6-12 which show contour and surface plots of absorbance spect ra and log-transformed absorbance spect ra col lected during the storm event in the Hatzic watershed. The surface in Figure 6-11 clearly illustrates increasing absorpt ion in two regions over the course of the storm event. The first region, from 200-225 nm, reflects increasing N 0 3 " concentrat ions assoc ia ted with the peaks in stream stage. The second , in the visible region, shows the greatest response during the second event and reflects the peak C D O M contributions descr ibed above. The input 178 Chapter 6 of C D O M during the second event is more clearly illustrated in Figure 6-12 which clearly shows the dec rease in S at that t ime. In both c a s e s , it can be seen that samp les col lected prior to the first event provided a basel ine against which future samp les could be compared . Further, qualitative ana lyses of these sur faces provide a rapid means of assess ing relative C D O M concentrat ion and composi t ion throughout the event. 179 Figure 6-11 - Time series of absorbance spectra for storm event in Hatzic watershed. Samples were collected every 3 hours starting at 09:00 on March 18 (Y-axis lines align with discrete spectra for each sample and correlate with the data points on Figure 6). Timing of three peaks Figure 6-12 - In-transformed spectra for the storm event captured in the Hatzic watershed. Note the clear decrease in spectral slope associated with the second peak in river stage, suggesting a shift to increased high-molecular-weight compounds in transported DOM. Chapter 6 The above illustrates that qualitative analysis of absorbance spectra provides useful information regarding water source , land-use type and land-use intensity beyond that which c a n be obtained from nutrient or bacterial analys is alone. A s demonstrated during the storm event in the Hatzic watershed, the utility of this technique improves when basel ine data are avai lable. Whi le it is c lear that further sampl ing is required to fully elucidate the potential of this technique, data col lected from the three watersheds in this study suggest that it is a useful tool for rapid assessmen t of water source and water quality. 6.3.8. Absorbance spectroscopy as a quantitative tool A s illustrated above, absorpt ion in the U V - C range shows excel lent correlation with N 0 3 " concentrat ion. This relationship is improved when the second derivative is used to minimize the inf luence of other absorbing spec ies in solution. B a s e d on the observed relationship, second-der ivat ive absorpt ion spect roscopy offers an effective, rapid, s imple and accurate technique for the determination of N 0 3 " in surface waters. The linear relationship between N 0 3 " and second-der iva t ive absorbance has been observed to deteriorate at concentrat ions approaching 40 mg-L" 1 (Suzuki and Kuroda, 1987). Whi le this may represent a chal lenge when directly monitoring effluent from wastewater treatment plants, it is not likely to limit the application of this technique in the majority of agricultural watersheds. For samples col lected in August , analysis for d issolved organic carbon (DOC) was conducted to a s s e s s the correlation with absorbance parameters (Table 6-6). A significant posit ive correlation between D O C and absorbance at several wavelengths was observed, however the strongest relationship was found between A 4 4 0 and D O C (rs = 0.782, P = 0.001, n = 15), suggest ing that this wavelength is a suitable indicator for d isso lved organic material. The variability in the relationship is due to the fact that light-absorb ing D O M represents only a portion of total D O C . Interferences from other absorbing compounds may also impact this relationship. 182 Chapter 6 Table 6-6 - Mean DOC concentration by land use and watershed for samples collected on August 31,2005 Watershed Land use N Mean DOC Concentration (mq-L1) H a t z i c F o r e s t e d 1 7.87 M i x e d 1 6.47 A g r i c u l t u r a l 2 10.16 E l k C r e e k F o r e s t e d 1 9.60 M i x e d 2 8.43 A g r i c u l t u r a l 3 11.15 U r b a n 1 13.32 S a l m o n A g r i c u l t u r a l 4 8.74 Absorbance at other wavelengths correlates well with traditional indicators of contamination, including D O C , P 0 4 3 ~ , N H 4 + and fecal col i form; however, these correlat ions were not consistent from site-to-site or between watersheds. Table 6-7 illustrates the absorbance wavelengths with the strongest correlat ions to traditional indicators of contamination for all watersheds. Signif icant, positive correlations are found for each water-quality variable when all watersheds are cons idered concurrently, and A 2 8 o shows the strongest correlation with all indicators (other than N 0 3 " ) . A s descr ibed above (Sect ion 6.3.3), absorpt ion in this range is attributed to the TZ-K* electron transition in subs tances which are commonly precursors to, or components of, humic materials. Absorbance at this wavelength, particularly when normal ized to D O C concentrat ion, has been used extensively as an indicator of the aromaticity and molecular weight of humic substances (Chin et a l . , 1994; Westerhoff and Anning, 2000; C h e n et al . , 2003 ; Volk et a l . , 2005). This normal ized value is referred to as speci f ic UV absorbance or ( S U V A ) . A s C D O M re leased by agricultural amendments (plant residue, manure, etc.) has consistently higher molecular weight than soi l-derived C D O M , (Ohno et al . , 2006) S U V A at 280 nm ( S U V A 2 8 0 ) has the potential to serve as an effective indicator of agricultural influence. In the present study, A 2 8 o and S U V A 2 8 o were strongly correlated (rs = 0.964, P < 0.001, n = 15), suggest ing that A 2 8 0 is also a good indicator of aromatic C content and molecular weight in the absence of D O C measurements . 183 Chapter 6 Table 6-7 - Correlations between nutrient and bacterial indicators and absorbance at specific wavelengths. Watershed Variable Absorbance n Correlation (rs) Indicator All Ammon ium A280 1 146 0.322 ** Nitrate A 2 2 4 115 0.999** Orthophosphate A 2 8 0 144 0.572** F C (log) A 2 8 0 108 0.412* T C (log) A 2 8 0 117 0.214** Elk Creek Ammon ium A 2 8 q | 37 0.795** Nitrate A 2 2 4 37 0.994** Orthophosphate A 2 8 0 38 0.576** F C (log) A 2 80 ) 37 0.518** T C (log) A 2 2 4 37 -0.492** Hatzic Ammon ium A280 78 0.342** Nitrate A 2 2 4 50 0.945** Orthophosphate A 2 8 0 78 0.247* F C (log) A 2 8 0 48 0.610** T C (log) A 2 2 o 52 0.668** Sa lmon Ammon ium A 2 8 0 31 0.606** Nitrate A 2 2 4 1 28 0.969** Orthophosphate A 2 2 4 ^ 28 -0.525** F C (log) A 2 2 4 23 0.250 T C ( l o g ) A 2 8 0 28 -0.339 A224 1 - 2"° der ivat ive abso rbance at 224 nm *-P<0.05 ** - P < 0.01 The correlations between A 2 8 0 and nutrient and fecal coliform concentrat ions descr ibed above suggest that A 2 8 0 is in fact a good indicator for relative agricultural inf luence as it detects contributions of higher molecular weight C D O M . It is possib le that the inconsistencies in the relat ionships descr ibed in the above table would be at least partially addressed through the use of S U V A 2 8 o as it accounts for the bias introduced by higher D O C concentrat ions. Table 6-7 also illustrates that there is variability in the correlations between A 2 8 o and nutrient and bacterial concentrat ions ac ross watersheds, despite significant correlations within watersheds. A s illustrated graphical ly in Figure 6-13, when cons idered on a watershed sca le , the only consistent correlation is between 2 n d -der ivat ive absorbance at 224 nm and N 0 3 " . 184 Figure 6-13 - Absorbance values vs. a) nitrate, b) ammonium and c) fecal coliform concentrations for all three watersheds. Note the strong, consistent correlation between 2nd derivative absorbance at 224 nm compared to correlations for NH4+ and fecal coliform, which vary by watershed. Chapter 6 The var iance in the degree and cons is tency of correlations between absorbance va lues and N H 4 + , P 0 4 3 " and bacterial concentrat ions is likely due to the fact that absorbance can result from several compounds in solution, thus complicat ing the relationship between absorbance values and the parameters of interest (Stumwohrer et al . , 2003). The concentrat ion of interfering compounds is dependent upon site-specif ic var iables ( land-use type, soil type, local vegetat ion, etc.), and as a result, so is the impact they have on these relationships. This explains the considerable variability in correlations from different sites and watersheds. However, as observed by Brookman (1997) once a relationship between absorbance va lues and parameters of interest are obtained for a given site, this rapid technique serves as a useful indicator of significant changes in water quality. Indeed, Langergraber et a l . (2004) demonstrated an absorpt ion system for rapid detection of surface water contamination events based on this principle. The system required a "learning per iod" during which site-specif ic calibration or basel ine absorbance spectra could be col lected. O n c e basel ines were establ ished, alarm levels were then def ined based on the degree of variation from basel ine levels, and the probability of var iance as determined during the learning period. The author is currently unaware of any instances where such sys tems have been utilized to track agricultural influence on water quality. 6.4. Conclusions The objectives of this chapter were to: 1) determine if absorbance spect roscopy could serve as a useful technique for the quantitative and/or qualitative assessmen t of agricultural inf luence on water quality, and 2) determine if this technique could a lso provide insight regarding water source and f lowpaths. B a s e d on the results descr ibed above, the following conc lus ions can be drawn. 1) Absorbance spectroscopy is a rapid and accurate technique for determining N03~ concentrations in filtered, bulk water samples It was demonstrated above that second-der ivat ive spect roscopy could be used to accurately and rapidly determine N 0 3 " concentrat ions in bulk surface water and groundwater samples . This technique offers several advantages over the Qu ikchem methods used for N 0 3 " quantification in the present study. Firstly, absorbance techniques require minimal sample treatment (filtering) and no addit ional reagents. Second ly , ana lyses are rapid (< 1 minute) and require less than 5 ml of sample. Finally, absorbance 186 Chapter 6 spectra provide a great deal of information regarding water source and C D O M composi t ion which can be used for both qualitative and quantitative ana lyses of land-use influence. 2) Absorbance values, particularly A280, are an effective indicator of relative agricultural influence on water quality Absorbance at 280 nm and SUVA 2 8o provide an indication of molecular weight of C D O M in water samp les . These parameters therefore have significant potential as indicators of agricultural influence as C D O M derived from manure and plant res idues is general ly of higher molecular weight than soi l-derived C D O M . This is supported in the present study by observat ion of consistent, significant correlat ions between A 2 8 0 and nutrient and bacterial concentrat ions. O n c e basel ine data were avai lable, absorbance values provided a useful indicator of relative agricultural inf luence on water quality in each watershed. 3) Absorbance spectra from filtered, bulk water samples provide useful qualitative information regarding water-quality and water source Visual inspection of absorbance spect ra provided a rapid technique for the determination of water source (groundwater vs . headwater stream vs. agricultural slough), and relative organic matter concentration and composi t ion. This supported the assessmen t of spatial trends in C D O M and agricultural inf luence, as well as the detection of changes in water quality and contributions to streamflow over the course of a storm event in the Hatz ic watershed. 187 Chapter 7 7. Fluorescence spectroscopy as a tool to detect agricultural influence on water quality 7.1. Introduction A s descr ibed in Chapter 6, the absorbance propert ies of C D O M provide an indication of their concentrat ion, as well as a preliminary indication of the relative amounts of fulvic vs . humic acids, both of which are useful in detect ing agricultural influence. However, due to the overlap of absorption peaks arising from several chromophores in solution, it can be difficult to identify and/or quantify individual C D O M components, particularly at relatively low concentrat ions. F luorescence spect roscopy may be used to address this chal lenge as many components of C D O M are f luorophores (compounds that re-emit absorbed radiation) which can be used to further character ize the concentrat ion and composi t ion of the C D O M load of surface waters and groundwaters. A s descr ibed in Chapter 2, there are two primary groups of f luorophores in C D O M , referred to as "protein-like" and "humic-l ike". Protein-l ike f luorescence refers to f luorescence peaks observed when samples are excited in the ultraviolet range (excitat ion/emission pairs 220/305 nm and 220/350 nm). T h e s e peaks are similar to those observed for the aromatic amino ac ids tyrosine (220/305) and tryptophan (220/350), and are therefore referred to as tyrosine-l ike and tryptophan-like f luorescence (Lakowicz, 1999; Blough and Del Vecch io , 2002). Humic-l ike f luorescence is attributed to aromatic organic compounds assoc ia ted with humic and fulvic materials (Blough and Del Vecch io , 2002), and is therefore commonly referred to as humic-l ike and fulvic-like f luorescence. F luorescence spectroscopy offers many of the s a m e advantages over traditional water quality assessmen t techniques as absorbance spectroscopy. F luorescence techniques require smal l sample vo lumes (< 5 ml), are relatively rapid (1-20 minutes per sample , depending on the resolution of the scan , although longer scans are possible) and are non-destructive. Further, because individual f luorophores have unique excitation and emiss ion wavelengths, f luorescence spect roscopy al lows the detection of individual C D O M compounds in solution. Finally, f luorescence techniques are far more sensit ive than absorpt ion, thus al lowing detection of C D O M compounds at low concentrat ions (Lakowicz, 1999). A s outl ined in Chapter 2, several studies have utilized f luorescence spect roscopy to a s s e s s organic-matter dynamics in marine (e.g., Cob le , 1996; e.g., B o e h m e et al . , 2004), nearshore (e.g., C h e n 188 Chapter 7 and Gardner , 2004; Jaffe et al . , 2004), wastewater (e.g., Reyno lds and A h m a d , 1997; Westerhoff et al . , 2001) and freshwater (e.g., Baker , 2002c; Ka tsuyama and Ohte, 2002) environments. Of particular re levance to the present study, Baker (2002b) examined the f luorescence properties of several types of isolated farm wastes (pig and cattle slurries, s i lage liquor and sheep barn wastes) . However, this study represents the first a imed specif ical ly at assess ing the link between C D O M f luorescence and nutrient and bacterial concentrat ions in surface waters across multiple watersheds and land uses , in order to determine the potential of f luorescence spect roscopy as a tool to detect agricultural runoff. The purpose of this chapter is to conduct a preliminary assessmen t of the utility of f luorescence spect roscopy to further character ize agriculturally-derived organic matter in filtered, bulk water samp les . Specif ical ly, this chapter a ims to: 1) evaluate relationships between land use and the type and concentrat ion of f luorophores in surface waters, 2) a s s e s s the potential for f luorescence spect roscopy as a tool to detect agricultural effluent in surface waters and 3) a s s e s s the relationships between f luorescence parameters and nutrient and bacterial concentrat ions to evaluate the potential of f luorescence spect roscopy as a proxy for contaminants of concern. 7.2. Methods Fluorescence data were col lected following the laboratory methods outl ined in Chapter 3. Ana lyses were conducted on a subset of the s a m e samples for which absorbance data were col lected (N = 67), including those representing the storm event in the Hatzic watershed (see Chapter 6 for collection dates, and assoc ia ted meteorological condit ions). The duration of f luorescence scans ranges from less than one minute to several hours, depending on the desired signal to noise ratio (to reduce the noise in a scan , the averaging time at each excitat ion/emission pair is increased). A s the objective of this study was to conduct a preliminary assessmen t of the technology as a tool to detect potentially minute dif ferences in f luorescence signals, a scan t ime of approximately 20 minutes was chosen to ensure a reasonable signal to noise ratio while still permitting the analysis of numerous samp les per day. A s a result, ana lyses were conducted only on se lected stations in the Hatzic, Elk Creek and Sa lmon watersheds. Resul ts are presented either as f luorescence intensity va lues or as exci tat ion-emission matr ices ( E E M ' s ) . Each E E M is a collection of emiss ion scans col lected every 5 nm at success ive ly longer excitation wavelengths between 220-450 nm. The resulting plot represents an interpolated surface for 47 189 Chapter 7 separate emiss ion scans , with f luorescence measured every 2 nm (between 230-600 nm), and contains 8,742 data points. A s with absorbance, f luorescence intensity is given in arbitrary units as no calibration to a standard was conducted. Six f luorescence peaks were initially cons idered in this analysis (Table 7-1). Humic-l ike f luorescence has previously been measured at two excitation wavelengths. The first, in the U V (220-260 nm), and the second , in the visible range (340-360 nm), have been referred to previously as "Peak A " and "Peak C " , respectively (Coble, 1996). For this study, only the latter was used to reflect humic-l ike f luorescence, as less work has been done to character ize the former in freshwater environments (Baker and Spencer , 2004). Table 7-1 - Wavelengths for fluorescence peaks assessed in this study. Fluorescence Peak Excitation/Emission Pair Humic-l ike 370-390/460-480 Fulvic-l ike 320-340/410-430 Tryptophan 1 (T 2 2 0 ) 220/340-350 Tryptophan 2 (T 2 8 0 ) 275-280/340-350 Tyrosine 1 (Tyr 2 2 0 ) 220-225/300-310 Tyrosine 2 (Tyr 2 8 0 ) 270-280/300-310 Protein-l ike f luorescence (tryptophan-like and tyrosine-like) was also measured as a potential indicator of agricultural inf luence, as it has been demonstrated to be a good proxy for wastewater effluent (Petrenko et a l . , 1997; Baker , 2001) and l ivestock waste (Baker, 2002b). Both amino acids were initially measured at 220 nm ( T 2 2 0 and Ty r 2 2 0 ) and 280 nm ( T 2 8 0 and Ty r 2 8 0 ) excitation. A compar ison of f luorescence intensities of the two tryptophan peaks indicated that T 2 2 0 and T 2 8 0 were significantly positively correlated (rs = 0.860, P < 0.001, n = 42). The correlation between T y r 2 2 0 and T y r 2 8 0 was also significant, but not as strong (rs = 0.594, P < 0.001, n = 42). This is a result of interference caused by the R a m a n peak for water in the s a m e wavelength range as T y r 2 8 0 . B e c a u s e of this interference, only T y r 2 2 0 was cons idered in this study. 7.3. Results and discussion 7.3.1. EEM features Figure 7-1 shows a typical contour plot for an E E M from HV-18 , a site under significant agricultural influence in the Hatzic watershed. This plot illustrates f luorescence centres typically observed 190 Chapter 7 in samples col lected from agriculturally-influenced sites, as well as the excitat ion/emission ranges used for determining f luo