1

1

2

The Effects of Multiple Beneficial Management Practices on 3

Hydrology and Nutrient Losses in a Small Watershed in the 4

Canadian Prairies 5

6

Sheng Lia, Jane A. Elliottb*, Kevin H.D. Tiessenac, James Yarotskid, David A. Lobba, Don N. 7

Flatena 8

a Department of Soil Science, University of Manitoba, Winnipeg, MB, Canada 9

b National Hydrology Research Centre, Environment Canada, Saskatoon, SK, Canada 10

c International Development Research Centre, Ottawa, ON, Canada 11

d

13

Agri-Environment Services Branch, Agriculture and Agri-Food Canada, Regina, SK, Canada 12

14

ACKNOWLEDGEMENTS 15

Financial support for this study was provided by the Agriculture and Agri-Food Canada (AAFC) 16

through the Watershed Evaluation of Beneficial Management Practices (WEBs) project. The authors 17

would like to thank B. Turner, D. Cruikshank and D. Robinson for their assistance with sample collection 18

and land use survey, D. Gallen and S. Dumanski for compiling the runoff and water quality data, and R. 19

Rickwood for his help on GIS data processing and mapping. 20

21

22

* Corresponding author. Tel.: +1-306-975-5738; Fax: +1-306-975-5143; E-mail: Jane.Elliott@EC.GC.CA (J.A. Elliott)

2

The Effects of Multiple Beneficial Management Practices on 23

Hydrology and Nutrient Losses in a Small Watershed in the 24

Canadian Prairies 25

26

ABSTRACT 27

Most beneficial management practices (BMPs) recommended for reducing nutrient 28

losses from agricultural land have been established and tested in temperate and humid 29

regions. Previous studies on the effects of these BMPs in cold-climate regions, especially at 30

the small watershed scale, are rare. In this study, runoff and water quality were monitored 31

from 1999 to 2008 at the outlets of two sub-watersheds in the South Tobacco Creek 32

watershed in Manitoba, Canada. Five BMPs—a holding pond below a beef cattle 33

overwintering feedlot, riparian zone and grassed waterway management, grazing 34

restriction, perennial forage conversion, and nutrient management—were implemented in 35

one of these two sub-watersheds, beginning in 2005. 36

We determined that > 80% of the N and P in runoff at the outlets of the two sub-37

watersheds were lost in dissolved forms, ≈ 50% during snowmelt events and ≈ 33% during 38

rainfall events. When all snowmelt- and rainfall-induced runoff events were considered, the 39

five BMPs collectively decreased Total N (TN) and Total P (TP) exports in runoff at the 40

treatment sub-watershed outlet by 41% and 38%, respectively. The corresponding 41

reductions in flow-weighted mean concentrations (FWMCs) were 43% for TN and 32% for 42

TP. In most cases, similar reductions in exports and FWMCs were measured for both 43

dissolved and particulate forms of N and P, and during both rainfall and snowmelt-induced 44

runoff events. Indirect assessment suggests that retention of nutrients in the holding pond 45

could account for as much as 63% and 57%, respectively, of the BMP-induced reductions 46

in TN and TP exports at the treatment sub-watershed outlet. The nutrient management 47

BMP was estimated to have reduced N and P inputs on land by 36% and 59%, respectively, 48

in part due to the lower rates of nutrient application to fields converted from annual crop 49

to perennial forage. Overall, even though the proportional contributions of individual 50

3

BMPs were not directly measured in this study, the collective reduction of nutrient losses 51

from the five BMPs was substantial. 52

53

echnological advances over the past century have allowed agriculture to become more productive. 54

However, many modern farming practices, such as the use of synthetic fertilizers and pesticides 55

and the enclosure of livestock in pastures and barns, have contributed to some degree of 56

environmental degradation, including the decline of water quality (Coote and Gregorich, 2000). In the 57

Prairie province of Manitoba in Canada, deteriorating water quality in Lake Winnipeg—the tenth largest 58

freshwater lake in the world—has been partly attributed to excessive nutrient loading (Lake Winnipeg 59

Stewardship Board, 2006). In order to protect the health of Lake Winnipeg, actions must be taken to 60

reduce nutrient losses from difference sources, including those from agriculture. 61

For decades, beneficial management practices (BMPs) have been developed and promoted to 62

reduce nutrient losses from agricultural land (Schnepf and Cox, 2006). Many of these BMPs were 63

established in temperate and humid regions where nutrient losses from agricultural lands are mainly in 64

particulate form and associated with rainfall runoff and erosion. In cold-climate regions such as much of 65

the Canadian Prairies, annual precipitation falls predominantly in the form of rainfall. However, snowfall 66

typically accounts for 20% of the total annual precipitation and snow accumulates throughout the winter 67

months – melting when temperatures increase in the spring. In addition, snowmelt often occurs on frozen 68

soil with almost no infiltration, and, therefore, surface runoff is encouraged. Consequently, the magnitude 69

of snowmelt-induced runoff often exceeds rainfall-induced runoff, and snowmelt runoff events often last 70

longer than typical rainfall runoff events (Nicholaichuk, 1967; Granger and Gray, 1990; Chanasyk and 71

Woytowich, 1986; Granger et al., 1984; Little et al., 2007). During the snowmelt period, low infiltration 72

rates also result in prolonged saturated condition on the soil surface, encouraging the release of dissolved 73

nutrients (e.g., Bechmann et al., 2005; Ontkean et al., 2005; Little et al., 2006; Ulen et al., 2007). 74

Although snowmelt runoff is usually less erosive than rainfall runoff (due to the absence of raindrop 75

splash to initiate the transport of soil particles), snowmelt-induced soil losses have been reported to 76

exceed rainfall-induced soil losses in western Canada because of the large runoff volumes and low 77

infiltration rates in snowmelt events (e.g., Chanasyk and Woytowich, 1987; van Vliet and Hall, 1991; 78

McConkey et al., 1997). 79

Due to the different mechanisms between snowmelt and rainfall runoff, the effectiveness of a given 80

BMP in reducing nutrient losses in a cold-climate region may be different from that in a temperate-81

climate region. For example, Salvano et al. (2009) tested the performance of three P risk indicators (Birr 82

and Mulla’s P Index, a preliminary P risk indicator for Manitoba, and a preliminary version of Canada’s 83

National Indicator of Risk of Water Contamination by Phosphorus) using water monitoring data from 14 84

T

4

watersheds in southern Manitoba, Canada. It was determined that all three P risk indicators performed 85

poorly and need to be modified to suit the cold-climate, soils and landscapes in southern Manitoba. 86

Similarly, numerous studies have shown that adoption of conservation tillage is effective in reducing 87

particulate P, but may increase dissolved P in the surface runoff (e.g., Baker and Laflen, 1983; Sharpley et 88

al., 1994; Bundy et al., 2001; Daverede et al., 2003). In temperate-climate regions, particulate P typically 89

dominates the total P loss and the reduction in particulate P more than offsets any increases in dissolved P. 90

Therefore, the overall effect of conservation tillage is a decrease of total P loss. However, also in southern 91

Manitoba, Tiessen et al. (2010) reported that total P export from a field under conservation tillage was 92

greater than that from its paired conventionally tilled field, because P loss from these fields were 93

dominated by dissolved P in snowmelt runoff, which was greater under conservation tillage than under 94

conventional tillage system. Similar results have been reported by researchers in Scandinavia (Ulen et al., 95

2010). Clearly, the effects and anticipated benefits from agricultural BMPs developed in temperate-96

climate regions need to be re-evaluated for cold-climate regions. 97

Previous field studies on the effects of BMPs in cold-climate regions have primarily been 98

conducted at the plot or in-field scale and many used the changes in soil nutrient concentrations due to 99

BMP implementation to indicate its impact on nutrient losses from the field (e.g. Uusi-Kamppa, 2005; 100

Vaananen et al., 2006; Miller et al., 2010b). Although these studies are crucial for understanding 101

individual processes, it has been recognized that plot or in-field scale research often fails to capture the 102

complexities and interactions among BMPs, biophysical settings and land use within a watershed 103

(Schnepf and Cox, 2006; Hoffmann et al., 2009). As a result, it can be difficult to relate results at the plot 104

or in-field scale to overall improvements in receiving water quality for a region. Unfortunately, watershed 105

scale studies on the effects of BMPs on nutrient losses in cold-climate regions are still rare. Moreover, 106

previous watershed scale studies in cold-climate regions have only dealt with single BMPs (e.g., 107

Sheppard et al., 2006; Eastman et al., 2010; Miller et al., 2010a; Tiessen et al., 2010; 2011), whereas in 108

reality, multiple BMPs are often required and applied by producers to different fields in a watershed. We 109

are not aware of any studies carried out to quantify the effects of multiple BMPs at the watershed scale in 110

cold-climate regions. 111

One difficulty of watershed scale studies is due to the temporal variability of climate and, therefore, 112

hydrology. Over short study periods (e.g., two to three years), the effects of various BMPs are often 113

overwhelmed by the natural variability of water quality due to climate and hydrology. To overcome this 114

difficulty, a paired watershed design has been recommended, in which two paired watersheds are 115

examined simultaneously and BMP treatments are applied to only one watershed (Spooner et al., 1985; 116

Clausen et al., 1996). To test the BMP effect, these authors recommend using an analysis of covariance 117

(ANCOVA), in which the control watershed (without the implementation of BMPs) serves as a reference 118

5

to account for temporal variations in water quality due to climate and hydrology, allowing for the effect of 119

the BMP(s) on changes in water quality in the BMP treatment watershed to be quantified. Guidelines 120

established by the United States Environmental Protection Agency (USEPA) for applying ANCOVA 121

recommend the use of a single covariate, the water quality variable measured at the control watershed, to 122

predict the matched variable measured at the treatment watershed (USEPA, 1993, 1997a, 1997b). This 123

simple ANCOVA method has been successfully applied in many studies to examine the BMP effects on 124

watershed water quality (e.g., Meals, 2001; King et al., 2008; Tiessen et al., 2010). In their classic paper 125

about the statistical power of pairing, Loftis et al. (2001) identified the need to incorporate other 126

covariates into the ANCOVA (termed multivariate ANCOVA, herein). Shilling and Spooner (2006) 127

successfully used a multivariate ANCOVA to examine the effects of watershed scale land use change on 128

stream nitrate concentrations in the Walnut Creek watershed in Iowa, USA. Similarly, in New York State, 129

Bishop et al. (2005) used three hydrologic variables as covariates in their study to evaluate the effects of a 130

suite of BMPs on stream water phosphorus. The authors concluded that by incorporating additional 131

hydrologic variables, the statistical power of the ANCOVA can be significantly increased and the 132

minimum detectable treatment effect can be greatly reduced— therefore, the effect of BMP can be 133

detected in a relatively short period. 134

The objective of this study was to use simple and multivariate ANCOVAs to quantify both the 135

seasonal and overall effects of multiple BMPs recommended for use in Manitoba—a holding pond 136

downstream of a beef cattle overwintering feedlot, riparian zone and grassed waterway management, 137

grazing restriction, perennial forage conversion and nutrient management—to reduce nutrient exports and 138

concentrations in surface runoff to receiving waters at small watershed scale under cold-climate 139

conditions typical of the Canadian Prairies. 140

141

142

MATERIALS AND METHODS 143

Study Sites 144

To validate the performance of selected BMPs in a watershed setting, Agriculture and Agri-Food 145

Canada (AAFC) launched the Watershed Evaluation of Beneficial Management Practices (WEBs) project 146

in 2004 (AAFC, 2007a). A core component of the WEBs project was to quantify the biophysical impacts 147

of BMPs on environment factors such as nutrient losses to waterbodies. The effects of selected BMPs 148

were examined in seven watersheds across Canada. This paper reports on the results for one of these 149

seven watersheds, the South Tobacco Creek (STC) watershed, located near the town of Miami in 150

southwest Manitoba, Canada and within the drainage area of Lake Winnipeg (Fig. 1a). Background 151

6

information about the STC watershed has been given in detail by Tiessen et al. (2010, 2011). The 152

dominant soils in the region are Dark Grey Chernozems (Mollisols), mostly clay loam in texture, formed 153

on moderately to strongly calcareous glacial till that overlay shale bedrock (Soil Classification Working 154

Group, 1998). Dominant landscapes in this region are undulating to hummocky landscapes (AAFC, 155

2007b). The climate is classified as sub-humid continental with short cool summers and long cold winters. 156

The long term mean annual precipitation is ≈ 550 mm with 25 to 30 % occurring as snowfall. The mean 157

annual temperature is ≈ 3°C and the monthly average temperatures are below zero from November 158

through March (Environment Canada, 2011). Most of the land in the STC watershed is used for 159

agricultural production, including cereal crops, oilseeds, perennial forages and livestock. 160

Two small sub-watersheds, the Madill and Steppler sub-watersheds, within the larger STC 161

watershed were examined in this study. The two sub-watersheds are located approximately 3 km apart 162

and both are situated in the headwaters of the STC watershed (Fig. 1a). The Madill sub-watershed has a 163

drainage area of 207 ha (Fig. 1b). During the experimental period (1999 to 2008), no BMP was 164

implemented and land use (mainly annual crop land) has remained largely unchanged. The Steppler sub-165

watershed has a drainage area of 205 ha, most of which is operated by a single producer. The farm site is 166

located inside the sub-watershed and consists of farm buildings and a cattle feedlot for over-167

wintering/feeding (1.9 ha, Fig. 1c). The producer operates a mixed farm; growing cereal grains and 168

oilseeds and managing a beef cattle herd of approximately 100 cows. The farm is divided into several 169

fields separated by two small intermittent watercourses traversing the farm. Fields are seeded to either 170

cereal grains or oilseeds on a rotational basis. Five BMPs—a holding pond downstream of a beef cattle 171

overwintering feedlot, riparian zone and grassed waterway management, grazing restriction, forage 172

conversion and nutrient management—were initiated in the Steppler sub-watershed in 2005 (Table 1, Fig. 173

1c) to reduce the environmental impact of agricultural practices. These BMPs are recommended for use in 174

Manitoba by the local government or soil and water conservation groups (e.g., MSFAC, 2007; MAFRI, 175

2009). 176

Overall, the Madill and Steppler sub-watersheds were similar in size, climate, landscapes and land 177

uses. The main differences between these two sub-watersheds are that: i) most annual cropped fields in 178

the Madill sub-watershed were under conservation tillage, whereas those in Steppler sub-watershed were 179

mainly under conventional tillage (i.e., primary tillage involving the use of a heavy duty chisel plow); ii) 180

the riparian areas of the Madill sub-watershed were wide and mostly vegetated with trees, whereas those 181

of the Steppler sub-watershed were narrower and contained pasture lands without trees; and iii) the 182

streams in the Madill sub-watershed were incised in the landscape, whereas those in the Steppler sub-183

watershed were less incised and shallow. Despite these differences, however, the similarity between the 184

two watersheds and the ANCOVA method allows for the Madill sub-watershed to be used as a control to 185

7

test the effects of BMPs implemented at the Steppler sub-watershed (the treatment watershed). 186

187

Sample Collection and Laboratory Analysis 188

Monitoring stations were established at the outlets of the Madill and Steppler sub-watersheds in 189

1999. These stations—MSC for the Madill (control) sub-watershed and MST for the Steppler (treatment) 190

sub-watershed—correspond to the inflow water monitoring stations described by Tiessen et al. (2011) in 191

their examination of the effects of small reservoirs on retaining sediments and nutrients. Since the same 192

raw data were used, detailed descriptions about sample collection and laboratory analysis can be found in 193

Tiessen et al. (2011). Briefly, water levels in the two reservoirs were monitored on continuous basis using 194

electronic water-level recorders. The water levels, in conjunction with the hydraulic parameters of each 195

reservoir, were used to calculate the inflow and outflow hydrographs (hourly data). Water quality samples 196

were collected using an auto-sampler (Sigma 800SL) triggered using a float system. During low flow 197

events, additional samples were collected manually and used to augment the auto-sampler collected 198

samples. Water quality sampling was conducted between 1999 and 2008 (i.e., the study period). However, 199

no water quality samples were taken in 2003. Construction of the holding pond below the cattle 200

overwintering site and the widening of the riparian areas were carried out in 2005, so that data collected 201

in 2005 were excluded from the analysis. Overall, there were five years (1999 – 2002 and 2004) of data 202

for the period before the BMP implementation (the Pre-BMP period) and three years (2006 – 2008) of 203

data for the period after the BMP implementation (the Post-BMP period). 204

After sample collection, sample bottles were extracted from the auto-sampler, packed on ice and 205

sent to the Fisheries and Oceans Canada's Freshwater Institute Laboratory in Winnipeg, Manitoba for 206

analyses of total phosphorus (TP), total dissolved phosphorus (TDP), particulate phosphorus (PP), total 207

nitrogen (TN), total dissolved nitrogen (TDN), particulate nitrogen (PN), ammonia (NH3), nitrate and 208

nitrite (NOx), and particulate organic carbon (POC). Methods used for the laboratory analyses were 209

described in detail by Tiessen et al. (2011). 210

211

Data Preparation 212

Hydrographs of the two sub-watersheds were split into paired runoff events following a procedure 213

similar to that described by Tiessen et al. (2011). The time durations of each pair of events were kept the 214

same for the two watersheds. Each runoff event was determined as either a snowmelt event or a rainfall 215

event based on the climate data (temperature and precipitation) recorded at a nearby Environment Canada 216

weather station (Miami-Orchard, 49°22’ N, 98°17 W). Diurnal fluctuations of runoff during snowmelt 217

period were not considered separate runoff events. Similarly, a single rainfall runoff event could include 218

8

multiple peaks with lower flow rates between the peaks. Therefore, a typical runoff event in both spring 219

and summer would have a rising limb starting from a flow rate of zero, one to several peaks with large 220

flow rates and a falling limb ending at a flow rate of zero. Initially, a total of 71 events were defined for 221

the entire study period (8 years). However, four rainfall events (one in 2000 and three in 2004) were 222

unmatched (zero flow flux on one sub-watershed). These unmatched events were considered to be the 223

result of differences in precipitation between the two sites, not the effects of BMPs and, therefore, were 224

eliminated from the analyses. In addition, two rainfall events in 2007 were not sampled and were also 225

eliminated from the analyses. Consequently, there were a total of 65 paired events (19 snowmelt and 46 226

rainfall events) in the analyses. 227

Three hydrologic variables—flow volume ratio (VolR), average flow rate (AvgFR) and peak flow 228

rate (PkFR)—were calculated as suggested by Bishop et al. (2005). Flow volume ratio (dimensionless), 229

which characterizes the imbalances in hydrology for matched events between the two sub-watersheds, 230

was calculated as the flow volume on the treatment sub-watershed divided by that on the control sub-231

watershed for paired events. Unit flow volume (UFV, m3 ha-1), calculated as the flow volume divided by 232

the catchment area, was used for comparisons between the sub-watersheds. Average flow rate (L s-1), 233

which reflects the intensity of runoff event, was calculated as the event flow volume divided by the 234

duration of the event. Peak flow rate (L s-1), which indicates the magnitude of runoff event, was 235

determined as the maximum flow rate in the hydrographs for the duration of that event. Nutrient 236

concentrations at hourly intervals were estimated from actual sample concentrations through interpolation 237

between sampling times (e.g., Bishop et al., 2005; Tiessen et al., 2011). Nutrient exports (also termed 238

fluxes or loads, g ha-1) were calculated for each event as the sum of products of hourly nutrient 239

concentrations and UFV. Flow-weighted mean concentrations (FWMC, mg L-1) were calculated for each 240

event as the total nutrient exports divided by total flow volume. 241

242

Statistical Analysis 243

Both the hydrologic variables and the nutrient loss variables were highly skewed (skewness 244

normally greater than 1.5). The natural log transformation was used to normalize the data as commonly 245

applied in watershed studies (e.g., USEPA, 1993) and reduced the skewness of most of the data to 246

between -0.5 and 1. All statistical analyses described below were performed on the transformed data. 247

The effects of BMPs on the hydrologic variables and nutrient export and FWMC were initially 248

examined using two simple ANCOVA models as recommended by USEPA (1993, 1997a, 1997b): 249

iiii XPrdY ε+++= cba [1] 250

iiiii IntXPrdY εdcba ++++= [2] 251

9

where i is the event index; Yi and Xi are the parameters measured on the treatment and control watersheds, 252

respectively, for event i; Prdi is the dummy variable created to indicate the experimental period of event i 253

[0 for the Pre-BMP (calibration) period and 1 for the Post-BMP (treatment) period]; εi is the residual error; 254

and Inti is the interaction term, calculated as: 255

( )mXPrdInt iii −= [3] 256

where m is the average of Xi during the post-BMP period. Equations 1 and 2 are equivalent to the reduced 257

and full ANCOVA models used in many paired watershed studies (e.g., USEPA, 1993; Grabow et al., 258

1999; Tiessen et al., 2010). 259

The two simple ANCOVA models did not account for the effects of hydrologic variations on the 260

nutrient loss variables and, therefore, the effects of BMPs may have been overwhelmed by noise created 261

due to the hydrologic differences between the two sub-watersheds or the temporal variations of hydrology. 262

To reduce this noise, a procedure similar to that described by Bishop et al., (2005) was adopted for the 263

multivariate ANCOVA. A multivariate regression model was used to predict each nutrient loss variable 264

(nutrient export or FWMC) at the treatment sub-watershed (dependent variable, Y). The complete model 265

was: 266

iiiiiiii PkFRAvgFRVolRIntXPrdY εgfedcba +++++++= [4] 267

in which, the same nutrient loss variable measured at the control sub-watershed (X), the interaction 268

between the treatment (BMP implementation) and X (Int), and the three hydrologic variables (VolR, 269

AvgFR and PkFR) were used as covariates (independent variables). Not all covariates in Eq. 4 were 270

always necessary or beneficial for testing the effect of BMPs. To determine the effectiveness of the 271

hydrologic covariates, Equations 1 and 2 were used to test for effects of BMP implementation on the 272

hydrologic covariates. In addition, we examined correlations among the covariates and between the 273

covariates and nutrient loss variables. The multivariate ANCOVA was conducted as a multivariate 274

regression analysis using the Reg procedure in SAS. Data for the snowmelt and rainfall events were 275

analyzed, separately and in combination (all events). In each case, covariates (excluding Prd and X) were 276

omitted if: i) they were significantly affected by the BMPs; or ii) if they did not make a significant 277

contribution to the model by increasing the determination coefficient (R2) of the model (the Maxr option 278

in the Reg procedure in SAS); or iii) if their influences were not significant (t-test, P ≤ 0.10). The 279

remaining covariates, together with Prd, X and the intercept term (a), constituted the best multivariate 280

ANCOVA model that: i) predict the nutrient loss variables on the treatment sub-watershed (i.e. Yi) most 281

effectively; and ii) describe the effects of BMPs most accurately. 282

The best multivariate ANCOVA model for each nutrient loss variable was used to calculate a 283

BMP-induced average percent reduction (%Rd) of that variable, following the method described by 284

Grabow et al. (1999) and Bishop et al. (2005): 285

10

( ) ( )( ) 100Yexp

YexpYexp%Rd0

10 •−

= [5] 286

where 0Y and 1Y were predictions made with a Prd value of 0 and 1, respectively, using the best 287

multivariate ANCOVA model and subsequently averaged across the entire study period. The 90% 288

confidence intervals for the percent reductions (%Rd90U and %Rd90L for the upper and lower confidence 289

intervals, respectively) were determined using the same method as that used to calculate %Rd, with the 290

value of b being replaced by its 90% confidence intervals, which in turn was computed in SAS from the 291

standard error of b. For models whose interaction terms (Int) were found to be significant, 292

the %Rd, %Rd90U and %Rd90L values were each calculated from two “filled-in” event datasets, one for no-293

BMP and one for BMP implemented, to account for the effects of interactions between BMP and event 294

magnitude (Bishop et al. , 2005). Due to the high variability inherent in the data, a P = 0.10 was used as 295

the significance threshold for all statistical analyses (Hansen et al., 2000; Teissen et al., 2011). 296

297

Evaluation of the Holding Pond and Nutrient Management BMPs 298

The reductions in nutrient exports and FWMCs determined by the multivariate ANCOVA were an 299

overall estimation for all five BMPs. Due to the interactions between BMPs at different scales, assessing 300

the effects of individual BMPs on nutrient export at the watershed outlet was difficult. For example, the 301

riparian management and grazing restriction BMPs were implemented throughout much of the treatment 302

sub-watershed and their impacts could not be separated from those of the other BMPs (Fig. 1c). The 303

perennial forage conversion BMP is being evaluated on two pairs of fields in the sub-watershed but we 304

anticipate that several more years of data collection will be required before the assessment of this BMP is 305

complete. Therefore, in this study, we focused the evaluation on the holding pond and the nutrient 306

management BMPs. 307

The feedlot area did not receive runoff water from the upstream catchment area and the runoff from 308

the feedlot area was directed to the holding pond via ditches. The holding pond was designed to retain all 309

runoff water from the feedlot. The retained water was used to irrigate a nearby forage field outside the 310

treatment sub-watershed. To evaluate the contributions of the holding pond BMP towards the changes in 311

water quality at the outlet of the treatment sub-watershed, a monitoring station was established at the inlet 312

of the holding pond (MSH) in 2005 after the holding pond was built (Fig. 1c). Runoff entering the 313

holding pond was monitored and water quality samples were collected and analyzed using methods 314

similar to those used for the MSC and MST water samples. It must be recognized that before the holding 315

pond was built, not all runoff and nutrient exports from the feedlot area was able to reach MST. From the 316

feedlot area to MST, the stream length was 2.3 km (Fig. 1c). During transport, runoff and nutrients may 317

11

be retained or lost through various processes (e.g., sedimentation, diffusion and leaching) in the riparian 318

zone and in the stream. In addition, nutrients may be transformed to different forms (e.g., from particulate 319

to dissolved or gaseous forms and from NH3 to NOx, or vice versa). Therefore, the effects of the holding 320

pond can only be assessed indirectly. The reductions in runoff and nutrient exports at MST due to the 321

holding pond (%Rdh) were approximated by: 322

100kEE

kE%Rdht

hh •

+= [6] 323

where Et and Eh are the total runoff (m3) or nutrient export (g) measured at MST and MSH, respectively; 324

and k is the fraction of runoff or nutrient export from the feedlot (MSH) that is delivered directly to MST, 325

whereas 1-k represents the fraction that is lost, in the absence of the holding pond. The value of factor k 326

ranges from 0 to 1 and may be different for flow volume and nutrient exports or vary temporarily. The 327

exact values of factor k for different variables were unknown. However, the feedlot was adjacent to a 328

stream and net losses in flow volume and nutrients in streams were usually small (Fig. 1c). Therefore, the 329

values of factor k were likely close to 1 for most variables examined in this study. When k = 1, the %Rdh 330

reaches its maximum value (denoted as %Rdhmax), which is also a measure of potential contribution of the 331

feedlot towards the runoff or nutrient exports at MST. The contribution of holding pond towards the 332

reductions of nutrient exports at MST (%Ctbh) was approximated by: 333

100%Rd%Rd%Ctb h

h •= [7] 334

where %Rdh is the reduction of nutrient export due to the holding pond and %Rd is the reduction of 335

nutrient export observed at MST (Eq. 5). Flow-weighted mean concentrations at MSH (FWMCh) and 336

MST (FWMCt) were used to calculate a ratio (RFWMC) to characterize the differences in FWMCs between 337

the two water monitoring stations: 338

t

hFWMC FWMC

FWMCR = [8] 339

The nutrient management BMP was extensively applied in the treatment sub-watershed. To assess 340

the effects of the nutrient management BMP, a simplified budget analysis was carried out to estimate the 341

nutrient balances on cropped fields in the two sub-watersheds: 342

B = If,m – Rh [9] 343

where B is the N or P balance (kg ha-1 yr-1); If,m is the N or P inputs from fertilizer and manure 344

applications (kg ha-1 yr-1) and Rh is the N or P removal due to crop harvesting (kg ha-1 yr-1). Fertilizer and 345

manure applications and crop yields were recorded for all fields in both sub-watersheds throughout the 346

entire study period. The amount of fertilizer applied was converted to net fertilizer-N and -P inputs based 347

on the fertilizers’ N and P contents, respectively, taking into account the time and method of application 348

12

(MSFAC, 2007). The amount of manure applied was also converted to net N and P inputs based on 349

generalized manure-N and -P concentrations in Manitoba, respectively, taking into account the manure 350

type and time and method of application (MAFRI, 2009). Similarly, crop yield was converted to N and P 351

removals due to crop harvesting based on generalized values for typical N and P concentrations in 352

Manitoba crops (MSFAC, 2007). N and P balances were calculated as net N and P inputs subtracted by N 353

and P removals, respectively. The N and P inputs, removals and balances were determined on an annual 354

basis for each field in the two sub-watersheds, which was categorized as either an annual-cropped field or 355

a perennial forage field. The data were then averaged within each sub-watershed for given crop categories 356

and also for all cropped fields. The annual N and P input, removal and balance data were averaged for the 357

Pre-BMP and Post-BMP periods and a two-way t-test was used to test the significance level of the 358

difference between the two periods. 359

Reductions in nutrient balances in cropped fields result in nutrient reductions on land, either in soil 360

profile or on soil surface. However, nutrient reductions on land are not the same as nutrient reductions at 361

the watershed outlet. Therefore, the nutrient budget analysis can only be used as an indirect assessment of 362

the management BMP on the reduction of nutrient loss at the sub-watershed outlet. It also should be noted 363

that there could be substantial errors associated with the calculated values in this simplified nutrient 364

budget given that N and P concentrations in manure and crop were not directly measured and some 365

sources and processes of N and P inputs and removals (e.g., N inputs from precipitation and biological N 366

fixation in perennial legume forage crops) were not taken into account. However, most such errors were 367

systemic biases for both monitoring periods and should have little effect on the differences between Pre-368

BMP and Post-BMP periods. Therefore, even though the absolute values for the nutrient budget within 369

each period may not be exact, our estimates of the differences (or changes) between periods, should be 370

valid. 371

372

373

RESULTS AND DISCUSSION 374

Hydrologic Variables 375

In the Steppler (treatment) sub-watershed and the Madill (control) sub-watershed, the three 376

hydrologic variables shared similar patterns in the comparisons between snowmelt events and rainfall 377

events, and between Pre-BMP period and post-BMP period (Table 2 and Fig. 2a). For example, on an 378

event basis, except for the median UFV values in the control sub-watershed, the median values of the 379

three hydrologic variables for the snowmelt events were all greater than the respective ones for the 380

rainfall events (Table 2). However, there were more rainfall events than snowmelt events so that, in some 381

13

cases, the total or annual average UFVs for rainfall events were greater than those for snowmelt events 382

(Fig. 2a). Also, except for the UFV for rainfall events at the control sub-watershed, the three hydrologic 383

variables all decreased from the Pre-BMP to the Post-BMP period. These general similarities between the 384

two sub-watersheds, especially those for the Pre-BMP period, strengthened the foundation for this paired 385

watershed study. However, there were also some noticeable differences between the two sub-watersheds. 386

For example, the median values of the event UFV and AvgFR at the control sub-watershed were 387

consistently greater than the those at the treatment sub-watershed (Table 2). The annual total UFV was 388

greater at the treatment sub-watershed in some years but was lower in some other years, compared to that 389

at the control sub-watershed (Fig. 2a). Also, at the treatment sub-watershed, the total UFV for snowmelt 390

events in the entire study period was greater than that for rainfall events, whereas at the control sub-391

watershed, opposite trends were observed. These differences in hydrology reflect the differences in land 392

use and topography between the two sub-watersheds and suggest that using hydrologic variables as 393

additional covariates could be beneficial in examining the effects of BMPs. 394

The effects of hydrology on nutrient exports were evidenced in that the annual total nutrient exports 395

followed a similar pattern as that of the annual total UFV (Fig. 2). In fact, for the treatment sub-watershed, 396

all correlation coefficients between nutrient exports and hydrologic variables were significant at P ≤ 0.10 397

(Table 3). The correlations between nutrient exports and the three hydrologic variables were greatest for 398

AvgFR, followed by PkFR and VolR. For nutrient flow-weighted mean concentration (FWMC), the 399

effects of the three hydrologic variables were different for snowmelt and rainfall events. For snowmelt 400

events, most nutrient FWMCs (except for PP-, PN- and POC-FWMC) were not significantly (P > 0.10) 401

correlated with the hydrologic variables, probably due to the small sample size. For rainfall events, most 402

nutrient FWMCs (except for PP-, PN- and POC-FWMC) were significantly (P ≤ 0.10) correlated with 403

AvgFR and PkFR but not with VolR. The exceptions of PP-, PN- and POC-FWMC from these general 404

trends suggest the strong impact of the three hydrologic variables on the FWMCs of nutrients in 405

particulate forms (Table 3). The lack of correlation between VolR and nutrient FWMCs can partially be 406

explained because event flow volume has been used in computing the nutrient FWMCs. In addition, the 407

effects of the three hydrologic variables may overlap as they were significantly correlated with each other 408

(P ≤ 0.10, Table 3). It is also possible that some other hydrologic variables (such as the timing of flow 409

events) have played an important role in the nutrient loss process. 410

Based on the two simple ANCOVA models (i.e., Eqs. 1 and 2), during rainfall events, coefficients 411

for X for all three hydrologic variables were significant at P ≤ 0.01, indicating that the variation of these 412

hydrologic variables observed at the treatment sub-watershed can be well explained by those observed at 413

the control sub-watershed (Table 4). However, this was not the case for snowmelt events. This 414

hydrological disconnection between the two watersheds for snowmelt events is likely the result of large 415

14

variability of snowmelt-induced hydrologic conditions. The coefficients for Prd were non-significant (P > 416

0.10) for VolR and AvgFR, in either rainfall or snowmelt events, suggesting that BMP implementation 417

had no significant effects on VolR and AvgFR (Table 4). The decreases of VolR and AvgFR that were 418

observed at the treatment sub-watershed were likely due to the temporal variation of precipitation since 419

similar decreases in VolR and AvgFR occurred at the control sub-watershed, where no BMP has been 420

implemented (Table 2). In contrast, there was a significant decrease of PkFR (P ≤ 0.01) due to BMP 421

implementation for rainfall events (Table 4). When both rainfall and snowmelt were pooled together (all 422

events) for the simple ANCOVA analyses, the results were similar to those of the rainfall events, i.e., 423

BMP implementation had significant (P ≤ 0.10) impact on PkFR but not on VolR or AvgFR. In addition, 424

multivariate analyses with the complete model (i.e., Eq. 4) showed that for snowmelt events, the effects of 425

PkFR were non-significant (P > 0.10) for all nutrient exports and FWMCs (data not shown). 426

Consequently, PkFR was dropped out from the best multivariate ANCOVA model used to examine the 427

effects of BMPs. 428

429

Nutrient Exports 430

For both snowmelt and rainfall events, phosphorus and nitrogen exports in dissolved form were 431

much greater than in particulate form (Table 2). For the entire study period, TDP and TDN exports were 432

5.0 and 7.9 times of the PP and PN exports, respectively, on the treatment sub-watershed (Fig. 2). On the 433

control sub-watershed, the ratios were 4.7 and 7.4 for TDP / PP and TDN / PN, respectively. Although 434

there were more rainfall events than snowmelt events, per event nutrient exports for snowmelt events 435

were much greater than for rainfall events. As a result, on both sub-watersheds during the entire study 436

period, the total exports of TDP and TDN in snowmelt events were greater than those in rainfall events 437

and the total exports of PP and PN in snowmelt events were similar to those in rainfall events (Fig. 2). 438

This highlights the importance of nutrient loss in dissolved form in the study area, especially during 439

snowmelt events, a characteristic that has been identified previously in cold-climate regions (e.g., Tiessen 440

et al. 2010; Ulen et al., 2010). Except for those for the rainfall events at the control sub-watershed, 441

median event nutrient exports all decreased after BMP implementation (Table 2). The fact that these 442

decreases in nutrient exports occurred at the control sub-watershed, where no BMP has been implemented, 443

suggests that there were differences in precipitation and hydrology between the Pre-BMP and Post-BMP 444

periods. However, there was also a general trend for the magnitude of decreases in both the snowmelt and 445

rainfall events to be greater at the treatment sub-watershed than at the control sub-watershed, indicating a 446

reduction of nutrient exports due to BMP implementation on the treatment sub-watershed (Table 2). 447

For the full ANCOVA models used to analyze results of nutrient exports (Eq.2), the interaction 448

term (Int) in these models was non-significant (P > 0.10) in all cases (data not shown) and there were 449

15

very few differences between the reduced model (Eq. 1) and the full model (Eq. 2). Therefore, the 450

following discussion will be focused on the reduced model (Table 5). For snowmelt events, the reduced 451

models for TP, TDP, TN and TDN and the coefficients of X in the models for TP, TDP, and NOx were 452

non-significant (P > 0.10), likely due to the large variability of nutrient exports and the small sample size 453

for snowmelt events. In all other cases (i.e., other nutrient export variables for snowmelt events and all 454

nutrient export variables for rainfall events and for all events), the reduced model and the coefficients of 455

X were significant at P ≤ 0.10 (Table 5). This indicates that nutrient exports observed at the treatment 456

sub-watershed can be explained by the model, in particular, by the respective nutrient exports observed at 457

the control sub-watershed. However, Prd was non-significant in all cases, indicating that the reduced 458

model (as well as the full model) was not able to detect the effects of BMPs on nutrient exports. This can 459

be clearly seen when the TP and TN exports for all events at the treatment sub-watershed is plotted 460

against that at the control sub-watershed (Fig. 3a and 3b). For both TP and TN exports, the paired 461

regression lines for the simple ANCOVA analysis were not much different, which suggests that BMP 462

implementation at the treatment sub-watershed had little impact on TP and TN exports. However, since 463

both the Pre-BMP and Post-BMP periods were short in this study, it is possible that the BMP effects were 464

masked by the variations in hydrology from event to event. The low R2 values of the regression lines also 465

indicated the large variation of the data. 466

Compared to their respective simple ANCOVA models (Table 5), the best multivariate ANCOVA 467

models (Table 6) had much greater R2 values (most at least doubled), indicating that nutrient exports at 468

the treatment sub-watershed are better explained after incorporating additional hydrologic variables. The 469

presence of the covariates in the best multivariate ANCOVA model reflected the importance of the 470

covariates. VolR was in each of the best multivariate ANCOVA models for nutrient exports and was 471

significant at P ≤ 0.10 (Table 6). In fact, except for the models for particulate nutrient exports (PP, PN 472

and POC) and NH3 exports in rainfall events, where AvgFR had a strong impact, VolR was significant at 473

P ≤ 0.001. This indicates the strong impact of VolR on nutrient exports. In contrast, Int was not in any of 474

the best multivariate ANCOVA models for nutrient exports, indicating that the effects of the interactions 475

between Prd and X on nutrient exports were negligible, which agreed with the results of the simple 476

ANCOVA models discussed earlier. Interestingly, although AvgFR correlated with nutrient exports better 477

than VolR did (Table 3), AvgFR was not in the best multivariate ANCOVA models for most nutrient 478

exports except for those for PP, PN, POC and NH3 exports in rainfall events (Table 6). Most importantly, 479

the Prd was a significant factor (P ≤ 0.10) in two thirds of the best multivariate ANCOVA models and 480

where it was non-significant (P > 0.10), the coefficient for Prd was negative in value, indicating a 481

consistent trend of reduction due to BMP implementation. The differences in nutrient export reductions 482

between snowmelt and rainfall events and between dissolved and particulate forms were not obvious, 483

16

mainly because only a few paired comparisons were valid (i.e., Prd was significant in the models of the 484

same nutrient export for both the rainfall and snowmelt events). For all runoff events combined, the 485

average reductions of nutrient exports due to BMP implementation ranged from 38% to 45% (Table 6). In 486

particular, on average, TP and TN exports have reduced by 38% and 41%, respectively. At the 90% 487

confidence level, TP export reduction was within the range of 20% to 51% and TN export reduction was 488

within the range of 20% to 57%. 489

490

Nutrient Flow-Weighted Mean Concentrations 491

The overall trends of the median values for event nutrient FWMCs were similar to those for event 492

nutrient exports (Table 2). The FWMCs of dissolved forms of phosphorus and nitrogen (i.e., TDP and 493

TDN) were much greater than those of particulate forms (i.e., PP and PN) in both snowmelt and rainfall 494

events at both sub-watersheds. Also, with only two exceptions, the median values for event nutrient 495

FWMCs in snowmelt events were greater than their respective values in rainfall events. Nutrient FWMCs 496

all decreased at the treatment sub-watershed after BMP implementation (Table 2). However, reductions of 497

FWMCs were also observed on the control sub-watershed for many nutrient variables and, therefore, 498

FWMC reductions at the treatment sub-watershed cannot be attributed to the effects of BMPs without 499

further evidence (i.e., ANCOVA). 500

With only a few exceptions, the results of the two simple ANCOVA models (results for the full 501

model were not shown and results for the simple model were shown in Table 5) for nutrient FWMCs 502

were similar to those for nutrient exports in that: i) the models were all significant at P ≤ 0.10; ii) the 503

interaction term (Int) was mostly non-significant in the full ANCOVA models; and iii) X was mostly 504

significant (Table 5). These results indicate that nutrient FWMCs at the treatment sub-watershed (Y) can 505

be explained by nutrient FWMCs at the control sub-watershed (X) in most cases and the interactions 506

between Prd and X were mostly non-significant (P > 0.10). However, in most cases, the influence of Prd 507

was significant (P ≤ 0.10), especially with the reduced model (Table 5). For rainfall and snowmelt events, 508

two thirds of the best multivariate ANCOVA models for nutrient FWMCs were, in fact, the reduced 509

ANCOVA model, which did not have any hydrologic covariate (Table 6). In contrast, those best 510

multivariate ANCOVA models for nutrient exports all had at least one hydrologic covariate. For all 511

events, most of the best multivariate ANCOVA models for nutrient FWMCs had one hydrologic covariate 512

(AvgFR or VolR). However, the increases in R2 values from the simple ANCOVA models to the best 513

multivariate ANCOVA models were small, compared to those for the nutrient exports. These further 514

proved that hydrologic covariates were less critical to nutrient FWMCs than to nutrient exports. 515

Nevertheless, for those nutrient FWMCs which did have a hydrologic covariate in the best multivariate 516

models, R2 values did improve from the simple ANCOVA models to the multivariate ANCOVA models, 517

17

and more importantly, the impact of Prd (P value for its coefficient in the t-test) generally increased. This 518

suggests that it was still beneficial to incorporate these hydrologic variables in these models. 519

The percent reduction values calculated from the best multivariate ANCOVA models for nutrient 520

FWMCs agreed well with those for the nutrient exports, especially for cases where these reductions for a 521

specific nutrient form were both significant (P ≤ 0.10) for nutrient export and FWMC (Table 6). This 522

consistency between the reductions of nutrient FWMCs and exports was expected because flow volume 523

was not significantly (P > 0.10) affected by BMP implementation (Table 4). The percent reduction values 524

for snowmelt and rainfall events were not much different from each other for the same nutrient FWMCs. 525

Neither were there many differences between the percent reduction values of dissolved and particulate 526

forms of nutrients. Overall, for all runoff events, the average reduction of nutrient FWMC due to 527

implementation of the five BMPs ranged from 32% to 55% (Table 6). In particular, on average, TP-528

FWMC and TN-FWMC have been reduced by 32% and 43%, respectively. At the 90% confidence level, 529

TP-FWMC reduction was within the range of 17% to 44% and TN-FWMC reduction was within the 530

range of 28% to 55%. 531

532

Contributions of the Holding Pond 533

Data collected at the inlet of the holding pond (MSH) showed that runoff from the feedlot site 534

overall could contribute a maximum of 4% towards the runoff at the sub-watershed outlet (Table 7); 535

whereas the area of the feedlot site was only 1% of the total area of the sub-watershed (Table 1). This 536

greater runoff contribution from the feedlot relative to the drainage area was likely due to the low 537

infiltration rate and limited water storage on the feedlot site, an effect reported in many previous studies 538

(e.g., Miller et al., 2004). Interestingly, when analyzed separately, the potential contributions of the 539

feedlot runoff were much greater in rainfall events than in snowmelt events (Table 7). This could be 540

explained because during rainfall events, other areas in the sub-watershed had much greater infiltration 541

rates than the feedlot. Alternatively, during snowmelt events, soil was mostly frozen and, therefore, 542

differences in infiltration rates between the feedlot and the rest of the sub-watershed were small. 543

Nevertheless, the reduction in runoff volume due to the holding pond was small in all cases, which agreed 544

with the non-significant (P > 0.10) effect of BMPs on UFV in the simple ANCOVA (Table 4). 545

Flow-weighted mean concentrations of the nutrients measured at MSH were much greater than 546

those measured at MST (ratios all greater or equal to 4, Table 7). As a result, although the runoff volume 547

from the feedlot was only a small portion of that at MST, the calculated maximum reductions due to the 548

holding pond were substantial. The maximum reductions for rainfall events were much greater than the 549

respective reductions for snowmelt events. The greater reductions for rainfall events was likely due to the 550

greater potential contribution of runoff from the feedlot in rainfall events than in snowmelt events (12% 551

18

versus 3%), since the ratios of FWMCs between MSH and MST within rainfall events and within 552

snowmelt events were not that much different (Table 7). For rainfall events, the maximum reductions due 553

to the holding pond (Table 7) exceeded the average reductions at MST determined by the multivariate 554

ANCOVAs (Table 6), indicating that some of the nutrients leaving the feedlot have been lost before they 555

reached MST. Such losses of nutrients likely have also occurred in snowmelt events, but could have been 556

limited by cool temperatures and frozen stream banks. In addition, there could be transformations of 557

nutrients, as evidenced in that there were lack of responses to the large %Rdhmax values of NH3 and 558

particulate form nutrients (PP, PN and POC) in the multivariate ANCOVAs of MST data (Table 6 and 559

Table 7). 560

For all events, assuming that all TP and TN leaving the feedlot reached MST before the holding 561

pond was built (i.e., k = 1.0 in Eq. 6), the reductions of TP and TN due to the holding pond were 24% and 562

24%, respectively, which were 63% and 57% of the average reductions in TP and TN exports, 563

respectively, at MST determined by the multivariate ANCOVAs (Table 7). In other words, the holding 564

pond may have contributed a maximum of 63% and 57%, respectively, of the TP and TN export 565

reductions observed at the treatment sub-watershed outlet. Even if we assume that half of the TP and TN 566

leaving the feedlot did not reach MST before the holding pond was built (i.e., k = 0.5 in Eq. 6), the 567

holding pond still contributed 36% and 33%, respectively, of the TP and TN export reductions observed 568

at MST. Considering the small area of the feedlot (1% of the sub-watershed, Table 1), the contributions of 569

the holding pond towards the reductions of nutrient exports at MST were very large. However, the 570

holding pond alone cannot explain all the reductions of nutrient exports at MST, and at least 37% of the 571

TP and 43% of the TN export reductions observed at MST must be due to the other BMPs. 572

573

Effects of the Nutrient Management 574

The nutrient budget analysis showed that when all cropped fields were taken into account, there 575

were no significant differences (P > 0.10) between the Pre-BMP and Post-BMP periods in N and P inputs 576

at the control sub-watershed but at the treatment sub-watershed, N and P inputs significantly (P ≤ 0.10) 577

decreased by 36% and 59% (i.e., 26 kg ha-1 yr-1 and 5 kg ha-1 yr-1), respectively, from the Pre-BMP to 578

Post-BMP period (Fig. 4a, 4b). Part of these decreases in N and P inputs were due to the perennial forage 579

conversion since there was very limited application of synthetic fertilizer and no manure applied to the 580

perennial forage conversion fields (data not shown). However, when only continuously annual cropped 581

fields were taken into account, similar patterns of N and P inputs at the two sub-watersheds were 582

observed—between Pre-BMP and Post-BMP periods, there were no significant differences (P > 0.10) at 583

the control sub-watershed but there was a significant decrease (P ≤ 0.10) of P input at the treatment sub-584

watershed (Fig. 4c, 4d). In spite of the reductions in N and P application rates, crop yields remained 585

19

similar and, therefore, differences in nutrient removals between the Pre-BMP and Post-BMP periods were 586

non-significant (P > 0.10) in all cases (Fig. 4e, 4f, 4g, 4h). 587

Both nutrient inputs and removals contributed to the status of nutrient balances. At both sub-588

watersheds, there was a surplus of N but a deficit of P in the Pre-BMP period (Fig. 4i, 4j, 4k, 4l). This 589

pattern was likely due to the low application rate of P fertilizer relative to that of N fertilizer (data not 590

shown). BMP implementation at the treatment sub-watershed had significantly (P ≤ 0.10) reduced the N 591

and P inputs, resulting in deficits in the N and P balances, whereas at the control sub-watershed, the 592

changes in N and P balances were non-significant (P > 0.10) (Fig. 4i, 4j, 4k, 4l). The nutrient budget 593

deficits at the treatment sub-watershed have not affected the yields significantly, even on annual cropped 594

fields, largely due to crop uptake of nutrients from substantial reserves of soil nutrients that were 595

measured in soil tests (data not presented). Therefore, the overall reductions in the nutrient balances were 596

driven by the decreases of N and P inputs. In particular, there was a significant reduction in P balance at 597

the treatment sub-watershed, which was due to the significant decrease of P input on annual cropped land. 598

The results from the annual cropland at the treatment sub-watershed seem to suggest that on fields with 599

fertile soil, nutrient management can be used as a BMP to reduce the nutrient inputs to the environment 600

without negatively affecting the yield, at least in a short term. It is clear that the nutrient management 601

BMP has reduced nutrient inputs to the treatment sub-watershed in this study and this reduction in 602

nutrient inputs likely has contributed to the reductions in nutrient losses at the sub-watershed outlet, but at 603

this time we cannot quantify the contributions of the decreased nutrient inputs to the observed nutrient 604

loss reductions at either the field or watershed scale. Nutrient source tracking that is currently underway 605

in the STC watershed should provide the data required to link the reduction in inputs to that at the sub-606

watershed outlet. 607

608

Implications 609

The five BMPs examined in this study collectively reduced nutrient exports and FWMCs regardless 610

of event type and nutrient form. However, in practice, it may be difficult to convince producers to adopt 611

the full suite of five BMPs all at once. A recommendation of one or two of the more effective BMPs may 612

be better received by producers. The major source of nutrient loss, the dissolved nutrients in snowmelt 613

events, has not been well targeted in these BMPs. However, among the five BMPs, some are likely to be 614

more effective in reducing nutrient losses in the particulate form (e.g., forage conversion) whereas some 615

are likely to be more effective in reducing nutrient losses in the dissolved form (e.g., grazing restriction). 616

Studies are underway in the STC watershed to quantify the effects of individual BMPs on reducing 617

nutrient losses in different forms and at different scales. Hopefully, these studies will provide a basis for 618

recommending the adoption of individual BMPs. They will also lead to the enhancement of current BMPs 619

20

and development of new BMPs that preferentially reduce losses of dissolved N and P, and will eventually 620

increase the overall efficiency of nutrient loss reduction. 621

622

623

CONCLUSIONS 624

Nitrogen and phosphorus in runoff at the outlets of the two sub-watersheds were mainly in 625

dissolved form (> 80% for both N and P) and were mainly lost during snowmelt events (from 52% to 626

65%). Collective effects of the five BMPs—holding pond below a beef cattle overwintering feedlot, 627

riparian zone and grassed waterway management, grazing restriction, perennial forage conversion, and 628

nutrient management—on hydrology were mostly non-significant but the implementation of these BMPs 629

resulted in a significant reduction in nutrient losses to surface water in the treatment sub-watershed. In 630

particular, when all runoff events were considered, the BMPs resulted in decreases of TN and TP exports 631

in runoff at the treatment sub-watershed outlet by 41% and 38%, respectively. The corresponding 632

reductions in FWMCs were 43% for TN and 32% for TP. In most cases, similar reductions in exports and 633

FWMCs were determined for N and P in dissolved and particulate forms, or when rainfall and snowmelt 634

runoff events were considered separately. 635

In the treatment sub-watershed, retention of nutrients in the holding pond could account for as 636

much as 63% and 57%, respectively, of the BMP-induced reductions in TN and TP. Improvements to 637

nutrient management reduced N inputs onto arable cropland by 36% and P inputs by 59%, in part due to 638

the reduced rates of nutrient application to fields converted from annual crop to perennial forage. Yet, the 639

decrease of N and P inputs have not significantly affected crop yields and the N and P removals due to 640

crop harvesting were maintained at similar levels in the pre- and post-BMP period. As a result, there were 641

significant decreases in N and P balances. It is clear that the holding pond and nutrient management 642

BMPs have contributed substantially to the nutrient loss reductions observed at the outlet of the treatment 643

sub-watershed. Overall, even though the proportional contributions of each of the five BMPs were not 644

individually measured in this watershed scale study, the collective reduction of nutrient losses from these 645

BMPs was substantial. 646

647

648

649

650

651

652

653

21

REFERENCES 654

Agriculture and Agri-Food Canada (AAFC). 2007a. Watershed Evaluation of BMPs (WEBs). 655 http://www4.agr.gc.ca/AAFC-AAC/display-afficher.do?id=1185217272386&lang=eng (verified Jan. 656 18, 2011). 657

Agriculture and Agri-Food Canada (AAFC), 2007b. Soil Landscape of Canada (SLC) Version 3.1.1. 658 1:1,000,000 scale digital soil map for the agricultural regions of Canada. Canadian Soil Information 659 System. Ottawa Ontario. 660

Baker, J.L., and J.M. Laflen. 1983. Water quality consequences of conservation tillage. J. Soil Water 661 Conserv. 38:186-193. 662

Bechmann, M.E., P.J.A. Kleinman, A.N. Sharpley, and L.S. Sposito. Freeze thaw effects on phosphorus 663 loss in runoff from manured and catch-planted soils. J. Environ. Qual. 34: 2301-2309. 664

Bishop, P.L., W.D. Hively, J.R. Stedinger, M.R. Rafferty, J.L. Lojpersberger, and J.A. Bloomfield. 2005. 665 Multivariate analysis of paired watershed data to evaluate agricultural best management practice 666 effects on stream water phosphorus. J. Environ. Qual. 34:1087–1101. 667

Bundy, L.G., T.W. Andraski, and J.M. Powell. 2001. Management practice effects on phosphorus losses 668 in runoff in corn production systems. J. Environ. Qual. 30:1822–1828. 669

Chanasyk, D.S., and C.P. Woytowich. 1986. Snowmelt runoff from agriculture land in the Peace River 670 region. Can. Agric. Eng. 28:7–13. 671

Clausen, J.C., W.E. Jokela, F.I. Potter, III, and J.W. Williams. 1996. Paired watershed comparison of 672 tillage effects on runoff, sediment, and pesticide losses. J. Environ. Qual. 25:1000–1006. 673

Coote, D.R., and L.J. Gregorich L.J. (eds.) 2000. The health of our water—toward sustainable agriculture 674 in Canada. Research Planning and Coordination Directorate, Research Branch,Agriculture and Agri-675 Food Canada, Ottawa, Ont. 676

Daverede, I.C., A.N. Kravchenko, R.G. Hoeft, E.D. Nafziger, D.G. Bullock, J.J. Warren, and L.C. 677 Gonzini. 2003. Phosphorus runoff: Effect of tillage and soil phosphorus levels. J. Environ. Qual. 678 32:1436–1444. 679

Eastman, M., A. Gollamudi, N. Sta¨mpfli, C.A. Madramootoo, and A. Sarangi. 2010. Comparative 680 evaluation of phosphorus losses from subsurface and naturally drained agricultural fields in the Pike 681 River watershed of Quebec, Canada. Agricultural Water Management 97:596–604. 682

Environment Canada, 2011. National Climate Data and Information Archive. Available online at: 683 www.climate.weatheroffice.gc.ca (verified Jan. 18, 2011) 684

Grabow, G.L., J. Spooner, L.A. Lombardo, and D.E. Line. 1999. Detecting water quality changes before 685 and after BMP implementation: Use of SAS for statistical analysis. NWQEP Notes, Number 93. 686 NCSU Water Quality Group. North Carolina State Univ., Raleigh, NC. 687

Granger, R.J., and D.M. Gray. 1990. A net radiation model for calculating daily snowmelt in open 688 environments. Nord. Hydrol. 21:217–234. 689

Granger, R.J., D.M. Gray, and G.E. Dyck. 1984. Snowmelt infiltration to frozen Prairie soils. Can. J. 690 Earth Sci. 23:669–677. 691

Hansen, N.C., S.C. Gupta, and J.F. Moncrief. 2000. Snowmelt runoff, sediment and phosphorus losses 692

under three different tillage systems. Soil Tillage Res. 57:93–100. 693

Hoffmann, C.C., C. Kjaergaard, J. Uusi-Kamppa, H.C.B. Hansen, and B. Kronvang. 2009. Phosphorus 694 retention in riparian buffers: review of their efficiency. J. Environ. Qual. 38:1942-1955. 695

22

King, K.W., P.C. Smiley Jr., B.J. Baker, and N.R. Fausey. 2008. Validation of paired watersheds for 696 assessing conservation practices in the Upper Big Walnut Creek watershed, Ohio. J. Soil Water 697 Conserv.63:380–395. 698

Lake Winnipeg Stewardship Board. 2006. Reducing nutrient loading to Lake Winnipeg and its watershed. 699 Our collective responsibility and commitment to action—Report to the Minister of Water 700 Stewardship. 78 pp. 701

Little, J.L., S.C. Nolan, and J.P. Casson. 2006. Relationships between soil-test phosphorus and runoff 702 phosphorus in small Alberta watersheds. 150 pp. In Alberta Soil Phosphorus Limits Project. Volume 703 2: Field-scale losses and soil limits. Alberta Agriculture, Food, and Rural Development, Lethbridge, 704 AB, Canada. 705

Little, J.L., S.C. Nolan, J.P. Casson, and B.M. Olson. 2007. Relationships between soil and runoff 706 phosphorus in small Alberta watersheds. J. Environ. Qual. 36:1289–1300. 707

Loftis, J.C., L.H. MacDonald, S. Streett, H.K. Iyer, and K. Bunte. 2001. Detecting cumulative watershed 708 effects: the statistical power of pairing. J. Hydrol. 251:49–64. 709

Manitoba Soil Fertility Advisory Committee (MSFAC). 2007. Manitoba soil fertility guide. 710 http://www.gov.mb.ca/agriculture/soilwater/nutrient/fbd02s00.html (verified Jan. 18, 2011). 711

Manitoba Agriculture, Food and Rural Initiatives (MAFRI). 2009. Manure management facts: calculating 712 manure application rates. 713 http://www.gov.mb.ca/agriculture/soilwater/nutrient/pdf/mmf_calcmanureapprates_factsheet.pdf 714 (verified Jan. 18, 2011). 715

McConkey, B.G., W. Nicholaichuk, H. Steppuhn, and C.D. Reimer. 1997. Sediment yield and seasonal 716 soil erodibility for semiarid cropland in western Canada. Can. J. Soil Sci. 77:33–40. 717

Meals, D.W. 2001. Water quality response to riparian restoration in an agricultural watershed in Vermont, 718 USA. Water Science and Technology 43:175–82. 719

Miller, J.J., D. Chanasyk, T. Curtis, T. Entz, and W. Willms. 2010a. Influence of streambank fencing with 720 a cattle crossing on riparian health and water quality of the Lower Little Bow River in Southern 721 Alberta, Canada. Agricultural Water Management 97:247–258. 722

Miller, J.J., T.W. Curtis, E. Bremer, D.S. Chanasyk, and W.D. Willms. 2010b. Soil test phosphorus and 723 nitrate adjacent to artificial and natural cattle watering sites in southern Alberta. Can. J. Soil Sci. 90: 724 331–340. 725

Miller, J.J., B.P. Handerek, B.W. Beasley, E. C.S. Olson, L.J. Yanke, F.J. Larney, T.A. McAllister, B.M. 726 Olson, L.B. Selinger, D.S. Chanasyk, and P. Hasselback. 2004. Quantity and Quality of Runoff from 727 a Beef Cattle Feedlot in Southern Alberta. J. Environ. Qual. 33:1088–1097. 728

Nicholaichuk, W. 1967. Comparative watershed studies in southern Saskatchewan. Trans. ASAE 10:502–729 504. 730

Ontkean, G.R., D.S. Chanasyk, and D.R., Bennett. 2005. Snowmelt and growing season phosphorus flux 731 in an agricultural watershed in south-central Alberta, Canada. Water. Qual. Res. J. Canada. 40:402–732 417. 733

Salvano, E., D.N. Flaten, A.N. Rousseau, and R. Quilbe, 2009. Are current phosphorus risk indicators 734 useful to predict the quality of surface waters in Southern Manitoba, Canada? J. Environ. Qual. 735 38:2096–2105. 736

Schilling, K.E., and Spooner J. 2006. Effects of watershed-scale land use change on stream nitrate 737 concentrations. J. Environ. Qual. 35:2132–2145. 738

23

Schnepf, M., and C. Cox, eds. 2006. Environmental Benefits of Conservation on Croplands: The Status of 739 Our Knowledge. Ankeny, IA: SWCS. 326 pp. 740

Sharpley, A.N., S.C. Chapra, R. Wedepohl, J.T. Sims, T.C. Daniel, and K.R. Reddy. 1994. Managing 741 agricultural phosphorus for protection of surface waters: issues and options. J. Environ. Qual. 742 23:437–451. 743

Sheppard, S.C., M.I. Sheppard, J. Long, B. Sanipelli, and J. Tait. 2006. Runoff phosphorus retention in 744 vegetated fi eld margins on fl at landscapes. Can. J. Soil Sci. 86:871–884. 745

Soil Classification Working Group. 1998. The Canadian System of Soil Classification. Agriculture and 746 Agri-Food Canada, Publ. 1646 (revised). 187 pp. 747

Spooner, J., R.P. Mass, S.A. Dressing, M.D. Smolen, and F.J. Humenik. 1985. Appropriate designs for 748 documenting water quality improvements from agricultural NPS control programs. p. 30-34. In 749 Perspectives on nonpoint source pollution. EPA 440/5-85-001. 750

Tiessen, K.H.D., J.A. Elliot, J. Yarotski, D.A. Lobb, D.N. Flaten, and N.E. Glozier. 2010. Conventional 751 and conservation tillage: influence on seasonal runoff, sediment, and nutrient losses in the Canadian 752 Prairies. J. Environ. Qual. 35:964-980. 753

Tiessen, K.H.D., J.A. Elliot, M. Stainton, J. Yarotski, D.A. Lobb, and D.N. Flaten. 2011. The 754 effectiveness of small-scale headwater storage dams and reservoirs on stream water quality and 755 quantity in the Canadian Prairies. Journal of Soil and Water Conservation 66:158–171. 756

Ulen, B., M. Bechmann, J. Folster, H.P. Jarvie, and H. Tunney. 2007. Agriculture as a phosphorus source 757 for eutrophication in the north-west European countries, Norway, Sweden, United Kingdom and 758 Ireland: a review. Soil Use and Management 23:5–18. 759

Ulen, B., H. Aronsson, M. Bechmann, T. Krogstad, L. Øygarden, and M. Stenberg. 2007. Soil tillage 760 methods to control phosphorus loss and potential side-effects: a Scandinavian review. Soil Use and 761 Management 26:94–107. 762

USEPA. 1993. Paired watershed study design. EPA 841-F-93-009. USEPA Office of Water, Washington, 763 DC. 764

USEPA. 1997a. Techniques for tracking, evaluating, and reporting the implementation of nonpoint source 765 control measures: Agriculture. EPA 841-B-97-010. USEPA Office of Water, Washington, DC. 766

USEPA. 1997b. Linear regression for nonpoint source pollution analyses. EPA 841-B-97-007. USEPA 767 Office of Water, Washington DC 768

Uusi-Kamppa, J. 2005. Phosphorus purification in buffer zones in cold climates. Ecological Engineering 769 24: 491–502. 770

Vaananen, R., M. Nieminen, M. Vuollekoski, and H. Ilvesniemi. 2006. Retention of phosphorus in soil 771 and vegetation of a buffer zone area during snowmelt peak flow in southern Finland. Water, Air, and 772 Soil Pollution 177:103–118. 773

van Vliet, L.J.P., and J.W. Hall. 1991. Effects of two crop rotations on seasonal runoff and soil loss in the 774 Peace River region. Can. J. Soil Sci. 71:533–544. 775

776

777

24

LIST of FIGURES 778

779

Figure 1. Locations of the study sites and land uses on the sites: a) Shaded relief map of the South 780

Tobacco Creek (STC) watershed showing locations of the two sub-watersheds and their land uses 781

in 2002; b) Land use at the Madill (control) sub-watershed in 2008; and c) Land use at the Steppler 782

(treatment) sub-watershed in 2008 and BMPs implemented at the Steppler (treatment) sub-783

watershed since 2005. MSC, MST and MSH are water monitoring stations at the outlets of the 784

control and treatment sub-watersheds and the inlet of holding pond, respectively. BMP1 to 5 are 785

holding pond, riparian management, grazing restriction, forage conversion and nutrient 786

management BMP, respectively (see Table 1 for details). 787

Figure 2. Annual and period total Unit Flow volume (UFV) and P and N exports in different forms and in 788

runoff events of different types measured at the outlet of the Madill (control) sub-watershed (MSC) 789

and the outlet of the Steppler (treatment) sub-watershed (MST). 790

Figure 3. Simple regression analyses of total phosphorous (TP) and total nitrogen (TN) exports and flow-791

weighted mean concentrations (FWMCs) for all runoff events, showing the effects of BMP 792

implementation without accounting for the effects of hydrologic covariates (*, ** and *** denote 793

that the regression is significant at P ≤ 0.05, 0.01and 0.001, respectively). 794

Figure 4. Gross N and P inputs (from fertilizer and manure applications), removals (due to crop 795

harvesting) and balances (= inputs - removals) estimated for cropped fields in the Madill (control) 796

and Steppler (treatment) sub-watersheds. The column height indicates the average value for the 797

period and the vertical bar indicates the full range of the data in the period. The number above the 798

column is the difference (+ for increase and - for decrease) in average values between the Pre-BMP 799

and Post-BMP periods. Significance level of the difference is determined using a two-way t-test 800

(NS, †, * and ** denote non-significant and significant at P ≤ 0.10, 0.05 and 0.01, respectively). 801

802

Tabl

e 1.

Ben

efic

ial M

anag

emen

t Pra

ctic

es (B

MPs

) im

plem

ente

d at

the

Step

pler

(tre

atm

ent)

sub-

wat

ersh

ed d

urin

g th

e Po

st-B

MP

perio

d.

‡

Perc

enta

ge o

f the

tota

l cat

chm

ent a

rea

of th

e St

eppl

er su

b-w

ater

shed

.

      

Tabl

e 2.

Eve

nt m

edia

n va

lues

for t

he h

ydro

logi

c va

riabl

es, n

utrie

nt e

xpor

ts a

nd fl

ow-w

eigh

ted

mea

n co

ncen

tratio

ns (F

WM

Cs)

at t

he o

utle

ts o

f the

Mad

ill

(con

trol)

and

Step

pler

(tre

atm

ent)

sub-

wat

ersh

eds‡

.

‡

Abb

revi

atio

ns: U

FV, u

nit f

low

vol

ume;

Avg

FR, a

vera

ge fl

ow ra

te; P

kFR

, pea

k flo

w ra

te; T

P, to

tal p

hosp

horu

s; T

DP,

tota

l dis

solv

ed p

hosp

horu

s; P

P, p

artic

ulat

e ph

osph

orus

; TN

, tot

al n

itrog

en;

TDN

, tot

al d

isso

lved

nitr

ogen

; PN

, par

ticul

ate

nitro

gen;

NH

3, am

mon

ia; N

Ox,

nitra

te a

nd n

itrite

; PO

C, p

artic

ulat

e or

gani

c ca

rbon

.

Tabl

e 3.

Cor

rela

tion

coef

ficie

nts b

etw

een

the

hydr

olog

ic v

aria

bles

and

bet

wee

n th

e hy

drol

ogic

var

iabl

es a

nd n

utrie

nt e

xpor

ts a

nd fl

ow-w

eigh

ted

mea

n co

ncen

tratio

ns (F

WM

Cs)

at t

he o

utle

t of t

he S

tepp

ler (

treat

men

t) su

b-w

ater

shed

‡§.

‡

Abb

revi

atio

ns: A

vgFR

, ave

rage

flow

rate

; PkF

R, p

eak

flow

rate

; TP,

tota

l pho

spho

rus;

TD

P, to

tal d

isso

lved

pho

spho

rus;

PP,

par

ticul

ate

phos

phor

us; T

N, t

otal

nitr

ogen

; TD

N, t

otal

dis

solv

ed

nitro

gen;

PN

, par

ticul

ate

nitro

gen;

NH

3, am

mon

ia; N

Ox,

nitra

te a

nd n

itrite

; PO

C, p

artic

ulat

e or

gani

c ca

rbon

; Vol

R, f

low

vol

ume

ratio

. §

Bol

d-fa

ced

corr

elat

ion

coef

ficie

nts a

re si

gnifi

cant

at P

≤ 0

.10.

Tabl

e 4.

Res

ults

of t

he si

mpl

e A

naly

ses o

f Cov

aria

nce

(AN

CO

VA

) mod

els f

or th

e hy

drol

ogic

var

iabl

es‡.

‡

Abb

revi

atio

ns: V

ol, r

unof

f flo

w v

olum

e; A

vgFR

, ave

rage

flow

rate

; PkF

R, p

eak

flow

rate

; Prd

, stu

dy p

erio

d; In

t, th

e in

tera

ctio

n te

rm in

the

mod

el.

†,

*, *

*, *

** d

enot

e si

gnifi

canc

e le

vel o

f P ≤

0.1

0, 0

.05,

0.0

1 an

d 0.

001,

resp

ectiv

ely.

Tabl

e 5.

Res

ults

of t

he A

naly

ses o

f Cov

aria

nce

(AN

CO

VA

) usi

ng th

e re

duce

d m

odel

s for

nut

rient

exp

orts

and

flow

-wei

ghte

d m

ean

conc

entra

tions

(FW

MC

s)‡.

‡

Abb

revi

atio

ns: T

P, to

tal p

hosp

horu

s; T

DP,

tota

l dis

solv

ed p

hosp

horu

s; P

P, p

artic

ulat

e ph

osph

orus

; TN

, tot

al n

itrog

en; T

DN

, tot

al d

isso

lved

nitr

ogen

; PN

, par

ticul

ate

nitro

gen;

NH

3, am

mon

ia; N

Ox,

nitra

te a

nd n

itrite

; PO

C, p

artic

ulat

e or

gani

c ca

rbon

; Prd

, stu

dy p

erio

d.

†, *

, **,

***

den

ote

sign

ifica

nce

leve

l of P

≤ 0

.10,

0.0

5, 0

.01

and

0.00

1, re

spec

tivel

y.

Tabl

e 7.

Res

ults

of t

he b

est m

ultiv

aria

te A

naly

ses o

f Cov

aria

nce

(AN

CO

VA

) mod

els f

or n

utrie

nt e

xpor

ts a

nd fl

ow-w

eigh

ted

mea

n co

ncen

tratio

ns (F

WM

Cs)

and

th

e es

timat

ed re

duct

ions

due

to B

MP

impl

emen

tatio

n‡.

‡

Abb

revi

atio

ns: T

P, to

tal p

hosp

horu

s; T

DP,

tota

l dis

solv

ed p

hosp

horu

s; P

P, p

artic

ulat

e ph

osph

orus

; TN

, tot

al n

itrog

en; T

DN

, tot

al d

isso

lved

nitr

ogen

; PN

, par

ticul

ate

nitro

gen;

NH

3, am

mon

ia; N

Ox,

nitra

te a

nd n

itrite

; PO

C, p

artic

ulat

e or

gani

c ca

rbon

; Prd

, stu

dy p

erio

d; In

t, th

e in

tera

ctio

n te

rm in

the

mod

el; V

olR

, flo

w v

olum

e ra

tio; A

vgFR

, ave

rage

flow

rate

. †,

*, *

*, *

** d

enot

e si

gnifi

canc

e le

vel o

f P ≤

0.1

0, 0

.05,

0.0

1 an

d 0.

001,

resp

ectiv

ely.

§ A

slig

htly

diff

eren

t met

hod

was

use

d to

acc

ount

for t

he si

gnifi

cant

eff

ects

of t

he in

tera

ctio

n te

rm (I

nt).

See

text

for d

etai

ls.

Tabl

e 6.

Max

imum

redu

ctio

ns o

f run

off a

nd n

utrie

nt e

xpor

ts a

t the

Ste

pple

r (tre

atm

ent)

sub-

wat

ersh

ed o

utle

t (M

ST) d

ue to

the

hold

ing

pond

BM

P an

d th

e ra

tios

of fl

ow-w

eigh

ted

mea

n co

ncen

tratio

ns b

etw

een

thos

e m

easu

red

at th

e ho

ldin

g po

nd in

let (

MSH

) and

thos

e m

easu

red

at th

e tre

atm

ent s

ub-w

ater

shed

ou

tlet‡

§.

‡

Abb

revi

atio

ns: V

ol, r

unof

f flo

w v

olum

e; T

P, to

tal p

hosp

horu

s; T

DP,

tota

l dis

solv

ed p

hosp

horu

s; P

P, p

artic

ulat

e ph

osph

orus

; TN

, tot

al n

itrog

en; T

DN

, tot

al d

isso

lved

nitr

ogen

; PN

, par

ticul

ate

nitro

gen;

NH

3, am

mon

ia; N

Ox,

nitra

te a

nd n

itrite

; PO

C, p

artic

ulat

e or

gani

c ca

rbon

. §

Bol

d-fa

ced

num

bers

are

thos

e re

duct

ions

that

are

sign

ifica

nt a

t P ≤

0.1

0 in

the

best

mul

tivar

iate

AN

CO

VA

mod

els (

Tabl

e 7)

.

  

 

b)M

adill

(con

trol)

sub

wat

ersh

eda)

Sou

thTo

bacc

oC

reek

wat

ersh

ed

Figu

re 1

!!!!!!

!!!!!!

!!!!!!!

!!!!!!!!

!!!!!!!!!!!!!

!!!!!!!!!!!!!!

!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!

!!

!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!

!!!!!!!!!!!

!!!!!!!!!!

!!!!!!!!!

!!!!!!!!

!!!!!!!

!!!!!!!

!!!!

!

!!!!!!!!!

!!!!!!!!!

!!!!!!

!!!!!!!!!!

!!!!!!

!!

!!!!!!!!

!!!!!!!!

!!!!!!!

!!!!!!

!!!!!!

!!!!!

!!!

!!!!

!!!!!

!!!!!!

!!!!

!!

!!

!!!

!!!

!!!

!!!

!

MSC

MSC

b) M

adill

(con

trol)

sub-

wat

ersh

eda)

Sou

th T

obac

co C

reek

wat

ersh

ed

Man

itoba

Man

itoba

Sask

aSa

ska--

tche

wan

tche

wan

Albe

rta

Albe

rta

!!!!!

!!!!!

!!!!!!

!!!!!!!

!!!!!!!!!

!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!!!!!!!!!

!!!!!!!!!!!!!

!!!!!!!!

!!!!!!!!!!

!!!!!!

!!!!!!!!!!!

!!!!

!!!!!!!!

!!!

!!!!!!!

!!!!!!!

!!!!!!

!!!!!!

!!!!!!!

!!!!!!

!!!!!!

!!!!!!!

!!!!!!!

!!!!!!!!

!!!!!!!!

!!!!!!!!!

!!!!!!!!!!

!!!!!!!!!!

!!

!!!!!!!!!!!

!!!!!!!!!!!!!!

!!!!!!!!!!!!!!!

!!!

!!!!!!!!!

!!

!!!

!!

!!

!!!

!!

!!!

!!!

!!!

!!

!!

!!

!!

!0

0.5

10.

25km

Mad

ill s

ubM

adill

sub

--wat

ersh

edw

ater

shed

!!

!!

!!!

!!!

!!

!!!

!!!

!!!

!!!

!!

!

!!!

!!

!!

!!

!!!

!!!

!!

!!

!!!

c) S

tepp

ler (

treat

men

t) su

b-w

ater

shed

BM

P3,5

MST

MST

³50

0 m

!!!!

!!!!!

!!!

!!

!!!!!!!

!!!!!!!

!!!!

!!!!!

!!!!!!

!!!!!!!!

!!!!!!!!

!!!!!!!!

!!!!!!!!

!!

!!!

!!

!!!!

BM

P3,4

BM

P3,4

BM

P3,4

BM

P3,4

BM

P3,5

BM

P3,5

BM

P2,3

BM

P2,3

BM

P1B

MP1

BM

P2,3

BM

P2,3

³St

eppl

er s

ubSt

eppl

er s

ub--w

ater

shed

wat

ersh

ed

400

m

!!!

!!

!!!!!

!!!!!!

!!!!!!!

!!!!!!!!

!!!!!!!!!

00.

51

0.25

km

MSH

MSH

300

m

Annu

alcr

opFo

rage

Gra

zing

land

or

Pas

ture

Tree

sW

ater

bo

dyO

ther

sSt

ream

Figu

re 2

Period Total UnVolume (M3 h

al Total Unit Flow lume (M3ha-1)

a)

1000

2000

3000

Snow

Rain

4000

8000

1200

0nit Flow ha-1)

PeriodE

AnnuaVol

sphorus 1)

b)

0

1000

0

1200

Snow

‐TDP

Snow

‐PP

Rain‐TDP

6000

Rain‐PP

d Total PhosphoruExport (g ha-1)

Annual Total PhosExport (g ha-

0

400

800

02000

4000

Period TotalExport (g

us

otal Nitrogen rt (g ha-1)

A

c)

00

4000

6000

8000

1000

0Snow

‐TDN

Snow

‐PN

Rain‐TDN

2000

0

3000

0

4000

0Ra

in‐PN

1999

2000

2001

2002

2004

2006

2007

2008

Pre

-P

ost-

Ent

ire

l Nitrogen g ha-1)

Annual ToExpor

0

2000

4000

MSCMST

MSCMST

MSCMST

MSCMST

MSCMST

MSCMST

MSCMST

MSCMST

01000

0

MSCMST

MSCMST

MSCMST

1999

2

000

20

01

2002

2

004

200

6

2007

20

08

----

----

----

-Pre

-BM

P p

erio

d --

----

----

---

---P

ost-B

MP

per

iod

---

Pre

BMP

perio

d

Pos

tBM

P pe

riod

Ent

ire

stud

y pe

riod

a)TP

expo

rt[Ln(gha

‐1)]

b)TN

expo

rt[Ln(gha

‐1)]

Figu

re 3

Y= 0.72

X + 0.28

R²= 0.23

***

510Y= 0.88

X ‐0

.11

R²= 0.31

***

510a) TP expo

rt [Ln(g ha

1 )]

b) TN export [Ln(g ha

1 )]

Y = 0.79

X ‐0

.11

R²= 0.55

***

‐50

‐20

24

68

10

Y = 0.75

X + 0.28

R²= 0.54

***

‐50

‐20

24

68

10

tershed value (Y) 

‐20

24

68

102

02

46

810

Y= 0.41

X ‐0

.45

R²= 0.32

***

01Y = 0.44

X + 0.62

R²= 0.21

**23

c) TP FW

MC [Ln(mg L‐1 )]

(treatment) sub‐wa

d) TN FWMC [Ln(mg L‐1 )]

Y = 0.75

X ‐0

.47

R²= 0.34

**‐2‐1

Y = 0.57

X ‐0

.07

R²= 0.23

*01

Steppler (

‐3

‐3‐2

‐10

1

‐1

‐10

12

3

Madill (con

trol) sub

‐watershed

 value

 (X) 

Pre‐BM

P (calibratio

n) period data point

Post‐BMP (BMP im

plem

entatio

n) period data point

Pre‐BM

P(calibratio

n)pe

riodregressio

nline

Post‐BMP(BMPim

plem

entatio

n)pe

riodregressio

nline

PreBM

P (calibratio

n) period regressio

n line

Post

BMP (BMP im

plem

entatio

n) period regressio

n line

a)Pinpu

tallcropp

edfie

lds

b)Ninpu

tallcropp

edfie

lds

c)Pinpu

tannu

alcrop

pedfie

lds

d)Ninpu

tannu

alcrop

pedfie

lds

Figu

re 4

a) P inpu

t, all cropp

ed fields

b) N inpu

t, all cropp

ed fields

c) P inpu

t, annu

al cropp

ed fields

d) N inpu

t, annu

al cropp

ed fields

g ha‐1yr‐1

+7NS

‐26†

+0NS

‐5**

+14N

S‐16N

S+4

NS

‐4*

e) P re

moval, all crop

ped fie

lds

f) N re

moval, all crop

ped fie

lds

g) P re

moval, ann

ual cropp

ed fields

h) N re

moval, ann

ual cropp

ed fields

Kg

+12N

S+20N

S+2

NS

+1NS

2NS

14NS

+0NS

‐0NS

Kg ha‐1yr‐1

+12N

S+20N

S‐2

NS

‐14N

S

i) P balance, all crop

ped fie

lds

j) N balance, all crop

ped fie

lds

k) P balance, ann

ual cropp

ed fields

l) N balance, ann

ual cropp

ed fields

ha‐1yr‐1

‐4NS

‐45*

‐1NS

‐6**

+16N

S‐12N

S+3

NS

‐4†

Pre‐BMP

Post‐BMP

Post‐BMP

Kg 

Pre‐BMP

Pre‐BMP

Post‐BMP

Post‐BMP

Pre‐BMP

Pre‐BMP

Post‐BMP

Post‐BMP

Pre‐BMP

Pre‐BMP

Post‐BMP

Post‐BMP

Pre‐BMP

P

P

P

Madill 

(con

trol)

Step

pler

(treatmen

t)

P

P

P

P

P

P

P

P

P

P

P

P

P

Madill 

(con

trol)

Step

pler

(treatmen

t)Madill 

(con

trol)

Step

pler

(treatmen

t)Madill 

(con

trol)

Step

pler

(treatmen

t)