UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Evaluation of the WEPP hillslope profile model for estimating runoff and soil loss for two sites in western… Mansoor, Kayyum 1998

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

Item Metadata

Download

Media
831-ubc_1998-0528.pdf [ 6.55MB ]
Metadata
JSON: 831-1.0088648.json
JSON-LD: 831-1.0088648-ld.json
RDF/XML (Pretty): 831-1.0088648-rdf.xml
RDF/JSON: 831-1.0088648-rdf.json
Turtle: 831-1.0088648-turtle.txt
N-Triples: 831-1.0088648-rdf-ntriples.txt
Original Record: 831-1.0088648-source.json
Full Text
831-1.0088648-fulltext.txt
Citation
831-1.0088648.ris

Full Text

E V A L U A T I O N O F T H E W E P P H I L L S L O P E P R O F I L E M O D E L F O R E S T I M A T I N G R U N O F F A N D S O I L L O S S F O R T W O SITES IN W E S T E R N C A N A D A By K A Y Y U M M A N S O O R B.Sc.(Agr.), McGi l l University, 1994 A T H E S I S S U B M I T T E D I N P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F M A S T E R O F S C I E N C E in T H E F A C U L T Y O F G R A D U A T E S T U D I E S D E P A R T M E N T O F S O I L S C I E N C E We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y O F B R I T I S H C O L U M B I A May, 1998 © Kayyum Mansoor, 1998 In presenting this thesis in partial fulfillment of the requirements for an advanced degree at the University of British Columbia, I agree that the library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Stsg-M LC The University of British Columbia Vancouver, Canada DE-6 (2/88) Abstract A B S T R A C T replace empirically based soil loss estimators for use with action agencies involved with based on the fundamentals of soil physics, hydrology and biomass prediction designed to he Water Erosion Prediction Project (WEPP) is a process based erosion prediction model soil and water conservation and environmental planning and assessment. To better assess and predict soil erosion, the WEPP Hillslope Soil Erosion model (version 95.7) was evaluated with measured runoff and soil loss data from two representative agricultural erosion monitoring sites located in distinctive climatic regions in Western Canada. The required climate, soil, slope and management data was carefully compiled to represent actual conditions. A sensitivity analysis was performed to assess model response to changes in parameters, and calibrations were performed to improve model efficiency. Original and calibrated model results were compared to annual and event based observations of runoff and soil loss. The WEPP model responded to changes in land use practices and it efficiently predicted annual runoff for a site with a temperate climate and soil loss for a fallow plot in a colder climate when adjustments to the soil detachment parameters were incorporated. Otherwise, the model did not comply with measured runoff and soil loss due to errors in estimation associated with weaknesses in the winter hydrology component, plant-growth and residue decomposition parameters, and the erosion component. If enhancements in the model components and a more comprehensive built-in database can improve soil loss predictions, the WEPP model can potentially be used as a powerful new tool for applications related to forestry, mine-reclamation, construction sites and rangelands in addition to agricultural situations to assess the impacts of on and off-site soil erosion and sedimentation. ii Table of Contents T A B L E O F C O N T E N T S A B S T R A C T n T A B L E O F C O N T E N T S in L I S T O F T A B L E S vi L I S T O F F I G U R E S vm L I S T O F S Y M B O L S xi A C K O W L E D G E M E N T S xm C H A P T E R 1 : I N T R O D U C T I O N 1 1.1 ISSUE ADDRESSED 3 1.2 OBJECTIVES 4 C H A P T E R 2 : L I T E R A T U R E R E V I E W 5 2.1 SEVERITY OF SOIL EROSION 5 2.2 W A T E R EROSION IN WESTERN CANADA 6 2.2.1 PEACE RIVER 7 2.2.2 LOWER FRASER V A L L E Y 9 2.4 MODELS USED TO ASSESS SOIL EROSION 10 2.4.1 UNIVERSAL SOIL LOSS EQUATION 11 2.4.2 T H E REVISED UNIVERSAL SOIL Loss EQUATION (RUSLE) 12 2.4.3 EROSION/ PRODUCTIVITY IMPACT CALCULATOR (EPIC) 12 2.4.4 SYSTEME HYDROLOGIQUE EUROPEEN SEDIMENT COMPONENT (SHESED) 13 2.4.5 W A T E R EROSION PREDICTION PROJECT ( W E P P ) 14 C H A P T E R 3 : W E P P M O D E L : S T R U C T U R E A N D C O M P O N E N T S 2 0 3.1 W E P P INPUT FILES 20 3.1.1 CLIMATE FILE 20 3.1.2 CROPPING/MANAGEMENT FILE 21 iii Table of Contents 3.1.3 SOIL FILE 21 3.1.4 SLOPE FILE 22 3.2 W E P P M O D E L COMPONENTS 22 3.2.1 CLIGEN: CLIMATE GENERATION COMPONENT 23 3.2.2 IRRIGATION COMPONENT 24 3.2.3 WINTER HYDROLOGY COMPONENT 24 3.2.4 WATER B A L A N C E AND PERCOLATION COMPONENT 25 3.2.5 INFILTRATION AND HYDROLOGY COMPONENT 26 3.2.6 EROSION COMPONENT 28 3.2.7 P L A N T GROWTH COMPONENT 31 3.2.8 RESIDUE DECOMPOSITION COMPONENT 32 3.2.9 SOIL COMPONENT 32 3.3 W E P P OUTPUT FILES 34 3.4 LIMITS OF APPLICATION 35 C H A P T E R 4: D A T A S O U R C E S U T I L I Z E D 3 6 4.1 CLIMATOLOGICAL DATA 36 4.2 M A N A G E M E N T DATA 40 4.3 SOIL AND SLOPE DATA 45 4.4 SOIL LOSS AND RUNOFF DATA 46 C H A P T E R 5 : S E N S I T I V I T Y A N D C A L I B R A T I O N 5 0 5.1 SENSITIVITY ANALYSIS 50 5.2 CALIBRATION 53 5.2.1 BEAVERLODGE CALIBRATION 55 5.3.2 ABBOTSFORD CALIBRATION 56 C H A P T E R 6 : M O D E L T E S T I N G A N D A N A L Y S I S 5 9 6.1 DATA ASSESSMENT PROCEDURE 59 6.2 BEAVERLODGE ANALYSIS 62 6.2.1 Y E A R L Y ANALYSIS 62 I V Table of Contents 6.2.2 E V E N T A N A L Y S I S : 70 6.2.3. W INTER/ S U M M E R A N A L Y S I S 86 6.3 ABBOTSFORD ANALYSIS 91 6.2.1 Y E A R L Y A N A L Y S I S 91 6.3.2 E V E N T A N A L Y S I S 101 6.3.3 B R E A K - P O I N T A N A L Y S I S 119 C H A P T E R 7: C O N C L U S I O N 123 B I B L I O G R A P H Y 125 A P P E N D I X 1: D E T A I L E D C R O P P I N G P R A C T I C E S F O R B E A V E R L O D G E 131 A P P E N D I X 2: D E T A I L E D S O I L L O S S (t ha"1) A N D R U N O F F (mm) O B S E R V E D A T B E A V E R L O D G E 133 A P P E N D I X 3: D E T A I L E D C R O P P I N G P R A C T I C E S F O R A B B O T S F O R D 135 A P P E N D I X 4: D E T A I L E D S O I L L O S S (t ha'1) A N D R U N O F F (mm) O B S E R V E D A T A B B O T S F O R D 137 List of Tables L I S T O F T A B L E S Table 2.1 Table 2.2 Table 2.3 Table 4.1 Table 4.2 Table 4.3 Table 4.4 Table 4.5 Table 4.6 Table 4.7 Table 4.8 Table 4.9 Table 4.10 Table 4.11 Table 4.12 Table 5.1 Table 5.2 Table 5.3 Table 5.4 Table 5.5 Table 6.1 Table 6.2 Table 6.3 Table 6.4 Global extent of land affected by wind and water erosion (Oldeman, 1992 in Lai , Ed . , 1994) 5 Statistical parameters obtained for Animal Science watershed study (Savabi et al., 1995 (I)) 17 Summary of W E P P vs. U S L E runoff and soil loss predictions (Kramer and Alberts, 1992) 18 Monthly climate statistics for Beaverlodge, Monthly climate statistics for Abbotsford Criteria for substituting events at the Abbotsford site .37 .39 39 Cropping practices at Beaverlodge. Treatments are conventional except when indicated with brackets. 41 Cropping practices at Abbotsford. Treatments are conventional except when indicated with brackets. 43 Soil and slope properties for Beaverlodge Soil and slope properties for Abbotsford _ Annual runoff data for Beaverlodge (mm yr-1) Annual soil loss data for Beaverlodge (T ha' 1 yr - 1)_ Annual runoff data for Abbotsford (mm yr'1) Annual soil loss data for Abbotsford (T ha' 1 yr"1) Summary of measured events Sensitivity analysis (based on Beaverlodge fallow plots). Calibration criteria Calibration procedure Calibrated soil parameters for Beaverlodge, Calibrated soil parameters for Abbotsford _ Statistical summary of soil loss (SL) (T ha"1 yr"1) and runoff (RO) (mm yr' 1) for Beaverlodge-Original uncalibrated data, based on yearly data .45 .46 .47 .47 .48 .49 .49 -51 .54 -54 .55 -56 62 Statistical summary of predicted soil loss (SL) (T ha"1 yr'1) and runoff (RO) (mm yr' 1) for Beaverlodge - Calibrated Kb, based on yearly data 66 Statistical summary of predicted soil loss (SL) (T ha-1 yr-1) and runoff (RO) (mm yr-1) for Beaverlodge - Calibrated K b and tc, based on yearly data 67 Statistical summary of predicted soil loss (SL) (T ha' 1 yr'1) and runoff (RO) (mm yr"1) for Beaverlodge - Calibrated Kb and Kn based on yearly data 69 vi List of Tables Table 6.5 Statistical summary of predicted soil loss (SL) (T ha'1) and runoff (RO) (mm) for Beaverlodge-Original uncalibrated data, based on event data 71 Table 6.6 Statistical summary of predicted soil loss (SL) (T ha"1) and runoff (RO) (mm) for Beaverlodge - Calibrated Kb, based on event data 78 Table 6.7 Statistical summary of predicted soil loss (SL) (T ha' ) and runoff (RO) (mm) for Beaverlodge - Calibrated Kb and T c, based on event data 80 Table 6.8 Statistical summary of predicted soil loss (SL) (T ha'1) and runoff (RO) (mm) for Beaverlodge - Calibrated Kb and K„ based on event data 85 Table 6.9 Statistical summary of snowmelt vs. rainfall induced soil loss (SL) (T ha'1) and runoff (RO) (mm) for Beaverlodge - Uncalibrated data, based on event data 86 Table 6.10 Statistical summary of snowmelt vs. rainfall induced soil loss (SL) (T ha"1) and runoff (RO) (mm) for Beaverlodge - Calibrated Kb and rc, based on event data 89 Table 6.11 Statistical summary of soil loss (SL) (T ha' 1 y r 1 ) and runoff (RO) (mm yr' 1) for Abbotsford-Original uncalibrated data, based on yearly data 91 Table 6.12 Statistical summary of calibrated soil loss (SL) (T ha' 1 yr'1) and runoff (RO) (mm yr' 1) for Abbotsford-Calibrated Kb, based on yearly data 95 Table 6.13 Statistical summary of calibrated soil loss (SL) (T ha' 1 yr"1) and runoff (RO) (mm yr"1) for Abbotsford-Calibrated Kb and T„ based on yearly data 97 Table 6.14 Statistical summary of calibrated soil loss (SL) (T ha' 1 yr'1) and runoff (RO) (mm yr' 1) for Abbotsford-Calibrated Kb and K„ based on yearly data 99 Table 6.15 Statistical summary of calibrated soil loss (SL) (T ha' 1 yr' 1) and runoff (RO) (mm yr"1) for Abbotsford-Calibrated Kb and Kt, based on yearly data 100 Table 6.16 Statistical summary of calculated soil loss (SL) (T ha'1) and runoff (RO) (mm) for Abbotsford-Uncalibrated data, based on event data 102 Table 6.17 Statistical summary of calibrated soil loss (SL) (T ha"1) and runoff (RO) (mm) for Abbotsford-Calibrated Kb, based on event data 109 Table 6.18 Statistical summary of calibrated soil loss (SL) (T ha'1) and runoff (RO) (mm) for Abbotsford-Calibrated Kb and T c , based on event data 112 Table 6.19 Statistical summary of calibrated soil loss (SL) (T ha-1) and runoff (RO) (mm) for Abbotsford-Calibrated Kb and K r , based on event data 118 Table 6.20 Statistical summary of calibrated soil loss (SL) (T ha'1) and runoff (RO) (mm) for Abbotsford-Calibrated Kb and Kit based on event data 119 Table 6.21 Precipitation characteristics for selected break-point events 119 Table 6.22 Statistical summary of soil loss (SL) and runoff (RO) efficiency for single peak and break point data for selected events : 121 vii List of Figures L I S T O F F I G U R E S Figure 4.1 Site locations for W E P P simulation study 36 Figure 4.2 Climatological characteristics for Beaverlodge 38 Figure 4.3 Climatological characteristics for Abbotsford 40 Figure 6.1 Comparison of measured (M) and predicted (P) total runoff (mm) and soil loss (t ha-1) at Beaverlodge using W E P P calculated K b , xc, K r and K i 63 Figure 6.2 Comparison of measured (M) and predicted (P) annual runoff (mm yr' 1) at Beaverlodge using W E P P calculated Kb, x„ Kr and Kt 64 Figure 6.3 Comparison of measured (M) and predicted (P) total soil loss (t ha' 1 yr' 1) at Beaverlodge using W E P P calculated Kb, rc, Kr and Kt 65 Figure 6.4 Comparison of measured (M) and predicted (P) total runoff (mm) and soil loss (t ha"1) at Beaverlodge using predicted Kb and rc and W E P P calculated Kr and Kt 68 Figure 6.5 Comparison of measured (M) and predicted (P) annual soil loss (t ha' 1 yr'1) at Beaverlodge using predicted Kb and xc and W E P P calculated Kr and Ki 69 Figure 6.6 1:1 comparison of measured and predicted event runoff (mm) at Beaverlodge using W E P P calculated Kb, t„ Kr and Kt 72 Figure 6.7 1:1 comparison of measured and predicted event soil loss (t ha"1) at Beaverlodge using W E P P calculated Kb, x„ Kr and Kt 73 Figure 6.8 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Beaverlodge plot 0 using W E P P calculated Kb, rc, Kr and Kt 74 Figure 6.9 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Beaverlodge plot 2 using W E P P calculated Kb, T„ Kr and Kt 74 Figure 6.10 Comparison of measured and predicted cumulative runoff (mm) at Beaverlodge using W E P P calculated Kb, xc, Kr and Kt 75 Figure 6.11 Comparison of measured and predicted cumulative soil loss (t ha'1) at Beaverlodge using W E P P calculated Kb, Tc, Kr and Kt 76 Figure 6.12 Comparison of measured and predicted runoff (mm) and soil loss (t ha'1) for all Beaverlodge events using W E P P calculated Kb, T„ Kr and Kt 77 Figure 6.13 Cumulative frequency distribution of measured and predicted runoff (mm) and soil loss (t ha'1) for all Beaverlodge events using W E P P calculated Kb, x„ Kr and tf, 77 Figure 6.14 1:1 comparison of measured and predicted event runoff (mm) at Beaverlodge using predicted Kb and W E P P calculated rc, Kr and Kt 78 Figure 6.15 Comparison of measured and predicted cumulative runoff (mm) at Beaverlodge using predicted Kb and W E P P calculated xc, Kr and K( 79 Figure 6.16 1:1 comparison of measured and predicted event soil loss (t ha"1) at Beaverlodge using predicted Kb and rc and W E P P calculated Kr and Kt 81 Figure 6.17 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Beaverlodge plot 0 using predicted Kb and rc and W E P P calculated Kr and Kt 82 viii List of Figures Figure 6.18 Temporal distribution of runoff (mm) and soil loss (t ha"1) for Beaverlodge plot 2 using predicted K„ and xc and W E P P calculated Kr and Kt 82 Figure 6.19 Comparison of measured and predicted cumulative soil loss (t ha"1) at Beaverlodge using predicted Kb and rc and W E P P calculated Kr and Kt 83 Figure 6.20 Comparison of measured and predicted runoff (mm) and soil loss (t ha"1) for all Beaverlodge events using predicted Kb and TC W E P P calculated Kr and Kt 84 Figure 6.21 Cumulative frequency distribution of measured and predicted runoff (mm) and soil loss (t ha'1) for all Beaverlodge events using predicted Kb and rc and W E P P calculated Kr and Ki 84 Figure 6.22 Comparison of measured and predicted rain vs. snow induced event runoff (mm) at Beaverlodge using W E P P calculated Kb, rc, Kr and Kt 87 Figure 6.23 Comparison of measured and predicted rain vs. snow induced event soil loss (t ha'1) at Beaverlodge using W E P P calculated Kb, tc, Kr and Kt 88 Figure 6.24 Comparison of measured and predicted rain vs. snow induced runoff (mm) and soil loss (t ha'1) for all Beaverlodge events using predicted Kb, and TC, and W E P P calculated Kr and Ki 90 Figure 6.25 Comparison of measured and predicted total runoff (mm) and soil loss (t ha'1) at Abbotsford using W E P P calculated Kb, xc, Kr and K t _ 92 Figure 6.26 Comparison of measured and predicted annual runoff (mm yr"1) at Abbotsford using W E P P calculated Kb, r„ Kr and Kt 93 Figure 6.27 Comparison of measured and predicted total soil loss (t ha' 1 yr"1) at Abbotsford using W E P P calculated Kk, r„ Kr and Kt 94 Figure 6.28 Comparison of measured and predicted annual runoff (mm yr'1) at Abbotsford using calibrated Kb and W E P P calculated x„ Kr and Kt 96 Figure 6.29 Comparison of measured and predicted total runoff (mm) and soil loss (t ha"1) at Abbotsford using calibrated Kb and rc and W E P P calculated Kr and Kt 97 Figure 6.30 Comparison of measured and predicted annual soil loss (t ha' 1 yr'1) at Abbotsford using calibrated Kb and rc and W E P P calculated Kr and Ki 98 Figure 6.31 1:1 comparison of measured and predicted event runoff (mm) at Abbotsford using W E P P calculated Kb, rc, Kr and Kt 102 Figure 6.32 1:1 comparison of measured and predicted event soil loss (t ha'1) at Abbotsford using W E P P calculated Kb, x„ Kr and Kt 103 Figure 6.33 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Abbotsford plot 1 using W E P P calculated Kb, Tc, Kr and Kt 105 Figure 6.34 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Abbotsford plot 2 using W E P P calculated Kb, TC, Kr and Kt 105 Figure 6.35 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Abbotsford plot 5 using W E P P calculated Kb, rc, Kr and Kt 106 Figure 6.36 Comparison of measured and predicted cumulative runoff (mm) at Abbotsford using W E P P calculated Kb, TC, Kr and Kt 107 ix List of Figures Figure 6.37 Comparison of measured and predicted cumulative soil loss (t ha'1) at Abbotsford using W E P P calculated Kb, T„ Kr and Kt 108 Figure 6.38 Comparison of measured and predicted runoff (mm) and soil loss (t ha'1) for all Abbotsford events using W E P P calculated Kb, rc, Kr and Kt 108 Figure 6.39 Cumulative frequency distribution of measured and predicted runoff (mm) and soil loss (t ha'1) for all Abbotsford events using W E P P calculated Kb, x„ Kr and Kt 109 Figure 6.40 1:1 comparison of measured and predicted event runoff (mm) at Abbotsford using calibrated Kb and W E P P calculated tc, Kr and Kt 110 Figure 6.41 Comparison of measured and predicted cumulative runoff (mm) at Abbotsford using calibrated Kb and W E P P calculated rc, Kr and Kt 111 Figure 6.42 1:1 comparison of measured and predicted event soil loss (t ha"1) at Abbotsford using calibrated Kb and xc and W E P P calculated Kr and 113 Figure 6.43 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Abbotsford plot 1 using calibra-ted Kb and % and W E P P calcula-ted Kr and Kt 114 Figure 6.44 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Abbotsford plot 2 using calibra-ted Kb and rc and W E P P calcu-lated Kr and Kt 114 Figure 6.45 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Abbotsford plot 5 using calibra-ted Kb and rc and W E P P calcula-ted Kr and Kt 115 Figure 6.46 Comparison of measured and predicted cumulative soil loss (t ha'1) at Abbotsford using calibrated Kb and rc and W E P P calculated Kr and Kt 116 Figure 6.47 Comparison of measured and predicted runoff (mm) and soil loss (t ha"1) for all Abbotsford events using calibrated Kb and TC W E P P calculated Kr and Kt 117 Figure 6.48 Cumulative frequency distribution of measured and predicted runoff (mm) and soil loss (t ha'1) for all Abbotsford events using calibrated Kb and TC and W E P P calculated Kr and Kt 117 Figure 6.49 Break-point hydrographs for selected events 120 x List 6fSymbols L I S T O F S Y M B O L S a 0 AUSLE a P CuSLB CEC CLAY CU D 8 Dc Df A dur e\ ET EU ez F fi ^nozzle 7 G Gs h I i h h KuSLE Kb Ke Kt Kiadj Kr K-sat L LuSLE m n OM output — outputi output2 PuSLE significance level (0.10 and 0.05 used) the soil water content in the root zone for a given day initial soil water content in the root zone predicted soil loss (USLE) depth-discharge coefficient raindrop-induced turbulence coefficient cover-management factor (USLE) cation exchange capacity proportion of clay in soil sample indicator if ponding occurs within an interval cumulative amount of percolation loss below the root zone slope of the saturated vapor pressure curve at mean air detachment capacity by rill flow rill detachment or deposition rate delivery of interrill sediments to rills duration of precipitation event saturated vapor pressure cumulative evapotranspiration daily potential evapotranspiration vapor pressure Kpa cumulative infiltration depth infiltration rate adjustment factor to account for sprinkler irrigation nozzle psycometric constant sediment load in flow soil heat flux depth of flow interception of precipitation by vegetation time index effective rainfall intensity impact variation ratio of maximum rainfall intensity to average rainfall intensity soil erodibility factor (USLE) baseline conductivity parameter effective saturated hydraulic conductivity interrill erodibility parameter adjusted interrill erodibility rill erodibility parameter saturated hydraulic conductivity slope length slope length factor (USLE) depth-discharge exponent sample size (number of years in simulation) organic matter content of soil sample average of output] and output2 associated output for input varablej associated output for input varable2 supporting practices factor (USLE) m m meq/lOOg % m m kg s"1 m"2 kgs"1 m kg s"1 m"2 h KPa m MJ m2 d"1 m m s"1 kg s"1 m"2 MJ m'2 d"1 m m m s mm hr'1 m s"1 kg s m"4 kg s m 4 s in 1 mmhr"1 m % xi List of Symbols p cumulative infiltration m prep precipitation mm Q cumulative amount of surface runoff m 9 dimentionless interrill erosion parameter -q discharge per unit width of the plane m 3 m"1 s" Qa computed value of event, / -0a soil moisture deficit mm"1 subsurface lateral flow to the drain tiles in Qm mean of measured values -Q m i measured value of event, /' _ r coefficient of correlation _ rv rainfall rate m s"1 R the cumulative rainfall depth m RuSLE rainfall-runoff erosivity factor (USLE) -r2 coefficient of determination -R2 Nash-Sutcliffe efficiency coefficient -Rn net radiation MJ m"2 d RJ daily net solar radiation iy Rs rill spacing m S sensitivity parameter -Se snow water content; (+ for snowmelt, - for snow accumulation) m SuSLE slope gradient or steepness factor (USLE) -Sat saturation of sample at the beginning of the simulation period % SDRm sediment delivery ratio -&ir interrill runoff rate m s"1 Srunqff sensitivity of parameter to runoff -Ssoil loss sensitivity of parameter to soil loss -sw standard deviation of the differences -T test statistic for r-test -T • 1 air average air temperature C critical sheer stress of soil Pa T transport capacity in the rill kg s"1 m" critical sheer stress parameter Nm"2 T 1 ce sediment transport capacity at end of slope kg s"1 m" t. total time when rainfall rate exceeds infiltration temperature s y sheer stress of flow acting on soil particles Pa u time s tP ratio of time of rainfall peak/total duration -tr effective runoff duration s wind speed m s"1 V cumulative rainfall excess depth m variable - the average of varablet and variable^ variable! least value of sensitivity input -variable2 greatest value of sensitivity input -Vf effective fall velocity for the sediment m s"1 w rill width m w mean of differences -X distance down slope m W average capillary potential m xii Acknowledgements A C K O W L E D G E M E N T S I am indebted to many people for the successful completion of this thesis. First, I would like to thank my committee members, Laurens J.P. Van Vliet, Soil Scientist, of the Pacific Agri-Food Research Centre for providing me with valuable data and supervision for the duration of this project, Michael D. Novak, Associate Professor of Soil Science for his attentive editing and for providing valuable guidance as a faculty advisor, and Sie-Tan Chieng, Associate Professor of Bio-Resource Engineering, for his creative contributions and stimulating course lectures in BIOE 560. I would also like to thank the UBC Department of Soil Science, Professor Andy Black and fellow colleagues in the Soil Physics & Biometeorology group as I learned much from these very dedicated individuals. A special thanks to Zoran Nesic for his creative and numerous programming tips and to Rick Kettler, for his help in preparing this final manuscript. Finally, my deepest appreciation goes to my family, Dad, Mum, Farook, Afsana, Sohail and Aamir for their guidance and support, and for encouraging me to pursue my goals, to a very special "Chach" for inspiring me to reach higher and farther, and to my wife and best friend, Rukhsaana. To my parents. xiii Chapter 1 Introduction C H A P T E R 1: I N T R O D U C T I O N As populations continue to grow, we are faced with the dilemma of maintaining non-renewable resources, specifically the top few centimeters of soil. Conservation of this most valuable resource is essential if we are to ensure food production for future generations. Therefore soil conservation planning will be an essential tool to effectively meet this goal. Erosion prediction is the most widely used and effective tool for soil conservation planning. Since it is economically impractical to measure soil erosion under all environmental, management and topographic conditions, erosion prediction is essential to predict the impact of erosion under various influences. The widely used Universal Soil Loss Equation (USLE) pioneered by Wischmeier and Smith (1965) has served the needs of soil conservation planners well. However, due to its empirically based nature, it is difficult to apply the USLE for applications outside the range of conditions for which it was developed. Physically based erosion models such as the Water Erosion Prediction Project (WEPP) incorporate actual mechanisms controlling erosion. They represent a synthesis of individual components which affect erosion including the complex interactions between various factors and their spatial and temporal variabilities (Nearing et al., 1994). The purpose of this study is to evaluate the effectiveness of the WEPP Hillslope model (version 95.7) in predicting soil loss and runoff for two sites in Western Canada. The Hillslope Profile Version was originally developed as a replacement of the Universal Soil Loss Equation (USLE) to be used by action agencies and organizations involved with soil and water conservation planning and environmental impact assessment by agencies, such as the United States Department of Agriculture's (USDA) Soil and Water Conservation Service (SWCS), USDA's Forest Service, United States Department of Interior's (USDI) Bureau of Land Management (Lane and Nearing, 1989). The distinct feature of the WEPP model is that it is process based model which incorporates the fundamentals of hydraulics and erosion mechanics, with few empirical considerations. It allows for both spatial and temporal estimates of erosion and deposition from hillslopes which range from simple and uniform to complex and non-uniform. The WEPP model can be Page 1 Chapter 1 Introduction applied to cropland, rangeland, forests and other applications, and it also incorporates sprinkler and furrow irrigation. The WEPP model, which is currently in its second decade of development incorporates the latest in erosion prediction science to allow the user to better estimate sheet, rill and ephemeral gulley detachment, as well as determine sediment yield and sediment particle characteristics (Flanagan et al., 1991). This new generation of erosion prediction technology is a distributed parameter, continuous simulation, erosion prediction model. The distributed parameters includes rainfall amounts and intensity, soil textural qualities, plant growth parameters, residue decomposition parameters, effects of tillage implements on soil properties and residue amounts, slope shape, steepness and orientation and soil erodibility parameters (Flanagan and Livingston, 1995). Continuous simulation means that the computer program simulates a number of years with each day having a different set of input climate data; based on daily climate data, if rain occurs, it may or may not create a runoff or erosion event. If runoff is predicted to occur, soil loss, sediment deposition, sediment delivery off-site and sediment enrichment for the event will be calculated and added to series of sum-totals. At the end of a simulation period, average values for detachment, deposition, sediment delivery and enrichment are calculated. The entire set of parameters important when predicting erosion are updated on a daily basis including soil roughness, surface residue cover, canopy height, canopy cover, soil moisture, etc. By emulating the physical processes involved in soil erosion, the model mimics the actual steps involved from rain induced detachment, sediment transport in flow to deposition. Continuously updating parameters (88 in total) daily, gives the user the advantage of determining temporal distributions of important parameters. Overall, the WEPP model is potentially a powerful tool for conservation planning. The model estimates when and where potential soil loss can occur for a given hillslope and given management system, and it provides a rapid method for evaluating various soil conservation options. Before this technology can be Page 2 Chapter J Introduction implemented by agencies involved with soil and water conservation, thorough evaluation of the model is necessary and critical for acceptance of the model. Validation requires matching model generated data with existing measured data of soil loss and runoff, the key parameters used in resource management planning. Also, validation is the most critical step of scientific acceptability. Since WEPP is a conservation planning tool, land management decisions will be based in part on results predicted by the model. Hence it is critical that soil and runoff losses predicted by WEPP be thoroughly evaluated to the best existing information on measured rates of soil erosion. 1.1 ISSUE ADDRESSED Since water erosion is one of the main causes of soil degradation in British Columbia (Novak and van Vliet, 1993), the WEPP model might provide an essential management tool to predict runoff and soil losses and to suggest management strategies in order to control and reduce water erosion on agricultural land. To facilitate the applicability of the hillslope version of the WEPP, validation was carried out to determine the accuracy of its predictions. The model is to be tested against data from two representative erosion monitoring sites located in two distinctive climatic regions in Western Canada. The first site was at the Northern Agricultural Research Centre (NARC) located at Beaverlodge, Alberta, in the Peace River region and is representative of Canada's north-west Prairie region. The soil at this site is an Esher clay loam (Dark-Gray Solod). Data from this site was monitored on six plots from 1979-1995. The Peace River region is ranked as having the most severe water erosion risk in Western Canada (Novak and van Vliet, 1993) followed by the second site, Mount Lehman, near Abbotsford, in the Lower Fraser Valley of South-Western British Columbia which exhibits a medium-high erosion risk. The Mount Lehman site is representative of the upland region of the Lower Fraser Valley and the soil is a Whatcom silt loam (Luvisolic Humo-Ferric Podzol). The Mount Lehman site had six erosion monitoring plots which were monitored at two different locations, Matsqui I (1989-1991) and Matsqui II (1991-1994). Both experiments at Mount Lehman were conducted on identical Page 3 Chapter 1 Introduction sites with similar conditions, so that for practical purposes, it was considered to be one site. In this thesis, the Mount Lehman site will be referred to as "Abbotsford". Plant-management, slope and soil data have been collected and soil loss and runoff have been measured at these sites. The climate input data were obtained from the closest Atmospheric and Environment Service (Environment Canada) climate stations. 1.2 OBJECTIVES This research project involved evaluating the accuracy and applicability of the WEPP model in predicting runoff and soil losses for the Beaverlogde and Abbotsford sites. The objectives of this validation study of the WEPP model were: 1. To compile soil, slope, management and climate data required by WEPP input files from 13 years of data collected at Beaverlodge erosion plots and 5 years of data collected from the Abbotsford erosion plots. 2. To evaluate the accuracy of the model in predicting runoff and soil loss from different cropping systems at two sites through a sensitivity analysis, calibration, model efficiency and error analysis. 3. To assess the applicability of the WEPP model for estimating runoff and soil loss in Western Canada. Page 4 Chapter 2 Literature Review C H A P T E R 2: L I T E R A T U R E R E V I E W The following section is a literature review which reflects on soil erosion issues and soil erosion models. It outlines the global and national extent of soil erosion and local concerns of soil erosion in the Peace River and Lower Fraser Valley regions. It also discusses applications of the WEPP model in the Peace River Region, gives a few brief descriptions and relative comparisons of other existing soil erosion prediction technology and it discusses current literature related to WEPP components and predictions. 2.1 SEVERITY OF SOIL EROSION The degradation of the top few centimeters of soil which supports and maintains all life is a globally destructive phenomenon. Over the past three thousand years, mankind has observed the destruction of hundreds of millions of hectares of once productive farmland in China, Korea. North Africa, the middle east, and currently to a lesser extent, in North America. The effects of intensive cultivation, overgrazing, salination and deforestation has promoted and accelerated global soil erosion (Table 2.1). Topsoil which takes thousands of years to develop can easily be degraded in a few years through misuse (Agricultural Institute of Canada, 1986). Table 2.1 Global extent of land affected by wind and water erosion (Oldeman, 1992 in La i , Ed . , 1994) Land area affected by Erosion (10 ha) Region Water Erosion Wind Erosion Africa 227 186 Asia 441 222 South America 123 42 Central America 46 5 North America 60 35 Europe 114 42 Oceana 83 16 World 1094 548 Erosion is a group of processes whereby earth and rock materials are loosened , dissolved or removed from the Earth's surface by the processes of weathering, solution, corrosion and transportation (Larson et al., Page 5 Chapter 2 Literature Review 1983). Before cultivation of crops, erosion was an essential process in the formation geomorphological landforms, but today's fundamental issue is the sustained acceleration of erosion resulting from human activities (Larson et al., 1983). In our present day, farmers are increasingly under pressure to maximize production and profit, with little regard to catastrophic long-term effects of "mining" the soil for immediate returns. Intensification leads to the destruction of this vital resource and in addition it generates serious off-site impacts of fertilizer, pesticide and sediment transported in runoff which is a contributor to watershed pollution. Aside from the serious environmental and social consequences of erosion, economic impacts must be examined. It has been estimated that soil degradation in Canada costs $3.0 million per day or $1.3 billion annually (Science Council of Canada, 1985). Losses associated with soil degradation exceeded $20-$25 per hectare of agricultural land in Canada, or 38% of net farm income and the cumulative costs of soil degradation could reach $42 billion from 1985-2005 (Science Council of Canada, 1985). In Western Canada, two areas have been identified as most susceptible to water erosion, namely the Peace River District and the Lower Fraser Valley (Science Council of Canada, 1985). The combination of slowly impermeable soils, long slopes, summer fallow, high snowfall and rapid spring runoff and intense summer storms all contribute to the high risk of soil erosion in the Peace River Region (Soil at Risk, 1984). Severe water erosion in the Lower Fraser Valley of British Columbia has been influenced by the impact of heavy fall and winter rains on soils used to grow side row crops which are left unprotected after fall cultivation. Unlike most other agricultural regions of Canada, the soils in the Lower Fraser Valley do not freeze in the winter, hence they are potentially exposed to rainfall impact throughout the year. 2.2 W A T E R EROSION IN WESTERN CANADA Page 6 Chapter 2 Literature Review The severity of water erosion in the Peace River Region and the Lower Fraser Valley of Western Canada was described by Novak and van Vliet (1983). According to them, the Peace River Region has the highest erosion risk followed by a medium erosion risk in the Lower Fraser Valley. Detailed studies in van Vliet (1989), van Vliet and Hall (1991), van Vliet (1992), van Vliet et al. (1993), van Vliet (1994), Kulis and van Vliet (1994) and van Vliet and Hall (1995) document the extent of erosion in the Peace River and Lower Fraser Valley Regions. 2.2.1 P E A C E RIVER The seriousness of soil erosion in the Peace River region was first recognized by Albright (1939) who concluded, "if it is not studied, some fine farming country will head for ruin, production will decline, operating costs will rise and there will be washed-out farmers to rehabilitate in the next generation" (in van Vliet, 1989). A study conducted by van Vliet et al. (1993) examined the effects of various tillage treatments on seasonal runoff and soil loss on barley (Hordeum vulgare L.) conducted near Dawson Creek in the Peace River Region of British Columbia. It was concluded that conventional tillage resulted in greater rainfall and runoff than a reduced tillage and zero tillage treatments with no significant differences observed between reduced and zero tillage treatments. Zero tillage and reduced tillage treatments reduced annual soil loss by 81% and 53% respectively in comparison to a conventional tillage treatment. Despite high amounts of runoff particularly during snowmelts, zero-tillage treatments had the lowest soil loss. A six year study (1983-1989) conducted by van Vliet and Hall (1991) investigated the effects of two crop rotations in the Peace River Region. Rotation 1 consisted of summer fallow-canola-barley and rotation 2 consisted of summer fallow-canola-barley-barley underseeded to red fescue. Over the six years, the cumulative runoff and soil loss for the replicated treatments of rotation 1 averaged 229 mm and 5 t ha"1 respectively. This translates into a 25% higher runoff and 79% higher soil loss than rotation 2. Runoff induced from snowmelt accounted for 90 and 96% of the total annual runoff and 39 and 89% of total Page 7 Chapter 2 Literature Review annual soil loss from rotations 1 and 2 respectively. Another soil loss study by van Vliet et al. (1993) in the Peace River Region from 1987-1991 concluded that rainfall was the major contributing factor in soil loss since it contributed 75% of the total four year annual soil loss. The variations in precipitation characteristics can account for the differences. For example, snowmelt induced runoff from the same Beaverlodge erosion plots were observed to produce 99% of total annual soil loss in 1980 and 19% of total annual soil loss in 1982 with a mean fraction of snowmelt induced runoff as a percentage of annual soil loss of 53% from 1980-1984 (van Vliet et al., 1986). In 1982, a record rainfall of 94 mm produced an average soil loss of 27 t ha"1 which accounted for 78% of total soil loss in 1982 (van Vliet et al., 1986). A preliminary investigation of the WEPP model applied to the Peace River was conducted by van Vliet in 1993. The 92.2 version of the WEPP model was tested in continuous mode from 1981-1989 and in single simulation mode for four events to represent break-point precipitation data from a record high measured soil loss event to a low measured soil loss and runoff event. Soil detachment parameters of interrill erodibility (K,), rill erodibility (Kr) and critical sheer stress (Tc) were computed by the model. Since no measurements of the saturated hydraulic conductivity (Ksat) were available, simulations were performed at 10.0 and 0.5 mm hr"1. While in continuous simulation mode, WEPP over predicted rainfall runoff and under predicted snow melt runoff on all three plots when the Ksm was initially assigned to 10 mm hr"'. Using a Ksat of 0.5 mm hr "' improved snowmelt runoff, but it consistently over- predicted runoff for continuous fescue plots. When measured values of Ksa, (20-75 mm hr _1) were applied to the model in single storm mode, the model greatly under predicted soil loss and runoff. In order to match observed and predicted runoff and soil loss, calibrated Ksa, and rc were used to optimize predictions, however, despite the similarity of the soils, optimized values varied by an order of magnitude for both plots. In a study conducted by Hayhoe et al. (1993), WEPP was used to estimate snow depth, soil frost depth and freeze-thaw cycles in order to estimate snowmelt runoff for the Peace River Region. From this study, it was concluded that WEPP needed to better adapt itself for regions with cold climates since it could not make useful predictions of freeze-thaw cycles and snow depth. WEPP overestimated snow depth, frost depth predictions were erratic, and freeze-thaw cycles did no correspond to the number observed at the 0.5 Page 8 Chapter 2 Literature Review m depth (Hayhoe et al., 1993). Later versions of the WEPP model have improved winter hydrology components. Izaurralde et al. (1994) investigated WEPP and EPIC (Erosion/Productivity Impact Calculator) for assessing different management scenarios on soil erosion for the Province of Alberta. The EPIC model (Williams et al., 1990) was developed to predict the relationship between soil erosion and soil productivity. EPIC simulations agreed with spring melt data but for summer events, WEPP predictions were closer and less variable than EPIC predictions. Both WEPP and EPIC were assessed in this study for limitations and advantages. EPIC required less data than WEPP and it also provided estimates on wind erosion and water quality. However, due to EPIC's empirical nature, the availability of obtaining input requirements can be a limitation. The complexity and detail of WEPP appears to exhibit long-term potential for more applied uses. WEPP offers a detailed output, but it also requires a substantial amount of input data. Various studies investigating the effects of crop rotations, tillage practices and applicability of the WEPP model for estimating runoff soil loss and moisture budgets in the Peace River Region have been examined. Local public interest of the erosion problem in the Peace River Region has promoted awareness of the issue and soil conservation measures, Therefore an assessment of soil erosion for this region is beneficial for farmers and regional soil conservation farmers. 2.2.2 LOWER FRASER V A L L E Y Erosion control has been an issue in the Lower Fraser Valley since this region exhibits the second highest degree of water erosion risk in British Columbia (Novak and van Vliet, 1983). Studies examining erosion control practices in the Lower Fraser Valley have been investigated by van Vliet and Hall (1995), van Vliet (1994) and Kulis and van Vliet (1994). Van Vliet and Hall evaluated the effects of planting direction and slope steepness on runoff and soil loss from eight erosion monitoring plots planted with brussels sprouts in the upland area of the Lower Fraser Page 9 Chapter 2 Literature Review Valley. Since there is no information on the effectiveness of across slope farming practices in the region, farmers are reluctant to adopt erosion control practices. Doubling of the slope from 5 to 10% resulted in a doubling of runoff and soil loss, but the effects of cultivation and planting direction on runoff were not conclusive, plots on the adjacent previously uncultivated field experienced an 85% lower soil loss, 72% lower runoff and a doubling in yields than plots in the continuously cultivated fields. The previously uncultivated plots promoted higher rates of infiltration and they maintained a well aggregated soil structure (Kulis and van Vliet, 1995). An Evaluation of the erosion control practices on eight strawberry plots for three years was conducted by van Vliet (1994). Barley planted as a winter cover crop reduced soil loss and runoff by 78 and 43% respectively. Strawberries planted across the rows were also effective erosion control measures since this reduced soil loss by 97% and it reduced runoff by 84%. Accumulation of water in the cross-cultivated fields was responsible for a 28% reduction in yield due to root rot disease. Since erosion can effect the soil's production potential, it was recommended to use a winter crop as a preferred erosion control practice. No previous studies on the WEPP model applied to the Lower Fraser Valley have been investigated. However, due to the severity of erosion in the Lower Fraser Valley it is necessary to investigate and evaluate erosion control practices; The WEPP model may potentially provide to be a useful tool in assessing management effects on erosion. 2.4 MODELS USED TO ASSESS SOIL EROSION Hydrologic models can be classified as deterministic or stochastic in nature. Deterministic, or process based models such as the WEPP are based on the actual physical processes governing soil detachment, transport and deposition. Stochastic models are based on historically observed trends and use empirical relationships to determine outcome. The merits and limits of various erosion prediction models are discussed below. Page 10 Chapter 2 Literature Review 2.4.1 UNIVERSAL SOIL Loss EQUATION Since the late 1960's, the Universal Soil Loss Equation (USLE), originally developed by Wischmeier and Smith (1960; 1965; 1978) has been the most widely applied soil erosion model (Kirkby, 1980). Due to its simplicity, the USLE has been extensively used as the "workhorse of erosion prediction and conservation planning technology in the U.S. and even worldwide" (Renard et al., 1994 (II)). In the USLE, the predicted soil loss (A), is the product of: A - RllSLE Kl/SLE Li/SLE SuSLE ClISLE PuSLE [2-1] RUSLE z - rainfall-runoff erosivity factor KuSLE z = soil erodibility factor L USLE = = slope length factor SlISLE z - slope gradient or steepness factor ClJSLE -= cover-management factor PuSLE z = supporting practices factor Van Vliet (1989) modified the RUSLE and LUSLE factor for the Peace River Region to account for climatic and topographic conditions. The rainfall-runoff erosivity factor, RUSLE was modified to account for snowmelt runoff in addition to rainfall runoff, and the slope gradient factor LU$LE was modified to account for the very long slopes commonly experienced in the Peace River Region. Since the modified factors were based on observed site-specific data, the USLE provided to be a reasonable means to assess relative soil loss in the Peace River Region (van Vliet, 1989). The USLE has been criticized for being ineffective in applications outside the range of conditions for which it was developed for (Nearing et al., 1994 in Lai, Ed. 1994). The simplicity and ease of use of the USLE has served as a good estimator since it predicts an average annual soil loss based on the RUSLE, KUSLE, LUSLE, SUSLE, CUSLE and PUSLE factors, but it cannot assess spatial and temporal variability of soil loss on a hill slope, it does not account for off-site impacts which can be a critical environmental concern (Clark, Page 11 Chapter 2 Literature Review 1985). Also the simplicity of its components cannot express the complex interactive nature of the soil erosion process. 2.4.2 T H E REVISED UNIVERSAL SOIL Loss EQUATION (RUSLE) The revision of the USLE, namely the RUSLE is an improvement over it's predecessor (Renard et al., 1994 (I)). RULSE uses the same fundamental structure (A - RUSLE KUSLE LUSLE SUSLE CUSLE PUSLE) as the USLE, but the algorithms used to calculate individual factors have been changed significantly in the RUSLE mainly due to improvements and accessibility of computerized technology (Renard et al., 1994 (II)). The major improvements includes additions and modifications to the USLE factors: RUSLE factor is based on data from more weather stations and it uses a correction factor to account for rainfall impact on flat slopes striking water ponded on the surface; KUSLE is adjusted to account for seasonal changes such as freezing, thawing, soil moisture and soil consolidation; L V S L E factor assigns new equations based on ratio of rill to interrill erosion and it accommodates complex slopes; CUSLE factor accounts for prior land use, canopy and surface cover, surface roughness and soil moisture, and it divides each rotation into 15 day intervals which recalculates soil loss based on temporal variations; and the PUSLE factor is based on hydrologic soil groups and it also accommodates intercropping (Renard et al., 1994 (I)). The RUSLE is scientifically more advanced than the USLE, and it is a relatively simple approach to determine soil loss on a hillslope. However, due to it's empirically based approach, it can only be applied to conditions and locations governed by the RUSLE database. Based on the type of prediction required, the RUSLE can demonstrate to be a useful tool. 2.4.3 EROSION/ PRODUCTIVITY IMPACT CALCULATOR (EPIC) The EPIC model was developed in the mid-eighties consists of various sub-models designed to simulate the bio-physical behavior of agroecosystems as well as economic factors involved in production systems Page 12 Chapter 2 Literature Review (Izuarralde et al., 1994). Specifically, EPIC predicts, or estimates the long term relationship between erosion and productivity through two approaches which both involve plotting erosion on the x-axis verses a term referred to as the erosion/ productivity index (EPI) on the y-axis (Williams et al., 1990). A major strength of EPIC is that it can simulate the fate of nutrients and pesticides applied to an agricultural field and it assesses the impact of soil erosion on productivity. EPIC applies the Soil and Water Conservation Service (SCS) curve number method to compute effective runoff and it uses the original or modified USLE to predict erosion estimates from rainfall and runoff. When applied to the Peace River Region, EPIC was determined to better predict soil loss from spring melt, however WEPP simulations for the summer were closer to measured values and less variable than EPIC simulations (Izuarralde et. al., 1994). When executed without calibration in Arkansas pasture fields EPIC accurately predicted runoff and nutrient/ sediment transport with significant correlation with observed data (Edwards etal., 1994). EPIC has demonstrated itself to be an effective tool for assessing impacts of land use practices on erosion and subsequent off-site impacts associated with runoff. EPIC's main limitation is due to it's empirical nature because it can only be applied to areas for which it's parameters have been calibrated for. 2.4.4 SYSTEME HYDROLOGIQUE EUROPEEN SEDIMENT COMPONENT (SHESED) The Systeme Hydrologique Europeen Sediment Component (SHESED) is a spatially distributed erosion and sediment yield component used for the existing Systeme Hydrologique Europeen (SHE) hydrologic modeling system, for use at the catchment scale (Wicks and Bathurst, 1996). Hydrologic processes are modeled either by finite difference representations of partial differential equations or by empirical equations derived from independent research (Wicks et al., 1992). The SHESED component simulates soil erosion by rainfall impact, leaf drip and overland flow (without rilling), and the transport of eroded material by overland flow (Wicks et al., 1992). Page 13 Chapter 2 Literature Review Application of SHE and SHESED in Iowa to rainfall-induced sediment catchment events indicated a good reproduction of observed temporal variations in sediment yield by calibrating the raindrop soil erodibility coefficient and the overland flow soil erodibility coefficient (Wicks and Bathurst, 1996). A comparison study conducted in Wyoming showed a good correlation in terms of volume of sediment yield and net bed erosion for a 37 day period, however there were some discrepancies in the timing of simulated sediment discharge (Wicks and Bathurst, 1996). SHESED can also be a useful tool in determining sediment yield in conservation planning. Simulations indicate that SHESED can successfully distinguish between different classes of land use cover but detailed variations in characteristics such as the effects of rills and microtopography are not reflected in the model's processes (Wicks et al., 1992). To be effective, it is suggested that SHESED parameters have to be calibrated for individual locations by performing in-situ simulation studies (Wicks et al., 1992; Wicks and Bathusrst, 1996). Although SHESED can provide to be an effective tool, site-specific parameter calibration can be impractical, and long-term simulations have yet to be validated. 2.4.5 WATER EROSION PREDICTION PROJECT (WEPP) Since empirical models area based on inductive logic, it can only be applied to those range of conditions for which the parameters have been calibrated (Nearing et al., 1994(11)). It is often impractical economically and time-wise to implement and assess field experiments in order to develop empirical parameters. The WEPP is a continuous simulation processed oriented model based on the fundamentals of erosion and hydrologic science. The main advantage of WEPP over other forms of erosion prediction technology is that: WEPP can give spatial and temporal distributions of soil loss and downslope deposition effects; it provides on-site and off-site particle characteristics of sediment; and since it is processed-based, it can accommodate a wide range of conditions which may not be economical or practical to field test (Lane et al., 1989). A more detailed description of the model and it's components is discussed in Chapter 3. Page 14 Chapter 2 Literature Review The WEPP model has been in development since 1986. The objective of the WEPP program is to improve erosion prediction technology for action agencies involved with conservation planning and impact assessment (Flanagan et al., 1994). The model is a synthesis of graphical user interfaces and mathematical models which allow the user to better estimate sheet, rill and ephemeral gully detachment along with determining field sediment yield and sediment particle characteristics (Flanagan et al., 1994). The model computes and updates 88 parameters essential to hydrologic processes, plant growth and residue decomposition and effects on soil physical properties. Interactions between processes is extensively considered in WEPP which gives the model one of it's greatest strengths in being able to simulate complex processes on an infinite variety of conditions. WEPP can be executed on a continuous or single storm basis. While in continuous simulation, the model imitates processes essential to the erosion process as affected by slope, soil type, management effects and climate. Since WEPP's primary purpose is to predict erosion, the CLIGEN program is used to generate climate data based on statistical climate parameters which forecasts weather based on historical information. Other input data required to successfully execute the model can be obtained from the site modeled, or based on soil type, topography and management practices common to the site being investigated. Although most testing is limited to agricultural conditions, WEPP can potentially be utilized in applications related to construction sites, forestry and mine reclamation. Before the model is accepted scientifically, by the conservation planner and the land owner, a thorough evaluation of WEPP is required by validation of predicted results with existing erosion data. Since WEPP is conservation planning tool, land management decisions which are always associated with an economic and social cost will be based on the models results, therefore, it is critical that the model predictions be thoroughly evaluated (Nearing et al., 1994(1)). Initial testing of the model was conducted at the National Soil Erosion Research Laboratory (NSERL) in West Lafayette, Indiana. Approximately 2000 plot years of data from natural runoff plot events at eleven locations were selected to evaluate the models response (Nearing et al., 1994(1)). The data selected was the Page 15 Chapter 2 Literature Review most reliable from the current database of soil erosion taken from the National Repository of Soil Erosion data at the NSERL. Representative soil and slope information with historical climate and management information was compiled to build input files for WEPP. One of the model's most critical parameter is the saturated hydraulic conductivity. This parameter can be assigned to an initial baseline value (Kb) which the model internally alters to account for management effects or it can be assigned to an effective constant value (Ke) which the model does not update. By using the varying hydraulic conductivity (Kb), the model gave better predictions than maintaining a constant hydraulic conductivity (Ke) in all cases except for one. In the same study, comparison of the WEPP and RUSLE erosion factors was conducted. Empirically based models such as the RUSLE cannot account for interactions between factors because model predictions are linearly dependent on it's parameters. WEPP, due to it's process based interactive nature can factor into consideration effects of soil surface cover by plant residues which is an important factor in determining model output. A validation study in the Ukraine cited in Nearing et al. (1994(1)) showed excellent agreement between measured and predicted soil loss. A method to determine an equation for estimating Kb from SCS curve number approach was conducted by Risse et al. (1995). Since it is expensive and time consuming to determine values for effective parameters, an approach was developed to formulate an equation describing the Kb value based on the percent sand, percent clay and cation exchange capacity of the soil. The Kb values were calibrated so that the annual runoff predicted by WEPP was equal to the SCS curve number predictions for 43 soils. Estimated values obtained from the equation compared favorably to measured data. The WEPP predictions using the optimized and estimated values of Kb was better in terms of average error and model efficiency, and WEPP predicted values of Kb were shown to be superior to predictions obtained from the curve number approach. In a study conducted by Savabi et al. (1995 (IV)), hydrometeorological, soil, topography and vegetation data from a Texas rangeland site were used to test the WEPP rangeland hydrology model. Since WEPP is capable of estimating root zone soil water content and surface runoff, these two properties were compared to site data. Comparison between measured and simulated root zone and soil water content for soil plots produced a Nash-Sutcliffe coefficient of efficiency (R2) (Equation 6.1) of 0.99 and 0.84, respectively, Page 16 Chapter 2 Literature Review indicating very good model efficiency, plots with herbaceous vegetation produced R2 values of 0.46 and 0.53 for comparison of measured and simulated values of simulated storm runoff and soil water content respectively. Savabi et al. (1995(1)) conducted a study whereby a Geographical Resources Analysis Support System-Geographic Information System (GRASS-GIS) was used to parametize the WEPP model. GRASS was not integrated with the WEPP model, but was rather used to obtain the many input soil and slope parameters required by WEPP. The WEPP model was evaluated on the Animal Science watershed in the Indian Pine Natural Resources station in West Lafayette, Indiana. WEPP was tested using three scenarios: 1) entire watershed was represented by one hillslope with one overland flow element (OFE-region of homogenous soil and land cover); 2) watershed is simulated as one hillslope with three OFE's; 3) the watershed was delineated into three different regions to represent the three different flow paths encountered in the watershed, each with one OFE each. The percent deviation (% dev), standard error (Se) and Nash-Sutcliffe coefficient of efficiency (R2) for the experiment are listed below (Table 2.2) Table 2.2 Statistical parameters obtained for Animal Science watershed study (Savabi et al., 1995 (D) Scenario % dev* Se. R2 1. one hillslope/ one OFE 15.4 mm 5.13 mm -0.84 2. one hillslope/ three OFE's 17.3 mm 5.15 mm -0.85 3. three hillslopes/ one OFE 8.05 MM 3.76 M M 0.01 measured storm runoff for 15 events - 86.1 mm Applying GRASS-GIS maps to recognize watershed configurations and representative hillslopes improved WEPP's ability to predict storm runoff. The results obtained in this study are encouraging and it indicates that using a GRASS Geographic Information System to parametize WEPP is potentially a powerful tool. In wet climates, drainage can be an important component since drainage of water redistributes water in the root zone. Drainage can lower high water tables which reduce infiltration and it can alleviate wet soil moisture conditions which can reduce runoff and soil loss. The drainage component of the WEPP model is discussed in detail by Savabi (1993). The drainage routine in WEPP is heavily synthesized from the DRAINMOD model (Skaggs, 1987). The drainage component in WEPP uses the upper most soil layer to Page 17 Chapter 2 Literature Review compute infiltration and all layers are subjected to percolation to lower layers, evapotranspiration and flow to drainage tiles. A validation study discussed by Savabi (1993) showed that WEPP was able to simulate the effect of subsurface drainage on storm runoff in the Willamette Valley watershed in Oregon with close agreement with measured values. Simulated storm runoff and simulated peak runoff prior to drainage installation produced a 0.92 and 0.88 correlation respectively with measured values. After a drainage system installation, WEPP simulated storm runoff and peak storm runoff with a 0.97 and 0.93 correlation respectively against measured values. A study comparing WEPP and USLE predictions on a 24 year period (1954-1977) of conventional corn was carried out by Kramer and Alberts (1992). The results were statistically analyzed by crop stage periods of rough fallow (F), seed bed (SB), rapid growth (P1&2), reproduction and maturation (P3), residue (P4) and for the entire tillage year. The results of the simulations are tabulated in Table 2.3. Table 2.3 Summary of WEPP vs. USLE runoff and soil loss predictions (Kramer and Alberts, 1992) Period Coefficient of Efficiency (R2) WEPP runoff WEPP soil loss USLE soil loss F 0.84 -13.65 -11.66 SB 0.86 0.61 0.73 P1&2 0.80 0.79 0.36 P3 0.64 -0.38 -17.77 P4 0.30 0.58 0.13 tillage year 0.74 0.68 0.20 Variation of WEPP predicted soil loss and runoff was less than observed runoff and soil loss, but the WEPP variation resembled the observed variation better than the USLE variation. It was also observed that for extreme values, WEPP simulated runoff and soil loss was less than observed runoff and soil loss indicating a model bias in underestimating large event runoff and subsequent soil loss. Based on the results of the experiment, it was concluded that: WEPP simulated soil loss distributions are more representative of observed soil loss distributions than the USLE soil loss distributions; and WEPP simulated runoff and soil loss values are biased high for low runoff events and biased low for extreme events. Page 18 Chapter 2 Literature Review Numerous studies have examined individual components of WEPP, and comparison studies comparing WEPP to more traditional empirically based soil erosion models have been investigated. Due to the physical based nature of WEPP, WEPP outperformed other models. Because of it's comprehensive nature, WEPP supports many various management practices, which can be applied to almost any region for customized applications. Spatial and temporal variability's can be incorporated into the model to assess when and where soil loss is predicted to occur. However, due to its complexity, a few thoughts arise: complexity can improve predictions, but it also requires an extensive input database; inaccuracies in estimating critical parameters can greatly affect model output; multiple interactions between components are present, and non-linearly related; scenarios, soil and slope can be arranged infinitely, so how can one assess and convey model results?; if the model is intended to be used as a tool to enhance soil conservation decisions, is it sufficiently accurate? why does the scientific community persist on analyzing problems instead of solving them? If WEPP is to be scientifically accepted and to be used as an effective decision maker, model results need to be compared to existing data. The fundamental objective of this study is to determine the effectiveness of the Water Erosion Prediction Project model in predicting soil loss and runoff, if we to confidently apply the model for conditions in Western Canada. No civilization in history has been able to survive without the capacity to sustainable preserve it's food production. The great civilizations of the Mesopotanians and Babylonians which lie in the fertile crescent between the Tigris and Euphrates River flourished, but their eventual demise was promoted by the mismanagement of soil and water resources. Today's civilization has been able to mitigate this by enhancing food production by inputting fertilizer, pesticide, water and advancing cultural practices, but we are no less immune to the faults encountered by our ancestors. Soil conservation needs to be implemented by society and promoted through government if we are achieve a level of sustainable food production; to provide for future generation without sacrificing the needs of the present. Prediction of soil erosion can be an effective tool in promoting healthier soil conservation practices. Page 19 Chapter 3 WEPP Model: Structure and Components C H A P T E R 3: W E P P M O D E L : S T R U C T U R E A N D C O M P O N E N T S The WEPP model can conveniently be described in three categories: pre-processing, processing and post-processing or the input files, the model components and the output options respectively. This section will describe these various categories and their components to give the reader a better understanding of the Water Erosion Prediction Project model. Components critical to this simulation study will be described in detail. A complete description of the user summary and model components can be located in Flanagan and Livingston (1995). 3.1 W E P P INPUT FILES The four basic input data requirements for the WEPP Hillslope model include a climate data file, a slope file, a soil data file and a cropping/management data file. The model can also simulate irrigation by additional input files. The WEPP model is accompanied by satellite programs consisting on an interface and file builders while allow creation of new data files or modifications of existing data files. The climate file can be generated by the CLIGEN climate generator which allows the user to choose information from over 1000 weather stations in the United States. Parameters for the soil file can be obtained from measuring soil composite properties, and erodibility related parameters which can be measured by laboratory techniques or calculated. Slope files are primarily based on the actual geometry of the landscape in question. The cropping-management files require the largest number of input parameters which describe different plants, tillage implements, tillage, planting and harvest dates, management practices, etc. 3.1.1 CLIMATE FILE Climate file is easily generated using CLIGEN program (Nicks et al., 1995) either within the WEPP interface or outside of it, and user has the option of selecting from over 1000 weather stations in the United Page 20 Chapter 3 WEPP Model: Structure and Components States. To generate a climate file, the user needs to specify monthly means, standard deviation and skewness for precipitation, temperature, solar radiation and wind information. An example of the required input data can be viewed on Tables 4.1 and 4.2. Climate data can be inputted in either continuous or single storm mode, both of which can be segregated into actual break-point data or as a single peak hydrograph. Since this study deals with long term predictions, continuous simulation was applied, in single peak hydrograph mode and in break-point mode for selected events. For evaluation of the WEPP model for Canadian conditions, actual measured climate data is collected at stations closest to the erosion monitoring sites. Incomplete data was substituted by calculated data based on temporal considerations or by using generated data. A Canadian version of the CLIGEN program can be obtained from the AES. 3 .1 .2 CROPPING/MANAGEMENT FILE The cropping/management file contains the largest number of input parameters which describe the different plants, tillage implements, tillage sequences, management practices, etc. The primary inputs required for this file are the initial conditions, tillage, planting and harvesting dates, types of tillage implements used for surface effects, specific plant types, and residue management practices. Parameters relating to crops and tillage implements can be obtained from the built-in WEPP management database or plant parameters not found in the WEPP database may be extracted through the Crop Parameter Intelligent Database System (CPIDS) which can conveniently downloaded from the Internet. The user can build or edit existing cropping/management files either within the interface file builder or by using a text editor. 3 .1 .3 SOIL FILE The soil file requires parameters relating to and including soil texture, strength, erodibility and infiltration rates. For each OFE, information on soil properties up to a depth of 1.8 m and up to 8 different soil layers may be inputted. Accurate estimation of soil physical and hydrologic parameters is essential when operating the WEPP erosion prediction model. A key parameter for WEPP in terms of infiltration is the Green and Ampt effective conductivity parameter which is related but not equal to the value of the Page 21 Chapter 3 WEPP Model: Structure and Components saturated hydraulic conductivity of the soil. If certain soil parameters do not exist, they can be estimated. For cropland soils containing less than 30% sand, the baseline interill and rill erodibility parameters (Kt and Kr respectively), critical shear parameter (Tc) and effective hydraulic conductivity of the soil (Kb) can be calculated by the following equations (Flanagan and Livingston, 1995): Ki = 6054000-55130 CLAY kg s m"4 [3.1] Kr = 0.0069 + 0. 134e(020CLAr> s m"! [3.2] TC = 3.5 Nm"2 [3.3] Kb = 0.0066 e<244/CLAY> mmhr1 [3.4] Where CLAY refers to the percentage of clay found in the uppermost soil layer. Parameter estimations for soils containing more than 30% sand and for rangeland conditions are included in the WEPP user summary (Flanagan and Livingston, 1995). Initial estimation of soil parameters for the Beaverlodge and Abbotsford erosion sites were determined using equations 3.1-3.4. 3.1.4 SLOPE FILE The slope file can be easily built based on the actual geometry of the slope by either using the slope builder interface or by hand. The user has the added advantage of graphically previewing the slope shape. Slopes are built by indicating slope (in percentage) on the relative position on the point of measurement. For example, at 37% of the total slope length, the slope at that position is 14%. The WEPP model allows the user to simulate many types of non-uniformity's on a hillslope through the use of strips of OFE's. 3.2 W E P P M O D E L COMPONENTS The hillslope profile version is based on the best available science for predicting soil erosion on hillslopes. Relationships in the model are based on solid scientific theory and parameters in the model were derived from a broad range of experimental data (Lane and Nearing, 1989). Data can be processed independently Page 22 Chapter 3 WEPP Model: Structure and Components or interactively in specific WEPP model components, or sub-programs which are used to simulate specific processes. The components of the hillslope profile version include: 1. a climate generation components which provides daily weather information 2. an irrigation component which accommodates stationary sprinkler and furrow irrigation systems 3. a winter process component which simulates frost and thaw development in the soil, snow accumulation and snow melt 4. a daily water balance and percolation component 5. an infiltration and hydrology component based on a modified Green-Ampt infiltration equation which incorporates unsteady rainfall 6. a hillslope erosion (rill and interrill) and deposition component 7. a plant growth component for rangeland and cropland conditions 8. a residue decomposition component 9. a soil component for parameters which influence hydrology and erosion In addition to the erosion component, it also includes a stochastic climate generator used to generate daily climate information for a specified period, a hydrology component based on a modified Green-Ampt infiltration equation, solutions of the kinematic wave equation, a daily water balance component, a plant growth and decomposition component and an irrigation component. 3.2.1 C L I G E N : CLIMATE GENERATION COMPONENT Climate simulation for the WEPP is stochastically generated by the CLIGEN model (Nicks et al, 1995), which runs separately from the WEPP model. Daily rainfall, temperature, solar radiation and wind parameters are generated according to statistical parameters for the specific location. CLIGEN creates climate input data files for WEPP which contain daily values for rainfall amount (prep), duration (dur), time to peak (tp), intensity at peak (ip), maximum and minimum temperatures (tmax and tmin respectively), solar radiation (rad), wind speed and direction (w-vl and w-dir respectively) and dew point temperature Page 23 Chapter 3 WEPP Model: Structure and Components (tdew). Daily rainfall is desegregated into a single-peak storm pattern (time-rainfall intensity format) for use by the infiltration and runoff components of the model. 3.2.2 IRRIGATION COMPONENT The irrigation component of WEPP can simulate stationary sprinkler systems which accommodates solid set, slide roll and hand move systems and furrow irrigation systems which can simulate uniform inflow, surge and cut-back flow (Kottwitz, 1995). Sprinkler events are simulated as rainfall events of uniform intensity. Scheduling options for both systems are classified as depletion level and fixed-date. Depletion level scheduling determines the date and amount of irrigation based on available soil moisture depletion and the fixed-date option uses predetermined irrigation amounts and dates. 3.2.3 WINTER HYDROLOGY COMPONENT The winter hydrology component of WEPP is designed to simulate snow accumulation and density, snowmelt, soil frost and thaw. Winter hydrologic processes are critical aspects of landscapes located in areas north and south of 40 degrees latitude (Savabi et al., 1995(11)). Snow accumulation in the model predicts whether the hourly precipitation falling is in the form of rain or snow and it also accounts for changes in snow depth and density. The melt component estimates the amount of snowmelt occurring, also on a hourly basis, hence hourly temperature, precipitation and radiation needs to be calculated before simulating snow accumulation or melt on a frozen soil (Savabi et al., 1995(11)). The snow-melt process is determined by using solar radiation, air temperature and wind parameters. To determine the flow of heat into and out of the soil and subsequent changes to thaw and frost depths, WEPP applies simple heat flow theory with daily information on temperature, solar radiation, residue cover, plant cover and snow cover (Flanagan and Livingston, 1995). Page 24 Chapter 3 WEPP Model: Structure and Components Soil freezing and thawing influence soil properties such as hydraulic conductivity, erodibility and water holding capacity. It also computes changes in the soil and water infiltration capacity which affect the hydrology and water balance of the soil during the winter period. 3.2.4 WATER BALANCE AND PERCOLATION COMPONENT This component is designed to use input from climate, infiltration and crop growth components to estimate the soil/water content in the root zone and evaporation losses. The model maintains a continuous water balance on a daily basis by using the following equation (Savabi and Williams, 1995): 0 = 0,„ + (P-l) ±SrQ-ET-D-Qd [3.5] where: 0 = the soil water content for a given day (m) © i n = initial soil water content in the root zone (m) P - cumulative infiltration (m) / = interception of precipitation by vegetation (m) S7 = snow water content( + for snowmelt,-for snow accumulation) (m) Q = cumulative amount of surface runoff (m) D = cumulative amount of percolation loss below the root zone (m) Qd = subsurface lateral flow to the drain tiles (m) ET = cumulative evapotranspiration (m) The water balance and percolation component of the hillslope model is based on the water balance component of SWRRB (Simulator for Water Resources in Rural Basins) (Williams and Nicks, 1985) which has been modified for better estimating soil evaporation and percolation parameters. This component uses information provided by the weather generation component, infiltration component, and the plant growth component. A continuous balance of moisture is maintained in the root zone on a daily basis by the water balance component. Evapotranspiration is computed in two ways, depending on the data available in the climate input files. In the case where daily radiation, temperature, wind and dewpoint are available, the Page 25 Chapter 3 WEPP Model: Structure and Components WEPP model uses a the Penman equation with the original wind function, method (Penman, 1963 and Jensen 1974): Eu=-^-(Rn -G,) + -Z-j6.43(lJD + 0S3ul)(e°-el) [3.6] where: Eu - daily potential evapotranspiration (MJ m"2 d"1) S = slope of the saturated vapor pressure curve at mean air temperature 7 = psycometric constant Rn = net radiation (MJ m 2 d"1) Gs - soil heat flux (MJ m"2 d"1) uz - wind speed (m s"1) e°z = saturated vapor pressure (KPa) ez = vapor pressure For the case where solar radiation and temperature data are available, the model uses the Priestley-Taylor (1972) method: E =0.00128 - ^ J . [3.7] 58.3 5 + y where: R„l = daily net solar radiation (ly) For both methods, incoming net radiation is calculated by multiplying the incoming daily radiation by (1-A), where A is the albedo of the soil surface. Eu is converted to meter per day by dividing it by 2.501-2.361*10"3 Tair, where Tair is the average air temperature. The water balance component directly affects hydrology, irrigation, plant growth, residue decomposition. 3.2.5 INFILTRATION AND HYDROLOGY COMPONENT The hydrology component of WEPP can be classified into surface and sub-surface hydrology. The main purpose of the WEPP surface hydrology component is to determine the duration of excess rain, rainfall Page 26 Chapter 3 WEPP Model: Structure and Components intensity during the period, runoff volume and peak discharge rate for the erosion component. Secondarily, it provides the amount of water which infiltrates into the soil for the water balance and crop growth/ residue decomposition calculations which are then used to update the infiltration, runoff routing and erosion parameters (Stone et al., 1995). Infiltration is computed by using a modified version of the Green-Ampt Mein-Larson (GAML) model (Mein and Larson, 1973) which has been adjusted to accommodate unsteady rain by Chu (1978). The average infiltration rate,/ (m s"1) for an interval is computed as: where: F = cumulative infiltration depth (m) ?, = time (s) i — current time interval i-l — previous time interval Chu (1978) determined an indicator, C„ (m), which computes if ponding occurs within a given interval of rainfall intensity given no ponding occurs at the beginning of an interval as: C=R.-V,-\ [3.9] where: Rj - the cumulative rainfall depth (m) V, = cumulative rainfall excess depth(m) Ke - effective saturated hydraulic conductivity (m s"1) *P = average capillary potential (m) 6d - soil moisture deficit (m m"1) rp = rainfall rate (m s"1) Rainfall excess occurs when the infiltration rate is less than rainfall rate. The volume of rainfall excess is reduced to account for depressional storage, and runoff commences once the depressional storage has been Page 27 Chapter 3 WEPP Model: Structure and Components satisfied. Surface runoff is calculated by the kinematic wave equation or an approximation of the kinematic wave solutions obtained for a range of rainfall intensity distributions, hydraulic roughness and infiltration parameter values (Flanagan and Livingston, 1995). The kinematic equation for flow on a plane is computed by: d h d q — + - i U v [3.10] at ax with a depth discharge relationship: q = ahm [3.11] where: h - depth of flow (m) q = discharge per unit width of the plane (m3 m"1 s"1) a = depth-discharge coefficient m = depth-discharge exponent The sub-surface hydrology component models sub-surface lateral flow to the tiles and sub-surface drainage routines (Savabi et al., 1995(111)). Root zone soil water redistribution is a critical part of the hydrology for three primary reasons: soil water content affects rainfall and runoff events, root zone soil water content is used in the interaction between soil water and plant growth and soil water content is used in residue decomposition. 3.2.6 EROSION COMPONENT The erosion component of the WEPP model is governed by sediment continuity, detachment, deposition, sheer stress in rills and transport capacity. The basic equation used to describe the movement of sediment along a hillslope in the hillslope erosion component is described by steady state sediment continuity (Foster et al., 1995): Page 28 Chapter 3 WEPP Model: Structure and Components dG dx ••Df+Dt [3.12] where: G = sediment load in flow (kg s"1 m"2) x - distance down slope (m) Df - rill detachment or deposition rate (kg s"1 m"2) Dj = interill sediment delivery rate (kg s"1 m2) For ease of computations, Df and D, are calculated on a per rill area basis, consequently, G is solved on a per unit rill width basis, after which, soil loss is expressed in terms of soil loss per unit land area. Conceptually, interill soil erosion is the process whereby sediment is delivered to concentrated flow channels or rills, and interrill sediment is then deposited in the rill or is carried off the hillslope by rill flow. Interill soil erosion parameter, G'\s modeled as (Foster et al., 1995): [3.13] where: L = slope length (m) Tce = sediment transport capacity at end of slope (kg s"1 m"') te = total time when rainfall rate exceeds infiltration (s) tr = effective runoff duration (s) Delivery of interrill sediments to rills (DJ (kg s"1 m"2) is a given by (Foster et al., 1995): Di = KIADJ Ie Cv SDRRR FNOZZLE | ^ | [3.14] w where: Kiadj = adjusted interrill erodibility (kg s m4) le - effective rainfall intensity (m s"1) air = interrill runoff rate (m s"1) SDRRK = sediment delivery ratio Fnozzle = adjustment factor to account for sprinkler irrigation nozzle impact variation Page 29 Chapter 3 WEPP Model: Structure and Components Rs = rill spacing (m) w - rill width (m) The sediment delivery ratio (SDRRR) is a function of the random roughness of the soil surface, fall velocity of the individual particle size classes, and particle size distribution. Soil detachment is computed when the hydraulic sheer stress exceeds the critical sheer stress of the soil and when sediment load is less than transport capacity. Rill detachment and deposition is modeled as a function of the difference between sediment transport capacity (Tc - a function of hydraulic sheer acting on soil and a transport coefficient) and sediment load (G) present in rill flow as follows (Foster et al., 1995): Df=Dc [3.15] c / where: Dc = detachment capacity by rill flow (kg s"1 m"2) Tc = transport capacity in the rill (kg s"1 m"1) When the hydraulic sheer stress of the rill flow exceeds the critical sheer stress of the soil, detachment capacity is expressed as (Foster et al., 1995): Dc = Kr(rf-Tc) [3.16] where: Kr = rill erodibilty parameter (s m"1) Tf = sheer stress of flow acting on soil particles (Pa) Tc = critical sheer stress of soil (Pa) Rill deposition is computed when sediment load (G) is greater than sediment transport capacity (Tc) as follows (Foster et al., 1995): bVf Df= 1— [3.17] f q(Tc-G) Page 30 Chapter 3 WEPP Model: Structure and Components where: (5 = raindrop-induced turbulence coefficient Vf = effective fall velocity for the sediment (m s"1) q = flow discharge per unit width (m 3 m"1 s"1) WEPP also computes the effects of selective deposition of different sediment classes and estimates a sediment size distribution leaving the hillslope, and it also computes a sediment enrichment of the sediment. Sediment enrichment refers to the mass increase of more chemically active fine sediment particles (silt, clay, organic matter) due to selective deposition of coarse sediment. The primary erodibility parameters representing rill erodibility (Ki) and interrill erodibility (Kr) are based on extensive field studies which are specifically designed for the WEPP model, (Foster et al., 1995) and hence, does not rely on empirically based USLE based relationships for parameter estimation. Adjustments to soil detachment are made to incorporate the effects of canopy cover, ground cover and buried residue effects on soil detachment and transport, which are made possible by plant growth and residue decomposition parameters in WEPP. Since the erosion routines incorporate spatially varied water balance and infiltration, the model can compute erosion estimates for the case of non-uniform hydrology on slopes. 3.2.7 PLANT GROWTH COMPONENT The plant growth component simulates plant growth for cropland and rangeland conditions, it's main purpose being that it simulates temporal variations in plant and residue variables such as canopy cover, canopy height, root development and biomass produced which is removed during a harvest operation or ends up as surface residue material (Arnold et al., 1995). The cropland model is based on the EPIC plant growth sub-model which predicts biomass accumulation as a function of heat units and photosynthetically active radiation. Moisture and temperature stress are incorporated to reduce potential growth. Critical parameters computed in the crop growth model include growing degree days, mass of vegetative dry matter, canopy cover and height, root growth and plant basal area. The model also accommodates rotations and strip-cropping practices. Page 31 Chapter 3 WEPP Model: Structure and Components The rangeland growth model estimates the initiation of above and below ground biomass, and rangeland management practices such as herbicide application, burning and grazing can be simulated. The plant growth component provides information to the water balance component which allows estimation of daily water use by the plants and extraction of water in the upper layers. Information produced from canopy height and cover are utilized by the erosion component to estimate interrill soil detachment. Residue created by leaf-drop after senescence or residue created after harvest is relayed to the residue decomposition component. Since WEPP is a continuous simulation model, the plant growth component plays an important part in simulating the growth of plants and their impact on hydrologic and erosion processes. 3.2.8 RESIDUE DECOMPOSITION COMPONENT Decomposition of residue for croplands is based on a "decomposition day" approach. Each residue type is given an optimal rate for decomposition where environmental influences of temperature and moisture reduce the optimum rate (Stott et al., 1995). The residue decomposition component estimates decomposition of flat residue mass, standing material, submerged residue mass and dead root mass (Flanagan and Nearing, 1995). Based on the type and method residue and harvest management techniques, the model's decomposition component separates total residue mass at harvesting into standing and flat components. The also model tracks decomposition of residue from three previous harvests. WEPP can accommodate several types of residue management practices including residue removal, shredding, burning and contact herbicide. (Flanagan and Livingston, 1995) 3.2.9 SOIL COMPONENT Page 32 Chapter 3 WEPP Model: Structure and Components The soil component updates parameters which influence hydrology and erosion. The impacts of various soil properties and model parameters are computed within the soils component of the WEPP model. Tillage activity during a simulation acts to decrease the soil bulk density, increase the soil porosity, change soil roughness and ridge height, destroy rills, increase infiltration parameters and change erodibility parameters (Flanagan and Livingston, 1995). The four soil parameters which have the greatest influence over the hydrology portion of the erosion process are: random roughness, ridge height, bulk density and effective hydraulic conductivity (Alberts et al., 1995). Random roughness is often a consequence of tillage operations, but any operation which disturbs the soil creates soil roughness. The decay of random roughness following a tillage operation is predicted in the soil component from a relationship including a random roughness parameter and cumulative rainfall since last tillage (Flanagan and Nearing, 1995). Ridge height results when the soil is arranged in a particular way by a tillage implement, which can also be referred to as oriented roughness. Hydraulic resistance to overland flow and depressional storage of rainfall are positively correlated with random roughness. Tillage, rainfall weathering, freezing and thawing all account for changes in random roughness and bulk density. Bulk density reflects the total pore volume of the soil, and it is used to predict several infiltration parameters including wetting front suction. One of the critical parameters of the WEPP model, namely the effective hydraulic conductivity can be invoked to make adjustments to the baseline value. Interrill and rill erodibility parameters (Kj and Kr respectively) are also updated in the soil component. Interrill erodibilty is a measure of the soil's resistance to raindrop impact. Adjustments to this parameter are made to account for root biomass, freezing and thawing, canopy cover, residue cover, and surface sealing and crusting (Flanagan and Nearing, 1995). Since rangelands are not affected by tillage impacts, the interrill erodibility parameter for rangelands are made to account for freezing and thawing, and livestock compaction. Rill erodibility is a soils susceptibility to detachment by concentrated rill flow. It is also defined as the increase in soil detachment per unit increase in shear stress (xc) of clear water flow since critical shear stress is a threshold parameter at which a rapid increase in soil detachment occurs per unit Page 33 Chapter 3 WEPP Model: Structure and Components increase in shear stress. Similarly to interrill detachment, the rate of detachment in rills is influenced by tillage, living root biomass, incorporated residue, soil consolidation, freezing and thawing (Flanagan and Nearing, 1995). 3.3 W E P P OUTPUT FILES WEPP produces many types of different outputs in various quantities depending on the user's wishes: • erosion summary file • hillslope plot file • storm summary files • graphic data output file • crop yield output file • winter, water, soil, plant, range The most basic output contains runoff and erosion summary information which may be produced on a storm-by-storm, monthly, annual or average annual basis. Time integrated estimates of runoff, erosion, sediment delivery and sediment enrichment are contained in this output as well as spatial distribution of erosion on 100 equally spaced points along a hillslope. The sum totals of these values are divided by the number of years in simulation to give an average annual deposition of each point. The output in this file is segregated into on-site and off-site effects of soil erosion. The on-site effects contain average annual soil loss estimates over the areas of the hillslope experiencing soil loss and deposition along a hillslope, giving a table of detachment/deposition at 100 points along a hillslope. The off-site impacts of erosion includes the estimated average annual sediment delivery from the hillslope, including particle size distributions of detached sediment and an estimate of enrichment of the specific surface area of the sediment-which can be a key parameter to determine the various influences of various management systems on sediment and sediment borne pollutants reaching waterways (Flanagan and Livingston, 1995). Page 34 Chapter 3 WEPP Model: Structure and Components Abbreviated summary information for each runoff and soil loss event can also be generated showing rainfall, runoff, soil detachment, sediment deposition, sediment and delivery, and enrichment ratio. A large graphics file can also be created which can be accessed by a built-in graphics program which allows the user to plot several different variables. Detailed soil, plant, water balance, crop, yield and rangeland can also be produced for further detailed investigation. In addition, a graphical output of the hillslope profile can be produced showing the original slope profile superimposed on the profile determined after a model run showing points of detachment/ deposition. Overall, the output provides a potentially powerful tool for conservation planning because the model estimates when and where along a hillslope soil loss problems are likely to occur so that conservation measures can be implemented. The various data and graphical output allows easy interpretation of the results. Overall, the WEPP model provides a rapid method for evaluating various soil conservation programs (Flanagan and Livingston, 1995). 3.4 LIMITS OF APPLICATION Erosion predictions created by WEPP are applicable only to field-sized areas or conservation treatment units (Flanagan and Livingston, 1995). To represent a large area, it may be possible to simulate a portion of a hillslope, assuming there is a distinct physiographic feature. If many distinct shapes, slopes may be modeled independently, or as a single watershed simulation with the Watershed Interface. A recommended watershed size should be limited to less than 260 ha (640 acres) and the model should not be applied to areas with permanent channels since these scenarios can not currently be simulated in WEPP (Flanagan and Livingston, 1995). Page 35 Chapter 4 Data Sources Utilized C H A P T E R 4: D A T A S O U R C E S U T I L I Z E D Figure 4.1 Site locations for WEPP simulation study This section details the methodologies and procedures used to compile the input data for the WEPP model and measured data required for model comparison and analysis. The input data required to successfully execute the WEPP model were collected from two erosion monitoring sites located in Western Canada (see Figure 4.1). Thirteen years of data from six plots were used from Beaverlodge, Alberta, and six years of data were used from two identical sites in Mount Lehman, near Abbotsford, British Columbia. Since WEPP is a processed based model, it was crucial to compile the existing climate, management, soil and slope data to represent conditions at the sites as accurately as possible. 4.1 CLIMATOLOGICAL DATA Climatological data for Beaverlodge was obtained from the Atmospheric and Environmental Service (AES) of Canada located in Downsview, Ontario. Data was collected by the AES at the Agriculture Canada climatolological station at Beaverlodge (latitude 55°12' N, longitude 119° 24' W; elevation 695 m) located Page 36 Chapter 4 Data Sources Utilized within one kilometer of the plot site (Hayhoe et al. 1993). In general, the Peace River Region has a moderate, continental climate dominated by Polar Continental and Arctic air masses (van Vliet et al., 1984). More than 50% of the total precipitation falls during the summer months, mostly as local thunderstorms of high intensity (van Vliet et al., 1984). The mean daily temperature in January ranges from -9° to -18°C which accounts for a portion of the large accumulation of snow-cover during the winter. Accelerated erosion in the Peace river region is influenced by rapid snowmelts on frozen soils often caused by warm Chinooks and the occurrence of high-intensity rainstorms (van Vliet, 1991). Missing solar radiation data was substituted by 33 year (1960-1993) daily means. Discrepancies in rainfall characteristics of more than 4 mm between daily and accumulated five minute interval values were assumed to be incorrect or missing. Missing values were determined from actual break-point climatic records and hydrographs obtained from Beaverlodge. When no records or charts were available, data was generated by the CLIGEN program. Basic Statistical parameters for Beaverlodge can be viewed in Table 4.1, and a graphical representation of the climate data can be viewed in Figure 4.2. Table 4.1 Monthly climate statistics for Beaverlodge J F M A M J J A S O N D mean Prep (mm) 2.7 2.6 2.3 3.3 4.3 6.4 5.5 5.6 3.7 2.7 3.1 2.9 st. dev. Prep (mm) 3.1 2.8 2.5 4.7 6.1 10.8 8.4 8.2 5.0 3.4 3.3 3.7 skew Prcp(mm) 3.1 2.6 2.4 3.3 3.1 5.2 4.6 3.4 3.2 3.1 1.7 3.4 wet-wet Prob (%) 0.5 0.5 0.5 0.4 0.5 0.5 0.5 0.5 0.6 0.4 0.5 0.5 wet-dry Prob (%) 0.3 0.2 0.2 0.2 0.2 0.3 0.3 0.3 0.3 0.2 0.2 0.2 mean T m a x (°C) -8.6 -4.8 0.2 9.0 15.9 19.5 21.7 21.0 15.5 9.6 -1.6 -6.9 mean T ^ (°C) -17.9 -14.4 -9.8 -2.1 3.2 7.1 8.9 8.0 3.8 -0.7 -10.0 -16.0 st. dev. T m a x (°C) 11.1 9.7 7.8 6.0 5.4 4.3 4.2 5.1 5.9 6.6 8.6 10.2 st. dev. T ^ (°C) 10.7 9.5 7.8 4.5 3.6 3.0 2.7 3.3 4.0 5.2 8.2 10.0 mean Solar rad. 180 232 359 455 526 576 566 511 389 249 169 162 st. dev. Solar rad. 303 260 230 196 206 214 209 248 286 268 273 308 mean MN 0.0 0.0 0.0 6.0 7.4 10.4 13.0 12.8 6.9 4.3 3.2 0.0 mean T d e w (°C) -14.0 -14.7 -9.8 -3.7 0.6 5.7 8.9 8.2 3.9 -1.4 -9.9 -16.1 interval 5 38 81 10 108 499 12 89 53 33 25 2 T P (%) 0.01 0.05 0.13 0.14 0.25 0.78 0.79 0.88 0.94 0.97 1.0 1.0 Page 37 Chapter 4 Data Sources Utilized £ 100-50-1 • • • • • i i i 1 1 1 lUu dik,ilLJir  JUL .J. JlUL,I . 82 83 84 85 86 87 89 90 91 92 93 94 Mean Precipitation (per event) Mean Precipitation Durations (per event) j f m a m j j a s o n d Mean Daily Temperature = 3 o 1 2 Q 1 0 j f m a m j j a s o n d Mean Daily Solar Radiation Mean Daily Wind Velocity j f m a m j j a s o n d j f m a m j j a s o n d j f m a m j j a s o n d Figure 4.2 Climatological characteristics for Beaverlodge Climatological data for Abbotsford was also obtained from AES. Data was collected from the Abbotsford Airport (latitude 49°03' N, longitude 122° 22' W; elevation 54 m) located within nine kilometers of the plot site. Unlike Beaverlodge where convective rainfall can result in large spatial variability, the Lower Fraser Valley Region which includes Abbotsford is characterized by frontal rains which are more spatially uniform although orographic effects induce spatial variability. Rainfall in the upland area of the Lower Fraser Valley is characterized by a distinct six-month wet-period from October to March accounting for 70% of the total yearly precipitation (van Vliet and Hall, 1995). November and January are the wettest months, contributing 12.2 and 10.5 mm of mean precipitation per event respectively (Table 4.3). Snowfall contributed less than 6% to the annual precipitation (van Vliet and Hall, 1995). Page 38 Chapter 4 Data Sources Utilized Table 4.2 Monthly climate statistics for Abbotsford J F M A M J J A S O N D mean Prcp(mm) 10.5 9.8 8.1 8.0 6.5 6.7 6.7 6.2 7.6 9.7 12.2 9.2 st. dev Prep (mm) 12.0 12.1 8.1 8.6 7.9 8.1 9.1 7.8 10.3 11.8 12.5 10 skew Prep (mm) 2.5 2.9 1.7 2.0 2.2 2.3 3.0 2.0 2.3 2.6 2.1 2.4 wet-wet Prob (%) 0.8 0.8 0.8 0.7 0.7 0.7 0.5 0.5 0.5 0.7 0.8 0.8 wet-dry Prob (%) 0.3 0.3 0.3 0.3 0.3 0.3 0.2 0.1 0.2 0.3 0.5 0.4 mean T m a x (°C) 6.4 8.6 12.1 14.8 18.0 20.5 23.2 24.1 21.0 15.2 8.9 5.6 mean T m i n (°C) -0.2 0.7 2.6 4.8 7.6 10.1 11.7 11.5 8.7 5.4 2.3 -0.6 st. dev. T m a x ( °C) 4.2 4.3 3.5 4.2 4.5 4.3 4.4 4.1 4.4 4.1 4.0 4.3 st. dev. T ^ (°C) 4.7 4.3 2.9 2.8 2.7 2.2 2.1 2.5 2.9 3.5 4.3 4.5 mean Solar rad. 94 160 250 359 432 462 492 440 325 196 103 77 st. dev. Solar rad. 44 73 100 136 153 163 163 144 108 86 48 36 mean MN 7.0 6.9 7.3 7.9 7.5 8.4 7.5 8.3 8.7 10.0 8.2 6.4 mean T d e w (°C) -0.3 0.1 2.3 4.6 7.6 10.2 12.1 12.2 10.0 6.5 2.6 -0.8 interval 38 81 10 108 499 12 89 53 33 25 33 25 T P (%) 0.0 0.0 0.1 0.1 0.3 0.8 0.8 0.9 0.9 1.0 1.0 1.0 Missing solar radiation and wind variables were substituted by daily means determined from 13 years (1980-1993) of data at Abbotsford. Since rainfall was collected at the sites for the entire simulation period (1989-1994), precipitation data from Abbotsford were verified with break-point data to obtain the most accurate representation of on-site precipitation characteristics (dur, Tp and Ip) for WEPP input. Table 4.3 summarizes the criteria and amount of events substituted for. Table 4.3 Criteria for substituting events at the Abbotsford site Criteria Number All events greater than 30 mm 34 All missing events greater than 5 mm 76 All measured events 86 If large discrepancies were observed, data for previous and proceeding days were also verified. Precipitation data for measured events were verified with break-point data for several days prior to the events in order to obtain realistic antecedent soil moisture conditions. Climatological characteristics for Abbotsford are graphically represented in Figure 4.3. Page 39 Chapter 4 Data Sources Utilized Measured Precipitation for Abbotsford J l l ? 8 ° - | '60-1.3 40-a, 1 20 0-Li 90 94 15 -Mean Precipitation (per event) Mean Precipitation Durations (per event) ' 10-5 -o « 6 " o 4-j f m a m j j a s o n d j f m a m j j a s o n d Mean Daily Temperature Mean Daily Solar Radiation Mean Daily Wind Velocity j f m a m j j a s o n d j f m a m j j a s o n d j f m a m j j a s o n d Figure 4.3 Climatological characteristics for Abbotsford 4.2 M A N A G E M E N T DATA Alfalfa, which had been grown continuously at the Beaverlodge site since 1950, was disked in August 1979 in preparation for the start of the erosion study on plots 1 and 2 (van Vliet, 1992). Plots 0, 3, 4 and 5 were constructed in the following year (1980) by plowing, and disking in August 1980. For simulation purposes, initial condition files were created for each plot and were based on the day prior to the first day of simulation (January 1, 1981). The initial conditions were calculated by simulating the two years prior to the beginning of the 1981 calendar year and obtaining the results at the end of the two year simulation for the comparison simulation. Plots 0, 3, 4 and 5 were simulated as continuous alfalfa with Page 40 Chapter 4 Data Sources Utilized surface disturbance prior to the study period, and plots 1 and 2 had one year of canola and fallow respectively prior to the study period. All six plots were operational at the beginning of 1981. After four years in 1985, the Beaverlodge study changed from a crop comparison to an erosion control experiment that involved monitoring soil loss on badly eroded fallow plots 2 and 5 and measured recovery of these plots in terms of barley yield (van Vliet, 1992). Fall cultivation was eliminated on plot 0, plot 1 was cultivated across slope, plot 3 remained in continuous fescue and plots 2, 4 and 5 remained in continuous annual cropping. The erosion control experiment lasted for five years, after which the crop comparison experiment resumed as before (van Vliet, 1992). The main crops grown for the Beaverlodge study, which are also commonly cultivated in the Peace River Region, were barley (Hordeum vulgare L.), canola (Brassica rapa) and red fescue (Festuca rubra L.). Cropping-management values for fescue and barley were not available on the WEPP plant database and had to be substituted with bromegrass and spring wheat, respectively, as suggested by LJ.P. van Vliet. Common agricultural management practices used in the region were employed on the plots. All crops were implemented at a medium fertilization level. Cropping practices at Beaverlodge can be viewed in Table 4.4, with treatments indicated in brackets. Table 4.4 Cropping practices at Beaverlodge. Treatments are conventional except when indicated with brackets. PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 1981 fescue barley fallow fescue canola fallow 1982 fescue canola fallow fescue barley fallow 1983 fescue barley fallow fescue canola fallow 1984 fescue canola fallow fescue barley fallow 1985 barley barley (CS)2 fallow fescue barley barley 1986 barley barley (CS) barley fescue barley barley 1987 barley (SC)1 barley (CS) barley fescue barley barley 1988 barley (SC) barley (CS) barley fescue barley barley 1989 barley (SC) barley (CS) barley fescue barley barley 1990 fescue barley fallow fescue barley fallow 1991 fescue barley fallow fescue barley fallow 1992 fescue barley fallow fescue barley fallow 1993 fescue barley fallow fescue barley fallow SC means spring cultivation; CS means cross-slope cultivation Page 41 Chapter 4 Data Sources Utilized The Abbotsford study was conducted on two similar sites, Matsqui 1 (November 1989 - April 1991) and Matsqui 2 (June 1991 - December 1993). The experiment for the Matsqui 1 site measured the effects of planting direction and previous cultivation on water erosion under brussels sprouts (Brassica oleracea; cultivar Lunet) (van Vliet and Hall, 1995). This site consisted of two adjacent fields with the same silt loam soil (Whatcom), and had been in an annual cultivation of cole crops and strawberries since 1976 (Kulis and van Vliet, 1994). Replicated plots with two cropping management treatments on two different slopes were compared to two replicated plots on an adjacent field which had never been in annual cultivation. Runoff sampling did not commence until November 1, 1991, hence the first ten months were considered to be a break-in period for simulation purposes, with runoff and soil loss comparison commencing after the break-in period. The Matsqui 2 site experiment was a three year evaluation of erosion control practices on strawberries (Fragaria spp., Rosaceae; cultivar Sumas) (Kulis and van Vliet, 1994). This study consisted of two replicates of four treatments which included: control (no cover crop, no drainage), barley (Hordeum vulgare) winter cover crop, sub-surface drained and cross-slope treatment. The drained plots were observed to produce runoff sooner and longer than the other treatments, which could have possibly resulted from improper installation of the drainage system, resulting in obstruction of seepage and hence back flow creating wetter than normal conditions (van Vliet, 1994). Due to this, the sub-surface drainage treatment was eliminated from the WEPP simulations. In late April, 1991, a previous brussels sprout crop was plowed under. Although the original plot location (Matsqui 1) was relocated, both sites had similar soil, physiographic and management features. For simulation purposes, soil and slope data from Matsqui 2 were used to develop the soil and slope input files, since the required soils data were available for this site only, and surface soil characteristics were known to be very similar. Management data for both sites were amalgamated to produce the management input files for a crop rotation of two years of brussels sprouts followed by three years of strawberries. At each plot locations, management practices were performed according to common local practices. The cropping practices for Abbotsford can be found in Table 4.5. Page 42 Chapter 4 Data Sources Utilized Table 4.5 Cropping practices at Abbotsford. Treatments are conventional except when indicated with brackets. Plotl Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 1989 brussels brussels brussels brussels brussels brussels sprouts sprouts sprouts sprouts sprouts (CS) sprouts (CS) 1990 brussels brussels brussels brussels brussels brussels sprouts sprouts sprouts sprouts sprouts (CS) sprouts (CS) 1991 strawberries strawberries strawberries strawberries strawberries strawberries (C) (Q (CS) (CS) 1992 strawberries strawberries strawberries strawberries strawberries strawberries (C) (C) (CS) (CS) 1993 strawberries strawberries strawberries strawberries strawberries strawberries (C) (C) (CS) (CS) 1994 strawberries* strawberries* strawberries* strawberries* strawberries* strawberries* (C) (C) (CS) (CS) turned over in spring; C = cover crop; CS - cross slope cultivation Prior to planting brussels sprouts, the plots were disked twice, turnplowed, disked twice again, cultivated and rotovated. The brussels sprouts were planted 40 cm apart with 100 cm in-between the rows. During the growing season, the plots were cultivated with shovels three to four times to prevent weed emergence, and the crop was hand harvested in late November for the 1989 and 1990 growing seasons. Planting and cultivation created a 15-20 cm ridge for brussels sprouts which were accommodated in the WEPP surface effects by specifying the ridge height. The effects of micro-topography are an important consideration in WEPP's erosion component, therefore ridge heights and rill widths had to be incorporated in the surface effects sub-component of the cropping-management file. Due to a crop failure in 1990, the brussels sprouts crop was replanted on June 21. Surface effect scenarios (tillage, planting, harvesting, etc.) in WEPP are limited to ten effects per year, hence, double disking effects were simulated as one only in order to comply with this limitation. Strawberry crops followed a similar seed bed preparation to that of brussels sprouts. The initial strawberry crop was planted on May 2, 1991, with 45 cm between plants along a row and 110 cm between rows (van Vliet, 1994). Planting and cultivation produced a 15-20 cm ridge for the strawberries. The crop was routinely cultivated and rotovated in between the rows for weed management. Since strawberry is a perennial crop, it was hand harvested the first time on June 5, 1992, and subsequently on June 15, 1993. To examine the effectiveness of erosion control practices on strawberries, plots 1 and 3 were seeded with a spring barley cover crop in late summer which died off and was plowed under as green manure the Page 43 Chapter 4 Data Sources Utilized following spring. Since the current version of WEPP (v. 95.7) cannot accommodate intercropping, plots 1 and 3 were simulated to have a higher planting density, with rows planted 55 cm apart. The cross-slope scenarios for Abbotsford plots 5 and 6 were simulated in WEPP with a zero percent slope along the contour. Cropping parameters for brussels sprouts were not available from the WEPP cropping management database, but the required parameters were extracted from the CPIDS database with suggested parameter values for cole crops. Adjustments to brussels sprouts crop height and rooting depth were incorporated to reflect local conditions, as suggested by Mark Sweeney, agronomist and brussels sprouts specialist with the British Columbia Ministry of Agriculture and Food (BCMAF). Cropping parameters for strawberries could not be located in WEPP's plant management database and very limited information was available from the CPIDS data base. Based on a suggestion from S.T. Chieng that strawberries are physiologically similar to eggplants, plant parameters for strawberries were substituted by those for eggplants obtained from the CPIDS program. Changes in maximum strawberry canopy height, rooting depth, critical freezing temperature, limiting maximum temperature and plant stem diameter were adjusted to reflect local conditions at Abbotsford, as suggested by Bill Peters, agronomist and strawberry specialist at BCMAF, and by Tom Baumann, horticultural specialist at Fraser Valley College, British Columbia. Other parameters specifically measured for strawberries such as the leaf area index, below ground biomass, root-to-shoot ratio, and canopy cover was based on information observed from a one year study obtained from personal communication with James Bunch, conservation agronomist at the USDA National Plant Data Center located at Southern University in Baton Rouge, Louisiana. Since the majority of the strawberry plant parameters were substituted by parameters obtained for eggplants, which is an annual crop, adjustments to the root biomass production, critical freezing temperature and limiting maximum temperature had to be incorporated to simulate conditions for a perennial such as strawberries. Detailed information on tillage, harvesting and planting dates for Beaverlodge and Abbotsford can be found in Appendices 1 and 3, respectively. Page 44 Chapter 4 Data Sources Utilized 4.3 SOIL AND SLOPE DATA Physical and chemical soil parameters such as textural characteristics (% sand, % silt, % clay and % very fine (vf) sand), organic matter (OM) and cation exchange capacity (CEQ for all plots were obtained from site measurements obtained from L.J.P. van Vliet. These properties are important when describing the different soil horizons and when determining critical soil parameters including baseline interill soil erodibility (Kj), baseline rill erodibility (Kr), baseline critical sheer stress (Tc), and baseline effective conductivity (Kh) for use with the WEPP soil input file. As suggested by Flanagan and Livingston (1995), initial saturation was assumed to be 80% for the Beaverlodge site. Since soils in the Lower Fraser Valley region are poorly drained and water-logged in the beginning of the year (January 1st), the Abbotsford soils were postulated to have an initial saturation of 90%. Soils in the Beaverlodge study are characterized by an Esher series (Dark Gray Solod) well drained clay loam soil underlain by a clayey lacustrotill parent material (Kulis and van Vliet, 1994). A detailed description of the physical, chemical and slope properties of the Beaverlodge plots is presented in Table 4.6. Table 4.6 Soil and slope properties for Beaverlodge Property Unit PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Sand (%) 16.9 18.9 26.3 26.4 27.2 25.6 Silt (%) 43.2 45.8 39.4 40.4 43.6 45.0 Clay (%) 39.9 35.3 34.3 33.2 29.2 29.4 vf sand (%) 5.6 6.7 8.1 8.0 8.3 7.7 Sat (m m"1) 80 80 80 80 80 80 OM (%) 7.1 6.8 4.7 5.5 5.0 4.7 Albedo (%) 3.5 4.0 9.2 6.6 8.1 9.2 slope (%) 11.40 11.40 11.35 11.05 10.45 10.30 CEC (meq/lOOg) 29:40 29.40 29.40 29.40 29.40 29.40 Ki (kg s m"4) 3,854,313 4,107,911 4,163,041 4,223,684 4,444,204 4,433,178 Kr (s m"1) 0.0069 0.0070 0.0070 0.0071 0.0073 0.0073 (N m-2) 3.5 3.5 3.5 3.5 3.5 3.5 Kb (mm h"1) 2.038 2.349 3.736 3.757 3.929 3.589 Soils at the Abbotsford site are characterized by Whatcom and Ryder series (Luvisolic and Orthic Humo-Ferric Podzol respectively) imperfectly drained silt loam. The parent material at this site is comprised of Page 45 Chapter 4 Data Sources Utilized eolian deposits over glacial till (Kulis and van Vliet, 1994). A detailed description of the of the soil and slope properties for Abbotsford plots is presented in Table 4.7. Table 4.7 Soil and slope properties for Abbotsford Property Unit Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Sand (%) 20.0 14.0 17.0 13.0 20.0 14.0 Silt (%) 69.0 75.0 72.0 75.0 69.0 74.0 Clay (%) 11.0 11.0 11.0 12.0 11.0 12.0 vf sand (%) 7.0 5.0 6.0 5.0 7.0 5.0 Sat (m m"1) 90 90 90 90 90 90 OM (%) 8.6 9.0 7.9 9.1 6.7 6.9 Albedo (%) 1.9 1.7 2.5 1.6 4.1 3.8 slope (%) 10.00 10.00 10.00 10.00 10.00 10.00 CEC (meq/lOOg) 22.10 21.60 22.70 24.90 19.90 21.20 (kg s m"4) 5,447,570 5,447,570 5,447,570 5,392,440 5,447,570 5,392,440 Kr (s m"1) 0.0217 0.0217 0.0217 0.0191 0.0217 0.0191 (N m"2) 3.5 3.5 3.5 3.5 3.5 3.5 Kb (mm h 1) 2.749 1.873 2.247 1.633 2.841 1.889 The parameters Kh Kr, T c and Kb were derived form Equations 3.1 - 3.4 respectively. Soil albedo was determined from the following formula as a function of organic matter (OM) (Flanagan and Livingston, 1995) Albedo = 0 ° 4 f O M [4.1] e 4.4 SOIL LOSS AND RUNOFF DATA Observed runoff and soil loss for the Beaverlodge site was collected from standard 5.2 m x 22.1 m (0.0118 ha) Wischmeier plots constructed at the Agriculture Canada research station at Beaverlodge in 1979. Yearly runoff and soil loss values for the Beaverlodge erosion plots as presented in Tables 4.8 and 4.9 respectively, and detailed event data are in Appendix 2. Both annual runoff and soil loss presented in Tables 4.8 and 4.9 are based on the calendar year (January 1 to December 31). Page 46 Chapter 4 Data Sources Utilized Table 4.8 Annual runoff data for Beaverlodge (mm yr-1) PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 1981 69.0 155.9 154.9 45.9 89.2 52.0 1982 127.6 61.1 128.3 104.4 83.6 186.5 1983 50.4 33.7 66.5 67.2 23.7 89.6 1984 93.8 1.1 95.3 2.6 3.3 14.5 1985 2.7 1.1 5.6 30.4 1.6 6.8 1986 36.9 87.0 37.5 54.0 55.1 47.9 1987 28.4 28.9 54.8 32.7 60.3 63.4 1988 0.0 0.3 10.3 2.5 13.2 12.9 1989 12.2 4.1 6.3 10.4 9.0 14.8 1990 0.0 1.5 9.0 3.5 5.5 9.0 1991 22.1 10.3 38.2 17.6 14.8 34.7 1992 34.2 61.9 90.5 42.0 93.9 . 101.3 1993 0.5 10.6 41.0 0.8 26.8 29.2 Total: 477.8 457.5 738.2 414.0 480.2 662.6 Mean: 36.8 35.2 56.8 31.8 36.9 51.0 Table 4.9 Annual soil loss data for Beaverlodge (T ha"1 yr'1) PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 1981 3.06 0.97 5.98 3.37 1.30 3.17 1982 0.41 0.69 40.01 0.57 1.33 30.27 1983 0.02 0.10 8.24 0.05 0.08 9.12 1984 0.05 0.04 0.53 0.00 0.11 1.01 1985 0.01 0.01 0.24 0.02 0.01 0.19 1986 0.05 0.70 2.60 0.05 1.31 1.00 1987 0.29 0.62 4.24 0.09 5.03 4.86 1988 0.00 0.03 0.54 0.01 0.49 1.34 1989 0.02 0.09 0.14 0.01 0.06 0.14 1990 0.00 0.01 0.03 0.00 0.01 0.02 1991 0.06 0.61 9.30 0.02 0.54 9.32 1992 0.01 3.79 21.38 0.01 6.97 20.16 1993 0.00 2.66 12.33 0.00 4.66 12.80 Total: 3.98 10.33 105.53 4.22 21.88 93.39 Mean: 0.31 0.79 8.12 0.32 1.68 7.18 Runoff and soil loss data collected for Beaverlodge presented in Tables 4.8, and 4.9, respectively, and in Appendix 2 were determined by scrutinizing collected data, based on plot notes during and shortly after runoff events (van Vliet, 1992). Some collected events had to be eliminated due to certain considerations. For example, during the 1990 growing season, a small rainfall event produced runoff with a maximum soil loss less than 100 kg ha"1 for fallow plots, but this event was eliminated (1 out of 20 events), because it was in the designated break-in period used to reduce residual effects between experiments (van Vliet, 1992). A total of 87 measured events in 4748 days were used for the WEPP simulation of the Beaverlodge data. Table 4.12 shows a yearly breakdown of the number of events. Page 47 Chapter 4 Data Sources Utilized Annual runoff and soil loss data obtained from Abbotsford are presented in Tables 4.10 and 4.11, respectively, and by event in Appendix 4. The original experiment at Matsqui 1 consisted of ten plots with three treatments on two slopes (van Vliet and Hall, 1995). Since plots for Matsqui 1 and Matsqui 2 were combined to produce the Abbotsford management, soil and slope file, data from four plots (plots 3, 4, 7 and 8 in van Vliet and Hall, 1995) located on a slope with a five percent gradient were eliminated to maintain consistency. Data obtained from the two plots in the uncultivated field (plots 9 and 10 in van Vliet and Hall, 1995) were also eliminated as well as data obtained from plot 6 (van Vliet and Hall, 1995) due to excessive seepage which produced higher than anticipated runoff and soil loss. Data from a total seven plots were eliminated from Matsqui 1. Runoff and soil loss data obtained for plots 1, 2 and 5 (in van Vliet and Hall, 1995) were used for data analysis. From the original study conducted by Van Vliet and Hall (1995), data obtained from plots 1, 2 and 5 were replicated twice to produce the data presented in Table 4.10 and 4.11 for plots 1 & 2, 3 & 4 and 5 & 6 respectively. Runoff and soil loss data for 51 events were recorded for the Matsqui 1 site. A total of 19 events (33 %) had to be eliminated due to erroneous results and malfunctioning of equipment. Sixteen of 96 events were missing due to equipment malfunction, and were substituted for using replicated plots in the original experiment. Accumulated sediment deposited in the flume on September 19, 1990 and on April 30, 1991 were distributed (weighed) over previous events. Eight plots were originally used in the Matsqui 2 experiment. Runoff and soil loss data from 42 events were collected, but eight (19%) had to be eliminated due to malfunctioning of equipment. Data from the sub-surface drained plots, 3 and 5 (in van Vliet, 1994) was eliminated due to improper installation producing higher than anticipated runoff. Other events were eliminated due to null values obtained for replicated plots. A total of 64 runoff events were measured for the Abbotsford erosion plots in a period of 1583 days Table 4.12). The original plots used by van Vliet (1994) labeled plots 1, 2, 4, 6, 7 and 8 were re-labeled and presented as plots 1, 2, 3, 4, 5 and 6 respectively for this simulation study. Table 4.10 Annual runoff data for Abbotsford (mm yr"1) Page 48 Chapter 4 Data Sources Utilized Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 1989 212.1 212.1 132.7 132.7 172.9 172.9 1990 380.5 380.5 322.3 322.3 430.9 430.9 1991 287.1 271.5 197.3 216.3 106.4 107.3 1992 129.9 249.0 126.3 236.9 50.3 36.2 1993 108.3 197.1 88.0 193.7 23.2 22.3 1994 61.7 71.2 51.5 72.6 17.9 9.9 Total- 1179.6 1381.4 918.1 1174.5 801.6 779.5 Mean: 196.6 230.2 153.0 195.8 133.6 129.9 Table 4.11 Annual soil loss data for Abbotsford (T ha' 1 yr"1) . Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 1989 0.50 0.50 0.30 0.30 0.15 0.15 1990 0.47 0.47 0.44 0.44 0.28 0.28 1991 0.84 2.21 0.86 2.37 0.28 0.26 1992 1.67 12.81 1.62 11.91 0.39 0.36 1993 1.43 3.41 1.54 3.17 0.09 0.11 1994 0.73 1.87 0.48 1.89 0.04 0.05 Total: 5.63 21.27 5.24 20.08 1.23 1.21 Mean: 0.9 3.5 0.9 3.3 0.2 0.2 The number of measured events for both locations are summarized below in Table 4.12. Table 4.12 Summary of measured events Beaverlodge Abbotsford Jan. 1, 1981 - Dec. 31, 1993 Nov. 1, 1989 - Mar. 3, 1994 year 4748 days 1583 days 1981 8 -1982 8 -1983 10 1984 6 -1985 8 -1986 8 -1987 5 -1988 4 -1989 4 8 1990 1 15 1991 9 131 1992 10 14 1993 6 11 1994 - 3 Total: 87 64 7 events measured at Matsqui I and 6 events measured at Matsqui 2 Page 49 Chapter 5 Sensitivity and Calibration C H A P T E R 5: S E N S I T I V I T Y A N D C A L I B R A T I O N This section details the procedures and results obtained from the sensitivity analysis used to assess model response to input parameters and parameter calibration used to obtain an optimal agreement between predicted and measured runoff and soil loss. 5.1 SENSITIVITY ANALYSIS Prior to calibration, a sensitivity analysis was performed to determine model response to changes in specific parameters. Sensitivity of a model is defined as the ratio of the relative changes of model response to the relative change of model input parameters. A sensitivity analysis ranks the model parameters based on their contribution to overall model predictions (Tiscareno-Lopez et al., 1994). The sensitivity parameter (S) is given by (Nearing et al., 1989): output j - output 2 output— var iable, - var table 2 var iable— where variablet and variable2 are the least and greatest values of input used, respectively, and variable - is the average of variablet and variable^ outputt and output2 are the associated outputs for the input variables and output— is the average of the two outputs (Nearing et al., 1989). The sensitivity of a parameter is reflected by and is proportional to the magnitude of the sensitivity parameter. For example, a high absolute sensitivity value indicates a greater model response to changes in a parameter. A sensitivity analysis was performed for the Beaverlodge fallow plots 2 and 5 over 13 years, since the fallow plots are expected to yield the greatest variability of runoff and soil loss. Because this study deals Page 50 Chapter 5 Sensitivity and Calibration primarily with an assessment of soil loss and runoff, parameters which are known to be highly and have a dominating impact on runoff and soil loss were chosen for this analysis. Sensitivity was performed by altering the soil parameters K„ Kr, xa Kh, Albedo and Sat individually, while assigning constant values for parameters not being tested, and recording the total runoff and soil loss for 13 years resulting from changes in these parameters (see Table 5.1). Since measured data was not available for Kh Kr, T0 Kh, Albedo and Sat, values for these parameters had to be determined by applying equations as suggested by Flanagan and Livingston (1995). Because of the resulting uncertainties, a sensitivity analysis was performed to rank the parameters in terms of their response to change which is beneficial when calibrating the model since changes to parameters can be made according to their sensitivity. Table 5.1 Sensitivity analysis (based on Beaverlodge fallow plots) Parameter Value Constant runoff soil loss Srunoff ^soil loss Value Kt 500000 5000000 704.2 71.4 0 0.1516 10000000 704.2 94.1 0.0003 704.2 30.1 Kr 0.3000 0.0035 704.2 437.3 0 0.8729 0.1 704.2 392.0 10.0 4.0 704.2 15.1 0 -0.9445 0.1 2287.7 220.5 Kb 10.0 4.0 439.6 37.4 -0.6913 -0.7243 2% 701.3 83.8 Albedo 75% 12% 746.8 79.3 0.0331 -0.0291 0% 704.2 83.6 Sat 100% 54% 708.9 83.6 0.0033 0 Soil parameters to which runoff is most sensitive, ranked in order from most to least sensitive are: Kb, Albedo and Sat. The baseline conductivity (Kb) had a large influence in determining runoff predictions. Since this parameter is related to soil hydraulic conductivity (Flanagan and Livingston, 1995), it greatly influences infiltration rates. WEPP internally adjusts Kb in continuous simulation mode to reflect changes caused by management practices (Flanagan and Livingston, 1995). The surface reflectance (Albedo) and Page 51 Chapter 5 Sensitivity and Calibration initial saturation (Sat) had a near negligible impact. Albedo influences the evapotranspiration routine and the initial saturation influences the soil water content in the soil profile for the beginning of the simulation period. Subsequently, they both influence the water balance subroutine, but this is useful only when predicting small events. Since large events dominate long-term runoff predictions (Flanagan and Livingston, 1995), soil saturation and subsequent saturated hydraulic conductivity is reached in a relatively short period and excess rainfall (that which is greater than the soil saturated hydraulic conductivity) instantly translates into runoff. In a detailed sensitivity analysis conducted by Nearing et al., 1989, it was found that sensitivity to infiltration parameters was greater for low than for high intensity storms, since a greater relative proportion of the rainfall from high intensity storms becomes runoff. The same study ranked sensitive precipitation parameters which influenced runoff in order from most to least as follows: precipitation amount (prep), duration (dur), intensity (ip) and time to peak (tp). In a WEPP rangeland sensitivity study conducted by Tiscareno-Lopez et al. (1994), the three major sources of error in predicting runoff volume and peak runoff originated from: (1) errors in rainfall characteristics (prep, dur, tp, ip), (2) errors in estimating saturated hydraulic conductivity and (3) errors in estimating the watersheds antecedent moisture conditions (for single storm simulations). The critical sheer stress (Tc) and erodibility parameters (Kt and Kr) had no impact on runoff calculations since these parameters exclusively influence soil detachment. Sensitivity of soil loss to soil parameters ranked in order from most to least sensitive are: rc, Kr, Kb, Kit Albedo and Sat. Sheer stress (TC ), rill erodibility (Kr) and the baseline conductivity (Kb ) had a large influence in determining soil loss predictions, while the interrill erodibility parameter (Kt) had a moderate impact on soil loss. Albedo and initial saturation (Sat) had little or no impact on soil detachment. Similar findings were presented in Nearing et al. (1989). In addition soil texture, canopy and rill cover, slope length and gradient parameters were also found to be important to model predictions by Nearing et al. (1995). These latter parameters were not used in the sensitivity analysis since these parameters were based on actual measured data from the plot sites, and were more confident in their values than had they been estimated from the equations. Page 52 Chapter 5 Sensitivity and Calibration The Beaverlodge study examined specific scenarios with distinctive soil, management and climate characteristics, therefore an independent sensitivity analysis was performed to evaluate model response to parameter changes. Although it is simple to calculate sensitivity and the results can be easily assessed, it is important to realize that interactions between parameters is not assessed. However, it provides a useful guide towards determining the relative effects of parameters on output and when calibrating the model by focusing on these parameters that result in the greatest impact on runoff and soil loss predictions. 5.2 CALIBRATION Calibration involved modification of critical parameters determined from the sensitivity analysis. The parameters modified were selected based on their sensitivity values. The parameters modified include: sheer stress (TC ), rill erodibility (Kr), baseline conductivity (Kb) and interrill erodibility (Kt). Calibration was achieved by trying to obtaining a zero percent deviation given by _ , . . predicted value - measured value ,„„ % deviation = x 100 [5.2] measured value between measured and predicted values of total soil loss and runoff for the entire simulation period. Calibration was performed on total soil loss and runoff for the entire simulation period and not on individual events or single years because the variabilities between measurements and model predictions from multiple soil, slope, management and climatic factors invoke variable predictions which can involve a greater number of calibration runs. In addition, multiple calibrations are often impractical since it is very time consuming (i.e. it took 1200 model runs to calibrate the Beaverlodge plots) and if the transfer of this technology is to be effective, it has to be simple to use with minimal modifications to obtain confident results. Page 53 Chapter 5 Sensitivity and Calibration Parameters were modified in accordance with their effects on soil loss and runoff predetermined from the sensitivity analysis. Based on their influences (Table 5.2), manual modification was conducted in order to achieve a minimal deviation between measured and predicted soil loss and runoff. Table 5.2 Calibration criteria parameter symbol unit primary influence on baseline conductivity Kb mm hr 1 runoff, soil loss sheer stress Nm"2 soil loss rill erodibility Kr s m"2 soil loss interill erodibility Ki kg s m" soil loss Calibration for this study was conducted in a four step process. First, runoff predictions were manually matched for each plot by modifying the baseline conductivity (Kb); when minimal deviation between measured and predicted runoff was established, soil loss predictions were matched for each plot by manually modifying sheer stress (Tc), rill erodibility (Kr) and interrill erodibility (/Q parameters. Soil loss parameters were modified by altering individual parameters while maintaining the other calibration parameters at their predetermined values (for calculated values, please see Tables 4.6 and 4.7 for Beaverlodge and Abbotsford, respectively). For example, once a calibrated Kh is established for a plot, the model is then modified to calibrate T c, while maintaining values for Kr and Kt determined by applying Equations 3.2 and 3.1, respectively, and assigning the newly calibrated value to Kb. When a calibrated value for rc is established, the plot is then calibrated for Kr by assigning xc and Kt at their calculated values, and again assigning Kb to the new calibrated value. The plot is then calibrated for Kr, while following a similar procedure for maintaining other parameter values. A summary of the process is in Table 5.3. Table 5.3 Calibration procedure procedure modified altered constant parameter parameter parameter1 1. modify Kb Kb none Tc, Kn Kj 2. modify rc ?c Kb Kn Kj 3. modify Kr Kr Kb Tc Ki 4. modify Ki ^—;—: K Kb Tc, Kr refers to parameters calculated by Equations 3.1-3.4 Page 54 Chapter 5 Sensitivity and Calibration Calibration for Beaverlodge and Abbotsford was conducted to obtain a minimal divergence between measured and measured soil loss and runoff. 5.2.1 BEAVERLODGE CALIBRATION Calibration for Beaverlodge was accomplished by modifying four parameters for six plots in approximately 1200 runs. Predicted runoff and soil loss data was compared with accumulated soil loss and runoff collected from January 1, 1981 to December 31, 1993 to represent 13 years of data. Calibrated parameters for Beaverlodge are summarized in Table 5.4. Table 5.4 Calibrated soil parameters for Beaverlodge Parameter PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 calibrated Kb 2.054 2.454 3.705 2.110 4.238 3.745 calibrated xc 6.042 3.040 4.683 2.312 2.999 3.921 calibrated Kr 0.0003 0.0157 0.0035 0.2425 0.0124 0.0055 calibrated Kj 5,000' 1 7,025,000 5,000' 5,000' 10,950,000 5,000' calibration of these parameters did not converge to minimal deviation (approx. 0%) Calibration of parameters for the Beaverlodge plots returned a minimal deviation (approx. 0%) in most cases with the exception of those noted with a superscript in Table 5.4. Calibration of Kb, resulted in a minimal deviation of runoff only, while subsequent calibrations resulted in a minimal deviation of runoff and soil loss since the calibrated Kb values replaced the original calculated Kb, values for the subsequent xc, Kr and K, calibrations. For the first calibration, Kb, the average difference between calculated and calibrated parameters was -4.5%, and the average difference between calibrated and predicted xc, Krand Kt was 9.5%, 558% and -30% respectively. The difference in calibrated ^parameters ranged from -1% for plot 2 to 8% for plot 4 with the exception of plot 3 which had to be changed from a Kb of 3.737 mm hr 1 to 2.110 mm hr "' (-44% difference). Small adjustments to the Kh parameters were required to get a minimal deviation between measured and predicted cumulative runoff. Page 55 Chapter 5 Sensitivity and Calibration Variations in the sheer stress parameter (Tc ) ranged from -34% for plot 3 to 73% for plot 0. Since all the TC values were originally set to 3.5 (see Table 4.6) as suggested by Flanagan and Livingston (1995), no adjustments were made to account for the effects of soil physical and chemical parameters on sheer stress. The differences ranged from 2.312 N m"2 to 6.042 N m"2 for plots 3 and 0 respectively. Since soil properties for the plots are similar in nature, these variations could suggest that the erosion component in the WEPP model which utilizes this parameter may not be effective. The sensitivities of the erodibility parameters Kr and K{ (Table 5.1) suggests that the rill erodibility (Kr) has a significant impact on soil loss predictions since it's sensitivity parameter is comparable in magnitude to that of Kb and rc. The variability of the optimized Kr parameter ranged from 0.0003 s m"1 to 0.2425 to account for a variation of -96% for plot 0 to 3323% for plot 3 respectively. The high mean variation of 558% for the optimized Kr parameters could be accounted for the high degree of variability for plot 3. Optimization of the interrill erodibility parameter, K for plots 0, 2, 3 and 5 did not converge to produce a minimal deviation (approx. 0%) in soil loss. Results for plots 1 and 4 did converge by altering the Kj parameter by 71% and 146% respectively. Since the sensitivity analysis indicated that soil loss is not sensitive to Kt the lack of convergence is justified. Overall, adjustments had to made to made to the baseline conductivity (Kb), sheer stress (rc) and rill erodibility (Kr) parameters in order to achieve a minimal deviation between measured and predicted runoff and soil loss. Changes in the interrill erodibility parameter (K) did not result in significant changes. This complies with the results obtained from the sensitivity analysis. 5.3.2 ABBOTSFORD CALIBRATION Calibration for Abbotsford was accomplished by modifying the same four parameters (Kb, Tc, Kr and Kj) for six plots in approximately 600 runs. Predicted runoff and soil loss data were compared with accumulated soil loss and runoff collected from November 1, 1989 to February 28, 1994 to represent over four years of data. Calibrated parameters for Abbotsford are summarized in Table 5.5 Table 5.5 Calibrated soil parameters for Abbotsford Page 56 Chapter 5 Sensitivity and Calibration Parameter Plot 1 Plot2 Plot 3 Plot 4 Plot 5 Plot 6 calibrated Kb 1.350 1.006 2.044 1.291 0.11 0.11 calibrated T c 81 7.725 81 6.817 3.51 3.51 calibrated Kr 0.000011 0.000011 0.000011 0.0002 0.000011 0.00001 calibrated Kt 50001 1 ,-,—T-50001 50001 50001 50001 50001 calibration of these parameters did not converge to minimal deviation (approx. 0%) Calibration of the Abbotsford data was not as successful as that for the Beaverlodge data. Out of all 24 individually calibrated parameters for the six plots, only 6 parameters converged to produce a minimal deviation of 0%. All other parameters could not obtain minimal deviation between measured and predicted runoff and soil loss. Calibration of the Kb parameter was successful for plots 1-4 only. The average parameter had to be decreased by an average of 32% to match measured and predicted runoff. When the original unmodified parameters were inputted for the initial simulation, WEPP under-predicted runoff by a range of 8% for plot 3 to 82% for plot 5, with a mean under-prediction of 46%. This suggests that the WEPP method for estimating the effective baseline conductivity (Equation 3.4) may not be effective since the higher calculated Kb resulted in quicker infiltration rates, and hence, lower runoff values were obtained. Calibrated runoff for the cross slope plots, plots 5 and 6 did not converge to measured values but instead, the Kb was kept constant at 0.1 mm hr'1 for subsequent calibrations. Calibration of the sheer stress parameter, xc was successful for the control plots only. To obtain minimal deviation between total soil loss, rc for plots 2 and 4 had to be increased by 121 and 95% respectively. Subsequent calibrations of Kr and Kt were not successful with the exception of the calibrated Kr for plot 4. Interesting to note that calibrations for all parameters for the cross-slope plots did not obtain any significant minimal deviation from measured totals. It was found experimentally that WEPP responds to parameters within a certain range, and when set outside these limits, the model does not respond to or responds negligibly. Hence calibrated parameters were limited to a range to reflect the best variations in model response. Using a Kb of 0.1 for plots 5 and 6 obtained a minimal runoff deviation of -20 and -17% respectively. Modifying soil loss parameters, rc, and Kr for plots 5 and 6 did not produce any soil loss. Page 57 Chapter 5 Sensitivity and Calibration This suggests that WEPP responds well to the effects of ridging and contouring, however, the model does not respond to sediment loss that occurred from breakdown and transport of contoured ridges. The effects of the calibrated parameters on soil loss and runoff will be discussed in detail in the following chapter. Page 58 Chapter 6 Model Testing and Analysis C H A P T E R 6: M O D E L T E S T I N G A N D A N A L Y S I S This section details the procedures and results used for model testing. For both locations, the original calculated and optimized parameters were used for data analysis. Data was analyzed on a yearly and event basis. For Beaverlodge, the period from January, 1981, to December 31, 1993, was used in the comparison and for Abbotsford, the period from November 1, 1989, to March 2, 1994, was used primarily due to the availability of measured data for the stated periods. 6.1 D A T A ASSESSMENT PROCEDURE Data analysis incorporated several methods of assessment. The objective was to compare the original and calibrated predictions of soil loss and runoff with measured soil loss and runoff. This was done on annual and event basis for both locations. The yearly comparisons for Beaverlodge are summarized in Tables 6.1-6.5, and for Abbotsford in Tables 6.13-6.16. The variables sum and average and % dev. refer to the sums, the average annual values and the percent deviation (Equation 5.2) of runoff and soil loss from measured values of soil loss and runoff, respectively. Initial assessment of the annual calculated and calibrated results was made using a student's f-test for annual measured and predicted runoff and soil loss. Assuming a normal distribution of measured and predicted values, a f-test was conducted at the 0.10 significance level. The results are tabulated in Tables 6.1-6.5 for Beaverlodge and in Tables 6.13-6.17 for Abbotsford. The results of the f-test indicate the variations in the means. For example, if the measured mean (u,) is greater than the predicted mean (u2), we would reject H 0 : Uj = u2 in favor of Hi: fij > u2 if the test statistic T>= t(a; n-1) where: Page 59 Chapter 6 Model Testing and Analysis [6.1] where: w = mean of differences Sw - standard deviation of the differences n = sample size (number of years in simulation) a = significance level (0.10 and 0.05 used) and t(a; n-1) refers to the value of the upper percentage points of the student's f-Distribution (Hogg and Ledolter, 1987). For the Beaverlodge data, t(a; n-1) was set at 1.356 and for Abbotsford, t(a; n-1) was set at 1.476 for the 0.10 level of significance. To test the model efficiency, the Nash-Sutcliffe coefficient of determination (Nash and Sutcliffe, 1970) was applied. This is the recommended method to objectively assess performance of continuous simulation watershed models (ASCE Task Committee on definition of Criteria for Evaluation of Watershed Models, 1993). The term efficiency describes the association between measured and estimated values. The coefficient of efficiency (R2) is calculated as follows: R2 = 1-if [6.2] t,(Qmi-Q^)2 i=l where: Qmi - measured value of event, i Qci =• computed value of event, i Qm - mean of measured values It is analogous to the coefficient of determination, but not identical (Aitken, 1973). The term n n Qm ~ Qm )2 represents the initial or measured variation and the term ^ ( Qmi - Qci f represents the i=l i=l unexplained or residual variation; however, the residual variation is calculated using the actual observation values rather than the values from the best regression line between measured and predicted values (Risse et Page 60 Chapter 6 Model Testing and Analysis al., 1995). This is an important consideration since the model is comparing the predictions on a 1:1 line rather than the best-fit regression line. The Nash-Sutcliffe coefficient of efficiency, or the R2 efficiency coefficient is interpreted as follows: a value of 1 indicates a perfect fit; a value greater than zero but less than one indicates adequate model performance with accuracy increasing with increase of R2; a value of zero indicates that using the model results are no better than using the average measured value; and a value less than zero indicates that the model results are worse than the average measured value. In studies investigating the behavior of WEPP and other erosion prediction models by Kramer and Alberts (1992), Risse et al. (1993 (I)), Risse et al. (1993 (II)), Risse et al. (1994), Risse et al. (1995), and Savabi et al. (1995) all applied the R2 coefficient to determine model efficiency. As a final estimator, the coefficient of determination (r2) calculated from a regression analysis was applied since it is a good measure of the degree of association between measured and estimated values (Aitken, 1973). The r2 expresses the proportion of the total variation of estimated values which can be accounted for or explained by a linear relationship with the measured values (Wampole and Myers, 1989). For this study, the efficiency estimator, R2 was applied to all calculated and calibrated runoff and soil loss predictions on an event and yearly basis. The years correspond to the calendar years, and the measured events correspond to the sampling date. A sampling date may include accumulations of one or more intermediate events since the last sampling date. Therefore, for a particular sampling date, all events predicted from the day following the previous sampling date to the sampling date in question were accumulated and recorded for that sampling date. For example, if sampling occurred on March 3 and the previous sampling date was on February 17, all events calculated from February 18 to March 3 are summed up into March 3. Since sampling could not be conducted for each individual event, by using a sampling date index, all events can be accounted for. The cumulative frequency distribution was examined graphically to determine the distribution of all events for all plots for a particular simulation. From it, we can see if the model reflects the frequency distributions of the measured and predicted events. Page 61 Chapter 6 Model Testing and Analysis 6.2 BEAVERLODGE ANALYSIS Analysis for the Beaverlodge plots was conducted on a yearly basis to evaluate general model performance, on an event basis for a more critical examination of soil loss and runoff predictions, and the model was tested for it's winter/summer events in order to get a better understanding of the winter hydrology component of WEPP. The results of the original data and calibration studies for Beaverlodge plots is presented below. Appendix 2 details the runoff and soil loss measurements and predictions for Beaverlodge. 6.2.1 Y E A R L Y ANALYSIS Statistical results obtained from the original data and calibration analysis on a yearly basis for all the Beaverlodge erosion plots are summarized in Tables 6.1-6.4. Table 6.1 Statistical summary of soil loss (SL) (T ha"1 yr"1) and runoff (RO) (mm yr"1) for Beaverlodge-Original uncalibrated data, based on yearly data PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 SL RO SL RO SL RO SL RO SL RO SL RO Sum: 11.7 479.1 7.6 469.3 179.3 734.3 0.7 239.9 16.1 501.4 117.6 677.7 Average: 0.9 36.9 0.6 36.1 13.8 56.5 0.1 18.5 1.2 38.6 9.0 52.1 % dev. 194% 0% -26% 3% 70% -1% -83% -42% -26% 4% 26% 2% t(0.10,12) P1=H2 Hl<ll2 H1=P.2 Mr=^2 Pl>P-2 P,=P2 lll=P2 ^=V-2 P1=H2 R2 -8.58 -1.07 -0.55 -1.27 -0.76 -1.04 -0.14 -0.13 0.26 -1.32 -1.00 -0.01 r2 0.2% 0.1% 1.7% 1.9% 35.2% 0.6% 3.2% 23.0% 34.4% 0.1% 19.1% 25.1% Total runoff predictions for the plots without using any calibrated data agreed with measured values, with an overall mean deviation of -5.6% for all plots. The Fescue plots, plots 0 and 3 showed the largest variation in runoff, since the deviations varied from 0 to -42% respectively. Soil loss predictions varied in deviation from -83 to 194% for plots 3 and 0 respectively, with an overall mean soil loss deviation of 25.8% for all plots. This suggests that the model is generally over-predicting soil loss, which is certainly true for the case of the fallow plots 2 and 5 and for the fescue plot 0. Page 62 Chapter 6 Model Testing and Analysis The Mest values were also in general agreement between annual measured and predicted soil loss and runoff. For most cases, the means at the a = 0.10 significance did not differ. Total runoff and soil loss for all Beaverlodge plots can be viewed in Figure 6.1. Figure 6.1 Comparison of measured (M) and predicted (P) total runoff (mm) and soil loss (t ha-1) at Beaverlodge using W E P P calculated Kb , TC, K r and K i Total soil loss predictions for the replicated fescue plots 0 and 3 varied from 11.7 t ha"1 for plot 0 to only 0.7 t ha"1 for plot 3. This variation can be explained because unlike plot 3, plot 0 was not in continuous fescue for the entire simulation period since plot 0 was planted with barley during 1985-1989 (Table 4.4). Therefore the two fescue plots were not identically replicated. Interestingly though, the model predicted runoff better for plot 0 than for plot 3. High soil losses of 2.6 and 8.3 t ha"1 yr"1 were calculated for 1986 and 1987 respectively accounting for 93% of the total soil loss for the entire simulation period for plot 0. A comparison of annual measured and predicted runoff and soil loss for all 6 plots can be viewed in Figures 6.2 and 6.3 respectively. Page 63 Chapter 6 Model Testing and Analysis Plot 0 Plot 1 81 82 83 84 85 86 87 88 89 90 91 92 93 81 82 83 84 85 86 87 88 89 90 91 92 93 Plot 2 Plot 3 81 82 83 84 85 86 87 88 89 90 91 92 93 81 82 83 84 85 86 87 88 89 90 91 92 93 Plot 4 Plot 5 81 82 83 84 85 86 87 88 89 90 91 92 93 81 82 83 84 85 86 87 88 89 90 91 92 93 Figure 6.2 Comparison of measured (M) and predicted (P) annual runoff (mm yr"1) at Beaverlodge using W E P P calculated Kb, xc, Kr and Kt Annual runoff comparisons using the unmodified data demonstrate poor agreement between predicted and measured runoff, despite similarities in total runoff. Negative runoff efficiency parameters (R2) for all the plots indicate poor model efficiency in predicting runoff. Furthermore poor coefficients of determination (r2) for all plots indicate poor correlation between annual measured and predicted runoff (Table 6.1). In 1981, the second highest average annual runoff for all plots (94.4 mm) was measured, however, that year ranked 13th, or last, in average annual simulated runoff. The lowest average annual measured runoff of 4.8 mm for all plots was recorded in 1990, but it ranked second in simulated runoff (117.9 mm). In 1990, only one runoff event was recorded on March 2, but a record high rainfall of 110 mm was recorded on June 6 but did not result in measured soil loss and runoff. Poor correlation (r2) and poor efficiency (R2) between annual measured and predicted soil loss and runoff indicate little association between data sets. An improvement of the predictions is sought by calibrating parameters influencing runoff. Annual soil loss comparisons can be viewed in Figure 6.3. Page 64 Chapter 6 Model Testing and Analysis PlotO "a 10 JI MJ> Mf> Mf> JL MP Mf MJ^ M,P Mp M.P 81 82 83 84 85 86 87 Plot 2 89 90 91 92 93 • 50 Mp | p , M ^ l l l r f l M P MP M f i ^ f l M , M , M A r Y l ^ n 81 82 83 84 85 86 87 Plot 4 89 90 91 92 93 1 / 5 81 82 83 84 85 86 87 88 89 90 91 92 93 Plotl 6 0 81 82 83 84 85 86 87 88 89 90 91 92 93 Plot 3 "a 4 F 3 0 M 81 82 83 84 85 p flP M.P Mf Mf Mj-, Mp My Mp MP MJI, Mp M£. 87 88 89 90 91 92 93 Plot 5 81 82 83 84 85 86 87 89 90 91 92 93 Figure 6.3 Comparison of measured (M) and predicted (P) total soil loss (t ha'1 yr'1) at Beaverlodge using WEPP calculated Kb, r„ Kr and Kj Soil loss comparisons for individual plots (Figure 6.3) also indicate poor agreement between annual measured and predicted soil loss, which was found earlier for total soil loss and can be further verified by the negative efficiency and low correlation coefficients. By observation, it appears that plots 2, 4 and 5 have systematic errors due to the similarities of proportions in measured and predicted soil loss. However, poor efficiency coefficients quantitatively indicate poor predictions, with the exception of plot 4, since it has a positive efficiency coefficient of 0.26. This suggests that the annual soil loss predictions of plot 4 are a better estimator than the measured average annual soil loss. In 1982, annual predicted soil loss for the fallow plots 2 and 5 were 38.8 and 15.2 t ha"1 yr"1, respectively, despite having similar treatment. Since plot 2 was plowed under earlier and had an extra year of fallow in 1980, prior to the simulation, the residual effects of the buried biomass were not sufficient enough to prevent excessive soil loss in 1982. The relative similarities in runoff between plots 2 and 5 in 1982 suggests that the model takes into consideration the effects of buried residue from previous harvests. Although the model was not simulated in 1980 for long term purposes, the initial condition file was based on management practices two years prior to 1981, hence the residue decay function calculated less below Page 65 Chapter 6 Model Testing and Analysis ground biomass for plot 2 which can prevent soil loss by the addition of soil binding organic matter particles, accounting for the high soil loss predicted in 1982. Plots 1 and 4 have almost identical cropping and management scenarios, (Table 4.4), but between 1985 and 1989, plot 1 was cross-slope cultivated. A zero percent contour gradient was used in the contouring sub-section of the cropping-management file to simulate this. Due to the zero percent contouring gradient, rill detachment does not exist since water in rills cannot flow between two points of equal gradients. However, some interrill detachment and runoff from the bottom most ridges can contribute to interrill soil loss and runoff. Contouring scenarios modeled in WEPP can yield reduced soil loss and runoff predictions. To obtain similarities in runoff predictions, all the plots were calibrated to obtain a minimal difference between total runoff by altering the baseline conductivity, Kb since runoff was determined to be most sensitive to changes in this parameter. Statistical results from calibrating Kb can be viewed in Table 6.2. Table 6.2 Statistical summary of predicted soil loss (SL) (T ha' 1 yr"1) and runoff (RO) (mm yr _ 1) for Beaverlodge - Calibrated Kb, based on yearly data PlotO '. Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 SL RO SL RO SL RO SL RO SL RO SL RO Sum: 11.7 478.0 7.5 457.1 180.2 738.2 1.6 413.9 14.9 480.2 115.0 662.7 Average: 0.9 36.8 0.6 35.2 13.9 56.8 0.1 31.8 1.1 36.9 8.8 51.0 %dev. 194% 0% -27% 0% 71% 0% -62% 0% -32% 0% 23% 0% t(0.10,12) 111=112 U =P2 1*1=1*2 1*1=1*2 i*1<\*2 1*1=1*2 1*1=1*2 U1=P2 \*1=l*2 1*1=1*2 1*1=1*2 Pi=Pi R2 -8.58 -1.07 -0.51 -1.26 -0.77 -1.04 -0.12 -1.18 0.24 -1.27 -1.00 0.00 r2 0.2% 0.1% 1.9% 1.8% 35.4% 0.6% 0.6% 3.6% 33.6% 0.2% 18.1% 24.9% The calibration of Kb achieved a 0% difference in runoff for all the plots (Table 6.2). Due to good agreement for total runoff without calibration, calibrating improved average runoff deviations for all plots by only 5.5%. In some cases (i.e. Plots 1 and 3) it improved the efficiency, but overall efficiency values were still less than zero, indicating poor model performance. Optimizing the runoff worsened soil loss predictions by an average of 2%. For plots 0 and 2 the R2 coefficient remained the same at -1.07 and -1.04 respectively which was the result of slight adjustments made to the Kb value that had a negligible impact on runoff calculations, and hence did not improve model efficiency. Adjustments to the Kb parameter Page 66 Chapter 6 Model Testing and Analysis improved overall or total runoff, however, the small adjustments did not yield significant improvements to R2 and r2. On average, Kb decreased by 0.18 mm hr"1 from the original average of 3.23 mm hr"1 for all the plots. Decreasing the overall Kb resulted in less water infiltration and therefore in greater runoff and soil loss. Adjustments in the Kb increased overall soil loss deviation for all plots by only -2.1% from the original average deviation of 25.6% for all plots. Because of the slight improvement made to optimize the Kb, the modified annual runoff and soil loss values did not produce noticeable changes with the exception of the runoff efficiency for plot 5 and soil loss efficiency for plot 4. The slight improvement in runoff efficiency for plot 5 (fl2 = -0.01 before calibration, and 0.00 after calibration) indicates that calibrated runoff predictions for fallow plots are till no better than the mean annual measured soil loss. The positive soil loss efficiency coefficient achieved for plot 4 (R2 - 0.26) indicates that annual soil loss predictions are acceptable for row crops. Plot 4's coefficient of determination (r2) of 34.4% translates into a correlation coefficient (r) of 58% between measured and predicted annual soil loss. The implications of the optimised soil loss and runoff parameters on individual events will be discussed later. In order to improve soil loss predictions, the critical sheer stress parameter, T c was optimized to achieve a minimal difference between measured and predicted total soil loss. The results are summarized in Table 6.3. Table 6.3 Statistical summary of predicted soil loss (SL) (T ha' 1 yr'1) and runoff (RO) (mm yr' 1) for Beaverlodge - Calibrated Kb and tc, based on yearly data PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 SL RO SL RO SL RO SL RO SL RO SL RO Sum: 4.0 478.0 10.3 457.1 105.7 738.2 4.2 413.9 21.9 480.2 93.4 662.7 Average: 0.3 36.8 0.8 35.2 8.1 56.8 0.3 31.8 1.7 36.9 7.2 51.0 % dev. 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% t(0.10,12) U1=P.2 1*1=1*2 t*i=l*2 t*1=P2 \*1=\*2 U1=[12 1*1=1*2 1*1=1*2 f*1=H2 U1=U2 U1=P2 Pl=\*2 R2 -0.67 -1.07 -1.50 -1.26 -0.10 -1.04 -0.66 -1.18 0.20 -1.27 -0.76 0.00 r2 0.9% 0.1% 0.2% 1.8% 21.8% 0.6% 1.8% 3.6% 35.0% 0.2% 13.5% 24.9% Page 67 Chapter 6 Model Testing and Analysis Calibration of xc was successful since a minimal deviation of 0% between total measured and predicted soil loss was achieved for all plots (Table 6.3). Overall average soil loss prediction was improved by 26% for all plots. The rc is independent of infiltration and changes in Tc parameter did not effect runoff rates. Changes in total runoff and soil loss with calibrated Kb and TC for all plots are shown in Table 6.4. 140 800 J -600 o e D ei •a 4 0 0 cd O H 200 0 M P M P M P M P M P M p 120 J ioo o 80 I—J 5 60 £ 40 20 0 M P i—u—1 M P nn M P M P 1—IJ—1 M P M p 0 1 > 3 Plot 4 5 0 l > 3 Plot 4 5 Figure 6.4 Comparison of measured (M) and predicted (P) total runoff (mm) and soil loss (t ha"1) at Beaverlodge using predicted Kb and TC and W E P P calculated Kr and Kt Calibrating the TC improved total soil loss predictions. It significantly improved the R2 of plot 1, (in comparison to predictions observed with optimized Kb) and moderately improved the R2 coefficients of plots 3 and 4. A reduction in R2 was observed with plots 1, 3 and 4 with calibrated xc. Calibrating TC resulted in proportionate increases or reductions in annual soil loss which are shown in Figure 6.5. Page 68 Chapter 6 Model Testing and Analysis PlotO •SI ^ 3 0 6 0 81 82 83 84 85 86 87 88 89 90 91 92 93 IP [~IP H P H P H P Hf. HP MTI H p H P "A •a <§5 Plotl H P H P Fir rv H P H P MTI M -IP Plot 2 81 82 83 84 85 86 87 Plot 3 89 90 91 92 93 "B 50 DL HP ^ T l rfl HP Mf_ 81 82 83 84 85 86 87 Plot 4 89 90 91 92 93 I ' J 2 0 flP HP HP HP M r i HP HP HP HH M HP HT1 81 82 83 84 85 86 87 Plot 5 89 90 91 92 93 •a 10 3-2 3 0 C/5 ru- ru, M N M,P HL U "n M P Mn "n M P . J 50 M \P MP _MP. MP MP m _HL p Mp nn m "I M . i n . 81 82 83 84 85 86 87 89 90 91 92 93 81 82 83 84 85 86 87 89 90 91 92 93 Figure 6.5 Comparison of measured (M) and predicted (P) annual soil loss (t ha'1 yr'1) at Beaverlodge using predicted Kb and rc and WEPP calculated Kr and Kt Altering the xc improved soil loss predictions efficiency, R2 from an overall average of -1.8 for all plots with the uncalibrated data to an overall average efficiency coefficient of -0.58. The negative efficiency coefficients indicate that soil loss predictions are worse than using the mean of annual soil loss measurements for the Peace River. Subsequent calibration by optimizing Kr (Table 6.4) and Kt was successful in most cases since a minimal deviation of zero percent between total measured and predicted soil loss were achieved for most plots, but this did not result in significant efficiency or correlation improvements. Table 6.4 Statistical summary of predicted soil loss (SL) (T ha'1 yr'1) and runoff (RO) (mm yr'1) for Beaverlodge - Calibrated Kb and Kr, based on yearly data PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 SL RO SL RO SL RO SL RO SL RO SL RO Sum: 4.0 478.0 10.3 457.1 105.6 738.2 4.2 413.9 21.9 480.2 93.4 662.7 Average: 0.3 36.8 0.8 35.2 8.1 56.8 0.3 31.8 1.7 36.9 7.2 51.0 % dev. 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 0% 1(0.10,12) P-1=tl2 ill=H2 H1=H2 H1=»2 P1=H2 P-1=H2 H1=H2 H1=H2 R2 -0.65 -1.07 -2.01 -1.26 0.30 -1.04 -0.96 -1.18 0.00 -1.27 -0.42 0.00 r2 1.0% 0.1% 0.0% 1.8% 36.2% 0.6% 1.1% 3.6% 34.2% 0.2% 18.3% 24.9% Page 69 Chapter 6 Model Testing and Analysis Calibration of the Kr parameter improved the efficiency for plot 0 (Table 6.4), and most notably, for plot 2 since it achieved a positive coefficient of efficiency in comparison to the xc calibration (Table 6.3). The positive efficiency coefficient (0.30) for annual soil loss predictions for plot 2 indicates that the model performance for predicting annual soil loss for fallow crops in northerly regions is acceptable if adjustments are made to the rill erodibility parameter (Kr). When adjustments were incorporated to the xc, Kr and Kit the model responded with proportionate shifts in annual soil loss. A proportionate shift refers to commensurate increase or decrease in soil loss predictions, either by a decrease or increase shift of the original uncalibrated annual soil loss. Changes in soil detachment parameters did not reflect the proportionate changes observed with annual soil loss, indicating a non-systematic error in the model. Calibrations of total runoff soil loss for Beaverlodge improved the results, however, for most cases, negative soil loss and runoff efficiencies indicate poor model performance, with the exception of the positive soil loss efficiency coefficients obtained for plot 4 with out any modifications and for plot 2, obtained with modifications made to the rill erodibility parameter. With a slight adjustment to the Kb parameter, runoff predictions for the fallow plot 5 produced a positive coefficient of efficiency indicating that annual runoff predictions for fallow plots can be confidently determined by slightly increasing the Kb parameter. Overall results for the fallow plots with minor adjustments to Kb the and by decreasing the Kr by almost one half can result in confident predictions of annual runoff and soil loss. Since fallow plots are very susceptible to surface ponding and particle detachment the WEPP can be a useful tool in determining annual rates of soil loss and runoff for bare fields. 6.2.2 EVENT ANALYSIS The Beaverlodge event analysis is comprised of comparisons of all 87 measured and predicted soil loss and runoff events. The results were compared on a 1:1 line for the individual plots and for all the events for a Page 70 Chapter 6 Model Testing and Analysis particular calibration, they were temporally and cumulatively distributed and analyzed. The results were also analyzed by a cumulative frequency distribution graph to analyze relative soil loss and runoff distributions in comparison to measured data. For the event analysis, the efficiency estimator, fl2 and the coefficient of determination, r2 was applied to test the degree of association between measured and predicted values. The results of the efficiency and the determination coefficients are summarized in Tables 6.5-6.8. Results are based on the individual events for a particular plot and coefficients based on all the combined events for all plots are also summarized, according to their calibration criteria. Without any modified parameters, soil loss and runoff event data was shown to have insufficient efficiency due to the negative efficiency coefficients which can be viewed in Table 6.5. Table 6.5 Statistical summary of predicted soil loss (SL) (T ha'1) and runoff (RO) (mm) for Beaverlodge-Original uncalibrated data, based on event data PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 SL RO SL RO SL RO SL RO SL RO SL RO fl2:-14.67 -1.07 -1.14 -1.33 -2.52 -1.54 -0.10 -0.54 -0.36 -1.85 -2.60 -2.35 r2: 0.0% 0.0% 0.0% 0.1% 9.2% 1.4% 0.2% 1.9% 10.9% 0.0% 3.9% 0.0% Cumulative fl2 for : Soil Loss -2.30 Runoff -1.44 The "Cumulative fl2" in Tables 6.5-6.8 refers to model efficiency in predicting event based runoff and soil loss for all plots, incorporating all events which can be viewed in Figures 6.12 and 6.20. The negative efficiency coefficients obtained from event comparisons of field measurements and model predictions from the original unmodified data indicates poor model performance, and low values of the coefficient of determination indicates a weak association between measured and predicted runoff and soil loss on an event basis. A comparison plot of measured verses predicted runoff based on individual event comparisons for all six plots can be viewed in Figure 6.6. Page 71 Chapter 6 Model Testing and Analysis PlotO 50 100 Measured Runoff (mm) Plot 3 0 20 40 Measured Runoff (mm) Plot 2 0 50 100 Measured Runoff (mm) Plot 4 0 50 100 Measured Runoff (mm) 0 50 100 Measured Runoff (mm) Plot 5 r 100 Qi •O 0 50 100 Measured Runoff (mm) Figure 6.6 1:1 comparison of measured and predicted event runoff (mm) at Beaverlodge using W E P P calculated Kb, r„ Kr and Kt Poor observable correlation can be observed between predicted and measured runoff events in Figure 6.6. In almost all cases, runoff is predicted when no or little runoff was measured, and similarly, in other cases measured runoff corresponded to little or no predicted runoff. A similar trend was observed with predicted and measured soil loss which can be viewed in Figure 6.7. Page 72 Chapter 6 Model Testing and Analysis PlotO 0 2 4 Measured Soil Loss (t/ha) Plot 3 - X -1 1.5 Measured Soil Loss (t/ha) Plot 1 0 1 2 Measured Soil Loss (t/ha) Plot 4 K X X X 0 2 4 Measured Soil Loss (t/ha) Plot 2 S 60 20 40 60 Measured Soil Loss (t/ha) Plot 5 20 40 Measured Soil Loss (t/ha) Figure 6.7 1:1 comparison of measured and predicted event soil loss (t ha"1) at Beaverlodge using W E P P calculated Kb, x„ Kr and Kt Soil loss comparisons also have poor correlation and poor efficiencies. The R coefficients ranged from -14.67 for plot 0 to a high of -0.10 for plot 3. Graphs of the temporal trends in measured vs. predicted runoff and soil loss from the fescue plot, plot 0 and from the fallow plot, plot 2 can be viewed in Figures 6.8 and 6.9. Page 73 Chapter 6 Model Testing and Analysis PlotO 100 50 0 u 81 82 83 84 85 86 87 88 89 90 91 92 93 94 100 50 0 U .1 X 81 82 83 84 85 86 87 88 89 90 91 92 93 94 8 0 | 81 82 83 84 85 86 87 88 89 90 91 92 93 94 o —1 V, 0 £ 81 82 83 84 85 86 87 88 89 90 91 92 93 94 Figure 6.8 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Beaverlodge plot 0 using W E P P calculated Kb, z„ Kr and Kt S= 100 t* 50 § 0 I ) .11 Plot 2 81 82 83 84 85 86 87 88 89 90 91 92 93 94 i a ioo o I 50 t 0 X J L 81 82 83 84 85 86 87 88 89 90 91 92 93 94 50 0 2 81 82 83 84 85 86 87 88 89 90 91 92 93 94 3 50 o "8 0 • L i . & 81 82 83 84 85 86 87 88 89 90 91 92 93 94 Figure 6.9 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Beaverlodge plot 2 using W E P P calculated Kb, T c , Kr and Kt In both cases, runoff and soil loss do not appear to be temporally correlated. The fist four years of data collection (1981-1984) measured the highest consecutive average runoff of 75 mm for all the plots. Graphically it appears that the last four years (1990-1993) had the highest average predicted runoff (60 mm). A record high runoff event of 88.5 mm for plot 0 was observed on January 27, 1984, but only 12 mm runoff was predicted to occur for that period. Conversely, highest predicted soil losses for plot 0 was predicted on June 6, 1987 and on June 21, 1987, but 0.19 t ha"1 and 0.0 t ha"1 was recorded for the respective dates. Since plot 2 was in continuous fallow from 1981 to 1985 and from 1990-1994, higher runoff and soil losses was predicted to occur for these years. The years, 1981 and 1982 were recorded to have the highest annual runoff of 155 and 128 mm respectively, but WEPP under-predicted runoff by 93% and over-predicted runoff by 31% respectively. Page 74 Chapter 6 Model Testing and Analysis A record high soil loss of 53.3 t ha"1 was predicted for plot 2 on June 29, 1993, but only 0.8 t ha"1 of soil loss was recorded for this period. From June 21-28, 1993, a total of 116.5 mm of rain fell on the plots, after a prolonged dry period. L.J.P. van Vliet observed (personal communication) dry antecedent moisture conditions resulted in greater infiltration of and storage of precipitation which resulted in low runoff and subsequent soil loss. High soil losses were predicted for the aforementioned period, which could have resulted from using a daily single peak hydrograph. This does not account for variations in storm intensities, breaks, and long duration's (> 24 hours). A cumulative plot of measured verses predicted runoff and soil loss for unmodified Beaverlodge data can be observed in Figures 6.10 an 6.11. Plot 0 Plot l Plot 2 Figure 6.10 Comparison of measured and predicted cumulative runoff (mm) at Beaverlodge using W E P P calculated Kb, T c , Kr and Kt With the exception of plot 3, measured and calculated runoff appears to converge at the end of the simulation. This trend was also observed in the total runoff predictions in Figure 6.1. It seems however, that the model is predicting slower than the measured data, but the similarities in convergence showed reasonable predictions at the end of the simulation period for total cumulative runoff. Cumulative soil loss for the Beaverlodge plots without the unmodified data can be viewed in Figure 6.11. Page 75 Chapter 6 Model Testing and Analysis PlotO 10 0 — meas. • pred. n u 81 83 85 87 89 91 93 Plot 3 .33 '5 oo 2 3 i e „ u 81 83 85 87 89 91 93 Plotl S 10 3 I 5 u 81 83 85 87 89 91 93 Plot 4 ^20 o d is '5 « 10 M 5 3 1 I 01 u 81 83 85 87 89 91 93 Plot 2 81 83 85 87 89 91 93 Plot 5 o 100 50 0 u 81 83 85 87 89 91 93 Figure 6.11 Comparison of measured and predicted cumulative soil loss (t ha"1) at Beaverlodge using W E P P calculated Kb, xc, Kr and The rotation plots and the fallow plots are observed to have similar trends in measured vs. predicted temporal soil loss. The fescue plots, plots 0 and 3 have large gaps between measured and predicted soil loss. The high soil loss predicted for plot 0 in 1987 which accounted for 93% of total predicted soil loss resulted from intense precipitation on barley, which occurred 4-6 weeks after planting, therefore, the simulated low plant canopy did not prevent precipitation interception resulting in high soil losses, and furthermore, the residue from fescue planted two years prior did not influence soil conservation. In the first year of simulation for plot 3, a high soil loss of 3.1 t ha"1 was observed which accounted for 77% of total measured soil losses, was not reflected in model predictions. These 'jumps' explain the large disparities observed with cumulative soil loss for plots 0 and 3. To test the overall efficiency of the model, all measured and predicted runoff and soil loss events for Beaverlodge with the unmodified soil parameters were compiled in Figure 6.12. Page 76 Chapter 6 Model Testing and Analysis Runoff Comparison - All Plots Soil Loss Comparison - All Plots | Measured Runoff (mm) Measured Soil Loss (t/ha) | Figure 6.12 Comparison of measured and predicted runoff (mm) and soil loss (t ha'1) for all Beaverlodge events using W E P P calculated Kb, rc, Kr and Kt The efficiency coefficients for measured and predicted runoff and soil loss for all events from all plots are -1.68 and -3.12 respectively. This still suggests overall poor model performance, but efficiency coefficients were expected to improve with calibration of sensitive parameters. A cumulative frequency distribution for all the events is also shown in Figure 6.13. From it, we can observe if the model reflects the measured distribution of the events. e Runoff-All Plots Soil Loss - All Plots itributioi itributioi 5 « 0.5 Cumulative Dis o O In Cumulatii o — meas. • • • pred. Cumulative Dis o O In Cumulatii o Cumulative Dis o O In Cumulatii o 10° Runoff (mm) 102 Cumulative Dis o O In 10~2 10° 102 Soil Loss (t/ha) Figure 6.13 Cumulative frequency distribution of measured and predicted runoff (mm) and soil loss (t ha'1) for all Beaverlodge events using W E P P calculated Kb, tc, Kr and Kt From the cumulative frequency distributions, measured and predicted runoff appear to have the same appearance, but the right side shifts of predicted runoff and predicted soil loss implies that WEPP tends to over-predict both runoff and soil loss. Since the minimal calculated soil detachment is limited to 0.1 t ha"1 a data gap is observed with predicted and measured soil loss frequency distributions. From Figure 6.13 we can interpret that 50% of the measured soil loss events were under approximately 0.06 t ha"1, but 50% predicted soil loss events were under 1.01 ha"1, more than an order of magnitude difference. Page 77 Chapter 6 Model Testing and Analysis The calibrated Kb values (Table 5.5) used to obtain a minimal difference between measured and predicted runoff were applied to the event analysis to obtain to evaluate comparative changes in soil loss and runoff. The efficiency coefficients of the event analysis is summarized in Table 6.6. Table 6.6 Statistical summary of predicted soil loss (SL) (T ha"1) and runoff (RO) (mm) for Beaverlodge - Calibrated Kb, based on event data PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 SL RO SL RO SL RO SL RO SL RO SL RO R2: -14.67 -1.07 -1.10 -1.31 -2.54 -1.54 -0.18 -2.25 -0.29 -1.79 -2.53 -2.29 r2: 0.0% 0.0% 0.0% 0.1% 9.2% 1.4% 0.4% 0.6% 11.2% 0.0% 3.8% 0.0% Cumulative R2 for : Soil Loss -2.28 Runoff -1.56 In most cases, runoff efficiency coefficients improved with the exception of plot 3 since efficiency decreased from -0.54 with the unmodified parameters to -2.25 with the modified Kb parameters (Table 6.6). The negative coefficients in all cases are indicative of poor model performance suggesting that the model cannot accurately determine soil loss and runoff predictions for short-term periods. A 1:1 graph of measured and predicted runoff obtained from modifying the Kb can be observed in Figure 6.14. Plot 3 50 100 Measured Runoff (mm) Plot 4 0 50 100 Measured Runoff (mm) Plot 5 0 50 100 Measured Runoff (mm) Figure 6.14 1:1 comparison of measured and predicted event runoff (mm) at Beaverlodge using predicted Kb and W E P P calculated rc, Kr and Kt Page 78 Chapter 6 Model Testing and Analysis The minor adjustments made to the Kb values for all plots did not reflect on any significant observable between measured and predicted runoff and soil loss. On January 22, 1991, all six plots had the highest predicted runoff with an average of 121 mm but only an average of 1.8 mm of runoff was recorded for the same period. Although no rainfall of equal or greater magnitude was predicted for the same period, this high volume of runoff is due to a record high rainfall event of 110 mm which occurred in June 6, 1990. Since there was only one recorded event in 1990 on March 2, all subsequent predicted events were summed up until the following sampling date on January 22, 1991. Soil loss predictions from using a calibrated Kb did not improve much. Efficiency coefficients marginally improved for plots 1, 4 and 5, but they were less than one. Cumulative runoff predicted with the calibrated Z^can be observed in Figure 6.15. Plot 0 Plot 1 Plot 2 Figure 6.15 Comparison of measured and predicted cumulative runoff (mm) at Beaverlodge using predicted Kb and W E P P calculated r„ Kr and Kt Calibration of Kb allowed convergence of total measured and predicted runoff (Figure 6.15). Since cumulative predicted runoff is shifted to the right, this implies that the model is slower at predicting runoff. A closer fit between measured and predicted cumulative runoff implies that the model is predicting correct Page 79 Chapter 6 Model Testing and Analysis amounts of runoff at the for a specified period. The large gaps in Figure 6.15 indicate disparity between temporal measurements and predictions. Small alterations made to calibrate the Kb resulted in minimal adjustments to soil loss. Mean predicted soil loss deviations calculated from all plots increased by 2% from 26% with uncalibrated data. Calibration of Kh retained a soil loss deviation from measured soil loss at 194% (over-prediction) for plot 0. Overall, changes in Kb did not show any observable improvements in soil loss. The calibrated critical sheer stress parameter rc, for the Beaverlodge plots were analyzed in accordance with measured events. The TC does not have any effect on runoff, hence efficiency coefficients (R2) for runoff remained the same. Calibration of the Tc improved the efficiency coefficients for plots 2 and 5, and most notably, for plot 0 since the efficiency increased from -14.67 to -1.33 (Table 6.7). However, overall, the efficiency coefficients were still less than zero suggesting inadequate model performance. Table 6.7 Statistical summary of predicted soil loss (SL) (T ha'1) and runoff (RO) (mm) for Beaverlodge - Calibrated Kb and Tc, based on event data PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 SL RO SL RO SL RO SL RO SL RO SL RO R2: -1.33 -1.07 -2.04 -1.31 -1.29 -1.54 -1.10 -2.25 -0.90 -1.79 -2.02 -2.29 r2: 0.0% 0.0% 0.1% 0.1% 2.4% 1.4% 0.3% 0.6% 9.4% 0.0% 1.8% 0.0% Cumulative R2 for : Soil Loss -1.43 Runoff -1.56 A comparison of measured and predicted event soil loss computed from the calibrated Kb and TC can be observed in Figure 6.16. Page 80 Chapter 6 Model Testing and Analysis Plot 0 Plot 1 Plot 2 Measured Soil Loss (t/ha) Measured Soil Loss (t/ha) Measured Soil Loss (t/ha) Plot 3 Plot 4 Plot 5 Measured Soil Loss (t/ha) Measured Soil Loss (t/ha) Measured Soil Loss (t/ha) Figure 6.16 1:1 comparison of measured and predicted event soil loss (t ha"1) at Beaverlodge using predicted Kb and rc and W E P P calculated Kr and Kt High soil losses predicted for the fallow plots, plots 2 and 5 of 39.9 and 42.2 t ha'1 on June 29, 1993 did not coincide with measured soil losses of 0.80 and 0.43 t ha"1, respectively. Since all measured events from January 5 to June 29, 1993 were summed, the exaggeration of soil loss is partially due to an over-prediction of snowmelt-induced runoff on January 29, 1993 which resulted in 24.0 and 25.3 t ha"1 or 60 and 60%, respectively, of total predicted soil loss for plots 2 and 5 for the sampling period. Calibration of Kb and tc improved WEPP's efficiency in predicting soil loss from -14.67 to -1.33 for plot 0 and from -2.52 to -1.29 for plot 2, with comparison to the soil loss predicted with the original unmodified parameters. A temporal distribution of measured and predicted soil loss for plots 0 and 2 are shown in Figures 6.17 and 6.18, respectively. Page 81 Chapter 6 Model Testing and Analysis PlotO 100 50 0 81 82 83 84 85 86 87 88 89 90 91 92 93 94 <r 100 S, 50 1 0 X 81 82 83 84 85 86 87 88 89 90 91 92 93 94 CO 1 2 1 13 0.5 Vi i o LL. 81 82 83 84 85 86 87 88 89 90 91 92 93 94 •8 0.5 Vi •o 0 81 82 83 84 85 86 87 88 89 90 91 92 93 94 Figure 6.17 Temporal distribution of runoff (mm) and soil loss (t ha"1) for Beaverlodge plot 0 using predicted Kb and Tc and W E P P calculated Kr and A, 100 S, 50 I 0 L L h Plot 2 XL 81 82 83 84 85 86 87 88 89 90 91 92 93 94 tl 100 o | 50 t 0 X X 81 82 83 84 85 86 87 88 89 90 91 92 93 94 « 40 to 2 rs 20 a o S 81 82 83 84 85 86 87 88 89 90 91 92 93 94 -40 ~i 20 o Vi •a 0 81 82 83 84 85 86 87 88 89 90 91 92 93 94 Figure 6.18 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Beaverlodge plot 2 using predicted Kb and xc and W E P P calculated Kr and Kt As previously stated, runoff values predicted for plots 0 and 2 on January 22, 1991 of 117.7 and 127.2 mm, respectively, resulted from a record high rainfall of 110 mm, on June 11, 1990. Runoff losses for this period were not reflected in the measured data. Since sampling did not occur from March 2, 1990 to January 22, 1991, all intermediate predicted events were summed until January 22, 1991. Overall, there is no observable correlation between measured and predicted soil loss and runoff for plots 0 and 2. A cumulative distribution of measured and predicted soil loss calculated from the calibrated Kb and rc are shown in Figure 6.19. Page 82 Chapter 6 Model Testing and Analysis PlotO < § 4 H 2 0 |— meas. pred. u 81 83 85 87 89 91 93 Plot 3 3 i 3 | ° 81 83 85 87 89 91 93 Plot 1 10 o u 81 83 85 87 89 91 93 Plot 4 ^ 2 0 O J 15 '3 ^ 10 J2 5 I 0 ° 81 83 85 87 89 91 93 Plot 2 100 3 50 0 u 81 83 85 87 89 91 93 Plot 5 8 80 o = 60 ^ 4 0 > •a J3 20 3 0 u 81 83 85 87 89 91 93 Figure 6.19 Comparison of measured and predicted cumulative soil loss (t ha"1) at Beaverlodge using predicted Kb and xc and W E P P calculated Kr and Kt The fescue plots, plots 0 and 3 did not confirm any observable agreement (Table 6.7) between data due to the large gaps between cumulative measured and predicted soil loss calculated with the calibrated Kb and Tc. The main cause of the disparity between plots 2 and 5 resulted from a record high measured soil losses of 29.97 and 24.92 t ha"1, respectively, recorded on July 21, 1982. Only 8.0 and 7.2 t ha'1 soil loss were recorded for this period for plots 0 and 2, respectively. Figure 6.20 demonstrates the overall performance of WEPP in predicting event-based runoff and soil loss calculated from the calibrated Kb and rc, for the combined events from all the plots Page 83 Chapter 6 Model Testing and Analysis Runoff Comparison - All Plots Soil Loss Comparison - All Plots Measured Runoff (mm) Measured Soil Loss (t/ha) Figure 6.20 Comparison of measured and predicted runoff (mm) and soil loss (t ha'1) for all Beaverlodge events using predicted Kb and % and W E P P calculated Kr and K, Since calibration of xc did not influence runoff, the runoff comparisons on Figures 6.20 and 6.21 are the result of calibrated Kb. However, the combined calibration of Kb and Tc improved soil loss efficiency from -2.30 to -1.43. This improvement is still not acceptable for adequate model efficiency since the combined R2 is still less than zero. The cumulative frequency distributions of measured and predicted soil loss are presented in Figure 6.21. Runoff - All Plots Soil Loss - All Plots 10° 102 10"2 10° 102 Runoff (mm) Soil Loss (t/ha) Figure 6.21 Cumulative frequency distribution of measured and predicted runoff (mm) and soil loss (t ha"1) for all Beaverlodge events using predicted Kb and rc and W E P P calculated Kr and Ki The cumulative frequency distributions of soil loss do not indicate overall similarity in prediction since both curves appear to follow dissimilar patterns. From Figure 6.21, it is evident that WEPP tends to over-predict runoff and soil loss even when each plot was calibrated for Kb and rc. Page 84 Chapter 6 Model Testing and Analysis A third calibration procedure involved the calibration of the rill detachment parameter, Kr. The results of this calibration are presented in Table 6.8 Table 6.8 Statistical summary of predicted soil loss (SL) (T ha'1) and runoff (RO) (mm) for Beaverlodge - Calibrated Kb and Kn based on event data PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 SL RO SL RO SL RO SL RO SL RO SL RO R2: -1.29 -1.07 -2.56 -1.31 -0.47 -1.54 -2.45 -2.25 -1.09 -1.79 -1.49 -2.29 r2: 0.0% 0.0% 0.1% 0.1% 8.7% 1.4% 0.1% 0.6% 11.4% 0.0% 3.8% 0.0% Cumulative R2 for : Soil Loss -0.79 Runoff -1.56 Runoff was not affected by changes in Kr< since it is only a soil detachment parameter. Efficiencies for soil loss improved for plots 0, 2 and 5 in comparison to the original unmodified parameters and results obtained from the Kb and rc calibration. However, all improvements still produced negative efficiency coefficients indicating poor model performance. Overall soil loss predictions for all combined events did improve, since the R2 improved from -2.30 with the unmodified data to -0.79 with the adjusted Kb and Kr. This represented the most improved overall efficiency from all calibrations, suggesting that soil loss predictions from the Beaverlodge erosion plots can be improved by modifying the baseline conductivity (Kb) and rill erodibility (Kr) parameters, Alterations of the interill erodibility parameter, Kt did not return successful calibrations for all the Beaverlodge erosion plots, since the efficiency coefficients were still less than zero, indicating poor model performance. Soil loss predictions were not sensitive to changes in Ki as determined by the sensitivity analysis (Table 5.1) and parameter calibration (Table 5.4). Page 85 Chapter 6 Model Testing and Analysis 6.2.3. WINTER/ SUMMER ANALYSIS To test WEPP's winter hydrology component, predictions from the original unmodified and calibrated outputs were analyzed for snowmelt verses rainfall induced runoff. It was suggested by L.J.P. van Vliet that any predicted events from November 1 to April 30 be considered as snowmelt induced runoff, otherwise any recorded or predicted events outside this period will be considered as rainfall induced runoff and soil loss. In April, when snowmelt was still likely to occur, the average minimum and maximum temperatures for Beaverlodge were -2 and 9°C respectively. The results of the winter/summer analysis for snowmelt and rainfall induced runoff for the original uncalibrated predictions are tabulated in Table 6.9. Table 6.9 Statistical summary of snowmelt vs. rainfall-induced soil loss (SL) (T ha'1) and runoff (RO) (mm) for Beaverlodge - Uncalibrated data, based on event data Runoff PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Snow Rain Snow Rain Snow Rain Snow Rain Snow Rain Snow Rain Meas. 434.5 43.3 378.9 78.7 375.3 362.9 349.2 64.8 324.5 155.7 396.4 266.3 RO Pred. RO 302.9 176.2 257.2 212.1 329.5 404.8 130.8 109.1 272.1 229.3 306.9 370.8 %difj -30% 307% -32% 170% -12% 12% -63% 68% -16% 47% -23% 39% R2 -1.15 -4.49 -1.19 -12.42 -1.73 -1.19 -0.88 -0.39 -1.71 -4.16 -1.72 -7.79 Soil Loss PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Snow Rain Snow Rain Snow Rain Snow Rain Snow Rain Snow Rain Meas. SL 1.6 2.4 1.8 8.5 7.6 97.9 1.0 3.2 2.5 19.3 6.1 87.3 Pred. SL 0.4 11.3 1.4 6.2 43.2 136.1 0.0 0.7 1.8 14.3 20.0 97.6 %difj -75% 373% -22% -27% 467% 39% -100% -78% -29% -26% 229% 12% R2 -0.21 -23.26 -4.75 -1.11 -93.71 -2.46 -0.09 -0.12 -1.36 -0.46 -19.19 -2.85 WEPP under predicted snowmelt-induced runoff by -20% and over-predicted rainfall-induced runoff by 106%, on average for all plots,(Table 6.9). The efficiency coefficients for runoff were still less than zero for rainfall and snowmelt induced events indicating poor model performance. Soil loss predictions presented a different picture. The model over-predicted soil loss from snowmelt by an average of 78% and over-predicted soil loss from rain fall by 49%. The efficiency coefficients ranged from -93.7 for snowmelt induced soil loss for plot 2 to a high of only -0.09 for snow-induced runoff for plot 3, but since all the efficiency coefficients are less than zero, this indicates poor model performance in predicting event based Page 86 Chapter 6 Model Testing and Analysis runoff and soil loss from snowmelt or rainfall. A comparison of snowmelt verses rainfall-induced runoff calculated with the unmodified data for all the Beaverlodge plots is presented in Figure 6.22. SNOW: Plot 0 0 50 100 Meas. RO (mm) SNOW: Plot 2 100 0 50 100 Meas. RO (mm) SNOW: Plot 4 0 50 100 Meas. RO (mm) RAIN: PlotO 0 20 40 Meas. RO (mm) RAIN: Plot 2 0 50 Meas. RO (mm) RAIN: Plot 4 0 50 Meas. RO (mm) SNOW: Plot 1 0 50 100 Meas. RO (mm) SNOW: Plot 3 0 20 40 Meas. RO (mm) SNOW: Plot 5 0 50 100 Meas. RO (mm) RAIN: Plot 1 0 50 Meas. RO (mm) RAIN: Plot 3 0 20 40 Meas. RO (mm) RAIN: Plot 5 0 50 Meas. RO (mm) Figure 6.22 Comparison of measured and predicted rain vs. snow induced event runoff (mm) at Beaverlodge using W E P P calculated Kb, rc, Kr and Kt In general, there was no clear agreement present between measured and predicted data for both summer and winter hydrologic routines. A comparison of soil loss predictions for all plots from snowmelt and rainfall-induced soil loss is shown in Figure 6.23. Page 87 Chapter 6 Model Testing and Analysis SNOW: Plot 0 RAIN: Plot 0 SNOW: Plot 1 RAIN: Plot ] 0 0.5 1 Meas. SL (t/ha) SNOW: Plot 2 0 5 10 Meas. SL (t/ha) SNOW: Plot 4 0 0.5 Meas. SL (t/ha) 0 5 Meas. SL (t/ha) RAIN: Plot 2 0 50 Meas. SL (t/ha) RAIN: Plot 4 0 2 4 Meas. SL (t/ha) 0 0.5 1 Meas. SL (t/ha) SNOW: Plot 3 0 0.5 Meas. SL (t/ha) SNOW: Plot 5 0 5 Meas. SL (t/ha) 0 1 2 Meas. SL (t/ha) RAIN: Plot 3 0 1 Meas. SL (t/ha) RAIN: Plot 5 0 50 Meas. SL (t/ha) Figure 6.23 Comparison of measured and predicted rain vs. snow induced event soil loss (t ha"1) at Beaverlodge using W E P P calculated Kb, rc, Kr and Kt Soil loss predictions for fallow plots, 2 and 5 were over-predicted for snowmelt and rainfall events (Table 6.9). The model over-predicted rainfall induced runoff for the fallow plots by an average of 26%, and it over-predicted snowmelt induced soil loss for plot 2 by 467% and for plot 5 by 229% for an average over-prediction of 348%. Furthermore, the low efficiency coefficients determined from snowmelt induced events for plots 2 and 5 (R2 = -94.7 and -19.2 respectively), could possibly suggest a flaw in the winter hydrology routines since the model grossly over-predicts snowmelt-induced soil loss. By using calibrated values of Kb and xc, improvements in overall total model predictions were expected which would reflect on improvements in soil loss and runoff predictions for the winter and summer events. Page 88 Chapter 6 Model Testing and Analysis The results of the winter/summer analysis for snowmelt and rainfall induced runoff for the calibrated predictions are tabulated in Table 6.10. Table 6.10 Statistical summary of snowmelt vs. rainfall-induced soil loss (SL) (T ha"1) and runoff (RO) (mm) for Beaverlodge - Calibrated Kb and T„ based on event data Runoff PlotO Plotl Plot 2 Plot 3 Plot 4 Plot 5 Snow Rain Snow Rain Snow Rain Snow Rain Snow Rain Snow Rain Meas. 434.5 43.3 378.9 78.7 375.3 362.9 349.2 64.8 324.5 155.7 396.4 266.3 RO Pred. RO 302.2 175.8 251.1 206.0 331.6 406.6 278.4 135.5 260.4 219.8 299.2 363.5 %diff -30% 306% -34% 162% -12% 12% -20% 109% -20% 41% -25% 37% R2 -1.15 -4.48 -1.18 -11.93 -1.74 -1.19 -3.21 -1.10 -1.67 -3.87 -1.68 -7.61 Soil Loss PlotO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Snow Rain Snow Rain Snow Rain Snow Rain Snow Rain Snow Rain Meas. SL 1.6 2.4 1.8 8.5 7.6 97.9 1.0 3.2 2.5 19.3 6.1 87.3 Pred. SL 0.4 3.6 2.5 7.8 26.9 78.8 1.7 2.5 3.7 18.2 15.8 77.6 %diff -75% 51% 39% -9% 253% -20% 65% -22% 46% -6% 160% -11% R2 -0.21 -2.00 -7.30 -2.01 -43.78 -1.33 -5.61 -0.76 -4.35 -0.98 -13.58 -2.25 Calibration of Kb and TC improved runoff efficiency coefficients slightly between the original unmodified data (Table 6.9) and calibrated data (Table 6.10). Calibration also improved soil loss efficiency coefficients for the rain fall induced events for all plots 0, 2 and 5. With calibration, the average soil loss deviation from snowmelt was 81%, and -3% for rainfall influenced events. With the exception of plot 0, the model over-predicted snowmelt-induced soil loss. Soil loss efficiencies from snowmelt were highly negative for plots 1-5. Comparison plots of the events produced from snowmelt, calculated with the calibration in Kb and TC are presented in Figure 6.24. Page 89 Chapter 6 Model Testing and Analysis Snow Induced Runoff ^ i Q(B) Rain Induced Runoff oi 10° j 10° P 10° Measured Runoff (mm) Snow Induced Soil Loss ©D 10" g o o / 10" 10° Measured Runoff (mm) Rain Induced Soil Loss e © — 10" 10° Measured Soil Loss (t/ha) Measured Soil Loss (t/ha) Figure 6.24 Comparison of measured and predicted rain vs. snow induced runoff (mm) and soil loss (t ha"1) for all Beaverlodge events using predicted Kb, and tc, and W E P P calculated Kr and Kt There seems to be observable agreement between measured and predicted snow induced runoff and rainfall induced runoff and soil loss, since only positive values can be plotted on log-log type scale used in all plots for Figure 6.24, this does not reflect the true nature of the comparison. Actual efficiency coefficients are still less than zero (Table 6.10). Calibrations of Kh and Tcdid not significantly improve snowmelt induced soil loss, suggesting inadequate model performance for snowmelt and rainfall induced events. The overall average high soil loss deviation from snowmelt indicates a general over prediction in soil loss during snowmelt scenarios. In most cases, calibrations of Kb, Tc, Kr and Kj improved model predictions and efficiencies. However, poor efficiency coefficients (R2) on a yearly and event basis suggests that improvements need to be made to WEPP in order to improve soil loss and runoff predictions. The low negative efficiency coefficients and high over-predictions of snowmelt induced soil loss suggests weaknesses in WEPP's winter hydrology Page 90 Chapter 6 Model Testing and Analysis component. Since the Peace River Region is located in an extreme northerly agricultural production area, snowmelt plays an important role in soil loss and runoff predictions. For the second portion of this study, measured runoff and soil loss data from Abbotsford, Bri t ish Columbia was compared to unmodified and calibrated predictions, in the same form as the Beaverlodge analysis. Since the Lower Fraser Va l l ey o f Bri t ish Columbia where the Abbotsford erosion plots are located are not subjected to extreme winters, improvements in runoff and soil loss were expected for this region. 6.3 ABBOTSFORD ANALYSIS Analys is for the Abbotsford plots was conducted on an annual and event basis in continuous simulation mode to evaluate model performance in predicting annual runoff and soil loss. The model was also tested in continuous simulation mode with break-point type hydrograph data to evaluate the ability o f the model to predict soil loss and runoff based on break-point precipitation data for a few selected events. Detailed runoff and soil loss measurements for individual events can be viewed in Appendix 4. 6.2.1 YEARLY ANALYSIS Statistical data based on yearly analysis o f measured and predicted soil loss and runoff are presented in Tables 6.11 to 6.13. Table 6.11 Statistical summary of soil loss (SL) (T ha' 1 yr'1) and runoff (RO) (mm yr"1) for Abbotsford-Original uncalibrated data, based on yearly data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 S L R O S L R O S L R O S L R O S L R O S L R O Sum: 52.9 592.5 69.5 817.4 64.7 845.1 68.6 958.0 0.0 141.8 0.0 191.6 Average: 8.8 98.8 11.6 136.2 10.8 140.9 11.4 159.7 0.0 23.6 0.0 31.9 %dev. 839% -50% 227% -41% 1135% -8% 242% -18% -100% -82% -100% -75% t(0.10,5) fiKte P-1>P-2 P1=P2 H1>P.2 P-1<ll2 til=P2 Hl>H2 H1>H2 P1>P2 P-1>P-2 fl2-808.35 0.15 -14.87 -0.39 -875.44 0.86 -16.46 -0.21 -2.83 -0.38 -3.72 -0.20 r2 41.4% 98.9% 27.1% 88.2% 27.3% 93.0% 28.2% 67.9% N / A 66.1% N / A 74.9% Page 91 Chapter 6 Model Testing and Analysis F r o m the original uncalibrated data, runoff and soil loss predictions on average were 42% smaller and 374% greater. The model over-predicted soil loss by an average 987% for the winter cover-crop plots 1 and 3. Conversely, soil loss was under-predicted for contour plots 5 and 6. Estimations for the control plots 2 and 4, over-predicted soil loss by an average of 235% and under-predicted runoff by 25%. The positive efficiency coefficients obtained for annual runoff predictions for plot 1 and most notably, for plot 2 (R2 = 0.86) indicates that the model adequately predicts annual runoff, unfortunately, soil loss predictions for the same plots did not perform as wel l . A graph of measured and model predicted total runoff and soil loss calculated from the unmodified data is presented in Figure 6.25. 1600 1400 e 1200 e SB 1000 o ai 800 otal 600 H 400 200 0 tL t l 3 4 Plot 80 70 'rt •g 60 J 5 0 Ui 2 30 o H 20 10 0 M 3 4 Plot Figure 6.25 Comparison of measured and predicted total (1989-1994) runoff (mm) and soil loss (t ha"1) at Abbotsford using W E P P calculated Kb, rc, Kr and Kt From Figure 6.25, it is apparent that the model tended to under-predict runoff and over-predict soil loss with the exception of plots 5 and 6, where no soil loss was predicted. A comparison of annual measured and predicted runoff and soil loss for all 6 plots are presented on Figures 6.26 and 6.27 respectively. Page 92 Chapter 6 Model Testing and Analysis Plot 1 Plot 2 Plot 3 , . . _ , . , , 4 0 0 . , _ 89 90 91 92 93 94 89 90 91 92 93 94 89 90 91 92 93 94 Plot 4 Plot 5 Plot 6 89 90 91 92 93 94 89 90 91 92 93 94 89 90 91 92 93 94 Figure 6.26 Comparison of measured and predicted annual runoff (mm yr"1) at Abbotsford using W E P P calculated Kh, x„ Kr and Kt F r o m Figure 6.26, there is observable agreement between measured and predicted annul runoff. However, for the cross-slope plots 5 and 6, there appears to be a large discrepancy between measured and predicted runoff in 1990. Although the model under-predicted runoff by an average of 46% for all plots, the positive runoff efficiency coefficients for plots 1 and 3 suggests adequate model performance. The positive values can be translated into acceptable model performance since the predictions are more effective than the measured means. The high average coefficient of determination for all plots (r2) of 82% indicates that 82% of the total variation in predicted runoff values can be accounted for by a linear relationship with the annual measured data. The correlation coefficient, r (where: r = square root of r2) for the unmodified data ranges from 81% for plot 5 to 99.2% for plot 1. These results are encouraging. Similar trends can be observed with measured and predicted runoff, as presented in Figure 6.26, with the exception of the high runoff predicted for plots 5 and 6 in 1990. Correlation's between annual measured and predicted soil loss are not very apparent, as can be observed in Figure 6.27. Page 93 Chapter 6 Model Testing and Analysis Plot 1 "» 20 f, t/5 3 io 'S 0 0 0 np n P 89 90 91 92 93 94 Plot 4 >>30 "a ! » 3,o o O 0 0 p nf. PIP 89 90 91 92 93 94 Plot 2 '30 -20 ho p n P nL nP 89 90 91 92 93 94 Plot 5 l a 0 ' 4 f.0.3 8 0.2 S 0.1 on n M H P h p 89 90 91 92 93 94 Plot 3 ~ 3 0 § 20 ~. 1° 'o on 0 r u -89 90 91 92 93 94 Plot 6 .0.4 0.3 j o , ^ 0.1 on ^ .1 ^ * ( A M P P P M p OP n -89 90 91 92 93 94 Figure 6.27 Comparison of measured and predicted total soil loss (t ha' 1 yr"1) at Abbotsford using W E P P calculated Kb, x„ Kr and Kt Annual soil loss efficiencies determined from the unmodified data are highly negative which suggests poor model performance. The gross over-predictions of soil loss for plots 1 and 3 returned efficiency coefficients o f -808.35 and -875.44 respectively. Interesting to note that the coefficient of determination (r 2 ) was on average 39% which translates into a correlation (r 2) of 62%. This emphasizes the biased behavior o f the coefficient o f determination and the coefficient of correlation since the best fitted data does not necessarily reflect model efficiency, and it cannot reflect systematic errors in model predictions. T o improve runoff predictions, each individual plot was calibrated for baseline conductivity, in order to achieve minimal deviation between total measured and predicted runoff. The results o f this calibration are presented in Table 6.12. Page 94 Chapter 6 Model Testing and Analysis Table 6.12 Statistical summary of calibrated soil loss (SL) (T ha' 1 yr'1) and runoff (RO) (mm yr' 1) for Abbotsford-Calibrated Kb, based on yearly data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 S L R O S L R O S L R O S L R O S L R O S L R O Sum: 85.2 1179.3 98.8 1376.8 68.8 918.9 77.8 1174.6 OO 642.6 O 0 649.2 Average: 14.2 196.6 16.5 229.5 11.5 153.2 13.0 195.8 0.0 107.1 0.0 108.2 % dev. 1413% 0% 365% 0% 1214% 0% 287% 0% -100% -20% -100% -17% t(0.10,5) Mr<M2 Mi=/"2 Mt=M2 Hi=r>2 Mr<M2 Mi=/*2 i^=M2 Mr=/^ 2 J*i>^ 2 M<=M2 M»>M2 Mr=f*2 fl2-2114.7 0.84 -29.83 0.31 -991.66 0.85 -21.34 -0.42 -2.83 0.43 -3.72 0.43 r2 A9.\% 95.7% 29.2% 92.4% 27.9% 94.4% 28.6% 71.8% N / A 66.9% N / A 64.6% Calibration of Kb was successful for plots 1-4, but plots 5 and 6 could not attain a minimal deviation of 0%. The results o f the positive runoff efficiency coefficients for plots 1, 2, 3, 5, and 6 are very encouraging because this states that annual runoff predicted by W E P P is a better estimator than using the means of the measured value for the Lower Fraser Val ley . However, for plots 5 and 6, the Kb value had to be reduced to 0.1 m m hr"1 which is almost comparable to the Kb for a semi-impervious surface. Forcing the low Kb for the cross-slope plots did not achieve a minimal deviation of 0% between measured and predicted total runoff. The effects o f calibrating Kb for runoff and soil loss can be viewed in Figure 6.38. The low rates of predicted runoff and soil loss can be attributed to the 0% contour modeled along the r i l l , which promotes absolutely no r i l l runoff and subsequent rill-detached soil loss since water does not flow in a field of equal gradients. Total runoff predictions improved for all plots, due to the closeness of total measured and predicted runoff. The r-test results indicated no significant differences between measured annual and predicted runoff at the 0.10 significance level. In all cases, the Kb had to be decreased in order to increase runoff predictions. The increase in this parameter also increases soil loss since lower infiltration rates result in greater runoff volumes, which promotes more soil detachment. Therefore, soil losses increased by 139% on average. Figures 6.28 and 6.21 indicate the annual runoff and soil loss values predicted from Kb calibration. Page 95 Chapter 6 Model Testing and Analysis Plot 1 89 90 91 92 93 94 Plot 4 r r—i ~l n M 89 90 91 92 93 94 Plot 2 p M 89 90 91 92 93 94 Plot 5 <400 o 200 s oi 0 ML P LM 89 90 91 92 93 94 Plot 3 89 90 91 92 93 94 Plot 6 •400 o 200 3 0 n M E , P 11 89 90 91 92 93 94 Figure 6.28 Comparison of measured and predicted annual runoff (mm yr'1) at Abbotsford using calibrated Kb and W E P P calculated tc,Kr and Kt Calibration of Kb significantly improved the predicted runoff for al l plots, with the exception of plot 4 (Figure 6.28). The high r2 values translates into high correlation between measured and predicted runoff. F r o m Figure 6.28, it is obvious the model predicts relatively similar amounts of runoff with the calibrated Kb. Calibrations of Kb did not reflect on any observable improvement in annual predicted soil loss for the six Abbotsford Erosion plots. The efficiency coefficients based on annual soil loss for plots 5 and 6 were not adjusted since soil loss was not predicted for this calibration. Efficiency coefficients for plots 1-4 decreased, and for the winter cover-crop plots, efficiency decreased from an average of -842 with the unmodified data to -1553 with the calibrated Kb. The negativity still indicates poor model performance for predicting soil loss, or from bad to worse. T o improve soil loss predictions, all plots were calibrated for Tc, while maintaining Kb in order to achieve minimal deviation between observed runoff and soil loss. The results o f this calibration is presented in Table 6.13. Page 96 Chapter 6 Model Testing and Analysis Table 6.13 Statistical summary of calibrated soil loss (SL) (T ha' 1 yr'1) and runoff (RO) (mm yr' 1) for Abbotsford-Calibrated Kb and tc, based on yearly data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 S L R O S L R O S L R O S L R O S L R O S L R O Sum: 16.4 1179.3 21.2 1376.8 14.7 918.9 20.1 1174.6 0.0 642.6 0.0 649.2 Average: 2.7 196.6 3.5 229.5 2.5 153.2 3.4 195.8 0.0 107.1 0.0 108.2 %dev. 191% 0% 0% 0% 181% 0% 0% 0% -100% -20% -100% -17% t(0.10,5) Ml=M2 1*1=1*2 i*1=\*2 Ml=^ 2 \*1=\*2 \*1=\*2 1*1=1*2 \*1=\*2 i*1>l*2 Ml=^ 2 Mf>M2 l*i=\*2 fl2 -89.87 0.84 -1.92 0.31 -44.80 0.85 -1.80 -0.42 -2.83 0.43 -3.72 0.43 r2 49.9% 95.7% 22.8% 92.4% 39.2% 94.4% 28.3% 71.8% N / A 66.9% N / A 64.6% Only the control plots 2 and 4 were able to converge successfully, or were able to obtain a minimal deviation between measured and predicted soil loss. Alterations in the tc for the cross slope plots 5 and 6 did not result in improvements of soil loss predictions (Figure 6.29). 1600 1400 e 1200 E SB 1000 o I 800 1 600 400 200 0 M P M P M p M P M p M p 25 •« 20 i | 1 5 '5 I 1 0 o H 5 0 M P M P M p M P M M rV 1 2 3 4 Plot 5 6 1 2 3 4 Plot 5 6 Figure 6.29 Comparison of measured and predicted total runoff (mm) and soil loss (t ha'1) at Abbotsford using calibrated Kb and % and W E P P calculated Kr and Kt A maximum critical sheer stress, xc of 8 N m" 2 was used on the cover-crop plots, 1 and 3. This improved overall predictions, by improving the deviation from an average of 1314% with the calibrated Kb to an average of 186% with the calibrated Kb and Tc. This resulted in an over-prediction, but the f-test coefficients at the 0.10. significance level showed no significant differences between mean annual measured and predicted soil loss. A comparison of measured and predicted annual soil loss calculated with the adjusted TC can be viewed in Figure 6.30. Page 97 Chapter 6 Model Testing and Analysis Plot 1 |>10 ffl n-89 90 91 92 93 94 Plot 4 | 10 to '5 vi 0 n M -Li t 89 90 91 92 93 94 Plot 2 15 § 10 o d 5 I n M 89 90 91 92 93 94 Plot 5 £ 0 . 4 "a f.0.3 I 0.2 S 0.1 Vi n M _ J P _ T ] P 89 90 91 92 93 94 Plot 3 § 6 0 4 —1 1 2 00 n M M nf l H A H P a 89 90 91 92 93 94 Plot 6 0.4 0.3 0.2 3 0.1 Vi J .1 t .1 M P P P M p H P rip 89 90 91 92 93 94 Figure 6.30 Comparison of measured and predicted annual soil loss (t ha"1 yr'1) at Abbotsford using calibrated Kb and tc and W E P P calculated Kr and Kt For plots 1-4, more soil loss was observed from 1989 and 1991, than the remaining years, since this included 2 years of brussels sprouts and the first year of strawberries. After the cover crop was established, less soil loss was predicted for plots 1 and 3 than that for the control plots 2 and 4. Since W E P P could not simulate intercropping, the density of the strawberry plants were doubled in the management files to simulate the effects o f a cover crop. The increased residue modeled in-between the ridges predicted lower rates of runoff and soil loss for plots 1 and 3 in comparison to the control plots 2 and 4 during 1991-1994. Because of the large discrepancies in measured verses predicted soil loss in 1989, 1990 and 1992, this accounted for most of the deviations from the total observed soil loss for plots 1-4 , which resulted in very low efficiencies as judged by the R2 values. So i l loss was not predicted for the Abbotsford contour plots. A similar pattern was observed for the Beaverlodge contouring scenario modeled for plot 2 during 1986 to 1990. A s stated earlier, the lack of prediction can be attributed to the zero percent slope along the contours which can not promote r i l l induced runoff and soil loss since water cannot flow on a horizontal plane. Adjustments in the interrill erodibility parameter are expected to induce soil loss for plots 5 and 6. Page 98 Chapter 6 Model Testing and Analysis Unl ike the runoff predictions, there is no observable correlation between measured and predicted soil loss calculated with the modified xc parameter. Adjustment in the Tc improved model efficiencies for plots 1-4, but it did not alter predictions for plots 5 and 6, since the model still returned a zero value of predicted soil loss. The coefficient of determination, r2 did not improve for plots 2 and 4, suggesting a biased behavior of the r 2 coefficient. O n average, the efficiency coefficients for plots 2 and 4 increased from -25.6 to -1.9 which implies an improvement in the predictions, but it still does not justify model prediction adequacy since the efficiency coefficients are still less than zero, which means that annual soil loss observations are still better than using model predicted observations. Since soil detachment in the rills where most soil loss is predicted to occur is proportional to the r i l l erodibility parameter, Kr, changes in the r i l l erodibility parameters were subjected to calibration. The statistical results o f this calibration are tabulated in Table 6.14. Table 6.14 Statistical summary of calibrated soil loss (SL) (T ha"1 yr'1) and runoff (RO) (mm yr"1) for Abbotsford-Calibrated Kh and Kn based on yearly data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 S L R O S L R O S L R O S L R O S L R O S L R O Sum: 16.4 1179.3 21.4 1376.8 14.8 918.9 20.1 1174.6 0.0 642.6 0.0 649.2 Average: 2.7 196.6 3.6 229.5 2.5 153.2 3.4 195.8 0.0 107.1 0.0 108.2 %dev. 191% 0% 1% 0% 183% 0% 0% 0% -100% -20% -100% -17% t(0.10,5) P-1=P-2 Ul=p2 U1=P2 li\=P-2 P1=U2 Pl=ll2 U,=P2 Ul=P2 Ul>U2 ^1=^-2 fil>P2 P\=P2 fl2 -89.87 0.84 -1.93 0.31 -44.95 0.85 -1.94 -0.42 -2.83 0.43 -3.72 0.43 r2 A9.9% 95.7% 23.1% 92.4% 39.2% 94.4% 27.7% 71.8% N / A 66.9% N / A 64.6% From Table 6.14, only plots 2 and 4 resulted in a relative minimal degree of deviation in soil loss. Changes in soil loss predictions were observed with plots 1 and 2, but these changes reflected similarities between calibration of r i l l detachment and sheer stress. F rom Equations 3.16 and 3.17, it can be seen that detachment capacity of r i l l flow is a product of the r i l l erodibility parameter and the difference between the sheer stress of f low and the critical sheer stress of the soil . Hence r i l l detachment is directly proportional to r i l l erodibility parameter (Kr) and the soil's critical sheer stress (Tc), and as discussed earlier, both parameters behave in the same manner. Page 99 Chapter 6 Model Testing and Analysis Changes to the r i l l erodibility (Kr) parameter behaved similarly as that which was observed for changes in the sheer stress parameter, Tc. So i l loss efficiencies based on annual data when compared to the results obtained original unmodified data and the modified data with adjustments in the baseline conductivity parameter improved, but coefficient of efficiencies and determinations remained relatively the same. A final calibration examining soil loss predictions with changes in the interrill erodibility parameter was conducted, and the results are presented in Table 6.15. Table 6.15 Statistical summary of calibrated soil loss (SL) (T ha"1 yr"1) and runoff (RO) (mm yr"1) for Abbotsford-Calibrated Kb and Kh based on yearly data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 S L R O S L R O S L R O S L R O S L R O S L R O Sum: 69.7 1179.3 78.7 1376.8 55.2 918.9 61.0 1174.6 0.1 642.6 0.1 649.2 Average: 11.6 196.6 13.1 229.5 9.2 153.2 10.2 195.8 0.0 107.1 0.0 108.2 %dev. 1137% 0% 270% 0% 954% 0% 204% 0% -92% -20% -92% -17% t(0.10,5) Pl<P-2 P1=H2 P-\=P2 P-1=P2 Hl<tl2 P1=P2 Pt=H2 1*1=^2 ^>^2 P-1=fl2 Hl>^2 iH=^2 fl2-1418.2 0.84 -18.97 0.31 -665.46 0.85 -13.22 -0.42 -2.32 0.43 -3.03 0.43 r2 44.3% 95.7% 28.4% 92.4% 21.9% 94.4% 26.9% 71.8% 6.7% 66.9% 10.1% 64.6% Calibrations of the interrill erodibility parameter {Ki) for al l plots were not successful since the model results were not able to converge successfully with measured results (Table 6.15). The results from the sensitivity analysis presented in Table 5.1 indicates that soil loss is 6 times less sensitive to the Kr parameter than the Kr or tc parameters. Slight improvements in annual soil loss predictions were observed (Table 6.15) when compared to predictions from the unmodified (Table 6.13) or the Kb calibrated (Table 6.14) data for plots 1-4. A s expected, improvements were observed for the cross-slope plots since for the first time, soil loss was predicted (Table 6.15). Predictions improved slightly, since the deviation from the measured values averaged -92% for plots 5 and 6. Interril detachment is primarily a function of rainfall intensity and runoff (Equation 3.14). In W E P P , the cross-slope scenario was modeled with a user-specified zero percent horizontal slope. Theoretically, off-slope interill detachment is only predicted to occur on the bottom most r i l l since detachment form higher ridges is deposited onto lower ridges, unless ponding is greater than the vertical ridge height which results in massive overland sheet flow. However, in most cases not subjected to Page 100 Chapter 6 Model Testing and Analysis extreme events, off-site sediment detachment from the slope primarily originates from the sides of the bottom most ridge, and since the horizontal slope is modeled at 0%, concentrated flow in ri l ls is not expected to occur. Hence, soil loss from 0% contoured slopes are calculated as the result o f interill detachment o f the bottom most ridge, since changes in the interill erodibility parameter resulted in detachment. Overall , W E P P was successful in determining annual runoff for the Abbotsford erosion plots, and runoff predictions were improved with adjustments in the baseline conductivity parameter, which improved the efficiency coefficients to an acceptable level. Annual soil loss predictions are still not effective since the model over-predicted soil loss for plots 1-4, and in most cases predicted no soil loss for plots 5 and 6. So i l loss for the cover-crop plots was grossly over-predicting, despite similarities in runoff, but this can be attributed to either the large magnitude of error between annual measured and predicted soil loss experienced in the first two years before the establishment of the cover crop or it suggests that doubling the planting density cannot adequetly simulate the effect of cover-cropping. 6.3.2 EVENT ANALYSIS Similar ly to the Beaverlodge event analysis, the Abbotsford event analysis investigated qualitative and quantitative comparisons of individual measured and predicted runoff and soil loss events. Since the Lower Fraser Va l l ey is not extensively subjected to winter snowmelt, the winter hydrology component is not expected to influence soil loss predictions to the extent it influenced events for the Beaverlodge study. The results obtained from the event analysis were compared on a 1:1 line for the individual plots and for al l events in a particular calibration. Measured and predicted soil losses were also examined temporally for plots 1, 2 and 5. In addition, every soil loss and runoff events for ail plots were combined and compared to examine overall model event comparisons, and a cumulative frequency distributions were applied to examine the distributions of measured and predicted run off and soil loss. Final ly , the model was subjected to a few selected individual events to compare predictions resulting from a W E P P ' s single storm simulation Page 101 Chapter 6 Model Testing and Analysis for small to large events. The efficiency and determination coefficients for al l plots resulting from event comparisons of the unmodified and calibrated data are presented in Tables 6.18 to 6.22. Event comparisons were first compared according to the unmodified, or original W E P P calculated Kb, xc, Kr and Kt parameters. The model efficiencies obtained from the yearly analysis compared much better than those obtained for the event analysis as presented in Table 6.18. Table 6.16 Statistical summary of calculated soil loss (SL) (T ha'1) and runoff (RO) (mm) for Abbotsford-Uncalibrated data, based on event data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 S L R O S L R O S L R O S L R O S L R O S L R O fl2:-317.3 -0.1 -29.1 -0.3 -519.0 -0.9 -29.8 -1.2 -0.4 -0.2 -0.4 -0.1 r2: 0.0% 27.8% 0.9% 24.3% 0.2% 8.6% 1.1% 5.9% N / A 7.3% N / A 10.2% Cumulative fl2 for : So i l Loss -43.90 Runoff -0.39 The negative runoff efficiency coefficients in Table 6.16 for al l plots indicate poor model performance, but since they are not highly negative, calibrations may improve efficiencies. A comparison of event observed verses event predicted runoff calculated with the unmodified parameters is presented in Figure 6.31. Plotl Plot 2 Plot 3 0 50 100 Measured Runoff (mm) Plot 4 50 100 150 Measured Runoff (mm) 0 50 100 Measured Runoff (mm) Plot 5 0 20 40 60 Measured Runoff (mm) 1 150 O / 100 a S, S 50 c9© °/o •o T ^ s f O £ ( °- (X K o 8 o 50 100 150 Measured Runoff (mm) Plot 6 O <§>, 0 20 40 60 Measured Runoff (mm) Figure 6.31 1:1 comparison of measured and predicted event runoff (mm) at Abbotsford using W E P P calculated Kb, rc, Kr and Kt Page 102 Chapter 6 Model Testing and Analysis F r o m the Figure 6.31, there is observable agreement between measured and predicted runoff for plots 1-4 since measured and predicted runoff appear to be oriented around the 1:1 line. The actual runoff correlation coefficient (r) averages 4 1 % for plots 1-4 and 30% for plots 5 and 6. A large portion of runoff comparisons for plots 5 and 6 are oriented below the 1:1 line which indicate a tendency for the model to under-predict runoff for contoured plots. Very poor correlation was observed with soil loss predicted for the Abbotsford plots, as can be observed in Figure 6.32. Plot l Measured Soil Loss (t/ha) Plot 2 Measured Soil Loss (t/ha) Plot 3 0 10 20 Measured Soil Loss (t/ha) Plot 4 Plot 5 Plot 6 Measured Soil Loss (t/ha) Measured Soil Loss (t/ha) Measured Soil Loss (t/ha) Figure 6.32 1:1 comparison of measured and predicted event soil loss (t ha"1) at Abbotsford using W E P P calculated Kb, tc, Kr and Kt Page 103 Chapter 6 Model Testing and Analysis For plots 1-4, the model overpredicted soil loss, by an order of magnitude (Table 6.18). For the cross-slope plots, the model predicted zero soil loss when soil loss was expected to occur, which was also apparent with the yearly analysis. Correlations were not evident between measured and predicted soil loss for al l six plots. F r o m Figure 6.32, we can infer the model over-predicted soil loss for plots 1-4 due to the orientation of the measured and predicted soil loss along the y-axis. Conversely, for plots 5 and 6, the model under-predicted soil loss for the cross-slope plots because measured and predicted data were oriented along the x-axis. The closeness of the runoff predictions for plots 1-4 and the low association between measured and predicted soil loss suggests that W E P P ' s infdtration component is effective in determining runoff loss, but the soil erosion component may not be capable of predicting soil loss accurately. Temporal distributions of event measured and predicted runoff and soil loss for plots 1, 2 and 5, to represent a cover-crop plot, control plot and a across-slope plot, respectively, are presented in Figures 6.33-6.55. Page 104 Chapter 6 Model Testing and Analysis Figure 6.33 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Abbotsford plot 1 using WEPP calculated Kb, rc, Kr and Kt Figure 6.34 Temporal distribution of runoff (mm) and soil loss (t ha"1) for Abbotsford plot 2 using WEPP calculated Kb, rc, Kr and Kt Runoff events for plots 1 and 2 appear to be associated from the beginning of the simulation period to the middle of 1992. Smaller runoff events measured around 1993 were not reflected in model predictions. The high soil losses predicted for plot 1 and 2 did not coincide with measured values, and the relative magnitudes are apparent on Figures 6.33 and 6.34 due to the equivalent y-axis representing soil loss. T w o record high predicted soil losses for plot 1 o f 15.3 t ha"1 predicted on September 20, 1990 and 14.1 t ha"1 predicted on January 15, 1991, did not coincide with measured soil losses of 0.10 and 0.01 ha"1 for the same dates, respectively. Page 105 Chapter 6 Model Testing and Analysis Total measured runoff for plot 5 was 5.6 times greater than predicted runoff for the entire simulation period. One possible reason for the under-prediction is the model does not consider flow originating from the entire slope, but rather from the bottom most slope, since runoff and soil loss from upper slopes gets deposited onto lower slopes, resulting in small changes of sediment and runoff leaving the slope profile. There was no predicted soil loss, from the original unmodified data, which is also apparent from the fourth graph on Figure 6.35. It can also be interpreted that soil loss from the contour plots are negligible, since the maximum measured soil loss was only 150 kg ha"1 on September 26, 1992. Figure 6.35 Temporal distribution of runoff (mm) and soil loss (t ha"1) for Abbotsford plot 5 using W E P P calculated Kb, Tc, Kr and A-, Overal l temporal distributions of measured runoff appear to exhibit similar relationships with predicted runoff, but may differ in magnitude since the model tended to under-predict runoff without any calibrated data. There appears to be almost no correlation between measured and predicted soil loss, especially since model predicted soil losses were on average 5 times greater than measured values. The poor spoil loss efficiency coefficients determined from the event data indicated very weak soil loss predictions which can be attributed to the erosion component. Calibrations of the critical soil parameters are expected to improve predictions. A plot o f cumulative measured and runoff predicted with the unmodified data is presented in Figure 6.50. Page 106 Chapter 6 Model Testing and Analysis Plot 1 Plot 2 Plot 3 Figure 6.36 Comparison of measured and predicted cumulative runoff (mm) at Abbotsford using W E P P calculated Kb, tc, Kr and Kt From Figure 6.36, cumulative runoff for plots 1 and 2 appear to have a moderate correlation, since on closer examination the model predicts in a similar manner, but is under-predicting by what appears to be a scaling factor. Cumulative runoff for plots 3 and 4 appear to be correlated suggesting similar trends in measured and predicted runoff. For the contour plots 5 and 6, the under-prediction is evident due to the large gap in-between measured and predicted soil loss. Cumulative soil loss predicted from the unmodified parameters are presented in Figure 6.37. Page 107 Chapter 6 Model Testing and Analysis Figure 6.37 Comparison of measured and predicted cumulative soil loss (t ha') at Abbotsford using W E P P calculated Kb, t„ Kr and Kt F r o m Figure 6.37, it is apparent that W E P P over-predicted soil losses for plots 1-4, and there is no observable correlation between the measured and predicted data. For the across slope plots, there was no soil loss predicted, resulting in a gross under-prediction of soil loss. Overall event comparisons o f runoff and soil loss for the combined Abbotsford events, for all plots are presented in Figure 6.38. Runoff Comparison - All Plots Soil Loss Comparison - All Plots | Measured Runoff (mm) Measured Soil Loss (t/ha) | Figure 6.38 Comparison of measured and predicted runoff (mm) and soil loss (t ha'1) for all Abbotsford events using W E P P calculated Kb, T c , Kr and From Figure 6.38, there is observable association between measured and predicted runoff. The overall runoff and soil loss efficiencies for the unmodified data is -0.39 and -43.9 respectively. Quantitatively, this Page 108 Chapter 6 Model Testing and Analysis implies that average runoff event runoff and soil loss is a better indicator of expected event runoff and soil loss. Improvements from parameter calibrations are expected to improve runoff and soil loss efficiencies, as that which was observed with the yearly data. Cumulative frequency distributions of runoff presented in Figure 6.53 suggest similarities in runoff predictions since both measured and predicted curves have similar distributions. This implies that the model proportionately predicted events in the same manner as with observed events. One observable difference is with the high magnitude runoff. The model's highest predicted runoff event was approximately 110 mm, while the highest observed runoff was approximately 70 mm. Runoff - All Plots Soil Loss - All Plots | Runoff (mm) Soil Loss (t/ha) | Figure 6.39 Cumulative frequency distribution of measured and predicted runoff (mm) and soil loss (t ha') for all Abbotsford events using W E P P calculated Kb, x„ Kr and Kt The right shift o f the cumulative frequency distribution curve for predicted soil loss in Figure 6.39 implies that the model generally over-predicted soil loss. This complies with the predicted soil loss values found with the yearly and event based comparison of measured and predicted soil loss. The gross over-predictions of soil loss suggests inaccuracies in computed detachment. Due to this, soil loss predictions calculated wi th W E P P calculated soil parameters must be applied with caution. Calibration o f parameters influencing soil loss may improve soil loss predictions. Runoff and soil loss event efficiency and determination coefficients based on event data calculated with the calibrated Kb are presented in Table 6.17. Table 6.17 Statistical summary of calibrated soil loss (SL) (T ha"1) and runoff (RO) (mm) for Abbotsford-Calibrated Kh, based on event data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Page 109 Chapter 6 Model Testing and Analysis S L R O S L R O S L R O S L R O S L R O S L R O R2 : -835.3 ^03 ^609 -0.6 -595.2 To 39~I T5 ^OA 0 0 ^04 OO r2: 0.0% 31.8% 0.8% 27.0% 0.1% 9.0% 1.1% 6.9% N / A 7.4% N / A 6.4% Cumulative R2 for : So i l Loss -76.46 Runoff -0.48 Baseline conductivity calibration decreased the efficiency coefficients for plots 1-4, but slightly improved efficiency coefficients for plots 5 and 6 (Table 6.17). Runoff efficiency coefficients were expected to improve with the calibration of Kb for plots 1-4 but this was not the case. The 0.0 runoff efficiency results for the contour plots suggests that the model predictions are no better than using mean values for the measured events. However, the calibrated Kb values (Table 5.5) used to obtain a minimal deviation between measured and predicted runoff is unrealistic for agricultural conditions. A comparison of event measured runoff calculated with the calibrated Kb is presented in Figures 6.40. Plot l Plot 2 Plot 3 0 50 100 150 0 20 40 60 0 20 40 60 | Measured Runoff (mm) Measured Runoff (mm) Measured Runoff (mm) | Figure 6.40 1:1 comparison of measured and predicted event runoff (mm) at Abbotsford using calibrated Kb and WEPP calculated re, Kr and Kt There was no observable difference in event predicted runoff and soil loss for plots 1-4. This is further verified by changes in the efficiency coefficients for these plots since the decrease in efficiency decreased by an average of -0.23 from -0.63. There are observable shifts in for runoff towards the 1:1 line for the contour plots 4 and 5 in comparison to Figure 6.31. Although the Kb calibrations were not successful in Page 110 Chapter 6 Model Testing and Analysis minimiz ing the overall runoff deviation between measured and predicted values to zero percent, the best calibrated parameters caused an increase in event runoff predictions since measured and predicted runoff seem to converge towards the 1:1 line. Due to the increase in runoff, predicted soil loss also increased and this further decreased the efficiency coefficients (Table 6.17). The maximum predicted soil loss for plots 1 and 2 increased from approximately 15 and 21 t ha"1 to approximately 30 and 34 t ha"1 respectively. Overall , soil losses are not correlated since soil loss due to the low efficiency coefficients in Table 6.17. Overall soil loss efficiency decreased from -43.9 to -75.5, indicating an even further worsening of model efficiency. A comparison of cumulative runoff calculated with the calibrated baseline conductivity is presented in Figure 6.41. Plot 1 Plot 2 Plot 3 90 91 92 93 94 90 91 92 93 94 90 91 92 93 94 Figure 6.41 Comparison of measured and predicted cumulative runoff (mm) at Abbotsford using calibrated Kb and WEPP calculated Tc, Kr and Kt Cumulative runoff for plots 1-3 appear to exhibit strong association, due to the observable closeness of fit despite the low efficiency coefficients, but the average runoff correlation coefficient for these plots is 48%. plot 4 exhibits a moderate association due to the voids associated between measured and predicted runoff Page 111 Chapter 6 Model Testing and Analysis from 1991-1994. The association between cumulative measured and predicted runoff for plots 5 and 6 appear weak. Event comparisons from the calibrated Kb and T c are presented in Table 6.18. Table 6.18 Statistical summary of calibrated soil loss (SL) (T ha'1) and runoff (RO) (mm) for Abbotsford-Calibrated Kb and rc, based on event data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 S L R O S L R O S L R O S L R O S L R O S L R O R2: -17.4 -0.3 -1.6 -0.6 -13.9 -1.0 -1.5 -1.5 -0.4 0.0 -0.4 0.0 r2: 0.3% 31.8% 2.1% 27.0% 0.5% 9.0% 2.8% 6.9% N / A 7.4% N / A 6.4% Cumulative R2 f o r : So i l Loss -1.93 Runoff -0.48 Only plots 2 and 4 attained a minimal deviation is soil loss of 0%, and there were significant improvements in efficiencies for plots 1-4. Changes in rc did not impact soil loss calculations for plots 5 and 6, hence soil loss efficiency coefficients for these plots remained the same. Although changes in T c improved soil loss calculations, the coefficient of determination and correlation were still low with an average of 1.4% (R2) and 12% (t2). The effects of the soil loss calibrations based on the modifying the Kb and TC is presented in Figure 6.42. Page 112 Chapter 6 Model Testing and Analysis Plot 1 Plot 2 Plot 3 | Measured Soil Loss (t/ha) Measured Soil Loss (t/ha) Measured Soil Loss (t/ha) | Figure 6.42 1:1 comparison of measured and predicted event soil loss (t ha"1) at Abbotsford using calibrated Kb and Tc and W E P P calculated Kr and Kt From Figure 6.42, improvements are visible since there appears to be a small shift towards the 1:1 line for plots 1-4, when compared to Figure 6.32, however the shift is small and it does not reflect on any observable improvements in soil loss predictions, since the efficiency coefficients still remain at less than zero. This implies inadequate model performance in predicting event-based soil loss, despite the calibrations. Effects o f the calibrated Kb and rc on temporal distributions o f runoff and soil loss for plots 1, 2 and 5 can be observed in Figures 6.43, 6.44 and 6.45, respectively. Page 113 Chapter 6 Model Testing and Analysis Plot 1 100 g 100 3 g 0 ^ 2 00 £ o o J a 2 o 00 •s o -III Ul J l , l . 90 91 92 93 94 JlL III! ., 1 ll 1, . . , .1 90 91 92 93 94 i . . . i . . . 1, in 90 91 92 93 94 ]],l„ III 1 1 90 91 92 93 94 Figure 6.43 Temporal distribution of runoff (mm) and soil loss (t ha"1) for Abbotsford plot 1 using calibrated Kb and rc and W E P P calculated Kr and Kt Figure 6.44 Temporal distribution of runoff (mm) and soil loss (t ha 1 ) for Abbotsford plot 2 using calibrated Kb and % and W E P P calculated Kr and Kt F r o m Figure 6.43, there are no visible improvements in soil loss predictions, based on the calibrated Kb and Tc parameters for plot 1. Alterations in the Tc increased soil loss predictions, but there are no temporal correlation's between the increased predicted soil loss and the measured soil loss. This can also be indicated by the weak association between measured and predicted soil losses observed in Figure 6.42. From the 1:1 soil loss comparison figures and the temporal distributions of predicted soil loss, it can be generally stated that soil loss is predicted to occur when no soil loss is measured and when soil loss is observed, no soil loss is predicted to occur. Alterations in the Kb and rc did not effect soil loss predicted to occur during the simulation of plot 5 for the contoured plots since soil loss was not (Figure 6.45). This could be reasoned as stated Page 114 Chapter 6 Model Testing and Analysis previously that soil erosion for the contour plots is subjected to r i l l erosion only, therefore, changes in the Kb and TC w i l l not effect soil loss predictions Cumulative distributions of soil loss based on calibrated Kb and TC can be viewed in Figure 6.46. Figure 6.45 Temporal distribution of runoff (mm) and soil loss (t ha'1) for Abbotsford plot 5 using calibrated Kb and Tc and W E P P calculated Kr and AT, Page 115 Chapter 6 Model Testing and Analysis Figure 6.46 Comparison of measured and predicted cumulative soil loss (t ha"1) at Abbotsford using calibrated Kb and Tc and WEPP calculated Kr and Kt From Figure 6.46, there is no observable agreement between cumulated measured and predicted soil loss for al l the plots despite successful convergence between total measured and predicted soil loss for the control plots 2 and 4. The large voids found between the measured and predicted curves are indicative of weak model soil loss predictions, despite similarities in runoff predictions. A plot o f a l l the measured and simulated runoff and soil loss events calculated with modified Kb and xc parameters are presented in Figure 6.47. Page 116 Chapter 6 Model Testing and Analysis Runoff Comparison - All Plots Soil Loss Comparison - All Plots Measured Runoff (mm) Measured Soil Loss (t/ha) Figure 6.47 Comparison of measured and predicted runoff (mm) and soil loss (t ha"1) for all Abbotsford events using calibrated Kb and tc W E P P calculated Kr and Kt Since alterations in the Tc parameter did not effect runoff, the runoff comparisons are primarily the result of the calibrated Kb parameter. F rom Figure 6.47, there is observable correlation between measured and predicted runoff, however, little association exists between measured and predicted soil loss despite the calibrations, and an improvement in soil loss efficiency from -76.5 with the Kb calibrated data to -1.93 with the Kb and %c calibrated data. A cumulative frequency distribution based on calibration of the Kb and Tc parameters is presented in Figure 6.48. Runoff - All Plots Soil Loss - All Plots | Runoff (mm) Soil Loss (t/ha) | Figure 6.48 Cumulative frequency distribution of measured and predicted runoff (mm) and soil loss (t ha"1) for all Abbotsford events using calibrated Kb and % and W E P P calculated Kr and The closeness of fit o f the runoff curves means that the model, predicts runoff volumes in relatively same proportions as that which was observed in the field. The slight right shift of predicted runoff cumulative Page 117 Chapter 6 Model Testing and Analysis frequency distribution curve for predicted runoff volumes of less than 20 mm, suggests that the model over-predicts runoff for smaller events, but high model runoff predictions coincide with measured values with the calibrated Kb parameter. It is evident from previous calculations and from Figure 6.48, that the model tends to over predict soil loss. Improvements in soil loss efficiencies were sought by calibrations of the r i l l (Kr) and interrill (Kj) erodibility parameters. The results of the event comparisons are presented in Tables 6.19 and 6.20. Table 6.19 Statistical summary of calibrated soil loss (SL) (T ha"1) and runoff (RO) (mm) for Abbotsford-Calibrated Kb and K„ based on event data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 S L R O S L R O S L R O S L R O S L R O S L R O R2: -17.4 -0.3 -1.6 -0.6 -14.0 -1.0 -1.5 -1.5 -0.4 0.0 -0.4 0.0 r2: 0.3% 31.8% 2.0% 27.0% 0.5% 9.0% 2.3% 6.9% N / A 7.4% N / A 6.4% Cumulative R2 f o r : So i l Loss -1.94 Runoff -0.48 Modifications in the Kr parameter did not return successful calibrations (Table 6.16) for al l plots, but this parameter behaved similarly to the sheer stress parameter as discussed previously. F rom the calibrated Kb and Tc, a maximum soil loss of 4.3 t ha"1 was predicted on September 20, 1990 and the same amount of soil loss was predicted for the same period with the modified Kb and Kr. Plot 3 was the only plot with differences in soil loss predicted from modified Kb and TC and modified Kb and Kr. The final calibration involved modification of Kb and Kj. Since this parameter is not particularly sensitive, no plots returned successful calibrations. The results of the event comparisons for this calibration are presented in Table 6.20. Page 118 Chapter 6 Model Testing and Analysis Table 6.20 Statistical summary of calibrated soil loss (SL) (T ha"1) and runoff (RO) (mm) for Abbotsford-Calibrated Kb and Kt, based on event data Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 S L R O S L R O S L R O S L R O S L R O S L R O R2: -684.7 -0.3 -48.6 -0.6 -487.7 -1.0 -30.9 -1.5 -0.5 0.0 -0.5 0.0 r2: 0 .1% 31.8% 0.6% 27.0% 0.1% 9.0% 0.8% 6.9% 0.1% 7.4% 0.1% 6.4% Cumulative ft2 f o r : So i l Loss -61.63 Runoff -0.48 A s expected, adjustments in the Kt parameter did not influence soil loss to a great extent, but soil loss was predicted for the contour plots 4 and 5 when soil loss was not predicted to occur with previous calibrations. Only 0.1 t ha"1 of soil loss was predicted to occur on September 20, 1990, with adjustments made to the Kr parameter. 6.3.3 BREAK-POINT ANALYSIS The model was tested with break-point data to examine its effectiveness in predicting soil loss and runoff for six short-term periods. For previous calibrations and analysis, the precipitation characteristics were limited to a single peak hydrograph where only one maximum rainfall intensity can be defined. Therefore, it did not account for temporal variability in rainfall intensity which can greatly influence the infiltration component. S ix events were selected based on their relative contribution of rainfall since the previous sampling date. This was done to ensure that maximum amount rainfall for a between sampling periods was accounted for. The selected event dates and their characteristics are listed in Table 6.21. Table 6.21 Precipitation characteristics for selected break-point events Total Precipitation Duration Proportion of precipitation Event Date (mm) (days) since previous event 1 November 10, 1990 153 3 100.0% 2 December 5, 1990 70.6 3. 100.0% 3 November 9, 1990 146.8 3 87.2% 4 February 21, 1992 41 1 100.0% 5 January 26, 1993 97.7 3 96.7% 6 January 28. 1993 7.4 2 100.0% Page 119 Chapter 6 Model Testing and Analysis Precipitation dates selected represent a range of high to low magnitude rainfalls (Table 6.21). Since WEPP cannot execute break-point data for storms with durations greater than 24 hours, each day for a particular event had break-point precipitation data extracted and inputted into a continuous simulation storm file. Hydrographs for the selected events can be viewed in Figure 6.49. Event 1: November 10, 1989 10 & 5 a 0 0 10 20 30 40 50 60 Duration (h) Event 3: November 9, 1990 a 0 20 40 60 Duration (h) Event 5: January 26, 1993 J L L L M - n - n 20 40 Duration (h) 60 H O 3 0 " 0 c 0 Event 2: December 5, 1989 LUI I — r w T mi 20 40 60 Duration (h) Event 4: Februaury 21, 1992 5 10 15 Duration (h) Event 6: January 28, 1993 20 10 15 20 25 Duration (h) Figure 6.49 Break-point hydrographs for selected events Since the initial conditions for break-point events are critical to model output, individual initial condition files for all plots and every event (36 initial condition files in total) were developed by executing the model for a long term continuous simulation until the day prior to the break-point simulation events (Table 6.21). To examine the effectiveness of the break-point verses single peak method, model results were analyzed for soil loss and runoff efficiency on a plot basis for the break-point and single peak events with the unmodified and calibrated Kb and Tc data in Table 6.22. Page 120 Chapter 6 Model Testing and Analysis Table 6.22 Statistical summary of soil loss (SL) and runoff (RO) efficiency for single peak and break point data for selected events P l o t l Plot 2 Plot 3 Plot 4 Plot 5 S L R O S L R O S L R O S L R O S L R O Plot 6 S L R O R2 -509.3 Original uncalibrated data - daily single peak hydrograph 0.61 -198.9 0.71 -3068 0.67 -356.1 0.30 -0.61 -0.61 -0.61 -0.43 R2 -54.7 Calibrated Kb and % data - daily single peak hydrograph 0.56 -15.4 0.27 -180 0.64 -16.7 -0.01 -0.61 0.02 -0.61 0.03 R2 -447.9 Original uncalibrated data - break point -0.60 -173.5 -0.65 -2366 -0.46 -214.1 -0.62 -0.61 -0.69 -0.61 -0.50 R2 -12.1 Calibrated Kb and Tc data - break point -0.48 -4.3 -0.94 -34 -0.42 -4.3 -0.93 -0.61 0.07 -0.61 0.09 The efficiency coefficients presented in Table 6.22 represent the efficiency of the model between observed and predicted runoff and soil loss based on all six events. It is interesting to note that the best runoff efficiencies for plots 1-4 were obtained with the single peak hydrograph from the long term simulation, since all the efficiency coefficients were positive indicating proficient model performance. Since calibrations were based on total runoff and soil loss for the entire simulation period, the runoff efficiencies obtained with the modified Kb and Tc parameters did not reflect an improvement in runoff efficiencies for plots 1-4. However, it did improve soil loss efficiencies, but soil loss efficiencies were st i l l less than zero. Runoff efficiencies obtained with the break-point data were expected to improve in comparison to the single-peak hydrograph data but were less than zero, with the exception of cross-slope plots 5 and 6, but as stated earlier, the Kb adjustments made to plots 5 and 6 are impractical. So i l loss efficiencies obtained from the break point data with the calibrated Kb and xc parameters showed the greatest improvements due to the low magnitude of the soil loss efficiencies, but since the efficiency values for plots 1-4 were less than zero, this indicates inadequate model performance. The efficiency results obtained from the unmodified parameters for plots 1-4 with the single peak hydrograph are encouraging, and it suggests that the model can be applied to predict runoff losses with no parameter modifications. Year ly runoff predictions for the calibrated parameters returned positive efficiency coefficients, which implies that W E P P model can adequately predict yearly runoff. Event comparisons produced negative runoff efficiency coefficients, suggesting that the model may not adequately predict event based runoff. However, the results obtained from the break-point analysis suggests that the model accurately predict Page 121 Chapter 6 Model Testing and Analysis runoff for short periods (few days) by using single peak hydrograph data in continuous simulation mode. So i l loss predictions for both event and yearly analysis were poor, due to the negative efficiency coefficients determined for both. In general, W E P P tended to predict runoff events at an occurrence at the measured occurance due to the closeness of fit o f the cumulative frequency distributions of soil loss. However, but the model was less successful in determining soil loss since it tended to overpredict soil loss for plots 1-4. Page 122 Chapter 7 Conclusion C H A P T E R 7 : C O N C L U S I O N The overall objective of this study was to determine the effectiveness o f the Water Erosion Prediction Project soil erosion model in predicting soil loss and runoff for two distinctive climatic regions in Western Canada. The model was assessed by comparing total, annual and event based runoff and soil loss predictions with measured runoff and soil loss. In addition, snowmelt and rain fall induced events were also assessed and the model was tested with break-point hydrograph data. In conclusion, the results of this study have indicated that: 1. So i l loss predictions cannot be confidently accepted even with calibrated critical sheer stress and r i l l erodibility parameters. 2. Annual runoff predictions can be confidently predicted for hillslopes in climates which are not dramatically effected by severe winters, also W E P P can confidently predict runoff for short-term events without any modifications to soil parameters. 3. Improvements need to be made to the winter hydrology components since the model over-predicts soil loss during the winter snowmelt, and to the crop growth component since the model cannot accurately simulate above ground biomass production for perennial crops. 4. The W E P P model responds accordingly to changes in cultural practices, which allows it to simulate distinctive scenarios not previously allowable with other erosion models. However i f the model is to be applied confidently, the model needs to accurately predict soil loss and runoff with minimal parameter modification. The degradation of soil is a synthesis o f many naturally occurring factors and soil erosion in nature is a highly variable and complex process. Model ing of this natural phenomenon requires an extensive understanding the processes and interactions of the natural system, and the comprehensive nature of W E P P Page 123 Chapter 7 Conclusion is an attempt to model the individual rationalized natural influences effecting soil erosion. The W E P P is an extensive physically based model, which considers many processes and interactions fundamental to the soi l erosion process based on a mathematical understanding of erosion, hydrology and plant physiology. When addressing the issue of model acceptability and applicability, practical considerations must also be addressed such as accuracy o f predictions, ease of use and availability of required input data. If the model is to be used as an effective planning tool, it needs to produce accurate results with minimal modifications, since accurate calibrations can only be done for sites with measured data. Further recommendations are suggested to effectively transfer this technology. These recommendations include: 1. Improvements to the winter hydrology routines are needed, especially i f the model is to be excepted for Canadian conditions. 2. Crop parameterization needs to be simplified or a more robust crop growth database is needed. 3. Modif icat ion to the crop growth component is essential so the model can correctly simulate above ground biomass production for perennial crops. 4. M o r e long-term evaluations are required based on measured soil loss and runoff data from natural, agricultural and forest ecosystems. 5. Off-site sediment characteristics have to be evaluated to assess model accuracy in predicting fine sediments, which greatly impacts water quality. 6. A n additional study investigating soil loss and runoff under specific conditions based on extreme events determined from a flood frequency analysis, to assess maximum probable soil loss for a determined design storm. I f adjustments in the model can improve predictions, the W E P P model can potentially be a powerful tool for use with erosion conservation planning and impact assessment for Canadian and global conditions. Page 124 Bibliography B I B L I O G R A P H Y Aitken , A . P . 1973. Assessing systematic errors in rainfall-runoff models. J . of Hydrology, 20: 121-126 Alberts, E . E . , M . A . Nearing, M . A . Nearing, M . A . Weltz . L . M . Risse, F . B . Pierson, X . C . Zhang, J . M . Laflen and J.R. Simanton. 1995. Soi l component. In: U S D A - Water Erosion Prediction Project: Hi l ls lope profile and watershed model documentation. N S E R L Report N o . 10. U S D A - A R S National So i l Erosion Research Laboratory, West Lafayette, I N . Albright , W . D . , 1939. The menace of water erosion in the Peace. Scientific A g r i c , 19(5): 241-248. Arno ld , J .G . , M . A . Wel tz , E . E . Alberts and D . C . Flanagan. 1995. Plant growth component. In: U S D A -Water Erosion Prediction Project: Hil ls lope profile and watershed model documentation. N S E R L Report N o . 10. U S D A - A R S National So i l Erosion Research Laboratory, West Lafayette, I N . A S C E Task Committee on Definition of Criteria for Evaluation of Watershed Models . 1993. Criteria for evaluation of watershed models. Watershed Management Committee, Irrigation and Drainage Div i s ion . J . o f Irrig. and Drain. Eng. , 119(3), 429-442. Bingner, R . L . , C . K . Mutchler arid C E . Murphree. 1992. Predictive capabilities of erosion models for different storm sizes. T r a n s - A S A E , 35(2): 505-513. Boardman, J . , J . A . Dearing and I .D .L . Foster. 1990. So i l erosion studies; some assessments. In: So i l erosion on agricultural land. J . Boardman, I .D .L . Foster and J .A . Dearing, Eds. John W i l e y and Sons. pp. 660-672. Bonta, J . V . and R. Rao. 1992. Estimating peak flows from small agricultural watersheds. J . o f Irrig. and Drain. Eng. , 118(1) 122-137. Chaves, H . M . L . and M . A . Nearing. 1991. Uncertainty analysis o f the W E P P soil erosion model. Trans-A S A E , 34(6) 2437- 2444. Chu , S.T. 1978. Infiltration during an unsteady rain. Water Resources Res., 14(3) 461-466. Clark II, E . H . 1985. The off-site costs of soil erosion. J . of So i l and Water Cons., 40(1) 19-22. Coote, D .R . , J . Dumanski and J.F. Ramsey, Editors. 1981. A n assessment of the degradation of agricultural lands in Canada. Land Resource Research Institute Contribution No . 118. Research Branch, Agriculture Canada, Ottawa, Ont. Edwards, D . R . , V . W . Benson, J.R. Wi l l i ams , T . C . Daniel , J . Lemunyon and R . G . Gilbert. 1994. Use of the E P I C model to predict runoff transport of surface applied inorganic fertilizer and poultry manure constituents. T r a n s - A S A E , 37(2): 403-409. El l io t , W . J . , R . B . Fo lz , P .R . Robichaud and C . H . Luce. 1994. The experience of the U S D A Forest Service with W E P P . In: Current and Emerging Erosion Prediction Technology Symposium - Extended Abstracts. A p r i l 10-11, Norfolk Vi rg in ia . Flanagan, D . C . 1993. Evaluation of the W E P P deposition component. A S A E Paper N o . 93-2107. Flanagan, D . C (Editor). 1994. Water Erosion Prediction Project Erosion Prediction M o d e l : Vers ion 94.7 User Summary. N S E R L Report N o . 9. U S D A - A R S National Soi l Erosion Prediction Laboratory, West Lafayette, Indiana. Page 125 Bibliography Flanagan, D . C . and S J . Livingston, Editors. 1995. W E P P user summary. N S E R L Research Report N o . 11. U S D A - A R S National So i l Erosion Research Laboratory, West Lafayette, I N . Flanagan, D . C , J .C. Ascough II, M . A . Nearing and J . M . Laflen. 1994. The Water Erosion Prediction Project erosion model - A powerful new tool for conservation planning. In: Current and Emerging Erosion Prediction Technology Symposium - Extended Abstracts. A p r i l 10-11, Norfolk Vi rg in ia . Foster, G .R . , D . C . Flanagan, M . A . Nearing, L . J . Lane, L . M . Risse and S.C. Finkner. 1995. Hi l l s lope erosion component. In: U S D A - Water Erosion Prediction Project: Hil ls lope profde and watershed model documentation. N S E R L Report N o . 10. U S D A - A R S National So i l Erosion Research Laboratory, West Lafayette, I N . Ghidey, F . and E . E . Alberts. 1990. Residue decomposition and management component of W E P P . A S A E Paper N o . 90-2556. Green, W . H . and G . Ampt . 1911. Studies in soil physics. I. The flow of air and water through soils. J . of Agr i c . Sc i . , 4:1-24. Harris, T and J . Boardman. 1990. A rule based expert system approach to predicting waterborne soil erosion. In: So i l erosion on agricultural land. J. Boardman, I .D .L . Foster and J . A . Dearing, Eds. John W i l e y and Sons. pp. 401-412. Hayhoe, H . N . , R . G . Pelletier and L . J .P . van Vl ie t . 1993. Estimation of a snowmelt runoff in the Peace River region using a soil moisture budget. Can. J. of So i l Sc i . , 73: 489-501. Hogg, R . V and J . Ledolter. 1987. Engineering statistics. Macmi l l an Publishing Company, N e w Y o r k . Huang, C , and J . M . Bradford. 1993. Analysis of slope and runoff factors based on the W E P P erosion model. So i l Sc i . A m . J. , 57: 1176-1183. Izaurralde, C , S. Nolan, A . Jedrych, D . Purveen, D . Vandelwel , T. Goddard, J. Tajek and P. Dz ikowsk i . 1994. Testing W E P P and E P I C on Hillslopes Using Alberta Erosion Data. A Progress Report to C A E S A -Soi l Quality Committee. Jedrych, A . T . , C R . Wright and D . S . Vanderwel. 1995. C A E S A - So i l quality water erosion report. Prepared for C A E S A - So i l Quality Committee. Alberta Agriculture, Food and Rural Development. Jensen, M . E . Editor. 1974. consumptive use of water and irrigation requirements. Report by the Commis ion on Irrrigation and Water Requirements, Irrig. and Drain. D i v . A S C E . Kautza, T .J . , D . L . Schertz and G . A . Weesies. 1995. Lessons learned in R U S L E technology transfer and implementation. J . of Soi l and Water Cons., 50(5) 490-493. K i r k b y , M . J . 1980. Mode l ing water erosion processes. In: So i l erosion; Edited by M . J . K i r k b y and R . P . C . Morgan. John W i l e y and Sons L td . pp. 183-216. Kottwitz , E . R . 1995. Irrigation Component. In: U S D A - Water Erosion Prediction Project: Hi l l s lope profile and watershed model documentation. N S E R L Report N o . 10. U S D A - A R S National So i l Eros ion Research Laboratory, West Lafayette, I N . Kramer. L . J and E . E . Roberts. 1992. Frequency distributions of W E P P 92.24 predicted soil loss. A S A E Paper N o . 92-2641 Page 126 Bibliography Kul i s , Y . H . and L . P J . van Vl ie t . 1994. Water erosion effects on crop productivity in Bri t ish Columbia . Progress report. B . C . Land Resource Unit , Center for Land and Biologica l Resources Research, Agriculture Canada. Laflen, J . M . , D . C . Flanagan, J .C. Ascough II, M . A . Wel tz and J.J. Stone . 1994. The W E P P model and its applicability for predicting erosion on rangelands. in: Variabi l i ty in rangeland water erosion processes, S S S A Special Publication 38. Edited by: W . H . Blackburn, F . B . Pierson Jr., G . E . Schuman and R. Zartman. So i l Science Society of America , Inc. Madison, Wisconsin, pp. 11-22. Laflen, J . M . , L . J . Lane and F .R . Foster. 1991. W E P P : A new generation of erosion prediction technology. J . o f So i l and Water Cons., 51(1): 34-38. Laflen, J . M . , W . J . El l io t , J.R. Simanton, C S . Holzhey and K . D . K o h l . 1991. W E P P Soi l erodibility experiments for rangeland and cropland soils. J . of Soi l and Water Cons., 46(2):39-44. L a i , R . 1994. So i l erosion by wind and water: problems and prospects. In: So i l erosion research methods. Second Edi t ion. Edited by R. L a i . Soi l and Water Conservation Society, Ankeney, IO. pp. 1-10. Lane, L . J . and M . A . Nearing, Editors. 1989. U S D A Water Erosion Prediction Project: Hi l ls lope profile model documentation. N S E R L Report N o . 2, U S D A - A R S National So i l Erosion Research Laboratory, West Lafayette, Indiana. Larson, W . E . , F . J . Price and R . H . Dowdy. 1983. The threat of soil erosion to long term crop production. Science, 219:458-265. Mansoor, K . , L . J . P . van Vl ie t , M . D . Novak and S.T. Chieng. 1995. Evaluation of the W E P P hillslope erosion model for simulating soil loss and runoff for two sites in western Canada. Poster presented at the W E P P / W E P S Symposium, Des Moines , Iowa. M e i n , R . G . , and C L . Carson. 1973. Model ing infiltration during a steady rain. Water Resources Res., 9(2): 384-394. Nash, J .E . and J . V . Sutcliffe. 1970. River flow forecasting through conceptual models, part 1: Discussion of Principles. J. o f Hydrology, 10: 282-290. Nearing, M . A , L . A . Deer-Ascough, B . Y . L i u , S. Livingston and X . Zhang. 1994. Val idat ion sudies of the W E P P model - hillslope applications. In: Current and emerging erosion prediction technology symposium -Extended Abstracts. Norfolk Vi rg in ia . 65-70. Nearing, M . A . , L . Deer-Ascough and J . M . Laflen. 1990. Sensitivity analysis of the W E P P hillslope profile erosion model. T r a n s - A S A E , 33(3) 839-849. Nearing, M . A . , L . D . Ascough and H . M . L Chaves. 1989. W E P P model sensitivity analysis. In: U S D A Water Erosion Prediction Project: Hil ls lope profile model documentation. Edited by: Lane, L . J . and M . A . Nearing. N S E R L Report N o . 2, U S D A - A R S National So i l Erosion Research Laboratory, West Lafayette, Indiana. Nearing, M . A . , L . J . Lane and V . L . Lopes. 1994. Model ing soil erosion. In: So i l erosion research methods. Second Edi t ion. Edited by: R. L a i . So i l and Water Conservation Society, Ankeny, Iowa. pp. 127-158. Nicks , A . D . , L . J . Lane and G . A . Gander. 1995. Weather Generator. In: U S D A - Water Erosion Prediction Project: Hi l ls lope profile and watershed model documentation. N S E R L Report N o . 10. U S D A - A R S National So i l Erosion Research Laboratory, West Lafayette, I N . Page 127 Bibliography Novak, M . D . and L . J . P . van Vl ie t . 1983. Degradation effects of soil erosion by water and wind. In: So i l degradation in Bri t ish Columbia - Proceedings of the 8th B . C Soi l Science Workshop. B . C . Minis t ry o f Agriculture and Food, Victor ia , B . C . 46-70. Oldeman, L . R . 1992. Globa l extent of soil degradation. Bi-annual report, International So i l Reference and Information Center, Wageningen, The Netherlands: 19-36. Penman, H . L . 1963. Vegetation and hydrology. Tech. C o m . N o . 53, Commonwealth Bureau of Soils, Harpenden, England. Perrens, S.J. and N . A . Trustrum. 1985. Soi l erosion assessment for conservation policy making. J . o f So i l and Water Cons., 40(6) 491:495. Priestley, C . H . B . and R . J . Taylor. 1972. On the assesment of surface heat flux and evapotranspiration using large scale parameters. Monthly Weather Rev. 100: 81:91. Renard, K . G and V . A . Ferreira. 1993. R U S L E M o d e l Description and Database Sensitivity. J . o f Environ . Qual . , 22:258-266. Renard, K . G . , G . R . Foster, D . C . Yoder and D . K . M c C o o l . 1994. R U S L E revisited: Status, questions, answers and the future. J . of So i l and Water Cons., 49(3): 213-220. Risse, L . M . , M . A . Nearing and M . R . Savabi. 1994. Determining the Green-Ampt effective hydraulic conductivity from rainfall-runoff data for the W E P P model. T r a n s - A S A E , 37(2): 411-418. Risse, L . M . , M . A . Nearing, A . D . N icks and J . M . Laflen. 1993. Error assessment in the Universal soil loss equation. So i l Sc i . Soc. of A m . J. , 57: 825-833. Risse, L . M . , M . A . Nearing, M . R . Savabi and J . M . Laflen. 1993. Optimization of saturated hydraulic conductivity for W E P P . A S A E Paper N o . 93-2038. Risse, L . M . , M . A . Nearing and X . C . Zhang. 1995. Variabil i ty in Green-Ampt effective hydraulic conductivity under fallow conditions. J . of Hydrology, 169: 1-24. Savabi, M . R . 1993. Mode l ing subsurface drainage and surface runoff with W E P P . J. of Irrig. Drain. Eng. 119(5): 801-813. Savabi, M . R . and D . E . Stott. 1994. Plant residue impact on rainfall interception. T r a n s - A S A E , 37(4) 1093-1098. Savabi, M . R . , and J.R. Wi l l i ams . 1995. Water balance and percolation. In: U S D A - Water Erosion Prediction Project: Hil ls lope profile and watershed model documentation. N S E R L Report N o . 10. U S D A -A R S National So i l Erosion Research Laboratory, West Lafayette, I N . Savabi, M . R . , D . C . Flanagan, B . Hebel and B . A . Engel . 1995. Applicat ion of W E P P and G I S - G R A S S to a small watershed in Indiana. J . of Soi l and Water Cons., 50(5): 477-483. Savabi, M . R . , R . A . Young , G . R . Benoit, J . M . Witte and D . C . Flanagan. 1995. Winter hydrology. In: U S D A - Water Erosion Prediction Project: Hil ls lope profile and watershed model documentation. N S E R L Report N o . 10. U S D A - A R S National Soi l Erosion Research Laboratory, West Lafayette, I N . Savabi, M . R . , R . W . Skaggs and C A . Onstad. 1995. Subsurface hydrology. In: U S D A - Water Erosion Prediction Project: Hil ls lope profile and watershed model documentation. N S E R L Report N o . 10. U S D A -A R S National So i l Erosion Research Laboratory, West Lafayette, I N . Savabi, M . R . , W . J . Rawls and R . W . Knight. 1995. Water Erosion Prediction Project ( W E P P ) rangeland hydrology component evaluation on a Texas range site. J. of Range. Manage. 48(6): 535-41. Page 128 Bibliography Skaggs, R . W . 1978. A water management model for shallow water table soils. Rep. N o . 134, Water Resources Research Institute of the University of North Carolina, N C . So i l at risk. 1984. A report on soil conservation by the Standing Committee on Agriculture, Fisheries and Forestry, to the Senate of Canada. 1984. Stone, J.J., L . J . Lane, E . D . Shirley and M . Hernandez. 1995. Hil ls lope surface hydrology. In: U S D A -Water Erosion Prediction Project: Hil ls lope profile and watershed model documentation. N S E R L Report N o . 10. U S D A - A R S National So i l Erosion Research Laboratory, West Lafayette, I N . Stone, R . J . 1993. Mode l ing subsurface drainage and surface runoff with W E P P - A discussion. J . of Irrig. and Drain. Eng. , 119(5): 217-218. Storm, D . E . , B . J . Barfield and C T . Altendorf. 1994. C R E A M S / W E P P sediment deposition equation: A semitheoretical evaluation. T r a n s - A S A E , 37(4): 1105-1108. Stott, D . E . , E . E . Alberts and M . A . Weltz . 1995. Residue decomposition and management. In: U S D A -Water Erosion Prediction Project: Hil ls lope profile and watershed model documentation. N S E R L Report N o . 10. U S D A - A R S National So i l Erosion Research Laboratory, West Lafayette, I N . Tiscareno-Lopez, M . , M . A . Wel tz and V . L . Lopes. 1995. Assessing uncertainties in W E P P ' s soil erosion predictions on rangelands. J . of So i l and Water Cons., 50(5): 512-516. Tiscareno-Lopez, M . , V . L . Lopes, J.J. Stone and L . J . Lane. 1994. Sensitivity analysis o f the W E P P watershed model for rangeland applications - II. Channel Processes. T r a n s - A S A E , 37(1): 151-158. Toy , T .J . and W . R . Osterkamp. 1995. The applicability of R U S L E to geomorphic studies. J . of So i l and Water Cons., 50(5): 498-503. van Vl i e t , L . J . P . 1989. Water erosion prediction for soils in the Peace River Region of Bri t ish Columbia : Estimates using the universal soil loss equation. Technical Bul le t in 1989-3E, Land Resource Research Center Contribution 88-67. B . C . Land Resource Unit , Land Resource Research Centre, Vancouver, B . C . van Vl i e t , L . J . P . 1992. Beaverlodge and Dawson Creek soil erosion plots: soil loss and runoff data. Progress Report. B . C . Land Resource Unit , Centre for Land and Bio log ica l Resources Research, Agriculture Canada. van Vl ie t , L . J .P . and J .W. H a l l . 1991. Effects o f two crop rotations on seasonal runoff and soil loss in the Peace River region. Can. J . of Soi l Sc i . , 71: 533-544. van Vl ie t , L . J . P . and J .W. H a l l . 1995. Effects of planting direction of brussel sprouts and previous cultivation on water erosion in Southwestern Bri t ish Columbia, Canada. J. of So i l and Water Cons., 50(2) 188-192. van Vl ie t , L . J .P . , A . M . F . Henning and M . D . Novak. 1986. So i l erosion loss monitoring and prediction under semi-arid agriculture in the Peace River Region of N W Canada. In: Land evaluation for land-use planning and conservation in sloping areas. Edited by: W . Siderius. International Institue for Land reclamation and Improvement, Publication 40 - International Workshop, Enschede, the Netherlands. van Vl i e t , L . J .P . , R. K l ine and J .W. H a l l . 1993. Effects o f three tillage treatments on seasonal runoff and soil loss in the Peace River region. Can. J. of So i l Sc i . , 73: 469-480. van Vl i e t , L . J .P . . 1993. Preliminary testing of the W E P P Hil ls lope Profile erosion model with Erosion plot data from Beaverlodge, Alberta. Progress Report. B . C . Land Resource Unit , Center for Land and Bio log ica l Resources Research, Agriculture Canada. Page 129 Bibliography van Vl ie t . 1994. Evaluation of erosion control practices in strawberries in the Matsqui - Langley Upland Area of the Fraser Val ley : Final Report (1991-1994). B . C . Land Resource Unit , Centre for Land and Bio log ica l Resources Research, A g . Can. Walter, M . F . and F . N . Swader. 1976. Erosion from Agricultural Land. Agricultural Engineering Extension Bullet in 422, Department of Agricultural Engineering, N e w Y o r k State College of Agriculture and L i f e Sciences, Cornel l University, N Y . Wampole , R . E . and R . H . Myers . 1989. Probability and statistics for engineers and scientists. Fourth Edi t ion. Macmi l l an Publishing Company, N e w York . Wicks , J . M . and J .C. Bathurst. 1996. S H E S E D : a physically based, distributed erosion and sediment yield component for the S H E hydrological modeling system. J. of Hydrology, 175:213-238. Wicks , J . M . , J .C . Bathurst and C W . Johnson. 1992. Calibrating S H E soil-erosion model for different land covers. J . o f Irrig. and Drain. Eng. , 118(5) 708-732. W i l k e n , G . C 1991. Sustainable agriculture is the solution, but what is the problem. Board for International Food and Agricultural Development and Economic Cooperation ( B I F A D E C ) , Occasional Paper N o . 114. Agency For International Development, Washington, D . C . Wi l l i ams , J.R., P.T. Dyke , W . W . Fuchs, V . W . Benson, O . W . Rice and E . D . Taylor. EPIC-Eros ion / Productivity Impact Calculator. User Manual . U . S . Department of Agriculture Technical Bul le t in N o . 1768. 122pp. Wischmeier, W . H . and D . D . Smith. 1965. Predicting rainfall erosion losses. U S D A Agricultural Handbook 537. Science and Education Administration, U . S . Government Printing Office, Washington, D . C . Wischmeier, W . H . and D . D . Smith. 1965. Predicting rainfall-erosion losses from cropland east o f the Rocky Mountains - Guide for selection of practices for soil and water conservation. U S D A Agricultural Handbook 238. U . S . Government Printing Office, Washington, D . C . W u , T . H . , J . A . H a l l and A . V . Bonta. 1993. Evaluation of runoff and erosion models. J . of Irrig. and Drain. Eng. , 119(4): 364-382. Yoder , D . , and J. L o w n . 1995. The future of R U S L E : Inside the new Revised Universal So i l Loss Equation. J . o f So i l and Water Cons., 50(5) 484-489. Page 130 Appendix 1 A P P E N D I X 1 : D E T A I L E D C R O P P I N G P R A C T I C E S F O R B E A V E R L O D G E Treatments W E P P Description Depth C = Cultivated Chisel plow, straight with spike pts 10-12 c m C p = Chise l p low Chisel plow with soulters and straight chisel spike pt 10-15 c m Cs = Cultivated with shovels Chisel plow with sweeps D = Disced Disk, offset-finishing 7-9" spacing H = Harrowed Harrow-spike tooth 3 c m H S = Harvested (for use with management option) R = Rotovated Rotary tiller-secondary operation 3" deep 7.5-10 cm R A = Raked two passes with 'Harrow-spike tooth' 3 cm S = Seeded D r i l l , single disk opener (conventional) 2-3 c m T = Turnplowed Plow, Moldboard, 8" H o = Hoe Field cultivator, primary tillage-sweeps or shovels 6' Year Date P lo tO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 1981 19-May C ; D ; H C ; D ; H C ; D ; H C ; D ; H C ; D ; H C ; D ; H 20-May R A R A R A R A R A R A 21-May S S - S S -14-Aug - H S - - H S -15-Sep - C C - C C 1982 17-May - R; R A ; S R; R A - R; R A ; S R; R A 28-Jul H S - - H S - -23-Aug - H S - - - -2-Sep - - R; R A - - R ; R A 9-Sep - - - - H S -7-Oct - R - - R -1983 4 - M a y - R R - R R 12-May - R A ; S - - R A ; S -2-Aug H S - - H S - -22-Aug - - R; R A - - R; R A 23-Aug - H S - - H S -7-Oct - R - - R -1984 26-Apr 4 - M a y 10-Jul R; R A S R; R A R; R A S R; R A 2 -Aug H S - - H S - -16-Aug - H S - - H S -9-Oct C - - - - -10-Oct - R - - R -11-Oct - - Ho - - H o Page 131 Appendix 1 Year Date P lo tO Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 1985 6-May R ; S R ; S R; S R; S 4 - A u g - - H S - -13-Aug H S H S - H S H S 23-Aug - - R; R A - - -12-Sep R R - - R R 1986 12-May R ; H ; S R; H ; S R; H ; S R; H ; S R; H ; S 8-Aug - - - H S - -12-Aug H S H S H S - H S H S 21-Aug R; R A R; R A R; R A - R ; R A R ; R A 1987 5-May R R R R R 6-May S S S S S 6-Aug H S H S H S H S H S 17-Sep - R R R R 1988 28-Apr R ; S R ; S R ; S R ; S R ; S 19-Aug H S H S H S H S H S 17-Oct - C C ; C C ; C C ; C 1989 17-May R ; S R ; S R ; S R ; S R ; S 30-Aug H S H S H S H S H S 19-Sep - - C C C 28-Sep - C - - -1990 28-May R ; S R ; S R R ; S R ; S R 16-Aug - H S - - H S -17-Aug C - R; R A C - R; R A 20-Sep - C - - C -12-Oct - Cs Cs - Cs Cs 1991 8-May R A ; R ; S R ; S R R A ; R ; S R ; S R 14-Aug - H S - - H S -5-Sep - C R; R A - C R; R A 1992 4 -May R; R A R ; R A R; R A R ; R A 5-May S - S -11-Jun - R A - R A 7-Jul - R; R A - R ; R A 23-Jul H S - - H S - -10-Sep - H S - - H S -30-Sep - C R; R A - C R; R A 1993 22-Apr R; R A ; S R; R A R; R A ; S R; R A 19-Jul Hs - - Hs - -9-Sep - - R; R A - - R; R A 17-Sep - H S - - H S -1-Oct - R ; C - - R ; C -Page 132 Appendix 2 A P P E N D I X 2 : D E T A I L E D S O I L L O S S (t ha"1) A N D R U N O F F (mm) O B S E R V E D A T B E A V E R L O D G E PlotO Plotl Plot 2 Plot 3 Plot 4 Plot 5 Index D M Y SL RO SL RO SL RO SL RO SL RO SL RO 55 24 2 81 0.08 8.4 0.35 114.5 0.75 23.4 0.09 11.3 0.09 80.5 0.06 4.1 77 18 3 81 1.03 54.6 0.06 35.7 0.93 53.9 0.52 25.3 0.75 3.6 1.01 40.5 127 7 5 81 0.00 0.0 0.00 0.0 0.86 66.0 0.00 0.0 0.00 0.0 0.00 0.0 163 12 6 81 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.3 0.00 0.2 166 15 6 81 0.10 0.4 0.10 0.8 0.02 0.1 0.19 0.7 0.05 0.4 0.00 0.0 203 22 7 81 1.17 3.3 0.33 3.2 1.34 3.5 1.55 5.3 0.24 2.3 0.96 2.1 209 28 7 81 0.69 2.2 0.14 1.8 1.69 3.2 1.03 3.2 0.16 2.2 0.97 2.6 245 2 9 81 0.00 0.0 0.00 0.0 0.41 4.8 0.00 0.0 0.00 0.0 0.17 2.6 476 21 4 82 0.01 1.2 0.01 4.9 0.05 9.9 0.04 12.3 0.00 0.0 0.11 30.2 478 23 4 82 0.20 79.3 0.00 4.0 0.16 15.5 0.07 15.8 0.00 0.0 0.20 55.3 479 24 4 82 0.02 6.8 0.03 14.7 0.25 16.5 0.04 25.6 0.02 8.7 0.13 56.1 483 28 4 82 0.02 9.6 0.06 6.9 0.02 4.6 0.04 7.4 0.05 49.0 0.05 11.3 554 8 7 82 0.00 0.0 0.00 0.0 6.62 6.8 0.00 0.0 0.00 0.0 3.55 5.0 567 21 7 82 0.12 3.2 0.34 13.2 29.97 44.2 0.20 6.1 1.21 22.5 24.92 11.1 569 23 7 82 0.00 0.3 0.22 4.2 1.30 4.1 0.01 0.4 0.03 2.3 0.59 3.5 582 5 8 82 0.05 27.1 0.04 13.1 1.63 26.6 0.18 36.7 0.02 1.0 0.74 14.1 822 2 4 83 0.00 5.0 0.00 1.6 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 824 4 4 83 0.01 4.5 0.00 0.1 0.00 0.2 0.00 10.0 0.00 0.0 0.00 0.1 826 6 4 83 0.00 31.6 0.08 29.1 0.52 25.9 0.03 43.6 0.02 14.9 0.39 24.4 835 15 4 83 0.01 6.5 0.02 2.9 0.11 3.7 0.00 1.4 0.05 8.8 0.12 2.5 910 29 6 83 0.00 0.0 0.00 0.0 0.04 0.8 0.00 0.0 0.00 0.0 0.16 1.4 915 4 7 83 0.00 0.9 0.00 0.0 0.73 5.1 0.00 0.0 0.00 0.0 1.34 6.9 918 7 7 83 0.00 0.0 0.00 0.0 0.25 3.8 0.00 1.9 0.00 0.0 0.24 4.0 932 21 7 83 0.00 0.0 0.00 0.0 1.71 14.6 0.00 0.0 0.00 0.0 1.57 16.1 938 27 7 83 0.00 1.8 0.00 0.0 0.14 3.9 0.01 9.9 0.00 0.0 1.04 26.1 945 3 8 83 0.00 0.1 0.00 0.0 4.74 8.5 0.00 0.5 0.00 0.0 4.25 8.0 1122 27 1 84 0.05 88.5 0.00 0.0 0.13 88.0 0.00 0.0 0.00 0.0 0.00 0.0 1146 20 2 84 0.00 1.8 0.00 0.0 0.02 2.2 0.00 1.1 0.00 0.0 0.00 0.0 1174 19 3 84 0.01 3.5 0.00 0.0 0.00 0.0 0.00 1.5 0.00 0.0 0.00 0.0 1260 13 6 84 0.00 0.0 0.04 1.1 0.25 3.7 0.00 0.0 0.11 3.3 0.40 6.8 1288 11 7 84 0.00 0.0 0.00 0.0 0.09 0.7 0.00 0.0 0.00 0.0 0.20 1.0 1353 14 9 84 0.00 0.0 0.00 0.0 0.05 0.7 0.00 0.0 0.00 0.0 0.41 6.6 1520 28 2 85 0.00 1.8 0.01 0.7 0.01 0.7 0.01 17.4 0.00 0.0 0.07 5.0 1531 11 3 85 0.00 0.0 0.00 0.0 0.06 0.8 0.01 1.6 0.00 0.5 0.04 0.7 1534 14 3 85 0.00 0.0 0.00 0.0 0.06 1.1 0.00 1.3 0.00 0.5 0.02 0.5 1535 15 3 85 0.00 0.0 0.00 0.0 0.03 0.8 0.00 0.5 0.00 0.2 0.00 0.1 1538 18 3 85 0.00 0.0 0.00 0.0 0.02 1.8 0.00 1.6 0.00 0.5 0.00 0.0 1539 19 3 85 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.4 0.00 0.0 0.00 0.0 1547 27 3 85 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 1815 20 12 85 0.00 0.9 0.00 0.4 0.06 0.3 0.01 7.5 0.00 0.0 0.06 0.4 Page 133 Appendix 2 PlotO P l o t l Plot 2 Plot 3 Plot 4 PlotS Index D M Y S L R O S L R O S L R O S L | R O S L R O S L R O 1883 26 2 86 0.01 15.1 0.00 44.2 0.00 5.0 0.00 1.5 0.00 0.0 0.04 20.1 1884 27 2 86 0.00 0.1 0.00 0.1 0.00 1.0 0.00 0.9 0.00 0.8 0.00 0.4 1885 28 2 86 0.00 0.8 0.03 0.8 0.01 1.0 0.00 0.9 0.01 0.8 0.02 0.8 1887 2 3 86 0.02 4.8 0.40 5.9 0.90 8.2 0.01 14.4 0.60 5.4 0.04 0.6 1888 3 3 86 0.01 1.0 0.04 3.8 0.68 10.1 0.00 2.0 0.13 16.1 0.05 0.0 1911 26 3 86 0.00 0.0 0.03 3.0 0.01 1.3 0.01 11.1 0.03 8.3 0.03 6.8 1927 11 4 86 0.01 15.1 0.21 29.2 0.49 7.0 0.03 23.1 0.14 19.1 0.37 16.1 2023 16 7 86 0.00 0.0 0.00 0.0 0.50 4.0 0.00 0.0 0.41 4.5 0.44 3.0 2287 6 4 87 0.08 27.1 0.16 23.1 0.22 22.1 0.09 32.7 0.09 28.2 0.29 30.7 2348 6 6 87 0.19 0.8 0.44 3.3 2.69 7.0 0.00 0.0 3.68 7.0 2.79 7.5 2363 21 6 87 0.00 0.0 0.00 0.0 0.86 4.0 0.00 0.0 0.67 4.0 1.13 5.0 2405 2 8 87 0.03 0.5 0.02 2.5 0.47 21.1 0.00 0.0 0.57 19.6 0.61 19.1 2413 10 8 87 0.00 0.0 0.00 0.0 0.01 0.5 0.00 0.0 0.03 1.5 0.04 1.0 2620 4 3 88 0.00 0.0 0.00 0.0 0.00 0.0 0.01 2.2 0.01 3.6 0.00 0.0 2642 26 3 88 0.00 0.0 0.01 0.0 0.06 0.0 0.00 0.3 0.07 0.8 0.10 0.1 2718 10 6 88 0.00 0.0 0.03 0.3 0.45 8.5 0.00 0.0 0.39 8.0 1.24 12.1 2751 13 7 88 0.00 0.0 0.00 0.0 0.02 1.8 0.00 0.0 0.01 0.8 0.00 0.7 2949 27 1 89 0.01 3.6 0.09 0.6 0.06 2.0 0.00 4.0 0.01 1.7 0.03 4.5 3016 4 4 89 0.01 8.5 0.01 3.5 0.01 1.5 0.00 6.3 0.02 6.0 0.06 8.5 3119 16 7 89 0.00 0.0 0.00 0.0 0.03 0.8 0.00 0.0 0.01 0.3 0.03 0.5 3130 27 7 89 0.00 0.0 0.00 0.1 0.04 2.0 0.00 0.1 0.02 1.0 0.03 1.2 3348 2 3 90 0.00 0.0 0.01 1.5 0.03 9.1 0.00 3.5 0.01 5.5 0.02 9.1 3674 22 1 91 0.00 3.5 0.00 0.0 0.00 0.0 0.00 4.5 0.00 1.5 0.01 1.5 3691 8 2 91 0.00 8.0 0.03 5.0 0.12 9.1 0.00 5.0 0.02 5.0 0.18 5.0 3702 19 2 91 0.00 3.5 0.02 3.0 0.13 7.0 0.00 4.5 0.03 2.0 . 0.16 4.5 3731 20 3 91 0.00 4.5 0.00 0.0 0.00 0.0 0.00 3.5 0.00 0.0 0.00 0.0 3786 14 5 91 0.02 0.0 0.23 1.5 0.32 2.0 0.02 0.0 0.19 1.5 0.27 1.0 3821 18 6 91 0.00 0.0 0.02 0.3 0.44 3.0 0.00 0.0 0.03 0.7 0.37 2.5 v 3826 3 6 91 0.03 2.5 0.31 0.5 7.28 11.1 0.00 0.0 0.27 4.0 6.81 12.1 3843 10 7 91 0.00 0.0 0.00 0.0 0.18 1.0 0.00 0.0 0.00 0.0 0.17 1.0 3867 3 8 91 0.00 0.0 0.00 0.0 0.83 5.0 0.00 0.0 0.00 0.0 1.36 7.0 '4039 22 1 92 0.00 0.0 0.03 0.6 0.08 3.0 0.00 0.8 0.04 1.0 0.00 0.0 4076 28 2 92 0.01 34.2 0.15 38.2 1.36 34.2 0.01 39.2 0.30 49.3 2.15 53.3 4089 12 3 92 0.00 0.0 0.01 0.8 0.27 4.5 0.00 2.0 0.02 2.0 0.24 3.0 4173 4 6 92 0.00 0.0 0.86 2.7 3.44 4.5 0.00 0.0 1.59 4.4 2.22 3.3 4179 10 6 92 0.00 0.0 1.11 9.1 3.90 10.6 0.00 0.0 1.92 10.6 3.70 10.1 4207 8 7 92 0.00 0.0 1.12 5.0 4.72 8.5 0.00 0.0 1.91 8.5 5.09 9.1 4227 28 7 92 0.00 0.0 0.19 1.5 0.96 2.0 0.00 0.0 0.33 1.0 1.04 2.0 4236 6 8 92 0.00 0.0 0.33 4.0 5.12 10.1 0.00 0.0 0.47 3.0 4.57 10.1 4251 21 8 92 0.00 0.0 0.00 0.0 0.63 3.0 0.00 0.0 0.19 5.0 0.40 3.0 4265 4 9 92 0.00 0.0 0.00 0.0 0.91 10.1 0.00 0.0 0.21 9.1 0.77 7.5 4387 4 1 93 0.00 0.5 0.00 0.0 0.02 0.0 0.00 0.8 0.02 0.2 0.03 0.0 4563 29 6 93 0.00 0.0 0.00 0.0 0.80 14.6 0.00 0.0 0.19 8.5 0.43 4.5 4570 6 7 93 0.00 0.0 2.01 6.0 5.24 11.6 0.00 0.0 3.53 11.1 7.39 11.1 4577 13 7 93 0.00 0.0 0.65 4.5 4.93 9.1 0.00 0.0 0.92 7.0 3.69 8.5 4604 9 8 93 0.00 0.0 0.00 0.0 1.23 5.0 0.00 0.0 0.00 0.0 1.13 5.0 4619 24 8 93 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 4748 31 12 93 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 0.00 0.0 Page 134 Appendix 3 A P P E N D I X 3 : D E T A I L E D C R O P P I N G P R A C T I C E S F O R A B B O T S F O R D Treatments W E P P Description Depth C = Cultivated* Chisel plow, turned points or shovels 12.5 cm Cs = Cultivated with shovels Chisel plow with sweeps 5 cm D = Disced Disk, offset-finishing 7-9" spacing 7.5 c m H S = Harvested (for use with management option) R = Rotovated Rotary tiller-secondary operation 3" deep 5 cm S = Seeded (cover crop) D r i l l , single disk opener (conventional) 2-3 c m T = Turnplowed Plow, Moldboard, 8" 20 cm P = Planted Double disk planter 5 cm S = Seeding Seeding of cover crop H H = Hand harvested * cultivating - created 15-20 cm ridge for brussels sprouts and 20-25 cm ridge for strawberries Year Date Plot 1 Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 1989 Apr-10 D ; D D ; D D ; D D ; D D ; D D ; D Apr-14 T T T T T T May-11 D ; D D ; D D ; D D ; D D ; D D ; D May-12 C C C C C C Jun-2 C C C C C C Jun-7 R ; P R ; P R ; P R ; P R ; P R ; P Jun-30 Cs Cs Cs Cs Cs Cs Jul-22 Cs Cs Cs Cs Cs Cs Sep-21 Cs Cs Cs Cs Cs Cs Nov-21 H S H S H S H S H S H S 1990 Apr -6 D ; D D ; D D ; D D ; D D ; D D ; D Apr-12 T T T T T T May-3 D ; D ; D ; D ; D ; D ; D ; D ; D ; D ; D ; D ; M a y - 4 D D D D D D M a y - 1 2 R R R R R R M a y - 3 0 P P P P P P Jun-20 D ; R D ; R D ; R D ; R D ; R D ; R Jun-21 S S S S S S Jul-6 Cs Cs Cs Cs Cs Cs Jul-24 Cs Cs Cs Cs Cs Cs Aug-24 Cs Cs Cs Cs Cs Cs Nov-19 H S H S H S H S H S H S Page 135 Appendix 3 Year Date Plot 1 Plot 2 Plot 3 Plot 4 P l o t 5 Plot 6 1991 Apr-18 D ; D D ; D D ; D D ; D D ; D D ; D Apr -22 T T T T T T Apr-23 D ; D D ; D D ; D D ; D D ; D D ; D Apr-30 R R R R R R M a y - 2 P P P P P P M a y - 9 Cs Cs Cs Cs Cs Cs M a y - 3 0 Cs Cs Cs Cs Cs Cs Jun-20 R R R R R R Jul-21 Cs Cs Cs Cs Cs Cs Aug-27 Cs Cs Cs Cs Cs Cs Sep-5 S - S - - -1992 Apr-10 R R R R R R May-1 C C C C C C Jun-5 H H H H H H H H H H H H Jun-28 R R R R R R Aug-1 Cs Cs Cs Cs Cs Cs Sep-1 Cs Cs Cs Cs Cs Cs Sep-15 S - S - - -1993 Apr-19 R R R R R R May-1 C C C C C C Jun-1 C C C C C C Jun-15 H H H H H H H H H H H H Aug-1 Cs Cs Cs Cs Cs Cs Sep-1 Cs Cs Cs Cs Cs Cs Sep-15 S - S - - -Page 136 Appendix 4 A P P E N D I X 4 : D E T A I L E D S O I L L O S S (t ha"1) A N D R U N O F F (mm) O B S E R V E D A T A B B O T S F O R D Plotl Plot 2 Plot 3 Plot 4 Plot5 Plot 6 Index D M Y S L RO S L RO S L RO S L RO S L RO S L RO 1 1 11 89 0.00 1.00 0.00 1.00 0.00 2.30 0.00 2.30 0.00 0.00 0.00 0.00 7 7 11 89 0.02 23.40 0.02 23.40 0.01 6.10 0.01 6.10 0.00 3.05 0.00 3.05 10 10 11 89 0.02 53.80 0.02 53.80 0.07 40.30 0.07 40.30 0.06 36.00 0.06 36.00 11 11 11 89 0.02 21.80 0.02 21.80 0.02 21.80 0.02 21.80 0.01 10.58 0.01 10.58 21 21 11 89 0.00 0.00 0.00 0.00 0.01 9.00 0.01 9.00 0.01 34.07 0.01 34.07 28 28 11 89 0.01 7.10 0.01 7.10 0.01 5.00 0.01 5.00 0.02 23.59 0.02 23.59 35 5 12 89 0.33 43.70 0.33 43.70 0.18 27.50 0.18 27.50 0.04 50.24 0.04 50.24 42 12 12 89 0.11 61.30 0.11 61.30 0.01 20.70 0.01 20.70 0.01 15.36 0.01 15.36 70 9 1 90 0.06 33.60 0.06 33.60 0.06 33.60 0.06 33.60 0.03 41.60 0.03 41.60 77 16 1 90 0.00 1.80 0.00 1.80 0.02 6.40 0.02 6.40 0.00 2.95 0.00 2.95 91 30 1 90 0.01 4.60 0.01 4.60 0.03 13.10 0.03 13.10 0.00 5.39 0.00 5.39 98 6 2 90 0.05 65.20 0.05 65.20 0.05 65.20 0.05 65.20 0.02 26.04 0.02 26.04 105 13 2 90 0.04 59.30 0.04 59.30 0.09 66.90 0.09 66.90 0.10 67.83 0.10 67.83 133 13 3 90 0.05 16.20 0.05 16.20 0.05 16.20 0.05 16.20 0.05 65.90 0.05 65.90 153 2 4 90 0.06 1.90 0.06 1.90 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 324 20 9 90 0.10 1.70 0.10 1.70 0.06 1.40 0.06 1.40 0.03 0.80 0.03 0.80 357 23 10 90 0.02 1.80 0.02 1.80 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 364 30 10 90 0.01 8.80 0.01 8.80 0.01 5.50 0.01 5.50 0.01 7.40 0.01 7.40 374 9 11 90 0.00 64.90 0.00 64.90 0.00 69.40 0.00 69.40 0.00 65.60 0.00 65.60 378 13 11 90 0.01 29.90 0.01 29.90 0.04 22.10 0.04 22.10 0.01 23.30 0.01 23.30 392 27 11 90 0.03 38.30 0.03 38.30 0.02 17.10 0.02 17.10 0.02 33.90 0.02 33.90 399 4 12 90 0.00 36.20 0.00 36.20 0.00 5.40 0.00 5.40 0.00 35.20 0.00 35.20 406 11 12 90 0.02 16.30 .0.02 16.30 0.00 0.00 0.00 0.00 0.02 55.00 0.02 55.00 441 15 1 91 0.00 64.40 0.00 64.40 0.00 17.10 0.00 17.10 0.00 0.00 0.00 0.00 443 17 1 91 0.00 37.80 0.00 37.80 0.00 40.30 0.00 40.30 0.00 0.00 0.00 0.00 448 22 1 91 0.00 16.90 0.00 16.90 0.00 9.80 0.00 9.80 0.00 0.00 0.00 0.00 462 5 2 91 0.15 52.30 0.15 52.30 0.14 52.30 0.14 52.30 0.04 24.10 0.04 24.10 476 19 2 91 0.00 0.00 0.00 0.00 0.01 7.60 0.01 7.60 0.00 0.00 0.00 0.00 521 5 4 91 0.17 37.10 0.17 37.10 0.14 32.50 0.14 32.50 0.05 61.20 0.05 61.20 525 9 4 91 0.00 2.30 0.00 2.30 0.00 0.00 0.00 0.00 0.11 7.10 0.11 7.10 Page 137 Appendix 4 P l o t l Plot 2 Plot 3 Plot 4 Plot 5 Plot 6 Index D M Y S L R O S L R O S L R O S L R O S L R O S L R O 593 16 6 91 0.31 7.20 0.25 9.80 0.34 8.80 0.39 10.30 0.00 0.00 0.00 0.00 646 8 8 91 0.00 0.00 0.00 0.00 0.00 0.40 0.01 2.70 0.00 0.00 0.00 0.00 668 30 8 91 0.08 0.30 0.07 2.00 0.08 0.60 0.04 1.50 0.02 0.80 0.01 0.30 749 19 11 91 0.01 15.00 0.42 12.30 0.04 10.20 0.43 12.40 0.02 6.30 0.02 6.30 756 26 11 91 0.11 30.00 0.82 26.80 0.09 11.80 0.88 24.90 0.02 4.00 0.02 4.90 770 10 12 91 0.01 23.80 0.34 9.80 0.01 5.90 0.32 4.90 0.02 2.90 0.01 3.40 818 27 1 92 0.00 20.30 0.82 37.40 0.02 19.50 0.53 14.50 0.00 0.00 0.00 0.00 825 3 2 92 0.06 9.80 0.98 23.10 0.08 9.70 0.87 19.80 0.00 0.00 0.00 0.00 842 20 2 92 0.02 9.00 2.14 62.20 0.05 10.00 2.21 49.00 0.05 10.00 0.06 12.00 843 21 2 92 0.01 5.10 0.74 12.50 0.01 5.10 0.71 10.50 0.00 0.00 0.00 0.00 887 5 4 92 0.10 2.00 0.61 7.70 0.11 3.10 0.77 2.80 0.08 2.10 0.07 1.90 911 29 4 92 0.24 43.10 1.85 29.80 0.20 39.20 1.22 15.30 0.09 29.80 0.04 15.70 978 5 7 92 0.06 0.00 0.12 0.00 0.05 0.00 0.04 0.00 0.01 0.00 0.01 0.00 1042 7 9 92 0.05 1.00 0.07 2.90 0.00 0.00 0.05 2.50 0.00 0.00 0.00 0.00 1061 26 9 92 0.81 5.00 4.24 13.70 0.78 10.50 4.14 10.50 0.15 5.00 0.14 4.00 1088 23 10 92 0.20 5.20 0.25 6.40 0.22 6.10 0.18 23.30 0.00 0.00 0.00 0.00 1099 3 11 92 0.06 0.50 0.56 5.20 0.04 0.10 0.64 5.90 0.02 0.50 0.03 0.30 1106 10 11 92 0.00 0.50 0.03 1.70 0.00 0.60 0.02 9.80 0.00 0.90 0.00 0.30 1117 21 11 92 0.01 7.10 0.23 39.80 0.01 5.00 0.30 54.00 0.00 2.00 0.00 2.00 1125 29 11 92 0.03 21.30 0.18 6.60 0.03 17.40 0.25 19.00 0.00 0.00 0.00 0.00 1178 21 1 93 0.00 11.00 0.03 42.90 0.00 11.00 0.03 35.20 0.00 0.00 0.00 0.00 1183 26 1 93 0.00 8.10 0.03 5.90 0.00 2.50 0.02 10.00 0.01 13.10 0.01 10.30 1185 28 1 93 0.02 15.20 0.22 17.30 0.02 14.00 0.18 18.30 0.00 0.50 0.00 0.20 1239 23 3 93 0.06 24.30 0.52 34.80 0.04 15.70 0.46 31.00 0.00 0.00 0.00 0.00 1259 12 4 93 0.07 11.10 0.65 26.80 0.09 3.70 0.70 28.80 0.01 6.30 0.01 8.00 1280 3 5 93 0.41 5.80 0.41 6.60 0.46 6.00 0.38 7.20 0.00 0.00 0.00 0.00 1304 27 5 93 0.36 1.20 0.44 7.50 0.42 0.50 0.35 6.40 0.05 1.20 0.07 1.70 1322 14 6 93 0.43 2.00 0.59 5.40 0.41 4.40 0.62 6.50 0.01 0.40 0.01 0.40 1334 26 6 93 0.03 6.50 0.05 6.10 0.04 8.60 0.06 7.20 0.00 0.00 0.00 0.00 1347 9 7 93 0.00 0.50 0.00 2.30 0.00 2.50 0.01 3.10 0.00 0.00 0.00 0.00 1474 13 11 93 0.01 3.10 0.05 5.60 0.01 2.60 0.09 10.10 0.00 0.00 0.00 0.00 1496 5 1 94 0.05 19.50 0.43 35.90 0.04 16.50 0.29 29.90 0.01 1.70 0.02 1.70 1568 15 2 94 0.02 1.00 0.21 15.00 0.01 0.50 0.19 13.30 0.01 0.60 0.02 0.30 1583 2 3 94 0.72 60.70 1.66 56.20 0.47 51.00 1.70 59.30 0.03 17.30 0.04 9.60 Page 138 

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0088648/manifest

Comment

Related Items