UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Modification and evaluation of a computer nitrogen management model for the Fraser Valley of British… Willetts, Shawn E. 1994

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

Item Metadata

Download

Media
831-ubc_1994-0653.pdf [ 3.07MB ]
Metadata
JSON: 831-1.0087544.json
JSON-LD: 831-1.0087544-ld.json
RDF/XML (Pretty): 831-1.0087544-rdf.xml
RDF/JSON: 831-1.0087544-rdf.json
Turtle: 831-1.0087544-turtle.txt
N-Triples: 831-1.0087544-rdf-ntriples.txt
Original Record: 831-1.0087544-source.json
Full Text
831-1.0087544-fulltext.txt
Citation
831-1.0087544.ris

Full Text

MODIFICATION AND EVALUATION OF A COMPUTER NITROGENMANAGEMENT MODEL FOR THE FRASER VALLEY OF BRITISH COLUMBIAByShawn E. WillettsB.A.Sc. (Bio-Resource Engineering), University of British Columbia, 1991A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF APPLIED SCIENCEinTHE FACULTY OF GRADUATE STUDIESDEPARTMENT OF BIO-RESOURCE ENGINEERINGWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIA1994© Shawn E. Willetts, September 1994Signature(s) removed to protect privacyIn presenting this thesis in partial fulfillment of therequirements for an advanced degree at the University of BritishColumbia, I agree that the Library shall make it freely availablefor reference and study. I further agree that permission forextensive copying of this thesis for scholarly purposes may begranted by the head of my department or by his or herrepresentatives. It is understood that copying or publication ofthis thesis for financial gain shall not be allowed without mywritten permission.(SignatureDepartment of - Rsou2 £vvr/A/er1eM/CrThe University of British ColumbiaVancouver, CanadaDate 9. 9.5/Signature(s) removed to protect privacyABSTRACTIn an effort to better predict both soil water flow and soil nitrogen processes at siteswithin the Fraser Valley, the simplistic soil water algorithms of the Bulley and Cappelaere(1978) nitrogen management model were replaced by the more accurate algorithms ofRichard (1988). The modified model was then calibrated using field-measured water tablelevels for 1992. A sensitivity analysis was performed and model simulations were run forNovember and December 1991 and January through September 1993. Results from themodified and unmodified models were compared to field nitrogen and soil moisture datacollected at two sites in the Lower Fraser Valley.Incorporation of the Richard (1988) soil water algorithms in the Bulley and Cappelaere (1978) nitrogen management model improved model representation of soil waterregimes, but did not improve simulated soil nitrogen processes. Discrepancies betweenmodel predicted and field-measured values were attributed to spatial variability in fielddata, lack of field-measured parameters for use as model input, and poor model descriptions of soil nitrogen processes.11Table of ContentsAbstractList of TablesList of FiguresList of SymbolsAcknowledgementsSuperscription1 INTRODUCTION1.1 Objectives11viviiixlixvixvii132 LITERATURE REVIEW 42.1 Soil Nitrogen 42.1.1 Mineralization 42.1.2 Volatilization 52.1.3 Denitrification 62.1.4 Nitrogen Leaching 82.1.5 Manurial and Fertilizer Nitrogen 112.2 Soil Water and Solute Flow Modelling 142.2.1 Description of Flow Processes 142.2.2 Computer Models 171113 MODEL DESCRIPTION 263.1 General Model Description 273.2 Initialization 313.3 Water Movement and Solute Transport Routines 323.3.1 Boundary Flow Conditions 343.3.2 Downward Water Movement 353.3.3 Plant Uptake and Evaporation 433.3.4 Upward Movement of Water 443.3.5 Water Table Level 453.3.6 Solute Transport 453.4 Nitrogen Calculations 473.4.1 Mineralization and Volatilization 483.4.2 Crop Uptake 503.4.3 Denitrification 533.4.4 Nitrification 564 FIELD SITES AND EXPERIMENTS 584.1 Field Sites and Treatment Descriptions 584.2 Field Measurements 595 CALIBRATION AND SENSITIVITY ANALYSIS 645.1 Calibration 645.1.1 Agassiz 645.1.2 Sumas 665.2 Sensitivity Analysis 676 MODEL SIMULATIONS AND FIELD OBSERVATIONS 74iv6.1 Results 746.1.1 Agassiz 746.1.2 Sumas 1026.2 Discussion 1266.2.1 Soil Moisture 1266.2.2 Nitrogen 1287 Summary and Conclusions 1317.1 Recommendations 132Bibliography 134A Climate Calculation Program 145B Manure Production Routines 148C Model Simulation of Several Hypothetical Field Scenarios 149D Model Input Values 156VList of Tables2.1 Comparison of several nitrogen and water flow models 193.1 Model input required 293.2 Flow conditions determining OjEQ 394.1 Summary of field treatments 594.2 Agassiz and Sumas soil bulk density values 624.3 Tensiometer reading dates 634.4 Soil sampling dates for determination of moisture content and inorganic N 635.1 Sensitivity analysis. Model output values following a 1 year simulation.. 706.1 Average errors between predicted and measured nitrate N, ammonium N,and total inorganic N concentrations, Agassiz, 1992 956.2 Model denitrification and leaching losses, Agassiz, 1992 1006.3 Measured and predicted crop yields and N content, Agassiz, 1992 1006.4 Summary statistics, new model, Agassiz 1036.5 Summary statistics, old model, Agassiz 1046.6 Average errors between predicted and measured nitrate N, ammonium N,and total inorganic N concentrations, Sumas, 1992 1196.7 Model denitrification and leaching losses, Sumas, 1992 1236.8 Crop yields and N content, Sumas, 1992 1236.9 Summary statistics, new model, Sumas 1246.10 Summary statistics, old model, Sumas 125viC.1 Field scenario results. New model.151C.2 Field scenario results. Old model 154D.1 Model input values, Agassiz 1991-1993 157D.2 Model input values, Sumas 1992 158vii5.15.25.35.45.55.65.7List of FiguresNitrogen management model: Generalized flowchart.Water movement routines9DEQ determination processRelative reduction of plant dry matter production rateN concentration6265• 666771• 717272table levels, Agassiz 1991table levels, Agassiz 19936.3 Water table levels, measured vs. predicted, Agassiz 1991 and6.4 Water table levels, measured vs. predicted, Agassiz 1992.6.5 Comparison of measured and predicted soil water to a 90 cm depth for thenew and old models, Agassiz 1991 783.13.23.33.430334052vs. available soil4.1 Agassiz ‘(O) curvesCalibrated water table levels, Agassiz, 1992Estimated Sumas retention curveCalibrated water table levels, Sumas 1992Effect of Kdl on water table levelsEffect of &() curve on water table levelsEffect of Ck and 0cK on water table levelsEffect of initial water table depth on water table levels6.1 Water6.2 Water1993.75767778viii6.6 Comparison of measured and predicted soil water to a 90 cm depth for thenew and old models, Agassiz 1992 796.7 Soil moisture profiles, Sept. 30—Nov. 20, 1991, Agassiz 806.8 Soil moisture profiles, Dec. 4 and Dec. 18, 1991, Agassiz 816.9 Soil moisture profiles, Jan. 16—May 4, 1992, Agassiz 826.10 Soil moisture profiles, May 19—July 14, 1992, Agassiz 836.11 Soil moisture profiles, Aug. 17—Oct. 14, 1992, Agassiz 846.12 Soil moisture profiles, Oct. 28—Dec. 8, 1992, Agassiz 856.13 Comparison of measured and predicted soil water, 0—30 cm., Agassiz, 1991. 866.14 Comparison of measured and predicted soil water, 30—60 cm., Agassiz, 1991. 866.15 Comparison of measured and predicted soil water, 60—90 cm., Agassiz, 1991. 866.16 Comparison of measured and predicted soil water, 0—30 cm., Agassiz, 1992. 876.17 Comparison of measured and predicted soil water, 30—60 cm., Agassiz, 1992. 876.18 Comparison of measured and predicted soil water, 60—90 cm., Agassiz, 1992. 876.19 Model and field soil NO-N to 90 cm, C, Agassiz, 1992 896.20 Model and field soil NHt-N to 90 cm, C, Agassiz, 1992 896.21 Model and field soil inorganic N to 90 cm, C, Agassiz, 1992 896.22 Model and field soil NOW-N to 90 cm, FM300+SM300, Agassiz, 1992. . 906.23 Model and field soil NH-N to 90 cm, FM300+SM300, Agassiz, 1992. . 906.24 Model and field soil inorganic N to 90 cm, FM300+SM300, Agassiz, 1992 906.25 Model and field soil NOW-N to 90 cm, FM600, Agassiz, 1992 916.26 Model and field soil NHt-N to 90 cm, FM600, Agassiz, 1992 916.27 Model and field soil inorganic N to 90 cm, FM600, Agassiz, 1992 916.28 Model and field soil NOW-N to 90 cm, SM600, Agassiz, 1992 926.29 Model and field soil NH-N to 90 cm, SM600, Agassiz, 1992 926.30 Model and field soil inorganic N to 90 cm, SM600, Agassiz, 1992 92ix6.31 Model and field soil NO-N to 90 cm, SF200, Agassiz, 1992 936.32 Model and field soil NH-N to 90 cm, SF200, Agassiz, 1992 936.33 Model and field soil inorganic N to 90 cm, SF200, Agassiz, 1992 936.34 Soil nitrate N profiles: C, SF200, SM600 FM300+SM300, summer 1992,Agassiz 976.35 Soil nitrate N profiles: C, SF200, SM600, & FM300+SM300, autumn &winter 1992, Agassiz 986.36 Soil nitrate N profiles: F600, summer and winter 1992, Agassiz 996.37 Water table levels, measured vs. predicted, Sumas 1992 1066.38 Comparison of measured and predicted soil water to a 90 cm depth for thenew and old models, Sumas 1992 1066.39 Soil moisture profiles, April 8—June 2, 1992, Sumas 1086.40 Soil moisture profiles, June 16—Sept. 17, 1992, Sumas 1096.41 Soil moisture profiles, Sept. 23—Nov. 9, 1992, Sumas 1106.42 Soil moisture profiles, Nov. 23 and Dec. 7, 1992, Sumas 1116.43 Comparison of measured and predicted soil water, 0—30 cm., Sumas, 1992. 1126.44 Comparison of measured and predicted soil water, 30—60 cm., Sumas, 1992. 1126.45 Comparison of measured and predicted soil water, 60—90 cm., Sumas, 1992. 1126.46 Model and field soil NO-N to 90 cm, C, Sumas, 1992 1136.47 Model and field soil NH-N to 90 cm, C, Sumas, 1992 1136.48 Model and field soil inorganic N to 90 cm, C, Sumas, 1992 1136.49 Model and field soil NO-N to 90 cm, FM300+SM300, Sumas, 1992.. . . 1146.50 Model and field soil NHt-N to 90 cm, FM300+SM300, Sumas, 1992.. . . 1146.51 Model and field soil inorganic N to 90 cm, FM300+SM300, Sumas, 1992 1146.52 Model and field soil NO-N to 90 cm, FM600, Sumas, 1992 1156.53 Model and field soil NHt-N to 90 cm, FM600, Sumas, 1992 115x6.54 Model and field soil inorganic N to 90 cm, FM600, Sumas, 1992. 1156.55 Model and field soil NO-N to 90 cm, SM600, Sumas, 1992 1166.56 Model and field soil NH-N to 90 cm, SM600, Sumas, 1992 1166.57 Model and field soil inorganic N to 90 cm, SM600, Sumas, 1992 1166.58 Model and field soil NO-N to 90 cm, SF200, Sumas, 1992 1176.59 Model and field soil NH-N to 90 cm, SF200, Sumas, 1992 1176.60 Model and field soil inorganic N to 90 cm, SF200, Sumas, 1992 1176.61 Soil nitrate N profiles: C, SF200, SM600 & FM300+SM300, June 2, 1992,Sumas 1206.62 Soil nitrate N profiles: C, SF200, SM600, & FM300+SM300, autumn 1992,Sumas 1216.63 Soil nitrate N profiles: F600, June 2 and Oct. 13, 1992, Sumas 122C.1 Effect of soil type on water table level 153C.2 Effect of precipitation on water table level 155xiList of Symbolsa unsaturated hydraulic conductivity coefficientAman mass of manure applied/3 mineralization coefficientC1 half-life constant for soil carbonC93 crop growth stage coefficientC,d layer nitrate concentration following downward lossesC,1 layer nitrate concentration following all gains and lossesCj,q layer nitrate concentration following infiltration from aboveC,,. layer nitrate concentration following gain from layer belowCk saturated hydraulic conductivity coefficientCrasp crop response coefficientV linear density distribution functionDrz number of layers in root zonedTD adjusted denitrification time delaydTD,max maximum denitrificatiori time delaydenitrification coefficientEact actual evaporationEadj adjusted potential transpirationE0 difference between potential evapotranspiration and actualplant water uptakef drainable porosityf, f manurial organic N mineralization factorsfnm fraction of organic N not mineralizedf fraction of organic N remaining at beginning of timestepFHd denitrification coefficient accounting for soil pHFpHm mineralization coefficient accounting for soil pHFH nitrification coefficient accounting for soil pHFtempn nitrification coefficient accounting for temperatureFd denitrification coefficient accounting for soil moistureFwm mineralization coefficient accounting for soil moisturenitrification coefficient accounting for soil moisturexii7 fraction of solute not leached from layer1’ numerical calibration parameterg1,g2 initial and final crop growth indicesdaily crop growth coefficientgrowth index at middle of timestepgy yearly summation of daily crop growth coefficientsGdm crop dry matter gainGmm mass of ammonium gained through mineralization of manurialorganic NGms mass of ammonium gained through mineralization of soil organic Nç average gradient under which downward flow occursaverage gradient under which upward flow occursll root depth at current root density112 current total root depthlevel of the Fraser River above geodetic datuminitial soil organic N contentlevel of field water table above geodetic datumi,j layer number indicesdeep flow saturated hydraulic conductivityKeq, harmonic average of K(O) valuesK8 saturated hydraulic conductivityweighting factor used in determining ODEQdistance from Agassiz field site to Fraser RiverTD nitrification time delayTD,max maximum nitrification time delayN1 half-life constant for soil nitrate2N half-life constant for soil ammonium2Namm ammonium available for crop uptakeNavaii total N available for crop uptakeNd0 concentration of nitrate lost to denitrificationNdenit nitrate lost to denitrification over entire profileNenitp potential loss of nitrate to denitrificationNmin inorganic N additions to soilpotential loss of ammonium to nitrificationNorg organic N additions to soilplant N uptake rateNtk amount of N removed from soil by cropNH4,301 ammonium concentration averaged over depth to plough layerNO3,avail nitrate available to crop rootsNO3,801 layer nitrate concentrationxJnm soil matric potential1’wf suction ahead of wetting frontPdr amount of water available for drainage from a layerPnf maximum amount of water that can infiltrate a layeractual crop production ratioP0, maximum crop production ratioP80 potential snowmeltqjn depth of water flowing into layerqinf amount of water infiltrating layer j from j — 1qj,, net flow of water from layer j to j — 1zat deep flow from profileqpotup potential upward water flowQ lOd Qio factor for denitrificationQim Qio factor for mineralizationTCN denitrification rate adjusted for soil N and C conditionsTdenit denitrification rateTm mineralization ratermaxd maximum denitrification rateTmaxfl maximum nitrification rateadjusted nitrification rater, nitrification rateR.. relative reduction of dry matter production rateRdens root densityS,bl0 solute gained from layer belowsolute lost to layer belowsolute lost to layer aboveO volumetric moisture content of soil layerUi volumetric moisture content of first soil layer°ck saturated hydraulic conductivity coefficientO.DEQ dynamic equilibrium moisture content0DR volumetric moisture content achieved after constantdrainage°FL volumetric moisture content achieved after constantinfiltration0FL1 volumetric moisture content achieved after infiltrationand no drainage9FL2 volumetric moisture content achieved after infiltrationand drainageUi initial volumetric moisture content of layerO intermediate volumetric moisture content of layerxiv°min lowest moisture content at which crop water availabilityis not affectedOsEQ static equilibrium moisture contenttimestep durationUNQ3 nitrate lost from profile to plant uptakeUNH4 ammonium lost from profile to plant uptakeWavaii water available in a layer for drainageW potential plant water uptakeWravaii water available to plant rootsactual plant water uptakeXc, XN denitrification factors accounting for soil nitrate andC conditionsnitrification factorZz layer thicknessz distance from soil surface to the bottom of layerz20,z60 depths at which soil C availability is at a maximumand minimum, respectivelyZm distance from soil surface to middle of layerz, distance from soil surface to water tablexvAcknowledgementsThe author wishes to thank Dr. M. Novak for his supervision, Dr. S.T. Chieng forreviewing this work, and Drs. J. Paul and B. J. Zebarth for making available to theauthor the results of their field work.In addition, special thanks go to Dr. P.F. Richard for his comments and suggestions, to K. Beckett for her support and tolerance, and to my colleague S. Lan for hisconstructive criticism and printer access.The nitrogen management model used in this study was provided courtesy of theBritish Columbia Ministry of Agriculture, Fisheries and Food, Resource ManagementBranch.Financial support through the Canada-British Columbia Soil Conservation Agreementis gratefully acknowledged.xvi‘Beware the Jab berwock, my son!The jaws that bite, the claws that catch!Beware the Jubjub bird, and shunThe frumious Bandersnatch!’— Lewis Carroll, Jabberwocky.xviiChapter 1INTRODUCTIONNitrogen, more than any other nutrient, is the primary factor influencing cereal crop yield(Addiscott et al., 1991). To maximize yield, agricultural lands are commonly amendedthrough addition of nitrogen in the form of chemical fertilizer, manure, or both. Increasingly, the potential for pollution through excess land application of nitrogen is beingrecognized. This, in turn, has led to a heightened awareness of the need for betternitrogen management.Leaching of soil nitrate can have both economic and environmental repercussions. Inaddition to no longer being available for plant uptake, nitrogen leached from the rootzone as NO is a potential contaminant of ground and surface water. Recent experimentsin south coastal British Columbia suggest that there is a high risk of NO leaching inthe fall and winter months (Kowalenko, 1987b), and groundwater surveys conducted onaquifers in the Fraser Valley (Liebscher et al., 1992; Gartner Lee Ltd., 1993) have revealednitrate concentrations in excess of the recommended drinking water maximum of 10 mg/i(Heaith and Welfare Canada, 1989).In order to decrease nitrogen losses and increase nitrogen use efficiency, it has beenrecommended that further research be undertaken to improve nitrogen management practices (Bertrand et al., 1989). Numerous models have been developed in an attempt tobetter understand the soil N cycle and aid in fertilizer management decisions (see reviews by Tanji, 1982; Addiscott and Wagenet, 1985; de Willigen, 1991). However, asnitrogen cycling is a highly site specific phenomenon, extrapolation and adaption of1Chapter 1. INTRODUCTION 2nitrogen practices from one geographic area to another should be done with extremecaution (Kowalenko, 1987a). Therefore, while many models have been presented to simulate nitrogen movement in soils, it may be best to develop a new model for a particulargeographic region.A nitrogen management model was developed specifically for the south coastal region of British Columbia by Bulley and Cappelaere (1978). Originally programmed onthe UBC Amdhal 580 mainframe computer, it has since been adapted for use on IBMcompatible personal computers (McDougall et al., 1989a, 1989b; Price, 1990). Since itsinception the model has seen a variety of applications; its use facilitated the formulationof the Manure Management Guidelines (Bertrand and Bulley, 1985) and currently anattempt is being made to adapt the model to Manitoba conditions (Ranjan et al., 1993).The water balance portion of the model, as originally developed, employed a numberof simplifying assumptions. One significant simplification, that the water table remainsbelow the root zone for the duration of the year, could result in the calculation of erroneous moisture content values within the soil profile. As soil moisture content is a majorfactor in the loss and movement of nitrogen through both leaching and denitrification(Kowalenko, 1987a), a realistic description of the soil moisture regime within the modelis essential. Richard’s (1988) nutrient flow model, also developed for use in the FraserValley, contains a more comprehensive description of soil water flow; among other thingsit allows for a fluctuating water table. As this was felt to be a more realistic description ofsoil water behaviour, the water movement algorithms of the nitrogen management modelwere replaced by those of Richard (1988). The modified nitrogen management modelwas then calibrated and validated against data collected at two Fraser Valley field sites.Chapter 1. INTRODUCTION 31.1 Objectives.This study focussed on the water balance portion of the nitrogen management model.The overall objectives were:• To incorporate the Richard (1988) water movement routine into the nitrogen management model of Bulley and Cappelaere (1978).• To calibrate the modified model and perform a sensitivity analysis using modelhydrological parameters.• To validate the modified model using field data collected in a separate study byZebarth and Paul (1994).• To evaluate the effectiveness of the water routine modifications by comparing thesoil nitrogen and water predictions of the modified and unmodified models againstfield data.Chapter 2LITERATURE REVIEWA number of studies have been conducted that elucidate the various processes involved inthe cycling of soil nitrogen. Rather than delve into every facet of this cycle, this reviewwill focus primarily on the roles of soil moisture, fertilizer N, and manurial N.2.1 Soil NitrogenNinety to ninety-five percent of the total nitrogen in topsoils occurs in organic compounds (Wild, 1988). The inorganic nitrogen fraction is comprised mainly of six compounds: nitrate (NO), nitrite (NOr), exchangeable and nonexchangeable ammonium(NH), dinitrogen gas (N2), and nitrous oxide gas (N20) (Young and Aldag, 1982). Ofthese inorganic constituents, nitrate and ammonium ions are the two forms most readilyassimilated by the large majority of plants (Wild, 1988).2.1.1 Mineralization.Conversion of soil N from organic to inorganic forms occurs through the process of mineralization. This is essentially a two-stage biological process consisting of ammonificationand nitrification. Ammonification is the transformation of organic sources of N to NHt,and its subsequent oxidation to NO is referred to as nitrification (Stevenson, 1982).Nitrification is an aerobic process, illustrated below (Wild, 1988):NH + 1O2 NO + H20 + 2H (2.1)4Chapter 2. LITERATURE REVIEW 5— bact. —NO2 + O2 —* NO3 (2.2)In tests conducted with flooded organic soil columns, Reddy and Graetz (1981) foundthat under aerobic conditions NH was lost via nitrification and NH3 volatilization; underanaerobic conditions NH was lost only through NH3 volatilization.Nitrification is also influenced by fluctuations in temperature and moisture content.Harris (1988) reported that ammonium N fertilizer applied in late autumn may remainunchanged until spring, largely because nitrification is slow at temperatures below 4 or5°C. Experimenting with a clay loam soil, Kowalenko and Cameron (1976) found that at4 and 35 °C the optimum moisture content for riitrifiers was 25% (wt. basis), whereas at15 and 25 °C nitrification peaked at 35% moisture content or higher. Maidl and Fischbeck(1989) found mineralization occurring at temperatures near 0°C in a field receiving longterm applications of pig slurry, whereas a short term application site required temperatures above 4°C for mineralization to commence. Higher rates of mineralization werealso observed in the long term field than in the short term field. Kowalenko and Hall(1987) observed apparent net soil mineralization rates of —0.1 to 1.4 kg.ha’d and 0.8to 1.0 kg.had’ under broccoli and corn plots, respectively, during the growing seasonat Agassiz, B.C.2.1.2 Volatilization.Soil nitrogen, once converted from organic N to NH-N, can be lost to the atmosphereas volatilized NH3 (Pain, 1991):NHatm.solid soln.NHNH +H20 NH3 +1120 + H (2.3)Volatilization in the field depends to a certain extent on the rate at which NH3 is transported away from the soil surface, either by wind or evaporating water (Wild, 1988).Chapter 2. LITERATURE REVIEW 6Al-Kanani et al. (1991) investigated the effect of soil moisture on volatilization fromsurface applied N fertilizer; initially moist samples (b = —0.01 MPa) exhibited greaterNH3 volatilization than initially air-dry samples ( < —1.5 MPa). In this study thecumulative amount of ammonia volatilized depended logarithmically on the cumulativeamount of water evaporated.Volatilization losses increase as soil pH rncreases (Stevenson, 1982; Pain, 1991). Surface applications of farm yard manure or urea can cause localized increases in pH throughthe hydrolysis of urea; this may cause large quantities of nitrogen to be lost as NH3 (Wild,1988; Stevenson, 1982). Pain (1991) noted that volatilization losses are greatest from manures with a low C:N ratio. Poultry manure and slurry were susceptible to losses of 6.9and 37.5% of total N, respectively, whereas cattle and pig slurry lost approximately 18%of total N.2.1.3 Denitrification.Denitrification, the reduction of nitrate to nitrous oxide and nitrogen gas primarily bydenitrifying soil bacteria, can be represented schematically as (Harris, 1988):NO — NO —+ NO —+ N20 — N2 (2.4)Denitrifiers are facultative anaerobes, and thus denitrification only proceeds under lowoxygen conditions. As nitrate is the main product resulting from the nitrification ofmanurial N (Smith and Peterson, 1982), denitrification can result in considerable Nlosses from manure applied to soil. The amount of nitrate lost will depend on sitespecific conditions such as the availability of oxygen, soil temperature, moisture content,pH, and readily available carbon sources (Stevenson, 1982).An environment conducive to denitrification would have a low oxygen status and aready supply of carbon and nitrate. Such conditions commonly exist in poorly drainedChapter 2. LITERATURE REVIEW 7soils, near a shallow water table, or just after periods of heavy rain, and numerous studieshave been conducted in an effort to better understand the effects and interactions of theseparameters. Reddy et at. (1980) applied‘5N03-N to saturated organic soil in test tubesand found that 97% of it was lost through denitrification. Further, creating a high oxygendemand, or increasing the concentration of N03-N in solution, increased the nitrate Nremoval rate. On a small silt loam field plot, Coulborn (1992) noted that soil moisturehad a measurable effect on denitrification; the more water present (as water filled porespace) the greater denitrification tended to be. Using soil columns, Coulborn (1992)also found that denitrification rates were higher on a silt loam than on an organic soil,though this may have been attributable to the low pH of the latter. Aulakh et at. (1991)determined that at 37% moisture content, gravimetric basis, a silty clay had 60% waterfilled pore space and a silty loam had 90% water filled pore space, promoting aerobic andanaerobic conditions in the two soils, respectively. Consequently, denitrification in thesilt loam was greater. On this basis they proposed that specifying soil moisture contentas water filled pore space rather than on a gravimetric basis gives a better indication ofdenitrification potential.Verdegem et at. (1981) postulated that in soils where a shallow groundwater table isprevalent, any NO left in the root zone following harvest will leach to the permanentlysaturated zone and be lost through denitrification. Studying nitrate transformations inthe vicinity of a shallow groundwater table, Meek et at. (1970) discovered that denitrification was rapid at and below the watertable; in the submerged zone the rate of denitrification was related to the concentration of NO and the quantity of soluble carbon leacheddown to the watertable. The findings of Yeomans et al. (1992), that denitrification indeep layers is limited by lack of carbon rather than lack of microorganisms, support this.Krajenbrink et at. (1989) suggested that deriitrification at depth is influenced more by indigenous soil organic matter within the aquifer than by dissolved organic carbon leachedChapter 2. LITERATURE REVIEW 8down to the saturated zone. Verdegem and Baert (1984) determined that accumulationof C at depth had little affect on rates of denitrification, and thereby argued that insubsoils with significant amounts of primary minerals chemical reduction of nitrate byferrous iron should be considered. Denitrification involving pyrite oxidation has also beensuggested by Krajenbrink et at. (1989).Because of inherent variability within the soil, phenomena observed in the laboratorymay not always manifest themselves in the field. The study of Heaney et at. (1992)found that laboratory measurements of NO lost through denitrification in Alberta soilsdid not correlate with losses measured in the field. They suggested that climatic, ratherthan inherent soil, chemical, or biological, variables controlled denitrification in the field.Denitrffication microsites, or “hot spots”, introduce further variability across a seeminglyuniform field. Christensen and Tiedje (1987) stated that hot spots are anaerobic regionsin the soil formed as a result of increased respiration of organic matter, and are usuallyassociated with local high availability of organic matter. Verdegem et at. (1981) identified“anaerobic soil pockets” in the surface layer of a heavy clay as being favourable fordenitrification. Fillery (1983) cited studies describing soil geometry and aggregate size,the diffusion rate of oxygen into aggregates, and soil moisture status as important factorsin the development of microsites within soil aggregates.Though denitrification causes valuable fertilizer and manurial N to be lost, whenexcessive amounts of nitrogen are applied denitrification may be considered desirablesince it reduces the nitrate load on groundwater (Meek et al., 1970; Smith and Peterson,1982; Stevenson, 1982).2.1.4 Nitrogen Leaching.Loss of nitrogen through leaching has been investigated for well over a hundred years(Lawes et at., 1882). While leaching of NH may occur in sandy soils (Stevenson, 1982;Chapter 2. LITERATURE REVIEW 9Wild, 1988), most nitrogen is lost as NO. As nitrate is found predominantly in the soilsolution, its behaviour in the profile is determined largely by the soil water regime.Temporal variations in precipitation and water fluxes will influence leachate qualityat a water table (Krajenbrink et al., 1989). Verdegem and Baert (1984) listed groundwater fluctuations and weather conditions among the factors contributing to the migrationof NO between the root and permanently saturated zones. They observed a bulk displacement of NO as a result of a dry spring followed by wetter periods in summer, andreported that a wet winter and spring period in the following year caused an importantleaching of nitrate. Over a two year period, White et al. (1983) found that leaching occurring in “wet” winter periods accounted for over 70% of the nitrogen lost; the remainderbeing lost during autumn. For coastal plain soils under corn crop, winter nitrate leachingwas found to exceed growing season leaching (Ritter et al., 1985). Hubbard et al. (1991),studying a different sandy coastal plain soil, determined that nitrate leaching was heavilyinfluenced by precipitation and irrigation. A period of unusually high rainfall, followingan application of inorganic N fertilizer, resulted in most of the N03-N being leachedbeyond the top 30 cm of soil within 1.5 months. MacGregor et al. (1974), in studyingtwo clay loam sites, concluded that downward movement of NO was season and rainfalldependent. Average rates of NO movement in Forman clay loam were found to be 1.7and 1.9 mni d1 over 10 and 15 year periods, respectively, and at the end of the 15 yearperiod appreciable quantities of nitrate were found at the 10 metre depth. Gast et al.(1978) found little evidence of nitrate movement below a depth of 2.2 metres on a clayloam soil, but suggested that this may have been attributable to low rainfall during thestudy period. For a Fraser Valley soil, Kowalenko (1987b) found significant NO leachingoccurred in winter months, whereas leaching during the growing season was minimal. Asa minor exception, some movement of nitrate was observed during the summer followingChapter 2. LITERATURE REVIEW 10a month of above average precipitation. Iqbal and Warkentin (1981) found that leaching losses increased with an increase in rainfall, and that nitrate leaching was greater insandy as opposed to clayey soils. The study of Devitt et al. (1976) found that soil profilecharacteristics heavily influenced nitrate movement; application of irrigation water wasalso found to promote downward movement of nitrates.A factor widely cited as affecting both water and solute movement in the soil profileis the phenomenon of channelling, or bypassing flow. This “short-circuiting” is the rapidflow of infiltrating water through large continuous pores in an unsaturated soil matrix(Dekker and Bouma, 1984). White et al. (1983) indicated that bypassing flow can significantly affect leaching as nitrate diffusing out of moist aggregates into macropore wateris rapidly transported down into the profile. Observing that fertilizer applied to clay soilcolumns after prewetting resulted in less nitrogen being lost in subsequent irrigations,Dekker and Bouma (1984) suggested that applying fertilizer after a rainfall event couldresult in a similar reduction of leaching losses. They explained that applying N fertilizerto a wet soil surface allows for better diffusion of the nitrogen into soil peds, and therebyless nitrogen is left on the surface vulnerable to mass flow in macropores. Christian et al.(1990) found that a reduction in autumn ploughing resulted in a reduction in NO leaching losses. Minimizing autumn ploughing left the continuous network of pores connectingtopsoil to subsoil intact; thus, drainage was assisted and NO within soil aggregates wasbypassed. Vinten et al. (1991) reported similar findings for a clay loam in Scotland.Bergstrom (1987) noted a large increase in leaching following the ploughing of a grassicy, due to both enhanced soil water flow and mineralization, and recommended that leysbe ploughed in the spring to reduce such losses.Variability in nitrate leaching was also investigated by Cameron et al. (1979). Theystated that differential leaching rates are a prime cause of nitrate variability in a field plot,and identified variability in the surface microrelief and variability in water storage andChapter 2. LITERATURE REVIEW 11flow characteristics as two main factors inducing differential leaching. A nonhomogeneoussurface microrelief could increase leaching at local depressions and cracks by divertingrainfall runoff to these areas from high points, and local differences in soii water storageand hydraulic properties would cause water flow to vary within the soil profile.It is commonly reported that rainfall after an extended dry period results in an increase in NO leaching losses. Presumably, mineralization occurs during the dry period,and with the application of water through rainfall or irrigation the nitrate produced isleached down into the profile. However, Meek et al. (1970) theorized that interspersingirrigation with periods of drying could bring carbon into solution and enhance denitrification. This suggests that intermittent precipitation can cause denitrification and leachingto proceed concurrently.2.1.5 Manurial and Fertilizer Nitrogen.Nitrogen is usually applied to land in the form of manure or chemical fertilizer, bothmethods having their attendant benefits and shortcomings. In terms of crop response,numerous studies have found no significant difference in crop yield when nitrogen wasapplied as properly handled manure or as chemical fertilizer (cited in Keeney, 1982.) Inadvocating one form over the other, Whitman (1855) divulged his preference toward theau naturel source when he implored his readers to “Behold this compost! behold it well!”Although, in all fairness, it must be admitted that the Haber process, which permittedfixation of atmospheric nitrogen and allowed for the synthesis of chemical fertilizers ona large scale, had not yet been discovered.Accurate application of a given amount of nitrogen is relatively straightforward whenonly chemical fertilizers are employed. Using manure alone, or a combination of manurial and manufactured sources, increases the uncertainty in determining the amount ofnitrogen available for crop uptake. Powison et al. (1989) cited the following difficultiesChapter 2. LITERATURE REVIEW 12associated with manure application:1. Handling and spreading problems make it difficult to apply a given quantity of Naccurately.2. NH3 volatilization during storage and application can result in large and variablelosses of N.3. Proportion of ammonium N (immediately available to crops) and organic N (available after mineralization) is variable and often unknown.4. Total amount of N which will become available for crop uptake, and the timing ofits release, is difficult to predict.5. Significant amounts of NO may be formed during times of low crop uptake andthus be susceptible to loss through leaching or denitrification.The inherent uncertainty in manurial N is possibly the most important reason behind itswidespread misuse in agricultural applications (cited in Jansson et a!., 1989).Time of application, type of nitrogen source and amount applied are all importantfactors in promoting efficient nitrogen use. Pain (1991) stated that the optimum timefor application of cattle slurry to most crops is late winter or early spring. Iqbal andWarkentin (1981), in comparing fall application of cattle slurry to sandy and clayey soilsat rates of 0, 260, 390 and 520 kg N ha’, found that the application rate had littleeffect on the amount of nitrate leached to drainage tiles. Results obtained by Vintenet a!. (1991) suggested that spring application of inorganic N fertilizer at recommendedrates had no significant effect on leaching losses. Gast et at. (1978) found that leachinglosses increased with increased amounts of urea applied in the spring to a clay loam soil;however, at the recommended application rate there was little increase in tile loss overChapter 2. LITERATURE REVIEW 13a check plot receiving minimal N. Using ‘5N labelled fertilizer solution Macdonald et al.(1989) determined that for a soil growing winter wheat, practically all nitrate availablefor over winter leaching originated from mineralisation of organic N rather than fromunused fertilizer applied in spring. They further observed that at harvest there was nosignificant difference in profile inorganic N content between plots receiving no fertilizerand those receiving N at rates of up to 234 kg ha1. Therefore, Macdonald et al. (1989)stated that limiting N fertilizer use would not much affect NO leaching. Powlson etal. (1989) declared that utilizing inorganic fertilizers in the spring to meet a crop’s Nrequirement will result in less nitrate pollution than the use of organic N sources.Green manuring, which is the use of crop residue as a source of nitrogen, is oftenaccomplished by ploughing in an existing crop. Vinten et al. (1991) found that the autumn incorporation of a leguminous green manure significantly increased nitrate leachingwhile contributing little to the nitrogen requirement of subsequent crops. Lab studies byAulakh ci al. (1991) demonstrated that incorporating legume residue into soil columnsgreatly promoted denitrification. Gast et al. (1978) found that use of soybean meal asan N source resulted in considerably less accumulation of NO within the 0—3 metreprofile than an equivalent amount of N applied as urea. Powison ci al. (1989) noted thatsoil inorganic N content during the winter months tended to be higher in plots amendedwith organic manure rather than chemical fertilizers. The study of Maidi and Fischbeck(1989) supported this.Despite the problems inherent in the utilization of manure N, the high cost of commercial fertilizers and the need for animal waste disposal will likely ensure its continueduse. In addition, application of manure does seem to offer benefits not realized throughuse of inorganic N sources alone. Crop and animal wastes applied to soil may reducenutrient losses by decreasing runoff and wind erosion (Legg and Meisinger, 1982). Sommerfeldt and Chang (1985) found that within the top 15 centimetres of an Alberta soil,Chapter 2. LITERATURE REVIEW 14organic matter increased while bulk density and drawbar draft decreased with increasingrates of manure application. Mathers and Stewart (1983) stated that while applicationof manure over a 14 year period decreased bulk density and increased hydraulic conductivity and soil organic matter on a clay loam, large applications could result in saltor high ammonia damage to seedlings and increase the potential for N leaching losses.In contrast, Darusman et al. (1991) found that after 20 years of applying inorganic Nfertilizers to silty and silty clay barns, the primary effect on the soils was an increase inacidity and NO and NH concentrations.2.2 Soil Water and Solute Flow Modelling.2.2.1 Description of Flow Processes.The basic equation describing steady state flow in saturated soil is Darcy’s law. For ahomogeneous, isotropic media in three-dimensions it is often written as:q=—KVq (2.5)where K is the hydraulic conductivity and q is the hydraulic head. Though developedempirically to describe one dimensional flow in sand columns, several attempts have beenmade to derive it from basic hydrodynamic principles (see the review by Bear, 1972).Richards (1931) extended Darcy’s law to unsaturated flow by expressing K as afunction of the soil matric potential, K(/). Further, by employing both Darcy’s relationand the continuity equation, he wrote a general flow equation for unsaturated soil as:= V (K(b)v(b + mg)) (2.6)where 0 is the soil moisture content and is the gravitational head. Richards’ equation ishighly nonlinear, and, except for simple geometries in homogeneous systems, it is usuallyChapter 2. LITERATURE REVIEW 15solved numerically via finite difference or finite element methods. Equation 2.6 can alsobe expressed as a 0-based diffusivity equation, but for numerical simulations it is usuallycast in terms of ib since this allows for incorporation of soil heterogeneity and saturatedregions (Milly, 1988).Solute movement is generally thought to parallel that of soil water. The behaviourof a solute that is subject to both adsorption and transformations within the soil can bedescribed with a form of the convection-dispersion equation such asöps Oc 8 öc öc+ 0. = —(D--)- q- + (2.7)where p is the soil bulk density, s the solute associated with the soil particles, c the concentration in the soil water, D an apparent diffusivity coefficient (combining mechanicaldispersion and diffusivity), and is a sink or source term (Nielsen et al., 1982; Wagenetand Rao, 1983; Addiscott and Wagenet, 1985). For a substance such as nitrate, whichdoes not react with the soil matrix to a large degree but is susceptible to chemical andbiological reactions, the first term may be negligible.Simulations of solute movement commonly use both the Richards and convection-dispersion equations in tandem; values for 0 and q are obtained by solving Richardsequation, and these values are then used in the convection-dispersion equation to describesolute behaviour (Addiscott and Wagenet, 1985).A conceptually simple alternative to the above approach is the description of soluteand soil water movement as plug, or piston, flow. This method assumes that all water in agiven volume of soil is “pushed out” by water flowing into the volume without any mixingoccurring. Using this analogy, a volume of water with a given solute concentration willmove through the soil as a “plug”, pushing along the water in front of it. The position ofthe front of this displacing “piston” of water can be estimated by a simple relation suchChapter 2. LITERATURE REVIEW 16as: =(2.8)where z,, is the depth to the piston front and Q is the quantity of water doing the displacing (Addiscott et al., 1991). This model works best in a sandy soil, but is unsatisfactoryin soils where the soil water is unequally mobile.The use of a water budget is a straightforward approach to describing soil water flow.This method is frequently based on capacity parameters such as the field capacity andwilting points. It is essentially a bookkeeping scheme whereby the quantity of waterin the profile is incremented by the amount of water flowing in, and decremented bythe water flowing out. The influx of water into the profile is usually calculated as thedifference between precipitation and evapotranspiration, and outflow can be determinedas any excess water over field capacity (De Jong, 1981). Solute transport can then besimulated as the product of the flux and the concentration (de Willigen, 1991).Further complicating the various descriptions of soil water movement is the mechanismof bypassing flow. While Richards’ equation offers an adequate representation of flowwithin the soil matrix, describing macropore flow and its interaction with matrix flow isnot as clear. Milly (1988) states that it is common to consider flow in macropores andthe soil matrix separately with a transfer term to describe the interaction between thetwo domains. The description of macropore flow itself is often approached on one of twolevels: as the study of an individual macropore and its interaction with the surroundingmatrix, or as a network of pores cut into the media (Milly, 1988). A combination of thetwo approaches is also feasible.Chapter 2. LITERATURE REVIEW 172.2.2 Computer Models.Soil water and nitrogen models vary in complexity from the conceptually simple (such asthe apportioning method of Beauchamp and Paul, 1989) to the comprehensive (McGillet al., 1980). Simulation models can be broadly classified as either deterministic orstochastic (France and Thornley, 1984; Addiscott and Wagenet, 1985). Deterministicmodels generate predictions based on the presumption that the outcome of soil processescan be determined as the result of explicitly defined, non-variable relations, whereasstochastic methods incorporate random elements or probability distributions to describeinherent variability in the field (France and Thornley, 1984; Addiscott and Wagenet,1985). The development of stochastic models, or the inclusion of stochastic features indeterministic models, evolved in recognition of the highly variable nature of deterministicmodel parameters in soils (Nielsen et al., 1982; Addiscott and Wagenet, 1985). Citing alack of appropriate field studies necessary for validation, Addiscott and Wagenet (1985)maintain that stochastic models have found use primarily as research tools.Deterministic models can be labelled as functional or mechanistic. Models that makeuse of simplistic, empirical descriptions of soil water and solute transport, such as waterbudgets and mass balances, are classed as functional, while those adhering more rigorously to classical soil water and solute flow theory, such as Richards’ equation or theconvection-dispersion equation, are labelled mechanistic. Table 2.1 presents a surveyof several deterministic models. While the models presented may have idiosyncrasiesbeyond those listed, the given conspectus lists the salient features of each system. Forpurposes of comparison some generalizations were made; for example, models incorporating Richards’ description of water flow were designated as such, while models with analternative approach were broadly grouped as “other.” The purpose of such liberties wasprimarily to differentiate models based on classical soil water flow and solute transportChapter 2. LITERATURE REVIEW 18theory from models with other modes of description. Explanatory notes for the headingsemployed follow Table 2.1.Mechanistic models are often touted as inherently superior to functional models sincethey are based on rigorous theory rather than empirical evidence or intuition. However,given their less rigorous derivation, functional models usually have fewer constraints thanmechanistic models. As such, France and Thornley (1984) contended that it is alwayspossible to find a functional model that gives a better fit to a given set of data thana mechanistic model. Smith and Ferreira (1989) compared three mechanistic and twofunctional unsaturated flow models to field data. They found that while both model typesperformed reasonably well in predicting field water content values, the functional modelwas a poor flux predictor and the dilfusivity based mechanistic model had difficultyhandling a layered soil. In comparing 14 functional and mechanistic nitrogen cyclingmodels, de Willigen (1991) concluded that the mechanistic models yielded predictions forsoil moisture content and mineralization that were no better, and often worse, than thosemade by functional models. For moisture content calculations, the poor performanceof the mechanistic models was attributed to a lack of detailed information about thesoil hydraulic properties. Difficulty in accurately determining &(O) and K() (or K(O))relations was also found to limit the usefulness of mechanistic models in the review by DeJong (1981). Addiscott and Wagenet (1985) stated that the predictions of mechanisticmodels have not yet been determined as being superior to those of simpler functionalmodels.The type of simulation model used to study a given field problem is best determinedby considering factors such as the intent of the study and site specific conditions. If aproject is research oriented and sufficient data is available, a mechanistic formulation maybe the favoured approach. However, faced with a paucity of data and a need for generalmanagement recommendations, researchers may well justify the use of a simple functionalChapter 2. LITERATURE REVIEW 19Table 2.1: Comparison of several nitrogen and water flow models.Water SoluteModel Reference Type Dim. Rich. Other C-D Other[1] Freeze 1969 W 1 X[2] Pikul et al. 1974 W 2 X[3] Bulley & Cappelaere 1978 B 1 X X[4] Broughton & Foroud 1978 W 1 X[5] McGill et al. 1981 N 1 X[6] Bhat et al. 1981 B 1 X X[7] Tubbs & Haith 1981 B 1 X X[8] Walley Hussein 1982 W 1 X[9] Kanwar et al. 1983 B 1 X X[10] Wagenet & Rao 1983 B 1 X X[ii] Mochoge 1984 N 1 X[12] Otoma & Kuboi 1985 B 1 X X[13) Van Ommen 1985 S 2 X[14] White 1985b B 1 X X[15] Whitmore t Addiscott 1986 B 1 X X[16] Madramootoo & Broughton 1987 W 2 X[17] Johnsson et al. 1987 N 1 X[18] Richard 1988 B 1 X X[19] Barraclough 1989 B 1 X X[20] Gysi 1990a B 1 X X[21] Addiscott & Whitmore 1991 B 1 X X[22] Cabon et al. 1991 B 1 X X[23] Eckersten & Jansson 1991 N 1[24] Hutson Wagenet 1991 B 1 X X[25] Kersebaum & Richter 1991 B 1 X X[26] Vereecken et al. 1991 B 1 X XChapier 2. LITERATURE REVIEW 20Upper B.C. Lower B.C.Reference Rain E.T. Snow Runoff Imperm. Deep Flow GWT Tiles[1] X X X[2] X X X X X[3) X X X[4] X X X X X X[5)[6] X X X[7] X X X X X[8] X X X X X[9] X X X X X X X[10) X X X X X X[11) X X[12] X X X[13] X X X[14] X X[15] X X X X[16] X X X X X X[17] X X X X X[18] X X X X X X X X[19] X X X[20] X X X[21) X X X X X[22] X X X X[23][24] X X X X X[25] X X X[26] X X X X XChapter 2. LITERATURE REVIEW 21N Processes Simulated Soil Prop.Reference Ammon. Nit. Volat. Denit. Immob. Crop Homog. Lyrd.[1] X[2] X[3] X X X X X X[4] X[5] X X X X X X[6] X X X X X X[7] X X X X[8] X[9] X X X X X X[10] X X X X X[11] X X[12] X X X X X[13] X[14] X X[15] X X X X[16] X[17] X X X X X X[18] X X X X X X X[19] X[20) X X X X[21] X[22] X X X X X[23] X[24] X X X X X X X[25] X X X X X[26] X X X X X XChapter 2. LITERATURE REVIEW 22Reference B.P. Hyst. Valid. Sens. Comments[1] X L See Freeze &; Banner 1970 for validation.[2] F Hysteresis incorporation possible.[3] N parameters may vary by layer.[4] F[5] F Uses H20 submodel, see Paustian 1981.[6] F[7] F N parameters may vary by layer.[8] F[9] X F[10] F X See also Tillotson and Wagenet 1982.[ii] L NO column tests with C-D Equation.[12] F 1 value represents net NO change.[13] X Accounts for solute adsorption & decay.[14] X L X See also White 1985a.[15] X F X See also Addiscott & Whitmore 1987,Whitmore & Addiscott 1987.[16] F See also Madramootoo et al. 1987.[17] F Requires water submodel.[18] X X F X[19] X F X See also Barraclough 1989b.[20] X F X See Gysi 1990b for validation.[21] X F X See also Lord & Bland 1991.[22] F[23] X F X Modified Johnsson et al. 1987.[24] Modified form of LEACHM.[25] F See also Richter et al. 1980,Richter et al. 1985.[26] FChapter 2. LITERATURE REVIEW 23Notes for Table 2.1 column headings.Type Denotes simulation of water movement (W), nitrogen processes (N), both nitrogenand water (B), or a generic solute routine (S).Dim. 1, 2, or 3 dimensions.Water Model employs a form of Richards equation (Rich.), or includes simple budgets,unsaturated forms of Darcy’s law, etc. (Other).Solute Convection-dispersion equation incorporated (C-D), or solute transport described as product of flux and concentration or other transport expression (Other).Upper B.C. Upper boundary conditions model accounts for: Rain = static or dynamicapplication of water either through precipitation or irrigation, E.T.= evaporation,transpiration, or both, Snow = snowfall accumulation and melting, Runoff = overland flow.Lower B.C. Lower boundary conditions model accounts for: Imperm.= impermeablelayer at bottom of profile, Deep Flow = percolation out of bottom of profile, GWT= accounts for a ground water table, either static, dynamic, or quasi-dynamic, Tiles= artificial drainage can be described.N Processes Simulated Ammonification, nitrification, volatilization, denitrification,immobilization, crop uptake. Several models utilized one mineralisation value toaccount for both ammonification and nitrification.Soil Prop. Properties constant within profile or vary by layer.B.P. Bypassing flow simulated.Hyst. Hysteresis described.Valid. Model validation from lab (L) or field data (F).Sens. Sensitivity analysis performed.Chapter 2. LITERATURE REVIEW 24model. For N cycle modelling de Willigen (1991) advocated the use of functional overmechanistic models. De Jong (1981) claimed that because of their required high inputlevel, and our limited knowledge of crop water uptake and transpiration processes, useof mechanistic models for estimating soil water conditions and crop water use for a largegroup of soils is not currently feasible. Echoing this lack of quantitative knowledge withrespect to crop response to nitrogen, van Keulen and Stol (1991) stated that practical application of simulation models for farm level fertilizer practices is premature, though theyassented that such models help identify the most important parameters and processesinvolved. Hutson and Wagenet (1991) also view modelling in more of a research, ratherthan a predictive, role. They stated that “unexplained differences between simulationand measurement advance knowledge because they challenge our current assumptionsand prompt new directions for research . . . Increasing our understanding and intuitionconcerning the interaction between various components of the soil system is far moreimportant than the exact reproduction of measured data.”Nitrogen management models are often described as an economical alternative toexpensive, site-by-site field testing when making fertilizer N recommendations. As fieldtests give only a random indication of soil N supply, the use of simulation models inan advisory capacity has great potential (Kersebaum and Richter, 1991). Modellingoutput should not, however, be accepted unquestioningly; a knowledge of the model’sshortcomings and limitations is necessary to fully evaluate simulation results. Warningagainst over reliance on modelling, Philip (1991) paraphrased an unidentified Frenchscientist’s caveat: “Modelling is rather like masturbation— a pleasurable and harmlesspastime just so long as you don’t mistake it for the real thing.” However, as even thesimplest model requires validation, and possibly calibration, with field data, it is unlikelythat models will supplant field testing; in fact, the continued growth in modelling willlikely spur development of more comprehensive and detailed site studies.Chapter 2. LITERATURE REVIEW 25The variation in nitrogen processes from region to region casts doubt on the portabilityof N cycling models. Nonetheless, models developed for use in a specific geographicallocation may facilitate fertilizer N management decisions pertaining to that particularregion. Since water plays a pivotal role in soil N transformations, any model attemptingto simulate such processes must realistically describe soil water flow. Prior to its advisoryuse, field validation and calibration, as well as the determination of a model’s weaknesses,are necessary. As functional models yield results comparable to those of mechanisticmodels, and require less field data, they are well suited for incorporation in a fertilizer Nmanagement scheme. The present study details the development, calibration, validation,and sensitivity analysis of a deterministic, functional soil nitrogen model for potentialuse as an advisory tool.Chapter 3MODEL DESCRIPTIONThe nitrogen management model, as originally developed by Bulley and Cappelaere(1978), linked mathematical descriptions of several soil processes in an attempt to better understand the mechanisms of nitrogen transformations and movement within thesoil. As the physical, biological, and chemical properties of a soil are variable, and theprocesses of nitrogen transformation and water movement are complex, approximationsand simplifications were often necessary in modelling actual soil processes. While allowing the overall project to remain tenable, such assumptions may also have limited theapplicability and accuracy of the model.One conceptual shortcoming of the original model was its assumption that the water table always remained below the root zone, and that the soil profile being modelledwas freely drained. Under the climatic conditions prevalent in south coastal BritishColumbia such an assumption is often invalid. To improve the model, a more realistic portrayal of the soil water regime was incorporated into the nitrogen managementmodel. Richard (1988) developed deterministic, mechanistic\functional soil water movement routines that allowed for a fluctuating water table, ponding of water on the soil’ssurface and subsequent runoff. It was expected that replacing the straightforward soilwater algorithms of the Bulley and Cappelaere (1978) nitrogen model with the routinesof Richard (1988) would improve the overall accuracy of the nitrogen model.An additional problem with the original nitrogen management model was discoveredduring the amalgamation of the two models. As programmed, the original nitrogen26Chapter 3. MODEL DESCRIPTION 27management model did not always conform to the law of conservation of mass withrespect to water, nitrate, and nitrogen in the simulated soil profile. For example, thefinal mass of water within the soil profile at the end of any given timestep must equalthe amount of water present in the profile at the beginning of the timestep plus anynet inflow of water to the profile during the timestep. While in accordance with thismass balance for the majority of calculations, the original nitrogen management modelperiodically violated this principle. This, however, was probably a programming errorrather than a conceptual error.The current version of the nitrogen management model, combining the Richard (1988)water movement algorithms and the Bulley and Cappelaere (1978) nitrogen algorithms,addresses two significant shortcomings present in the original nitrogen managementmodel; it allows for the movement of a water table within the root zone and complies withthe principle of mass conservation. The model was written in FORTRAN for executionon IBM compatible personal computers. A detailed description of the enhanced nitrogenmanagement model is now presented.3.1 General Model DescriptionThe simulated soil profile accounts for a volume of soil one hectare in area and 12.1metres deep. As the model is one-dimensional, the one hectare surface area is implicit inthe calculations. The 12.1 metre deep profile consists of one ten-centimetre layer abovethe soil surface to allow for ponding, 20 ten-centimetre layers below the surface layer,and one ten-metre layer at the bottom of the profile to accomodate a fluctuating watertable. Soil properties are assumed to be homogeneous throughout the profile.To simulate a given field, the model requires information concerning the soil properties, local climate, crop, and manure applied. This data is read in by the main programChapter 3. MODEL DESCRIPTION 28from specified files. Table 3.1 gives a description of the input required by the model. Values for parameters not read in from files, such as the unsaturated hydraulic conductivitycoefficient a, are initialized via BLOCK DATA statements within the model. The valuesfor potential evapotranspiration and relative humidity read in from the climate file in Table 3.1 are generated from a climate calculation program that operates in tandem withthe nitrogen management model. A description of this program is given in Appendix A.Standard results generated by the model include crop yield and crop N content,losses of N via denitrification, volatilization, and leaching, and gains of N from additionof ammonium in manure, mineralization of soil organic matter, and mineralization ofmanurial organic matter. The model is also capable of yielding output for other soilprofile variables; with minor modifications, daily water table levels and both soil moisturecontent values and NO concentrations at various depths in the profile can be recorded.A standard model run proceeds as follows. After the user specifies the time period forwhich the simulation is to be performed (generally one to five years), model variables areinitialized using data read from the soil, manure, crop, and climate files. For each one-daytimestep in the simulation period both water and nitrogen calculations are performed.First, water flow out of, into, and within the soil profile is calculated, and the volumetricmoisture contents of each layer and the water table level are adjusted accordingly. Solutetransport is calculated next; it is assumed to occur only through advection, and thusthe movement of NO in the profile parallels that of the soil water. Simulation of soilN processes such as mineralization, denitrification, nitrification, and crop uptake follownitrate movement, and new nitrate concentrations are assigned to each layer for use incalculations during the subsequent timestep. Results of the simulation are written tosummary files at the end of the time period and execution of the model stops. Figure 3.1represents the overall flow of data in the model.Chapter 3. MODEL DESCRIPTION 29Table 3.1: Model input required.File Variable DescriptionSoil Owp Volumetric moisture content (VMC) at wilting point°FC1, °FC2 VMC’s at lower and upper field capacitiesVMC at saturation°ck, Ck Capillary conductivity coefficientsFraction of soil from which NO is excludedPb Bulk densityz,1 Depth to plough layerH8 Soil organic N content as a percentage of dry matterM5 Initial mineralization rate of soil organic N‘/-m, 0 8 points on b(O) curveManure X Excretion factorCollection loss factorL8 Spreading loss factorL8 Storage loss factorN0 Manure organic N fractionN Manure inorganic N fractionMm Initial mineralization rate of manure NHm Fraction of manure remaining as soil humusFN Feed N ingestedApplication dates and rates of manure appliedCrop H1, HF Initial and final depths of rootsHL Depth to which roots fully occupy soilg1, g2 Initial and final crop growth indicesP Maximum crop production ratioRmin, Rmax Upper and Lower limits of optimum soil fertility rangeD50, P250 50% reduction in fertility, deficiency and excessUmin, Umax Minimum and maximum N uptake rates‘FLX Inflection point for N uptake curvePlanting and harvesting datesClimate Tmin, Tma, Ta,, Daily minimum, maximum, and average temperaturesRh Daily value for relative humidityF Daily precipitation valuePET Daily value for potential evapotranspirationChapter 3. MODEL DESCRIPTION 30Figure 3.1: Nitrogen management model: Generalized flowchart.Chapter 3. MODEL DESCRIPTION 313.2 InitializationThe first operation performed by the model, following the input of file data, is variableinitialization. Saturated hydraulic conductivity values for each layer are calculated bythe function COND using the relation= (3.1)where O is the saturated volumetric moisture content of the layer, Ck is an empiricalcoefficient, and °ck is the 0 value at which K is 1 cm day1 (after Warrick et al., 1977,and Nielsen et al., 1973).In the original version of the nitrogen management model, Equation 3.1 was alsoused to calculate unsaturated hydraulic conductivity values. However, as an expressionfor unsaturated hydraulic conductivity (after Gardner, 1960) was inherent in the routinesof Richard (1988), Equation 3.1 was retained only as a means of providing a saturatedconductivity value for the following expressionK(0) = K3(-) (3.2)where c is an empirical coefficient. Function COND has thus been relegated to calculatingonly a saturated hydraulic conductivity value, and could have been replaced entirely witha K3 value read in from the soil file. Nevertheless, in order to retain compatibility withoriginal soil files to the greatest extent possible, function COND was preserved.Subroutine LYRVMC sets the initial volumetric moisture content value of each layerbased on the soil moisture characteristic curve and the distance from the layer to thewater table. After setting an initial water table depth, the distance from both the topand the bottom of each layer to the water table is calculated, and the resulting distancesrepresent the matric potential /-‘m at both positions under zero flow conditions. Two 0Chapter 3. MODEL DESCRIPTION 32values corresponding to these ‘bm values are read from the (O) curve, and an averaged8 value is assigned to the layer.Initial values for several other variables, including concentrations of soil NO andNH, saturated 0 values, and the soil N mineralization rate, are set based on values readin from the input files.3.3 Water Movement and Solute Transport RoutinesA general overview of the routines comprising the water movement algorithms is shownin Figure 3.2. While serving primarily as a gate to subsequent water routines, WATBALalso calculates the daily value for actual evapotranspiration as the sum of the actualplant water uptake and soil evaporation losses. These values are returned by subroutineMOVEMT, which in turn calls WATERMVT, the core of the Richard (1988) water flowalgorithms.WATERMVT is a complex set of instructions; its order of operations can be outlinedas follows:1. Calculate boundary flow conditions.2. Calculate downward flow. This occurs in two stages:a. Unsaturated flow.1. Predict a water table based on deep flow boundary condition.ii. Predict 0 values for each layer based on prevailing soil moisture conditions.iii. Calculate flow from surface to predicted water table.iv. Recalculate 0 values for each layer.b. Saturated flow.Chapter 3. MODEL DESCRIPTION 33Calculate1. moisture availableto roots2. plant wateruptake3. evaporation4. surface pondingand runoffFigure 3.2: Water movement routines.Chapter 3. MODEL DESCRIPTION 34i. Calculate water flow from the predicted water table to the bottom of theprofile.ii. Correct the predicted water table based on final 9 values in the profile.3. Calculate plant water uptake and evaporation.4. Calculate upward flow.a. Predict 0 values for each layer above the water table.b. Calculate upward flow from the water table to the soil surface.c. Recalculate 0 values for each layer.5. Calculate a final water table level.The intricacies of each of the preceding steps are discussed in the following subsections.3.3.1 Boundary Flow Conditions.Addition of water through precipitation and snowmelt is calculated in subroutine SKYWTR. On days when the average air temperature is below freezing, all rainfall is designated as snow and allowed to accumulate on the soil surface. When the air temperatureis above freezing all precipitation, along with any water derived from snowmelt, is allocated as a depth of water to the 10 cm layer above the soil surface. Potential snowmeltis calculated using the expression of Kattelmann et al. (1985)Fsmow = Tav(0.0045 + 0.013P) + 0.00025 (3.3)where Tav is the average daily air temperature and P is the daily precipitation value.Losses through evaporation and plant uptake are determined following downward infiltration of moisture and are described in subsequent sections.Chapter 3. MODEL DESCRIPTION 35Water lost or gained through the bottom of the profile is calculated prior to thedownward infiltration of water added to the soil surface. While the Richard (1988)model allowed the user to input a constant value to describe such deep flow, in revisingthe nitrogen management model an attempt was made to link deep flow to the regionalgroundwater system. This computation is site specific; at the Agassiz site flow into andout of the profile was assumed to be dictated by the level of the Fraser River. SubroutineLATFLO calculates deep flow using Darcy’s lawqiat= (3.4)where K,i is the hydraulic conductivity of the soil between the Agassiz field site and theFraser River, /L is the distance from the site to the river, and Hfr and are the levelof the Fraser River and the field water table above geodetic datum, respectively. If qiat ispositive water is flowing from the soil profile to the river; if qiat is negative the directionof flow is reversed. Mean monthly water levels for the Fraser River were obtained fromthe Inland Waters Directorate (1992) and are initialized via routine RVRLVL.At present, the model uses the same routine to calculate deep flow from the Sumasprofile. However, Kdl, zL, and Hf,. values were estimated for the Sumas site by fittingthe model’s predicted water table to the water table observed in the field for 1992 (seeChapter 5 for details).3.3.2 Downward Water Movement.After determining flow conditions at the profile boundaries, downward movement of waterin the profile is simulated. Subroutine DOFLO redistributes moisture vertically withinthe profile. Downward flow from the soil surface to the water table proceeds by themechanism of unsaturated flow. Saturated flow then moves moisture from the watertable to the bottom of the profile, deep flow as calculated in Equation 3.4 is removedChapter 3. MODEL DESCRIPTION 36from the bottom layer, and any excess water remaining at the bottom of the profile isredistributed back up through the soil profile. An interim water table depth is thencalculated.Flow in the unsaturated region is determined by a ‘predictor-corrector’ type methodreferred to by Richard (1988) as ‘dynamic equilibrium’. Prior to calculating unsaturatedflow, a ‘dynamic equilibrium’ moisture content, ODEQ, is predicted for each layer abovean estimated water table. Drainage within the profile is based on these °DEQ values.The 9DEQ value for any given layer may attain one of four values:1. 9, the initial 0 in a layer.2. OsEQ, the static equilibrium 0 based on an estimated water table and the &(9)curve.3. 0irn, the 9 value achieved after continuous drainage with no infiltration.4. OFL, the 0 value achieved after continuous infiltration with or without drainage.The initial volumetric moisture content in the layer, 9, is the value of 0 at the beginningof the timestep. Determination of the other possible 0 values is outlined below.Prediction of water table. Before the ‘dynamic equilibrium’ moisture content is calculated a new water table level is predicted. Starting at the layer containing the presentwater table, the water available for drainage is expressed asWavaji =f(z—zt) (3.5)where f is the drainable porosity of the layer, approximated byf = — °FC1 (3.6)Chapter 3. MODEL DESCRIPTION 37and z, and z are the depths to the water table and to the bottom of this layer fromthe soil surface, respectively. If Wavaji is greater than or equal to the amount of waterflowing from the profile as calculated in Equation 3.4, the estimated water table is left atits current position. However, if Wavaji < qj, the estimated water table is moved downone layer, is decremented by Wa,,aji, and a new value of Wai,aji is calculated asWavaji = f (3.7)where I.z is the layer thickness. This process continues, moving down through the profile,until qit The first layer to satisfy this condition is where the predicted watertable is placed.Calculation of8SEQ. The layer’s °SEQ value is calculated in a fashion similar to thatused by subroutine LYRVMC. The distances from both the top and the bottom ofthe layer to the predicted water table are calculated; these figures represent the matricpotentials t/)m at both positions under zero flow conditions. Two 0 values correspondingto these m values are read from the &(0) curve, and an averaged 0 is designated thelayer’s °SEQ value.Calculation of ODR. If a layer receives no moisture from the layer above, it will drain toa moisture content of1K(l — c)Lt’\ 1—0DR — . (3.8)This expression is derived from integrating Darcy’s law for vertical flow under a unitgradient for a time step of Lit, using the expression for unsaturated conductivity presentedin Equation 3.2 (after Richard, 1988).Chapter 3. MODEL DESCRIPTION 38Calculation of9FL• When a layer receives infiltration from above and does not lose anymoisture to the layer below, it will attain a moisture content of6FL1 = 0 + (3.9)where 0 is the moisture content prior to infiltration and q2n is the depth of water flowinginto the layer. However, if a layer undergoes both infiltration and drainage, it will reacha moisture content at equilibrium with the unsaturated flux from above under a unitgradient, described by Richard (1988) as°FL2 = Os (KtY (3.10)after Rubin (1966).Determining 0DEQ. The ODEQ value assigned to a layer is dictated by three factors:1.) the flow conditions of the layer to which ODEQ is being assigned, 2.) the relativemagnitudes of that layer’s 01, 8sx, °DTh 0FL1 and °FL2 variables, and 3.) the 01 and OsEQvalues of the layer below that to which ODEQ is being assigned. Possible flow conditionsaddressed by the model are no drainage and no infiltration, drainage and no infiltration,both drainage and infiltration, and no drainage and infiltration. Table 3.2 presents theflow conditions implied by the status of variables 0, OsEQ, and °FL1 in the two layers;the layer being assigned a ODEQ value and the layer below it are designated J and J+1,respectively. The process of determining a ODEQ value is summarized graphically inFigure 3.3.Subject to neither drainage nor infiltration, with 0 < 0SEQ, the ODEQ value for thelayer will be set equal to Oi. However, if 01 0sEQ such flow conditions no longerexist; drainage is allowed to occur and ODEQ is set to the greater of 0DR and OsEQ. Thismaintains the layer’s moisture content at or above its equilibrium value.Chapter 3. MODEL DESCRIPTION 39Table 3.2: Flow conditions determining ODEQ.Variable Status Flow Conditions ImpliedLayer J Layer J+l01 <OsEQ 01 OSEQ • No infiltration to J.• No drainage from J to J+l.0FL1 <OSEQ 01 OSEQ • Infiltration to J.• No drainage from J to J+l.01 < OSEQ 9 < 0g • No infiltration to J.• Drainage from J to J+l possible.0FL1 <OSEQ 0 < OSEQ • Infiltration to J.• Drainage from J to J+l possible.01 °SEQ 01 OsEQ • No infiltration to J.• Drainage from J to J+l possible.°FL1 9SEQ 01 °SEQ • Infiltration to J.• Drainage from J to J+1 possible.01 OSEQ 01 <OSEQ • No infiltration to J.• Drainage from J to J+l possible.°FL1 °SEQ 01 < 0SEQ • Infiltration to J.• Drainage from J to J+l possible.Chapter 3. MODEL DESCRIPTION 40NFigure 3.3: ODEQ determination process.Chapter 3. MODEL DESCRIPTION 41If the layer is subject to drainage only, 0DEQ is set to 0DR In the common event thata layer undergoes simultaneous infiltration and drainage, the ODEQ value is determinedas a linear combination of the °DR and 8FL2 values, calculated as0DEQ = 0FL2 \ + (1— .\)ODR (3.11)where ) is a weighting factor determined byqjX=mint1,maxiO,1 II (3.12)\ IUFL2UDR}IZJJUnder conditions of infiltration with no drainage, where °FL1 <OsEQ, °FL1 is assignedto9DEQ• But, if 0FL1 OSEQ, the flow condition is altered and drainage occurs; an intermediate 0 value based on both drainage and infiltration is calculated using Equation 3.11,and ODEQ is set to the greater of 0JNT and 0SEQ• Once a ODEQ has been assigned to eachlayer, unsaturated, downward flow is calculated.Unsaturated Flow. The amount of water infiltrating into a layer (j) from the layer above(j — 1) is set toqinf = flhlfl(Pjmf,Pdr) (3.13)where Pdr is the amount of water available for drainage from the layer above, calculatedasPdr = max(O, (O—0DEQ,3—1)) (3.14)and Pinf is the maximum flux that can infiltrate the layer, as(3.15)where is the average gradient under which flow occurs. The gradient is set to unityunless the intrapedal zone of layer j — 1 becomes saturated; under this condition ç isChapter 3. MODEL DESCRIPTION 42calculated using an approximation for the Green and Ampt model for vertical infiltration,after Richard (1988), asc= i+i12 +1’1’ (3.16)where iI’wf is the suction ahead of the wetting front and 1’ is a numerical calibrationparameter expressed asr= (3.17)where z is the distance from the soil surface to the bottom of layer j.One idiosyncrasy of the Richard (1988) unsaturated flow routine is that the K3 valuemay inadvertently affect the determination of the gradient. Saturation of the intrapedalzone in layer j — 1, the criterion used to determine the gradient calculation method,is assumed to occur if Pdr exceeds the Pinf value calculated using a unit gradient inEquation 3.15. A consequence of this is that if the K3 value is sufficiently large, Pdr willrarely exceed pinj; under such conditions the infiltration portion of the model is reducedto a simple capacity flow model.Saturated Flow. Water movement below the water table is governed by Darcy’s lawunder a unit gradient. Flow from the bottom layer of the profile is calculated in Equation 3.4.If the moisture content of any layer exceeds O following the unsaturated and saturated flow events, the layer’s excess moisture 0— Os is cascaded upward to the overlyinglayer. Excess water reaching the soil surface is lost as runoff. This adjustment correctsexcessive downward flow predictions (Richard, 1988). An interim water table is set usingEquation 3.27.Chapter 3. MODEL DESCRIPTION 433.3.3 Plant Uptake and Evaporation.Availability of water for plant uptake is dependent on a layer’s current moisture contentand root density. For each layer, the water available to plant roots is calculated asWravaji = /.Z . 0 ‘D(Zm, H1,H2)(1 — D(0, Owp,0min)) (3.18)where Zm is the distance from the soil surface to the midpoint of the layer, H1 is theroot depth at the current root density calculated in Equation 3.48, H2 is the current rootdepth, and 0min represents the lowest moisture content at which crop water availabilityis not affected, determined as0min (1 — 3) . 0wP + 3 0FC1 (3.19)where 3 is an empirical coefficient. The function D(a, b, c) produces a linear densitydistribution as0 acVz 1 a<b (3.20)b<a<cPotential plant water uptake is set toW,0 = (3.21)where Drz is the number of layers in the root zone. Actual plant water uptake is thendetermined asl’V = min(1’Vt, Wravaji) (3.22)and removed from each layer within the root zone.The depth of water potentially lost to evaporation E0 is the difference betweenpotential evapotranspiration PET and actual plant water uptake W. Evaporation islimited to the layer above the soil surface and the first soil layer. If sufficient water isChapter 3. MODEL DESCRIPTION 44not present on the soil surface to satisfy the potential evaporative loss, moisture is alsoremoved from the first soil layer up to an amount (8— Owp) Liz. Actual evaporationloss is then calculated asE0 if 01 . LSz1 E0Eact = E0 if 0 Lz1 + (02— Owp) Lz2 E0 (3.23)01 .zz+(02—Owp).tz ifO1 .zi+(O2—Owp)./.z <E0where subscripts 1,2 denote the layers above the soil surface and the first soil layer,respectively. If W,7,, = PET no evaporative losses occur.3.3.4 Upward Movement of Water.Prior to calculating the upward movement of water into a layer, an equilibrium moisturecontent for that layer is predicted. This value is determined in an identical fashion to°SEQ described in DOFLO. If the layer’s current moisture content is equal to or exceedsthe equilibrium value no upward flow occurs. Otherwise the potential upward flow iscalculated using Darcy’s law at steady stateqpotup = Keqv up it (3.24)where is an equivalent hydraulic conductivity and,is the gradient. Keqv is theharmonic average of the K(0) values for each layer between the water table and the layerto which flow ascends, computed by the relationKequ = (3.25)2=3Wt K(9)1where subscripts j,jwt denote the layer to which flow rises and the layer containing thewater table, respectively. The gradient is calculated asg — 1m,j + =it /Xz) 3 26up— jwt ( . )Chapter 3. MODEL DESCRIPTION 45where 1I’m,j is the matric potential of layer j. Actual upward flow is set to the lesser of(OsEQ — O)zz and qpotup.3.3.5 Water Table Level.While the initial water table depth is set in LYRVMC, depths on subsequent days arecalculated using subroutine WTABLE. Starting at the bottom of the profile and movingupward, this routine places the water table in the first unsaturated layer it encountersusing the following linear relation:Zt=ZZ (3.27)After redistributing moisture in the soil profile a mass balance is performed on eachlayer and a net flux for each layer is calculated. These flux values qj,net denote the netflow from layer j to j + 1 and are used to determine solute transport upon returning tosubroutine MOVEMT.3.3.6 Solute Transport.As with water, nitrate is allowed to move both up and down in the soil profile. Whethersolute is transported from layer j to j + 1 or vice-versa is determined by the value.If qj,n is positive nitrate flow is down from j to j + 1; if is negative the direction isreversed. A qj,net value of zero indicates no net movement of nitrate between the layers.Downward leaching of solute from each layer in the profile with a positive qj,net value isdetermined prior to the calculation of any upward transport resulting from negative qj,netvalues.Chapter 3. MODEL DESCRIPTION 46Downward Leaching. The nitrate concentration of a layer following infiltration from thelayer above can be expressed as+ qj_,netCj_iCj,inf A i (3.28)-I— qi—1,where C is the NO concentration of the layer and the subscripts mit, j, and j—1 denotean initial value, a value for the current layer, and a value for a layer above the current,respectively (Iuitii Oand C values are dened as the 0 and C values for the layer priorto moisture redistribution in WATERMVT.) If the layer’s qj,net value is positive the massof solute it loses to the layer below j + 1 isSj,a, = (1— ) C3,1 q,,rt (3.29)where-y is an empirical coefficient designating the fraction of solute not susceptible toleaching. A layer’s final nitrate concentration following the downward leaching of solutethrough the profile will be—+ qj—i,net) — Sd,,.j,d—(3.30)where 5dn = 0 if q3,net 0.Upward Movement. Upward transport of nitrate is calculated following the assigning ofa CJ,d value to each layer. For all layers for which qj,net < 0, the mass of solute gainedfrom the layer below, j + 1, will be= (1—_y) . IqtI (3.31)where is the layer’s nitrate concentration following a gain from below, given asj+l,blo‘—‘j+l,up ‘—‘j+1,d + i2 A .32vj+1’-zj+Chapter 3. MODEL DESCRIPTION 47If the qj—i,net value is negative, the mass of solute layer j loses to the layer above, j — 1,is= (1— ) C3, Iqi—i,netl (3.33)where is the nitrate concentration of layer j following the addition of Si,blO, calculatedas in Equation 3.32.The final nitrate concentration in each layer following both upward and downwardtransport of solute is expressed assj,upC,f Cj,up—(3.34)3 z3Leaching losses of nitrate from the root zone are summed prior to the return to subroutineWATBAL.3.4 Nitrogen CalculationsThe nitrogen management model endeavours to describe a portion of the nitrogen cyclebeginning with manure production and ending with plant uptake. In its attempt totrace nitrogen two classes of nitrogen processes are simulated; those involving manureproduction and storage and those occurring within the soil. The manure production andstorage calculations are concerned primarily with herd size and management and are ofperipheral value to the model. However, as the model currently requires the input ofparameters pertaining to such calculations in the manure file, a brief description of theiroperation is presented in Appendix B.The simulation of nitrogen processes in the soil profile proceeds in the following order:mineralization and volatilization, crop uptake, denitrification, and nitrification. Mineralization and nitrification are assumed to occur only within the top 20 centimetres of soil,and denitrification takes place over a 60 centimetre depth. Following the simulation ofChapter 3. MODEL DESCRIPTION 48all soil N processes the profile nitrate and ammonium contents are updated using simplemass balances.3.4.1 Mineralization and Volatilization.The initial soil organic N content is determined asH30 = H3 . z1. Pb (3.35)where zq is the depth to the plough layer and H8 is the soil organic N content as apercentage of soil solids. This tacitly assumes that all organic N is within the top 20centimetres of soil. Inorganic and organic nitrogen additions to the soil are calculated asNrnjnArnanNjLs336org = man owhere Nmin and Norg are the masses of inorganic and organic nitrogen applied, respectively, L3 is the spreading loss factor, and Aman is the mass of manure applied. Using asimple mass balance, the volatilization losses can then be expressed asNvoiat = Aman N1. (1 — L8) (3.37)and are accounted for prior to the addition of a manure load to the soil. The total gainof NH by the profile within one timestep is the sum of the applied mineral nitrogen andthe ammonium gained through mineralization of soil and manurial organic N.Manurial N. The mass of ammonium gained through mineralization of manurial organicN, Gmm, is expressed asGmm = Norg(ftt—fnrn) (3.38)where fnm is the fraction of organic N not mineralized and ft is the fraction of organicN remaining at the beginning of the timestep (identical to the f,m determined in theChapter 3. MODEL DESCRIPTION 49previous timestep). The fraction fnm is calculated asfnrn = fr(1(1- ())exp (_fr Mrnrm (339)where f, is the portion of manurial organic N susceptible to mineralization, rm is themineralization rate, and fr is a factor expressing the relation between the amount ofmanurial organic N ultimately mineralized and the initial rates of mineralization of soiland manurial organic N asHmMmMs 4r MmM8 3.0The mineralization rate rm is calculated as71rm = 1Om r rpHmwhere Qi is a Qio factor for mineralization, FpHm is an empirical coefficient accountingfor pH effects, and Fwm is a factor accounting for soil moisture conditions, calculated as1—V(O1,O,OFcl) if 0< °FC1Fum = V(0,0FC2, 0) if 0 > °FC2 (3.42)1 if 0FC1 0 OFC2)where 0 is the volumetric moisture content of the top soil layer.After performing mineralization calculations for each load of manure applied, theprogram checks to see if the organic N remaining for a given load is equal to or lessthan the background soil organic N level. If it is, this application is dropped from futuremanurial mineralization calculations and the remaining manurial organic N is consignedto the soil humus.Soil N. Mineralization of soil organic N is performed by a simple routine. The mass ofNHt gained from the soil organic matter isGms = M3 H30 rm Lt (3.43)Chapter 3. MODEL DESCRIPTION 50where all variables are as previously defined.3.4.2 Crop Uptake.Subroutine CROP calculates both crop dry matter gain and nitrogen uptake. In calculating plant N uptake no preference is given to nitrate or amnlonium; any nitrogenremoved via crop uptake is subtracted equally from the soil NO and NH pools.Crop uptake and growth proceed in the following sequence: calculation of the dailygrowth coefficient and the daily root depth and density, determination of nitrogen available for plant uptake, and crop N uptake and dry matter gain.Growth Coefficient and Root Depth and Density. The daily growth coefficient g isexpressed as the product of the crop response coefficient Cr8p and the adjusted potentialtranspiration Eadj (McDougall et al., 1989a). Cresp is calculated asCresp= f’) (3.44)where g1 and g2 are the initial and final crop growth indices, respectively, and gy is thesum of all g values since the beginning of the year. The Eadj value is determined usingEact V(01,°FC2, 9)Ead,= 1 (3.45)I—IL/IRh being the relative humidity and Eact the actual evapotranspiration as calculated inEquation 3.23.Root density is calculated asRd3 = D(gmd,gl,g2) gmd = gy — 2L (3.46)where Ymd is an estimate of the growth index at the middle of the current timestep. Boththe current root depth H2 and the root depth at the current root density H1 are thenChapter 3. MODEL DESCRIPTION 51calculated asH2 = (H1— HF) Rdens + HF (3.47)andH1 = (H1— HL) Rdens + HL (3.48)respectively. Variables H1, HF, and HL are all read in through the crop file and denotethe initial crop root depth, the final crop root depth, and the depth to which roots fullyoccupy the soil, respectively.Availability of Nitrogen. Nitrate available to crop roots is represented as(Drz+1)N03,avaii = (O Ci . D(Zm,i, H1,H2) (3.49)i=2where Drz is the number of layers in the root zone, C is the layer’s solute concentration,and Zm is the depth of the middle of the layer from the soil surface. All ammonium inthe profile, Namm, is assumed to be available for plant uptake. Total nitrogen availablefor crop uptake Navaji is then expressed as a mean soil N concentration over the rootingdepth by the equationNavait = (Namm + N03,avaii) (H1 + H2) Pb (3.50)where Pb 1S the soil bulk density.Plant N Uptake. The amount of nitrogen removed from the soil by plants, Ntk, isexpressed as the product of dry matter gain 0dm and the nitrogen uptake rate Nuprt.Dry matter gain is calculated as the product of the actual crop production ratio andthe daily crop growth coefficient gd Pact is expressed asPact = Fact,tt — Cgs Fcrop (R—??) (3.51)Chapter 3. MODEL DESCRIPTION 52where C is a coefficient describing the current growth stage of the crop, P07, is themaximum crop production ratio, the subscript t— Lt denotes a value from the previoustimestep, and 7? is a function of ‘relative reduction of dry matter production rate’ versusthe available soil N concentration, presented graphically in Figure 3.4. This function isC0Figure 3.4: Relative reduction of plant dry matter production rate vs. available soil Nconcentration.characterized by the following expressions:exp{ln2. (v1ü — i) (i — P’min )}RmnNa.,ai= 1— exp {_1n2(NatRmax) }0D R R E50 mm max 50Available Soil NNavaji <Navaji > RmaRmin < Navaji Rma(3.52)where 7?. is the relative reduction coefficient, Rmn and Rmax are the lower and upperbounds on the optimum soil N range, respectively, and E50 and D50 are the concentrationsof soil N at which a 50% reduction in fertility occurs due to excessive or deficient soil Nconditions.The coefficient C9. is found using a linear density function as= D(gmd,gl,g2) (3.53)Chapter 3. MODEL DESCRIPTION 53where g1, g2, and g, are the initial, final, and midday crop growth indices. C hasan initial value of 1 and decreases to 0 as crop growth increases toward its maximum(McDougall et al., 1989a).The crop nitrogen uptake rate is determined asNupri = Umin + (Umax — Umin)(1 — exp {. (Navail)2}) (354)2 ‘FLXwhere Umin and Umax are the minimum and maximum soil N uptake rates, and ‘FLX isthe inflection point on the N uptake curve.The soil nitrogen loss following plant uptake is distributed evenly between the soilnitrate and ammonium reserves using the ratio of crop N uptake Ntk to total soilnitrogen available Naaii. The amount of nitrate lost from the profile is expressed asNO3 = IV t-3,avail .‘availand the amount of ammonium lost is— r uptkNH4 — IVammI V availwhere Namm is the mass of ammonium present in the profile.3.4.3 Denitrification.The denitrification rate is calculated by an expression identical to that used to calculatethe mineralization rate, and is written asrdenit= Q(O.1Tav1).F,, . FHd (3.57)where Qiod is a Qio factor for denitrification, Fi-j-d is an empirical coefficient accountingfor pH effects, and Fd is a factor accounting for soil moisture conditions, calculated as( \2i I if 8 > 6FC1=— (3.58)(0 if8<OFciChapter 3. MODEL DESCRIPTION 54When performing the denitrification calculation for each layer the model takes intoconsideration the following factors:1. A denitrification rate at 10°C and optimum pH adjusted for soil moisture andtemperature conditions, calculated as rdenjt in Equation 3.57.2. A maximum denitrification rate when neither soil NO nor carbon are limiting,adjusted for soil NO and carbon conditions.3. A maximum time delay for soil denitrification adjusted for actual denitrificationrates.The adjusted time delay for denitrification is first calculated asdTD,max— rcN,t_tdTD = (3.59)rmaxdwhere dTD,max is the maximum time delay, r the denitrification rate adjusted for soilnitrate and carbon conditions, rmaxd the maximum rate of denitrification when neithersoil NO nor carbon are limiting, and the subscript t — /t denotes a value from theprevious timestep. Values for rcN are determined usingXNrcN=XXXX (3.60)where is an empirical coefficient and XN,Xc are factors accounting for soil nitrate andcarbon conditions, calculated asXN= NO3,8011 (3.61)andy C,011 V(Zm, Z20,z60)F-,L’iwhere N1 and C1 are ‘half-life’ constants for soil nitrate and carbon, respectively, NO38011is the concentration of nitrate in the soil layer, C8011 is a coefficient related to soil carbonChapter 3. MODEL DESCRIPTION 55content, Zm is the depth to the middle of the layer, and z20 and z60 are the depths atwhich soil carbon availability is at a maximum and a minimum, respectively (assumedto be 20 and 60 centimetres). The concentration of nitrate in the soil layer is expressedNO3,801 = (3.63)where C is the NO concentration of the soil solution.After its initial calculation, the time delay dTD is subject to further modificationsbased on the test criteria outlined below:0 ifdTD0dTD = t if dTD rdenit Lt (3.64)otherwiseIf the final dTD value is set to rdenjt zt, then a new value for rcN is determined usingrritrcN = rcN,,_ + (3.65)(LTD ,marand the potential loss to denitrification is then calculated asNdenitp = TCN rdenit• t + (rcN,t_t — rcN). (3.66)After calculating the potential denitrification loss Ndenitp, the expressionNd0 = mm (NO3,01 Nctenitp) (3.67)ensures that the maximum amount of nitrate lost to denitrification is limited to theconcentration of nitrate present in the soil. The total amount of nitrate lost over theprofile is then the sum of the loss from each layer, given asDTZ+1Ndt= 1=2(Nd,1 . Pb) (3.68)Chapter 3. MODEL DESCRIPTION 563.4.4 Nitrification.Nitrification proceeds in a fashion similar to that of denitrification. The nitrification rate,adjusted for soil moisture and temperature conditions, is= Ftempn Fpin Fwn (3.69)where FH is an empirical coefficient accounting for pH effects, and Ftempn and arefactors accounting for soil temperature and moisture conditions, respectively. Ftempn isdescribed as a ‘weak quadratic’ by McDougall et al. (1989a), and can be expressed asT+1 T0Ftempn=—lT.<0 (3.70)0 T<—1whereTx0.lTavl (3.71)and cr is an empirical coefficient. The factor is calculated as1— V(01,Owp,0FC1) if 01 < °FC1= V(01,°FC2, 03) if Oi > °FC2 (3.72)1 if °FC1 0 OFC2)An initial time delay value for nitrification TD is determined as—TD = TD,max (3.73)where rmajn is the maximum rate of nitrification when ammonium is not limiting, TD,macis the maximum time delay required to reach the maximum rate of nitrification afterheavy application of NHt to a soil containing no previous NH at 10°C, and r is theadjusted rate of nitrification accounting for soil NH concentrations, calculated as—Chapter 3. MODEL DESCRIPTION 57and YN is a factor expressed asNH4,301YN = 3.75IV 4,minwhere N,mn is a ‘half life’ constant and NH4,01 is the soil ammonium concentrationaveraged over the depth to the plough layer, z,1, asTTHammtV 4,soil —PbThe time delay value TD is adjusted according to the relation0 flTD<OTD = • Lt if TD rt• L.t (3.77)TD otherwiseand if TD is set to rm t a new r value is calculated asrmaxn r• Ltr = + (3.78)72TD,maxThe potential loss of ammonium to nitrification is calculated as= r,-, r• Lt + (r,t_t — r) (3.79)and the actual concentration of ammonium lost from the soil above the plough pan is setto the minimum of and NH4,301 multiplied by the bulk density and the depth tothe plough layer.Chapter 4FIELD SITES AND EXPERIMENTSTo assess the accuracy of the new version of the nitrogen management model, and toevaluate the effectiveness of the improved soil water algorithms, output from both theold and new versions of the model have been compared to field data gathered at twolocations in south coastal British Columbia. Field data was collected under a separatestudy, described in Zebarth and Paul (1994).4.1 Field Sites and Treatment DescriptionsAn experiment to investigate the influence of the time of manure application on crop yieldand soil nitrogen processes was conducted at two field sites in the Fraser Valley from fall1991 to fall 1992. The Agassiz site was characterized by a medium textured soil, while theSumas site was located on a coarse textured soil (Zebarth et al., 1993). Six treatmentsin four replications were used at each site consisting of a control receiving no manure orfertilizer, fall application of 300 kg N ha’ as manure plus spring application of 300 kgN ha’ as manure, fall application of 600 kg N ha’ as manure, spring application of300 kg N ha’ as manure plus spring application of 100 kg N ha’ as inorganic fertilizer,spring application of 600 kg N ha’ as manure, and spring application of 200 kg N ha’as inorganic fertilizer (Zebarth et al., 1993). Table 4.1 presents a summary of the fieldtreatments.Application of manure occurred on 3 October 1991, 7 April 1992, and 30 September1992 at the Agassiz site, and 2 October 1991, 21 April 1992, and 28 September 199258Chapter 4. FIELD SITES AND EXPERIMENTS 59Table 4.1: Summary of field treatments.Fall 1991 Spring 1992 Fall 1992Treatment Manure Fert. Manure Fert. Manure Fert.______________kg N/ha kg N/ha kg N/ha kg N/ha kg N/ha kg N/haC 0 00 00 0FM300+5M300 300 0 300 0 300 0FM600 600 0 0 0 600 0SM300+SF100 0 0 300 100 0 0SM600 0 0 600 0 0 0SF200 0 0 0 200 0 0at the Sumas site. Liquid dairy manure was surface applied with a vacuum tank liquidmanure spreader and incorporated within 24 hours by discing or ploughing (Zebarth etat., 1993). Inorganic fertilizer was surface applied at Agassiz on 5 May 1992, and atSumas or’ 14 May 1992. Plot size at the Agassiz site was 6 x 30 m, whereas the Sumasplots were 9 x 30 m.Silage corn was planted at both sites. Sumas was seeded with Pioneer 3901 on 13May 1992 and hand harvested on 9 September 1992. The Agassiz site was planted on 6May 1992 with Funks 4066 and hand harvested on 17 September 1992 (Zebarth et al.,1993).4.2 Field MeasurementsWater table levels were measured using one well at each site. Wells were constructedby lining an augered hole (outside diameter 175 mm) with a two-metre section of pipecontaining 12.5 mm diameter perforations. A float was lowered down the well to the freewater surface, and a Stevenson recorder connected to the float allowed for continuousmonitoring of water table fluctuations. Recording of water table levels was not continuousChapter 4. FIELD SITES AND EXPERIMENTS 60for the study period; wells were removed for several weeks during spring planting and fallharvest and were not always reinstalled at the same locations and depths. In addition,well measurements were susceptible to errors such as freezing soil, frost heaving andsloughing of the well walls. Agassiz well depths varied between 1.37 and 1.87 metres,while the maximum Sumas well depth was approximately 1.25 metres.Soil cores at depths 0—30, 30—60, 60—90 and 90—120 cm were collected from the Agassizsite on 14 May 1992 for use in determining the soil moisture retention curve. Moisturecontents were determined at tensions of 0, 0.25, 0.5, 1.0, 3.3, 10, 30, and 150 m ofwater. Measurements for 0—1 m of water were performed with Tempe cells and a tensiontable, while measurements at tensions exceeding 1 m of water were conducted with apressure chamber. Figure 4.1 presents the soil moisture retention curves for each depthand the average curve used in model simulations. As a soil moisture characteristic curvewas not measured for the Sumas site, one was approximated during model calibration.The estimated Sumas ib(O) curve and a description of its derivation are presented inSection 5.1.The cores used in plotting the i,b(O) curve for Agassiz were also used to determine bulkdensity values for the soil. For the Sumas site p values were derived from denitrificatioricores collected by Paul (1993). Table 4.2 presents bulk density values at each depth andthe average Pb used in model simulations. Paul (1993) also provided estimates of thesoil organic matter and carbon content at each site. In February 1992 the soil organicmatter content for the Agassiz and Sumas sites were 8.6% and 2.5%, respectively. Carboncontent of the soil organic matter was estimated at 40% for each site, and a C:N ratio of11 was assumed (Jenkinson, 1988).Tensiometers were installed at both sites at depths of 30, 60, 90, 120, 150 and 180cm, with four replicates per depth. Readings were taken on the dates specified in Table 4.3. Soil samples were obtained from both sites at depths of 0—15, 15—30, 30—60, andChapter 4. FIELD SITES AND EXPERIMENTS 6160—90 cm on the dates specified in Table 4.4. These samples were used to determineboth soil moisture content and inorganic N content. Moisture content was determinedgravimetrically, while samples used in determining NO and NH contents were prepared by suspending 20 g of moist soil in 100 ml of 2M KC1 extract, agitating for onehour, and filtering through #40 Whatman filter paper. The filtrate was then analysedcolorimetrically using flow injection analysis.Climatic data for the Agassiz site was collected at the meteorological station runby the Agriculture Canada Research Station in Agassiz, and Sumas climatological datawas obtained from the Environment Canada weather station at the Abbotsford Airport,approximately 20 kilometres from the study site.Saturated and unsaturated hydraulic conductivity values were not measured at eithersite; values used in model simulations were estimated during model calibration. Detailsare given in Section 5.1.Chapter 4. FIELD SITES AND EXPERIMENTS 620.7 -406_____:: 1350.3- I I0 2 4 6 8 10 12 14Tension (- m of water)0Cm . 30-60 cm. 60-90cm. 90+ cm. AverageFigure 4.1: Agassiz i/.’(O) curves. Inset graph illustrates the dry end of the curves.Table 4.2: Agassiz and Sumas soil bulk density valies.Agassiz SumasDepth (cm) Pb (g cm3) Depth (cm) Pb (g cm3)0—30 1.07 0-15 1.1030—60 1.32 15-30 1.2560—90 1.36 30+ 1.4090+ 1.42Average 1.29 Average 1.33Chapter 4. FIELD SITES AND EXPERIMENTS 63Table 4.3: Tensiometer reading dates.Site Year Reading dateAgassiz 1991 Oct. 9, Oct. 21, Nov. 6, Nov. 20, Dec. 4, Dec. 191992 Jan. 15, Feb. 12, Mar. 11, Mar. 25, May 21, May 26June 1, June 15, July 13, Aug. 19, Sept. 14, Oct. 14Oct. 26, Nov. 10, Nov. 24, Dec. 10Sumas 1991 Oct. 10, Oct. 24, Nov. 5, Nov. 19, Dec. 3, Dec. 181992 Jan. 14, Feb. 10, Mar. 10, Mar. 24, Apr. 2, May 28June 3, June 16, July 14, Aug. 18, Sept. 8, Oct. 13Oct. 26, Nov. 9, Dec. 7Table 4.4: Soil sampling dates for determination of moisture content and inorganic N.Site Year Sampling DatesAgassiz 1991 Sept. 30, Oct. 21, Nov. 6, Nov. 20, Dec. 4, Dec. 181992 Jan. 16, Apr. 6, Apr. 22, May 4, May 19, June 1June 15, July 14, Aug. 17, Sept. 21, Sept. 24, Oct. 14Oct. 28, Nov. 10, Nov. 24, Dec. 8Sumas 1991 Oct. 1, Oct. 22, Nov. 5, Nov. 19, Dec. 3, Dec. 171992 Jan. 14, Apr. 8, May 5, May 20, June 2, June 16July 15, Aug. 18, Sept. 17, Sept. 23, Oct. 13, Oct. 27Nov. 9, Nov. 23, Dec. 7Chapter 5CALIBRATION AND SENSITIVITY ANALYSIS5.1 CalibrationAs this project focussed on the water balance portion of the model, the criteria usedto calibrate the model to each field site were the predicted water table levels. Visualcomparison of the plotted field and predicted water table levels for 1992 was used todetermine the extent of calibration required. Variables manipulated during calibrationwere the Kdl, Hfr and iL values from subroutine LATFLO. Saturated and unsaturatedhydraulic conductivity parameters, Ck, °ck and cv, were also estimated.5.1.1 Agassiz.Parameters Ck, 8ck and cv were chosen such that the soil conductivity characteristics ofthe new model resembled those of the old. To ensure that saturated and unsaturatedhydraulic conductivity values for both models were realistic, the magnitudes of K3 andK(O) were checked against those of a similar soil. As the Guelph loam of Campbell (1974)had a curve similar to that measured on the Agassiz soil, Guelph loam hydraulicconductivity parameters were modified such that the K3 and K(O) values of the two soilswere comparable. Similarity between new model and old model hydraulic conductivityfunctions was maintained. The resultant Agassiz K3 value of 1.5 x 10_i m d—’ washalf that given for the Guelph loam, and values of K(O) were within the same range ofmagnitude as those derived by Campbell (1974).64Chapter 5. CALIBRATION AND SENSITIVITY ANALYSIS 65Simulation of the 1992 water table levels at the Agassiz site was performed using26-year, monthly average Hf values from the Inland Waters Directorate (1992), adistance measured on a topographical map, and a Id1 value determined by trial anderror. Final calibrated and observed 1992 water table levels for Agassiz are shown inFigure 5.1. The maximum error between the measured and simulated water table levelswas 0.88 metres on day 327, and the average error was +0.27 metres. Average error wascalculated asE =— oil (5.1)where E is the average error, p, and o are the predicted and observed daily water tablelevels, respectively, and W is the number of observations..1.I0...I1280 311 342ITJulian DayFigure 5.1: Calibrated water table levels, Agassiz 1992. Flat sections of the field plotindicate periods during which the well was dry, and gaps represent periods when watertable levels were not recorded. Bars denote daily precipitation.Chapter 5. CALIBRATION AND SENSITIVITY ANALYSIS 66Figure 5.2: Estimated Sumas retention curve. ‘Field’ legend denotes paired /‘ and 0values collected on the following dates: 16 June, 18 August, 13 October, 9 Novemberand 7 December, 1992.5.1.2 Sumas.As a (0) curve was not measured on this soil, one was created during calibration bymodifying a Rehovot sand b(0) curve (Mualem, 1976) in accordance with ‘field measured’ 0 and values. The ‘field measured’ /i(0) curve was approximated by matchingtensiometer and soil moisture content values measured at the Sumas site on the samedates. (It is noted that the cores used to determine soil moisture contents were not necessarily collected from the areas of the field where the tensiometers were located. Thus,paired 0 and /‘ values are best viewed as a tough estimate of the wet end of the Sumas0) curve.) Figure 5.2 gives the Rehovot b(0) curve, ‘field measured’ 0) values, andthe b(0) curve derived for Surnas.Both the deep flow and hydraulic conductivity variables were estimated during thefitting of the model to the field water tables. Values for Ck, 0ck and o, as well as the threeTension (m of water)Chapter 5. CALIBRATION AND SENSITIVITY ANALYSIS 67A. .lnn— Field — New Model:NJr rAt\M40-1.5i1111 1.1.]..__•.__- .1 32 63 94 125 156 187 218 249 280 311 342Julian DayFigure 5.3: Calibrated water table levels, Sumas 1992. Flat sections of the field plotindicate periods during which the well was dry, and gaps represent periods when watertable levels were not recorded. Bars denote daily precipitation.deep flow variables Kdz, Hf,., and zL, were determined by trial and error. Final calibratedand observed 1992 water table levels for Sumas are shown in Figure 5.3. Maximum andaverage errors between the calibrated and field measured values were 0.68 metres on day179, and ±0.17 metres, respectively. The final K, value of 1.13 m d’ was reasonablefor this coarse textured, sandy soil (Bear, 1972).5.2 Sensitivity AnalysisA sensitivity analysis was performed in order to assess the effects of several variableson model output. Testing was limited to the following factors: the soil (O) curve, thesaturated and unsaturated hydraulic conductivity coefficients Gk, 0ck and , the hydraulicconductivity used in calculating deep flow from the profile Kdl, bulk density Pb, and thefinal root depth of the crop HF. As the purpose of this study was to examine the soil waterChapter 5. CALIBRATION AND SENSITIVITY ANALYSIS 68regime, analysis was restricted primarily to hydrological parameters. The modificationof factors used in the nitrogen calculations, though affecting output such as crop yieldand losses to denitrification, was not expected to substantially affect the simulated soilhydraulic behaviour.Variables under consideration were modified within a 40% range of values used instandard input files for the Agassiz site (standard values being those derived after calibration of the water table; see Appendix D). It was felt that this range made manifestthe effects of extreme soil values while still maintaining the factors within a realisticrange. However, a word of caution is in order. Because of the relationship between thesoil moisture retention curve and the saturated and unsaturated hydraulic conductivities, modifying one of these values in isolation can result in unrealistic conditions; anexample using a 40% decrease in the &(O) curve is presented. The standard Agassiz soilfile contains O and Owp values of 0.54 and 0.13, and Ck and °CK values of 45 and 0.48,respectively. Shifting only the i&(O) down by 40% results in values of 0.32 and 0.08 forO and 9wp. While this decreased &(O) curve may still represent a real soil, using O andEquation 3.1 results in a saturated hydraulic conductivity value on the order of iO cmThis K8 value renders the soil impervious (Bear, 1972). The generation of such anunrealistic soil condition while using realistic soil properties illustrates the caution thatmust be exercised in altering model variables; it may be best to modify all soil characteristics simultaneously as opposed to single value manipulatioll. Nevertheless, for thepurposes of this analysis variables were modified singly; this allowed the robustness ofthe model to be tested. The results of trials where several factors were modified at onceare presented in Appendix C.A five-year simulation was run for each modified variable using Agassiz site conditionsand 1992 climate data as the benchmark. The effects of variable modification followingChapter 5. CALIBRATION AND SENSITIVITY ANALYSIS 69year one of the five-year simulation are presented in Table 5.1. Overall trends in subsequent years exhibited similar patterns of response. For display purposes the actual modelresults were rounded to the nearest whole number; therefore, small percentage changesin model output may not be reflected in the corresponding table values.Figures 5.4, 5.5, and 5.6 illustrate the effects of altering the Kdl, i&(O), and Ck, °CKvalues on water table height. While not assessed against standard model output suchas crop yield and leaching losses, the effect of varying the initial water table depth ispresented in Figure 5.7. Modification of the unsaturated hydraulic conductivity coefficient a, the soil bulk density Pb, and the final crop rooting depth HF had minimaleffects on water table levels, although a 40% increase in HF lowered the water table byapproximately 10 centimetres for a brief period in July.Changes in the Kdl value primarily affected the level of the water table and not theshape of the curve. Alteration of either the &(O) curve or the Ck, 0CK parameters affectedboth the shape and level of the water table. With the exception of the +40% setting,the initial water table depth only affected the water table levels during the first threemonths of the year. Regardless of the starting water table depth all plots in Figure 5.7converge by the end of one year. The common water table level past this point was aresult of the stabilizing conditions that the regional hydrology imparts on the field site.Alteration of most model variables produced a near linear response in model output.Thus, if a parameter increase of 20% had an effect on an output value, an increase of40% produced an effect of approximately twice the magnitude of that obtained by the20% increase. A notable exception was the nonlinear response to alteration of b(O) andCk, 0CK values; this was due in part to the unrealistic conditions generated by modifyingthese parameters individually.Crop yield was affected most by changes in the &(O) curve and Ck, OCK values.Mineralization of both manurial and soil organic N were sensitive to b(O) and Ck, 0cKChapter 5. CALIBRATION AND SENSITIVITY ANALYSIS 70Table 5.1: Sensitivity analysis. Model output values following a 1 year simulation.Value Crop Mineralization LossesYield N Cont. Manure Soil Denit. Leach.%zBmrk. 15 174 42 141 74 6+40 18 (17) 207 (19) 44 (5) 155 (10) 9 (-88) 10 (70)+20 18 (17) 207 (19) 44 (5) 156 (11) 9 (-87) 8 (46)-20 4 (-71) 82 (-53) 10 (-77) 16 (-89) 63 (-15) 0 (-100)-40 12 (-18) 140 (-19) 25 (-39) 83 (-41) 45 (-39) 0 (-100)Ck, °ck+40 17 (11) 197 (14) 28 (-33) 99 (-30) 11 (-85) 0 (-100)+20 5 (-69) 84 (-51) 10 (-75) 17 (-88) 67 (-10) 0 (-100)-20 17 (15) 204 (17) 44 (5) 155 (10) 10 (-87) 12 (113)-40 17 (15) 204 (17) 44 (5) 156 (10) 8 (-89) 13 (136)a+40 13 (-16) 151 (-13) 38 (-10) 120 (-15) 96 (30) 3 (-43)+20 13 (-11) 158 (-9) 39 (-7) 127 (-10) 85 (15) 4 (-25)-20 16 (6) 184 (6) 43 (2) 147 (4) 53 (-28) 7 (32)-40 17 (10) 195 (12) 43 (3) 151 (7) 30 (-59) 10 (75)Kdz+40 15 (1) 176 (2) 42 (0) 143 (1) 72 (-3) 6 (11)+20 15 (1) 175 (1) 42 (0) 142 (1) 73 (-1) 6 (7)-20 15 (-1) 171 (-2) 42 (0) 138 (-2) 77 (4) 4 (-27)-40 14 (-4) 165 (-5) 41 (-2) 128 (-9) 90 (21) 3 (-55)Pb+40 15 (-1) 158 (-9) 42 (0) 197 (40) 103 (39) 8 (45)+20 15 (1) 166 (-4) 42 (0) 169 (20) 89 (20) 7 (23)-20 15 (-1) 181 (4) 42 (0) 113 (-20) 60 (-20) 4 (-23)-40 14 (-5) 188 (8) 42 (0) 85 (-40) 45 (-39) 3 (-46)HF+40 14 (-9) 153 (-12) 42 (0) 141 (0) 83 (12) 0 (-95)+20 14 (-5) 163 (-6) 42 (0) 141 (0) 79 (7) 2 (-73)-20 16 (6) 187 (8) 42 (0) 141 (0) 68 (-8) 10 (80)-40 17 (12) 202 (16) 42 (0) 141 (0) 61 (-17) 22 (286)Chapter 5. CALIBRATION AND SENSITIVITY ANALYSIS 71—— Relative Plot Position:—40%_1J5. —20%—1• +40%, —1.5—211111111111 liii.,... 1111111111111111111111111 IllIllIllIll 111111111111 31 61 91 121 151 181 211 241 271 301 331 361Julian DayFigure 5.4: Effect of ‘(dl on water table levels.—3.5 •.......,31 61 91 121 151 181 211 241 271 301 331 361Julian DayFigure 5.5: Effect of ‘(O) curve on water table levels.Chapter 5. CALIBRATION AND SENSITIVITY ANALYSIS 72—31 31 61 91 121 151 181 211 241 271 301 331 361Julian DayFigure 5.6: Effect of Ck and °CK on water table levels.0.50 —I Relative Plot Position—0.5—40%-20%SInd.—1• +20%+40%—1.5—2.5 I 1111111111•1,, liii..,,, liii..., 111111111111111111 I 111111111111 31 61 91 121 151 181 211 241 271 301 331 36_Julian DayFigure 5.7: Effect of initial water table depth on water table levels.Chapter 5. CALIBRATION AND SENSITIVITY ANALYSIS 73values. Modifying Pb affected the mineralization of soil organic N; this was a result ofthe role Pb plays in calculating initial soil organic N content in Equation 3.35.Model predictions for denitrification were susceptible to changes in çb(O), Ck, 0CK, a,and Pb values. Modification of HF and ‘cdl also affected denitrification, but to a muchlesser extent. Loss of nitrate from the root zone responded to every variable modification.In summary, all model output was sensitive to fluctuations in both the soil moisture characteristic curve and saturated hydraulic conductivity coefficients. Due to theircontrolling influence on the soil water regime, modification of these two parameters hadthe greatest impact on water table levels and soil 0 values. As soil moisture content iscentral to most model calculations, all model results were subsequently affected. Withsome exceptions, particularly leaching and denitrification, the model was not as sensitiveto alteration of a, Kdj, Pb, and HF as it was to alteration of b(0), Ck and °ckChapter 6MODEL SIMULATIONS AND FIELD OBSERVATIONSThis chapter details the application of both the original and modified manure management models to the field sites described in Chapter 4. Predicted values of soil water,nitrogen and crop yield from both models are compared with field measurements foreach site. Model input parameters used to simulate the two field sites are listed inAppendix D.It is noted that the new model functions on a daily basis and the old model functionson a weekly basis. As the old model could not produce daily soil water and nitrogenpredictions, weekly predictions nearest the specified plotting dates were used in the comparative plots. Old model predictions were always within two days of the designateddates.6.1 Results6.1.1 AgassizAt the Agassiz site, sufficient input data existed to permit model simulation of soil waterand nitrogen processes over the three-year period 1991—1993. However, complete fielddata was not available for this period; for the fall of 1991 and the year 1992 both soilmoisture and nitrogen data were available, but for 1993 data was limited to field watertable measurements.74Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 75Soil Moisture.Water Table Levels. The original model assumed a water table depth well below thestandard crop root zone of 1.30 metres and did not simulate water table fluctuations. Forthis reason, only new model predictions were plotted against field measurements. Plotsfor 1991 and 1993 are presented in Figures 6.1 and 6.2, respectively. While water tablelevels were recorded continuously by the field instrument, nly average daily levels wereused in the field plots.Water table levels for the fall of 1991 were measured and simulated from 20 November(day 324) through until 31 December (day 365), a 42 day period. In general, though themodelled variations of the water table paralleled those of the field, the magnitudes of thesimulated and measured daily water table levels did not correspond. Relative to fieldvalues, model predicted water levels were too low for the first 21 days and too high forthe remaining 21 days. The maximum error between the measured and simulated watertable levels was 0.79 metres on day 340, and the average error was ±0.37 metres. TheSC04.Julia DayFigure 6.1: Water table levels, Agassiz 1991. Bars denote daily precipitation.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 76Figure 6.2: Water table levels, Agassiz 1993. Flat sections of the field plot indicateperiods during which the well was dry, and gaps represent periods when water tablelevels were not recorded. Bars denote daily precipitation.S0001)Julian Daymaximum daily field water table fluctuation was 0.55 metres on day 339.Measurement of 1993 water table levels occurred from 1 January (day 1) throughuntil 13 September (day 256). The soil was frozen during days 14—24, and frost heavingand soil sloughing near the end of January combined to decrease the depth of the well byapproximately 0.16 metres. As for 1991, the 1993 simulated and measured water tablelevels were of the same general shape, though not of the same magnitude. For the firsttwo-thirds of the comparison period the model-predicted values were usually below thoseof the field, whereas for the remainder of the period model levels generally exceeded thefield measured values. The maximum error between the measured and simulated watertable levels was 0.98 metres on day 83, and the average error was ±0.38 metres. Themaximum daily field water table fluctuation was 0.35 metres on day 27.Figure 6.3 plots the water table levels predicted by the new model against the levelsmeasured in the field for 1991 and 1993, and Figure 6.4 presents the same plot for theChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS2.477calibrated water table of 1992. Visually, the fit for 1992 is better than that for 1991 and1993; the average error for 1991 and 1993 is 0.08 m greater than that for 1992.Total Water to 90 Centimetres. Figures 6.5 and 6.6 plot the total depths of water withinthe top 90 centimetres of soil predicted by both models against the total depths measuredin the field for 1991 and 1992. Data for the plots was obtained using the average fieldmoisture content values presented in the soil moisture profiles (see Figures 6.7 — 6.12).For 1991, the new model estimates were better than those of the old on two of the sixdays plotted. Both new and old models were better at predicting high moisture contentsthan low. The maximum and average errors between measured and predicted values, forthe new and old models, respectively, were 3.12 and ±1.18 centimetres, and 2.85 and±1.13 centimetres. By comparison, the maximum and average ranges of field measuredmoisture content values were 1.20 centimetres and ±0.83 centimetres, respectively.For the sixteen 1992 days plotted, new model predictions were noticeably better thanAA A&AAAAAAAAAAAAAAA —AA0.8 1.2 1.6Measurea (in. below soil surface)Figure 6.3: Water table levels, measured vs. predicted, Agassiz 1991 and 1993. Averageerror is ±0.38 metres.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 782.2-AAiLAAAA*18- AA AAAA AAA AA A1.4- AA AAAAA AA1• A A A AAAAAA0.6’ AAA AA0.2- AA—0.20 0.4 0.8 1.2 1.6 2 2.4Measured (in. below soil surface)Figure 6.4: Water table levels, measured vs. predicted, Agassiz 1992. Average error is+0.30 metres.tA -42- A40 AC)34 ><Old ANew><3232 34 36 38 40 42 44Measured (cm. of water)Figure 6.5: Comparison of measured and predicted soil water to a 90 cm depth for thenew and old models, Agassiz 1991.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 79Figure 6.6: Comparison of measured and predicted soil water to a 90 cm depth for thenew and old models, Agassiz 1992.old model predictions on six days; on two days new and old model predictions were similar. New model estimates were better than old model estimates at low moisture contents,while the old model was better than the new at predicted high moisture contents. Themaximum and average errors between measured and predicted values, for the new andold models, respectively, were 4.00 and ±1.87 centimetres, and 8.10 and ±2.13 centimetres. Maximum and average ranges of field measured moisture content values were 5.83centimetres and ±1.65 centimetres, respectively.Soil Moisture Profiles. All 1991 soil moisture profiles, given in Figures 6.7 through 6.8,were for wet, fall months. Overall, there was a reasonable fit between the estimates ofboth models and the field measured values. The new model appeared to be marginallybetter than the old at capturing the pattern of moisture distribution in the soil, while theold model appeared to better reproduce the absolute values of field measured moisturecontents.4240S383638 40Measured (cm. of water)Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 80Volumetric Moisture Content Volumetric Moisture Content0.3 0.4 0.5 0.6 0.3 0.4 0.5 0.60- 0 0--0 -- ua ‘ -15- n.ç 15- c.--- )c -30- ()Et 30- ci. >- c5 n.ç o.45 Ql 45- C Nc’60 4c . 50- 0thi -75 t0 75-0s -cr E90 6 ci 90- 6SEPTEMBER 30 OCTOBER 21Volumetric Moisture Content Volumetric Moisture Content0.3 0.4 0.5 0.6 0.3 0.4 0.5 0.6I I I I0- l.O 0- I4J 5IcJILI15 15--.• 4. - 4J•30 30--.f-. . --4. -60- 60-C Y1 L’ -75 • l 75-•-0 • 1>90 () C I t’ 90-NOVEMBER 6 NOVEMBER20Field Maximum Field Minimum New Model Old Model0-—--- ----0-----Figure 6.7: Soil moisture profiles, Sept. 30—Nov. 20, 1991, Agassiz.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 81p°.D.xxD.J.‘4,14,DECEMBER 4Volumetric Moisture Content0.4 0.5 0.50•ci.D.tp>àj.j.cij.tqri.J.J.r.f1n,IDECEMBER18Field Maximum Field Minimum New Model Old Model0----0----Figure 6.8: Soil moisture profiles, Dec. 4 and Dec. 18, 1991, Agassiz.The 1992 soil moisture profiles are presented in Figures 6.9 through 6.12. Agreementbetween the output of both models and the field measured values was reasonable, butthe new model appeared to be marginally better at imitating the behaviour of water inthe field.In general, the new model tended to overpredict the moisture content of the 0—15centimetre soil depth during the months of May through September, and underpredictit during the remainder of the year. Excepting the months of October through January,when its prediction were low, the new model produced a reasonable estimate of the soilmoisture content at the 15—60 centimetre depth. At the 60—90 centimetre depth thenew model predictions were also quite good, except for August and September when thepredicted moisture contents were high. Temporal plots of soil moisture content by depth,presented in Figures 6.13 through 6.18, support these observations.Volumetric Moisture Content0.3 0.4 0.5 0.6 0.301530.C45.60-75 -go -goChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 82Volumetric Moisture Content Volumetric Moisture Content0.3 0.4 0.5 0.6 0.3 0.4 0.5 0.8I I I I0- uca 0• t’cic.• oq.15- Etl 15 n• D.• C.30- L&. 30• r*n. aa.045 crj. 45 Lr aa-cu. aso- qua 60 C 1> a- im+- U75 75 ° •- U-lEF Cgo- l 90 CJANUARY16 APRIL 6Volumetric Moisture Content Volumetric Moisture Content0.3 0.4 0.5 0.6 0.3 0.4 0.5 0.6I I I I0- U’ao 0-- i-ap --EUaQ nç15 15-- a -.• i’a4 . - po.30- l’Dd U 30- L! do a- a- U.- L’ o4 • - a.45. t’EJQ • 45- l Ua-- ro4 a -60- iib a 60- 4 •-- C- o-odi75 75- C c*- üØ-oi- o -go cxI go- oAPRIL 22 MAY4Field Maximum Field Minimum New Model Old ModelDFigure 6.9: Soil moisture profiles, Jan. 16—May 4, 1992, Agassiz.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 83Volumetric Moisture Content Volumetric Moisture Content0.2 0.3 0.4 0.5 0.2 0.3 0.4 0.5I I I I0- Q L ‘ 0- ID Q 1>- G• 1>- 9 1> 9 0• 1>15- 9 15 9 GB 1>- 0•- “t90u30- CG B 30- C4 B- 1> up . cci.- 1> EJI • gLIØ. £,45- lB- v_ ci9 ct60- 09. 60 CI0-pt4 c.75- t’ ol 75- 1>01-EJt* -90- 01* 90______________MAY19 JUNE1Volumetric Moisture Content Volumetric Moisture Content0.2 0.3 0.4 0.5 0,2 0.3 0.4 0.50- c 0 0 OBO 1>- c 9 0.9 1>- 0. I’15- 9— 15 OB 91>- 1>0.-“9 G. i530- ID—,, 1>0. 30 iiII-ND D9 G•DI. 459 DI.-9 D’ E1>B60- 60- 0)• <p01>.-ci, • j> d ai75- thB3 75---0 -90- ODII 90-JUNE15 JULY14Field Maximum Field Minimum New Model Old ModelFigure 6.10: Soil moisture profiles, May 19—July 14, 1992, Agassiz.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 84Volumetric Moisture Content Volumetric Moisture Content0.3 0.4 0.5 0.6 0,3 0.4 0.5 0.6____________________________________I0 C aQ L 0- u L?c a l. uc 1?0 •9 E’ C l15 C a ç 15 0 •930 0 30,45- 9’ 0 • .45-•-C’ •80 a • 60-cb- 0 a75- 90 • 75. Q •• 0• 9 0 I90- CC • 90-0 0 • ‘AUGUST 17 SEPTEMBER21Volumetric Moisture Content Volumetric Moisture Content0.3 0.4 0.5 0.5 0.3 0.4 0.5 0.6I I I0- 0 II ‘ 0-9 i t-•.15- 9 — L 15-- C.•0- (30- a 30-- C L’ a. atNi.• 9 C L’ a. 9t’o.9 0’. 45--® 0L’I-0 l’ I 9L’Da60-dy C L’ a 60’ pJ’0a• o a l’-• 0 a- oió75 u a i’ 75- 019- 0 a--90 a l- caçgo-a o a 9°-_____________SEPTEMBER24 OCTOBER 14Field Maximum Field Minimum New Model Old Model• D----0-----Figure 6.11: Soil moisture profiles, Aug. 17—Oct. 14, 1992, Agassiz.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 85Volumetric Moisture Content Volumetric Moisture Content0.3 0.4 0.5 0.6 0.3 0.4 0.5 0.6I I0 0-16 15-4, — >4>n.•. d,D.30 L’ d • 30- L’ d10U•.E’ O - LcD.45 LI • .45-U.-60 U • 60- >q)D.75 C 75-Dt -90 90-OCTOBER 28 NOVEMBER10Volumetric Moisture Content Volumetric Moisture Content0.3 0.4 0.5 0.6 0.3 0.4 0.5 0.6I I I0 0 —>i.15 15 It,> LwL 4,t. 430 30 DUI DI US46- %45 U U- .4,u. 0•60- 60 0 U- J.75- ci 76- c4- 090-_________________90_________________NOVEMBER24 DECEMBER 8Field Maximum Field Minimum New Model Old Model• D--Figure 6.12: Soil moisture profiles, Oct. 28—Dec. 8, 1992, Agassiz.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 8614.514__________________/‘- /_13.513. 12.5120rj11.5—11•— Field—New -+--Old10.5275 294 316 324 338 352.Julian DayFigure 6.13: Comparison of measured and predicted soil water, 0—30 cm., Agassiz, 1991.15— —14.5 — —1413.5C)1312.5 — —1211.5’11•— Field — New ‘- Old275 294 316 324 338 352Julian DayFigure 6.14: Comparison of measured and predicted soil water, 30—60 cm., Agassiz, 1991.Is.14.514113.5 —1312.5 —12 — —11.511— FNewOk110.5275 294 316 324 338 352Julian DayFigure 6.15: Comparison of measured and predicted soil water, 60—90 cm., Agassiz, 1991.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 87—I\I \14— —11312hi — —0ti)— Fieki—New -4+01d818 • 125 • 167 • 265 302 • 343Julian DayFigure 6.16: Comparison of measured and predicted soil water, 0—30 cm., Agassiz, 1992.1016 125 167 • 265 302 343Juliazi DayFigure 6.17: Comparison of measured and predicted soil water, 30—60 cm., Agassiz, 1992.151413 ———12— — —10— Field—New -4+-Old9 , • • • • • • • • •16 125 167 265 302 343Julian DayFigure 6.18: Comparison of measured and predicted soil water, 60—90 cm., Agassiz, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 88Nitrogen.As information concerning the application of nitrogen to the site prior to the fall of1991 was lacking, and data concerning the soil nitrogen regime for 1993 was unavailable,nitrogen processes were only investigated for the 1992 year. Although there were six fieldtreatments, only five were simulated because the model was unable to mimic the springapplication of 300 kg ha’ of manure plus 100 kg ha1 of inorganic fertilizer. Treatmentplots were located on one part of the field for fall 1991 through September 1992, andmoved to another part of the field for the remainder of the study.Soil Nitrate N and Ammonium N Concentrations. Variations in soil NO-N and NH-Nconcentrations, within the surface 90 centimetres, are plotted against time for each ofthe five treatments: control (C), spring and fall application of manure at 300 kg N ha’(FM300+SM300), fall application of manure at 600 kg N ha1 (FM600), spring application of manure at 600 kg N ha1 (SM600), and spring application of inorganic fertilizerat 200 kg N ha’ (SF200). Graphs illustrating the variations in total inorganic N to a90-centimetre depth, expressed as the combination of NOW-N and NH-N values, are alsopresented. Plots for the five treatments are given in the following figures: C, Figures 6.19,6.20, and 6.21; FM300+SM300, Figures 6.22, 6.23, and 6.24; FM600, Figures 6.25, 6.26,and 6.27; SM600, Figures 6.28, 6.29, and 6.30; and SF200, Figures 6.31, 6.32, and 6.33.Field data for the plots was collected on the same dates as those used in the soil moistureprofiles.The NO-N plots, given in Figures 6.19, 6.22, 6.25, 6.28, and 6.31, demonstrate howpoorly both the old and new models performed with respect to predicting soil nitrate Nconcentrations at various times throughout the year. Neither model was able to generatethe field measured NO-N concentrations, though both models displayed patterns innitrate N variations over time similar to those observed in the field. Although the newChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 89160 —— FiedNew*01d140120 — — —‘1000L) 50 100 150 200 250 300 350Julian DayFigure 6.19: Model and field soil NO-N to 90 cm, C, Agassiz, 1992.0 50 100 150 200 250 300 350Julian DayFigure 6.20: Model and field soil NH-N to 90 cm, C, Agassiz, 1992.180—— Fie1dNew+0IdI-.140 — — —z 19n —0 —u050 100 150 200 250 300 350Julian DayFigure 6.21: Model and field soil inorganic N to 90 cm, C, Agassiz, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 90— d—Nw-’+-0250200 —— —150 —00 5b 100 150 200 250 300 350Julian DayFigure 6.22: Model and field soil NO-N to 90 cm, FM300+SM300, Agassiz, 1992.C 50 100 150 200 250 300 350Juliaii DayFigure 6.23: Model and field soil NHt-N to 90 cm, FM300+SM300, Agassiz, 1992.300— Fie’d—New -‘4--Old‘.250— —I0 50 100 150 200 250 300 350Julian DayFigure 6.24: Model and field soil inorganic N to 90 cm, FM300+SM300, Agassiz, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 92350— FIeld—New-4-Okl300‘?250,‘200 —Eido io i1o 160 180 200 220 240 260 280Julian DayFigure 6.28: Model and field soil NO-N to 90 cm, SM600, Agassiz, 1992.250— Field—New—‘4-Old80 100 120 140 160 180 200 220 240 260 2Julian DayFigure 6.29: Model and field soil Nfl-N to 90 cm, SM600, Agassiz, 1992.400-— YIeldNew -*4-Old350 —-. 30080 100 120 140 160 180 200 220 240 260 280Julian DayFigure 630: Model and field soil inorganic N to 90 cm, SM600, Agassiz, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 93600— Feld—New-4-0Id500’400N —30020O — — —100080 ido io i0 leO 10 200 220 2iO 260 280Julian DayFigure 6.31: Model and field soil NO-N to 90 cm, SF200, Agassiz, 1992.200180 — FdNew -‘4-Old1601140 —12080 100 120 140 160 180 200 220 240 260 280Julian DayFigure 6.32: Model and field soil NHt-N to 90 cm, SF200, Agassiz, 1992.800— dNew—4-0Id700N600Z 500400300200—100-8o ido io i4o 160 180 200 220 240 260 280Julian DayFigure 6.33: Model and field soil inorganic N to 90 cm, SF200, Agassiz, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 94model did capture the peaks in NO-N concentrations marginally better than the old,and the old mimicked fall reductions in NO — N better than the new, the nitrate Npredictions of the two models were virtually identical. Both the model and the fieldplots had increased NO-N concentrations in the spring and fall, and decreased nitrateN concentrations in the summer and winter. Initial model estimates of soil NO-Nconcentrations were much lower than those measured in the field, but by April the modeland field concentrations almost coincided.Figures 6.20, 6.23, 6.26, 6.29, and 6.32 present the NH-N plots for the fivetreatments. The new and old model plots were almost indistinguishable, and both failedto accurately predict field ammonium N concentrations throughout the year. Initialmodel estimates of soil NHt-N were greater than those measured in the field, and, exceptfor the SF200 treatment, tended to remain high over the year. Field observed patterns insoil NH-N concentrations, such as the spring and fall peaks and the summer and wintertroughs, were captured by the model, though not in a magnitude equal to that of thefield.Total inorganic N plots, as a combination of the NO-N and NH-N data, are presented in Figures 6.21, 6.24, 6.27, 6.30, and 6.33. As per the comments on thenitrate and ammonium plots, both the new and old model estimates were similar, andneither model accurately reproduced the magnitudes of the field measured values. Plotsfor both models displayed the spring and fall increases and summer and winter decreasesin inorganic N evidenced in the field. It should be noted that for the SF200 treatment,Figure 6.33 did not show an increase in soil inorganic N equivalent to the amount ofinorganic fertilizer applied on 5 May 1992. This was a result of the spreading loss calculations performed by the model; of the 200 kg N ha1 applied, 67 kg was assumed to belost through volatilization during application.Field measurements suggested that the extreme, spring peaks in field arnmoniumChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 95Table 6.1: Average errors between predicted and measured nitrate N, ammonium N, andtotal inorganic N concentrations, Agassiz, 1992.Treatment NO (kg ha’) NHt (kg ha1) Inorganic N (kg ha1)___________New Old New Old New OldC 36.74 45.11 7.13 7.39 34.16 41.70FM300+SM300 59.66 72.32 35.16 35.53 47.17 47.45FM600 66.11 73.70 39.97 38.72 47.51 50.64SM600 109.79 124.57 41.40 45.37 75.18 86.04SF200 133.61 143.75 29.84 27.86 157.03 164.09N, nitrate N, and inorganic N concentrations for the SF200 treatment, as shown inFigures 6.31 through 6.33, may well have been attributable to inconsistencies in measurement rather than exceptional field processes. On 4 May (day 125), one day priorto fertilization, the total inorganic N concentration on the SF200 treatment was 98.9kg ha1. This measurement compares favourably with the 108.5 kg ha’ concentrationmeasured on the C plot. By 19 May (day 140), the total inorganic N concentration onthe SF200 and C plots were 557.5 and 134.5 kg ha1, respectively. As only 200 kg ha’of inorganic N was added to the SF200 plot, this indicates a gain of 258.6 kg N ha’through mineralization of soil organic matter; for the same time period the C plot gainedonly 26 kg N ha—i. Therefore, the large discrepancy in mineralization was probably dueto measurement error.Average errors between new and old model predicted and field measured NO-N,NHt-N, and total inorganic N concentrations are given in Table 6.1. For nitrate N andtotal inorganic N, average errors of the new model were less than those of the old onall treatments. In terms of amnionium N, new and old model errors were similar on alltreatments.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 96Soil Nitrate N Profiles. Over 80 soil NO-N profiles were available for the Agassiz site;of these, ten were chosen to illustrate the seasonal variations in nitrate N distributionwithin the surface 90 centimetres of soil. Figures 6.34 and 6.35 present the early summerand autumn and winter profiles, respectively, for the C, FM300+SM300, SM600, andSF200 treatments, and Figure 6.36 plots both summer and winter profiles for the F600treatment. As June 1 was the date of the extreme NO-N concentration for the SF200treatment, the June 15 profile has been substituted.The new and old model summer profiles, while underpredicting the amount of nitrateN measured in the field, did reproduce the patterns of nitrate N distribution within thesoil. Nitrate N concentrations for both the models and the field were greatest in the top30 centimetres of soil and decreased with depth. New model nitrate N predictions weremarkedly better than those of the old in the top 30 centimetres of soil; below this depththere were no noticeable differences in the profiles of the two models.Autumn and winter profiles for the two models were similar. Both models tended toerr on the low side when predicting nitrate N concentrations within the soil, especially atdepths of greater than 30 centimetres. Above this level the new and old model estimatesof nitrate N in the top 30 centimetres were reasonable for the three December profiles,but were approximately one-half of the field values for the plots of September. The fieldobserved trend of increasing nitrate N concentrations with depth was also captured bythe two models.Soil NO-N Losses. A summary table of the masses of nitrate-N lost by both modelsvia denitrification and leaching during the year 1992 is presented in Table 6.2. Masseslisted in the leaching column represent nitrate N leached beyond the 1.30-metre standardroot zone. Leaching losses from the old model were considerably greater than thoseaccumulated by the new model, while the new model lost noticeably more NOW-N throughChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 97N03 kg/ha N03 kg/ha0 20 40 60 80 100 120 140 o 20 40 60 80 100 120 140I I I I I I I0 o ‘. 0 cL’i.15 I K 1530 30••••80 K 60 )• K1 K 1 K75 • 75. •••90 U 90 KCONTROL JUNE 1 SPRING 200, JUNE 15NO3kg/ha N03 kg/hao20 40 60 60 100 120 140 0 20 100 120 140I I I I I I I I I0- Q ‘• 0 0 F’ •- ‘U 1I-‘UIs- ç t’ U IS ç fl’ U- ih.1’ • U- •30- U 30- U•-‘ K•• •45-p U•-f K•-:eQ-s • 60-,> U- U-.- U- .75. U 75. •- U U- K-90- K go-SPRING 600, JUNE 1 SPRING & FALL 300, JUNE 1Field New Model Old Model• . eFigure 6.34: Soil nitrate N profiles: C, SF200, SM600 & FM300+SM300, summer 1992,Agassiz.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 98N03 kg/ha N03 kg/ha0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140I I I I I I I I I I0V I 0-VI-V.15 I 15I B301 30 BV1 ••45. I, U.45 .I•60- > B 60> B•. .. 175-q’ B 75 U.V.B90-ê B 90 ICONTROL, DECEMBER 8 SPRING 200, SEPTEMBER 21NO3kg/ha NO3kg/ha0 20 40 80 80 100 120 140 0 20 40 60 80 100 120 140I I I I I I I I0- B 0c • I•15- 1530 30d.t• BB B• 45q’’• I:p, a• 4r a80I B eod’ a-. aV-a75• I 75- BV-4. IV •- B90- a 90- Lé aSPRING 600, SEPTEMBER 21 SPRING & FALL 300, DECEMBER 8Field New Model Old ModelVVVVVVVVVV VVV. 9Figure 6.35: Soil nitrate N profiles: C, SF200, SM600, & FM300+SM300, autumn &winter 1992, Agassiz.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 99o 20 200- Q 0-i15- 15- q.cb-p.’30_45- T60 60-75 75. 190 90 t’OField New Model Old Model•-- -.t---- eFigure 6.36: Soil nitrate N profiles: F600, summer and winter 1992, Agassiz.denitrification. In terms of the total mass of nitrate N lost from the profile the predictionsof the two models were comparable, but old model losses were greater than new modellosses.Crop Yields.Table 6.3 presents a summary of the crop yields and N contents for Agassiz, 1992. Cropyields and crop-N predictions for the two models were similar; model yields and N estimates were less than field values for all treatments except SM600. Average errors betweenestimated and measured crop yields and N contents, for the new and old models, respectively, were ±3.32 t ha’ (20%) and ±4.26 t ha’ (26%), and ±70.13 kg-N ha1 (34%)and ±78.51 kg-N ha’ (37%).40N03 kg/ha60 80 100 120 140N3kgTha40 60 80 100 120 140I I Iaaaaaaaaa.aaaaaaaFALL 600, JUNE 1aaaaaaFALL 600, DECEMBER 8Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 100Table 6.2: Model denitrification and leaching losses, Agassiz, 1992.Treatment Denitrification (kg ha’) Leaching (kg ha’) Total (kg ha’)___________New Old New Old New OldCONTROL 73.0 52.0 6.0 62.3 79.0 114.3FM300+SM300 104.1 72.8 6.5 71.5 110.6 144.3FM600 97.3 74.1 6.2 77.7 103.5 151.8SM600 110.3 73.8 7.2 70.9 117.5 144.7SF200 79.7 52.7 5.8 54.1 85.5 106.8Table 6.3: Measured and predicted crop yields and N content, Agassiz, 1992.Treatment Crop yield (t ha1) Crop N (kg-N ha1)__ ________Field New Old Field New OldCONTROL 16.02 10.00 8.30 182.82 82.80 68.50FM300+SM300 18.14 16.60 15.10 254.80 193.50 192.00FM600 16.35 10.00 8.30 182.68 82.80 68.50SM600 17.61 19.30 18.10 259.50 297.40 301.30SF200 17.19 16.20 15.20 261.96 210.40 211.50Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 101Statistical Evaluation.The statistical evaluation method of Loague and Green (1991) was employed to objectively evaluate the effect the Richard (1988) water routines had on model performance.Test statistics were calculated as follows (Loague and Green, 1991):Maximum Error (ME),ME=maxIP1—O i= (6.1)Root Mean Square Error (RMSE),‘P O.2 0.5 100RMSE = ——=— (6.2)n 0Coefficient of Determination (CD),CD — YZ7’_1(O, — O)2 6 3- (P_O)2 (•)Modelling Efficiency (EF),EF—- O)2 —1(P - Q.)2)64— 1(O_O)2 (.)Coefficient of Residual Mass (CRM):CRM-(>1O-F)n(.)where P and O are the predicted and observed values, respectively, 0 is the mean ofthe observed data, and n is the number of samples. Evaluation standards based on thesestatistics have not yet been determined. However, if the predicted and measured valueswere all equal, the values of ME, RMSE, CD, EF and CRM would be 0, 0, 1, 1, and0, respectively. A negative EF value indicates that use of the model predicted values isworse than simply using the observed mean (Loague and Green, 1991).Summary statistics were calculated for the total amounts of moisture, nitrate N,ammonium N, and inorganic N in the soil to a depth of 90 centimetres, and are presentedChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 102in Tables 6.4 and 6.5, for the new and old models, respectively. Soil moisture statisticsare for the years 1991 and 1992, and nitrogen statistics are for 1992.Nitrogen test statistics for the old and new models were similar. Overall, the newmodel test statistics were better than the old, though neither model yielded particularlygood results. Excepting the SF200 NH and F600 total inorganic N statistics, all of theold and new model EF values were negative. These test statistics indicated that the newmodel simulations of 1992 nitrogen processes were marginally better than the simulationsof the old model.For soil moisture in 1991, new model values for RMSE, EF and CRM were betterthan old model values. As neither models’ EF values were negative, use of either models’predictions would have been no worse than using the observed mean. In 1992, all of thenew model test statistics for soil moisture were better than the old. The EF value ofthe old model was negative, indicating that use of its predictions would have been worsethan using the observed mean. On the basis of these test statistics, it appeared that thesoil moisture routines of the new model were marginally better at mimicking the 1991and 1992 soil water regimes than those of the old.6.1.2 SumasInsufficient climatic data at the Sumas site limited model simulation to the 1992 year.As well, field measured parameters necessary for model emulation were not available.This paucity of field data reduced the simulation of soil water movement to an exercisein ‘curve-fitting’, as discussed in Chapter 5. Nevertheless, as model nitrogen algorithmswere not altered during the calibration process, the results of the 1992 Sumas simulationsare presented in order to investigate the limitations of the model water and nitrogenroutines.It should be noted that the soil input files used by the old and new models wereChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 103Table 6.4: Summary statistics, new model, Agassiz. Ideal values of ME, RMSE, CD, EF,and CRM are 0, 0, 1, 1, and 0, respectively.Treatment ME (kg ha’) RMSE (%) CD EF CRMNitrateCONTROL 86.04 49.89 1.35 -0.46 0.30FM300+SM300 145.14 52.00 0.84 -0.68 0.38FM600 147.06 57.22 0.96 -0.32 0.43SM600 171.66 55.09 0.59 -0.83 0.50SF200 440.25 92.80 1.30 -0.49 0.66AmmoniumCONTROL 32.56 48.92 0.68 -1.66 -0.18FM300+SM300 68.80 183.08 0.02 -34.69 -1.44FM600 123.81 234.72 0.05 -15.54 -1.47SM600 99.71 154.70 0.19 -1.88 -1.17SF200 101.26 75.94 3.16 0.44 0.36Inorganic NCONTROL 83.86 37.13 2.20 -0.18 0.21FM300+SM300 132.84 36.11 1.05 -0.16 0.12FM600 118.51 35.61 0.62 0.42 0.12SM600 151.12 34.48 0.70 -0.11 0.27SF200 541.51 88.22 1.71 -0.22 0.59Moisture ME (cm. of 1120) RMSE (%) CD EF CRM1991 3.12 3.68 1.80 0.73 -0.011992 4.00 5.40 1.56 0.43 0.01Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 104Table 6.5: Summary statistics, old model, Agassiz. Ideal values of ME, RMSE, CD, EF,and CRM are 0, 0, 1, 1, and 0, respectively.Treatment ME (kg ha) RMSE (%) CD EF CRMNitrateCONTROL 103.38 60.47 0.91 -1.15 0.37FM300+SM300 185.33 61.23 0.63 -1.34 0.47FM600 148.06 63.43 0.89 -0.62 0.47SM600 219.40 63.60 0.50 -1.44 0.56SF200 482.64 100.82 1.15 -0.76 0.71AmmoniumCONTROL 32.55 48.88 0.69 -1.65 -0.21FM300+SM300 73.70 182.37 0.02 -34.42 -1.45FM600 118.33 224.52 0.05 -14.14 -1.42SM600 92.11 160.84 0.18 -2.11 -1.28SF200 88.25 70.74 3.69 0.52 0.31Inorganic NCONTROL 100.29 45.58 1.42 -0.78 0.25FM300+SM300 168.73 40.69 0.99 -0.48 0.20FM600 123.49 37.58 0.69 0.35 0.15SM600 167.00 39.81 0.63 -0.48 0.31SF200 570.89 92.97 1.62 -0.36 0.62Moisture ME (cm. of H2O) RMSE (%) CD EF CRM1991 2.85 3.89 0.47 0.69 0.031992 8.17 7.55 0.42 -0.12 0.05Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 105slightly different. After estimating several soil variables, as discussed in Chapter 5, itwas discovered that the old model failed to function with the new Sumas soil file. Toavoid this problem the original Sumas soil file, which was provided with the old model,was used to generate old model predictions.Soil Moisture.Water Table Level. Figure 5.3 presents the field measured and model simulated watertables. As for Agassiz, the old model assumed a water table well below the standard rootzone of 1.30 metres, and, as such, was not plotted. It is interesting to note that, evenafter extensive calibration, the model was incapable of reproducing the sudden drops inwater table level that occurred near days 31, 166 and 196.Figure 6.37 plots the calibrated water table levels against the levels measured in thefield. The fit is comparable to that given in Figure 6.4 for the calibrated Agassiz watertable, though the Sumas error of ±0.17 metres is approximately one-half the Agassizerror.Total Water to 90 Centimetres. Figure 6.38 plots the total depths of water within thetop 90 centimetres predicted by both the old and new models against the total depthsmeasured in the field for 1992. Data for the plots was obtained using the average fieldmoisture content values presented in the soil moisture profiles (see Figures 6.39— 6.42).For the fourteen 1992 days plotted, the new model predictions were high on four daysand low on 10, while the old model predictions were low on all 14 days. The new modelestimates of total soil moisture to 90 centimetres were noticeably better than the oldmodel estimates, which tended to be one-fifth to one-half of the field measured values.The maximum and average errors between measured and predicted values, for the newand old models, respectively, were 7.77 and ±3.85 centimetres, and 25.72 and ±21.57Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 1061.2AA0.6 A A A___AA AAA A AAA4 AAA A A 4E A A&02.AAAA AI I I I0 0.2 0.4 0.6 0.8 1 1.2 1.4Measured (in. below soil surface)Figure 6.37: Water table levels, measured vs. predicted, Sumas 1992. Average error is+0.17 metres.I><d ANewlAAAA.4-40 20SV105>(0 I I I I0 5 15 20 25 35Measured (cm. of water)Figure 6.38: Comparison of measured and predicted soil water to a 90 cm depth for thenew and old models, Sumas 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 107centimetres. Maximum and average ranges of field measured moisture content valueswere 7.92 centimetres and ±2.30 centimetres, respectively.Soil Moisture Profiles. Soil moisture profiles for 1992, to a 90 centimetre depth, are presented in Figures 6.39 through 6.42. Agreement between the new model output and thefield measured values was reasonable, while the old model predictions were consistentlylow. The new model tended to underpredict the moisture content of the 0—15 centimetresoil depth during the months of April through June, and overpredict it from Septemberthrough December. Excepting the profiles of June 16 and September 23, when the predictions were high, the new model estimates of the soil moisture content at the 15—60centimetre depth were consistently low. At the 60—90 centimetre depth the new modelpredictions were generally low from April through September and high for the remainderof the year. Temporal plots of soil moisture content by depth, presented in Figures 6.43through 6.45, support these observations.Nitrogen.Treatments simulated at the Sumas site were identical to those at Agassiz: CONTROL,FM300+SM300, FM600, SM600, and SF200. Treatment plots were located on one partof the field for fall 1991 through September 1992, and moved to another part of the fieldfor the remainder of the study.Soil Nitrate N and Ammonium N Concentrations. Soil NO-N, NH-N and total inorganic N concentrations, within the surface 90 centimetres, were plotted against time foreach of the five treatments. Graphs for the five treatments are presented in the followingfigures: C, Figures 6.46, 6.47, and 6.48; FM300+SM300, Figures 6.49, 6.50, and 6.51;FM600, Figures 6.52, 6.53, and 6.54; SM600, Figures 6.55, 6.56, and 6.57; and SF200,Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 108Volumetric Moisture Content Volumetric Moisture Content0 0.1 0.2 0.3 0.4 0.5 0.6 o 0.1 0.2 0.3 0.4 0.5 0.6I I I I0 0 0- 0>4).4)!)’ U.4’ >01 . 4’. [II15 9 >01 15. 4)Dl 4) > ci.30 !)> Dl 30 (1) L> ]4) C4) It> 4)45 4) .045. 4)4) it>. 4) >o.Ij). 4)60 4) I1’ 60 4) ci.>u• 4) t> ci.4) U • 4) I ci.75. 1’ l> 1] I 75. 4) L> DU4) t> 0 I 4) It> UI4) t>D. 4)90 0 LLJ I 90 0 ci.APRIL 8 MAY5Volumetric Moisture Content Volumetric Moisture Content0.1 0.2 0.3 0.4 0.5 0.8 0.1 0.2 0.3 0.4 0.5 0.6I I I I I I I0 0 [> UI 0 0 1> Dl4) It> Dl.4) t> UI4) Dl 4) It> Dl15 4) ti 15 <..v’o..V—U.30 Ct) It’ ci. 30 Ct) It’ ci•4) ci-. 4) o.4) 31 4) >01.045 4) L’0.4) ItO.60- 4) . 60 4) It’ciI- It’ Dl 4) ‘- Ct) 1>01 4) 1> c75- Ct) (> DI 75- 4) 1>- 4) It’D- 4) It’ C.- 4) 1>UI- 4)90- C!) 1>0. 90- 0MAY20 JUNE 2Field Maximum Field Minimum New Model Old Model• DFigure 6.39: Soil moisture profiles, April 8—June 2, 1992, Sumas.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 109Volumetric Moture Content Volumetric Moture Content0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.60- Q ci 0-0 L-<P ..P[-15- 9 f’ cii 15- 9 cii4) >. ci. - 4) L> ci.4 t> ci.- 4) > ci.30 6 cii 30- [> U.- b- 4)- 4) >ci•.045. <P .045 4)( c’- <P4) c., . 4)60 4) 60- 4)- 4)- 4) > ci.-- c. ci.75. 4 754) c.- 4) cliC)-90- 0 90- CD ci4JUNE 16 JULY15Volumetric Motur. Content Volumetric Moisture Content0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6I I I I I I I I I I0- 0 0- 9 cit• 4)- 4)-c ti 4)15- 15- p i-4 oi- 4)4) r’ci.-4 ci.30-4) l> cii 30- 6 a .- c1. -Ic B4) a>..Ø a’45<P 454) a><p a>.-<p ci>4) a>i.-4) ci60 4) 50-4) cii> Bi> ci.-4 ‘ ci.i> ci.-4) ‘ ci.75 4) 1> ciB 75- 4) i, cii(P 1 ci• 41 ci•4) ci. 4)90- CD 1> cii 90 0 ciLSAUGUST18 SEPTEMBER17Field Maximum Field Minimum New Model Old ModelcFigure 6.40: Soil moisture profiles, June 16—Sept. 17, 1992, Sumas.015•304560759000153045607590015-30-45.t..80-75-90-00-15-30-F560-7590Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONSVolumetric Moisture Content0.1 0.2 0.3 0.4 0.5I I I I0.6110Volumetric Moisture Content0.1 0.2 0.3 0.4 0.5 0.6C]. L>() ci. >ci. I>U. r4, ci.4, ci. C>4, D 1>C ciSC>ci.C>4,4,4,4, LU.c C>cimC>U.C>.Da4, ni.C>ci.C>SEPTEMBER 23Volumetric Moisture Content0.1 0,2 0.3 0.4 0.5 0.6I IC4,4) té4, ci.4,4,4,c.U.4)4)4,4,4, ‘0.ci.4)rJ.C>OCTOBER13Volumetric Moisture Content0.1 0.2 0.3 0.4 0,5 0.6I I I Io04,4)4, —4,4’4,4)4, c.•4, ci*4, UPDR4)DiC D•t0 ciBC>4, 0•4, j.4, ci.4,4’,D •0 C>U •4, tc1rci4,4, rfl..4, U•iC>ci.t.ci .t4) 0 •C>4, ci .I,aOCTOBER 27Field Maximum Field Minimum New Model Old Model• D-Figure 6.41: Soil moisture profiles, Sept. 23—Nov.NOVEMBER 99, 1992, Sumas.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 111Volumetric Moisture Content Volumetric Moisture Content0 0.1 0.2 0.3 0.4 0.5 0.6 0.1 0.2 0.3 0.4 0.5 0.6I I I I I I I0- 0 c.1> 0- Q--- c- c15- 9 1>Efl 15- 9 2- 4> 1> - 4>--cj30- 30- di--1>0•-45 45- DI- 4, - 4, >oa4>- 4> c.60 C) 80- 4)- 4,-4, > cu011> [a75. u. ‘ 75- d>0.1>-4> DI 1>-90 D• ‘ 90- 0NOVEMBER23 DECEMBER 7Field Maximum Field Minimum New Model Old Modeln- ----0----•Figure 6.42: Soil moisture profiles, Nov. 23 and Dec. 7, 1992, Sumas.Figures 6.58, 6.59, and 6.60. Field data for the plots was collected on the same dates asthose used in the soil moisture profiles.Nitrate N plots, given in Figures 6.46, 6.49, 6.52, 6.55, and 6.58, demonstratethe inability of the old and new models to reproduce field measured soil NO-N concentrations. New model predicted values were generally greater than those of both the oldmodel and the field, while old model estimates tended to be lower than field values. Thegeneral shapes of the old and new model nitrate N plots over the year were similar tothose observed in the field; both the model and the field plots had increased NOV-N concentrations in the spring and fall, and decreased nitrate N concentrations in the summerand winter. Initial model estimates of soil NO-N concentrations were much lower thanthose measured in the field; by April old model and field concentrations almost coincided,while new model predictions were high.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 11212 — ——— Field — New —‘a-- Old0154 231 287 • 328Julian DayFigure 6.43: Comparison of measured and predicted soil water, 0—30 cm., Sumas, 1992.16 I — Field—New--’4--Ohi I99 154 231 287 328Julian DayFigure 6.44: Comparison of measured and predicted soil water, 30—60 cm., Sumas, 1992.16— Field — New -‘+- Old99 154 231 287 328Julian DayFigure 6.45: Comparison of measured and predicted soil water, 60—90 cm., Sumas, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS908070. 60, 504039.2010’50 100 150 200 250 300 350Juliaii DayFigure 6.46: Model and field soil NO-N to 90 cm, C, Sumas, 1992.40135p30’2520I150 50 100 150 200 250 300Julian DayFigure 6.47: Model and field soil NHt-N to 90 cm, C, Sumas,110’________________________— F1ed—New ‘+Old3501992.2050 100 150 2ó0 250 3ó0 30Julian DayFigure 6.48: Model and field soil inorganic N to 90 cm, C, Sumas, 1992.113JIj0I\— __ ——Field—New -‘4-OldChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 114250_________________________— I — FieldNew-)4--OId I200I—‘ 15010050 —0 I I I0 50 100 150 200 250 300Julian DayFigure 6.49: Model and field soil NO-N to 90 cm, FM300+SM300, Sumas, 1992.100908060‘‘ 50z40— _ —3020 —10 I — Field — New -‘f- OldI I I0 50 100 150 200 250 300Julian DayFigure 6.50: Model and field soil NHt-N to 90 cm, FM300+SM300, Sumas, 1992.250I200Z 150•0100050— I FieldNew-4-0h1 I0 I I I I0 50 100 150 200 250 300 350Julian DayFigure 6.51: Model and field soil inorganic N to 90 cm, FM300+SM300, Sumas, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 115180________________________160 I — Field—140120,-100Z 8060 —40200 I I0 50 100 150 200 250 300 350Julian DayFigure 6.52: Model and field soil NO-N to 90 cm, FM600, Sumas, 1992.180____ ____ ___ _160_ _140120100806040200 I0 50 100 150 200 250 300 350Julian DayFigure 6.53: Model and field soil NHt-N to 90 cm, FM600, Sumas, 1992.300_ ____ ____ __250200C)15010050U I I0 50 100 150 200 250 300 350Julian DayFigure 6.54: Model and field soil inorganic N to 90 cm, FM600, Sumas, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 116300____________________________— Field—New -‘01d250200150100 120 140 160 180 200 220 240 260 280Julian DayFigure 6.55: Model and field soil NO-N to 90 cm, SM600, Sumas, 1992.180_ __ __ _ _I — FieldNew401 I080 idO 120 140 160 180 200 220 240 260 280Julian DayFigure 6.56: Model and field soil NH-N to 90 cm, SM600, Sumas, 1992.350____ __ _I — Fleld—New-’4--Okt II —/080 IdO 10 ilo 10 180 2ó0 220 240 2d0 280Julian DayFigure 6.57: Model and field soil inorganic N to 90 cm, SM600, Sumas, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 117300 i — —--‘--° i250200508otoo 120 140 160 180 200 220 240 260 280Julian DayFigure 6.58: Model and field soil NO-N to 90 cm, SF200, Sumas, 1992.250_______________________I — Fie1d—New-I--0Id I200BO100 120 140 160 180 200 20 240 260 280Julian DayFigure 6.59: Model and field soil NH-N to 90 cm, SF200, Sumas, 1992.450 — —— FieldNew-’401a I1 400‘350300v 250050 Ido 10 1Áo 10 180 260 220 2.iO 260 250Julian DayFigure 6.60: Model and field soil inorganic N to 90 cm, SF200, Sumas, 1992.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 118Figures 6.47, 6.50, 6.53, 6.56,and 6.59 present the NHt-N plots for the five treatments. Plots of the old and new models were similar; both failed to accurately predictfield arnmonium N concentrations throughout the year. Initial model estimates of soilNH-N were greater than those measured in the field, and, except for the SF200 and Ctreatments, tended to remain high over the year. Field observed patterns in soil NH-Nconcentrations, such as the spring and fall peaks and the summer and winter troughs,were captured by the models.Total inorganic N plots, as a combination of the NO-N and NHt-N data, are presented in Figures 6.48, 6.51, 6.54, 6.57, and 6.60. As for the ammonium N plots, new andold model estimates were similar; neither model accurately reproduced the magnitudesof the field measured values. Both model plots displayed the spring and fall increases andsummer and winter decreases in inorganic N observed in the field. It should be notedthat for the SF200 treatment, Figure 6.60 did not show an increase in soil inorganic Nequivalent to the amount of inorganic fertilizer applied on 14 May 1992. This was a resultof the spreading loss calculations performed by the model; of the 200 kg N ha1 applied,67 kg was assumed to be lost through volatilization during application.The SF200 treatment exhibited extreme, spring peaks in field ammonium N, nitrateN, and inorganic N concentrations (see Figures 6.58 through 6.60). Between 5 May (day126) and 20 May (day 141) the SF200 plot gained 202.0 kg ha’ of inorganic N throughmineralization, while the C plot shows a mineralization gain of 36.5 kg N ha’. As forAgassiz, this mineralization discrepancy was probably due to measurement error.Average errors between new and old model predicted and field measured NO-N,NH-N, and total inorganic N concentrations are given in Table 6.6. For nitrate N,average errors of the new model were greater than those of the old on all treatments,while for ammonium N new model errors exceeded old model errors on all treatmentsexcept FM600. New model errors for total inorganic N were greater than those of theChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 119Table 6.6: Average errors between predicted and measured nitrate N, ammonium N, andtotal inorganic N concentrations, Sumas, 1992.Treatment NO (kg ha’) NH (kg ha1) Inorganic N (kg ha’)___________New Old New Old New OldC 31.73 21.44 10.75 9.27 25.77 25.25FM300+SM300 66.30 45.96 34.84 34.02 67.52 35.87FM600 41.45 28.20 37.55 42.05 67.91 30.93SM600 78.45 76.23 59.76 49.49 37.53 39.31SF200 82.54 80.74 16.23 15.32 91.79 96.06old on all treatments except for SF200 and SM600.Soil Nitrate N Profiles. Seventy-five soil NO-N profiles were available for the Sumassite; ten were chosen to illustrate seasonal variations in nitrate N distribution withinthe surface 90 centimetres of soil. Figures 6.61 and 6.62 present the early summer andautumn profiles, respectively, for the C, FM300+SM300, SM600, and SF200 treatments,and Figure 6.63 plots both summer and fall profiles for the F600 treatment.New and old model summer profiles, while underpredicting the amount of nitrate Nmeasured in the field, did reproduce the patterns of nitrate N distribution within thesoil. Nitrate N concentrations for both the models and the field were greatest in the top30 centimetres of soil and decreased with depth. New model nitrate N predictions weremarkedly better than those of the old in the top 30 centimetres of soil; below this depththere were no real differences in the profiles of the two models.Autumn profiles for the new model were closer than those of the old to field determinedprofiles; the new model produced a good description of nitrate N distribution while theold model tended to underpredict field NO values. The field observed trend of decreasingnitrate N concentrations with depth was also captured by the new model. Overall, theChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS40N03 kg/ha60 80 100 120 140I I120N03 kg/ha40 60 80 100 120 1400 200-) 1.), .15- 9 F’ •4) -a30- i U45- F’•- ‘a60- ‘U75- Igo- IIIUa.IIaIo 20-0- )15-9-4)30-->60 -- U- I75- I- U- ago- Uo 20140SPRING 200, JUNE 2N03 kg/ha40 60 60 100CONTROL JUNE 2N03 kg/ha20 40 60 80 100 120I I Irs’9UUci IaDa.aa120 14000-15-30 -£45-60-75 -90 -01530 --4560-75 -90-Ua.IUc aac.aUaaaSPRING 600, JUNE 2Field New Model-••->Old ModeleSPRING & FALL 300, JUNE 2Figure 6.61: Soil nitrate N profiles: C, SF200, 5M600 & FM300+SM300, June 2, 1992,Sumas.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 121N03 kg/ha NO3kgIhao 20 40 60 80 100 120 140 0 20 40 60 80 IX 120 140I I I I I0 I 0-0 W’-4:1 I15 I 15- 930.4> 30--IJI.45 9E-..4>.80 tt •75. 75*goCONTROL, OCTOBER 13 SPRING 200, SEPTEMBER 17N03 kg/ha NO3kgIhao 20 40 60 100 120 140 20 40 so so 100 120 140I I I I I I I I1’ 0 Q —S—. S r:.15- q S 15 9- . t 1•30- 30 O-80-tI 6075- iq 75 p_____________g0SPRING 600, SEPTEMBER 17 SPRING & FALL 300, OCTOBER 17Field New Model Old Model• eFigure 6.62: Soil nitrate N profiles: C, SF200, SM600, g FM300+SM300, autumn 1992,Sumas.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 122N03 kg/ha N03 kg/hao20 40 60 80 100 120 140 0 20 40 60 80 100 120 140I I I I I0- ‘ 0-?--)--?15-9 15-9 ‘-- . .L’-ci30- ti> • 30- Ô-I,-I,45- I 45-cpI.-60- I 60--I-_-I-75- I 75-- I- -.-I- I90- 90- •FALL 600, JUNE 2 FALL 600, OCTOBER 13Field New Model Old Model• •••-• eFigure 6.63: Soil nitrate N profiles: F600, June 2 and Oct. 13, 1992, Sumas.new model’s autumn profiles were better than its June profiles.Soil NO-N Losses. Model denitrification and leaching losses during the year 1992 aresummarized in Table 6.7. Masses listed in the leaching column represent nitrate N leachedbeyond the 1.30-metre standard root zone. Leaching losses from the old model were anorder of magnitude greater than those accumulated by the new model, while the newmodel lost much more NO-N through denitrification. Old model estimates of the totalmass of nitrate N lost from the profile were three to four times larger than the estimatesof the new model.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 123Table 6.7: Model denitrification and leaching losses, Sumas, 1992.Treatment Denitrification (kg ha1) Leaching (kg ha’) Total (kg ha’)___________New Old New Old New OldCONTROL 27.1 0.4 10.4 107.0 37.5 107.4FM300+SM300 37.9 0.6 12.2 194.8 50.1 195.4FM600 30.6 0.6 11.7 190.5 42.3 191.1SM600 43.4 0.6 12.7 195.1 56.1 195.7SF200 31.2 0.4 10.7 123.2 41.9 123.6Table 6.8: Crop yields arid N content, Sumas, 1992.Treatment Crop Yield (t ha’) Crop N (kg-N ha’)___ ________Field New Old Field New OldCONTROL 11.60 4.60 2.30 93.02 35.30 15.70FM300+SM300 19.34 10.10 6.60 211.82 124.70 89.80FM600 11.57 4.60 2.30 102.11 35.30 15.70SM600 20.58 12.60 8.0 290.77 214.40 160.50SF200 18.92 10.30 7.30 264.29 155.10 135.00Crop Yields.A summary of the crop yields and N contents for Sumas, 1992, is presented in Table 6.8.The crop yields and crop-N predictions of the new model were greater than those of theold model, but both of the models’ yields and N estimates were less than field valuesfor all treatments. Average errors between estimated and measured crop yields and Ncontents, for the new and old models, respectively, were ±7.96 t ha’ (51%) and ±11.10t ha (70%), and ±79.44 kg-N ha1 (47%) and ±109.06 kg-N ha1 (64%).Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 124Table 6.9: Summary statistics, new model, Sumas. Ideal values of ME, RMSE, CD, EF,and CRM are 0, 0, 1, 1, and 0, respectively.Treatment ME (kg ha’) RMSE (%) CD EF CRMNitrateCONTROL 60.73 112.21 0.30 -2.56 -0.91FM300+SM300 141.53 97.05 1.79 -0.52 -0.24FM600 103.67 97.76 0.41 -0.64 -0.69SM600 183.74 75.08 4.23 -0.07 0.32SF200 203.43 103.47 4.60 -0.25 0.42AmmoniumCONTROL 25.44 51.07 0.71 -2.47 0.13FM300+SM300 64.94 134.31 0.07 -11.79 -1.02FM600 97.20 192.32 0.03 -24.23 -1.095M600 133.71 232.31 0.11 -6.97 -1.75SF200 50.54 29.84 1.53 0.91 0.0.07Inorganic NCONTROL 55.85 53.21 0.61 -0.84 -0.45FM300+SM300 152.14 76.33 0.68 -0.67 -0.46FM600 195.29 115.28 0.15 -3.16 -0.83SM600 74.30 25.78 2.04 0.82 -0.11SF200 231.68 68.79 5.05 0.47 0.27Statistical Evaluation.As the simulation of the Sumas soil water regime was basically a ‘curve-fitting’ exercise,calculation of soil moisture test statistics would have been meaningless. However, summary statistics were calculated for the total amounts of nitrate N, ammonium N, andinorganic N within a 90-centimetre soil depth for 1992. Tables 6.9 and 6.10 present thetest statistics for the new and old models, respectively.Overall, the new model test statistics were better than the old for the SF200, SM600Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 125Table 6.10: Summary statistics, old model, Sumas. Ideal values of ME, RMSE, CD, EF,and CRM are 0, 0, 1, 1, and 0, respectively.Treatment ME (kg ha’) RMSE (%) CD EF CRMNitrateCONTROL 44.45 74.18 1.51 -0.56 0.66FM300+SM300 172.97 85.90 2.25 -0.19 0.45FM600 73.70 69.31 1.95 0.17 0.39SM600 209.82 81.58 1.96 -0.27 0.49SF200 239.96 118.45 1.74 -0.64 0.69AmmoniumCONTROL 25.77 47.05 0.96 -1.95 -0.05FM300+SM300 61.68 134.56 0.07 -11.84 -1.04FM600 113.38 223.28 0.03 -33.01 -1.29SM600 113.06 187.18 0.15 -4.18 -1.45SF200 50.49 29.62 1.69 0.91 0.12Inorganic NCONTROL 60.96 49.59 1.85 -0.59 0.47FM300+SM300 121.79 45.51 2.92 0.40 0.03FM600 89.13 48.44 0.40 0.27 -0.20SM600 96.76 30.16 2.51 0.75 0.09SF200 269.75 79.82 3.11 0.29 0.45Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 126and FM300+SM300 treatments. On these treatments the test statistics indicated thatthe new model simulations of 1992 nitrogen processes were marginally better than theold model simulations.6.2 Discussion6.2.1 Soil Moisture.As evidenced by the field observations, the old model assumption of a deep water tablepersisting below the 1.30 metre root zone was erroneous. At Agassiz the water table roseabove this level during the wet fall and winter months of 1991—1993, and at Sumas thewater table was generally above this level throughout the 1992 year. The new modelwas able to reproduce the general shapes of the field measured water tables at both theAgassiz and Sumas sites for the years 1991—1993. Model accuracy during the simulationperiods was reasonable; 68% of the predicted daily water table levels for Agassiz werewithin 0.45 metres of the field measured levels. This suggested that the discrepanciesbetween model and field values may have been attributable to inadequate field datarather than limitations in the new model water routines.Peaks in the simulated water table levels occasionally lagged behind those observedin the field by several days. This delay in the rise of the simulated water table maybe attributed to the deeper water table levels assumed by the model at the onset ofthe precipitation events that yielded the peaks. In general, the new model was able tosimulate the rapid rise of the field water table in response to precipitation events, butwas unable to duplicate the sudden, subsequent decreases. This implied that drainagefrom the model profile was restricted. As drainage of the model profile was influenced toa large extent by the deep flow values set in subroutines LATFLO and RVRLVL, modelwater table behaviour suggested that the assumptions upon which the deep flow valuesChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 127were calculated needed improvement. For the Agassiz site the assumption of the FraserRiver as the dominant influence in the regional hydrologic system may have been invalid,and the average river levels used in routine RVRLVL may not have been representativeof the prevailing river levels for 1991—1993. The Sumas site was drained by ditchesadjacent to the field rather than by a single, regional, hydrologic feature, and as such itwas imperfectly represented by the calibrated LATFLO and RVRLVL subroutines. Hadthe drainage ditch levels been recorded, a Hooghoudt type drainage equation could havebeen employed to better reproduce the local groundwater flow patterns.The accuracy of the water table levels predicted by the new model was comparableto that of the models developed by Chao (1987) and Richard (1988), and the non-bypassing flow model of Eckersten and Jansson (1991). However, the model developedby Broughton and Faroud (1978), and the bypassing flow model of Eckersten and Jansson(1991), appeared to better simulate the observed water table levels for their respectivefield sites.At both field sites the new model estimates of the total amount of water within thesurface 90 centimetres of soil were quite good. Average errors between measured andpredicted values and the average ranges of field moisture content measurements werecomparable; for 1991 and 1992 at Agassiz the average errors were greater than the rangesin field measured values by 0.35 and 0.22 centimetres, respectively, while at Sumas theaverage error was 1.55 centimetres greater than the field measured range. For the Agassizsite the average errors of the old model were greater than the field measured ranges by0.30 and 0.48 centimetres for 1991 and 1992, respectively. This indicated that there wasno real difference between old and new model predictions of total water to 90 centimetresfor 1991 and 1992. Old model estimates for Sumas were much worse than those of thenew for 1992; this may be attributable to the more realistic water regime simulated bythe new model. However, the increased accuracy of the new model may also have beenChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 128a result of the extensive calibration performed on the new model and its use of a morerepresentative soil file rather than a consequence of the new water routines.In general, the new model produced soil moisture profiles that were reasonable representations of the field conditions observed at both sites. Old model profiles for Agassizwere similar to those of the new model, while for Sumas the old model consistently Underpredicted field moisture contents. Agreement between the new model predicted andfield measured soil moisture contents was best at the Agassiz site for the 15—60 centimetre soil depth; this was probably a result of the selected &(O) curve. As the texture ofthe Agassiz soil varied greatly with depth, the i/,(0) curve used by the new model wasdetermined using the average of four measured b(O) curves (see Chapter 4). This averagecurve was closest in shape and magnitude to that of the 30—60 centimetre soil depth.Comparisons of new model and old model results to field measured values suggestthat the incorporation of the Richard (1988) water movement routines into the modelhas improved the model representation of the soil water regime. However, as manyof the new model predictions were similar to those of the old, the improvement wasmarginal. One possible reason why the improvement was not greater than that observedwas variability in and lack of field data. In order to have taken full advantage of theRichard (1988) routines extensive field data was required; lack of field measured data,especially for the K(O) function, may have reduced the benefits of the new model waterroutines.6.2.2 Nitrogen.The soil nitrate and ammonium concentration plots, presented in Figures 6.19— 6.33 and6.46— 6.60, highlight the deficiencies in the model nitrogen routines. At the Agassiz site,old model and new model plots for NO, NHt, and total inorganic N plots were nearlyidentical, whereas at Sumas the new model plots differed noticeably from those of theChapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 129old. Since the old model and new model simulated soil water regimes were also similarfor Agassiz and different for Sumas, this difference in model soil N representations forSumas is a direct result of the change in soil water routines.Field nitrate and ammonium plots at both sites and for all treatments displayed peaksin N03 and NH during the early summer and autumn months. The summer and fallincreases were due to mineralization and nitrification of soil organic matter and appliedmanure. Subsequent decreases in summer nitrate concentrations may have been due toplant uptake, leaching or denitrification, whereas leaching and denitrification were responsible for the winter decreases in NO concentrations. As both models mimicked theshapes of these peaks but not the magnitudes, the indication was that model mineralization, nitrification, denitrification and crop-N uptake routines were deficient. Movementof the treatment plots from one location on the field to another in the fall of 1992 mayhave also contributed to the discrepancies between model predicted and field measuredsoil nitrogen concentrations, particularly on the FM300+SM300 and FM600 treatments.Except for the SF200 and C treatments, the new model estimates of ammoniumconcentrations were generally greater than the peak values measured at the Agassiz site,while the nitrate concentration predictions were generally lower than the peak values.Therefore, if the model peaks in total inorganic N would have been close to the fieldmeasured peaks, the conclusion would have been that the nitrification routines were theprimary cause of error. An examination of the total inorganic N plots revealed that thiswas not the case; field measured inorganic N totals generally surpassed those of the model.As the field gains in inorganic N were greater than the model simulated gains, both themodel mineralization and nitrification routines were identified as needing improvement.Poor agreement between the predicted and measured nitrogen concentrations at theSumas site support this statement; even though the soil water regime was reproducedthere were still large discrepancies between predicted and measured soil nitrogen values.Chapter 6. MODEL SIMULATIONS AND FIELD OBSERVATIONS 130The old model assumption of a freely drained profile may have led to high estimatesof nitrate leaching losses and low estimates of denitrification losses. New model estimatesof nitrate leaching appeared to be unrealistically low. Nitrate loss mechanisms presentedin Tables 6.2 and 6.7 were the processes affected most by the new model and old modelwater routines. The different nitrate and ammonium plots for each field site indicatedthat the model nitrogen routines were responsive to changes in soil moisture conditions.Chapter 7Summary and ConclusionsThe objectives of this study were to incorporate the Richard (1988) water movementalgorithms into the Bulley and Cappelaere (1978) nitrogen model; calibrate the modelusing field data gathered at two sites in the Fraser Valley; perform a sensitivity analysison the modified model; and evaluate the effectiveness of the modifications by comparingsoil nitrogen and water predictions of the modified and unmodified models to field data.The new model was calibrated using the 1992 water table levels at both sites. Visualinspection determined the extent of calibration required. The sensitivity analysis waslimited to the variables (O), Ck, °ck, a, Kdl, p and HF. Variables were modified withina 40% range of the calibrated variables. All model output was sensitive to changes in Ckand °ckThe Richard (1988) water routines improved model representation of the field watertable levels. Water table shapes during the years 1991—1993 were reproduced, and modelaccuracy was reasonable. New model estimated water table levels were generally within0.45 metres of the field measured levels. The new model was able to simulate the rapid riseof the field water table in response to precipitation events, but was unable to duplicatethe sudden, subsequent decreases. The old model assumption of a deep water tablepersisting below the 1.30 metre root zone was erroneous.At the Agassiz site there was no real difference between old model and new modelestimates of total water within the surface 90 centimetres of soil. By visual inspectionthe new model was judged to be marginally better than the old model at reproducing the131Chapter 7. Summary and Conclusions 132distribution of soil moisture within the profile. Discrepancies between model and fieldvalues may have been due to inadequate field data. Improvements in model representationof the soil water regime at Sumas may have been a result of the extensive calibration anduse of more realistic soil variables in the new model simulations rather than a consequenceof the new water routines.Old model and new model plots for NO, NH, and total inorganic N were virtuallyidentical for the Agassiz site, whereas for Sumas the new model plots differed noticeablyfrom those of the old. Both models reproduced the trends in nitrate and ammoniumvariations observed in the field, but did not reproduce the measured concentrations.New model predicted crop yields and crop-N contents were better than those of the oldmodel, but did not agree with the field measured values. This suggested that modelmineralization, nitrification, denitrification and crop N uptake routines were deficient.The old model assumption of a freely drained profile may have led to high estimatesof nitrate leaching losses and low estimates of denitrification losses. New model estimatesof nitrate leaching appeared to be unrealistically low. The leaching and denitrificationnitrate loss mechanisms were the model processes most affected by changes in the modelwater routines.The incorporation of the Richard (1988) water routines appeared to improve modelrepresentation of the soil water regime, but no significant improvement in simulated soilnitrogen processes was observed.7.1 Recommendations.Recommendations concerning the modified model are as follows:1. Mineralization, nitrification, denitrification, and crop uptake routines require further development. Realistic and up-to-date descriptions of soil nitrogen and cropChapter 7. Summary and Conclusions 133growth processes may improve the accuracy of model predictions.2. Further field studies are required to fully validate the model. In addition, morecomprehensive field studies would allow for better estimation of site specific soilmoisture variables such as the &(O) curves and K(O) functions. Field measuredvalues for factors used in model nitrogen calculations, such as mineralization anddenitrification rate coefficients, would also improve the predictive capacity of themodel. Modification of the model to accept distinct soil moisture and nitrogenparameters for each soil layer would be ideal.3. Field measured values for the initial January water table level and soil nitrate andammonium concentrations should be read in through the soil file for each simulatedsite.4. A more comprehensive deep flow description that accounts for differences in regionalhydrology should be developed. However, as every simulated site will be unique,this may be impractical. One possible solution would be to allow the user to specifythe amount of deep flow leaving the profile each day; daily deep flow values couldthen be determined during the calibration of the model for a specific site.5. Model N algorithms should be modified to allow for the application of two or moretypes of fertilizer.Bibliography[1] Addiscott, T.M., and R.J. Wagenet. 1985. Concepts of solute leaching in soils: areview of modelling approaches. Journal of Soil Science. 36:411-424.[2] Addiscott, T.M., and A.P. Whitmore. 1987. Computer simulation of changes in soilmineral nitrogen and crop nitrogen during autumn, winter and spring. J. agric. Sci.,Camb. 109:141-157.[3] Addiscott, T.M., and A.P. Whitmore. 1991. Simulation of solute leaching in soilsof differing permeabilities. Soil Use and Management. 7:94-102.[4] Addiscott, T.M., A.P. Whitmore, and D.S. Powison. 1991. Farming, Fertilizersand the Nitrate Problem. C.A.B International. Wallingford, U.K.[5] Al-Kanani, T., A.F. Mackenzie, and N.N. Barthakur. 1991. Soil water and ammoniavolatilization relationships with surface-applied nitrogen fertilizer solutions. SoilSci. Soc. Am. J. 55:1761-1766.[6] Aulakh, M.S., J.W. Doran, D.T. Walters, and J.F. Power. 1991. Legume residueand soil water effects on denitrification in soils of different textures. Soil Biol.Biochem. 23:1161-1167.[7] Barraclough, D. 1989a. A usable model of nitrate leaching 1. The model. Journalof Soil Science. 40:543-554.[8] Barraclough, D. 1989b. A usable model of nitrate leaching 2. Application. Journalof Soil Science. 40:555-562.[9] Bear, J. 1972. Dynamics of Fluids in Porous Media. American Elsevier PublishingCo. NY, NY.[10] Beauchamp, E.G., and J.W. Paul. 1989. A simple model to predict manure Navailability to crops in the field. In Hansen, J.A., and K. Hendriksen (eds), Nitrogenin Organic Wastes Applied to Soils. Academic Press, Toronto. 140-150.[11] Bergstrom, L. 1987. Nitrate leaching and drainage from annual and perennial cropsin tile drained plots and lysimeters. J. Environ. Qual. 16:11-18.[12] Bertrand, R.A., and N.R. Bulley. 1985. Manure Management Guidelines. BritishColumbia Ministry of Agriculture and Food.134Bibliography 135[13] Bertrand, R.A., H. Sasaki, and D. Moon. 1989. Soil and Water Strategy for BritishColumbia. British Columbia Ministry of Agriculture and Fisheries.[14] Bhat, K.K.S., T.H. Flowers, and J.R. O’Callaghan. 1981. A model for the simulation of the fate of nitrogen in farm wastes on land application. In Brogan, J.C. (ed),Nitrogen Losses and Surface Run-off. ECSC, EEC, EAEC. Brussels-Luxembourg.222-246.[15] Bloom, E.P. 1988. Function date 29. The Turbo C Trilogy. Windcrest Books, BlueRidge Summit, P.A. pg. 432.[16] Broughton, R.S., and N. Foroud. 1978. A model to predict water table depths forflat lands. Canadian Agricultural Engineering. 20:81-86.[17] Bulley, N.R., and B. Cappelaere. 1978. A Dynamic Simulation of Nitrogen Movement on Livestock Farms. Presented at the 1978 Annual Meeting, Paper No. 78-211,Canadian Society of Agricultural Engineers.[18] Cabon, F., G. Girard, and E. Ledoux. 1991. Modelling of the nitrogen cycle in farmland areas. Fertilizer Research. 27:161-169.[19] Cameron, D.R., C.G. Kowalenko, and C.A. Campbell. 1979. Factors affecting nitrate nitrogen and chloride leaching variability in a field plot. Soil Sci. Soc. Am. J.43:455-460.[20] Campbell, G.S. 1974. A simple method for determining unsaturated conductivityfrom moisture retention data. Soil Science. 117:311-314.[21] Chao, E.C.Y. 1987. Water Table Depth Simulation for Flat Agricultural Land Under Subsurface Drainage and Subirrigation Practices. M.A.Sc. Thesis. The Facultyof Graduate Studies, Department of Bio-Resource Engineering, U.B.C. Vancouver,B.C.[22] Christensen, S., and J.M. Tiedje. 1988. Denitrification in the field, analysis ofspatial and temporal variability. In Jenkinson, D.S., and K.A. Smith (eds), NitrogenEfficiency in Agricultural Soils. Elsevier, NY. 295-301.[23] Christian, D., M. Goss, R. Howse, D. Powison, and T.J. Pepper. 1990. Leaching ofnitrate through soils. IACR Report for 1989. Lawes Agricultural Trust. 67-68.[24] Colbourn, P. 1992. Denitrification and N20 production in pasture soil: the influenceof nitrogen supply and moisture. Agriculture, Ecosystems and Environment. 39:267-278.Bibliography 136[25] Darusman, L.R. Stone, D.A. Whitney, K.A. Janssen, and J.H. Long. 1991. Soilproperties after twenty years of fertilization with different nitrogen sources. SoilSci. Soc. Am. J. 55:1097-1100.[26] Dekker, L.W., and Bouma J. 1984. Nitrogen leaching during sprinkler irrigation ofa Dutch clay soil. Agricultural Water Management. 9:37-45.[27] De Jong, R. 1981. Soil Water Models: A Review. Agriculture Canada ResearchBranch. LRRI No. 123.[28] Devitt, D., J. Letley, L.J. Lund, and J.W. Blair. 1976. Nitrate-nitrogen movementthrough soils as affected by soil profile characteristics. J. Environ. Qual. 5:283-288.[29] de Willigen, P. 1991. Nitrogen turnover in the soil-crop system; comparison offourteen simulation models. In J.J.R. Groot et al. (eds), Nitrogen Turnover in theSoil-Crop System. Kiuwer, London. 141-149.[30] Fillery, I.R.P. 1983. Biological denitrification. In Freney, J.R., and J.R. Simpson(eds), Gaseous Loss of Nitrogen from Plant-Soil Systems. Martinus Nijhoff/Dr W.Junk Publishers, The Netherlands. 33-64.[31] France, J., and J.H.M. Thornley. 1984. Mathematical Models in Agriculture. Butterworths, Toronto. 12-13.[32] Freeze, R.A. 1969. The mechanism of natural ground-water recharge and discharge1. One dimensional, vertical, unsteady, unsaturated flow above a recharging ordischarging ground-water flow system. Water Resources Research. 5:153-171.[33] Freeze, R.A., and J. Banner. 1970. The mechanism of natural ground-water rechargeand discharge 2. Laboratory column experiments and field measurements. WaterResources Research. 6:138-155.[34] Gardner, W.R. 1960. Soil water relations in arid and semi-arid conditions. UNESCO. 15:37-61.[35] Gartner Lee Limited. 1993. Fraser Valley Ground Water Monitoring Program Phase1 Report. Prepared for Ministry of Health, British Columbia, April 1993.[36] Gast, R.G., W.W. Nelson, and G.W. Randall. 1978. Nitrate accumulation in soilsand loss in tile drainage following nitrogen applications to continuous corn. J.Environ. Qual. 7:258-261.[37] Gysi, C. 1990a. Modelling and measurement of the nitrogen cycle in a vegetablefield in Switzerland. 1. A soil-plant model for the nitrogen cycle. Z. Pflanzerierniihr.Bodenk. 153:181-187.Bibliography 137[38] Gysi, C. 1990b. Modelling and measurement of the nitrogen cycle in a vegetablefield in Switzerland. 2. Validation of the nitrogen model with field data. Z. Pflanzenerniihr. Bodenk. 153:189-196.[39] Harris, P.J. 1988. Microbial transformations of nitrogen. In Wild, A. (ed), Russell’sSoil Conditions and Plant Growth. 11th edn. Longman, London. 608-651.[40] Health and Welfare Canada. 1989. Guidelines for Canadian Drinking Water Quality. 4rth ed. Minister of National Health and Welfare Canada.[41] Heaney, D.J., M. Nyborg, E.D. Solberg, S.S. Maihi, and J. Ashworth. 1992. Over-winter nitrate loss and denitrification potential of cultivated soils in Alberta. SoilBiol. Biochem. 24:877-884.[42] Hubbard, R.K., R.A. Leonard, and A.W. Johnson. 1991. Nitrate transport on asandy coastal plain soil underlain by plinthite. Transactions of the ASAE. 34:802-808.[43] Hutson, J.L. and R.J. Wagenet. 1991. Simulating nitrogen dynamics in soils usinga deterministic model. Soil Use and Management. 7:74-78.[44] Inland Waters Directorate. 1992. Historical Water Levels Summary to 1990 BritishColumbia. Water Resources Branch, Water Survey of Canada. Ottawa.[45] Iqbal, M.M., and B.P. Warkentin. 1983. Nitrogen and phosphorus content of flowfrom drains fertilized with cow manure slurry. Plant and Soil Science. 63:517-521.[46] Jansson, P.—E., R.S. Antil, and G. Ch. Borg. 1989. Simulation of nitrate leachingfrom arable soils treated with manure. In Hansen, J.A., and K. Hendriksen (eds),Nitrogen in Organic Wastes Applied to Soils. Academic Press, Toronto. 150-166.[47] Jenkinson, D.S. 1988. Soil organic matter and its dynamics. In Wild, A. (ed),Russell’s Soil Conditions and Plant Growth. 11th edn. Longman, London. 564-607.[48] Johnsson, H., L. Bergstrom, P.-E. Jansson, and K. Paustian. 1987. Simulated nitrogen dynamics and losses in a layered agricultural soil. Agriculture, Ecosystemsand Environment. 18:333-356.[49] Jury, W.A., and C.B. Tanner. 1975. A modification of the Priestley and Taylorevapotranspiration formula. Agron. J. 67:840-842.[50] Kanwar, R.S., H.P. Johnson, and J.L. Baker. 1983. Comparison of simulated andmeasured nitrate losses in tile effluent. Transactions of the ASAE. 26:1451-1457.Bibliography 138[51] Kattelman, R.C., N.H. Berg and M.K. Pack. 1985. Estimating regional snow waterequivalent with a simple simulation model. Water Resources Bulletin. 21:273-280.[52] Keeney, D.R. 1982. Nitrogen management for maximum efficiency and minimumpollution. In Stevenson, F.J. (ed.), Nitrogen in Agricultural Soils. American Societyof Agronomy, Madison, WI. 605-649.[53] Kersebaum, ICC., and J. Richter. 1991. Modelling nitrogen dynamics in a plant-soilsystem with a simple model for advisory purposes. Fertilizer Research. 273-281.[54] Kowalenko, C.G. 1987a. An Evaluation of Nitrogen Use in British Columbia Agriculture. Agriculture Canada Technical Bulletin No. 1987-3e.[55] Kowalenko, C.G. 1987b. The dynamics of inorganic nitrogen in a Fraser Valley soilwith and without spring or fall ammonium nitrate applications. Can. J. Soil Sci.67:367-382.[56] Kowalenko, C.G., and D.R. Cameron. 1976. Nitrogen transformations man incubated soil as affected by combinations of moisture content and temperature andadsorption-fixation of ammonium. Can. J. Soil Science. 56:63-70.[57] Kowalenko, C.G., and J.W. Hall. 1987. Nitrogen recovery in single- and multiple-harvested direct-seeded broccoli trials. J. Amer. Soc. Hon. Sci. 112(1):4-8[58] Krajenbrink, G.J.W., L.J.M. Boumans, and C.R. Meinardi. 1989. Hydrochemicalprocesses in the top layer of groundwater under pasture land. In Hansen, J.A., andK. Hendriksen (eds), Nitrogen in Organic Wastes Applied to Soils. Academic Press,Toronto. 317-333.[59] Lawes, J.B., J.H. Gilbert, and R. Warington. 1882. On the amount and compositionof the rain and drainage waters collected at Rothhamsted. Part III. J. R. Agric.Soc. Engl. Ser. 2, 18:1-71.[60] Legg, J.O., and J.J. Meisinger. 1982. Soil nitrogen budgets. In Stevenson, F.J.(ed.), Nitrogen in Agricultural Soils. American Society of Agronomy, Madison,WI. 503-566.[61] Liebscher, H., B. Ru, and D. McNaughton. 1992. Nitrates and Pesticides in the Abbotsford Aquifer Southwestern British Columbia. Environment Canada. July 1992.[62] Loague, K., and R.E. Green. 1991. Statistical and graphical methods for evaluating solute transport models: overview and applications. Journal of ContaminantHydrology. 7:51-73.Bibliography 139[63] Lord, E.I., and G. Bland. 1991. Leaching of spring-applied fertilizer nitrogen: measurement and simulation. Soil Use and Management. 7:110-114.[64] Lowe, P.R. 1977. An approximating polynomial for computation of saturation vaporpressure. J. Appi. Meteorol. 16:100-103.[65] Macdonald, A.J., D.S. Powison, P.R. Poulton, and D.S. Jenkinson. 1989. Unusedfertiliser nitrogen in arable soils—its contribution to nitrate leaching. J. Sci. Food.Agric. 46:407-419.[66] MacGregor, J.M., G.R. Blake, and S.D. Evans. 1974. Mineral nitrogen movementinto subsoils following continued annual fertilization for corn. Soil Sci. Soc. Amer.Proc. 38:110-112.[67] Madramootoo, C.A., and R.S. Broughton. 1987. A computer simulation model ofsurface and subsurface flows from agricultural areas. Canadian Water ResourcesJournal. 12:30-43.[68] Madramootoo, C.A., R.S. Broughton, 5.0. Prasher, and F. Peyrow. 1987. Development and testing of an agricultural land drainage simulation model. Computersand Electronics in Agriculture. 2:15-30.[69] Maidl, F.X., and G. Fischbeck. 1989. Effects of long-term application of slurry onsoil nitrogen mineralization. J. Agronomy and Crop Science. 162:310-319[70] Mathers, A.C., and B.A. Stewart. 1983. Manure effects on crop yield and soilproperties. Presented at the 1983 ASAE Summer Meeting, June 26-29, Bozeman,Montana. Paper no. 83-2120.[71] McDougall, N.R., W.T. Murrie, and D.T. Price. 1989a. Familiarization with theStructure, Function and Execution of the Manure Management Model. ARDSAProject Number 11025 “Nitrogen in South Coastal B.C.”Report to Soils and Engineering Branch, B.C. Ministry of Agriculture and Fisheries. Abbotsford, B.C.[72] McDougall, N.R., W.T. Murrie, and D.T. Price. 1989b. Report on Upgrades Required to be Implemented in the Manure Management Model. ARDSA ProjectNumber 11025 “Nitrogen in South Coastal B.C.” Report to Soils and Engineering Branch, B.C. Ministry of Agriculture and Fisheries. Abbotsford, B.C.[73] McGill, W.B., H.W. Hunt, R.G. Woodmansee, J.O. Reuss, and K.H. Paustian.1980. Formulation, process controls, parameters and performance of PHOENIX: Amodel of carbon and nitrogen dynamics in grassland soils. In Frissel, M.J., and J.A.van Veen (eds), Simulation of Nitrogen Behaviour of Soil-Plant Systems. Pudoc.The Netherlands. 171-191.Bibliography 140[74] Meek, B.D., L.B. Grass, L.S. Willardson, and A.J. Mackenzie. 1970. Nitrate transformations in a column with a controlled water table. Soil Sci. Soc. Amer. Proc.34:235-239.[75] Milly, P.C.D. 1988. Advances in modeling of water in the unsaturated zone. In E.Custodio et al. (eds), Groundwater Flow and Quality Modelling. D. Reidel Publishing Co. Dordrecht, Holland. 489-514.[76] Mochoge, B.O. 1984. Simulation of nitrate movement in undisturbed soil columns.Agriculture, Ecosystems and Environment. 11:105-115.[77] Mualem, Y. 1976. A Catalogue of the Hydraulic Properties of Unsaturated Soils.Research Project No. 442, Technion, Israel Institute of Technology, Haifa, Israel.[78] Nielsen, D.R., J.W. Biggar, and K.T. Erh. 1973. Spatial variability of field-measured soil-water properties. Hilgardia. 42(7) :215-260.[79] Nielsen, D.R., J.W. Biggar, and P.J. Wierenga. 1982. Nitrogen transport processesin soil. In Stevenson, F.J. (ed.), Nitrogen in Agricultural Soils. American Societyof Agronomy, Madison, WI. 423-448.[80] Otoma, S., and T. Kuboi. 1985. Model simulation of solute leaching and its application for estimating the net rate of nitrate formation under field conditions.Journal of Hydrology. 82:193-209.[81] Pain, B.F. 1991. Improving the utilisation of slurry and farm effluents. In Mayne,C.S. (ed), Management Issues for the Grassland Farmer in the 1990’s. BritishGrassland Society. Antrim, N.J. 121-133.[82] Paul, J.W. 1993. Personal communications. Agriculture Canada, Agassiz ResearchStation, Agassiz, B.C., Canada.[83] Paustian, K. 1981. Simulation modelling. In Rosswall, T. (ed), Ecology of ArableLand: The Role of Organisms in Nitrogen Cycling Progress Report 1981. A Contribution to the UNESCO Program Man and the Biosphere. Uppsala, Sweden. 221-239.[84] Philip, J.R. 1991. Soils, natural science, and models. Soil Science. 151:91-98.[85] Pikul, M.F., R.L. Street, and I. Remson. 1974. A numerical model based on coupledone-dimensional Richards and Boussinesq equations. Water Resources Research.10:295-302.[86] Powlson, D.S., P.R. Poulton, T.M. Addiscott, and D.S. McCann. 1989. Leachingof nitrate from soils receiving organic or inorganic fertilizers continuously for 135years. In Hansen, J.A., and K. Hendriksen (eds) Nitrogen in Organic Wastes Applied to Soils. Academic Press, Toronto. 334-345.Bibliography 141[87] Price, D.T. 1990. User’s Guide to the Manure Management Model for IBM-PCCompatible Microcomputers. ARDSA Project Number 11025 “Nitrogen in SouthCoastal B.C.” Report to Soils and Engineering Branch, B.C. Ministry of Agricultureand Fisheries. Abbotsford, B.C.[88] Priestley, C.H.B., and R.J. Taylor. 1972. On the assessment of surface heat fluxand evaporation using large scale parameters. Monthly Weather Review. 100:81-92.[89] Ranjan, R.S., T. Karthigesu, and N.R. Bulley. 1993. Application of a manure-N management model under Manitoba conditions. Presented at the 1993 ASAEInternational Summer Meeting, Spokane, Washington. June 20-23. Paper No. 93-4022.[90] Reddy, K.R., and D.A. Graetz. 1981. Use of shallow reservoir and flooded organicsoil systems for waste water treatment: nitrogen and phosphorus transformations.J. Environ. Qual. 10:113-119.[91] Reddy, K.R., P.D. Sacco, and D.A. Graetz. 1980. Nitrate reduction in an organicsoil-water system. J. Environ. Qual. 9:283-288.[92] Richard, P.F. 1988. A Computer Analysis of the Flow of Water and Nutrients inAgricultural Soils as Affected by Subsurface Drainage. Ph.D. Thesis. The Facultyof Graduate Studies, Resource Management Science, U.B.C. Vancouver, B.C.[93] Richards, L.A. 1931. Capillary conduction of liquids through porous mediums.Physics. 1:318-333.[94] Richter, J., A. Nuske, M. Boehmer, and J. Wehrman. 1980. Simulation of nitrogenmineralization and transport in loess-parabrownearthes plot experiments. Plantand Soil. 54:329-337.[95] Richter, J., H. Nordmeyer, and K. C. Kersebaum. 1985. Simulation of nitrogenregime in bess soils in the winter half-year: comparison between field measurementsand simulations. Plant and Soil. 83:419-431.[96] Ritter, W.F., A.E.M. Chirnside, and R.W. Scarborough. 1985. Leaching of nitratesunder irrigation on coastal plain soils. Paper presented at the ASAE Winter Meeting, Dec. 17-20, 1985. Chicago, Illinois. Paper no. 85-2613.[97] Rubin, J. 1966. Theory of rainfall uptake by soils initially drier than their fieldcapacity and its applications. Water Resource Research. 2:739-749.[98] Smith, R.E., and V.A. Ferreira. 1989. Comparative evaluation of unsaturated flowmethods in selected USDA simulation models. In Morel- Seytoux, H.J. (ed), Unsaturated Flow in Hydrologic Modeling. Kluwer Academic Publishers. Boston. 391-412.Bibliography 142[99] Smith, J.H. and J.R. Peterson. 1982. Recycling of nitrogen through land applicationof agricultural, food processing, and municipal wastes. In Stevenson, F.J. (ed.),Nitrogen in Agricultural Soils. American Society of Agronomy, Madison, WI. 791-832.[100] Sommerfeldt, T.G., and C. Chang. 1985. Changes in soil properties under annualapplication of feedlot manure and different tillage practices. Soil Sci. Soc. Am. J.49:983-987.[101] Stevenson, F.J. 1982. Origin and distribution of N in soil. In Stevenson, F.J. (ed.),Nitrogen in Agricultural Soils. American Society of Agronomy, Madison, WI. 1-42.[102] Tanji, K.K. 1982. Modeling of the soil nitrogen cycle. In Stevenson, F.J. (ed.),Nitrogen in Agricultural Soils. American Society of Agronomy, Madison, WI. 721-772.[103] Tubbs, L.J., and D.A. Haith. 1981. Simulation model for agricultural non-point-source pollution. Journal WPCF. 53:1425-1433.[104] Tillotson, W.R., and R.J. Wagenet. 1982. Simulation of fertilizer nitrogen undercropped situations. Soil Science. 133:133-143.[105] van Keulen, H., and W. Stol. 1991. Quantitative aspects of nitrogen nutrition incrops. In J.J.R. Groot et al. (eds), Nitrogen Turnover in the Soil-Crop System.Kluwer, London. 151-160.[106] Van Ommen, H.C. 1985. Systems approach to an unsaturated- saturated groundwater quality model, including adsorption, decomposition and bypass. AgriculturalWater Management. 10:193-203.[107] Verdegem, L., and Baert L. 1984. Losses of nitrate nitrogen in sandy and clayeysoils. Pedologie. 34-3:235-255.[108] Verdegem, L., 0. Van Cleemput, and J. Vanderdeelen. 1981. Some factors inducingthe loss of nutrients out of the soil profile. Pedologie. 31-3:309-327.[109] Vereecken, H., M. Vanclooster, M.Swerts, J. Diels. 1991. Simulating water andnitrogen behaviour in soils cropped with winter wheat. Fertilizer Research. 27:233-243.[110] Vinten, A.J.A., R.S. Howard, and M.H. Redman. 1991. Measurement of nitrateleaching losses from arable plots under different nitrogen input regimes. Soil Useand Management. 7:3-14.Bibliography 143[111] Wagenet, R.J., and B.K. Rao. 1983. Description of nitrogen movement in the presence of spatially variable soil hydraulic properties. Agricultural Waste Management.6:227-242.[112] Walley, W.J., and D.E.D.A. Hussein. 1982. Development and testing of a generalpurpose soil-moisture-plant model. Hydrological Sciences Journal. 27:1-17.[113] Warrick, A.W., G.J. Mullen, and D.R. Nielsen. 1977. Predictions of the soil waterflux based upon field-measured soil-water properties. Soil Sci. Soc. Amer. 41:14-19.[114] White, R.E. 1985a. The analysis of solute breakthrough curves to predict water redistribution during unsteady flow through undisturbed structured clay soil. Journalof Hydrology. 79:21-35.[115] White, R.E. 1985b. A model for nitrate leaching in undisturbed structured clay soilduring unsteady flow. Journal of Hydrology. 79:37-51.[116] White, R.E., S.R. Wellings, and J.P. Bell. 1983. Seasonal variations in nitrate leaching in structured clay soils under mixed land use. Agricultural Water Management.7:391-410.[117] Whitman, Walt. 1855. The Compost. Leaves of Grass. Self published.[118] Whitmore, A.P., and T.M. Addiscott. 1986. Computer simulation of winter leachinglosses of nitrate from soils cropped with winter wheat. Soil Use and Management.2:26-30.[119] Whitmore, A.P., and T.M. Addiscott. 1987. Applications of computer modelling topredict mineral nitrogen in soil and nitrogen in crops. Soil Use and Management.3:38-43.[120] Wild, A. 1988. Plant nutrients in soil: nitrogen. In Wild, A. (ed), Russell’s SoilConditions and Plant Growth. 11th edn. Longman, London. 652-694.[121] Yeoman, J.C., J.M. Bremner, and G.W. McCarty. 1992. Denitrification capacityand denitrification potential of subsurface soils. Commun. Soil Sci. Plant Anal.23(9&zJO) :919-927.[122] Young, J.L. and R.W. Aldag. 1982. Inorganic forms of nitrogen in soil. In Stevenson,F.J. (ed.), Nitrogen in Agricultural Soils. American Society of Agronomy, Madison,WI. 43-66.[123] Zebarth, B.J., J.W. Paul, 0. Schmidt, and R. McDougall. 1993. Efficient ManureN Use in Silage Corn Production. Paper presented at the ASAE/CSAE SummerMeeting, June 20-23, 1993. Spokane, Washington. Paper no. 934024.Bibliography 144[124] Zebarth, B.J. and J.W. Paul. 1994. Influence of the rate and time of liquid dairy manure application on leaching and denitrificatiori in south coastal British Columbia,Final report for Project #S2201, funded under the Canada-B.C. Soil ConservationProgram. Pacific Agriculture Research Centre (Agassiz) Technical Report No. 99,in press.Appendix AClimate Calculation ProgramThis subprogram prepares the climate file read by the nitrogen management model basedon the following values: daily minimum, maximum, and average temperatures (Tmin,Tmax, Tav), sunshine percent hours (S%), total precipitation (F) and the latitude of thesimulated site (l). Its primary function is to determine daily values for relative humidityand potential evapotranspiration. A point summary of the calculations performed foreach day is presented below.• The Julian day of the year J is calculated based on the day’s date and the algorithmpresented in Bloom (1988).• A mean relative humidity value is estimated as the ratio of the saturated vaporpressure at Tmin to the saturated vapor pressure at Tav, calculated asSVP(Tmin)Rh= SVP(Tav) (A.1)where SVP(T) is the saturated vapor pressure at the specified temperature T. Saturated vapor pressure is determined using a sixth order polynomial approximationexpressed asSVP(T)=W0+T.(W12-f-T.(3456))) )(A.2)where W0—W6 are constant coefficients (Lowe, 1977). If Tav 0 saturated vaporpressure over ice is calculated; otherwise the saturated vapor pressure over wateris determined.145Appendix A. Climate Calculation Program 146• Tav is converted to Kelvins, TK.• The total extraterrestrial solar radiation Er is calculates asEr = 0.0864 Ce Sc (cos ir cos A8d sin A3 + A3 sin ir sin A3d) (A.3)where 0.0864 is a factor used to convert to MJ m2 day’, C8 is the extraterrestrialsolar constant, 4. is the latitude converted to radians, A3d is the solar declinationangle for the Julian day J in radians calculated asA3d = 0.41 . sin(0.0172 (284 + J)) (A.4)S is the standard correction for solar declination angle expressed as= I + 0.033 . cos(0.0172 . J) (A.5)and A3 is the arc of the solar track for the Julian day in radians expressed asA3 = arccos(— tan(lr) tan(A3d)) (A.6)• The estimated shortwave irradiance at the surface, corrected for atmospheric absorption effects, is determined usingR8 = Er (0.23 + 0.0053 S%) (A.7)• Emissivity of the vegetation surface is estimated as= 0.261 exp (—7.77. i04 Ta2,) (A.8)• Net irradiance at the surface is estimated as the sum of the net shortwave andlongwave radiation, corrected for cloudiness, as(A.9)where ,3 is the short wave reflectivity, ç is the clear sky atmospheric emissivity,and o is the Stefan-Boltzmann constant.Appendix A. Climate Calculation Program 147• If RN < 0, or Ta < 0, then PET = 0. Otherwise, potential evapotranspirationin mm/day is calculated using the Jury and Tanner (1975) simplification of thePriestley and Taylor (1972) method, expressed asP = ((1 + C• (1 — Rh)). (0.41 + 0.014•Tav) R• 1000) t (A.10)where C8 is the characteristic site coefficient and Lh is the latent heat of vaporization of water with linear temperature correction to 0°C, in J/g, determinedbyLh = 2499 — 2.386 Tav (A.11)Appendix BManure Production RoutinesTwo routines, MNPROD and AMNPRD, are used to calculate manure production andthe herd size required to amass the amount of manure applied.MNPROD. This routine calculates the amount of manurial N produced betweentwo dates, d1 and d2, for one ‘unit of operation’ asNman = d2Fjq,: . X . . L3 (B.1)i=d1where FN is the feed N ingested by one ‘unit of operation’, X2 is the excretion coefficient,L is the collection loss coefficient, and L3 is the storage loss coefficient. If d1 and d2designate two subsequent manure application dates, the number of ‘units of operation’required to produce the amount of manure applied on day d2 in the timespan d2— di isexpressed asu = Aman (B.2)manAMNPRD. This routine calculates the amount of manure produced over the entireyear by one ‘unit of operation’ and the number of ‘units of operation’ required to supplythe total amount of manure applied to the soil. It determines a value for manurial Nproduced over the entire year by calling subroutine MNPROD and then calculates thenumber of ‘units of operation’ required in a similar fashion to Equation B.2.148Appendix CModel Simulation of Several Hypothetical Field ScenariosWhile the sensitivity analysis highlighted the effects of single variable manipulation onmodel output, it did not afford an examination of model behaviour for various field sites.In order to illustrate trends in model output under different field conditions, severalinput files were selected to produce a ‘typical’ field scenario. This consisted of a dairymanure file, a climate file with average precipitation, a silty clay loam soil, and a corncrop. Three simulations were run using this combination of files, each with a differentamount of manure applied. Using these three simulations as a benchmark, each of thefiles comprising the ‘typical’ field scenario was replaced, in turn, by another input fileof the same type. Substituting standard input files in this fashion allowed the model tosimulate many different field sites.Though hundreds of field simulations were possible, this study was limited to thefollowing simulations:Bench Benchmark runs consisting of the following files:• Climate: average precipitation, 173.2 cm yr1.• Soil: silty clay loam.• Crop: corn.• Manure: dairy; 250, 500, 1000 kg N ha’ yr’.Wet Three simulations as per the benchmark run but with a wet year (233.1 cm yr1)substituted for the average climate file.149Appendix C. Model Simulation of Several Hypothetical Field Scenarios 150Dry Three simulations as per the benchmark run but with a dry year (131.3 cm yr1)substituted for the average climate file.Sand A sand file replacing the silty clay loam file in the three runs.Clay A clay loam file substituted for the silty clay loam file in the three runs.Swine Use of swine manure rather than dairy; three runs.Poult Use of a poultry manure rather than dairy; three runs.Grass A grass crop substituted for the corn crop; three runs.Rasp A raspberry crop substituted for the corn crop; three runs.No C No crop planted; three runs.No M No manure applied; one run.The results of each simulation on yearly model output for a one year run are tabulatedin Table C.1.An increase in the amount of manure applied resulted in a subsequent increase inmanure mineralization, volatilization, denitrification, and leaching losses for all scenarios.For the poultry manure runs the amount of ammonium gained by mineralization ofsoil organic matter also increased slightly with an increase in the amount of manureapplied. This is a consequence of two variables in the poultry manure file, Mm, theinitial mineralization rate of manure N, and Hm, the fraction of manure remaining as soilhumus. The Mm value in the poultry file is an order of magnitude larger than the valuein the swine manure file, and two orders of magnitude larger than the value in the dairymanure file. In addition, the Hm value for the poultry manure is slightly larger than boththe values in the swine and dairy files. As a result, the organic nitrogen in the poultryAppendix C. Model Simulation of Several Hypothetical Field Scenarios 151Table C.1: Field scenario results. New model.Manure Crop NH Mineralztn. LossesAppid. Yld. N Cont. Appid. Man. Soil Volat. Denit. Leach.kg kg kg kg kg kg kg kgh h Ii l250 15 168 83 40 191 42 92 9BENCH 500 18 261 166 80 191 84 107 91000 20 421 333 160 191 168 148 12250 14 158 83 39 187 42 102 16WET 500 17 254 166 79 187 84 117 171000 20 412 333 158 187 168 163 22250 16 184 83 40 196 42 70 4DRY 500 20 281 166 80 196 84 79 41000 21 442 333 160 196 168 111 6250 11 123 83 37 116 42 1 93SAND 500 15 218 166 75 116 84 1 1051000 18 376 333 149 116 168 2 155250 17 562 83 43 1104 42 201 59CLAY 500 12 452 166 86 1104 84 261 891000 6 220 333 172 1104 168 353 145250 17 224 113 70 191 13 97 9SWINE 500 20 364 225 140 191 25 120 101000 13 423 450 281 191 50 259 26250 17 236 113 62 192 13 89 8POULT 500 20 375 225 124 193 25 111 101000 10 343 450 248 195 50 282 32250 19 233 83 40 192 42 59 5GRASS 500 22 344 166 80 192 84 65 61000 21 532 333 160 192 168 94 7250 5 121 83 40 192 42 116 40RASP 500 5 173 166 80 192 84 151 521000 5 214 333 160 192 168 242 93250 0 0 83 41 194 42 175 12NO C 500 0 0 166 81 194 84 228 161000 0 0 333 162 194 168 319 24NO M 0 9 71 0 0 191 0 79 8Appendix C. Model Simulation of Several Hypothetical Field Scenarios 152manure is mineralized faster than that in either the swine or dairy manure, and more ofit is relegated to the soil humus following manurial N mineralization. This increases thesoil organic N content, H80, and thereby increases the gain of ammonium through soilorganic N mineralization according to Equation 3.43.The effect on crop yield of increased amounts of applied manure depended on theother conditions specified. For the corn crop an increase in the amount of manure appliedgenerally resulted in higher crop yields, whereas the raspberry and grass crops respondedonly slightly to larger applications of manure. Excessive nitrogen, in particular the 1000kg ha1 application to the clay loam soil, served to decrease crop yield. Manure type hada minor effect on crop yield but significantly affected gains of NH through manurial Nmineralization and losses of nitrogen through volatilization, denitrification and leaching.Increasing the yearly precipitation increased both denitrification and leaching whiledecreasing slightly the gains through mineralization of both soil and manure. Decreasingprecipitation had the opposite effects; denitrification and leaching decreased while mineralization gains increased. Crop yields decreased slightly with increased precipitationand increased slightly with decreased precipitation.The soil type affected all model output except volatilization and NHt applied. Leaching losses were greatest on the sand and least on the silty clay loam. Denitrification losseson the clay loam exceeded those on both the sand and silty clay loam.Both denitrification and leaching losses were affected by the type of crop planted.Denitrification losses were least under the grass crop and greatest when no crop wasplanted. Nitrate lost to leaching was greatest under the raspberry crop; this result,however, is misleading and deserves a word of explanation. What the model presents asthe leaching loss for a given simulation is the sum of the nitrate passing out of the rootzone. The root zone, in turn, is dictated by the type of crop planted; for a raspberry cropthe root zone is 90 centimetres, while for all of the other crops, including the zero cropAppendix C. Model Simulation of Several Hypothetical Field Scenarios 153S04..).0f-i4..)file, the root zone is set to 130 centimetres. Thus the leaching losses under a raspberrycrop, with its shallower rooting depth, will always exceed the losses from a crop with adeeper rooting depth. This effect is also evident in the final rooting depth modificationsin the sensitivity analysis.Fluctuations in water table height were also observed under the different field simulations. Figures C.1 and C.2 illustrate the effect of the soil and climate files on thewater table levels. While the amount and type of manure applied had no effect on watertable height, crop type did mildly influence the level of the water table in the summermonths. This is a result of the varying rooting depths and was also noted in the sensitivityanalysis.To further investigate the effects of modifying the soil water movement routines thesame field scenarios were simulated using the original version of the nitrogen managementmodel; Table C.2 presents the results of the old model simulations.In general, the old model lost more total nitrate, as the sum of denitrification andJulian DayFigure C.1: Effect of soil type on water table level.Appendix C. Model Simulation of Several Hypothetical Field Scenarios 154Table C.2: Field scenario results. Old model.250 13 163 83 36 160 42 32 74BENCH 500 16 256 166 71 160 84 37 831000 16 377 333 142 160 168 57 149250 12 155 83 37 171 42 42 106WET 500 17 256 166 74 171 84 49 1121000 17 383 333 149 171 168 74 181250 13 167 83 35 162 42 36 40DRY 500 16 261 166 69 162 84 42 441000 16 374 333 138 162 168 71 79250 8 110 83 32 93 42 0 113SAND 500 11 190 166 64 93 84 0 1351000 12 308 333 127 93 168 1 216250 6 249 83 44 1139 42 38 421CLAY 500 3 132 166 88 1139 84 42 4851000 1 35 333 175 1139 168 46 542250 15 219 113 68 160 13 34 78SWINE 500 19 356 225 135 160 25 43 1001000 8 301 450 271 160 50 98 378250 16 238 113 62 161 13 31 66POULT 500 17 357 225 124 162 25 44 1011000 6 232 450 248 163 50 104 433250 15 207 83 34 152 42 20 44GRASS 500 17 308 166 68 152 84 22 481000 17 493 333 136 152 168 32 67250 6 125 83 37 166 42 42 120RASP 500 6 178 166 73 166 84 54 1651000 5 208 333 146 166 168 82 323250 0 0 83 38 175 42 70 199NO C 500 0 0 166 76 175 84 88 2801000 0 0 333 152 175 168 116 445Manure Crop NH Mineralztn.Appld. Yld. N Cont. Appid. Man. Soil Volat. Denit. Leach.g t ka kaLossesNOM 0 7 64 0 0 160 0 29 69Appendix C. Model Simulation of Several Hypothetical Field Scenarios 155leaching losses, than the new model under the same conditions. Leaching losses predictedby the old model are substantially higher than those predicted by the new model, whereasthe new model yields greater values for denitrification than the old. Crop yields, crop Ncontent, and mineralization gains are similar for both models with the new model valuesbeing only slightly higher than the values for the old model. Ammonium applied andvolatilization losses are identical for both model since they are affected only by the typeand amount of manure applied.S-0.5‘V‘V91 121 151 181Julian DayFigure C.2: Effect of precipitation on water table level.Appendix DModel Input ValuesA listing of the input file values used to generate the model results given in Chapters 5and 6 is presented below in Tables D.1 and D.2. Variable descriptions are given inTable 3.1. As the full 1993 year was not simulated, annual precipitation and potentialevapotranspiration values are not listed.Parties interested in the nitrogen management model should direct all inquiries tothe British Columbia Ministry of Agriculture, Fisheries and Food, Resource ManagementBranch. The model can be compiled to run on any IBM compatible PC with a minimumof 640 kB of RAM. However, as execution on PC-XT or 286 machines is time consuming,the minimum operating system recommended is a 386-based machine with a math coprocessor.156Appendix D. Model Input Values 157Table D.1: Model input values, Agassiz 1991-1993.File Variable Value(s)Soil 0wp 0.130FC1, 0FC2 0.41, 0.430.54°ck, Ck 0.48, 45ST 0.02pb 1.29 (g cm3)z7,1 20 (cm)H3 0.3(%)M3 0.00007 (d)a 21.0Manure X 0.700.900.67L3 0.95, 0.9, 0.87, 0.83, 0.8(Jan., Feb., Mar., Apr., May-Dec.)N0 0.70 (1991), 0.55 (1992-3)N, 0.30 (1991), 0.45 (1992-3)Mm 0.002 (d1)Hm 0.4FN 12.52 (kg . wk’)Crop H1, HF 20, 130 (cm)HL 50 (cm)g1, g2 10, 70 (cm 1120 transpired)P 170 (kg dry matter yield. cm H2O)Rmin, Rmax 20, 100 (ppm soil N)D50, E50 15, 50 (ppm soil N)Umin, UmaT 0.5, 4.0 (kg N . kg dry matter1)‘FLX 30 (ppm soil N)ClimateAnnual total PPT 156.0 (1991), 164.4 (1992) centimetresAnnual total PET 49.7 (1991), 54.4 (1992) centimetresAppendix D. Model Input Values 158Table D.2: Model input values, Sumas 1992.File Variable Value(s)Soil Owp 0.11°FC1, °FC2 0.29, 0.330.42°ck, Ck 0.0, 11.26S 0.02Pb 1.33 (g . cm3)z1, 20 (cm)H5 O.2(%)M5 0.00007 (d’)11.0Manure X 0.70L 0.90L3 0.67L3 0.95, 0.9, 0.87, 0.83, 0.8(Jan., Feb., Mar., Apr., May-Dec.)N0 0.61N, 0.39Mm 0.002 (d1)Hm 0.4FN 12.52 (kg . wk’)Crop H1, HF 20, 130 (cm)HL 50 (cm)g1, g2 10, 70 (cm 1120 transpired)P 170 (kg dry matter yield cm H20’)Rmin, Rmax 20, 100 (ppm soil N)D50, E50 15, 50 (ppm soil N)Umin, Umac 0.5, 4.0 (kg N . kg dry matter’)‘FLX 30 (ppm soil N)ClimateAnnual total PPT 146.0 centimetresAnnual total PET 53.9 centimetres

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:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0087544/manifest

Comment

Related Items