Development of a Model to Predict Spray Deposition in Air-Carrier Sprayer Applications

MISSING IMAGE

Material Information

Title:
Development of a Model to Predict Spray Deposition in Air-Carrier Sprayer Applications
Physical Description:
1 online resource (191 p.)
Language:
english
Creator:
Larbi,Peter A
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Agricultural and Biological Engineering
Committee Chair:
Salyani, Masoud
Committee Members:
Christman, Mary C
Ehsani, M Reza
Kiker, Gregory
Beck, Howard W
Ingley, Herbert A

Subjects

Subjects / Keywords:
airblast -- analysis -- canopy -- citrus -- citrussprayex -- compartment -- deposition -- dispersion -- drift -- evaluation -- expert -- ground -- model -- pesticide -- prediction -- sensitivity -- simulation -- spray -- sprayer -- system -- validation
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre:
Agricultural and Biological Engineering thesis, Ph.D.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
A simulation model was developed to predict on-target spray deposition when using an airblast sprayer. It was developed as a compartment model where, at a discrete location in the sprayer?s travel direction, the plume or spray cloud is handled as passing through several connected compartments of equal thickness but increasing cross section, in the direction of application. With this approach, the spaces between the sprayer and tree and within the canopy were divided into small elements to achieve the needed simplification for simulation purposes. The model, which accounts for evaporation, drift, and ground deposition, simulates spray mass dispersion, assuming no slip between spray droplets and airstream and no contribution to sprayer air velocity from spray droplets. The tree canopy equations account for foliage distribution within a canopy in the direction of spray application, which simultaneously represents resistance to spray transport resulting in deposition. With the incorporation of maximum deposition, it is possible to account for spray runoff from leaves. Two field experiments were conducted to validate the model in two parts, namely: dispersion and deposition. The dispersion experiment setup consisted of a Polyvinyl Chloride pipe structure that provided a grid of five target distances from sprayer outlet and four sampling heights. Absorbent paper targets were used to sample airborne spray from a conventional airblast sprayer. In the deposition experiment, twenty 3-tree plots in an orange grove were sprayed and leaves sampled from the middle tree on two sampling lines at two sampling heights and four canopy depths. Ground samples were also collected. In both experiments, spray treatments consisting of combinations of two nozzles (Albuz ATR Lilac and Albuz ATR Blue nozzles) and two forward speeds (2.4 and 4.8 km/h) were applied using a conventional airblast sprayer. In both experiments, samples were analysed by fluorometry and the total leaf area of each sample from the deposition experiment measured with an area meter. The results showed good agreements between the model output and the experimental data with best predictions in the first case yielding modeling efficiency (EF) and correlation coefficient (r) of 78% and 0.90 for airborne spray and 50% and 0.62 for ground deposits, respectively. The second case yielded EF = 61% and r = 0.92 for canopy deposition, but ground deposition data was overly under-predicted by more than three orders of magnitude and was not considered appropriate. Despite shortcomings, the model has potential for meeting the set objectives, and has been implemented in an expert system to assist spray applicators in making decisions critical for efficient spray operations.
General Note:
In the series University of Florida Digital Collections.
General Note:
Includes vita.
Bibliography:
Includes bibliographical references.
Source of Description:
Description based on online resource; title from PDF title page.
Source of Description:
This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility:
by Peter A Larbi.
Thesis:
Thesis (Ph.D.)--University of Florida, 2011.
Local:
Adviser: Salyani, Masoud.

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2011
System ID:
UFE0043176:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 DEVELOPMENT OF A MODEL TO PREDICT SPRAY DEPOSITION IN AIR CARRIER SPRAY ER APPLICATIONS By PETER AKO LARBI A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREME NTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 201 1

PAGE 2

2 201 1 Peter Ako Larbi

PAGE 3

3 This dissertation is dedicated to my wife Auriette my parents, Emmanuel and Joan, and my brother Daniel for outstanding motivation, love, and support.

PAGE 4

4 ACKNOWLEDGMENTS This dissertation represents not only my efforts, as the author of the work, but also the contribution of a few other individuals whose efforts cannot go unmentioned. These people offered a variety of support cutting across spiritual, physical, emotional, technical, and professional. I am thus indebted to them to acknowledge their parts. Firstly, I thank my God for counting me worthy to be found among the living at this moment of completing my doctoral work. He granted me pea ce when the going was tough and strength to push through. Without His grace, I could not have made it to this point. Secondly, I am very grateful to the University of Florida for granting me an assistantship, without which I could not have started my PhD, and the opportunity to pursue my studies through its outfit. I have really enjoyed the experience an d the value it has added to me. Thirdly I express great thanks to Dr Masoud Salyani, my PhD advisor and advisory committee chair, for the opportunity to w ork with him and learn from him, and for h is patience in guiding me through my work in a professional manner. His critical assessments and queries challenged me to think deeper about the work and better pursue my objectives. Furthermore I acknowledge the efforts of the rest of my advisory committee members Dr. Herbert Ingley, Dr. Howard Beck, Dr. Mary Christman, Dr. Greg Kiker, and Dr. Reza Ehsani in advising me during my consultations with them. Each of their contributions was invaluable and very sign ificant in creating that quest for knowledge and in pursuit of my research objectives. Moreover I would like to thank Mr. and Mrs. Byron Herlong for granting me a Herlong Endowed Scholarship through the College of Agricultural and Life Sciences as an addi tional support for m y studies. I would like to thank my colleagues and friends, Dr. Michael Miyittah, Kofikuma Dzodzi, Ramin Shamshiri, and Ahmed Al Juma i li, for sitting in at my mock orals qualifying exam

PAGE 5

5 critiquing my presentation, and helping me in my preparations. I would also li ke to thank my lab mates, Roy D. Sweeb, Dr. Lav Khot, Guiseppe Vanella, and Ahmed Al Jumaili for maintaining good working environment and tremendous working relationships, and for useful discussions regarding my work. I would l ike to appreciate Dr. James Jones, for providing technical consultation Special thanks to Dr. Reza Ehsani and Dr. Steve Futch extension specialist at the CREC for the opportunity to use their tractor and all terrain vehicle at a critical poin t in my wo rk to complete my experiments My heartfelt gratitude goes to GAPWAY Grove Corp for the opportunity to conduct one of my experiments in their orchard Could I leave my family out, who consistently offered me moral support and stood by me in daily prayers f or my success? Great thanks to my parents, Mr. and Mrs. Larbi, my brothers, Paa Kwasi, Daniel, James, David, and Michael, and all other relatives who shared in my challenges and struggles. Finally, I thank my wife, Auriette, for providing domestic support and company, for her tremendous motivation and prayers, for her love and support, and for sharing with me my deepest joys and sorrows. Indeed, I could not have made it without her

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMEN TS ................................ ................................ ................................ ............... 4 LIST OF TABLES ................................ ................................ ................................ ........................... 9 LIST OF FIGURES ................................ ................................ ................................ ....................... 10 LIST OF SYMBOLS ................................ ................................ ................................ ..................... 14 ABSTRACT ................................ ................................ ................................ ................................ ... 18 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .................. 20 Background ................................ ................................ ................................ ............................. 20 Problem Statement ................................ ................................ ................................ .................. 21 Objectives ................................ ................................ ................................ ............................... 24 2 LITERATURE REVIEW ................................ ................................ ................................ ....... 25 Spray Application Planning ................................ ................................ ................................ .... 25 Sprayer Calibration ................................ ................................ ................................ ................. 26 The Spray Application System ................................ ................................ ............................... 28 Factors Affecting Spray Efficiency ................................ ................................ ................. 30 Sprayer Design and Application Parameters ................................ ................................ ... 30 Spray application rate ................................ ................................ ............................... 32 Air volume rate ................................ ................................ ................................ ......... 32 Air velocity ................................ ................................ ................................ ............... 33 Ground speed ................................ ................................ ................................ ............ 34 Spray Physical Properties ................................ ................................ ................................ 34 Weather Conditions ................................ ................................ ................................ ......... 36 Wind speed and atmospheric stability ................................ ................................ ...... 36 Air temperature and relative humidity ................................ ................................ ..... 37 Solar radiation and rainfall ................................ ................................ ....................... 37 Tree Canopy Characteristics ................................ ................................ ............................ 38 Target Location ................................ ................................ ................................ ............... 39 Field Asses sment of Spray Deposition and Drift ................................ ................................ ... 40 Tracers ................................ ................................ ................................ ............................. 41 Spray Deposit Sampling ................................ ................................ ................................ .. 41 Laboratory Measurements and Analyses ................................ ................................ ......... 42 Efficient spray Application /Loss Mitigation Techniques ................................ ...................... 43 Modeling Spray Di spersion and Deposition ................................ ................................ ........... 44 Significant Spray Modeling Contributions ................................ ................................ ............. 44 Plume Models ................................ ................................ ................................ .................. 45

PAGE 7

7 Wake Models ................................ ................................ ................................ ................... 46 Random Walk Models ................................ ................................ ................................ ..... 48 Turbulent Jet Flow Models ................................ ................................ .............................. 49 Other Important Models ................................ ................................ ................................ .. 49 Meteorological Measurements in Model Evaluation ................................ .............................. 51 Review of Selected Modelin g Approach ................................ ................................ ................ 52 Review of Some Expert Systems ................................ ................................ ............................ 55 3 MODEL DEVELOPMENT AND SIMULATION STUDY ................................ .................. 57 System Description ................................ ................................ ................................ ................. 59 Plume Geometry ................................ ................................ ................................ ..................... 60 Assumptions ................................ ................................ ................................ .................... 61 Spray Mass Compartments ................................ ................................ .............................. 61 Air/Spray Velocity ................................ ................................ ................................ ........... 61 Droplet Size Distribution ................................ ................................ ................................ 62 Spray Dispersion ................................ ................................ ................................ ............. 63 Spray mass inflow/outflow ................................ ................................ ....................... 64 Droplet evaporation ................................ ................................ ................................ .. 66 Potential spray drift ................................ ................................ ................................ .. 67 Ground deposition ................................ ................................ ................................ .... 68 Canopy deposition ................................ ................................ ................................ .... 68 Model Simulation ................................ ................................ ................................ ................... 7 0 Consideration for Time Step and Compartment Thickness ................................ ............ 70 Model sc ale ................................ ................................ ................................ .............. 71 Sensitivity to temporal and spatial difference ................................ .......................... 71 Sampling time ................................ ................................ ................................ .......... 72 Optimum time step ................................ ................................ ................................ ... 72 Approach Used in Solving Droplet Evaporation ................................ ............................. 73 4 MODEL VALIDATION ................................ ................................ ................................ ........ 84 Methods ................................ ................................ ................................ ................................ .. 85 Dispersion Experiment ................................ ................................ ................................ .... 85 Fluorometric analysis of samples ................................ ................................ ............. 86 Data analysis ................................ ................................ ................................ ............ 87 Prediction of dispersion data ................................ ................................ .................... 87 Deposition Experiment ................................ ................................ ................................ .... 88 Fluorometric analysis of samples ................................ ................................ ............. 89 Leaf area measurements ................................ ................................ ........................... 89 Data analy sis ................................ ................................ ................................ ............ 90 Prediction of deposition data ................................ ................................ .................... 90 Results and Discussion ................................ ................................ ................................ ........... 91 Dispersion Experiment ................................ ................................ ................................ .... 91 Comparison of Results with Model Predictions ................................ .............................. 92 Deposition Experiment ................................ ................................ ................................ .... 93 Comparison of Results with Model Predictions ................................ .............................. 94

PAGE 8

8 5 SENSITIVITY STUDY FOR DECISION SUPPORT ................................ ......................... 116 Methods ................................ ................................ ................................ ................................ 117 Global Sensitivity Analysis ................................ ................................ ........................... 117 Local Sensitivity Analysis ................................ ................................ ............................. 118 Results and Discussion ................................ ................................ ................................ ......... 118 Global Sensitivity Analysis ................................ ................................ ........................... 118 Local Sensitivity Analysis ................................ ................................ ............................. 120 Relationship between pressure and flow rate ................................ ......................... 120 Relationship between flow rate and volume median diameter .............................. 120 Relationship between airflow rate and air velocity ................................ ................ 121 Relationship between air velocity, volume median diameter and airborne spray mass ................................ ................................ ................................ .................... 121 Relationship between air velocity, volume median diameter, and ground deposit ................................ ................................ ................................ ................. 122 Relationship between air velocity, volume median diameter, and Canopy Deposit ion ................................ ................................ ................................ ........... 122 6 EXPERT SYSTEM DESIGN AND DEVELOPMENT ................................ ....................... 131 Development of the Expert System ................................ ................................ ...................... 132 General Structure ................................ ................................ ................................ ........... 132 Detailed Description of Expert System ................................ ................................ ......... 133 Spray calibration ................................ ................................ ................................ .... 134 Spray timing ................................ ................................ ................................ ........... 136 Spray implementation ................................ ................................ ............................ 137 Evaluation of Expert System ................................ ................................ ................................ 138 Results and Discussion ................................ ................................ ................................ ......... 140 Test Run ................................ ................................ ................................ ......................... 140 Evaluation ................................ ................................ ................................ ...................... 140 7 CONCLUSIONS ................................ ................................ ................................ .................. 171 APPENDIX A CHAPMAN RICHARDS PARAMETERS ................................ ................................ ......... 173 B SPRAYER AIR VELOCI TY MEASUREMENTS ................................ .............................. 176 C DEFINITIONS OF FLOWCHART SYMBOLS ................................ ................................ .. 178 LIST OF REFERENCES ................................ ................................ ................................ ............. 179 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ....... 190

PAGE 9

9 LIST OF TABLES Table page 2 1 Classification of sprays. ................................ ................................ ................................ ..... 56 4 1 Summary of spray treatments used in experiments. ................................ .......................... 95 4 2 Weather parameters recorded for various spray treatments in the dispersion experiment. ................................ ................................ ................................ ......................... 96 4 3 Actual sprayer speeds for various spray treatments in the dispersion experiment measured over 30.5 m. ................................ ................................ ................................ ....... 96 4 4 Weather parameters recorded for various spray treatments in the deposition experiment. ................................ ................................ ................................ ......................... 96 4 5 Actual sprayer speeds for various spray treatments in the deposition experiment measured over 15.2 m. ................................ ................................ ................................ ....... 96 4 6 Prior and posterior distributions for parameters estimated with the GLUE method. ........ 97 4 7 Measures of agreement between measured and predicted values. ................................ .... 97 5 1 Global sensitivity analysis factor levels. ................................ ................................ .......... 123 5 2 Regression coefficients for fan and cone nozzles where VMD = A ( FR B ). ...................... 123 5 3 Intercept, of polynomial relating airborne spray mass to droplet size for different air velocities. ................................ ................................ ................................ .................... 123 5 4 Intercept, of polynomial relating ground spray mass to droplet size for different air velocities. ................................ ................................ ................................ .................... 123 5 5 Coefficients of polynomial relating air velocity, total flow rate per side and dimensionless spray mass deposit. ................................ ................................ ................... 124 6 1 Summary of evaluation feedback categories. ................................ ................................ .. 142 6 2 Effect of volume application rate with medium foliage tree canopies shown by ES. ..... 142 6 3 Effect of foliage density with medium foliage tree canopies shown by ES. ................... 143 6 4 Effect of ground speed with medium foliage tree canopies shown by ES. ...................... 143 6 5 Effect of missing trees with medium foliage tree canopies shown by ES. ...................... 143 A 1 Summary results of Chapman Richards slope and inflexion parameter estimates. ......... 174

PAGE 10

10 LIST OF FIGURES Figure page 3 1 Spray model system diagram. ................................ ................................ ............................ 74 3 2 Spray model Forrester diagram ................................ ................................ .......................... 75 3 3 Schematic of a spray cloud from a rectangular cross section (stationary) source (adapted from Beychok, 2005) ................................ ................................ .......................... 76 3 4 Three dimensional view of imaginary compartments ................................ ....................... 76 3 5 Spray partition in transit. A) Schematic of spray transport from one compartment to another. B) Free body spray mass balance of a compartment outside tree canopy. C) Free body spray mass balance of a compartment inside tree c anopy. ............................... 77 3 6 Simulated airborne spray mass with distance using different time steps at various compartment thicknesses. ................................ ................................ ................................ .. 77 3 7 Simulated airborne spray mass with distance using different compartment thicknesses at various time steps. ................................ ................................ ....................... 78 3 8 Simulated airborne spray mass with time using different compartment thicknesse s at various time steps. ................................ ................................ ................................ .............. 79 3 9 Simulated airborne spray mass with time using different time steps at various compartment thicknesses. ................................ ................................ ................................ .. 80 3 10 Trend of maximum sampling time for different compartment thicknesses. ...................... 81 3 11 Relationship between sprayer air velocity, compartment thickness, and simulation time step. ................................ ................................ ................................ ............................ 81 3 12 Simulated airborne spray mass with time using optimum time step/compartment thickness combinations. ................................ ................................ ................................ ..... 82 3 13 Simulated spray mas s inflow (cumulative over time) with distance using optimum time step/compartment thickness combinations. ................................ ............................... 82 3 14 A schematic of approach used to solve temporal spatial evaporation. .............................. 83 3 15 Plot of evaporating droplet size over time. ................................ ................................ ........ 83 4 1 Field setup for dispersion experiment. ................................ ................................ ............... 98 4 2 Plot of concentration against Turner 111 fluorescence at 3X. ................................ ........... 99 4 3 Experimental setup for the deposition experiment. ................................ ......................... 100

PAGE 11

11 4 4 Plot of concentration against Turner 111 fluorescence at 1X. ................................ ......... 101 4 5 Airborne and ground dye deposit from different spray treatments in Dispersion Experiment at di fferent target distances from sprayer outlet. ................................ .......... 102 4 6 Airborne dye deposit for different spray treatments in Dispersion Experiment at different sampling heights. ................................ ................................ ............................... 103 4 7 Mean deposit from different spray treatments in Dispersion Experiment. ...................... 104 4 8 Mean deposit from different nozzles and speeds in the Dispersion Experi ment. ............ 105 4 9 Measured and predicted dye deposit at different heights for all spray treatments in the dispersion experiment. ................................ ................................ ............................... 106 4 10 Predicted dye deposit and residuals from the first and second simulation setup versus measured values. ................................ ................................ ................................ .............. 107 4 11 Measured and predicted ground dye deposit for all spray treatments in the Dispersion Experiment at different target distance from sprayer outlet. ................................ ........... 108 4 12 Predicted ground deposit and residuals from the first and second simulation setup versus measured values. ................................ ................................ ................................ ... 109 4 13 Canopy and ground dye deposition from different spray treatments in Deposition Experiment at different canopy depths. ................................ ................................ ........... 110 4 14 Canopy dye deposition for different spray treatments in Deposition Experiment at different sampling heights. ................................ ................................ ............................... 111 4 15 Mean deposition from different spray treatments in Deposition Experimen t. ................. 11 2 4 16 Mean deposit from different nozzles and speeds in the Deposition Experiment. ............ 113 4 17 Measured and predicted dye dep osition versus distance at two heights from all spray treatments in the deposition experiment. ................................ ................................ ......... 114 4 18 Predicted dye deposition and residuals for the deposition experiment versus measured val ues. ................................ ................................ ................................ .............. 115 5 1 Matrix of two factor interaction graphs of input factors. In each graph, the first factor levels are indicated by the horizontal axis, the second factor levels by the legend of ea ch row, and the response by the vertical axis. ................................ .............................. 125 5 2 Factorial sensitivity indices (contributions) of significant main factors and interactions based on a 35 complete factorial design and it s analysis of variance. ......... 126 5 3 Relationship between nozzle operating pressure and flow rate. ................................ ...... 126

PAGE 12

12 5 4 Relationship between air v elocity and regression coefficients A (top) and B (bottom) for fan nozzles. ................................ ................................ ................................ ................. 127 5 5 Relationship between air velocity and regression coefficients A (top) and B (bottom) for cone nozzles. ................................ ................................ ................................ .............. 128 5 6 Simulated airborne spray mass for different droplet sized sprays. ................................ .. 129 5 7 Relationship between droplet size and airborne spra y mass. ................................ ........... 129 5 8 Relationship between droplet size and ground spray deposit. ................................ ......... 130 5 9 Simulated spray deposition in an assumed can opy with uniform medium density foliage obtained with different droplet sizes. ................................ ................................ ... 130 6 1 Structure of ES with arrows showing the direction of information flow. ........................ 144 6 2 Detailed structure of the expert system. ................................ ................................ ........... 144 6 3 Base flowchart of the ES. ................................ ................................ ................................ 145 6 4 GUI for sprayer speed calculation. ................................ ................................ .................. 146 6 5 Flowchart for sprayer speed calculation. ................................ ................................ ......... 147 6 6 GUI for nozzle selection. ................................ ................................ ................................ 148 6 7 Flowchart for nozzle selection. ................................ ................................ ........................ 149 6 8 Flowchart for nozzle selection (cont.). ................................ ................................ ............ 150 6 9 Flowchart for nozzle selection sub function. ................................ ................................ ... 151 6 10 Flowchart for nozzle comparison. ................................ ................................ .................... 152 6 11 Flowchart for Flo w Rate Sub. ................................ ................................ .......................... 153 6 12 GUI for sprayer output calculation. ................................ ................................ ................. 154 6 13 Flowchart for sprayer output calculation. ................................ ................................ ........ 155 6 14 GUI for Spray Timing. ................................ ................................ ................................ ..... 156 6 15 Flowchart for Spray Timing. ................................ ................................ ............................ 157 6 16 Decisio n structure for spray timing. ................................ ................................ ................. 158 6 17 GUI for Dispersion and Deposition simulations. ................................ ............................. 159

PAGE 13

13 6 18 Flowchart for Dispersion and Depo sition simulations. ................................ .................... 160 6 19 Flowchart for Dispersion and Deposition simulation ( cont .). ................................ .......... 161 6 20 Flowchart for Dispersion and D eposition simulation (cont.). ................................ .......... 162 6 21 ............... 163 6 22 Resp ondent agreement with ES answer to spray timing scenario. ................................ .. 163 6 23 Evaluation of program content of CitrusSprayEx. ................................ ........................... 164 6 24 Eval uation of presentation of CitrusSprayEx. ................................ ................................ 165 6 25 Evaluation of effectiveness of CitrusSprayEx. ................................ ................................ 166 6 26 Evaluation of user appe al and suitability of CitrusSprayEx. ................................ ........... 167 6 27 Evaluation of program response of CitrusSprayEx. ................................ ........................ 168 6 28 valuation of ease of use of CitrusSprayEx. ................................ ................................ ...... 169 6 29 Evaluation of user interface and media quality of CitrusSprayEx. ................................ .. 170 A 1 Relationships of 1 and 2 withVMD. ................................ ................................ .............. 174 A 2 Droplet size sprectra and error using relationships of 1 and 2 withVMD. .................... 175 B 1 Sampling grid on sp rayer air outlet ................................ ................................ .................. 176 B 2 Mean air velocity at sample grid points ................................ ................................ ........... 177

PAGE 14

14 LIST OF SYMBOLS height of compartment [m] cross sectional area of compartment [m 2 ] width of compartment [m] retention capacity of leaves [m 3 /m 2 ] concentration [mg/kg or parts per million (ppm)] drag coeffi cient of canopy rate of change [s 1 ] droplet diameter [m] mean droplet diameter of a droplet size class [m] diffusion coefficient of vapor in air [m 2 /s] fraction of total liquid volume contained in droplets of diameter less than a certain diameter fraction of total liquid volume contained in droplets of diameter less than a certain diameter rate function of mass flow [kg/ s] fraction of total liquid volume contained in droplet diameter class leaf area density vertical structural coefficient height of air outlet above ground level [m] height of lower boundary of spray cloud in compartment k above ground level [m] height of tree [m] mass of spray in compartment [kg] molecular mass of air [g/mol] mean droplet mass in droplet size class [kg] molecular mass of vapor diffusing from droplet [g/mol]

PAGE 15

15 number of droplet size classes number of droplets number of drop lets in droplet size class air pressure [N/m 2 ] radius of tree canopy [m] Reynolds number row spacing [m] ground speed of sprayer [m/s] Schmidt number width of sprayer [m] time [s] delay time [s] dry bulb temperature [ C] wet bulb temperature [ C] mean air velocity [m/s] instantaneous droplet velocity in the y direction [m/s] instantaneous droplet velocity in the downward direction [m/s] volume of compartment [m 3 ] mean droplet volume in droplet size class [m 3 ] liquid volumetric flow rate [m 3 /s] Volume Median Diameter [m] horizontal distance from the sprayer outlet to a given poin t in the direction of spray application[m]

PAGE 16

16 horizontal distance from the canopy boundary to a point within the canopy in the direction of spray application [m] horizontal distance in the forward direction of spr ayer [m] horizontal distance in the forward direction of sprayer inside the canopy [m] vertical distance upward from ground level [m] maximum leaf area density [m 2 /m 3 ] leaf area density range [m 2 /m 3 ] leaf area density [m 2 /m 3 ] asymptote of the Chapman Richards function slope of the Chapman Richards function inflection parameter of the Chapman Richards function vapor pressure difference [N/m 2 ] depth of compartment before tree canopy [m] depth of compartment inside tree canopy [m] top angle of spray cloud to the horizontal [] bottom angle of spray cloud to the horizontal [] dynamic viscosity of spray liquid [kg/m s] density of air [kg/m 3 ] density of liquid [kg/m 3 ] fraction of exposed leaves fraction of spray compartment within tree canopy boundary Subscripts index of initial compartment

PAGE 17

17 droplet size class on target deposition drift evaporation ground deposition inflow compartment index outflow spray runoff outside tree canopy inside tree canopy

PAGE 18

18 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requiremen ts for the Degree of Doctor of Philosophy DEVELOPMENT OF A MODEL TO PREDICT SPRAY DEPOSITION IN AIR CARRIER SPRAY ER APPLICATIONS By Peter Ako Larbi August 201 1 Chair: Masoud Salyani Major: Agricultural and Biological Engineering A simulation model wa s developed to predict on target spray deposition when using an air blast sprayer It was developed as a compartment model where, at a discrete location in the connec ted compartments of equal thickness but increasing cross section, in the direction of application. With this approach, the space s between the sprayer and tree and within the canopy were divided into small elements to achieve the needed simplification for s imulation purposes. The model which accounts for evaporation, drift, and ground deposition, simulates spray mass dispersion, assuming no slip between spray droplets and airstream and no contribution to sprayer air velocity from spray droplets. The tree ca nopy equations account for foliage distribution within a canopy i n the direction of spray application, whic h simultaneously represents resistance to spray transport resulting in deposition. With the incorporation of ma ximum deposition, it is possible to ac count for spray runoff from leaves. Two field experiments were conducted to validate the model in two parts, namely: dispersion and deposition. The dispersion experiment setup consisted of a Polyvinyl Chloride pipe structure that provided a grid of five ta rget distances from sprayer outlet and four sampling heights Absorbent paper targets were used t o sample

PAGE 19

19 airborne spray from a conventional airblast sprayer. In the deposition experiment, t wenty 3 tree plots in an orange grove were sprayed and leaves samp led from the middle tree on two sampling lines at two sampling heights and four canopy depths Ground samples were also collected. In both experiments, spray treatments consisting of combinations of two nozzles (Albuz ATR Lilac and Albuz ATR Blue nozzles ) and two forward speeds (2.4 and 4.8 km/h) were applied using a conventional airblast sprayer. In both experiments, samples were analysed by fluorometry and the total leaf area of each sample from the deposition experiment measured with an area meter. The r esults showed good agreements between the model output and the experimental data with best predictions in the first case yielding modeling efficiency ( EF ) and correlation coefficient ( r ) of 78% and 0.90 for airborne spray and 50 % and 0.6 2 for ground deposi ts, respectively. The second case yielded EF = 61% and r = 0.92 for canopy deposition but ground deposition data was overly under predicted by more than three orders of magnitude and was not considered appropriate. Despite shortcomings the model has pote ntial for meeting the set objectives, and has been implemented in an expert system to assist spray applicators in making decisions critical for efficient spray operations

PAGE 20

20 CHAPTER 1 INTRODUCTION Background A erial and ground sprayers are both used in trea ting disease and pest infestations in tree crops. Aerial sprayers have been mostly used in forest applications where there are no clear rows, although they are also used in row crops. Ground sprayers have been used mostly for orchard situations where the t rees are planted in rows and there is ample space to drive a trailed or mounted sprayer through. In a survey conducted (Summerhill et al., 1989) among 312 commercial citrus grove owners and managers with 4 hectares or more orchard size from 21 major citrus producing counties in Florida, 90% sprayed for pest. Out of this, 98% used ground sprayers. The remaining used aerial sprayers 11% for fixed wing aircrafts and 4% for helicopters. The situation has not changed much over the years although several advan cement s in spray equipment have taken place This indicates that ground spraying is the most employed mode of pest management in Florida citrus production making ground sprayers of critical importance to the citrus industry. For citrus applications, air c arrier ground sprayers (mostly air blast sprayers) have played, and continue to play, a critical role ( Futch and Atwood, 2009 ) Air carrier spraying involves the atomization and transport of spray droplets with the aid of an airstream usually high volume high velocity airstream towards an intended target(s). It may be categorized into two different types, namely air assist and air blast spraying, according to the manner in which atomization is achieved. In air assist atomization, high velocity air is us ed to enhance pressure atomization at low liquid flow rates. In air blast atomization, a liquid jet or sheet is exposed to air flowing at high velocity which breaks up the liquid into droplets and carries them along. A major difference

PAGE 21

21 is that air assist u ses relatively small quantities of air at high velocities whereas air blast uses large amounts of air at much lower velocities (<100 m/s) While pesticide use in crop management remains a necessary aspect of Florida citrus production (Salyani, 1994 b ), spra y applications continue to suffer from various sources of inefficiency leading to off target loss of applied pesticides. In citrus and other tree crop applications, the target of interest is the tree canopy. Effectiveness and efficiency in spray technology for agrochemical applications has been an issue for many years owing to the fact that, time and again, a large amount of the spray material misses the target and is lost. But the prime objective of pesticide spray application is to get the spray material to the target pest area, and the more the material that gets to the target the better the chances of curbing the pest disease situation. Greater efficiencies need to be aspired in all spray applications as pesticide lost in the air by spray drift or to the ground as ground deposit impacts the enviro nment and health significantly. It is important to know the right volume of pesticide to be applied per area under specific conditions. Pesticide labels recommend volume per area for applied spray which is often referred to as dilute spraying (Pfeiffer, 2002). Failure to apply the right amount of water and spray material results in applying either inadequate or excessive dosage, each of which is not desired. Under dosing implies that the full power of the chemical is not utilized and the pest situation is not brought under control. On the other hand, over dosing implies that excess spray material than required is applied, resulting in waste. This situation can be improved by effectively planning each spray applicat ion to the extent that losses can be minimized. Problem Statement Spray application off target losses are a great concern to Florida citrus. A r ecent stud y (Salyani et al., 2007) show that about 17.9 % to 25.7% of the total spray volume applied to citrus

PAGE 22

22 ca n be lost in air carrier applications Although this is an improvement over several years of research and development it is still a big problem, converting to loss of several dollars. This adds to the cost of labor provided by the spray applicator, fuel to run spray equipment, maintenance of spray equipment, including downtime, and overhead. Pesticide lost during an air carrier spray application ends up in the environment. Direct ground deposition and indirect ground deposition through runoff of the pestici de from leaf surfaces increase pesticide burden on the soil. Drift deposit on the ground far beyond the target may also add to ground deposit. Apart from potentially affecting microorganisms in the soil, rain runoff may carry pesticide deposits into surfac e water bodies where they are not needed and pollute them. This may affect aquatic life to some extent and cause death in extreme cases. Potentially, pesticide deposit may also seep down the soil and cause underground water pollution. Added to these is the problem of volatilization of on target deposit, chemical and microbiological degradation in causing dissipation of pesticide available for pest control (Wolters et al 2004). Pesticide spray application losses may become hazardous to human health. Most p esticides are toxic chemical formulations that when lost through drift poses health risks to both humans and animals. Pesticide lost by drift may be inhaled directly by bystanders without protective gear Drift losses may also be carried long distances to neighboring human settlements creating air pollution. The foregoing discussion and more, indicate that the effect of pesticide spray application losses go beyond just the application site and affect people who are not even directly involved with the activi ty. Policy makers have expressed great concern and the Environmental Protection Agency (EPA) has guidelines to control the misapplication of pesticides. Buffer zones have been

PAGE 23

23 established to prevent spray application to areas within a certain radius around aquifers and surface water bodies (Huitink et al. 1990 ; Miller and Salyani, 2006) However, these efforts are unable to eliminate pesticide spray losses wi th their corresponding effects. There is the advocacy for adopt ing biological control and other non chemical management practices and to minimize or eliminate the use of chemicals However, these other practices cannot alone fight the battle against pest s in citrus because they are relatively less available, slow acting, and not as efficient as chemical control and thus Florida citrus necessarily will continue to depend on pesticide spray application in order to survive. In this case, there is the need to improve the application efficiency and minimize the losses that occur. Researchers have made effort s to measure off target losses and identify their causes via mathematical procedures, experiments in wind tunnels, and field tests (Parkin and Wheeler, 1996 ; Miller et al., 2003 ) Moreover several spray professionals have been working over the years to de velop systems that would improve delivery of the spray material to the heart of the plant canopies (Pai et al. 2009) Yet, losses still occur in spray applications and the implications regarding them remain of great concern (Hernndez Hernndez et al., 2 007) Thus more remains to be done to further minimize these losses. Experiments to study the factors influencing on target deposition, with corresponding losses, and their interactions have succeeded in increasing understanding about the spray application system (Miller et al., 2003) However, limitations of cost, equipment, and experimental design make it difficult to assess every scenario and spray operation planning continues to be done with much uncertainty about resulting deposition. A further step be yond quantifying deposition and identifying the influential factors would be to develop systems that will aid in predicting deposition based on previous experimental data. It is hypothesized that a

PAGE 24

2 4 spray deposition model can adequately predict on target de position from an air blast sprayer in citrus applications. Computer model simulations serve a complementary purpose to actual field experiments and may guide research as to what is or is not important to measure. S everal aerial and ground boom spray models have been developed and are in use, but these are not easily adaptable to air blast spray application to citrus trees because of difference in their phenomena. Computer models may also assist decision making at planning stage for a spray operation or real t ime decision making during an operation and thus help minimize application errors and off target losses. Decision support systems have been developed and are successfully in use in different areas of agriculture, but these tools are developed specific to a problem domain and therefore not adaptable to citrus air carrier spray application. It is thus also hypothesized that a model based decision support system for spray applications can help the grower to minimize appli cation errors and spray losses. Object ives Based on the above, the new thinking is an attempt to model the interactions between the factors influencing spray efficiency with a goal of predicting deposition before an actual spraying operation is carried out. Thus the main objectives of this res earch were: 1. To develop a computer simulation model to predict spray deposition in citrus trees from an air blast sprayer 2. To validate the model by field experiments. 3. To develop a n expert system based on the model to assist citrus growers in planning spray o perations. 4. To evaluate the expert system by end users.

PAGE 25

25 CHAPTER 2 LITERATURE REVIEW Spray Application Planning The need for chemical application to citrus trees remains unsatisfied as long as pests remain a challenge to the Florida citrus industry. In pl anning a spray operation it is important to select the proper combination of sprayer type, application parameters including droplet size, pesticide properties, tree canopy characteristics, and weather conditions. Inability to do so will result in applicat ion errors, and the efficiency of the spray application can be seriously hampered Application errors can stem from erroneous tank mix concentration or erroneous sprayer output per unit area/tree (Salyani 2003 a ) which can be moderated by proper calibrati on. Proper calibration ensure s that the sprayer is performing as expected and can potentially improve the application effectiveness greatly. In a typical citrus orchard, trees are planted in rows. In most cases, the tree distance is selected such as will a ccommodate full grown trees in the row. The row distance is chosen such that a tractor mounted with equipment such as a sprayer, a tree canopy trimmer, or a tree shaker, can drive in between two rows. Small and medium trees tend to be single standing tree s, with some canopies overlapping. Larger trees tend to almost completely overlap, forming hedgerows. Given that all conditions are perfect (an unrealistic situation) it is desirable that the whole amount of spray applied to tree rows deposit on the tree canopies. However, this is not expected knowing that off target losses are inevitable. Response to spray treatment of single standing tree rows is somewhat different from that of hedgerows. Due to spaces between single standing tree canopies, there is hig h chance of spray drift beyond the target row. This is not the case with hedgerows where the canopy row is a continuous wall. Spray capture is better for hedgerows than for single standing tree rows.

PAGE 26

26 Sprayer Calibration Every sprayer needs to be calibrated before it can be used to effectively apply pesticides. Sprayer calibration is often done at the beginning of the spraying season and also before any specific spray application. An effective calibration will achieve the desired application rate (AR) thus avoiding overdose or underdose applications which are undesirable. In this regard, there are established procedures and calculations involved ( Salyani 2003a) which this study considers good practice and a first step in minimizing spray wastage due to mis application. The calibration process helps in the selection of nozzles. Based on the desired AR operating pressure, droplet size, and sprayer compatibility, the spray nozzles can be selected le with every sprayer and it is common not to find a nozzle with the calculated nozzle flow rate at the desired operating pressure. In such a case, a close match is selected and either the operating pressure or t he ground speed is adjusted to give the desi red AR The nozzle arrangement is done so as to match the size, shape, and density of the target canopy. Nozzles and air guide vanes are oriented so as to rightly direct the spray cloud in the target canopy location. Depending on the orchard condition (sma ll trees), some of the nozzles may be shut off or plugged. Due to ground condition resulting in slippage the actual ground speed (GS) of a tractor does not match up with the indicated speed on the speedometer. The weight added due to the trailed sprayer c an contribute much to the already lowered speed, thus the need to calibrate the tractor for GS The measurement of GS is done on a ground surface similar to the orchard condition with the sprayer, half filled, attached to the tractor. The time taken to tra vel a known distance (or pass a number of trees with known tree distance) is recorded and used to divide the known distance. GS measurement could be enhanced by using a speed radar or a GPS (geographical positioning system) unit installed on the tractor.

PAGE 27

27 T o measure the spray liquid flow rate (FR) the entire nozzle discharge is collected for a specified time at the desired operating pressure. The FR i s obtained as the nozzle discharge divided by the spray time Alternatively, the sprayer is run with the ent ire nozzles discharging for a specified length of time, and then the tank filled back to the level at start of spraying. The volume of water filled back into the tank is divided by the spray time to obtain FR Whereas the capacity of spray nozzles is based on water, it is possible to use actual spray mix other than water. In such a case, the FR is multiplied by a correction factor Aside the above stated procedures, the past three decades have seen several developments in sprayer calibration techniques. The se techniques range from electronic systems that generate signals relating to spray liquid FR and AR to devices that show whether or not there is equal FR through multiple outlet lines. But each of the dev ices and techniques has had limitation s In some c ases, ordinary water has been used instead of actual pesticide due to issues of economy, safety, and environmental pollution related to using actual pesticides without any provision correction Also, s ome techniques have relied on electronic circuits where measurement is affected by sensor accuracy and reliability. Furthermore, incorporation of devices in spray equipment has limited the use or trial with older sprayers to assess the utility of such devices. While the above limitations remain, some other tec hniques have proved u seful with one nozzle at a time whereas others are usable only on boom sprayers. In some other cases, the devices have even not been able to determine steady state FR Salyani and Serdynski (1993) describe a calibration device they dev eloped and explain its usefulness for different agricultu ral and industrial applications, and claim an advantage over other calibrators in measuring FR from any type and number of hydraulic nozzles They claim that there is minimum likelihood of breakdown due to the absence of electronic circuitry making

PAGE 28

28 it useful in both developed and developing countries with different levels of agricultural mechanization. The Spray Application System Droplet size may be considered the most important factor in air carrie r spray application. It influences the processes that occur during spray application to citrus tree targets. Consequently, atomization the process of breaking down bulk liquid into multiplicity of tiny droplets of the spray liquid by using a nozzle or at omizer is critical to the overall spray application. D ifferent types and designs of nozzles exist for agricultural purposes. These include, but not limited to full cone and hollow cone nozzles, flat fan nozzles, and air induction nozzles. The size and ge ometry of the nozzle or atomizer used, the physical properties of the dispersed phase, and the physical properties of the continuous phase (surrounding air) are the main factors that affect atomization. The atomization process, considering the type of noz zle or atomizer used, determines the droplet size spectrum of the spray produced. Agricultural sprays have been categorized (ASAE Standard S572) by droplet size into different classes rang ing from aerosols to co a rse (Table 2 1). Droplet sizes used in agric ultural applications are mostly less than 500 m. Most nozzles are color coded according to the classification of the spray they produce. Reichard et al. (1977) measured droplet size distributions of several disc core nozzles used in airblast applications and report ed that different nozzles produced different spectrums. A similar study measured droplet size spectra of fan and hollow cone nozzles and assessed the effects of FR, nozzle orientation, and airspeed on droplet size spectrum (Yates et al., 1985). I n another study, Salyani (1998) characterized the distribution patterns of two rotary atomizers Although the general classifications give an idea of the spray quality, Hoffmann and Kirk (2005) argued out that spray applicators should not rely on only the rather on the overall droplet spectra, and suggested future modification of ASAE Standard S572.

PAGE 29

29 Subsequent to atomization, the entrained spray droplets are transported by a high volume high velocity airst ream towards the target tree canopy. The spray cloud (a passive mixture of spray droplets and air) expands due to drag and decelerates with distance from the source. Due to turbulence and gravity force in the spray cloud, some droplets tend to become detra ined from the spray cloud. These detrained droplets may either fall to the ground by gravity or be drifted away by prevailing winds. Depending on the droplet size and the intensity of the wind, the droplets may either be carried far or near. Droplet lifeti me while aloft depends on the rate of droplet evaporation which begins as soon as the droplet emerges from the nozzle and continues as long as the droplet remains aloft or even after it has settled on the canopy or the ground. Cunningham et al. ( 1962 ) mea sured evaporation loss proportions as high as 0.4 at 11 m from an airblast sprayer. The rate of evaporation i s highly influenced by prevailing weather conditions including air temperature, relative humidity, and wind speed. Generally, a high temperature an d a low relative humidity will result in high droplet evaporation rate. Increasing wind speed will also increase droplet evaporation rate. Since d roplet lifetime decreases with decreasing droplet size larger droplets remain aloft longer than smaller dropl ets. The size of small droplets of water based sprays decreases rapidly due to evaporation (Reichard et al., 1977) mak ing them highly susceptible to the least wind current. Therefore, ideally, spray droplets need to be large enough so as to still reach th e target after some evaporation, yet small enough to be transported by the airstream to give satisfactory coverage on the target. As the spray cloud passes through the target canopy, some of the spray droplets are intercepted by the canopy components (leav es, stems, branches twigs and fruits ) and deposit on them. For the most part, the amount of spray material that deposits on the canopy depends on the

PAGE 30

30 canopy characteristics (Hall et al., 1991) (shape, size, and density). Moderately dense canopies usual ly allow good penetration and give good spray capture Sparse canopies capture less spray and most of the spray droplets end up traveling beyond the actual target canopy. Highly dense canopies provide greater resistance to spray transport and are difficult to penetrate. Inefficient penetration results in spray runoff, an indirect loss to the ground. Beyond the target canopy the droplets that do not deposit on the canopy may be drifted away by the wind, deposit on other canopies, or deposit on the ground (Sa lyani et al., 2007) All the partitions of the spray discharged that do not end up on the target tree canopies are considered to be off target losses. These losses are of great concern to the applicator, the grower, the researcher, regulatory bodies, and p olicymakers. Factors Affecting Spray Efficiency The two main losses resulting from a typical citrus spray application are: (i) losses to the ground directly or indirectly ( due to runoff from leaves) a lso k nown a s ground deposit and; (ii) losses to the air a lso k nown a s drift material. The factors influencing these have been found to include sprayer design and application parameters, spray physical properties, meteorological conditions, and tree canopy characteristics. In reality, spray deposition and drift phenomena are complex. Sprayer Design and Application Parameters Several types and sizes of air carrier ground sprayer designs exist with different levels of utility and performance. The type of s pray equipment influences deposition and drift to a ma rked extent. Traditional airblast sprayers discharge the spray radially, but sprayers with tower configurations discharge it horizontally toward the target tree canopy. Sprayers with tower configuration have shown lower drift potential above canopy than ra dially discharging airblast sprayers (Salyani and Farooq 2004). Most airblast application rates range between 200 and 5000

PAGE 31

31 L/ha (Whitney et al., 198 9 ; Salyani, 2000). This wide range is achieved by different combinations of nozzle type, size, and number, o perating pressures, and ground speeds. Droplet size decreases with increasing flow rate and increasing atomizer rotational speed (Salyani, 1998) which increases the drift potential of the spray. Atomizer rotational speed is synonymous to nozzle pressure. D roplet size is inversely proportional to both, whichever is used. Th e range of application rates can be split into low volume applications w here more concentrated material than recommended by product label is applied at low spray volume rate and high v olume applications where dilute material is applied. In terms of the rate of active ingredient application rate, both low volume and high volume spraying apply the same rate but low volume spraying involves relatively smaller droplet sizes as against lar ger sizes with high volume spraying. In some studies, low volume ground sprayers produced the highest airborne drift ( Salyani and Cromwell 1992a & 1992b ). Salyani (1997) found the higher deposition in low volume application to be due to re duced runoff fro m leaf surface. Considering the field situation, s pray application parameters influence deposition and drift greatly (Kirk and Hoffmann, 2002) These may include, but not limited to, type and number of nozzles, nozzle pressure or atomizer rotational speed, nozzle angle, spray volume rate, air volume rate, air speed, and ground speed. Hobson et al. (1993) studied the effects of nozzle size (and flow rate) on spray drift for 80 and 110 nozzles at operating pressure of 300 kPa. The effects of application par ameters on spray deposition in high and low volume applications have also been studied ( Koo et al. 2000). Salyani (2000 c ) applied spray using different nozzle arrangements, disc core combinations, operating pressures, and ground speeds to establish the combination of these parameters that produces the highest deposition efficiency. He did not find substantial effect of d isc size on deposition, but established significant interaction of disc size

PAGE 32

32 with number of nozzles and ground speed. He concluded that for low volume application s, reducing the number of nozzles and using smaller disc and core sizes can improve deposition better than s p raying at higher ground speeds. Svensson et al (2003) measured the effects of sprayer speed, fan outlet air velocity, a nd fan jet characteristics on the velocity of the spray jet in the center and beyond apple trees. Other application parameters have been studie d and their effects identified. Spray application r ate In the literat ure, spray application rate spray volume rate and spray volume have been used synonymously to refer to the volume of spray liquid applied per ground area. In this document, spray application rate will be used. Spray application rate significantly affects both deposition and off target losses (Hoffmann and Salyani, 1996) Deposition was found to increase with decreasing application rate, and significantly more deposits were observed on the outside of the canopy than at the inner locations ( Salyani 1995 ; Salyani and Hoffmann, 1996a) However, application rate was not found to have significant effect on ground deposition when comparisons were made among five sprayers (Salyani et al. 2007). In fact, Salyani and Farooq (2003) did not find spray volume rate alone to have significant effect on mean deposition but intera ction with air volume rate had. Air v olume r ate Air volume rate has been found to significantly affect both canopy penetration and spray deposition. Studies have been done to determine the effect of air volume rate on penetration and d eposition of spray within citrus tree canopies, and the interaction between air volume rate and spray application rate o n spray coverage (Salyani and Farooq, 2003). Salyani (1995) reported increased variability in mean deposition at higher speeds using hig h air volume sprayers. Salyani and Farooq (2003) also did not find significant interaction of air volume rate and target height on

PAGE 33

33 mean deposition. Airblast sprayers typically discharge constant airflow (Whitney et al., 198 9 ). However, Pai et al. (2009) ad justed airflow with an electro mechanical system to explore the potential for spray loss reduction and concluded that it is possible to reduce airblast spray application losses by reducing the airflow rate. Air v elocity Air velocity and velocity profiles h ave been studied for different spray ers. Salyani and Hoffmann (1996 a) studied air and spray distribution of an air carrier sprayer equipped with an axial flow fan, a diaphragm pump, and 10 hydraulic nozzles per side and differentiated air velocity profile s of a stationary and a moving sprayer. No correlation between air velocity and deposition was observed. Sidahmed (1997) developed a transport model for droplets in fan sprays near the nozzle, and an empirical equation for effective drag coefficient. Using experimental drop size/velocity data from three nozzles for model testing and comparison with an earlier model he found droplet movement to be influenced by constant, but not necessarily equ al, effective drag coefficient. Sidahmed (1999) followed up with an evaluation of the model proposed by Sidahmed (1996) using data from an Aerometrics Phase/Doppler Particle Analyzer that used the light dispersed by spherical particles to simultaneously measure size and velocity. He argued out that air entrainment has an opposing effect to viscous drag on smaller droplets: Whereas drag tends to retard the speed of smaller droplets faster than larger ones, the entrained air on the other hand, tends to accelerate smaller droplets faster than larger ones. He also concluded that neglecting unaccounted not the case for larger droplets. Salyani and Hoffmann (1996a) also established air velocity distribution and deposition relationship.

PAGE 34

34 Ground s peed The effects of sprayer speed have been studied under different settings (Salyani, 1995; Salyani 1999b; Svensson et al, 2003). Salyani (1995) did not find ground speed to have significant effect on deposition from ground speed tests with two engine dri ven air blast sprayers that were equipped with axial flow fans and disc/core nozzles. Although leaf samples from locations near the sprayer showed slight increase in deposition with increasing ground speed, similar trends were not observed at other locatio ns. He concluded that there was no significant effect of ground speed on deposition and thus, high speeds could be combined with high air volume sprayers to reduce spray application time. Also l ow speed applications do not n ecessarily increase deposition. Spray Physical Properties The mean size and size distribution of the droplet are dependent on several variables including liquid properties, nozzle geometry, and operating variables ( Bouse, 1990; Liu, 2000). Spray physical properties play a major role in determining droplet size and distribution. When the injection pressure is high a low ambient pressure and/or a small orifice will produce small sized droplets. Increasing liquid viscosity and/or surface tension hampers the breakup of the liquid. Usually, h igh viscosity results in poor atomization and produces larger droplets. Surface tension tends to resist distortion of the liquid surface. Liquid density also influences spray penetration since spray behavior is determined by the ki netic energy of the liqui d jet. As already stated, droplet size affects drift and evaporation Generally, droplets less than 100 m in diameter have lifetime of less than the time (14 s) required for them to fall 2m at 30 C and 50% RH (Willis and McDowell, 1987). Droplets less th an 80 m in diameter can rapidly lose their initial momentum after leaving the nozzle (Hobson et al., 1993). These spray droplets then become susceptible to diffusion due to local atmospheric turbulence and advection due to

PAGE 35

35 crosswind, and therefore become detrained from the spray cloud. Droplets of 100 m in diameter can vanish in about 20 s at 60% RH (Holterman et al., 1994). Regardless of wind velocity, droplets less than 50 m in diameter almost always drift away. However, negligible drift has been obser ved from 200 m sized droplets even in 10 m/s speed winds (Thompson and Ley, 1983). At the leaf interface, the spray mixture formulation characteristics also influence deposition amount (Brazee et al.; 1991) The chemical nature of pesticide, adjuvants, pH and droplet size affect spray retention by leaves. Apart from the leaves intercepting the droplets, the droplet retention capacity would determine whether droplets will settle on a leaf. Adhesion between liquid and solid molecules and the degree of sprea ding are characterized by the contact angle between spray droplets and leaf surface. Research at the plant interface show s that sometimes droplets bounce off the leaf surface upon impaction. This droplet rebound is a significant limiting factor of effecti ve spray application and retention. Some researchers have attributed this rebound to kinetic energy, surface tension and degree of wetting between the leaves and the droplets. However, it has been reported that droplets < 100 m in diameter are generally a lmost certain to be retained on leaves irrespective of other factors. Another significant limiting factor of effective spray application and retention is runoff from leaf surface. This is also influenced by droplet size Small droplets g i ve better uniformi ty in deposit distribution than larger ones (Salyani, 2003b). Larger and heavier droplets have a greater tendency to runoff than smaller droplets. Larger droplets may roll off the leaf surface depending on the surface characteristics and orientation of the leaf, and also on the spreading ability of the spray mixture. Smaller droplets may not have enough momentum to roll. However there is the tendency for droplets to coalesce into larger droplets. This becomes very serious when the

PAGE 36

36 droplets are already larg e because much larger drops form with enough momentum to roll off. Smaller droplets may not form large enough drops when they coalesce and may not contribute much to runoff. Weather Conditions A discussion on spray deposition and drift can not be complete without mention of weather factors. Although weather factors cannot be controlled, understanding their influence in the system plays a major role in planning and implementing spray operations. Spray deposition and drift are influenced extensively by prevai ling weather conditions at the time of spraying (Hoffmann and Salyani, 1996) The turbulence in the atmosphere affects spray transport; both horizontal and vertical components of wind velocity are crucial in spray drift phenomenon. Thus spraying under unst able weather conditions is not recommended Salyani and Hoffmann (1996b) characterized the influence of weather parameters on spray application and determined how application time, spray volume rate, and their interaction affect deposition. Spraying at dif ferent times (days and nights) and different months, they found both application time and volume to have significant effect on deposition in each month. Deposition changed with changing weather conditions although individual weather parameters were not sig nificant. Gil et al (2005) assessed spray drift during vine pesticide application under several influential conditions. Thompson and Ley (1983) found drift deposits to increase with wind spee d except near the spray source. Wind s peed and a tmospheric s tab ility Miller (1980) has stated emphatically that when wind velocity surpasses 4 m/s (10 mph) it is not advisable to apply aerial treatments because horizontal advection becomes important even for larger droplets. Thompson and Ley (1983) found that drift de posits increased with atmospheric stability but did not find noticeable effects for small evaporating drops. Hobson et

PAGE 37

37 al. (1993) found an approximate linear relationship between spray drift and wind speed. They also established that the rate of drift decr ease s with decreasing wind Hewitt et al. (1997) found a substantial decrease in downwind deposition rates with a decrease from 2.2 to 0.9 m /s in the crosswind speed. For deposition rates in the first 24 meters spray to b (2005) also found drift to be influenced by droplet size and wind speed. Vertical drift for low speed was more significant than horizontal flow. Air t emperature and r elative h umidity Air temperature and rel ative humidity influence the rate of evaporation of droplets and thus affect deposition. Evaporation is accompanied with heat and mass transfer. Thompson and Ley (1983) observed that drift deposit either remained constant or reduced with evaporation up to a few tens of meters distance away. However, they found noticeable increase beyond 100 m. Hoffmann and Salyani (1996) measured higher depositions at nighttime, where there is lower temperature and higher relative humidity, than at daytime, where there is h igher temperature and lower relative humidity. Solar r adiation and r ainfall Research indicates that weather factors that affect the rate of pesticide loss from leaf surfaces also include rain and sunlight. These do not directly affect spray output during a pplication but rather contribute to supplementary loss of spray material after final deposition. Solar radiation affects deposit decay due to overexposure (Salyani and Cromwell, 1992b). Salyani (2003b) found solar radiation to have significant effect on de posit fluorescence reduction ; l onger exposure of deposits significantly reduced the deposit fluorescence. He also found that rainfall significantly affected the amount of remaining deposits on sprayed targets. Here too, droplet size plays a significant rol e

PAGE 38

38 Tree Canopy Characteristics In citrus and other tree crop spray applications, spray penetration through the canopy is of key importance. This is necessary for uniform distribution of the spray material inside the canopy. Since pests can hide anywhere in the canopy, especially under leaves and other cool locations inside the canopy, i t is necessary that the active ingredient in the spray gets deposited uniformly throughout the canopy system, as much as possible, and remain there for some time to control t he situation. However, canopy characteristics can place a great limitation on the application and hamper deposition. Canopy characteristics like shape, size, and density play a major role in spray deposition and drift. Generally, deposition decreases with canopy depth (Salyani and Farooq, 2003) because the mass flux of the spray reduces with the distance from the sprayer outlet However, denser canopies often create greater resistance to airflow and spray penetration and reduce deposition drastically In su ch a case, a lot of runoff may be observed at the canopy boundary nearest to the sprayer. For single trees or tree rows with gaps, the spray cloud may go around the tree and drift beyond the row. As an advantage, trees planted as h edgerows offer a wider ar ea for spray capture and tend to reduce drift than rows with gaps between trees (Salyan i and Cromwell, 1992a & 1992b). In order t o achieve best deposition results there is the need to have adequate penetration into the canopy to allow the spray material to deposit within the canopy. Studies have been done to assess the possibility of penetrating even the densest canopy. Modifications have been made to equipment designs and different application parameter combinations have been studied to detect combinations that give the optimum deposition. Svensson et al (2003) measured the penetration of a sprayer air jet into an apple tree canopy and found fan positions to have significant effect on air velocities within the canopies at 4.2 m elevation. But the effect wa s less significant at the 1.8

PAGE 39

39 m elevation. Characteristics of airflow of converging fan positions alternatives were found to enhance penetration and improve deposition. Flow above the tree canopy was also reduced with airflow from converging air jets. They also found air penetration to improve by high air ou tput power at low travel speed. As the spray droplets penetrate the tree canopy the characteristics of the leaves also play an important role in influencing efficient application. Generally, l esser conta ct ang les result in greater spreading an d water based solutions have lesser contact angles, thus better wetability, but on older leaves than young leaves Also, lower surfaces are usually less wetted than upper surfaces. Target Location Target tree canopy distance has been found to have significant effect on deposition as well. The farther the target distance the smaller the air velocity, and the lesser spray material that reaches it as well as the resulting deposition. Salyani and Cromwell (1992a, 1992b) observed a decrease in deposition with downwind distance. In an aircraft application, Thompson and Ley (1983) also found drift deposit to increase with release height except near source and found a rough proportionality. Salyani and Hoffmann (1996a) found air velocity to reduce speedily with increasing distance from sprayer outlet in both stationary and moving sprayers. However, velocity was comparatively lower for the moving sprayer. Deposition also decreased with increasing distance from the sprayer at al l sampling heights and spray volume rates. Salyani and Farooq (2003) attributed the observed significantly lower deposition at higher sample heights to farther target distances. They found that interaction of air volume rate and target height did not have significant effect on mean deposition. Considering a whole grove spraying condition, Salyani and Cromwell (1992a) have concluded that the most contribution to the overall off target deposits is from the last one or two rows downwind side of the grove.

PAGE 40

40 Fiel d Assessment of S pray D eposition and D rift Drift measurement has been discussed extensively (Miller, 2003). Two ways of collecting drift material are by collecting the material at ground level or by sampling the air carrying the material. An established me thod of quantifying spray drift at ground level is by using horizontal targets (collection surfaces). Examples of horizontal targets include Petri dishes, cut pieces of filter paper, rectangular plastic cards, chromatography paper, and similar targets. Air sampling can be achieved by using high volume air samplers low volume air samplers, cascade impactors, and rotary samplers. Different researchers have used different equipment, combinations of operational variable and application parameters, and differen t sampling methodologies to study spray deposition and/or drift. Basic assessment techniques involve the use of targets to sample the spray. Ebert and Downer (2006) made a monumental review on several of such experiments and concluded that every spray dist ribution experiment needs to measure both retention and biological effects. However, this would mean that all these experiments should use real pesticides, which would not be in the interest of the environment unless the measurements are made during an act ual pesticide application so that the environment is not contaminated unnecessarily. Stoughton et al. (1997) assessed long range spray drift prediction accuracy using lidar measurements. The field experiments were conducted to monitor aerially applied pest icide transport above an oak forest in a near neutral planetary boundary layer. They took micrometeorological data which they used to run the FS C BG (Forest Service Cramer Barry Grim) model and University of Connecticut Spray Transport (UCST) models for com parison. They found both models to predict more than 99.5% of the deposition of the total material complete by 200 m. Both models predicted that droplet diameters greater than 100 m all fallout. However, earlier, Bache and Johnstone (1992) had found that droplets as large as 250 m could

PAGE 41

41 be drifting. The lidar measurements tended to favor the results of Bache and Johnstone (1992). Threadgill and Smith (1975) studied the influence of selected physical and meteorological parameters on drift potential of non pesticidal spray s applied from ground sprayers. Tracers In order to minimize environmental contamination through the use of real pesticides that are toxic, tracer dyes have been used instead in most experiments. A few examples include Salyani (1993, 1999 a 2003 b ), Salyani and Cromwell (1992a, 1992b), Brown and Taher, (1999) and Farooq and Salyani (2002). Different colors including blue, green/yellow, orange, red, and black exist for dyes with detection levels referred to as parts per billion (ppb) this is one part of active dye found in one billion parts of water. Visual detection of fluorescent dyes is possible in the range of 100 ppb. With a black ultra violet (UV) light or a fluorometer, detection in the 10 ppb or <1.0 ppb ranges can be achieved, respec tively ( Lab Safety Supply Inc. 2007 ). However, depending on the specific water conditions, detection level may vary. Spray Deposit Sampling Deposition of spray material on the target canopies is primarily the objective of any agricultural spray operation. To measure deposition, samples are taken from a canopy that have been sprayed or from the ground and the material captured is measured in the laboratory. Both natural (actual leaves) and artificial collectors have been used to capture spray. Sometimes arti ficial collectors have been used where it is impractical to use the actual leaves. Salyani and Hoffmann (1996a) measured spray deposition on leaf and paper towels to compare their efficiencies as samplers. They found that, whereas deposition on the paper t argets increased significantly with increasing air velocity, there was no correlation in the case of the leaves. This indicates the inadequacy of the paper towel sampler s in representing leaf capture.

PAGE 42

42 Farooq and Salyani (2002) captured spray droplets with absorbent cotton ribbon. Gil et al (2005) used PVC to capture spray droplets and shade netting to simulate vine crops. Kromekote cards produced clear stains with rhodamine dye where droplets had evaporated (Brown and Taher, 1999). Mylar sheets were used to quantify deposition from an aerial application (Miller, 1980; Barry et al., 1992; Salyani and Cromwell, 1992a & 1992b). Filter paper may also be used (Salyani and Cromwell, 1992a & 1992b). The time lapse between spray application and collection of the s pray samplers can influence results significantly. Willis and McDowell (1987) have stated that the accuracy in measuring pesticide interception by plants significantly depends on the pace of target collection and pesticide stabilization/extraction. Most dy es are biodegradable and the longer their exposure to solar radiation the less accurate the measured values are likely to be. Salyani and Cromwell (1992b) reported faster decay of fluorescent deposits on Mylar targets than on filter paper and attributed i t to the porosity and dye absorbance of the latter as against the smoothness of the former. Salyani (2000 b ) observed less than 10% degradation in fluorescence of the tracer after it was exposed for one hour to 2.9 MJ/m 2 of solar radiation. To minimize degr adation, Farooq and Salyani (2002) collected samples after 15 to 45 min, that is, as soon as the leaves dried. Brilliant Sulphoflavine (BSF) has been found to reproduce the atmospheric transport of pesticides and could be the most suitable fluorescent trac er (Gil et al., 2005; Gil and Sinfort; 2005). Threadgill and Smith (1975) used Kromekote paper card to collec t droplet number and size data. Laboratory Measurements and Analyses In the laboratory, different methods have been used to measure the tracer depos ition including fluorometry (Salyani and Farooq, 2003; Gil et al 2005), colorimetry (Salyani, 1995), and i n some cases too spectrometry. All these methods have their advantages and shortfalls. Generally, colorimetry is not as fast as fluorometry. Other m ethods include using an optical

PAGE 43

43 comparator ( Ekblad et al. 1987; Brown and Taher, 1999 ) to count dried droplets. Although this method is laborious it avoids problems w ith overlapping droplet stains. Efficient spray Application /Loss Mitigation Techniques A lot of research has gone on over the past years to measure the losses resulting from drift and identify their causes via mathematical procedures, wind tunnel experiments, and field tests. Salyani and Cromwell (1992a & 1992b) conducted field tests (for both aerial and ground sprayers) to quantify spray drift from typical Florida citrus applications and find out the contribution of spraying each row to the overall drift. With the ultimate objective to make air assisted spray applications more efficient variou s aspects of the spray application system have been studied to understand the contributions from different factors. Salyani (1995) identified practices capable of minimizing inefficiencies and environmental contamination. Previous studies have compared the performance of different types of sprayers based on resulting on target deposition, airborne drift, or fallout. Salyani and Cromwell (1992a and 1992b) quantified fallout and airborne drift deposits in aerial and ground sprayers. Salyani (1995) characteriz ed spray deposition and drift from common ground sprayers. Salyani (1997) compared deposition performance of PTO driven with engine driven sprayers. Salyani and Hoffmann (1996) studied air and spray distribution from an air carrier sprayer equipped with an axial flow fan, a diaphragm pump, and 10 hydraulic nozzles per side. For the most part, agricultural spray droplets with sizes above 140 microns initial diameter are less prone to drift (Smith and Burt, 1970). Threadgill and Smith (1975) determined drople t size and specific gravity effects on drift potential. They also determined the effects of horizontal and vertical components of wind speed on drift potential. According to Threadgill and Smith (1975), greater distances (than 400 m ) may let considerable d rifting spray material to evaporate thus escaping sampling They found that drift

PAGE 44

44 potential of sprays decreased with increasing atmospheric stability and increasing droplet size. Drift above 140 microns was not significant for a wide range of atmospheric c onditions. They also found that drift p otential increased with increasing vertical wind component and de creas ing h orizontal wind speed component. Modeling Spray Dispersion and Deposition The cost of conducting experiments over and over again to verify find ings or make new ones is high and the time involved is great. However, there are few to several constraints that prevent researchers from explor ing all possible scenarios. But complementary to experiments have been mathematical and simulation models that h ave been developed based on findings from these experiments. Models serve a better purpose of assessing scenarios through simulation that otherwise are hard to achieve with real experiments. Threadgill and Smith (197 5 ) argue out for an alternative to expen sive and lengthy use of research in evaluating pesticides as being analog, digital, or hybrid simulation. Nevertheless, the success in developing a physical model of a system cannot be achieved without some basic understanding of the physics and dynamics o f the system one is dealing with. Although in some cases one may not fully understand all the complexities that are involved, a good amount of information about the system is tantamount to adequately approximating the behavior of the system. In most cases in science and engineering, a lot of assumptions are necessary to reduce the limitations in order to make some good attempts at describing a provides some information that c an be built up on to understand the system more. Significant Spray Modeling Contributions Modeling and simulation of agricultural spray application systems have been used to understand the dynamics of the system and identify important variables at play in t he system. These have served as basis for further experiments in order to describe the system more precisely

PAGE 45

45 and thus make accurate predictions about the system. However, different researchers have used different approaches at different scales of the syste for one situation or the other Although some authors have used approaches that tend to simplify the system, others have rather delved deeper to account for greater details Plume Models To maximize the benefits associated with the utilization of chemical pesticides, Miller (1980) developed a mathematical model to predict aerial pesticide drift, given experimental data, spray formu lation and nozzle characteristics. He analyzed twenty eight different datasets of deposition observations. He describes a numerical assessment of the spatial distribution of active material both on target and downwind of the target area. Stability effect w as pronounced for smaller droplets. Larger and heavier droplets tend to experience less vertical dispersion. In most agricultural applications the source of the spray cloud is a moving source and at field scale may be viewed as a traveling point source. C alland er and Unsworth (1982) described the design and operation of a traveling point source that they employed to simulate droplet dispersion and deposition in crops. They argued out that the maximum distance downwind of a moving point line source where ob served amount represents an infinite line source depends on the length of the line and also on the standard deviation of the wind direction. Steinke and Yates (1989 a ) modified Gaussian models to improve predictions by including additional assumptions to th e basic Gaussian model assumption. The Gaussian model was observed to o ver predict the downwind concentration up to three orders of magnitude By u sing all available data they could improve prediction s to below one order of magnitude after correction. In addition, the Gaussian model could r eproduc e the shape of the curve well

PAGE 46

46 Wake Models Wake models simulate the behavior of spray as influenced by the wake of an aircraft. A summary of results obtained from a mechanical model is provided by Atias and Weihs (1985). The model was developed to study the behavior of an agricultural airplane wake close to the ground and the movement of sprays under the influence of this wake. Specifically, they studied the distribution of different initial droplet size sprays and evaluated the influence of relative humidity and initial droplet velocity. They concluded that the model could assist in design ing the location of the spraying nozzles along the wing span. Partnership between t he U.S. Department of Agriculture Forest Serv ice and the U.S. Army to develop computer models that predict the deposition and drift of aerially applied spray has been ongoing for several years The se models could p redict the environmental fate of pesticides released from aircrafts, and could account for atomization t hrough settling of the pesticide particles on surfaces. Significant among these is the AGDISP (AGricultural DISPersal) model This model has been validated under diverse settings and utilized extensively (Ekblad et al., 1987; Hoffmann et a l., 2007) Ekblad et al. (1987) conclude d that spray behavior predictions are adequate for drop let s between 100 and 600 m released from a helicopter of speeds greater than the transition speed. Hoffmann et al. (2007) instilled confidence in AG DISP users regarding its accuracy but cautioned regarding its use with canopies with more than 80% enclosure. The FSCBG model is a Gaussian line source model that predicts downwind dispersion by using the near wake results of AGDISP model. It could predict the effects of weather, evaporation, canopy penetration, ground deposition and on canopy deposition. Teske et al. (1990) validated the FSCBG computer simulation model for spray behavior in Douglas Fir seed

PAGE 47

47 orchards. They determined the success of an aeria l spray application system to protect Douglas fir cones and seeds from harmful insects. Mickle (1987) reviewed the strengths and weaknesses of AGDISP, Wiehs Atias, FSCBG, and PKBW (Picot, Kristmanson, Basak Brown, Wallace) models as related to ultra low vo lume (ULV) pesticide spraying and determined how operational parameters affect d eposition patterns near field and drift deposition among the se models. All t hese models are use ful in the predicti on of deposit and drift from aerial application s Since the At ias Weihs model modeled the total mass it was not included in the inter model comparison. They conclude that incorporating AGDISP into FSCBG can fortify FSCBG in terms of its WAKE VORTEX code but note d the extensive increase in computing time that may rule out its usefulness when applied to ULV spray season. Provided the agreement in shape, the Cramer dispersion model could be valuable in evaluating the effects of operational spraying on off target deposits. Also PKBW model effectively modeled the deposit s hape to 400 m but with an over prediction. He states an advantage of PKBW as the potential to directly compare spray cloud movement and dispersion with spray mapping results. Barry et al (1992) describe d the inputs and mode of operation of the FSCBG model and provided summar y of model validation The FSCBG model predicted well following the data downwind until spray cloud dispersion and growth characteristics downwind limited prediction accuracy They concluded that the model could potentially lead to a r educed need for expensive field experiments and the determination of buffer zone and drift potential of pesticides. Parkin and Wheeler (1996) modeled the influence of spray induced vortices on droplet movement in wind tunnels. Having considered airborne sp ray movement downwind of a flat fan hydraulic nozzle to be influenced by trailing wind vortices, t hey studied the mechanism causing

PAGE 48

48 spray displacement, and the dimensions suitable for wind tunnel experiments. They observed trailing vortices in spray clouds in all tests and non occurrence of vertical displacement for airborne spray at wind speeds 2 m / s. They conclude d that small cross section wind tunnels are unsuitable for quantifying drift potential. Random Walk Models Random walk procedure is one approach used to model spray transport of aerially applied spray material where the paths of indi vidual droplets are tracked from release until deposition. Reid (1979) undertook a new look at this procedure as applied to vertical dispersion in the neutral surface layer and compared the integrated cell volumes with numerical solutions f or 500 trajector ies. A nalytical and numerical concentrations were strongly correlated and the vertical distribution variance assumed by Gaussian profiles seemed fairly valid. However, an optimum source height resulting in a maximum standard deviation increased with downwi nd distance. He found a good agreement between his Markov Chain simulation results of vertical dispersion in the neutral surface layer and experimental data. Thompson and Ley (1983) also describe d a random walk model of evaporating drops they used to estim ate spray drift. Hobson et al. (1993) studied the effects on spray drift of nozzle type, angle, and operating pressures for boom mounted hydraulic nozzles operating over a range of meteorological and crop conditions using a random walk approach. To reduce dependence on pesticides or amount u sed with associated emission and drift, Holterman et al. (1994) developed the two dimensional random walk simulation model IDEFICS (IMAG program for Drift Evaluation from Field sprayers by Computer Simulation) This co uld predict spray drift resulting from a crosswind from a conventional boom sprayer using flat fan nozzles. They observed a good agreement between predicted emission to soil and experimental results although there was an overestimation of airborne drift 5 m downwind which

PAGE 49

49 they attributed to difficulty in catching very small droplets. They acknowledged the continued difficulty in quantifying the effects of boom height, forward speed, an d wind velocity on entrainment. Turbulent Jet Flow Models Sprays behave l ike jets because both expand and reduce speed as they move away from the discharge source Abramovich (1963) has dealt extensively with the concepts of spray, jets, and turbulence in his work, which serves a great purpose as a theoretical background for un derstanding turbulent jet theory. The need to understand this turbulence theory is paramount in modeling spray transport, as in transport to target canopy in air assisted spr aying of fruit and tree crops. Sidahmed and Brown ( 2001 ) studied and evaluated a c omputational fluid dynamic code (FLUENT TM ) in simulating the airflow from free round jets and compared the simulation results the two. Delele et al. (2007) d evel oped a CFD model that could predict the droplet dispersion from a cross flow air assisted sprayer account ing for nozzle characteristics and the liquid atomization and v alidated the model predictions with field data. Other Important Models Evaporation rat e is important in spray modeling and most spray models incorporat e models for evaporation. Teske and Hill (1995) used modeling approach to describe the evaporation rates of agricultural spray materials. They considered the assumption that physical properti es of the evaporating spray material primarily behave like water. Moyle et al. (2006) studied and modeled the evaporation of water droplets under laboratory conditions. Miller et al. (1996) adapted a model to predict entrained air velocities to compare wit h experimental results. Using the model they determined entrained air velocities caused by droplet

PAGE 50

50 movement from an agricultural flat fan nozzle through the air in vertical and radial directions inside the spray by measuring the velocities of small tr acer droplets. They used a PMS laser imaging probe to measure the distributions of droplet volume median diameter (VMD) and mean liquid velocity in the spray. Radial air velocity component were observed to de cline with distance from nozzle and angle from t he centerline. Spray pattern varied significantly with distance below nozzle, and the VMD was largest at the spray edges, which was consistent with other evidence that the edges at the spray sheet contain larger droplets. Vieri and Spugnoli (1996) develope d a computer model to control both spray and airstream suitability. The model could predict droplet behavior relat ive to temperature, humidity, and wind variations. They defined working thresholds to avoid dangerous and expensive waste of chemicals and w i th this model they hoped to ensure necessary pesticide deposition on whole canopy sides. It was possible to indicate minimum spray dimensions to avo id evaporation losses. They concluded that it wa s easily transferrable to and implementable on a tractor ins trument to use in spraying operations. Spray modeling at the spray plant interface has been attempted by some researchers. Brazee et al. (1991) modeled the spray impaction on foliage using a partitioned energy method. They noted that the target foliage pro vides significant drawback to effective application and retention of spray material upon impaction on. Giles et al (1991) experimentally characterized the flow field produced by a stationary dual source air nozzle near the nozzle. Using a He Ne laser the y provide d an accurate reference axis for all velocity measurements and determined turbulence intensity for all tests at each sampling point along the flow centerline. They compared their results with mathematical models and previous experimental data and conclude d that mass

PAGE 51

51 flow rate is proportional to the product of centerline velocity and the squ are of the downstream distance. Hewitt et al (1997) provided, as examples, comparisons of possible trends that would be expected if specific variables within th e control of many applications are varied. They reported an effective 48% decrease in deposition rate at 0 m using full swath adjustment. At boom resulted a deposition profile like t hat 26 m An approximate 50% decrease in deposition rate was observed at 15 m downwind from the field edge when the emission height decreased fro m 3 m to 1.5 m At this distance downwind, using boom length of 60% resulted in an approximate 55% decrease in the deposition rate compared to a 75% boom length. A decrease in boom length from 80% to 50% gave a decrease in effective swath width from 22 to 17 m Meteorological Measurements in Model Evaluation Spray models, with real time meteorological data that drive them, can be useful in assisting informed on site spray management decisions during the spraying operation. Miller et al. (1994) provide d info rmation on the meteorological measurements needed to run FSCBG, AGDISP, and VALDRIFT (Valley Drift) models. They indicate d that the weather data necessary to support an experiment strongly depends on the equipment design. VALDRIFT solves the transport, dif fusion, and deposition of an inert cloud emitted from multiple point and/line sources in a valley atmosphere where these sources may be at ground level or elevated and the release rate from each source can vary with time. Thistle et al. (1996) provide d the technical description of VALDRIFT 1.0 and its validation Allwine et al. (1997) describe d the model where the dominant physical process governing atmospheric transport and diffusion are treated explicitly in the model in a highly parameterized fashion.

PAGE 52

52 Xu et al (1997) present ed and validated a stochastic transport model used to analyze the turbulent dispersion of spray droplets from air assisted sprayers. Brown and Taher (1999) soybean plant canopy under conditions of steady wind and sprayer motion. They modeled the process of dynamic transport of droplets from the nozzle to the target, accounting for the effects of sprayer motion, wind, and evaporation. They validated the virtu al nozzle and transport model data from experiment in a canopy under controlled conditions in a wind tunnel. Teske et al (2004 a ) discusse d the use of experimental data in validating a simple model for ground sprayer simulation. Recently, Richardson and Th istle (2006) performed further validation of AGDISP using results from a field study designed to measure spray deposition profiles with in a young radiata pine canopy. Review of Selected Modeling Approach The dynamics of the spray system consist of the comp lex interaction of the processes of atomization or spray droplets formation, spray transport with change in mass flux as a result of expansion of the spray cloud, deposition on the intended target as a result of interception, ground deposition and spray dr ift and deposition beyond the intended target. Deposition on the intended target, for instance a tree canopy, consists of spray penetration, distribution within the canopy, impaction and retention on leaves, deposit formation, and movement on or into the p lant for physiological effect. These components have varying degrees of importance to the entire system and different authors have looked at them either in isolation or as composite. Some authors have worked on aerial application whereas others have dwelle d on ground applications. However, the approaches used to tackle one type of application could have same or similar relevance to the other.

PAGE 53

53 walk model (Reid, 1979) divides the atmosphere of concern into cells, which are the smallest volumes f or which concentration can be resolved. The total time particles spen d by all the particles in each cell is proportional to the concentration in that cell. Practical calculations are done at short time increments with the number of increments all particles spend walk model assumes that trajectories of individual drops can be represented by discrete displacements. These displacements are determined partially by a random selection from appropri ate Gaussian distributions of turbulent air velocity. The model of Miller (1980) describes deposition of spray droplets, which he considers to be from a volume source. In this model he has assumed negligible diffusive transport (passive droplet diffu s ion) no horizontal wind velocity change over height above grade, cloud movement corresponding to local wind velocity and direction, and no evaporation, coagulation, or deformation. The model accounts for the effects of spray droplet size distribution, spray f ormulation, active material proportion in released cloud, spray emission height, and swath width and length. Atias and Weihs (1985) model assumes mono size spherical drops, no collision or other secondary inter drop occurrences, no change in flow field due to spray presence, and no evaporation effect on drag force. The model describes mass transfer through an equation that accounts for diffusion and boiling effects The empirical Gaussian plume model is based on geometrical similarity. This approach has bee n employed for several engineering p urposes with good satisfaction. Milton Hill model (Teske and Hill, 1995) works on some very basic but significant assumptions: that evaporation remains water like when spray materials are added to the tank mix

PAGE 54

54 and that n o matter what is added to the tank mix, as long as the continuous phase is water, the sprayed material will behave like water when it evaporates. The one dimensional model of Miller et al. (1996) accounts for kinetic energy exchange between the droplets an d the air. The model assumes mono dispersed spray and neglects the effect of gravity, which is reasonable where the difference between droplet and air velocity is large and droplet diameter is small. The model also neglects the frictional resistance to air movement caused by viscosity. VALDRIFT considers the atmosphere of concern as consisting of flow tubes aligned along the valley axis. A one dimensional species conservation equation is solved along the valley axis of each flow tube with interactions betwe en flow tubes handled through source sink terms in individual conservation equations. According to Allwine et al. (1997), the conservation of species equation for each flow tube is integrated using fully explicit finite difference scheme consisting of forw ard Euler differencing in time, upwind differencing for advection, and central differencing for diffusion, and it requires initial conditions to solve. Vieri Spugnoli model (Vieri and Spugnoli, 1996) assumes that the canopy offers an aerodynamic resistance proportional to the square of airspeed and that the fluid density undergoes an exponential decay as a function of the canopy width. Xu et al (1997) model us ed transmission probability to account for the impaction and deposition of droplets on a crop surf ace at a point in space. The model assumes that velocity fluctuations are isotropic with a Gaussian probability density distribution. They simulated the 3 D turbulent flow field created by the air jet and sprayer movement and used the results as inputs for solving the stochastic transport model to determine spray distributions in the field. The model assumes the initial velocity of droplets to be that of surrounding air jet and represents the trees as cubes, a simplification that reduces accuracy of predict ions.

PAGE 55

55 Review of Some Expert Systems In a knowledge rich world where large numbers of decisions are made by managers concerning resource allocation, process control, and production efficiency, several decision options may exist for a given problem. Generall y, these decisions range from simple to complex. The use of knowledge of varying kinds and quantities are involved in these decisions and managers can derive benefits from the use of technology known as expert systems (ES) (Nikolopoulos, 1997). These syste ms make use of artificial intelligence programming techniques in solving well defined problems, and store and process certain kinds of knowledge much faster than humans, particularly in situations where human experts are not available to assist in deciding on the feasible options. These enhance efficient decision making process, and failure to exploit such opportunities, can place a manager at some disadvantage. Several ES have been developed over the years to meet various agricultural management needs. Ozk an (1987) utilized a decision support system in identifying defective nozzles and determining errors in speed and application rate. Plant (1989) developed an integrated ES for decision support in agricultural management. Initial development consisted of a package of modules for cotton and another for peaches management. Beck et al. (1989) developed an ES to advise Florida soybean farmers in managing four key pests of soybean. Freeman and Ayers (1989) developed an ES to assist farmers in tractor selection fr om a list of tractors that best suit a user specified operation. Freeland and Howard (1990) developed an ES to assist the selection and use of ground sprayer nozzles. The system could select up to three nozzles given a user specified application. Han et al (1991) developed an ES for diagnosing technical problems of agricultural sprayers. VanDevender et al. (1994) developed an integrated management support system that can provide recommendations for controlling weeds in rice production. Mansingh et al. (200 7)

PAGE 56

56 developed an expert system for Jamaica coffee pest and disease management. Landers and Gil (2009) evaluated and modified the DOSAVIA software, which was developed by Gil and Planas (2003) to optimize pesticide volume rates in vineyard applications. Gil and Escol (2009) presented the design details of the DOSAVIA software. These tools are important contributions to the domains for which they were developed. Several other expert systems have been developed for different management purposes in a variety of domains. The ES developed in this study also aims at making a significant contribution to air carrier spraying of citrus. Table 2 1. Classification of s prays. Classification Size Range (m) Aerosols < 50 Mists 51 100 Fine 101 200 Mediu m 201 400 Coarse >500 *Obtained from Willis and McDowell (1987).

PAGE 57

57 CHAPTER 3 MODEL DEVELOPMENT AND SIMULATION STUDY Spray penetration into a target tree canopy is critical in orchard air assisted spray applications. Often, spray off target losses oc cur due to several interacting factors. According to literature, these factors include pesticide properties, sprayer design, spray application parameters, tree and orchard characteristics, and weather parameters. The complex nature of the interactions of t hese factors makes spray deposition very unpredictable. Several research efforts have gone into attempting to understand the complex interactions of the influential factors via wind tunnel experiments, open field experiments, and in orchard stu dies (Salyan i and Cromwell, 1992a & 1992b ; Holterman et al., 1994; Salyani, 1998; Koo et al., 2000; Svensson et al., 2003), with very insightful findings. These provide a basis for formulating spray deposition models that can increase knowledge about the application s ystem and thus contribut e to off target loss reduction. Until recently, most agricultural spray models were created for aerial spraying (Miller, 1980; Callander and Unsworth, 1982; Atias and Weihs, 1985; Ekblad et al., 1987; Mickle, 1987; Steinke and Yates 1989; Bilanin et al.,1989; Teske et al., 1990; Barry et al., 1992; Parkin and Wheeler, 1996); very little had been attempted in the area of ground air assisted tree spraying (Sidahmed and Brown, 2001; Brown and Sidahmed, 2001; Walklate et al., 2002; Faro oq and (1990) reviewed and classified aerial spray models. Although the phenomenon of ground air assisted orchard spraying is fairly different from that of aerial fo rest/orchard spraying, several concepts in the lat ter are relevant to the former. In air blast spray applications, atomization of the spray liquid is achieved by hydraulic pressure which forces the spray liquid through the small orifice of the nozzles. With the pressure

PAGE 58

58 alone, the spray can travel up to or beyond a meter. However, the air blast serves to carry the spray droplets to the intended target tree canopies and could carry the spray to several meters away. Random walk procedures have been used to mod el spray transport of aerially applied sprays where the paths of individual droplets are tracked from release until deposition (Reid, 1979; Thompson and Ley, 1983; Holterman et al., 1994). This procedure may be applied to airblast spray cautiously since dr oplet velocities are much higher in airblast sprays. Spray penetration into the canopy is aided by the opening of the leaves as the turbulent spray entrained airstream (spray cloud) passes through. Da Silva et al. (2006) proposed a Lagrangian model to simu late spray behavior within vine canopies. Some spray droplets deposit on intercepting leaves within the canopy. The amount of deposition has been used as an indicator of the penetration (Juste et al., 1990; Harvey et al., 1990). The spray may be discharged either radially using conventional airblast sprayers, or horizontally using tower sprayers. Air assisted sprays behave similarly to air jets in that they both expand and reduce speed as they move away from the point of discharge (Abramovich, 1963). Lately computational fluid dynamics (CFD) has been explored as a practical means of elucidating sprayer design characteristics and operational adjustments in a more controlled manner (Xu et al., 1998). Both semi empirical relationships and fundamental conservat ion laws have been used to describe the development of air jets (Walklate et al., 1996; Delele et al., 2005). free round jets and compared the simulation results to Abram turbulent jets. They found significant agreement between the two. Delele et al. (2007) developed a CFD model that could predict the droplet dispersion from a cross flow air assisted sprayer. It

PAGE 59

59 could account for velocity varia tion at the fan outlet and the type, position, size, and direction of nozzles and the liquid atomization. Farooq and Salyani (2004) modeled the spray transport, penetration into and deposition on a citrus tree canopy. They handled the spray transport from a tower air carrier sprayer as a control volume problem, dividing the spray transport into two phases: transport from sprayer to the tree canopy boundary, and droplet displacement within the canopy. However, the horizontal transport of a spray cloud discha rged from a vertical source with a rectangular cross section may also be viewed as passing through several connected bounded spaces (compartments). This idea is somehow similar to the moving control volume and could be an alternative modeling approach. The current model is based on the compartment modeling approach. It has been developed to predict on target deposition from a typical air blast sprayer, configured with radial air outlet. Such sprayers are commonly used in citrus orchard applications. This cha pter presents the technical details and summary of simulation studies of the model System Description The model is developed using a systems approach. The system diagram is shown in Figure 3 1, where the components are Spray Application Orchard Character istics and Citrus Tree Structure. Spray Application is a management operation. Orchard Characteristics consist of row design, spacing, and number, tree spacing and number per row, and number of missing trees. Citrus Tree Structure includes tree height, sk irt height, canopy radius, and foliage density. Application parameters, consisting of nozzle number, pressure and flow rate, tank mix properties, sprayer airflow rate and speed, and ground speed, are fed into the system as inputs. Spray application to the citrus trees is influenced by orchard condition and the tree structure.

PAGE 60

60 Weather factors, external to the system, influence spray application to the trees. The outputs from the system are droplet evaporation, on target deposit, ground deposit, and drift In Figure 3 2, the tank liquid is transformed from source into spray droplets in transit and transported towards the target canopy. During transport from the source to the canopy boundary, where droplets undergo size reduction due to evaporation, part of the spray deposits on the ground while part drifts away due to turbulence in the spray cloud that causes spray droplets to become detrained. While passing through the canopy, part of the spray may be deflected or drifted due to the canopy shape and density, p art may deposit on intercepting foliage, and part may again deposit directly on the ground or indirectly through runoff from the leaves. The remaining spray emerges from the other side of the canopy boundary and may deposit on the ground, drift away, or de posit on the next canopy rows (Salyani et al., 2007). Nevertheless, the model presented in this chapter is restricted to the immediate rows next to the sprayer. Plume Geometry The dispersion of spray cloud discharged horizontally from a continuous source of vertical rectangular cross section may be visualized as a truncated rectangular pyramid (Figure 3 3). In reality the cross section of the spray cloud expands (in both vertical and horizontal dimensions) and the spray decelerates with distance due to dr ag. As the spray cloud moves away from the source it may be further viewed as passing through several compartments (imaginary slices of sectional area and volume corresponding to the expansion of th e spray cloud. At a certain distance from the source the spray cloud is intercepted by the ground, resulting in ground deposits. Between the spray source and the point of interception, there is additional ground deposit resulting from spray cloud turbulenc e.

PAGE 61

61 Assumptions Formulating the model involved the following assumptions: (1) c ontinuous flow The spray cloud formation is continuous and both airflow and spray liquid discharge at the sprayer outlet are constant for sprayer distance travelled during sim ulation time ; (2) s pray cloud is incompressible ; (3) s lippage (relative motion between the spray droplets and the airstream) is negligible, i.e., droplets move at the same speed of the transporting airstream ( Xu et al. 1997, Lal et al., 2010) ; (4) i nertial spray diffusion in the direction of spray discharge is negligible relative to the diffusion by the airstream, i.e., sprayer airflow is the primary source of droplet movement (Lal et al., 2010) ; (5) t here is no reflection of spray cloud from the ground sur face, i.e., the amount of spray that reaches the ground deposits on the ground ; (6) p hysio chemical properties of the spray liquid do not change during droplet transport ; (7) h omogenous distribution of droplet s within each spray compartment; and (8) t he sp ray cloud width expands in the horizontal dimension, perpendicular to spray direction, by a factor of 0.1 of the distance from the sprayer outlet (Rajaratnam, 1976). Spray Mass Compartments With reference to the 3 dimensional view of the first three compar tments of spray cloud from a rectangular cross section spray source (Figure 3 4), the cross sectional area and volume ( ) of the k th compartment are given by the following formulae : ( 3 1 ) k ( 3 2 ) Air/Spray Velocity As the spray cloud moves away from the sprayer outlet, there is a variation in the horizontal velocity at the cross section of each compartment. It is assumed that the relationship

PAGE 62

62 between the centerline air velocity at source and that at a distance away is applicable to mean spray velocity at the entry cross section of compartments (Farooq and Salyani, 2004). Thus, the mean air velocity at the entry cross section of the k th compartment ( ) is written as: k n ( 3 3 ) where, the mean spray velocity, defined as ( 3 4 ) where, is the ground speed of the sprayer (Farooq and Salyani, 2004). It should be noted that Equation 3 3 is applicable to the air in the free space up to the tree canopy boundary, or where there is no tree at all. Droplet Size Distribution Droplet size distribution of agricultural spray differs from o ne application to another (Reich ard et al., 1977) The American Society of Agricultural and Biological Engineers (ASABE) has grouped these sprays into different categories ( ASABE Standard S 572; Teske et al., 2004 b ). Data for the various categories, contained in the AGricultural DISPersa l model software (AGDISP Version 8.21), were fitted with the Chapman Richards fun ction (Blank and Krantz, 2005) in SigmaPlot software (Systat Software Inc., San Jose, CA), in the form ( 3 5 ) where D is droplet diameter in m F is the fraction of the total liquid volume contained in droplets of diameter less than D and 0 1 2 are positive regression parameters to be estimated. Parameter 0 is the asymptote (which is set to 1), 1 is the slope, and 2 establishes the

PAGE 63

63 point of inflection. These were plotted against volume median diameter (VMD) and tabulated for determ ination by interpolation, with the following approximate relationships: ( 3 6 ) ( 3 7 ) Therefore, with the VMD known, the droplet distribution reconstruction could be obtained by using Equations 3 5 through 3 7, and the range of droplet size established. This size range is divided into a desired number of classes ( n ), with the midpoint (class mean droplet size ( )) as the representative droplet size ( Reichard et al., 1977; Bluman, 2001). This approach discretizes the size of droplets generated with the VMD information. The fraction of the total liquid volume contained in a droplet size class ( ) is calculated as the difference between the values of F ( F is cumulative) for the upper and lower class limits. The mass of the mean droplet ( ) can be obtained as, ( 3 8 ) where, is t he density of the spray liquid. The rate of droplet generation at the sprayer outlet is the summation of droplet generation rates of all mean droplets, i.e., ( 3 9 ) ( 3 10 ) where is the volumetric flow rate of the spray liquid. Spray Dis persion The mass of spray liquid in a given compartment is a contribution of masses of all droplets in the compartment. In the transfer of spray from one compartment to another (Figure 3 5A), not

PAGE 64

64 all the droplets get to move to the subsequent compartment. Due to turbulence in the spray cloud some of the droplets may become detrained from the spray cloud (a lso k nown a s fallout). These droplets may either drift away by the wind or deposit on the ground. However, the proportion of droplet fallout that deposit s on the ground and that portion which drifts are both variable. These are represented in Figure 3 5B, based on which the rate of change of mass of spray in the compartment up to the tree canopy boundary is derived as ( 3 11 ) While passing through the tree canopy, spray droplets undergo further size reduction due to evaporation. Also, part of the spray is captured by the canopy to result in deposition, part deposits directly on the ground or indirectly through runoff from leaves due to excess deposition. Each leaf has a finite holding capacity which if exceeded results in spray runoff. These are represented in Figure 3 5C, based on which the rate of change of mass of spray in a compartment inside the tree canopy is derived as (3 12 ) where ( horizontal distance from the canopy boundary to a point within the canopy ) c an be ca lculated from Equation 3 13 (3 13 ) Equation 3 11 and 3 12 attempt to account for all the spray material that is discha rged. However, in reality, this is very difficult to achieve (Miller et al., 2003). Spray m ass i nflow / o utflow The rate of liquid mass inflow into Compartment 0 (immediately at the sprayer outlet) can be obtained as

PAGE 65

65 ( 3 14 ) To obtain the mass inflow of dye (or active ingredient), the insid e of the summation of Equation 3 1 4 is multiplied by the dye concentration in solutio n ( C l ). For the purpose of simulation, the sprayer is modeled to move in discrete steps in the forward direction. The length of each discrete sprayer outlet. Onl y one side spray application is simulated and mirrored unto the other. The two dimensional spray dispersion from the sprayer outlet is repeated in each step, with a delay time of ( 3 15 ) which determines the amount of spray that is released into the first compartment, of each step using Equation 3 1 4 The rate of mass inflow to Compartment 1 is a function of the mass remaining in Compartment 0 in the previous time step, and so on. The total spray mass over compartment volume of the previous compartment determines the source spray density and the cross sectional area of the current compartment de termines the amount of spray that can enter per time unit. Therefore, the rate of mass inflow to compartment k is given by k n ( 3 16 ) and is the rate of mass outflow from compartment k 1 (the preceding compartment). The rate of mass inflow to the first compartment inside the canopy is the rate of mass outflow from the last compartment at the tree canopy boundary multiplied by the fractio n of compartment volume ( ) that falls within the canopy boundary:

PAGE 66

66 ( 3 17 ) After that, the rate of mass inflow to any compartment inside the canopy is expressed as j m (3 18 ) Droplet e vaporation The rate of spray mass evapo rated from any compartment depends on the overall rate of droplet evaporation. The rate of decrease of the mean droplet diameter of a droplet size class due to evaporation with respect to time, as suggested by Thompson and Ley (1983), is given by ( 3 19 ) where and are Schmidt number and Reynolds number, respectively where the simplicity of the treatment is maintained by ass uming that the transfer of mass occurs similarly to pure water droplets until only the non volatile proportion remains. Thus, f or water based sprays, (Thompson and Ley, 1983) ( 3 20 ) Assuming spherical droplets, t he rate of volume decrease of the mean droplet of a droplet size class w ith r espect t o its diameter is given by ( 3 21 ) The rate of change of droplet mass is obtained by multiplying the product of equation 1 9 and 20 by the spray liquid density, i.e.,

PAGE 67

67 ( 3 22 ) where, ( Thompson and Ley, 1983 ) is the vapor pressure difference in Pa, is the dry bulb temperature in C, and is the wet bulb temperature in C. The total rate of decrease of drople t mass over all droplet size classes in a compartment is ( 3 23 ) Thus, the rate of mass evaporation is given by ( 3 24 ) Potential s pray d rift Spray mass drift is a function of droplet size, wind speed, and wind direction. The rate of spray ma ss drift from any compartment k is the summation of masses drifted from all droplet size classes, i.e., k n ( 3 25 ) ( 3 26 ) where is the instantaneous droplet velocity in the y direction due to the resultant force acting on it by the interaction between spray airstream and ambient wind. It is further assumed that the air velocity is so high such that buoyancy effect on droplet, resulting in upward drift, is ignored.

PAGE 68

68 Ground d eposit ion Direct ground deposition occurs by direct intercepti on of the spray cloud by the ground surface and fallout due to spray cloud turbulence. The vertical distance from the ground to the lower boundary of the k th compartment (upward direction is positive) can be determined as ( 3 27 ) where is the height of the air outlet above ground level, and is the vertical distance between the ground and the lower boundary of the spray cloud. Where > for a given compartment, ground deposition occurs by interception in that compartment. The cumulative rate is obtained as the fraction of m ass inflow that is captured by the ground (Figure 3 3): ( 3 28 ) Ground deposition due to fallout is a function of the droplet mass and gravi ty, and is assumed as ( 3 29 ) ( 3 30 ) where is the instantaneous droplet velocity in the downward direction due to the resultant force acting on it by the interaction between spray airstream and gravity. Canopy d epos it ion As the spray cloud passes through the canopy, further decline of the air velocity occurs due to momentum absorption by the leaves. The air velocity at the entry cross section of the j th compartment inside the canopy can be obtained from the equation suggested by Da Silva et al. (2006):

PAGE 69

69 j m ( 3 31 ) where is the air velocity at the canopy boundary and is the drag coefficient of the canopy. The occurrence of deposition depends on the leaf density within the canopy and the retention capacity of the leaves. The rate of spray mass deposition within the canopy is proposed to be given by ( 3 32 ) where , the fraction of exposed leaves has to be estimated The leaf area density proposed by Farooq and Sa lyani (2004) as a modified Gaussian distribution function is modified as in the sprayer forward direction and expressed as: ( 3 33 ) for a single tree and (3 34 ) for a hedgerow, wh ere is the radius of the tree canopy, is the height of the tree, and the maximum leaf area density , the leaf area density range, and the variance of change in leaf area density along the canopy diameter, are to be estimated Equa tion 3 33 assumes that the tree is symmetric about the vertical axis through the trunk. The maximum deposition that can occur is limited by the retention capacity of leaves, This may be expressed as ( 3 35 )

PAGE 70

70 Any further deposition that occurs would result in spray runoff. Model Simulation The purpose of this step was to numerically solve the set of eq uations presented in the preceding text to result in a time series behavior of the state variables. T o achieve this, the model equations were converted into finite difference equations by Euler Method This technique is equivalent to the Taylor series expa nsion with only the first derivative term included explicitly (Jones and Luyten, 1998). Although relatively less accurate than other methods like the Trapezoidal and second order Runge Kutta m ethods, it was selected for its simplicity and ease of use Thi s was to reduce the computer memory needed and the simulation time required Consideration for Time Step and Compartment Thickness Similarly to every time dependent model simulation, t emporal discretization is an important element in this model The choice t ) value depends on the timescale of the model and how fast the system chang es with time. A slow changing system requires a t whereas a rapidly t t may reduce the number of ti me points (and the amount of data) for a fixed sampling time, an inappropriately t introduces system instabilities, resulting in erroneous outcomes. On the other hand, a t may smoothen data but will also increase computational time. However, a will give a smooth data at a relatively moderate computational time. Spatial discretization is also essential in numerically approximating the solution of an otherwise continuous space dependent model. This may be achieved by using finite dif ferencing which is the case for the model presented here. x ) has an effect t in a time dependent model. For model s that depend on both time and space, like this one, it is often necessar t and x that will optimize the output. In some complex models, the best practical approach is to use

PAGE 71

71 trial and error. The rest of this section t x on system behavior and p ropose s criteria for selecting optimum t x Model s cale The dispersion of spray discharged from one side of an airblast sprayer through a single tree or tree row at one forward step location can last within a few seconds, and depends on the sprayer air velocity. For instance, model observation has shown that a discharged spray mass dispersed within 2 s through a 5 m distance in an open space (Larbi and Salyani, 2010). Citrus trees are planted at a maximum of 7.6 m row spacing making the distance fr om sprayer outlet through a single tree within about 7 m Also, sprayer air velocities range between 2 0 and 70 m/ s (Yates et al., 1985) making the bulk of a discharged spray to pass the space within very few seconds t needs to be app ropriately small for stability. Sensitivity to t emporal and s patial d ifference Several model simulations were run to observe model output from different combinations t x Figure 3 6 shows compariso n between simulation results for spray dispersion using t x Each curve represents airborne spray mass with distance from sprayer outlet, cumulative over time. It can be observed that there is no difference between curves obtained wi t x x = 0.5 and 1.0 m, the curves t = 0.0005 and 0.0008 s are highest at the near side and lowest at the far side, t x = 0.2 and 0.1 m, the curves t = 0.0005 and 0.0008 s are highest throughout, followed by curves obtained with t = 0.002 and 0.004 s, respectively, and the curves are closer at the far side. Additionally, it can x the faster the curves tend to zero. Figure 3 7 x t Similar trends can be observed. The curves are not different at the near side. However, it can be

PAGE 72

72 generally inferred that for increasing indicating larger predicted airborne spray mass The plots of spray dispersion with time, cumulative over distance up to 5 m, for different x t are shown in Figure 3 8 The time point corresponding to the pea ks of the curves indicate the point where the sprayer just makes a discrete forward move, that is, where x Fi gure 3 9 also shows that these heights Sampling t ime The sampling time used for simulating the model determines the computational time, for a t x and U a 0 t the maximum sampling time that may be used is finite. Figure 3 10 shows the maximum sampling time for different combinations of x and U a 0 t = 0.002 s. Time points beyond the maximum sampling time pr oduce indeterminate data. Optimum t ime s tep Analy x and U a 0 The proposed relationship is represented by the following equation: (3 36 ) The plot of Equation 3 3 6 is shown in Figure 3 11 Using the proposed selection criteria 3 1 2 shows a comparison between simulated airborne spray mass for the various combinations over time. Essentially, they produce the same curve. However, the height of the peak observed around 0.1 s

PAGE 73

73 increases with increa x. Figure 3 1 3 shows simulated airborne spray mass (cumulative over time ) with distance ly, the height of the curves decreases with may not give significantly different values. But, there is a pro for choosing one over the other. Selecting 0.5 m will give fewer points in space and thus less time would be involved in computation. However, in another case, 0.5 m will give too few points such that 0.1 m would be a better choice. Approach U sed in S olving D roplet E vaporation To solve for droplet evaporation the rates of decrease of diameter and mass for all the mean droplets were initialized to zero. Using Equation s 3 20 and 3 2 2 these rates were calculated for each time point and then th e dropl et sizes and masses, respectively, were obtained by Euler integration over the sampling time defined This corresponds to temporal evaporation of the spray as represented by the first column of the grid shown in Figure 3 14 The mean time it takes f or droplets to travel across each compartment was calculated and used in Equation s 3 20 and 3 22 to obtain these rates for each compartment. This corresponds to spatial evaporation of the spray as represented by the first row of the grid shown in Figure 3 14 The rates of decrease of diameter and mass of the mean droplets at each temporal spatial grid point were obtained by resolving the rates at the corresponding time point and compartment This is represented by the blue arrow cutting diagonally down the grid (Figure 3 14 ). Since it was very difficult to precisely simulate the change s in droplet size and mass in the compartments while atomization is ongoing into the first compartment due to their complexity it was necessary t o place a constraint on the st art of evaporation. Thus, in the simulation, evaporation is made to

PAGE 74

74 start at time where the sprayer has just advanced past the current forward location and there is no release of spray into the first compartment. This means that all the mean droplets begin evaporation at their original sizes and can be tracked adequately without the complexi ty of dealing with other sizes. Figure 3 15 demonstrates results obtained with droplets of sizes of 76, 229, 382, and 535 m using the approach described above It can be seen that the 76 m droplets completely evaporated within a little over 1 s, implying that they lose mass quickly and are more prone to drift. Figure 3 1. Spray model system diagram

PAGE 75

75 Figure 3 2. Spray model Forrester dia gram

PAGE 76

76 Figure 3 3. Schematic of a spray cloud from a rectangular cross section (stationary) source (adapted from Beychok, 2005) Figure 3 4. Three dimensional view of imaginary compartments

PAGE 77

77 Figure 3 5. Spray partition in transi t. A) Schematic of spray transport from one compartment to another. B) Free body spray mass balance of a compartment outside tree canopy. C) Free body spray mass balance of a compartment inside tree canopy. Figure 3 6 Simulated airborne spray mass with distance using different time steps at various compartment thicknesses. C Distance from Sprayer, m Distance from Sprayer, m

PAGE 78

78 Figure 3 7. Simulated airborne spray mass with distance using different compartment thicknesses at variou s time steps.

PAGE 79

79 Figure 3 8. Simulated airborne spray mass with time using different compartment thicknesses at various time steps.

PAGE 80

80 Figure 3 9. Simulated airborne spray mass with ti me using different time steps at various compartment thicknesses.

PAGE 81

81 Figure 3 10. Trend of maximum sampling time for different compartment thicknesses. Figure 3 11. Relationship b etween sprayer air velocity, compartment thickness, and simulation time step.

PAGE 82

82 Figure 3 12. Simulated airborne spray mass with time using optimum time step/compartment thickness combinations. Figure 3 13. Simulated spray mass inflow (cumulative over time) with distance using optimum time step/compartment thickness combinations.

PAGE 83

83 Figure 3 14. A schematic of approach used to solve temporal spatial evaporation. Figure 3 15 Plot of evaporating droplet size over time.

PAGE 84

84 CHAPTER 4 MODEL VALIDATION In order to gain confidence in their use, a ll models need to be validated (Buranathiti et al., 2006 ; Macal, 2005 ) Primarily, model validation refers t performance, and the only objective test is with data that has not been used in developing the model. Thus, data that has been used in estimating model parameters or, otherwise, to perform cross validation cannot be u sed to quantify model performance (Wallach, 2006) The basic question that need s to be answered is whether a model embodie s and appropriately reproduces the behaviors of the real world system. Validation ensures that the model meets its anticipated require ments. Eventually, the intent of validation is to make the model used ( Craig, 2004 ) At the end of the process is really not a validated model, rather, a model that has passed all the validation tests. There is also at the end a better understanding of the suitability for tackling a range of important questions (Bilanin et al., 1989 ; Richardson and Thistle, 2006 ; Hoffmann et al., 2007 ) Discussions of the methods and equations used in model validation can be found in Wallach (2006) A critical first step in model validation is comparing model predictions with experimental data (Duan et al. 1992; Farooq and Salyani 2004 ) This helps to identify problems with the model and offer ide a s for improvement. Graphs are really valuable for providing a quick visual summary of comparison between model and real data. There are also numerical comparisons which may be simple measures of agreement or difference, normalized measures, decomposable measures that provide extra sources of error, or mea sures that a re based on model quality threshold. For all of these, there is no single method that is considered best. Therefore, it is essential to use a combination of these methods to explore the predictive quality of a model.

PAGE 85

85 The objective of th is stud y w as to validate the model presented in Chapter 3 establish confidence in its usage, and provide basis for improving its predictive quality In meeting this objective, the model was validated in two parts, namely, dispersion and deposition. The dispersion part focused on spray transport in a n open space without a tree which could represent the space between sprayer outlet and tree canopy boundary or a missing tree scenario, whereas the deposition part concentrated on spray deposition within the target tre e canopy. T wo field experiments were conducted to collect appropriate data to compare with model predictions. The data was not used in guiding model develop ment and thus was considered appropriate as test data. The rest of this chapter will describe the ex periment al procedures and results, and attempt to use different methods to explore the predictive quality of the model. Methods Dispersion Experiment The experimental setup ( Figure 4 1 ) consisted of a structure made of polyvinyl chloride (PVC) pipes that f ormed a spatial grid of five target distances (D1 = 1.6, D2 = 2.6, D3 = 3.6, D4 = 4.6, and D5 = 5.6 m) from sprayer outlet and three heights (H1 = 0.6, H2 = 1.8, and H3 = 3.0 m) for sampling airborne spray with vertical targets Horizontal targets were als o used at H1 to compare with vertical targets, and at ground level (0 m). Spray t argets consisted of absorbent paper ( Kimberly Clark, Irvin, TX ) folded, and held onto plastic / acetate cards (5. 4 cm X 8.9 cm) by rubber bands. Spray applications of a solution of a water soluble fluorescence dye, Pyranine 10G ( Keystone Aniline Corporation, Chicago, IL), at a concentration of 100 ppm were made with a PTO driven tractor trailed airblast sprayer (PowerBlast 500 Sprayer, Rears Manufacturing Co., Eugene, Ore.) attac hed to a Ford 7610 tractor (Ford Motor, Dearborn, MI) with PTO rpm of 540 at 2100 engine rpm. The sprayer had twelve nozzles per side and spray applications were made

PAGE 86

86 from the right hand side with all nozzles opened. S pray treatments consisted of combinati ons of two Albuz ATR hollow cone nozzles (Lilac and Blue) and two sprayer speeds ( S low = 2.4 km/h and F ast = 4.8 km/h, nominal) in four replications (Rep 1 to 4) This resulted in sixteen spray runs. Summary of the spray treatments is shown in Table 4 1. T he applications were made on October 6,7, 8, and 11, 2010 in the early mornings up to 10:30 am. Samples were collected with a pair of tweezers into pre coded sealable plastic bags after each spray run and temporarily stored in a n ice chest before sending t o the lab for storage in a refrigerator ( Salyani and Hoffmann, 1996 a ; Salyani, 2000 b ; Salyani et al., 2000) This was done to minimize degradation of the dye deposits and to maintain the integrity of the original samples (Hoffman n and Kirk, 2005) To mini mize cross contamination of unsprayed targets from one spray run to another, the target holders, as well as the PVC structure, were wiped dry with a r a g before setting unsprayed targets for the next run Also, each target was underlain on the target holder s with clean sh eets of paper to avoid cross contamination A weather station with CR10 data logger (Campbell Scientific, Logan, UT) was installed on the South eastern side of the PVC structure to record weather data. The data (Table 4 2) included dry bulb temperature (T db ) wet bulb temperature (T wb ), relative humidity (RH), wind speed (W spd ) and wind direction (W dir ), and was recorded at a frequency of 1 Hz The start and run time s of each spray application were recorded (Table 4 3 ) over a span of 30. 5 m and used to extract the relevant parts of the weather data. The spray samples were analyzed by fluorometry. Fluorometric analysis of s amples Fluorometric analysis of the samples (Salyani 2000b) was done by rep lication A variable wash volume of deionized ( DI) water, as washing liquid, was used for all the samples. A Turner 111 fluorometer ( Turner Designs, Sunnyvale, CA) was used to analyze the samples.

PAGE 87

87 Data a nalysis The Turner 111 fluorometer reads the fluorescence of each solution, say in Flx, at different magnitudes (1X, 3X, 10X, and 30X). Sample data read at different magnitu des were used to establish pair wise ratios between the different magnitudes. The ratios were sorted in ascending order and averaged, leaving out 25% above and below. Since most of th e data could be read at the 3X magnitude on the instrument, all the data read on this instrument were converted to 3X using the established ratios. Standard solutions of different concentrations obtained from a stock solution of concentration equal to the expected tank concentration were used to establish a relationship between the fluorescence readings at 3X and actual concentrations (in ppb) of the solutions. This relationship is shown in Figure 4 2 and is given as: ( 4 1 ) T he Flx data was converted to concentrations and corrected by subtracting concentration of solutions obtained from blank unsprayed targets and then converte d to pyranine dye deposit using Equation 4 2 (4 2) where, is dye deposit in mg, is concentration of the solution in ppb, is the wash volume in mL, and is the dilution factor. The dye deposit values were normalized by sprayer speed. Mean, standard deviation (SD), and coefficient of va riation (CV) values were calculated for each spray treatment. Prediction of dispersion data The appropriate parameters from the dispersion experiment were used as input for the model to simulate the spray treatments. Time step of 0.005 s and compartment wi dth of 0.2 m

PAGE 88

88 were used, based on criteria defined in Chapter 3. The height of the bottom of the sprayer air outlet above the ground was measured to be 0.52 m. Two approaches (Pred. 1 and Pred. 2) were used in the simulation. In Pred. 1, the spray discharge d was handled as from a partitioned lower, mid, and upper sprayer outlet to conform to the sprayer outlet sections. In Pred. 2, the spray was handled as a whole, with the vertical distance from the bottom of the air outlet to the upper boundary of the spray cloud estimated to be 2.32 m, with and of 65 and 34, respectively. Initial air velocity of the sprayer was measured to be 48.6 m/s, and ten discrete droplet size classes were used in the simulation. Predictions from the simulation were compared with results from the field experiments. Deposition Experiment Twenty plots each of three continuous trees were selected in a Hamlin orange grove in Lake Alfred, FL. The grove was made up of 4.6 m high hed ge r ow trees with 7.6 m between rows and 3.8 m within rows. The rows were set in East West direction and each row had a w idth of 5.5 m The plots were grouped into five blocks, each having four of the selected 3 tree plot s Before the experiment, leaf samples were collected from the unsprayed trees to serve as a baseline ( Derksen and Gray, 1995) F our spray treatments consisting of combinations of two Albuz ATR hollow cone nozzles (Lilac and Blue) and one of two sprayer speeds (Slow = 2.4 km/h and Fast = 4.8 k m/h, nominal), were randomly applied to plots within each block (representing replication). The spray liquid was a solution of Pyranine 10G fluorescence dye at a concentration of 500 ppm. Spray applications were made with a PTO driven tractor trailed airbl ast sprayer (PowerBlast 500 Sprayer, Rears Manufacturing Co., Eugene, Ore.) attached to a Ford 7740 tractor (Ford Motor, Dearborn, MI) with PTO rpm of 540 at 1900 engine rpm.

PAGE 89

89 After each application, three to seven leaves were randomly sampled from the midd le tree (Salyani and Whitney, 1990; Salyani, 2000b) at two sampling heights (H1 = 1.2 m and H2 = 2.4 m) and four canopy depths (D1 = 0 m, nearside canopy boundary, D2 = 0.6 m, D3 = 4.9 m and D4 = 5.5 m). Ground samples were also collected using 8.9 x 5.4 cm 2 plastic cards placed on plastic trays. The setup is shown in Figure 4 3 The samples were collected into pre coded sealable plastic bags (a pair of tweezers was used to collect ground samples) after each spray run and temporarily stored in an ice chest before being sent to the lab for storage in a refrigerator. This was done to minimize degradation of the dye deposits and to maintain the integrity of the original samples before analyzing them. To minimize cross contamination of unsprayed ground targets from one spray run to another, the trays were wiped dry with a r a g before setting unsprayed targets for the next run. Also, each ground target was underlain on the trays with clean sheets of paper A weather station (same as used in the dispersion experime nt) was installed around the middle of the grove to record weather data (Table 4 4) The start and run times of each spray application was recorded (Table 4 5 ) over four trees ( 15.2 m) and used to extract the relevant parts of the weather data. Fluorometri c a nalysis of s amples The samples were analyzed with a Turner 111 fluorometer by replication. A variable wash volume of DI water, as washing liquid, was used for all the samples. Leaf a rea m easurements After the fluorometric analysis of each sample, the le aves were dried with paper towel and the total number of leaves contained in the sample bag recorded. The total one sided leaf area was measured with an area meter (Delta T D evices, England).

PAGE 90

90 Data a nalysis The Turner 111 fluorescence readings were converte d to 1X using pre established ratios. Standard solutions obtained from a stock solution of concentration equal to the expected tank concentration were used to establish a relationship between the fluorescence reading at 1X and actual concentration of the s olutions. This relationship is shown in Figure 4 4 and is given as: ( 4 4) Using this relationship, the data was converted t o concentrations in ppb. The concentrations were converted to pyranine dye deposit in mg using Equation 4 1. Deposition was calculated as a ratio of dye deposit to two times the total one sided leaf area. The deposition values were corrected by subtracting the average deposition of the baseline, and normalizing by sprayer speed to the nominal speeds. The mean, SD and CV values were calculated for each spray treatment. Prediction of deposition data The appropriate parameters from the experiment were used as input into the model to simulate the spray treatments. Time step of 0.005 s and compartment width of 0.2 m were used, based on criteria defined in Chapter 3 The height of the bottom of the sprayer air outlet above the ground was measured to be 0.52 m. Th e vertical distance from the bottom of the air outlet to the upper boundary of the spray cloud (used in handling the spray as a whole) was estimated to be 2.32 m, with and of 65 and 34, respectively. Initi al air velocity of the sprayer was measured to be 48.6 m/s, and ten discrete droplet classes were used in the simulation. To predict spray deposition inside the canopy, five parameters related to the canopy were estimated using Monte Carlo sampling and the GLUE (Generalized Likelihood Uncertainty Estimation) approach ( Beven and Binley, 1992 ). Table 4 6 shows a summary of the prior and adjusted posterior parameter distributions. Retention capacity of leaves was assumed to be 1.25

PAGE 91

91 L/m based on Farooq and Sal yani (2004). These were used to simulate the model and the predictions adjusted Predictions from the simulation were compared with results from the field experiments. Results and Discussion Dispersion Experiment Figure 4 5 (top) shows mean dye deposit for the different spray treatments at different target distances. From this point, samples collected from vertical targets will be referred to as airborne spray and those collected on the ground as ground deposit Overall, mean dye deposit decreased with dist ance from sprayer outlet as expected ( Salyani and Whitney, 1990; Salyani, 2000 c ). Apart from at 1 m from sprayer outlet, B lue S low gave the highest mean dye deposit at all target distances Dye deposit was more variable (higher CV values) in B lue than in L ilac nozzle Mean d eposit decreased and deposit variabili ty increased with travel speed. Ground dye deposit did not show any obvious trends with distance from sprayer outlet and it was highly variable (Figure 4 5 ). However, B lue nozzles gave higher ground deposit than L ilac nozzle and S low than F ast speed Also, at all target distances, variability increased with travel speed for L ilac nozzle but decreased for B lue nozzle At the 0.6 m height (not shown), vertical targets gave higher deposit (averagely abo ut 100% more) than horizontal targets. This shows that horizontal targets cannot be substituted for vertical targets in sampling spray for dispersion measurement purposes. After all, since the spray droplets are travelling horizontally, it is crucial to sa mple airborne spray with vertical targets. For all the spray treatments, dye deposit was highest at the mid target height (1.8 m) and similar for the lower and upper target heights (Figure 4 6 ). At all heights, dye deposit decreased with increase in travel speed and B lue S low treatment gave the highest deposit. For both L ilac and B lue nozzles variability increased with travel s peed at the lower target height; but this was

PAGE 92

92 not consistent with the mid and upper heights. Overall, B lue nozzle gave higher depos it than L ilac nozzle and S low than F ast speed (Figure s 4 7 and 4 8 ). Variability in tracer deposit was not affected by travel speed for L ilac nozzle but it increased for B lue nozzle Ground deposit also decreased with sprayer travel speed but its variabi lity increased. Multiple regression of weather parameters on mean dye deposit revealed that overall, weather conditions had significant effect on mean deposition ( p = 0.0285) at the 95% confidence level. However, individual weather parameters were not sign ificant in agreement with the results of Hoffmann and Salyani (1996) Comparison of R esults with M odel P redictions Airborne Spray Figure 4 9 show s the plots of dye deposit versus distance from sprayer outlet at different heights for all four spray treat ments for measured (markers, Meas. Rep. 1 4) and model predicted (lines) data. Although Pred.1 (dash lines) was computationally more involving, it could not accurately capture the trend in the data for all three heights. Pred. 2 (solid lines), which is a simplification over Pred.1, captured the trend in the data much better. Plots of predicted dye deposit and residuals versus measured data (Figure 4 10) show improved data with Pred. 2. The summary of some measures of agreement between measured and predic ted values are shown in Table 4 7 These measures were computed using the entire data from each replication, not the means over replications. Overall, for airborne spray samples, the bias shows that Pred. 1 under predicted, whereas Pred. 2 over predicted b ut with a smaller magnitude. However, bias is not a sufficient summary of the model error since the outliers make its interpretation ambiguous. For both RMSE and MAE Pred. 2 gave a smaller value implying a better approach than Pred. 1. Finally, using Pred 2, the modeling efficiency ( EF ) is 78% with correlation coefficient ( r ) of

PAGE 93

93 0.8958, indicating that the model is a much better predictor than using the average of measured values as the predictor. Further analysis showed that considering data from individ ual distances from the sprayer outlet, EF value ranged from 71.7% (for 2 m) to 82.1% (for 4 m). Also, considering data from individual heights, the EF values were 71.5, 83.6, and 75.2 % for 0.6 1.8 and 3.0 m heights, respectively. In addition, consider ing means over replications from individual treatments, EF values were 76.7, 69.9, 79.1, and 82.3% for Lilac Slow, Lilac Fast, Blue Slow, and Blue Fast, respectively. Ground Deposit Figure 4 1 1 shows the plots of mean ground deposit for all four spray tre atments at the various distances from the sprayer outlet for measured and model predicted data. The model capture d the trend in the ground deposit data reasonably well for all two approaches used. However, Pred. 2 gave a better prediction, with lower magni tude under prediction, than Pred. 1. The plot of predicted versus measured data and the residuals ( Figure 4 12 ) also show an improved data with Pred. 2 over Pred. 1. With the exception of RMAE (Table 4 7 ), which ignored measured values equal to zero, Pred. 2 had smaller values than Pred. 1 for all the errors calculated. EF and r both increased (from 3 2% and 0. 63 to 50% and 0.72, respectively). Thus ultimately, the model is a much better predictor than using the average of the measured values as the predicto r. Deposition Experiment Consistent with previous results from literatur e (Salyani and Whitney, 1990; Salyani, 2000 c ) canopy dye deposition decreased with canopy depth for all spray treatments (Figure 4 1 3 ). This could be due to reduced mass flux of the s pray and/or spray capture by foliage at depths nearer to the spray side. Also, at all canopy depths, deposition increased with ground speed in agreement with earlier observations of Salyani (2000c) This may have be en due to

PAGE 94

94 decrease in runoff from leaves. However, ground deposition was highly variable and did not show any obvious trend s similarly to results from the dispersion experiment although it is normally expected to reduce with distance Nevertheless, a possible explanation could be that since the row trees were hollow, that the inner wall of the canopy at the far side to the sprayer could have provided relatively greater resistance to the already reduced velocity spray and caused droplets to lose momentum and fall to the ground. For all spray trea tments, canopy deposition decreased and deposition variability increased with height (Figure 4 1 4 ). Figure s 4 1 5 and 4 16 show that canopy deposition was higher in Fast speed than S low speed This could have been due to reduced runoff from leaves. Also, de position was higher in Lilac than in Blue nozzle which could be explained by increased runoff due to high volume However, ground deposition increased with both speed and volume (nozzle) The fraction of the volume applied that deposited on the ground seem s to be on the high side. This could be attributed to the fact that, in some cases, the row trees used were not completely solid walls and therefore reduced spray capture by the canopy. Multiple regression of weather parameters on deposition showed that we ather condition did not have significant effect on mean deposition at the 95% confidence level This could be due to the fact that the ranges of the mean parameters for all five parameters recorded were fairly narrow such that changes in their values could not significantly influence the results. Comparison of R esults with Model P redictions Canopy Deposition Figure 4 17 show s measured and predicted mean dye deposit ion on the tree canopy for all spray treatments Generally, it can be seen that using the est imated parameters the model predicts the data quite well. Predictions for B lue S low and B lue F ast spray treatments appeared to overestimate at the near side but doing quite well at the far side The plot of predicted data and residuals versus measured data (Figure 4 18 ) show a good scatter of the data

PAGE 95

95 around the one to one line Overall, the model gave a bias of 0.03 implying that the model generally over predicted the data. The root mean squared error of prediction given the estimated parameters ( ) was 0.23. On the whole, the predictions resulted in EF = 61.3% and r = 0.9214. This shows that the model is a much better predictor of the measured data than just using the average of the measured data as the predictor. Ground Deposition Ground deposition predictions were very poor. The model under predicted to more than three orders of magnitude and the prediction data considered to be inappropriate and therefore results not shown. There could be several explanations to this shortcoming. Firstly, due to the closeness of the canopy boundary (0.46 m from sprayer outlet) the concept of spray interception could not yield good data. By the angle used, interception should have occurred under the canopy which was prevented by the canopy wall. Mo reover, since the canopy skirt height was very variable, and in one or more cases there were unavoidable openings into the canopy that ideally should have been closed, this could have affected the results. Furthermore, it could also be that most of the spr ay droplets that passed through the canopy or under the skirt lost their momentum very rapidly and thus deposited before they could leave the canopy section. Table 4 1. Summary of spray treatments used in experiment s Treatment Nozzle Operating pressure G round speed VMD Flow/ side Application Rate bar km/h m L/min L/ha Lilac Slow Lilac 150 2.4 97 5.87 383 Lilac Fast Lilac 150 4.8 97 5.87 192 Blue Slow Blue 240 2.4 109 50.89 3323 Blue Fast Blue 240 4.8 109 50.89 1661 *Calculated based on data ob tained from Albuz ATR Application Rate Chart, Nuyttens et al. (2006), and Van De Zande (2008)

PAGE 96

96 Table 4 2 Weather parameters recorded for various spray treatments in the dispersion experi ment Treatment T db (C) T wb (C) RH (%) W spd (m/s) W dir () L ilac S low 17.5 5. 2 15.8 3.7 85.6 12. 2 1.28 0.7 219.4 50.4 L ilac F ast 22.3 1. 4 19.2 1.5 74.0 3.0 1.42 0.6 205.2 0.1 B lue S low 19.3 4.5 17.6 3.1 85.1 13.3 1.18 0. 4 2 20.8 76. 7 B lue F ast 20.5 2.1 18.4 2. 4 82.2 4.5 1.32 0.6 268.2 30. 9 Table 4 3 Actual sprayer speeds for various spray treatments in the dispersion experiment measured over 30.5 m Treatment Rep 1 Rep 2 Rep 3 Rep 4 Run Time Speed Run Time Speed Run Time Speed Run Time Speed s km/h s km/h s km/h S km/h Lilac Slow 41.62 2.6 41.75 2.6 32.06 3.4 41.40 2. 7 Lilac Fast 19.53 5.6 19.53 5.6 15.15 7.2 19.37 5. 7 Blue Slow 42.46 2. 6 42.06 2.6 41.81 2.6 42.09 2.6 Blue Fast 19.62 5.6 19.68 5. 6 15.4 6 7.1 19.75 5. 6 Table 4 4. Weather parameters recorded for various spray treatments in the deposition experiment. Treatment T db (C) T wb (C) RH (%) W spd (m/s) W dir () Lilac Slow 18.2 8.9 16.3 1.9 56.5 11.5 2.22 0.7 153.5 73.5 Lilac Fast 21 .1 2.4 16.1 1.6 59.3 12.5 2.46 0.8 151.3 71.5 Blue Slow 21.3 2.2 16.0 1.6 57.7 13.4 2.71 0.6 139.9 73.8 Blue Fast 21.5 1.7 16.3 1.5 57.6 11.5 2.48 0.8 152.7 77.4 Table 4 5. Actual sprayer speeds for various spray treatment s in the deposition experiment measured over 15.2 m Treatment Rep 1 Rep 2 Rep 3 Rep 4 Rep 5 Run Time Speed Run Time Speed Run Time Speed Run Time Speed Run Time Speed s km/h s km/h s km/h s km/h s km/h Lilac Slow 19.25 2.9 21.06 2 .6 20.65 2 .7 20.84 2 .6 21.37 2.6 Lilac Fast 9.06 6.1 11.87 4.7 12.12 4.5 11.46 4.8 11.65 4.7 Blue Slow 20.53 2.7 19.71 2 .7 20.15 2 .7 20.84 2 .6 20.61 2 .7 Blue Fast 8.68 6.3 11.59 4.7 11.56 4.7 11.68 4.7 11.9 4.7

PAGE 97

97 Table 4 6. Prior and posterior distributions for parameters estimated with the GLUE method Parameter Distribution Prior Parameters Posterior Parameters Normal 3.5 0.5000 3.0 0.0225 Normal 4.0 0.7000 3.4 0.0230 Normal 0.6 0.2000 0.61 0.003 5 Normal 0.5 0.3300 0.10 0.0022 Normal 0.25 0.2100 0.32 0.0334 = maximum leaf area density in m2/m3, = leaf area density range in m2/m3, = variance of change in leaf area density along the canopy diameter, = fraction of exposed leaves = the drag coefficient of canopy Table 4 7 Measures of agreement between measured and predicted values Airborne Spray Ground Deposit Equation Pred. 1 Pred. 2 Pred. 1 Pred. 2 0.0048 0.0028 0.0033 0.0021 0.0002 0.0002 0.0002 0.0002 0.0151 0.0135 0.0126 0.0109 0. 0095 0.0083 0.0075 0.0064 0.621 (62%) 0.555 (56%) 1.005 (101%) 0.865 (86.5%) 0.870 1.131 1.104 1.183 0.728 (73%) 0.783 (78%) 0.320 (32%) 0.496 (50%) 0.8747 0.89 58 0.6267 0.7185 Based on Wallach (2006); MSE = Mean Squared Error; RMSE = Root Mean Squared Error; MAE = Mean Absolute Error; RRMSE = Relative Root Mean Squared Error; RMAE = Relative Mean Absolute Error; EF = Modeling Efficiency; r = Correlation Coeff icient.

PAGE 98

98 Figure 4 1. Field setup for dispersion experiment.

PAGE 99

99 Figure 4 2 Plot of concentration against Turner 111 fluorescence at 3X.

PAGE 100

100 Figure 4 3 Ex perimental setup for the deposition experiment.

PAGE 101

101 Figure 4 4 Plot of concentration against Turner 111 fluorescence at 1X.

PAGE 102

102 Figure 4 5 Airborne and ground dye deposit from different spray treatments in Dispersion Experiment at different target distances from sprayer outlet. Distance from Sprayer Outlet, m Dye Deposit, mg

PAGE 103

103 Figure 4 6 Airborne dye deposit for different spray treatments in Dispersion Experiment at different sampling heights. Dye Deposit, mg

PAGE 104

104 Figure 4 7 Mean deposit fr om different spray treatments in Dispersion Experiment. Spray Treatment Dye Deposit, mg

PAGE 105

105 Figure 4 8 Mean deposit from different nozzles and speeds in the Dispersion Experiment.

PAGE 106

106 Figure 4 9 Measured and pre dicted dye deposit at different heights for all spray treatments in the dispersion experiment. Distance from Sprayer Outlet, m Distance from Sprayer Outlet, m Dye Deposit, mg

PAGE 107

107 Figure 4 1 0 Predicted dye deposit and residuals from the first and second simulation setup versus measured values. Measured Airborne Dye Deposit, mg Predicted Airborne Dye Deposit, mg

PAGE 108

108 Figure 4 1 1 Measured and predicted ground dye deposit for all spray treatments in the Dispersion Experiment at different target distance from sprayer outlet. D Blue Fast Distance from Sprayer Outlet, m

PAGE 109

109 Figure 4 12 Pred icted ground deposit and residuals from the first and second simulation setup versus measured values. Predicted Airborne Dye Deposit, mg

PAGE 110

110 Figure 4 13 C anopy and ground dye deposit ion from different spray treatments in Deposition Experiment at different canopy depths. Canopy Depth ID

PAGE 111

111 Figure 4 1 4 C anopy dye deposit ion for different spray treatments in Deposition Experiment at different sampling heights. Dye Deposition, mg/cm 2 Target Height, m

PAGE 112

112 Figure 4 1 5 Mean deposit ion from different spray treatments in Deposition Experiment. Spray Treatment Mean Dye Deposition, m g/cm 2

PAGE 113

113 Figure 4 1 6 Mean deposit from different nozzles and speeds in the Deposition Experiment. Mean Dye Deposition, mg/cm 2

PAGE 114

114 Figure 4 1 7 Measured and predicted dye deposition versus distance at two heights from all spray treatments in the deposition experiment. Mean Dye Deposition, g/cm 2 Canopy Depth m Canopy Depth m

PAGE 115

115 Figure 4 18 Predicted dye deposition and residuals for the deposition experiment versus measured values.

PAGE 116

116 CHAPTER 5 SENSITIVITY STUDY FOR DECISION SUPPORT Sensitivity analysis of a model is performed to study how t input variables). This provides vital guidance during model development, as well as understanding model behavior when applied to m aking predictions (Monod et al., 2006; Hoffmann et al., 2007 ). Sensitivity analysis relies strongly on simulation outputs and is related to computer experiments (Sacks et al., 1989; Welch et al., 1992). A computer experiment uses a set of computer simulati ranges. A computer experiment may primarily seek to optimize model response, visualize model behavior, explore approximation by a simpler model, or estimate the mean, variance or pro bability to surpass some given threshold (Koehler and Owen, 1996). For a given model, equations, parameters, and input variables all have some associated variability or uncertainties. Initial choices can be made without a complete understanding of the cons equences. Thus Craig et al., (1998) and Craig (2004) performed sensitivity analysis on a model to expound on the effects of some major parameters on spray drift. Estimation procedures or literature reviews may yield certain parameter values whose precision is constrained by inadequate data. In some other cases, parameters may vary from one specific situation to another. The sensitivity of a model can be defined in several ways, and with respect to a single factor or several factors. Either a local sensitivi ty analysis (LSA) or a global sensitivity analysis (GSA) may be done. LSA is derived from the derivatives of the predicted output with respect to the input and shows how rapidly the output increases or decreases locally about the given values of the input. GSA provides a better view of model behavior since it explores the sensitivity of model output where several input factors change in their entire uncertainty domain. Methods of

PAGE 117

117 GSA including one at a time, complete factorial design, and fractional factor ial design methods have been explored by Monod et al. (2006). For simulated experiments, there is no measurement error, at least for a deterministic model. Consequently, there is no residual variance, and thus no need to replicate the same scenarios or int roduce blocking. This is a major difference from designed real field experiments, where replication and blocking are crucial. Although less valuable than GSA when it comes to exploring the uncertainty of several input factors of model output, LSA can be us ed to investigate the role of some model parameters or input variables, and it may consist of simply observing model response. In this chapter, both GSA and LSA are used to study the response of the model presented in this dissertation. The purpose of the input factors within given ranges with respect to the model output, mean deposition. The focus of the LSA was on study ing the response of some system variables of in terest, a irborne spray mass, canopy deposition, and ground deposition to some other system variables based on literature or the model. The purpose was to establish response equations relating them for rapid decision support purposes. The results of this study were meant to be used as basis for defining rules for the development of an expert system developed in C hapter 6 The rules would particularly be used in the engine of a what if analyzer that would be part of the expert system. The rest of this chapter descri bes into details the methods and results of the sensitivity stud ies Methods Global Sensitivity Analysis A set of simulation runs was done, varying certain inputs, to investigate the sensitivity of the model to variations of these inputs. One target tree w as considered and the sprayer outlet was adjacently near the center of the tree. It consisted of varying three levels each of spray volume rate, sprayer air velocity, ground speed, target canopy distance from sprayer outlet, and canopy

PAGE 118

118 density. Details of these factors are summarized in Table 5 1. In each run, the resulting total wet spray liquid mass deposited was recorded. These were done as a 3 5 complete factorial experiment, resulting in an overall 243 runs. A global sensitivity analysis (GSA) was perfo rmed on the data obtained. GSA explores the sensitivity of model output where several input factors change in their entire uncertainty domain. The data was analyzed using the proc glm procedure in SAS for Windows (SAS 9.2 TS Level 1M0, SAS Institute Inc., Cary, NC, USA.). The model sensitivity to the factors was determined by sensitivity indices, which are the sum of squares proportions of the total sum of squares from the analysis of variance (ANOVA) tables. The factor with the highest sensitivity index c ontributes the most to explaining the variations in the response variable, and vice versa. Local Sensitivity Analysis A set of model input and related variables were selected. These variables consisted of direct inputs (liquid flow rate, VMD, and maximum l eaf area density) and an indirect input (operating pressure). Firstly, established relationships among the variables were identified from literature. Secondly, a series of simulation runs was done, varying the direct inputs, to observe the response of the model to changes of these inputs. This was done one variable separately at a time. Results and Discussion Global Sensitivity Analysis Figure 5 1 shows a matrix of two factor interaction graphs of the input factors. Each legend applies to all the graphs in the adjacent row. In each graph, the first factor levels are indicated by the horizontal axis, the second factor levels by the legend, and the response value by the vertical axis. For the actual plots, each line represents one level plot of the second fact or while each marker point on a line represents a plot of one level of the first factor. Parallel line

PAGE 119

119 plots represent no interaction between factors. Conversely, nonparallel lines indicate a possibility of an interaction. In the figure, there seems to be little or no interaction between the factors although there are evident main effects. Mean deposition generally increased with spray volume rate. Increase in deposition could be attributed to the fact that, at higher spray volume rates, more spray mass is available to be intercepted by leaves inside the canopy. At all spray volume rates, deposition increased with air velocity but decreased with sprayer ground speed, target canopy distance, and canopy density. At all air velocities, mean deposition decreased with ground speed, target canopy distance, and canopy density. Mean deposition generally decreased with ground speed, target canopy distance, and canopy density. The presence of significant main effects and interactions are confirmed by Figure 5 2. Figure 5 2 shows the plot of the sensitivity indices of the significant main factors and interactions. Insignificant main factors and interactions are not shown. Each shaded bar represents the proportion of the variability in mean deposition that is explained by the corresponding factor. The unshaded bars show the cumulative proportion of the response explained by the factors, beginning (from the top) with the factor with the largest contribution. It can be observed that target canopy distance contributes the mos t (about 43%) to variation in mean deposition, followed by spray volume rate (about 25%), and then, canopy (foliage) density (about 15%). The only significant interactions were between volume rate and canopy distance, and between volume rate and canopy den sity. All others were not significant and are not shown in the figure. Overall, the significant main factors and interactions explain about 95% of the variabilit y in simulated mean deposition.

PAGE 120

120 Local Sensitivity Analysis Relationship between p ressure and f l ow r ate The relationship between pressure and flow rate of the nozzles has been established (Salyani, 2003 a ) as ( 5 1 ) which can be re written as (5 2) where, are initial and final operating pres sure, respectively, and are initial and final flow rates, respectively. Based on Equ ation 5 1 or 5 2 an increase in operating pressure of 10% will result in an increase in flow rate of 4.9 %. Similarly, a de crease in operating pressure of 10% will result in a decrease in flow rate of 5.1 %. The plot of Equation 5 2 is shown in Figure 5 3 Relationship between f low r ate and v olume m edian d iameter The relationship between nozzle flow rate and VMD of the nozzles has been established by Yates et al. (1985) as : (5 3 ) The values of the regression coefficients A and B for nozzles set at 0 to the airflow direction extracted from the results of Yates et al. (1985) are summarized in Table 5 2 Figures 5 4 and 5 5 show the plots of A and B versus air velocity, and the relationships are given, respectively, by the following polynomials : (5 4 )

PAGE 121

121 (5 5 ) for fan nozzles, and: (5 6 ) (5 7 ) for cone nozzles. Relationsh ip between a irflow rate and a ir v elocity By the law of conservation of mass, t he air velocity is related to the airflow rate , by the cross sectional area , of the air outlet by the equation: (5 8) (5 9) Thus an increase 10% i ncrease in airflow rate will cause a 10% increase in air velocity. Relationship between a ir v elocity, v olume m edian d iameter and a irborne s pray m ass Twenty simulation runs were done for five droplet sizes (VMD = 50, 150, 250, 350, and 450 m) at four air v elocities (22.5, 30, 40, and 67.2 m/s). Figure 5 6 shows simulated cumulative airborne spray mass for sprays of different VMDs at the 22.5 m/s air velocity. All five air velocities gave the same curves with different intercept The result s show that a irbor ne spray mass increased with increasing droplet size. The plot of the dimensionless airborne spray mass versus droplet size for u a = 30 m/s is shown in Figure 5 7 Overall, the relationship is established as : (5 10 ) where the intercept, can be obtained by interpolation from Table 5 3

PAGE 122

122 Relationship between a ir v elocity, v olume m edian d iameter and g round d eposit Figure 5 8 shows the plot of the dimensionless ground deposit corresp onding to the simulations run in the previous section versus droplet size. Overall, the relationship is established as: (5 11 ) where the intercept, can be obtained by interpolation fro m Table 5 4 Relationship between a ir v elocity, v olume m edian d iameter, and Canopy Deposition Figure 5 9 shows the plot of simulated mean spray deposit ion in an assumed tree canopy with uniform foliage obtained with different droplet sizes. The simulation considered an ideal case of zero wind, a condition that is actually attainable in reality. Although actual trees do not have uniform foliage densities, this was chosen to have a basis for comparing deposition from different sized droplets and different air velocities. The figure shows that higher deposition could be achieved with smaller droplets than with larger droplets although in the presence of wind smaller droplets offer a greater chance for drift From simulations over all droplet sizes and all air velocities, the relationship between deposition, droplet size and air velocity is established by the following polynomial : (5 1 2 ) where is dimensionle ss canopy spray deposition, and the coefficients can be obtained by interpolation from Table 5 5

PAGE 123

123 Table 5 1. Global sensitivity analysis factor levels. Factor Label Level Factor Low Medium High A 625 1250 1875 B Sprayer Air velocity (m/s) 15 30 45 C Sprayer ground speed (m/s) 1.5 3 4.5 D 0.4 1.2 2.0 E 2 /m 3 ) 1.5 3.0 4.5 004). Table 5 2. Regression coefficients for fan and cone nozzles where VMD = A ( FR B ). Nozzle Type Airspeed Regression Coefficients m/s A B Fan 22.5 306.4 0.4084 Fan 44.7 299.4 0.3653 Fan 67.2 235.4 0.1695 Cone 22.5 278.5 0.3259 Cone 44.7 24 3.6 0.3296 Cone 67.2 206.7 0.1615 Table 5 3. Intercept , of polynomial relating airborne spray mass to droplet size for different air velocities. Airspeed Intercept m/s 22.5 1.3305 30 1.5338 40 1.6451 67.2 1.7631 Table 5 4. Intercept , of polynomial relating ground spray mass to droplet size for different air velocities. Airspeed Intercept m/s 22.5 0.6873 30 0.8665 40 0.971 67.2 1.18993

PAGE 124

124 Table 5 5. Coefficients of polynomial relating air velocity, total flow rate per side and dimensionless spray mass deposit. Air Velocity Total Flow Rate / Side Polynomial Coefficient m/s L/min 22.5 5.28 4.8609 0.0205 0.0001 0.0183 0.0139 22.5 33.36 4.6165 0.0206 0.0001 0.0173 0.0127 22.5 61.44 4.4411 0.0218 0.0001 0.0208 0.0157 22 .5 89.52 4.3671 0.0251 0.0001 0.0275 0.0218 22.5 117.50 4.6861 0.0247 0.0001 0.0269 0.0210 30.0 5.28 3.2646 0.0247 0.0001 0.0260 0.0207 30.0 33.36 3.4517 0.0291 0.0002 0.0351 0.0292 30.0 61.44 3.3586 0.0247 0.0001 0.0260 0.0204 30.0 89.52 3.2643 0.023 6 0.0001 0.0232 0.0179 30.0 117.50 3.2493 0.0245 0.0001 0.0249 0.0192 40.0 5.28 2.7547 0.0241 0.0001 0.0245 0.0190 40.0 33.36 2.8165 0.0266 0.0001 0.0296 0.0239 40.0 61.44 2.8221 0.0255 0.0001 0.0272 0.0217 40.0 89.52 2.7616 0.0243 0.0001 0.0253 0.020 0 40.0 117.50 2.7955 0.0253 0.0001 0.0271 0.0215 67.2 5.28 2.4607 0.0259 0.0001 0.0278 0.0220 67.2 33.36 2.5225 0.0272 0.0001 0.0312 0.0256 67.2 61.44 2.4724 0.0260 0.0001 0.0281 0.0223 67.2 89.52 2.4686 0.0257 0.0001 0.0283 0.0227 67.2 117.50 2.5036 0.0268 0.0001 0.0305 0.0248

PAGE 125

125 Figure 5 1. Matrix of two factor interaction graphs of input factors. In each graph, the first factor levels are indicated by the horizontal axis, the second factor levels by the legen d of each row, and the response by the vertical axis.

PAGE 126

126 Figure 5 2 Factorial sensitivity indices (contributions) of significant main factors and interactions based on a 35 complete factorial design and its analysis of variance. Figure 5 3 Relationship between nozzle operating pressure and flow rate. Pressure Ratio Flow Rate Ratio

PAGE 127

127 Figure 5 4 Relationship between air velocity and regression coefficients A (top) and B (bottom) for fan nozzles. Air Velocity, m/s Regression Coefficient A Regression Coefficient B

PAGE 128

128 Figure 5 5 Relationship between air velocity and regression coefficients A (top) and B (bottom) for cone nozzles. A ir Velocity, m/s Regression Coefficient A Regression Coefficient B

PAGE 129

129 Figure 5 6 Simulated airborne spray mass for different droplet sized sprays. Figure 5 7 Relationship between droplet size and airborne spray mass. Distance from Air Outlet, m Airborne Spray Mass Dimensionless Airborne Spray Mass Droplet Size, m

PAGE 130

130 Figure 5 8 Relationship between droplet size and gro und spray deposit. Figure 5 9 Simulated spray deposition in an assumed canopy with uniform medium density foliage obtained with different droplet sizes. Dimensionless Ground Deposit Droplet Size, m Mean Deposit, g Assumed Canopy Depth, m

PAGE 131

131 CHAPTER 6 EXPERT SYSTEM DESIGN AND DE VELOPMENT During a cit rus spray application many factors interact to influence the final spray coverage and deposition on the target tree canopies. These are: tree and orchard characteristics, pesticide properties, sprayer design, spray application parameters, and weather param eters. The interacting complex behavior of these factors adds to the unpredictability of spray deposition. Spray off target losses in citrus applications amount to about 18 26% (Salyani et al., 2007). Added to these losses are application errors, which c an be either mixing errors (erroneous spray mix concentration) or calibration errors (erroneo us spray volume per unit area). In a field study, 152 private and commercial pesticide applicators were surveyed to assess the accuracy of application rates (Rider and Dickey, 1982). The survey revealed that only a quarter of the applications were within 5% of their desired application rates. Calibration errors contributed to about 60% under application to > 90% over application. Grisso et al. (1989) surveyed 103 pr ivate herbicide applicators in Nebraska to assess the extent and cost of misapplication of herbicides. This survey revealed that 30% of the applications were within 5% of the desired application rates, 44% were under applied, and 26% were over applied. The se application errors and off target losses might be costly because they could increase production cost due to crop damage, cause environmental pollution (air, land, and/or water), and potentially impact human and animal health. In order to curtail spray a pplication errors and off target losses there is a need for both effective planning and effective implementation of the spray operations. An effective plan requires that the applicator carefully follows the sprayer calibration procedures and formulae used for the necessary calculations and also considers all the factors affecting spray coverage and

PAGE 132

132 deposition. Weather factors guide spray application timing, and orchard and tree canopy characteristics direct nozzle arrangement and application parameter setti ngs. The knowledge gained through experiments over many years have been used to improve equipment designs, sprayer configuration, material properties, and orchard conditions, all geared towards attaining better spray coverage and increased on target deposi tion. Many years of experience with citrus orchard spraying have contributed to developing a particular way of practice by experienced spray applicators. This experience can be represented as expert knowledge (EK) and have been documented for reference and knowledge gaining. However, this knowledge has not been adequately packaged in a form that can be very readily available to spray applicators. Thus, where no expert is involved in planning a spray application scheme it could be difficult to adequately uti lize the documented EK to achieve an optimum application An attempt to package this knowledge in a way that will enhance planning is vital in helping spray applicators achieve the efficiency which they aspire. Thus, this chapter presents the structure and development of an ES to help citrus pesticide applicators in planning their spray operations. The specific objectives were to develop an ES that would provide assistance and recommendations for sprayer calibration, advise on spray timing, and recommend sp ray parameters that could optimize on target deposition for a user defined operation. Development of the E xpert S ystem General Structure The general structure of the ES is illustrated in Figure 6 1. The User Interface (UI) provides access to the ES, allow ing a user to make input, operat e controls, and receiv e feedbacks. The Inference Module (IM) is a collection of decision rules representing the knowledge domain. The Spray Model is the simulation model presented in Chapter 3 The IM passes values from UI a nd results from model to the Knowledge Base (KB) to check against

PAGE 133

133 rules. Based on the conditions met in the rules an advice/suggestion /recommendation is passed to the IM from the Advice Module (AM) and is displayed by the UI for the user. The AM and KB ar catalogs, and personal communication with experts. Components of the ES inside the dash lined rectangle have been integrated into a composite engine and implemente d in MATLAB Two human experts consulted in developing the ES are Dr. Masoud Salyani, Professor of Agricultural and Biological Engineering and Mr. Roy D. Sweeb, Senior Engineering Technician, at the Citrus Research and Education Center, University of Flor ida. The various parts of the ES were developed under the direct advice of Dr. Salyani, the main expert, where Mr. Sweeb critiqued the design features of the system. Flow charts have been developed to illustrate the logic used in programming the system an d will be described in the following text Detailed Description of E xpert S ystem The detailed structure of the ES is shown in Figure 6 2. The ES ( CitrusSprayEx ) consists of two parts, namely: spray planning and spray implementation. Under spray planning t he ES seeks to assist a user to minimize the chances of spray application errors. It assists in sprayer calibration and spray timing decision making pertaining to the suitability of spray application under different anticipated weather conditions at the intended time for spray application. Sprayer calibrati on guidance are based on established procedures ( Smith, 1990; Salyani 2003 ; Spraying Systems Co., 2008 ) and consist of establishing sprayer travel speed, selecting nozzles appropriate for an intended spray application, and establishing sprayer output Advices on spray timing are based on potential droplet evaporation and drift S pray implementation comprises dispersion and deposition simulations for an orchard application to output canopy deposition, g round deposition, and drift as percentages of the total volume discharged. Here, dispersion concerns spray transport in an open space without a tree

PAGE 134

134 based on the spray model and deposition concerns spray deposition on a citrus tree canopy The output is previous simulation run and the resulting output in a current run. This gives the user a quick idea on how to improve a user defined operation. About under M ain GUI (graph ical user interface) contains information on the ES package. The base flowchart of the ES implementation is shown in Figure 6 3.The symbols are explained in Appendix C Spray c alibration When speed calculation starts the ES gets user input from the G UI ( Figure s 6 4 and 6 5 ). Based on the user preferred units system the ES decides which formula to use according to the method of speed measurement Calibration for speed consists of two methods, namely: Known Distance Method (KDM) and Tree Passed Method (TPM ). KDM consists of running the sprayer over a known distance (D) and recording the run time (T). TPM consists of running the sprayer at constant speed (S) through a mid row of an orchard with known tree spacing (TS) and recording the number of trees passed (TP) and run time. At the end of the calculations, the results are display ed in the selected unit system. The ES selects nozzle based on calculated nozzle flow rate to achieve a desired application rate. At the start of the selection process, the ES gets user input from the GUI (Figure 6 6) and then calculates a correction factor (CF) as the square root of the specific gravity (SG) of the spray solution (Figure s 6 7 and 6 8 ). By default the spray solution is water, with SG = 1.0. Next, the ES decides which units system to use for the selection based on user preference. The KB contains nozzle flow rate s within an optimal pressure range of 6 15 bar for citrus application s Thus, if the operating pressure (Puser) indicated by the user falls outside this rang e the ES notifies the user and does not continue If Puser is within range, then the ES sets a reference

PAGE 135

135 pressure (Pref) which corresponds to pressure used to manually select nozzles from a and then calculates SO from AR and RS. Next, the ES calculates the upper nozzle output (UNO, which is the total output from entire upper nozzles on one side) and lower nozzle output (LNO) based on how the user wants the spray partitioned. Typically, in Florida, spray is applied to citru s trees with two thirds applied to the upper part of the canopy and one third to the lower part Other non uniform partitions may be used However, most applicators do not bother since some sprayers have their manifold already set to discharge in that form at. For instance, a sprayer with twelve nozzles per side will have eight (2/3) in the upper section and four (1/3) in the lower section. In some situations, it becomes necessary to shut off a number of the upper nozzles to adjust for tree size demanding a change in the nozzle configuration per side. These notwithstanding, most applicators also make do with a uniform discharge from both upper and lower sections. The ES allows for both cases, and calculates the flow rate per upper nozzle (RPUN) and rate per l ower nozzle (RPLN) by dividing UNO and LNO by the upper nozzle number (UNN) and lower nozzle number (LNN), respectively, and adjusting by the CF. The ES makes a sub function call to Select Nozzle Sub with inputs as Pref and FR which are used to search thr ough the range of nozzles to select the most appropriate. The flow chart of the Select Nozzle Sub is shown in Figure 6 9 Figure 6 10 shows the flowchart for nozzle comparison. This performs comparative analysis for the expected output from the recommended nozzles, against the desired output, and recommends adjustments that will achieve the desired. At the start, the ES loads nozzle selection data file, which contains inputs and outputs from the latest nozzle selection calculations, and then decides which un its system to use. Next, it assigns SO to the desired flow rate (DFR) per side and then calls Flow Rate Sub function, which searches the KB for nozzle and operating pressure

PAGE 136

136 (P) that match the recommended (Pref) and then returns the corresponding flow rate s (Figure 6 11 ), to obtain the reference flow rates for both upper and lower nozzles (UNFRref and LNFRref, respectively). It calculates the expected flow rates for both upper and lower nozzles (UNFRuser and LNFRuser, respectively) and the expected total fl ow rate. It then calculates a new operating pressure (Prec) and a new ground speed (Srec) that separately will achieve DFR, and gives recommendation. If TFR = DFR, then Prec = Puser and Srec = Suser. In that case, no change should be made. Otherwise, Prec or Srec is recommended. Upon starting sprayer output calculation, the ES gets user input from the G UI and decides whether to perform calculations in Metric or English units according to user preference (Figures 6 12 and 6 13) It then decides on which meth od the calculations should be based. Calibration for sprayer output (SO) and application rate (AR) consists of two different methods, namely: Known Area Method (KAM) and Fixed Position Method (FPM). There are two distinct methods under KAM, namely: Actual Area Known Method (AAKM) and Row Tree Distance Known Method (RTKM). Depending on the specif ic method specified by the user, the ES chooses which equations to use, and then displays the results. Spray t iming Figure s 6 1 4 and 6 15 show the Spray Timing GUI a nd flow chart respectively The ES gives recommend ation s based on the number of parameter s for which the user has provided 6 1 6 ) shows how the ES arrives at a rec ommendation given a set of weather parameters. The three weather parameters of interest are air temperature (TEMP), relative humidity (RH) and wind speed (WS). These have two critical limits (L = critical low; H = critical high) and a normal (N) range for input values. TEMP has a range of 5 35 C, RH has 50 99%, and WS has 0.5 5 m/s. In addition, gives further comments based on rain expectancy within 2 h of application.

PAGE 137

137 Vertical lines connected to the ranges of input values indicate single entry by a u ser, i.e., for instance, a user has enter ed TEMP value only. The ES comments based on that single entry. Horizontal lines indicate multiple (two or three parameters) entries and the nodes repr esent inclusion of a parameter. Spray implementation The GUI for dispersion and deposition simulations showing an ongoing simulation is shown in Figure 6 17. Figure 6 1 8 shows the first part of the flowchart. When the simulation begins, if a prior run had been run, the ES loads the data file for the previous run and as signs the input and output values to some variable. Next, the ES gets user input and checks against acceptable ranges. If any input falls out of range, the ES notifies the user and ends the simulation. If the all the input values are within range, the ES V MD assigns an ASABE droplet size category to it and then finds the regression coefficients for the Chapman Richards function. Next it calculates an upper limit for the droplet size range and then defines the minimum droplet size class midpoint, which is us ed to obtain the other droplet sizes based on the preferred number of classes. With these, the ES obtains the droplet size distribution for the simulated spray. Next (Figure 6 1 9 ), the ES establishes spray compartment s and tree canopy properties and then continues appropriately (Figure 6 20 ) until it displays the simulation results. If a previous simulation run was made, then the ES also displays the changes made to the input values as well as corresponding changes expected in the out put. Test runs To dem onstrate the trends in the ES predictions, the spray implementation GUI was used to simulate several spray applications. The following initial inputs were used: airflow rate = 20 m 3 /s; ground speed = 2.4 km/h; tree height = 5 m; skirt height = 0 m; canopy diameter = 4 m; foliage density = medium; number of trees per row = 100; number of rows = 20; tree spacing = 4 m; row spacing = 6 m; number of missing trees = 0; temperature = 25C; relative

PAGE 138

138 humidity = 60%; and wind speed = 1 m/s. First, to observe the ef fect of application rate on percent canopy deposition, ground deposition, and spray drift, four nozzles (D2 23, D3 25, D4 45, and D5 46 ( Spraying Systems Co., Wheaton, I L ) ) were used. The resulting application rates represent typical rates used in citrus applications. Second, only D3 25 nozz le was used with low, medium, and high foliage densities to observe the effect of foliage density. Third, only D4 25 nozzle was used at three ground speeds of 1.6, 2.4, and 4.8 km/h to observe the effect of ground speed. Fourth, only D4 45 nozzle was used at ground speed of 2.4 km/h with 0, 400 (20%) and 1000 (50%) missing trees to observe the effect of missing trees Evaluation of E xpert S ystem The purpose of evaluating the ES was to assess its utility and extent of its capabilities or limitations. This wa s done to obtain feedback from users who are considered to be knowledgeable in citrus pesticide spray application to use as basis of improving the ES as well as making it popular In the final outcome, the exercise was intended to instill confidence in the intended users regarding use of the system. To a chieve this, an integrated questionnaire comprising field scenario s and evaluation feedback was developed Three major scenarios were presented where the user was required to perform calculations both manual ly and using the ES for comparison between the two The evaluation feedback questions were divided into s even categories Details of the questionnaire are provided in the following text. Spray c alibration scenario (a) A pes ticide applicator wants to know the output per side, the application rate, and the ground speed of his sprayer. He drives through one row middle of an orchard and passes 5 trees in 14.7 s, applying 34 L of water on both sides. The sprayer has 12 nozzles pe r side. The orchard has tree x row spacing of 4.6 m x 6 m, respectively. (i) What is the output/side in L/min? (ii) What is the ground speed in km/s? (iii) What is the application rate in L/ha? (b) Later, the applicator wants to select the appropriate nozz les to give a desired rate of

PAGE 139

139 gravity of 1.25. In addition, he intends to turn off the two uppermost nozzles on each side to conform to mean tree size and apply 2/3 (67%) of the output to the upper canopy and 1/3 (33%) to the lower canopy. (i) Select appropriate upper nozzle. (ii) Select appropriate lower nozzle. (iii) What value must ground speed be changed to in order to attain the desired application rate? (iv) Wh at value must operating pressure be changed to in order to attain the desired application rate? Spray t iming scenario A pesticide applicator wants to apply spray under the following conditions: (a) Dry bulb temperature = 22C, Relative humidity = 60%, Win d speed = 2 m/s; (b) Dry bulb temperature = 18C, Relative humidity = 76%, Wind speed = 5 m/s; and (c) Dry bulb temperature = 32C, Relative humidity = 48%, Wind speed = 2.4 m/s. Spray implementation scenario Assume that a spray applicator intends to appl y spray to a citrus orchard using a conventional radially discharging airblast sprayer. The sprayer is equipped with 12 open nozzles per side at operating pressure of 12 bar and delivers an output of 6 L/min per side. Also, the sprayer travels at a ground speed of 2.5 km/h. The airflow from the two sides (each of 0.13 m width and 1.45 m length) is 18 m 3 /s. The trees (or rows) have a mean height of 4.5 m, skirt height of 0.5 m, canopy diameter (or row width) of 5 m, and have medium foliage density. There are 25 trees per row spaced at 4.6 m and 10 rows in all spaced at 6 m, with overall 10 missing trees. The prevailing weather conditions are as follow: Temperature = 22 C (72 F), Relative humidity = 70%, Wind speed = 2 m/s. (a) Estimate percent: (i) canopy d eposition (ii) ground deposition (iii) drift. (b) What will happen to canopy deposition, ground deposition, and drift (i.e. increase or decrease) if you: (i) use larger nozzles (i.e. increase nozzle output)? (ii) use smaller nozzles (i.e. decrease nozzle output)? (iii) use a higher operating

PAGE 140

140 pressure? (iv) use a lower operating pressure? (v) drive at a faster speed? (vi) drive at a slower speed? The ES was evaluated by five people with varying years of experience in spray technology: 0, 1, 2, 8, and 21 yea rs of experience. Each of the evaluators first answered the questions asked under the scenario problems manually using suggested formulae and procedures attached to the questionnaire and then by using the ES to answer the questions. Finally, each evaluato r evaluated the ES based on seven categories (Table 6 1) and scored various aspects of the ES on a scale of 1 to 5: 1 = Poor; 2 = Fair; 3 = Good; 4 = Very Good; and, 5 = Excellent. Results and Discussion Test R un The results are summarized in Tables 6 2 th rough 6 5 Generally it was observed that increasing volume application rate decreased canopy deposition, increased ground deposition, and decreased spray drift (Table 6 2 ) This is due to increased runoff and increased droplet size. Also increasing foliag e density increased canopy deposition, increased ground deposition, and decreased spray drift (Table 6 3 ). For effect of ground speed, increasing ground speed increased canopy deposition, decreased ground deposition, and decreased spray drift (Table 6 4 ) However, the reverse of the spray drift trend was expected. This indicates a limitation and calls for adjusting the formulation of the ES to give the expected trend. The results also show, as expected, that increasing percentage of missing trees decreased canopy deposition, increased ground deposition and increased drift. Overall, these results show a good indication that the ES could be improved to adequately serve the purpose for which it was developed. Evaluation Figure 6 21 shows the results for t he spray calibration scenario part one. On average, respondents were in about 87% agreement with the ES. Despite availability of formulae, one

PAGE 141

141 respondent (representing 20%) miscalculated speed and application rate. This could represent a real situation whe re a spray applicator might miscalculate application rate resulting in application errors. For the second part, all respondents selected the right upper nozzles in agreement with the ES but one respondent selected a wrong lower nozzle. Also, one and three respondents agreed with the ES on adjustments to ground speed and operating pressure respectively This was due to ina ccurate application of the formulae. Given suggested ranges of weather parameter values, evaluators were in 100% agreement with the ES (Figure 6 22 ). H owever, some evaluators raised concerns that the ES was somewhat too conservative with the ranges. While the ES recommends the suitability of spraying based on mean values, one evaluator suggested that for wind speed, it would be better to use both mean an d gust values. It was not expected that any evaluator would be able to accurately estimate percent canopy deposition, ground deposition, and drift under the spray implementation scenario In fact, one evaluator simply stated that this was not possible. How ever, evaluators selected different nozzles to start with and changed nozzles, operating pressure and ground speed to observe the general Each evaluator used different selections to observe the ES output. Overall, for those evaluators who went through this part, they were about 67% in agreement with the ES. There were discussions about the ES not showing the expected trends for some input value changes observed although the general trends in all three seemed acceptable This indicates the need for further refinement of the ES formulation in order to improve this aspect or rather an opportunity to study the extent of each input variable on the output Figures 6 23 through 6 29 show the results for the evaluation feedback based on the various categories used. On average, 80% of the for program content (Figure 6 23 ). For presentation (Figure 6 24 ), 55% of

PAGE 142

142 the 25 ), 30% of the respondents scored the ES as 26 ), 50% of the Overall, based on program response (Figure 6 27 ), 55% of the respondents scored the ES 28 ), 70% on average scored the quality (Figure 6 2 9 ), m all evaluators as a percentage of the maximum score (Table 6 1 ), the ES scored 89%. This is a good indication that the ES has the potential of serving its objective. Table 6 1 Summary of evaluati on feedback categories. Category No of Q uestions Maximum Score Total Score % Score Program Content 2 50 48 96 Presentation 4 125 110 88 Effectiveness 2 50 43 86 User Appeal & Suitability 4 100 85 85 Program Response 4 100 89 89 Ease of Use 4 10 0 94 94 User Interface and Media Quality 3 75 63 84 TOTAL 24 600 532 89

PAGE 143

143 Table 6 2 Effect of volume application rate with medium foliage tree canopies shown b y ES Nozzle Application Rate L/ha Canopy Deposition % Ground Deposition % Spray Drift % D2 23 700 67 2 31 D3 25 1300 62 12 26 D4 45 2500 53 24 23 D5 46 5600 41 41 18 Table 6 3 Effect of foliage density with medium foliage tree canopies shown by ES Foliage Density Canopy Deposition % Ground Deposition % Spray Drift % Low 43 10 47 Medium 62 12 26 High 86 12 2 25 nozzles were used at ground speed of 2.4 km/h Table 6 4 Effect of ground speed with medium foliage tree canopies s hown by ES Ground speed km/h Application Rate L/ha Canopy Deposition % Ground Deposition % Spray Drift % 1.6 3000 47 22 31 2.4 2000 55 20 25 4.8 1000 69 12 19 2 5 nozzles were used. Table 6 5 Effect of missing trees with medium foliage tree canopies shown by ES Number of Missing Trees Canopy Deposition % Ground Deposition % Spray Drift % 0 53 24 23 400 (20%) 42 19 39 1000 (50%) 27 12 61 45 nozzles were used at ground speed of 2.4 km/h.

PAGE 144

144 Figure 6 1. Structure of ES with arrows showing the direction of information flow. Figure 6 2. Detailed structure of the expert system. Main Graphic User Interface

PAGE 145

145 Figure 6 3. Base flowchart of the ES

PAGE 146

146 Figure 6 4. GUI for sprayer speed calculation.

PAGE 147

147 Figure 6 5 Flowchart for sprayer speed calculation.

PAGE 148

148 Figure 6 6. GUI for nozzle selection.

PAGE 149

149 Figur e 6 7. Flowchart for nozzle selection.

PAGE 150

150 Figure 6 8 Flowchart for nozzle selection (cont.).

PAGE 151

151 Figure 6 9 Flowchart for nozzle selection sub function.

PAGE 152

152 Figure 6 10 Flow chart for nozzle comparison.

PAGE 153

153 Figure 6 1 1 Flowchart for Flow Rate Sub.

PAGE 154

154 Figure 6 12. GUI for sprayer output calculation.

PAGE 155

155 Figure 6 13 Flowchart for sprayer output calculation.

PAGE 156

156 Figure 6 14. GUI for Spray Timing.

PAGE 157

157 Figure 6 1 5 Flowchart for Spray Timing.

PAGE 158

158 Figure 6 1 6 Decision structure for spray timing.

PAGE 159

159 Figure 6 17. GUI for Dispersion and Deposition simulations.

PAGE 160

160 Figure 6 1 8 Flowchart for Dispersion and Deposition simulation s

PAGE 161

161 Figure 6 1 9 Flowchart for Dispersion and Deposition simulation ( cont .).

PAGE 162

162 Figure 6 20 Flowchart for Dispersion and Deposition simulation (con t.).

PAGE 163

163 Figure 6 2 1 Respondent agreement with ES answer to spray er calibration scenario (a). Figure 6 2 2 Respondent agreement with ES answer to spray timing scenario.

PAGE 164

164 Figure 6 2 3 Evaluation of program content of CitrusSprayEx.

PAGE 165

165 Figure 6 2 4 Evaluation of presentation of CitrusSprayEx

PAGE 166

166 Figure 6 2 5 Evaluation of effectiveness of CitrusSprayEx.

PAGE 167

167 Figure 6 2 6 Evaluation of user appeal and suitability of CitrusSprayEx.

PAGE 168

168 Figure 6 27 Evaluation of program response of CitrusSprayEx.

PAGE 169

169 Figure 6 28 Evaluation of ease of use of CitrusSprayEx.

PAGE 170

170 Figure 6 29 Evaluation of user interface and media quality of CitrusSprayEx.

PAGE 171

171 CHAPTER 7 CONCLUSIONS A computer model was developed to simulate the dispersion and deposition of spray discharged from an air carrier sprayer used in applying pesticides in citrus orchards. The model handled the spray as travelling through several connected compartments with increasing cross section, representing the expansion of the spray cloud, and accounted for dr oplet evaporation, potential drift, and ground deposit. The model has been solved with MATLAB using Euler integration method. Two field experiment s have been carried out to validate the model in two parts, namely: dispersion and deposition. The dispersion experi ment focused on spray transport and dispersion in an open space. The follow ing conclusions could be drawn: (1) The results of the field experiment compar ed well to prev ious studies and are repeatable; (2) Partitioning of the spray discharged radiall y into lower, mid, and upper sections in the simulations, to conform to the model could predict airborne spray dispersion with a modeling efficiency ( EF ) of 74% and ground deposit of 12%, and correlation coeffic ient ( r ) of 0.88 and 0.57, respectively; (3) However, p rojecting the middle and upper sections to align with the bottom section, and handling the spray discharged as a whole gave the best adjusted predictions with EF = 78% and r = 0.90 for airborne spray a nd EF = 19% and r = 0.62 for ground deposit ; (4) Handling the dispersion of the spray discharged from one side of an airblast sprayer as a whole gives acceptable results and is better and less involving than dividing the spray radially into sections corres ponding to p artitions of the actual sprayer; and (5) Although the model performed well in predicting ground deposit, the high variability in the measured data makes it s modeling difficult. The model will need to be improved to give better predictions to ac count for th is variability.

PAGE 172

172 T he deposition experiment included a target tree canopy and quantified deposition inside the canopy The following conclusions apply: (6) Like the dispersion experiment, results from deposition experiment also compared well to p ast studies; (7) Parameter estimation using Monte Carlo sampling approach and maximum likelihood estimation method yielded a good set of parameters that were adjusted and used in simulating the deposition data; (8) The simulated spray was handled as from o ne outlet based on conclusion from the dispersion experiment; (9) Model predictions compared well with the field data for on canopy deposition with EF = 61% and r = 0.92; and (10) However, ground deposition was overly under predicted to more than three ord ers of magnitude. This could be due to the fact that since the canopy was very close, the concept of spray interception by the gro und was inadequate in explaining the ground deposition data. Thus, the ground deposition equations would need to be improved i n the future. support for spray applicators. Based on confidence gained in the model, the results have been extended in the design and development of an expert syste m (ES) to assist spray applicators in planning their operations effectively. The ES algorithm has been implemented in MATLAB The following can be concluded: The ES can guide spray applicators in performing sprayer calibration procedures and calculations, recommend the suitability of spray application under given weather conditions, predict output of a spray operation, suggesting changes that could improve canopy deposition. The ES has been evaluated by five respondents with years of experience in spray ap plication ranging from 0 21 years. Overall, the ES was scored at 89%. Although this is high, evaluators raised minor issues that are worth addressing in future improvements. This will instill the needed confidence for its utilization by the intended end us ers.

PAGE 173

173 APPENDIX A CHAPMAN RICHARDS PARAMETERS Data for various agricultural spray categories, contained in the AGricultural DISPersal model software (AGDISP Version 8.21), were fitted with the Chapman Richards function (Equation [3 5]), forcing the asymptot e, 0 to 1. The slope, 1 and the inflexion parameter, 2 estimatess were obtained for all categories. Table A 1 shows the summariz ed results of the data fitting. These parameters were plotted against VMD to establish relationships. Figure A 1 shows the plot of 1 against VMD fitted with the power relationship: ( A 1) and the plot of 2 against VMD with the straight line relationship: ( A 2) The estimates of 1 and 2 from Equation [ A 1] and Equation [ A 2] are also contained in Table A 1 and the droplet size spectra for the different spray categories using t hese second estimates are shown in Figure A 2. The plots of the error difference between the cumulative volume fraction estimates obtained with 1 and 2 from the individual fitted data and that obtained with 1 and 2 from Equation [ A 1] and Equation [ A 2], respectively are shown in Figure A 3.

PAGE 174

174 Table A 1. Summary results of Chapman Richards slope and inflexion parameter estimates. VMD Estimates from individual data fitting Estimates from parameter VMD relationship ASABE Spray Category ID m 1 2 1 2 Aerosol to Fine A2F 47.69 0.0388 4.2419 0.0321 3.3903 Very Fine VF 81.17 0.0155 2.1129 0.0208 3.6481 Very Fine to Fine VF2F 136.61 0.0159 5.2614 0.0136 4.0750 Fine F 179.37 0.0095 3.4676 0.0109 4.4042 Fine to Medium F2M 253.39 0.0086 5.4766 0. 0082 4.9742 Medium M 292.57 0.0073 5.1755 0.0073 5.2759 Medium to Coarse M2C 339.06 0.0069 6.3696 0.0065 5.6339 Coarse C 383.78 0.0058 5.621 0.0059 5.9782 Coarse to Very Coarse C2VC 424.26 0.0052 5.9395 0.0054 6.2898 Very Coarse VC 472.93 0.0049 6.202 9 0.0050 6.6646 Very Coarse to Extremely Coarse VC2XC 519.02 0.0049 7.5181 0.0046 7.0196 Figure A 1. Relationships of 1 and 2 withVMD y = 0.746 x 0.814 Chapman Chapman 1 0.05 0.04 0.03 0.02 0.01 0.00 7 6 5 4 3 2 1 y = 0.077 x +3.023 1 0 100 200 300 400 500 600 Volume Median Diameter, m

PAGE 175

175 Figure A 2. Droplet size sprectra and error using relationships of 1 and 2 withVMD Error Cumulative Volume Frac tion 1.0 0.8 0.6 0.4 0.2 0.0 0.06 0.04 0.02 0.00 0.02 0.04 0.06 0.08 Droplet Size, m 0 200 400 600 800 1000 1200 1400 Droplet Spectra Error VC XC VC C VC C M C M F M F VF F VF A VF

PAGE 176

176 APPENDIX B SPRAYER AIR VELOCITY MEASUREMENTS An experiment was done to determine the mean air velocity from the right side air outlet of a PTO driven tractor trailed sprayer (PowerBlast 500 Sprayer, Rears Manufacturing Co., Eugene, Ore.) attached to a Ford 7740 tractor with engine rpm of 1900 and 540 pto rpm. The air outlet had been marked out int o ten points along the profile and five points across the air outlet (Figure B 1 ), forming a sampling grid of fifty points. The tractor was operated at full throttle in a stationary position in an open field. Air velocity pressure measurements in inches of water column (in. W.C.) were made with a pitot tube. Three replications were made for each sample point. The velocity pressure readings (in. W.C.) were converted to velocity (mph). Summary of the average velocities for the three replications are presented in Figure B 2. Consistently, no data were obtained at grid points (51, 2.5), (51, 3.5), and (51, 4.5) and are assigned zero in Figure B 2. The overall mean air velocity at the air outlet wa s calculated as 108.6 mph (48. 6 m/s). Figure B 1. Sampling grid on sprayer air outlet 3 9 15 21 27 33 39 45 51 55 a b 1 5 1 5 a 3 9 15 21 27 33 39 45 51 55 Sample Points Air outlet 0 .5 1 .5 2.5 3 .5 4 .5

PAGE 177

177 Figure B 2 Mean air velocity at sample grid points

PAGE 178

178 APPENDIX C DEFINITIONS OF FLOWCHART SYMBOLS

PAGE 179

179 LIST OF REFERENCES Abramovich, G.N. 1963. The Theory of Turbulent Jets Ca mbridge, Mass.: M.I.T. Press Allwine, K.J., X. Bian, C.D. Whiteman, and H.W. Thistle. 1997. VALDRIFT A Valley Atmospheric Dispersion Model. J. App. Meteorol. 36: 1076 1087. Atias, M. and D. Weihs. 1985. On the Motion of Spray in the Wake of an Agricultur al Aircraft. Atomization and Spray Tech. 1:21 36. Bache, D.H. and D.R. Johnstone. 1992. Microclimate and Spray Dispersion. Ellis Horwood, New York. Barry, J.W., M.E. Teske, J.A. Rafferty, B.S. Grim, and P.J. Skyler. 1992. Predicting Spray Drift in Complex Terrain. ASAE Paper No. 92 1085. St. Joseph, Mich.: ASABE. Beck, H.W., P. Jones, and J.W. Jones. 1989. SOYBUG: An Expert System for Soybean Insect Pest Management. Agric. Syst 30: 269 286. Beven, K.J. and A.M. Binley. 1992. The Future of Distributed Mode ls: Model Calibration and Uncertainty Prediction. Hydrol. Process. 6(3): 279 298. Beychok, M.R. 2005. Fundamentals of Stack Gas Dispersion 4 th Ed. ISBN: 0964458802. Bilanin, A.J., M.E. Teske, J.W. Barry, and R.B. Ekblad. 1989. AGDISP: The Aircraft Spray D ispersion Model, Code Development and Experimental Validation. Trans. ASAE 32(1): 327 334. Blank, B.E. and S.G. Krantz. 2005. Calculus: Single Variable 1 st Ed. ISBN 10: 1931914591. 461. Bluman, A.G. 2001. Elemental Statistics: A Step by Step Approach 4 th Ed. ISBN 0 07 231694 2. Bouse, L. F., I. W. Kirk and L. E. Bode. 1990. Effect of Spray Mixture on Droplet Size. Trans. ASAE 33(3): 783 788. Brazee, R.D., D.R. Reichard, M.J. Bukovac, and R.D. Fox. 1991. A Partitioned Energy Transfer Model for Spray Impa ction on Plants. J. Agric. Eng. Res. 50:11 24. Brown, R. B. and Sidahmed, M. M. 2001. Simulation of Spray Dispersal and Deposition from a Forestry Airblast Sprayer Part II: Droplet Trajectory Model. Trans ASAE 44(1):11 17. Brown, R.B. and M.D. Taher. 199 9. Modeling Pesticide Deposition in a Plant Canopy Using a Virtual Nozzle. ASAE Paper No. 99 1113. Toronto, ON. Buranathiti, T., J. Cao, W. Chen, L. Baghdasaryan, and Z.C. Xia. 2006. Approaches for Model Validation: Methodology and Illustration on a Sheet Metal Flanging Process. J. Manuf. Sci. Eng. 128 (2): 10 p.

PAGE 180

180 Callander, B.A. and M.H. Unsworth. 1982. Simulation of a Line Source by a Moving Point Source of Droplets II. Practice. Atmos. Environ. 16(8): 1829 1833. Craig, I.P. 2004. The GDS model a rapid computational technique for the calculation of aircraft spray drift buffer distances. Comput. Electron. Agric. 43 (2004) 235 250. Craig, I., N. Woods and G. Dorr. 1998. A Simple Guide to Predicting Aircraft Spray Drift. Crop Prot. 17(6): 475 482. Cunningha m, R.T., J.L. Brann, and G.A. Fleming. 1962. Factors Affecting the Evaporation of Water from Droplets in Airblast Spraying. J. Econ. Ent. 55(2):192 199. Da Silva, A., C. Sinfort, C. Tinet, D. Pierrat, and S. Huberson. 2006. A Lagrangian Model for Spray Beh aviour within Vine Canopies. Aerosol Science 37: 658 674. Delele, M.A., A De Moor, B. Sonck, H. Ramon, B.M. Nicola, and P. Verboven. 2005. Modelling and Validation of the Air Flow Generated by a Cross Flow Air Sprayer as Affected by Travel Speed and Fan S peed. Biosystems Eng. 92(2):165 174. Delele, M.A., P. Jaeken, C. Debaer, K. Baetens, A.M. Endalew, H. Ramon, B.M. Nicola, and P. Verboven. 2007. CFD Prototyping of an Air Assisted Orchard Sprayer Aimed at Drift Reduction. Comput.Electron. Agric. 55: 16 27 Derksen, R.C. and R.L. Gray. 1995. Deposition and Air Speed Patterns of Air Carrier Apple Orchard Sprayers. Trans. ASAE 38(1):5 11. Duan B., Yendol W. G. and Mierzejewski K. (1992). Statistical Comparison of The AGDISP Model With Deposit Data. Atmos. En viron. 26A: 1635 1642. Ekblad, R.B., J.W. Barry, A. Bilanin, and R. Hauser. 1987. Experimental Verification of an Aerial Spray Model in Complex Terrain. ASAE Paper No. 87 1540. St. Joseph, Mich.: ASABE. E bert, T.A. and R.A. Downer. 2006. A Different Look a t Experiments on Pesticide Distribution. Crop Prot. 25:299 309 Farooq, M. and M. Salyani. 2002. Spray Penetration into the Citrus Canopy from Two Air carrier Sprayers. Trans. ASAE 45(5):1287 1293. Farooq, M. and M. Salyani. 2004. Modeling of Spray Penetrat ion and Deposition on Citrus Tree Canopies. Trans. ASAE 47(3):619 627. Freeland, R.S. and K.D. Howard. 1990. TIP An Expert System for Ground Sprayer Nozzle Selection Appl. Eng. Agric. 6(6): 697 700. Freeman, S.A. and P.D. Ayers. 1989. An Expert System f or Tractor Selection. Appl. Eng. Agric. 5(2): 123 126.

PAGE 181

181 Futch, S.H. and R. Atwood. 2009. All About Airblast Applications: Understanding how airblast sprayers work will help you minimize the use of pesticide. FDACS CUE Series. Gil, Y., and C. Sinfort. 2005. Emission of Pesticides to the Air During Sprayer Application: A Bibliographic Review. Atmos. Environ. 39:5183 5193. Gil, Y., C. Sinfort, B. Bonicelli, V. Bellon Maurel, and A. Vallet. 2005. Methodology for Assessment of Drift from Radial Sprayers in Vineyar d Applications. Information and Technology for Sustainable Fruit and Vegetable Production FRUTIC 05. Montpellier, France Giles, D.K., E.B. Salem, and N. Bacheri. 1991. Turbulent Jet Flow Characteristics of a Dual Port, Air Atomization Spray Nozzle. J. Agri c. Eng. Res. 49: 133 149. Grisso, R.D., E.C. Dickey, and L.D. Schulze. 1989. The Cost of Misapplication of Herbicides. Appl. Eng. Agric. 5(3): 344 347. Hall, F.R., J.A. Cooper, and D.C. Ferree. 1991. Orchard geometry and pesticide placement. Brit. Crop Pro t. Coun. Mono 46:171 176. Han, Y. J., J. Boomaerts, S. Nugroho, G. D. Christenbury and F. J. Wolak. 1991. Development of an Expert System for Sprayer Diagnostics. Appl. Eng. Agric. 7(5):520 524 Harvey, B.M.R., S. Watson, J.E. Bailie, A. Saunders, and I. McMordie. 1990. Pre Harvest Retting of Flax with Glyphosphate: Effects of Novel Methods of Spray Application on Penetration of the Canopy and Uniformity of Deposition. Ann. Applied Biology 116(1): 111 118. Hernndez Hernndez, C.N.A., J. Valle Mora, A. S antiesteban Hernndez, and R. Bello Mendoza. 2007. Comparative ecological risks of pesticides used in plantation production of papaya: Application of SYNOPS indicator. Sci. Total Environment 381:112 125 Hewitt, A.J., D.L. Valcore, D. Esterly, and M.E. Tes Scenarios with the AgDRIFT Model. ASAE Paper No. 97 1073, St. Joseph, Mich.: ASABE. Hobson, P.A., P.C.H. Miller, P.J. Walkate, C.R. Tuck, and N.M. Western. 1993. Spray Drift from Hydraulic Nozzles: The Use of a Computer Simulation Model to Examine Factor Influencing Drift. J. Agric. Eng. Res. 54:293 305. Hoffmann W. C., B. K. Fritz and D. E. Martin. 2007. AGDISP Sensitivity to Crop Canopy Characterization. Trans ASAE Vol. 50(4): 1117 1122. Hoffmann, W.C. and I.W. Kirk. Trans. ASAE 48(1):5 11. Hoffmann, W. C. and M. Salyani. 1996. Spray Deposition on Citrus Canopies under Different Meteorological Conditions. Trans ASAE Vol. 39(1): 17 22.

PAGE 182

182 Holterman, H.J., H.A.J Porskamp, and J.F.M. Huijsmans. 1994. Modeling Spray Drift from Boom Sprayers. Comput. Electron. Agric. 19: 1 2. Huitink, G. IV., J. T. Walker and T. L. Lavy. 1990. Downwind Deposition of 2, 4 Dichlorophenoxyacetic acid Herbicide (2, 4 D) in Invent Emuls ion. Trans ASAE 33(4): 1051 56. Jones, J.W. and Luyten, J.C. 1998. Simulation of Biological Processes. In Agricultural Systems Modeling and Simulation Peart, R.M. and Curry, R.B. (eds). Marcel Dekker Inc. ISBN 0 827 0041 4. Juste, F., S. Sanches, R. Iba nez, L. Val, and C. Garcia. 1990. Measurement of Spray Deposition and Efficiency of Pesticide application in Citrus Orchards. J. Agric. Eng. Res. 46(3): 187 196. Kirk, I. W. and W. C. Hoffmann. 2002. Operational Factors Influence Spray Drift and Deposition from Helicopters. ASAE Paper No. 02 AA06. St. Joseph, Mich.: ASABE. Koehler, J. R. and A. B. Owen. 1996. Computer Experiments. In Handbook of Statistics Volume 13 Ed. S. Ghosh and C. R. Rao. Elsevier Science B.V: 261 308. Koo, Y.M., M Salyani and J.D. Whitney. 2000. Spray Variable Effects on Abscission of Orange Fruit for Mechanical Harvesting. Florida Agric. Experiment Station Journal Series No. R 07247, Vol. 43(5): 1067 1073. Lab Safety Supply Inc. 2007. Dye Tracers. Ezfacts Document Number: 199 Lal, S., A. Kushari, J.C. Kapoor, and S. Maji. 2010. Modelling of Externally Mixed Air Blast Atomizer. International Journal of Dynamics of Fluids 6(1):25 40. Landers, A. and E. Gil 2009 Software to Determine the Optimal Volume Rate for Pesticide Application in Vineyard. Annual Report of 2009/08 and07. Viticulture Consortium East.Geneva. C ornell University, NYSAES. Larbi, P.A. and M. Salyani. 2010. Spray Model to Predict Deposition in Air Carrier Sprayer Applications. ASABE Paper No. 1008319. St. Joseph, Mich. : ASABE. Liu, H. 2000. Science and Engineering of Droplets: Fundamentals and Applications. Noyes Publications. Norwich, NY Macal, C.M. 2005. Model Verification and Validation. Workshop on "Threat Anticipation: Social Science Methods and Models"The Universi ty of Chicago and Argonne National Laboratory. April 7 9, 2005. Chicago, IL Mansingh, G., H. Reichgelt, and K.M.O. Bryson. 2007. CPEST: An Expert System for the Management of Pests and Diseases in the Jamaican Coffee Industry. Expert Syst. Appl. 32: 184 19 2.

PAGE 183

183 Meroney, R.N. 1990. Review and Classification of Complex Terrain Models for Use with Integrated Pest Management Program Spray Models. Colorado State University Civil Engineering Memorandum 89 90 RNM 1, Prepared for Forest Service Technology and Developm ent Program, U.S. Dept. of Agriculture, Forest Service, Missoula, Montana Mickle, R.E. 1987. A Review of Models for ULV Spraying Scenarios. In Proc. Symposium of the Aerial Application of Pesticides in Forestry. AFA TN 18. NRC No. 29197: 179 188, Ottawa, O N. Miller, C.O.M. 1980. A Mathematical Model of Aerial Deposition of Pesticides from Aircraft. Environ. Sci. Technol. 14(7): 824 831. Miller, D. R., E. W. Huddleston, J. B. Ross and W. E. Steinke. 2003. Airblast Spray Partitioning in a Mature Pecan Orchar d. Trans. ASABE Vol. 46(6): 1495 1501. Miller, D., J. Rafferty, R. Sanderson, H. Thistle, and D. Whiteman. 1994. Meteorological Measurements for Spray Drift Modeling. Draft Report of the National Spray Model Advisory Committee. Miller, P. 2003. The Measu rement of Spray Drift. Pesticide Outlook (Oct): 205 209. Miller, P.C.H., M.C. Buttler Ellis, C.R. Tuck. 1996. Entrained Air and Droplet Velocities Produced by Agricultural Flat Fan Nozzles. Atomization Sprays 6(6): 693 707. Miller, W. M. and Salyani, M. 2006. Stewardship Monitoring and Control of Aldicarb Application to Florida Citrus. Appl. Eng. Agric Vol. 22(3): 351 356. Monod, H. C. Naud, and D. Makowski. 2006. Uncertainty and Sensitivity analysis for crop models. In: Working with Dynamic Models: Eva luation, Analysis, Parameterization, and Applications Ed. D. Wallach, D. Makowski, and J.W. Jones. Elsevier. New York. ISBN 10: 0 444 52135 6: 55 99. Moyle, A. M., P. M. Smidansky and D. Lamb. 2006. Laboratory Studies of Water Droplet Evaporation Kinetics The 12th Conference on Cloud Physics. Madison, WI. Nigg, H. N. and J. H. Stamper. 1983. Exposure of Spray Applicators and Mixer Loaders to Chlorobenzilate Miticide in Florida Citrus Groves. Arch. Environ. Contam. Toxicol 12, 477 482. Nikolopoulos, C. 1997. Expert System: Introduction to First and Second Generation and Hybrid Knowledge Based Systems. Marcel Dekker Inc., New York, NY Nuyttens, K. Baetens, M. De Schampheleire, and B. Sonck. 2006. PDPA Laser Based Characterisation of Agricultural Sprays. Agricultural Engineering International: the CIGR Ejournal III, Manuscript PM 06 024. Ozkan, H.E. 1987. Sprayer Performance Evaluation with Microcumputers. Appl. Eng. Agric. 3: 36 41.

PAGE 184

184 Pai, N., M. Salyani and R. D. Sweeb. 2009. Regulating Airflow of Orchard Airblast Sprayer Based on Tree Foliage Density. Trans. ASABE 52(5): 1423 1428. Parkin, C.S. and P.N. Wheeler. 1996. Influence of Spray Induced Vortices on the Movement of Drops in Wind Tunnels. J. Agric. Eng. Res. 63(1): 35 44. Pfeiffer, D.G. 2002. Airb last Sprayer Calibration. [Retrieved on 1/14/2008 from http://www.virginiafruit.ento.vt.edu/calib.html] Plant, R.E. 1989. An Integrated Expert Decision Support System for Agricultural Management. Agric. Syst 29: 49 66. Rajaratnam, N. 1976. Turbulent Jet s In Developments in Water Science, Volume 5 series Amsterdam, The Netherlands: Elsevier Scientific. Reichard, D.L., H.J. Retzer, L.A. Liljedahl, and F.R. Hall. 1977. Spray Droplet Size Distributions Delivered by Airblast Orchard Sprayers. Trans. ASAE 20( 2):232 237, 242. Reid, J.D. 1979. Markov Chain Simulations of Vertical Dispersion in the Neutral Surface Layer for Surface and Elevated Releases. Boundary Layer Meteorol. 16: 3 22. Richardson, B. and H.W. Thistle. 2006. Measured and Predicted Aerial Spray Interception by a Young Pinus Radiata Canopy. Trans. ASABE 49(1): 15 23. Rider, A.R. and E.C. Dickey. 1982. Field Evaluation of Calibration Accuracy for Pesticide Application Equipment. Trans. ASAE 25(2): 258 260. Sacks, J., W.J. Welch, T.J. Mitchell, and H.P. Wynn. 1989. Design and Analysis of Computer Experiments. Statistical Science 4(4):409 435. Salyani, M. Calibration of Airblast Sprayers. Florida Agricultural Experiment Station Journal Series No. N 01255. Circular 1435 Salyani, M. Fruit Abscission Sp rays for Mechanical Harvesting of Oranges. ISBN 3 00 008305 7 Salyani M. 1988. Droplet Size Effect on Spray Deposition Efficiency on Citrus Leaves. Trans. ASAE 31(6):1680 1684 Salyani, M. 1993. Degradation of Fluorescent Tracer Dyes Used in Spray Applicat ions. Pesticide Formulations and Application Systems 13:215 226. Salyani, M. 1994a. Spray Deposition and Drift from Air Carrier Sprayers. AgEng Conference, Report No. 94 D 146, Milano 23 26 September 1994 Salyani, M. 1994b. Spray Deposition and Drift in Ci trus Applications. Proceedings of the 2nd Environmentally Sound Agriculture Conference, April 22 24, 1994. Orlando, FL

PAGE 185

185 Salyani, M. 1994c. Spray Technology Research for Orchard Applications. Acta Hortic. 372:67 74. Salyani, M. 1995. Spray Deposition and Dri ft from Airblast Sprayers used in Citrus Applications. Citrus Industry Magazine, January, 1995 Salyani, M. 1997. Performance of Sprayers in Florida Citrus Production. 5th Industrial Conference of Fruit, Nut, and Vegetable Production Engineering, Davis, CA Salyani, M. 1998. Effects of Flow Rate and Rotational Speed on Performance of Two Rotary Atomizers. In Pesticide Formulations and Applic. Sys.: 18 th Vol., ASTM 1347, J.D. Nalevaja, G.R. Goss, and R.S. Tann, Eds., ASTM. Salyani, M. 1999a. A Technique for St abilizing Droplet Spots on Oil Sensitive Paper. Trans. ASAE 42(1)45 48. Salyani, M. 1999b. Spray Volume Rate Errors in Intermittent Operation of Hydraulic Nozzles. Appl. Eng. Agric 15(1):31 34. Salyani, M. 2000a. Effective Application of Pesticides in Flo rida Citrus Production. In Proc. Intl. Soc. Citricult. IX Congr 248 251 Salyani, M. 2000b. Methodologies for Assessment of Spray Deposition in Orchard Applications. ASAE Paper No. 00 1031. St. Joseph, Mich.: ASABE. Salyani, M. 2000c. Optimization of Dep osition Efficiency for Airblast Sprayers. Trans. ASAE 43(2):247 253. Salyani, M. 2003a. Calibration of Airblast Sprayers. UF/IFAS Extension, Circular 1435. Salyani, M. 2003b. Droplet Size Affects Durability of Spray Deposits, Pesticide Formulations and App lication Systems. 23rd International Symposium, ASTM STP 1449, G Volgas, R. Downer, and H. Lopez, Eds., ASTM International, West Conshohocken, PA. Salyani, M. and R.P. Cromwell. 1992a. Drift Losses from Citrus Spray Applications. In Proc. Florida State Hor t. Soc. 105: 13 18. Salyani, M. and R.P. Cromwell. 1992b. Spray Drift from Ground and Aerial Applications. Trans. ASAE 35(4): 1113 1120. Salyani, M. and R. P. Cromwell. 1993. Adjuvants to Reduce Drift from Handgun Spray Applications. In Pesticide Formulati ons and Application Systems: 12th Volume, ASTM STP 1146, B. N. Devisetty, D.G Chasin, and P. D. Berger, Eds., American Society for Testing and Materials, Philadelphia. Pp 364 376 Salyani, M. and M. Farooq, 2003. Sprayer Air Energy Demand for Satisfactor y Spray Coverage in Citrus Applications. In Proc. Florida State Hort. Soc 116:298 304.

PAGE 186

186 Salyani, M. and M. Farooq. 2004. Drift Potential of Citrus Air Carrier Sprayers. In Proc. Florida State Hort. Soc. 117:130 135. Salyani, M. and R. D. Fox. 1994. Perform ance of Image Analysis for Assessment of Simulated Spray Droplet Distribution. Trans. ASAE 37(4): 1083 1089. Salyani, M and R. D. Fox. 1999. Evaluation of Spray Quality by Oil and Water Sentitive Papers. Trans. ASAE Vol. 42(1): 37 43. Salyani, M. and W .C. Hoffmann. 1996a. Air and Spray Distribution from an Air Carrier Sprayer. Appl. Eng. Agric. 12(5):539 545. Salyani, M. and W.C. Hoffmann. 1996b. Effects of Application Time and Spray Volume on Deposition. In Proc. of Florida State Horticultural Society Paper 109:46 50 Salyani, M. and W. M. Miller. 2005. Precision Application Technology for Monitoring Soil Applied Pesticides in Florida Citrus Production. Information and Technology for Sustainable Fruit and Vegetable Production. FRUTIC 05. Montpellier, Fra nce. Salyani, M. and J. Serdynski. 1990. Development of a Sensor for Spray Deposition Assessment. Trans. ASAE 33(5):1464 1468. Salyani, M. and J.W. Serdynski. 1993. A Device and Method for Sprayer Calibration. Appl. Eng. Agric. 9(1):29 32. Salyani, M. and J.D. Whitney. 1990. Ground Speed Effect on Spray Deposition inside Citrus Trees. Trans. ASAE 33(2):361 366. Salyani, M. and J. D. Whitney. 1991. Effect of Oscillators on Deposition Characteristics of an Airblast Sprayer. Trans. ASAE 34(4):1618 1622. Saly ani, M., M. Farooq, and R.D. Sweeb. 2007. Spray Deposition and Mass Balance in Citrus Orchard Applications. Trans. ASABE 50(6): 1963 1969. Salyani, M., S.L. Hedden, and G.J. Edwards. 1987. Deposition Efficiency of Different Droplet Sizes for Citrus Sprayin g. Trans. ASAE 30(6):1595 1599. Salyani, M., Y.M. Koo, and R.D. Sweeb. 2000. Spray Application Variables Affect Air Velocity and Deposition Characteristics of A Tower Sprayer. In Proc. Florida State Hort. Soc. 113:96 101. Salyani, M., W. M. Miller, S. Buch anon, and R. D. Sweeb. 2007. Managing Aldicarb Application with GPS/GIS Systems. J. ASTM Int 4(5):7 p. Salyani, M., R.D. Sweeb, and M. Farooq. 2006. Comparison of String and Ribbon Samplers in Orchard Spray Applications. Trans. ASABE 49(6):1705 1710.

PAGE 187

187 SDTF 1997. A Summary of Ground Application Studies. Spray Drift Task Force. Contact David R. Johnson at Stewart Agricultural Research Services, Inc., Macon, MO. SDTF. 2001. A Summary of Tank Mix and Nozzle Effects on Droplet Size. Spray Drift Task Force. Cont act David R. Johnson at Stewart Agricultural Research Services, Inc. Macon, MO. Sidahmed, M. M. 1996. A Theory for Predicting the Size and Velocity of Droplets from Pressure Nozzles. Trans. ASAE 39(2):385 391. Sidahmed, M.M. 1997. A Transport Model for Nea r Nozzle Fan Sprays. Trans. ASAE 40(3):547 554 Sidahmed, M.M. 1999. Drop Size/Velocity Correlations at Formation of Sprays from Fan Nozzles. Trans. ASAE 42(6):1557 1564. Sidahmed, M.M. and R.B. Brown. 2001. Simulations of Spray Dispersal and Deposition fro m a Forestry Airblast Sprayer Part I: Air Jet Model. Trans. ASAE 44(1): 5 10. Sidahmed, M. M., M. D. Taher and R. B. Brown. 2005. A Virtual Nozzle for Simulation of Spray Generation and Droplet Transport. Biosystems Eng. 92(3): 295 307. Smith, D. B. and E. C. Burt. 1970. Effects of The Size of ULV Droplets on Deposits within Cotton Foliage Both inside And Immediatel y Downwind from A Treated Swath. J. of Econ. Entomol. 63(5):1400 1405. Smith, T.J. 199 0. Orchard Airblast Sprayer Calibration, Adjustment and Operation. Corporate Extension. EB1575. Washington State University. Spraying Systems Company. 2008. TeeJet Technologies. Catalog 50A. Wheaton, IL 60187. Steinke, W.E. and W.E. Yates. 1989 a Modifying Gaussian Models to Obtain Improved Drift Predictions ASAE Paper No. 891525. St. Joseph, Mich.: ASABE. Steinke, W.E, and W.E. Yates. 1989 b Quantification of Aerosol Dispersion from a Pesticide Application. ASAE Paper No. 891044. St. Joseph, Mich.: ASABE. Stoughton, T.E., D.R. Miller, X. Yang, K.M. Ducharme 1997. A Comparison of Spray Drift Predictions to Lidar Data Agric.For Meteorol 88:15 26. Stover, E.D. 2002. Sensor Controlled Spray Systems for Florida Citrus. UF/IFAS Extension EDISHS 872. Stover, E., J. Hebb, R. Sonoda, and M. Salyani. 2004. Airblas t Application of Copper Fungicide to Grapefruit Does Not Affect Windscar. HortScience 39(3): 516 519. Stover, E., D. Scotto, and J. Salvatore. 2002a. Spray Applications to Citrus: Survey of Current Practices in the Indian River Area. IFAS Fact Sheet HS 127

PAGE 188

188 Stover, E., D. Scotto, C. Wilson, and M. Salyani, 2002 b Spray Applications to Citrus: Overview of Factors Influencing Spraying Efficacy and Off target Deposition. IFAS Fact Sheet HS 851. Stover, Ed., C. Wilson, D. Scotto and M. Salyani. 2004. Pesticide Spraying in Indian River Grapefruit: III. Opportunities for Improving Efficacy and Efficiency while Reducing Off target Deposition. Hort. Tech 14(4):564 574. Summerhill, W.R., J.L. Knapp, J.W. Noling, G.D. Israel, and C.L. Taylor. 1989. Citrus Pest Mana gement. University of Florida, IFAS, Coop. Ext. Ser. PE 6. Svensson, S.A., R.D. Brazee, R.D. Fox, and K.A. Williams. 2003. Air Jet Velocities in and Beyond Apple Trees from a Two Fan Cross Flow Sprayer. Trans. ASAE 46(3): 611 621. Teske, M.E., J.W. Barry, and R.B. Ekblad. 1990. Canopy Penetration and Deposition in a Douglas Fir Seed Orchard. ASAE Paper No. 901019. St. Joseph, Mich.: ASABE. Teske, M.E., N.B. Birchfield, and S.L. Bird. 2004. Development and Validation of a Mechanistic Ground Sprayer Model. AS AE Paper No. 041036. Ottawa, ON. Teske, M.E., A.J. Hewitt, and D.L. Valcore. 2004. Suggested Revisions to ASAE Standard S572 Aug99. ASAE Paper No. 041091. St. Joseph, Mich.: ASABE. Teske, M.E. and R.L. Hill.1995. The Evaporation Rates of Agricultural Spray Material. In Proc. 8th Annual Conference on Liquid Atomization and Spray Systems Troy, MI. Thistle, H.W., K.J. Allwine, C.D. Whiteman, X. Bison, and J.W. Barry. 1996. Validation of the VALDRIFT 1.0 Complex Terrain Pesticide Dispersion Model. ASAE Paper N o. 961079. St. Joseph, Mich.: ASABE. Thompson, N. and A.J. Ley. 1983. Estimating Spray Drift using a Random walk Model of Evaporating Drops. J. Agric. Eng. Res 28:419 435. Threadgill, E. D. and D.B. Smith. 1975. Effects of Physical and Meteorological Para meters on the Drift of Controlled Size Droplets. Trans. ASAE (1975):51 56. Van De Zande, J.C. 2005. IDEFICS. In Waikoloa, Hawaii. Van De Zande, J.C., H.J. Holterman, and M. Wenneker. 2008. Nozzle Classification for Drift Reduction in Orchard Spraying: Identification of Drift Reduction Class Threshold Nozzles. Agricultural Engineering International: the CIGR Ejournal Vol. X.. Manuscri pt ALNARP 08 0013. VanDevender, K.W., T.A. Costello, J.A. Ferguson, B.A. Huey, N.A. Salton, R.J. Smith, Jr., R.S. Helms. 1994. Weed Management Support System for Rice Producers. Appl. Eng. Agric 10(4): 573 578.

PAGE 189

189 Vieri, M. and P. Spugnoli. 1996. Containing Evaporation Losses in Pesticide Application: Mistblower Set up Related to Threshold Meteorological Conditions in Mediterranean Areas. AgEng Conf. Report No. 96A 146. Madrid, Spain Walklate, P.J., J.V. Cross, G.M. Richardson, R.A. Murray, and D.E. Baker. 2002. Comparison of Different Spray Volume Deposition Models Using LIDAR Measurements of Apple Orchards. Biosystems. Eng. 82(3): 253 267. Wallach, D. 2006. Evaluating Crop Models. In Working with Dynamic Crop Models: Evaluation, Analysis, Parameterization, and Applications Ed. D. Wallach, D. Makowski, and J.W. Jones. Elsevier. New York. ISBN 10: 0 444 52135 6: 11 50. Welch, W.J., R.J. Buck, J. Sacks, H.P. Wynn, T.J. Mitchell, and M.D. Morris. 1992. Screening, predicting and computer experiments. Technometr ics 4: 15 25. Whitney, J. and M. Salyani. 1991. Deposition Characteristics of Two Air Carrier Sprayers in Citrus Trees. Trans. ASAE 34(1):47 50. Whitney, J.D., M. Salyani, D.B. Churchill, J.L. Knapp, J.O. Whiteside, and R.C. Littell. 1989. A Field Investig ation to Examine the Effects of Sprayer Type, Ground Speed, and Volume Rate on Spray Deposition in Florida Citrus. J. Agric. Eng. 42:275 283. Willis, G.H. and L.L. McDowell. 1987. Pesticide Persistence on Foliage. Rev. Environ. Contam. Toxicol. 100:23 73. Wolters, A., M. Klein, and H. Vereecken. 2004. An Improved Description of Pesticide Volatilization: Refinement of the Pesticide Leaching Model (PELMO). J. Environ. Qual. 33: No. 1629 1637. Xu, Z.G., P.J. Walklate, and P.C. Miller. 1997. Evaluation of a Sto chastic Model for Spray Transport Prediction from Air Assisted Sprayers. Aspects of Applied Biology 48:195 201. Xu, Z.G., P. J. Walklate, S. G. Rigby, and G. M. Richardson. 1998. Stochastic Modelling of Turbulent Spray Dispersion in the Near field of Orcha rd Sprayers. J. Wind Eng. Ind. Aerodyn. 74 76 : 295 304. Yates, W. E., R. E. Cowden, and N. B. Akesson. 1985. Drop Size Spectra from Nozzles in High Speed Airstreams. Trans ASAE 28: 405 410, 414.

PAGE 190

190 BIOGRAPHICAL SKETCH Peter Ako Larbi was born on 1979 in G hana and has twelve years of progressive and responsible engineering training and experience. His high school educati on which had a concentrated in science, with physics, chemistry, and additional m athematics, prepared him for his engineering training in h is advanced education. He graduated from the Kwame Nkrumah University of Science and Technology (KNUST), Kumasi, Ghana in 2003 with a Bachelor of Science degree with honors in Agricultural Engineering. During his third year (i.e. 2002) he had a three month industrial training at Cocoa Processing Company, Tema, Ghana as a Trainee Engineer attached to the Maintenance Section. He gained a lot of practical experience in handling faults and solving problems within industry. After his first degree in 2003, he wor ked with Trans Ghana Communications Ltd., Accra, for three months as a Marketing Officer while awaiting his National Service and graduation. He gained a lot of experience in human relations due to the wide range of clients he encountered. Peter graduated i n February 2004 while doing his national service at the Technology Consultancy Centre, KNUST, where he worked as a Research Assistant/ Agricultural Engineer for ten months He gained a lot of practical experience in designing and developing food processing equipment, and developed a full sense of responsibility as an engineer. Here he also helped the centre a lot in solving most of their computer software and hardware problems. During his national service, Peter led an international engineering project team to develop the project proposal over nine months and won a Mondialogo Engineering Award (organi z ed by UNESCO and DaimlerChrysler). The team received one of the twenty one awards and also one of t he five Special Jury Recognition Awards at the Mondialogo Symposium held in Berlin, Germany in May 2005.

PAGE 191

191 At the end of his national service, Peter enrolled into a two year masters programme in the Department of Agricultural Engineering, KNUST as one of the pioneers of postgraduate studies in the department in October 2004. During the period of the programme c omplete d a three month Advanced Course in Computer Aided Designing and Drafting (CADD) Engineering at the Centre for the Development of Advanced Comput ing (C DAC), Mohali, India. On his return, he became part of a feasibility project team of three Ghanaian and three Swedish students that studied the development of oil palm value chain in Ghana within the framework of the President Special Initiative on O il Palm. This became the basis of his research for his masters degree. Through hard work and perseverance Peter completed in May 2006, but graduated in June 2007, with a Master of Science degree in Food and Post Harvest Engineering. Meanwhile, having recei ved the award money to initiate the start up of the Jatropha Energy Development Project, Peter led his team to start the implementation in January 2006. He then led the team to further register the company, Sustainable Technologies for Agricultural Resourc es Ltd in November 2006. Peter came to the University of Florida in August 2007 and received his Ph D in Spray Application Technology from the Agricultural and Biological Engineering Department. Peter and his wife, Auriette, have a beautiful baby girl, M ireille.