Citation
Detection and Protection of Triticale Seed from Infestation in a Storage System

Material Information

Title:
Detection and Protection of Triticale Seed from Infestation in a Storage System
Creator:
Khedher-Agha, Mahmoud
Place of Publication:
[Gainesville, Fla.]
Florida
Publisher:
University of Florida
Publication Date:
Language:
english
Physical Description:
1 online resource (101 p.)

Thesis/Dissertation Information

Degree:
Doctorate ( Ph.D.)
Degree Grantor:
University of Florida
Degree Disciplines:
Agricultural and Biological Engineering
Committee Chair:
LEE,WON SUK
Committee Co-Chair:
BUCKLIN,RAY A
Committee Members:
TEIXEIRA,ARTHUR A
TURNER,ALLEN E
BLOUNT,ANN RACHEL SOFFES
Graduation Date:
12/19/2014

Subjects

Subjects / Keywords:
Drying ( jstor )
Eggs ( jstor )
Infestation ( jstor )
Insects ( jstor )
Larvae ( jstor )
Mortality ( jstor )
Rice ( jstor )
Triticale ( jstor )
Wavelengths ( jstor )
Weevils ( jstor )
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
drying -- emc-curve -- germination -- infestation -- mortality -- nir -- sorption-isotherm -- spectral-signature -- water-activity -- weevils
Genre:
bibliography ( marcgt )
theses ( marcgt )
government publication (state, provincial, terriorial, dependent) ( marcgt )
born-digital ( sobekcm )
Electronic Thesis or Dissertation
Agricultural and Biological Engineering thesis, Ph.D.

Notes

Abstract:
The goal of this project was to develop improved methods for protecting triticale seed from infestation during storage. Triticale is the one of the most insect susceptible small grains during post-harvest storage, and serves as a model grain to test methods of insect control in seed storage. This was done through two main goals and one sub goal. The first main goal was to protect triticale seed from infestation using drying methods. This done through finding the effects of three different drying temperatures and three different drying times on seed viability and insect mortality for different storage times. The results showed that total mortality occurred for 8% moisture content and that seed germination was not affected. Also, drying prevented a new generation of insects from emerging due to the low moisture content of the seed. The economic outcome of this research has the potential to significantly lower the cost by as much as 90% compared with chemical treatments for small grain seed storage currently used for insect control. The second main goal, was to develop a method of identifying infestations by detecting and predicting the degree of infestation of triticale seed infested with rice weevils using near-infrared (NIR) spectroscopy. The detection was for 11 degrees of infestation with six growth stages from egg to adult inside the seed. The data were analyzed using several methods such as stepwise multiple linear regression, exhaustive model and regression tree. The results led to the conclusion that the late growth stages could be detected more accurately than early infestation. The model that produced the lowest mean square differences was stepwise. A sub goal was to measure sorption isotherms curves for triticale and use the results to develop a method to dry triticale seed. A best fit equation was developed describing these relationships. The results showed that the modified Chung-Pfost equation represents this relationship most accurately. Overall, this study showed great potential for using drying treatments and near-infrared spectroscopy for protection and detection of the infested seed in the different growth stages of the rice weevil. ( en )
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.
Thesis:
Thesis (Ph.D.)--University of Florida, 2014.
Local:
Adviser: LEE,WON SUK.
Local:
Co-adviser: BUCKLIN,RAY A.
Statement of Responsibility:
by Mahmoud Khedher-Agha.

Record Information

Source Institution:
UFRGP
Rights Management:
Copyright Khedher-Agha, Mahmoud. Permission granted to the University of Florida to digitize, archive and distribute this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.
Resource Identifier:
974007359 ( OCLC )
Classification:
LD1780 2014 ( lcc )

Downloads

This item has the following downloads:


Full Text

PAGE 1

DETECTION AND PROTECTION OF TRITICALE SEED FROM INFESTATION IN A STORAGE SYSTEM By MAHMOUD KAMAL AHMED KHEDHER AGHA A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF T HE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2014

PAGE 2

© 2014 Mahmoud Kamal Ahmed Khedher Agha

PAGE 3

To my f ather, m other, w ife, s isters, and b rothers

PAGE 4

4 ACKNOWLEDGMENTS All prai se is due to the Creator , the Lord of the worlds, the Beneficent, the Merciful, Master of the Day of Judgment . G reat thank s to my advisors , Dr. Won Suk Lee and Dr. Ray Bucklin , for all their support and encouragement that led to the completion of this long journey. I am grateful to them for sharing their knowledge and experience and the pleasure for having be en under their supervision . Also, I would like to thank my committee members , Dr. Teixeira , Dr. Blount and Dr. Turner , for all their help, guidance and support that lead to the completion of this work. Also , I would like to thank Dr. Mankin in the U . S. Department of Agriculture, Agricultural Research Service , Center for Medical, Agricultural, and Veterinary Entomology (USDA , ARS , CMAVE) , in Gainesville, FL for his help and support . W ithout his support this work could not have be en done . T hank s to Ms. Betty Weaver, in USDA, AR S, CMAVE, Gainesville, FL for her help. I would like to show my appreciation to Dr. Paul Armstrong and Dr. Mark Casada USDA, ARS, C GAHR, SPIERU in Manhattan, KS for their help in this research and for sharing their experience . I would like to thank Dr. Bliznyuk for his s tatistic al advice . In the mean time , I want to thank Mr. Chuan , a PhD candidate in Statistic Department , for help ana lyzing part of the data. I would like to thank Dr. Steven Sargent, Ms. Adrian Berry, and Ms. Kim Cordasco, in the Horticultural Sciences Department at UF , for all of their help and support. Thanks to Dr. Donald H. Huber, and Mr. James H. Lee, Jr., in the H orticultural Sciences Department at UF for all their help and support. Thanks also are extended to Mr. Jim Colee in the IFAS Statistical Consulting Unit at UF for his help in design ing the experiment and advice with SAS program ing . I give m any thanks to Ly le Buss, Insect Identification Laboratory, Entomology & Nematology Department,

PAGE 5

5 University of Florida, for all his help with insect photos and identification. I would like to thank Dr. Susana Goggi Agronomy of the Department , Ms . Kim North and Mr . Mike Stah r , Seed Testing Laboratory at Iowa State University , for all their help sharing their knowledge of seed germination. Moreover, I want to thank all the staff, colle a g u es and friends in our Agricultural and Bi ological Engineering Department at University of Florida for their help and support in this long journey. I will not forget to thank my wife and my family for their continuous support and help with this long journey . Finally, great thanks to the Iraqi Ministry of Higher Education and the A gricultural and B iological E ngineering Depar tment at U niversity of Florida for their financial support.

PAGE 6

6 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 8 LIST OF FIGURES ................................ ................................ ................................ ........ 10 ABSTRACT ................................ ................................ ................................ ................... 12 CHAPTER 1 INTRODUCTION ................................ ................................ ................................ .... 14 Triticale Seed ................................ ................................ ................................ .......... 14 Insect Infestation ................................ ................................ ................................ ..... 15 2 OBJECTIVES ................................ ................................ ................................ ......... 18 Main Goal ................................ ................................ ................................ ............... 18 The Objectives of the Research ................................ ................................ .............. 18 3 LITERATURE REVIEW ................................ ................................ .......................... 19 Seed Infestation ................................ ................................ ................................ ...... 19 Insect Detection ................................ ................................ ................................ ...... 19 Spectroscopy Theory ................................ ................................ ....................... 20 Spectrophotometer Measurements Errors ................................ ........................ 22 Sorption Isotherms ................................ ................................ ................................ .. 24 Seed Drying ................................ ................................ ................................ ............ 25 4 DETECTION AND PREDICTION OF TRITICALE I NFESTATION VIA NEAR INFRARED SPECTRAL SIGNATURE ................................ ................................ .... 27 Detection Overview ................................ ................................ ................................ . 27 Materials and Methods ................................ ................................ ............................ 28 Triticale Seed Samples ................................ ................................ ..................... 28 Seed Infestation ................................ ................................ ............................... 28 NIR Spectral Measurements ................................ ................................ ............ 30 Spectral Data Analysis ................................ ................................ ..................... 30 Regression ................................ ................................ ................................ . 32 Classification ................................ ................................ .............................. 33 Results ................................ ................................ ................................ .................... 33 Multiple Linear Regression with Stepwise Model Search ................................ . 33 Exhaustive Search and Predictive Quality of the Linear Regression Models ... 34 Regression Trees ................................ ................................ ............................. 35

PAGE 7

7 Detection Summary ................................ ................................ ................................ 37 5 SORPTION ISOTHERMS FOR TRITICALE SEED ................................ ................ 48 Sorption Overview ................................ ................................ ................................ .. 48 Material and Methods ................................ ................................ ............................. 49 Sampling and Testing ................................ ................................ ....................... 49 Analysis Method ................................ ................................ ............................... 51 Results and Discussion ................................ ................................ ........................... 52 Sorption Summary ................................ ................................ ................................ .. 54 6 THE EFFECT OF DRYING CONDITIONS ON TRITICALE SEED GERMINATION AND INFESTATION ................................ ................................ ..... 59 Seed Drying ................................ ................................ ................................ ............ 59 Material and Methods ................................ ................................ ............................. 60 Monitor and Control Dryers and Environmental Chambers. ............................. 62 Preliminary Drying Test ................................ ................................ .................... 63 Moisture content ................................ ................................ ............................... 64 Seed Germination ................................ ................................ ............................ 64 Mortality Test ................................ ................................ ................................ .... 65 Drying Power Consumption and Cost ................................ .............................. 66 Energ y Statement ................................ ................................ ............................. 66 Methodology ................................ ................................ ................................ ..... 67 Data Analysis ................................ ................................ ................................ ... 68 Results and Discu ssion ................................ ................................ ........................... 68 Moisture Content ................................ ................................ .............................. 69 Seed Germination ................................ ................................ ............................ 69 Insect Mortalit y ................................ ................................ ................................ . 70 Energy Consumption ................................ ................................ ........................ 71 Energy Cost ................................ ................................ ................................ ...... 72 Energy Statement ................................ ................................ ............................. 72 Drying Summary ................................ ................................ ................................ ..... 73 7 CONCLUSIONS ................................ ................................ ................................ ..... 91 Summary ................................ ................................ ................................ ................ 91 Future Work ................................ ................................ ................................ ............ 92 APPENDIX: DRYING COST CALCULATION ................................ ............................... 94 LIST OF REFERENCES ................................ ................................ ............................... 96 BIOGRAPHICAL SKETCH ................................ ................................ .......................... 100

PAGE 8

8 LIST OF TABLES Table page 4 1 Planned and measured degree of infestation ( DI) of rice weevils inside triticale seed. ................................ ................................ ................................ ...... 38 4 2 Growth stages with different infestation percentages. ................................ ........ 38 4 3 S pectral measurement dates of each growth stage of rice weevils inside triticale seed. ................................ ................................ ................................ ...... 38 4 4 Stepwise multiple linear regression output evaluation results for each growth stage. ................................ ................................ ................................ .................. 41 4 5 Stepwise multiple linear regression selection parameters for each growth stages. ................................ ................................ ................................ ................ 42 4 6 Best subsets of size le ss than 5 for each stage. ................................ ................. 44 4 7 Mean values (Standard error) of the RMSEs of the entire cross validation for each size (< 5). ................................ ................................ ................................ ... 44 4 8 Wavelengths selected by regression tree and corresponding RMSEs. .............. 47 4 9 RMSE comparison of different selection methods for each growth stage. .......... 47 5 1 Salt types and corresponding relative humidities at each temperature. .............. 56 5 2 EMC prediction equations constants for triticale ................................ .................. 56 6 1 The age of the rice weevil adults during the experiment and the staring date for each replication. ................................ ................................ ............................ 74 6 2 Drying temperatures, drying time for drying test. ................................ ................ 74 6 3 The effects of different drying temperatures, dryi ng times and moisture content on seed viability and insect mortality. ................................ ..................... 80 6 4 The effect of drying temperature and drying time on triticale seed moisture content (%) wb. ................................ ................................ ................................ ... 81 6 5 The effect of drying temperature and drying time on triticale seed germination (%). ................................ ................................ ................................ ..................... 82 6 6 The effect of drying temperat ure and triticale seed moisture content on insect mortality%. ................................ ................................ ................................ .......... 85

PAGE 9

9 6 7 Tukey Comparison Lines for Least Squares Means ( LSMEAN ) of Treatments for the new generation adults. ................................ ................................ ............ 87 6 8 The effect of drying temperature and drying time on energy consumption (kWh). ................................ ................................ ................................ ................. 88 6 9 The effect of drying temperature a nd drying time on energy cost ($/tonne). ...... 89 6 10 Comparison of infestation control treatments cost vs drying treatments cost. .... 89 6 11 Explanation of the points on the psychometric chart. ................................ ......... 90

PAGE 10

10 LIST OF FIGURES Figure page 3 1 The effect of grain moistur e and temperature on seed and insect growth. . ........ 26 4 1 Triticale seed sample s in a sample holder . ................................ ......................... 39 4 2 Rice weevils Larvae 2 nd instar growth stage inside triticale seed . ..................... 39 4 3 The method and steps to generate triticale seed samples with different growth stages of rice weevils. ................................ ................................ ............. 40 4 4 Absorb ance for triticale seed with different infestation percentages of rice weevils, using average spectral of six growth stages. ................................ ........ 40 4 5 Absorbance for triticale seed with different growth stages of rice weevils, using average spectra of 33 samples for each growth stages. ........................... 41 4 6 Observed vs. predicted DI for triticale seed infested with rice weevils of two different gr owth stages. . ................................ ................................ ..................... 43 4 7 Boxplots of test data RMSE across 1000 rounds of cross validation for different growth stages . ................................ ................................ ...................... 45 4 8 Reflectance of Egg larvae 1 st instar after smoothing. ................................ ......... 46 4 9 Zoomed in Egg larvae 1 st reflectance graph (reflectan ce of wavelength 400 411). ................................ ................................ ................................ ................... 46 4 10 Tree structures for two different growth stages . ................................ .................. 47 5 1 Six desiccating jars with saturated salt solution and three triticale samples each, in an environmental chamber at 35° C . ................................ ..................... 55 5 2 Tri ticale seed samples with saturated salt solution inside desiccating jars . ........ 55 5 3 Residuals for prediction EMC equations for triticale seed. ................................ . 57 5 4 Sorption isotherm curves comparing model predicted values with those observed at te ................................ ................... 58 6 1 The seed dryers . ................................ ................................ ................................ . 75 6 2 Drying curve for triticale seed at three drying temperatures. .............................. 76 6 3 Vacuum seed counter apparatus . ................................ ................................ ....... 77 6 4 Germination paper . ................................ ................................ ............................. 78

PAGE 11

11 6 5 The seed germinator . ................................ ................................ ......................... 79 6 6 Vacuum insect counter apparatus with separation sieves . ................................ . 79 6 7 Distribution of MC for each treatment for drying triticale seed. ........................... 81 6 8 Distribut ion of germination percentages with each treatment of drying triticale seed. ................................ ................................ ................................ ................... 82 6 9 Distribution of germination percentages with each treatment means of drying triticale seed. ................................ ................................ ................................ ...... 83 6 10 Triticale seed after germination with control treatment without drying . ............... 84 6 11 Triticale seed after germination with treatment of drying on 45 ° C for 48 h . ........ 84 6 12 Distribution of mortality perc entages with each treatment means of triticale seed drying. ................................ ................................ ................................ ........ 85 6 13 Distribution of the new generation of rice weevil adults for each drying treatment mean of triticale seed. ................................ ................................ ........ 86 6 14 The seed sample with drying treatments left side, and without drying treatments with emerging of new generation of rice weevil adults on the right side . ................................ ................................ ................................ .................... 87 6 15 Distribution of the energy consumption for each drying treatment means. ......... 88 6 16 Dis tribution of the energy cost ($/tonne) for each drying treatment means. ....... 89 6 17 Psychometric chart with saturated curve represent 20% MC for triticale and drying points. ................................ ................................ ................................ ...... 90

PAGE 12

12 Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy DETECTION AND PROTECTION OF TRITICALE SEED FRO M INFESTATION IN A STORAGE SYSTEM By Mahmoud Kamal Ahmed Khedher Agha December 2014 Chair: Won Suk Lee Cochair: Ray A. Bucklin Major: Agricultural and Biological Engineering The goal of this project was to develop improved methods for protecting triticale seed from infestation during storage . Triticale is the one of the most insect susceptible small grain s during post harvest storage, and serve s as a model grain to test methods of insect control in seed storage. T his was done through two main goals and one sub goa l . The first main goal was to protect triticale seed from infestation using drying methods. This done through finding the effect s of three different drying temperatures and three different drying times on seed viability and insect mortality for dif ferent storage time s . The results show ed that total mortality occurred for 8% moisture content and that seed germination was not affected . Also, drying prevented a new generation of insect s from emerging due to the low moisture content of the seed. The eco nomic outcome of this research has the potential to significantly lower the cost by as much as 90% compare d with chemical treatments for small grain seed storage currently used for insect control.

PAGE 13

13 The second main goal , was to develop a method of identifyin g infestations by detect ing and predict ing the degree of infestation of triticale seed infested with rice weevils using n ear infrared (NIR) spectroscopy. T he detection was for 11 degrees of infestation with six growth stages from egg to adult inside the se ed. The data w ere analyzed using several methods such as stepwise multiple linear regression , exhaustive model and regression tree . T he results led to the conclusion that the late growth stages could be detected more accurately than early infestation. The model that produced the lowest mean square differences was stepwise . A sub goal was to measure sorption isotherms curve s for triticale and use the results to develop a method to dry triticale seed. A best fit equation was developed describing these relatio nships . The results show ed that the modified Chung Pfost equation represents this relationship most accurately. Overall, th is study showed great potential for using drying treatments and near infrared spectroscopy for protection and detection of the infest ed seed in the different growth stages of the rice weevil.

PAGE 14

14 CHAPTER 1 INTRODUCTION Triticale Seed Triticale, Triticosecale , a hybrid of wheat and rye, is a relatively new crop that was developed in 1888 as a potential crop for animal feed. Global product ion of triticale in 2012 was 13.7 million metric tons, cultivated in about 60 countries (FAOSTAT, 2012) . Meanwhile, USA production of triticale in 2012 was 78 thousand metric tons (NASS, 2012) , the dollar value of this production is $94.6 mil lion (Danny Mixon, personal communication, November 14 th , 2014) . Triticale has high protein content as a grain and as forage. Also, triticale forage has a high yield of dry matter. Triticale is vulnerable to grain insect infestations while in storage. Trit icale is infested easily since it has a soft seed coat (Salmon, Mergoum, & Gomez Macpherson , 2004) . Researchers observed that triticale wa s sticky and wa s less resistant to insects than wheat . In addition, storage insects prefer softer grain be cause it is susceptible to attack. And triticale is attractive to insects because of its high protein content. At the North Florida Research and Education Center (NFREC) in Quincy, FL triticale seed ha s been grown to produce varieties that produce a high v egetati ve yield for use as animal feed. Seed needs to be kept viable in a storage area for one or more years. Seed storage in north Florida has many issues, but the main issue in Florida is the hot humid environment that is suitable to grow weevils and at the same time makes the seed more susceptible to mold.

PAGE 15

15 Insect Infestation The NFREC has identified significant insect infestations in their seed storage area, especially with triticale seed. In this center, triticale seed was heavily infested with rice wee vils which are common harmful grain insect. Rice weevils, Sitophilus oryzae (L.) , are one of the main insects that cause wide damage to grain. Weevils are internal feeders and they complete their life cycle inside the seed from egg until adult stage withou t a visual sign, until the insect emerges leaving the empty seed behind. The ideal environment for the insects is between 23°C to 30°C and above 50% relative humidity (RH). Their life cycle time depends on these conditions, and the growth of these insects is exponential within this range of ambient conditions. Each insect can lay ap proximately four eggs a day and within forty days the adult is ready to start a new lifecycle (for instance 100 insects after forty days will produce 400 insects each day for thi s ambient environment). The weather in Florida is considered to be an ideal ambient environment for these insects for almost eight months of the year. Insect infestation of triticale seed causes large storage losses that can exceed 50% according to an indu stry source (Bill Smith, personal communication, November 14th, 2014) (Dobie & Kilminster, 1978) . These poor defense characteristics lead to the thought of protecting these seed by developing a method for early detection. Early detection can reduce infestation losses r emarkably by removing the infested grain. Rice weevil is a common infestation insect in Florida, which is the location of this research , but w eevils are one of most damaging insects to small grain globally. Their lifecycle which develops from egg to the ad ult stage inside the seed in a month or more, depending on the

PAGE 16

16 differentiating infested seed from non infested until the adult emerges from the seed. At that point , the seed is totally damaged and severe damage can spread to the entire storage area. Therefore, the emphasis of this research was to find effective, nonchemical methods to detect and protect triticale seed from infestation. Identifying infestation in early growth stages and at lower infestation percentages prevents large losses and prevents the spread of the infestation to all grain in a storage area. Using near infrared (NIR) spectroscopy to i dentify infestation is one of the potential methods for detecting infest ation. After identifying infestation , a separation needs to be done, where new batch es of seed need to be tested before entering the storage area to give an idea about the infestation percentage and whether to be considered as sound seed or separated as in fested grain. After that, the next step is storing the seed in a manner that provides resistance fr o m any external sources of infestation that can start a new infestation. This can be done using the procedure of drying which prevents new insect s from start ing a new cycle because the low grain moisture content make s it hard for the insect to survive without available moisture. In the other direction, drying must be limited to let the germ survive inside the seed so that it can germinate in the future. Dryin g can be done using many combinations of different treatments to show the effect of low to high drying temperatures. Once the infestation is i dentified , the problem can be fixed using drying method s as will be explained later . Many method s have been used t o identify the damage from infestation. The acoustic method was used by Herrick and Mankin. ( 2012 ) . This

PAGE 17

17 method can identify the larvae stage under the right conditions . Also, an X ray method identified the infestation clearly but was highly cost ly (Mankin & Hagstrum, 2011) . Spectral signature has been used for many crops but has not been done for triticale. After the identifying the infestation, action must be taken to control the infesta tion and to stop the spread to the entire storage area. Before adding a new batch of grain, the storage area should be inspected carefully. Drying the seed could have a wide potential t o help prevent further infestation. The focus of this research was on u sing the spectral signature method to identify the infestation.

PAGE 18

18 CHAPTER 2 OBJECTIVES Main Goal The main goal of this research wa s to find an acceptable method that protect s triticale seed from insects without using chemicals as well as developing meth od s of detecting the infestation without destroying seed. There are two steps for achieving this research goal . The first step is to determine the status of the seed that is going to be stored in order to take appropriate action s based on the seed conditio n. The second step is preventing infestation or control ling the infestation, by preventing insects that are within the seed f rom acting normally ( feeding) by dehydrating the insect through drying the seed its elf to optimum moisture content . In addition , dr ying gives additional resistance to the seed as it protects from other insects and potential mold. The O bjectives of the R esearch The main objectives of the r esearch are first; create a method to determine the infestation degree in the seed based on measu ring the spectral reflectance. Second, e stablish sorption isotherms for triticale seed at three different temperatures, and to develop mathematical expressions to predict equilibrium moisture content (EMC) for the seed at a specified relative humidity (RH) and temperature (T) of the storage environment . Third, find the effects of drying conditions on seed germination and on insect mortality, and determine optimum storage time under optimum drying condition.

PAGE 19

19 CHAPTER 3 LITERATURE REVIEW Seed Infestation T riticale ( X Triticosecale Wittmack ) is a robust, disease resistant wheat rye hybrid that is well adapted to dro ught and difficult soils (Salmon et al., 2004) Modern cultivars provide humans and farm animals a source of essential amino acids and vitamin B (Tsen, 1974) . The production of triticale around the world in 2012 was 1 3.7 million metric tons (FAOSTAT, 2012) . Meanwhile, in the U.S., the production from triticale in 2012 w as 78 thousand metric tons (NASS, 2012) . Moreover, Triticale is a robust, disease resistant wheat rye hybrid that is well adapted to dro ught and difficult soils (Salmon et al., 2004) . Modern cultivars provide humans and farm animals a worthy source of essential amino acids and vitamin B (Tsen, 1974) . However, the combination of a soft seed coat and high protein content makes triticale in storage highly vulnerable to ins ects, including the rice weevil (Dobie & Kilminster, 1978) . These vulnerabilities could be reduce d by developing a method for early detection and targeting of insect infestations in stored triticale. Insect Detection Early detection of rice weevil s, Sitophilus oryzae (L.) is difficult because the larvae feed while hidden inside the kernels and there are no visible external indicators of damage until the adult emerges (as shown in Figure 4 1). Nonvisual methods to detect insects inside grain kernels include acoustic sensing, X ray, nuclear magnetic reson ance, and infrared spectroscopy (Mankin & Hagstrum, 2011) ; (Singh, Jayas, Paliwal, & White, 2009) . For this research, the detection of rice weevil larvae and later stages in triticale was studi ed by applying near infrared (NIR) spectroscopy, which

PAGE 20

20 measures reflectance characteristics of the kernel at different infestation stages. It was o f interest in this study to develop a spectroscopic method that predicts percentage of infestation of different instars in a triticale sample based on measurements of spectral reflectance at wavelengths where the greatest differences between undamaged and insect damaged kernels occur. In previous studies, Peiris et al. (2010) found that undamaged and Fusariaum head blight damaged kernels of wheat could be distinguished by comparing the second derivative near infrared spectra of the kernels in bands at 1160 1220 nm and 1395 1440 nm. They predicted that undamaged and damaged kernels could be distinguished for other small grains as well. Singh et al. (2009) found that starch molecules showed absorption between 1100 1300 nm in wheat kernels. They used a binary classification model to differentiate between undamaged kernels and kernels with damage caused by rice weevil and three other stored product insects . T he accuracy of this method ra nged from 73 to 100%, depending on the insect (Singh et al., 2009) . Siuda, Grabowsk i, Lenc, Ralcewicz, and Spychaj Fabisiak (2010) investigated infrared measurements of starch content for four classes of infestation of Fusarium in wheat: DI C (control), DI 1 (light infestation less than 15%), DI 2 (strongly infested up to 50%), and DI 3 (very strong infestation more than 50%). They found that the thousand kernels weight (TKW), Hagberg falling number (HFN), protein and starch content, and sedimentation value (SV) decreased as the degree (percentage) of i nfestation (DI) increased (Siuda et al., 2010) . Spectroscopy Theory The Beer Lambert Law deals with absorbance, and transmittance of light through materials. The law governs the relation ship between absorbance of light and intensity

PAGE 21

21 by stating that a linear relation exists between absorbance with concentration. There is also a linear relation between the absorbance with path length, as expressed by Equation (3 1). (3 1) Where: A= absorbance, = molar absorptivity (L /mol. Cm), b= path length of the sample (cm), c= concentration of the complex in solution (mol /L). The law is expressed in terms of absorbance because it is easier to deal with than transmittance which has a n exponential relation ship with intensity as shown in Equation (3 2). (3 2) Where: T = Transmittance (%). The Beer Lambert Law for transmittance is given by Equation (3 3). (3 3) The relation given by Equatio n (3 3) is exponential. In addition, the transmittance is the ratio between the intensity of the light after entering the material (I) and the intensity of the light before entering the material (Io) as shown in Equation (3 4). (3 4 ) The law can also be written in alternative ways as shown in Equation (3 5). (3 5) The intensity is the amount of light that absorbed from the materiel as the light goes through for each unit of this material concentration (Kazakevich, 2010) , (Wellbeing) , (Clark, 2007).

PAGE 22

22 Spectrophotometer M easurements Errors To better understand how to analyze the data and to interpret the result s of measurement s , it was necessary to analyze the errors that could affect the outco me data. There are three main sources of errors that the spectrometer could have: instrument errors, sample errors, and operator errors (Williams & Norris, 2001) . First I nstrument E rror : t here are two related important scales that the instrument used, which are the wavelength and the photometric scales. Wavelength scale s in biological materials usually use the narrow band filter to determine a wavelength that is directed to the object or by using gratings . Gratings act like mirror s t hat divide the light into different wavelength s to produce one wavelength each time. L ight has many wavel engths and the instrument use s only one at a time. In order to eliminate the other wavelength s, a filter is used first and then a grating is used . In both methods there are very important factors that need to be kept in the mind which are efficiency, purit y, and stability. The most important one with respect to time is stability. If the wavelength is not stable over time a different wavelength will be produced each time the instrument is operated , which will affect the efficiency of the instrument. In order to overcome variation of in the wavelength, a calibration procedure can be used that can reduce these changes to the minimum. S tability is affected by many factors . I nstrument wear is the first and most important factor. Second, the mounting of the instru ment to eliminate vibration is necessary . If there is any vibration then the stability of the instrument will be affected . T hird, thermal expansion and contraction that change dimension s of the parts inside the instrument affect stability. F o u rth a small a mount of dust inside the instrument can cause an error in the lenses. Meanwhile, temperature inside the instrument affects

PAGE 23

23 the stability of the wavelength . A s the temperature change s by 1°C the wavelength will change s by 0.5 nm (Williams & Norris, 2001) . The photometric scale depends on linearity of the measurement, whe n the source is from stray lights. In other word, n onlinearity occurs from the noise of stray light other than from the light source. Stray lights are innate with the filters and with the grating . Also, light comes from the gap between the sample and the chamber or from any other gaps can reach the detector. To overcome this error, these gaps must be blocked . Also, error from the reflectance from the surface of the sample create s a nonlinearity which is hard to overcome (Williams & Norris, 2001) . There are two type of noise , ra ndom and systematic. The random noise cause d by the detector or leakage can be reduced by using a method called Coadding signals. The c oadding signals method makes the systematic no i se grow. However, s ystematic noise is a big problem f or the data so it mus t be kept as low as possible (Williams & Norris, 2001) . The effect of heat in the room on the instrument: The NIR instrument is very sensitive to change in the surrounding temperature because each part of the instrument is affected by th e s e chang es . It is necessary to keep a fan inside the instrument on and to keep the surrounding room temperature stable (Williams & Norris, 2001) . The Cary instrument has the ability to keep the temperature stable and the operator must turn it on for 30 minute before the test is run . The sample cover : The s ample cover is very important because if the cover i s not clean or there are scratch es in it, there will be error. Also , the type of cover affects the data because the detector can receive wrong reflectance s. For example glass and

PAGE 24

24 Quartz are different in many reflectance properties . Quartz is better because it overcome s these errors by minimize the scattering for the reflectance while the light penetrat e quartz (Williams & Norris, 2001) . Relative humidity: Humidity should be controlled during the calibration. R ecommended humidit ies differ from one instrument to an other because each company uses different method s of calibration and many other factors change. Sample variation : Also, the variability in the insect stages inside the seed cause s a difference in reflectance . Second S ample E rror: Seed methods of sampling are an error source due to variation in the seed size, variation in color, the length of the light path. Also, t here are many issues affecting the result caused by t he sample: the moisture cont ent , foreign material with the seed, and temperature. Third O perator E rror : t here are many factors affecting the result s which include the number of samples, the choosing of the sample, the accuracy of the operator, how the sa mple is stored before and after the preparation, how the sample is loaded , and how the area is cleaned . Sorption Isotherms In this study, the modified Henderson (MHE) , modified Chung Pfost (MCPE) , and modified Oswin equations (MOE) were chosen because the y are widely used to express EMC versus RH relationships for agricultural grains, and are included in ASABE Standard D245.6, Moisture relationships of plant based agricultural products (ASABE 2012a). F statistic, coefficient of determination (R 2 ) and stan dard error (SE) were used to evaluate suitabili ty of the prediction equations. Also, residual plots were considered to assi st comparisons of models (Godbolt, Danao, & Eckhoff, 2013; Igathinathane, Womac, Sokhansanj, & Pordesimo, 2005; Viswanathan, Jayas, & Hulasare, 2003) .

PAGE 25

25 Godbolt et al. (2013); Igathinathane et al. (2005) used nonlinear regression with SAS for the prediction of the equilibrium moisture content (EMC) equatio manner similar to the method used in this study. Khedher Agha, Lee, Bucklin, Teixeira, and Blount (2014) established equilibrium moisture content (EMC) relationships for triticale seed, where they found that the bes t equation that represents the relation between EMC and ERH was the modified Chung Pfost equation. Seed Drying S eed s have some ability to survive under extreme conditions . One of these extreme conditions is high temperature. In order to perform drying , ai r with high temperature need s to flow through the seed . S o drying can a ffect seed viability . Baker, Paulsen, and Zweden (1993) f ound that drying corn seed with temperatures of 40.5/46°C and 35/40.5°C d id not affect the germination percentage which was 98%. Increasing the drying temperature (from 35/40.5°C to 40.5/46°C) increase d d rying capacity by 28%, and increase d energy usage but reduc ed drying time (Baker et al., 1993) . Also, inc reas ing the drying temperature would increase the productivity of the dryer b y reducing the time required to dry grain (Baker et al., 1993) . Baker et al. (1993) conclude d that energy usage was not affected by increasing the drying temperature. Ondier, Siebenmorgen, and Mauromoustakos (2010) conclude that reducing the RH of the drying air reduce d the drying time significantly. Hall (1970) explained the effect of MC and t emperature on safe storage condition s , by using regions a s shown in Figure 3 1 where seed start s to germinate with high MC with above 10°C. The safe region for grain storage will be with low MC and low temperature, wh ich limit the insect population and mol d growth .

PAGE 26

26 Figure 3 1. The effect of grain moisture and temperature on seed and insect growth . Source: D. W. Hall . Handling and Storage of Food Grains in Tropical and subtropical Areas . Rome: FAO, (1970).

PAGE 27

27 CHAPTER 4 DETECTION AND PREDICTION OF TRITICA LE INFESTATION VIA NEAR INFRARED SPECTRAL SIGNATURE Detection Overview Early detection of rice weevil can be difficult because the larvae feed hidden inside the kernels and there are no visible external indicators of damage until the adult emerges (as show n in Figure 4 1). For this research, the detection of rice weevil larvae and later stages in triticale was considered by applying near infrared (NIR) spectroscopy, which measures reflectance characteristics of the kernel at different infestation stages. It was of interest to develop a spectroscopic method that predicts percentage of infestation of different instars in a triticale sample based on measurements of spectral reflectance at wavelengths where the greatest differences between undamaged and insect d amaged kernels occur. Figure 4 2 A shows the size of the larvae 2 nd instar compare d with triticale seed kernel, the color matching between them, and the way that the instar is track ed though it is growing development from an egg near the seed surface to th e 2 nd instar larvae in the middle of the kernel (Figure 4 2 B) . Although there have been many studies for detection of various insect infestations in wheat, no similar studies have been conducted with triticale. An objective of this study was to investigat e the difference between the infe sted and sound kernels using NIR spectroscopy, and also identify the best wavelengths that determine degree of infestation.

PAGE 28

28 Materials and M ethods Triticale Seed Samples The triticale seed , variety Trical 342 , used for this experiment was grown at the North Florida Research and Education Center (NFREC), Quincy, Florida, USA from October to May 2012. This seed was harvested, then thrashed and cleaned at NFREC. The seed was sieved for five minutes using a Ro Tap sieve shaker ( Octagon 2000, 2.36, 1.00, 0.71, and 0.50 mm opening size. Seed Infestation Infested kernels were obtained by adding 1200 adult rice weevils to 350 g triticale seed, hold ing them for 5 d in glass jars in a conditionin 60% 65% relative humidity. Figure 4 3 shows a flow diagram of the procedure that has been followed to prepare the seed infestation to be tested. The jars were covered with a fine screen and two sheets of filter paper to allow the air to exchange without allowing insects to escape from or get in these jars. The rice weevils were obtained from a colony reared at the U. S. Department of Agriculture Agricultural Research Service Center for Medical, Agricultural, and Veterinary Entomology (USDA , ARS , CMAVE) in Gainesville, FL, USA. Then, the adults were removed from the jars and the infested seed (seed with eggs only) was mixed with sound (healthy) seed to create 11 different degrees of infestation (DI), where the mixing was on weight basis between sound and infested seeds. The actual mixing ratios were from 0% to 100% with increments of ten percent, however, when the degree of infestation was measured manually, the actual ratios were different than planned, as shown in Table 4 1 .

PAGE 29

29 Then, 60 g aliquots of each mixing ratio (including 0% DI or control) were placed in transparent plastic containers, each with a cover containing a fine metal screen at its center. The infested seeds were kept in the containers for 40 d ays to create diff erent life cycle stages of egg, larva, pupa, and pre emerge adult (adult inside seed). Those stages represent a storage environment similar to an actual storage situation, where six different stages were grown over time until adults emerged from the seed, as shows in Table 4 2 . The zero percentage was sound seed, representing the experiment control or uninfested seed. The seed samples were infested with 11 different infestation percentages from 0% to 62% with three replications for a total of 33 samples fo r each growth stage. These were replicated six times to produce the six growth stages from egg to pre emergence adult. The sound seed was tested using the same procedure as for the uninfested seed. These samples provided a wide variety of degrees of infest ation that represented the normal infestation conditions as shown in Table 4 2 and Table 4 3. The separation between the stages was done by estimating the stages based on the mean duration of each instar by counting the days after the weevil adults were pl aced with the seed (Perez Mendoza, Throne, & Baker, 2004; Sharifi & Mills, 1971) . The actual percentage of infestation for each category of DI was measured using the manual counting basis procedure (Siuda et al., 2010) , where infested seed was observed visually through counting one by one the emerged adults and the seed that had emerged holes, during the emergence o f the rice weevil adults from the seed at the end of the experiment.

PAGE 30

30 The counting results for both seed holes and emerged adults were compared to determine if two eggs numerical exist ed in one seed . Equation 4 1 was used to calculate the degree of infesta tion (DI) as percentage: (4 1) NIR Spectral Measurements Diffuse reflectance was measured for triticale seed sam ples of 15 g kernels with known infestation percentage placed in a sample holder as shown in Fig ure 4 1. The spectral measurements were conducted using a spectrophotometer (Cary 500, Varian Inc., Palo Alto, CA, USA), with mercury lamp, which have a wide ra nge of wavelength in the visible (VIS) and near infrared (NIR) ranges as light source. An integrating sphere (DRA CA 5500, Labsphere Inc., North Sutton, NH, USA) with an interior coating of polytetrafluoroethylene (PTFE) was attached to the spectrophotome ter. The warm up time of the lamps was 30 min before any measurements and then a reference standard test using a standard sample with PTFE was done each testing day. The wavelength range used for this experiment was from 400 nm to 2500 nm with one nm incre ments. As shown in Table 4 2 and Table 4 3 . The seed was placed on a sample measurement holder, which had a diameter of 38 mm with a quartz glass cover to keep the seed in a fixed position during the test without affecting the measurement, as shown in Fi g ure 4 1. Spectral Data Analysis Spectral data were smoothed using the Savitzky Golay method using 3rd order polynomial in MATLAB R2010a (Mathworks Inc. Natick, MA, USA) to remove the noise from the data. The frame size for the smoothing was set to 41 poin ts. After smoothing,

PAGE 31

31 the data were analyzed using a suite of linear regression and classification methods. Continuous response (percentage of infestation), multiple linear regression (MLR) with variable selection and regression trees (CART) were considered . In the case of discrete responses (using 4 labels for the degree of infestation after binning, where the bin was chose n according to infestation percentages, with zero% infestation as 1st bin, 6.3% to 18.7 % infestation as 2nd bin, 24% to 41.4% as 3rd bi n and 50.3% to 62.5 as 4th bin) . O rdinal logistic regression and classification trees were used to give classification results. Due to the large number of highly correlated reflectance values for neighboring wavelengths used as predictors, dimension reduct ion of the predictor space was conducted to aggregate the reflectance values over 10 wavelength consecutive non overlapping bins. The original wavelengths ranged from 400 to 2500 nm. By taking the mean of reflectance for each 10 wavelengths, 210 resulting reflectance values were acquired (only the last one original wavelength, 2500 nm, being discarded), named as The absorbance curves for triticale seed with different infestation percentages are shown in Fig ure 4 4 using average of 18 spectra from six growth stages and three replications each. In Fig ure 4 4 the water, starch and protein absorbance bands (Pena, 2004) are marked . The curve shows also there is a distinctive difference in the region between 1400 and 1900 nm, caused by the water and starch absorbance bands. A large number of reflectance values showed large differences between the growth stages of rice weevils infesting the triticale seed using an average spectrum of eleven infestation percentages. The wavelengths between 900 to 1150 nm show

PAGE 32

32 moderate differences due to the water and starch absorption bands as shown in Fig ure 4 5 . Larger differences appeared between 1200 to 1400 nm due to the combined effect of the water and starch absorption bands together on the same points. Also, the wavelengths between 1450 to 1850 nm and 1900 to 1980 nm exhibit a big difference among the sta ges, particularly at 1500 and 1900 nm, due to the overlap of the water and starch absorption bands. The starch and water represent about 60% (Pena, 2004) and 13.5%, respectively as the late growth stages consumed more substrate, which contained a high perc entage of starch and water. There was a slight difference among the growth stages for the wavelengths between 2089 to 2200 nm due to the smaller percentage of protein in the seed approximately 12.5% (Pena, 2004) , compared with starch . Regression For each insect growth stage, the response variable was the percentage of infestation while the covariates were the reflectance values averaged across the 10 wavelength bins. Model selection was coupled by fitting the lin ear regression models. A heuristic (stepwise) model was carried out in SAS (Proc GLM SELECT) using a five fold cross validation method to validate the selection represented in the result later on as cross validation the predicted residual sum of squares st atistic (CV PRESS). The l parameter values, SLE = 0.15 and SLS = 0.10, were selected according to values suggested by SAS as a starting point, then a trial and error method was used to obtain the best prediction. The selection results were corroborated using exhaustive model enu meration in R package

PAGE 33

33 'leaps' (http://www.r project.org). The out of sample model performance was assessed using five fold cross validation (Hastie, Tibshirani, & Friedman, 2008) , with random approximate 20% splits (randomly splitting the sample into 5 approximately equal size groups, then using each of the 5 groups as validation data and the other 4 as training data). Regression tree models (implemented in R package 'tree') were validated by the same cross validation technique. Classification To apply classification algorithms, the group categories can be defined by binning the percentages of infestation into categories. Both the 11 original and 4 binned infestation levels were considered in ordinal logistic regression models. Variable selection methods for binned reflectances, as well as dimension reduction approaches based on principal components regression were used to define covariates. Unfortunately, the small number of data points and replications were insufficient to train the classifier to achiev e acceptable out of sample misclassification rate. Because of this, the focus was on regression, rather than classification based procedures . Results M ultiple Linear Regression with Stepwise Model Search SAS (PROC GLM SELECT) was used to search for subset s of predictors achieving high R 2 values and low root mean square error (RMSE) values based on validation data set (Table 4 4). The l arvae 4th instar (L4) yielded a highest value of R 2 of 0.988 and a lowest RMSE value (3.2%), which leads to linear models y ielding lower RMSE values compared with other methods as explained later on. The L4 stage used many prediction parameters as shown in Table 5, where 16 parameters were used. Table 4 4 shows that the late stages of L4, p upae, and adult in was yielded higher R 2

PAGE 34

34 and lower RMSEs than the other early stages. This suggested that there was a relationship between the size of the insect inside the seed and the chemical composition of the seed. As the size of the insects increased, they ate more from the seed, which caused a reduction of the seed mass and changed the relative chemical composition of the seed. This change was detected using the spectral signature. The prediction was more accurate compared with the early growth stages such as egg L1, L2 and L3 that cons umed less from the seed. Similar wavelengths were chosen as parameters in many growth stages as shown in Table 4 5, where the W40 appeared in all stages and other wavelengths such as W213, W217 and W247 appeared in two stages only. It was observed that man y wavelengths were used to produce this prediction, as shown in Table 4 5. The relationships between predicted and observed degree of infestation for triticale seed infested with rice weevils of two different growth stages are shown in Figure 4 6 . Larvae 4th instar produced the best prediction with a high R 2 of 0.988, as shown in Table 4. Meanwhile, the other stages also yielded a good prediction with high R 2 values above 0.94. Exhaustive Search a nd Predictive Quality o f t he Linear Regression Models Functi on 'regsubsets' from 'leaps' package in R was used for variable selection (best subset size < 5) using exhaustive search with aggregated reflectances as predictors. The best models are summarized in Table 4 6, which shows that W40 is a significant predicto r for all stages (except Adult In, with three subset size and egg and larvae 1 st with two subset size). Predictive performance on a left out dataset was assessed using a five fold cross validation with random approximate 20% splits. This was carried out 10 00 times, with

PAGE 35

35 the results summarized in Table 4 7. Also, in this Table many differences were observed between the early verses late growth stages, where the late stages yielded a lower RMSE than early stages when four variables were used. As the subset si ze increased and the growth stage developed, the RMSE decreased. The lowest RMSE value was 4.9% from the larger subset and the last growth stage. Increasing the size of the best subset in excess of four was computationally expensive, since there were 210 b inned reflectances. To mitigate this problem, the number of candidate variables was reduced from 210 to 50 by prescreening the ones that had the greatest correlations (in absolute value) with the response. Again 1000 rounds of five fold cross validation we re performed. Boxplots of the test RMSEs for each stage are shown in Figure 4 7 . It was observed that test RMSE for all stages dropped considerably as the subset size increased from 1 to 4. After that, most stages became stable in terms of the test RMSE. T his corroborates the earlier finding from the stepwise search that the models achieving better out of sample predictions were larger models; that is, the models with ten or so carefully chosen predictors were not overfitting the data. Regression Trees Regr ession tree models were fitted in R (library 'tree') with the percentage of infestation as a numerical response and aggregated reflectances as predictors. The wavelengths used to define splits are listed in Table 4 8. Performance of the tree models is comp arable to that of linear regression models with variable selection. This explanation for the regression tree result shown in Figure 4 10 and in Table 4 8 were presented to show summary details with simplicity.

PAGE 36

36 Table 4 8 shows that the early growth stages h ad a higher RMSE than late growth stages due to the difference in chemical composition as explained previously in multi ple linear regression section. Also, there were some exceptions as the pupa stage had a higher RMSE than the other early stages. This cou ld be due to the characteristic of the pupa stage due to the cocoon materials that affect the reflectance results, which their activities were stopped and that could affect the prediction. Also, Table 4 8 shows that the wavelength W40 (an average of 400 to 410 nm) was selected for all the stages. To take a deeper look at the variables being selected, Figure 4 8 shows the overall smoothed reflectance of egg larvae 1st instar, the lines are average of 3 replications. The curve shows 11 percentages of infestat ion from 0%, until 62%. Figure 4 9 shows the zoomed in graph of wavelength 400 412, which is approximately in correspondence to the W40 in the model selection, regression tree, and best subset selection. It can be see n that curves show an obvious positive relationship between reflectance and infestation percentages. Close inspection of the other stages give the same conclusion and this might also be the reason of W40 being the most significant predictor for all stages. Figure 4 10 shows selected wavelengths for two different growth stages (Egg L1 and p upae), and the degree of infestation predictions were given near the chosen wavelength and in the end of the tree as explained with more details for the RMSE values for regression tree in Table 4 8. Table 4 9 c ompares the performance of different selection methods using RMSE, and shows that the GLM SELECT yielded the lowest RMSE compared with the other methods for all the stages, where the RMSE value was as low as 3.7 for Adult In stage,

PAGE 37

37 and highest was 6.4 RMSE value for Egg L1 stages. It also shows that the second method of selection was the exhaustive search, which produced an RMSE lower than the regression tree method for all the stages except for L2 stage which was the highest. Also, this table shows that th e early growth stages had higher RMSE values than the late growth stages for all the methods that used in this research, this lead to the late growth stages predictions having had less error compared with the early stages. This would be related to the bigg er size of the insect at the late growth stages and the amount of the inner seed materials that had been eaten by the late growth stages. Detection Summary The main aim of this research was to develop prediction models of the infestation degree for tritic ale seed infested with rice weevils of different growth stages using the spectral signature. It was found that heuristic (stepwise) model produced the best prediction model that yielded lowest RMSE and with a high R 2 of 0.988. This model was able to predi ct the infestation well for each growth stage separately. In addition, the prediction model for all the stages yielded an RMSE of 10%. The second method was the exhaustive search that produced RMSE lower than the regression tree method. Different growth s tages affected the selection accuracy, where the late growth stages yielded a higher RMSE than the early growth stages for all the methods and almost all the growth stages. This led to a conclusion that the insect size affected the prediction accuracy. The wavelengths between 400 to 410 nm were selected as a one of few wavelengths for each growth stage and for all the methods.

PAGE 38

38 Table 4 1. Planned and measured degree of infestation (DI) of rice weevils inside triticale seed. Degree of infestation c ategory (DI %) Planned 0 10 20 30 40 50 60 70 80 90 100 Measured 0 6.3 11.4 18.7 24.0 30.8 36.3 41.4 50.3 59.0 62.5 Table 4 2. Growth stages with different infestation percentages. Growth stages Infestation percentage (%) Egg Larvae 1st instar (Egg L1) 0, 6.3, 11.4, 18.7, 24.0, 30.8, 36.3, 41.4, 50.3, 59.0, 62.5 Larvae 2nd instar (L2) Larvae 3rd instar (L3 Larvae 4th instar (L4) Pupae Pre emerge adult Table 4 3. Spectral measurement dates of each growth stage of rice weevils inside t riticale seed. Date Growth s tages 6/29/2012 Adult mixed with sound seed 7/4/2012 Adult taken out (only seed with egg left in cans) 7/7/2012 Egg Larvae 1st instar 7/16/2012 Larvae 2nd instar 7/18/2012 Larvae 3rd instar 7/22/2012 Larvae 4th instar 7/ 25 26/2012 Pupae 8/1/2012 Adult inside seed (pre emerge adult) 8/4/2012 Adult outside

PAGE 39

39 A B Figure 4 1. Triticale seed samples in a sample holder ( Photo courtesy of author , Mahmoud Khedher Agha) . A) Z ero% infestation as a control . B) 62.5% infe station of larvae 2nd instar growth stage of rice weevils. A B Figure 4 2 . R ice weevils Larvae 2 nd instar growth stage inside triticale seed A) shows the size of the larvae vs. triticale seed, B) larvae 2 nd instar. Source : Lyle Buss . Larvae . August 18, 2011. Gainesville, University of Florida Entomology Department .

PAGE 40

40 Figure 4 3 . The method and steps to generate t riticale seed samples with different growth stage s of rice weevils. Figure 4 4 . Absorbance for triticale seed with different infestation percentages of rice weevils, using average spectral of six growth stages.

PAGE 41

41 Figure 4 5 . Absorbance for triticale seed with different growth stages of rice weevils, using average spectra of 33 samples for each growth stages. Table 4 4 . Stepwise multip le linear regression output evaluation results for each growth stage. Stage R 2 CV PRESS RMSE Number of predictors Egg L1 0.945 1219.7 6.4 14 L2 0.951 1010.9 6.0 14 L3 0.947 1274.1 6.1 13 L4 0.988 357.2 3.2 16 Pupa 0.957 906.8 5.1 10 Adult In 0.976 46 3.3 3.7 9 All Stages 0.725 28523 10.9 16

PAGE 42

42 Table 4 5 . Stepwise multiple linear regression selection parameters for each growth stages. Stages Parameters Egg L1 W40 W41 W65 W80 W114 W143 W145 W197 W153 W196 W199 W207 W217 W222 L2 W40 W41 W61 W62 W68 W 88 W108 W109 W113 W114 W116 W133 W194 W217 L3 W40 W41 W45 W46 W71 W72 W79 W88 W212 W237 W239 W246 W247 L4 W40 W41 W43 W47 W51 W55 W97 W100 W104 W107 W113 W122 W125 W128 W137 W247 Pupa W40 W49 W57 W59 W74 W76 W156 W189 W197 W209 Adult In W40 W47 W108 W1 16 W117 W122 W127 W132 W133 All stages W40 W41 W42 W54 W57 W59 W64 W137 W146 W202 W210 W211 W213 W215 W240 W244

PAGE 43

43 A B Figure 4 6 . O bserved vs . predicted DI for triticale seed infested with rice weevils of two d ifferent growth stages . A ) Egg L1 . B ) Pupae growth stage . F ive fold cross validation was used with the Ste pwise S elect method, where the straight line represent s the prediction line and the dashed line is a diagonal line .

PAGE 44

44 Table 4 6 . Best subsets of size less than 5 for each stage. Subset si ze 1 2 3 4 Egg Larvae 1 st W40 W46 W48 W40 W103 W113 W40 W79 W103 W113 Larvae 2 nd W40 W40 W41 W40 W74 W78 W40 W41 W75 W78 Larvae 3 rd W40 W40 W41 W40 W106 W112 W40 W86 W107 W112 Larvae 4 th W40 W40 W41 W40 W61 W89 W40 W61 W103 W113 Pupae W40 W40 W193 W40 W157 W197 W40 W48 W159 W190 Adult In W40 W40 W51 W93 W107 W112 W40 W78 W107 W112 Table 4 7 . Mean v alues (Standard error) of the RMSEs of the entire cross validation for each size (< 5). Subset size 1 2 3 4 Egg Larvae 1 st 15.0 (0.40) 13.0 (0.35) 11 .7 (0.41) 10.4 (0.39 ) Larvae 2 nd 18.3 (0.65) 15.8 (0.64 ) 13.8 (0.51 ) 13.1 (0.63) Larvae 3 rd 14.3 (0.34) 13.1 (0.38) 12.0 (0.31 ) 10.6 (0.38 ) Larvae 4 th 11.9 (0.23) 10.0 (0.24 ) 8.7 (0.18 ) 7.4 (0.16) Pupae 14.2 (0.33 ) 9.6 (0.26 ) 8.0 (0.19) 6.0 (0 .10 ) Adult In 8.6 (0.15 ) 6.5 (0.11 ) 5.7 (0.09 ) 4.9 (0.09)

PAGE 45

45 Figure 4 7 . Boxplots of test data RMSE across 1000 rounds of cross validation for different growth stages, where X axis is the subset size and Y axis is the RMSE: a) Egg L1, b) L2, c) L3, d ) L4, e) Pupae and f) Adult In .

PAGE 46

46 Figure 4 8 . Reflectance of Egg larvae 1 st instar after smoothing . Figure 4 9 . Zoomed in Egg larvae 1 st reflectance graph (reflectance of wavelength 400 41 1 ).

PAGE 47

47 A B Figure 4 10 . Tree structures for two different growth stages , A ) Egg L1, B ) Pupae. Table 4 8 . Wavelengths selected by regression tree and corresponding RMSEs. Stage Wavelengths selected RMSE Egg Larvae 1 st W40 W41W86 W70 10.5 Larvae 2 nd W44 W69 W40 W218 12.7 Larvae 3 rd W40 W78 W82 W242 10.5 La rvae 4 th W40 W197 W56 W139 7.6 Pupae W40 W111 W233 W88 11.2 Adult In W40 W97 W44 7.0 Table 4 9 . RMSE c omparison of different selection methods for each growth stage. Method of prediction GLM s elect Regression t ree Subset s election Size 4 Stages R MSE RMSE RMSE Egg L1 6.4 10.5 10.4 L2 6.0 12.7 13.1 L3 6.1 10.5 10.6 L4 3.2 7.6 7.4 Pupa 5.1 11.2 6.0 Adult In 3.7 7.0 4.9

PAGE 48

48 CHAPTER 5 SORPTION ISOTHERMS FOR TRITICALE SEED Sorption Overview Triticale, Triticosecale , is a hybri d of wheat and rye. Triticale in Florida is grown mainly for forage used as animal feed. The University of Florida's North Florida Research and Education Center (NFREC) in Quincy conducts long term research using triticale as winter season forage appropria te for beef cattle, particularly if combined with ryegrass. However, drying and storing seed in the warm, humid climate of North Florida is problematic because seed viability is reduced at the high drying temperatures, which are required to dry grain under conditions of high humidity. At the same time, the combination of high humidity and warm temperatures during storage provides an optimum climate for ra pid growth of mold and insects. Determining and controlling optimum drying and storage conditions can he lp redu ce these losses. Little information is available in the literature relating to drying and storing triticale. A sorption isotherm is a curve on a graph showing the relationship between equilibrium moisture content of the grain (EMC) and relative humi dity (RH) at constant temperature. It is influenced by temperature for most seeds. This curve is an important physical property of the seed, which indicates the target level of moisture content that must be reached during drying in order to achieve the equ ilibrium relative humidity needed to reduce losses from mold growth and insect infestation, while maintaining seed viability. The objective of this study was to establish the sorption isotherms for triticale seed at three different temperatures, and to dev elop mathematical expressions to predict equilibrium moisture content (EMC) at a specified relative humidity (RH) and temperature (T) of seed.

PAGE 49

49 In this study, the modified Henderson (MHE) , modified Chung Pfost (MCPE) , and modified Oswin equations (MOE) were chosen because they are widely used to express EMC versus RH relationships for agricultural grains, and are included in ASABE Standard D245.6, Moisture relationships of plant based agricu ltural products (ASABE 2012a). F statistic, coefficient of determina tion (R 2 ) and standard error (SE) were used to evaluate suitabili ty of the prediction equations. Also, residual plots were considered to assi st comparisons of models (Godbolt et al., 2013; Igathinathane et al., 2005 ; Viswanathan et al., 2003) . Godbolt et al. (2013); Igathinathane et al. (2005) used nonlinear regression with SAS for the prediction of the equilibrium moisture content ilar to the method used in this study. Material a nd Methods Sampling and T esting Triticale seed samples of variety Trical 342 , harvested in May 2012, were obtained from the NFREC. The seed was clean ed using standard U. S. sieves. Manual inspection was also used to insure the purity of the seed from foreign materials. A static method was used in this experiment to measure the EMC, where the seed samples were randomly distributed into six desiccating jars with three samples (replications) in each jar ( Figura and Teixeira, 2007) as shown in Figure 5 1, and Figure 5 2 . The dimensions of the glass desiccation jars were 24 cm inner diameter and 23 cm inner height. Then, six saturated salt solutions ( Greenspan, 1977) were placed in the jars to control RH. The ASTM standard method of using saturated salt solutions to control RH in a small chamber was chosen for this work (ASTM, 2012) ; Greenspan, 1977). This approach is the method most frequently cited for studies using the static method in a recent review of the literature involving equilibrium moisture content relationships of rice

PAGE 50

50 grain (Choi, Lanning, & Siebenm orgen, 2010) and is the basis of most of the studies used to develop the information provided in ASAE D245.6, (2012a). The salts were laboratory grades as listed in T able 5 1, and were used to provide RHs from 11% to 84% at 23°C (Greenspan, 1977). After s eed samples were in place, jars were covered with a glass cover and sealed with a silicone based lubricant (Dow Corning corp. Midland, MI. 48640 USA) to keep the seed in a controlled RH environment. The seed samples and the saturated salt solutions were ke pt in the jars until each sample reached equilibrium with the controlled environment within the desiccator initial and moisture contents (MC) for the 1 st run, 2 nd run and 3 rd run were 23°C with 13.05% ±0.09%, 5°C with MC 13.35% ±0.02% and 35°C with MC 13.67 ± 0.02%, respectively. The samples were weighed frequently until mass differences between measurements taken a week apart were less than 0.01 g. Very little time elapsed when the jars were opened and closed to remove and return samples used for weighing and disturbances to the RH in jars were brought back to equilibrium by the salt solution. The time to reach equilibrium for the three temperatures was an average of 57 days , (65 days for 35°C and 69 days for 5°C but only 36 days for 23°C). This experiment was repeated t effect of different temperatures on the EMC. The runs with 5°C and 35°C were done inside environmental chamber to control the temperatures, as shown in Figure 5 1. The temperatures for the first, second and third runs were 23° C ± 1.0° C, 5° C ± 0.5° C and 34.5° C ± 0.5° C, respectively. It took a total of nine months to conduct the three runs with the three temperatures. All mass measurements were made with an electronic

PAGE 51

51 balance (M ettler AE 200, Mettler Toledo Inc., Columbus, OH.) with a capacity of 205 g, readability of 0.1 mg, and reproducibility of ± 0.1 mg as shown in Figure 5 2 B . After the samples reached equilibrium, the moisture contents (MC) were measured using the same pro cedure for wheat given by the ASAE Standard (2012b) using a natural convection oven [130° C for 19 hr.]. In addition, data loggers located both inside and outside the jars (HOBO RH/Temp H08 003 02, Onset Computer Corporation, Bourne, MA, USA) were used to monitor the temperatures and to check the RHs randomly. The time interval for the reading was 30 minutes. Analysis Method reference (Greenspan 1977) and the MC data measured us analyzed using Psi Plot (Poly Software International.com, Pearl Rive r, NY) and SAS (version 9.2, SAS Institute Inc. Cary, NC, USA). Modified Henderson, Modified Chung Pfost and Modified Oswin equations , as shown in the Equations (5 1 ), (5 2 ) and (5 3 ), respectively , were used to develop EMC vs. RH curves for triticale seed compatible with the ASABE Standard (2012a) Modified Henderson equation, (MHE) Modified Chung Pfost equation (MCPE) ( 5 2)

PAGE 52

52 Modified Oswin eq uation (MOE) ( 5 3) Where: RH = relative humidity (decimal), T = temperature (°C), MC D = dry basis moisture content (%) and A, B, C = constant parameters to be determined from the shape of the isotherms. Dat a were examined, tested and evaluated using standard error (SE), coefficient of determination obtained from the scatter plot for the predicted vs measured RH (R 2 ),residuals and F statistic, in a manner similar to the methods used by Igathinathane et al. (2005); Iguaz and Virseda (2007); Viswanathan et al. (2003) and ). Results and D iscussion The numerical values found for each of the constant parameters needed in each equation are shown in Tabl e 5 2. This table also presents the evaluation values for each equation, where MCPE gives the lowest SE and the highest F test and the scatter plot for the predicted vs measured gives the highest R 2 value, compared with the other equations. The non liner regression model provided the constants values for the three equations and that help develop ed EMC vs. RH curves . The Modified Henderson equation , Modified Chung Pfost equation and Modified Oswin equati on for triticale seed are shown in the Equations (5 5) and (5 6), respectively, where were compatible with the ASABE Standard (2012a) Modified Henderson equation, (MHE)

PAGE 53

53 Modified Chung Pfost eq uation (MCPE) ( 5 5 ) Modified Oswin equation (MOE) ( 5 6 ) Those results in Table 5 2 sho w that the MCPE equation produced lower error values than the other two equations. Also, T able 5 2 shows agreement among the SE and scatter plot R 2 and F statistic. The MCPE equation represents the relationships between EMC, T and Equilibrium Relative Humi dity (ERH) with the highest accuracy. The residuals for these equations were examined and found to be random. The highest and lowest residual values for MHE, MCPE, and MOE were (7.39%, 5.59%), (5.54%, 6.01%) and (6.28%, 5.74%), respectively, as shown i n F igure 5 3 . The MOE model yielded the second lowest values of 3.08 and 0.988 for SE and the scatter plot R 2 , respectively. The observation vs. prediction curve for triticale seed using The MHE and MOE yielded scatter plot R 2 values of 0.987 and 0.988, r espectively using the observed and predicted values for all the three temperatures, which indicates that these equations also represent the relation between RH and MC well. The scatter plot for the observed vs. predicted for the MCPE relationship yielded the highest the R 2 of 0.993, indicating that the representation by MCPE prediction equation gives the highest accuracy. As shown in F igure 5 4 , MHE produces reasonable predictions, where the prediction curves for 23°C and 35°C fell on the observation poin ts and the 5°C curve is near the observation points. The EMC curve for triticale seed using MCPE shows that

PAGE 54

54 the prediction curve falls on or very near the observation points for the three temperatures as shown in Figure 5 4 . The middle point in the Isother m curves was the starting point for the initial MC, where 13.35% was an average initial MC for three runs, as shown in the gray solid horizontal line in F igure 5 4 . So there is desorption from the 13.35% MC to the lower end of the curve, and adsorption fro m the initial MC to the upper part of the curve. The isotherm curves are divided into two parts. The first is the desorption region, where the drying process is governed by the lower part of the curve. The second is the adsorption region, where the hydrat ion process is governed by the upper part of the curve. The three model curves are smooth because mathematical models are represented by inherently smooth curves. The observed data do not contain enough points through the mid range to judge the smoothness at the adsorption desorption transition point. However differences between adsorption and desorption in the range of these data have been reported to be sufficiently small that it should not significantly affect the use of the results. Li (2012) conclude statistically significant differences between sorption and desorption at lower ranges of the absorption desorption curve, but not at middle and higher ranges for several varieties of wheat. Sorption Summary The relationships between equilibrium moisture co ntent and relative humidity of triticale seed were measured and then presented as sorption isotherms for different temperatures using non liner regression for prediction the EMC curve .

PAGE 55

55 Figure 5 1. Six desiccating jars with saturated salt solution and three triticale samples each, in an environmental chamber at 35° C ( Photo courtesy of author , Mahmoud Khedher Agha) . A B Figure 5 2 . Triticale seed samples with saturated salt solution inside desiccating jars ( Photo s courtesy of author , Mahmoud Khe dher Agha) . A) Z oom inside the jar . B) T he jars and the electronic scale during the measurements.

PAGE 56

56 Table 5 1 . Salt types and corresponding relative humidities at each temperature. Table 5 2 . aluation parameters. EMC Equation SE Scatter plot R 2 F statistic A B C Residual classification Modified Henderson 3.14 0.987 5,759 9.81E 06 2.509 69.4585 Random Modified Chung Pfost 2.39 0.993 10,000 701 0.1815 59.6692 Random Modified Oswin 3.08 0.98 8 5,988 15.3462 0.0707 3.591 Random

PAGE 57

57 Figure 5 3 . Residuals for prediction EMC equations for triticale seed. -8 -4 0 4 8 0 10 20 30 M Henderson -8 -4 0 4 8 0 10 20 30 Residual ERH (%) M Chung Pfost -8 -4 0 4 8 0 10 20 30 EMC (% d.b. ) M Oswin

PAGE 58

58 Figure 5 4 . Sorption isotherm 0 5 10 15 20 25 30 0 0.2 0.4 0.6 0.8 1 EMC ( % d.b.) Modifed Handerson 0 5 10 15 20 25 30 0 0.2 0.4 0.6 0.8 1 EMC ( % d.b.) Modified Chung Pfost 0 5 10 15 20 25 30 0 0.2 0.4 0.6 0.8 1 EMC ( % d.b.) ERH (decimal) Modified Oswin Observed 5°C Observed 23°C Observed 35°C Predicted 5°C Predicted 23°C Predicted 35°C

PAGE 59

59 CHAPTER 6 THE EFFECT OF DRYING CONDITIONS ON TRITICALE SEED GERMINATION AND INFESTATION Seed Drying Seed drying is a critical process since the seed s em bryo must to be protected from high temperature and dehydration that cause reduced viability of the seed. Triticale seed is very sensitive to the rice weevils infestation and therefore i t worked w e ll as a model grain to study insect control protection . I ns ects are also affected by drying operation because reduc ing the moisture content of the seed lead s to reduce d relative humidity around this seed that cause s dehydration of the insect. Rice weevil s live near the seed all the time and use the seed as a sourc e of food for all growth stages from an egg to a pre emerged adult . A nd , after emerg ence , the adult s live on or in the seed most of the time. The direct contact between the insect and the seed for all the insect life cycle s is the key reason for choosing d rying as a method to stop or prevent infestation. The difference between the insect and the seed is that the embryo ha s the ability to survive due to its inactive state , while the insect adult is active and crawls outside of the seed . The concept of equilibrium moisture content (Sorption Isotherm ) is fundamental to the drying operation where low RH air that contact s the seed and reduce s the MC of the seed . The drying process is governed by the characteristics of the seed and the time required to equi librium. Ondier et al. (2010) conclude d that drying rough rice with low temperature and low R H (drying range between 26 34°C, and 19 47% RH) did not affect the quality and the characteristics of the rice. The same drying concep t govern s both the insect and the seed . T he relationship between the ERH and EMC for triticale is shown in Figure 5 4.

PAGE 60

60 For example, after drying the seed to 8% MC, the ERH for the air was 19.2%, according to the Modified Chung Pfost equation (Equation 5 5) . This low RH is an extreme environment for the insect to live in and leads to insect dehydration. The RH is the ratio between the vapor pressure of the water vapor in the air to the vapor pressure of the water vapor at saturated air at same atmospheric pr essure and temperature (Brooker, Bakker Arkema, & Hall, 1992) . The difference between the ambient vapor pressure and the vapor pressure at saturation is the gradient that drives the drying process. At low RH conditions, the large difference s in ambient and saturated vapor pressures cause moisture to be removed from the high moisture content eggs and bodies of insects causing them to dehydrate and die. Beckett (2011) conclude d in his review that in order to protect storage product s , drying or cooling the product s is ultimately neede d. Hertert and Burris (1989) Material and Methods The Trical 342 Triticale used in the experiment was harvested, thrashed, and cleaned in the NFREC in Quincy FL. on May 2012. The seed was kept in a cold storage facility at 5° C until used in this experiment. T he seed was cleaned for the second time using an air cleaner and the seed samples were mixed to be uniform before starting the experiment in 2014. The samples were kept in room temperature for a few days before drying experiment was started . Rice weevils w ere reared at the USDA, ARS, CMAVE to provide the same age adults. Their age s range d between five to seven days , a nd each replication had the

PAGE 61

61 same day age. The age of these adults at the time of the experiment w as between 54 to 70 days, where each replicat ion had a certain age differen ce , as shown in Table 6 1. These insects were kept in a control led chamber with 23° C and 55% RH from emerging until the day of the experiment began . Two gravity convection incubators (Blue M, Model 100A, Thermal Product Solut ions (TPS), New Columbia, PA) were used as a low temperature dryer s for this experiment , as shown in Figure 6 1 . A Proportional integral derivative (PID) controller (Mypin temperature controller, Model TA4 RNR, China) was used to control the temperatures i n the incubators , as shown in Figure 6 1 , using J type thermocouple s (Omega Engineering Inc. , Stanford, CT) as temperature sensors for the controller. The thermocouple sensors were placed in the middle of the drying basket (tray) inside the seedbed . An i n line duct fan (model DB204, Suncourt Inc. Durant, Iowa) was mounted on top of the incubator with a plastic holder to exhaust the humid air from the incubator. The dimension s of the cylindrical fan duct w ere 9.5 cm in diameter and 15 cm in length. The air f low was an average of 1.5 m 3 / s. A two kg sample of triticale seed was placed in a basket . The basket was made from perforated screen with dimension s of 27.5, 28, and 8 cm for width, length and depth, respectively, and a 2 mm screen square opening s . The air was forced through the seedbed using the fan and polyester foam around the screen , as shown in Figure 6 1B . The depth of the seedbed was 2, 2.7 and 3.5 cm for different drying time of 24, 48 and varies (12, 36 and 96 hr.) h r., respectively. Three drying t emperatures of 35°C, 40°C and 45°C were used to dry the seed with different drying time of 12, 24, 36, 48 and 96 Hrs.

PAGE 62

62 Monitor and Control Dryers and Environmental C hambers. The drying experiment was monitor ed using data loggers (Lascar Electronics , Model EL USB 2 LCD+, Erie, PA) placed inside seedbed during the drying operation and in the air inlet of the dryer s . The logger record ed both temperature and RH at 5 minute interval s . Two of the environmental chamber s were also monitored using the same data log gers that were used in the drying test. Two environmental chambers were used as a storage area for the samples after drying. The first chamber , named ow H umidity C h (LH chamber) , was kept at 23°C and 35% RH to study the effect of a l ow humidity en vironment on seed germination and insect mortality, while keeping the seed after drying at the same condition without gaining moisture from the environment. Temperature in t he LH chamber was controlled using a temperature PID controller (Extech Instruments Corporation, model 48VFL, Nashua, NH) and using two light bulbs ( 60 watts ) as heating element s covered with aluminum foil to keep the place without direct light. Glycerin (vegetable glycerin , Dudadiesel.com, Madison , WI S ) with a 99.7% concentration was us ed to control RH inside the LH chamber. Also, totally dried seed (at 130°C for 19 hr.) was used to help keep stabilize RH inside the LH chamber. The second c h amber , named igh H umidity C h amber (HH chamber) , was kept at 23°C and 55% RH to study the effect of high humidity environment on seed germination and insect mortality, while keeping the insect s at optimum condition s without environment al effect s . T he t emperature of t he HH chamber was controlled using a digital temperature controller (AGPtek Controlle r STC 1000, China) using one light bulb ( 10 0 watts) as a heating element covered with aluminum sheet to keep the chamber without direct light. A d igital humidity controller (model WH8040, Willhi.com ,

PAGE 63

63 C hina ) was used as a RH controller inside the HH chamber . The RH controller operate d a small air pump (Metaframe , model Hush II, Japan) that pump ed air through deionized water kept in a glass flask (Pyrex boiling filtering flasks flat bottom long neck 1000 Ml, USA). A d ata logger (CR10x Campbell Scientific, In c. Logan, Utah 84321) was used to monitor the temperature of the germinator, driers, inlet air, LH c h amber, and HH c h amber. The temperature sensors used to monitored the air temperature with this logger were a type T thermocouple s with special limits of er ror (SLE) and 24 AWG ( Omega Engineering Inc. , Stanford, CT). These thermocouples were placed inside the dryer (under , inside, and above the seedbed ), in the inlet of the air, in the germinator, and in LH c h amber to monitor the temperature. The power consu mption was monitored using an electricity usage monitor (Kill A Watt, model P4400.01, P3 International corporation, China) that recorded the kWh for each drying operation to measure the electrical usage difference for each temperature. The electrical usage was measured for the dryer and the fan attached to the dryer, which was record ed during each drying period to monitor the power consumption. Preliminary Drying Test A drying test was done to find the drying time that g a ve desire d moisture content for eac h treatment. This was done by operating the dryer at each drying temperature and measur ing the MC frequently between 4 to12 h to find the drying time corresponding to the desired MC . The drying curve s for triticale seed for the three drying temperatures 35 , 40 and 45° C are s hown in Figure 6 2 .

PAGE 64

64 Moisture content The d irect method was used to find seed MC, using (Blue M, Model OV 18SA, Thermal Product Solutions (TPS).com, New Columbia, PA) a convection oven follow ing the procedure for wheat given by the ASA E Standard (2012b) to dry the seed at 130°C for 19 h . A d igital scale (Mettler PE 160, Hightstown, New Jersey made in Switzerland) with three digit readability , was used to measure the mass of samples. A d esiccator with dimensions (H x L) of 30.5 x 30.5 cm (Nalgene, Acrylic Desiccator Cabinets) with active Drierite (W.A. Hammond Drierite Co., Xenia, Ohio) as a desiccant, was used to keep the samples from reach ing room temperature without gaining moisture from the air. Seed Germination Seed germination test s were done using the procedure provided by the Association of Official Seed Analysts ( AOSA ) (AOSA, 2013) . Germination paper (Anchor paper, Hudson, WI) was used as a substratum to provide the optimum amount of moisture and air to the seed needed to germinate. The dimensions of the paper were 25.4 cm wide, 38.1 cm long, regular weight seed germination paper . The germination papers were s ubmerged in water then pressed by a wood roller on a tilt board to keep the minimum amount of water in the paper for required to provide the seed with optimum amount of air and moisture. The seed was counted using a vacuum apparatus with 50 small holes in a plastic board with two rows of 25 holes each arranged evenly and making a zigzag arrangement to avoid having parallel holes, and providing more space to the seed to grow. This apparatus is shown in Figure 6 3 . Two germinat ion paper s were placed under the seed, and then the seed was placed on the lower third of the paper, which was covered with another sheet. Then they were rolled together, as this arrangement is shown in Figure 6 4 . A plastic rubber band

PAGE 65

65 was placed around the lower end, and keeping the se ed from falling down. The paper was marked using i ndelible pencil ( Blu Blak Noblot Indelible 740 Med by Faber Castell . com). The roll was placed inside a plastic container covered with a plastic bag to prevent evaporation. The container was divided into 15 sections using fishing line, and was placed in a germinator (Seedburo Equipment Com. Model 548, Lab Tech Inc. Crystal Lake IL, 60039) as shown in Figure 6 5, keeping the seed in 15 to 20°C using ice cubes inside the germinator, which were changed regularl y. After seven days, the seed was counted manually to find the germination percentage. According to the AOSA recommendations, the germinated seed was inspected to find the normal vs. irregular germination, and each sample was repeated twice to be a total o f 100 seeds. For each germination result, a photo was taken for a record. Mortality T est One hundred adult rice weevils were placed with 200 g of dried seed in a transparent plastic container to measure the effec t of seed drying on rice weevil mortality. The containers ha d A fter placing the seed with the adults , t he container was kept in a controlled environmental chamber at 23°C . T he treatment samples were kept at 35% RH and 55% RH for the control sample s. Each week , mortality test s w ere conducted using two set s of U . S . standard sieves with 1 and 2 mm opening s for separatin g the insects from the seed. L ive insect s were counted manually using the vacuum insect counter apparatus as shown in Figure 6 6 . The dead insects were count ed manually using tweezers . They were placed on a white paper and monitored for 5 to 10 minutes to find if some insect s were act ing dead . After that the percentage of mortality was calculated , a photo was tak en for each sample for a record.

PAGE 66

66 This procedure was repeated each week for each sample with insects. After six weeks , the mortality percentage was calculated . During the same time period live insects were also counted and monitor ed to determine if a ny new generation s occurr ed . Drying Power Consumption and Cost The power consumption was monitored using an electricity usage monitor (Kill A Watt, model P4400.01, P3 International corporation, China) that recorded the kWh for each drying operation to measure the electrical usage difference for each temperature. The electrical usage was measured for the dryer and the fan attached to the dryer, which was record ed during each drying period to monitor the power consumption. The cost of the power was calculated using the Gainesville Regional Utility (GRU) fact sheet for calculation electric bill. The calculation include d energy use tier 2 with $0.042/kWh, electric fuel adjustment with $ 0.078/kWh , Florida gross receipts tax with 2.5% ( 0.025641 ) and Gainesville electric utility tax with 10%. This was done for 2kg of triticale seed to find the cost for t on of seed the mass was converted from kg to tone, to be the cost value ($/ton ne ). Energy Statement An analysis was conducted to evaluate the economic feasibility of drying triticale see d at low temperature. It was assumed that drying was done with air heated with propane as the fuel source. The fan ca pacity assumed to be 283 m 3 /min ( 10,000 cfm) and the initial air conditions entering the dryer were assumed to be 25°C and 79% for temperat ure and RH, respectively. These conditions are averages based on the weather conditions recorded at the Florida Automated Weather Network (FAWN)

PAGE 67

67 weather station (IFAS UF, 2014) for the typical harvest time of mid to late May in Quincy FL where triticale se ed used in this study was grown. The saturation line for the air exiting the seed after drying was estimated fro m the EMC curve from Figure 5 4 , represents by purple line with three points as shown in Table 6 11 , and Figure 6 1 7 . The RHs for the three dif ferent temperatures at desired MCs were read from Figure 5 4, for triticale seed assumed to be 20% MC at the time of harvesting. The horizontal line intersection with20% MC in Figure 5 4, for three temperatures of 5 °C , 23 °C, and 35 °C gives three RHs, 76 % , 80 % , and 82%,. By assuming that the air exiting the seedbed was in equilibrium with the seed, the line connecting these points gave the saturation curve at 20% MC based on the recommended procedures given by Brooker et al. (1992) and Loewer et al. (1994) . The air was assumed to be heated to 45°C before entering the seedbed as shown as a blue line in Figure 6 17 . Moisture was removed from the seed by the drying air as an adiabatic absorption process following the saturation line for 20% MC until air exite d the seedbed as shown as a red line in Figure 6 17. The energy needed to remove the amount of water required to lower the seed moisture content to 8% was calculated based on enthalpies from the psychometric chart. Calculations of drying costs are shown in the Appendix. Methodology The effect of drying temperature s , drying time s , and moisture content s on the seed germination index and insect mortality w ere observed. An i ncubator was used as a small seed dryer after some modifications to provide better ai r movement and to hold 2 kg of seed for drying. The seed was dried at three different temperatures (35, 40, and

PAGE 68

68 45 °C), as shown in Table 6 3, to test the effect of drying triticale seed on seed germination and insect mortality. The design for the drying e xperiment was a completely randomized design (CRD) with split split plot design with three factors of drying temperature, drying time, and moisture content, containing three replications with and without insects. The experimental unit was o ne hundred adult insects of the same age were placed with each 200 gram sample of seed. Data A nalysis The measured data obtained from drying experiment were analyzed using SAS program (version 9. 4 , SA S Institute Inc. Cary, NC, USA) with using PROC GLM code that calculat e d the analysis of variance ( ANOVA ) to test the pairwise comparison between the mean of significant treatments , if there was a significant differences by using Tukey Kramer method as a multiple comparison procedures . Results and D iscussion The d rying exp eriment was conducted with three drying temperatures matched with three drying times to find their effect on moisture content (MC), seed germination, insect mortality, power consumption, and energy cost. M oisture content was the main factor in control infe station, as mentioned previously by reducing seed MC causing dehydration of the insects that live with the seed . A p reliminary test was done to find the best drying time that provide d the ideal MC for prevent ing insects from normal growth. According to the results from th e preliminary test, as shown in Figure 6 2, the drying time s w ere observed then selected to be 24 and 48 h for each drying temperatures . T his was done to find the effect of drying temperature with fix ed drying time on seed germination and i nsect mortality . T he

PAGE 69

69 time was var ied for the third drying test to find the effect of drying temperature with M C on mortality and germination. The observed times from the preliminary test that give 9% MC were 12, 36, and 96 h for drying temperatures of 45, 40 and 35° C, respectively . Moisture Content The results for t riticale seed moisture content tests show ed that there was a high ly significant differ ence between the treatments with P value lower than 0.0001 as shown in Figure 6 7 . T he mean comparison bet ween the treatments is shown in Table 6 4. The initial MC for the seed was 13.85% and the lowest MC value obtained after drying was 7.9% at 45°C for 48 h as shown in Table 6 4. T he highest MC value s were observed for drying with 35°C for 24 h a nd for 45°C for 12 h . T ime s was short est for 45°C to remove water. Also, Table 6 4 shows that drying at 35°C did not produce a significant difference between treatments with different drying time, while drying at 40°C show ed treatment differing means for 24 and 48 h o nly. D ying at 45°C produced treatment means differe d for all the three drying time. Figure 6 7 shows that the increasing drying time cause d decreasing MC for each drying temperature . Also, Figure 6 7 shows that by increasing the drying temperature the MC decrease d when compare d with same drying time . D rying with 35°C for 24 h gives high er MC than 40° C which is lower MC than drying with 45° C. Figure 6 7 also shows that by increasing the drying time the variance between the samples that have the same treat ment was small er , as shown in the dying with 35°C for 96 h and 40°C for 48 h . Seed Germination The germination test showed that there were not significant differences between drying treatments with 0.59 P value . This result lead to the conclusion that dy ing at

PAGE 70

70 these temperatures did not affect the germination for triticale seed , as shown in Figure 6 8. Also, F igure 6 8 and Table 6 5 show that the control treatments did not differ from the drying treatment s. E ven in the range s within each treatment, the on ly obvious treatment difference was drying with 40°C for 48 h as shown in the middle of F igure 6 8 . S ix of the eight replications were close to (89%) with high and low values of 90% and 84%, respectively. Figure 6 9 shows that the distribution of the germi nation mean values for each drying treatment was random with a range within two to three percentages, where the highest germination percentage was 90.74% for control treatment with HHC and the lowest was 87.64% for drying with 35°C for 48 h . Figure 6 10 shows the real germination rate of t riticale seed befor e the drying treatment as a control treatment. Figure 6 11 shows the real germination of t riticale seed after the drying treatment at 45°C for 48 h as the highest treatment. From both Figures 6 10 and 6 11 it can be observe d that the diff erence before drying and after were insignificant. Insect Mortality The i nsect mortality result s for seven weeks after the drying treatments were completed shows that there were significant differences between drying t reatments on the mortality of rice weevil adult s , as shown in Figure 6 12 . Figure 6 12 shows that all drying treatment s cause d high mortality percentages above 99.66% after seven weeks . It was also observed that during two weeks after the drying treatment of 45°C for 48 h there was 92% mortality comp a re d with 15% for the control treatment. The storage in LH chamber without drying treatment as a control was also produced a high mortality value of 93% . E ven the control treatment , without any drying treatment and with HH

PAGE 71

71 chamber ( without low humidity chamber), obtained high mortality reach ed 73.73% , as shown in Table 6 6 . The high mortality percentages were caused by the age of the adults which reach ed an average of 110 d (almost four months) includ ing 49 d (7 weeks) ( time spent measuring mortality to reach total mortality) after complet ion of the drying test. These results were confusing at some level, but at the end of the period of counting mortality, a new generation of adults start ed emerg ing from the seed of some treatments. Results for t he new generation were recorded and analyzed . T here w ere high ly significant differences between treatment s , as shown in Figure 6 13 . T he significant difference s between the mean drying treatments w ere between the control (LH and HH at the same level) and the rest of the treatments (at the second level). Figure 6 13 shows, all the drying treatment s were lower than four new generation adult s , while the control treatments were between 200 to 400 new generation adults. This re sult explain s the effect of drying treatment in the long run, where within only two months the control treatment s showed triple the amount of new generation s while the dried treatment s still resist ed infestation of the new generation , as shown in Table 6 7 . Figure 6 14 shows the differences between the samples with and without drying treatment, where in both cans there were 100 rice weevil adults in one can (left side) dr i ed to 8.6% MC and the left can without drying treatment. A fter two months all insects were dead for the dried sample but other sample a new generation of insect s emerged with triple the amount of insect s, as shown clearly in Figure 6 14. Energy Consumption The energy consumption results shows the energy consumption differences between the drying treatments, as shown in Figure 6 15, where there was a high ly

PAGE 72

72 significant difference between the drying treatments with 0.0001 P value. Figure 6 15 shows that increasing the drying time increase d the power consumptio n for each drying temperature. Ta ble 6 8 shows that there was a significant difference between drying treatments, where the highest energy consumption was for drying at 35°C for 96 h. T his high consumption was due to the long period of dying . T he lowest one was drying with45°C for 12 h an d the medium consumption was 40°C for 36 h . Energy Cost The cost of the energy for drying operation was highly correlated to the energy consumption due to the electricity powered dryer , a s shown in Figure 6 16 . Figure 6 16 shows that there w ere significan t differences between treatments means, and also shows that the cost increased rapidly where drying time increased with the same drying temperature, but that the cost increased gradually when the drying temperature was increased for fix ed drying time. The highest cost occurred with 96 h drying at 35°C with 327 $/ tonne , and the lowest cost occurred by drying for only 12 h at 45°C with 58 $/ tonne , as shown in Table 6 9 . Energy Statement An estimate of economic feasibility was done based on the cost of infes tation losses compared with the costs of different treatments. Industrial sources were used to find the cost of the seed and the pesticide treatments, energy feasibility was calculated, as shows in Table 6 10 . T he average price of triticale seed was $1212. 54 / tonne (Danny Mixon, personal communication, November 14 th , 2014). T he production in 2012 for triticale seed was 1668.247 tonnes (70,727 bushels) for North Carolina , the nearest

PAGE 73

73 state that ha s statistics available f or 2012 (NASS, 2012) , The value of the North Carolina triticale crop was $2,022,816.22 / year. T ypical losses from insect infestation of triticale seed according to an industry source (Bill Smith, personal communication, November 14 th , 2014) exceed 50% of the seed after 90 days at ambient storage temperature . So without insecticide treatments the losses w ould be $1,011,408.11 / year for North Carolina . The cost of chemical application for seed treatment by insecticide s was $ 65.4/tonne/month ($7.5 /cwt/ month) according to a source from the seed industry (Dan ny Mixon, personal communication, November 14 th , 2014). The cost of drying seed at low temperature was $ 8.83/ tonne, as shown in Table 6 10 . Drying Summary The dying treatments significantly aid ed control of weevil infestation of triticale seed , and at t he same time d id not affect seed germination. The effect of the drying treatments was to reduce the seed moisture content to levels that produced mortality of the insects. E nergy cost and consumption were a ffected by the drying time more than the drying te mperature.

PAGE 74

74 Table 6 1. The age of the rice weevil adult s during the experiment and the staring date for each replication. Replication Adult age (days) Date of starting the test 1 54 to 59 September 2 nd 2014 2 59 to 64 September 7 th 2014 3 62 to 67 Se ptember 12 th 2014 Table 6 2. Drying temperatures, drying time for drying test. Drying t emperature (°C) Drying t ime ( h ) 35 24, 48, 96 40 24, 36, 48 45 12, 24, 48

PAGE 75

75 A . B Figure 6 1. The seed dryers ( Photo s courtesy of author , Mahmoud Khedh er Agha) , A) Front view with the exhaust fans in top of them . B) Drying chamber with a screen basket hold the seed and surrounded by the blue foam.

PAGE 76

76 Figure 6 2 . Drying curve for triticale seed at three drying temperatures. R² = 0.9379 R² = 0.9832 R² = 0.997 7 8 9 10 11 12 13 14 15 16 0 10 20 30 40 50 60 70 80 90 100 MC ( % db ) Drying Time (h) 40 C Drying 45 C Drying 35 C Drying Poly. (40 C Drying ) Poly. (45 C Drying) Poly. (35 C Drying)

PAGE 77

77 A B C Figure 6 3 . Vacuum seed counter apparatus ( Photo s courtesy of author , Mahmoud Khedher Agha) , A) Front view . B) Rear view . C) The apparatus connected to vacuum source.

PAGE 78

78 A B Figure 6 4 . Germinat ion paper ( Photo s courtesy of author , Mahmoud Khedher Agha) , A) B e fore rolling with 50 triticale seed s . B) After rolling.

PAGE 79

79 A B Figure 6 5 . The seed germinator ( Photo s courtesy of author , Mahmoud Khedher Agha) , A) F ront view of the germinator . B) I nside the germinator. Figure 6 6 . Vacuum insect counter appa ratus with separation sieves ( Photo courtesy of author , Mahmoud Khedher Agha) .

PAGE 80

80 Table 6 3 . T he effects of different drying temperature s, drying t imes and moisture content on seed viability and insect mortality . Drying temperature Drying time (h) Moisture content (%) wb Insects 35 24 9.71 with without 48 9.18 with without Vary ( 96) 9.04 with without 40 24 9.10 with without 48 8.56 with without Vary (36) 8.81 with without 45 24 8.56 wi th without 48 7.85 with without Vary (12) 9.69 with without 45 Control at high Rh 48 7.85 with without No drying at low Rh control 13.88 with without No drying at high Rh control 13.83 with without

PAGE 81

81 Table 6 4. The effec t of drying temperature and drying time on triticale seed moisture content (%) wb. Drying t emperature (°C) Drying t ime ( h ) No d rying 12 24 36 48 96 35 9.7 b 9.2 b c 9.1 c d 40 9.1 c 8.8 c d 8.6 d 45 9.7 b 8.6 d 7.9 e Contro l HHC d ryin g 7.9 e Control LHC 13.88 a Control HHC 13.83 a Figure 6 7. Distribution of MC for each treatment for drying triticale seed.

PAGE 82

82 Figure 6 8. Distribution of germination percentages with each treatment of drying tritica le seed. Table 6 5. The effect of drying temperature and drying time on triticale seed germination (%). Drying t emperature (°C) Drying t ime (Hr.) No d rying 12 24 36 48 96 35 90.25 87.6 4 89.7 4 40 89.85 90 88.4 9 45 89.1 3 89.4 9 88.6 9 Control HHC d rying 87.95 Control LHC 89.5 Control HHC 90.74

PAGE 83

83 Figure 6 9. Distribution of germination percentages with each treatment means of drying triticale seed.

PAGE 84

84 Figure 6 10. Triticale seed after germination with control treatment with out drying ( Photo courtesy of author , Mahmoud Khedher Agha) . Figure 6 11. Triticale seed after germination with treatment of drying on 45 ° C for 48 h ( Photo courtesy of author , Mahmoud Khedher Agha) .

PAGE 85

85 Table 6 6. The effect of drying temperature and trit icale seed moisture content on insect m ortality% . Drying temperature (°C) MC (%) No drying 7.9 8.6 9.1 9.7 35 99.67 ab 100 a 40 100 a 99.33 ab 45 99.67 ab 100 a 99.66 ab Control HHC Drying 98.67 ab Control LHC 93 b Control HHC 73.73 c Figure 6 12. Distribution of mortality percentages with each treatment means of triticale seed drying.

PAGE 86

86 Figure 6 13. Distribution of the new generation of rice weevil adults for each drying treatment mean of tritica le seed.

PAGE 87

87 Table 6 7. Tukey Comparison Lines for Least Squares Means ( LSMEAN ) of Treatments for the new generation adults . Differences level Number of new generation adults mean Treatments LSMEAN n umber A 323 CH 10 A 205 CL 12 B 4 35 24 1 B 3 CH 45 48 11 B 1 35 48 2 B 0.3 35 96 3 B 0 40 36 5 B 0 45 24 8 B 0 45 48 9 B 0 40 48 6 B 0 45 12 7 B 0 40 24 4 Note: the same letter are not significantly different Figure 6 14. The seed sample with drying treatments left side , and without drying tre atments with emerging of new generation of rice weevil adults on the right side ( Photo courtesy of author , Mahmoud Khedher Agha) .

PAGE 88

88 Figure 6 15. Distribution of the energy consumption for each drying treatment means. Table 6 8. The effect of drying te mperature and drying time on energy consumption (kWh). Drying t emperature (°C) Drying t ime ( h ) 12 24 36 48 96 35 1.3 f 2.6 c 4.9 a 40 1.4 f 2.1 d 2.8 c 45 0.9 g 1.7 e 3.4 b Control HHC Drying 7.9 e

PAGE 89

89 Figure 6 16. Distributio n of the energy cost ($/tonne) for each drying treatment means. Table 6 9. The effect of drying temperature and drying time on energy cost ($/tonne). Drying temperature (°C) Drying t ime (h ) 12 24 36 48 96 35 88 f 17 8 c 327 a 40 93 f 141 d 19 2 c 45 58 g 114 e 23 1 b Table 6 10. Comparison of infestation control treatments cost vs drying treatments cost. Triticale Seed Value ($/tonne) Pesticide Treatment Cost ($/tonne/month) Drying Treatment Cost ($/tonne) 1212.54 165.35 8.83

PAGE 90

90 Figure 6 17. Psychometric chart with saturated curve represent 20% MC for triticale and drying points . Table 6 11. Explanation of the points on the psychometric chart. Explanation of points Points Temperature DB °C (°F) RH % Blue line start A 25 (77) 79 Blue line end B 45 (113) 26 Red line end C 30 (86) 80 Purple line lower point 5 (41) 76 Purple line mid point 23 (73.4) 80 Purple line upper point 35 (95) 82

PAGE 91

91 CHAPTER 7 CONCLUSIONS Summary The main purpose of this research was to develo p a method that prevents triticale seed infestation by weevils , without using chemicals . Insect infestation caus es industry 50% losses if chemical treatments are not applied. Solving this problem was done through two main steps and one sub step . The f irst main step , a prediction model of insect infestation inside triticale seed was developed . The prediction of degree of triticale seed infest ation with rice weevils with different growth stages was done using the spectral signature. It was found that stepwis e regression produced the best model that yielded the lowe st error with a high R 2 of 0.988. This model was able to predict infestation well for each growth stage separately. In addition, a model for all the stages was found with an RMSE of 10%. The second method was the excessive search that produced a lower RMSE than the regression tree method. Also, different growth stages affected the selection accuracy, where the late growth stages gave a higher RMSE than the early growth stages for all the methods and almost all the growth stages. This led to a conclusion of the insect size affecting the prediction accuracy. The wavelengths between 400 to 410 nm were selected as important ones for each growth stage and for all the metho ds. The su b step was to find the relationships between equilibrium moisture content and equilibrium relative humidity of triticale seed that aid the third step of controlling infestation. This was studied using saturated salt techniques, and presented as sorption isotherms at different te mperatures. These isotherms were modeled using the modified Henderson (MHE), modified Chung Pfost (MCPE) and modified Oswin (MOE)

PAGE 92

92 equations. All three models represented this relationship very well with R 2 values over 0.98. The MCPE gave the best predictio n with a lowest standard error (2.4%) and a highest R 2 (0.99) value, followed closely by the MHE and MOE models. The highest and lowest residual values of ERH for MCPE were 5.5% and 6.0%, respectively. The second main step was done using seed drying trea tment s as a method of infestation control that dehydrate d the insects. The drying experiment showed that total insect mortality occurred with in seven weeks with drying temperatures 40, 45 C with 8.6% M C , while there was not an ef fect on germination percent ages. Also, drying treatments prevent ed the occurrence of new generation s of weevils in the seed. The economic feasibility indicated that there is a big potential for dryi ng seed at low temperature to replace pesticide application. Economic analysis of p otential costs involved with seed drying compare d with saving s from not using pesticide indicated that drying triticale seed for long term storage is desirable. Future Work Conducting drying test s followed by m onitored storage for longer time period of on e to three or more years would confirm the desirability of drying for the storage times required for commercializ ation . Control of the RH of the air in storage system is one of the key fact ors that help prevent insect infestation. D ifferent methods could b e used , first using an air condition ing unit for dehumidification by adjusting the cooling coil temperature below the dew point of the air. Second, using dried materials that have very low MC plac ed inside the storage area . Materials such as dried seed, dr y wood, and g unny or j ute bags could be used to store 100 kg of dry seed providing a barrier to moisture enter ing the seed.

PAGE 93

93 Conducting a control led atmosphere experiment w ould be a useful method to control infestation . This method has the fastest and big g est effects on insect mortality. Using the predicted wavelengths to develop a portable NIR spectrometer with a microcontroller and small display as an instrument that can analyze the samples on the go for detecting infestation, which can be used for triti cale and other type s of seeds. Apply the result s of drying experiment s at a prototype scale seed storage system to evaluate the effect s of drying on germination and insect mortality and evaluate the economic benefit s . If economic benefits are favorable, th en construct a full scale system and develop educational program s for industry.

PAGE 94

94 APPENDIX DRYING COST CALCULATION Heat balance equation (Brooker et al., 1992) Dry ma t ter (DM) Heat requirements (Loewer, Brid ges, & Bucklin, 1994)

PAGE 95

95

PAGE 96

96 LIST OF REFERENCES AOSA. (2013). AOSA Rules for t esting s eeds. Moline, IL Association of Official Seed Analysts, Inc. ASABE Standards. ( 2012 a ) . D245.6: Moisture relationships of plant based agricultural products. St. Joseph, M I .: ASABE. ASABE Standards. ( 2012 b ) . S352.2: Moisture measurement unground grain and seeds. St. Joseph, M I .: ASABE. ASTM. (2012). Standard p ractice for m aintaining c onstant r elative h umidity by m eans of a queous s olutions (Vol. West Conshohocken, PA . Baker, K. D., Paulsen, M. R., & Zweden, J. V. (1993). Temperature e ffects on s eed c orn d ryer p erformance. Applied Engineering i n Agriculture , 9 (1). Beckett, S. J. ( 2011). Insect an d mite control by manipulating temperature and moisture before and during chemical free storage. Journal of Stored Products Research, 47 (4), 284 292. doi: 10.1016/j.jspr.2011.08.002 Brooker, D. B., Bakker Arkema, F. W., & Hall, C. W. (1992). Drying and St orage Of Grains and Oilseeds : Springer. Choi, B. M., Lanning, S. B., & Siebenmorgen, T. J. (2010). A r eview of hygroscopic equilibrium studies applied to rice . Transactions of the ASABE , 53 (6), 14. Clark, J. ( 2007). The Beer Lambert Law. Retrieved Nove mber 30th, 2014, from http://www.chemguide.co.uk/analysis/uvvisible/beerlambert.html Dobie, P., & Kilminster, A. M. (1978). The susceptibility of triticale to post harvest infe station by Sitophilus zeamais Motschulsky, Sitophilus oryzae (L.) and Sitophilus granarius (L.). Journal of Stored Products Research, 14 (2 3), 87 93. doi: http://dx.doi.org/10.1016/0022 474X(78) 90003 6 FAOSTAT. (2012). Food and Agriculture Organization of the United Nations Statistics Division. Retrieved 08/29, 2014, from http://faostat3.fao.org/faostat gateway/go/to /download/Q/QC/E Figura, L. O., and A. A. Teixeira. ( 2007 ) . Food physics: physical properties measurement and applications. Springer Press, Heidelberg, Germany, 2007. 550 pages. Godbolt, C., Danao, M. C., & Eckhoff, S. R. (2013). Modeling of the equi librium moisture conten t (EMC) of Switchgrass. Transactions of the A SABE , 56 (4), 1495 1501.

PAGE 97

97 Greenspan, L. ( 1977 ) . Humidity fixed points of binary saturated aqueous solutions. Journal of Research of the National Bureau of Standards A. Physics and Chemistr y, 81A (1): 89 96. Hall, D. W. (1970). Handling and Storage of Food Grains in Tropical and subtropical Areas (Vol. 90): FAO, Rome. Hansen, R. ( 2011 ) . Triticale. Agricultural marketing resource center. Retrieved 0 2 / 07 , 201 2 , from http://www.agmrc.org/commodities__products/grains__oilseeds/triticale/ Hastie, T., Tibshirani, R., & Friedman, J. (2008). The e lements of s tatistical l earning d ata m ining, i nference, and p redi ction (Second ed.): Springer. Herrick, N. J., & Mankin., R. W. (2012 ). Acoustical detection of early instar Rhynchophorus ferrugineus (Coleoptera: Curculionidae) in Canary Island date palm, Phoenix cana riensis (Arecales: Arecaceae). Florida Entomologist, 95 (4), 983 990. Hertert, U., & Burris, J. S. (1989). Effect of drying and temperature on drying injury of corn seed . Canadian Journal Plant Science, 69 (July), 763 774. IFAS UF. (2014). Florida automated weather network . Retrieved November, 27th 2014, fr om University of Florida http://fawn.ifas.ufl.edu/data/reports/ Igathinathane, C., Womac, A. R., Sokhansanj, S., & Pordesimo, L. O. (2005). Sorption equilibrium moisture characteristics of selected co rn stover components. Transactions of the Asae, 48 (4), 1449 1460. Iguaz, A., & Virseda, P. (2007). Moisture desorption isotherms of rough rice at high temperatures. Journal of Food Engineering, 79 (3). doi: 10.1016/j.jfoodeng.2006.03.002 Kazakevich, Y. ( 2010). Molecular s pectroscopy Beer Lambert Law (textbook). Retrieved November 27, 2014, from Seton Hall University https://hplc.chem.shu.edu/NEW/Undergrad/Molec_Spectr/Lambert .html Khedher Agha, M. K., Lee, W. S., Bucklin, R. A., Teixeira, A. A., & Blount, A. R. (2014). Sorption isotherms for triticale seed . Transactions of the ASABE, 57 (3), 901 904. Li, X. ( 2012 ) . The hygroscopic properties and sorption isosteric heats of different Chinese wheat types. Journal of Food Research 1(2):82 98. Loewer, O. J., Bridges, T. C., & Bucklin, R. (1994). On farm drying and storage systems : American Society of Agricultural Engineers.

PAGE 98

98 Mankin, R., & Hagstrum, D. (2011). Acoustic monitorin g of insects. In T. W. P. D. W. Hagstrum, and G. Cuperus (Ed.), Stored product protection . Manhattan, KS: Kansas State Univ. Press. characteristics of triticale (X Tri ticosecale Wittmack) with HMW glutenin subunits 5+10. Journal of Cereal Science, 47 (1), 68 78. doi: 10.1016/j.jcs.2007.02.003 NASS. (2012). The 2012 Census of Agriculture. Retrieved 08/29, 2014, from http://quickstats.nass.usda.gov/results/744C6455 A9E7 3573 8462 57EB04DD7F57 Ondier, G. O., Siebenmorgen, T. J., & Mauromoustakos, A. (2010). Low temperature, low relative humidity drying of rough rice. Journal of F ood Engineering, 100 (3). doi: 10.1016/j.jfoodeng.2010.05.004 Peiris, K. H. S., Pumphrey, M. O., Dong, Y., Maghirang, E. B., Berzonsky, W., & Dowell, F. E. (2010). Near infrared spectroscopic method for identification offusariumhead blight damage and predi cti on of Deoxynivalenol in s ingle w heat k ernels. Cereal Chemistry, 87 (6), 511 517. doi: 10.1094/cchem 01 10 0006 Pena, R. J. (2004). Food uses of triticale. In M. Mergoum & H. Gomez Macpherson (Eds.), Triticale improvement and production (pp. 37 44). Rome : FAO. Perez Mendoza, J., Throne, J. E., & Baker, J. E. (2004). Ovarian physiology and age grading in the rice weevil, Sitophilus oryzae (Coleoptera: Curculionidae). Journal of Stored Products Research, 40 (2), 179 196. doi: http://dx.doi.org/10.1016/S0022 474X(02)00096 6 Salmon, D. F., Mergoum, M., & Gomez Macpherson, H. (2004). Triticale production and management. In M. Mergoum & H. Gómez Macpherson (Eds.), Triticale improvement and producti on (pp. 27 34). Rome FAO . Sharifi, S., & Mills, R. B. (1971). Developmental activities and behaviour of the rice weevil inside wheat kernels. . Journal of Economic Entomology, 64 (5), 1114 1118. Singh, C. B., Jayas, D. S., Paliwal, J., & White, N. D. G. (2009). Detection of insect damaged wheat kernels using near infrared hyperspectral imaging. Journal of Stored Products Research, 45 (3), 151 158. doi: 10.1016/j.jspr.2008.12.002 Siuda, R., Grabowski, A., Lenc, L., Ralcewicz, M., & Spychaj Fabisiak, E. (2 010). Influence of the degree of fusariosis on technological traits of wheat grain. International Journal of Food Science & Technology, 45 (12), 2596 2604. doi: 10.1111/j.1365 2621.2010.02438.x

PAGE 99

99 Tsen, C. C. H. (1974). Triticale: First Man Made Cereal . St. P aul, Minnesota: The American Association of Cereal Chemists. Viswanathan, R., Jayas, D. S., & Hulasare, R. B. (2003). Sorption isotherms of tomato slices and onion shreds . Biosystems Engineering, 86 (4), 465 472. doi: 10.1016/j.biosystemseng.2003.08.013 W ellbeing, F. o. H. a. (2009). Molecular Spectroscopy Beer's Law. Retrieved November 27, 2014, from Sheffield Hallam University http://teaching.shu.ac.uk/hwb/chemistry/tuto rials/molspec/beers1.htm Williams, P., & Norris, K. H. (2001). Near infrared technology in the agricultural and food industries (Second Edition ed.): American Association of Cereal Chemists,Inc. St. Paul, Minnesota,USA.

PAGE 100

100 BIOGRAPHICAL SKETCH Mahmoud w as born in Baghdad, Iraq in 1976. He receiv ed his Bachelor of Science in a gricultural m achinery in 1998 and then completed his Master of Science in a gricultural m echanization in 2001 from the College of Agriculture University of Baghdad , in Iraq . After tha t he worked as l ecturer in Agricultural Mechanization Department College of Agriculture University of Baghdad from 2002 until 2009 , where h e was involved in teaching several courses including postharvest equipment and technology, tractor mechanics, and tra ctor hydraulics . Also, he participated with his student program by assisting m aster student s with research topics in applied agricultural engineering . He ha s been a member of the Iraqi Agricultural Engineers Union since 1998. In 2003, h e participated in training for developing the educational methods of teaching and university staff. H e ha s been a member of the American S ociety of Agricultural and B iological Engineering ( ASABE ) since 2012. He is a lso a member of Alpha Epsilon at the Un iversity of Florida , an Honor Society of Agricultural and Biological Engineering since 2012. Mahmoud start ed his PhD stud ies in summer 2010 . He has been involved in projects such as working on a grain mass flow meter, using photo diode detectors, as part o f a class project. H e designed and fabricated a pneumatic specific gravity apparatus for potato es at the Hastings Research and Demonstration Site UF/IFAS in Hastings, FL. H e was awarded a Master of Engineering in a gricultural and b iological e ngineering in fall 2013. He has published three peer reviewed journal papers, two of them published in 2006 and 2007 in The Iraqi Journal of Agricultural Science and the other published in

PAGE 101

101 2014 in Transactions of the ASABE . A lso , he presented and published two conferen ce papers at the 2013 ASABE I nternational M eeting. He was awarded a PhD degree in the Agricultural and Biological Engineering from University of Florida in fall , 2014, in Postharvest Engineering and Technology , and Control Seed Infestation .