<%BANNER%>

Evaluating Vine-Kill Harvested Potatoes to Determine Effects of Harvest Wait Periods on Damage Resistance and to Detect ...

Permanent Link: http://ufdc.ufl.edu/UFE0045110/00001

Material Information

Title: Evaluating Vine-Kill Harvested Potatoes to Determine Effects of Harvest Wait Periods on Damage Resistance and to Detect Surface and Subsurface Damage with Spectral Measurement
Physical Description: 1 online resource (203 p.)
Language: english
Creator: Brecht, Michael Anthony
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: blackspot -- brown -- bruising -- damage -- fabula -- greening -- internal -- measurement -- nirs -- potato -- potatoes -- redlasoda -- resistance -- rot -- shatter -- skinning -- spectral -- spectroscopy -- sunscald -- vine-kill
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, M.E.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Any force exerted on a potato tuber during harvesting and packaging may result in mechanical damage.Mechanical damage and disease are major problems in the harvesting and movement of potatoes from the field to consumer. Being able to quickly identify and remove damaged potatoes during post-harvest handling could reduce storage costs for potatoes which never make it to market. Excessive deterioration in market quality of potatoes from mechanical damage and defects cut into the profits of the producers and shippers. Vine-kill is a practice which involves killing the upper portions of plants by spraying with herbicides or cutting the vines.After the upper portion of the plant has been killed, the potato tubers are left in the soil undergoes skin set which involves the periderm thickening and becoming more damage resistant. This thickening of skin also helps to prevent storage diseases and shrinkage due to water loss. However, leaving tubers in the ground can increase their exposure to brown rot, insects and other deteriorating conditions. The two cultivars used for this study were ‘Fabula’ and ‘Red La Soda’.   A study was carried out to simulate potato handling in a repeatable manner to investigate the damage thresholds of each cultivar at different points from harvest through packing. Potato tubers were collected 7 days before vine-kill,the day before vine-kill and then 7, 14 and 21 days after vine-kill. A set of potatoes from each cultivar was run along the same portion of a packing line each harvest to access skinning resistance, while the remaining tubers were used for drop tests. Drop tests were performed 1 day and 7 days after harvest; then stored to allow internal bruising and shatter to develop before being assessed. Tubers were dropped from the following heights; 30 cm, 30 cm twice, 60 cm, 60 cm twice, and 90 cm. Analysis of ‘Fabula’ data showed a clear decrease in skinning and drop damage between pre vine-kill and only 7 days of skin set; while ‘Red La Soda’ maintained consistently low skinning damage, but during drop testing there was a remarkable decrease in damage development after 7 days of skin set. A study was also carried out in order to collect spectral data of vine-killed potatoes for the purpose of grading and sorting both cultivars. Spectral measurements were made on undamaged potatoes after harvest, after simulating mechanical damage and then following 7 days of storage at 20°C (68°F) in order to allow internal and external damage to develop. Spectral data were also collected for tubers showing greening, brown rot, shatter, growth cracking, sunscald and insect damage. After collecting data for the wavelength range between 200-2500 nm for moisture content, it was decided to concentrate on the sensing range between 360-800 nm due to water having less effect on reflectance measurement. Analysis showed that undamaged reflectance properties of each tuber cultivar were altered in predictable ways depending on which defect was present. The change to the spectral reflectance caused by defects was found to be statistically significant using PLS and SMLR analysis; which would allow the potential of detection and elimination of damaged tubers on a packing line.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Michael Anthony Brecht.
Thesis: Thesis (M.E.)--University of Florida, 2012.
Local: Adviser: Lee, Won Suk.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0045110:00001

Permanent Link: http://ufdc.ufl.edu/UFE0045110/00001

Material Information

Title: Evaluating Vine-Kill Harvested Potatoes to Determine Effects of Harvest Wait Periods on Damage Resistance and to Detect Surface and Subsurface Damage with Spectral Measurement
Physical Description: 1 online resource (203 p.)
Language: english
Creator: Brecht, Michael Anthony
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2012

Subjects

Subjects / Keywords: blackspot -- brown -- bruising -- damage -- fabula -- greening -- internal -- measurement -- nirs -- potato -- potatoes -- redlasoda -- resistance -- rot -- shatter -- skinning -- spectral -- spectroscopy -- sunscald -- vine-kill
Agricultural and Biological Engineering -- Dissertations, Academic -- UF
Genre: Agricultural and Biological Engineering thesis, M.E.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Any force exerted on a potato tuber during harvesting and packaging may result in mechanical damage.Mechanical damage and disease are major problems in the harvesting and movement of potatoes from the field to consumer. Being able to quickly identify and remove damaged potatoes during post-harvest handling could reduce storage costs for potatoes which never make it to market. Excessive deterioration in market quality of potatoes from mechanical damage and defects cut into the profits of the producers and shippers. Vine-kill is a practice which involves killing the upper portions of plants by spraying with herbicides or cutting the vines.After the upper portion of the plant has been killed, the potato tubers are left in the soil undergoes skin set which involves the periderm thickening and becoming more damage resistant. This thickening of skin also helps to prevent storage diseases and shrinkage due to water loss. However, leaving tubers in the ground can increase their exposure to brown rot, insects and other deteriorating conditions. The two cultivars used for this study were ‘Fabula’ and ‘Red La Soda’.   A study was carried out to simulate potato handling in a repeatable manner to investigate the damage thresholds of each cultivar at different points from harvest through packing. Potato tubers were collected 7 days before vine-kill,the day before vine-kill and then 7, 14 and 21 days after vine-kill. A set of potatoes from each cultivar was run along the same portion of a packing line each harvest to access skinning resistance, while the remaining tubers were used for drop tests. Drop tests were performed 1 day and 7 days after harvest; then stored to allow internal bruising and shatter to develop before being assessed. Tubers were dropped from the following heights; 30 cm, 30 cm twice, 60 cm, 60 cm twice, and 90 cm. Analysis of ‘Fabula’ data showed a clear decrease in skinning and drop damage between pre vine-kill and only 7 days of skin set; while ‘Red La Soda’ maintained consistently low skinning damage, but during drop testing there was a remarkable decrease in damage development after 7 days of skin set. A study was also carried out in order to collect spectral data of vine-killed potatoes for the purpose of grading and sorting both cultivars. Spectral measurements were made on undamaged potatoes after harvest, after simulating mechanical damage and then following 7 days of storage at 20°C (68°F) in order to allow internal and external damage to develop. Spectral data were also collected for tubers showing greening, brown rot, shatter, growth cracking, sunscald and insect damage. After collecting data for the wavelength range between 200-2500 nm for moisture content, it was decided to concentrate on the sensing range between 360-800 nm due to water having less effect on reflectance measurement. Analysis showed that undamaged reflectance properties of each tuber cultivar were altered in predictable ways depending on which defect was present. The change to the spectral reflectance caused by defects was found to be statistically significant using PLS and SMLR analysis; which would allow the potential of detection and elimination of damaged tubers on a packing line.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Michael Anthony Brecht.
Thesis: Thesis (M.E.)--University of Florida, 2012.
Local: Adviser: Lee, Won Suk.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2012
System ID: UFE0045110:00001


This item has the following downloads:


Full Text

PAGE 1

1 EVALUATING VINE KILL HARVESTED POTATOES TO DETERMINE EFFECTS OF HARVEST WAIT PERIODS ON DAMAGE RESISTANCE AND TO DETECT SURFACE AND SUBSURFACE DAMAGE WITH SPECTRAL MEASUREMENT By MICHAEL ANTHONY BRECHT A THESIS PRESENTED TO THE GRADUATE SC HOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ENGINEERING UNIVERSITY OF FLORIDA 201 2

PAGE 2

2 201 2 M ichael A nthony B recht

PAGE 3

3 To my parents, Brenda and Jeffrey; and everyone that helped me make it to this point in my academic career

PAGE 4

4 ACKNOWLEDGEMENTS my committee chair who always encouraged me to work harder and helped me determine the best way to perform my spectral analysis. I thank Dr. Steve Sargent my committee co chair who provided me wit h advice, helped with the collection of samples for thi s work and provided me with access to the laboratories in the Horticultural Sciences Department I would also like to thank Dr. Ray Bucklin for ta king time to review my work and make suggestions. I thank Adrian Berry, Kim Cordasco, and Mildred Makani for their help with data collection and analysis.

PAGE 5

5 TABLE OF CONTENTS P age ACKNOWLEDGMENTS ................................ ................................ ................................ ............. 4 LIST OF TABLES ................................ ................................ ................................ ......................... 7 LIST OF FIGURES ................................ ................................ ................................ ....................... 8 LIST OF ABBREVIATIONS ................................ ................................ ................................ ...... 1 7 ABSTRACT ................................ ................................ ................................ ................................ 19 CHAPTER 1 LITERATURE REVIEW ................................ ................................ ................................ ...... 2 1 1.1. Introduction ................................ ................................ ................................ ..................... 2 1 1.2. The Potato ................................ ................................ ................................ ....................... 2 3 1.3. Ha rvesting A fter Vine Kill ................................ ................................ ............................. 2 4 1.3.1. Mechanical Vine killing ................................ ................................ ...................... 2 5 1.3.2. Chemical Vine killing ................................ ................................ .......................... 2 6 1.3.3. Combined Methods ................................ ................................ .............................. 2 7 1.4. Packaging and Sto rage ................................ ................................ ................................ .... 2 7 1.4.1. Preharvest Period ................................ ................................ ................................ 29 1.4.2. Cooling Per iod ................................ ................................ ................................ ..... 29 1.4.3. Long term Sto rage Period ................................ ................................ .................... 29 1.4.4. Marketing Period ................................ ................................ ................................ 3 0 1.5. Potato Quality Factors ................................ ................................ ................................ .... 3 1 1.5.1. Mechanical Damage of Potatoes ................................ ................................ .......... 3 1 1.5.1.1. Causes of mechan ical damage ................................ ................................ 3 3 1.5.1.2. Mechanical damage p revention ................................ ............................... 3 4 1.5.2. Environmental Physiological Defects of Potatoes ................................ ............... 3 6 1.6. Reflectance Measur ement ................................ ................................ ............................... 3 7 1.6.1. Effect of Water on Spectral Ch aracteristics ................................ ......................... 3 7 1.6.2. Application of Spectroscopy in Food Analysis ................................ ................... 39 1.6.3. Alternative Food Analysis: M achine Vision ................................ ........................ 4 3 1.7. Simulation of Mechanical Damage ................................ ................................ ............... 4 4 1.7.1. Damage Resistance of New Potatoes ................................ ................................ ... 4 5 1.7.2. Skinning Simulation ................................ ................................ ............................. 4 6 1.7.3. Impact Damage Simu lati on ................................ ................................ .................. 4 6 1.7.4. Detection and Evaluation of Mechanical Damage ................................ ............... 4 8 1.7.4.1. V ine kill wait period assessment ................................ ............................. 49 1.7.4.2. Spec tral measurem ent assessment ................................ ........................... 5 0 1.7.4.3. Assessment o b jectives ................................ ................................ ............. 5 6

PAGE 6

6 2 DESTRUCTIVE ASSESSMENTS ................................ ................................ ....................... 6 1 2.1. Materials and Methods ................................ ................................ ................................ ... 6 1 2.1.1. Plant Materials ................................ ................................ ................................ ..... 6 2 2.1.2. Potato Skinning Tests ................................ ................................ .......................... 6 3 2.1.3. Impact Drop Tests ................................ ................................ ................................ 6 3 2.1.4. Moisture Content Tests ................................ ................................ ........................ 6 5 2.1.5. Compression Rupture Force Tests ................................ ................................ ....... 6 6 2.2. Results and Discussion ................................ ................................ ................................ ... 6 6 2.2.1. Potato Skinning Tests ................................ ................................ .......................... 6 6 2.2.2. Impact Drop T ests ................................ ................................ ................................ 6 8 2.2.2.1. Fabula ................................ ................................ ................................ .... 6 8 2.2.2.2. Red la s od a ................................ ................................ ............................. 7 0 2.2.3. Moisture Content Tests ................................ ................................ ........................ 7 3 2.2.4. Com pression Rupture Force Tests ................................ ................................ ....... 7 4 2.3. Summary ................................ ................................ ................................ ......................... 7 5 3 NON DESTRUCTIVE SPECTRAL MEASUREMENT ASSESSMENT ........................... 9 7 3.1. Materials and Methods ................................ ................................ ................................ ... 9 7 3.1.1. Potato Plant Materials ................................ ................................ .......................... 9 7 3.1.2. Mechanical Damage Stimulation and Disorders ................................ .................. 9 8 3.1.2.1. Mechanical damag e simulation ................................ ................................ 9 8 3.1.2.2. Disease and dis orders ................................ ................................ ............... 99 3.1.3. Potato Sampling and Reflectan ce Measurement ................................ ................ 10 0 3.1.4. Determination of Impo rtant Wavelengths ................................ ......................... 10 2 3.1.4.1. Correlati on spectrum ................................ ................................ .............. 10 2 3.1.4.2. Partial least square s (PLS) regression ................................ .................... 10 3 3.1.4.3. Stepwise multiple lin ear regression (SMLR ) ................................ ......... 10 4 3.2. Results and Discussion ................................ ................................ ................................ 10 4 3.2.1. Effects of Water on Spectr al Characteristics ................................ ..................... 10 4 3.2.2. Correlation Coefficie nt Spectrum ................................ ................................ ...... 10 6 3.2.3. Partial Least Squares (P LS ) Regression ................................ ............................ 1 09 3.2.4. Stepwise Multiple Linear Reg ression (SMLR ) ................................ .................. 11 6 3.3. Summary ................................ ................................ ................................ ....................... 12 0 4 CONCLUSIONS ................................ ................................ ................................ .................. 1 9 5 REFERENCE LIST ................................ ................................ ................................ ................... 1 9 8 BIOGRAPHICAL SKETCH ................................ ................................ ................................ ..... 20 3

PAGE 7

7 LIST OF TABLES Table P age 1 1 Mechanical Damage of Potatoes ................................ ................................ ...................... 5 9 1 2 Classification for degrees of skinning by l ot ................................ ................................ ... 5 9 1 3 Physiol ogical Defects of Potatoes ................................ ................................ .................... 60 2 1 Jenkins Wehner Damage Rating Scale ................................ ................................ ............ 7 9 3 1 Number of Samples collected for each Variety and Disorder ................................ ....... 1 27 3 2 Factor Results for Fabula PLS analysis of sample sets ................................ ............... 1 5 2 3 3 Factor Results for Red La Soda PLS analysis of sample sets ................................ ..... 1 5 3 3 4 Factor Results for Combined Varieties PLS analysis of sample sets ............................ 1 5 3 3 5 Results of Fabula SMLR analysis of sample sets ................................ ....................... 1 8 6 3 6 Results of Red La Soda SMLR analysis of sample sets ................................ .............. 1 8 8 3 7 Results of Combined varieties SMLR analysis of sample sets ................................ ...... 1 9 0 3 8 ................................ ................................ 1 9 3 3 9 ................................ ......................... 1 9 4 3 10 Combined wavelength selection of sample sets ................................ .............................. 1 9 5

PAGE 8

8 LIST O F FIGURES Figure Page 1 1 Potato route on Hasting Pa ckaging Line ................................ ................................ ........... 6 1 2 1 ................................ ................................ ........... 7 9 2 2 Hand ........................ 80 2 3 Packing line washer section used for at har vest skinning test ................................ .......... 80 2 4 Drop test stand using s ling ................................ ................................ ................................ 8 1 2 5 d Fabula with damage (right ) ..... 8 1 2 6 Response of tissue to compression force. Compressing and critically failure tuber tissue (left). Photo of critically failed tuber tissue (right ) ................................ ........................... 8 2 2 7 skinning damage resulting from packing line handling as affected by harvest time (n=20 ) ................................ ................................ ................................ ........... 8 2 2 8 skinning (n=20 ) ................................ ................................ ................................ ................. 8 3 2 9 by harvest time (n=20 ) ................................ ................................ ................................ ...... 8 3 2 10 affected by skinning ( n=20 ) ................................ ................................ .............................. 8 4 2 11 ategory ) ....... 8 4 2 12 testing (n=10 per category ) ........ 8 5 2 13 testing (n=10 per category ) ...... 8 5 2 14 testing (n=10 per category ) ... 8 6 2 15 nt of samples with shatter from impact testing (n=10 per category ) ......... 8 6 2 16 esting (n=10 per category ) ........ 8 7 2 17 t tubers with skinning damage from impact testing (n=70 per catego ry ) ................................ ................................ ................................ ........... 8 7 2 18 per category ) ................................ ................................ ................................ ..................... 8 8

PAGE 9

9 2 19 tubers versus percent tubers with bruising damage from impact testing (n=70 per categor y ) ................................ ................................ ................................ ........... 8 8 2 20 (n=70 per category ) ................................ ................................ ................................ ........... 8 9 2 21 (n=70 per catego ry ) ................................ ................................ ................................ ........... 8 9 2 22 = 10 per category ) ................................ ................................ ................................ ........................... 90 2 23 = 10 per category ) ................................ ................................ ................................ ................................ ........... 90 2 24 Percent of samples with bruising from impact testing (n = 10 per category ) ................................ ................................ ................................ ................................ ........... 9 1 2 25 = 10 per category ) ................................ ................................ ................................ .......................... 9 1 2 26 Percent of samples with shatter from impact testing (n = 10 per category) ................................ ................................ ................................ ................................ ............ 9 2 2 27 = 10 per category ) ................................ ................................ ................................ ................................ ............ 9 2 2 28 Mass of tubers versus percent skinning damage from impact testing (n=70 per category ) ................................ ................................ ................................ ..................... 9 3 2 29 (n=70 per category ) ................................ ................................ ................................ ........... 9 3 2 30 testing (n=70 per categ or y ) ................................ ................................ ............................... 9 4 2 31 testing (n=70 per categ ory ) ................................ ................................ ............................... 9 4 2 32 Mass of tubers versus percent tubers with shatter damage from impact testing (n=70 per catego ry ) ................................ ................................ ............................... 9 5 2 33 Moisture content levels of skin and flesh at t ime of impact testing (n=5 ) ......... 9 5 2 34 mpact testing (n = 5 ) ................................ ................................ ................................ ................................ ............ 9 6

PAGE 10

10 2 35 ) ............................... 9 6 2 36 e failure loading (n=10 ) ..................... 9 7 2 37 Calculated impact energy versus potato mass for drop heights used ................................ 9 7 3 1 Fabula Internal Bruising ................................ ................................ ............................... 12 7 3 2 R ed La Soda Internal Bruising ................................ ................................ ..................... 12 7 3 3 Fabul a Shatter ................................ ................................ ................................ ............... 12 8 3 4 Red La Soda Shatter ................................ ................................ ................................ ..... 12 8 3 5 Fabula Greening ................................ ................................ ................................ ........... 12 8 3 6 Red La Soda Greening ................................ ................................ ................................ 12 9 3 7 Fabula Sunscald ................................ ................................ ................................ ............ 12 9 3 8 Red La S oda Sunscald ................................ ................................ ................................ .. 12 9 3 9 Fabula Brown Rot ................................ ................................ ................................ ........ 1 30 3 10 Red La Soda Brown Rot ................................ ................................ ............................... 1 30 3 11 Fabula Growth Cracking ................................ ................................ .............................. 1 30 3 12 Red La Soda Gro wth Cracking ................................ ................................ .................... 1 3 1 3 13 Red La Soda Insect Damage ................................ ................................ ........................ 1 3 1 3 14 Cary 500 Scan Spectrophotometer (Varian, Inc. Palo Alto, CA ................................ .... 13 2 3 15 Cary 500 Sample measurement port ................................ ................................ ............... 13 3 3 16 Integrating sphere with white pol ytetrafluoroethylene coating ................................ ...... 13 3 3 17 PTFE calibration disk ................................ ................................ ................................ ..... 13 4 3 18 Fabula Moisture Content Reflectance spectrum (200 nm to 2500 nm ) ....................... 13 4 3 19 Red La Soda Moisture Content Reflectance Spectrum (200 nm to 2500 nm ) ............. 13 5 3 20 Fabula Correlation Coefficient between tuber moi sture content and wavelength ....... 13 5 3 21 Red La Soda Correlati on Coefficient between tuber moi sture content and wavelength ................................ ................................ ................................ ................................ .......... 13 6

PAGE 11

11 3 22 Fabula Internal Bruising v s. Undamaged 6 days storage ................................ ............ 13 6 3 23 Red La Soda Internal Bruising v s. Undamaged 7 days storage ................................ .. 13 7 3 24 Fabula Internal Bruising Correlation coefficients between reflectance at each wavelength and internally damaged tissue concentration ................................ ............... 13 7 3 25 Red La Soda Internal Bruising Correlation coeffi cients between reflectance at each wavelength and internally damaged tissue concentration ................................ ............... 13 8 3 26 Combined Internal Bruising Correlation coefficients between reflectance at each wavelength and internally damaged tissue concentra tion ................................ ............... 13 8 3 27 Fabula Shatter v s. Undamaged Day of Harvest ................................ .......................... 13 9 3 28 Red La Soda Shatter vs. Undamaged Day of Harves t ................................ ................ 13 9 3 29 Fabula Shatter Correlation coefficients be tween reflectance at each waveleng th and Shatter concentration ................................ ................................ ................................ ...... 1 40 3 30 Red La Soda Shatter Correlation coefficients between reflectance at each wavelengt h and Shatter concentration ................................ ................................ ................................ 1 40 3 31 Combined Shatter Correlation coefficients between reflectance at each waveleng th and Shatter concentration ................................ ................................ ................................ ...... 1 4 1 3 32 Fabula Greening v s. Undamaged Day of Harvest ................................ ...................... 1 4 1 3 33 Red La Soda Greening v s. Undamaged Day of Harvest ................................ ............. 14 2 3 34 Fabula Greening Correlation coefficients between reflectance at each wavelengt h and Greening concentration ................................ ................................ ................................ ... 14 2 3 35 Red La Soda Greening Correlation coefficients between refl ectance at each wavelengt h and Greening concentration ................................ ................................ ............................ 14 3 3 36 Combined Greening Correlation coefficients between reflectance at each wavelength and Greening concentration ................................ ................................ ................................ ... 14 3 3 37 Fabula Sunscald v s. Undamaged Day of Harvest ................................ ....................... 14 4 3 38 Red La Soda Sunscald v s. Undamaged Day of Harvest ................................ ............. 14 4 3 39 Fabula Sunscald Correlation coefficients between reflectance at each wavelength and Sun Damage concent ration ................................ ................................ ............................. 14 5

PAGE 12

12 3 40 Red La Soda Sunscald Correlation coefficients between reflectance at each wavelength a nd Sun Damage concentration ................................ ................................ ...................... 14 5 3 41 Combined Sunscald Correlation coefficients between reflectanc e at each wavelength a nd Sun Damage concentration ................................ ................................ ............................. 14 6 3 42 Fabula Brown rot v s. Undamaged Day of Harvest ................................ ..................... 14 6 3 43 Red La Soda Brown rot v s. Undamaged Day of Harvest ................................ ........... 14 7 3 44 Fabula Brown Rot Correlation coefficients between reflectance at each wavelength and Brown rot conc entration ................................ ................................ ................................ 14 7 3 45 Red La Soda Brown Rot Correlation coefficients between reflectance at each wavelength and Brown rot co ncentration ................................ ................................ ....... 14 8 3 46 Combined Brown Rot Correlation coefficients between reflectance at each wavelength and Brown rot concentration ................................ ................................ ........................... 14 8 3 47 Fabula Growth Cracking v s. Undamaged Day of Harvest ................................ ......... 14 9 3 48 Red La Soda Growth Cracking v s. Undamaged Day of Harvest ................................ 14 9 3 49 Fabula Growth Cracking Correlation coefficients between reflectance at each wavelength and G rowth Cracking concentration ................................ ............................ 1 50 3 50 Red La Soda Growth Cracking Correlation coefficients between reflectance at each wavelength and Growth Cracking concent ration ................................ ............................ 1 50 3 51 Combined Growth Cracking Correlation coefficients between refle ctance at each wavelength and G rowth Cracking concentration ................................ ............................ 1 5 1 3 52 Red La Soda Insect Damage v s. Undamaged Day of Harvest ................................ .... 1 5 1 3 53 Red La Soda Insect Damage Correlation coefficients between refle ctance at each wavelength and Gr owth Cracking concentration ................................ ............................ 15 2 3 54 Damage Prediction Using PLS Fabula Internal Bruising Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ ............ 15 4 3 55 Damage Prediction Using PL S Red La Soda Internal Bruising Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ ............ 15 4 3 56 Damage Prediction Using PLS Combined Internal Bruising Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ ............ 15 5 3 57 B Coeffic ients Fabula I nternal Bruising Reflectance ................................ .................. 15 5

PAGE 13

13 3 58 B Coefficients Fabula Internal Bruisin g 1st Derivative Reflectance ........................... 15 6 3 59 B Coefficients Red La Soda I nternal Bruising Reflectance ................................ ......... 15 6 3 60 B Coefficients Red La Soda Internal Bruisin g 1st Derivative Reflectance ................. 15 7 3 61 Damage Prediction Using PLS Fabula Shatter Reflectance (left) 1st Derivative (right) ................................ ................................ ................................ ................................ .......... 15 7 3 62 Damage Prediction Using PLS Red La Soda Shatter Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ .............................. 15 8 3 63 Damage Prediction Using PLS Combined Shatter Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ ................................ ......... 15 8 3 6 4 B Coefficients Fabula Shatter Reflectance ................................ ................................ .. 15 9 3 65 B Coefficients Fabula Shatte r 1st Derivative Reflectance ................................ ........... 15 9 3 66 B Coefficients Red La Soda Shatter Reflectance ................................ ........................ 1 60 3 67 B Coefficients Red La Soda Shatte r 1st Derivative Reflectance ................................ 1 60 3 68 Damage Prediction Using PLS Fabula Greening Reflectance (l eft), 1st Derivative (right) ................................ ................................ ................................ .............................. 1 6 1 3 69 Damage Prediction Using PLS Red La Soda Greening Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ .............................. 1 6 1 3 70 Damage Prediction Using PLS Combined Greening Reflectance (left), 1st Derivative (right ) ................................ ................................ ................................ .............................. 16 2 3 71 B Coefficients Fabula G reening Reflectance ................................ ............................... 16 2 3 72 B Coefficients Fabula Greenin g 1st Derivative Reflectance ................................ ....... 16 3 3 73 B Coefficients Red La Soda Greening Reflectance ................................ ..................... 16 3 3 74 B Coefficients Re d La Soda Greeni ng 1st Derivative Reflectance ............................. 16 4 3 75 Damage Prediction Using PLS Fabula Sunscald Reflectance (l eft), 1st Derivative (right) ................................ ................................ ................................ ................................ ......... 16 4 3 76 Damage Prediction Using PLS Red La Soda Sunsc ald Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ .............................. 16 5

PAGE 14

14 3 77 Damage Prediction Using PLS Combined Sunscald Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ .............................. 16 5 3 78 B Coefficients Fabula Sunscald Reflectance ................................ ............................... 16 6 3 79 B Coefficients Fabula Sunscal d 1st Derivative Reflectance ................................ ........ 16 6 3 80 B Coefficients Red La Soda Sun scald Reflectance ................................ ..................... 16 7 3 81 B Coefficients Red La Soda Sunscal d 1st Derivative Reflectance .............................. 16 7 3 82 Damage Prediction Using PLS Fabula Brown Rot Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ .............................. 16 8 3 83 Damage Prediction Using PLS Red La Soda Brown Rot Reflectance (l eft) 1st Derivative (right ) ................................ ................................ ................................ ............ 16 8 3 84 Damage Prediction Using PLS Combined Brown Rot Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ .............................. 16 9 3 85 B Coefficients Fabula Brown Rot Reflectance ................................ ............................ 16 9 3 86 B Coefficients Fabula Brown Rot 1st Derivative Reflectan ce ................................ .... 1 70 3 87 B Coefficients Red L a Soda Brown Rot Reflectance ................................ .................. 1 70 3 88 B Coefficients Red La Soda Brown Ro t 1st Derivative Refle ctance ........................... 1 7 1 3 89 Damage Prediction Using PLS Fabula Growth Cracking Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ ............ 1 7 1 3 90 Damage Prediction Using PLS Red La Soda Growth Cracking Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ ............ 17 2 3 91 Damage Prediction Using PLS Combined Growth Cracking Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ ............ 17 2 3 92 B Coefficients Fabula Growth Cracking Reflectan ce ................................ .................. 17 3 3 93 B Coefficients Fabula Growth Crackin g 1st Derivative Reflectance .......................... 17 3 3 94 B Coefficients Red La Soda Growth Cracking Reflectance ................................ ........ 17 4 3 95 B Coefficients Red La Soda Growth Crac kin g 1st Derivative Reflectance ................ 17 4 3 96 Damage Prediction Using PLS Red La Soda Insect Damage Reflectance (l eft), 1st Derivative (right ) ................................ ................................ ................................ ............ 17 5

PAGE 15

15 3 97 B Coefficients Red La So da Insect Damage Reflectanc e ................................ ............ 17 5 3 98 B Coefficients Red La Soda Insect Damag e 1st Derivative Reflectance .................... 17 6 3 99 st Derivative (right) ................................ ................................ ................................ ............ 17 6 3 100 st Derivative (right) ................................ ................................ ................................ ............ 17 7 3 101 Damage Prediction Using SMLR Combined Internal Bruising Reflectance (left), 1st Derivativ e (right) ................................ ................................ ................................ ............ 17 7 3 102 (right) ................................ ................................ ................................ .............................. 17 8 3 103 tive (right) ................................ ................................ ................................ .............................. 17 8 3 104 Damage Prediction Using SMLR Combined Shatter Reflectance (left), 1st Derivative (right) ................................ ................................ ................................ .............................. 17 9 3 105 ive (right) ................................ ................................ ................................ .............................. 17 9 3 106 Derivative (right) ................................ ................................ ................................ ............ 1 80 3 107 Damage Prediction Using SMLR Combined Greening Reflectance (left), 1st De rivative (right) ................................ ................................ ................................ .............................. 1 80 3 108 (right) ................................ ................................ ................................ .............................. 18 1 3 109 st Derivative (right) ................................ ................................ ................................ ............ 18 1 3 110 Damage Prediction Using SMLR Combined Sunscald Reflectance (left), 1st Derivative (right) ................................ ................................ ................................ .............................. 18 2 3 111 1st Derivative (right) ................................ ................................ ................................ .............................. 18 2 3 112 Derivative (right) ................................ ................................ ................................ ............ 18 3

PAGE 16

16 3 113 Damage Prediction Using SMLR Combined Brown Rot Reflectance ( left), 1st Derivative (right) ................................ ................................ ................................ .............................. 18 3 3 114 (right) ................................ ................................ ................................ .............................. 18 4 3 115 nce (left), 1st Derivative (right) ................................ ................................ ................................ ............ 18 4 3 116 Damage Prediction Using SMLR Combined Cracking Reflectance (left), 1st Derivative (right) ................................ ................................ ................................ .............................. 18 5 3 117 e Reflectance (left), 1st Derivative (right) ................................ ................................ ................................ ............ 18 5

PAGE 17

17 LIST OF ABBREVIATIONS a acceleration B B coefficient CCD Charge Couple Device CI Confidence interval CVTEST Cross Validation d Distance F Impacting load g gravitational force h Drop height h2 Rebound height I Intensity m Mass NIRS Near infrared spectroscopy p Sample proportion PCR Principal components regression PDI Potato damage index PLS Partial least squares PRESS Predicted Residual Sum of Squares PTFE Polytetrafluoroethylene r Correlation coefficient R 2 Coefficient of determination RH Relative Humidity RM SD Root mean square difference RPD ratio of prediction to deviation

PAGE 18

18 RSS Residual sum of squares SDBI Surface damage and bruising index SMLR Stepwise multiple li near regression SEC Standard error of calibration SEP Standard error of prediction UV Ultraviolet v Velocity VNIS Visible and Near Infrared Spectroscopy W Weight

PAGE 19

19 Abstract of Thesis Presented to the Graduate School of the University of Florida in Partial Fulfillment of t he Requirements for the Degree of Master of Engineering EVALUATING VINE KILL HARVESTED POTATOES TO DETERMINE EFFECTS OF HARVEST WAIT PERIODS ON DAMAGE RESISTANCE AND TO DETECT SURFACE AND SUBSURFACE DAMAGE WITH SPECTRAL MEASUREMENT By Michael Anthony Brecht December 201 2 Chair : Lee M ajor : Agricultural and Biological Engineering Any force exerted on a potato tuber during harvesting and packaging may result in mechanical damage. Mechanical damage and disease are majo r problems in the harvesting and movement of potatoes from the field to consumer. Being able to quickly identify and remove damaged pota toes during post harvest handling could reduce storage costs for potatoes which never make it to market. Excessive deter ioration in market quality of potatoes from mechanical damage and defects cut into the profits of the producers and shippers. V ine kill is a practice which involves killing the upper portion s of plants by spraying with herbicides or cutting the vine s Afte r the upper portion of the plant has been killed, the potato tubers are left in the soil undergo es skin set which involves the periderm thicken ing and becoming more damage resistant. This thickening of skin also helps to prevent storage diseases and shrink age due to w ater loss However, leaving tu bers in the ground can increase t heir exposure to brown rot, insects and other deteriorating conditions. The two cultivars used for this study were Fabula and Red La Soda A study was carried out to simulate potato handling in a repeatable manner to investigate the damage thresholds of each cultivar at different points from harvest through packing Potato

PAGE 20

20 tubers were collected 7 days before vine kill, the day before vine kill and then 7 14 and 21 days after v ine kill A set of potatoes from each cultivar w as run along the same portion of a packing line each harvest to access skinning resistance, while the remaining tubers were used for drop tests. Drop tests were performed 1 day and 7 days after harvest ; th e n stored to allow internal bruising and shatter to develop before being assessed. Tubers were dropped from the following heights; 30 cm, 30 cm double 60 cm, 6 0 cm double and 90 cm Analysis of Fabula data showed a clear decrease in skinning and drop dama ge between pre vine kill and only 7 days of skin set; w hile Red La Soda maintai ned consist e ntly low skinning damage, but during drop testing there was a remarkable decrease in damage development after 7 days of skin set A study was also carried out in o rder to collect spectral data of vine killed potatoes for the purpose of grading and sorting both cultivars. Spectral measurements w e re made on undamaged potatoes af ter harvest, after simulating mechanical damage and then following 7 days of storage at 20 C ( 68F ) in order to allow internal and external damage to develop. Spectral data w ere also collected for tubers showing greening, brown rot, shatter growth cracking, sunscald and insect damage. After collecting data for the wavelength range between 200 2 500 nm for moisture content it was decided to concentrate on the sensing range between 360 800 nm due to water having less effect on reflectance measurement and equipment pricing of spectrometers for packing houses Analysis showed that undamaged reflecta nce properties of each tuber cultivar were altered in predictable ways depending on which defect wa s present. The change to the spectral reflectance caused by defects was found to be statistically significant using PLS and SMLR analysis; which would allow the potential of detection and e limination of damaged tubers on a pack ing line

PAGE 21

21 CHAPTER 1 LITERATURE REVIEW 1.1 Introduction The United States is the fourth largest potato producing country in the world with a yearly production of around 20 million metric tons, while the yearly production of potatoes i s approximately 325 million metric tons for the entire world. The United States however i s the leader in metric tons produced per hectare with production averaging around 44.6 metric tons/hectare compared t (FAO, 2008) The average Florida potato yield is 48 metric tons/hecta re, with a marketable yield of 41.5 metric tons/hectare (Hutchinson and Gergela 2007). This means there is an average of 9% loss of prod uct during packaging and storage. Only about one third of potatoes grown in the United States are consumed fresh. Around 60 percent of annual U.S. potato output is processed into frozen products (such as frozen fries and wedges), chi ps, dried products and starch production while 6 percent is re used as seed potato for future production Each American eats more than 54 kg of potatoes every year (FAO, 2008) While Florida is not one of the top ten potato producing states in the United States by volume produ cing only 2% of the annual supply in the U.S.; it ranked 7 th for the high value winter and early spring potato production in which Florida produces one third of the winter/ early spring crop supply $1 6 0 million in annual revenues for the state due to its profitable production window ( USDA 2008) Mechanical damage in agricultural production is a major cause of low grade quality (USDA, 2008). The mechanization of agricultural processes in the U.S. has favored the development of increased mechanical damage on produce. Mechanical damage that occur s o n

PAGE 22

22 potato es inc ludes skinning, internal black spot, pressure bruising and shatter. Skinning is an injury in which the tuber periderm is remo ved by friction. Internal black spot is a discoloration of the tissue below the periderm due to an impact on the potato tuber surface damaging internal cells ; these damaged cells then generate the black pigment called melanin A pressure bruise is a depression of the tuber surface that o ccurs due to a constant external pressure at the point of contact between adjacent tubers or storage surfaces. Shatter is the result of a mechanical impact that causes internal or external splitting (cracking) of the tuber. M echanical damage diminishes the income of potato farmers by reducing the quality and quantity of potatoes that are harvested from their fields. The elimination of produce that has been mechanically damaged during harvest and reduction of mechanical damage caused during processing would help improve the quality of product and increase profits by eliminating the storage costs associated with storing these damaged potatoes. Potatoes that have been stressed from mechanical damage or disease also reduce the income of farmers by increasing the load requirements of cooling and moisture maintenance of storage areas. The above ground portion of potato plants is killed prior to harvest in order to promote tuber periderm development (i.e. Potato tubers that have been allowed to underg o skin set after vine kill tend to have a much lower chance of developing mechanical damage. But the extended time spent underground exposes the tubers to other major causes of lowered grade quality such as soft rot and other decays due to various pathogen s, black heart freezing, and insect damage. These damaging defects have a more likely chance of occurring the longer a tuber is left in the ground before and after vine kill.

PAGE 23

23 1.2 The Potato Potatoes ( Solanum tuber o sum ) belong to the (Solanaceae) of flowering plants and share its genus Solanum with at least a thousand other species, including tomatoes and eggplants These herbaceous annuals grow up to 100 cm tall and produce a tuber, which is referred to as the potato, that is so rich important food crop behind maize, wheat and rice (FAO, 2008). As a potato plant grows, sugars are produced in the leaves that are transported to and then stored in its underground stems as starch. These s tems are separate unique organs, distinct from the roots. As these stems thicken at the ir distal end s from storing starch, they begin to form tubers (potatoes). The number of tubers present in a fully matured potato plant depends on the availability of nut rients and moisture in the soil. The potato that is grown today is a single domesticated botanical species of potato that contains just a fragment of the genetic diversity developed by farmers in the central Andes Mountains region over the course of millen nia ; that contains thousands of varieties with diverse sizes, shapes, skin or flesh color and other sensory characteristics. The potato varieties that are hundred s pecies of wild potatoes, which are found in the Americas. There are seven recognized potato species and 5,000 varieties grown in the Andes (FAO, 2008). The end of a growing and stems to die down to soil level and its tubers to detach from their stolons (FAO, 2008). Once the plant begins to naturally die back at the end of the season, the tubers undergo a process of skin set which allows them to toughen u p for the harsh cold w eather so that they can later re grow into new plants when environmental conditions improve.

PAGE 24

24 1. 3 Harvesting A fter Vine kill Harvesting after v in e kill is a practice that involves killing the upper portions of the potato plants by either spraying them wit h herb icides by mechanically cutting or a combination of both methods; the potato tubers are then left in the soil to allow their skins to set (Hutchinson and Stall 2007) S kin set helps prevent storage diseases, skinning and shrinkage due to water loss. However, the longer tubers are left in the ground the more exposure they receive from the fungal incrustation called black scurf decay s such as brown rot, damage from insects eating portions of the tuber and rot caused by heavy rains and other harmful e nvironmental conditions Vine kill is also performed in order to prevent spread of virus diseases by aphid infestations, to kill and to prevent spreading of late blight spores on foliage to prevent spreading and reduce late blight tuber rot infection, to control tuber size by terminating above ground plant growth in order to take advantage of early market demand, and to improve removal of vine growth that interferes with harvest (WPC, 2 011 ). After the vines are killed, sunscald and greening (i.e., chloroph yll synthesis in the tuber epidermis) can affect tubers that are not sufficiently covered with soil and are no longer protected by vegetative cover To allow for less problematic harvesting, potato vines are currently killed 2 to 3 weeks before the potatoe s are scheduled to be harvested Production scale then determines the harvesting method that will be used; potato tubers can be harvested using a spading fork, a plough or commercial harvesters that unearth the entire plant and remove the soil from the tub ers (FAO, 2008) Tubers harvested from living vines are more likely to be severely skin ned and bruise d during harvest ; t he ir heavy green foliage may also interfere with harvest machinery (Olson and Simonne 2006). Immature tubers from living or very recen tly killed vines are typically more susceptible to the skinning and mechanical injury that can take place during harvest and sorting.

PAGE 25

25 These immature tubers also tend to have higher sugar and lower starch concentrations than mature tubers. According to Hutc hinson and Stall (2007), the maturity of tubers at harvest is an important aspect in the production quality of Florida f resh market potatoes. Tubers that have reached suitable maturity have a better level of skin set, enhanced bruising resistance and impro ved storage life. Vine killing before harvest has an additional benefit wherein it improves tuber release from potato vines In its area of origin and most of the major commercial growing regions in temperate areas t ubers are able to natu rally mature as t he potato plant senesces due to changes in weather ; h owever, improved production methods and farming in areas without distinct seasonal changes allow potato vines to remain healthy and green longer into the growing season preventing tubers from maturing an d beginning the skin set process This is the reason why t uber maturation is artificially induced by killing the potato vines at ground level Harvesting is usually done 14 to 21 days after vine kill in order to allow the periderm to set and to reduce the amount of skinning and scuffing. Senescent vines and vines in poor condition due to disease or other factors do not have to be killed as long before harvest as vigorous vine s in order for the periderm to be set (Olson and Simonne 2006). D ue to the natura l process by which tubers begin to mature when harsh conditions begin to kill off plant foliage above the surface in order to survive until favorable conditions return 1. 3.1 Mechanical Vine Killing Mechanical vine killing generally refers to the destruct ion of potato vines by disintegration of the vines by beating, chopping or burning Mechanical vine killing is often used to terminate tuber growth, to ease harvesting by destroying vine structure, or to take advantage of an early market (Murphy, 1968). Fl ail mowing and rolling are the two of the most popularly used

PAGE 26

26 methods of mechanical vine kill for potato harvesting It is suggested that these methods be used 14 to 21 days preceding the planned harvest period to insure ample time for tubers maturation an d skin set Since revolving flails or blades can damage tubers and create air flow that can expose tubers by moving loose soil, c are should be taken when using mechanical methods to avoid disturbing the soil so that tubers close to the surface are not expo sed, leading to greening sun scald or mechanical damaged (Hutchinson and Stall 2007). Burning potato vines is a method of destroying vines that utilizes a series of burners directed to obtain a uniform heat or flame coverage on the vines. Complete inciner ation often requires more than one flaming and can be more expensive than chemical methods. Burning of the vines also destroys organic residues and nitrogen needed for soil maintenance (Murphy, 1968). Other methods of mechanical vine killing include pullin g and steaming. 1. 3. 2 Chemical Vine Killing The most common method of vine kill ten ds to be chemical desiccation. This method involves the application of agricultural grade chemicals to kill the potato vines over a varying number of days ; depending on the type used it can take 1 day and a single application or 2 3 weeks and several applications to achieve full vine kill According to Hutchinson and Stall (2007), chemical vine desiccants should not be applied during cool and damp or hot and dry weather in or der to avoid stem end discoloration from rapid vine killing Vine kill can also be improved for actively growing plants by splitting application of the chemical desiccant; this can also help skin set development Split application can only be implemented i permits for it; this is performed by an application of desiccant at less than full rate followed by a second application several days later (Hutchinson and Stall 2007). Chemical vine kill is usually

PAGE 27

27 used to terminate vine growth ove r time, to prevent oversized tubers, to harden tuber sk ins, and kill late blight spore and prevent spread of tuber rot (Murphy, 1968). 1. 3 .3 Combined Methods Combining mechanical and chemical methods to perform vine kill can improve the efficiency of vine desiccation and speed up the tuber maturatio n process of skin set (WPC, 20 11 ) The use of a roller in order to bend and partially kill vines ; whil e spraying chemical desiccant results in improved foliage and stem coverage a nd more efficient chemical vine kill Combining the methods allows for fewer applications of chemicals in order to achieve a faster complete vine kill. Another advantage of combining v ine rolling with chemical application is that it may close cracks in the potato row that results in a re duction of incidence s of tuber greening and sun scald after vine desiccation by preventing tuber contact with sunlight (Hutchinson and Stall 2007). For a very rapid vine kill, flail vine shredders can be combined with a chemical application after a short w ait period to allow debris to fall off of the remaining vine before application. This flailing reduces the chemical desiccant ap plications required for completing vine kill and resu lts in a very fast rate of vine kill whi ch allows the vine kill procedure t o be timed closer to the harvest date and thus extends the length of time available for the tubers to bulk up and mature physiologically (WPC, 2011 ). 1. 4 Packaging and Storage During commercial production, the p acki ng line process begins when trucks bring tubers from the field and dump freshly harvested potatoes onto the packing line in feed elevator, which takes the fresh potatoe s through an initial wash area utilizing overheard watering and brushes to move the potatoes forward (Fig. 1 1). Once the potato es are washed they go through a sorting

PAGE 28

28 area where potatoes with defect s, injuries and/or disease s are removed. The potatoes that make it past the sort table are then taken by elevator to an accumulator before entering the primary wash area. In the primar y wash area, potatoes are moved forward by brushes again while being watered from above before entering a sponge roller area that takes them through the sizer. The sizer separates potatoes by using slowly increasing gaps to drop different sized potatoes on to divided sections. Once the potatoes have passed the sizer, they enter the drying area where most of the surface moisture from the two washes is removed before entering the grading table where the final inspection of potatoes is performed before proceedi ng to the packing area where potatoes are put into their storage containers. Newly har vested tubers are living tissue making proper storage essential. When storing potatoes, facilities need to be designed to keep the potatoes alive and healthy, and must b e able to slow the natural tuber senescence which includes the conversion of starch to sugar ( WPC, 2011 ) It is important to store potatoes in a dark, well ventilated environment with an optimum relative humidity of 90% at a temperature of 3.9 to 10 C (39 to 50 F) depending on the variety These storage conditions are intended to prevent greening and losses in weight and quality Several distinct storage phases exist which will ideally allow potatoes to be stored up to 10 months (Calverley, 1998) Mathew an d Hyde (1997) found that potato tubers which had been stored over 5 months rarely show any signs of blackspot bruising. The four main storage phases are curing, cooling, long term storage and m arketing. The best practice for each storage phase depend s on t uber conditions weather and projected use of the crop. New potatoes grown in Florida do not undergo the curing process because they are produced for the table stock market and sold shortly after harvest.

PAGE 29

29 1. 4 .1 Pre Harvest Period Most storage problems be gin in the field before harvest begins. Potatoes from healthy vines have been shown to be much more resistant to storage decay than those harvested from vines that had been weakened from physiological stress or diseases. Good storage practices and manageme nt can help maintain the tubers, but poor quality produce that has already been damaged will not improve during storage The storage facility shoul d be prepared well in advance of harvest in order to allow time for proper cleaning and disinfection. 1 .4 .2 C ooli ng Period The greatest amount of water loss (shrink) occurs after harvest is complete and potato tubers are placed in to storage The potato pile temperature is initially maintained in the range of 4.4 to 12.8 C ( 40 to 55 F) depending on pile condition s and ultimate use Tubers destined to become chips are generally stored at 10 to 12.8C (50 to 55F), while french fries are stored at 7.2 to 8.9C (45 to 48F), and finally fresh market and seed potatoes are maintained near 4.4C (40F) (Bohl and Johnson 2010). Afterward the potatoes should be warmed to 10C (50F) to prevent bruising damage before being removed from storage or nee d be cooled to the long term storage temperature at a rate of 2 to 3C (4 to 5 F) per week, s ince rapid cooling of potatoes h as a tendency to cause colorati on problems in processed potatoe s (Bohl and Johnson 2010). The ventilation used during c ool ing is determined by the cooling requirements, the need for fresh air to remove the products of respiration and for maintaining relat ive humidity (Brook et al., 1995) Ventilating is mainly done to maintain the pile at the desired temperature, but ventilation may be required more frequently if there is excess condensation on the tuber surfaces.

PAGE 30

30 An excess of surface moisture can encourag e soft rot, so continuous ventilation is often recommended when condensation is present. 1. 4 .3 Long term Storage Period The main objective for potatoes destined for long term storage is to maintain a consistent ideal environment for the entire storage per iod to maintain tuber quality while facilitating marketing Recommended storage temperatures are determined by condition, as well as, intend ed use Sugar levels in potato tubers can increase during storage due to conversion from starch, which is favored at lower temperatures Soluble sugar content greater than 0.5% is very undesirable for potatoes intended for processing, especially frying, beca use the sugars are caramelized at cooking temperatures. It is generally accepted that no more than a 1 C ( 1 .8 F) d ifference should be maintained between the top and bottom of the pile (Brook et al 1995). S torage temperatures can be maintained more uniformly if conti nuous ventilation i s used; especially during periods when outdoor temperatures are considerable low er A relative humidity above 95% is recommended for long term storage. 1. 4 .4 Marketing period Reconditioning after long term storage is a procedure that improves the color of processed potatoes. This is accomplished by increasing s torage temperatures to 10 to 18 C (50 to 64 F) for 2 to 4 weeks before marketing. These higher temperatures increase tuber respiration rates which reduce s the sugar amount of soluble sugars that may have accumulated in the tubers during storage by using those sugars as substrate The reduced sugar levels improve chip and fry coloration when cooked. These higher storage temperatures also increase shrinkage, possibility

PAGE 31

31 of rot and may break the dormancy of sprouting in tubers. Thus it is recommended that processing occur within 1 mon th of reconditioning (Brook et al., 1995) 1. 5 Potato Quality Factors A potato tuber s size, shape, and physical appearance, as well as the pre sence or absence of diseases defects and damage all contribute to potato quality. Pathogens affect all parts of the potato plant, and disease can develop at any time during the potato production and storage season. These pathogens can cause extreme reductions in yield and quality of a potato harvest individually an d as complexes (Stevenson et al. 2001). Potato tu bers can also display a number of defects and disorders that are not the result of disease or insects. Some of t hese external and internal defects and disorders are caused by mechanical damage. External tuber damage can reduce the marketability due to unfa vorable reactions by consumers and also cause reductions in processing quality and storability. Internal tuber damage usually goes undetected until after tubers are cut and inspected unless it is extensive enough to reach the surface of the tuber. This typ e of damage can also result in sever e reductions in crop quality, marketability and storability. 1.5 .1 Mechanical Damage of Potatoes Any mechanical force exerted on a potato tuber during harvesting and packaging may result in mechanical damage. Damage to potato tubers due to mechanical forces is among the most important causes of loss of quality reported throughout the world (Peters, 1996). This mechanical damage is a major problem in the harvesting and movement of potatoes from the field to consumer. Exce ssive deterioration in market quality of potatoes caused by mechanical damage can cut into the profits of the producers and shippers (Nylund et al. 1955) Mechanical damage costs the potato industry by increasing storage losses due to shrinkage and diseas e,

PAGE 32

32 increasing labor costs for trimming and inspecting, increasing the cost of raw product through greater trim losses, lowering the quality of final product, increasing the incidence of disease and decreasing shelf life, and finally by reducing the appeal of fresh potatoes to wholesale and retail customers (Thornton and Bohl 1998). Fresh market potato consumer standards are usually associated with visual characteristics such as shape, firmness and exterior appearance being free from defects and disorders. Mechanical damage can occur in several different forms including minor skin abrasions ( skinning ) external shatter, i nternal shatter, cuts and black spot bruising These main types of damage and their causes are listed ( T able 1 1 ) Mechanical damage is one of the most important causes of loss of potato quality reported throughout the world. In a U S study by Preston and Glynn (1995) it was reported that for the total potato production in the United States valued at $2 billion, losses of at least $125 mill ion occurred between potato fields and the consumer Skinning is the most common type of mechanical damage and if severe enough can result in USDA grade losses for the product ( T able 1 2 ) While s lightly skinned tubers are able to heal with proper storag e, these tubers can still result in increased shrinkage rates from water loss and from tubers using stored energy to repair the damage E arly and later blight infection can sometimes occur on these skinned tubers that have lost their protective periderm (Ste venson et al. 2001). Other mechanical damage, such as bruising and shatter can also result in a loss of grade when present at moderate to s erious levels that detract from the edible or marketing quality R emoval of injured tissue from bruised and shatter ed tubers causes l oss es of 5 to 10 percent of the total weight of the potato (USDA, 2008).

PAGE 33

33 1.5 .1 .1 C auses of mechanical d amage M echanical damage to potato tubers can occur at any of several stages of production, especially during ha rvesting and sorting operations; which vary by the variety of potato being processed. Blahovec (2005) analyzed the effects of different cultivation conditions on the only slightly depen ded on cultivation regime ; implying that a varieties damage resistance could impact mechanical damage more than conditions during the growing season In o ne American investigation it was reported that 70% of the damage done to potato tubers occurred at ha rvest and the remaining 30% occurred during transport and storage (Peters, 1996). Thornton and Bohl (1998) reported that there are four factors that have a major influence on the amount and severity of mechanical damage: 1) soil conditions, 2) tuber condit ion, 3) equipment maintenance, adjustment, operation and modification and 4) magnitude of tuber impact. Chiputula et al. (2009) harvesting and packaging operations; th e damage found were skinning, external shatter, cuts, internal shatter and blackspot bruising, that were mainly caused by harvesting operations. Skinning can often be the result of rough handling of immature potato tubers which causes the skin to be easi ly scuffed or rubbed off. Black spot bruising occurs when a potato tuber undergoes a significant impact against a solid object damaging cells in the tissue just beneath the skin without actually breaking the skin ; that results in the damaged tissu e develop ing a black pigment over the course of 24 to 48 hours called melanin Shatter bruising occurs when impacts cause cracks or splits in the potato tuber skin and/or extend ing into the underlying tissue (Thornton and Bohl, 1998).

PAGE 34

34 Sk inning and cutting are ofte n caused by mechanical harvesters as they dig the tubers out of the soil D umping and tumbling of potatoes on mechanical equipment during harvest ing and unloading onto packaging lines can also lead to skinning and cutting, as well as, external shatter, int ernal shatter, and blackspot bruising. Nylund et al. (1955) investigated the causes of mechanical damage during harvesting and handling operations and found that diggers, trucks, unloading and bin to shipping bag operations all contribute to mechanical dam age which can lower the U.S. grade number from 1 to 2 or less. Mechanical damage can be linked to drop height and potato tuber size due to the increase in impact force that is created by increasing either of those factors Thornton and Bohl (1998) reported that research has shown the greater the drop height the larger the force, and the more likely that a bru ise will occur, especially when drops are greater than 6 inches and the tuber is supported underneath b y a hard metal roller or plate. 1.5 1. 2 Mechani cal damage p revention Prevention of mechanical damage often begins before planting with good plowing etiquette to avoid the formation of clods in wet, heavy textured soil and the removal of rocks, so that when it comes time to harvest there is less debris to come in contact with potato tubers on harvester conveyors (Thornton and Bohl, 1998) The next step is to plant in time to allow potato plants to reach desired maturity before performing vine kill and harvesting at the end of the season. Vine killing is a vital step in mechanical damage prevention; allowing skin set to toughen tuber skins and allowing tubers to reach proper maturity increases resistance to skinning and shatter bruising. Thornton and Bohl (1998) reports that providing adequate fertilizati on of potassium and calcium enhances the ability of potato tubers to heal wounds and also reduces susceptibility to black spot bruising.

PAGE 35

35 At the end of the season is it very important to try and harvest under as near ideal temperature and soil moisture cond itions as possible, in order to maximize the number of b ruise free potato tubers acquired (Thornton and Bohl 1998). When potato tubers become dehydrated and limp the y are more susceptible to black spot bruising, while being over hydrated and crisp potato tubers to shatter bruise more often. If soil moisture contents d rop below 50 percent after vine kill, tubers become more likely to dehydrate excessively and irrigation should be used to supplement a lack of rainfall. Generally, as tuber temperature increases, less bruising occurs. However, i t is considered ideal to harvest and handle potato tubers when temperatures are between 10 C and 15.6 C (50 F and 60 F) because temperatures above 18 C (65 F) promote storage decay (Thornton and Bohl, 1998). Dry, sandy soil can separate from potato tubers too quickly as they move over the harvester conveyors which need to carry soil to the end of the secondary chains to operate properly An under loaded conveyor increases tuber irritation and can lead to higher levels of mechanical damage. Thornton and Bohl (1998) suggested that light irrigation should be applied to soil before potato harvest in order to partially moisten the soil and overcome problems with excess soil separation on conveyors. Furthermore, if irrigation is not an o ption, it was recommended that the forward speed of harvesters should be increased in order to increase the load of tubers and soil on primary and secondary conveyors. On the other hand it wa s suggested that when soil is too mois t at harvest, the forward speed should be decreased in order to reduce the load of tubers on the primary and secondary conveyors to offset an increase in bed agitation of potato tubers.

PAGE 36

36 1.5 .2 Environmental Physiological Defects of Potatoes The potato is a host to many pathogenic organisms that include bacteria, fungi, viruses, viroids, nematodes, and phytoplasmas ; however fungi, bacteria and viruses are the domina n t causes of potato diseases (Bohl and Johnson 2010) These pathogens can cause extreme red uctions in yield and quality of a potato harvest individually an d as complexes Adverse environmental conditions and certain agronomic factors can also negatively affect the health of the potato plant and quality of harvest (Stevenson et al., 2001). Some of the common defects that occur in Florida potato fields and their causes are listed ( T able 1 3 ) Controlling pathogens and preventing environmentally caused defects is central to successful, quality potato production. Many pathogens enter potato tubers during harvest when tuber periderms are damaged by mechanical harvesting. These pathogens go unnoticed until they begin to decay the affected tubers during storage. Other pathogens are transmitted by insects serving as the mechanical vector or their feedin g providing an entry point for pathogen invasion into the tuber flesh (Bohl and Johnson, 2010). The general practices to prevent and contr ol diseases suggested by Bohl and Johnson (2010) include: 1) Handling seed tubers sanitarily 2) following a regular and rigorous sanitation program, 3) irrigating uniformly and adequately but not excessively, 4) controlling aphids, leafhoppers, and nematodes, 5) harvesting and handling tubers gently, 6) never harvesting when tuber temperatures are below 7 C (45 F) or above 30 C (86 F) and 7) providing an environment conductive to wound healing, followed by proper temperature, humidity, and aeration during storage. Defects caused by environmental conditions cannot be controlled by growers, however, proper management c an help promote uniform growth and minimize the impact of environmental stresses. This includes establishing a uniform stand, monitoring soil moisture, irrigating in a timely manner, using soil tests to help fertilize for

PAGE 37

37 reasonable yield goa ls and using proper storage conditions to reduce losses (Bohl and Johnson, 2010) 1.6 Reflectance Measurement V isible and n ear infrared spectroscopy ( VNIS ) is a nondestructive method of product evaluation that uses the phenomena associated with visible light and near i nfrared reflectance and transmission the instrumentation, laboratory analysis, and data processing to detect differences in the physical and compositional status of the measured samples When a material is exposed to visible light it may be selectively r eflected, absorbed or transmitted; rendering wavelength dependence to the emergent radiation that is perceived as color. Similarly, near infrared radiation may be selectively reflected, absorbed or transmitted by a sample. This phenomenon is well known, pr oviding a basis for structural study and quantitative analysis using transmission spectroscopic techniques (Dahm and Dahm 2001). According to Yee and here are four advantages that have contribute d to the popularity of near infrared s pe ctroscopy ( NIR S) : 1) t he fact that samples take very little preparation and a spectrum can be acquired quickly. 2) The process is non destructive and t he same sample can be retained for other analytical procedures or returned to the population. 3) Analysis can be performed to determine more than one constituent from a single scan or spectrum 4) There is no dependence on highly skilled personnel to operate the instrumentation required to perform scans 1.6.1 Effect of Water on Spectra l C haracteristics It i s well known that water strongly affects the reflectance spectra of sample materials at various wavelengths. Yee (1999) reported that spectral analysis is widely used for moisture determination in food, but different chemical constituents of produces pose

PAGE 38

38 problems for accurate measurement of moisture due to shifts of peak spectral responses and changes in peak spectral responses in the absorption bands. Water is the most important substance necessary for plant growth and water content varies extensively in potatoes Potato tubers have been reported by the potato industry to contain huge amounts of water (73.8% to 81%) according to Souci et al (2000). While experimenting with determining potato tuber water content using NIR diffuse reflection methods Elbatawi et al. (2008) found water content levels from as low as 40% to as high as 85% with average water contents around 81.7%. Water is a n excellent absorber of infrared energy at several wavelengths, which can strongly affect reflectance spec tra of potatoes Carter (1991) discovered that water content had primary and secondary effects on leaf spectral reflectance. These primary effects resulted from the radiative properties of water, while the secondary effects were those that could not be exp lained solely by these properties. Generally a s the water content of the leaves decreases, reflectance increased throughout the entire wavelength range measured (400 2500 nm) with the greatest sensitivity to water content being between 1,300 and 2,500 nm, especially near the water absorption bands at 1,450, 1,940, and 2,500 nm Additionally, Carter (1991) also determined that a secondary effect occurred between 400 720 nm that resulted in decreased absorption by pigments. (1999) stated th at the fundamental absorption of water occurs at approximately 3,840 nm of the infrared region with the first, second and third overtone at approximately one half, one third and one fourth of the fundamental (1920, 1280 and 960 nm, respectively), but in re ality the overtones will not follow this exact relationship due to losses and inefficiencies occurring between molecular interactions. By performing a full spectral analysis of the suspected water absorbing bands Yee (1999) was able to find strong evidence for 5 maximum wavelengths between 0.7 2.5 m at 0.76, 0.97, 1.19, 1.45, and 1.94 m.

PAGE 39

39 Muir et al (1999) determined that water loss from the tissue in and around an area could be attributed to pathogens such as common scab, dry rot, skin spot and gangrene ; b ut it could not be linked to the somewha t wetter rots associated with bacterial soft rot or blight because the tissue will not dry out to the same extent. Fassio and Cozzolino (2004) used a NIRS spectrum between 1100 and 2500 nm in order to predict the chemical composition of sunflowers along wi th their water content ; c oncluding that re calibration may be necessary for moisture and oil over different harvests due to climatic or agronomic conditions. Bull (1991) suggest ed that i f the reflectance at the reference wavelength is sensitive to the mois ture content of the sample the calibration will only be linear over a relatively small range of moisture contents. This may be inconvenient in a practical instrument. In this case, it is preferable to choose closely matched absorption and reference wavele ngths for which the reflectance is relatively insensitive to moisture content. 1.6.2 Application of Spectroscopy in Food Analysis Nondestructive detection such as, photoelectric detection, electromagnetic characteristics analysis, NI R S X ray analysis, com puter vision and so forth, have been used more frequently in food and agricultural indus tries for product inspection and evaluation in order to provide reliably objective assessment at a rapid rate with an economic price tag (Jin et al 2009). After deter mining a wide range of parameters present in food products s pectroscopic technique s began to receive worldwide attention. main advantages are its speed, little to no sample preparation, and the absence of chemical use However, well designed calibration processes must be performed in order to allow for the best model predictions of specific parameters of interest in the desired food product (Singh et al. 2005) Porteus et al. (1981) reports that the interaction of only 3 causative agencies c an account almost entirely for the range of spectral types recorded

PAGE 40

40 from tubers suffering from a number of defects through the a pplication of the statistical techniques of factor ana lysis and discriminant analysis. Using the information from only a few wav elengths, i ndividual defects can be detected instrumentally with about 80%. The detection of some diseases can even be determined before their presence causes effects that are visible to the eye. With further instrumentation it would be possible to constru ct a device that could recognize a number of defects in tubers with an acceptably low failure rate. Numerous nondestructive sensing techniques have been studied for predicting firmness, sugar or soluble solids, and other quality attributes of apples and ot her fresh agricultural products. Light based sensing technologies, especially NIRS, offer great potential for predicting i nternal quality of fresh fruits and has the potential to nondestructively measure multiple quality attributes of agricultural products simultaneously (Lu and Ariana 2003 and Kang et al 2004 ). The nondestructive technology NIRS has received significant consideration as a means for detection of fruit quality. The rapid nondestructive technology NIRS is fairly easy to use with online and offline applications. NIRS also has the potential to simultaneously evaluate numerous quality traits of apples In order to be adapted for online sorting applications a sensing system must be able to obtain spectral information from a product rapidly Adv ances in charge coupled device (CCD) technology allows for r apid measurements using a spectrophotometer that obtains spectral information from all wavelengths of a spectral range in the visible and short NIR region from around 400 up to 1100 nm simultaneou sly (Lu and Ariana 2002). Porteous and Muir (1986) used light ranging in wavelength from 650 to 1680 nm to assess seed potatoes versus human assessors. M achines used to estimate the extent of disease on tubers were more reliable than human assessors ; due to the estimates varying widely between different assessors and between repeated evaluations of tubers by the same assessors. The optical

PAGE 41

41 methods investigated provide a basis for the design of quality grading machines which could compete with inspectors i n accuracy of sorting. Liew et al (2001) developed a method of using fiber optic spectroscopy to detect green and red fluorescent proteins in vitro and in vivo in order to select appropriate markers for identifying genetically engineered plants as a step towards improving public perception of bio safety over the current methods of screening plants based on antibiotic resistance. Muir et al. (1999) reported using optical spectral reflectance in the visible to near infrared range (400 nm to 2000 nm) to take measurements of 11 different varieties of potatoes grown in Scotland. These data w ere used to compare reflectance properties of disease free tubers with the properties of tubers exhibiting several types of artificially introduced disease defects Using the first 3 eigen vectors they were able to account for 96.1% of the variance in measurements. The varieties selected to represent the range of skin and flesh colors were considered in groups defined by their combination of brown, pink, yellow or white skin w ith yellow or white flesh. Lefebvre et al. (1995) reported using a combination of grey scale, f lourometry and infrared techniques to detect sprouting of potato tubers. Flourometry was used to emit an excitation wavelength at 488 nm and a band pass filter a round the emission value of 680 nm for successfully detection of sprouting Internal discoloration is one of the major defects of potatoes. O ne of the internal discolorations in potato characterized by a discolored cavity in the center of potato tubers is called Hollow heart Hollow heart previously ha d not been detectable without cutting. Kang et al (2004) set out to detect Hollow heart using a robust VIS /NIR transmittance measurement technique for determining specific gravity or dry matter in potato tub ers nondestructively with 87% accuracy. They used b oth quantitative and qualitative modeling methods were employed to

PAGE 42

42 develop models for predicting and classifying internal quality in potatoes based on specific gravity. Clark et al (2003) used a clinical MRI system to preselect 240 apples to establish a uniform range of disorder from 0 to 100%. The sample apples containing the internal browning brown heart as the percentage of browned tissue increased. The visible and near infrared spectra of bruised and intact spots on Jonagold apples were acquired from 400 to 1700 nm with spectrophotometers and by using PLS analysis a total classification accuracy of more than 90% could be achieved for detecting bruises on Jonagold apples (Xing et al., 2005). Hyperspectral imaging can be used to distinguish variations between the normal or abnormal parts of a sample using the spectra of image pixels. The technology uses the pixels from an im age to provide a spectrum using a combination of imaging, spectrometric, and radiometric techniques. Multispectral image analysis is a faster technique based on a discrete spectral analysis at a few wavelengths as opposed to the continuous spectral analysi s performed by the hyperspectral imaging technology. Mehl et al (2002) used t he se techniques to create a 100% accurate linear discriminant model to separate normal and abnormal conditions of Gala apples while their models for Red Delicious and Golden Del icious apples showed limited classification accuracy with results below 70% for normal and abnormal apples and Golden Delicious showing results of 85% accuracy. Lu (2003) used an imaging spectroscopy system to successfully acquire scattering images from a pple fruit over the spectral region between 500 nm and 1000 nm. Imaging spectroscopy goes beyond conventional imaging and spectroscopy to acquire both spectral and spatial

PAGE 43

43 information from an object simultaneously. As a result, it greatly expands the capab ility to identify or detect some subtle or minor features in an object. Equally successful studies have been performed on a wide range of defects for many fruits and vegetables. Therefore previous research results indicate that there should potentially be no limitations for application if spectroscopic techniques for the evaluation of other varieties of potatoes and defects. 1.6.3 Alternative Food Analysis: Machine Vision Machine vision systems are widely applied in automatic sorting, since they can give c onsistent and objective evaluation s compared to human inspection. Defect detection by machine vision systems is strongly dependent on the light intensity difference between sound and bruised surfaces making bi colored products difficult to screen (Xing et al., 2005). Computer vision technology has the advantages of real time, objectivity, low cost and high precision; which can effectively detect the exter nal features of potatoes Jin et al (2009) reported collected results for correct classification rate o f defects, correct recognition rate of defects and correct inspection rate of potatoes based on Fixed Intensity Interception ( FII), which operates under a controlled lighting condition and setting a CCD camera in an automatic shooting mode, that were 92.1% 91.4% and 100% respectively. The results of defect detection on yellow skinned potatoes showed this approach was fast, valid and convenient. Davenel et al (1988) developed and tested on line, a system for automatic detection of surface defects on Golde n Delicious apples on a conveyor system used for automatic colo r grading. While the fruit wa s rolling, a solid state camera took four pictures, and so was able to view most of the surface. The tests on line showed that 69% of the fruit were correctly grade d, but 26% were classified immediately above or below the right grade.

PAGE 44

44 Grading operations on packing lines are predominantly performed by trained human inspectors who inspect potatoes by looking or feeling for certain quality characteristic Human inspecto rs have a few disadvantages, which include inconsistency, a shortage of labor and slow huge production volumes. This is why it is necessary to convert grading over to a utomation in order to provide a more consistent product quality and handle large r volumes of materials The appropriate hardware and software for produce handling and grading is required to incorporate machine vision and automation into a completely automated inspection station for packing lines Potato motion can adversely affect s hape analysis and contribute any errors in classification (Heinemann et al., 1996). Tao et al (1995) trained a machine vision system to distinguish between good and greened potatoes and yellow and green es. The method created proved highly effective for color evaluation and image processing by using the HSI (Hue, Saturation and Intensity) col or system. An accuracy of over 90% in the inspection of potatoes and apples was achieved by applying multivariate d iscriminant techniques to the features that were represented with hue histograms. 1.7 Simulation of Mechanical Damage There are different degrees of mechanical damage that can be sustained at any stage of production from pre harvest operations, harvesting and subsequent handling operations when the product is graded, packed and transported for market or storage at a new location (Calverley, 1998). This damage includes anything from skinning and bruising to deep cuts; if mechanical injury is serious enough the product may be rejected during grading resulting in losses. Any damage to the tuber skin can also lead to physiological deterioration and entry of decay pathogens which can result in losses later on. Impact testing can be used to evaluate the potential for damage to occur under different conditions. Impact testing is the evaluation of an object's

PAGE 45

45 ability to resist high rate loading of energy absorbed in a collision with a t est object at high velocity (Instron, 2011). 1.7 .1 Damage Resistance of New Potat oes Impact resistance is one of the most important biological properties of agricultural products. Bajema and Hyde (1998) reported that the two components of impact sensitivity are : bruise threshold and bruise resistance. Bruise threshold of a given tuber mass and impact surface is the drop height at which bruising starts to develop Bruise resistance of a given tuber mass is the bruise energy per unit of bruise volume; while the inverse is the bruise size f or a given bruising drop height. Bruising is infl uenced by tuber mass, impact height slowdown distance and impact deceleration Impacts on a hard surfac e result in much higher impact dec elerat ion and short slowdown distance while i mpacts on soft or padded surfac es result in lower peak impact de ce lerat ion and lowered slowdown distances (Bajema and Hyde 1998) It was determined by Mathew and Hyde (1997) that the type of impact damage that occurred in potato tubers and the size of bruising damage were influenced by drop height. The percent of tuber sampl es that were damaged increased as the drop height increased for samples of 250 30 g (8 to 10 oz.) but when the drop height increased beyond 200 mm (8 in.), the number of samples with black spot bruising decreased and more critical types of tissue failure began to appear; at 450 mm (18 in.), blackspot damage occurrence dropped to zero. The mechanical damage of potato tubers can result in b lackspot bruising when sufficient damage occurs that cause s a mixing of substrate and an enzyme (tyrosine and polypheno l oxidase) that

PAGE 46

46 results in the black discoloration, melanin. D iscoloration of tissue can only occur if the substrate and/or the enzyme are present in adequate quantities. 1.7 .2 Skinning Simulation In order to determine the effects of changing conditions o n the skinning of potato tubers, tubers need to be handled in a similar manner each time and repeatable experiments must be performed. Hall (1953) used a method of skinning that involved washing tubers before placing them in a skinning apparatus consisting of a garbage can lined with in mesh hardware cloth. This can was then rotated for 3 minutes on rolls at about 10 11 revolutions per minute. Afterward each tuber sample was rated individually using a rating system of 1 (for least skinning) to 5 (for most skinning). 1.7 .3 Impact Damage Simulation When seeking to obtain the s tatistical bruise thresholds of fruits and vegetables; groups of uniform sized samples must be dropped from various known heights onto identical impact surfaces and the percent of sampl es in each group that show bruising must be noted (Bajema and Hyde 1998) The biggest disadvantage related to the testing of falling sample s is that heavier samples will impact wi th greater energy when dropped and samples must be weighed so their mass may be considered; according to the well known formula, F = ma where F = force, m = mass, and a = acceleration Fluck and Ahmed (1973) suggested that impact velocities less than 250 mm/s should be considered slow loading since a drop height of only 3.2 mm re sults in an impact velocity of 250 mm/s The equation for impact velocity is given by equation (1 1 ). (1 1)

PAGE 47

47 where h is the travelled distance of a falling object and g is the gravitational force I mpact velocities have historically been considered to range from 250 mm/s to as high as 40 m/s for agricultural products Thus, most impa cts occurring during handling of fruits and vegetables are in the fast loading rate range; and since bruising is un common at less than 3.2 mm drop height, dynamic loading need not be considered in evaluating bruising. Fluck and Ahmed (1973) found that the measured parame ter, peak force, increased with increasing drop height and increasing mass that was dropped This supports the hypothesis that peak force and resulting internal stresses are critical elements in the incidence of impact damage due to stress concentration at the tissue failure point. A large potato tuber dropped from the same height as a small tuber has more kinetic energy and will more likely sustain greater damage during handling even though smaller tubers typically have smaller radii of cur vature. Tubers with small radii of curvature are also particularly prone to damage because impact forces are concentrated upon a smaller area of the tuber surface In addition cell s ize increases as potato tuber size increases Larger cells tend to have t hinner cell walls, resulting in possibly weaker tissue in larger specimens (Bajema and Hyde 1998) The equation for impact force is given by equation ( 1 2). (1 2) Where m is the mass of the free falling potato tuber d is t he slowdown distance determined by the flex in potato flesh at impact and v is the impact velocity.

PAGE 48

48 1.7 .4 Detection and Evaluation of Mechanical Damage Finding mechanical damage means finding the physiological responses of potato tissue to physical damage Tuber tissue responds to damage by attempting to heal itself which results in increased respiration and ethylene production at the site of injury. I f the damage is sufficient enough to cause cellular disruption that mixes the substrate and enzyme involv ed in oxidation of phenolics ( i.e. tyrosine and polyphenol oxidase) black discoloration develops due to melanin production ( Mathew and Hyde 19 9 7). Bruise damage does not show up imme diately after impact. Thornton and Bohl (1998) reported that black spot bruising appears in damaged tissue over the course of 24 to 48 hours as tissue begins to progress from grey to black over time; this makes it difficult to determine how much damage was done or will develop at time of impact Blahovec (2006 ) report ed that b lackspot bruising occurs more often when impacts involve tubers of greater size, poor hydration causing limpness, potassium deficiency, warmer harvest temperature, less significant drop heights and old or very mature tubers Thornton and Bohl (1998) report ed that s hatter bruising occurs when impacts are great enough force to generate cracks or splits in the potato tuber skin and/or underlying tissue. Shatter bruising is more likely to result from impacts involving tubers of greater size, crisp texture (i.e. well hydrated tissue) cooler harvest temperature, higher drop heights and immature tubers. The decision making process involved in grading and packaging top quality potatoes is based heavily on product appearance, which requires reliance on human visua l interpretation. The majority of the actions involved in this labor intensive task are highly repetitive and boring, with accurate grading depending greatly on the effectiveness and efficiency of human operators which Muir et al. (1999) report ed can be as low as 8% and at best 60%. This makes d etection and evaluation of mechanical damage using human graders difficult particularly if the damage is

PAGE 49

49 below the surface or still developing. This has led to efforts to develop special techniques or instrumentatio n to accurately and nondestructively detect and measure such damage. Some of the non destructive methods for detecting and evaluating internal defects of fruits and vegetables include x ray absorption devices MRI, and spectral based sensing 1.7 .4.1 Vine kill wait period assessment A popular damage assessment system created by Robertson (1970) is called the Potato Damage Index which requires that tuber be washed and dipped in a paracresol, a red stain that reveals wherever the surface of the tuber is dam aged; the tubers are th en examined and separated into four categories: 1. Undamaged No staining; 2. Scuffed Skin broken but no flesh damage; 3. Peeler Flesh damage to a depth not greater than 1.5 mm; 4. Severe Flesh damage deeper than 1.5 mm. The potato damage ind ex (PDI) is calculated as shown in equation ( 1 3): (1 3) A newer index created by McGechan (1980) which accounts for both surface damage and internal bruising called the surface damage and brui sing index (SDBI) is calculated as shown in equation ( 1 4): (1 4) Many mechanical damage studies carried out on potatoes have involved varieties in growing areas that produce potatoes for storage ; these storage varieties are usually not marketed

PAGE 50

50 as es The reaction to impact force that potatoes display often depends on the variety, growing conditions and pre harvest practices. Also, s ince new potatoes d o not undergo the curing phase of storage used for wound healing ; results from damage studies conducted using storage varieties cannot always be applied to the handling of new potatoes since potatoes react differently to impact forces after harvest and aft er curing. It is common practice for researchers to design their own scales for the experiments and evaluation methods for mechanical damage. damage threshold and harvesting and packing operation components which contribute significantly to mechanical damage. Results showed that components affected each variety to different degrees and the main form of mechanical damage Research like this is rare; t herefore there is a need to investigate mechanical damage with the different new potato varieties and handling systems used in Florida potato production 1.7 .4.2 Spectral measurement assessment A primary interest of reflectanc e spectroscopy is the rapid and nondestructive assessment of certain aspects of a material. The objective is to accurately record the wavelength dependent nature of absorption relative to a stable standard (Dahm and Dahm 2001). According to the Beer Lambe rt Law, the concentration of an absorber is directly proportional to the sample absorbance as shown in equation ( 1 5): (1 5 ) with I 0 being the intensity of the incident radiation and I t the intensi ty of the transmitted radiation

PAGE 51

51 A calibration model is developed by using the NIR spectra and known analytical information in order to calculate the regression equation to create a model that can be used to predict future outcomes S everal parameters are evaluated to create a calibration model : factors, loadings, and scores It is important to avoid under or over fitting when choosing the number of factors to be used. The prediction model will be unreliable if an insufficient number of factors are used; si nce useful information may be left out of the model. However, more uncertainty will be included in the calibration set if too many factors are chosen, which can cause errors in prediction to occur The number of factors used will generally be determined by using the minimum Predicted Residual Sum of Squares ( PRESS ) value associated with factors in order to increase accuracy Scores are used to check the sample ho mogeneity and possible clusters; while loadings are used to interpret how the variables are wei ghted in principal component space ( Hatchell 1999) Crowe and McNichol (1985) consider the critical step of any calibration the identification of a suitable set of four or five wavelengths where the combined changes in optical density correlate highly wit h the constituent of interest. The successful prediction of defects or disease using VNIS is largely based on identification of significant wavelengths that will act as markers This is one of the main issues of spectral based sensing technology, because t he success of a calibration model heavily depends on the selected wavelengths. Min et al (2004) described correlation coefficient spectrum as the simplest method to determine important wavelengths between absorbanc e and the desired measurements. The corre lation coefficient (r) 1, with a value of 1 indicating that the two variables X and Y are directly related with one increasing with the other for a linear equation data points lie on a line for which the variables Y and X are in versely related. No linear correlation exists between variables for data points which have a

PAGE 52

52 value of 0 The variables X and Y are used to represent the wavelengths and damage for this project. The initial introduction of d erivatizing spectral data as a c oncept was introduced in the 1950s when its many advantages f or qualitative and quantification analysis were presented (Owen, 1995) A fir st order derivative is the rate of change of absorbance with respect to wavelength, as shown by equation (1 6): (1 6) where a spectrum is expressed as The 1 st derivativ e is a plot of equation 1 6 using the absorption envelope versus wavelength. Derivative spectra l data can be produced by proc essing the spectrophotometer output. Using the derivative spectra would be beneficial because it can increase the detection sensitivity of minor spectra l features and reduce the error caused by the overlap of the analytic spectral band by interfering bands of other species in the sample. Partial least squares (PLS) regression in its simplest form is a l inear model that specifies the relationship between a depend ent variable Y (response) and a set of predictor variables X PLS was initially introduced by Wo ld ( 1975) and has been heavily promoted in chemometrics literature and more recently in other fields as well Lingaerde and Christophersen (2000) referred to the PLS method as a linear model that estimates parameters when predictor variables are close to being collinear PLS can be characterized in terms of the scaling (shrinkage or expansion) that occurs with each eigenvector of the predictor correlation matrix. Helland (2001) considered PLS to be a two step method where the first step reduces the matrix dimensions (m

PAGE 53

53 and n) which represent the values of the row and column; while the second step identifies latent structure models in the data matrix. PLS procedures work by extracting successive linear combinations of the predictors underlying the Y and X va riables until ideally explaining response variation and predictor variation with minimized response prediction error ; though the number of prediction functions can never exceed the minimum of the number of Y variables and X varia bles. In contrast to principal components regression ( PCR ) which extracts factors t o explain as much predictor variables variation as possible without considering the response variables, the PLS method explains both the response and predictor variations b y balancing the two ob jectives and seeking the factors that explain the variation (SAS, 20 09 ). According to Hatchell (1999), validity must be tested. This can be done by separating a data set into two groups ; one group is used for calibration and the second group is used for validation. C ross validation can be performed i f there are not enough samples by leaving one sample out at a time and using the rest of the samples to build a calibration model and predict ing the sample that was excluded Cros s validation has the advantage of excluding the prediction sample from the calibration model, unlike calibration with a full data set In order to complete a PLS model the number of factors must be chosen using the PRESS statistic in PLS The factors are c hosen using cross validation, in which the data sets are divided into two or more groups. T he model is fit to all groups except one ; the capability of the model is then checked to predict responses for the group omitted. T his is repeated for each group in order to measure the overall capability of a given form of the model. The PRESS statistic is based on the residuals generated by this process. The optimal number of factors for PRESS is generally ob tained when factors are minimized with a smaller PRESS val ue indicates a better model prediction ( Sundberg,

PAGE 54

54 199 9) S electing the number of factors for which the absolute minimum PRESS exists is not always the best choice since a lower number of factors may have a significantly close PRESS value T he CVTEST (Cros s Validation) option in SAS PLS, which is a statistical comparison suggested by van der Voet (1994) can be used in order to perform a test on the significance of the PRESS v alue s for each number of factors. When interpreting PLS outputs Esbensen (2002) sta tes that; X loadings are used to signify common variations in the spectral data, while X weights signify the fluctuations in the spectra that resemble the regression constituents. Thus, h igh X l oadings and X weights are used in order to isolate important w avelengths However, the main purpose of PLS regression is to build a linear model with the B coefficient taken from the traditional regression equation ( 1 7 ) : (1 7 ) with Y being n cases by m variables response matrix, X n cases by p variables predictor matrix, and B a p by m regression coefficient matrix The B coefficient provides a complete picture of the most significant wavelengths. Wav elengths with high er B value s are able to provide better results for a calibration model, and are considered to be more significant wavelengths. The B coefficient is calculated using PLS loadings and weights as shown in equation ( 1 8 ): (1 8 ) with w being the X weight, p the X loading, and q the Y weight. The reliability of calibration models can be assessed by using the coefficient of determination (R 2 ) found comparing predicted damage concentration and true concentration,

PAGE 55

55 standard er ror of calibration (SEC), standard error of prediction (SEP) which signifies the accuracy of the results root mean square difference (RMSD) and ratio of prediction to deviation (RPD) as shown in ASTM ( 1997). The SEC, SEP, RMSD and RPD were of determined b y the equations ( 1 9 through 1 1 2 ) : (1 9 ) (1 10) (1 1 1 ) (1 12) with n being the number of samples p the number of independent variables in the calibration model e i the difference between actual N concentration and predicted N concentration in the i th sample and the mean value of e i Stepwise mu ltiple linear regression (SMLR) is a form of forward regression that aims to obtain a robust model using a minimum number of characteriz ing variables obtained from a large set of potentially useful variables that permits re examination at every step of the variables incorporated into the model in previous steps (Yee 1999) Each forward can be followed by one or more backward of 0.1. The stepwise selection process terminates if no further variable can be added to the model or if the variable just entered into the model is the only variable removed in the subsequent backward elimination (Min 2005). SMLR starts with

PAGE 56

56 no variables in the model and the basic method of obtaining optimal predictors starts with each x variable in the data set; then determines the regression model relating to the y variable and calculates the corre sponding regression coefficient which is the residua l sum of squares ( RSS ) in order to assess the correlation of the predictor. The most highly corre lated variable is then selected as the first variable in the regression model; denoted as x s 1 Then r egression models with both the pre selected variable, x s1 and each of the remaining x variables are determined using MLR and the corresponding RSS value c alculated for each model. The x variable that produces the highest RSS when used in combination with x s 1 then becomes the second selected variable and is deno ted, x s 2 The process is then repeated until the stepwise selection reaches the desired number of variables selected by the stepwise regression routine equal to the desired number of variabl es as specified by the operator (Yee 1999) In SMLR, over fitting could be a problem because too many wavelengths might be selected by the stepwise procedure (Min 2005). 1.7.4.3 Assessment o bjectives The main objective of the vine kill wait period assessments were to determine the effects of time elapsed between vine kill and harvest on resistance to mechanical damage incurred during the harvesting, grading and packing of Florida new potatoes This was accomplished by evaluating the damage caused by controlled impacts on tubers dropped from several heights that commonly occur during the harvesting and processing of potatoes. In addition to the main o bjective several sub objectives were: Determine the physical properties of potato tubers at various time s after vine killing had been performed on the plants;

PAGE 57

57 Determine the change in skinning re sistance that occurs as potato tubers were allowed to underg o skin set in the soil; Determine the effect of short term storage on the damage resistance of potato tubers at various times after vine kill. The two main objective s of the spectral measurement assessments were to d etermine the important wavelengths in th e electromagnetic spectrum to assess if physical damage had occurred internally in the two varie etermine the important wavelengths in the electromagnetic spectrum that can be used to detect if internal bruising, gree ning, scar ring or other possible external defects were present The sub objectives were to: content change to assist in the detection of damage. Develop a calibration model for damage prediction.

PAGE 58

58 Table 1 1 Mechanical Damage of Potatoes Damage Type Description Causes Skinning (Feathering) The skin layer (periderm) of the potato is partially or completely removed, exposing the underlying tuber flesh. Immature tubers, high nitrogen availability at harvest, and excessively wet soils. Shatter b ruise A visible crack forms in the tuber surface which can extend deep into the tuber flesh and causes discoloration around the crack. Ruptures caused by an impact that result s in tissue separation and drying out. Cool harvesting temperatures and exposure to low humidity favor shatter. Blackspot b ruise Caused by significant impact to the tuber which results in the cell membrane of tissue within a tuber rupturing and causes t he formation of a black pigment. Impacts caused by large tuber size or drops, poor tuber hydration, potassium deficiency, significant curvature and old or very mature tubers. Pressure b ruise Flattened areas or indentations on the tuber surface. Intern ally a gray to black discoloration in the flesh occurs. The result of tissue damage due to a continuous weight being exerted onto the surface. Favored by dehydration of tubers, storage humidity below 90% and excessive pile height. Cut The separation o f a portion of the potato tuber flesh due to an external force. The result of tubers coming in contact with the edges of mechanical harvesting devices such as digging blades and the sides of conveyors or sorting tables. Compiled from Compendium of Pota to Diseases 2001. The American Phytopathogical Society. St. Paul, Minnesota. Table 1 2 C lassification for degrees of skinning by lot Skin r ating Classification Practically no skinning Not more than 5 percent of the potatoes in the lot have more t han one Slightly skinned Not more than 10 percent of the potatoes in the lot have more than one Moderately skinned Not more than 10 percent of the potatoes in the lo t have more than one Badly skinned More than 10 percent of the potatoes in the lot have more than one Compiled from USDA Standards for Grades of Potatoes. 2011. USDA AMS Washington, D.C.

PAGE 59

59 Table 1 3 Physiological defects of p otatoes Disorder Description Causes Greening Light enhanced chlorophyll production that results in a green skin. This greening is harmless, but compounds called glycoalkaloids which are harmful to humans also increase in tissue. Sunlight, indirect daylight or artificial lights enhances chlorophyll production. Tubers developing near the soil surface may be directly exposed to sunlight or through cracks in the soil surface caused by tuber growth. Sunscald Tuber surface takes on a blistered, metallic appearance. Underlying tissues may become watery and turn brown. Making tubers predisposed to rot in storage. Caused by frequent or prolonged exposure of tubers to intense sunlight and high temper ature. Threshold tuber flesh temperature for sunscald is 43C (109F) and can occur in tubers as deep at 2.5 cm. Growth Cracking Shallow to moderately deep fissures in the surface tissue of tubers, usually following the long axis of the tuber. Caused by internal pressure that exceeds the tensile strength of surface tissues during tuber enlargement. Often caused by an uneven availability of soil moisture and rapid water intake. Early Blight Lesions are dark, circular to irregular in shape and sunken in surrounded by a raised purplish gray border with dry, leathery flesh that is usually brown. Caused by spores of the fungus Alternaria solani but does infect tubers before harvest. Infection occurs when contaminated soil comes into contact with wounds made during harvest. Late Blight Infection is initially somewhat superficial, but lesions can eventually extend several centimeters into the tuber. Lesions are reddish brown, dry and granular. Caused by the pathogen Phytophthora infestans Tuber infe ction occurs whenever the potato comes in contact with the pathogen in the soil, often being transported by rain or irrigation. Dry Rot Shallow lesions are visible as small brown areas after around 1 month of storage with infection slowly enlarging in a ll directions. Periderm over the lesion sinks and wrinkles. Tuber i nfection occurs at wound sites by the pathogens of Fusarium spp. present in infested soil and contaminated equipment. Brown r ot Vascular tissue of infected tubers has a distinct grayis h brown and the discoloration may extend into the pith or cortex. Sticky exudate may form at eyes or stolon Caused by the pathogen Ralstonia solanacearum infecting the potato plant and spreading into the tubers through the stolon. Insect d amage Insec ts bore into tubers superficially, produce tunnels or scratch the skin facilitating penetration of pathogens in the soil. Caused by insects in the soil eating portions of the tuber Compiled from Compendium of Potato Diseases 2001. The American Phytopa thogical Society. St. Paul, Minnesota.

PAGE 60

60 Figure 1 1 Potato route on Hastings P ackaging L ine

PAGE 61

61 CHAPTER 2 DESTRUCTIVE ASSESSMENT 2 .1 Materials and Methods A total of four tests were performed in order to evaluate the two potato varieties chosen. All sample sets for the two varieties were collected at the same time and the following four tests were performed ; d rop impact tests skinning tests, moisture content tests and compression rupture force tests. Two of the tests were performed in order to determine r esistance of the potato periderm to mechanical harvest These two tests, s kinning and drop impact, were simulated in order to estimate the resistance to skinning and bruise damage change s that occur over time after vine kill for tubers Moisture content an d compression rupture force tests were performed in order to help with the evaluation of changes between harvests. T ubers used for skinning tests were evaluated for the degree of mechanical damage the day damage testing was applied. Impacted tubers were e valuated for mechanical damage after six days of cold room storage at 20 C and 80% RH in order to allow damage to develop One skinning test was performed for each variety on the day of harvest, while two drop impact test s were carried out for each variety during every harvest period. The first test was done one day after harvest to determine damage thresholds and the second test was done 7 days after harvest to determine the effect of short term storage on impact thresholds Two moisture content tests and rupture force tests were performed on each variety corresponding with the drop impact testing times. T he properties to be determined over the course of this study we re changes in moisture content, skin toughness and resistance to the forms of mechanical da mage that often occur during and after harvest ; skinning, internal shatter, external shatter and blackspot bruising.

PAGE 62

62 2.1.1 Potato Plant Materials Sample tubers of the two new potato Fabula and Red La Soda 1 ) were hand harvested out of the fields of the U niversity of F lorida Research and Demonstration Site in Hastings, Florida and were transported to the Postharvest Horticulture Laboratory of the University of Florida in field lugs. The variety h a s a very high yield and v ery large size tuber with pale yellow flesh and oval barrel shape It also has an extremely smooth, light yellow colored skin with shallow eyes. It was described to have good internal bruising resistance, as well as fairly good to very good resistance to many viruses that effect tubers Samples were determined to have a mass between 90 and 470 grams during the study The other variety ha s a white to cream colored tuber flesh a round to oblong shape and smooth deep red color ed skin; eyes are of medium to deep depth and well distributed. It was described to have good skinning resistance, good yield potential and a relatively low specific gravity compared to other red skinned varieties and a general disease resistance requiring standard disease control programs be followed. Samples were determined to have a mass between 80 and 325 grams during the study. Initial f ield samples for 7 days before vine kill were collected May, 18 2010. Samples for the day before vine kill we re collected on May, 24 20 10. The 7 14 and 21 days after vine kill samples were collected June, 1 st 8 th and 15 th of 2010 respectively All samples were collected during late morning to noon hours. Normally, potatoes would be harvested mechanically and transported to a packingho use. But in order to reduce the number of immeasurable impacts and scuffing that can take place during this process, it was decided that physically harvesting tuber samples by hand and s orting for defects in the field would be used for this research (Figur e 2 2 )

PAGE 63

63 2.1.2 Potato Skinning Tests Samples for different harvesting periods around vine kill were tested f or mechanical skinning damage using the cleaning section of the packing line present at the UF Hasting s farm ( Figure 2 3 ) this portion of the pack aging line included a rubber conveyor to lift potato tubers up from the receiving hopper onto the wash area, which had rotating rollers to move tubers along under the drenching tubes and abrasively clean the tubers A sample group of (n=20) potato tubers f rom each variety were run along the cleaning section in order to determine day of harvest skinning that would occur for each harvest period. The s amples designated for the packing line skinning test were dumped onto the packing line conveyor, where they we re elevated and dropped onto the washer/brush bed. Brush rolls in the wash area conveyed the tubers toward the conveyor w hich led to the sorting table. The samples were collected at thi s point for each run to maintain consistency. A n initial assessment of s kinning mechanical damage was done using a nondestructive method immediately after returning from the fields in Hastings, Florida. Due to the difficulty of accurately measuring percent skinning of a tuber, a s ub jective scale rating was used to assess the skinning damage called Jenkins Wehner method ( Table 2 1 ) The p ercent of tubers damaged was compared between harvest period s in order to compare the changes in mechanical damage done as tubers were allowed to mature. 2.1.3 Impact Dr op Tests A test was performed in order to simulate the mechanical damage caused by all the tumbling and dropping during harvesting and handling operations. Individual t ubers were held in a sling in order to control the point of impact and prevent rotation ( Figure 2 4 ) the bud end was

PAGE 64

64 chosen due to it s smaller curved area of impact Tubers of each variety (n=10 for each drop height) were dropped from various heights (30 cm, 60 cm, and 90 cm and 120 cm) onto a hard surface in order to perform comparable dama ge to tubers from each harvest period and single drop tests were performed in order to evaluate the damage caused by one time impacts, while double impact tests were performed in order to determine if extra damage would be accumulated from repeated impact s on the same location. White chalk dust was spread over the impact area in order to accurately indicate the point of impact for later damage assessment. The procedure was repeated for each variety using various drop heights and tubers of different mass. D rop tests were performed on separate sample groups the day after harvest and 7 days after harvest. The drop heights performed were: 30 cm, 30 cm double drop, 60 cm, 60 cm double drop, 90 cm and 120 cm for the last two harvest period s The internal and ext ernal mechanical damage assessment was done after tubers were stored in a 20C (68F) and 80% RH cold room for six days in order to allow mechanical damage to develop. On the seventh day, the potato tubers were assessed for external and internal mechanical damages in order to identify changing trends caused by vine kill harvesting times. External mechanical damage assessment was done visually inspecting for skinning and external shatter around the point of impact and measuring the size of the skinned area o r length of shatter if present. Internal mechanical damage assessment was done using a destructive method. Potato tubers were sliced along the bud end at the point of impact (Figure 2 5). Once sliced, tubers were visually inspected for development of black spot bruising and internal shatter (which would lead to external shatter if permitted to continue developing). Internal shatter was assessed by measuring the depth and length of the damage and these diameters were used to calculate the cross sectional area of the internal shatter. Blackspot bruising was assessed by measuring the

PAGE 65

65 length, width and height of damage, as well as, how deep below the surface damage developed. This assessment generally required additional cutting of tuber to allow for more accurat e measurement The se three dimensions were used to calculate the volume of blackspot damage to the tuber. The presence of d amage, as well as, the volume or cross sectional area of this damage was plotted using bar graphs in order to compare the severity of d amage at each point of harvest. 2.1.4 Moisture Content Tests During each harvest samples were collected in order to perform moisture content tests on the tuber skin and flesh. This was done in order to compare results between harvests to identify if mo isture content level differences and similarities influenced the results of mechanical damage testing between harvests The moisture content levels of potato tubers were shown in many studies to affect the occurrence of shatter and blackspot bruising. Alum inum sample trays were weighed before t he skin of severa l tubers was peeled off to accumulate a significant weigh t in each sample tray for dehydrating Next chunks of tuber flesh were sliced up and weighed and recorded in individual sample trays a lso Once the skin and flesh samples for both varieties were collected they were placed in a hot air drying oven kept at 60C for 2 weeks to allow for complete removal of moisture. Percent m oisture content was calculated using wet basis by dividing the total mass o f water found from oven drying by the total initial wet mass of material and multiplying the result by 100 to get percent Samples were collected to coincide with each drop test performed, so moisture content tests were performed the day after harvest and 7 days after harvest as well.

PAGE 66

66 2.1.5 Compression Rupture Force Tests The compression strengt h of the two tuber varieties was tested at each point of harvest except for the 7 days before vine kill group Cylindrical core s were taken from potato tubers usi ng a metal punch to create 2.22 cm ( 0.875 in .) diameter, and 2.54 cm ( 1 in .) tall samples. Using an Instron Universal Testing Instrument (model 4411 Instron, Norwood, MA ) a constantly increasing force was applied to samples up to 500 N until samples crit ically failed under the load ruptured diagonally (F igure 2 6 ) Compression strength was calculated by dividing the load at critical failure by the cross sectional area resisting the load and reported in kilop ascal ( k Pa). Just like the previous two tests sa mples were run 1 day after harvest and 7 days after. 2.2 Results and Discussion 2.2.1 Potato Skinning Tests Incidences: vine kill, a significant extent of skinning showed up on tu ber samples used ( Figure 2 7 ) However, after vine kill was performed there was a sharp decline in the number of tuber samples which displayed over 10% skinning or feathering area from 60% to 0% with just 7 days of skin set ( Figure 2 8 ) As the weeks after vine kill increased there was an insignificant change in the percent of potato tubers which displayed skinning damage between weeks which indicates that 7 days of waiting allowed skins to set enough for harvesting operations. In contrast to the results f tubers before vine kill resulted in many tuber samples with little to no skinning, which was to be expected from the skinning Figure 2 9 ) After vine kill was

PAGE 67

67 perfo rmed there was little to no change in the number of tuber samples which displayed significant skinning or feathering for any period of wait for skin set ( Figure 2 9 ) As the days after vine kill increased there was an insignificant change in skinning incid ents on tubers which suggested that 7 days of waiting allowed skins to set enough for harvesting operations. Severity: significant change between pre vine kill and post vine kill tub ers ( Figure 2 8 ) harvested after vine kill showed little or no skinning damage, while the majority of those harvested before vine kill showed skinning. The USDA standards dictate that because of the quantity of tubers skinned and the sever ity of skinning for 7 days before vine kill the harvest kill faired a little better due t o less sever e resulting in more than 10 percent of the tubers being skinned more than 25% of their total kill ended up with the best results due to most skinned tubers only suffering 3 6% skinning which allowed them to pass with the periods, 14 and 21 days after vine kill resulted in somewhat significant skinning results on a few insignifica nt changes between pre vine kill and post vine kill tubers ( Figure 2 10 ) The skinning damage that occurred on the two harvests before vine kill were insignificant by the USDA initiated on t he potato tubers. The potato tubers harvested 7 and 14 days after vine kill ended up

PAGE 68

68 ays after vine kill, resulted in such insignificant skinning results on a few tubers which qualified them for the ing during skinning simulation. 2.2.2 Impact Drop Tests The following type s of mechanical damage were observed in the impact drop tests for of impact testing for both varieties displayed effects similar to those reported by Thornton and Bohl (1998) in which greater drop heights resulted in a larger force that caused bruising to occur more often. Incidences: Two types of external damages (skinning and external shatter) were pact damage, while only one type of internal damage (blackspot bruising) was observed. The most common form of mechanical damage 11 and 12), followed by blackspot bruising (Figure 2 13 and 14), and only o ne incident of shatter (Figure 2 15 and 16); similarly, Chiputula et al. (2009) reported the major mechanical damage found during sample collection from would be expe cted with two varieties that have more resistance to bruising than skinning However, the results of impact testing performed by Chiputula et al. (2009) resulted in many incidences of external shatter, internal shatter and black spot bruising, but oddly fo und no cases

PAGE 69

69 of skinning kill appeared to be unaffected by drop height, with most tubers skinning regardless of height. The three harvests after vine kill had a large drop in skinning frequency of 20 40% for all drop heights and often required a significant drop of 90 cm or consecutive drops at 30 cm and 60 cm in order to develop skinning damage, and only in decreasing significance as time after vine kill increased ( Figure 2 11 ) harvest period 7 days after vine kill. Blackspot bruising was not observed day after picking in two harvests, 7 days before vine kill and 14 days after vine kill. Storage of 7 days fo r those harvested 7 days after vine kill allowed tubers to completely resist bruising, ( Figure 2 13 ) Bruising occurrence dropped 20 30% 7 days after vine kill from the day before vine kill and 30 40% for the following last two harvest periods. Incidences of bruising for the periods that displayed bruising tended to decrease over time and ended up only resulting from impacts with more force exerted on the tuber. The only occurrence of external shatter was during the last harvest period from one of the tuber s used in a 60 cm double drop test. These results vary greatly from the results of impact testing done by Chiputula et al. (2009) in which a drop above 90 cm was required for any damage to appear. These results involved a large occurrence of internal brui sing and shatter development, which was likely encouraged by the much higher moisture content levels around 92 collected and tested. Severity: The severity of skinning damage from drop impacts was much more substantial during the two harvests befor e vine kill ( Figure 2 12 ) Skinning damage area dropped

PAGE 70

70 significantly in comparison to the day before vine kill harvest with only 7 days of vine kill skin set with drops of 90 cm being required to skin a remarkable area of the impact zone. Skinning damage for the remaining harvests and extra 7 days storage were even further reduced for the minimal dimensions except for one incident during the 7 days after vine kill harvest when a bruise formed with a volume over 1 cubic cm which was an extreme outlier when compar ed to all over bruising volumes ( Figure 2 14 ) tuber during the 21 days after vine kill harvest was severe enough to cause 2.54 cm of damage to the exterior of the tuber, but was only 1/3 the length of the tuber making it less serious damage according to the USDA. Mass: ne how much of a role tuber mass played in damage development. Skinning occurrence versus tuber mass ( Figure 2 17 ) skinning area versus tuber mass ( Figure 2 18 ) bruising occurrence versus tuber mass ( Figure 2 19 ) bruising v olume versus tuber mass ( Figur e 2 20 ) and shatter occurrence versus tuber mass ( Figure 2 21 ) When comparing the mechanical damage that plays a slightly larger role in damage development than mass. Skinning and bruising incidents tend to occur at a higher rate when the drop height was 60 cm or more regardless of tuber mass as can be seen in ( Figure 2 17 and 19 ) The skinning area damage was also more significantly dependent upon drop height tha n tuber mass ( Figure 2 18 )

PAGE 71

71 Incidences: Two types of external damages (skinning and external shatter) were damage (blackspot bruising and intern al shatter) were observed. The most common form of 22 and 23), followed by blackspot bruising (Figure 2 24 and 25), than external and internal shatter (Figure 2 26 and 27) which dam aged tubers to an extent that would lower the USDA grade the most significantly ; t his was the same result for impact testing by Chiputula et al. (2009) one week after harvest for the var Skinning incidents for the two group s kill appeared to be unaffected by severity of drop height, with large quantities of tubers developing skinning regardless of height; over 60% for each category 7 days before vine kill and over 40% for each ca tegory the day before vine kill developed skinning damage. With just 7 days of storing the tubers harvested the day before vine kill skinning damage decreased 40 50% for lower drop heights while skinning dropped less for higher drop heights (20 40%). The t hree harvests after vine kill showed a significant drop of over 40% less skinning damaged than previous test periods caused from the drop impact force. Storage for these harvests had a less significant influence on skinning occurrence than the day before v ine kill due to already low skinning rates with skinning damage actually increasing slightly for some drop heights ( Figure 2 22 ) harvests, 7 days before vine kill and 14 days after vine kill, as well as after storage of the 14 days after harvest ( Figure 2 24 ) Incidences of bruising for the periods that displayed it tended to decrease over time after vine kill at lower drop heights and ended up only resulting from impacts

PAGE 72

72 with more force exerted on the tuber of double 60 cm and 90 cm in the last testing period (21 days after vine kill) and not at all during the 14 days after vine kill harvest. Shatter tended to and 60 cm double drop and only very sparingly ( Figure 2 26 ) iputula et al. (2009) 93%) in the study. Severity: The severity of skinning damage from drop impacts was much more substantial during the two harvests before vine kill and increased for higher drops ( Figure 2 23 ) 7 days of storage at day before vine kill harvesting caused skinning area damage to decrease significantl y to around half the previous skinning area. Skinning area damage dropped significantly with only 7 days of vine kill skin set with any impact skinning being slight and insignificant. Blackspot nsions at lower impact heights, while higher drop heights resulted in much more significant damage volume up to five times a s large as smaller drop heights (F igure 2 25 ) Time after vine kill caused bruising volume to drop much more significantly 14 days a fter vine kill, with only 7 days slightly dropping average volume an insignificant amount of 10% at the 90 cm drop height. Shatter that occurred on double drop at 60 cm and 90 cm; with often as much as half the length of the sample tubers getting up to 3 cm or more shatter damage ( Figure 2 27 ) Increased harvest time after vine kill was not able to significant ly affect the amount of shatter

PAGE 73

73 Mass: ured in order to determine how much of a role tuber mass played in damage development. Skinning occurrence versus tuber mass ( Figure 2 28 ) skinning area versus tuber mass ( Figure 2 29 ) bruising occurrence versus tuber mass ( Figure 2 30 ) bruising volume versus tuber mass ( Figure 2 31 ) and shatter occurrence versus tuber mass ( Figure 2 32 ) When comparing the mechanical damage that height plays a slightly larger rol Skinning and shatter incidents tended to occur at a higher rate when the drop height was 60 cm or more regardless of tuber mass ( Figure 2 28 and 32 ) while bruising results appeared to be random wi th more bruising occurring with tubers of smaller mass at lower drop heights ( Figure 2 30 ) The skinning area and bruising volume damage were also more dependent upon drop height than tuber mass ( Figure 2 29 and 31 ) 2.2.3 Moisture Content Tests The moist ure content levels for both varieties were determined at each point of harvest ( Figure 2 33 ) ( Figure 2 34 ) the skin varied between 77% and 84% moisture content while the skin varied between 82% and 90%. iputula et al. (2009) the two tuber varieties used were harvested from the same soil conditions, which would attribute the differences in moisture content to their different physical properties and cause the potato tubers to resist damage differently. The moisture content of tuber skins tended to be slightly higher at

PAGE 74

74 down for both varieties. More incidences of skinning occurred during testing when the skin moistu re content of have a slight effect on the occurrence of skinning damage which occurred during skinning testing, but skinning damage caused by drop impacts was still influenced by the difference in moisture content between skin and flesh. r flesh was above 79% at higher drop heights. On the other hand, incidences of blackspot bruising in tubers appeared to be unaffected by moisture content levels of tuber flesh, which could mean ing regardless of moisture content. tuber flesh was near its highest. On the other end of the moisture scale incidences of blackspot rs occurred at higher rates during periods in which tuber flesh was below 85% moisture content. These results appear to follow the guidelines for mechanical damage compiled by Stevenson et al. (2001), which states that poor tuber hydration encourages bruis ing, while better hydration encourages shatter as a result of tubers having a more solid structure. 2.2.4 Compression Rupture Force Tests ( Figure 2 35 ) while the results for ( Figure 2 3 6 ) The results of rupture force testing for both varieties oddly follow a

PAGE 75

75 similar trend as moisture content skin results over the weeks of harvest with higher skin moisture content meaning higher required rupture force, except for the very last sample set The standard deviation for both varieties showed that failure loadings were close enough to have no significant difference, but could possibly show how much load the potato varieties could hold from the bottom of a potato pile. 2.3 Summary The potato var the results of following types of mechanical damage to some degree; skinning, blackspot bruising, and external shatter. The results of testing agree with the conclusions of Blahovec (2005) and Chiputula et al. (2009) that mechanical damage outcomes only slightly depended on cultivation regime and a potato varieties damage resistance could impact mechanical damage more than conditions during the growing season. skinning, blackspot bruising and external shatter. The main grade decreasing mechanical dama ge absence of skinning from impact simulation could be due to samples being collected after 3 4 weeks of skin set have been allowed. Four types of mechanical dama ge were observed during

PAGE 76

76 main grade decreasing mechanical damage that occurred in the impact drop testing for the nal sh atter; which can be attributed to high moisture content levels in the flesh of tuber samples. Incidences and severity of mechanical damage generally increased with drop height for both varieties, with blackspot bruising occurrence being replaced by e xternal shatter at higher moisture content flesh resulted in higher inc the guidelines of Stevenson et al. (2001). The variation of damage between the two varieties could also be associated with the damage resistance and threshold differences referenced by Bajema and Hyde ( 1998) which had an effect on mechanical damage results during testing. It was observed that impact forces experienced by potato tubers greatly depended on drop height and mass of tubers ( Figure 2 37 ) skinning simulation on the cleaning portion of a packaging line, while drop impact testing the skinned area in a ffected samples. This skinning associated with rubbing against surfaces, but is susceptible to skinning which can occur when a small amount of surface area is exposed to large external forces from impact which breaks the bond b etween the skin and flesh. The results of skinning and impact testing before and after vine stated that tubers harvested from living vines were more likely to be severely skinn ed and/or

PAGE 77

77 this could be attributed to the varieties resistance to skinning. Both varieties experienced minor to extreme drops in mechanical damage occurrence wit h just 7 days of skin set after vine kill, depending on how much mechanical damage physically occurred before vine kill. These results coincide with those found by Hutchinson and Stall (2007), in which proper tuber maturity at harvest improves skin set and bruising resistance in order to produce high quality fresh market potatoes. The significant drop in mechanical damage occurrence with just 7 days of skin set from vine killing should allow farmers to harvest earlier than the standard 2 3 weeks if weather conditions could cause crop left in the soil longer to be exposed to circumstances that would threaten tuber health.

PAGE 78

78 Table 2 1 Jenkins w ehner d amage r ating s cale Skinning s cale Min imum s kinning Maximum s kinning Equivalent USDA r ating 0 0% 0% Practical ly no skinning 1 0.1% 3% Practically no skinning 2 3.1% 6% Practically no skinning 3 6.1% 12% Slightly skinned 4 12.1% 25% Slightly skinned 5 25.1% 50% Moderately skinned 6 50.1% 75% Badly skinned 7 75.1% 87% Badly skinned 8 87.1% 99.9% Badly skinn ed 9 100% 100% Badly skinned Figure 2 1 Fabula (left) and Red La Soda (right)

PAGE 79

79 Figure 2 2 Hand harvesting potato tubers prior to vine kill Red La Soda on left, Fabula on right. Figure 2 3 Packing line washer section used for at harvest skinning test

PAGE 80

80 Figure 2 4 Drop test stand using sling. Figure 2 5 A sliced Red La Soda with damage (left), and sliced Fabula with damage (right).

PAGE 81

81 Figure 2 6 Response of tissue to c ompression f orce. Compressing and critically fail ure tuber tissue (left). Photo of c ritica l l y failed tuber tissue (right). Figure 2 7 Fabula Average s kinning d amage resulting from packing line handling as affect ed by harvest time (n=20). 0 5 10 15 20 25 7 Days Before Day Before 7 Days After 14 Days After 21 Days After % Percent Area with Skinning Damage Harvest Period Average

PAGE 82

82 Figure 2 8 Fabula Percent s kinning according to USDA grade standards for potato as affect ed by skinning (n=20). Figure 2 9 Average skinning damage resulting from packing line handling as affect ed by harvest time (n=20). 0 10 20 30 40 50 60 70 80 90 7 Days Before Day Before 7 Days After 14 Days After 21 Days After % Percent of Potatoes with Damage Harvest Period 10 % Skinning 25% Skinning 50% Skinning 0 5 10 15 20 25 7 Days Defore Day Before 7 Days After 14 Days After 21 Days After % Percent Area with Skinning Damage Harvest Period Average

PAGE 83

83 Figure 2 10 Percent skinning according to USDA grade standards for potat o as affect ed by skinning (n=20). Figure 2 11 Fabula Percent of samples with skinning from impact testing (n=10 per category ). 0 10 20 30 40 50 60 70 80 90 7 Days Defore Day Before 7 Days After 14 Days After 21 Days After % Percent of Potatoes with Damage Harvest Period 10 % Skinning 25% Skinning 50% Skinning 0 10 20 30 40 50 60 70 80 90 100 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage %Percent of Samples with Skinning Harvest Period and Drop Height 30 cm 30 cm Double 60 cm 60 cm Double 90 cm

PAGE 84

84 Figure 2 1 2 Fabula Average skinning area of tubers from impact testing (n=10 per category). Figure 2 1 3 Fabul a Percent of samples with bruising from impact testing (n=10 per category). 0 0.5 1 1.5 2 2.5 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage Average Skinned Area (cm 2 ) Harvest Period and Drop Height 30 cm 30 cm double 60 cm 60 cm double 90 cm 0 10 20 30 40 50 60 70 80 90 100 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage % Percent of Samples with Bruising Harvest Period and Drop Height 30 cm 30 cm Double 60 cm 60 cm Double 90 cm

PAGE 85

8 5 Figure 2 1 4 Fabula Average bruising volume of tubers from impact testing (n=10 per category ). Figure 2 1 5 Fabula Percent of samples with shatter from impact testing (n=10 per category). 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage Average Bruising Volume (cm 3 ) Harvest Period and Drop Height 30 cm 30 cm double 60 cm 60 cm double 90 cm 0 10 20 30 40 50 60 70 80 90 100 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage % Percent of Samples with Shatter Harvest Period and Drop Height 30 cm 30 cm Double 60 cm 60 cm Double 90 cm

PAGE 86

86 Figure 2 1 6 Fabula Average shatter length of tubers from impact testing (n=10 per category). Figure 2 17 Fabula Mass of tubers versus percent tubers with skinning damage from impact testing (n=70 per category). 0 0.5 1 1.5 2 2.5 3 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage Average Shatter Length (cm) Harvest Period and Drop Height 30 cm 30 cm double 60 cm 60 cm double 90 cm 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 100 150 200 250 300 350 400 % Percent Potatoes with Skinning Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm

PAGE 87

87 Figure 2 18 Fabul a Mass of tubers versus average skinning damage area from impact testing (n=70 per category). Figure 2 19 Fabula Mass of tubers versus percent tubers with bruising damage from impact testing (n=70 per category). 0 0.5 1 1.5 2 2.5 3 3.5 100 150 200 250 300 350 400 Average Skinning Area (cm2) Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 100 150 200 250 300 350 400 % Percent Potatoes with Bruising Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm

PAGE 88

88 Figure 2 20 Fabula Mass of tubers versus average bruising damage volume from impact testing (n=70 per category). Figure 2 21 Fabula Mass of tubers versus percent tubers with shatter damage from impact testing (n=70 per category). 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 100 150 200 250 300 350 400 Average Bruising Volume (cm3) Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 100 150 200 250 300 350 400 % Percent Potatoes with Shattering Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm

PAGE 89

89 Figure 2 22 Percent of samples with sk inning from impact testing (n=10 per category). Figure 2 23 category). 0 10 20 30 40 50 60 70 80 90 100 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage % Percent of Samples with Skinning Harvest Period and Drop Height 30 cm 30 cm Double 60 cm 60 cm Double 90 cm 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage Average Skinning Area (cm 2 ) Harvest Period and Drop Height 30 cm 30 cm double 60 cm 60 cm double 90 cm

PAGE 90

90 Figure 2 24 Percent of samples with bruising from impact testing (n=10 per category). Figu re 2 25 category). 0 10 20 30 40 50 60 70 80 90 100 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage % Percent of Samples with Bruising Harvest Period and Drop Height 30 cm 30 cm Double 60 cm 60 cm Double 90 cm 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage Average Bruising Volume (cm 3 ) Harvest Period and Drop Height 30 cm 30 cm double 60 cm 60 cm double 90 cm

PAGE 91

91 Figure 2 2 6 Percent of samples with shatter from impact testing (n=10 per category). Figure 2 2 7 ers from impact testing (n=10 per category). 0 10 20 30 40 50 60 70 80 90 100 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage % Percent of Samples wuth Shatterring Harvest Period and Drop Height 30 cm 30 cm Double 60 cm 60 cm Double 90 cm 0 1 2 3 4 5 6 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage Average Shatter Length (cm) Harvest Period and Drop Height 30 cm 30 cm double 60 cm 60 cm double 90 cm

PAGE 92

92 Figure 2 28 Mass of tubers versus percent skinning damage from impact testing (n=70 per category). Figure 2 29 Mass of tubers versus average skinning damage area from impact testi ng (n=70 per category). 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 90 140 190 240 290 % Percent Potatoes with Skinning Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm 0 0.5 1 1.5 2 2.5 3 3.5 90 140 190 240 290 Average Skinning Area (cm2) Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm

PAGE 93

93 Figure 2 30 Mass of tubers versus percent tubers with bruising damage from impact testing (n=70 per category). Figure 2 31 Mass of tubers versus average bru ising damage volume from impact testing (n=7 0 per category). 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 90 140 190 240 290 % Percent Potatoes with Bruising Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm 0 0.5 1 1.5 2 2.5 3 3.5 90 140 190 240 290 Average Bruising Volume (cm3) Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm

PAGE 94

94 Figure 2 32 Mass of tubers versus percent tubers with shatter damage from i mpact testing (n=70 per category). Figure 2 3 3 Fabula Moisture c ontent levels of skin and flesh at time of impact testing (n= 5 ). 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 90 140 190 240 290 % Percent Potatoes with Shattering Mass (g) 30 cm 30 cm Double 60 cm 60 cm Double 90 cm 50 55 60 65 70 75 80 85 90 95 100 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage % Percent Moisture Content Harvest Period Skin Flesh

PAGE 95

95 Figure 2 3 4. (n=5). Figure 2 3 5 Fabula Average compression rupture force failure loading (n=10). 50 55 60 65 70 75 80 85 90 95 100 7 Days Before Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage % Percent Moisture Content Harvest Period Skin Flesh 0 50 100 150 200 250 300 350 400 450 Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage Failure Load (N) Harvest Period Average

PAGE 96

96 Figure 2 3 6 Red La Soda Average compression rupture force failure loading (n=10). Figure 2 3 7 Calculated i mpact energy versus potato mass for drop heights used. 0 50 100 150 200 250 300 350 400 450 Day Before Day Before + 7 Days Storage 7 Days After 7 Days After + 7 Days Storage 14 Days After + 7 Days Storage 14 Days After + 7 Days Storage 21 Days After 21 Days After + 7 Days Storage Failure Load (N) Harvest Period Average 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 75 125 175 225 275 325 375 425 475 Impact Energy (J) Mass (g) 30 cm 60 cm 90 cm 120 cm

PAGE 97

97 CHAPTER 3 NON DESTRUCTIVE SPECTRAL MEASUREMENT ASSESSMENT 3.1 Materials and Methods Visible and n ear infrared reflectance technology was utilized in order to collect spectr al measurements of damaged, diseased and undamaged potatoes for comparison between the range of 360 and 800 nm These data were compared in order to locate the differences in spectral measurements between damaged and undamaged potatoes and create calibrati on models for detecting damage in Fabula and ed La Soda cultivars. 3.1.1 Potato Plant Materials Sample tubers of two new potato Fabula and Red La Soda were hand harvested on May 13 th and 31 st June 7 th and 14 th 2011 and July 1 st 2012 from the fields of the U niversity of F lorida Research and Demonstration Site in Hastings, Florida and were transported to a laboratory at UF in field lugs for data collection (Table 3 1) The variety h ad a very high yield and very large size tuber with pale yellow flesh, and oval barrel shape It also had an extremely smooth, light yellow colored skin with shallow eyes. It was described to have good internal bruising resistance, as well as fairly good to very good r esistance to many viruses that a f fect tubers. The other variety ha d a white to cream colored tuber flesh a round to oblong shape and smooth deep red colored skin; eyes are of medium to deep depth and well distributed. It was described to have good skinning resistance, good yield potential and a relatively low specific gravity compared to other red skinned varieties and a general disease resistance requiring standard disease control programs be followed.

PAGE 98

98 Initial sample set s of 40 tubers for each variety were collected for eva luation on May 13 th 2011 in order to determine the effects of moisture content levels on spectral reflectance before testing began on reflectance measurement of damaged samples Samples for reflectance measurement were collected on May 31 st June 7 th and June 14 th 2011 in order to allow potato tubers to naturally change condition after vine kill for a range of values to be expected at harvests and allow diseases to develop on infected tubers before harvesting and again on July 1 st 2012 in order to increa se sample variation for evaluation The disorders and disease d samples collected included; internal blackspot bruising, shatter, brown rot, greening, sunscald, growth cracking, and insect damage. All samples were collected du ring late morning to noon hours Potato tubers were harvested by hand in order to prevent mechanical damage being induced before initial spectral reflectance could be measured in a lab oratory 3.1.2 Mechanical Damage Stimulation and Disorders In order to provide undamaged potato tuber samples for comparison to mechanical damaged tuber, reflectance measurements were taken before damage was induced in order to provide sample data with 0% damage; mech anical damage was the n induced in the group of samples before they were used for measureme nt and evaluation. Samples of tubers that had disease or disorders were collected in the fields when they were found and sorted into their respective sample groups for each tuber variety 3.1.2.1 Mechanical damage simulation Undamaged p otato tuber samples for each variety were selected from those harvested in Hastings Florida Freshly harvested tubers were mechanically damaged by exposing them to an impact force by means of dropping them from a height of 120 cm which previous research

PAGE 99

99 showed would cause blackspot bruising (Fig. 3 1 and 2) and/or external shatter (Fig. 3 3 and 4) Once dropped, sample tubers were stored in a 20C (68F) and 80% RH cold room for six days in order to allow mechanical damage to develop. After the sixth day, the potato tubers were assessed for external mechanical damages, in order to identify samples with shatter to exclude from internal damage detection measurements and use for external mech anical damage detection with a spectrophotometer ( Cary 500, Varian Inc., Palo Alto, CA ) External mechanical damage assessment was done visually by looking for skinning and external shatter around the point of impact Samples with no external evidence of mechanical damage were used for internal mechanical damage detection with the spectropho tometer 3.1.2.2 Diseases and disorders Samples of potato tubers with g reening, s unscald, b rown rot, g rowth cracking, and i nsect damage (for Red La Soda only) were collected in the fields at Hastings at the same time as undamaged samples over the weeks and separated into their respective damage categories. These diseases and disord ers were previously described ( Table 1 3 ) In addition to field harvested samples of potato tubers with greening samples were created by leaving undamaged potato tubers expos ed to indirect sunlight for 2 weeks in order to allow greening to develop naturally over time (Fig. 3 5 and 6 ) The number of samples with sunscald naturally increased as time after vine kill increased, due to unprotected tubers near the surface being expo sed to prolonged levels of intense sunlight (Fi g. 3 7 and 8 ) Potato tubers that developed brown rot were limited to those that were exposed to the pathogen before the parent vine was killed and continued to develop the damage related to the disease as har vest weeks went on (Fig. 3 9 and 10 ) Growth cracking occurred in samples while they were growing and exposed to uneven availability of soil moisture which caused them to split and heal along points of surface tension caused by this occurrence

PAGE 100

100 (Fig. 3 11 a nd 12 ) (Fig. 3 13 ) 3.1.3 Potato Sampling and Reflectance Measurement All reflectance measurements were c ollected using a Car y 500 spectrophotometer (Varian, Inc.) with an integrating sphere from 200 nm to 2500 nm with 1 nm increments as shown in (Fig 3 14). The spectrophotometer was allowed to warm up for 30 minutes before any samples were taken in order to allow the light sou rce to stabilize. The sample measurement port had a circular area of 1,134 mm 2 with a diameter of 38 mm as shown in (Fig. 3 15) T he coating material inside the integrating sphere was white polytetrafluoroethylene (PTFE) which has the capability to diffuse a transmitting light nearly perfectly and maintain optical properties constantly over a wide range of wavelengths from the ultraviolet up to the near infrared range as shown in (Fig. 3 16) UV and mercury lamps were used as light sources A 50 mm diameter PTFE calibration disk was used to cover the sampling port and obtain the optical reference standard for the system before spectral measurement of s amples were made at each sampling period as shown in (Fig. 3 17) Potato tuber samples for reflectance meas urement were collected on May 31 st June 7 th and June 14 th 2011. Undamaged p otato tubers from each variety were initially measured on each day of harvest in order to create a reference range of data for comparison to samples of damaged, diseased and potat o disorders. These potato tuber samples were marked with a 40 mm diameter circle at the point of initial measurement and then damaged at this point to induce mechanical damage development and reflectance measurements were made once again before samples wer e stored for 6 days to allow damage to develop. Then samples were separated into groups with

PAGE 101

101 external shatter and no outside visible damage and refle ctance measurements were made for both groups Once samples with no visible damage were measured, the tub ers were sliced along the marked portion where the point of impact occurred Once sliced, tubers were visually inspected for development of internal blackspot bruising Blackspot bruising was assessed by determining the percent of the area measured by the 38 mm diameter sampling port that bruising was present to assist in determination of reflectance changes from undamaged samples This assessment generally required additional cutting of tuber to allow for more accurate measurement. Samples with disease dam age and disorders, which were collected in the fields, were also measured in the spectro photo meter on each day of harvest. However, the additional greening samples which had greening induced on undamaged potato tubers were initially measured and marked wit h a 40 mm diameter circle at the point of measurement before greening was developed by exposure to sunlight. After greening developed, samples were measured in the spectro photo meter again to allow for comparison of samples before and after greening develop ed. Initial samples of Fabula for moisture content evaluation were collected on May 13, 2011 in order to determine the effects of moisture content levels on reflectance before testing began on reflectance measurement of damaged and diseased samples. Coll ected s amples were measured for reflectance in the spectro photo meter before evaluations of moisture content levels i n the tuber s could be determined due to the destructive nature of cutting and drying in the process of determining the moisture content Thi s was done in order to evaluate the effect of tuber moisture content level s which could influence the results of spectral reflectance among tubers with varying moisture contents The moisture content levels of agricultural produces were

PAGE 102

102 shown in many studi es to affect the reflectance of light for various reasons and spectral ranges Aluminum sample trays were weighed using an electronic balance (BX 4200H, Shimadzu Corporations, Philippines) before the area of reflectance measurement was cut off and weigh ed in each sample tray to prepare for dehydrating. Once the samples were collected they were placed in a hot air drying oven kept at 60C for 2 weeks to allow for complete removal of moisture. Percent moisture content was calculated by mass for tubers by div iding the initial wet weight minus the final dry weight by initial wet weight and multiplying th e result by 100 to get percent. 3.1.4 Determination of Important Wavelengths The data sample set s for each potato tuber variety and defect were separated and a ssessed individually on the reflectance values and first derivative using three methods; correla tion coefficient (r) spectrum, partial least squares regression and stepwise multiple linear regression 3.1.4.1 Correlation coefficient spectrum As previously stated the correlation coefficient spectrum is the simplest method used to determine important wavelen gths between spectral reflectance and the desired measurements. The Pearson product moment correlation 1 wi th a that the data points have a high linear dependence and a value of 0 indicating that there is no linear correlation between the variables. Correlation coefficients were calculated by using the SAS CORR procedure (SAS, 2009).

PAGE 103

103 3.1.4.2 Partial least squares (PLS) regression As identified earlier, PLS regression in its simplest form is a linear m odel that specifies the linear relationship between a dependent response variable Y, and a set of predictor variables X Wold (1975) initially introduced the analysis tool to the field of chemometrics and it has been spreading to other fields of study ever since. PLS is based on linear transition from a large number of original descriptors to a small number of descriptive factors to pro vide an optimal model for prediction. In order to complete the PLS model the number of factors must be chosen using the Predicted Residual Sum of Squares (PRESS) statistic in PLS Samples are divided into calibration and validation sets with 2/3 of the s amples being used for calibration and 1/3 used for validation. The factors are chosen using cross validation, in which the data set s are divided into two or more groups and t his is repeated for each group in order to measure the overall capability of a giv en form of the model. The PRESS statistic is based on the residuals generated by this process. The optimum number of factors for the PLS models are generally ob tained when factors are minimized with a smaller PRESS value s indicating better model prediction (Sundberg, 1999) Often selecting the number of factors where the absolute minimum PRESS exists may not be the best choice since a lower number of factors may have a significantly close PRESS value Cross validation is a method specified by van der Voet (1994) which uses randomized based model comparison testing to compare test models with different numbers of extracted factors against the model that minimizes the PRESS. PLS regressions can be calculated by using th e SAS PLS procedure with cross validatio n performed using the CVTEST option in SAS to test the significant differences in the PRESS values (SAS, 2009).

PAGE 104

104 3.1.4.3 Stepwise multiple linear regression (SMLR) As explained earlier, SMLR starts with no variables in the model and the basic method of ob taining the optimal predictors starts with each x variable in the dat um set; determining the most highly correlated regression coefficient in the regression model relating to the y variable which becomes the first selected variable denoted x s1 and the resi dual sum of squares RSS to assess the correlation of the predictor used in the model. Then r egression models with both the pre selected variable, x s1 and each of the remaining x variables are determined using MLR and the corresponding RSS value are calcul ated for each model. The x variable that produces the highest RSS when used in combination with x s1 then becomes the second selected variable and is deno ted, x s2 This process repeats until the desired number of variables selected by the stepwise regressio n routine is equal to the desired number of variabl es as specified by the operator (Yee 1999) SMLR was calculated by using the SAS PROC REG model selection to perform the stepwise regression method with a threshold value of 0.05 signific ance to be included in the model, once wavelength selection is unable to find a variable that meets this requirement analysis ends 3.2 Results and Discussion 3.2.1 Effects of Water on Spectral Characteristics Moisture content distributions collected for experiments, which are similar to the skin moisture content previously determined during mechanical damage assessment at 82% to 89% for

PAGE 105

105 freshly harvested new potatoes and variation between varieties. The range of water content can range between 73.8% and 81% according to Souci et al. (2000) and Elbatawi et al. (2008) found water content levels from as high as 85% with average water contents around 81.7%. Fi gure 3 18 2500 nm with various moisture contents, while figure 3 20 tance versus moisture content between 200 and 2500 nm, while figure 3 potato samples. s to vary greatly because of the moisture content level s of potato tubers above the 850 nm spectral range, as shown by the extremely high correlation (r) coefficient betwe en moisture and reflectance (F igure 3 20 and 21). The correlation of moisture content for both varieties show results similar to those found by Singh (2005) who determined that the peaks in the wavelength ranges of 738 837, 914 1120, and 1354 1456 nm correspond to the changes of water content in an unknown variety of potato samples The slight difference in peaks caused by water could be due t o the physical experimentation and attributed to losses and inefficiencies occurring between molecular al properties of samples. The result of moisture content testing versus reflectance spectrums helped influence the decision to run experiments between the range of 360 nm and 800 nm M oisture content ha d a reduced secondary effect on the range of waveleng ths between 360 nm and 800 nm This

PAGE 106

106 decision was made in order to avoid error s caused by widely varying moisture contents found naturally in potatoes depending on pre harvest and post harvest conditions. 3.2.2 Correlation Coefficient Spectrum The results f or reflectance measurements and correlation coefficients of disorders and diseases (Internal Bruising, Shatter, Greening, Sunscald, Brown Rot, Growth Cracking and La ( Figure 3 21 thru 3 53 ) While the sample sizes used in the evaluati on of each variety are shown ( Table 3 1 ) Correlation coefficient spectrums for both varieties were created from the correlation between re flectance and the percent of damaged area within the measured dimensions of the tuber for each disorder and disease. The Pearson product moment correlation coefficient denoted by (r) lies between 1 and 1 with a value of 1 or 1 indicating high linear depe ndence and a value of 0 indicating that there is no linear correlation between the variables (SAS, 2009) Areas of reflectance for externally visible sample sets were highly correlated with the presence of damage for both varieties and the combination of b oth while internal bruising was only slightly correlated for a wide range of wavelengths. However, even the lower correlation coefficient of 0.4 found for internal bruising with combined varieties would be considered significant in the research done by Mi n et al. (2004) who used correlation coefficients with r larger than 0.3. I nternal bruising coefficients between 360 800 nm for both varieties and combination of the two (Figure s 3 24 thru 3 26) 701 nm were the mos varieties showed the highest correlation above 600 nm up to 800 nm. These results can be

PAGE 107

107 attributed to the melanin development that occurs when sufficient damage cause s cellular disruption according to Mathew and Hyde (1997). S hatter coefficients between 360 800 nm for both varieties and combination of the two (Figures 3 29 thru 3 31) 6 40 nm were highly between 360 and 660 nm with |r|>0.8, while the combination of both showing a reduced correlation of latter between 360 and 660 nm with |r|>0.6. These results can be tied to the simil ar wavelength reflectance caused by the starch present in the tubers of all varieties. G reening coefficients between 360 800 nm for both varieties and combination of the two (Figures 3 34 thru 3 36) nm were extremely correlated with |r| >0.8 |r|>0.4 while samples were extremely correlated between 5 90 and 7 0 0 nm with | r| >0. 8. When combining the two varieties the correlation between reflecta nce and greening was most highly correlated between 400 480 nm and 600 700 nm with |r|>0.6. The reduced significance of color which could already have some greens and other pigments associated with it naturally in those reflected by chlorophyll which causes the greening. S unscald coefficients between 360 800 nm for both var ieties and combination of the two (Figures 3 39 thru 3 41) 3 60 480 nm and 6 20 710 nm with |r| >0. 6, while the combi nation of both varieties showed slightly reduced correlation

PAGE 108

108 reflection are associated with the burning of pigments in the tubers which causes yellow and brownish color ation in the tubers. B rown rot coefficients between 360 800 nm for both varieties and combination of the two (Figures 3 44 thru 3 46) 360 and 800 n m that were extremely correlated around |r| >0.8. A co mbination of both varieties had a reduction in correlation between the wavelengths 360 600 nm with |r|<0.7, while the rest of the spectrum stayed above |r|>0.8. These results could mean that the main pigment formation caused by brown rot lies above the 600 present in the 360 G rowth cracking coefficients between 360 800 nm for both varieties and combination of the two (Figures 3 49 thru 3 51) nm were extremely correlated with |r| >0. again had reflectance between 360 4 60 nm and 620 800 nm extremely correlated with |r| >0. 6. These results are due exhibits a stretching appearance and less extreme scarring formation. I nsect damage coeffic ients lie between 360 800 (Figure 3 53) The results show a reflectance between 360 480 nm and 640 800 nm extremely correlated with |r| >0.8. These results show a significant correlation range similar to those for sunscald and growth c racking, which could make differentiating between the three types of damage difficult.

PAGE 109

109 Each set of data for Fabula potato disorders has a range of values between 360 nm and 800 nm with a |r| value greater than 0.4; these results display that there is a c orrelation between wavelength values and percent area damaged being scanned which be used to calibrate a model in order to predict damage. 3.2.3 Partial Least Squares (PLS) Regression In order to complete the PLS model, the number of factors must be chosen using the Predicted Residual Sum of Squares (PRESS) statistic in PLS, predictive modeling using cross validation and minimizing the error of the prediction model by checking how the number of factors affects the results Sundberg (1999) states that the op timal number of factors for PRESS is generally ob tained when factors are minimized and smaller PRESS sizes usually develop better fit to data. Results of PLS analysis for Fabula ( Table 3 2 Table 3 3 ) and a c ombination of both varieties ( Table 3 4 ) The tables depict the extracted factors, dependent variables, PRESS, SEC, SEP RPD and RMSD for selected factor values of each varieties disorder sample set. T he regression relationship between predicted damage values versus actual damage valu es for the number of factors chosen for each variety and the combination of both sample sets using reflectance and 1 st derivative evaluation, as well as the B coefficients determined for each wavelength using PLS in order to show the most significant wavel engths from the reflectance spectrum measured (Figures 3 54 thru 98) P of both (Figures 3 54 thru 3 60) had relatively low wavelength selection variation with R 2 higher t han 0.96 for all prediction models using a various number of extracted factors which would

PAGE 110

110 always favor 1 st smallest error was at 15 factors and obtained a SEP of 0.991% with sta ndard wavelengths, while 1 st needed 11 factors to obtain a SEP of 1.245% with standard wavelengths, while 1 st derivative results only needed four factors in orde r to obtain a SEP of 0.960%. Combining the two varieties resulted in an increase of prediction error by increasing the SEP to 2.847% with 15 factors used, while 1 st derivative SEP was reduced to 1.487% with seven factors. For internal bruising prediction o f each variety the most prediction error occurred in samples with no damage at all, while combining them resulted in error in prediction as high as 9% in healthy samples and varying error in damaged samples. Internal bruising which was associated with the production of melanin had its h ighest B coefficients at 799, 795, 787, 768, 735, 732, 714, 709, 692, 464, 425, 400, 394, 390, 384, 374, 373, 371 and 365 nm; while 1 st derivative results with five with 11 factors they were 371, 363, 365, 370, 380, 390, 397, 714, 766, 782, 790, 770 and 787 nm; while 1 st derivative results with four factors were 683, 687, 588, 566, 462, 465 and782 nm For the combination of varieties with 15 factors they were 793, 790, 782, 762, 752, 747, 710, 392, 383, 378, 372, 373 and 370 nm; while 1 st derivative results with seven factors were 728, 709, 793, 787, 676, 660, 420 and 400 nm. This shows that the most important wavelengths for model prediction were in the same general area for both varieties. P (Figures 3 61 thru 3 67) had relatively low wavelength selection variation wit h R 2 higher than 0.97 for all prediction models using a various number of extracted factors which would always

PAGE 111

111 favor 1 st was obtained with four factors at a SEP of 1.239 % with standard wavelengths, while 1 st needed four factors to obtain a SEP of 0.936 % with standard wavelengths, while 1 st derivative results only needed t wo factors in order to obtain a SEP of 1.261%. Combining the two varieties resulted in an increase of prediction error by increasing the SEP to 1.678% with five factors used, while 1 st derivative SEP results were 1.783% with two factors. For shatter predic tion the most prediction error occurred in samples with no damage or very high damage with error in prediction as high as 15% for extremely damage samples for combined spectral data of the two varieties. Shatter wavelengths were the result of missing skin surface area and increased reflectance from potato starch. The highest B four factors were between 500 516, 475 486, 445 455, 360 400 nm; while 1 st derivative results with two factors were betw een 360 371, 380 425, 436 448, 455 465, 470 480 and 482 510 nm. 384, 423 462 and 514 579 nm; while 1 st derivative results with two factors were between 360 371, 397 405, 427 436, 470 480 and 488 520 nm. For the combination of varieties with five factors they were 360 397, 440 460 and 473 484 nm; while 1 st derivative results with two factors were 768, 752, 722, 713, 540, 449 466 and 553 nm. This shows that the most important wavelengths for model pred iction were in the same general area for both varieties. P (Figures 3 68 thru 3 74) had relatively low wavelength selection variation with R 2 higher than 0.98 for all predi ction models using a various number of extracted factors which would always

PAGE 112

112 favor 1 st derivative prediction using fewer factors in all situations. was at seven factors with a SEP of 0.017% for standard wavelengths, while 1 st derivative results to obtain a SEP of 1.881% with standard wavelengths, while 1 st derivative results only needed two factors in order to obtain a SEP of 0.771%. Combining the two varieties resulted in an increase in factors by requiring seven to get an SEP of 1.511%, while 1 st derivative improved SEP results to 1.003% with three factors. For greening prediction there was almost no error at all in the prediction m odel due to the easily detected increase in chlorophyll which affected the wavelength reflectance. Greening wavelength changes were associated with the increase in chlorophyll, solanine and other photosynthesis chemicals. The highest B coefficients calcula ted by the PLS procedure 1 st derivative results with two factors were 767, 768, 758, 748, 722, 670, 530, 506, 478, 463 and tors they were between 618 698 nm; while 1 st derivative results with two factors were between 680 715, 725, 735, 744, 750, 797, 530, 497, 480, 475, 463 and 461 nm. For the combination of varieties with seven factors they were between 670 683 nm; while 1 st derivative results with three factors were 725, 748, 798, 473, 442, 406 and 397 nm. This shows that the most important wavelengths for model prediction were the ones that contributed P (Figures 3 75 thru 3 81) had relatively low wavelength selection variation with R 2 higher than 0.97 for all prediction models using a various number of extracted factors which would always favor 1 st derivative pre

PAGE 113

113 minimized at eight factors in order to obtain a SEP of 1.154% with standard wavelengths, while 1 st needed nine factors to obtain a SEP of 1.4 % with standard wavelengths, while 1 st derivative results only needed three factors in order to obtain a SEP of 0.957%. Combining the two varieties resulted in an increase in factors by requiring 15 to get an S EP of 2.028%, while 1 st derivative improved SEP results to 1.476% with six factors. For sunscald prediction the most prediction error occurred once again in samples with no damage at all which resulted in error in prediction as high as 10% with combined sp ectral data of the two varieties. Sunscald was the burning of potato surface area by exposure to high temperatures and sun light. The highest B coefficients factors were between 798 788, 785, 780, 777 397, 371, 365 362 nm; while 1 st derivative results with four factors were 798, 797, 788, 787, 780, 748, 726, 722, 718, 711, 682, 610, 606, 600 and and 360 nm; wh ile 1 st derivative results with three factors were 725, 711, 687, 540, 527 and 512 nm. For the combination of varieties with 15 factors they were 800, 792, 788, 780, 770, 766, 764, 746 and 722 nm; while 1 st derivative results with six factors were 797, 790 788, 777, 774, 750, 720, 713, 708 and 680 nm. This shows that the most important wavelengths for model prediction were in the same general area for both varieties. P (Fi gures 3 82 thru 3 88) had relatively low wavelength selection variation with R 2 higher than 0.98 for all prediction models using a various number of extracted factors which would always favor 1 st derivative prediction using fewer factors in all situations. factors to minimize error with a SEP of 1.34% with standard wavelengths, while 1 st derivative

PAGE 114

114 factors to obtain a SEP of 1.267% with stand ard wavelengths, while 1 st derivative results only needed four factors in order to obtain a SEP of 0.822%. Combining the two varieties resulted in an increase in factors by requiring 15 to get an SEP of 1.878%, while 1 st derivative SEP results were 2.314% with four factors. For brown rot prediction the error was evenly distributed for the two varieties, but combining them caused prediction error focused around samples with no damage which was likely caused by the differences between the skin coloration of t he varieties. Brown rot caused by the bacterium Ralstonia solanacearum causes tuber flesh to rot and turn a brownish grey color which resulted in t he highest B coefficient s calculated by the PLS 375, 36 6 and 365 nm; while 1 st derivative results with three factors were 787, 725, 676, 677, 646, 608, 607, 594, 589, 576, 566, 533, 470 nm; while 1 st derivative results with four factors were 799, 791, 790, 777, 774, 736, 726, 713, 709, 688, 673, 670, 657, 533 and 507 nm. For the combination of varieties with 15 factors they were 793, 790, 782, 762, 747, 383, 392, 372, 373 and 370 nm; while 1 st derivative results with se ven factors were 797, 790, 683, 660, 530, and 423 nm. This shows that the most important wavelengths for model prediction were in the same general area for both varieties. P of both (Figures 3 89 thru 3 95) had relatively low wavelength selection variation with R 2 higher than 0.97 for all prediction models using a various number of extracted factors which would always favor 1 st derivative prediction using fewer factors in all factors to minimized error at a SEP of 2.086% with standard wavelengths, while 1 st derivative

PAGE 115

115 to obtain a SEP of 2.120% with st andard wavelengths, while 1 st derivative results only needed five factors in order to obtain a SEP of 1.071%. Combining the two varieties resulted in an increase in factors by requiring 15 to get an SEP of 3.039%, while 1 st derivative improved the SEP resu lts to 1.949% with six factors. For growth cracking prediction the error was evenly distributed for the two varieties, but combining them caused prediction error of up to 15% focused around samples with no damage which was likely caused by the differences between the cracking developments of the two varieties. Growth cracking which was the result of tuber suberization to repair splitting during growth had its highest B coefficients factors at 361, 363, 366, 371, 373, 382, 390, 394, 787, 790 and 795 nm; while 1 st derivative results with four factors were 790, 787, 756, 733, 729, 722, 711, 677, 645, 643, 631, 602, 536, 722, 713, 685, 662, 566, 533, 516, 458 and 402 nm; while 1 st derivative results with five factors were 622, 433, 625 and 633 nm. For the combination of varieties with 15 factors they were 800, 797, 789,787, 782, 778, 752, 738, 734, 722, 704, 404, 398, 38 0, 378 nm; while 1 st derivative results with six factors were 789, 787, 753, 730, 686, 665, 579, 540, 475, 442, 413 and 371 nm. This shows that the most important wavelengths for model prediction were in the same general area for both varieties. P rediction (Figures 3 96 thru 3 98) which had relatively low wavelength selection variation with R 2 higher than 0.99 for all prediction models using a various number of extracted factors which would always favor 1 st derivat ive prediction using fewer factors in all situations. to obtain a SEP of 1.216% with standard wavelengths, while 1 st derivative results only needed

PAGE 116

116 three factors in order to obtain a SEP of 0.988%. For insect damage prediction the error was evenly distributed due to low sample numbers which were primarily extremely damaged. Insect damage which was the result of tuber suberization from tuber matter being eaten; had its highest B coefficients calculated by the P at 800, 797, 792, 787, 783, 777, 750, 527, 507, 465, 380, 376, 375, 371 and 363 nm; while 1 st derivative results with three factors were 793, 787, 770, 740, 738, 700, 690, 670, 657, 580, 574, 568, 519, 512, 50 2, 436 and 420 nm. Since all sample disorders had significant wavelengths with a high B coefficient value between the rand of 360 460 and/or 700 800 spectral wavelength scans could be concentrated on these areas in order to calibrate models for prediction of damage faster on a packaging line. Similarly, Xing et al. (2005) observed pronounced spectral changes for apples containing the 900 nm. 3.2.4 Stepwise Multiple Linear Regression (SMLR) SMLR wa s used as a second method of wavelength significance examination. The results of statistical analysis for Table 3 5 ) Table 3 6 ) and the c ombination of both varieties ( Table 3 7 ) The stepwise option of SAS sele cted the most highly correlated wavelengths for each Fabula sample set until the accuracy of new wavelengths gave diminishing returns below a threshold significance of 0.05. Prediction models racy of model calculations (Figure 3 99 thru 3 117). selected wavelengths 795, 791, and 435 nm resulted in a selection variation R 2 of 0.9202.

PAGE 117

117 Stepwise selection picked six wavelengths to predict Fabula internal bruising 1 st derivative, the selected wavelengths 408, 742, 531, 456, 482, and 435 nm resulting in a R 2 2 of 0.9824; while the 1 st derivative used eight wavelengths 640, 412, 705, 648, 650, 786, 781 and 484 nm for a R 2 of 1. Combining the two varieties resulted in a wavelength selection of no significant variables that met the 0.05 threshold; while 1 st derivat ive results selected 10 wavelengths 682, 667, 716, 427, 589, 558, 791, 576, 540 and 443 nm for a R 2 of 0.9711. versus 788 and 784, combining the two varieties int ernal bruising data failed to give results with enough significance to include. Using the 1 st derivative however, resulted in wavelengths near the previously individually selected wavelengths being used. Stepwise selection picked eight wavelengths to predi selected were 432, 438, 383, 379, 382, 386, 376 and 471 nm for a R 2 of 0.9996; the 1 st derivative selection of wavelength chose 11 which were 499, 500, 629, 504, 688, 613, 722, 790 690, 648 and 654 nm for a R 2 nm to produce a variation R 2 of 0.9999; while the 1 st derivative used nine wavelengths 502, 625, 788, 576, 585, 471, 642, 621 and 479 nm for a R 2 of 0.9999. Combining the two varieties resulte d in the selection of the five wavelengths 378, 362, 388, 395 and 422 nm for a R 2 of 0.9945; while 1 st derivative results selected 14 wavelengths 672, 388, 390, 437, 582, 365, 366, 451, 387, 703, 570, 541, 649 and 733 nm for a R 2 of 0.9988. Combining the s hatter data for both varieties resulted in a selection of wavelengths which appeared to be more influenced by the st derivative results were equally influenced to the wavelength selection of both varieties being i n the same general area.

PAGE 118

118 wavelength 702 and 503 nm resulted in a R 2 of 0.9932. Stepwise selection picked four wavelengths to predict 1 st derivative, the selected wavelengt hs 688, 510, 533 and 625 nm resulting in a R 2 variation R 2 of 0.9978; while the 1 st derivative used four wavelengths 650, 391, 673 and 730 nm for a R 2 of 0.9999. Combining the two var ieties resulted in the selection of the three wavelengths 681, 672 and 517 nm for a R 2 of 0.9927; while 1 st derivative results selected 10 wavelengths 697, 758, 578, 763, 363, 711, 738, 405, 747 and 562 nm for a R 2 of 0.9999. Combining greening data result ed in wavelength selection in the general area of those selected separately, while 1 st w avelength 446 and 432 resulted in a R 2 of 0.979. Stepwise selection picked six wavelengths to predict Fabula sunscald 1 st derivative, the selected wavelengths 518, 792, 625, 421, 650 and 543 nm resulting in a R 2 of 1. hs 435 and 433 nm to produce a variation R 2 of 0.9852; while the 1 st derivative used six wavelengths 750, 608, 425, 530, 444 and 675 nm for a R 2 of 1. Combining the two varieties resulted in the selection of the seven wavelengths 786, 773, 589, 588, 629, 7 87 and 758 nm for a R 2 of 0.9968; while 1 st derivative results selected six wavelengths 729, 430, 483, 433, 428 and 553 nm for a R 2 of 0.9932. Combining the sunscald data resulted in wavelength selection which varied greatly from those selected for both va rieties separately, which were both between 446 and 432 nm. Combined 1 st derivative wavelength selection matched both varieties very closely. Stepwise selection picked two wavelengths to predict Fabula brown rot; the selected wavelength 689 and 688 nm re sulted in a R 2 of 0.9969. Stepwise selection picked six

PAGE 119

119 wavelengths to predict Fabula brown rot 1 st derivative, the selected wavelengths 518, 792, 625, 421, 650 and 543 nm resulting in a R 2 607 nm to p roduce a variation R 2 of 0.9824; while the 1 st derivative used six wavelengths 631, 781, 776, 629, 442 and 615 nm for a R 2 of 0.9996. Combining the two varieties resulted in the selection of the three wavelengths 708, 487 and 712 nm for a R 2 of 0.9931; whi le 1 st derivative results selected 15 wavelengths 567, 579, 433, 662, 717, 759, 405, 439, 737, 671, 434, 718, 390, 482 and 689 nm for a R 2 of 0.9999. Combing the data for brown rot resulted in very different st derivative selection however resulted in a wide selection of wavelengths that were close to some of the results for both varieties. Stepwise selection picked seven wavelengths to predict Fabula growth cracking, the selected wavelengths 545, 516, 517, 54 7, 477, 543 and 486 nm resulting in a R 2 of 0.9995. Stepwise selection picked six wavelengths to predict Fabula growth cracking 1 st derivative, the selected wavelengths 510, 769, 528, 453, 529 and 473 nm resulting in a R 2 of 0.9996. e wavelength 381 nm to produce a variation R 2 of 0.8793; while the 1 st derivative used four wavelengths 622, 433, 625 and 633 nm for a R 2 of 0.9978. Combining the two varieties resulted in the selection of the six wavelengths 670, 461, 673, 400, 401 and 66 9 nm for a R 2 of 0.9906; while 1 st derivative results selected 18 wavelengths 475, 460, 651, 520, 440, 603, 579, 459, 427, 770, 685, 638, 387, 512, 381, 479, 556 and 800 nm for a R 2 of 1. Combining data for growth cracking resulted in wavelength selection that was not similar to either variety, while 1 st derivative selection chose wavelengths across the spectrum. selected wavelength 800 nm resulted in a R 2 of 0.8891. Stepwi se selection picked two

PAGE 120

120 wavelengths to predict 1 st derivative, the selected wavelengths 549 and 586 nm resulting in a R 2 of 1. 3.3 Summary This chapter was focused on the objective to determine important wavelengths in the electromagnetic spectrum that cou ld help develop a real time damage sensor for the potato Fabula potato samples were measured from 200 nm to 2500 nm using a spectrophotometer in a laboratory environment in order to deter mine the effects of moisture content on the reflectance found in tubers. While examining results, the reflectance spectrum for Fabula was found to be greatly influenced by moisture content levels above 900 nm; this was similar to results found by Yee and (1999), where they discovered evidence for 5 maximum wavelengths between 700 2500 nm (at 760, 970, 1190, 1450 and 1940 nm) which were heavily influenced by moisture content in organic samples and stated that physical properties of organic sampl es could influence the location of moisture content wavelength peaks. Bull (1991) suggests that if reference wavelengths are sensitive to the moisture content of the sample that calibration may become linear over a relatively small range of moisture conten ts and states it is preferable to choose reference wavelengths for which reflectance is relatively insensitive to moisture content. These results which naturally varied between 78% and 92.3% moisture content across the two varieties be less expensive for packing house purposes led to the decision to use the range of the electromagnetic spectrum from 360 nm to 800 nm for disorder de tection in potato tuber samples

PAGE 121

121 The reflectance of Fabula and Red La Soda samples were measured from 360 nm to 800 nm with a spectrophotometer in a laboratory environment. A correlation coefficient spectrum, PLS regression and SMLR procedures were then used to determine important wavelengths for de tecting; while PLS was also used for prediction of percent damage present in samples. PLS regression yielded good results, while SMLR and correlation coefficient gave alternative information into the significance of some wavelength ranges in damage detecti on (Tables 3 8 thru 3 10) For internal bruising, both varieties showed significant correlation coefficient above 0.4 between 600 800 nm, though when combined this dropped to just above 0.2 for the same range. PLS and SMLR analysis of both varieties result ed in many wavelength selections within the same range as high correlation wavelengths and 1 st derivatives separately and combined selected several wavelengths to have a high significance between 360 500 nm. These results indicate that the physical changes associated with internal bruising that occur have the highest effect on reflectance in this range. Melanin was known to effect wavelengths around 475 nm which indicates that the two varieties have different physical changes that affect the results, while combining the varieties targets wavelengths around melanin as the most significant. For shatter damage, both varieties show significant correlation coefficient above 0.4 between 360 660 nm 490 nm while a c ombination of the two varieties maintained significant correlation through the entire same range. PLS and SMLR analysis determined that the most significant wavelengths were between 360 600 nm for both varieties and the combined data, as well as a few valu es in the up 700 nm range u s ing 1 st derivative. The results of analysis show where both varieties have similar reflectance due to the presence of potato tuber starches present in the flesh.

PAGE 122

122 For greening damage, had significant correlation coeffici ent above 0.4 for the entire range, while significant corr elation between 510 570 nm and 590 700 nm above 0.4. Combining the two varieties resulted in a correlation coefficient range above 0.4 between 376 505 nm and 580 715 nm. PLS and most significant wavelengths between 760 800 nm, 702 nm, 503 nm and between 360 385 nm. Using 1 st derivatives, wavelengths between 440 avelengths between 618 698 nm and 1 st derivatives in the range on 400 800 nm. Combining the data resulted in most significant wavelengths being between 670 683 nm and at 517 nm. The 1 st derivative data determined wavelengths between 700 800 nm and 360 580 nm to be highly significant. Many of the wavelengths determined to be significant were within the photosynthetically active region of 450 nm to 670 nm which had the most significant influence in the visible light range; solanin would have had an effect on significant reflectance changes outside this range. For sunscald damage, showed significant correlation above 0.4 between 368 and 800 nm while significant correlation between 360 514 nm and 611 800 nm. Combining of both va rieties resulted in a slightly reduced range of significant wavelengths between 360 477 nm and 618 800 nm. PLS and SMLR analysis determined the most significant 800 nm and 360 440 nm, while 1 st derivative waveleng ths were between 600 800 nm for PLS and 500 found to have most significant wavelengths between 640 800 nm and 360 366 nm for PLS and 433 nm and 435 nm for SMLR, while 1 st derivative had values in the range between 420 725 nm. Combining the data resulted in the most significant wavelengths being between 720 800 nm and 580 630 nm with SMLR. The 1 st derivative data determined significant wavelengths to be

PAGE 123

123 between 680 800 nm with PLS and between 430 555 nm with SMLR. The refle ctance changes in sunscalded samples would be caused by the radiative burning caused by high temperatures and sun exposure, which caused chemical changes in the tuber surface and subsurface areas. and combining the two varieties showed a significant correlation coefficient above 0.4 for the entire range. PLS and SMLR analysis determined the 380 nm using PLS and 680 690 nm with SMLR. Using 1 st derivative data wa velengths between 420 800 nm were found with both 800 nm and 360 390 nm. 1 st derivative analysis selected wavelengths between 500 800 nm. Combining the two sets of data resulte d in wavelength selection between 700 800 nm and 360 400 nm, while the 1 st derivative data selecting most significant wavelengths between 400 800 nm. Brown rot caused by the bacterium Ralstonia solanacearum had an effect on reflectance at both ends of the measured range associated with the rotting of tuber flesh which causes brown grey discoloration and the development of creamy pus. For growth cracking, 800 nm with significant correlation above 0.4, w showing significance between 360 495 nm and 590 800 nm. PLS and SMLR analysis determined the 400 nm and 785 800 nm using PLS opposed to SMLR which determined the range between 470 550 nm to be most significant. The 1 st derivative wavelengths were found to be between 450 800 nm, while the 1 st derivative of data determined th e wavelengths to be between 430 640 nm. Combining the data resulted in most significant wavelengths being between 660 800 nm and 375 400 nm. The 1 st derivative data

PAGE 124

124 selected wavelengths from the entire spectrum measured for both methods. All reflectance ch anges were affected by the suberization of tuber flesh in order to repair damage. Reflectance changes would have also been influenced by the development of H 2 O 2 accumulation, which causes suberization to occur. For insect damage between 360 51 0 nm and 58 0 800 nm was significantly between 750 800 nm and 360 530 nm. While 1 st derivative wavelengths were between 420 800 nm. Insect damage repair was perform ed by the same suberization process as growth cracking and resulted in similar wavelength ranges being most significant on reflectance changes. Each set of data for Fabula ranges of spectral wavelength with a correlation coefficient value |r| greater than 0.4 between 360 nm and 800 nm; this shows that there is a correlation between wavelength values and percent area damaged being scanned which can be used to calibrate a model in order to predict damage. PLS and SMLR analysis showed t hat all sample disorders had significant wavelengths with a high coefficient of significance between either 360 50 0 nm or 6 00 800 nm, if not both. This would mean that spectral wavelength scans could be concentrated on these areas in order to calibrate mod els for prediction of damage faster on a packaging line. Lu and Arianna (2002) referenced that advances in CCD technology have enabled rapid acquisition of spectral information in the visible and short NIR region (400 1100 nm) which would allow be very adv antageous for packaging line applications using the wavelengths found to accurately predict the presence of tuber disorders. However, as Singh et al. (2005) stated, spectroscopic techniques require a set of well designed calibration processes in order to a chieve the best models to predict specific parameters

PAGE 125

125 of interest in a food product. Thus more data would need to be collected in order to allow for more accurate prediction of specific disorders and whether or not suffered damage would warrant removal fro m the packaging line due to models having greater error associated with undamaged healthy tubers.

PAGE 126

126 Table 3 1 Number of samples collected for each variety and d isorder. Fabula Red La Soda Disorder Damaged Healthy Damaged Healthy Bruising 14 29 1 5 24 Shatter 1 7 79 17 77 Greening 13 20 13 20 Sunscald 12 79 12 77 Brown rot 13 79 15 77 Growth c racking 14 79 14 77 Insect d amage 5 77 Fabula 3 1 Fabula Internal Bruising Figure 3 2 Red La Soda Internal Bruising

PAGE 127

127 Figure 3 3 Fabula Shatter Figure 3 4 Red La Soda Shatter Figure 3 5 Fabula Greening

PAGE 128

128 Figure 3 6 Red La Soda Greening Figure 3 7 Fabula Sunscald Figure 3 8 Red La Soda Sunscald

PAGE 129

129 Figure 3 9 Fabula Brown Rot Figure 3 10 Red La Soda Brown Rot Figure 3 11 Fabula Growth Cracking

PAGE 130

130 Figure 3 12 Red La Soda Growth Cracking Figure 3 13 Red La Soda Insect Damage

PAGE 131

131 Figure 3 14. Cary 500 Scan Spectro photo meter (Varian, Inc. Palo Alto, CA)

PAGE 132

132 F igure 3 15. Cary 500 Sample measur ement port Figure 3 16. Integrating sphere with white polytetrafluoroethylene coating

PAGE 133

133 Figure 3 17 PTFE calibration disk Figure 3 18 Fabula Moisture Content Reflecta nce spectrum (200 nm to 2500 nm) 0 10 20 30 40 50 60 70 80 200 700 1200 1700 2200 % R Wavelength (nm) 72.2% 74.8% 77.4% 78.9% 79.4% 79.7% 79.9% 80.2% 80.6% 81.1% 81.9% 82.9% 84.9%

PAGE 134

134 Figure 3 19 Red La Soda Moisture Content Refle ctance Spectrum (200 nm to 2500 nm) Figure 3 20 Fabula Correlation c oefficient between tuber moisture content and wavelength 0 10 20 30 40 50 60 70 80 200 700 1200 1700 2200 % R Wavelength (nm) 81.0% 82.6% 83.6% 85.2% 86.1% 86.4% 86.5% 86.9% 87.2% 87.8% 88.5% 89.3% 91.1% -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 200 700 1200 1700 2200 Correlation Coefficient Wavelength (nm)

PAGE 135

135 Figure 3 21 Red La Soda Correlation c oefficient between tuber moisture content and wavelength Figure 3 22 Fabula Intern al Bruising vs. Undamaged 6 days storage -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 200 700 1200 1700 2200 Correlation Coefficient Wavelength (nm) 0 10 20 30 40 50 60 70 80 360 410 460 510 560 610 660 710 760 % Reflectance Wavelength (nm) Average Undamaged (29 samples) Average Bruised (14 samples)

PAGE 136

136 Figure 3 23 Red La Soda Internal Bruising vs. Undamaged 7 days storage Figure 3 2 4. Fabula Internal Bruising Correlation coefficients between reflectance at each wavelength and internall y damaged tissue con centration. 0 10 20 30 40 50 60 70 80 360 410 460 510 560 610 660 710 760 % Reflectance Wavelength (nm) Average Undamaged (24 samples) Average Bruised (15 samples) -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm)

PAGE 137

137 Figure 3 25 Red La Soda Internal Bruising Correlation coefficients between reflectance at each wavelength and internall y damaged tissue concentration. Figure 3 26 Combined Internal Bruising Correlation coefficients between reflectance at each wavelength and internall y damaged tissue concentration. -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm) -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r ) Wavelength (nm)

PAGE 138

138 Figure 3 2 7 Fabula Shatter vs. Undamaged Day of Harvest Figure 3 28 Red La Soda Shatte r vs. Undamaged Day of Harvest 0 10 20 30 40 50 60 70 80 360 400 440 480 520 560 600 640 680 720 760 800 % Reflectance Wavelength (nm) Average Shatter (17 samples) Average Undamaged Fabula (79 samples) 0 10 20 30 40 50 60 70 80 90 360 400 440 480 520 560 600 640 680 720 760 800 % Reflectance Wavelength (nm) Average Shatter (17 samples) Average Undamaged Red (79 samples)

PAGE 139

139 Figure 3 29 Fabula Shatter Correlation coefficients between refl ectance at each wavele ngth and Shatter concentration. Figure 3 30 Red La Soda Shatter Correlation coefficients between reflectance at each wavelen gth and Shatter concentration. -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm) -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm)

PAGE 140

140 Figure 3 31 Combined Shatter Correlation coefficients between reflectanc e at each wavelen gth and Shatter concentration. Figure 3 32 Fabula Greening vs. Undamaged Day of Harvest -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r ) Wavelength (nm) 0 10 20 30 40 50 60 70 80 90 360 410 460 510 560 610 660 710 760 % Reflectance Wavelength (nm) Average Undamage Fabula (20 samples) Average Greened Fabula (13 samples)

PAGE 141

141 Figure 3 33 Red La Soda Greening vs. Undamaged Day of Harvest Figure 3 34 Fabula Greening Correlation coefficients between reflectance a t each wavelength and Greening concentration. 0 10 20 30 40 50 60 70 80 360 410 460 510 560 610 660 710 760 % Reflectance Wavelength (nm) Average undamaged (20 samples) Average Greened Red La Soda (13 samples) -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm)

PAGE 142

142 Figure 3 35 Red La Soda Greening Correlation coefficients between reflectance at each wavelength and Greening concentration. Figure 3 36 Combined Greening Correlation coefficients between reflectance at e ach wavelength and Greening concentration. -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm) -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r ) Wavelength (nm)

PAGE 143

143 Figure 3 37 Fabula Sunscald vs. Undamaged Day of Harvest Figure 3 38 Red La Soda Sunscald vs. Undamaged Day of Harvest 0 10 20 30 40 50 60 70 80 360 400 440 480 520 560 600 640 680 720 760 800 % Reflectance Wavelength (nm) Average Sun Damage (12 samples) Average Undamaged (79 samples) 0 10 20 30 40 50 60 70 80 360 400 440 480 520 560 600 640 680 720 760 800 % Reflectance Wavelength (nm) Average Sun Damage (12 samples) Average Undamaged Red (77 samples)

PAGE 144

144 Figure 3 39 Fabula Sunscald Correlation coefficients between reflectance at eac h wavelength and Sun Damage concentration. Figure 3 40 Red La Soda Sunscald Correlation coefficients between reflectance at each wavelength and Sun Damage concentration. -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm) -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm)

PAGE 145

145 Figure 3 41 Combined Sunscald Correlation coefficients between reflectance at e ach wavelength and Sun Damage concentration. Figure 3 42 Fabula Brown rot vs. Undamaged Day of Harvest -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r ) Wavelength (nm) 0 10 20 30 40 50 60 70 80 360 400 440 480 520 560 600 640 680 720 760 800 % Reflectance Wavelength (nm) Average Brown Rot (13 samples) Average Undamaged (79 samples)

PAGE 146

146 Figure 3 43 Red La Soda Brown rot vs. Undamaged Day of Harvest Figure 3 44 Fabula Brown Rot Correlation coefficients between reflectance a t each wavelength and Brown rot concentration. 0 10 20 30 40 50 60 70 80 360 400 440 480 520 560 600 640 680 720 760 800 % Reflectance Wavelength (nm) Average Brown Rot (15 samples) Average Undamaged (77 samples) -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm)

PAGE 147

147 Figure 3 45 Red La Soda Brown Rot Correlation coefficients between reflectance at each wavelength and Brown rot concentration. Figure 3 46 Combined Brown Rot Correlation coefficients between reflectanc e at each wavelength and Brown rot concentration. -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm) -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r ) Wavelength (nm)

PAGE 148

148 Figure 3 47 Fabula Growth Cracking vs. Undamaged Day of Harvest Figure 3 48 Red La Soda Growth Cracking vs. Undamaged Day of Harvest 0 10 20 30 40 50 60 70 80 360 400 440 480 520 560 600 640 680 720 760 800 % Reflectance Wavelength (nm) Average Growth Cracking (14 samples) Average Undamaged (79 samples) 0 10 20 30 40 50 60 70 80 360 400 440 480 520 560 600 640 680 720 760 800 % Reflectance Wavelength (nm) Average Growth Cracking (14 samples) Average Undamaged Red (77 samples)

PAGE 149

149 Figure 3 49 Fabula Growth Cracking Correlation coefficient s between reflectance at each wavelength and Growth Cracking concentration. Figure 3 50 Red La Soda Growth Cracking Correlation coefficients between reflectance at each wavelength and Growth Cracking concentration. -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm) -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm)

PAGE 150

150 Figure 3 51 Combined Growth Cracki ng Correlation coefficients between reflectance at each wavelength and Growth Cracking concentration. Figure 3 52. Red La Soda Insect Damag e vs. Undamaged Day of Harvest -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r ) Wavelength (nm) 0 10 20 30 40 50 60 70 80 360 400 440 480 520 560 600 640 680 720 760 800 % Reflectance Wavelength (nm) Average Insect Damage (5 samples) Average Undamaged Red (77 samples)

PAGE 151

151 Figure 3 53 Red La Soda Insect Damage Correlation coefficients between reflec tance at each wavelength and Growth Cracking concentration. Table 3 2 Factor r esults for f ab ula PLS analysis of sample sets Disorder Factors Dep. variables Root m ean SEC (%) SEP (%) RMSD (%) RPD R 2 Current (%) Total (%) PRESS Calib Valid Calib. V alid Valid Internal b ruising 15 1.22 98.82 1.672 1.190 0.991 0.994 3.047 0.993 I.B. 1st d er. 5 1.18 99.68 0.782 0.835 0.706 0.698 4.277 0.996 Shatter 4 1.25 98.58 0.138 1.517 1.239 1.235 2.717 0.995 Sh. 1st d er. 2 0.66 99.56 0.099 0.852 0.696 0.694 4 .838 0.998 Greening 7 0.02 100.00 0.029 0.022 0.017 0.017 101.4 1.000 Gr. 1st d er. 2 1.36 99.96 0.111 0.352 0.285 0.281 6.051 1.000 Sunscald 8 0.47 96.05 0.302 1.422 1.154 1.150 2.495 0.984 Sun. 1st d er. 4 1.31 99.37 0.318 0.743 0.598 0.601 4.815 0.996 Brown r ot 8 0.14 99.04 0.136 1.746 1.340 1.386 3.123 0.991 B.R. 1st d er. 3 0.93 99.72 0.166 0.910 0.745 0.741 5.617 0.998 Growth c racking 9 0.50 97.25 0.300 3.102 2.086 2.507 2.373 0.978 G.C. 1st d er. 4 0.51 99.68 0.262 0.931 0.784 0.752 4.737 0.997 -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 Correlation coefficient (r) Wavelength (nm)

PAGE 152

152 Table 3 3 Factor r esults for red la s oda PLS analysis of sample sets Disorder Factors Dep. variables Root m ean SEC (%) SEP (%) RMSD (%) RPD R 2 Current (%) Total (%) PRESS Calib. Valid. Calib. Valid. Valid. Internal b ruising 11 1.346 98.91 0.712 1.411 1.245 1.228 2.005 0.988 I.b. 1st d er. 4 2.681 99.09 0.938 1.144 0.960 0.996 2.600 0.993 Shatter 4 0.580 99.70 0.068 0.964 0.936 0.785 3.500 0.998 Sh. 1st d er. 2 0.328 99.89 0.089 1.269 1.261 1.034 2.601 0.996 Greening 5 0.095 99.03 0.261 2.399 1. 881 1.913 2.776 0.989 Gr. 1st d er. 2 2.282 99.97 0.089 0.965 0.771 0.770 4.516 1.000 Sunscald 9 0.717 98.65 0.246 1.630 1.400 1.393 2.137 0.986 Sun. 1st d er. 3 2.348 99.25 0.271 1.114 0.957 0.952 3.126 0.995 Brown r ot 10 0.230 98.80 0.215 1.625 1.267 1 .323 3.622 0.991 B.R. 1st d er. 4 0.401 99.71 0.208 1.004 0.822 0.818 5.583 0.998 Growth c racking 11 0.709 97.60 0.356 2.663 2.120 2.162 2.406 0.995 G.C. 1st d er. 5 0.512 99.57 0.294 1.312 1.071 1.065 3.572 0.994 Insect d amage 12 0.804 97.74 0.555 1.638 1.216 1.292 1.354 0.992 I.D. 1st d er. 3 5.326 98.32 0.422 1.247 0.988 0.983 1.666 0.998 Table 3 4 Factor r esults for combined v arieties PLS analysis of sample sets Disorder Factors Dep. variables Root Mean SEC (%) SEP (%) RMSD(%) RPD R 2 Current ( %) Total (%) PRESS Calib. Valid. Calib. Valid. Valid. Internal b ruising 15 1.922 93.65 1.128 3.420 2.847 2.857 1.378 0.966 I.B. 1st d er. 7 0.441 98.77 0.942 1.760 1.487 1.470 2.639 0.990 Shatter 5 0.205 98.78 0.126 2.070 1.678 1.686 2.802 0.995 Sh. 1st d er. 2 23.27 97.83 0.550 2.396 1.783 1.952 2.637 0.974 Greening 7 0.191 99.43 0.110 1.907 1.511 1.521 3.251 0.997 Gr. 1st d er. 3 1.989 99.78 0.107 1.263 1.003 1.007 4.898 0.998 Sunscald 15 0.402 97.65 0.337 2.500 2.028 2.020 2.048 0.977 Sun. 1st d er. 6 0.835 98.46 0.343 1.821 1.476 1.472 2.814 0.989 Brown r ot 15 0.201 99.30 0.084 2.358 1.878 1.872 3.064 0.992 B.r 1st d er. 4 0.899 99.07 0.193 2.842 2.314 2.314 4.210 0.999 Growth c racking 15 0.318 98.24 0.279 3.654 3.039 2.953 2.340 0.979 G.C. 1st d er. 6 0.765 98.87 0.325 2.400 1.949 1.939 3.649 0.990

PAGE 153

153 Figure 3 54 Damage Prediction Using PLS Fabula Internal Bruising Reflectance (left), 1 st Derivative (right) Figure 3 55 Damage Prediction Using PLS Red La Soda Internal Bruising Reflect ance (left), 1 st Derivative (right) R = 0.9928 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9959 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9878 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.993 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage

PAGE 154

154 Figure 3 56 Damage Prediction Using PLS Combined Internal Bruising Reflectance (left), 1 st Derivative (right) Figure 3 57 B Coefficients Fabula Internal Bruising Reflectance R = 0.9659 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9898 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage -1.5 -1 -0.5 0 0.5 1 1.5 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 155

155 Figure 3 58 B Coefficients Fabula Internal Bruising 1 st Derivative Reflectance Figure 3 59 B Coefficients Red La Soda Internal Bruising Reflectance -0.06 -0.04 -0.02 0 0.02 0.04 0.06 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) -1.5 -1 -0.5 0 0.5 1 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 156

156 Figure 3 60 B Coefficients Red La Soda Internal Bruising 1 st Derivative Reflectance Figure 3 61 Damage Prediction Using PLS Fa bula Shatter Reflectance (left), 1 st Derivative (right) -0.04 -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 0.05 0.06 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) R = 0.9947 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9984 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage

PAGE 157

157 Figure 3 62 Damage Prediction Using PLS Red La Soda Shatter Reflectance (left), 1 st Derivative (right) Figure 3 63 Damage Prediction Using PLS Combined Shatter Reflectance (left), 1 st Der ivative (right) R = 0.9979 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9962 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9953 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9744 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage

PAGE 158

158 Figure 3 64 B Coefficients Fabula Shatter Reflectance Figure 3 65 B Coefficients Fabula Shatter 1 st Derivative Reflectance -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 0.025 0.03 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) -0.01 -0.008 -0.006 -0.004 -0.002 0 0.002 0.004 0.006 0.008 0.01 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 159

159 Figure 3 66 B Coefficients Red La Soda Shatter Reflectance Figure 3 67 B Coefficients Red La Soda Sha tter 1 st Derivative Reflectance -0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 0.025 0.03 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) -0.008 -0.006 -0.004 -0.002 0 0.002 0.004 0.006 0.008 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 160

160 Figure 3 68 Damage Prediction Using PLS Fabula Greening Reflectance (left), 1 st Derivative (right) Figure 3 69 Damage Prediction Using PLS Red La Soda Greening Reflectance (left), 1 st Derivative (right) R = 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9891 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9999 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage

PAGE 161

161 F igure 3 70 Damage Prediction Using PLS Combined Greening Reflectance (left), 1 st Derivative (right) Figure 3 71 B Coefficients Fabula Greening Reflectance R = 0.9965 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9977 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 0.12 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 162

162 Figure 3 72 B Coefficients Fabula Greening 1 st Derivative Reflectance Figure 3 73 B Coeffi cients Red La Soda Greening Reflectance -0.008 -0.006 -0.004 -0.002 0 0.002 0.004 0.006 0.008 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) -0.02 -0.015 -0.01 -0.005 0 0.005 0.01 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 163

163 Figure 3 74 B Coefficients Red La Soda Greening 1 st Derivative Reflectance Figure 3 75 Damage Prediction Using PLS Fabula Sunscald Reflectance (left), 1 st Derivative (right) -0.008 -0.006 -0.004 -0.002 0 0.002 0.004 0.006 0.008 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) R = 0.9843 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9961 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage

PAGE 164

164 Figure 3 76 Damage Predi ction Using PLS Red La Soda Sunscald Reflectance (left), 1 st Derivative (right) Figure 3 77 Damage Prediction Using PLS Combined Sunscald Reflectance (left), 1 st Derivative (right) R = 0.9859 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9949 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9769 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9886 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage

PAGE 165

165 Figure 3 78 B Coefficients Fabula Sunscald Reflectance Figure 3 79 B Coefficients Fabula Sunscald 1 st Derivative Reflectance -0.3 -0.2 -0.1 0 0.1 0.2 0.3 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) -0.03 -0.02 -0.01 0 0.01 0.02 0.03 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 166

166 Figure 3 80 B Coefficients Red La Soda Sunscald Reflectance Figure 3 81 B Coefficients Red La Soda Sunscald 1 st Derivative Reflectance -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) -0.025 -0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 0.025 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 167

167 Figure 3 82 Damage Prediction Using PLS F abula Brown Rot Reflectance (left), 1 st Derivative (right) Figure 3 83 Damage Prediction Using PLS Red La Soda Brown Rot Reflectance (left), 1 st Derivative (right) R = 0.9912 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9977 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9931 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9974 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage

PAGE 168

168 Figure 3 84 Damage Prediction Using PLS Combined Brown Rot Reflectance (left) 1 st Derivative (right) Figure 3 85 B Coefficients Fabula Brown Rot Reflectance R = 0.9922 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.988 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage -0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 169

169 Figure 3 86 B Coefficients Fabula Brown Rot 1 st Derivative Reflectance Figure 3 87 B Coefficients Red La Soda Brown Rot Reflectance -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 170

170 Figure 3 88 B Coefficients R ed La Soda Brown Rot 1 st Derivative Reflectance Figure 3 89 Damage Prediction Using PLS Fabula Growth Cracking Reflectance (left), 1 st Derivative (right) -0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 0.025 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) R = 0.9779 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.997 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage

PAGE 171

171 Figure 3 90 Damage Prediction Using PLS Red La Soda Growth Cracking Reflectance (left), 1 st Derivative (right) Figure 3 91 Damage Prediction Using PLS Combined Growth Cracking Reflectance (left), 1 st Derivative (right) R = 0.9947 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9944 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9785 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9899 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage

PAGE 172

172 Figure 3 92 B Coefficients Fabula Growth Cracking Reflectance Figure 3 93 B Coefficients Fabula Growth Crackin g 1 st Derivative Reflectance -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) -0.025 -0.02 -0.015 -0.01 -0.005 0 0.005 0.01 0.015 0.02 0.025 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 173

173 Figure 3 94 B Coefficients Red La Soda Growth Cracking Reflectance Figure 3 95 B Coefficients Red La Soda Growth Cracking 1 st Derivative Reflectance -1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) -0.03 -0.02 -0.01 0 0.01 0.02 0.03 0.04 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 174

174 Figure 3 96 Damage Prediction Using PLS Red La Soda Insect Dama ge Reflectance (left), 1 st Derivative (right) Figure 3 97 B Coefficients Red La Soda Insect Damage Reflectance R = 0.9915 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9976 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage -1.5 -1 -0.5 0 0.5 1 1.5 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm)

PAGE 175

17 5 Figure 3 98 B Coefficients Red La Soda Insect Damage 1 st Derivative Reflectance Figure 3 99. Damage Prediction Using SMLR Fabula Internal Bruising Reflectance (left), 1 st Derivative (right) -0.03 -0.02 -0.01 0 0.01 0.02 0.03 360 410 460 510 560 610 660 710 760 B Coefficient Wavelength (nm) R = 0.9905 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.9972 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 176

176 Figure 3 100. Damage Prediction Using SMLR Red La Soda Internal Bruising Reflectance (left), 1 st Derivative (right) Figure 3 101. Damage Prediction Using SMLR Combined Internal Bruisin g Reflectance (left), 1 st Derivative (right) R = 0.9968 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9997 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predicted % Damage Actual % Damage R = 0.7386 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9977 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 177

177 Figure 3 102. Damage Prediction Using SMLR Fabula Shatter Reflectance (left), 1 st Derivative (right) Figure 3 103. Damage Prediction Using SMLR Red La Soda Shatter Reflectance (left), 1 st Derivative (right) R = 0.9999 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9999 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9998 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 178

178 Figure 3 104. Damage Prediction Using SMLR Combined Shatter Reflectance (left), 1 st Derivative (right) Figure 3 105. Damage Prediction Using SMLR Fabula Greening Reflectance (left), 1 st Derivative (right) R = 0.9995 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9992 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9977 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 179

179 Figure 3 106. Damage Predict ion Using SMLR Red La Soda Greening Reflectance (left), 1 st Derivative (right) Figure 3 107. Damage Prediction Using SMLR Combined Greening Reflectance (left), 1 st Derivative (right) R = 0.9989 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 1 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9974 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9999 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 180

180 Figure 3 108. Damage Prediction Using SMLR Fabula Sunscald Re flectance (left), 1 st Derivative (right) Figure 3 109. Damage Prediction Using SMLR Red La Soda Sunscald Reflectance (left), 1 st Derivative (right) R = 0.9958 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9991 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9988 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9997 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 181

181 Figure 3 110. Damage Prediction Using SMLR Combined Sunscald Reflectance (left), 1 st Derivative (r ight) Figure 3 111. Damage Prediction Using SMLR Fabula Brown Rot Reflectance (left), 1 st Derivative (right) R = 0.9977 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9976 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.999 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9999 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 182

182 Figure 3 112. Damage Prediction Using SMLR Red La Soda Brown Rot Reflectance (left), 1 st Derivative (right) Figure 3 113. Damage Pr ediction Using SMLR Combined Brown Rot Reflectance (left), 1 st Derivative (right) R = 0.9967 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9999 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9991 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9994 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 183

183 Figure 3 114. Damage Prediction Using SMLR Fabula Cracking Reflectance (left), 1 st Derivative (right) Figure 3 115. Damage Prediction Using SMLR Red La Soda Crack ing Reflectance (left), 1 st Derivative (right) R = 0.9989 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9995 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9815 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9993 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 184

184 Figure 3 116. Damage Prediction Using SMLR Combined Cracking Reflectance (left), 1 st Derivative (right) Figure 3 117. Damage Prediction Using SMLR Red La Soda Insect Damage Reflectance (left), 1 st De rivative (right) R = 0.9981 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9996 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9996 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage R = 0.9997 0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0 0.2 0.4 0.6 0.8 1 Predict % Damage Actual % Damage

PAGE 185

185 Table 3 5 Results of f abula SMLR analysis of sample sets Disorder Step Variable e ntered Partial R s quare Model R s quare F v alue Pr > F Internal b ruising 1 w795 0.534 0.534 13.7 0.003 2 w791 0.296 0.830 19.2 0.0011 3 w435 0.090 0.920 11.3 0.0072 I.b 1st d er. 1 w408 0.675 0.675 24.9 0.0003 2 w742 0.186 0.861 14.8 0.0027 3 w531 0.095 0.956 21.5 0.0009 4 w456 0.027 0.983 14.9 0.0039 5 w482 0.012 0.995 21.0 0.0018 6 w435 0.003 0.999 15.2 0.0059 Shatter 1 w432 0.030 0.951 11.3 0.0033 2 w438 0.029 0.980 25.6 <.0001 3 w383 0.008 0.988 11.2 0.0038 4 w379 0.007 0.992 15.4 0.0011 5 w382 0.004 0.996 15.9 0.0011 6 w386 0.002 0.998 13.9 0.002 7 w376 0.001 0.999 15.7 0.0014 8 w471 0.001 1.000 15.6 0.0017 Shatter 1st d er. 1 w499 0.942 0.942 324.0 <.0001 2 w500 0.036 0.978 31.0 <.0001 3 w629 0.009 0.987 12.2 0.0026 4 w504 0.005 0.992 10.0 0.0056 5 w688 0.004 0.996 14.0 0.0018 6 w613 0.002 0.997 10.7 0.0051 7 w722 0.001 0.998 7.7 0.0149 8 w790 0.001 0.999 8 .2 0.0134 9 w690 0.001 1.000 12.2 0.0045 10 w648 0.000 1.000 54.5 <.0001 11 w654 0.000 1.000 9.3 0.0121 Greening 1 w702 0.986 0.986 775.6 <.0001 2 w503 0.007 0.993 10.6 0.0087 Greening 1st d er. 1 w688 0.990 0.990 1056.0 <.0001 2 w510 0.010 0.99 9 143.6 <.0001 3 w533 0.001 1.000 67.4 <.0001 4 w625 0.000 1.000 18.1 0.0028 Sunscald 1 w446 0.954 0.954 247.4 <.0001 2 w432 0.025 0.979 13.3 0.0039 Sunscald 1st d er. 1 w512 0.928 0.928 154.2 <.0001 2 w610 0.056 0.984 38.1 <.0001 3 w474 0.008 0 .992 9.1 0.0129 4 w737 0.006 0.998 23.2 0.0009

PAGE 186

186 Table 3 5 Continued Disorder Step Variable e ntered Partial R s quare Model R s quare F v alue Pr > F 5 w449 0.002 0.999 16.8 0.0034 6 w417 0.000 1.000 5.8 0.0467 7 w711 0.000 1.000 30.3 0.0015 8 w611 0.000 1.000 30.7 0.0026 Brown r ot 1 w689 0.994 0.994 1950.4 <.0001 2 w688 0.003 0.997 8.3 0.0163 Brown rot 1st d er. 1 w518 0.976 0.976 453.6 <.0001 2 w792 0.014 0.990 13.8 0.004 3 w625 0.006 0.996 15.1 0.0037 4 w421 0.002 0.999 13.8 0.006 5 w650 0.001 1.000 23.8 0.0018 6 w543 0.000 1.000 63.3 0.0002 Growth c racking 1 w545 0.949 0.949 221.0 <.0001 2 w516 0.034 0.983 21.9 0.0007 3 w517 0.007 0.990 7.2 0.0232 4 w547 0.005 0.995 8.5 0.017 5 w477 0.003 0.997 7.3 0.0273 6 w543 0.0 02 0.999 8.5 0.0226 7 w486 0.001 1.000 8.7 0.0256 Growth c racking 1st d er. 1 w510 0.965 0.965 330.9 <.0001 2 w769 0.020 0.985 15.2 0.0025 3 w528 0.010 0.995 21.3 0.001 4 w453 0.003 0.998 15.2 0.0036 5 w529 0.001 0.999 6.7 0.032 6 w473 0.001 1 .000 8.6 0.0221

PAGE 187

187 Table 3 6 Results of r ed la s oda SMLR analysis of sample sets Disorder Step Variable e ntered Partial R s quare Model R s quare F v alue Pr > F Internal b ruising 1 w774 0.137 0.904 17.2 0.001 2 w776 0.030 0.935 5.1 0.045 3 w763 0.0 34 0.944 6.6 0.026 4 w788 0.022 0.965 6.3 0.031 5 w784 0.017 0.982 8.7 0.016 I.b. 1st d er. 1 w640 0.716 0.716 32.7 <.0001 2 w412 0.151 0.867 13.7 0.003 3 w705 0.072 0.939 13.0 0.004 4 w648 0.033 0.972 11.7 0.007 5 w650 0.021 0.993 28.8 0.001 6 w786 0.006 1.000 111.7 <.0001 7 w681 0.000 1.000 14.8 0.006 8 w484 0.000 1.000 13.3 0.011 Shatter 1 w568 0.996 0.996 4203.3 <.0001 2 w398 0.001 1.000 38.6 <.0001 3 w588 0.000 1.000 15.4 0.001 4 w403 0.000 1.000 6.5 0.022 5 w514 0.000 1.000 11.7 0.004 Shatter 1st d er. 1 w502 0.985 0.985 1250.1 <.0001 2 w625 0.010 0.995 37.6 <.0001 3 w788 0.003 0.998 27.8 <.0001 4 w576 0.001 0.999 11.0 0.004 5 w585 0.000 0.999 9.3 0.008 6 w471 0.000 1.000 9.5 0.008 7 w642 0.000 1.000 8.9 0.011 8 w621 0.000 1.000 11.8 0.005 9 w479 0.000 1.000 12.5 0.005 Greening 1 w680 0.988 0.988 873.1 <.0001 2 w675 0.010 0.998 46.8 <.0001 Greening 1st d er. 1 w650 0.991 0.991 1173.4 <.0001 2 w391 0.008 0.999 58.0 <.0001 3 w673 0.001 1.000 24.9 0.001 4 w730 0.000 1.000 31.8 0.001 Sunscald 1 w435 0.952 0.952 199.9 <.0001 2 w433 0.033 0.985 20.0 0.002 Sunscald 1st d er. 1 w750 0.034 0.990 31.0 0.000 2 w608 0.005 0.995 8.0 0.022 3 w425 0.003 0.998 10.7 0.014 4 w530 0.002 0.999 16.2 0.005 5 w444 0.001 1.000 37.9 0.001

PAGE 188

188 Table 3 6 Continued Disorder Step Variable e ntered Partial R s quare Model R s quare F v alue Pr > F 6 w675 0.000 1.000 12.6 0.017 Brown r ot 1 w665 0.974 0.974 477.5 <.0001 2 w607 0.009 0.982 6.1 0.030 Brown rot 1st d er. 1 w631 0.982 0.982 717.9 <.0001 2 w781 0.011 0.993 17.8 0.001 3 w776 0.004 0.997 12.0 0.005 4 w629 0.002 0.998 11.3 0.007 5 w442 0.001 0.999 6.6 0.030 6 w615 0.001 1.000 8.9 0.018 Growth c racking 1 w381 0.879 0.879 87.4 <.0001 Growth cracking 1st d er. 1 w622 0.955 0.955 254.1 <.0001 2 w433 0.024 0.979 12.1 0.005 3 w625 0.016 0.995 29.1 0.000 4 w633 0.003 0.998 13.7 0.005 Insect d amage 1 w800 0.889 0.889 24.1 0.016 Insect damage 1st d er. 1 w549 0.981 0.981 150.8 0.001 2 w586 0.020 1.0 00 3573.2 0.000

PAGE 189

189 Table 3 7 Results of c ombined varieties SMLR analysis of sample sets Disorder Step Variable e ntered Partial R s quare Model R s quare F v alue Pr > F Internal b ruising No variable met the 0.0500 significance level for entry into the mod el. I. b. 1st d er. 1 w682 0.359 0.359 15.1 0.001 2 w667 0.180 0.539 10.1 0.004 3 w716 0.247 0.786 28.9 <.0001 4 w427 0.044 0.831 6.3 0.020 5 w589 0.044 0.874 7.9 0.010 6 w558 0.029 0.903 6.6 0.018 7 w791 0.029 0.932 8.8 0.007 8 w576 0.018 0.9 49 6.9 0.016 9 w540 0.014 0.963 7.2 0.015 10 w443 0.008 0.971 4.9 0.040 Shatter 1 w378 0.855 0.855 242.1 <.0001 2 w362 0.021 0.978 36.9 <.0001 3 w388 0.006 0.984 15.0 0.000 4 w395 0.003 0.990 9.4 0.004 5 w422 0.005 0.995 32.1 <.0001 Shatter 1 st d er. 1 w672 0.803 0.803 167.0 <.0001 2 w388 0.098 0.900 39.2 <.0001 3 w390 0.034 0.934 19.8 <.0001 4 w437 0.030 0.964 32.1 <.0001 5 w582 0.014 0.978 22.6 <.0001 6 w365 0.007 0.984 14.9 0.000 7 w366 0.004 0.988 10.6 0.003 8 w451 0.006 0.994 31.6 <.0001 9 w387 0.002 0.995 11.0 0.002 10 w703 0.001 0.996 9.3 0.005 11 w570 0.001 0.997 11.4 0.002 12 w541 0.001 0.998 9.0 0.006 13 w649 0.000 0.998 8.0 0.009 14 w733 0.000 0.999 9.3 0.005 Greening 1 w681 0.942 0.942 392.0 <.0001 2 w672 0.047 0.989 98.4 <.0001 3 w517 0.004 0.993 10.7 0.003 Greening 1st d er. 1 w697 0.964 0.964 640.9 <.0001 2 w758 0.020 0.984 28.4 <.0001 3 w578 0.011 0.995 50.2 <.0001 4 w763 0.002 0.997 17.1 0.001 5 w363 0.000 0.999 12.2 0.002 6 w711 0.000 1.0 00 8.0 0.011 7 w738 0.000 1.000 9.3 0.007

PAGE 190

190 Table 3 7. Continued Disorder Step Variable e ntered Partial R s quare Model R s quare F v alue Pr > F 8 w405 0.000 1.000 9.8 0.007 9 w747 0.000 1.000 5.1 0.040 10 w562 0.000 1.000 8.3 0.012 Sunscald 1 w786 0.083 0.918 22.2 0.000 2 w773 0.054 0.972 40.1 <.0001 3 w589 0.005 0.985 7.2 0.014 4 w588 0.003 0.992 7.3 0.014 5 w629 0.002 0.994 7.0 0.016 6 w787 0.002 0.996 9.5 0.006 7 w758 0.001 0.997 6.3 0.023 Sunscald 1st d er. 1 w729 0.730 0.730 65.0 <.0001 2 w430 0.171 0.902 40.0 <.0001 3 w483 0.062 0.964 37.9 <.0001 4 w433 0.021 0.985 27.9 <.0001 5 w428 0.007 0.991 16.2 0.001 6 w553 0.002 0.993 4.9 0.039 Brown r ot 1 w708 0.976 0.976 1044.5 <.0001 2 w487 0.007 0.982 9.2 0.006 3 w71 2 0.011 0.993 37.9 <.0001 Brown r ot 1 st d er. 1 w567 0.810 0.810 110.6 <.0001 2 w579 0.106 0.915 31.1 <.0001 3 w433 0.039 0.954 20.4 0.000 4 w662 0.022 0.976 21.2 0.000 5 w717 0.008 0.984 10.1 0.004 6 w759 0.004 0.988 6.8 0.017 7 w405 0.005 0.9 92 11.9 0.003 8 w439 0.003 0.996 13.6 0.002 9 w737 0.001 0.998 11.3 0.004 10 w671 0.001 0.999 13.4 0.002 11 w434 0.001 1.000 13.6 0.002 12 w718 0.000 1.000 10.4 0.006 13 w390 0.000 1.000 12.2 0.004 14 w482 0.000 1.000 5.8 0.032 15 w689 0.00 0 1.000 9.7 0.009 Growth c racking 1 w670 0.041 0.857 7.4 0.011 2 w461 0.106 0.963 72.4 <.0001 3 w673 0.016 0.980 19.0 0.000 4 w400 0.007 0.986 10.6 0.004 5 w401 0.003 0.989 5.4 0.030 6 w669 0.003 0.991 6.4 0.019

PAGE 191

191 Table 3 7. Continued Dis order Step Variable e ntered Partial R s quare Model R s quare F v alue Pr > F Growth c racking 1st d er. 1 w475 0.471 0.471 23.2 <.0001 2 w460 0.270 0.741 26.0 <.0001 3 w651 0.130 0.871 24.2 <.0001 4 w520 0.055 0.926 17.2 0.000 5 w440 0.034 0.961 19.1 0.000 6 w603 0.015 0.976 13.1 0.002 7 w579 0.006 0.982 7.2 0.015 8 w459 0.006 0.988 8.5 0.009 9 w427 0.003 0.991 5.6 0.029 10 w770 0.002 0.993 5.6 0.030 11 w685 0.002 0.995 5.1 0.038 12 w638 0.002 0.996 7.1 0.018 13 w387 0.001 0.999 8.5 0.0 12 14 w512 0.001 0.999 12.0 0.005 15 w381 0.000 1.000 21.1 0.001 16 w479 0.000 1.000 6.5 0.029 17 w556 0.000 1.000 5.6 0.043 18 w800 0.000 1.000 14.0 0.005

PAGE 192

192 Table 3 election of sample sets Disorder Correlation coef. B c oefficient SMLR Internal Bruising 622 701 nm 365, 371, 373, 374, 384, 390, 394, 400, 425, 464, 692, 709, 714, 732, 735, 768, 787, 795, 799 nm 435, 791, 795 nm I.B. 1st Der. 530, 605, 650, 666, 730, 790 nm 408,435, 456, 531, 742 nm Shatter 360 420, 493 640 nm 360 400, 445 455, 475 486, 500 516 nm 376, 379, 382, 383, 386, 432, 438, 471 nm Sh. 1st Der. 360 371, 380 425, 436 448, 455 465, 470 480 482 510 nm 499, 500, 504, 613, 629,648, 654, 688, 690, 722, 790 nm Greening 380 710 nm 360, 371, 379, 384, 760, 762, 767, 774, 800 nm 503, 720 nm Gr. 1st Der. 440, 463, 478, 506, 530, 670, 722, 748, 758, 767, 768 nm 510, 533, 625, 688 nm Sunscald 490 700 nm 362 365, 371, 397, 777, 780, 788 798 nm 432, 446 nm Sun. 1st Der. 599, 600, 606, 610, 682, 711, 718, 722, 726, 748, 780, 787, 788, 797, 798 nm 417, 449, 474, 512, 610, 611, 711, 737 nm Brown Rot 360 800 nm 365, 366, 370 375 nm 688, 689 nm B.R. 1st Der. 440, 470, 533, 566, 576, 589, 594, 607, 608, 646, 676, 677, 725, 787 nm 421, 518, 5 43, 625, 650, 792nm Growth Cracking 360 800 nm 361, 363, 366, 371, 373, 382, 390, 394, 787, 790, 795 nm 477, 486, 516, 517, 543, 545, 547 nm G.C. 1st Der. 473, 438, 536, 602, 631, 643, 645, 677, 711, 722, 729, 733, 756, 787, 790 nm 473, 510, 528, 529 769 nm

PAGE 193

193 Table 3 oda avelength s election of sample sets Disorder Correlation coef. B c oefficient SMLR Internal Bruising 600 800 nm 363, 365, 370, 371, 380, 390, 397, 714, 766, 770, 782, 787, 790 nm 763, 774, 776 784 78 8 nm I.B. 1st Der. 462, 465, 566, 588, 683, 687, 782 nm 412, 484, 640, 648, 650, 681, 705, 786 nm Shatter 360 660 nm 360 384, 423 462, 514 579nm 398, 403, 514, 568, 588 nm Sh. 1st Der. 360 371, 397 405, 427 436, 470 480, 488 520 nm 471, 479, 502, 576, 585, 621, 625, 642, 788 nm Greening 510 570 nm 618 698 nm 675, 680 nm Gr. 1st Der. 461, 463, 475, 497, 530, 680 715, 725, 735, 744, 750, 797 nm 391, 650, 673, 730 nm Sunscald 360 480, 620 710nm 360, 363, 366, 644, 650, 683, 686, 799, 800 nm 433, 435 nm Sun. 1 st Der. 512, 527, 540, 687, 711, 725 nm 425, 444, 530, 608, 675, 750 nm Brown Rot 360 800 nm 361, 378, 382, 384, 416, 787, 800 nm 607, 665 nm B.R. 1st Der. 507, 533, 657, 670, 673, 688, 709, 713, 726, 736, 774, 777, 790, 791, 799 nm 442, 615, 629, 63 1, 776, 781 nm Growth Cracking 360 460, 620 800nm 402, 458, 516, 533, 566, 662, 685, 713, 722, 746, 750, 754, 777, 790, 791 nm 381 nm G.C. 1st Der. 433, 622, 625, 633 nm 433, 622, 625, 633 nm Insect Damage 360 480, 640 800nm 363, 371, 375, 376, 380, 4 65, 507, 527, 750, 777, 783, 787, 792, 797, 800 nm 800 nm I. D. 1st Der. 420, 436, 502, 512, 519, 568, 574, 580, 657, 670, 690, 700, 738, 740, 770, 787, 793 nm 549, 586 nm

PAGE 194

194 Table 3 10. Combined wavelength s election of sample sets Disorder Correlatio n coef. B c oefficient SMLR Internal Bruising 600 800 nm 370, 373, 378, 383, 392, 710, 747, 752, 762, 782, 790, 793 nm No wavelengths selected I.B. 1st Der. 400, 420, 660, 676, 709, 728, 787, 793 nm 427, 442, 443, 540, 558, 576, 589, 667, 682, 716, 791, nm Shatter 360 800 nm 360 397, 440 460, 473 484 nm 378, 362, 388, 395, 422 nm Sh. 1st Der. 449 466, 553, 713, 722, 752, 768 nm 365, 366, 387, 388, 390, 437, 451 541 570, 582, 649, 672, 703 733 nm Greening 400 480, 600 700nm 670 683 nm 517, 672, 681 nm Gr. 1st Der. 397, 406, 442, 473, 725, 748, 798 nm 363, 405, 562, 697, 711, 738, 747, 758, 763 nm Sunscald 360 480, 620 710nm 722, 746, 764, 766, 770, 780, 788, 792, 800 nm 588, 589, 629, 758, 773, 786, 787 nm Sun. 1st Der. 680, 708, 713, 720 750, 774, 777, 788, 790, 797 nm 428, 430, 433, 483, 553 729 nm Brown Rot 360 600 nm 370, 372, 373, 383, 392, 747, 762, 782, 790, 793 nm 487, 708, 712 nm B.R. 1st Der. 423, 530, 660, 683, 790, 797 nm 390, 405, 433, 434, 439, 482, 567, 579, 662, 671, 689, 717, 718, 737, 759 nm Growth Cracking 360 460, 620 800nm 378, 389, 398, 404, 704, 722, 734, 738, 752, 778, 782, 787, 789, 797, 800 nm 400, 401, 461, 669, 670, 673 nm G.C. 1st Der. 371, 413, 442, 475, 540, 579, 665, 686, 730, 753, 787, 789, 787 nm 381, 387, 427, 440, 459, 460, 475, 479, 512, 520, 556, 579, 603, 638, 651, 685, 770, 800 nm

PAGE 195

195 CHAPTER 4 CONCLUSIONS Four types of damage were observed in the impact drop tests of the two varieties ; skinning, black spot external and internal shatter The major type of external mechanical damage in the harvesting and packaging operations for Fabula was skinning; while the major type of damage for Red La Soda was external shatter. Fabula and Red La Soda before vin e kill showed average percentages of potatoes detected with skinning damage at 77.8% and 5% with Red La Soda maintaining a fairly consistent low skinning rate and severity throughout the study. Fabula on the other hand started out with a very high ski nning rate which dropped signif icantly for harvest periods following vine kill with severity of skinning dropping dramatically as a result of only 7 days of wait The highest USDA skinning S gh skinning resistance throughout packaging line skinning, while Fabula 7 days before vine kill which significantly lowered skinning which oc curred. During testing, i ncidences and severity of mechanical damage generally increased with drop height for both varieties, with blackspot bruising occurrence switching to external shatter at of data indicated that low Impacts of greater he ight resulted in more severe damage for all types of damage, but did not always result in larger groups of tubers developing bruising or shatter damage. Both varieties showed resistance to the development of bruising and shatter the 7 days before vine kill

PAGE 196

196 was performed, but showed their greatest incidence and severity the day before vine kill was damage after vine kill and with greater severity; which lowered as the tim e after vine kill increased. The significant drop in mechanical damage which occurred with just 7 days of skin set after vine killing would allow farmers to harvest earlier than the standard 2 3 weeks if desired; harvesting earlier than the standard time w ould result in a small, but noticeable, increase in damage. But if wea ther conditions or insect presence could result in tuber plants left in the soil being exposed to circumstances that would threaten tuber health ; it is advised that potato tubers be harv ested after a shorter period of skin set A wide range of s pectroscopic techniques have been widely used for analysis of food products Spectral reflectance in the range of 360 800 nm were collected for potato samples of and then analyzed using the correlation coefficient spectrum, PLS and SMLR in order to find important wavelengths to establish calibration models for predicting the potato tuber disorders; internal bruising, shatter, greening, sunscald, brown rot, g rowth cracking and insect damage. The range used for disorder evaluation (360 800 nm) was chosen after determining that this range had a lower dependence of moisture content after analyzing data in the range between 200 800 nm) which had high correlation coefficients above 0.4. Analysis of the data using PLS and SMLR, endorsed the reduction of the wavelength ranges necessary for prediction of damage to 360 48 0 nm and 600 800 nm. This would mean that spectral wavelength scans could be concentrated on these areas in order to calibrate models for prediction of damage faster on a packaging line. Prediction models created with PLS allowed for fairly accurate prediction of damaged tubers with SEP below 3%

PAGE 197

197 varieties. Each disorder, except for growth cracking and insect damage, had a unique cause of reflectance cha nges associated with damage that occurred. Internal bruising was associated with the production of melanin which effected wavelengths around 475 nm for both varieties. Shatter damage resulted in an increase of exposure to potato tuber starches below the su rface of the skin which was missing. Greening reflectance was changed by the increased presence of chlorophyll, solanine and other photosynthesis chemicals; which had significant effects on the photosynthetically active range of 450 nm to 670 nm. Sunscald was associated with the radiative burning of tuber flesh which caused chemical changes in the surface and subsurface. Brown rot was influenced by the bacterial rot that caused brown grey discoloration and the development of creamy pus. Growth cracking and insect damage reflectance changes were both the result of H 2 O 2 development that resulted in tuber suberization at the point of damage. The study demonstrated the potential for predicting damage and disorders in potat oes using VIS/NIR spectral data, but res ults of prediction models would need to be improved with additional data points in order to reduce the error which occurred when predicting damage levels on healthy undamaged samples The experiment showed that the selected wavelength ranges in the VIS/NIR range could be used for successful prediction of damage and disorders in potato samples and used to develop rapid and objective damage prediction systems for fresh market Fabula and Red La Soda potatoes.

PAGE 198

198 REFERENCE LIST ASAE. 1994. ASAE Standards, 41 st. Edition. St. Joseph, MI: ASAE ASTM Standards 1997. E131 95: Standard terminology relating to molecular spectroscopy. In Annual Book of ASTM Standards. West Conshohocken, PA: ASTM. Bajema, R.W., and G. M. Hyde 1998. Instrumented pendulum for impact characterization of whole fruit and vegetable specimens, Transactions of the ASAE 41(5): 1399 1405. Blahovec, J. 2005. Impact induced mechanical damage of Agria potato tubers. Transactions of the ASAE 51 (2): 39 43. Blahovec, J. 2006. Shape of bruise sp ots in impacted potatoes. Postharvest Biology and Technology 39 (3) : 278 284. Bohl, W.H., and S.B. Johnson. 2010. Commercial Potato Production in North America USDA Handbook No. 267. Washington, D.C.: USDA. Brook R.C., R. J. Fick, and T. D. Forbush. 199 5. Potat o Storage Design and Management American Journal of Potato Research 72(8): 463 480. Brown, G.K., C.L. Burton, S.A. Sargent, N.L. Schulte Pason, E.J. Timm, and D.E. Marshall. 1989. Assessment of apple damage on packing line, Transactions of the AS AE 5(4): 475 484. Bull, C.R. 1991. Wavelength selection for near infrared reflectance moisture meters. Journal of Agricultural Engineering Research 49: 113 125. Calverley D.J.B. 1998. Storage and Processing of Roots and Tubers in the Tropics Rome, Ita ly: FAO. Carter, G. A. 1991. Primary and secondary effects of the water content on the spectral reflectance of leaves. American J. of Botany 78(7): 916 924. Chiputula, J. 2009. Evaluating Mechanical Damage of Fresh Potato during Harvesting and Postharves t Handling. MS Thesis. Gainesville, Florida: University of Florida, Department of Agricultural and Biological Engineering. by transmission NIR spectroscopy. Post harvest Biology and Technology 28: 87 96. Crowe, I., and J. McNichol. 1985. The use of principal components in the analysis of near infrared spectra. Applied spectroscopy 39(4): 275 288. Dahm, D. J. and K. D. Dahm. 2001. The Physics of Near Infrared Sca ttering. In Near Infrared Technology in the Agricultural and Food Industries 1 18. St. Paul, Minnesota: American Associ ation of Cereal Chemists, Inc.

PAGE 199

199 Davenel, A., C. Guizard, T. Labarre and F. Sevila. 1988. Automatic detection of surface defects on fruit using a vision system. J. Agric. Eng. Res. 41(1): 1 9. Elbatawi, I. E., M. T. Ebaid, and B. E. Hemeda. 2008. Determination of Potato Water Content Using NIR Diffuse Reflection Method. Misr Journal of Agricultural Engineering: 25(4): 1279 1292. Esbensen, K. H. 2002. Multivariate Data Analysis in Practice. 5th ed. Woodbridge, NJ: CAMO Technologies. FAO. 2008. International Year of the Potato Food and Agriculture Organization of the United Nations Available at: http://www.potato2008.org. Accessed 24 Febr uary 2011 Fassio, A., and D. Cozzolino. 2004. Non destructive prediction of chemical composition in sunflower seeds by near infrared spectroscopy. Industrial Crops and Products 20: 3 21 329. Fluck, R. C., and E. M. Ahmed. 1973. Impact testing of fruits an d vegetables. Transactions of the ASAE 16(4): 660 666. Georgios, P., H. G. Diofantos, T. Kyriacos and T. Leonidas. 2010. Spectral vegetation indices from field spectroscopy intended for evapotranspiration purposes for spring potatoes in Cyprus. Proc. SPIE 7824: 782410.Washington: SPIE Publications. Ghazvi, M. and S. Houshmand. 2010. Effects of Mechanical Damage and Temperature on Potato Respiration Rate and Weight Loss. World Applied Sciences Journal 8 (5): 647 652 Halderson, J.L., L. C. Haderlie, and A Skrobacki. 1988 Mechanical vine killing of potatoes. American Journal of Potato Research 65(7): 415 423. Springe r New York. Hall, C.B. (1953). Prevention of Skinning of Potatoes. Proceedings of the Florida State Horticultural Society 66: 124 126. Hatc hell, D. C 1999. Technical Guide 3rd Ed Boulder, CO: Analytical Spectral Devices Inc. Heinemann, P. H., N. P. Pathare, C. T. Morrow. 1996. An automated inspection station for machine vision grading of potatoes. Machine Vision and Applications 9(1): 14 19. Helland, I. S. 2001. Some theoretical aspects of partial least square s regression. Chemometrics and Intelligent Laboratory Systems 58(2): 97 107. Hutchinson, C. M. and W. M. Stall. 2007 Potato Vine Killing or Desiccation. Gainesville, Fl.: IFAS Ext ension University of Florida. Available at: http://edis.ifas.ufl.edu/hs181. Accessed 20 March 2011

PAGE 200

200 Hutchinson, C.M. and D. Gergela 2007 Average Yearly Potato Production Statistics for Potatoes Produced at the UF/IFAS PSREU Hastings Farm. Gainesville, F l. IFAS Extension University of Florida. Available at: http://edis.ifas.ufl.edu/pdffiles/HS/HS36000.pdf Accessed 20 March 2011. Instron Impact testing. 2011. http://www.instron.us/wa/home/default_en.aspx. Accessed 13 December 2011. Jenkins, S.F., T.C. W ehner. 1983 A System for the Measurement of Foliar Diseases of Cucumbers. North Carolina State University. Cucurbit Genetics Cooperative Report 6 (5):10 12. Jin, J., J. Li, G. Liao, X. Yu, L. C. Viray. 2009. Methodology for Potatoes Defects Detection wit h Computer Vision. In International Symposium on Information Processing 346 351. Huangshan, P. R. China: Academy Publisher. Kang, S., K.J. Lee, W. Choi, L.R. Son, D.S. Choi, and G. Kim. 2004. A near infrared sensing technique for measuring the quality of potatoes. ASAE Paper No : 033137. St. Joseph, MI: ASAE. Lefebvre, M., T. Zimmerman C. Baur, P. Gugerli, T. Pun. 1995. Potato Operation: automat ic detection of potato diseases. Proc. SPIE 2345, 2. Washington: SPIE Publications. Liew, O. W., A. K. Asundi, J. W Chen, C. Yiwen, Y. ShangJuan, and Y. G. Hoon. 2001. Spectroscopy Detection of Green and Red Fluorescent Proteins in Genetically Modified Plants Using a Fiber Optics System. Proc. SPIE 4416: 276. Washington: SPIE Publications. Lingaerde, O. L., and N Christophersen. 2000. Shrinkage structure of partial least squares. Scandinavian J. Statistics 27(3): 459 473. Lu, R. and D. Ariana. 2002. A near infrared sensing technique for measuring internal quality of apple fruit. Applied Engineering in Agricultur e 18(5): 585 590. Lu, R. 2003. Imaging spectroscopy for assessing internal quality of apple fruit. ASAE Paper No: 036012. St. Joseph, MI: ASAE. Mathew, R., and G.M. Hyde. 1997. Potato impact damage thresholds. Transactions of the ASAE 40(3): 705 709. Mc Gechan, M.B. 1980. An investigation into the damage sustained by different varieties of potatoes during riddling to remove soil. Journal of Agricultural Engineering Research 25: 345 353. Mehl, P.M., K. Chao, M. Kim, and Y.R. Chen. 2002. Detection of defec ts on selected apple cultivars using hyperspectral and multispectral image analysis. Applied Engineering in Agriculture 18(2): 219 226.

PAGE 201

201 Min, M. 2005. Determination of significant wavelengths and prediction of nitrogen content for citrus. Transactions of t he ASAE 48(2): 455 461 Min, M 2004. The effect of water and variety on nitrogen sensing of citrus leaf. ASAE Paper No. 041080. St. Joseph, MI: ASAE. Muir A. Y., D. W. Ross, C. J. Dewar and D. D. Kennedy. 1999. Defect and disease detection in potato tubers Proc. SPIE 3543: 199. Washington: SPIE Publications. Murphy, H. 1968 Potato vine killing. American Journal of Potato Research 45(12): 472 478. Noordam, J.C., G.W. Otten, A.J.M. Ti mmermans, B.H. van Zwol. 20 00. High speed potato grading and quality inspection based on a color vision system. Proc. SPIE 3966 : 206. Washington: SPIE Publications. Nylund, R.E., Hemphill, P., Lutz J. M. and Sorenson H 1955. Mechanical damage to potatoes during harvesting and h andling operations in the red river valley of Minnesota and North Dakota. American Journal of Potato Research 32(7) : 237 247 Olson, S.M. and E. Simonne. 2005 2006 Chapter 35: Potato Production in Florida. In Vegetable Production Handbook for Florida : 30 5 318. Lincolnshire, Ill: Vance Publishing Owen, A.J. 1995. Uses of Derivative Spectroscopy. Waldbronn, Germany. Agilent Technologies Available at: http://www.chem.agilent.com/en us/search/library/Pages/default.aspx. Accessed 10 June 2012. Peters, R. 19 96 Damage of Potato Tubers, a review. Potato Research 39(4): 479 484 Porteous, R.L., A.Y. Muir, R.L. Wastie. 1981. The identification of diseases and defects in potato tubers from measurements of optical spectral reflectance. Journal of Agricultural En gineering Research 26(2) : 151 160. Porteous, R. L. and A. Y. Muir. 1986. Some measurements of the performance of an experimental quality grading machine for potatoes. Potato Research 30 (1): 23 34. Preston, D., and M. Glynn, 1995 Bruising evaluation of different cultivars on machines. Spudman 33: 16 23. Robertson I.M., 1970. Assessment of damage in potato tubers. Dept. Note SIN/60 Scottish Penicuik, Scotland: Institute of Agricultural Engineering. SAS. 20 09 SAS/STAT User's Guide Ver. 9.2 2nd E d itio n Cary N.C.: SAS Institute, Inc. Singh, B. 2005. Visible and Near Infrared Spectroscopic Analysis of Potatoes. MS Thesis. Montreal, Canada : McGill University, Department of Bioresource Engineering.

PAGE 202

202 Souci, S. W., W. Fachmann and H. Kraut. 2000. Food co mposition and nutrition tables 639 640 Stuttgart, Germ any: Medpharm, Scientific Publishers. Stevenson, W. R., R. Loria G. D. Franc D. P. Weingartner. 2001 Compendium of Potato Diseases St. Paul, Minnesota: The American Phytopathological Society. Su ndberg, R. 1999. Multivariate calibration: Direct and indirect regression methodology. Scandinavian J. Statistics 26(2): 161 207. Tao, Y., P. H. Heinemann, Z. Varghese, C. T. Morrow, H. J. L. Sommer. 1995. Machine vision for color ins pection of potatoes a nd apples. Transactions of the ASAE 38(5): 1555 1561. Thornton, M., and W. Bohl 1998. Preventing Potato Bruise Damage University of Idaho, Moscow, ID : Agricultural Publications. USDA. 2008. United States Standards for Grades of Potatoes Washington, D C L USDA USDA. 2008. U.S. Potato Statistics. National Agricultural Statistics Database Washington, DC: USDA National Agricultural Statistics Service. Available at: www.nass.usda.gov. Accessed 20 March 2011. Van Der Voet, H. 1994. Comparing the Predictive Accuracy of Models Using a Simple Randomization Test. Chemometrics and Intelligent Laboratory Systems 25: 313 323. Western Potato Council. 20 11 Guide to Commercial Potato Prod uction on the Canadian Prairies Portage la Prairie MB R1N 3B9: Western Potat o Council. Wold, H. 1975. Soft Modeling by Latent Variables; the Nonlinear Iterative Partial Least Square s Approach. In Perspectives in Probability and Statistics London: Academic Press. Wold, S., H. Martens, and H. Wold. 1983. The multivariate calibra tion method in chemistry solved by the PLS method. In Proc. Conference in Matrix Pencils : 286 293 Xing, J., V.V. Linden, M. Vanzeebroeck, and J. De Baerdemaeker. 2005. Bruise detection on Jonagold apples by visible and near infrared spectroscopy. Food Co ntrol 16: 357 361. band Spectral Measurement for Moisture Content. Tamaki Campus New Zealand : University of Auckland Computer Science Department

PAGE 203

203 BIOLOGICAL SKETCH Michael Anthony Brecht was born in 1985 in Gai nesville, Florida, where he was raised. After graduating from Buchholz High School in 2003 he o btained an Associate of Arts from Santa Fe Community College in 2005 and transferred to the University of Florida to get a biological e ngineering in 2009. In 2010, he came back to the University of Florida as a m aster of Engineering degree in agricultural and biological e ngineering in December 201 2