EPIDEMIOLOGY AND MANAGEMENT OF

PHOMOPSIS CANE AND LEAF SPOT OF GRAPE

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of

Philosophy in the Graduate School of The Ohio State University

By

Mizuho Nita, M.S.

*****

The Ohio State University

2005

Dissertation Committee: Approved by

Professor Laurence V. Madden, Advisor

Professor Michael A. Ellis ______

Associate Professor Anne E. Dorrance Adviser

Professor Patrick E. Lipps Plant Pathology Graduate Program

ABSTRACT

Phomopsis cane and leaf spot is a disease of grape (Vitis spp.) caused by

Phomopsis viticola (Sacc.). The fungus survives winter in grape cane tissues that were

infected in previous years. In the spring, conidia are splashed by rain onto new growth

where they infect the plant tissues. The fungus can infect many parts of the grape,

including shoots, rachises, leaves, and fruits, and infection typically takes place when the

tissues are immature. The disease is considered to be monocyclic. The control of the

disease has usually been done by either selective pruning of diseased canes (reducing

inoculum) or preventative spraying of protectant fungicides onto new tissues. To extend

our understanding of the epidemiology and control the disease, studies were conducted to:

1) evaluate a disease warning system by applying fungicides and fungicide-adjuvant

combinations based on predicted infection periods utilizing measured weather conditions;

2) determine efficacy of a dormant fungicide spray program for controlling the disease in

its early stage of development (spring); 3) correlate commercial control practices and environmental conditions with disease incidence based on a state-wide survey of

commercial vineyards; and 4) determine spatial pattern of the disease in small (within a

vine) and large (among vines) spatial scales using a range of spatial analyses. With the

warning system, control was often equal to that obtained with a 7-day calendar-based

protectant program, but with fewer fungicide applications. However, fungicides and

ii fungicide-adjuvant combinations used with the warning system did not show curative activity in a controlled-environment study. A dormant fungicide application provided consistent, yet only moderate, control of the disease; however, growers who applied a dormant application or spring protectants tended to have less disease incidence based on the results of the commercial survey. Analyses of spatial patterns revealed that the disease tended to aggregate within a vine, but clustering of the disease among vines was not common. Based on the control achieved with the warning system, the moderate control by dormant applications, and management-practice information from the survey, early season application of protective fungicides was shown to be a key factor for successful management of P. viticola.

iii

To my friends and family

iv ACKNOWLEDGMENTS

I deeply appreciate my advisor, Dr. Laurence V. Madden, for intellectual support,

insightful opinions, his professionalism, patience, and flexible mind to guide me

throughout the graduate program.

I wish to thank my co-advisor, Dr. Michael A. Ellis, for his expert opinions, and providing various opportunities to become familiar with the field of plant pathology and beyond.

I also thank Dr. Anne E. Dorrance for various encouragements and challenges, which enriched my graduate program.

I thank Dr. Patrick E. Lipps for his generosity, expert opinions, and careful review of my project.

I feel very lucky to have Mr. Lee Wilson in our lab as a research associate.

Without his help on numerous technical and other issues, my program could not be the same.

Other people I should mention are my colleagues Mr. Angel Rebollar-Alviter and

Dr. Edgar Huitema for their friendship and advice, Mr. Bill Bardall and Mr. Bob James for their help with the research in the field and greenhouses, and summer helpers, Ms.

Maria Holdcraft, Ms. Taylor Hawkins, and Ms. Heather Reed, for their hard work.

v VITA

February 24, 1972 ⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅ Born – Okegawa, Japan

1994 ⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅ B.A. in Geography, Southern Illinois University at Carbondale.

2002 ⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅ M.S. in Plant Pathology, The Ohio State University

2002 - present ⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅⋅ Graduate Research Associate, The Ohio State University

PUBLICATIONS

Refereed Journal Articles 1. Ellis, M. A., Nita, M. and Madden, L. V. (2000). "First report of Phomopsis fruit rot of strawberry in Ohio." Plant Disease 84: 199. (Disease note) 2. Madden, L. V., Turechek, W. W. and Nita, M. (2002). "Evaluation of generalized linear mixed models for analyzing disease incidence data obtained in designed experiments." Plant Disease 86: 316-325. 3. Nita, M., Ellis, M. A. and Madden, L. V. (2003). "Effects of temperature, wetness duration, and leaflet age on infection of strawberry foliage by Phomopsis obscurans." Plant Disease 87: 579-584. 4. Nita, M., Ellis, M. A. and Madden, L. V. (2003). "Reliability and accuracy of visual estimation of Phomopsis leaf blight of strawberry." Phytopathology 93: 995-1005.

vi Technical, Non-refereed Journal Articles 1. Ellis, M. A., Nita, M., Madden, L. V. and Wilson, L. L. (2001). "Evaluation of fungicides and fungicide combinations for control of gray mold, 2000." Fungicide and Nematicide Tests 56: SMF45. 2. Nita, M., Madden, L. V. and Ellis, M. A. (2000). "Evaluation of fungicides and fungicide combinations for control of leaf blight, 1999." Fungicide and Nematicide Tests 55: 133. 3. Nita, M., Madden, L. V. and Ellis, M. A. (2001). "Evaluation of fungicides and fungicide combinations for control of leaf blight, 2000." Fungicide and Nematicide Tests 56: SMF46. 4. Nita, M., Rebollar-Alviter, A., Wilson, L. L. and. Ellis, M. A (2005). "Evaluation of fungicides for control of grape black rot and powdery mildew, 2004." Fungicide and Nematicide Tests 60: SMF030. 5. Rebollar-Alviter, A., Nita, M., and Ellis, M. A. (2005). "Evaluation of strobilurin fungicides to control strawberry leaf blight, 2003, 2004.” Fungicide and Nematicide Tests 60: SMF036.

Abstracts in Journal 1. Nita, M., Ellis, M. A. and Madden, L. V. (2001). "Functional relationship between estimated and actual disease severity." Phytopathology 91: S66. 2. Nita, M., Ellis, M. A., Wilson, L. L. and Madden, L. V. (2001). "Effects of temperature, wetness duration, and leaf age on infection of strawberry foliage by Phomopsis obscurans." Phytopathology 91: S66. 3. Nita, M. and Madden, L. V. (2000). "Evaluation of visual disease severity assessments of strawberry leaf blight." Phytopathology 90: S56. 4. Nita, M. and Madden, L. V. (2002). "Comparison of the spatial pattern of two foliar diseases of strawberry." Phytopathology 92: S59. 5. Nita, M., Madden, L. V., Wilson, L. L. and Ellis, M. A. (2003). "Evaluation of a disease prediction system for Phomopsis cane and leaf spot of grape." Phytopathology 93: S65. 6. Nita, M., M. A. Ellis and L. V. Madden (2004). "Evaluation of the efficacy of dormant applications of lime sulfur and fixed copper to control Phomopsis cane and leaf spot of grape." Phytopathology 94: S76. 7. Nita, M., M. A. Ellis, L. L. Wilson and L. V. Madden (2004). "Evaluation of fungicides and fungicide-adjuvant mixtures for post-infection control of Phomopsis cane and leaf spot of grape." Phytopathology 94: S76. 8. Mideros, S. X., S. Constanzo, M. Nita, S. K. St. Martin and A. E. Dorrance (2004). "Screening of disease susceptibility genes to Phytophthora sojae in soybean line Ox-20-8." Phytopathology 94: S70. 9. Nita, M., M. A. Ellis, and L. V. Madden (2005). "Spatial patterns of Phomopsis leaf and cane spot of grape at different spatial scales in Ohio’s commercial vineyards." Phytopathology 95: S76.

vii 10. Nita, M., M. A. Ellis, L. L. Wilson and L. V. Madden (2004). "Evaluation of new and current management strategies to control Phomopsis cane and leaf spot of grape." Phytopathology 95: S128.

FIELDS OF STUDY

Major Field: Plant Pathology

viii TABLE OF CONTENTS

Abstract ………………………………………………………..………………… ii Dedication ………..……………………………………………………………… iii Acknowledgments ……………………………………………………………….. iv Vita ………………………………………………………………………………. v List of Tables ……………………………………………………………………. xii List of Figures …………………………………………………………………… xvi

Chapters Pages 1. Introduction ……………………………………………………………… 1

Reference ……………………………………………………………… 10

2. Evaluation of a disease warning system for Phomopsis cane and leaf spot of grape: a field study

Introduction ………………………………………………………………. 12

Materials and Methods ..…………………………………………………… 15 The disease warning system …………………………………………… 15 Treatments …………………………………………………………….. 17 Disease assessment …………………………………………….……… 19 Analysis of data ……………………………………………………….. 19

Results ……………………………………………………………………... 20 Preliminary study (2001) ……………………………………………… 20 2002 …………………………………………………………………… 19 2003 …………………………………………………………………… 21 2004 …………………………………………………………………… 23 Rachis infection ……………………………………………………….. 25

Discussion ………………………………………………………………… 26

References ……………………………………………………………….... 44

ix 3. Evaluation of the curative and protectant activity of fungicides and fungicide- adjuvant mixtures on Phomopsis cane and leaf spot of grape: a controlled environment study

Introduction ……………………………………………………………….... 47

Materials and Methods …………………………………………………….. 49 Host preparation ……………………………………………………….. 49 Inoculum Preparation …………………………………………………. 50 Inoculation methods …………………………………………………… 50 Fungicides and adjuvants ……………………………………………… 51 Post-inoculation application study …………………………………….. 52 Pre-inoculation application study ……………………………………… 53 Re-isolation of P. viticola ……………………………………………… 53 Assessment methods and analysis of data …………………………….. 54

Results ……………………………………………………………………… 55 Post-inoculation application study ……………………………………… 55 Post-inoculation application study ……………………………………… 55 Re-isolation of P. viticola ………………………………………………. 56

Discussion …………………………………………………………………... 56

References ………………………………………………………………….. 68

4. Dormant application of fungicide and its effects on disease intensity and inoculum production

Introduction………………………………………………………………….. 70

Materials and Methods …………………………………………………….. 72 Collection of splashed rain water ……………………………………… 75 Sporulation on cane pieces ……………………………………………. 75

Results …………………………………………………………………….. 76 Disease incidence and severity – ‘Concord’ ………………………….. 76 Disease incidence and severity – ‘Catawba’ ………………………….. 78 Collection of spores in splashed rain water …………………………… 79 Sporulation on cane pieces ……………………………………………. 79

Discussion ………………………………………………………………… 80

References ………………………………………………………………… 93

x 5. Disease incidence of Phomopsis cane and leaf spot of grape in commercial vineyards in Ohio

Introduction………………………………………………………………….. 95

Materials and Methods …………………………………………………….. 97 Hierarchical analysis …………………………………………………… 98 Development of risk models: Predictor variables ……………………… 98 Stepwise logistic regression models …………………………………… 101

Results …………………………………………………………………….. 103 Hierarchical analysis ….………………………………………………. 103 Grower survey …………………………………………………………. 105 Preliminary evaluation of predictor variables …………………………. 105 Stepwise logistic regression models ……………………….…………… 106

Discussion ………………………………………………………………… 108

References ………………………………………………………………… 123

6. Spatial distribution of Phomopsis cane and leaf spot symptoms in commercial vineyards in Ohio

Introduction………………………………………………………………….. 124

Materials and Methods …………………………………………………….. 126 Discrete distributions analyses ………..……………………………….. 127 Binary power law analysis ……………………………………………… 129 SADIE ………………………………………………………………….. 130 Spatial covariance ……………………………………………………… 132

Results ……………………………………………………………………… 134 Disease incidence ……………………………………………………… 134 Discrete distributions analyses ………………………………………… 135 Binary power law analysis ……………………………………………… 136 SADIE ………………………………………………………………….. 138 Spatial covariance ……………………………………………………… 138

Discussion ………………………………………………………………… 140

References ………………………………………………………………… 153

Bibliography ……………………………………………………………………… 156

xi LIST OF TABLES

Page Table 2.1. Fungicides and adjuvant used in experiments and doses 32

Table 2.2. Evaluation of warning-system-based and calendar-based 33 protectant spray programs for control of Phomopsis cane and leaf spot of grape (caused by Phomopsis viticola) in 2001

Table 2.3. Evaluation of warning-system-based and calendar-based 24 protectant spray programs for control of Phomopsis cane and leaf spot of grape (caused by Phomopsis viticola) in two vineyards (‘Catawba’ and ‘Seyval’) in 2002

Table 2.4. Evaluation of warning-system-based and calendar-based 35 protectant spray programs for control of Phomopsis cane and leaf spot of grape (caused by Phomopsis viticola) in two vineyards (‘Catawba’ and ‘Seyval’) in 2003

Table 2.5. Evaluation of warning-system-based and calendar-based 36 protectant spray programs for control of Phomopsis cane and leaf spot of grape (caused by Phomopsis viticola) in two vineyards (‘Catawba’ and ‘Seyval’) in 2004

Table 2.6. Evaluation of warning-system-based and calendar-based 37 protectant spray programs for control of Phomopsis cane and leaf spot of grape rachis infections (caused by Phomopsis viticola) in the ‘Catawba’ vineyard in 2003 and 2004

Table 3.1. Fungicides and doses used in experiments 61

Table 3.2. Effects of various fungicides and fungicide-adjuvant mixtures 62 applied before or after inoculation of Phomopsis viticola on disease symptom development on grape leaves and internodes

Table 3.3. F-test statistics for contrasts of treatment means for leaf or 63 internode disease severity.

xii Table 3.4. Effects of fungicides applied before inoculation of Phomopsis 64 viticola on disease symptoms on grape leaves and internodes (Experiment V)

Table 3.5. Effects of fungicides-adjuvant combinations applied before 65 inoculation of Phomopsis viticola on disease symptoms on grape leaves and internodes (Experiment VI)

Table 3.6. Re-isolation of Pomopsis viticola from inoculated vines in 66 controlled environment studies for post-infection activity of fungicides and fungicide adjuvant mixtures (Experiments II-IV)

Table 4.1. Efficacy of dormant applications of calcium polysulfide and fixed 86 copper, and their application timing, on disease incidence and severity of Phomopsis cane and leaf blight in a ‘Concord’ vineyard in 2003 and 2004 in Ohio

Table 4.2. Efficacy of dormant fungicide applications on disease incidence of 87 rachises of Phomopsis cane and leaf spot in a ‘Concord’ vineyard in 2003 and 2004 in Ohio

Table 4.3. Efficacy of a dormant application of calcium polysulfide and of a 88 calendar-based mancozeb application on disease incidence and severity of Phomopsis cane and leaf spot in the ‘Catawba’ vineyard in 2003 and 2004 in Ohio

Table 4.4. Efficacy of dormant fungicide applications on rachis disease 89 incidence of Phomopsis cane and leaf spot in a ‘Catawba’ vineyard in 2003 and 2004 in Ohio

Table 4.5. Mean number of spores per milliliter of splashed rain per rain 90 event in the first and second month after applications of dormant fungicides in spring

Table 4.6. Mean number of mature pycnidia with cirrhi of Phomopsis viticola 91 per cm2 of grape cane from the ‘Concord’ vineyard collected during the winter following dormant application of fungicides

Table 5.1. Location of sampled sites for Phomopsis cane and leaf spot of 114 grape disease incidence from 2002 to 2004 in Ohio

Table 5.2. Mean disease incidence (%) of Phomopsis cane and leaf spot of 115 grape per vineyard in percentage during 2002-2004 in Ohio for leaves and cane internodes.

xiii Table 5.3. Potential predictor variables considered in development of risk 116 model for Phomopsis cane and leaf spot of grape on leaves and cane internodes in commercial vineyards in Ohio over 3 years

Table 5.4. Test statistics for the effect of region (R), farm within region (G), 117 and vineyard within farm (V) on incidence of Phomopsis cane and leaf spot of grape on leaves and cane internodes in Ohio over 3 years

Table 5.5. 2 118 Estimated variances for farm within region (σ G ), vineyard within 2 2 farm (σ V ), and site within vineyard (σ S ) for incidence of Phomopsis cane and leaf spot of grape on leaves in internodes in Ohio over 3 years

Table 5.6. Kendall correlation coefficients (Ks)between weather or 119 management variables and indicator variables (D20, and D40) for the risk of Phomopsis cane and leaf spot of grape on leaves and cane internodes in Ohio over 3 years

Table 5.7. Variables identified with stepwise logistic regression analysis with 120 different subsets of variables for two different risk levels of Phomopsis cane and leaf spot of grape on leaves and cane internodes in commercial vineyards in Ohio over 3 years (D40 and D20)

Table 5.8. Logistic linear models developed for predicting Phomopsis cane 121 and leaf spot risks (D40 or D20), together with overall prediction accuracy, sensitivity, and specificity of models.

Table 6.1. Summary statistics [minima, maxima, and three quartiles (25% 145 percentile (Q1), median, and 75% percentile (Q3))] for number of diseased individuals (leaves or internodes; DI), estimated beta- binomial parameters p) (expected probability of diseased leaves or internodes) and θˆ (heterogeneity of disease incidence), index of dispersion (D), SADIE’s index of aggregation (Ia), and first-order autocorrelation (r1) for Phomopsis cane and leaf spot of grape in commercial vineyards from 2002 to 2004

xiv

Table 6.2. Percent of vineyards where: the beta-binomial (BBD) distribution 146 provided a good fit to number of diseased leaves and internodes; where the binomial (BIN) distribution provided a good fit to number of diseased leaves and internodes; where the beta- binomial provided a significantly better fit than the binomial based on the likelihood ratio statistic (LRS); and where the index of dispersion (D) was significantly larger than 1.

Table 6.3. Estimated parameters of the binary power law (equation 2) and 147 their standard errors for Phomopsis cane and leaf spot of grape on leaves and internodes in commercial vineyards on Ohio from 2002 to 2004

Table 6.4. Percentage of vineyards where SADIE indicated aggregation, 148 where a likelihood ratio statistics (LRS) indicated that a random effects model with spatial structure fitted better than a model with independence of data, and where a LRS measured that nugget effect model fitted better than a no-nugget model

xv LIST OF FIGURES

Page Figure 1.1 A, α-conidia of Phomopsis viticola observed under compound 6 microscope (100x), B, β-conidia of P. viticola observed under scanning electron micrograph, and C, Cirrhus of P. viticola coming out from a pycnidium observed under scanning electron micrograph.

Figure 1.2 A, Typical leaf symptom and B, internode symptom of Phomopsis 7 cane and leaf spot of grape

Figure 1.3. Disease cycle of Phomopsis cane and leaf spot of grape. 8

Figure 1.4. A conceptual diagram of Phomopsis cane and leaf spot disease 9 system drawn based on convention of Odum’s energy flow diagram (Odum and Odum 2000). The system’s boundary is defined with a solid blue line. Light-blue circles represent factors outside of the system. Green icons represent energy “producer” (grape) and energy “consumer” (P. viticola). Yellow boxes are countable variables and red arrow-shaped boxes represent processes in the system. Each arrow represents a flow of energy or “information”, and ones highlighted with light-blue are discussed in the chapters (indicated by dark-blue boxes).

Figure 2.1. Predicted number of lesion per leaf based on equation 1, a model 38 developed by Erincik et al. (9). Curved lines are predictions for combinations of temperature and wetness duration. Horizontal lines are thresholds categories tested with the disease warning system.

Figure 2.2. Plot of leaf wetness events and average temperature during the 39 events, applied warning-system-based treatments (responded infection events are shown with arrows), and theoretical coverage of the fungicide application (a 7-day period). Data are for ‘Catawba’ vineyard in A, 2002, B, 2003, and C, 2004.

xvi Figure 2.3. Efficacy of the warning-system-based treatment of mancozeb with 40 Regulaid and calendar-based treatment of mancozeb relative to untreated control. Bars represent percent control: PC = 100[(C- T)/C], where C is disease intensity (incidence or severity) for the untreated control and T is disease intensity for the treatment (either calendar-based mancozeb application or warning-system- based mancozeb + Regulaid application on predicted medium infection events).

Figure 2.4. Efficacy of the warning-system-based treatment of mancozeb with 41 Regulaid and calendar-based treatment of mancozeb relative to untreated control. Bars represent percentage of disease control by treatments per spray: PCPS=PC/(# of sprays), where PC is percent control (Figure 2.3), for the calendar-based protectant schedule with mancozeb and for predicted medium infection event application of mancozeb + Regulaid.

Figure 3.1. Time-flow diagram of post-inoculation application of fungicides 67 and a list of wetness period and drying times of each experiment. a indicates fungicide at 150% labeled dose

Figure 4.1. Number of P. viticola conidia per milliliter of splashed rain water 92 in the ‘Concord’ vineyard, and amount of rainfall recorded in 2003 and 2004 (vertical bars). Arrows indicate the date of the spring dormant fungicide application of calcium polysulfide. Solid lines represent mean number of conidia observed per rain event for the fall-and-spring application of calcium polysulfide, and dotted lines represent the mean for the unsprayed control.

Figure 5.1. A map of major grape growing regions in Ohio. Regions surveyed 122 in this study are circled. The map was obtained from http://www.ohgrapes.org/.

Figure 6.1. Frequency distribution of: A, B, the estimated beta-binomial 149 parameter p) (estimated average probability of disease incidence) for leaf and internode disease incidence of Phomopsis cane and leaf spot of grape; D, E, the estimated beta-binomial aggregation ) parameter θ ; G, H, the index of dispersion, D; and J, K, SADIE’s index of aggregation, Ia. C, F, I, L, the frequency distributions of ˆ the difference in pˆ ,θ , D , and I a for leaves (L) and internodes (I)

xvii

Figure 6.2. Relationship between the logarithm of the observed variance of 150 Phomopsis cane and leaf spot disease incidence among sampling units (V) and the logarithm of the variance assuming disease incidence follows a binomial distribution (eq. 1). In the equation, pˆ is the estimate of the mean disease incidence and n is the number of leaves or internodes assessed per sampling unit (n = 15). Each point represents a single year x vineyard combination. Solid line represents the least squares fit to the data, and dashed

line represents the binomial [log(Vobs ) − log(npˆ(1− pˆ )) ] line.

Figure 6.3. The relationship between estimated aggregation parameter of the 151 beta-binomial distribution (θˆ ) and disease incidence ( pˆ , estimated expected disease probability), for Phomopsis cane and leaf spot of grape in commercial vineyards in Ohio over 3 years. Each point represents a single year x vineyard combination. The solid line represents a theoretical relationship between θˆ and pˆ based on the results from the binary power law (eq. 2)

Figure 6.4. Examples of Phomopsis cane and leaf spot spatial pattern results 152 for two commercial vineyards in Ohio A, B, Frequency distribution of diseased leaves (bars) and expected frequency for binomial (short-dash) and beta-binomial (dot) distributions, C, D, disease incidence contour maps (darker the color, higher the disease incidence) E, F, Semivariograms

xviii CHAPTER 1

INTRODUCTION

Phomopsis cane and leaf spot is a disease of grape (Vitis spp.) caused by the fungus Phomopsis viticola (Sacc.). Previously, this disease was also known as “dead-arm disease” in American literature (5, 7, 17), but it was eventually discovered that this could be caused by two pathogens, Eutypa lata (causal agent of Eutypa dieback on grape) and P. viticola (4). Phomopsis cane and leaf spot is common in the U.S. and other grape growing regions around the world (8, 9, 10). Up to 30% loss of the crop has been reported in Southern Ohio grape vineyards (7).

The genus Phomopsis is taxonomically classified in a group of fungi named Fungi

Imperfecti (= Deuteromycete, a conidial fungus that lacks, or has unknown, sexual stage)

(24). Since a characteristic of the genus Phomopsis is a pycnidium (a fruiting body), it is further classified into Sphaeropsidales (2, 24). In the many of known cases, Phomopsis species are considered as the anamorphs (asexual stage) of Diaporthe species (Phylum

Ascomycota, Order Diaportheles) (2, 23). There has been several studies on identification of a teleomorph (sexual stage) of P. viticola; however, conclusive results on sexual stage of P. viticola have yet to found (2, 11, 12, 16, 19, 20, 21).

1 More than 60 species in this genus are known to be plant pathogenic, and many of these cause economically important plant diseases, such as Phomopsis seed rot on soybean, caused by Diaporthe phaseolorum var. sojae, Diaporthe phaseolorum var. caulivora, and Phomopsis longicolla (19, 20). Typical symptoms caused by Phomopsis species are: cankers; fruit rots; stem end rots; root rots; blight; spots; decay; dieback; damping off; and mummification (23). A typical morphological characteristic of

Phomopsis is a “dark eustromatic or pycnidial conidoma” (1, 23). One of the differences between Phomopsis and the closely related genus Phoma is that Phomopsis species produce two different types of pycnidial conidioma (i.e., asexual spores produced from a fruiting body pycnidium), α-conidia and β-conidia. In contrast, Phoma species produce only one type of conidia (23). In the literature, α-conidia are described as “hyaline, nonsptate, and elliptic, fusiform-elliptic, or oblong-elliptic” (23), and β-condia are described as “hyaline, nonsptate, filiform, and hamate or curved” spore (23).

In the case of P. viticola, α-conidia (Figure 1.1A) are hyaline, nonseptate oblong- elliptic in shape with a size of 6.3-11.2 x 1.7-2.8 µm, and β-conidia (Figure 1.1B) are hyaline, nonseptate, filiform and curved in shape with a size of 20-25 x 1 µm (23).

Function of the β-conidia for P. viticola is not known (23).

The fungus can infect many parts of the grape including, shoots, rachises, leaves, and fruits, and infection typically takes place when plant tissues are immature (15, 22).

Infected leaves show small irregular, or round, pale green-to-yellow spots with dark centers (Figure 1.2A). Brown to black necrotic irregular shaped lesions develop on the infected canes and rachises (Figure 1.2B). Infected fruits appear as brown shriveled 2 berries near harvest (7, 14, 18, 21). Infections on the canes and rachises weaken the plant and may cause premature fruit drop (14, 18). Infections on fruits directly decrease yield and fruit quality (7).

The disease is considered to be a monocyclic (i.e., one complete disease cycle per season) (Figure 1.3). The fungus survives winter in grape cane tissues that were infected in previous years (7, 14). In spring, numerous pycnidia develop on infected canes, and a gelatinous masses of spores (α- or β-conidia), called cirrhi, exude out from the pycnidia

(Figure 1.1C). Conidia are splashed by rain onto new growth where they can infect the plant tissues. Free water on the plant surface is required for germination of Phomopsis conidia (14, 18).

The damage caused by the disease can be direct or indirect. Fruit infection reduces yield and increases labor costs involved in selecting disease-free berries.

Premature dropping of fruits due to breaking of rachises and shoots also decreases yield

(17). Also, heavily infected canes may not produce new shoots, or new shoots have a high risk of being infected. As with most plant diseases, it is assumed that yield is also reduced due to reduction in radiant energy absorbed or radiation use efficiency by diseased leaves (14).

Selective pruning and protective fungicide applications are commonly used to control the disease (3). Protective fungicides, such as mancozeb or captan, applied on a

7-10 day calendar based schedule on green tissues and fruits have been used for disease control (14, 18). Research has indicated that the fungus becomes active early in the growing season, when only about 2.5-cm of the new growth of the grape tissues are

3 present (6, 14). Close proximity between the source of inoculum (infected canes from the main trunk) and the susceptible tissues (new growth of vines) may also favor infection in the early growing season.

The overall objective of this dissertation was to extend the knowledge and understanding of various aspects of disease epidemiology and control of Phomopsis cane and leaf spot of grape, so that more effective and efficient management strategies that consist of fewer number of fungicide applications can be developed. A series of studies addressed different aspects of disease management and epidemiology, which is summarized by the diagram in Figure 1.4.

The following chapters describe:

2) Test of a disease warning system for Phomopsis on grape under field conditions,

where fungicides are applied based on weather conditions;

3) Examination of potential curative activity of fungicides or fungicide-adjuvant

combinations;

4) Determination of the efficacy of dormant fungicide spray program for

controlling the disease in its early stage of the development;

5) Examination of disease incidence in commercial vineyards in Ohio, obtained

from a state-wide survey over 3 years; and

6) Examination of spatial pattern of the disease in commercial vineyards using

datasets from the survey.

4 In this thesis, several epidemiological approaches were utilized, including: a warning-system evaluation; field and controlled-environment experiments; survey samplings; development of risk models; and spatial analyses. Results from warning system evaluation (Chapter 2) and curative activity study (Chapter 3) indicated that

Phomopsis cane and leaf spot can be controlled with fewer applications of protectant fungicides. The moderate, but consistent, efficacy of dormant fungicide application

(Chapter 4), and the results from management-practice evaluations (Chapter 5) indicated the importance of dormant and early season fungicide applications. Phomopsis on grape was found in all the vineyards surveyed (Chapter 5), and the disease exhibited moderate level of aggregation at smaller (within vine) spatial scales, but there were little evidence for larger (among vines) spatial scale aggregation of this disease (Chapter 6).

5

A

B

C

Figure 1.1 A, α-conidia of Phomopsis viticola observed under compound microscope (100X), B, β-conidia of P. viticola, observed under scanning electron micrograph, and C, Cirrhus of P. viticola emerging from a pycnidium, observed under scanning electron micrograph.

6 A

B

Figure 1.2 A, Typical leaf symptom and B, internode symptom of Phomopsis cane and leaf spot of grape, caused by Phomopsis viticola

7 Dissemination and inoculation establishment

survival

Figure 1.3. Pictorial representation of disease cycle of Phomopsis cane and leaf spot of grape.

8

Figure 1.4. A conceptual diagram of Phomopsis cane and leaf spot disease system drawn based on convention of Odum’s energy flow diagram (13). The system’s boundary is defined with a solid blue line. Light-blue circles represent factors outside of the system. Green icons represent energy “producer” (grape) and energy “consumer” (P. viticola). Yellow boxes are countable variables and red arrow-shaped boxes represent processes in the system. Each arrow represents a flow of energy or “information”, and those highlighted with light-blue are discussed in the chapters (indicated by dark-blue boxes).

9 REFERENCE

1. Agrios, G. N. 1997. Plant Pathology. 4th Academic Press, San Diego. 2. Alexopoulos, C. J., Mims, C. W., and Blackwell, M. 1996. Introductory mycology. Fourth edition John Wiley & Sons, Inc, New York. 3. Bergamin Filho, A., Carneiro, S. M. T. P. G., Godoy, C. V., Amorim, L., Berger, R. D., and Hau, B. 1997. Angular leaf spot of Phaseolus beans: relationships between disease, healthy leaf area, and yield. Phytopathology- 87 5:506-515. 4. Carter, M. V. 1960. Further study on Eutypa armeniacea Hansf. & Carter. Austrarian Journal of Agricultural Research 11:498-504. 5. Cucuzza, J. D., and Sall, M. A. 1982. Phomopsis cane and leaf spot disease of grape vine: Effect of chemical treatments on inoculum level, disease severity, and yield. Plant Disease 66:794-797. 6. Ellis, M. A., Welty, C., Funt, R. C., Doohan, D., Wiliams, R. N., Brown, M., and Bordelon, B. ed. 2004. Midwest Small Fruit Pest Management Handbook. Bulletin 861 Ohio State University Extension, Columbus, OH. 7. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2001. Effect of growth stage on susceptibility of grape berry and rachis tissues to infection by Phomopsis viticola. Plant Disease 85 5:517-520. 8. Lal, B., and Arya, A. 1982. A soft rot of grapes caused by Phomopsis viticola. Indian Phytopathology 35 2:261-264. 9. Malathrakis, N. E., and Baltzakis, N. G. 1976. Control of Dead-arm of Grapes. Poljoprivredna znanstvena smotra - Agriculture Conspectus Scientificus 39 49:261-269. 10. Mostert, L., Crous, P. W., and Petrini, O. 2000. Endophytic fungi associated with shoots and leaves of Vitis vinifera, with specific reference to the Phomopsis viticola complex. Sydowia 52 1:46-58. 11. Mostert, L., Denman, S., and Crous, P. W. 2000. In vitro screening of fungicides against Phomopsis viticola and Diaporthe perjuncta. South African Journal for Enology and Viticulture 21 2:62-65. 12. Niekerk, J. M. v., Groenewald, J. Z., Farr, D. F., Fourie, P. H., Halleen, F., and Crous, P. W. 2005. Reassessment of Phomopsis species on grapevines. Australasian Plant Pathology 34 1:27-39. 13. Odum, H. T., and Odum, E. C. 2000. Modelling for all Scales: an introduction to System Simulation. Academic Press, San Diego, CA. 14. Pearson, R. C., and Goheen, A. C. ed. 1988. Compendium of Grape Diseases. APS Press, St. Paul, MN. 15. Pezet, R., Pont, V., and Girardet, F. 1983. Microcycle conidiation in pycnidial cirrhii of Phomopsis viticola Sacc. induced by elemental sulfur. Canadian Journal of Mycrobiology 29:179-184. 16. Phillips, A. J. L. 1999. The relationship between Diaporthe perjuncta and Phomopsis viticola on grapevines. Mycologia- 91 6:1001-1007. 17. Pine, T. S. 1959. Development of the grape dead-arm disease. Phytopathology 49:738-743.

10 18. Pscheidt, J. W., and Pearson, R. C. 1989. Effect of grapevine training systems and pruning practices on occurrence of Phomopsis cane and leaf spot. Plant Disease 73 10:825-828. 19. Rawnsley, B., Wicks, T. J., Scott, E. S., and Stummer, B. E. 2004. Diaporthe perjuncta does not cause Phomopsis cane and leaf spot disease of grapevine in Australia. Plant Disease 88 9:1005-1010. 20. Scheper, R. W. A., Crane, D. C., Whisson, D. L., and Scott, E. S. 2000. The Diaporthe teleomorph of Phomopsis taxon 1 on grapevine. Mycological Research 104 2:226-231. 21. Schilder, A. M. C., Erincik, O., Castlebury, L., Rossman, A., and Ellis, M. A. 2005. Characterization of Phomopsis spp. infecting grapevines in the Great Lakes regions of north America. Plant Disease 89:755-762. 22. Sergeeva, V., Nair, N. G., Barchia, I., Priest, M., and Spooner-Hart, R. 2003. Germination of beta-conidia of Phomopsis viticola. Austrarian Plant Pathology 32:105-107. 23. Uecker, F. A. 1998. A world list of Phomopsis names with notes on nomenclature, morphology and biology. Mycologia memoir No.13 J. CRAMER, Berlin. 24. Ulloa, M., and Hanlin, R. T. 2000. Illustrated Dictionary of Mycology. The American Phytopathological Society, St. Paul, MN.

11 CHAPTER 2

EVALUATION OF A DISEASE WARNING SYSTEM

FOR PHOMOPSIS CANE AND LEAF SPOT OF GRAPE:

A FIELD STUDY

INTRODUCTION

Phomopsis cane and leaf spot of grape, caused by the fungus Phomopsis viticola

(Sacc.), is a common grape disease in the U.S. and grape growing regions around the world (4, 20, 22, 26). The fungus overwinters as mycelia or possibly as immature pycnidia in infected bark tissue. In spring, pycnidia mature and emerge through grape epidermal cells, and then spores (α- and β-conidia) in a gelatinous mass (cirrhi) exude from the pycnidia. α-conidia are splashed by rain to healthy tissues where infection may occur. Ability to cause infection by β-conidia is unknown (29).

The fungus can infect many parts of the grape including shoots, rachises, leaves, and fruits; and young, immature tissues are most susceptible to infection (8, 24). Because of the nature of rain splash dispersal and distance between vines, spread of the disease

12 usually takes place within the vine rather than from vine-to-vine (24). Disease development is more prominent in spring (8), probably because of high inoculum production at this time period (4) and the relatively closer proximity of inoculum sources

(older infected canes on the main trunks) to susceptible young tissues (shoots growing from the main trunks).

Symptoms of Phomopsis leaf spot are small irregular, or round shaped, pale green-to-yellow spots with dark centers (24). Dark irregular shaped lesions develop on infected canes and rachises. Close to harvest, infected fruits appear as brown shriveled berries with visible pycnidia (24). Premature dropping of fruits due to breaking of the rachises and shoots also contribute to yield loss (24). Up to 30% crop loss has been reported in southern Ohio due to fruit rot (8).

Control of the disease is usually accomplished by the use of protectant fungicides such as captan, mancozeb and folpet, applied on a calendar-based schedule (24). In order to improve the efficacy of fungicide applications, studies on the effect of environmental conditions (e.g., temperature and leaf wetness duration) on the infection process were recently conducted in Ohio, and a model using temperature and leaf wetness duration was developed to predict infection (9). A disease warning (or predictive) system was developed using this model to predict potential disease intensity of Phomopsis cane and leaf spot throughout a season based on measured temperature and leaf wetness duration following rainfall.

A disease warning system can be used to reduce the number of applications of fungicide by eliminating unnecessary sprays, and also to identify periods for disease development (18, 35). Several warning systems for major grape diseases such as downy

13 mildew (Plasmopara viticola) (19), powdery mildew (Unicinula necator) (3, 11), and black rot (Guignardia bidwellii) (5, 11), have been developed and used, but none has been developed previously for Phomopsis cane and leaf spot. In the new warning system, fungicide treatments are applied after predicted infection events (i.e., periods with high potential for infection). Ideally, fungicides with curative activity should be used, if available (5, 6). However, many successful warning systems use only protectant fungicides (e.g., 1, 15). Presumably, post-infection applications of protectants limit subsequent infections after initial infections have occurred, or limit sporulation

(depending on fungicide’s mode of action). There are no labeled fungicides with known curative activity against P. viticola. However, in some pathosystems, a protective fungicide mixed with an adjuvant demonstrates curative (or post-infection) activity because adjuvant facilitates the uptake of fungicide by plant tissues (16, 30).

The objectives of this study were to: evaluate the disease warning system for

Phomopsis cane and leaf spot under field conditions by determining the efficacy of warning-system-based treatments consisting of fungicides and fungicide-adjuvant mixtures; and determine which prediction threshold should be used to base fungicide applications.

14 MATERIALS AND METHODS

The disease warning system. A mathematical model (equation 1) to predict disease intensity from average temperature (T) and wetness duration (W) during an inoculation period, assuming inoculum is available, was developed by Erincik et al. (9):

y = α ⋅t β ⋅(1− t)γ ⋅W δ (1) where y is disease intensity (disease severity per internode or number of lesions per leaf),

Tmin and Tmax are estimated potential minimum (5 °C) and maximum (35.5 °C) temperatures, t = (T-Tmin)/(Tmax-Tmin), and α (=1.8), β (=1.5), γ (=1.7), and δ (=1.1) were parameters estimated from their controlled-environment experiments (9). In the disease warning system, equation 1 was used with parameters for leaf disease severity of cultivar

'Seyval', that is highly susceptibility to the disease. Using equation 1 and assuming high inoculum density, predicted infection events were classified into three categories: light (y

< 30 lesions per leaf); moderate (30 lesions < y < 90 lesions); and high (y > 90 lesions).

Basically, these three categories indicate relative expected disease intensity based on a length of wetness duration and average temperature during the time period (Figure 2.1).

The specific values of y are not important since they are functions of inoculum density as well as environmental conditions. Thus, it is the relative value of y for different environmental conditions that matters.

Environmental conditions used in the warning system: temperature, leaf wetness, and precipitation, were monitored using electronic sensors attached to a datalogger

(Model CR23X micrologger, Campbell Scientific, Logan, Utah). Temperature was measured with a thermister (Model 107, Campbell Scientific, Logan, Utah) located 15 within the grape canopy (~1.7-m above ground). In order to protect it from direct sun light, a piece of plastic (half-pipe shaped, 2.5-cm wide, 12.5-cm long) was placed ~ 1-cm above the sensor. Wetness duration was measured with two flat circuit-board sensors

(Model 237, Campbell Scientific, Logan, Utah) located inside the grape canopy. Rainfall was measured with a tipping bucket rain gauge (Model TR-525M, Campbell Scientific,

Logan, Utah) located next to the datalogger.

Prior to each growing season, leaf wetness sensors were calibrated based on measured voltage changes (i.e., changes in resistance) with the degree of wetness, so that the data-logger recorded leaf wetness with a 0-100 linear scale (0 = dry, 100 = entire surface of the sensor is wet). When the rainfall sensor recorded precipitation (> 0.25- mm3 of rain) and one of the two wetness sensors scored more than 50, this was considered as the beginning of a wetness duration. The end of wetness duration was the time when the both sensors scored less than 50. However, if the conditions for the beginning of a new wetness event were met within a 4-h period from the end of a previous wetness event, the second wetness duration was combined with the first one to create a single wetness event. If rain was not detected, any recorded wetness was considered to be dew and was not used since the system assumes that rainfall is necessary to disseminate spores to infection sites.

Experiments were conducted over 4 yr, and in three of these years, two separate locations were used. Each experiment was a randomized complete block design with four replications. Because many vines experienced a high degree of disease severity in previous years, and some vines were inoculated in prior years, we assumed abundant inoculum was present. In 2001, a small preliminary trial was conducted in an

16 experimental vineyard near Wooster, OH, consisting of 16 vines (4 treatments and 4 replications each) of the cultivar Catawba (V. labrusca, planted in 1996, trained to the umbrella Kniffin system). There was 2.1 m between vines within rows and 3 m between rows. In 2002-04, a larger part of this vineyard as well as another vineyard near Wooster,

OH, of the cultivar Seyval (Vitis interspecific hybrid, planted in 1986, trained to the single cordon system) was used in separate experiments. Number of treatments varied each year: 36 vines (9 treatments and 4 replications) of ‘Catawba’ and ‘Seyval’ in 2002;

52 vines (13 treatments and 4 replications) of ‘Catawba’, and 40 vines (10 treatments and

4 replications) of ‘Seyval’ in 2003 and 2004.

Treatments. Fungicides studied in one or more years were: benomyl (Benlate,

DuPont, Wilmington, DE); mancozeb (Dithane, Dow Agrosciences LLC, Indianapolis,

IN); calcium polysulfide (liquid lime sulfur, Platte Chemical CO., Greeley, CO); thiophanate-methyl (Topsin-M, Cerexagri, Inc., King of Prussia, PA); azoxystrobin

(Abound, Syngenta Crop Protection, Inc, Greensboro, NC ); and myclobutanil (Nova,

Dow Agrosciences LLC, Indianapolis, IN). Adjuvants studied were JMS Stylet-Oil (JMS

Flower Farms, Inc, Vero Beach, FL), and Regulaid

(polyoxyethylenepolypropoxypropanol alkyl 2-ethoxyethanol/dihydroxy-propane, Kalo

Laboratories, Inc., Kansas City, MO). Stylet-Oil, a highly refined paraffinic oil, can be used for control of insects, mites, aphid-transmitted plant viruses, and various powdery mildew fungi (2, 21, 23, 30). Regulaid is a nonionic spreader-activator, often used for improving efficacy of foliar plant growth regulators (32). Doses of fungicides and adjuvants used in experiments are summarized in Figure 2.1.

17 Fungicide treatments were applied to vines in 946 L of water per hectare with an

11.3-L CO2-pressurized hand sprayer operated at 274 kPa pressure. Vines were sprayed to the point of runoff. Model-based treatments were applied after a predicted infection event. Each fungicide application was assumed to provide 7 days of protectant activity; thus, infection events that occurred within 7 days after application were not responded.

In 2001, benomyl (Benlate) mixed with 1% Stylet-Oil was applied based on the warning system. The benomyl + Stylet-Oil treatment was applied after predicted light or moderate infection periods. Mancozeb (Dithane) was applied on a 7-day protectant schedule for comparison. The warning system was operated from 10-cm shoot growth

[Eichhorn-Lorenz (E-L) stage 9] through the fruit set (E-L stage 29) (24).

In 2002, the fungicides thiophanate-methyl (Topsin-M) and mancozeb were each sprayed following a 7-day protectant schedule, and the same fungicides with an adjuvant, either Stylet-Oil or Regulaid, were applied after predicted light or moderate infection periods. The warning system was operated from 2.5-cm shoot growth (E-L stage 7) to the end of the bloom (E-L stage 27). Initiation of the warning system was shifted to an earlier growth stage due to the observation of symptoms on 10-cm shoots in the previous year.

In 2003 and 2004, mancozeb and calcium polysulfide (lime sulfur) were each sprayed following a 7-day protectant schedule. Model-based applications of mancozeb, thiophanate-methyl, myclobutanil (Nova), and azoxystrobin (Abound) were applied with an adjuvant (Regulaid) after predicted moderate infection periods. Mancozeb and thiophanate-methyl were also applied with Stylet-Oil after predicted moderate infection

18 periods. Mancozeb + Regulaid was also applied at predicted light and high infection periods. The warning system was operated from 2.5-cm shoot growth (E-L stage 7) to the end of the bloom (E-L stage 27).

Disease assessment. Disease severity and incidence were evaluated visually in late-June or early-July of each year. Specific dates were: 1 July 2001; 26 June 2002 for

‘Catawba’; 1 July 2002 for ‘Seyval’; 2 July 2002 for ‘Catawba’; 16 July 2003 for

‘Seyval’; 6 July 2004 for ‘Catawba’; and 8 July 2004 for ‘Seyval’. The five basal internodes and leaves on 10 randomly selected shoots per vine were evaluated. Direct estimation of percentage of diseased area was used for internode disease severity assessments. Disease severity of leaves was assessed by estimating the number of lesions on each leaf using a scale with seven levels (0 = no lesions; 1 = 0 to 10; 2 = 10 to 25; 3 =

25 to 50; 4 = 50 to 75; 5 = 75 to 100; and 6 = more than 100 lesions per leaf). Disease incidence was based on percentage of infected internodes or leaves per shoot using the same internodes or leaves used in the severity estimate.

Disease incidence for rachis-infection was assessed on 16 September 2003 and 1

September 2004. Rachis infection was not assessed in 2001 or 2002. Five rachises were randomly chosen from each vine and examined for presence of visual symptoms.

Analysis of data. Data were analyzed using a linear mixed model with PROC

MIXED of SAS in order to determine effects of treatments on disease intensity.

Treatments were considered as fixed effects and blocks as random effects. Values of internode and leaf disease incidences, and that of internode disease severities were

19 transformed using an angular transformation [ arcsin( proportion) ], and a square-root transformation was used for lesion counts on leaves. Statistical analyses and treatment comparisons were based on least square means of transformed values, which were then back-transformed after the analysis to obtain the presented means.

RESULTS

Preliminary study (2001). Fungicide application was started on 26 April and ended on 8 June. A total of four applications were made in response to predicted light infections events, three were made for moderate infection periods, and seven were made for the calendar-based 7-day protectant program (Table 2.2). The average time between the initiation of a predicted infection event and application of model-based treatments was 45 h.

Spraying benomyl with an adjuvant based on the warning system provided reasonable control of Phomopsis cane and leaf spot, with fewer number of fungicide applications compared to a 7-day protectant program (P≤0.05) (Table 2.2). Disease incidence and severity were significantly lower than in the untreated control, but were significantly higher than for the 7-day protectant program (P≤0.05). Generally, the effect of spray timings (predicted light or moderate infection events) was not significant. The trend was the same in both disease incidence and severity, although incidence was higher.

20 2002. Initiation of treatment application was on 23 April, and application were terminated on 28 June. A total of five fungicide applications were made in response to predicted infection periods (both light and moderate) compared to eleven in a calendar- based protectant program (7-day interval) in both the ‘Catawba' and 'Seyval' vineyards.

During this season, there were no predicted light infection events, i.e., all predicted events (when plants were not already protected) were either for moderate and/or high infection (Figure 2.2A). Thus, the number of fungicide applications were the same in the light and moderate treatments (Table 2.3). It took 49 h, on average, from the initiation of a predicted infection event until the application of treatments.

During 2002, mancozeb applications with an adjuvant (either Stylet-Oil or

Regulaid) based on predicted infection periods controlled the disease as well as a calendar-based 7-day schedule program, with fewer numbers of applications in the

‘Catawba’ vineyard (Table 2.3). All mancozeb treatments resulted in lower disease than in the control (Tables 2.3 and 2.4). Model-based treatments using thiophanate-methyl were not effective in controlling the disease, regardless of timing of applications, except for internode disease severity on both cultivars (Table 2.3). Moreover, a calendar-based protectant program of thiophanate-methyl generally provided control which was not significantly different from that of the untreated control (P≤0.05) (Table 2.3). Overall, the trends in both disease incidence and severity were similar, although with disease severity, differences among treatments were smaller. Generally, vines in the 'Catawba' vineyard had higher internode disease intensity (i.e., both incidence and severity) than vines in the 'Seyval' vineyard, whereas leaf disease intensity was higher in the ‘Seyval’ vineyard (Table 2.3).

21

2003. The first fungicide application was applied on 29 April for both vineyards, and the last application was made on 26 June (‘Catawba’ vineyard) and on 3 July

(‘Seyval’ vineyard). For the 'Catawba' vineyard, a total of six, three and two applications were made in response to predicted light, moderate, and high infection events, respectively (Table 2.4, Figure 2.2B). There were nine fungicide applications made in the 7-day interval calendar-based protectant program. The warning system was used for an additional week for the 'Seyval' vineyard due to relatively slow development of vines.

Five, four, and three applications were made for predicted light, moderate, and high infection periods, respectively (Table 2.4). An average response time between the beginning of the predicted infection event and application of treatment was 30 h. A calendar-based protectant treatment resulted in ten fungicide applications (Table 2.4).

In 2003, there were differences in number of applications made between the predicted light and moderate infection events, and these differences generally were reflected in disease incidence and severity (Table 2.4). There was a tendency for incremental increases of disease intensity from light, to moderate, and to high infection- event-treatments. However, means were not significantly different between the light and moderate treatments (Table 2.4). Mean disease intensity for mancozeb + Regulaid applications based on predicted high disease infection events was significantly higher than for treatments based on either light or moderate disease infection event (Table 2.4).

When mancozeb alone was applied based on predicted medium infection events, it

22 provided moderate control. However, in the 'Catawba' vineyard, it resulted in significantly lower disease intensity than that obtained with mancozeb + Regulaid treatment (P≤0.05) (Table 2.4).

Applications of thiophanate-methyl + Regulaid based on predicted medium infection events in the ‘Catawba’ vineyard provided similar level of control to that obtained with mancozeb + Regulaid application based on the same conditions (Figure

2.4); however, thiophanate-methyl + Regulaid provided significantly less control than mancozeb + Regulaid in the ‘Seyval’ vineyard (Table 2.4).

Use of Regulaid with fungicides generally provided significantly better control than use of Stylet-Oil (Table 2.4). Other disease-warning-based treatments provided similar control to thiophanate-methyl + Stylet-Oil on ‘Catawba’ vines and to thiophanate- methyl + Regulaid on ‘Seyval’ vines (Table 2.4). Comparison between calendar- and disease-warning-based applications of calcium polysulfide showed that the latter provided significantly less control (Table 2.4). As found in 2002, the 'Catawba' vineyard generally had higher internode disease intensity than the 'Seyval' vineyard in 2003.

2004. The first fungicide application was on 6 May and the last was on 17 June.

Due to relatively high temperature during the last week of April, vines grew more rapidly than the normal, and thus the first fungicide application was made at close to 10-cm shoot growth (E-L stage 9) instead of at 2.5-cm growth (E-L stage 7) as intended. A total of three fungicide applications were made in response to predicted light and moderate infection periods, compared to seven with the protectant program in both the 'Catawba'

23 and 'Seyval' vineyards (Figure 2.2C). Model-based treatments applied based on predicted infection light and medium events had the same number of applications, even though, some applications were on different days (Figure 2.2C). Predicted high infection periods in the 'Catawba' vineyard resulted in one application compared to two applications in the

‘Seyval’ vineyard, due to differences in rainfall patterns between the two vineyards

(Table 2.5). An average response time from the initiation of the predicted infection event and application of treatment was 37 h.

Results from 2004 season were similar to the previous 2 yr, but disease intensity was relatively high, apparently due to frequent rains with relatively high temperature in early May, and later initiation of sprays. For instance, there was a rain event on 26 April, a predicted medium infection event, with no protection by fungicides (Table 2.5, Figure

2.2C). In the ‘Catawba’ vineyard, mancozeb + Regulaid applied based on predicted light and medium infection events had significantly lower disease intensity than that in mancozeb + Regulaid treatment based on predicted high infection events, but means for the light- and medium-infection-event treatments were not different (Table 2.5). On the other hand, mean disease values between light-, medium-, and high-infection-event treatments were not significantly different, in general, in the ‘Seyval’ vineyard; however, the vines received two applications for the high events instead of one with the ‘Catawba’ vineyard (Table 2.5). Although most of warning-system-based treatments applied at either predicted light or medium infection events had significantly lower disease intensity than that in the untreated control, with one exception, the model-based treatments resulted in significantly higher disease intensity than that in a 7-day calendar-based schedule of mancozeb. The exception was for mancozeb with Regulaid based on

24 predicted medium infection events in the ‘Catawba’ vineyard that provided a similar degree of control to that of the calendar-based protectant application of mancozeb (Table

2.5).

Although there was significantly less disease than in the untreated control, calendar- and model-based treatments of calcium polysulfide provided relatively poor control (Table 2.5). There was generally higher internode disease intensity in the

‘Catawba’ vineyard than in the 'Seyval' vineyard, but there was higher leaf disease intensity in the ’Seyval’ than in the ‘Catawba’ vineyard (Table 2.5).

Rachis infection. Rachis disease incidence was measured in 2003 and 2004

(Table 2.6). Results of rachis infection on ‘Catawba’ were similar to the results from leaf and internode disease intensity, although magnitude of control with the warning system was less (Tables 2.4-2.6). Typically, a 7-day protectant schedule showed good control

(low disease intensity), and warning-system-based treatments applied based on predicted light or moderate infection events exhibited moderate-to-poor control (Figure 2.6), even though leaf and internode disease control was good. With warning-system-based mancozeb + Regulaid treatments, disease incidence for the predicted high infection events tended to have higher disease intensity than for the medium or light predicted infection events. Several treatments, thiophanate-methyl + Regulaid (medium event treatment), myclobutanil + Regulaid (medium event treatment), and calcium polysulfide

(medium event treatment, and calendar-based 7-day treatment) showed good-to-moderate

25 control in 2003, but efficacy of these treatments was lower in 2004. However, in both years, there were few rachis infections on 'Seyval' and disease intensity did not vary among any of the treatments (data not shown).

DISCUSSION

Over 4 years, a disease warning system for Phomopsis cane and leaf spot of grape was tested in Ohio by applying fungicides and fungicide-adjuvant mixtures based on predicted infection events. Our findings suggest that spraying grape vines with certain protectant fungicide-adjuvant combinations in response to predicted light or medium infection events can generally provide control comparable to that obtained with a calendar-based 7-day protectant schedule, yet with substantially fewer fungicide applications (Tables 2.2-2.6). For example, when mancozeb + Regulaid was applied based on either predicted low or medium infection events, disease severity and incidence were similar to that obtained from a standard 7-day protective spray program of mancozeb in both ‘Catawba’ and ‘Seyval’ vineyards in 2002 and in 2003, and in the

‘Catawba’ vineyard in 2004 (Tables 2.3-2.6).

In addition to mancozeb, fungicides with different modes of actions mixed with

Regulaid were examined for their efficacy when used with the warning system. Although there were some yearly fluctuations in the degree of control, benzimidazols [benomyl

(Benlate), thiophanate-methyl (Topsin-M)], a DMI-fungicide [myclobutanil (Nova)], and a strobilurin [azoxystrobin (Abound)] all resulted in poorer control of P. viticola

26 compared with that obtained with mancozeb treatments. In a separate controlled environment study, a pre-inoculation application of mancozeb, thiophanate-methyl, and azoxystrobin provided good control (Chapter 3); however, mancozeb often provided better control than the other two in the field (Tables 2.3-2.6).

The other fungicide studied was calcium polysulfide (lime sulfur). One of advantages of calcium polysulfide is that it is registered for use in organic grape production (7). Applications based on predicted medium infection events resulted in the degree of control similar to that obtained with thiophanate-methyl, myclobutanil, and azoxystrobin. That is, it provided moderate-to-poor control (Tables 2.3-2.6). In the

‘Catawba’ vineyard, a calendar-based 7-day schedule application of calcium polysulfide was also examined in 2003 and 2004 (Tables 2.3-2.4), which tended to provide significantly better control than model-based applications in 2003 (Table 2.4), but not in

2004 (Table 2.5).

In the 'Catawba' vineyard, the warning-system-based treatment of mancozeb alone based on predicted medium infection events often had significantly higher mean disease intensity than for the model-based application of mancozeb + Regulaid applied based on predicted medium infection events (Tables 2.4 and 2.5). In the 'Seyval' vineyard, this trend was not as clear, and disease intensity between these two treatments was not significantly different (Tables 2.4 and 2.5). Although not striking, the differences in disease intensity for the treatments with or without an adjuvant may indicate effects of the curative activities or other beneficial effects of fungicide-adjuvant mixtures.

Adjuvants are often added to herbicide spray in order to increase efficacy of herbicides by facilitating penetration and movement of chemical in plant tissues (14, 30). In a

27 similar manner, use of adjuvant could enhance penetration of fungicidal chemicals into plant tissues which could give some level of curative activity to preventative fungicides

(12, 13, 25, 30). Examples include: Baycor 25W (bitertanol, Bayer South Africa) mixed with Agridex (paraffin base petroleum oil, polyoxythylated polyol fatty acid ester and polyol fatty acid ester constituents, Bayer South Africa) for apple scab (Venturia inaequalis) control (28); and 1-(4-chlorobenzyl)-4-phenylpiperidine (Fluorochem Ltd.,

UK) and Dobanol 91-6 (alcohol ethoxylate, Shell Chemicals, UK) for powdery mildew on barley (Erysiphe graminis) control (13).

There are other beneficial properties of adjuvants in addition to possible curative activities. For example, application of mancozeb with Stylet-Oil is known to improve its rain-fastness (ability to retain on the plant surface) (17); also adjuvants, in general, improve spray efficiency by reducing drift (34) and/or increasing coverage by creating finer spray droplets (33, 34). Thus, increased efficacy observed in fungicide-adjuvant mixture applications in the field (Tables 2.3-2.5) is probably due to some combination of these effects. Other studies show that some adjuvant-fungicide combinations improve efficacy, depending on the pathosystem (2, 13, 30). In our experiments, mancozeb +

Regulaid tended to provide better control than other fungicide-adjuvant combinations

(Tables 2.3-2.5). Moreover, when mancozeb or thiophanate-methyl was applied with

Regulaid based on the warning-system, both consistently provided significantly better control than when the fungicide was applied with Stylet-Oil (Tables 2.3-2.6). This may indicate that, with tested fungicides, Regulaid facilitates better adhesion or coverage of fungicide than Stylet-Oil does.

28

Disease development was slightly different in the two vineyards. The 'Catawba' vineyard generally had higher disease intensity than the 'Seyval' vineyard (Tables 2.3-2.5).

There was generally higher internode disease intensity in the ‘Catawba’ vineyard whereas higher leaf disease intensity was observed in the ‘Seyval’ vineyard. Moreover, differences in internode disease intensity means were large with ‘Catawba’ vines, whereas differences of leaf disease intensity among treatment means were high in

‘Seyval’ (Tables 2.3-2.5). These results are probably due to differences in resistance between cultivars and plant tissues, but there is no information on this (24). Inoculum level could also account for the differences, because the ‘Catawba’ vineyard has been known to have higher Phomopsis incidence than the ‘Seyval’ vineyard prior to the experiments (Ellis, unpublished). Also differences in vine training system, umbrella

Kniffin for ‘Catawba’ and single cordon for ‘Seyval’, or their growing habits, (e.g.,

‘Catawba’ tends to grow more vigorously at the beginning of the season than ‘Seyval’), may have affected the dispersal of conidia to healthy tissues. In addition, weather conditions between these two vineyards varied, sufficiently to have different fungicide application numbers in 2003 and 2004 for the warning-system based treatments (Tables

2.4 and 2.5).

Overall, disease severity was reduced more than incidence by the effective treatments. An example would be the mancozeb + Regulaid disease-warning-system treatment on ‘Catawba’ applied based on predicted medium infection events in 2003 and

2004. The reduction in disease incidence compared to that in untreated control was between 41 and 53% in 2003- 2004 whereas reduction in disease severity was 84 to 89 %

29 (Tables 2.4 and 2.5). Similar results were found in the calendar-based treatments. The large reductions in internode disease severity are especially noteworthy, because infected canes were likely to become the source of the inoculum in subsequent years (24).

In general, based on three thresholds for spraying based on the warning system

(low, medium, and high), use of predicted medium infection event provided the most degree of control with the fewest number of sprays. Moreover, use of fungicides with

Regulaid is recommended over sprays with fungicide alone, since use of mancozeb with

Regulaid showed better control overall (Tables 2.3-2.5). The performance of this specific treatment versus a calendar-based protectant spray program of mancozeb can be seen by calculating percent control for each treatment: PC = 100[(C-T)/C], where C is disease intensity (incidence or severity) for the untreated control and T is disease intensity for the treatment (either calendar- or warning-system-based).

Figure 2.3 shows that percentage of control achieved by mancozeb + Regulaid applied based on predicted medium infection events and that by mancozeb applied on the calendar-based 7-day schedule across 3 years and two vineyards. The trend was that the calendar-based treatments provided better percent control than the warning-system-based treatments, with average percent control of 55 and 80 % for incidence and severity, respectively, with calendar-based applications, and 36 and 60 % for incidence and severity, respectively, with warning-system-based applications. However, there was a two-to-three fold difference in the number of fungicide applications between the warning-system- and calendar-based treatments (Tables 2-5). Thus, percent control per

30 spray was calculated as: PCPS=PC/(# of sprays) for the calendar-based protectant schedule application of mancozeb and for the predicted medium infection event application of mancozeb + Regulaid.

Percent of disease control per application was higher for the warning-system based treatment than that for the calendar-based treatment across the 3 years and two vineyards (Figure 2.4). Average percent control per spray was 6 and 9 for incidence and severity, respectively, with calendar-based applications, and 10 and 18 for incidence and severity, respectively, with warning-system-based applications. This indicates that some of applications made for the calendar-based treatment were probably not contributing to the overall reduction of the disease.

Efficacy of treatments on rachis infection of P. viticola was examined in 2003 and

2004. In both 2003 and 2004, application of fungicides ceased at the end of the bloom for all treatments. Thus, these experiments provided information on purely early season fungicides on end-of-season control of rachis infection. In the ‘Catawba’ vineyard, the trend of disease control on rachis disease was similar to that found for leaf and internode disease. Calendar-based 7-day protective applications of mancozeb provided good protection, but all disease-warning-system treatments had poorer control (Table 2.6).

Notable differences between the 2 years were for the efficacy of thiophanate-methyl +

Stylet-Oil, myclobutanil + Regulaid, and calcium polysulfide applied based on predicted medium infection events, and calcium polysulfide on a calendar-based 7-day program.

These treatments exhibited a two-to-three fold increase of disease incidence from 2003 to

2004 (Table 2.6). Since overall disease incidence was higher in 2004, treatments with less effectiveness were apparent in disease intensity of rachises. In the ‘Seyval’ vineyard,

31 disease incidence was lower in both years (23.7% and 40.0%, in 2003 and 2004, respectively), and treatment effects were not apparent (Nita, unpublished). The different rachis infection results between the ‘Catawba’ and ‘Seyval’ vineyards could be due to different degrees of resistance, local weather conditions, or differences in amount of inoculum. Thus, although the disease warning system generally provided good control on leaves and internodes, rachis infection control was less satisfactory.

Work in NY, indicates that the critical period to control the Phomopsis cane and leaf spot with protectant fungicide program is very early in the growing season, starting from bud-break (Wilcox, personal communication). One study indicated that fruit rot was significantly decreased by two applications of captan during pre-bloom (at 2.5-cm and 12.7-cm shoot growth or at E-L stages 7 and 12) (27). Also, observations of spore production in a study in California indicate that there are high inoculum doses present in the early spring, followed by a steady decline over the course of the season (4). Results from these studies indicate the importance of fungicide coverage during the early growth stages of grape. This may explain the higher disease intensity observed in 2004, where vines were unprotected between E-L stages 5-7 due to rapid growth, compared to earlier years (Tables 2.3-2.6). Also, this indicates that later season sprays with the calendar- based treatment may not have been needed considering the higher percentage of disease control per spray by the warning-system-based treatments than the calendar-based treatments (Figure 2.4).

Overall, the calendar-based fungicide applications and the warning system can provide a similar degree of control, but the latter system with fewer sprays (Tables 2.3-

2.6). The economic importance of reducing the number of fungicide applications by

32 using a disease forecasting or warning system has been discussed in many articles (e.g., 1,

10). Other advantages of use of a warning system include beneficial environmental effects of reduced fungicide application, reduced exposure to humans, and slower development of fungicide resistance by pathogens (1, 10, 18, 19, 31, 35).

33

Fungicide (trade name) Abbreviationa Doseb/ha a.i.c /ha Mancozeb (Dithane) M 4.5 kg 3.60 kg Thiophanate-methyl (Topsin-M) T 1.7 kg 1.19 kg Myclobutanil (Nova) N 0.3 kg 0.12 kg Azoxystrobin (Abound) A 1.1 m3 0.25 m3 Calcium polysulfide (Lime sulfur) LS 4.7 m3 1.36 m3 JMS Stylet-Oil S 3.8 m3 3.69 m3 Regulaid R 2.4 m3 2.17 m3

a Used in Tables 2.3-2.6 b Dose of product (not active ingredient)/ha c Active ingredient/ha

Table 2.1. Fungicides and adjuvant used in experiments and doses

34

Incidence Severity Leaf Node Leaf Node Treatmenta Eventb #%c % Ld % Benomyl + Stylet-Oil Light 3 13.3 be 64.7 b 1.7 b 2.0 b Benomyl + Stylet-Oil Med 3 12.9 b 54.2 b 1.1 b 1.2 b Mancozeb 7-d 7 3.4 c 5.9 c 0.2 c 0.1 c Control --f 50.2 a 81.7 a 10.5 a 5.0 a

a Fungicide doses are described in Table 2.1. b Treatments were applied based on either the warning system in respect to predicted light, med (medium), or high predicted infection events, based on observed weather conditions, or based on a calendar-based protectant program (7-d) c % = back-transformed value of mean disease incidence or severity. Disease severity was estimated by visual assessment of proportion of internode covered by lesions. Disease incidence was the proportion of leaves or internodes showing visible symptoms of Phomopsis cane and leaf spot to healthy leaves or internodes. Disease incidence or severity were angular transformed, [ arcsin( proportion) ], before analysis with Proc MIXED of SAS. d L = back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. e Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. f Untreated control

Table 2.2. Evaluation of warning-system-based and calendar-based protectant spray programs for control of Phomopsis cane and leaf spot of grape (caused by Phomopsis viticola) in 2001

35

‘Catawba’ ‘Seyval’ Incidence Severity Incidence Severity Leaf Node Leaf Node Leaf Node Leaf Node Trta Eventb #c %d %d Le %d #c %d %d Le %d

M + R Light 5 9.3 df 88.8 b 0.1 f 1.2 d 5 88.9 b 13.0 d 5.1 bc 0.1 d

M + S Light 5 14.7 cd 42.2 c 0.2 ef 0.4 ef 5 65.3 cd 13.1 d 2.1 f 0.1 d

M + S Med 5 15.4 cd 32.9 cd 0.3 df 0.4 ef 5 99.7 a 24.2 cd 7.6 a 0.2 cd

T + S Light 5 40.6 ab 99.5 a 1.6 a 4.2 c 5 86.4 b 48.7 b 4.5 bcd 0.8 b

T + S Med 5 36.2 ab 100.0 a 1.7 a 4.3 c 5 86.5 b 48.5 b 3.8 ce 0.8 b

T 7-d 11 45.7 b 97.7 a 1.5 ab 6.7 b 11 91.1 b 59.8 b 6.0 ab 0.9 b

M 7-d 11 19.6 cd 22.6 d 0.4 df 0.1 f 11 54.5 d 21.5 cd 1.8 f 0.1 cd

C --g 0 35.9 ab 74.7 b 1.0 bc 1.3 d 0 68.9 cd 52.3 b 3.5 de 0.9 b

a Trt = treatment. Fungicide doses and abbreviations are described in Table 2.1. C = untreated control b Treatments were applied based on either the warning system in respect to predicted light, med (medium), or high predicted infection events, based on observed weather conditions, or based on a calendar-based protectant program (7-d) c # = number of fungicide applications d % = back-transformed value of mean disease incidence or severity. Disease severity was estimated by visual assessment of proportion of internode covered by lesions. Disease incidence was the proportion of leaves or internodes showing visible symptoms of Phomopsis cane and leaf spot to healthy leaves or internodes. Disease incidence or severity were angular transformed, [arcsin( proportion) ], before analysis with Proc MIXED of SAS. e L= back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. f Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. g Untreated control

Table 2.3. Evaluation of warning-system-based and calendar-based protectant spray programs for control of Phomopsis cane and leaf spot of grape (caused by Phomopsis viticola) in two vineyards (‘Catawba’ and ‘Seyval’) in 2002

36

‘Catawba’ ‘Seyval’ Incidence Severity Incidence Severity Leaf Node Leaf Node Leaf Node Leaf Node Trta Eventb #c %d %d Le %d #c %d %d Le %d

M + R Light 6 24.7 ef 30.0 d 0.5 fg 0.2 g 5 64.3 cf 35.7 de 0.6 c 0.4 bc

M + R Med 3 33.8 de 40.5 d 1.3 ef 0.6 ef 4 58.4 c 30.0 e 0.7 c 0.2 c

M + R High 2 45.5 cd 77.8 ab 3.5 c 2.2 b 3 86.2 b 50.5 bcd 1.2 b 0.6 bd

M + S Med 3 53.5 bc 78.6 ab 3.5 bc 1.4 bc - - - - -

M Med 3 42.5 cd 60.4 c 3.6 bc 1.4 cd 4 60.4 c 39.5 ce 0.6 c 0.4 bc

T+ R Med 3 33.8 de 39.5 d 1.6 de 0.5 f 4 84.8 b 57.4 ab 1.3 ab 0.6 d

T+ S Med 3 59.4 b 71.7 bc 4.1 bc 1.4 c - - - - -

N + R Med 3 51.5 bc 68.0 bc 5.7 b 1.7 bc 4 87.5 b 52.5 ac 1.3 ab 0.8 d

A + R Med 3 51.5 bc 65.2 bc 3.1 c 0.8 ef 4 82.6 b 43.5 bce 1.1 b 0.5 bc

LS Med 3 46.5 bd 60.4 c 2.8 cd 0.8 de 4 86.2 b 53.5 ac 1.2 b 0.8 d

LS 7-d 9 24.7 e 37.6 d 1.2 ef 0.5 f - - - - -

M 7-d 9 13.1 f 10.5 e 0.3 g 0.0 g 10 56.4 c 36.6 de 0.6 c 0.4 bc

C --g 0 69.9 a 86.2 a 12.4 a 4.3 a 0 96.8 a 67.1 a 1.5 a 1.2 a

a Trt = treatment. Fungicide doses and abbreviations are described in Table 2.1. C = untreated control b Treatments were applied based on either the warning system in respect to predicted light, med (medium), or high predicted infection events, based on observed weather conditions, or based on a calendar-based protectant program (7-d) c # = number of fungicide applications d % = back-transformed value of mean disease incidence or severity. Disease severity was estimated by visual assessment of proportion of internode covered by lesions. Disease incidence was the proportion of leaves or internodes showing visible symptoms of Phomopsis cane and leaf spot to healthy leaves or internodes. Disease incidence or severity were angular transformed, [ arcsin( proportion) ], before analysis with Proc MIXED of SAS. e L = back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. f Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. g Untreated control

Table 2.4. Evaluation of warning-system-based and calendar-based protectant spray programs for control of Phomopsis cane and leaf spot of grape (caused by Phomopsis viticola) in two vineyards (‘Catawba’ and ‘Seyval’) in 2003

37

‘Catawba’ ‘Seyval’ Incidence Severity Incidence Severity Leaf Node Leaf Node Leaf Node Leaf Node Trta Eventb #c %d %d Le %d #c %d %d Le %d

M + R Light 3 65.9 gf 63.2 ef 3.0 ef 1.8 de 3 86.0 df 39.5 bd 8.6 d 0.4 cd

M + R Med 3 51.0 f 53.3 f 2.0 fg 1.5 e 3 89.3 cd 50.8 ab 11.3 cd 0.6 bc

M + R High 1 90.5 b 94.4 b 9.3 b 6.5 b 2 84.6 d 54.8 ab 10.0 d 0.9 ab

M + S Med 3 80.2 cd 74.3 de 4.8 d 1.6 de - - - - - M Med 3 77.4 de 73.0 de 4.3 de 1.8 de 3 91.7 bd 44.5 bc 8.7 d 0.3 de

T+ R Med 3 65.2 ef 80.1 cd 3.1 ef 2.4 d 3 94.7 bc 47.9 b 14.8 bc 0.4 ce

T+ S Med 3 90.2 bc 80.5 cd 7.1 c 2.4 d - - - - -

N + R Med 3 71.1 de 89.3 bc 4.5 d 3.7 c 3 95.5 bc 43.5 bc 13.7 bc 0.3 de

A + R Med 3 74.3 de 78.9 cd 3.8 de 2.4 d 3 95.3 bc 44.0 bc 13.7 bc 0.5 cd LS Med 3 91.1 b 95.3 b 7.7 bc 5.6 b 3 97.1 ab 31.0 cd 15.4 b 0.3 de

LS 7-d 7 88.6 bc 93.6 b 7.6 bc 6.1 b - - - - - M 7-d 7 52.1 fg 33.3 g 1.7 g 0.7 f 7 71.4 e 25.9 d 4.7 e 0.2 e C --g 0 98.1 a 100.0 a 12.8 a 11.7 a 0 99.4 a 66.1 a 27.8 a 0.9 a

a Trt = treatment. Fungicide doses and abbreviations are described in Table 2.1. C = untreated control b Treatments were applied based on either the warning system in respect to predicted light, med (medium), or high predicted infection events, based on observed weather conditions, or based on a calendar-based protectant program (7-d) c # = number of fungicide applications d % = back-transformed value of mean disease incidence or severity. Disease severity was estimated by visual assessment of proportion of internode covered by lesions. Disease incidence was the proportion of leaves or internodes showing visible symptoms of Phomopsis cane and leaf spot to healthy leaves or internodes. Disease incidence or severity were angular transformed,arcsin( proportion) , before analysis with Proc MIXED of SAS. e L= back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. f Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. g Untreated control

Table 2.5. Evaluation of warning-system-based and calendar-based protectant spray programs for control of Phomopsis cane and leaf spot of grape (caused by Phomopsis viticola) in two vineyards (‘Catawba’ and ‘Seyval’) in 2004

38

2003 2004 incidence incidence Treatmenta Eventb %c %c M + R Light 5.3 ee 11.6 ef M + R Med 55.5 ab 24.6 de M + R High 80.8 a 65.8 bc M + S Med 11.3 de 34.7 ce M Med 39.5 ae 39.0 ce T+ R Med 28.6 bcde 29.1 de T+ S Med 15.4 bde 44.5 cd N + R Med 5.3 e 34.2 ce A + R Med 15.4 bde 28.6 de LS Med 45.0 ad 84.6 ab

LS 7-d 8.0 de 55.5 bcd M 7-d 5.3 e 1.3 f C --e 71.4 ac 97.1 a

a Trt = treatment. Fungicide doses and abbreviations are described in Table 2.1. C = untreated control b Treatments were applied based on either the warning system in respect to predicted light, med (medium), or high predicted infection events, based on observed weather conditions, or based on a calendar-based protectant program (7-d) c % = back-transformed value of mean disease incidence. Disease incidence was the proportion of rachises showing visible symptoms of Phomopsis cane and leaf spot to healthy leaves or internodes. Disease incidence were angular transformed, [ arcsin( proportion) ], before analysis with Proc MIXED of SAS. d Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. e Untreated control

Table 2.6. Evaluation of warning-system-based and calendar-based protectant spray programs for control of Phomopsis cane and leaf spot of grape rachis infections (caused by Phomopsis viticola) in the ‘Catawba’ vineyard in 2003 and 2004

39

140 5 h 120 10 h 15 h High 100 20 h

80

Moderate 60

40

Light 20 Number of lesions per leaf 0

5 101520253035 Temperature (oC)

Figure 2.1. Predicted number of lesion per leaf based on equation 1, a model developed by Erincik et al. (9). Curved lines are predictions for combinations of temperature and wetness duration. Horizontal lines are thresholds categories tested with the disease warning system.

40

50 30 Light Light Light Light Light Med Med Med Med Med

25 C) 40 o

20 30 A 15 20 2002 10

10 5 Average temperature ( temperature Average Leaf wetness duration (hr) duration wetness Leaf 0 0 50 30 Legend Light Light Light Light Light High leaf wetness

25 C)

Med o 40 Med Med High Average temperature 20 30 Light Model-based Med treatments 15 High applied 20 B Responded 2003 10 infection event Length of 10 fungicide 5 coverage (7-day) Average temperature ( temperature Average Leaf wetness duration (hr) duration wetness Leaf 0 0 50 30 Light Light Light Med Med High C)

25 o 40 Med C 20 30 2004 15 20 10

10 5 Average temperature ( temperature Average Leaf wetness duration (hr)

0 0 4/25 5/5 5/15 5/25 6/4 6/14 6/24 Day of the year

Figure 2.2. Plot of leaf wetness events and average temperature during the events, applied warning-system-based treatments (responded infection events are shown with arrows), and theoretical coverage of the fungicide application (a 7-day period). Data are for ‘Catawba’ vineyard in A, 2002, B, 2003, and C, 2004. 41

100.0 Mancozeb+ Regulaid (Med) 90.0 Mancozeb (Calender-based)

80.0

70.0

60.0

50.0

40.0 % diseasecontrol 30.0

20.0

10.0

0.0 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 Catawba Seyval Catawba Seyval Catawba Seyval Catawba Seyval Leaf, incidence Node, incidence Leaf, Severity Internode, Severity

Figure 2.3. Efficacy of the warning-system-based treatment of mancozeb with Regulaid and calendar-based treatment of mancozeb relative to untreated control. Bars represent percent control: PC = 100[(C-T)/C], where C is disease intensity (incidence or severity) for the untreated control and T is disease intensity for the treatment (either calendar-based mancozeb application or warning-system-based mancozeb + Regulaid application on predicted medium infection events).

42

30.0 Mancozeb+ Regulaid (Med) Mancozeb (Calender-based)

25.0

20.0

15.0

10.0 % disease control per spray 5.0

0.0 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 2002 2003 2004 Catawba Seyval Catawba Seyval Catawba Seyval Catawba Seyval Leaf, incidence Node, incidence Leaf, Severity Internode, Severity

Figure 2.4. Efficacy of the warning-system-based treatment of mancozeb with Regulaid and calendar-based treatment of mancozeb relative to untreated control. Bars represent percentage of disease control by treatments per spray: PCPS=PC/(# of sprays), where PC is percent control (Figure 2.3), for the calendar-based protectant schedule with mancozeb and for predicted medium infection event application of mancozeb + Regulaid.

43

REFERENCES

1. Beresford, R. M., and Manktelow, D. W. L. 1994. Economics of reducing fungicide use by weather-based disease forecasts for control of Venturia inaequalis in apples. New Zealand Journal of Crop and Horticultural Science 22 2:113-120. 2. Calpouzos, L. 1966. Action of oil in the control of plant disease. Annual Review of Phytopathology 4:369-390. 3. Correiar, B. R. 1999. Use of automated weather stations and the Gubler-Thomaz model for control of powdery mildew (Uncinula nacator) in California grapevines. Summa Phytopathologica 25 1:70-73. 4. Cucuzza, J. D., and Sall, M. A. 1982. Phomopsis cane and leaf spot disease of grape vine: Effect of chemical treatments on inoculum level, disease severity, and yield. Plant Disease 66:794-797. 5. Ellis, M. A., Madden, L. V., and Wilson, L. L. 1986. Electronic grape black rot predictor for scheduling fungicides with curative activity. Plant Disease 70 10:938-940. 6. Ellis, M. A., Madden, L. V., and Wilson, L. L. 1984. Evaluation of an electronic apple scab predictor for scheduling fungicides with curative activity. Plant Disease 68:1055-1057. 7. Ellis, M. A., and Nita, M. 2004. Disease management guidelines for organic grape production in the Midwest. Plant Pathology Department Series 121 The Ohio State University OARDC/OSUE, Wooster, OH 8. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2001. Effect of growth stage on susceptibility of grape berry and rachis tissues to infection by Phomopsis viticola. Plant Disease 85 5:517-520. 9. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2003. Temperature and wetness-duration requirements for grape leaf and cane infection by Phomopsis viticola. Plant Disease 87:832-840. 10. Funt, R. C., Ellis, M. A., and Madden, L. V. 1990. Economic analysis of protectant and disease-forecast-based fungicide spray programs for control of apple scab and grape black rot in Ohio. Plant Disease 74 9:638-642. 11. Gadoury, D. M. 1993. Integrating management decisions for seveal pests in fruit production. Plant Disease 77:299-302. 12. Gent, D. H., Schwartz, H. F., and Nissen, S. J. 2003. Effect of commercial adjuvents on vegitable crop fungicide coverage, absorption, and efficacy. Plant Disease 87:591-597. 13. Grayson, B. T., Price, P. J., and Walter, D. 1997. Effects of adjuvants on the performance of a novel powdery mildew fungicide, 1-(4-chlorobenzyl)-4- phenylpiperidine. Pesticide Science 51 2:206-212. 14. Grayson, B. T., Webb, J. D., and Pack, S. E. 1993. Investigation of an emulsifiable oil adjuvant and its components on the activity of a new grass herbicide by factorial experimentation. Pesticide Science 37 2:127-131.

44 15. Grunwald, N. J., Romero Montes, G., Lozoya Saldana, H., Rubio Covarrubias, O. A., and Fry, W. E. 2002. Potato late blight management in the Toluca valley: field validation of SimCast modified for cultivars with high field resistance. Plant Disease 86 10:1163-1168. 16. Holloway, P. J., Ellis, M. C. B., Webb, D. A., Western, N. M., Tuck, C. R., Hayes, A. L., and Miller, P. C. H. 2000. Effects of some agricultural tank-mix adjuvants on the deposition efficiency of aqueous sprays on foliage. Crop Protection 19 1:27-37. 17. Kudsk, P., Mathiassen, S. K., and Kirknel, E. 1991. Influence of formulations and adjuvants on the rainfastness of maneb and mancozeb on pea and potato. Pesticide Science 33 1:57-71. 18. Madden, L. V., and Ellis, M. A. 1988. How to develop plant disease forecasters. 191-208 in: Experimental techniques in plant disease epidemiology Kranz, J. and Rotem, J. Springer-Verlag, Berlin 19. Madden, L. V., Ellis, M. A., Lalancette, N., Hughes, G., and Wilson, L. L. 2000. Evaluation of a disease warning system for downy mildew of grapes. Plant Disease 84 5:549-554. 20. Malathrakis, N. E., and Baltzakis, N. G. 1976. Control of Dead-arm of Grapes. Poljoprivredna znanstvena smotra - Agriculture Conspectus Scientificus 39 49:261-269. 21. McGrath, M. T., and Shishkoff, N. 2000. Control of cucurbit powdery mildew with JMS Stylet-Oil. Plant Disease 84 9:989-993. 22. Mostert, L., Crous, P. W., and Petrini, O. 2000. Endophytic fungi associated with shoots and leaves of Vitis vinifera, with specific reference to the Phomopsis viticola complex. Sydowia 52 1:46-58. 23. Northover, J., and Schneider, K. E. 1996. Physical modes of action of petroleum and plant oils on powdery and downy mildews of grapevines. Plant Disease 80 5:544-550. 24. Pearson, R. C., and Goheen, A. C. 1988. Compendium of Grape Diseases. 25. Percich, J. A., and Nickelson, L. J. 1982. Evaluation of several fungicides and adjuvant materials for control of brown spot of wild rice. Plant Disease 66 11:1001-1003. 26. Phillips, A. J. L. 2000. Excoriose, cane blight and related diseases of grapevines: a taxonomic review of the pathogens. Phytopathol. Mediterr. 39:341-356. 27. Pscheidt, J. W., and Pearson, R. C. 1989. Effect of grapevine training systems and pruning practices on occurrence of Phomopsis cane and leaf spot. Plant Disease 73 10:825-828. 28. Schwabe, W. F. S., and Jones, A. L. 1983. Apple scab control with bitertanol as influenced by adjuvant addition. Plant Disease 67:1371-1373. 29. Sergeeva, V., Nair, N. G., Barchia, I., Priest, M., and Spooner-Hart, R. 2003. Germination of β conidia of Phomopsis viticola. Austrarian Plant Pathology 32:105-107. 30. Steurbaut, W. 1993. Adjuvants for use with foliar fungicides. Pesticide Science 38:85-91.

45 31. Sutton, T. B. 1996. Changing options for the control of deciduous fruit tree diseases. Annual Review of Phytopathology 34:527-547. 32. Thomson, L. A. 1986. A guide to agricultural spray adjuvants used in the united states. Thomson publications, Fresno, CA 33. Webb, D. A., Holloway, P. J., and Western, N. M. 1999. Effects of some surfactants on foliar impaction and retention of monosize water droplets. Pesticide Science 55 3:382-385. 34. Western, N. M., Hislop, E. C., Bieswal, M., Holloway, P. J., and Coupland, D. 1999. Drift reduction and droplet-size in sprays containing adjuvant oil emulsions. Pesticide Science 55 6:640-642. 35. Zadoks, J. C. 1984. A quarter century of disease warning, 1958-1983. Plant Disease 68 4:352-355.

46 CHAPTER 3

EVALUATION OF THE CURATIVE AND PROTECTANT

ACTIVITY OF FUNGICIDES AND FUNGICIDE-ADJUVANT

MIXTURES ON PHOMOPSIS CANE AND LEAF SPOT OF GRAPE:

A CONTROLLED ENVIRONMENT STUDY

INTRODUCTION

The fungus Phomopsis viticola (Sacc.) is the causal agent of Phomopsis cane and leaf spot of grape. This disease was previously known as "dead-arm" in American literature (2, 8). Various parts of the vine, such as shoots, rachises, leaves, and fruits are susceptible to infection, especially when tissues are immature (8, 19). Leaf symptoms are small irregular, or round shaped, yellow spots with dark centers. Infected canes and rachises show dark necrotic irregular shaped lesions. Infected fruits appear as brown shriveled berries near harvest (19, 20). Up to 30% loss of the crop has been reported in

Southern Ohio grape growing regions due to fruit rot (8).

47

The fungus survives winter in grape cane tissues that were infected in previous years. In spring, the pathogen on infected canes produces numerous pycnidia. Conidia are splashed by rain onto new growth where they can infect plant tissues. Typically, control of the disease is obtained with selective pruning of previously infected canes and calendar-based applications of protective fungicides such as mancozeb or captan (7, 19).

A disease-warning system for Phomopsis cane and leaf spot was developed based on the results of Erincik et al. (9). The relationship between wetness duration (W), air temperature (T), and level of infection was determined in controlled environment studies.

Essentially, prediction of cane and leaf spot infection level, assuming inoculum is available, is made with an empirical non-liner model using monitored leaf wetness and air temperature (9). By incorporating the warning system into a fungicide application program, growers may be able to reduce the number of applications, or improve application timing (4, 14). Thus, growers could improve disease control while saving money and reducing environmental impacts from unnecessary fungicide applications.

Warning systems have been successfully used for other grape diseases (14).

Ideally, the fungicide used with this type of warning systems should have curative activity (i.e., control the pathogen after penetration of host tissue), although warning systems can also be used to optimize scheduling of protectant fungicides (4, 5, 14, 30).

Field experiments based on use of the warning system with application of fungicide- adjuvant mixtures have been conducted through 2001-2004 growing seasons, and results have indicated advantages of using the system (17).

48 There are no known curative fungicides effective against Phomopsis cane and leaf spot of grape. However, some fungicide-adjuvant mixtures are known to enhance curative effects for some pathosystems, possibly by increasing uptake of the fungicide into the plant (10, 16, 24, 26). Studies on herbicide-adjuvant mixtures are relatively common, but not many studies have been done to determine effects of fungicide-adjuvant mixtures on plant disease control (26). The objective of this study was to evaluate fungicides and fungicide-adjuvant mixtures for their curative activity against Phomopsis cane and leaf spot under controlled environment conditions.

MATRIALS AND METHODS

Host preparation. Grape vines (cultivar 'Seyval') were grown from rooted cuttings in #1 trade size pots (17.1-cm diameter, 19.7-cm deep, 3.3-L in volume) for 3-6 mo. Grapes were grown in a greenhouse where temperature ranged from 18 to 30 °C.

Plants were fertilized with a slow-releasing granule fertilizer (13 g/pot of Osmocote 14-

14-14, Scotts Company, Marysville, OH) once every 3 mo. Young leaves and internodes were used in the experiment because of ontogenic resistance expressed by older leaves

(8). Leaves less than 3-cm in length and less than 1 wk old were chosen per shoot and were identified by tags. A total of five young leaves and five internodes between these leaves were selected per vine for each treatment.

49 Inoculum preparation. An isolate of P. viticola originally obtained from M.E.

Palm, USDA/APHIS, Beltsville, MD, which was used in our previous studies (9), was utilized in this study. This isolate was shown to be aggressive (8). Cultures were grown in an incubator at 22 °C under continuous fluorescent light at 58 µE m-2 s-1.

Pathogenicity of the isolate was maintained by inoculating autoclaved grape leaf tissue placed on PDA (Potato Dextrose Agar, in a 100 x 15 mm plastic Petri dish) with a mycelia plug from a previously grown culture. After 2-3 wk, plugs from the edge of actively growing mycelia from the PDA with grape tissue were transferred onto new Petri dishes with PDA.

Cultures of P. viticola grown on the PDA for approximately 21 days were used for preparation of inoculum. A Petri dish of the fungus with abundant pycnidia was flooded with ~15 ml of distilled water, left undisturbed for 2 min, then the surface of

PDA was gently rubbed with a paint brush. The suspension was filtered through four layers of cheesecloth to collect alpha spores. Beta conidia were observed on very few occasions, but not counted as active spores since their ability to cause infection is unclear

(25). Spore density was adjusted to 5 x 106 /ml using a hemacytometer.

Inoculation methods. Inoculum was sprayed onto grape leaves and internodes until run off with an electronic atomizer (with pressure of 51.7 kPa). After inoculation, plants were placed in a mist chamber where high RH and free surface moisture were maintained with a mister. After application of treatments, plants were returned to the greenhouse for 4 wk. Plants were watered on the soil surface to avoid wetting leaves and internodes.

50 The mist chamber measured 2.3 x 1.4 x 1.0 m (WxDxH), and the wall and ceiling consisted of polyethylene sheeting, supported by PVC pipe (2.54-cm diameter). The chamber could hold ~20 potted plants in 3.3-L pots. An atomizer with a fluid nozzle

(0.56 kg/cm3 of water, #2850 SS; Spraying Systems, Co., Spokane, WA) and an air nozzle (4.2 kg/cm3 of air, #70) was placed near the top of the chamber, and provided 90 ml of water for 90 s every 180 s. There was a polyethylene sheet between the nozzles and plants which ensured steady drift of mists to the chamber and prevented accumulation of water on inoculated tissues. The chamber was located indoors, and temperature was maintained at approximately 22 (±2) °C and relative humidity was maintained around 100%. There were high pressure sodium lights above the chamber providing continuous light at 47 µE m-2 s-1. Temperature and wetness duration was monitored using WatchDog data-logger model 130 (Spectrum Technologies, Inc.

Plainfield, IL).

Fungicides and adjuvants. Fungicides studied were: mancozeb (Dithane, Dow

Agrosciences LLC, Indianapolis, IN); calcium polysulfide (liquid lime sulfur, Platte

Chemical CO., Greeley, CO); thiophanate-methyl (Topsin-M, Cerexagri, Inc., King of

Prussia, PA); azoxystrobin (Abound, Syngenta Crop Protection, Inc, Greensboro, NC ); and myclobutanil (Nova 40W, Dow Agrosciences LLC, Indianapolis, IN). Adjuvants examined were: JMS Stylet-Oil (JMS Flower Farms, Inc, Vero Beach, FL), and Regulaid

(polyoxyethylenepolypropoxypropanol alkyl 2-ethoxyethanol/dihydroxy-propane, Kalo

Laboratories, Inc., Kansas City, MO). Stylet-Oil is a highly refined paraffinic oil that can

51 be used for control of insects, mites, aphid-transmitted plant viruses, and powdery mildews (3, 15, 18). Regulaid is a nonionic spreader-activator, often used for improving efficacy of foliar plant growth regulartors (27).

Post-inoculation application study. Four separate experiments were conducted, each consisting of a single wetness duration and drying time to produce an infection period (Figure 3.1). Experiment I consisted of a 24-h wetness period plus a 24-h drying time. Experiments II and III had a 20-h wetness and a 4-h drying time, and experiment

IV had a 16-h wetness plus a 4-h drying time. Labeled doses of fungicides were used in experiments I, II and IV (Table 3.1), and 1.5 times the labeled dose (150%) was used in experiment III (Figure 3.1).

Post-inoculation treatments (mancozeb, thiophanate-methyl, azoxystrobin, and myclobutanil) in combination with adjuvant (JMS Stylet-Oil or Regulaid), and mancozeb and lime sulfur alone were sprayed onto the plants after the inoculation and drying period

(Table 3.1). Protectant (pre-inoculation) sprays of mancozeb or liquid lime sulfur were applied with a hand atomizer at least 3 h, but no more than 6 h, prior to inoculation.

Plants were dry at the time of inoculation. After post-inoculation treatment or after the wetness period (for pre-inoculation treatments), grapes were placed in a greenhouse.

52 Pre-inoculation application study. Two additional experiments were performed using only pre-inoculation application of fungicides. Previously described fungicides

(Table 3.1) were applied with or without adjuvant using a hand atomizer. Fungicides were applied 3-6 h before inoculation to allow fungicides to dry. Fungicide, fungicide- adjuvant combinations, and doses were as same as in the post-inoculation experiments

(Table 3.1). After inoculation, vines are placed in the mist chamber for a 22.5-h, then moved to a greenhouse in order to achieve a ~24-h leaf wetness duration. In experiment

V, fungicides were applied alone, and in experiment VI, fungicides-adjuvant mixtures were applied.

Re-isolation of P. viticola. In order to verify symptoms observed were actually due to infection of P. viticola, six experimental repetitions from the post-inoculation studies (Experiments I-IV) were randomly chosen, and re-isolation of P. viticola from infected canes was attempted. After assessment of visual symptoms, six 5-cm pieces of internodes were collected. Then canes were surface disinfested by soaking in 95% EtOH for 5 sec, then 0.5% sodium hypochlorite (10% Clorox) for 30 sec; after soaking, canes were washed in distilled water for 1.5 min, followed by 2 min drying on sterile paper towels. After disinfestation, cane pieces were placed on Petri dishes with acidified PDA

(APDA), and incubated for 72 h under continuous light continuous light at 48 µE m-2 s-1; then, P. viticola was identified based on colony morphology. When distinctive characteristics were not observed at 72 h, the culture was kept for a week or until fruiting bodies formed.

53 Assessment methods and analysis of data: Four weeks after inoculation, disease severity and incidence on internodes and leaves were visually assessed. Direct estimation of percentage of visually diseased area was used for internode disease assessments.

Disease severity of leaves was assessed by estimating the number of lesions on each leaf using a scale with seven levels (0 = no spots, 6 = more than 100 spots [=lesions]).

Each experiment was a completely randomized design with three replications.

The five leaves and internodes per vine were considered sub-samples. Data were analyzed using a linear mixed model (PROC MIXED in SAS) in order to determine effects of treatment on the disease severity. Contrasts were used for comparing groups of treatments. For example, the inoculated-control mean was compared with the mean of the post-inoculation treatments. Likewise, the means for the pre-inoculation fungicide treatments were compared with the means for the post-inoculation treatments. Disease severity values were transformed using an angular transformation ( arcsin( proportion) ) for internode disease severity, or a square-root transformation for lesion counts on leaves.

Mean severity numbers presented in tables were back-transformed from the least-square means.

54 RESULTS

Post-inoculation application studies. Vines not treated with fungicides had 76-

93 lesions per leaf and 3-12% disease severity on internodes (Table 3.2) in experiments I-

IV. Pre-inoculation application of mancozeb or calcium polysulfide significantly reduced severity in all experiments compared with the inoculated control (Tables 3.2 and 3.3).

On average, severity was reduced by 96 % with pre-inoculation application of fungicides.

Results were very similar for the different experiments even though wetness periods and fungicide dose varied.

Disease severity with post-inoculation application of fungicides was not significantly different from the inoculated control in most experiments. Furthermore, application of mancozeb with adjuvant did not reduce severity compared to treatment without adjuvant in most experiments (Table 3.3). The two adjuvants gave very similar results (Table 3.3), and the mean for Stylet-Oil was not different from the mean for

Regulaid.

Pre-inoculation application studies. Vines not treated with fungicides had, on average, 29-58 lesions per leaf and ~3% disease severity on internodes (Tables 3.4 and

3.5). Pre-inoculation application of fungicides, with exception of myclobutanil (Nova), significantly reduced disease severity in both experiments compared with the inoculated control (Tables 3.4 and 3.5). Although there tended to be reductions in disease severity, adjuvants applied without fungicide did not significantly reduce severity compared with the inoculated control (Table 3.4). Thiophanate-methyl (Topsin-M) provided more

55 control when used alone (Experiment V) (Table 3.4), than when used with Regulaid

(Experiment VI) (Table 3.5). Also, pre-inoculation applications of liquid lime sulfur showed moderate control of P. viticola (Tables 3.2 and 3.4).

Re-isolation of P. viticola. P. viticola was re-isolated from plants in all treatments, but number of successful re-isolation was less from the pre-inoculation treatments and azoxystrobin (Abound) + Regulaid post-inoculation treatment than the others (Table 3.6).

The lower recovery rate from pre-inoculation treatments was likely because of the relatively high degree of control achieved with the fungicides.

DISCUSSION

A series of experiments were conducted to determine possible curative activity of fungicides against Phomopsis viticola when used with certain adjuvants. The use of fungicides with adjuvants has increased efficacy and curative activity for some diseases.

Examples include Baycor 25W (bitertanol, Bayer South Africa) and Agridex (paraffin base petroleum oil, polyoxythylated polyol fatty acid ester, and polyol fatty acid ester constituents, Bayer South Africa) for apple scab control (24), and 1-(4-chlorobenzyl)-4- phenylpiperidine (Fluorochem Ltd., UK) and Dobanol 91-6 (alcohol ethoxylate, Shell

Chemicals, UK) for powdery mildew control (11). However, we found no evidence of curative activities against P. viticola from any of the tested fungicides or fungicide- adjuvant combinations in any of those experiments.

56 Leaf and internode disease severity for all post-inoculation treatments were not different from that in the inoculated control, even when fungicides were used at 150% of labeled doses (Experiment III) or when fungicides were applied within 20 h of the start of the infection period, or 4 h from the end of the wetness period (Experiment IV).

Exceptions were leaf disease severities for mancozeb, calcium polysulfide, and myclobutanil-Regulaid treatments in Experiment IV, but leaf disease severities for these treatments were still high and significantly different from pre-inoculation treatments, and internode disease severities of these treatments were not significantly different from the inoculated control (Table 3.2).

Excellent disease control was obtained with pre-inoculation application of some fungicides, such as mancozeb or azoystrobin, with experiments that varied in length of the wetness periods for infection (Tables 3.2-3.5). However, because there was no evidence of curative activities from any fungicide-adjuvant combinations we tested, suppression of disease symptoms in the field when fungicides were applied based on the disease warning system (17) probably was due to protective activity of the treatment rather than curative activity. Although not tested here, adjuvants may have other types of beneficial effects in the field. Rainfastness (retention capacity on the plant surface) of mancozeb was shown to be improved with addition of adjuvant (crop oil or sicker products) (12). Also, adjuvants could improve efficiency of fungicide spray by reducing fungicide drift or improving target acquisition (29).

Applications of calcium polysulfide (lime sulfure) on P. viticola when used as a protectant provided the moderate control provided (Tables 3.2 and 3.4). Although reduction in disease severity with protective application of calcium polysulfide pre-

57 application tended to be less than that with the mancozeb or azoxystrobin applications

(Tables 3.2 and 3.4), differences were not significant. Disease severity for calcium polysulfide pre-inoculation application was significantly different (P≤0.05) from that for the inoculated control. Since calcium polysulfide is registered for use in organic grape production (6), calcium polysulfide may be useful to organic grape growers for control of

P. viticola.

Not all fungicides tested were effective when used as protectants. Myclobutanil

(Nova) did not reduce the disease severity whether used before or after inoculation, or whether used alone or with an adjuvant (Tables 3.2-3.5). This agrees with other studies

(28). However, myclobutanil has activity against P. viticola in vitro, in that spore germination is inhibited on PDA amended with 63 ppm (1/2 of the recommended dose) of the fungicide (Madden, unpublished). Moreover, inconsistent results with thiophanate-methyl (Topsin-M) was found when used as a protectant (Tables 3.4 and 3.5).

Thus, an additional experiment was performed where thiophanate-methyl was applied with and without Regulaid, and either before or after inoculation (data not shown). None of these treatments resulted in a reduction in disease severity relative to the no-fungicide control (Nita, unbublished). It may be possible that thiophanate-methyl requires specific environment, host, or pathogen conditions to be effective against P. viticola, or simply, it is less effective.

Adjuvant alone may provide some control on P. viticola in the field (22, 23).

Stylet-Oil has been commonly used to control powdery mildew on grape (caused by

Uncinula nacator) (3, 18, 19), as well as various other diseases, including yellow

Sigatoka on banana (caused by Mycosphaerella fijiensis) (1) and cucurbit powdery

58 mildew (caused by Erysiphe polygoni) (15). There are several hypotheses on how these adjuvants suppress disease organisms. Various studies suggested that highly refined paraffinic mineral oil, such as Stylet-Oil, is neither fungicidal nor able to suppress sporulation (1, 15, 18), but other studies show inhibition of conidial germination of E. polygoni which causes powdery mildew on cucumber (3). Studies on powdery mildew and downy mildew on grape (18), as well as on cucurbit powdery mildew (15) indicate that Stylet-Oil suppresses the lesion growth temporary, but with time, fungi can re-grow and its sporulation ability is not affected. In the same studies (15, 18), the authors suggested that Stylet-Oil may reduce dissemination of spores by forming a thin film on plant surfaces, without disrupting sporulation. In addition, other investigators suspect that surfactants, such as Stylet-Oil and Regulaid, trigger changes in plant physiology, including plant defense responses (1, 13, 18).

When pre-inoculation application of Stylet-Oil and Regulaid were tested (Table

3.4), there appeared to be reductions in mean disease severity; however, no significant reduction was observed. Moreover, the adjuvants did not provide any disease control when used with fungicide after inoculation (Table 3.2). Based on our results, any beneficial effect of these products in the field (22, 23) would not be due to direct effects on infection.

Although it was not recorded for all plants in the different experiments, symptoms on tissues receiving treatments with adjuvant frequently showed deviation from regular P. viticola symptom expression. Typically, lesions of P. viticola have sharp margins, but lesions resulting after application of treatments with adjuvant often showed less distinctive, brownish margins (19). One study (13) showed that there was increased

59 ethylene production from various plant species upon application of various adjuvants, including Regulaid. The authors speculated that the reason for increased effectiveness of herbicide-adjuvant combinations relative to simple use of the herbicide alone could be related to the ethylene production and phytotoxicity due to adjuvant. Phytotoxicity due to

Regulaid has been reported (21), and Stylet Oil on grapes was shown to cause phytotoxicity when applied frequently (1, 19). Consistent success of re-isolation of P. viticola (Table 3.6) suggests that symptoms observed were due to infection by the pathogen, and not due solely to phytotoxic effects from adjuvants. The use of adjuvant, however, could have caused atypical symptoms observed in some treatments.

60

Fungicide (trade name) Dosea/ha a.i.b /ha Dose (150%)c/ha Mancozeb (Dithane) 4.5 kg 3.60 kg 6.8 kg Thiophanate-methyl (Topsin-M) 1.7 kg 1.19 kg 2.6 kg Myclobutanil (Nova 40W) 0.3 kg 0.12 kg 0.5 kg Azoxystrobin (Abound) 1.1 m3 0.25 m3 1.7 m3 Calcium polysulfide (Lime sulfur) 4.7 m3 1.36 m3 7.1 m3 JMS Stylet-Oil 3.8 m3 3.69 m3 5.7 m3 Regulaid 2.4 m3 2.17 m3 3.6 m3

a Dose of product (not active ingredient), applied with 946L of water per ha b Active ingredient c In Experiment III only

Table 3.1. Fungicides and doses used in experiments

61

Experiment I II III IV Leaf Node Leaf Node Leaf Node Leaf Node Treatmenta #b %c #b %c #b %c #b %c Mancozeb+ Regulaid 66.2 ad 5.0 ab 78.8 a 12.2 a 56.9 a 1.4 ac 63.9 ab 5.6 ab Mancozeb+ Stylet Oil 59.8 a 3.0 bc 57.4 ab 2.2 bcd 82.9 a 2.7 ac 66.9 ab 8.4 ab Topsin-M + Regulaid 68.3 a 3.8 ac 57.2 ab 3.3 ad 42.2 a 0.7 ac 57.4 ab 4.0 ac Topsin-M + Stylet Oil 70.5 a 5.1 ab 77.4 a 0.7 bcd 62.1 a 1.8 ac 55.6 ab 7.5 a Nova + Regulaid 83.1 a 6.7 ab 69.6 a 3.5 ad 67.4 a 3.1 ac 45.1 bc 5.2 ac Abound + Regulaid 56.9 a 3.7 ac 94.6 a 4.9 ab 67.0 a 3.2 ac 48.2 ac 5.3 ac Mancozeb 63.5 a 8.8 ab 85.1 a 2.8 bcd 94.0 a 1.7 ac 30.6 c 7.0 a Lime Sulfur 75.9 a 8.3 ab 99.1 a 2.1 bcd 70.4 a 2.3 ab 45.7 bc 10.6 a Lime Sulfur (pre-inoculation) 4.1 b 0.3 c 25.7 b 0.2 cd 1.4 b 0.2 bc 0.8 d 0.2 bc Mancozeb (pre-inoculation) 0.5 b 0.3 c 0.4 c 0.1 d 1.1 b 0.1 c 2.9 d 0.2 c Control 77.0 a 12.2 a 92.7 a 4.7 ac 84.1 a 3.2 a 76.2 a 7.3 a

a Doses are described in Table 3.1 b # = back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. c % = back-transformed value of mean disease severity. Disease severity was estimated by visual assessment of proportion of internode covered by lesions, then angular transformed, arcsin( proportion) , before analysis with Proc MIXED of SAS. d Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data.

Table 3.2. Effects of various fungicides and fungicide-adjuvant mixtures applied before or after inoculation of Phomopsis viticola on disease symptom development on grape leaves and internodes

62

Experiment I II III IV Contrast Leaf Node Leaf Node Leaf Node Leaf Node Control vs. pre-inoculation treatments 45.2*a 17.8** 34.2** 6.0** 46.4** 5.3* 93.6** 80.8** Control vs. post-inoculation treatments 1.0 3.8 1.4 0.2 0.7 0.1 5.7* 0.05 Pre- vs. post-inoculation treatments 82.0** 15.4** 55.8** 9.9** 88.9** 10.5** 193.1** 17.0** Post-inoculation treatments: 0.02 3.1 3.2 0.7 1.5 0.3 6.8** 1.0 with adjuvant vs. without adjuvant Post-inoculation treatments: 0.01 0.03 0.01 6.0** 1.9 1.5 0.0 1.0 Stylet Oil vs. Regulaid Mancozeb: with vs. without adjuvant 0.02 2.0 1.5 1.2 1.2 0.4 11.1** 0.0 a ** = significant at P=0.01, * = significant at P=0.05

Table 3.3. F-test statistics for contrasts of treatment means for leaf or internode disease severity.

63

Leaf Node Treatmenta #b %c Mancozeb 6.2 abd 0.4 a Topsin-M 1.0 a 0.1 a Nova 7.1 a c 0.6 a Abound 2.2 ab 0.3 a Lime Sulfur 4.6 ab 0.4 a Stylet Oil 19.2 bc 2.6 b Regulaid 9.9 a c 1.0 b Control 29.4 c 2.6 b

a Doses are described in Table 3.1 b # = back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. c % = back-transformed value of mean disease severity. Disease severity was estimated by visual assessment of proportion of internode covered by lesions, then angular transformed, arcsin( proportion) , before analysis with Proc MIXED of SAS. d Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data.

Table 3.4. Effects of fungicides applied before inoculation of Phomopsis viticola on disease symptoms on grape leaves and internodes (Experiment V)

64

Leaf Node Treatmenta #b %c Mancozeb 0.02 dd 0.4 cd Mancozeb + Regulaid 0.02 d 0.1 d Mancozeb + Stylet Oil 0.1 d 0.3 d Topsin-M + Regulaid 43.6 a 1.0 bc Topsin-M + Stylet Oil 55.4 a 2.0 ab Nova + Regulaid 16.4 b 0.1 d Abound + Regulaid 6.5 c 0.1 d Control 58.2 a 2.9 a

a Doses are described in Table 3.1 b # = back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. c % = back-transformed value of mean disease severity. Disease severity was estimated by visual assessment of proportion of internode covered by lesions, then angular transformed, arcsin( proportion) , before analysis with Proc MIXED of SAS. d Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data.

Table 3.5. Effects of fungicides-adjuvant combinations applied before inoculation of Phomopsis viticola on disease symptoms on grape leaves and internodes (Experiment VI)

65

Treatmenta %b Mancozeb + Regulaid 58.9 ac Mancozeb + Stylet Oil 47.6 ab Mancozeb 71.3 a Topsin-M + Regulaid 37.1 a c Topsin-M + Stylet Oil 41.1 a c Nova + Regulaid 68.7 a Abound + Regulaid 8.4 bc Lime Sulfur 77.8 a Lime Sulfur (pre-inoculation) 4.0 c Mancozeb (pre-inoculation) 10.9 bc Control 62.5 a

a Doses are described in Table 3.1 b % = back-transformed value of the percentage of canes with P. viticola mycelia. Incidence of re- isolation of P. viticola was examined visually, then angular transformed, arcsin( proportion) , before analysis with Proc MIXED of SAS. c Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data.

Table 3.6. Re-isolation of Pomopsis viticola from inoculated vines in controlled environment studies for post-infection activity of fungicides and fungicide adjuvant mixtures (Experiments II-IV)

66

Application of inoculum

Wetness period (e.g., 24 hr)

Move to a greenhouse

Drying time (e.g., 24 hr)

Application of fungicide hr) 48 (e.g., of incubation Length

Experiment Incubation Wetness period Drying time

I 48 hr 24 hr 24 hr II 24 hr 20 hr 4 hr a III 24 hr 20 hr 4 hr IV 20 hr 16 hr 4 hr

Figure 3.1. Time-flow diagram of post-inoculation application of fungicides and a list of wetness period and drying times of each experiment. a indicates fungicide at 150% labeled dose

67 REFERENCES

1. Calpouzos, L. 1966. Action of oil in the control of plant disease. Annual Review of Phytopathology 4:369-390. 2. Cucuzza, J. D., and Sall, M. A. 1982. Phomopsis cane and leaf spot disease of grape vine: Effect of chemical treatments on inoculum level, disease severity, and yield. Plant Disease 66:794-797. 3. Dell, K. J., Gubler, W. D., Krueger, R., Sanger, M., and Bettiga, L. J. 1998. The efficacy of JMS Stylet-Oil on grape powdery mildew and Botrytis bunch rot and effects on fermentation. American Journal of Enology and Viticulture 49 1:11-16. 4. Ellis, M. A., Madden, L. V., and Wilson, L. L. 1986. Electronic grape black rot predictor for scheduling fungicides with curative activity. Plant Disease 70 10:938-940. 5. Ellis, M. A., Madden, L. V., and Wilson, L. L. 1984. Evaluation of an electronic apple scab predictor for scheduling fungicides with curative activity. Plant Disease 68:1055-1057. 6. Ellis, M. A., and Nita, M. 2004. Disease management guidelines for organic grape production in the Midwest. Plant Pathology Department Series 121 The Ohio State University OARDC/OSUE, Wooster, OH 7. Ellis, M. A., Welty, C., Funt, R. C., Doohan, D., Wiliams, R. N., Brown, M., and Bordelon, B. ed. 2004. Midwest Small Fruit Pest Management Handbook. Bulletin 861 Ohio State University Extension, Columbus, OH 8. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2001. Effect of growth stage on susceptibility of grape berry and rachis tissues to infection by Phomopsis viticola. Plant Disease 85 5:517-520. 9. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2003. Temperature and wetness-duration requirements for grape leaf and cane infection by Phomopsis viticola. Plant Disease 87 7:832-840. 10. Gent, D. H., Schwartz, H. F., and Nissen, S. J. 2003. Effect of commercial adjuvents on vegitable crop fungicide coverage, absorption, and efficacy. Plant Disease 87:591-597. 11. Grayson, B. T., Price, P. J., and Walter, D. 1997. Effects of adjuvants on the performance of a novel powdery mildew fungicide, 1-(4-chlorobenzyl)-4- phenylpiperidine. Pesticide Science 51 2:206-212. 12. Kudsk, P., Mathiassen, S. K., and Kirknel, E. 1991. Influence of formulations and adjuvants on the rainfastness of maneb and mancozeb on pea and potato. Pesticide Science 33 1:57-71. 13. Lownds, N. K., and Bukovac, M. J. 1989. Surfactant-induced ethylene production by leaf tissue. Journal of the American Society for Horticultural Science 114 3:449-454. 14. Madden, L. V., Ellis, M. A., Lalancette, N., Hughes, G., and Wilson, L. L. 2000. Evaluation of a disease warning system for downy mildew of grapes. Plant Disease 84 5:549-554. 15. McGrath, M. T., and Shishkoff, N. 2000. Control of cucurbit powdery mildew with JMS Stylet-Oil. Plant Disease 84 9:989-993.

68 16. Morrison, L. S., and Russell, C. C. 1976. Timing of fungicide - adjuvant mixtures for control of rose blackspot. Plant Disease Reporter 60 7:634-636. 17. Nita, M., Madden, L. V., Wilson, L. L., and Ellis, M. A. 2003. Evaluation of a disease prediction system for Phomopsis cane and leaf spot of grape. Phytopathology 93 6:S65. 18. Northover, J., and Schneider, K. E. 1996. Physical modes of action of petroleum and plant oils on powdery and downy mildews of grapevines. Plant Disease 80 5:544-550. 19. Pearson, R. C., and Goheen, A. C. ed. 1988. Compendium of Grape Diseases. American Phytopathological Society, St. Paul, MN 20. Pezet, R., Pont., V., and Girardet, F. 1983. Microcycle conidiation in pycnidial cirrhii of Phomopsis viticola Sacc. induced by elemental sulfur. Canadian Journal of Mycrobiology 29:179-184. 21. Raese, J. T., and Drake, S. R. 1993. Effects of preharvest calcium sprays on apple and pear quality. Journal of Plant Nutrition 16 9:1807-1819. 22. Schilder, A. M. C., Gillett, J. M., and Sysak, R. W. 2000. Evaluation of fungicides for control of phomopsis cane and leaf spot of grape, 1999. Fungicide and Nematicide Tests 55:105. 23. Schilder, A. M. C., Gillett, J. M., Sysak, R. W., and Wise, J. C. 2002. Evaluation of environmentally friendly products for control of fungaldiseases of grapes. Fordergemeinschaft Okologischer Obstbau e.V. (FOKO); Weinsberg; Germany, Weinsberg, Germany 163-167. 24. Schwabe, W. F. S., and Jones, A. L. 1983. Apple scab control with bitertanol as influenced by adjuvant addition. Plant Disease 67:1371-1373. 25. Sergeeva, V., Nair, N. G., Barchia, I., Priest, M., and Spooner-Hart, R. 2003. Germination of β conidia of Phomopsis viticola. Austrarian Plant Pathology 32:105-107. 26. Steurbaut, W. 1993. Adjuvants for use with foliar fungicides. Pesticide Science 38:85-91. 27. Thomson, L. A. 1986. A guide to agricultural spray adjuvants used in the united states. Thomson publications, Fresno, CA 28. Travis, J. W., and Hed, B. 2001. Evaluation of fungicides for control of powdery mildew and phomopsis canse and leaf spot on grapes, 2000. Fungicide and Nematicide Tests 56:SMF30. 29. Western, N. M., Hislop, E. C., Bieswal, M., Holloway, P. J., and Coupland, D. 1999. Drift reduction and droplet-size in sprays containing adjuvant oil emulsions. Pesticide Science 55 6:640-642. 30. Zadoks, J. C. 1984. A quarter century of disease warning, 1958-1983. Plant Disease 68 4:352-355.

69 CHAPTER 4

DORMANT APPLICATION OF FUNGICIDE AND ITS EFFECTS

ON PHOMOPSIS CANE AND LEAF SPOT ON GRAPE DISEASE

INTENSITY AND INOCULUM PRODUCTION

INTRODUCTION

Phomopsis cane and leaf spot is a disease of grape caused by the fungus

Phomopsis viticola (Sacc.) (16). Various tissues of the vine, such as leaf, cane, rachis, and fruit can be infected by P. viticola. Infected leaves show small irregular, or round, pale green-to-yellow spots with dark centers. Brown to black necrotic irregular shaped lesions develop on the infected canes and rachises. Infected fruits appear as brown shriveled berries near harvest (16, 18). Infections on the canes and rachises weaken the plant and may cause premature fruit drop. Infections on fruits directly decrease yield and fruit quality. The disease is common in the U.S. and grape growing regions around the world (9, 12, 15). Up to 30% loss of the crop has been reported in Southern Ohio grape vineyards (5).

70 The fungus survives winter in grape cane tissues that were infected in previous years. In spring, numerous pycnidia develop on infected canes, and a gelatinous mass of spores (α- or β-conidia) called cirrhi exudes out from the pycnidia. Conidia are splashed by rain onto new growth where they infect the plant tissues. The function of the β- conidia is not known (17, 22).

Selective pruning and protective fungicide applications are commonly used to control the disease (16, 18). Protective fungicides, such as mancozeb or captan, applied on a 7-10 day calendar based schedule have been used for disease control on green tissues and fruits (4, 16). Research has indicated that the fungus becomes active early in the growing season, when only about 2.5-cm of the new growth of the grape tissues are present (1, 12, 18). Close proximity between the source of inoculum (infected canes from the main trunk) and the susceptible tissues (new growth of canes) may also favor infection in the early growing season. However, it may not always be practical to apply fungicides at this time because the target plant tissue area is very small and rapidly growing. On the other hand, if fungicides are applied after this window of infection, disease intensity could be high no matter how many fungicide applications are made afterwards (Chapters 1 and 2). Under favorable weather conditions, the fungus could become established in grape tissues regardless of the use of standard protectant fungicides.

Thus, the use of dormant applications of fungicide, which may aid in controlling very early season infection, needs to be studied. Instead of applying the fungicide to newly produced tissue, a dormant application of fungicide targets the main trunk of the vine. Dormant applications of fungicide are thought to reduce overall level of the 71 inoculum before the beginning of a growing season (6). Examples of effective control provided with dormant application of fungicide includes: fire blight on apple and pear

(caused by Erwinia amylovora) (8), powdery mildew on grape (caused by Uncinula necator) (6), anthracnose on grape (caused by Elsinoe ampelina) (4, 16) and raspberry

(caused by Elsinoe veneta) (2, 4). The potential effects of dormant fungicide applications against P. viticola have been reported in limited studies; however, the fungicide used may no longer be available (12), or experiments were not repeated (7, 21). The objective of the current study was to determine the efficacy of dormant applications of fungicide against P. viticola under field conditions in Ohio. Also, the effects of dormant applications of fungicide on sporulation were examined.

MATERIALS AND METHODS

A vineyard of the cultivar 'Concord' (Vitis labrusca), planted in 1970, was used for this experiment. In addition, a portion of a vineyard of the cultivar 'Catawba' (Vitis labrusca), planted in 1995, was used for an additional experiment. In both vineyards, vines were trained to an umbrella Kniffin system, and distance between vines was 2.1 m, with 3 m between rows. The two vineyards were at the Ohio Agricultural Research and

Development Center, Wooster, OH, and about 100-m apart from each other. Both the

‘Concord’ and ‘Catawba’ vineyards were known to have relatively high incidence of

Phomopsis cane and leaf spot in previous years (Ellis, unpublished).

72 In the 'Concord' vineyard, treatments consisted of different fungicide types and application timings. Calcium polysulfide (liquid lime sulfur) or copper oxychloride sulfate (fixed copper, COCS) were applied to dormant vines: in the fall (after leaf drop and after formation of periderm of canes, early-November); in the spring (at bud swell, mid-April); or in both fall and spring. A control consisted of unsprayed vines. Calcium polysulfide was applied at 95 L per ha, and fixed copper was applied at 3.4 kg per ha in

948 L of water. Treatments were applied using a handgun sprayer at a pressure of 689 kPa. Vines were sprayed to the point of runoff. Fall and spring applications for the 2003 season were applied on 7 November 2002 and 17 April 2003, respectively; fall and spring applications for the 2004 season were applied on 6 November 2003 and 20 April 2004, respectively. No other fungicides were applied in either vineyard.

In the 'Catawba' vineyard, a spring dormant application of calcium polysulfide was compared with a calendar-based protectant program (7-d interval) of mancozeb, and an unsprayed control. The experiment was conducted in 2003 and 2004. Dates of dormant application were the same as in the ‘Concord’ vineyard, and the calendar-based applications of mancozeb started on 29 April and ended on 26 June in 2003 (consisted of

9 sprays), and it started on 6 May and ended on 17 June in 2004 (7 sprays).

Experimental design was a randomized complete block, replicated four times in both experiments in each year. In the 'Concord' vineyard, each replication consisted of three adjacent vines, and the center vine was assessed for disease in order to minimize influence of chemical drift on disease development. Also, the same treatments were applied to the same vines for the two consecutive years. In the 'Catawba' vineyard, each replication consisted of a single vine, and treatments were randomized each year.

73 Levels of disease intensity (incidence and severity) were visually assessed on 27

June 2003 and 28 June 2004 in both vineyards. The five basal internodes and leaves on

10 randomly chosen shoots per vine were evaluated (50 internodes and leaves per vine).

Direct estimation of percentage of diseased area was used for internode disease severity assessments. Disease severity of leaves was assessed by estimating the number of lesions on each leaf using a scale with seven levels (0 = no lesions; 1 = 0 to 10; 2 = 10 to 25; 3 =

25 to 50; 4 = 50 to 75; 5 = 75 to 100; and 6 = more than 100 lesions). Disease incidence was based on percentage of infected internodes or leaves per shoot using the same internodes or leaves used in severity estimate. Disease incidence on rachises was evaluated on 16 September 2003 and 1 September 2004. Five rachises were randomly chosen from each vine and examined for presence of visual symptoms.

Data were analyzed using a linear mixed model with PROC MIXED of SAS

(SAS institute, Inc., Cary, NC) in order to determine effects of treatments on disease intensity. Because the treatment locations were not randomized between years, a mixed- model spatial analysis was performed to detect and correct for any correlation of the disease values in neighboring vines. Using the methods described in Little et al. (10) and

Schabenberger and Pierce (20), a linear mixed model was fitted using a spatial first-order autoregressive model for the residuals. A comparison of the log likelihood statistics (10), and also the AIC statistics, for models with and without spatial correlation component revealed that results were not affected by treatment location in the vineyard. Thus, results for the simpler linear mixed model are reported. Treatments were considered as fixed effects and blocks as random effects. Values of internode and leaf disease incidences, and that of internode disease severities were transformed using an angular

74 transformation, ( arcsin proportion ), and a square-root transformation was used for lesion counts on leaves. Statistical analyses and treatment comparisons were based on least square means of transformed values, which were then back-transformed after the analysis to obtain the presented means.

Collection of splashed rain water. To assess the effects of a dormant treatment on sporulation in the field during the spring, spore traps, consisting of a 20-cm-diameter funnel attached to a 2-L plastic bottle, were placed under the center vine of the three-vine plot, starting at bud-break to the end of bloom. A trap was placed in each replication of treatments receiving both fall and spring applications of lime sulfur and in the unsprayed control. Traps were placed under the main trunk where the canes emerged out from the main trunk. After each rain event, rainwater in the container was collected and the number of conidia was counted. Collected water was mixed constantly with a magnetic stir plate and six 20-µl samples per replication were collected and conidia were visually counted using a hemocytometer. Some samples were plated on PDA to confirm growth morphology. Mean number of conidia collected per month was analyzed using the linear mixed model, with treatment and month as fixed factors.

Sporulation on cane pieces. To assess the effects of treatments on subsequent sporulation, three canes per vine were collected at the end of the growing season, after formation of periderm tissues on the canes (4 November 2003 and 10 February 2005).

These 1-yr-old canes, with or without symptoms, were randomly chosen from the main trunk of each treatment replication. Each cane was cut into three 10-cm sections, then 75 surface disinfested for 5 sec with 70% EtOH, followed by 1 min of 0.5% sodium hypochlorite (10% Clorox), and 2 min of rinse with deionized H2O. Canes were then placed in a moist chamber at ~20 °C for 3 wk to induce fungal sporulation. The moist chamber consisted of a 15-cm plastic Petri dish with a moistened paper towel at the bottom, covered with a 0.5-cm mesh steel screen to prevent the canes from touching the paper towels directly. The lids were sealed with Parafilm.

Sporulation of P. viticola was assessed by counting the number of pycnidia with cirrhi per square centimeter of cane. A total of nine 1-cm2 sections per replication (36 per treatment) were assessed. A linear mixed model was used to analyze the efficacy of the treatments on sporulation with treatment as a fixed effect factor.

RESULTS

Disease incidence and severity – ‘Concord’. In 2003, spring application of calcium polysulfide (lime sulfur) or fixed copper (COCS) both provided a significant

(P≤0.05) level of control of Phomopsis cane and leaf spot (Table 4.1). Leaf disease incidence was decreased by 30% by a spring application of calcium polysulfide, and cane internode incidence was decreased by 50% compared to the unsprayed control. Leaf severity was decreased by 42% with spring application of either fungicide, and internode severity was decreased by 84% and 64% with spring application of calcium polysulfide and spring application of fixed copper, respectively (Table 4.1).

76 Generally, there were no significant differences in disease levels between the spring only application and the fall-and-spring application of calcium polysulfide; however, the fall-and-spring application of fixed copper generally resulted in significantly higher disease intensity compared to the spring-only application of the same fungicide (Table 4.1). Overall, a fall-only application of fungicide resulted in disease incidence and severity not different from the unsprayed control.

In 2004, there was higher disease incidence and severity, but percentage control provided, especially for disease severity, by the treatments was similar to that found in

2003. Spring application of calcium polysulfide provided better control than did the fall application, resulting in 28% and 25% reduction in disease incidence of leaves and internodes, respectively, and 70% and 75% reduction in disease severity of leaves and internodes, respectively, compared to that with the unsprayed control. Spring application of fixed copper provided 22% and 6% reduction in disease incidence of leaves and internodes, respectively, and 62% and 56% reduction in disease severity of leaves and internodes. However, only leaf disease incidence and severity for the spring-only application were significantly lower than that in unsprayed control (Table 4.1). The fall- and-spring dormant application of fixed copper provided poorer control than just the spring application of the same fungicide, as in the 2003 season. Again, fall-only application was not effective for either fungicide.

Overall, rachis incidence in the ‘Concord’ vineyard was less in 2003 than in 2004

(Table 4.2). The spring dormant application of calcium polysulfide was the only treatment that provided significant level of control in 2003 (Table 4.2). Although most other treatments tended to have lower means than the control, there was high variability,

77 and no significant differences among these means. Results from 2004 showed that the spring and the fall-and-spring dormant application of calcium polysulfide, and fall-and- spring dormant application of fixed copper resulted in significantly lower incidence of rachis infection than that in the unsprayed control. Nonetheless, even with the best treatment, at least 44% of the rachises were infected in 2004.

Disease incidence and severity – ‘Catawba’. For the experiment in the

'Catawba' vineyard, the results were consistent in both years (Table 4.3). The dormant application of calcium polysulfide in spring provided moderate level of control, with disease incidence and severity of leaves and internodes significantly lower (P≤0.05) than that found in the unsprayed control. The percent control of disease severity ranged from

75% to 98% (Table 4.3). However, the calendar-based protectant applications of mancozeb provided significantly lower disease incidence and severity than did the dormant fungicide treatment (Table 4.3). Percent control based on severity ranged from

87% to 100%.

On the evaluation on rachis infections in the ‘Catawba’ vineyard, there were no significant differences (P≤0.05) among treatment means in 2003 (Table 4.4), even though leaf and internode disease incidence and severity were lower when treated with dormant application of fungicide or with mancozeb on a protective schedule (Table 4.3). Even though only 13% and 11% of leaves and internodes were infected when treated with mancozeb, respectively, 45% of rachises were infected with the same treatment. In 2004, both the calendar-based applications of mancozeb and the dormant application of calcium 78 polysulfide resulted in significantly lower rachis incidence means than that found in unsprayed control (Table 4.4). Again, at least 44% of the rachises were infected with the best treatment in 2004.

Collection of spores in splashed rain water. In 2003, the peak of spore discharge was relatively early (mid-April) in the field, with a secondary peak around late

May (Figure 1). In 2004, the peak was around mid-May (Figure 1). In 2003, there was a significant interaction of treatment and sampling month on spore density (P≤0.05). The control had a significantly higher mean than treated vines for the first month after spring application, but not for the second month (Table 4.5), suggesting a short time effect of the fungicide.

In 2004, month and treatment both had significant effects (P≤0.05), but the interaction was not significant (P≤0.05). Thus comparisons are limited to the main-effect means. Mean spore density trapped under treated vines was significantly less than that found for unsprayed vines (Table 4.5), and the mean was higher in the first month compared with the second. Because of the lack of interaction, effectiveness of the fungicide did not depend on sampling month.

Sporulation on cane pieces. In 2003, there were significant reductions (P≤0.05) in density of mature pycnidia on the canes collected at the end of the season that were treated with a calcium polysulfide dormant application in the ‘Concord’ study (Table 4.6).

Reductions in number of mature pycnidia were about 55% compared to the control, regardless of timing of application. Means for fixed copper treatments were not significantly different from that in the unsprayed control (Table 4.6). 79 Results from 2004 season were similar to that in 2003. Dormant application of calcium polysulfide in both fall and spring, and also the spring application of fixed copper, resulted in mean density of mature pycnidia significantly less than that in the unsprayed control (Table 4.6). Although there was about a 57% reduction in the mean density of mature pycnidia, the mean for the spring application of calcium polysulfide was not significantly different from that in the unsprayed control (Table 4.6).

DISCUSSION

Over 2 years, the efficacy of dormant applications of two different fungicides

(calcium polysulfide and fixed copper) was examined under field conditions for control of Phomopsis cane and leaf spot of grapes. The results from the ‘Concord’ vineyard indicate that calcium polysulfide application generally suppressed the disease on leaves and internodes better than did fixed copper application when applied in the spring (Table

4.1). With this application timing of calcium polysulfide, mean disease incidence or severity on leaves and internodes were reduced relative to that obtained with the unsprayed vines. However, in some cases, there were no significant differences between the means for the different dormant treatments, but means for spring application of calcium polysulfide were significantly different from the unsprayed control (e.g., internode incidence in 2003)(Table 4.1).

80 In the 'Catawba' vineyard, the calendar-based mancozeb application had consistently better control than did the dormant spring application of calcium polysulfide that also significantly reduced disease incidence and severity (Table 4.3). This showed the importance of protection of the canes and internodes during the spring in order to control Phomopsis cane and leaf spot.

Fall applications of both chemicals did not provide significant effects on the disease control (Table 4.1), and would not be recommended for disease management.

With calcium polysulfide, the fall-and-spring application tended to have lower disease intensity on leaves and canes compared to that with spring-only application, but the differences were not significant or consistent (Table 4.1) which suggests that the contribution of the fall application of the fungicide to the degree of disease control was low. Applications of fixed copper in the fall-and-spring resulted in significantly higher disease incidence and severity than that with application just in the spring for some years and measurement of disease (Table 4.1). Since the treatment was applied 2 years in row to the same vines, it is possible that one particular vine with higher inoculum doses affected the results (data not shown). Because of poorer and inconsistent control, a dormant application of fixed copper would not be recommended for management of P. viticola.

Often, disease severities of treated vines were significantly reduced by fungicide, whereas incidences were not, or percent reduction in incidence was much less than percent reduction in severity (Tables 4.1 and 4.3). The large reductions in internode

81 disease severity were especially noteworthy, because infected canes are the source of the inoculum in subsequent years (16, 18). However, observations from incubation of canes showed that mature pycnidia can be formed on non-symptomatic canes (Nita, unpublished).

Since a dormant application of a protectant fungicide does not protect susceptible plant tissues produced after application, other mechanisms must underlay the measured control. One explanation for the positive effect of dormant application of lime sulfur is in reduction of spore production that can potentially infect new tissues. In our study, reduction of conidia density in splashed rain water was observed during spore sampling in the spring, where significantly fewer conidia were collected under treated vines compared to the control vines (Table 4.5). However, there was still a substantial amount of spore release with dormant fungicide applications during both years (Figure 1, Table

4.5).

Since spores were collected with rain water, absence of significant rain events from 27 April to 7 May 2004, and relatively less rainfall in 2004 than 2003 may have affected the monthly spore total results (Figure 1, Table 4.5). Also, the variability of rain water collection should be noted. Although spore traps were placed in a relatively small field (~25 m x 15 m), there was large variation in the amount of rain water collected in each trap. Sometimes, a spore trap collected more than 1 L of rain water while another trap located ~2-m away collected no water (Nita, unpublished). This may due to slight difference of angle of the funnel of the traps, raindrop trajectory, or rainfall heterogeneity.

82 To provide stronger evidence of the effects of the dormant fungicide application on the sporulation of P. viticola, more consistent sampling methods or a larger number of sampling sites are needed.

A study conducted in California by Cuccuza et al. (1) showed that the peak of inoculum production of P. viticola is in early-April, followed by a gradual decline over the course of the season. Also, grape tissues seem to be more susceptible to P. viticola infection during the early stages of growth (19). Relative proximity of the source of the inoculum (infected old canes) and the susceptible young tissues in early growth stages in the spring increases the chance of infection because the conidia are splashed by rain, and, in many cases, the distance traveled by dispersal of spores is short (i.e., < 1m) (11, 13,

14). The limited range of spore dispersal was also indicated by the lack of spatial autocorrelation of disease in neighboring vines within vineyards (Chapter 6). Thus, the reduction of spore release early in season by the dormant fungicide application may have played an important role in reduction in disease incidence and severity we observed.

This tendency of reduction of sporulation activity by the dormant application of fungicide was also discussed previously (1, 12).

A potential longer term effect of dormant application of fungicides could be seen in lowered inoculum density on canes in the spring following fungicide application prior to the growing season. Calcium polysulfide treatments significantly reduced the density of mature pycnidia with cirrhi, compared with the control (Table 4.6). Fixed copper fungicide was clearly less effective as a dormant application compared with lime sulfur in this regard. Even though there were significant reductions in density of mature pycnidia with both the spring and fall-and-spring dormant application of calcium polysulfide, there

83 were still 5-6 pycnidia per cm2 on average, which could produce numerous conidia

(Table 4.6). Since overall disease intensity on leaves and internodes increased from 2003 to 2004 when the same treatments were applied to the same vines, the dormant application of fungicide alone clearly does not provide satisfactory long-term control of P. viticola. This lack of long term effects by the dormant fungicide application was discussed previously (1, 12).

Spring and fall-and-spring dormant application of calcium polysulfide also resulted in lower rachis infection in the 'Concord'-vineyard study and in one year in the

'Catawba'-vineyard study (Tables 4.2 and 4.4). Reduction in the disease by this fungicide applied in spring 51-84% in the 'Concord' vineyard. More variation in disease control was found in the ‘Catawba’ vineyard, and it affected the mean separation results (Table

4.4). In general, better control was provided by the calendar-based protectant schedule.

Since previous studies (18, 19) indicated that application of protective fungicide (e.g., captan) from 2.5-cm to 12.5-cm growth of grape was key to control Phomopsis cane and leaf spot, it was noteworthy to see a significant reduction of rachis disease incidence provided by the spring, or fall-and-spring, dormant application of calcium polysulfide

(Tables 4.2 and 4.4).

Overall, there is apparent benefit in applying lime sulfur early in the spring in order to control Phomopsis cane and leaf spot. In general, dormant application of lime sulfur around bud-swell provided moderate disease control, around 25-50% and 42-84% in disease incidence and severity, respectively (Tables 4.1 and 4.3). Other limited studies

(7, 21) showed very similar results. There was, however, variation among years and experiments, and the magnitude of control is not sufficient to recommend this treatment

84 alone for season-long control of Phomopsis cane and leaf spot. However, in this study, even a calendar-based 7-day interval application of protective fungicide provides only

57-88 % control of disease incidence (Table 4.3). Because frequent rains in the spring can prevent the timely application of protective fungicides, an application of a dormant fungicide can provide some level of control during this critical period when plant tissues would be otherwise unprotected. A dormant fungicide application in combination with a standard protective fungicide program may lead to better control of Phomopsis cane and leaf spot of grape than either program alone. The economics of using dormant applications of fungicides in relation to the amount of disease control needs to be studied.

Dormant applications of fungicide may have greatest potential in organic grape production systems where the choices of fungicides are currently very limited. Calcium polysulfide, as well as fixed copper, is registered for use with organic production of grapes (3). A study showed that the dormant application of calcium polysulfide (at 284

L/ha) can also reduce powdery mildew intensity on grape by reducing viable cleistothecial numbers (6). In addition, protective foliar applications of lime sulfur (4.5 kg/ha) applied weekly can be used as an effective control for P. viticola (Chapter 2), and Uncinula necator in grape production (3, 16).

85 Leaf Node Leaf Node Treatmenta Timingb %c PCd %c PCd #e PCd %c PCd Calcium polysulfide Fall 77.3 af -44.9 77.3 abf 17.4 4.8 af -67.5 1.4 abf 36.0 Calcium polysulfide Spring 37.5 bc 29.7 47.1 bc 49.7 1.7 bc 41.5 0.4 c 84.0 Calcium polysulfide both 47.1 c 11.8 36.4 c 61.1 1.2 c 58.1 0.4 c 84.0 Fixed copper Fall 58.4 ac -9.6 92.8 a 0.8 3.2 ac -12.1 2.0 ab 12.9 Fixed copper Spring 66.1 c -24.0 54.6 bc 41.6 1.7 bc 41.5 0.8 bc 64.0 Fixed copper Both 37.5 ab 29.7 92.8 a 0.8 4.0 ab -38.4 2.0 a 12.9 Control --g 53.33 ac 93.56 a 2.89 ac 2.25 a

2004 Incidence Severity Leaf Node Leaf Node Treatmenta Timingb %c PCd % PC #e PC % PC Calcium polysulfide Fall 91.07 abf 6.6 95.04 ab 4.7 6.25 ab 35.0 4.00 ab 30.6 Calcium polysulfide Spring 69.93 b 28.3 74.89 bc 24.9 2.89 b 69.9 1.44 bc 75.0 Calcium polysulfide both 71.18 b 27.0 66.13 c 33.7 3.24 b 66.3 0.81 c 85.9 Fixed copper Fall 81.94 ab 15.9 98.41 ab 1.4 5.76 ab 40.1 3.61 ab 37.3 Fixed copper Spring 76.10 b 21.9 93.56 ac 6.2 3.61 b 62.4 2.56 ac 55.6 Fixed copper Both 91.07 ab 6.6 98.81 a 1.0 8.41 a 12.5 4.00 ab 30.6 Control --g 97.48 a 99.77 a 9.61 a 5.76 a

a Calcium polysulfide (lime sulfur) was applied at rate of 94.7 L/ha and fixed copper (COCS) was applied at 3.4 kg/ha b Fall and spring application for the 2003 season was applied at 7 November 2002 and 17 April 2003, respectively, and fall and spring application for 2004 season was applied at 6 November 2003 and 20 April 2004, respectively. c % = back-transformed value of mean disease incidence or severity. Disease severity was estimated by visual assessment of proportion of internode covered by lesions. Disease incidence is a proportion of leaves or internodes showing visible symptoms of Phomopsis cane and leaf spot to healthy leaves or internodes. Disease incidence or severity were angular transformed,[ arcsin( proportion )], before analysis with Proc MIXED of SAS. d PC=Percent control= 100[(C-T)/C], where C is disease intensity (incidence or severity) for the unsprayed control and T is disease intensity for the treatment e # = back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. f Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. g unsprayed control

Table 4.1. Efficacy of dormant applications of calcium polysulfide and fixed copper, and their application timing, on disease incidence and severity of Phomopsis cane and leaf blight in a ‘Concord’ vineyard in 2003 and 2004 in Ohio

86

2003 2004 Treatmenta Timingb #c PCd #c PCd Calcium polysulfide Fall 33.2 abe 33.6 75.4 ac 15.3 Calcium polysulfide Spring 8.0 b 84.0 44.0 c 50.6 Calcium polysulfide Both 14.6 ab 70.8 44.5 c 50.0 Fixed copper Fall 34.2 ab 31.6 92.0 a -3.4 Fixed copper Spring 19.6 ab 60.8 84.6 ab 4.9 Fixed copper Both 14.6 ab 70.8 55.5 bc 37.6 Control --f 50.0 a 89.0 a

a Calcium polysulfide was applied at 95 L and fixed copper was applied at 3.4 kg per ha in 948L of water b Fall and spring application for the 2003 season was applied at 7 November 2002 and 17 April 2003, respectively, and fall and spring application for 2004 season was applied at 6 November 2003 and 20 April 2004, respectively. c Back-transformed value of mean disease incidence. Disease incidence is a proportion of rachises showing visible symptoms of Phomopsis cane and leaf spot to healthy leaves or internodes. Disease incidence were angular transformed,[arcsin( proportion )], before analysis with Proc MIXED of SAS. d PC=Percent control= 100[(C-T)/C], where C is disease incidence for the unsprayed control and T is disease incidence for the treatment e Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. f Unsprayed control

Table 4.2. Efficacy of dormant fungicide applications on disease incidence of rachises of Phomopsis cane and leaf spot in a ‘Concord’ vineyard in 2003 and 2004 in Ohio

87

2003 Incidence Severity Leaf Node Leaf Node Treatmenta Timingb %c PCd %c PCd #e PCd %c PCd Calendar Mancozeb 7-d 13.1 cf 81.0 10.5 cf 87.8 0.3 cf 96.8 0.0 bf 100.0 Dormant Calcium polysulfide Spring 33.8 b 51.0 28.2 b 67.3 1.1 b 88.4 0.1 b 97.7 Control --g 69.0 a 86.2 a 9.5 a 4.3 a

2004 Incidence Severity Leaf Node Leaf Node Treatmenta Timingb %c PCd % PC #e PC % PC Calendar Mancozeb 7-d 52.2 cf 46.8 33.3 c 66.7 1.7 c 86.7 0.7 c 94.0 Dormant Calcium polysulfide Spring 68.8 b 29.9 81.0 b 18.9 3.2 b 75.0 2.2 b 81.2 Control --g 98.1 a 100.0 a 12.8 a 11.7 a

a Calcium polysulfide (lime sulfur) was applied at rate of 94.7 L/ha, mancozeb was applied at 4.5 kg/ha in 948L of water b Spring application of calcium polysulfide were applied at 17 April 2003, and 20 April 2004. Application of mancozeb were made from 29 April 2003 to 26 June 2003 (9 applications), and from 6 May 2004 to 17 June 2003 (7 applications). c % = back-transformed value of mean disease incidence or severity. Disease severity was estimated by visual assessment of proportion of internode covered by lesions. Disease incidence is a proportion of leaves or internodes showing visible symptoms of Phomopsis cane and leaf spot to healthy leaves or internodes. Disease incidence or severity were angular transformed,[ arcsin( proportion )], before analysis with Proc MIXED of SAS. d PC=Percent control= 100[(C-T)/C], where C is disease intensity (incidence or severity) for the unsprayed control and T is disease intensity for the treatment e # = back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. f Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. g unsprayed control

Table 4.3. Efficacy of a dormant application of calcium polysulfide and of a calendar- based mancozeb application on disease incidence and severity of Phomopsis cane and leaf spot in the ‘Catawba’ vineyard in 2003 and 2004 in Ohio

88

2003 2004 Treatmenta Timingb #c PCd #c PCd Calendar Mancozeb 7-d 44.6 ae 47.5 44.0 b 52.5 Calcium polysulfide Spring 72.6 a 14.5 64.3 b 30.6 Control --f 84.9 a 92.6 a a Calcium polysulfide (lime sulfur) was applied at rate of 94.7 L/ha, mancozeb was applied at 4.5 kg/ha in 948L of water b Spring application of calcium polysulfide were applied at 17 April 2003, and 20 April 2004. Application of mancozeb were made from 29 April 2003 to 26 June 2003 (9 applications), and from 6 May 2004 to 17 June 2003 (7 applications). c PC=Percent control= 100[(C-T)/C], where C is disease intensity (incidence or severity) for the unsprayed control and T is disease intensity for the treatment d # = back-transformed value of the number of lesions per leaf. Disease severity was estimated by visual assessment of lesion numbers based on a 7-class scale (0=no lesion, 6=more than 100 lesions) per leaf, then square-root transformed before analysis with Proc MIXED of SAS. e Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. f unsprayed control

Table 4.4. Efficacy of dormant fungicide applications on rachis disease incidence of Phomopsis cane and leaf spot in a ‘Catawba’ vineyard in 2003 and 2004 in Ohio

89

Sporesa 2003 Time Treatmentb Mea Month 1c Month 2c n Calcium polysulfide 3056 bd 3426 b 3194

Control 9744 a 1542 b 6178 (unsprayed) Mean 6161 2434 4654 2004 Time Treatmentb Month 1c Month 2c Total Calcium polysulfide 4130 1637 3071 ae

Control 9029 2800 6750 b (unsprayed) Mean 6580 af 2183 b 4776

a Spores were collected with traps that were located next to the main trunk. Rain water was collected after each rain event, and then six samples were taken from the corrected water to count number of conidia per ml using a hemocytometer. b Calcium polysulfide (lime sulfur) was applied at 95 L per ha in 348L of water. Fall and spring application for 2003 season was applied at 7 November 2002 and 17 April 2003, respectively, and fall and spring application for 2004 season was applied at 6 November 2003 and 20 April 2004, respectively. c Month 1 in 2003 and 2004 represents periods between 17 April to 17 May 2003 and 20 April to 20 May 2004, respectively, and month 2 in 2003 and 2004 are between 17 May to 17 June 2003 and 20 May to 20 June 2004, respectively. d Means within a column or row for 2003 followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. e Means within a column for 2004 followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares month means determined with a linear mixed model fitted to the data. f Means within a row for 2004 followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares month means determined with a linear mixed model fitted to the data.

Table 4.5. Mean number of spores per milliliter of splashed rain per rain event in the first and second month after applications of dormant fungicides in spring

90

2003a 2004a Treatmentb Timingc #d PCe #d PCe Calcium polysulfide Fall 5.3 bcdf 54.3 8.0 abf 19.2 Calcium polysulfide Spring 4.9 cd 57.8 4.2 ac 57.6 Calcium polysulfide Both 4.6 d 60.3 2.1 c 78.8 Fixed copper Fall 8.7 ab 25.0 5.4 ac 45.5 Fixed copper Spring 7.8 acd 32.8 3.2 bc 67.7 Fixed copper Both 8.4 ab 27.6 5.3 ac 46.5 Control --g 11.6 a 9.9 a

a Canes were sampled at 2 November 2003 and 10 February 2005 b Calcium polysulfide (lime sulfur) was applied at rate of 94.7 L/ha and fixed copper (COCS) was applied at 3.4 kg/ha in 948L of water c Fall and spring application for the 2003 season was applied at 7 November 2002 and 17 April 2003, respectively, and fall and spring application for 2004 season was applied at 6 November 2003 and 20 April 2004, respectively. d # = back-transformed value of mean number of mature pycnidia with chirri per cm2. Number of mature pycnidia per cm2 was angular transformed,[ arcsin( proportion)], before analysis with Proc MIXED of SAS. e PC=Percent control= 100[(C-T)/C], where C is disease incidence for the unsprayed control and T is disease incidence for the treatment f Means within a column followed by same letter are not significantly different (P≤0.05) based on multiple comparisons of least-squares treatment means determined with a linear mixed model fitted to the data. g unsprayed control

Table 4.6. Mean number of mature pycnidia with cirrhi of Phomopsis viticola per cm2 of grape cane from the ‘Concord’ vineyard collected during the winter following dormant application of fungicides

91 80000 40 CP 2003 Control 60000 Rainfall 30

40000 20

20000 10

0 0 03 03 03 03 03 03 03 03 03 03 03 03 03 03 5/ 0/ 5/ 0/ /5/ 0/ 5/ 0/ 5/ 0/ /4/ /9/ 4/ 9/ 4/1 4/2 4/2 4/3 5 5/1 5/1 5/2 5/2 5/3 6 6 6/1 6/1

Conidia / ml 80000 40 2004 Rainfall (ml) CP Control 60000 Rainfall 30

40000 20

20000 10

0 0 04 04 04 04 04 04 04 04 04 04 04 04 04 04 5/ 0/ 5/ 0/ /5/ 0/ 5/ 0/ 5/ 0/ /4/ /9/ 4/ 9/ 4/1 4/2 4/2 4/3 5 5/1 5/1 5/2 5/2 5/3 6 6 6/1 6/1

Date

Figure 4.1. Number of P. viticola conidia per milliliter of splashed rain water in the ‘Concord’ vineyard, and amount of rainfall recorded in 2003 and 2004 (vertical bars). Arrows indicate the date of the spring dormant fungicide application of calcium polysulfide (CP). Solid lines represent mean number of conidia observed per rain event for the fall-and-spring application of calcium polysulfide, and dotted lines represent the mean for the unsprayed control.

92 REFERENCES

1. Cucuzza, J. D., and Sall, M. A. 1982. Phomopsis cane and leaf spot disease of grape vine: Effect of chemical treatments on inoculum level, disease severity, and yield. Plant Disease 66:794-797. 2. Ellis, M. A., Converse, R. H., Wiliams, R. N., and Williamson, B. ed. 1991. Compendium of raspberry and blackberry diseases and insects. APS Press, St. Paul, MN 3. Ellis, M. A., and Nita, M. 2004. Disease management guidelines for organic grape production in the Midwest. Plant Pathology Department Series 121 The Ohio State University OARDC/OSUE, Wooster, OH 4. Ellis, M. A., Welty, C., Funt, R. C., Doohan, D., Wiliams, R. N., Brown, M., and Bordelon, B. ed. 2004. Midwest Small Fruit Pest Management Handbook. Bulletin 861 Ohio State University Extension, Columbus, OH 5. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2001. Effect of growth stage on susceptibility of grape berry and rachis tissues to infection by Phomopsis viticola. Plant Disease 85 5:517-520. 6. Gadoury, D. M., Pearson, R. C., Riegel, D. G., Seem, R. C., Becker, C. M., and Pscheidt, J. W. 1994. Reduction of Powdery Mildew and Other Diseases by Over- the-Trellis Applications of Lime Sulfur to Dormant Grapevines. Plant Disease 78:83-87. 7. Gubler, W. D., and Baldwin, L. G. 2000. Evaluation of fungicides for the control of Phomopsis cane and leaf spot of grapes, 1999. Fungicide and Nematicide Tests 55:97. 8. Jones, A. L., and Aldwinckle, H. S. ed. 1990. Compendium of apple and pear diseases. APS press, St Paul, MN 9. Lal, B., and Arya, A. 1982. A soft rot of grapes caused by Phomopsis viticola. Indian Phytopathology 35 2:261-264. 10. Littell, R. C., Milliken, G. A., Stroup, W. W., and Wolfinger, R. D. 1996. SAS system for mixed models. SAS Institute Inc., Cary, NC 11. Madden, L. V. 1992. Rainfall and the dispersal of fungal spores. Advances in Plant Pathology 8:39-79. 12. Malathrakis, N. E., and Baltzakis, N. G. 1976. Control of Dead-arm of Grapes. Poljoprivredna znanstvena smotra - Agriculture Conspectus Scientificus 39 49:261-269. 13. McCartney, H. A., and Fitt, B. D. L. 1998. Dispersal of foliar fungal plant pathogens: Mechanisms, gradients and spatial patterns. 138-160 in: The epidemiology of plant diseases Jones, D. G. Kluwer, Dordrecht, the Netherlands 14. Mcdonald, O. C., and McCartney, H. A. 1987. Calculation of splash droplet trajectories. Agricultural and Forest Meteorology 39:95-110. 15. Mostert, L., Crous, P. W., and Petrini, O. 2000. Endophytic fungi associated with shoots and leaves of Vitis vinifera, with specific reference to the Phomopsis viticola complex. Sydowia 52 1:46-58.

93 16. Pearson, R. C., and Goheen, A. C. ed. 1988. Compendium of Grape Diseases. APS Press, St. Paul, MN 17. Pezet, R., Pont, V., and Girardet, F. 1983. Microcycle conidiation in pycnidial cirrhii of Phomopsis viticola Sacc. induced by elemental sulfur. Canadian Journal of Mycrobiology 29:179-184. 18. Pscheidt, J. W., and Pearson, R. C. 1989. Effect of grapevine training systems and pruning practices on occurrence of Phomopsis cane and leaf spot. Plant Disease 73 10:825-828. 19. Pscheidt, J. W., and Pearson, R. C. 1989. Time of infection and control of Phomopsis fruit rot of grape. Plant Disease 73 10:829-833. 20. Schabenberger, O., and Pierce, F. J. 2001. Contemporary statistical models for the plant and soil sciences. CRC Press, Boca Raton, FL 21. Schilder, A. M. C., Gillett, J. M., and Sysak, R. W. 2000. Evaluation of fungicides for control of phomopsis cane and leaf spot of grape, 1999. Fungicide and Nematicide Tests 55:105. 22. Sergeeva, V., Nair, N. G., Barchia, I., Priest, M., and Spooner-Hart, R. 2003. Germination of β conidia of Phomopsis viticola. Austrarian Plant Pathology 32:105-107.

94 CHAPTER 5

DISEASE INCIDENCE OF PHOMOPSIS CANE AND LEAF SPOT

OF GRAPE IN COMMERCIAL VINEYARDS IN OHIO

INTRODUCTION

Phomopsis cane and leaf spot is a disease of grape caused by the fungus

Phomopsis viticola (Sacc.) (14). Various tissues of the vine, such as leaf, cane, rachis, and fruit, can be infected by P. viticola. Infected leaves show small irregular, or round, pale green-to-yellow spots with dark centers. Brown to black necrotic irregular shaped lesions develop on the infected canes and rachises. Infected fruits appear as brown shriveled berries near harvest (14, 15). Infections on the canes and rachises weaken the plant and may cause premature fruit drop. Infections on fruits directly decrease yield and fruit quality. The disease is common in the U.S. and grape growing regions around the world (8, 10, 11). Up to 30% loss of the crop has been reported in Southern Ohio grape growing regions (6).

95 Selective pruning and protective fungicide applications are commonly used to control the disease (14, 15). Protective fungicides, such as mancozeb or captan, applied on a 7-10 day calendar-based schedule have been used for disease control (5, 14).

Research has indicated that the fungus becomes active early in the growing season, when only about 2.5-cm of the new growth of the grape tissues are present (3, 10, 15). During this period, the fungus, which had previously survived winter in infected old canes, produce pycnidia that contain numerous α-conidia. Rain splash of these α-conidia onto healthy susceptible tissues (such as newly developed leaves and internodes) may result in new infections early in the season. The close proximity between the source of inoculum

(infected canes from the main trunk) and the susceptible tissues (new growth on canes) may also favor infection in the early growing season compared to later. It has been shown that infection is dependent on ambient temperature and wetness duration (7).

Despite the fact that the disease is believed to be very common in Ohio (Ellis personal communication), there is no quantitative information about how widely the disease is distributed among the vineyards in the state, the mean incidence of the disease, and how growers’ management practices affect incidence of the disease. The objective of this study was to determine the distribution of the disease in Ohio, and to identify factors associated with incidence of Phomopsis cane and leaf spot of grape in commercial vineyards. A state-wide survey was conducted to collect data from growers in different parts of the state over 3 years. Then, effects of geographical, weather-related, or management-related factors on the disease were examined. A hierarchical linear mixed model (9) was used to quantify the variation in observed disease incidence among regions, farms within regions, vineyards within farms, and sampling sites within vineyards.

96 Generalized linear model analysis was used to develop risk models for disease from the survey data (17). Through construction of risk models, relationship between disease incidence and potential predictor variables were evaluated.

MATERIALS AND METHODS

Incidence of grape leaves and internodes infected by P. viticola was obtained though a state-wide survey from 2002 to 2004 in Ohio. Five major regions were selected within the state: 1) Lake Erie east (North-East), 2) Lake Erie west (North-Central), 3)

Central, 4) Ohio River Valley (South-West), and 5) Ohio Heartland area (Central-East)

(Table 5.1) A sixth region (Appalachian) was established after this study was done. The naming of regions is adapted from the Ohio Wine Production Association. Within each region, two-to-three commercial farms were selected randomly from a list of known farms, and two-to-three vineyards were randomly selected within each farm (Table 5.1).

Majority of cultivars examined were French-hybrid (Vitis intraspecific hybrid), except one location with American cultivars (Vitis labrasca). Clutivar was not used in selecting farms or vineyards. Locations of each vineyard were recorded with a GPS unit

(Magellan SporTreck Color, Thales Navigation, Inc, Santa Clara, CA).

In each vineyard, disease incidence was assessed on a regularly-spaced 8x8 grid of grape vines, which were selected systematically for the analysis of spatial pattern of disease (Chapter 6). Typically, there was ~3m between vines within rows, and ~2m

97 between rows. Three shoots per vine were randomly selected, and the first five leaves and internodes from the basal part of each shoot were visually assessed for Phomopsis symptoms. Thus, each sampling site (vine) consisted of 15 leaves and internodes.

Hierarchical analysis. A hierarchical linear mixed model was used to determine effects of region, farm, vineyard and sampling site on disease incidence. The model can be written as:

y = µ + R + G +V + S (1) where y is angular transformed disease incidence (i.e., arcsin( proportion) ) for leaves or internodes, R is effect of region, G is effect of farm within a region, V is effect of vineyard within a farm, and S is effect of site (vine) within a vineyard. All variables were considered random-effect factors, except region which was considered as a fixed-effect factor. The effect of region was determined with an F test. The random effects were evaluated with chi-square tests based on differences of 2x-log-likelihoods for sequential fits of the model (9). For example, effect of G term was determined by calculating 2x the differences of log-likelihoods for the model with and without the G term.

Development of risk models: Predictor variables. Ambient weather data were obtained from Ohio Agricultural Research and Development Center (OARDC) weather station records available on-line (12). Data from nearby weather stations from the surveyed vineyards were used. Both hourly and daily weather information were used to calculate sums, averages, and durations (h) of each specified condition.

98 Since there were no previous general attempt to relate epidemics of Phomopsis cane and leaf spot to weather data, except for predictions of infection for individual inoculation periods (see Chapter 2), a range of weather variables available from weather station datasets were taken into consideration. Because of splash dispersal nature of the disease (14), and requirements of leaf wetness and appropriate temperature for infection

(7), weather information regarding temperature, rain, and relative humidity (as a substitute for unavailable leaf wetness) were analyzed.

Initially, weather data from 1 April to 30 June of each year were considered for examination, and various time windows were used (e.g., 1-mo periods) to construct variables. Ultimately, variables representing weather in the last 10 days of April and the first 10 days of May were used for the model development, because of low correlation values of variables in other time periods with the disease incidence, and also because of our current understanding of the disease cycle. P. viticola spolurates greatly in early spring (3), and grape tissues are most susceptible when young (7). Also, previous studies indicated the importance of early season application of fungicides for control of the disease (14, 16). Since grape growth in Ohio generally initiates in late April and growth is vigorous in the first 2-3 weeks of a growing season, focusing on this time period was reasonable. Moreover, in terms of potential management variables (see below), almost all growers provided good fungicide coverage starting in the second-half of May; thus differences among surveyed vineyards were nil for disease management starting at mid-

May. All variables were scaled between 0 and 1 by dividing by the observed maximum to ensure uniformity of the variable scale.

99 Weather variables used are summarized in the first part of Table 5.2. Also, the infection model for Phomopsis cane and leaf spot (discussed in Chapter 2) was used to classify weather conditions into two categories of potential predicted infection events.

The model developed by Erincik et al. (7), to predict disease intensity from average temperature (T) and wetness duration (W) during an infection period (defined as wetness event triggered by rain), assuming inoculum was available, is given as:

z = α ⋅t β ⋅(1− t)γ ⋅W δ (2) where z is disease intensity (disease severity per internode or number of lesions per leaf),

Tmin and Tmax are assumed minimum (5 °C) and maximum (35.5 °C) temperatures, t = (T-

Tmin)/(Tmax-Tmin), and α (=1.8), β (=1.5), γ (=1.7), and δ (=1.1) are parameters estimated from their controlled-environment experiments and evaluated using inoculation in the field (7).

Since there was no leaf wetness records available in weather station datasets, precipitation durations were used to estimate leaf wetness duration by adding 1.5 h to the duration of precipitation events. The 1.5 h accounted for an average drying time of leaves. Using equation 2, predicted infection events were classified into two categories: light infection event (z>0 lesions per leaf, i.e., any potential infection events); or moderate infection event (z>30 lesions). DL and DM were then defined as the number of separate light and moderate, respectively, infection periods over the relevant 10-day time window (Table 5.2). Since the amount of inoculum in vineyards was unknown, only the relative value of z is relevant for predictions.

100

In addition to weather variables, age of vineyards, cultivar, and fungicide application records obtained from growers were also considered as potential predictor variables (“management variables”). Mail-in surveys were sent to growers each year to obtain fungicide application records. Based on the records, the sum of days with effective fungicide coverage was estimated for each vineyard (over the relevant time window) assuming that a protectant fungicide application was effective for 10 days after a spray. Only fungicides considered to be effective against P. viticola were considered (5,

14). This variable was also scaled between 0 and 1 by dividing by 10.

Stepwise logistic regression models. Since the economic threshold of this disease incidence is not known, two arbitrary thresholds of incidence were used to indicate a (relatively) high incidence of Phomopsis leaf and cane spot. Thresholds were either 40% or 20% disease incidence on leaves or internodes. Two disease variables were thus defined, D20 and D40. D20=1 if incidence is above 20%, and D20=0 otherwise; D40=1 if incidence is above 40% and D40=0 otherwise. For analysis purpose, incidence above 40% can be considered relatively poor control of the disease. Incidence below 40% (and even more so, below 20%) can be considered as an adequate control of the disease.

Logistic regression analysis was used to select variables and model the relationship between D20, D40, and predictor variables. The dependent variable for logistic regression is binary (i.e., above or below the threshold of disease incidence) with assumed binomial distribution of errors. The predicted response for a logistic model is

101 the logit of the estimated probability of D20=1 or D40=1 (i.e., “high” incidence) with a given set of predictor variables (1). Thus, logistic models predict the risk of disease, and can be considered as risk assessment models.

Each weather variable and management variable was examined using a logistic regression model and also examined using nonparametric correlation coefficients.

Kendall correlation coefficient was calculated for each predictor variable and D20 and

D40, and used for initial selection of variables. An achieved significance level (P) of <

0.1 for a correlation was used as a guideline for subsequent variable selection. A stepwise logistic regression was then used for selection of subset of predictor variables in models. The “stepwise” method in PROC LOGISTIC of SAS (SAS Institute, Cary NC) was used in the model selections. The criterion for acceptance was that the estimated coefficient for the variable was significant at P ≤ 0.10 .

Since the DL and DM predictor variables based on the infection model; (eq. 2) contain summarized weather conditions (i.e., temperature and precipitation duration), they were highly correlated with other precipitation variables. Thus, the DL and DM variables were examined separately from other weather variables in some logistic regression analyses. Overall, five separate stepwise logistic analyses were performed with combinations of variables: 1) weather condition variables; 2) the infection model variables; 3) management practice variables; 4) both weather and management variables; and 5) both infection model and management variables (see Table 5.3 for variable description). Separate analyses were done for leaves and internodes. From these

102 analyses, several candidate models were identified, and then these models were evaluated for prediction accuracy and interpretation based on known aspects of the epidemiology of the disease.

Each logistic model can be written as linear combinations of predictor variables, specifically:

⎛ p ⎞ ln⎜ ⎟ = p* = β 0 + β1 X 1 + β 2 X 2 +K (3) ⎝1− p ⎠ where p is the probability that D20=1 or D40=1, p* is the logit of p, Xi are predictor variables, and βi are parameters (coefficients). p can be obtained by back-transformation of the logit ( p =1/(1+ e− p * ) ). If predicted p is greater than 0.5, a high risk (e.g., D40=1) is predicted. Overall accuracy was determined by calculating the percent of all cases

(vineyards) where observation and prediction agreed. Two other forms of accuracy were determined, sensitivity and specificity. Sensitivity is the percent of high risk cases (i.e.,

D20 or D40=1) correctly predicted. Specificity is the percent of low risk cases (i.e., D20 or D40=0) correctly predicted. For each model, max-rescaled R2, -2ll (-2x log- likelihood), and standard errors for each estimated parameter were also calculated.

RESULTS

During 2002, 20 different vineyards from 11 different farms were evaluated for leaf and internode disease incidence. Due to the difficulty of the assessment (leaf symptoms due to an abiotic disorder on cultivar Chambourcin), leaf disease incidence data from one of the vineyards was omitted from the dataset; thus 19 and 20 vineyards for

103 leaf and internodes were evaluated, respectively. The mean disease incidence per vineyard varied from 4% to 87%, and the median disease incidence across all the vineyards was 38% and 57% for leaf and internode, respectively (Table 5.2). In 2003, the same number of vineyards was assessed, and the mean disease incidence per vineyard varied from 1% to 76%. Median disease incidence across all the vineyards was 42% and

41% for leaf and internode, respectively.

In 2004, 16 and 17 vineyards were assessed, for leaf and internode disease incidence, respectively, from 10 different farms. The number of vineyards and farms decreased due to a renovation of vineyards by one farm. Mean disease incidence per vineyard varied from 12% to 86%, and median disease incidence for all vineyards for leaves was 59% and that for internodes was 61%. Thus, the disease was found in all vineyards surveyed in all years, with wide range of disease incidence found over 3 years.

There was no overall trend in incidence with year.

Hierarchical analysis. Results from a linear mixed model indicated that disease incidence was significantly affected (P≤0.05) by farm within region, and vineyard within farm (Table 5.3), with higher variance in disease among sites within the vineyard than among vineyards or among farms (Table 5.4). However, regions were not significantly different (Table 5.3), indicating that incidence of disease did not systematically vary across the state. This suggests that the differences in observed disease incidence in different vineyards or farms were not due to regional scale variables such as climate, but probably due to much smaller spatial scale factors such as local weather conditions, landscape, specific management methods, or location of initial inoculum.

104

Grower survey. Of the10 growers asked to respond to the mail-in survey, eight responded in all 3 yr. One vineyard was experimental vineyard that received no fungicide applications. Thus, datasets from 18 and 17 vineyards, for internodes and leaves, respectively, in 2002 and 2003, and from 14 and 13 vineyards for internodes and leaves, respectively, in 2004 were used for analyses.

Preliminary evaluation of predictor variables. The Kendall correlation coefficient between mean disease incidence of leaves and internodes across all vineyards was 0.31, 0.59, and 0.54, for 2002, 2003, and 2004, respectively (Table 5.6). Thus, there was only moderate correlation in disease of internodes and leaves, suggesting the separate analyses for the two disease variables.

Kendall correlation coefficients were obtained between calculated prediction variables (Table 5.5) and risk index variables (D20 or D40). The variables most positively correlated with leaf disease risk were TA (average temperature in °C from 21-

30 April) and DLA (sum of predicted light infection events from 21-30 April). Variables most negatively correlated with leaf disease risk were PPTDM (sum of days with precipitation from 1-10 May) and DOR (dormant application of fungicide) for in both

D20 and D40.

For internodes, CA (sum of days with fungicide coverage from 21-30 April) and

RHA (Daily average relative humidity from 21-30 April) were most positively correlated with disease risks, and CM (sum of days with fungicide coverage from 1-10 May), and

DOR (dormant application of fungicide) were most negatively correlated with risks.

105 Age of vine (variable AGE) was not significant in most of cases (Table 5.5), probably because of relatively old vine age of most surveyed vineyards (average vineyard age was 31 years, with youngest vineyard of 10 years old). Considering disease incidence was found in all vineyard surveyed, there appeared to have been sufficient time to distribute P. viticola throughout vineyards.

Stepwise logistic regression. Stepwise logistic regression indicated several possible variable combinations to predict risk (Table 5.7). Generally, variables indicated by stepwise logistic regression were the ones with high Kendall correlation coefficients with D20 and D40 (Table 5.6). In general, environmental conditions in late April, especially temperature and relative humidity, fungicide coverage during early May, and dormant application of fungicide were the main contributing factors to high risk. Models with a single type of variable, either weather, predicted infection events, or management factors, tended to have lower R2 values than models with combinations of different types of variables (Table 5.7). Moreover, models with predicted infection event variable (see eq. 2) alone or in combination with management variables had relatively low R2 values

(0.23-0.50) compared with models with other variables (Table 5.7). Even though models based on D20 as the response variable tended to have fewer number of weather predictor variables than models based on D40 as the response variable, R2 values tended to be higher for the former.

It is well established that stepwise regression methods should only be used to identify potential subsets of predictor variables, and not to derive a single final model (1).

There may be alternative models with slightly different combinations of variables that

106 result in high accuracy on the current or similar datasets, or which are more logical in terms of current understanding of the system. Thus, alternative models were constructed by adding or subtracting individual variables to the models identified with stepwise regression (Table 5.8). In addition, a model with the same predictor variables (labeled a

“generic” model) was fitted to the leaf and internode data for both response variables

(D20 and D40). The general model is based on average temperature and relative humidity from 21 to 30 April (variables TA and RHA, respectively) and dormant application of fungicide (variable DOR). Estimated parameters, sensitivity, specificity, and accuracy of predicting epidemics of each model are listed in Table 5.8.

In general, variables with high Kendall correlation coefficients (Ks) were significant in the logistic regression models, and models containing the variables with the highest Ks had high sensitivity and specificity, and overall accuracy (Tables 5.6 and 5.8).

However, R2 values of models did not seem to have a systematic relationship to overall accuracy (Table 5.8). A high R2 value corresponded to either high specificity or sensitivity, but not necessary both, and did not always correspond to a high overall accuracy.

High overall accuracy was found for some models, but sometimes this value could be misleading, because either sensitivity or specificity could be low. For example, the P model for internode incidence (D40) had a 67% overall accuracy (Table 5.8), but failed to correctly predict any of the low risk (i.e., D40=0) cases (specificity = 0%). In this example, all observations were predicted as being high risk (D40=1).

107 Proposed models that were aimed at predicting D40 (i.e., incidence >40%) resulted in averages of 70% and 65% overall accuracy for leaves and internodes, respectively (Table 5.8). The lowest overall accuracy of those shown was 41% (‘generic’ model), and highest overall accuracy was 83% (M and P+M models). Models based on a single type of variable, such as weather (W) or predicted infection event (P), tended to have lower overall accuracy than the models with a combination of weather and management variables. An exception was the situation of predicting D40 for internodes.

Despite the high specificity (100%) for the W+M model (Table 5.8), sensitivity was low

(28%), resulting in low overall accuracy (52%).

Models developed to predict D20 had higher overall accuracy in general, compared with models to predict D40 (Table 5.8). Overall model accuracy was 90% for leaves and 89% for internodes, on average (Table 5.8). For prediction of leaf disease risk, weather variables provided high overall accuracy (94%), with high sensitivity (98%), and specificity (75%), whereas other combinations of variables tended to have lower specificity, ranging 25 to 38%. CA (fungicide coverage from 21 to 30 April) alone provided a high overall accuracy (89%), with 100% sensitivity and 89% specificity

(Table 5.8).

For prediction of internode disease risk, dormant application of fungicide tended to be key factor. A model with only DOR provided high 91% overall accuracy, with high sensitivity (91%) and specificity (89%) (Table 5.8). Weather and predicted-event variables also had high overall accuracy (87%); however, their specificity was low (22%).

With low specificity, many cases with D20=0 were predicted to be high risk (D20=1).

108

DISCUSSION

Phomopsis cane and leaf spot of grape has been known to be present in Ohio’s commercial vineyards for many years (Ellis personal communication), either as this name or as a component of the “dead-arm disease” complex of grape, caused by either P. viticola or Eutypa lata (2). In this study, Phomopsis on grape was found in all vineyards surveyed in all years, with a mean disease incidence per vineyard ranging from 4% to

86% (Table 5.1). Median incidence over all years was 48%, and there was no trend in disease incidence over years (Table 5.1).

A hierarchical linear mixed model was used to quantify the variation in disease incidence at different geographical scales in Ohio. Variation in disease incidence was greatest in the lowest strata, that is, sites (vines) within vineyards (Table 5.4). Moreover, regions did not affect disease incidence in the state (P≤0.05) (Table 5.3). Thus, there was no evidence of overall difference in mean disease incidence among regions of the state, indicating that large-scale factors, such as climate and landscape, were not substantially affecting disease. Results also suggest that sufficient inoculum (in the form of infected canes) is present in the state to cause infections in all sampled regions.

Within regions, there was significant variation in disease among farms, vineyards within farms, and sampling sites within vineyards. Thus, there was evidence that smaller-scale factors, such as local weather patterns, varieties, soil conditions, location of inoculum, amount of inoculum per vineyard, and disease management practices could be affecting Phomopsis disease incidence. Although the greatest variation in disease, in

109 2 general, was at the site (vine) scale, as measured by the estimated σ s , there was still significant variation between vineyards and farms. The logistic regression analysis and nonparametric correlation analysis was conducted to determine if environmental or management practices could be associated with the variation in disease incidence among those vineyards and farms.

Since disease incidence on a continuous percentage scale (leaves or internodes) had substantial variation within vineyards, incidence was classified into categories prior to the analysis in order to more directly relate potential predictor variables to disease risk.

Similar analytical approaches, for instance, have been taken for the risk of gray leaf spot of maize (13) and Fusarium head blight of wheat (4). A high risk event is defined as mean disease incidence being above the disease threshold (either 40% or 20%, defined as

D40 and D20, respectively). Two different risk levels were examined because there was limited information on economic threshold of the disease.

In general, results from logistic regression analyses showed that combinations of weather variables, especially average temperature or relative humidity during 21-30 April, and dormant application of fungicide, can predict risk satisfactorily (Table 5.6). The positive coefficients for the weather variables TA, TM, and RHA in one or more logistic models indicated that risk of disease incidence increased with increasing TA, TM, and

RHA. Thus, increasing temperatures in late-April and early-May, and increasing relative humidity in late-April may favor the active growth of mycelia and sporulation of the fungus in the infected tissue, as well as infection of leaves and internodes.

110 However, models comprised of just weather variables (W models) or predicted infection events (P models) alone tended to have imbalance in prediction of risk; that is, models had either low sensitivity or specificity even when overall accuracy was reasonably high (Table 5.6). One source of variability in the relation between disease and weather is the distance between vineyards and weather stations. On-site measurements may have resulted in less variable model results.

Management variables, on the other hand, predicted risks well for both D40 and

D20 models, even if other weather variables were not used in the model (Table 5.6). The most important variable for risk prediction was DOR for most models (D20 or D40, for leaf or internode). Because of the estimated negative coefficient for DOR, use of a dormant application of fungicide significantly reduced the risk of disease. For example, with the leaf model for D40, the estimated probability of D40=1 (i.e., incidence >40%) is

0.57 when a dormant fungicide application is not applied and 0.17 when applied.

Based on the fungicide application records, growers applied either lime sulfur

(calcium polysulfide) at the rate of 95 L per ha in 948L of water or fixed copper (copper oxychloride) at a rate of 2.3 kg per ha in 948L of water. These treatments were commonly applied to control grape powdery mildew (caused by Uncinula nacator) or anthracnose (caused by Elsinoe ampelina) (5, 14). Other research (Chapter 4) with field experiments indicated that a dormant application of fungicide (where lime sulfur was applied at a rate of 95L per ha in 948L of water) provided consistent yet only moderate

(around 25-50% and 42-84% decrease in disease incidence and severity, respectively) control of Phomopsis cane and leaf spot.

111 In all cases, based on the grower survey, growers who applied a dormant application of fungicide also started their regular-season protectant fungicide applications on new plant tissue earlier than the others did (Nita, unpublished). These growers had better protectant coverage of plants during both time windows considered in the analysis

(and even later time windows which is not considered in the presented models) compared with other growers. Thus, DOR was positively correlated with CA and CM variables (all are management variables), and only one of the variables could generally be utilized in the model (1).

If DOR was replaced by either CA or CM in a logistic model, a significant coefficient was found, and the log-likelihood statistics of the models were similar, indicating a comparable fit. However, in most of the cases, models with DOR had better overall accuracy (Nita, unpublished) than those using CA or CM instead. Therefore, DOR was maintained in the many of the models considered. These results indicated the successful control of Phomopsis cane and leaf spot obtained with early-season fungicide applications.

Disease incidence was relatively high over the 3 years of the survey, and common throughout the state even though all growers were using protectant fungicide programs

(at least during some of the growing season). Unlike Phomopsis on grape, other economically important grape diseases in Ohio, such as black rot (caused by Guignadia bidwellii), were not observed commonly in the 3-yr survey, although no formal assessment of incidence of these diseases were made (Nita, unpublished).

112

Thus, it appears that Phomopsis on grape is the most difficult disease to control using current management practices. Successful management of Phomopsis on grape achieved by some farms clearly indicates improvements in control are possible for

Phomopsis on grape. Early season application of fungicide, dormant application of fungicide (Chapter 4), and the improvement of fungicide timing with the use of the disease warning system (Chapter 2) can increase efficacy and efficiency of control of

Phomopsis cane and leaf spot. Studies are also needed on the effect of this disease on fruit yield and quality, as well as long-term effects on infected vines.

113

Region Grower Longitude Latitude County Central A 39°45’N 82°49’W Fairfield B 40°08’N 82°33’W Licking C 40°28’N 83°00’W Marion South D 39°20’N 84°10’W Warren E 38°38’N 83°38’W Adams North-East F 41°54’N 80°37’W Ashtabula G 41°45’N 80°57’W Ashtabula North-West H 41°38’N 82°49’W Ottawa I 41°39’N 82°49’W Ottawa J 41°39’N 82°47’W Ottawa North-Central K 40°45’N 81°54’W Wayne

Table 5.1. Location of sampled sites for Phomopsis cane and leaf spot of grape disease incidence from 2002 to 2004 in Ohio

114

2002 2003 2004 Region Farma Variety Leaf Node Leaf Node Leaf Node Central A Vignol 47.5 48.9 49.3 40.4 67.7 59.3 Central A Vidal Blanc 29.2 54.5 38.7 45.2 55.3 50 Central A Chambourcin --b 17.5 -- 40.8 -- 41.7 Central B Seyval 50.2 61.3 63.4 44 -- -- Central B Rayon d’Or 50.4 66.6 58 46.6 -- -- Central B Catawba 38.6 86.7 54.6 75.1 -- -- Central C Vidal 42.4 37.5 33.5 30.8 37.2 31.1 Central C Chanceller 64.1 64.5 53.9 59.9 60.3 61.5 Central C Seyval 64.4 39.2 23.8 27.3 62.5 53.8 South D Vidal 20.3 19.2 44.4 34.6 20.8 19.2 South E* Vidal 25.4 56.1 68.7 73.9 76.4 72.3 North-East F Chardonel 3.9 11.6 7.2 1.4 15 11.7 North-East F Cabernet Sauvignon 5.9 13.8 20.5 14.8 28.4 34.9 North-East G* Concord 13.5 55.5 19.4 29 30.5 34.3 North-West H Niagra 17.6 58.5 29.8 31.7 56.1 85.9 North-West I Concord 6 36.6 13.5 19.8 22.9 42.2 North-West J Catawba 19.9 73.3 24.9 23.4 41.8 83.2 North-Central K Clinton 61.3 73.1 34.8 48.5 65 61.5 North-Central K Clinton 61 55.2 27.7 31.3 58.2 59.2 North-Central K Reliance 37.9 72.9 46.9 41.5 73.5 66.3 North-Central K Ives 30 82.5 34.6 66.7 60.4 65.4 MEAN 34.5 53.4 37.4 39.4 48.9 51.9 a Management information from farms with asterisks was not provided. Farm K is an experimental station with no fungicide applied during the season. b Missing data

Table 5.2. Mean disease incidence (%) of Phomopsis cane and leaf spot of grape per vineyard in percentage during 2002-2004 in Ohio for leaves and cane internodes.

115

Symbol Description Weather variable PPTS Sum of precipitation (mm) from 21 to 30 April (PPTSA), or from 1 to 10 May (PPTSM) PPTD Sum of days with precipitation (day) from 21 to 30 April (PPTDA), or from 1 to 10 May (PPTDM) PPTH Sum of hours with precipitation (hr) from 21 to 30 April (PPTHA), or from 1 to 10 May (PPTHM) T Average temperature (C) from 21 to 30 April (TA), or from 1 to 10 May (TM) RH Average relative humidity (%) from 21 to 30 April (RHA), or from 1 to 10 May (RHM) RH90 Sum of hours with RH >90% from 21 to 30 April (RH90A), or from 1 to 10 May (RH90M) Infection model variable DL Sum of predicted low (or higher) infection events from 21 to 30 April (DLA), or from 1 to 10 May (DLM) DM Sum of predicted moderate (or higher) infection events from 21 to 30 April (DMA), or from 1 to 10 May (DMM) Management variable C Sum of days with fungicide coverage from 21 to 30 April (CA), or from 1 to 10 May (CM), based on the assumption that a protectant fungicide provides coverage for 10 days DOR Dormant application of fungicide (0 if not used, 1 if used) Risk indicator variable D20 Risk indicator variable. D20=1 if observed disease incidence > 20%, and 0 otherwise D40 Risk indicator variable. D40=1 if observed disease incidence > 40%, and 0 otherwise Others AGE Age of vineyard (years)

Table 5.3. Potential predictor variables considered in development of risk model for Phomopsis cane and leaf spot of grape on leaves and cane internodes in commercial vineyards in Ohio over 3 years

116

Disease Year Ra Gb V Leaves: 2002 2.89 316.8* 184.7* 2003 3.74 129.1* 103.1* 2004 1.38 59.2* 234.3*

Internodes: 2002 1.29 153.1* 304.1* 2003 3.67 135.8* 123.2* 2004 3.22 103.3* 152.8* a F-statistic (asterisks indicates significant effect (P≤0.05)) b Likelihood ratio statistic, Asterisks indicates significant effect (P≤0.05) based on a chi-square test

Table 5.4. Test statistics for the effect of region (R), farm within region (G), and vineyard within farm (V) on incidence of Phomopsis cane and leaf spot of grape on leaves and cane internodes in Ohio over 3 years.

117

2 2 2 Disease Year σˆ G σˆV σˆ S Leaves: 2002 0.010 0.026 0.029 2003 0.005 0.015 0.041 2004 0.034 0.009 0.035

Internodes: 2002 0.030 0.021 0.047 2003 0.001 0.026 0.055 2004 0.016 0.017 0.042

2 2 Table 5.5. Estimated variances for farm within region (σ G ), vineyard within farm (σ V ), 2 and site within vineyard (σ S ) for incidence of Phomopsis cane and leaf spot of grape on leaves in internodes in Ohio over 3 years

118

D40a D20a Variableb Leaf Node Leaf Node c c c c PPTSA 0.15(0.21) -0.01(0.96) -0.02(0.84) -0.17(0.15)

PPTSM -0.04(0.70) 0.00(0.99) 0.17(0.15) 0.06(0.61)

PPTDA 0.14(0.25) -0.04(0.75) -0.13(0.3) -0.13(0.29)

PPTDM -0.30(0.02) -0.15(0.22) -0.01(0.93) -0.20(0.11)

PPTHA 0.13(0.34) 0.01(0.92) -0.12(0.33) -0.14(0.25)

PPTHM -0.13(0.26) -0.15(0.22) 0.04(0.75) 0.12(0.31)

TA 0.32(0.01) 0.09(0.43) 0.41(<0.01) 0.22(0.06)

TM 0.19(0.11) 0.14(0.24) 0.47(<0.01) 0.24(0.04)

RHA 0.20(0.09) 0.25(0.03) -0.14(0.22) 0.12(0.33)

RHM -0.13(0.26) -0.07(0.54) 0.13(0.27) -0.01(0.91)

RH90A 0.19(0.11) 0.18(0.12) 0.09(0.45) -0.07(0.54)

RH90M -0.12(0.30) -0.04(0.75) 0.01(0.90) 0(0.98)

DLA 0.40(<0.01) 0.30(0.02) 0.4(<0.01) 0.15(0.22)

DLM -0.19(0.13) -0.10(0.42) 0.03(0.82) 0.07(0.59)

DMA 0.01(0.96) -0.16(0.26) 0.03(0.83) -0.24(0.07) -0.15(0.27) -0.08(0.57) -0.15(0.28) 0.17(0.21) DMM

CA -0.13(0.32) -0.40(<0.01) -0.29(0.03) -0.45(<0.01)

CM -0.27(0.04) -0.52(<0.01) 0(0.99) -0.41(<0.01) DOR -0.34(0.01) -0.57(<0.01) -0.28(0.04) -0.72(<0.01) AGE 0.07(0.55) -0.06(0.62) 0.15(0.18) -0.23(0.05) a D40=1 if incidence >40%, and 0 otherwise. D20=1 if incidence >20%, and 0 otherwise. b List of abbreviations of variables are in Table 5.3. c Numbers in parentheses are significance levels for correlation coefficients

Table 5.6. Kendall correlation coefficients (Ks) between weather or management variables and indicator variables (D20, and D40) for the risk of Phomopsis cane and leaf spot of grape on leaves and cane internodes in Ohio over 3 years

119

Predictor variables selecteda Weather Infection Management Weather and Predicted events D40b Only events only Only Management and management Leaves: PPTHA DLA DOR TA DLA TA (0.23) (0.15) RHA DOR RHA PPTDM (0.35) (0.50) DOR (0.56) Internodes: PPTDA DLA CM PPTDA DLA RHA (0.26) DOR TA CM RHM (0.42) RHA DOR (0.54) PPTDM (0.50) CM (0.72) Predictor variables selected Weather Predicted Management Weather and Predicted events D20b Only events only Only Management and management Leaves: TA DLA CA TA DLA TM DLM (0.11) TA DLM (0.63) (0.49) (0.63) (0.49)

Internodes: TA n/a DOR RHA DOR RHA (0.60) DOR (0.60) (0.25) (0.71) a List of abbreviations of variables are in Table 5.3. Number in parentheses is maximum-rescaled R2 value. b D40=1 if incidence >40%, and 0 otherwise. D20=1 if incidence >20%, and 0 otherwise.

Table 5.7. Variables identified with stepwise logistic regression analysis with different subsets of variables for two different risk levels of Phomopsis cane and leaf spot of grape on leaves and cane internodes in commercial vineyards in Ohio over 3 years (D40 and D20).

120

Max- Overall rescaled accuracy Sensitivity Specificity Modela Model equationb R2 (%)c (%)d (%)e D40 Leaves: W -28.1 + 10.8TA + 21.6RHA 0.41 68.5 69.2 67.9 P -1.1 + 3.2DLA 0.23 66.7 65.4 67.9 M 0.3-1.9DOR 0.15 63.0 92.3 35.7 W+M -35.4 + 13.0TA + 28.3RHA - 2.7DOR 0.54 79.6 84.6 75.0 P+M -0.8+3.5DLA-2.0DOR 0.35 74.1 61.5 85.7 Generic -38.1 + 13.3TA + 28.3RHA - 2.7DOR 0.54 66.7 34.6 96.4

Internodes: W -14.6 + 12.6RHA + 4.5RHM 0.17 64.8 72.2 50.0 P 0.0 + 2.3DLA 0.11 66.7 100.0 0.0 M 1.6 - 3.2CM – 2.1DOR 0.42 83.3 91.7 66.7 W+M -34.5 + 5.3TA + 34.2RHA - 6.2PPTDA - 3.4DOR 0.64 51.9 27.8 100.0 P+M 0.8 + 3.4DLA - 3.8CM - 2.1DOR 0.50 83.3 91.7 66.7 Generic -25.1 + 3.8TA + 22.2RHA - 4.1DOR 0.59 40.7 11.1 100.0 D20 Leaves: W -22.5+5.0TA+27.1TM 0.63 94.4 97.8 75.0 P -1.3+11.6DLA+2.8DLM 0.49 88.9 97.8 37.5 M 2.0-4.0CA 0.12 88.9 100.0 88.9 W+M -21.8+4.5TA+26.7MA-2.8CA 0.64 94.4 97.8 75.0 P+M -1.3+11.6DLA+2.8DLM 0.49 88.9 97.8 37.5 Generic -20.2+16.3TA+11.8RHA-1.0DOR 0.54 88.9 97.8 37.5

Internodes: W -15.6+8.5TA+12.2RHA 0.25 87.0 100.0 22.2 P 0.0+3.1DLA+2.1DLM 0.15 87.0 100.0 22.2 M 3.7-4.4DOR 0.60 90.7 91.1 88.9 W+M -17.9+25.6RHA-6.8DOR 0.71 88.9 95.6 55.6 P+M 3.2+2.0DA-4.4DOR 0.62 90.7 91.1 88.9 Generic -69.5+25.0TA+67.4RHA-12.9DOR 0.82 92.6 97.8 66.7 a Type of models based on variables used. W=weather-related variables, P=infection-model-based variables (see eq. 2), M=management-related variables, Generic=same variables for both risk variables and disease types (leaves and internodes). b Logisitc linear model were developed using variables listed in Table 5.3. Model variables were rescaled before analyses by dividing by the maximum observed value which is 14.3, 17.7, 76.8, 10, 10, and 1 for TA, TM, RHA, DA, DM, and DOR, respectively. c Percentage of correctly classified cases d Percentage of correctly classified high risk cases (where observed D40 or D20 =1) e Percentage of correctly classified low risk cases (where observed D40 or D20 =0)

Table 5.8. Logistic linear models developed for predicting Phomopsis cane and leaf spot risks (D40 or D20), together with overall prediction accuracy, sensitivity, and specificity of models.

121

Figure 5.1. A map of major grape growing regions in Ohio. Regions surveyed in this study are circled. The map was obtained from http://www.ohgrapes.org/.

122 REFERENCES

1. Allison, P. D. 1999. Logistic regression using the SAS system: Theory and application. SAS Institute, Cary, NC. 2. Carter, M. V. 1960. Further study on Eutypa armeniacea Hansf. & Carter. Austrarian Journal of Agricultural Research 11:498-504. 3. Cucuzza, J. D., and Sall, M. A. 1982. Phomopsis cane and leaf spot disease of grape vine: Effect of chemical treatments on inoculum level, disease severity, and yield. Plant Disease 66:794-797. 4. De Wolf, E. D., Madden, L. V., and Lipps, P. E. 2002. Risk assessment models for wheat Fusarium head blight epidemics based on within-season weather data. Phytopathology 93:428-435. 5. Ellis, M. A., Welty, C., Funt, R. C., Doohan, D., Wiliams, R. N., Brown, M., and Bordelon, B. ed. 2004. Midwest Small Fruit Pest Management Handbook. Bulletin 861 Ohio State University Extension, Columbus, OH. 6. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2001. Effect of growth stage on susceptibility of grape berry and rachis tissues to infection by Phomopsis viticola. Plant Disease 85 5:517-520. 7. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2003. Temperature and wetness-duration requirements for grape leaf and cane infection by Phomopsis viticola. Plant Disease 87:832-840. 8. Lal, B., and Arya, A. 1982. A soft rot of grapes caused by Phomopsis viticola. Indian Phytopathology 35 2:261-264. 9. Littell, R. C., Milliken, G. A., Stroup, W. W., and Wolfinger, R. D. 1996. SAS system for mixed models. SAS Institute Inc., Cary, NC. 10. Malathrakis, N. E., and Baltzakis, N. G. 1976. Control of Dead-arm of Grapes. Poljoprivredna znanstvena smotra - Agriculture Conspectus Scientificus 39 49:261-269. 11. Mostert, L., Crous, P. W., and Petrini, O. 2000. Endophytic fungi associated with shoots and leaves of Vitis vinifera, with specific reference to the Phomopsis viticola complex. Sydowia 52 1:46-58. 12. OARDC 2005. OARDC weather system. OARDC (Ohio Agricultural Research and Development Center), http://www.oardc.ohio-state.edu/centernet/weather.htm. 13. Paul, P. A., and Munkvold, G. P. 2004. A model-based approach to preplanting risk assessment for gray leaf spot of maize. Phytopathology 94:1350-1357. 14. Pearson, R. C., and Goheen, A. C. ed. 1988. Compendium of Grape Diseases. APS Press, St. Paul, MN. 15. Pscheidt, J. W., and Pearson, R. C. 1989. Effect of grapevine training systems and pruning practices on occurrence of Phomopsis cane and leaf spot. Plant Disease 73 10:825-828. 16. Pscheidt, J. W., and Pearson, R. C. 1989. Time of infection and control of Phomopsis fruit rot of grape. Plant Disease 73 10:829-833. 17. Yang, X. B. 2003. Risk assessment: concepts, development, and future opportunities. Plant Health Progress:doi:10.1094.

123

CHAPTER 6

SPATIAL DISTRIBUTION OF PHOMOPSIS CANE AND LEAF

SPOT SYMPTOMS IN COMMERCIAL VINEYARDS IN OHIO

INTRODUCTION

Phomopsis cane and leaf spot is a common disease of grape in the U.S. and grape growing regions around the world (2, 15, 19). The fungus, Phomopsis viticola (Sacc.), is the causal agent of the disease (19). Various tissues of the vine, such as leaf, cane, rachis, and fruit can be infected by P. viticola. Infected leaves show small irregular, or round, pale green-to-yellow spots with dark centers. Brown to black necrotic irregular shaped lesions develop on the infected canes and rachises. Infected fruits appear as brown shriveled berries near harvest (15, 19). Infections on the cane and rachis weaken the plant and may cause premature fruit drop (23). Infections on fruits directly decrease yield and fruit quality. Up to 30% loss of the crop has been reported in Southern Ohio grape growing regions (4).

124 Typically, selective pruning and protective fungicide applications are used to control the disease (19, 22, 23). Protective fungicides, such as mancozeb or captan, applied on a 7-10 day calendar-based schedule, have been used for disease control (19,

23).

The fungus survives winter in grape cane tissues that were infected in previous years. In spring, numerous pycnidia rise from infected canes, and a gelatinous mass of spores (α- and β-conidia) called cirrhus exudes out from the pycnidium. Conidia are splashed by rain onto new growth where they infect the plant tissues. Function of the β- conidia is not well known (27).

Recently, several studies were conducted to reveal various aspects of Phomopsis cane and leaf spot epidemiology and control, such as timing of fungicide applications to protect plant tissue from infection (4, 23; Chapter 2), elucidation of the relationship between environmental conditions (especially wetness duration and temperature) and disease intensity (5), and pre-season application of fungicide to help manage the disease

(Chapter 4). However, spatial distribution of the disease was not studied extensively.

Spatial distributions of other major grape diseases, such as downy mildew (caused by

Plasmopara viticola) (13), phytoplasma (leafhopper and planthopper transmitted prokaryotes) (18), Eutypa dieback (caused by Eutypa lata) (8, 17), and Pierce’s disease

(caused by Xylella fastidiosa) (32), have been studied.

Knowledge of the spatial attributes of plant diseases benefits researchers in various ways, including: determining efficient sampling methods (12); identifying the conditions for disease development (29, 34); and selecting proper statistical methods for analyzing effects of treatments on disease (9, 25). Various statistics tools are available to 125 describe spatial patterns. Approaches based on discrete probability distributions are useful for determining spatial patterns of disease incidence at or below the sampling-unit scale (11, 34). Spatial autocorrelation approaches and SADIE can be used to determine spatial patterns at multiple spatial scales, starting at the scale of the sampling unit (21, 24,

34).

The objective of the study was to determine spatial patterns of Phomopsis cane and leaf spot of grape within commercial vineyards. Our hypothesis was that long-term development of disease in this perennial crop would lead to aggregation of disease. A state-wide survey on occurrence of Phomopsis cane and leaf spot of grape was conducted in Ohio over 3 years to determine incidence of disease at multiple sites in each vineyard

(Chapter 5). These data were the basis for the statistical analysis.

MATERIALS AND METHODS

Distribution of grape canes and leaves infected by Phomopsis viticola was evaluated though a state-wide survey over 3 years, 2002, 2003, and 2004. Five major grape growing regions were selected within the state: 1) Lake Erie east (North-East), 2)

Lake Erie west (North-Central), 3) Central, 4) Ohio River Valley (South-West), and 5)

Ohio Heartland area (Central-East). The naming of regions is adapted from Ohio Wine

Production Association. Within each region, two-to-three commercial growers were selected randomly, and two-to-three vineyards were randomly selected within each grower’s farm. Locations of each vineyard were recorded with a GPS unit (Magellan

SporTreck Color, Thales Navigation Inc., Santa Clara, CA).

126 In each vineyard, an 8x8 grid of grape vines was selected systematically. When the vineyard was smaller than 8x8, the maximum possible numbers of rows was sampled.

Typically, there was ~3m between vines in rows, and ~2m between rows. Three shoots per vine were randomly selected, and first five leaves and internodes from the basal part of each shoot were visually assessed for Phomopsis symptoms. Thus, each sampling site

(vine) consisted of 15 leaves and internodes.

Analysis based on discrete distributions. The small-scale spatial pattern of smaller (< sampling unit size) was characterized by comparing the observed frequency distribution of diseased leaves and internodes to predictions from the binomial and beta- binomial distributions (6, 11). The binomial distribution has a single parameter, p, which represents the (constant) probability of a leaf or internode being diseased. The beta- binomial distribution has two parameters, p and θ. Parameter p represents the expected probability of a leaf or internode being diseased and θ is a measure of variation, or heterogeneity, of disease between sampling units. If θ = 0, variation is equal to that obtained with a binomial distribution, and thus, θ = 0 is interpreted as a random pattern.

As θ increases above 0, there is increasing variation among sampling units, and corresponding increasing similarity of disease status of observations (leaves of internodes) within sampling units. Thus, θ is considered as an index of aggregation. A good fit of the binomial distribution is suggestive of a random distribution of diseased leaves and internodes, and a good fit of the beta-binomial distribution is suggestive of heterogeneity of diseased plant tissues (6, 11), and hence aggregation of disease.

127 The computer program BBD (11) was used to fit the binomial and beta-binomial distributions to observed counts of diseased observations per sampling unit (leaves or internodes). By using BBD, estimates of disease incidence ( pˆ ) for a single vine and the heterogenity parameter (θˆ ) were obtained using the maximum likelihood method (11).

In addition, chi-square goodness-of-fit tests were calculated for each distribution.

Classes of disease incidence were pooled so that there expected frequency was at least 5.

A chi-square goodness-of-fit test cannot be performed when the number of incidence classes is equal to or is less than the degrees of freedom for the test [the latter being given as (number of disease classes after pooling) – (number of parameters) – 1]. A log- likelihood test statistics (LRS) was also calculated to test whether the beta-binomial distribution describes the observation better than did the binomial distribution.

The likelihood-ratio statistics is calculated as follows:

LRS = 2[ll(x; pˆ,θˆ) − ll(x; pˆ)] in which ll(x; pˆ,θˆ ) and ll(x; pˆ ) are the log-likelihoods for the beta-binomial and binomial distributions, respectively, and x is a vector of the observed number of diseased individuals for the sampling units.. The LRS has a chi-square distribution with 1 df under the null hypothesis of no difference between the two log-likelihoods; in other words, a significant result indicates that the beta-binomial distribution provides a better fit than the binomial (34).

128

The BBD program was also used to calculate the index of dispersion, D, to describe the degree of heterogeneity at the sampling unit or lower scale, but without the assumption of a beta-binomial distribution (11). D = 1 indicates randomness, and D > 1 indicates heterogeneity (aggregation).

Binary power law analyses. Hughes and Madden (6, 7) developed the binary power law to represent the relationship between the observed sample variance of diseased individuals [leaves or internodes (Vobs)] and the theoretical variance for the binomial

distribution [Vbin = np(1− p ) ]. The equation, transformed with logarithms, for count data is,

ln(Vobs ) = ln(Ax ) + b ⋅ ln(npˆ(1− pˆ)) (1)

where ln(Ax) and b are intercept and slope of a straight line, respectively, pˆ is the estimate of the mean leaf or internode disease incidence, and n is the number of leaves or internodes assessed per sampling unit (n=15 in this study). Ax and b describe the spatial distribution of the observed disease incidence. When Ax = 1 [or ln(Ax) = 0] and b = 1, there is random pattern (as represented by a binomial distribution) over all datasets.

When Ax >1 and b = 1, diseased individuals are aggregated (at the scale of the sampling unit or smaller), but the degree of aggregation does not depend on p (mean disease incidence). When both Ax and b are greater than 1, the diseased individuals are

129 aggregated and the degree of aggregation changes with changes in p . When A x= 1 and b

= 1, θ = 0; when Ax >1 and b = 1, θ > 0 (but not affected by pˆ ), and when Ax >1 and b >

1, θ varies as a function of p .

If the data can be described by the beta-binomial distribution, theθ parameter of the beta-binomial distribution can be used to describe the parameters of the binary power law.

θ = [a − f ( p) / n]/[ f ( p) − a] (2)

b−2 (1−b) where a is Ax n and f ( p) = p(1− p) for disease incidence datasets with counts (6,

34).

SADIE. To determine larger scale (sampling unit and above) patterns across the vineyards, data were analyzed using SADIE (Spatial Analysis by Distance IndicEs) methodology (20, 21) and with spatial autocorrelation (described below). SADIE uses the coordinates of sampling units and the counts of diseased individuals per unit to evaluate the spatial arrangement of diseased individuals, conditional on the heterogeneity of disease between sampling units.

The key concept in SADIE is the idea of a "distance to regularity" (Da) (20, 21).

If one pictures an observed field as a grid that consists of m x n quadrats or sampling units (e.g., 8 x 8 lattice of grape vines), the algorithm moves observed diseased individuals incrementally in each quadrat to other quadrats in order that all quadrats

130 ultimately have the same number of diseased individuals. The total distance moved by all individuals to achieve the uniformity or 'regularity' of diseased individuals throughout the field is given as Da. The moves are determined with a transportation algorithm (21).

In order to determine if Da is large or small, the values of the counts in each quadrat are randomly reassigned to the locations, and the distance to regularity is recalculated. This is repeated numerous times (the number of randomizations can be defined by a user). From the frequency distribution of the distance values of the randomized cases, percentiles can be calculated (e.g., 2.5%, 5%, etc.). Aggregated arrangements have large distances to regularity, while regular arrangements tend to have smaller distances to regularity. One can perform a test of spatial randomness at a specified probability level (e.g., P = 0.05) by determining the proportion of randomizations with distances larger (smaller) than observed Da. The average distance to regularity from the randomization is denoted as Ea (i.e., Ea is the average Da from the randomizations).

If the distance to regularity of the observed sample (Da) is not significantly different from the randomized cases (Ea), there would be no evidence against the null hypothesis of a random spatial pattern. If the distance to regularity of the observed sample (Da) is significantly larger than the distances to regularity of the randomized cases, then the null hypothesis is rejected in favor of the alternative hypothesis of an aggregated,

(i.e., clumped or clustered) arrangement of disease incidence. Alternatively, if the distance to regularity is sufficiently small relative to the randomizations, then the null hypothesis is rejected in favor of the alternative hypothesis that indicates a regular or uniform pattern of observed disease incidence.

131 An index of aggregation (Ia) based on the SADIE methodology can be written as:

Ia = Da/Ea. Ia =1 suggests a spatially random pattern; Ia > 1, a more aggregated pattern; and Ia < 1, a more regular pattern. However, Ia is not directly used for determining significance; only the percentages of distances from the randomizations are used. For each year, vineyard, and disease assessment (leaves and internodes), SADIE analysis was performed with 78,000 randomizations. There are other indices calculated by SADIE, but Xu and Madden (37, 38) showed that results were very similar for each index.

Spatial covariance. The model to estimate spatial autocorrelation or covariance for a given dataset (vineyard, year, and disease measurement) is:

xi = µ + ei (3)

th where xi is disease incidence in the i sampling unit (i=1, 2, …, 64), µ is a constant

(mean), and ei is the corresponding error (residual). The variance of x is given as

2 Var(xi)=σ , and covariance of x at two locations, i and j (where j=1, 2, …, 64), is given by

Cov(xi, xj)=σij. It is assumed that σij can be defined based on distance between sampling

2 units, dij, which can be written as Cov(xi, xj)=σij=σ f(dij), where f(•) is a distance function that equals 1 at dij=0 and goes to 0 at large dij. Using the mixed modeling approach described by Little et al. (9) and Schabenbergar and Pierce (25), isotropic spatial covariance models were fitted to the observed data from each vineyard using maximum likelihood.

The correlation of disease incidence between neighboring sampling units is

2 defined as Corr(xi, xj)=rij=Cov(xi, xj)/Var(xi)=σij/σ =f(dij). As is typical for spatial analysis, it is assumed that the covariance or correlation depends only on the distance 132 between locations i and j. Thus, σij and rij are written as σ|i-j| and r|i-j|, respectively, where the single subscript indicates the Euclidian distance between i and j. The variance, covariances and correlations were determined with PROC VARIOGRAM of SAS (SAS

Insititute, Cary NC).

The semivariogram [γ(dij)] can be calculated from the variances and covariances

2 2 as γ(dij)=σ -σij=σ [1-f(dij)]. In general, observed γ(dij) goes to 0 as dij goes to 0, and

2 approaches σ as dij becomes large. However, if there is an apparent discontinuity in γ(dij)

2 at small distance, a so-called nugget variance σ 0 can be defined, so that

2 2 2 2 Var(xi ) = σ + σ 0 , and as before, γ(dij)=σ -σij=σ [1-f(dij)], and

2 2 2 2 givingγ (d ij ) = σ [1− f (dij )] + σ 0 . The variance,σ + σ 0 , is known as the sill, and

2 2 when σ 0 > 0 , then σ is known as the partial sill.

There are three terms that describe properties of semivariograms: the nugget, sill,

2 and range. The presence of the nugget effect (σ 0 > 0 ) is representative of very small scale patterns, and the sill represents variation of independent observations (far enough apart so that there is no influence). The lag-distance where the semivariance reaches the sill (or is within a certain small fraction of it) is called the range. The range indicates the hypothetical maximum distance of the spatial dependency of disease incidence. Different models were fitted to describe the relationship between semivariance (or covariance) and distance between samples. The spatial models examined were: exponential, spherical, power, and Gaussian, as described in Little et al. (9) and Schabenbergar and Pierce (25).

These models were fitted with and without the nugget term. The AIC (Akaike's

Information Criterion) statistics, which shows the degree of fit of the model to the 133 observed data based on log-likelihood, was used to compare model fits. Also, a model that assumes no spatial covariance (zero- or independent-covariance; i.e., σij=0) was fitted to the observed data, and compared against the results for the spatial covariance models using likelihood ratio statistics (LRS, with 1 df) test and AIC.

RESULTS

Disease incidence. In 2002, 20 different vineyards, from 11 different growers, were evaluated for leaf and internode disease incidence. Due to difficulty of the assessment (other symptoms on leaves), leaf disease incidence data from one of the vineyards were omitted from the dataset; thus, 19 and 20 vineyards for leaf and internodes were evaluated respectively. The mean vineyard disease incidence

(represented by pˆ , the estimated parameter of the beta-binomial distribution) varied from

0.05 to 0.87, and the median disease incidence for leaves and internodes for all vineyards was 0.38 and 0.57, respectively (Table 6.1).

In 2003, the same number of vineyards was assessed, and the disease incidence per vineyard varied from 0.03 to 0.76 (Table 6.1). Median disease incidence for all the vineyards was 0.42 and 0.41 for leaves and internodes, respectively. In 2004, 16 and 17 vineyards were sampled, for leaves and internodes disease incidence, respectively, corresponding to 10 different growers. The number of vineyards and growers decreased due to a renovation of vineyards by one grower. Disease incidence per vineyard varied from 0.17 to 0.91, and median disease incidence for all vineyards for leaves was 0.59 and for internodes was 0.61 (Table 6.1). 134 Comparison of pˆ (estimated mean value) showed that there were some vineyards where leaf disease incidence was much less than internode disease incidence (Figure 6.1).

Across 3 years, the median difference of pˆ for leaves and internodes ( pˆ L − pˆ I ) was -0.03.

This difference may be due to the time of assessment. That is, in experimental plots, it was observed that the oldest leaves, which have the highest disease severity, defoliate first (Nita, unpublished)

Discrete distributional analyses. Maximum likelihood estimation of the beta- binomial parameters, which is an iterative process, was successful in 50 out of 54 cases

(93%) of leaf disease incidence, and in 56 out of 57 (98%) cases of internode disease incidence over the 3 years. For all five cases where maximum likelihood estimation was not possible, the estimated θ value was at or very close to 0. The moment estimate of p was used for these five cases. The median θˆ value was larger than 0 in all years and assessed tissues (Table 6.1, Figure 6.1). Internode disease incidence tended to have higher θˆ than that of leaf disease incidence (Figure 6.1) with a median difference

ˆ ˆ (θ L −θ I ) of -0.06. The estimate of θ was significantly greater than 0 for 63% (34 out of

54) of the datasets for leaf disease incidence, and 90% (51 out of 57) of the datasets for internode disease incidence (Table 6.2), based on t-tests (7).

A chi-square goodness-of-fit test of the beta-binomial distribution to observed disease incidence indicated that one could not reject the null hypothesis of a good fit in

85% and 88% of the cases for leaf and internode disease incidence, respectively (P>0.05)

(Table 6.2). Moreover, a chi-square goodness-of-fit test of the binomial distribution

135 revealed that in 56% (leaf) and 30% (internode) of the cases, one could not reject the null hypothesis of a good fit (P>0.05) (Table 6.2). In other words, there were many more cases where the observed disease incidence was described by the beta-binomial distribution than by the binomial distribution. Likewise, the likelihood ratio statistic

(LRS) showed that the observed data were better fitted by the beta-binomial distribution than the binomial distribution (P≤0.05) in 67% and 91% of the cases for leaf and internode disease incidence, respectively (Table 6.2).

The index of dispersion, D, ranged from 0.93 to 6.87, and a median D across all vineyards and years was 2.17 and 2.85 for leaf and internode disease incidence, respectively (Table 6.1, Figure 6.1). For leaf disease incidence, 78% of D values were significantly larger than 1 (P≤0.05), indicative of aggregation of the disease incidence at a small spatial scale (Table 6.2). For internode disease incidence, D was significantly larger than 1 (P≤0.05) in 98% of the cases (Table 6.2), indicating more of a tendency of higher aggregation with internode disease incidence than leaf disease incidence (Figure

6.1).

When compared for each vineyard and year, the index of dispersion (D) for leaf disease incidence tended to be less than that for internode incidence (Figure 6.1). The median difference of D values for leaves and internode across 3 years (DL-DI) was -0.71.

Binary power law analyses. The binary power law provided a good fit to the data for both leaf and internode when all years were combined, (Table 6.3, Figure 6.2), but in some individual years and disease incidence measurements, the coefficient of determination, R2, was very low and the estimated slope, b, was close to or less than 1 136 (Table 6.3). Intercepts, ln(Ax), were significantly larger than 0 (i.e., Ax > 1) for leaf disease incidence in 2003, and in 2002 and 2003 for internode disease incidence. The estimated slope, b, was significantly larger than 1 in 2002 for leaf disease incidence, and in 2003 for internode disease incidence (Table 6.3). The difficulty in applying the binary power law was due to the fairly narrow range of pˆ values in a single year, with medians near 0.5 (Table 6.1); thus, a high percentage of points were at the far right of the log-log graph (Figure 6.2) where ln[np(1-p)] is at its maximum, making estimates of b unreliable.

The estimated slope parameter for pooled data was significantly greater than 1

(P≤0.05) (Table 6.3, Figure 6.2) for both leaf and internode disease incidence; moreover, the intercept was significantly different from 0 for internode disease incidence. However, when b > 1, simple interpretation of Ax is difficult. Most of variances in points were above the binomial line for both leaf and internode datasets (Figure 6.3) except at very low incidence ( pˆ < 0.2 ). These results indicated that less overdispersion, or spatial heterogeneity, with lower extremes of disease incidence.

The estimated parameters from the binary power law (pooled over years) were used to determine θˆ from pˆ for leaf and internode disease incidence using equation 2

(Figure 6.3). For both leaf and internode disease incidence, the curve generally followed the trend of θˆ : pˆ relationship observed in the datasets. As found by others (6, 7, 34), there was considerable variation of θˆ around pˆ ≈ 0.5 . The θˆ tended to increase as pˆ increased, and at pˆ ≈ 0.5 , θˆ started to decrease as pˆ increased. The relatively low R2 from the binary power-law regression was also responsible for some of the variation inθˆ .

137 SADIE. SAIDE's index of aggregation, Ia, was calculated successfully for all cases in an effort to characterize spatial patterns of disease at scales larger than the sampling unit. The Ia values for the vineyards varied from 0.76 to 2.14, and a median Ia over 3 years and all vineyards was 1.06 for leaf disease incidence (Table 6.1, Figure 6.1).

For internode disease incidence, vineyard Ia ranged from 0.78 to 1.58, and the median over 3 years and all vineyards was 1.11. The median difference of Ia values for leaves and internode across 3 years (IaL-IaI) was 0.04 (Figure 6.1).

The vineyards with significant aggregation (corresponding, in general, to Ia > 1)

(P≤0.05) complied only 11% and 16% of the total datasets for leaf and internode disease incidence, respectively (Table 6.4). Results indicated that only a few vineyards showed evidence of larger scale (i.e., between vine scale) spatial aggregation of the disease incidence (example is in the right-hand column of Figure 6.4).

Spatial covariance. In only a few cases did a spatial covariance model fit the data better than a model without spatial covariance. For example, the exponential model fitted the datasets better than the “independence” , or zero-covariance, model only in 13% and 9% of cases overall, for leaf and internode, respectively (Table 6.4), as indicated by the LRS test result (P≤0.05) and lower AIC by the spatial covariance model. Results for the spherical model are also shown in Table 6.4, and results were similar to those found for the exponential model. None of spatial covariance models tested (exponential, power, spherical, and Gaussian) had noticeably lower AIC values compared to the others.

138 The first-order spatial autocorrelation (lag distance of 1; r1) was determined for each dataset based on the semivariogram. Values ranged from 0.02 to 0.14 (Table 6.1).

Overall median r1 was 0.04 for leaves and 0.06 for internodes (Table 6.1)

Spatial covariance models with a nugget variance term were also fitted to the datasets. The results showed that in only 2% of the cases for leaf disease incidence, did the exponential model with nugget term fit better than the no-nugget model (Table 6.4).

For internode disease incidence, the model with nugget term never fitted better than a model without a nugget term (Table 6.4). Thus, variation close to origin (adjacent sampling unit) was not different from 0. Majority of the vineyards were represented by a relatively flat semivariogram or one in which there was no systemic trend described by the model that tested (example is in the left-hand column of Figure 6.4).

Overall, a model without spatial covariance had similar log-likelihood and AIC values compared with a covariance model, indicating incorporation of the spatial covariance did not improve fit of the model to most datasets. Thus, in general, there was only weak evidence of larger scale spatial structures of disease incidence in a few datasets. When both leaf and internode results were pooled, eight out of 11 cases that were identified with significant improvement in log-likelihood (P≤0.05) with the exponential covariance model compared with spatial “independence” model were also identified by SADIE as being aggregated. However, eight out of 16 cases with significant aggregation based on SADIE results did not have evidence of aggregation based on spatial covariance results.

139

DISCUSSION

Spatial distribution of Phomopsis cane and leaf spot of grape at different spatial scales (within vines and vine-to-vine) was examined using datasets obtained from a state- wide survey of commercial vineyards in Ohio from 2002-2004. Discrete distributional analyses was used to define the statistical distribution of disease incidence in each sampled dataset in order to characterize the underlying disease incidence pattern in the vineyards at a scale where spatial referencing was not possible (11). Results from fitting the beta-binomial distribution, and calculating the index of aggregation (D) (Tables 6.1 and 6.2) indicated that there was evidence of aggregation of the disease incidence at smaller (i.e., within a vine or grape canopy) spatial scales in many of the vineyards over 3 years of survey. There also was a general tendency of higher aggregation for diseased internode than diseased leaves, but this could be due to, in part, defoliation of the older leaves on shoots prior to disease assessment.

The binary form of the power law (6, 7) summarized the overall small-scale pattern of incidence of Phomopsis cane and leaf spot of grape. With the pooled data, the estimate of b of the binary power law was significantly greater than 1 (Table 6.3, Figure

6.2), further indication of small-scale aggregation of the disease incidence over all datasets (30, 31), but with degree of aggregation (characterized by θˆ of the beta-binomial distribution) changing with incidence (Figure 6.3). There was considerable variability in

θˆ of the beta-binomial and pˆ , but the general trend in relation to disease incidence was 140 that expected from the binary power law. Given that spore dispersal is by rain splash

(19), one would expect small-scale aggregation of diseased individuals (10, 16), and results are consistent with the hypothesis of this study.

Although there was strong evidence of small-scale aggregation of moderate magnitude, there was little evidence of larger-scale aggregation for most vineyards. A linear random effects model with spatial covariance structure fitted to the data, and analysis with the recently developed SADIE methodology (3, 21, 26, 34, 38), revealed that the majority of datasets exhibited random patterns. Only 14 % of datasets showed significant aggregation with SADIE, and 66% of Ia values from the SADIE analysis were less than 1.5. Of the datasets with significant aggregation, Ia ranged from 1.15 to 2.14, which is indicative of moderate aggregation (Figure 6.1). Moreover, for the spatial autocorrelation analysis, only 15% of vineyards had any evidence of aggregation (Table

6.4).

Advantage of using SADIE is that it is specifically designed for situations where discrete individuals (such as diseased leaves) exist in patches with relatively clear boundaries, instead of where there are smooth changes in disease intensity on a continuous scale (21, 35). Thus, this methodology may be more useful than spatial autocorrelation analysis (or random effects covariance modeling) for characterizing counts of diseased individuals. Because eight more datasets were identified as being aggregated based on SADIE compared with spatial covariance analysis, SADIE may have greater statistical powering in detecting patterns of disease. However, Xu and

Madden (36) show that SADIE assesses aggregation in terms of absolute as well as relative pattern of patches, whereas only the relative positions of patches are relevant

141 with spatial covariance. That is, patches of disease near a field edge will have a larger Ia than when the same patches are in the field center. Thus, the SADIE-based analyses in this study were likely detecting some situations where there was clustering of disease near an edge or corner of a vineyard.

The spatial-covariance-model analysis detected about half of the cases in which aggregation of disease incidence was identified by SADIE, but in three cases, the former analysis identified aggregation in vineyards that were not identified with SADIE methods

(Figure 6.4). One of advantages of the use of the spatial covariance models is that they can be extended to analyses of fields where treatments are applied based on an experimental design (25; Chapter 4). If there is spatial autocorrelation of disease incidence observed, researchers can analyze effects of treatments on means with higher accuracy and precision (9, 25).

Overall, this study indicated that the spatial distribution of Phomopsis cane and leaf spot of grape showed a tendency of aggregation at the vine-level, with only a few situations with larger scale clustering. The disease cycle is considered monocyclic; conidia of P. viticola are disseminated by rain splash; the source of inoculum, infected old canes, were typically located near the main trunk; and distance between vines are fairly far (2-3m between vines and rows); thus, long distance dissemination of the conidia may not happen frequently (1, 10, 16). In contrast to our results, Munkvold (17) found larger scale patterns of Eutypa dieback in California vineyards using semivariograms (8), and greater large-scale aggregation of diseased vines was found in commercial vineyards with perithecia than those without.

142 Using the same data, Hughes et al. (8), showed that there were also small-scale patterns of this disease based on a discrete-distribution analysis, similar to the beta- binomial approach used here. Interestingly, greater small-scale aggregation was found in

California vineyards without perithecia than those with perithecia (opposite of the larger- scale results). In a study with grape downy mildew in Ohio over 3 years, Madden et al.

(13) also found small-scale patterns of incidence of disease leaves with θˆ being related to pˆ , and bˆ of the binary power law greater than 1. The estimated θˆ values in the current study were similar to these reported for grape downy mildew (13) and to those calculated from the Eutypa datasets (17). In the latter case, θˆ values were estimated from the reported intra-cluster correlation ( ρˆ ), using θˆ = ρˆ /(1− ρˆ ) (14).

Diseases of other crops with splash dispersal of pathogen inoculum that cause disease on other hosts, such as leaf blight of strawberry (caused by Phomopsis obscurans)

(34) and leaf spot of strawberry (caused by Sphaerotheca macularis) (33) showed similar spatial patterns in commercial fields to that found here, that is, more small-scale aggregation and less large scale aggregation. Both of these diseases have a polycyclic disease cycle in contrast to the monocyclic cycle of Phomopsis on grape. However, with

Stagonospora nodorum on winter wheat in New York (28), which is caused by splash- dispersed pathogen with a polycyclic disease cycle, larger scale spatial scale aggregation could be observed with favorable weather conditions, as well as small scale patterns.

Phomopsis cane and leaf spot of grape is known to be a part of the “grape dead- arm” disease complex, which has been common for a long time in U.S. grape production

(15, 19). Given that many of vineyards examined in this study were relatively old (> 15

143 yrs or older), it is possible that the disease was present for a long time. With sufficient time for disease spread throughout the vineyards to occur, there would be no large patches of disease and gaps with no disease, which is consistent with the results for

SADIE and spatial covariance analysis. Because of the stochastic and short range nature of splash dispersal (10, 16), small-scale patterns (within vines) would remain, for leaves and internodes, even if all vines were infected, as also found here. Thus, even though all growers were applying fungicides for disease control, a fairly high percentage of leaves and internodes were infected in most vineyards. However, some of the old vineyards had relatively low disease incidence, which was probably achieved by proper disease management strategies or other factor (Chapter 5).

144

Leaves: BBDa SADIE ˆ ˆ Statistics DI p θ D Ia r1 2002 Minimum 0.69 0.05 0.00 0.88 0.76 0.02

Q1 2.73 0.18 0.01 1.14 0.95 0.03 Median 5.69 0.38 0.07 1.92 1.10 0.04

Q3 7.90 0.53 0.11 2.45 1.21 0.01 Maximum 9.81 0.66 0.16 2.90 1.57 0.09 2003 Minimum 1.42 0.10 0.00 1.07 0.71 0.03

Q1 3.73 0.25 0.05 1.80 0.97 0.04 Median 6.26 0.42 0.08 2.16 1.08 0.05

Q3 8.19 0.55 0.16 2.90 1.19 0.02 Maximum 10.30 0.69 0.28 4.23 1.71 0.08 2004 Minimum 3.11 0.21 0.00 0.65 0.85 0.02

Q1 5.00 0.34 0.04 1.51 0.97 0.03 Median 8.82 0.59 0.07 1.92 1.13 0.04

Q3 10.81 0.73 0.11 2.54 1.28 0.01 Maximum 13.40 0.89 0.68 6.87 2.14 0.14

Internodes: BBDa SADIE ˆ ˆ Statistics DI p θ D Ia r1 2002 Minimum 1.89 0.13 0.03 1.44 0.83 0.03

Q1 5.96 0.40 0.07 1.95 0.93 0.04 Median 8.48 0.57 0.14 2.70 1.05 0.06

Q3 10.27 0.69 0.19 3.35 1.19 0.02 Maximum 13.00 0.87 0.40 5.19 1.37 0.10 2003 Minimum 0.47 0.03 0.03 1.60 0.85 0.04

Q1 4.49 0.30 0.11 2.45 1.03 0.06 Median 6.14 0.41 0.15 2.84 1.11 0.07

Q3 7.29 0.49 0.20 3.51 1.33 0.01 Maximum 11.27 0.76 0.53 6.06 1.58 0.13 2004 Minimum 2.53 0.17 0.00 0.91 0.77 0.03

Q1 6.12 0.41 0.10 2.33 0.94 0.04 Median 9.18 0.61 0.11 2.44 1.00 0.05

Q3 11.03 0.74 0.17 3.08 1.32 0.01 Maximum 13.64 0.91 0.46 5.98 1.74 0.12 a The beta-binomial distribution results ( pˆ and θˆ )

Table 6.1. Summary statistics [minima, maxima, and three quartiles (25% percentile (Q1), median, and 75% percentile (Q3))] for number of diseased individuals (leaves or internodes; DI), estimated beta-binomial parameters p) (expected probability of diseased leaves or internodes) and θˆ (heterogeneity of disease incidence), index of dispersion (D), SADIE’s index of aggregation (Ia), and first-order autocorrelation (r1) for Phomopsis cane and leaf spot of grape in commercial vineyards from 2002 to 2004.

145

Year pˆ a BBDb BINb LRSc Dd Leaves: 2002 0.38 78.9(19) 73.7(19) 52.6(19) 68.4(19) 2003 0.42 94.7(19) 47.4(19) 78.9(19) 84.2(19) 2004 0.59 81.3(16) 50.0(16) 68.8(16) 81.3(16) Overall 0.46 85.2(54) 57.4(54) 66.7(54) 77.8(54)

Internodes: 2002 0.52 85.0(20) 25.0(20) 90.0(20) 100(20) 2003 0.41 90.0(20) 20.0(20) 95.0(20) 100(20) 2004 0.58 88.2(17) 47.1(17) 88.2(17) 94.1(17) Overall 0.50 87.7(57) 29.8(57) 91.2(57) 98.2(57) a Median of the expected proportion of diseased leaves or internodes (see Table 6.1) b Percent of cases where the null hypothesis was accepted (P>0.05) based on a chi-square test. c Likelihood ratio statistic. Percent of cases where the log-likelihood for the beta-binomial was significantly larger than for the binomial based on a chi-square test (P≤0.05) d Percent of vineyards where D>1 based on chi-square test (P≤0.05)

Table 6.2. Percent of vineyards where: the beta-binomial (BBD) distribution provided a good fit to number of diseased leaves and internodes; where the binomial (BIN) distribution provided a good fit to number of diseased leaves and internodes; where the beta-binomial provided a significantly better fit than the binomial based on the likelihood ratio statistic (LRS); and where the index of dispersion (D) was significantly larger than 1.

146

2a ˆ b ˆ c ˆ d ˆ e Year R ln(Ax ) SE[ln(Ax )] b SE(b) Leaves: 2002 0.81 0.01 0.20 1.58* 0.18 2003 0.28 0.87* 0.40 0.91 0.34 2004 0.14 0.89 0.59 0.81 0.52 pooled 0.55 0.28 0.19 1.36* 0.17

Internodes: 2002 0.38 0.81* 0.37 1.12 0.33 2003 0.83 0.71* 0.17 1.37* 0.14 2004 0.59 0.45 0.33 1.42 0.30 pooled 0.63 0.62* 0.15 1.33* 0.13 a Coefficient of determination b Estimated intercept of the best fitting line based on least squares regression. Number with “*”

indicates the estimated intercept is significantly larger than 0 (P≤0.05) (i.e., Ax > 1) c Standard error of the estimated intercept d Estimated slope of the best fitting line based on least squares regression. Number with “*” indicates the estimated intercept is significantly larger than 1 (P≤0.05) e Standard error of the estimated slope

Table 6.3. Estimated parameters of the binary power law (equation 2) and their standard errors for Phomopsis cane and leaf spot of grape on leaves and internodes in commercial vineyards on Ohio from 2002 to 2004

147

Percentage a b c d e Year pˆ SADIE LRSexp-ind LRSsph-ind LRSnugget Leaves: 2002 0.38 5.3(19) 10.5(19) 10.5(19) 5.3(19) 2003 0.42 10.5(19) 10.5(19) 10.5(19) 0(19) 2004 0.59 18.8(16) 18.8(16) 18.8(16) 0(16) Overall 0.46 11.1(54) 13.0(54) 13.0(54) 1.9(54)

Internodes: 2002 0.52 10.0(20) 0.0(20) 5.0(20) 0(20) 2003 0.41 20.0(20) 15.0(20) 15.0(20) 0(20) 2004 0.58 20.0(20) 11.8(17) 6.0(17) 0(17) Overall 0.50 17.5(57) 8.8(57) 8.8(57) 0(57) a Median estimate of the expected proportion of disease (see Table 6.1) b Percentage of vineyards with aggregation. Based on a randomization tests for each dataset (21) c Percentage of vineyards with significant results based on the difference of log-likelihoods for exponential spatial covariance model and independence model d Percentage of vineyards with significant results based on the difference of log-likelihoods for spherical spatial covariance model and independence model e Percentage of vineyards with significant results based on the difference of log-likelihoods for exponential model with and without nugget effect

Table 6.4. Percentage of vineyards where SADIE indicated aggregation, where a likelihood ratio statistics (LRS) indicated that a random effects model with spatial structure fitted better than a model with independence of data, and where a LRS indicated that a nugget effect model fitted better than a no-nugget model

148

Leaf Internode Difference 12 12 18 16 10 10 AB14 C 8 8 12 10 6 6 8 4 4 6

Frequency 4 2 2 2 0 0 0 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 -0.2 0.0 0.2 0.4 0.6 0.8 1.0 1.2 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 ) ) pL pI pˆ L − pˆ I 40 20 18 18 16 16 30 DE14 F 14 12 12 10 20 10 8 8 6 6

Frequency 10 4 4 2 2 0 0 0 -0.2 0.0 0.2 0.4 0.6 0.8 -0.1 0.0 0.1 0.2 0.3 0.4 0.5 0.6 -0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3 ) ) ) ˆ θL θ I θL −θI

30 16 18 14 16 25 GH12 14 I 20 10 12 10 15 8 8 6 10 6

Frequency 4 4 5 2 2 0 0 0 0246801234567 -4 -3 -2 -1 0 1 2 D L DI DL − DI 30 14 25

25 12 JKL20 D 10 20 8 15 15 6 10 10 4 Frequency 5 5 2

0 0 0 0.40.60.81.01.21.41.61.82.02.22.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8 2.0 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2 0.4 0.6 0.8 1.0

Ia L Ia I Ia L − IaI

Figure 6.1. Frequency distribution of: A, B, the estimated beta-binomial parameter p) (estimated average probability of disease incidence) for leaf and internode disease incidence of Phomopsis cane and leaf spot of grape; D, E, the estimated beta-binomial ) aggregation parameter θ ; G, H, the index of dispersion, D; and J, K, SADIE’s index of ˆ aggregation, Ia. C, F, I, L, the frequency distributions of the difference in pˆ ,θ , D , and I a for leaves (L) and internodes (I).

149

log(V ) = 0.28 +1.36log(npˆ(1− pˆ)) 3 obs R2 = 0.55

2

1

0 log(observed variance) log(observed Leaf

3 log(Vobs ) = 0.62 +1.33log(npˆ(1− pˆ)) R2 = 0.63

2

1

0 log(observed variance) log(observed Internode

-1.0 -0.5 0.0 0.5 1.0 1.5

log(binomial variance)

Figure 6.2. Relationship between the logarithm of the observed variance of Phomopsis cane and leaf spot disease incidence among sampling units (V) and the logarithm of the variance assuming disease incidence follows a binomial distribution (eq. 1). In the equation, pˆ is the estimate of the mean disease incidence and n is the number of leaves or internodes assessed per sampling unit (n = 15). Each point represents a single year x vineyard combination. Solid line represents the least squares fit to the data, and dashed line represents the binomial [log(Vobs ) − log(npˆ(1− pˆ )) ] line.

150

0.6 Leaf

0.4 ˆ θ Col 3 vs Col 4 0.2 Plot 1 Regr

0.0

0.6 Internode

0.4 θˆ 0.2

0.0

0.01 0.1 1 pˆ

Figure 6.3. The relationship between estimated aggregation parameter of the beta- binomial distribution (θˆ ) and disease incidence ( pˆ , estimated expected disease probability), for Phomopsis cane and leaf spot of grape in commercial vineyards in Ohio over 3 years. Each point represents a single year x vineyard combination. The solid line represents a theoretical relationship between θˆ and pˆ based on the results from the binary power law (eq. 2) 151

Vineyard #7 Vineyard #17 leaf 2003 leaf 2004

14 14 ABθ=0.02 12 θ=0.06 12 I =2.14 10 Ia=1.05 10 a 8 8 6 6 Counts 4 4 2 2 0 0 01234567891011121314151617 01234567891011121314151617 Disease incidence Disease incidence

CD

0.14 0.070

0.12 EF0.065

0.10 0.060

0.08 0.055

0.06 0.050

Semivariance 0.04 0.045

0.02 0.040 0246810 0246810 Lag distance Lag distance

Figure 6.4. Examples of Phomopsis cane and leaf spot spatial pattern results for two commercial vineyards in Ohio A, B, Frequency distribution of diseased leaves (bars) and expected frequency for binomial (short-dash) and beta-binomial (dot) distributions, C, D, disease incidence contour maps (darker the color, higher the disease incidence) E, F, Semivariograms.

152 REFERENCES

1. Campbell, L. C., and Madden, L. V. 1990. Introduction to Plant Disease Epidemiology. John Wiley & Sons, NewYork. 2. Cucuzza, J. D., and Sall, M. A. 1982. Phomopsis cane and leaf spot disease of grape vine: Effect of chemical treatments on inoculum level, disease severity, and yield. Plant Disease 66:794-797. 3. Dallot, S., Gottwald, T., Labonne, G., and Quiot, J.-B. 2003. Spatial pattern analysis of Sharka disease (Plum pox virus strain M) in peach orchards of southern France. Phytopathology 93:1543-1552. 4. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2001. Effect of growth stage on susceptibility of grape berry and rachis tissues to infection by Phomopsis viticola. Plant Disease 85 5:517-520. 5. Erincik, O., Madden, L. V., Ferree, D. C., and Ellis, M. A. 2003. Temperature and wetness-duration requirements for grape leaf and cane infection by Phomopsis viticola. Plant Disease 87:832-840. 6. Hughes, G., and Madden, L. V. 1992. Aggregation and incidence of disease. Plant Pathology 41 6:657-660. 7. Hughes, G., and Madden, L. V. 1993. Using the beta-binomial distribution to describe aggregated patterns of disease incidence. Phytopathology 83 7:759-763. 8. Hughes, G., Munkvold, G. P., and Samita, S. 1998. Application of the logistic- normal-binomial distribution to the analysis of Eutypa dieback disease incidence. International Journal of Pest Management 44 1:35-42. 9. Littell, R. C., Milliken, G. A., Stroup, W. W., and Wolfinger, R. D. 1996. SAS system for mixed models. SAS Institute Inc., Cary, NC. 10. Madden, L. V. 1992. Rainfall and the dispersal of fungal spores. Advances in Plant Pathology 8:39-79. 11. Madden, L. V., and Hughes, G. 1994. BBD - computer software for fitting the beta-binomial distribution to disease incidence data. Plant Disease 78 5:536-540. 12. Madden, L. V., and Hughes, G. 1999. Sampling for plant disease incidence. Phytopathology 89 11:1088-1103. 13. Madden, L. V., Hughes, G., and Ellis, M. A. 1995. Spatial heterogeneity of the incidence of grape downy mildew. Phytopathology 85 3:269-275. 14. Madden, L. V., Hughes, G., and Munkvold, G. P. 1996. Plant disease incidence: inverse sampling, sequential sampling, and confidence intervals when observed mean incidence is zero. Crop Protection 15 7:621-632. 15. Malathrakis, N. E., and Baltzakis, N. G. 1976. Control of Dead-arm of Grapes. Poljoprivredna znanstvena smotra - Agriculture Conspectus Scientificus 39 49:261-269. 16. McCartney, H. A., and Fitt, B. D. L. 1998. Dispersal of foliar fungal plant pathogens: Mechanisms, gradients and spatial patterns. 138-160 in: The epidemiology of plant diseases Jones, D. G. Kluwer, Dordrecht, the Netherlands. 17. Munkvold, G. P., Duthie, J. A., and Marois, J. J. 1993. Spatial patterns of grapevines with Eutypa dieback in vineyards with or without perithecia. Phytopathology 93 12:1440-1448. 153 18. Orenstein, S., Zahavi, T., Nestel, D., Sharon, R., Barkalifa, M., and Weintraub, P. G. 2003. Spatial dispersion patterns of potential leafhopper and planthopper (Homoptera) vectors of phytoplasma in wine vineyards. Annals of Applied Biology 142 3:341-348. 19. Pearson, R. C., and Goheen, A. C. ed. 1988. Compendium of Grape Diseases. APS Press, St. Paul, MN. 20. Perry, J. N. 1998. Measures of spatial pattern for counts. Ecology 79 3:1008-1017. 21. Perry, J. N., Bell, E. D., Smith, R. H., and Woiwod, I. P. 1996. SADIE: software to measure and model spatial pattern. Aspects of Applied Biology 46:95-102. 22. Pscheidt, J. W., and Pearson, R. C. 1989. Effect of grapevine training systems and pruning practices on occurrence of Phomopsis cane and leaf spot. Plant Disease 73 10:825-828. 23. Pscheidt, J. W., and Pearson, R. C. 1989. Time of infection and control of Phomopsis fruit rot of grape. Plant Disease 73 10:829-833. 24. Reynolds, K. M., and Madden, L. V. 1988. Analysis of epidemics using spatio- temporal autocorrelation. Phytopathology 78:240-246. 25. Schabenberger, O., and Pierce, F. J. 2001. Contemporary statistical models for the plant and soil sciences. CRC Press, Boca Raton, FL. 26. Schmale, D. G., III, , Shah, D. A., and Bergstrom, G. C. 2005. Spatial patterns of viable spore deposition of Gibberella zeae in wheat fields. Phytopathology 95:472-479. 27. Sergeeva, V., Nair, N. G., Barchia, I., Priest, M., and Spooner-Hart, R. 2003. Germination of β conidia of Phomopsis viticola. Austrarian Plant Pathology 32:105-107. 28. Shah, D., Bergstrom, G. C., and Ueng, P. O. 2001. Foci of Stagonospora nodorum blotch in winter wheat before canopy development. Phytopathology 91:642-647. 29. Shah, D. A., Pucci, N., and Infantino, A. 2005. Regional and varietal differences in the risk of wheat seed infection by fungal species associated with fusarium heaf blight in Italy. European Journal of Plant Pathology 112:13-21. 30. Taylor, L. R. 1961. Aggregation, variance and the mean. Nature 189:732-735. 31. Taylor, L. R., Perry, J. N., WoiwoId, P., and J., T. R. A. 1988. Specificity of the spatial power-law exponent in ecology and agriculture. Nature 32:721-722. 32. Tubajika, K. M., Civerolo, E. L., Ciomperlik, M. A., Luvisi, D. A., and Hashim, J. M. 2004. Analysis of the spatial patterns of Pierce’s disease incidence in the lower San Joaquin Valley in California. Phytopathology 94:1136-1144. 33. Turechek, W. W., and Madden, L. V. 2000. Analysis of the association between the incidence of two spatially aggregated foliar diseases of strawberry. Phytopathology 90 2:157-170. 34. Turechek, W. W., and Madden, L. V. 1999. Spatial pattern analysis of strawberry leaf blight in perennial production systems. Phytopathology 89 5:421-433. 35. Winder, L., Alexander, C. J., Holland, J. M., Woolley, C., and Perry, J. N. 2001. Modelling the dynamic spatio-temporal response of predators to transient prey partches in the field. Ecology Letters 4:568-576. 36. Xu, X., and Madden, L. V. 2003. Considerations for the use of SADIE statistics to quantify spatial patterns. Ecography 26 6:281-291.

154 37. Xu, X. M., and Madden, L. V. 2005. Interrelationships among SADIE indices for characterizing spatial patterns of organisms. Phytopathology 95:874-883. 38. Xu, X. M., and Madden, L. V. 2004. Use of SADIE statistics to study spatial dynamics of plant disease epidemics. Plant Pathology 53 1:38-49.

155

BIBLIOGRAPHY

Agrios GN. 1997. Plant Pathology. San Diego: Academic Press

Alexopoulos CJ, Mims CW, Blackwell M. 1996. Introductory Mycology. New York: John Wiley & Sons, Inc

Allison PD. 1999. Logistic regression using the SAS system: Theory and application. Cary, NC: SAS Institute

Beresford RM, Manktelow DWL. 1994. Economics of reducing fungicide use by weather-based disease forecasts for control of Venturia inaequalis in apples. New Zealand Journal of Crop and Horticultural Science 22:113-20

Bergamin Filho A, Carneiro SMTPG, Godoy CV, Amorim L, Berger RD, Hau B. 1997. Angular leaf spot of Phaseolus beans: relationships between disease, healthy leaf area, and yield. Phytopathology- 87:506-15

Calpouzos L. 1966. Action of oil in the control of plant disease. Annual Review of Phytopathology 4:369-90

Campbell LC, Madden LV. 1990. Introduction to Plant Disease Epidemiology. NewYork.: John Wiley & Sons

Carter MV. 1960. Further study on Eutypa armeniacea Hansf. & Carter. Austrarian Journal of Agricultural Research 11:498-504

Correiar BR. 1999. Use of automated weather stations and the Gubler-Thomaz model for control of powdery mildew (Uncinula nacator) in California grapevines. Summa Phytopathologica 25:70-3

Cucuzza JD, Sall MA. 1982. Phomopsis cane and leaf spot disease of grape vine: Effect of chemical treatments on inoculum level, disease severity, and yield. Plant Disease 66:794-7

156

Dallot S, Gottwald T, Labonne G, Quiot J-B. 2003. Spatial pattern analysis of Sharka disease (Plum pox virus strain M) in peach orchards of southern France. Phytopathology 93:1543-52

De Wolf ED, Madden LV, Lipps PE. 2002. Risk assessment models for wheat Fusarium head blight epidemics based on within-season weather data. Phytopathology 93:428-35

Dell KJ, Gubler WD, Krueger R, Sanger M, Bettiga LJ. 1998. The efficacy of JMS Stylet-Oil on grape powdery mildew and Botrytis bunch rot and effects on fermentation. American Journal of Enology and Viticulture 49:11-6

Ellis MA, Converse RH, Wiliams RN, Williamson B, eds. 1991. Compendium of raspberry and blackberry diseases and insects. St. Paul, MN: APS Press. 100 pp.

Ellis MA, Madden LV, Wilson LL. 1984. Evaluation of an electronic apple scab predictor for scheduling fungicides with curative activity. Plant Disease 68:1055-7 Ellis MA, Madden LV, Wilson LL. 1986. Electronic grape black rot predictor for scheduling fungicides with curative activity. Plant Disease 70:938-40

Ellis MA, Nita M. 2004. Disease management guidelines for organic grape production in the Midwest. Wooster, OH: The Ohio State University OARDC/OSUE. 35 pp.

Ellis MA, Welty C, Funt RC, Doohan D, Wiliams RN, et al, eds. 2004. Midwest Small Fruit Pest Management Handbook. Columbus, OH: Ohio State University Extension. 234 pp.

Erincik O, Madden LV, Ferree DC, Ellis MA. 2001. Effect of growth stage on susceptibility of grape berry and rachis tissues to infection by Phomopsis viticola. Plant Disease 85:517-20

Erincik O, Madden LV, Ferree DC, Ellis MA. 2003. Temperature and wetness-duration requirements for grape leaf and cane infection by Phomopsis viticola. Plant Disease 87:832-40

Funt RC, Ellis MA, Madden LV. 1990. Economic analysis of protectant and disease- forecast-based fungicide spray programs for control of apple scab and grape black rot in Ohio. Plant Disease 74:638-42

Gadoury DM. 1993. Integrating management decisions for seveal pests in fruit production. Plant Disease 77:299-302

Gadoury DM, Pearson RC, Riegel DG, Seem RC, Becker CM, Pscheidt JW. 1994. Reduction of Powdery Mildew and Other Diseases by Over-the-Trellis Applications of Lime Sulfur to Dormant Grapevines. Plant Disease 78:83-7

157

Gent DH, Schwartz HF, Nissen SJ. 2003. Effect of commercial adjuvents on vegitable crop fungicide coverage, absorption, and efficacy. Plant Disease 87:591-7

Grayson BT, Price PJ, Walter D. 1997. Effects of adjuvants on the performance of a novel powdery mildew fungicide, 1-(4-chlorobenzyl)-4-phenylpiperidine. Pesticide Science 51:206-12

Grayson BT, Webb JD, Pack SE. 1993. Investigation of an emulsifiable oil adjuvant and its components on the activity of a new grass herbicide by factorial experimentation. Pesticide Science 37:127-31

Grunwald NJ, Romero Montes G, Lozoya Saldana H, Rubio Covarrubias OA, Fry WE. 2002. Potato late blight management in the Toluca valley: field validation of SimCast modified for cultivars with high field resistance. Plant Disease 86:1163-8

Gubler WD, Baldwin LG. 2000. Evaluation of fungicides for the control of Phomopsis cane and leaf spot of grapes, 1999. Fungicide and Nematicide Tests 55:97

Holloway PJ, Ellis MCB, Webb DA, Western NM, Tuck CR, et al. 2000. Effects of some agricultural tank-mix adjuvants on the deposition efficiency of aqueous sprays on foliage. Crop Protection 19:27-37

Hughes G, Madden LV. 1992. Aggregation and incidence of disease. Plant Pathology 41:657-60

Hughes G, Madden LV. 1993. Using the beta-binomial distribution to describe aggregated patterns of disease incidence. Phytopathology 83:759-63

Hughes G, Munkvold GP, Samita S. 1998. Application of the logistic-normal-binomial distribution to the analysis of Eutypa dieback disease incidence. International Journal of pest management 44:35-42

Jones AL, Aldwinckle HS, eds. 1990. Compendium of apple and pear diseases. St Paul, MN: APS press

Kudsk P, Mathiassen SK, Kirknel E. 1991. Influence of formulations and adjuvants on the rainfastness of maneb and mancozeb on pea and potato. Pesticide Science 33:57-71 Lal B, Arya A. 1982. A soft rot of grapes caused by Phomopsis viticola. Indian Phytopathology 35:261-4

Littell RC, Milliken GA, Stroup WW, Wolfinger RD. 1996. SAS system for mixed models. Cary, NC: SAS Institute Inc.

158 Lownds NK, Bukovac MJ. 1989. Surfactant-induced ethylene production by leaf tissue. Journal of the American Society for Horticultural Science 114:449-54

Madden LV. 1992. Rainfall and the dispersal of fungal spores. Advances in Plant Pathology 8:39-79

Madden LV, Ellis MA. 1988. How to develop plant disease forecasters. In Experimental techniques in plant disease epidemiology, ed. J Kranz, J Rotem, pp. 191-208. Berlin: Springer-Verlag

Madden LV, Ellis MA, Lalancette N, Hughes G, Wilson LL. 2000. Evaluation of a disease warning system for downy mildew of grapes. Plant Disease 84:549-54

Madden LV, Hughes G. 1994. BBD - computer software for fitting the beta-binomial distribution to disease incidence data. Plant Disease 78:536-40

Madden LV, Hughes G. 1999. Sampling for plant disease incidence. Phytopathology 89:1088-103

Madden LV, Hughes G, Ellis MA. 1995. Spatial heterogeneity of the incidence of grape downy mildew. Phytopathology 85:269-75

Madden LV, Hughes G, Munkvold GP. 1996. Plant disease incidence: inverse sampling, sequential sampling, and confidence intervals when observed mean incidence is zero. Crop Protection 15:621-32

Malathrakis NE, Baltzakis NG. 1976. Control of Dead-arm of Grapes. Poljoprivredna znanstvena smotra - Agriculture Conspectus Scientificus 39:261-9

McCartney HA, Fitt BDL. 1998. Dispersal of foliar fungal plant pathogens: Mechanisms, gradients and spatial patterns. In The epidemiology of plant diseases, ed. DG Jones, pp. 138-60. Dordrecht, the Netherlands: Kluwer

Mcdonald OC, McCartney HA. 1987. Calculation of splash droplet trajectories. Agricultural and Forest Meteorology 39:95-110

McGrath MT, Shishkoff N. 2000. Control of cucurbit powdery mildew with JMS Stylet- Oil. Plant Disease 84:989-93

Morrison LS, Russell CC. 1976. Timing of fungicide - adjuvant mixtures for control of rose blackspot. Plant Disease Reporter 60:634-6

Mostert L, Crous PW, Petrini O. 2000a. Endophytic fungi associated with shoots and leaves of Vitis vinifera, with specific reference to the Phomopsis viticola complex. Sydowia 52:46-58

159 Mostert L, Denman S, Crous PW. 2000b. In vitro screening of fungicides against Phomopsis viticola and Diaporthe perjuncta. South African Journal for Enology and Viticulture 21:62-5

Munkvold GP, Duthie JA, Marois JJ. 1993. Spatial patterns of grapevines with Eutypa dieback in vineyards with or without perithecia. Phytopathology 93:1440-8

Niekerk JMv, Groenewald JZ, Farr DF, Fourie PH, Halleen F, Crous PW. 2005. Reassessment of Phomopsis species on grapevines. Australasian Plant Pathology 34:27- 39

Nita M, Madden LV, Wilson LL, Ellis MA. 2003. Evaluation of a disease prediction system for Phomopsis cane and leaf spot of grape. Phytopathology 93:S65

Northover J, Schneider KE. 1996. Physical modes of action of petroleum and plant oils on powdery and downy mildews of grapevines. Plant Disease 80:544-50

OARDC. 2005. OARDC weather system. http://www.oardc.ohio- state.edu/centernet/weather.htm

Odum HT, Odum EC. 2000. Modelling for all Scales: an introduction to System Simulation. San Diego, CA: Academic Press

Orenstein S, Zahavi T, Nestel D, Sharon R, Barkalifa M, Weintraub PG. 2003. Spatial dispersion patterns of potential leafhopper and planthopper (Homoptera) vectors of phytoplasma in wine vineyards. Annals of Applied Biology 142:341-8

Paul PA, Munkvold GP. 2004. A model-based approach to preplanting risk assessment for gray leaf spot of maize. Phytopathology 94:1350-7

Pearson RC, Goheen AC, eds. 1988. Compendium of Grape Diseases. St. Paul, MN: APS Press. 94 pp.

Percich JA, Nickelson LJ. 1982. Evaluation of several fungicides and adjuvant materials for control of brown spot of wild rice. Plant Disease 66:1001-3

Perry JN. 1998. Measures of spatial pattern for counts. Ecology 79:1008-17

Perry JN, Bell ED, Smith RH, Woiwod IP. 1996. SADIE: software to measure and model spatial pattern. Aspects of Applied Biology 46:95-102

Pezet R, Pont V, Girardet F. 1983. Microcycle conidiation in pycnidial cirrhii of Phomopsis viticola Sacc. induced by elemental sulfur. Canadian Journal of Mycrobiology 29:179-84

160 Phillips AJL. 1999. The relationship between Diaporthe perjuncta and Phomopsis viticola on grapevines. Mycologia- 91:1001-7

Phillips AJL. 2000. Excoriose, cane blight and related diseases of grapevines: a taxonomic review of the pathogens. Phytopathol. Mediterr. 39:341-56

Pine TS. 1959. Development of the grape dead-arm disease. Phytopathology 49:738-43

Pscheidt JW, Pearson RC. 1989a. Effect of grapevine training systems and pruning practices on occurrence of Phomopsis cane and leaf spot. Plant Disease 73:825-8

Pscheidt JW, Pearson RC. 1989b. Time of infection and control of Phomopsis fruit rot of grape. Plant Disease 73:829-33

Raese JT, Drake SR. 1993. Effects of preharvest calcium sprays on apple and pear quality. Journal of Plant Nutrition 16:1807-19

Rawnsley B, Wicks TJ, Scott ES, Stummer BE. 2004. Diaporthe perjuncta does not cause Phomopsis cane and leaf spot disease of grapevine in Australia. Plant Disease 88:1005-10

Reynolds KM, Madden LV. 1988. Analysis of epidemics using spatio-temporal autocorrelation. Phytopathology 78:240-6

Schabenberger O, Pierce FJ. 2001. Contemporary statistical models for the plant and soil sciences. Boca Raton, FL: CRC Press

Scheper RWA, Crane DC, Whisson DL, Scott ES. 2000. The Diaporthe teleomorph of Phomopsis taxon 1 on grapevine. Mycological Research 104:226-31

Schilder AMC, Erincik O, Castlebury L, Rossman A, Ellis MA. 2005. Characterization of Phomopsis spp. infecting grapevines in the Great Lakes regions of north America. Plant Disease 89:755-62

Schilder AMC, Gillett JM, Sysak RW. 2000. Evaluation of fungicides for control of phomopsis cane and leaf spot of grape, 1999. Fungicide and Nematicide Tests 55:105

Schilder AMC, Gillett JM, Sysak RW, Wise JC. 2002. Evaluation of environmentally friendly products for control of fungaldiseases of grapes. 10th International Conference on Cultivation Technique and Phytopathological Problems in Organic Fruit Growing and Viticulture., pp. 163-7. Weinsberg, Germany: Fordergemeinschaft Okologischer Obstbau e.V. (FOKO); Weinsberg; Germany

Schmale DG, III, , Shah DA, Bergstrom GC. 2005. Spatial patterns of viable spore deposition of Gibberella zeae in wheat fields. Phytopathology 95:472-9

161 Schwabe WFS, Jones AL. 1983. Apple scab control with bitertanol as influenced by adjuvant addition. Plant Disease 67:1371-3

Sergeeva V, Nair NG, Barchia I, Priest M, Spooner-Hart R. 2003a. Germination of b conidia of Phomopsis viticola. Austrarian Plant Pathology 32:105-7

Sergeeva V, Nair NG, Barchia I, Priest M, Spooner-Hart R. 2003b. Germination of beta- conidia of Phomopsis viticola. Austrarian Plant Pathology 32:105-7

Shah D, Bergstrom GC, Ueng PO. 2001. Foci of Stagonospora nodorum blotch in winter wheat before canopy development. Phytopathology 91:642-7

Shah DA, Pucci N, Infantino A. 2005. Regional and varietal differences in the risk of wheat seed infection by fungal species associated with fusarium heaf blight in Italy. European Journal of Plant Pathology 112:13-21

Steurbaut W. 1993. Adjuvants for use with foliar fungicides. Pesticide Science 38:85-91 Sutton TB. 1996. Changing options for the control of deciduous fruit tree diseases. Annual Review of Phytopathology 34:527-47

Taylor LR. 1961. Aggregation, variance and the mean. Nature 189:732-5

Taylor LR, Perry JN, WoiwoId P, J. TRA. 1988. Specificity of the spatial power-law exponent in ecology and agriculture. Nature 32:721-2

Thomson LA. 1986. A guide to agricultural spray adjuvants used in the united states. Fresno, CA: Thomson publications

Travis JW, Hed B. 2001. Evaluation of fungicides for control of powdery mildew and phomopsis canse and leaf spot on grapes, 2000. Fungicide and Nematicide Tests 56:SMF30

Tubajika KM, Civerolo EL, Ciomperlik MA, Luvisi DA, Hashim JM. 2004. Analysis of the spatial patterns of Pierce’s disease incidence in the lower San Joaquin Valley in California. Phytopathology 94:1136-44

Turechek WW, Madden LV. 1999. Spatial pattern analysis of strawberry leaf blight in perennial production systems. Phytopathology 89:421-33

Turechek WW, Madden LV. 2000. Analysis of the association between the incidence of two spatially aggregated foliar diseases of strawberry. Phytopathology 90:157-70

Uecker FA. 1998. A world list of Phomopsis names with notes on nomenclature, morphology and biology. Berlin: J. CRAMER

162 Ulloa M, Hanlin RT. 2000. Illustrated Dictionary of Mycology. St. Paul, MN: The American Phytopathological Society

Webb DA, Holloway PJ, Western NM. 1999. Effects of some surfactants on foliar impaction and retention of monosize water droplets. Pesticide Science 55:382-5

Western NM, Hislop EC, Bieswal M, Holloway PJ, Coupland D. 1999. Drift reduction and droplet-size in sprays containing adjuvant oil emulsions. Pesticide Science 55:640-2

Winder L, Alexander CJ, Holland JM, Woolley C, Perry JN. 2001. Modelling the dynamic spatio-temporal response of predators to transient prey partches in the field. Ecology Letters 4:568-76

Xu X, Madden LV. 2003. Considerations for the use of SADIE statistics to quantify spatial patterns. Ecography 26:281-91

Xu XM, Madden LV. 2004. Use of SADIE statistics to study spatial dynamics of plant disease epidemics. Plant Pathology 53:38-49

Xu XM, Madden LV. 2005. Interrelationships among SADIE indices for characterizing spatial patterns of organisms. Phytopathology 95:874-83

Yang XB. 2003. Risk assessment: concepts, development, and future opportunities. Plant Health Progress:doi:10.1094 Zadoks JC. 1984. A quarter century of disease warning, 1958-1983. Plant Disease 68:352-5

163