Source: WEST VIRGINIA UNIVERSITY submitted to NRP
MANAGEMENT OF FIRE BLIGHT WITH IMPROVED DISEASE FORECASTING AND PLANT GROWTH REGULATORS
Sponsoring Institution
National Institute of Food and Agriculture
Project Status
COMPLETE
Funding Source
Reporting Frequency
Annual
Accession No.
0200620
Grant No.
(N/A)
Cumulative Award Amt.
(N/A)
Proposal No.
(N/A)
Multistate No.
(N/A)
Project Start Date
Jul 1, 2004
Project End Date
Sep 30, 2009
Grant Year
(N/A)
Program Code
[(N/A)]- (N/A)
Recipient Organization
WEST VIRGINIA UNIVERSITY
886 CHESTNUT RIDGE RD RM 202
MORGANTOWN,WV 26505-2742
Performing Department
PLANT & SOIL SCIENCES
Non Technical Summary
MARYBLYT, a widely used system to forecast fire blight in apple orchards, has been found to have predictive weaknesses and sometimes is difficult to use. This study will take historical data plus data collected during this study and incorporate the data sets in refined models that allows for more accurate predictions of fire blight or the probability of fire blight occuring in a given orchard.
Animal Health Component
80%
Research Effort Categories
Basic
20%
Applied
80%
Developmental
(N/A)
Classification

Knowledge Area (KA)Subject of Investigation (SOI)Field of Science (FOS)Percent
2121110116090%
2161110116010%
Goals / Objectives
Revise and validate MARYBLYT to generate a system of risk points based on a quantitative measurement of risk of infection (i.e. a probability). Using the system of risk points; determine optimal threshold values for minimizing blossom blight. Modify and validate MARYBLYT to account for varietal susceptibility and inoculum pressure. Develop a prevalence model for estimating inoculum pressure.
Project Methods
Historical data will be collected on weather and fireblight occurance. This data will be used to revise MARYBLYT. Logistic regression will be used to calculate risk points for various levels of risk factors. Receiver Operating Curve analysis will be used to evaluate the performance of MARYBLYT. Field plots will be used to validate model performace. Varieties of known resistance will be numerically classified. Logistic regression will be used to incorporate this data in the model. Inoculum potential will be estimated based on the number of overwintering cankers. A model for estimating inoculum pressure will be develolped using data from direct inoculation of cultivars under field conditions.

Progress 07/01/04 to 09/30/09

Outputs
OUTPUTS: Maryblyt version 7 for Windows was finalized in August, 2009. This grower-freindly software is the most widely used forecasting system for fire blight of apple and pear. The software and information about he epidemiological concepts upon which the software is based have been disseminated using 1) email update to colleagues, growers, agents, and consultants (list culled from a world-wide list of fire blight professionals); 2) presentations at professional meetings; 3) presentations at grower meetings; 4) written publication in Plant Health Progress; and 5) a web page for delivering the program to potential users. PARTICIPANTS: Alan R. Biggs, Professor of Plant Pathology and Extension Specialist, West Virginia University; William W. Turechek, Research Scientist, USDA-ARS, Fort Pierce, FL; collaborators include Donna M. Kocak, computer programmer, Gary Lightner, computer programmer, other university plant pathologists tested preliminary versions of the final software. The lead scientist (Biggs) will continue to provide training and extension presentations to producers on availability and use of the software. TARGET AUDIENCES: 2010 will be the first year of practical use of this product. We are encouraging producers to become familiar with it prior to the growing season, and several favorable comments have been received. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
Fire blight is one of the most destructive diseases of apples and pears. Throughout most of its history, the disease has been frequently described as "characteristically erratic and unpredictable". Young orchards with no history of the disease could suffer severe tree losses in a single season, while other, more established orchards seemed to escape infection even though the disease may have been common the previous year. The destructiveness of fire blight and the highly sporadic nature of damaging epidemics have encouraged growers to adopt intensive management practices that depend heavily on the use of protective antibiotic sprays applied at frequent intervals every year. This approach, although generally adequate, seldom affords complete control, sometimes fails, and often seems excessive given the amount of disease that subsequently develops in comparable untreated orchards. Many aspects of fire blight have been described since its discovery. More recently, some of the specific requirements governing the infection process, and how these affect the progress of epidemics, have been described. The Maryblyt personal computer software was an early attempt to integrate much of what was known about the disease in apples and pears into a comprehensive model. The primary focus has been to construct a program that fruit growers can use as an aid in making decisions for controlling fire blight in apple and pear orchards. Because of its accuracy, Maryblyt has also found use in research, teaching and extension programs. No one has attempted to improve the software since 1998, and has only been available as a DOS version. This project attempted to improve on Maryblyt by using ROC analysis to determine the relative importance of the individual parameters in the model. We had hypothesized that certain parameters could be more important, and, if so, a model with appropriately weighted parameters could be more effective. Our research determined that the original Maryblyt model predicted fire blight significantly better than chance. We also learned that we could not improve the predictions of the exisiting model without adding more parameters. We determined that additional parameters would detract from the programs ease of use. We awere able to re-write some of the mathematical functions to providec for accurate estimates of the epiphytic infection potential during the 80 degree-day blossom life function. We were able to produce and compile for distribution a version of the software that is compatible with current Windows operating systems (XP, Visat, Windows 7).

Publications

  • Biggs, A. R., and Turechek, W. W. 2010. Fire blight of apples and pears: Epidemiological concepts comprising the Maryblyt forecasting program. Online. Plant Health Progress doi:xx.xxxx/PHP-2009-xxxx-xx-RS. (in press).


Progress 01/01/08 to 12/31/08

Outputs
OUTPUTS: Knowledge generated from these experiments have been shared with Research/Extension personnel at sister institutions. Results have been shared with the scientific community in the form of a research publication. Results have been shared with growers at fruit extension meetings. Model validation assures researchers and growers that disease predictions are valid and that the models used provide the basis for good management decisions. PARTICIPANTS: William W. Turechek and Tim R. Gottwald, U.S. Department of Agriculture, Agricultural Research Service, Ft. Pierce, FL 34945, assisted with data analyses and manuscript preparation. TARGET AUDIENCES: Target audiences include growers/producers and research/extension personnel at sister institutions. Results from this project have been shared with both groups at regular intervals. PROJECT MODIFICATIONS: Nothing significant to report during this reporting period.

Impacts
Maryblyt predicted the onset of shoot blight symptoms within 48 hours of their appearance in 3 of the 4 years studied, and this report is the first to validate the shoot blight portion of the Maryblyt model. In the third year of the study, the first occurrence of shoot blight was observed 6 days later than the predicted date. The basis for shoot blight predictions in the Maryblyt model is the presence of blossom blight or canker blight and 375 accumulated degree days (base 4.4 C) after bud break (i.e., green tip). The required degree day accumulation is based on the appearance of adults of the putative vector white apple leaf hopper. No variations in insect occurrence or development rate were noted, therefore the delay in shoot blight appearance in year 3 could not be explained by differences in vector populations or their rate of development. Additional research is needed to gain a better understanding of the role of insects and weather in the development of shoot blight epidemics. Our observations of blossom blight development in relation to Maryblyt predictions are in agreement with previous studies. Of the four cultivars examined, disease incidence and rate of disease increase was greatest in York. This observation agrees with previous reports of high susceptibility of the cultivar York to E. amylovora. However, Fuji has been reported to be as highly susceptible as York, and our results do not support this ranking. The mathematical description of these fire blight epidemics by the logistic model has been associated with a polycyclic disease cycle, which is typified by the secondary spread of inoculum. For the cultivar York, the apparent infection rate for shoot blight ranged from 0.053 to 0.20/day in the year in which shoot blight was most severe. These values show that disease incidence can increase dramatically, easily doubling in 3 to 5 days when environmental conditions and host susceptibility are highly favorable. It should be noted that these infection rates were observed under a disease management strategy that was considered optimum at the time (i.e. copper applied at the 1-cm stage of leaf development, applications of streptomycin during the bloom period, as determined with the Maryblyt model, and removal and destruction of infected plant tissues on a weekly basis).

Publications

  • Biggs, A.R., Turechek, W. W., and Gottwald, T. R. 2008. Analysis of fire blight shoot infection epidemics on four apple cultivars. Plant Disease 92:1349-1356.


Progress 01/01/07 to 12/31/07

Outputs
Significant differences between the forecasts of Cougarblight and Maryblyt were found with pairwise contrasts in historical data sets from highly susceptible cultivars. Youden's index was used to determine the optimal cutpoint of both forecasters, Maryblyt and Cougar Blight, and the greatest sensitivity and specificity for Maryblyt coincided with the use of the highest risk threshold for predictions of infection. In field studies, analysis of the canker margin ratings resulting from inoculations made on three dates revealed a significant cultivar x date interaction. Incidence of cankers was low on the cultivar Red Delicious relative to the other two cultivars, and fewer cankers developed on Golden Delicious when compared with Rome Beauty. It should be noted that these experiments were conducted under near-drought conditions. In a previous study, Red Delicious exhibited uniformly determinate canker ratings from inoculations made at different times during the season and Golden Delicious and Rome Beauty exhibited cankers with increasingly indeterminate margins as inoculations were made later in the growing season. This pattern of canker margin development was not generally observed in the present study, with the exception of Rome Beauty, the cultivar regarded as most highly susceptible to E. amylovora. Based on canker incidence and canker margin appearance, Rome Beauty was the most susceptible cultivar, followed by Golden Delicious, and Red Delicious, which was almost completely resistant. Cultivar differences in mean canker rating were significant. The trend to form increasingly indeterminate cankers was linear for Rome Beauty, based on regression of the mean canker ratings against inoculation date. For Red Delicious and Golden Delicious, the slope of the linear regression was not significantly different from 0. These results generally confirm those earlier investigators who found that inoculation of young, nursery-grown trees in midseason (July) produced cankers with determinate margins, whereas those inoculated later in the season (September) produced cankers with indeterminate margins. They recovered E. amylovora from intact canker margin surfaces in 15% of the 130 cankers examined. The bacterium was recovered more frequently from cankers with indeterminate margins than from those with determinate margins. In the present study, recovery of E. amylovora was infrequent (6.7%) and limited to cankers initiated in July and August, but not in June.

Impacts
Comparison of models demonstrated the utility of Maryblyt for predicting disease on susceptible cultivars in the Mid-Atlantic region. Improved understanding of the production of initial inoculum and the effect of cultivar on the production of initial inoculum could lead to improved disease prediction with the Maryblyt program.

Publications

  • Dewdney, M. M., Biggs, A. R., and Turechek, W. W. 2007. A statistical comparison of the blossom blight forecasts of Maryblyt and Cougarblight with receiver operating characteristic (ROC) curve analysis. Phytopathology 97: 1164-1176.


Progress 01/01/06 to 12/31/06

Outputs
Fire blight infections were created by artificially inoculating the apple cultivars 'Rome' and 'Golden Delicious' in order to study the production of primary inoculum in the spring of the following year. Thirty shoots on three trees each of the two cultivars were inoculated in June, July, and August, 2005, for a total of 180 overwintering cankers. In spring, 2006, cankers were swabbed and the swabs rubbed on to CCT medium for detection of the fire blight pathogen. At bloom, only 4% of cankers exhibited external populations of the pathogen as detected with the swab method. Destructive sampling of the cankers after the bloom period yielded pathogen detection in 8% of cankers. There were no differences between the two cultivars in proportion of cankers with detectable pathogen, either internal or external. Inoculations were repeated in 2006, with plans to sample cankers in spring, 2007. Inoculations of apple cultivar 'Red Delicious' failed to produce fire blight infections and overwintering cankers for the second consecutive season.

Impacts
Improved understanding of the production of initial inoculum and the effect of cultivar on production of initial inoculum could lead to improved disease prediction with the Maryblyt program.

Publications

  • Dewdney, M. M., Biggs, A. R., and Turechek, W. W. 2007. A statistical comparison of the blossom blight forecasts of Maryblyt and Cougarblight with receiver operating characteristic (ROC) curve analysis. Phytopathology 97: (accepted for publication).


Progress 01/01/05 to 12/31/05

Outputs
Overwintering fire blight cankers were initiated on the apple cultivars 'Rome' and 'Golden Delicious' for studies on the production of primary inoculum. Thirty shoots on three trees each of the two cultivars were inoculated in June, July, and August, 2005, for a total of 180 overwintering cankers. Inoculation on a third cultivar, 'Red Delicious', failed to produce fire blight infections and overwintering cankers. A Windows version of the Maryblyt program for predicting fire blight was tested in 2005 and will be available commercially in February 2006.

Impacts
Improved understanding of the production of initial inoculum and the effect of cultivar on production of initial inoculum could lead to improved disease prediction with the Maryblyt program.

Publications

  • Lightner, G., Biggs, A. R., and Turechek, W. 2005. Maryblyt for Windows version 6 (research version) pre-release. Computer Software.


Progress 01/01/04 to 12/31/04

Outputs
Two hundred and forty-three historical data sets of environmental data and disease incidence for individual orchard blocks were collected from co-operators in British Columbia, England, Michigan, New York, Quebec, Vermont, Washington State, and West Virginia. Data sets were classified as either "cases" if fire blight was observed that season or "controls" if fire blight was not observed. For each data set, the weather data were entered into Maryblyt and Cougarblight. The frequency of cases and controls for each set of single and combined threshold values was recorded for both models. For each possible threshold value tested, the prediction was recorded and subsequently compared to field observations. Cougarblight was evaluated with the use of the rain threshold (Cougarblight and rain) and without it (Cougarblight). For each threshold value, the numbers of true negatives, false negatives, false positives and true positives were recorded. Receiver operating characteristic (ROC) curves were created for both forecasters as a plot of the sensitivity (= TP/# of cases) versus 1 minus specificity (= FP/#of controls). ROC curve analysis helps to determine whether a forecaster is performing better than chance as well as to help determine optimum cutpoints. Data were analyzed as a whole, then partitioned by region and cultivar susceptibility and the analysis was rerun to determine the effects of these variables. The area under the curve (AUC) was calculated for each curve. A 95% confidence interval (CI) was calculated around the AUC via bootstrapping and the models were compared. The frequency distributions of cases and controls in the data set "All Data" for each model were plotted as a function of their tested threshold values. The majority of the cases were classified by value 7 (infection) in Maryblyt. The cases were more evenly distributed across the values for Cougarblight with or without the rain threshold. With Cougarblight and Cougarblight and rain, the controls were similarly distributed to the cases. However, with Maryblyt the controls were more evenly spread among the different threshold values than the cases. The models performed similarly for all regions and cultivar susceptibility levels. Based on the overlapping 95% confidence intervals there were no significant differences between Cougarblight and Maryblyt with any data set.

Impacts
Our analyses show that Maryblyt and Cougarblight have equivalent predictive capacity. In general Cougarblight generates fewer false positives than Maryblyt, improving specificity. We determined that the Cougarblight category "high risk" appears equivalent to Maryblyt category "infection" and that the Cougarblight category "extreme" has good specificity but is not particularly sensitive. It is possible that both models would be improved by the inclusion of cultivar susceptibility.

Publications

  • No publications reported this period