Link to home

Predicting Infection Risk of Hop by Pseudoperonspora humuli

October 2009 , Volume 99 , Number  10
Pages  1,190 - 1,198

David H. Gent and Cynthia M. Ocamb

First author: United States Department of Agriculture--Agricultural Research Service, Forage Seed and Cereal Research Unit, and Oregon State University, Department of Botany and Plant Pathology, Corvallis, OR 97331; and second author: Oregon State University, Department of Botany and Plant Pathology, Corvallis 97330.

Go to article:
Accepted for publication 10 June 2009.

Downy mildew, caused by Pseudoperonospora humuli, is one of the most destructive diseases of hop. Weather factors associated with infection risk by P. humuli in the maritime region of western Oregon were examined for 24- and 48-h periods and quadratic discriminant function models were developed to classify periods as favorable for disease development on leaves. For the 24-h data sets, the model with superior predictive ability included variables for hours of relative humidity >80%, degree-hours of wetness, and mean night temperature. The same variables were selected for the 48-h data sets, with the addition of a product variable for mean night temperature and hours of relative humidity >80%. Cut-points (pT) on receiver operating characteristic curves that minimized the overall error rate were identified by selecting the cut-point with the highest value of Youden's index. For the 24- and 48-h models these were pT = 0.49 and 0.39, respectively. With these thresholds, the sensitivity and specificity of the models in cross validation by jackknife exclusion were 83.3 and 88.8% for the 24-h model and 87.5 and 84.4% for the 48-h model, respectively. Cut-points that minimized the average costs associated with disease control and crop loss due to classification errors were determined using estimates of economic damage during vegetative development and on cones near harvest. Use of the 24- and 48-h models was estimated to reduce average management costs during vegetative development when disease prevalence was <0.31 and 0.16, respectively. Using economic assumptions near harvest, management decisions informed by the models reduced average costs when disease prevalence was <0.21 and 0.1 for the 24- and 48-h models, respectively. The value of the models in management decisions was greatest when disease prevalence was relatively low during vegetative development, which generally corresponds to the normally drier period from late spring to midsummer in the Pacific Northwest of the United States.

The American Phytopathological Society, 2009