|Turechek, William - CORNELL UNIVERSITY|
Submitted to: Phytopathology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: May 24, 2004
Publication Date: October 1, 2004
Citation: Turechek, W., Mahaffee, W.F. 2004. Spatial pattern analysis of hop powdery mildew in the pacific northwest: implication for sampling. Phytopathology. 94(10):1116-1128. Interpretive Summary: This research was conduted to determine the pattern of hop powdery mildew disease spread in commerical hop yards in order to develop efficient sampling stratagies to be used by researchers and growers. Statistical analysis of 3 years of data indicated that hop powdery mildew was nearly randomly distributed with no discernable aggegration (foci)suggesting that epidemics are initiated from a well distributed or readily dispersable overwintering population. Accurate estimates of hop powdery mildew disease levels will require observations at the plant level from several rows randomly selected through the field, since disease could occur anywhere in the yard.
Technical Abstract: The spatial pattern of hop powdery mildew (HPM) was characterized using 3 years of disease incidence (DI) data. Commercial hop yards with a history of HPM were divided into 20 row strata (H) with 1 row from each strata randomly sampled. Binomial and beta binomial frequency distributions were fit to N sampling units observed in each row and to 'N sampling units observed in each yard. Distributional analyses indicated that DI was better characterized by the beta-binomial than the binomial distribution in 25% and 47% of the data sets at the row and yard scales, respectively, according to a log-likelihood ratio test. Median values of the beta-binomial parameter theta, a measure of small-scale aggregation, were near 0 at both sampling scales indicating DI was close to being randomly distributed. Variability in DI among rows sampled in the same yard usually increased with mean incidence at the yard scale. Spatial autocorrelation analyses, a measure of large-scale aggregation patterns, indicated that DI was not correlated between sampling units over several lag distances. Results of covariance analysis showed that heterogeneity of DI was not dependent upon variety, region, or time of year when sampling was conducted. Hierarchical analysis showed that DI at the sampling unit scale (proportion of plants with theta 1 diseased leaves) increased as a saturation-type curve with respect to incidence at the leaf level and was described by a binomial function modified to account for the effects of heterogeneity through an effective sample size. Use of these models may permit sampling at the plant scale while allowing inferences to be made at the leaf scale. HPM was nearly randomly distributed with no discernable foci suggesting epidemics are initiated from a well distributed or readily dispersible overwintering population.