|Dewdney, Megan - CORNELL UNIVERSITY|
|Biggs, Alan - WEST VIRGINIA UNIV|
Submitted to: Phytopathology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: November 2, 2006
Publication Date: November 1, 2007
Citation: Dewdney, M., Biggs, A., Turechek, W. 2007. A statistical comparison of the reliability of the blossom blight forecasts of MARYBLYT and cougarblight with receiver operating characteristic (ROC) curve analysis. Phytopathology. 97:1164-1176. Interpretive Summary: Fire blight is one of the most important diseases of commercial apple and pear production. Growers rely on the fire blight forecasters MARYBLYT or Cougarblight to perfectly time applications of bactericides to control disease to produce marketable fruit. A statistical comparison of the forecast accuracy of two common forecasters was performed with receiver operating characteristic (ROC) curve analysis and historical data. ROC analysis is used widely in the medical field for evaluation of diagnostic tests; here we treat plant disease forecasters as a diagnostic tool. Historical data were grouped into geographic regions and cultivar susceptibilities. It was found that the forecasters performed equivalently in the geographic regions of eastern and west coasts of North America and on moderately susceptible cultivars. However, Cougarblight performed better with data collected in England and MARYBLYT performed better with very susceptible cultivars; this reason this occurs is not fully understood. Results reported in this paper will be useful to growers and cooperative extension agents interested in controlling fire bight.
Technical Abstract: Blossom blight forecasting is an important aspect of fire blight, caused by Erwinia amylovora, management for both apple and pear. A comparison of the forecast accuracy of two common fire blight forecasters, MARYBLYT and Cougarblight, was performed with receiver operating characteristic (ROC) curve analysis and historical data. The rain threshold of Cougarblight was analysed as a separate model termed ‘Cougarblight and rain’. Data were used as a whole and then grouped into geographic regions and cultivar susceptibilities. Frequency distributions of cases and controls, orchards with and without disease respectively, in all data sets overlapped. MARYBLYT, Cougarblight and ‘Cougarblight and rain’ all predicted blossom blight infection better than chance (p = 0.05). It was found that the blossom blight forecasters performed equivalently in the geographic regions of the east and west coasts of North America and moderately susceptible cultivars based on the 95% confidence intervals (CI) and pairwise contrasts of the area under the curve (AUC). The AUC was equal to the probability that a forecaster will correctly classify two randomly paired orchards, one with fire blight and the other without. There were no significant differences (p > 0.05) among the three-way contrasts of the AUC’s in any data sets. The pairwise contrasts between Cougarblight and MARYBLYT as well as ‘Cougarblight and rain’ and MARYBLYT for the England data set were significantly different (p = 0.0316) and (p = 0.0419) where the two forms of Cougarblight had greater AUC’s. The pairwise contrast between ‘Cougarblight and rain’ and MARYBLYT for the very susceptible cultivars data set was also significant (p = 0.0269). The AUC for MARYBLYT was greater than that for ‘Cougarblight and rain’. Youden’s index was used to determine the optimal cutpoint of both forecasters. MARYBLYT cutpoint 7 was found to be optimal in most data sets. No one Cougarblight cutpoint was found to be optimal among the data sets.