Comparison of monitoring- and weather-based risk indicators of botrytis leaf blight of onion and determination of action thresholds

Botrytis leaf blight (Botrytis squamosa) is the key disease for scheduling fungicide sprays in many oniongrowing areas. Various disease predictors have been developed to identify the best time to initiate fungicide spray programs or to time spray intervals. However, individual predictors have not been rigorously evaluated based on their predictive accuracy. Receiver operating characteristic (ROC) curve analysis was used to evaluate the reliability of the following seven predictors at various damage thresholds: (1) the number of lesions on the oldest leaves; (2) the number of lesions on the youngest leaves; (3) the airborne conidia concentration (ACC); (4) the sporulation index (SI); (5) the inoculum production index (IPI); (6) the infection probability (IP); and (7) a disease severity value (DSV). Data on disease intensity were concurrently collected in sprayed and unsprayed plots. The analysis was conducted with data from the entire season (185 samplings) and for data after the critical disease level of 1 lesion per leaf was reached (107 samplings). At damage thresholds of 1 or 5 lesions per leaf, predictors based on biological monitoring generally were more reliable at predicting disease risk than weather-based predictors. At a damage threshold of 10 lesions per leaf, most pairwise comparisons showed that there were no significant differences among the areas under the ROC curves (AUCs) for any predictors except for the AUC for ACC which was significantly higher and for and IP that were significantly higher and significantly lower, respectively, than AUC for most other predictors. The difference between the two types of predictors was smaller when only data after the first fungicide spray were analysed. Best action thresholds were 8.06 to 13.79 conidia per m3 of air, 2.73 to 3.56 lesions per oldest leaf, and 0.16 to 0.43 lesions per youngest leaf. The most reliable weather-based predictor was SI at an action threshold of 82.55 to 86.46 SI value. Best time to initiate the spray program could be predicted using monitoring-based predictors at the lowest action threshold. The interval between sprays could be best predicted using monitoring-based predictors at a higher threshold combined with SI values above 80.