Calibrating a disease forecaster for Elsinoe spot in organic apples using receiver operating characteristic curve analysis


Abstract: Elsinoe spot (Elsinoë pyri), which is a minor disease of apple and pear worldwide, is not a problem in New Zealand apple orchards under conventional synthetic fungicide regimes, but can be problematic in “organic” orchards where lime sulphur and copper fungicides are used. Elsinoë pyri overwinters as ascomata in fallen leaves and as acervuli on unharvested fruit on trees. Epidemics develop from midsummer to leaf fall following rainy weather. A disease forecaster to identify conditions causing high infection risk was developed to aid decisions about lime sulphur spraying. The risk model was derived from data on infection under controlled temperature and wetness conditions. The model outputs a risk indicator variable (0-10) for individual wet periods and indicator values exceeding given thresholds allow adjustment of timing and rate of lime sulphur applications to maximise disease control and minimise phytotoxicrisk. The forecaster was field calibrated using potted ‘Royal Gala’ trap trees exposed in an unsprayed Elsinoe-infected orchard to ‘trap’ infection and identify individual infection periods. Preliminary and revised versions of the forecaster’s risk algorithm were compared using receiver operating characteristic (ROC) curve analysis. Actual infection from the trap plant data and the forecaster’s prediction of infection were used to determine conditional probabilities for true positive, false positive, true negative and false negative predictions. Both algorithms gave significantly better than random prediction of infection events and the revised algorithm, which included a rainfall criterion, was substantially better than the preliminary algorithm.

Cookie Consent with Real Cookie Banner