Use of survival analysis to assess management options for ray blight in Australian pyrethrum fields

The effects of fungicide, cultivar and plant density on the time-to-death of pyrethrum flowers affected by ray blight (caused by Phoma ligulicola var. inoxydablis) in Australia were analysed using nonparametric Kaplan-Meier (KM) estimates and accelerated failure time (AFT) models with a Weibull probability distribution. Analyses using KM estimates and AFT models yielded similar results. The median survival time (T) for all flowers in the fungicide trial was estimated at 53 days [95% confidence interval (CI) = 43-53] in 2000 and 60 days (CI = 51-60) in 2001. In both years, all fungicides tested except copper oxychloride significantly (P [less-than or equal to] 0·0495) increased the duration of flower survival compared with nontreated plots. Significant variation (P < 0·0001) was noted between years and among four cultivars in terms of flower survival, with T values for different cultivars ranging from 41 to 81 days, and averaging 69 days (CI = 60-69) in 2005 and 64 days (CI = 56-64) in 2006 for all cultivars. Planting at a quarter the density currently recommended increased flower survival by 41·8% (χ² = 29·19; P < 0·0001), but did not increase yield. Linear regression indicated that defoliation severity accounted for at least 94% of variation in median survival time. Improved management may be achieved via an integrated strategy incorporating these factors.