Novel action thresholds of a logistic regression model to forecast dollar spot on bentgrasses

Abstract

AbstractDollar spot (Clarireedia jacksonii) susceptibility varies among bentgrasses (Agrostis spp.). This trial assessed the ability of five action thresholds of the risk index (RI) of a logistic regression model to forecast dollar spot incidence on bentgrass fairway turf grown on a sandy loam in North Brunswick, NJ. Dollar spot incidence was assessed over 128 observation periods (May to Nov. annually) over 3 yr and related to five action thresholds for prediction accuracy. Action thresholds included a RI threshold of 20% (RI 20%), interpreting the change in the RI over time (RI slope), adjusting the RI upwards to maximize accuracy (RImax), combining RI 20% with RI slope, and combining RImax with RI slope. The RI 20% accurately predicted disease on 63 to 66% of observation periods for highly susceptible creeping bentgrass (A. stolonifera L.) cultivars Shark, Penncross, and Independence; other action thresholds improved accuracy by only 10% or less for these cultivars. Prediction accuracy on lower susceptibility cultivars (007, Declaration, and colonial bentgrass Capri; A. capillaris L.) was substantially improved with four novel action thresholds by reducing over‐predictions compared to RI 20%. Accuracy was improved by as much as 32% when the RImax and RI slope were combined for Declaration compared to RI 20%. On low susceptibility cultivars, turf managers may be able to use action thresholds that incorporate a RI greater than 20% and/or RI slope to reduce fungicide inputs and maintain acceptable dollar spot control compared to RI 20%.