Predicting Stripe Rust Severity on Winter Wheat Using an Improved Method for Analyzing Meteorological and Rust Data

Publisher Website
Google Scholar
Abstract
An improved method was used to determine more precisely than previous methods the relationship of meteorological factors and stripe rust (caused by Puccinia striiformis) on winter wheat [Triticum aestivum L. em Thell], cultivars Gaines, Nugaines, and Omar at Pullman, WA. A computer program WINDOW was written and used to analyze metereological data for 1967-1984 in segments of 21-65 days beginning on 29 July of each year and ending on 24 July of the following year. Metereological factors were used as the independent X-variables in multiple regression with disease index (DI) used as the dependent y-variable. For each cultivar, four statistical models (two two-variable and two three-variable) provided more accurate predictions than either the local or regional models previously used in the Pacific Northwest. The three-variable models had adjusted R2 = 0.73-0.88, and were 89-100% accurate for predicting rust severity. Contingency quadrants were used to evaluate accuracy of predicted DI versus actual DI. Winter temperature and spring precipitation factors were included in the proposed three-variable models and were positively correlated with DI. Two models for each cultivar were "predictive" in that they could have been used early enough in the season to allow application of fungicides if severe disease had been predicted. The number of days with maximum temperature greater than 25 C was important in each full-season model. For Gaines and Nugaines (cultivars with high-temperature, adult-plant resistance), high temperatures were necessary for their resitance. The frequency of this factor from 21 April to 26 June was highly correlated (r = -0.88 and -0.90) with DI. However, for Omar, a cultivar without resistance, that factor was not important until June. Model validation included making DI predictions for 1985 and 1986, years not used in model development. The models should be used with caution whenever input data exceeds the range of the modeled data.