VIEW ARTICLE

A Method of Predicting Epidemic Development of Wheat Leaf Rust. M. G. Eversmeyer, Research Technician, ARS, USDA Cooperative Investigations, Crops Research Division, ARS, USDA, and Kansas Agricultural Experiment Station, Manhattan 66502; J. R. Burleigh, Research Plant Pathologist, ARS, USDA Cooperative Investigations, Crops Research Division, ARS, USDA, and Kansas Agricultural Experiment Station, Manhattan 66502. Phytopathology 60:805-811. Accepted for publication 4 December 1969. DOI: 10.1094/Phyto-60-805.

A stepwise multiple regression computer program was used to formulate equations to predict per cent disease severities of wheat leaf rust (Puccinia recondita f. sp. tritici). Equations were generated for winter and spring wheats by analysis of biological and meteorological data from 24 winter wheat and 16 spring wheat locations. Equations were in the form Y_i = b₁X_1i + …. + b_nX_ni + K, where X_ni values (independent variables) were measures of disease severity, weekly urediospore numbers, cumulative urediospore numbers, average maximum temperature, average minimum temperature, hr of free moisture as dew or rain per day and days of precipitation, respectively, recorded during the seven days immediately preceding the date of prediction. The dependent variable (Y_i) was the log₁₀ or log_e x/1 – x transformation of per cent disease severity recorded 8 days after the date of prediction. Analysis of data from individual locations was of little value in determining the independent variables to be used in prediction. When data from all winter wheat locations were combined, however, coefficients of determination (R²) indicated that variation in disease severity, minimum temperature, and either hr of free moisture or days of precipitation explained over 70% of the variation in epidemic development. R² values indicated that in spring wheats, weekly or cumulative urediospore numbers could be substituted for disease severity in the prediction model. Partial regression coefficients (b) from 1967 and 1968 combined data were used to predict leaf rust severities at successive wheat growth stages up to hard dough for several 1968 locations. Data from those locations were not used to calculate b values used in prediction. The average variation between actual and predicted severities was 5% when predictions were made between wheat growth stages, boot to milk.