VIEW ARTICLE

Linear Models Applied to Variation in Numbers of Cereal Rust Urediospores. V. A. Dirks, Research Associate, College of Veterinary Medicine, University of Minnesota, St. Paul 55101; R. W. Romig, Research Plant Pathologist, Cooperative Rust Laboratory, Department of Plant Pathology, University of Minnesota, St. Paul 55101, Present address of junior author: Northrup, King and Company, Eden Prairie, Minnesota 55343. Phytopathology 60:246-251. Accepted for publication 3 September 1969. DOI: 10.1094/Phyto-60-246.

We developed three models to account for the variation in the cumulative numbers of Puccinia graminis and P. recondita urediospores trapped/cm² on microscope slides in the Mississippi River Basin during 1953, 1962, 1963, 1964, and 1965. These were linear multiple regression models derived from (i) a combination of five biological and six climatological variables; (ii) the biological variables alone; or (iii) the climatological variables alone. The biological variables were developed from regression components associated with the numbers of urediospores trapped. The climatological variables were derived from temperature and precipitation records. The best model for P. graminis was one combining both biological and climatological variables. In the spring wheat area, a significant portion of the variation in final cumulative number of urediospores could be accounted for as early as 2 weeks before heading. However, in the winter wheat area, estimates made 2 weeks before heading were significant only for the 7- and 14-day predictions. For P. recondita, the model with only biological variables accounted for a highly significant portion of the variation in the final cumulative number of urediospores 2 weeks before heading. The model based on the six climatological variables was unsatisfactory for both P. graminis and P. recondita.