Evaluation of the Ryegrass Stem Rust Model STEMRUST_G and Its Implementation as a Decision Aid

• Published in: Phytopathology 2015-01, Vol.105 (1), p.35-44
• Main Authors: Pfender, W F, Coop, L B, Seguin, S G, Mellbye, M E, Gingrich, G A, Silberstein, T B
• Format: Article
• Language: English
• Subjects: Basidiomycota - drug effects; Basidiomycota - physiology; Computer Simulation; Decision Support Techniques; Fungicides, Industrial - pharmacology; Hordeum vulgare; Lolium - drug effects; Lolium - microbiology; Models, Theoretical; Northwestern United States; Plant Diseases - microbiology; Plant Diseases - prevention & control; Plant Stems - drug effects; Plant Stems - microbiology; Seasons; Seeds - drug effects; Seeds - microbiology; Time Factors; Triticum aestivum
• ISSN: 0031-949X; 1943-7684
• DOI: 10.1094/PHYTO-06-14-0156-R

STEMRUST_G, a simulation model for epidemics of stem rust in perennial ryegrass grown to maturity as a seed crop, was validated for use as a heuristic tool and as a decision aid for disease management with fungicides. Multistage validation had been used in model creation by incorporating previously validated submodels for infection, latent period duration, sporulation, fungicide effects, and plant growth. Validation of the complete model was by comparison of model output with observed disease severities in 35 epidemics at nine location-years in the Pacific Northwest of the United States. We judge the model acceptable for its purposes, based on several tests. Graphs of modeled disease progress were generally congruent with plotted disease severity observations. There was negligible average bias in the 570 modeled-versus-observed comparisons across all data, although there was large variance in size of the deviances. Modeled severities were accurate in >80% of the comparisons, where accuracy is defined as the modeled value being within twice the 95% confidence interval of the observed value, within ±1 day of the observation date. An interactive website was created to produce disease estimates by running STEMRUST_G with user-supplied disease scouting information and automated daily weather data inputs from field sites. The model and decision aid supplement disease managers' information by estimating the level of latent (invisible) and expressed disease since the last scouting observation, given season-long weather conditions up to the present, and it estimates effects of fungicides on epidemic development. In additional large-plot experiments conducted in grower fields, the decision aid produced disease management outcomes (management cost and seed yield) as good as or better than the growers' standard practice. In future, STEMRUST_G could be modified to create similar models and decision aids for stem rust of wheat and barley, after additional experiments to determine appropriate parameters for the disease in these small-grain hosts.