A neural network model to predict wheat stem sawfly cutting in solid-stemmed wheat cultivars
Brian L Beres, Bernard D. Hill, Héctor A Cárcamo, Janet J. Knodel, David K. Weaver, and Richard Cuthbert. A neural network model to predict wheat stem sawfly cutting in solid-stemmed wheat cultivars, 2016 Joint Annual Meeting of the Canadian Society of Agronomy and the Canadian Society of Horticultural Sciences, Montreal, QC, Canada. July 24-26. Pg. 22.
The wheat stem sawfly, Cephus cinctus Norton (Hymenoptera: Cephidae), has been a major pest of wheat (Triticum aestivum L.) in the northern Great Plains of North America for more than 100 years. The combined losses from stem boring and eventual cutting by the larva causes the stem to topple to the ground where it is usually not recovered at harvest. The use of solid-stemmed cultivars helps mitigate crop losses and can also affect the survivorship of C. cinctus. The efficacy of ‘resistance’ is based on the plant’s ability to develop pith in the culm of the stem, which is influenced greatly by interactions between the genotype and the environment in which it is grown. Precipitation-related weather interferes with photoperiod and results in reduced pith expression or solidness in solid-stemmed wheat. Given the extensive hectares planted to solid-stemmed wheat, a model that can accurately predict pith expression could serve as a vital quality assurance tool to prevent losses by alerting producers if in-season precipitation patterns have caused less than ideal pith expression in a solid-stemmed cultivar. Neural Network (NN) models are used to make predictions for complex, non-linear systems with many co-related variables. Our objective was to improve upon past models that used regression analyses by deploying a NN model to predict in-season stem cutting of wheat by wheat stem sawfly. Results indicate that stem cutting is influenced by the precipitation within a 6-wk period from May 25 to July 5. These results were successfully deployed in a model that should assist with predictions of potential late season stem cutting. Deployment of this NN model as a transferable executable file should help facilitate predictions of stem cutting by wheat stem sawfly in any given year. Such predictions would alert producers to implement management strategies to reduce losses from wheat stem sawfly.