Author
MEHRA, L - North Carolina State University | |
Cowger, Christina | |
OJIAMBO, P - North Carolina State University |
Submitted to: APS Annual Meeting
Publication Type: Proceedings Publication Acceptance Date: 8/14/2015 Publication Date: N/A Citation: N/A Interpretive Summary: Stagonospora nodorum blotch (SNB) is a major fungal disease of wheat. Pre-planting factors such as previous crop, tillage, host genotype, disease history, and location of a field affect disease intensity, but the relative disease risk due to each of those factors has not been quantified. Experiments were conducted at several locations in North Carolina from 2012 to 2014. Data on disease were related to the pre-planting factors, using several kinds of models: multiple regression (MR), classification and regression tree (CART), and logistic regression. Previous crop, tillage type, location, and cultivar resistance were the most important predictors of AUDPC and DS. The ability of each model to correctly predict whether disease would be low or high based on the field experiment data was observed. Once optimized, these models can provide important information on the risk of disease occurrence during the season and thereby guide growers in making pre-planting management decisions for SNB. Technical Abstract: Stagonospora nodorum blotch (SNB) caused by Parastagonospora nodorum, is a major disease of wheat. Pre-planting factors such as previous crop, tillage, host genotype, disease history, and location of a field affect disease intensity. However, the risk of SNB due to these factors has not been quantified. Experiments were conducted at several locations in North Carolina from 2012 to 2014. For modeling purposes, unique cases were randomly divided into training, validation, and test dataset and this process was repeated 15 times. Multiple regression (MR), classification and regression tree (CART), and logistic regression approaches were used to relate area under disease progress curve (AUDPC) and maximum disease severity (DS) to pre-planting factors. Generally, MR models explained 73% and CART models explained 68% of the variation in AUDPC. When DS was dichotomized into low and high disease level, logistic regression correctly classified 61 to 85% cases in test datasets. Similarly, when MR and CART were used to predict DS and then dichotomized into low and high disease level, they correctly classified 73 to 94% and 76 to 91% cases in test datasets, respectively. Previous crop, tillage type, location, and cultivar resistance were the most important predictors of AUDPC and DS. Once optimized, these models can provide important information on the risk of disease occurrence during the season and thereby guide growers in making pre-planting management decisions for SNB. |