Hierarchical Bayesian Generalized Linear Models for Predicting Stagonospora Nodorum Blotch in Winter Wheat. Part II: Validation, Yield Loss Quantification, and Fungicide Profitability

Translating plant disease forecasts into profitable management decisions requires quantifying model prediction accuracy, yield response variability, and economic uncertainty. A Bayesian hierarchical framework employing Beta regression was recently used to develop models to predict severity of Stagonospora nodorum blotch in winter wheat. However, performance of these models (validation, yield loss quantification and stochastic fungicide profitability analysis) has not been conducted. Using 394 disease cases from 22 environments (2012-2014), the most complex weather-based model had the lowest root mean square error (RMSE = 11.7%) and near-nominal calibration predictive interval coverage probability (PICP = 100%), although systematic underestimation at high severity (mean error = -16.8% at high disease epidemic scenarios) was observed. Covariate shift analysis across 1,104 synthetic deployment environments identified divergence in some weather predictors. Population-level yield loss was modest (-13.6 kg ha⁻¹ per 1% severity increase) but with extreme environment-specific heterogeneity, indicating that disease severity alone is a poor predictor of yield impact in this pathosystem. A multi-trait Bayesian meta-analytic model jointly estimated disease suppression and yield response to a single fluxapyroxad + pyraclostrobin application at Feekes Growth Stage 8-10, which provided disease protection (46% efficacy) and a severity-independent benefit (+151.3 kg ha⁻¹). Economic simulations showed that breakeven severity shifted from ~5% at low application costs to ~15% at moderate costs under typical wheat prices . Return on investment was more uncertain under higher disease pressure and premium market conditions, where potential gains were also the largest. The integrated framework developed here provides a template for risk-informed decision support system.