Hierarchical Bayesian Generalized Linear Models for Predicting Stagonospora Nodorum Blotch in Winter Wheat. Part I: Model Development and Evaluation

Reliable and uncertainty calibrated prediction of plant disease epidemics are key for informing disease management decisions. Such performance can be attained through hierarchical models designed to integrate field-level management data with higher-level environmental processes that drive epidemic development. In this study, a Bayesian hierarchical framework employing Beta regression was developed and validated to predict severity of Stagonospora nodorum blotch (SNB) in winter wheat. Nine competing models (M1–M9) combined individual-level management covariates (residue management, cultivar reaction class, disease onset) with group-level predictors derived from up to 117 pre-anthesis weather variables, a key biofix for fungicide decision-making. Inference for this high dimensional weather dataset, generated through window-pane analysis and factor analytic environmental loadings, was stabilized with R2D2 shrinkage priors. The full model (M9; 3 categorical + 117 weather predictors) achieved the highest accuracy ( RMSE : 3.6%, MAE : 3.1%) and outperformed an XGBoost benchmark ( RMSE : 4.6%). A more parsimonious model (M7; 3 categorical + 42 weather predictors) retained interpretability while maintaining robust performance ( RMSE : 3.9%, MAE : 3.2%). Predictive intervals were well-calibrated ( PICP ≈ 97%), but models slightly under-predicted severe epidemics (>15% severity), likely reflecting the constraints of relying exclusively on pre-anthesis weather in the absence of pathogen data. Simulations probing model behavior revealed how asymptote bias and coefficient uncertainty are influenced by group-effect variability, environmental relatedness, and sample size. These results demonstrate that uncertainty aware hierarchical models can resolve the multi scale drivers of SNB, and thereby strengthening risk assessment and management decision in wheat production in the south Atlantic region of the U.S.

Authors

Institutions

Publication Details

Journal
Phytopathology
Published
2026-10-31
DOI
https://doi.org/10.1094/phyto-12-25-0384-r
Primary Topic
Smart Agriculture and AI
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
article

Hierarchical Bayesian Generalized Linear Models for Predicting Stagonospora Nodorum Blotch in Winter Wheat. Part I: Model Development and Evaluation

Peter S. Ojiambo, Vinicius C. Garnica
Phytopathology
Smart Agriculture and AI
article

Hierarchical Bayesian Generalized Linear Models for Predicting Stagonospora Nodorum Blotch in Winter Wheat. Part I: Model Development and Evaluation

Peter S. Ojiambo, Vinicius C. Garnica
article en

Abstract

Reliable and uncertainty calibrated prediction of plant disease epidemics are key for informing disease management decisions. Such performance can be attained through hierarchical models designed to integrate field-level management data with higher-level environmental processes that drive epidemic development. In this study, a Bayesian hierarchical framework employing Beta regression was developed and validated to predict severity of Stagonospora nodorum blotch (SNB) in winter wheat. Nine competing models (M1–M9) combined individual-level management covariates (residue management, cultivar reaction class, disease onset) with group-level predictors derived from up to 117 pre-anthesis weather variables, a key biofix for fungicide decision-making. Inference for this high dimensional weather dataset, generated through window-pane analysis and factor analytic environmental loadings, was stabilized with R2D2 shrinkage priors. The full model (M9; 3 categorical + 117 weather predictors) achieved the highest accuracy ( RMSE : 3.6%, MAE : 3.1%) and outperformed an XGBoost benchmark ( RMSE : 4.6%). A more parsimonious model (M7; 3 categorical + 42 weather predictors) retained interpretability while maintaining robust performance ( RMSE : 3.9%, MAE : 3.2%). Predictive intervals were well-calibrated ( PICP ≈ 97%), but models slightly under-predicted severe epidemics (>15% severity), likely reflecting the constraints of relying exclusively on pre-anthesis weather in the absence of pathogen data. Simulations probing model behavior revealed how asymptote bias and coefficient uncertainty are influenced by group-effect variability, environmental relatedness, and sample size. These results demonstrate that uncertainty aware hierarchical models can resolve the multi scale drivers of SNB, and thereby strengthening risk assessment and management decision in wheat production in the south Atlantic region of the U.S.

Phytopathology
North Carolina State University (US), The Ohio State University (US)
Openalex Percentile: Top 14%
Smart Agriculture and AI
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon