Predicting crop disease severity using real time weather variability through machine learning algorithms.

Area of Science:

• Agricultural Science
• Plant Pathology
• Data Science

Background:

• Wheat diseases, specifically yellow rust and powdery mildew, pose significant threats to crop yield.
• Accurate disease severity prediction is crucial for effective crop management and yield optimization.

Purpose of the Study:

• To integrate meteorological variables with machine learning for predicting wheat disease severity.
• To evaluate the performance of artificial neural networks (ANN) and other machine learning models in forecasting yellow rust and powdery mildew.

Main Methods:

• Field experiments conducted over two growing seasons with varying sowing dates.
• Weekly disease severity assessments combined with real-time meteorological data.
• Analysis using Artificial Neural Networks (ANN), Random Forest (RF), and regularized regression models (Elastic Net, Lasso, Ridge).

Main Results:

• ANN models achieved high predictive accuracy for both yellow rust (R²=0.93 validation) and powdery mildew (R²=0.95 validation).
• Random Forest models also demonstrated strong performance (R²=0.93 and R²=0.90 validation, respectively).
• Evapotranspiration, temperature, wind speed, and humidity were identified as key meteorological factors influencing disease incidence via Principal Component Analysis (PCA).

Conclusions:

• Machine learning, particularly ANN, shows significant potential for accurate wheat disease severity prediction.
• These predictive capabilities can form the basis of decision support systems for farmers.
• Informed decisions based on disease prediction can lead to optimized wheat production and reduced crop losses.