Integrating deep learning and field validation into a decision support system for Northern Corn Leaf Blight...

Area of Science:

• Plant Pathology
• Agricultural Science
• Artificial Intelligence

Background:

• Northern Corn Leaf Blight (NCLB), caused by Exserohilum turcicum, is a major global threat to maize production, leading to substantial yield losses.
• Accurate and timely detection of NCLB is crucial for effective disease management and minimizing economic impact on farmers.

Purpose of the Study:

• To develop and validate a maize-specific, web-based Decision Support System (DSS) for real-time NCLB detection and management.
• To integrate advanced deep learning models for automated NCLB diagnosis and provide fungicide recommendations.
• To assess the field efficacy and economic viability of the DSS-guided management strategy.

Main Methods:

• Evaluated thirteen machine learning and deep learning models for NCLB classification, selecting the VGG16 convolutional neural network for its superior performance.
• Developed a web-based DSS (https://maize-nclb.streamlit.app/) incorporating the VGG16 model for automated diagnosis and fungicide advisory.
• Conducted field trials to validate DSS-guided fungicide applications (Azoxystrobin + Difenoconazole) and assess disease incidence, yield, and cost-benefit ratio.

Main Results:

• The VGG16 model achieved 94.0% accuracy in NCLB detection, with high precision, recall, and F1-score (0.94), and an AUC-ROC of 0.93.
• Field validation showed DSS-guided fungicide application reduced NCLB incidence to 6.8% (vs. 67.4% in controls), achieving 90% disease reduction and a 35.4% grain yield increase.
• The management strategy demonstrated a favorable cost-benefit ratio of 1:2.49, with cumulative disease burden reduced by approximately 85%.

Conclusions:

• The developed maize-specific DSS, powered by deep learning (VGG16), provides a reliable tool for real-time NCLB detection and diagnosis.
• Integrating the DSS with field-validated fungicide recommendations significantly enhances NCLB management, leading to substantial disease reduction and improved crop yield.
• This approach demonstrates the potential of precision agriculture technologies for effective and economically viable disease control in maize production.