Assessing the Capability of a Spectral-Based Model in Detection of Northern Leaf Blight in Maize; A Case Study of Asili Maize Farm in Kiryandongo

Abstract

Maize (Zea mays L.) is the world’s most productive crop, with 1.12 billion metric tons harvested in 2018/19. Northern Leaf Blight is a devastating fungal disease caused by the pathogen Exserohilum turcicum that affects maize crops worldwide. It causes premature death of blighted leaves and results in significant yield reductions. 40-70% intensity of NLB and yield loss of 60% due to this disease have been reported in Uganda. Efficient detection of Northern Leaf Blight in maize is vital for effective disease management and maximizing crop productivity. This study aimed to evaluate the effectiveness of a spectral-based model for NLB detection, addressing the limitations of labor-intensive and error-prone visual examination methods. Through the utilization of Sentinel-2 imagery, thirteen vegetation indices were computed. Using the Random Forest Regressor, a spectral model was developed with NLB severity as the dependent variable and the six most suitable vegetation indices as predictors. Among the examined indices, Chlorophyll Red Edge, Leaf Area Index, Structured Insensitive Pigment Index, Normalized Difference Vegetation Index, Moisture Stress Index, and Green Normalized Difference Vegetation Index correlated most with the disease index. These findings demonstrate that these indices can serve as reliable indicators of NLB presence in maize, providing valuable tools for early detection and effective disease management. The model exhibited an impressive performance with a low Mean Squared Error (MSE) of 0.0656 and a high R-squared (R²) value of 0.857. These findings underscore the efficiency of the spectral-based model as a potential alternative to labor-intensive visual examination methods for accurate NLB detection in maize.