A novel composite Index for early-season maize mapping

Maize cultivation significantly contributes to global food security and sustains human livelihoods. Efficient early-season maize mapping is pivotal for forecasting production and informed pre-harvest decisions. Existing approaches rely on prolonged phenological data or available crop labels, limiting their applicability in areas lacking comprehensive data. Thus, an automated, dynamic, and accurate maize identification method for the early growing season is crucial. This study explores spectral bands to distinguish maize early in terms of water content and chlorophyll levels. A novel composite index for dynamic maize identification independent of labels was proposed. Utilizing this index with a multi-temporal Gaussian Mixture Model facilitated early-season maize mapping and identification. Assessments across diverse global regions revealed the method's robustness, consistently achieving 90% accuracy and F1-score. NDCI outperformed other indices, enhancing F1-score by up to 30%. NDCI-mGMM accurately generated maize maps two months pre-harvest, promising an F1 score of at least 77%. Operating autonomously from labels, this framework offers swift and precise maize identification in data-deficient regions, revolutionizing global food security and trade forecasts.