CROP YIELD MODEL BASED ON MAXIMUM VALUES OF CUMULATIVE VEGETATION INDICES

This research develops a precision modeling approach for cereal crop yield estimation utilizing remote sensing data within a secure information architecture framework. A two-tier model is proposed wherein the first tier conducts vegetation index dynamics modeling (NDVI, MTCI) through an adaptive modified Monod model based on contemporary differential equation systems, while the second tier performs yield prediction via linear regression and machine learning methodologies to accommodate nonlinear interdependencies. An efficient parametric identification algorithm for models is developed, accounting for their nonlinearity characteristics and employing the Levenberg-Marquardt gradient method for refined parameter optimization. A multi-tier information architecture incorporating blockchain technologies is proposed as a decentralized layer for ensuring data integrity and authenticity to mitigate cyber threats including data poisoning attacks and industrial espionage. An adaptive prediction algorithm based on observation window methodology is implemented, leveraging an ensemble of previously observed trajectories to maximize forecasting precision. Practical applicability is validated through numerical experiments on empirical vegetation index data from rice cultivation. The synthesized findings demonstrate the potential of the proposed methodology for addressing contemporary precision agriculture challenges, systematic food security monitoring, and strategic decision-making processes in the agricultural sector.