Satellite monitoring of rice canopy using artificial intelligence for sustainable soil management in Venezuela

Abstract

Efficient management of agricultural inputs and the preservation of soil properties are fundamental pillars for the sustainability of modern agri-food systems. This research designed and validated a methodology based on Artificial Intelligence (AI) aimed at estimating canopy cover in rice crops within the central-western region of Venezuela, using radiometric data from the Sentinel-2 satellite mission. The objective was to develop a non-invasive monitoring system capable of optimizing nitrogen fertilizer application, thereby mitigating environmental impact and promoting agricultural soil health. The methodology was based on training a Multilayer Perceptron (MLP) model, whose architecture was optimized to process multitemporal spectral signatures. The statistical rigor of the model was ensured through five-fold cross-validation, which reported strong performance metrics, including an average Root Mean Square Error (RMSE) of 0.0071 and a coefficient of determination (R²) of 0.9996. Additionally, an external verification phase was conducted, confirming the algorithm’s capacity to generalize across diverse phenological scenarios. The results conclude that the integration of artificial neural networks and multispectral remote sensing enables robust inference of biophysical variables, facilitating strategic decision-making in precision agriculture. This technical approach ensures greater production efficiency while positioning itself as a valuable tool for monitoring soil health and reducing the carbon footprint in large-scale cereal production.