Predição dos parâmetros de qualidade e quantidade de forragem

Abstract

This work presents the design and development of predictive models for the parameters pasture moisture content (PMC), crude protein (CP), neutral detergent fiber (NDF), green matter (GM), and dry matter (DM). The predictors were built using an extensive dataset composed of spectral data from satellite platforms such as Landsat-8, Sentinel-2, and MODIS, combined with climatic variables from eight regions. The study also presents the methodology employed, which includes techniques such as feature selection, PCA, and GridSearchCV, achieving adjusted R² values of up to 63% for moisture content, 60% for crude protein, 24% for neutral detergent fiber, 22% for green matter, and 34% for dry matter.