Sugarcane Yield Prediction Using Spatiotemporal Analysis on Sentinel-2 Imagery and Machine Learning Models

Abstract

Sugarcane is most prominent crop for sugar production and for bioenergy. For crop management and agricultural resource optimization crop yield prediction become more important. This study introduces remote sensing-based sugarcane yield forecasting framework for sentinel-2A satellite imagery with the help of vegetation indices and machine learning techniques. For analysis Sentinel-2A Multispectral data from duration November 2024 to December 2025 is acquired. Important preprocessing like cloud masking, Area of Interest(AOI) clipping is performed, with extraction of vegetation indices. For the prediction of yield, four models were implemented: Linear Regression, Random Forest Regression, and Support Vector Regression (SVR). Among which random forest model showed good capability (92.50% accuracy) in yield prediction. Harvested yield was 81 tons where as predicted yield was 74.03 tons. Correlation analysis was performed on vegetation indices and crop yields, where positive correlation was observed. Normalized Difference Vegetation Index (NDVI) showed the highest correlation among the all vegetation indices. The study demonstrated that proposed framework for sugarcane yield prediction with combination of vegetation indices and machine learning techniques proves to be good indicator.