Smart Classification of Summer Crops on the Google Earth Engine Platform

Abstract

The effective monitoring of agricultural areas holds significant importance for economic, social, and political decision-making. This study aims to develop a supervised classification model for identifying summer crops using machine learning techniques applied to multispectral images. The selected study area is the centre-south region of the Ventania system, located in the southwest of Buenos Aires province, Argentina. The Random Forest (RF) algorithm was employed to implement the classification model, utilizing spectral bands and indices derived from Sentinel-2 images available on the Google Earth Engine (GEE) platform. The model achieved an overall accuracy exceeding 94%, demonstrating high reliability in distinguishing maize, sorghum, sunflower, and grassland crops. These preliminary results indicate the model's potential as a valuable tool for agricultural management. Furthermore, the use of GEE enabled the automation of the entire process, reducing work time and enhancing efficiency.