Optimal Sample Size and Composition for Crop Classification with Sen2-Agri’s Random Forest Classifier

Sen2-Agri is a software system that was developed to facilitate the use of multi-temporal satellite data for crop classification with random forest (RF) classifier in an operational setting. It automatically ingests and processes Sentinel-2 LandSat 8 images. Our goal provide practitioners recommendations best sample size composition. The study area located Yaqui Valley Mexico. Using polygons more than 6000 labeled fields, we prepared sets training, which nine crops had equal or proportional representation, called Equal Ratio, respectively. Increasing training set improved overall accuracy (OA). Gains became marginal once total number fields approximated 500 40 45 per type. achieved slightly higher OAs Ratio given fields. However, recall F-scores individual tended be Equal. high wheat scenarios, ranging from 275 2128, produced accurate maximal 80 This resulted turn limited errors commission non-wheat crops. Thus, representation preferable yields better accuracies, even minority