Detection of Winter and Spring Crops based on Vegetation index time series with Random Forest algorithm: A Case Study of Bela Crkva Municipality, Serbia

Machine learning represents a scientific field that focuses on knowledge extraction from data. Random Forest is a machine learning algorithm based on supervised classification. The algorithm builds a large collection of independent decision trees, and as a final result, displays the class selected by the majority of decision trees. The application of the Random Forest algorithm in combination with Vegetation Indices has become a standard in remote sensing as a result of its high accuracy in classification and monitoring of vegetation conditions. Vegetation index time series enable the monitoring of temporal variations during vegetation periods. Detection of winter and spring crops was performed by employing the Normalized Difference Vegetation Index (NDVI) and the Normalized Difference Moisture Index (NDMI) in three different vegetation periods (end of the winter, spring and summer) from February to July 2025. The research area was the Bela Crkva Municipality, which covers 353 square kilometres located mostly on flat terrain in the northeastern part of Serbia. The vegetation indices were calculated using bands obtained from the ESA Copernicus Sentinel-2 satellite mission. The results obtained by the Random Forest algorithm indicate that the most important index for detecting winter and spring crops according to the model was summer NDVI (27.43%). In addition, the most important vegetation indices were from the summer period (46.03%), followed by spring (31.49%) and winter (22.48%). Winter crops covered 16.65% of the research area, while spring crops were identified across 33.02% of the area. To the best of our knowledge, this research represents one of the first attempts to integrate remote sensing and machine learning methodologies with vegetation index time series within the research area. The study indicates that vegetation indices acquired during the summer phenological stage provide the highest discrimination capability for distinguishing winter and spring crop areas.