Early detection of canola fields prior to the flowering stage has emerged as one of the main challenges in crop monitoring, as the spectral signature during early growth stages shows considerable overlap with similar crops. In this study, the capability of early detection of canola fields in the Parsabad region of Ardabil Province was evaluated using Sentinel-2 satellite imagery and the Random Forest algorithm. To this end, in the first step, the performance of Sentinel-2 and Landsat-8 imagery during the flowering period (April and May) was compared, confirming the superiority of Sentinel-2 with an overall accuracy of 0.94 and a Kappa coefficient of 0.87, compared to 0.84 and 0.63 for Landsat-8. In the second step, five spectral indices — NDVI, EVI, NDRE, CFI, and NDYI — were extracted and analyzed as time series for the growing season from October 2016 to May 2017. The NDRE index, requiring the Red Edge band, is exclusively computable from Sentinel-2 imagery and was therefore incorporated into the index set at the early detection stage. Classification results demonstrated that during the winter season (January to March), canola can be detected at least one to two months prior to flowering with an overall accuracy of 0.91 and a Kappa coefficient of 0.81. In contrast, autumn classification (October to December) yielded a Kappa coefficient of only 0.55 due to high spectral overlap with cereal crops. Furthermore, excluding March from the winter dataset and repeating the classification solely for January and February resulted in an accuracy decline to 0.88, highlighting the critical role of March in distinguishing canola from cereals. The findings of this study demonstrate that combining multi-index spectral time series, particularly NDRE, with the Random Forest algorithm within the Google Earth Engine platform constitutes an effective approach for regional-scale early detection of canola.