New data platforms have made satellite data freely available and farmers can now produce variable rate application (VRA) maps for nitrogen fertilisers and other inputs based on satellite images. The data platforms are currently attracting attention for site-specific weed management because they are free and user-friendly. Satellite imagery is mainly relevant for detecting large, dense weed patches that have unique spectral characteristics in low-resolution images. The objective of this study was to examine whether Sentinel-2 images are useful in the detection of Cirsium arvense and other green vegetation in pre-harvest cereals. It was hypothesised that there is a lower limit for detection and that weeds in large patches are easier to detect than weeds in small patches. Fifteen fields infested with C. arvense were used to evaluate the possibilities of utilising the normalised difference vegetation index (NDVI) from Sentinel-2 images as a weed classifier. High-resolution RGB images from unmanned aerial vehicles (UAV) were used as ground-truthing after classifying green pixels as weeds with C. arvense as the main contributor. The study showed that C. arvense-dominated weed populations were much less aggregated than previously reported. On average, 90 % of the weeds occurred on 50 % of the field area, in a range from 21 % to 72 %. The potential herbicide savings using Sentinel-2 images were in the range of 6 % to 46 %, averaging 24 %. The low weed aggregation combined with low image resolution limited the prospects of using Sentinel-2 images for weed detection. The study showed that UAV imagery offers much greater potential for herbicide savings due to higher image resolution, allowing the detection of individual C. arvense shoots.