The frequency and timing of mowing activity influence habitat quality, as well as the occurrence and diversity of organisms on agricultural grassland. Therefore monitoring mowing activity plays an important role in estimating land-use intensity, serving as an indirect indicator for monitoring biodiversity in agricultural landscapes.
Mowing events are not recorded in statistics, but they can be detected area wide from space. The development of mowing detection approaches based on time series of satellite data has been the subject of numerous scientific studies and international publications in recent years. However, a wide variety of different methods and data sources have been used. Whilst these studies have demonstrated the general suitability of satellite image time series in various regions, it is not possible to directly compare the results.
As part of the MonViA project, a comparison exercise was therefore launched, involving more than 35 researchers from nine European countries. The study, which has now been published, evaluated the accuracy of ten different grassland mowing detection approaches. It utilised a standardised dataset of satellite imagery and a unique, multi-year reference dataset covering a wide variety of regions across Europe.
The results confirmed the general suitability of the various approaches and also enabled a regionally differentiated analysis of their strengths and weaknesses. Whilst the rule-based approaches demonstrated good accuracy in most cases, the results from two AI-based approaches were more consistent across different regions and years. The combined use of optical and radar satellite data did not always lead to improved accuracy. These findings will be taken into account in the future revision of the proposed MonViA indicators for grassland use intensity, which are currently based on a rule-based approach.
The results have been published as open access and can be compared interactively on a website. Furthermore, it is possible to evaluate your own mowing detection approach using freely available satellite and reference data, and to compare them with the study’s results.
Publication: https://doi.org/10.1016/j.rse.2026.115466
Interactive results: https://modcix.thuenen.de
Data sets: https://huggingface.co/datasets/thunen-earth-observation/modcix