Grasslands are a source of goods and ecosystem services. It would therefore be helpful to monitor grass growth and estimate grass productivity indicators in order to optimize grassland management over time. Until today, farmers have had to cope with a lack of regular assessments of grass availability over time across the whole farm. In order to simplify and automate grass measurements, we propose to develop methods for estimating grassland biomass using remote sensing. The aim of this study is to assess the ability of Sentinel-2 remotely sensed data to estimate grassland height as measurements in order to provide farmers with information on the quantity of grass available per agricultural plot. We propose a generic data-driven methodology to identify 1) the set of features derived from Sentinel-2 remote sensing images and 2) a regression technique, in order to yield the best performances in estimating grassland height. Before selecting a subset of features, we generated 1, 935 partly new but potentially meaningful features derived from the spectral indices available. The study was conducted between 2017 and 2020 on 18 farms located in France. The model has been tested and evaluated using the data from 2017 to 2019. The average RMSE (resp. R2 ) is 1.78 ± 0.30cm (resp. 0.70 ± 0.12) on the test set. The RMSE is lower than 10 percent of the range width of the predicted values, indicating a very good assessment of grassland height and this is consistent with the precision required for the grassland management support service. The model has also been evaluated on the data from 2020. The correlation between measurements and es- timations is encouraging with R2 = 0.56 and RMSE = 2.1cm. The majority of the differences are between – 1cm and – 2 cm which are relevant according to grassland management.