The grassland mowing events detection product is an Earth Observation (E.O) change detection module that exploits satellite data along with the use of Artificial Intelligence (AI) algorithms, Machine Learning (ML) and Artificial Neural Network (ANN) algorithms. The product utilises work progressed by similar projects (e.g. SEN4CAP) with the main scope to monitor grassland activity and efficiently and precisely track the key dates when these cultivation events occurred. To achieve this objective, the product will combine Sentinel-1 and Sentinel-2 data, the Normalized Difference Vegetation Index (NDVI) time series, Fraction of Absorbed Photosynthetically Active Radiation (FAPAR), Fraction of Green Vegetation Cover (fCover) and Leaf Area Index (LAI) indices and potentially Very High Resolution (VHR) data, as provided by the Paying Agencies.
The product will provide Grassland Event Maps via a shapefile which will be exported and transferred via FTP or HTTP in an automated way. These Event Maps in the form of a shapefile of grassland mowing detection per parcel will encapsulate all the extracted information regarding the detected events, confidence levels and compliance with the respective mowing regulations.
The grassland mowing events detection tool produced by ENVISION aims at compliance checks. The tool provides alerts at various time points in the growing season and supports compliance checks with greening rules.
Evaluation can re-use the data provided by the tool in various operations. First, the tool can contribute or cross-validate data from other sources to identify natural grassland of context indicator C.05 ‘Land Cover’. CORINE class 321 ‘Grasslands’ are low productivity grasslands situated in rough, uneven ground, steep slopes, frequently including rocky areas or patches of other (semi-)natural vegetation, not mowed, fertilized or stimulated by chemicals that might influence the production of biomass. Mowing events indicate the classification of the land to ‘Permanent Grasslands’ (CORINE class 231), which includes grassland under intensive agricultural use. The same applies to context indicator C.19 ‘Farming in Natura 2000 areas’, which, among others, measures the share of agricultural area and natural grassland under Natura 2000. The tool can also support the estimation of context indicator C.40 and Impact indicator I.11 on ‘Enhancing carbon sequestration: Soil organic carbon in agricultural land’, and the result indicator R.14PR ‘Carbon storage in soils and biomass’.
The tool's adoption requires access to EO, the adaptation and application of the algorithms recognising changes in the fundamental physical properties (NDVI, LAI, and others) and their training to detect mowing events. The use of the tool assumes that the IT infrastructure is adequate and that the evaluator has the skills to use the data. In general, when using EO, several conditions may limit their utility and functionality. For this tool, the most critical limitation is the extent of inconclusive changes, i.e. parcels for which the mowing event is inconclusive. Inconclusive parcels may be due to specific EO factors such as cloudiness or the prevalence of small parcels and also may be due to difficulties in producing the algorithms to train and forecast mowing events.
When fully operational, the ENVISION platform will be open-source. However, access to the tool will be given primarily to the partners and selected customers identified by the project. Interested stakeholders and prospective users should contact the project holder.