NIVA’s Use Case 2 provides a solution to support farmers to submit GeoSpatial Aid Application – GSAA faster and with less errors by using automatically detected parcels boundaries layer, early crop detection results and prefilled data by robotized framework.

The Use Case 2 was developed by the NIVA partners NPA, SINERGISE and ITREE. The testing of the Use Case was done by NPA in Lithuania, by the NIVA partner FEGA in Spain and outside NIVA – in Hungary, Turkey and Ukraine.

What is the UC about?

Farmers use GSAA, which allows them to submit parcels’ boundaries, to specify crop types on the respective parcels and to provide additional information to Paying Agencies.

The objectives of the Use Case are:

  • development of the open-source algorithm prototype for the automated parcel preliminary boundary detection and delineation based on Sentinel-2 data;
  • test of Sen4cap algorithms for early stage crop classification, which are based on Sentinel-2 10m resolution images;
  • find and test a robotic process automation tool suitable for data harvesting from external registers.

Through the new tools farmers should be provided with additional data, necessary for submission of the GSAA, thus allowing them to avoid errors and to reduce their administrative burden. Therefore, Paying Agencies should identify less errors in GSAAs and thus reduce their administrative burden related to additional controls.

Testing results

UC2 preliminary parcel boundary detection tool or generated results with boundaries were tested by several testers:

  • Spanish Paying Agency tested generated parcel boundaries in Castilla y Leon and Andalusia regions;
  • Hungarian company Ulyssys Ltd installed, ran the tool and generated results across the whole country;
  • Representatives of Turkish Ministry of Agriculture and Forestry checked generated boundaries in Izmir region;
  • Sinergise provided delineated agricultural fields (seasons 2016-2022) for entire Ukraine to EO4UA initiative.


 A tool-algorithm prototype for automatic delineation:

  • Prototype developed based on Sentinel 2 10m resolution data;
  • Piloted on 1 m super-resolution data generated from Sentinel 2 data.

Prototype based on Sentinel 2 is not accurate enough to generate prefilled parcel boundaries data in GSAA, but its results could be used as an additional data to check parcel homogeneity and prevent farmers from major mistakes. To reach better results, the tool was used on Sentinel-2 1m super resolution image.  In a pilot zone 3 % more boundaries were identified based on super-resolution data, and the percentage of correctly detected boundaries increased dramatically from 15 % (only Sentinel 2) to 75% based on super-resolution.

Automaticaly detected and delineated parcel boundaries based on Sentinel-2 1m (yellow boundary) and Sentinel-2 10m (blue boundary) March 2022

A methodology for early stage crop type detection:

  • Run Sen4CAP crop type detection algorithms not in June-August, but in March-April​;
  • Classified: 80% of fields (small fields problem)​, 98% of land​ 99% of farmers.

During early stage classification 6 crop classes were generated by using Sen4CAP crop type algorithm based on 2019 GSAA parcels and Sentinel S2 images of March-April. Results can prevent mistakes in GSAA where permanent grasslands are declared as arable land, winter crops or winter rapeseed – as summer crops.

A robotic process automation tool – Robot Framework:

  • Open source tool identified and tested
  • White paper, delineating Robot Framework application,

By using tool in NPA 50 records were processed in90 seconds, while manually it would take about 20 – 120 seconds for one record only. Such automation of data harvesting was done 36 times faster and was less prone to errors. Robotic process automation tool can help prevent mistakes in GSAA and significantly reduce time, necessary for farmer to fill the GSAA


Starting from 2023 in Lithuania it is planned to:

  • use parcel boundary delineation tool by providing in GSAA a visual layer of suggested preliminary parcel boundary for farmers to submit precise GSAA and for Paying Agency to meet GSAA QA requirements​ and to check of parcel homogeneity and prevent farmers from mistakes, for example, artificial split of the parcel;
  • Further test early stage crop type/land use classification approach;
  • Move from licensed to open-source Robot Framework software solution.

In parallel it is planned to start using Sentinel-2 1m super resolution March-April and September-October period images on the whole territory of Lithuania  (from​

  • For yearly LPIS update (block/parcel boundaries, non-eligible area, land use change) to overcome the problem of 3-year LPIS orthophoto cycle (1/3 will be update with orthophoto and 2/3 with S2 1m image) ​;
  • Visually in GSAA system for submission of more accurate GSAA (parcel boundaries and non-eligible areas) by farmers​;
  • For execution of remote field checks for non monitorable requirements or yellow AMS parcels​;
  • For prevention of mistakes during LPIS/GSAA/AMS quality assessments​.


Examples how 1m helps to detect accurate LPIS land cover/boundary change

All new data-products will also be provided and visible in geotag app “NMA agro“ as to share new layers with farmers.

More information on the NIVA UC2 can be found on the NIVA website: and the source code is available on GitLab: