This project was selected through an SSGP competition run using Small Business Research Initiative (SBRI), supported by Innovate UK.
The project was defined to meet the SSGP specified challenge of “How can Copernicus Sentinel-1 data be used in fulfilling Defra’s monitoring requirements for CAP compliance?” In 2015, the CAP Basic Payment Scheme was introduced, which includes new greening requirements for crop diversification and establishment of Ecological Focus Areas (EFAs). EC regulations require national administrations (Rural Payments Agency in England) to undertake detailed checks on 5% of farmer applications each year, and this is now mainly carried out using remote sensing controls.
Most of the checks require very detailed mapping which is only possible using commercially available optical satellite data with a very high resolution of 1m or less. However, there are some which can use new Sentinel data, which have spatial resolutions of around 10m and are freely available for research and commercial use. The project has looked specifically at the potential of Sentinel-1, which is an all-weather radar satellite providing 12-day repeat coverage over most of Europe.
The application of Sentinel-1 for crop diversification checks has been the main development. The new rules require all but the smallest arable farmers to grow two or three different crops and satisfy overall percentage crop area requirements, and this means that the annual remote sensing controls now need to be able to verify the crop types grown as part of the regulatory activity.
In order to identify a range of different crop types using remote sensing it is necessary to use multi-temporal satellite data covering the main growing period. In 2015, the EC provided optical satellite data to national administrations to help undertake these new checks, but in England and other northern European countries there were practical difficulties in doing this because of image acquisition problems and cloud cover. The advantages of an alternative approach involving the use of Sentinel-1 data is that radar data are obtained independent of cloudy weather conditions and measurements are better calibrated and repeatable than those from optical satellites.
The project developed a new methodology and system for verifying declared crop types based on Sentinel-1 data. After pre-processing of farmer declared field parcels with time series Sentinel-1 data, an automated process is used to verify those with the most commonly grown arable crops, including winter wheat, winter barley, spring barley, oilseed rape, potatoes, sugar beet and maize. These comprise more than 80% of claimed fields across most of the country. In practice only a very small number of parcels with these crops were not able to be verified, which is in line with the fact that very few of the crops are incorrectly declared. RPA evaluated results against available ground data and some false data values supplied to test the system’s robustness, and this demonstrated very high accuracy levels.
There is a national requirement that if the percentage of permanent grassland relative to the area of agricultural land falls to below 5%, then farmers who have ploughed up permanent grassland may have to re-instate it. Identification of permanent grassland has been demonstrated using Sentinel-1, based on the fact that grass fields have a relatively consistent backscatter over the whole year in contrast to arable crops. However, as the main requirement on the UK is to maintain an overall percentage of permanent grassland rather than checking individual fields, further thought is required on how this application could be usefully developed.
A method was also developed for identifying parcels in the LPIS which have significant boundary errors. This involved finding parcels with a high variability in radar backscatter within Sentinel-1 data. This was tested successfully on small areas, but could be scaled up for national implementation. Linked to this, there was also some development of an internet application suitable for farmer mark-up of mapping errors.
Supporting objectives included more efficient processing of time series Sentinel-1 data, and exploring the potential of coherence products. A Sentinel-1 Stack Processor has been developed which automates many of the processes involved in producing registered and calibrated time series data.
The most significant outcome of the work has been to establish a methodology and system for verifying farmer declared crop types for crop diversification controls based on the use of Sentinel-1 data. The plan for Phase 2 is to develop a service which can be offered to the UK National Administrations to deal with crop diversification checks within their operational processes, which is more accurate and cost-effective than the approaches used in 2015. Successful integration of the Sentinel-1 Stack Processor will streamline production and help to ensure timely delivery of crop type verification results.