Геопространственная информация в поддержку устойчивых продовольственных систем

Use of Sentinel-1 imagery to support detection of land features

With the increasing availability of multi-temporal and multi-sensor remote sensing imagery, mapping land features is less and less dependent on the existence of satellite information and more on the ability to access and use this data in a common environment.

Cloud-computing platforms fill the gap between data and users, providing online analytical capabilities to interact with the images. Understanding the limitations and the necessary steps to get real Analysis-Ready Data (ARD) is often underestimated. This is particularly true when working with radar images which for many users are less intuitive than optical images.

From 4-6 March, FAO geospatial experts attended meeting at the offices of the European Space Research Institute (ESRIN), one of the five specialized centers of the European Space Agency (ESA). It was an opportunity to speak to other international experts from the World Wildlife Fund and the Phi-lab of ESA. The aim was to develop a production chain to pre-process Sentinel-1 imagery and get a summary of multi-temporal parameters, through cloud-computing platforms.

The Sentinel-1 mission is part of the Copernicus constellation to observe and monitor the earth and is composed of two satellites that include C-band imaging, operating in four exclusive imaging modes with different resolution, dual polarization capability and a very short revisit time. The advantage of Synthetic Aperture Radar (SAR) images is to acquire data not impeded by cloud cover, providing information during day and night in almost any weather condition. These data require certain techniques to be geometrically corrected, calibrated and corrected for topographic effects, particularly in mountainous regions.

“While the scientific community continues to investigate efficient techniques for automatic classification, the effort to use calibrated and comparable satellite imagery is somehow neglected” said Gianluca Franceschini, a FAO expert at the meeting who is interested to use Sentinel-1 images to detect rice fields and other crops in Lao People's Democratic Republic.

FAO Geospatial expert, Ece Aksoy, also attended the meeting. She highlighted the importance of finding ways to handle the weekly influx of satellite information. She believes that  the improved crop-classification results which are based on Sentinel 1 and 2 data could help in the creation of practical and accurate crop-maps for seasonal monitoring.

Ece Gultan, FAO Machine Learning expert explained that integrated information from radar and optical data offer more accurate classification. She said that by adding the ancillary data (such as soil properties, climate, terrain), Machine Learning applications can be significantly improved.

For more information on FAO Geospatial team’s work to produce geospatial information for sustainable food systems, visit our website: www.fao.org/geospatial