SATERRE, an Innovative Project to Provide Added Value Services to Agriculture Monitoring and Land-Use Planning Policy
Savinaud, Mickaël1; Dedieu, Gerard2; Guzzonato, Eric1; Herman, Pascal3; Lenormand, Pauline4; Thumerel, Bernard4

Based on the Copernicus philosophy and new economic model funded on Internet, the SATERRE project aims to provide a new methodological approach for added value service to farmers. The key point of this new approach is to integrate farmer and agricultural organization directly in the definition of a “terroir” indicator. This project is guided by the end user need contrary to the classical approach which imposed uniform product in a close source approach. The services developed during this project will be based on multi-temporal series of optical sensor as SPOT4-5 for example but aims to be used with Sentinel-2 products. This project was funded by CNES in the field of the last call of innovative applications in 2011. It was supported by four partners: Agri-Intranet, CS-SI, Aida and CESBIO and potential end-users.

One of the innovative aspects of this project is to integrate a living lab approach in its core. The living lab aims to build a user-centered ecosystem and produce innovation. It is also well adapted to operate in a territorial context which is the case for agriculture projects. Different meetings have been organized by SATERRE partners to create and expand the end user’s community. The requirements collected during these meetings will be used to define this geospatial service based on "terroir" indicator. This indicator should define the potential of the land based on geospatial analysis and under contextual constraints (legal restriction about water policy for example). Each farmer or agricultural organization can define their land used strategy upon this type of indicators and maps.

Another innovative aspect is the use of web services and open-source solutions to provide service adapted to each farmers and each territorial organization. This organization allows to adapt the cost of the service to each demand and each level of territory resolution. The computation of the different information will be done with the specific inputs and parameters provided by the users.

The geospatial analysis of the multi-temporal series will be done with the image processing methodologies provided by the CESBIO. The CESBIO have developed a strong experience to produce land cover maps based on temporal analysis and classification algorithms. This land cover map will be used to discriminate the different culture and compute vegetation indexes only on interesting area. This vegetation indexes could be segmented with Mean-Shift algorithm to extract different homogeneous region into a field with their corresponding attributes for example (figure 1). These attributes could inform on the potential of the different part of the field. These different regions can be correlated with exogenous data (yield map, hydrographic network and road network) to provide contextual information. This type of product can help the farmer to manage precisely the fertilizer introduction for example.

Figure 1: Segmentation of LAI computed on a multi-temporal acquisition over sunflower culture in the south west of France (with color mapping to enhance the display).

The great majority of the processing chains have been done with the Orfeo ToolBox (OTB). This open source library (CeCILL license, similar to GPL), remote sensing-oriented, image processing library has been initiated by the French Space Agency (CNES) in the frame of the ORFEO accompaniment program. OTB provides to its users an extensive set of algorithms and functionalities dedicated to remote sensing data exploitation. More specifically, it embeds approaches to handle large data using advanced streaming and multi-threading strategies. SATERRE project demonstrator will base its image processing developments on this project and its application framework which can be used in QGIS software or in client/server architecture. First demonstrator is based on QGIS to allow a first experimentation with end-users.

To conclude SATERRE project have collected through a living lab approaches the user needs for a "terroir" indicator in the field of agriculture. This indicator is based on the analysis of multi-temporal series like Sentinel-2. Moreover it will demonstrate a new business model which used web services and image processing to define the land potential on demand and with integration of farmer knowledge. In the context of regional and spatial planning policy, this project should give pertinent inputs to the regional authorities. For example land cover maps, vegetation potential of the land can be used to optimize and help by funding the culture rotation and selection .