DESMED

Etude de la désertification en Méditerranée.

Monitoring desertification in the Mediterrannean area.

Context

DESMED is an INRIA P3+3 project, funded for the period 2006-2007.

Partnership

• INRIA Clime project-team
• CNR-ISAC, Italy
• INSAT, Tunisa,
• University IbnTofail, Morocco,
• FORTH IACM, Greece.

Objectives

Land degradation is the set of processes, to a large extent caused by human activities, that set-up a pressure on soil quality, in terms of productivity, biodiversity, chemical and hydric properties. Land degradation mainly results from socio-economic pressures, notably the extension of built areas and intensive agriculture, that result in unadapted land use such as: deforestation, overgrazing, irrigation, use of fertilizers. Land degradation is a precursor of desertification. It is now recognized that over 2/3 of arid and semi-arid lands worldwide are degraded and threatened by desertification. This situation has economic consequences (loss of productivity, of ressources), as well ans ecological (floodings, sand winds) and sanitary (threats on food safety, water quality, propagation of vectors of diseases).

The fight against degradation and desertification requires costly measures to be implemented as soon as possible, when first evidences of degradation are detected. Satellite data are very useful for that purpose, as they allow the early detection of precursors of degradation (such as deforestation and overgrazing), and the mapping of land recovery. Satellite data furthermore allow to analyse land degradation at adapted scales, either in space (from local -land parcel- to regional -catchments- and global) and time (quick processes -forest cuts- and progressive transitions -erosion).

The DESMED project aims to monitor land degradation using time-series of low-resolution satellite images, adapted to the regional and global scales. Monitoring is based on the analysis of time-series of vegetation indices acquired by the AVHRR and MODIS sensors. The researches tend to determine which types of soil (healthy and degraded) can be detected at that resolution, and to devise mathematical tools for characterizing temporal profiles in view of mapping land use, land use changes, and precursors of desertification. The approach relies on a training performed on documented sites where high resolution data are available.

Achievements

• Design of an automated learning process, in view of supporting users in selecting the most appropriate satellite data and their preprocessing level, of determining the land use categories observable at satellite resolution, of characterizing temporal profiles, of selecting and training the most appropriate classification method.
• Design of a hierarchical ontological representation of processes, in view of sharing the knowledge with different types of users (scientists in land degradation, in remote sensing, operational end-users), for supporting them in their choices, in the evaluation of results, in the parameterization of operational monitoring chains from the results of learning.
• Application to test sites in Salento peninsula (Southern Italy) and Menzel Habib (Tunisia) for characterizing the dynamics of land use (study under way).

Material

Satellite monitoring of desertification (slides)

Also refer to the document section of this web site.

Contacts

• INRIA: Isabelle Herlin.

 

Last update: 28.10.2013