Sophie Bontemps

Human activities are affecting the environment and modifying the global system (climate, natural risk and biodiversity). On the basis of this fact, developing an ecosystems monitoring capacity has become a major concern to which the Heart Observation technologies are trying to respond.
Ecosystems monitoring, which was until now mainly carried out with the NOAA-AVHRR and SPOT-VEGETATION sensors, could nowadays be improved by using two new high temporal resolution sensors, MODIS on TERRA (NASA) and MERIS on ENVISAT (ESA). Actually, thanks to their better spatial resolution (300m) and higher number of spectral bands, they enable us to plan to reach a finer scale analysis and a better spatial patterns discrimination. However, working with such time series will generate an incredible amount of data and consequently will require new methods and algorithms to exploit as automatically as possible these information flows. This research takes place in this framework.
It aims to develop a generic change detection algorithm which would be based on a statistical comparison of two multispectral time series. This algorithm should detect, in an automatic way, any significant deviation of a spectro-temporal profile in comparison with the previous profile of the same interest area. The changes to detect are natural changes (inundation, exceptional snow coverage, new burned areasÂ…) or human-induced changes (clearings, plantations, extension of building areasÂ…).
The achievement of this objective will involve firstly the segmentation of moderate resolution images on a high number of spectral bands and secondly the development of a comparison method between n-dimensions temporal series.
This first objective will be extended conceiving and implementing a changes pre-interpretation automatic procedure thanks to the integration of several decision criteriaÂ’s associated with a GIS.

2. Research projects

• GlobCover

The GlobCover initiative develops and demonstrates a service for the generation of a global land cover map for the year 2005/2006. The service is designed in a way that further updates of the land cover map are possible at recurrent cost to run the developed system. GlobCover uses as its main data source the Envisat MERIS fine resolution (300m) mode data acquired between end 2004 and mid 2006. For maximum user benefit the thematic legend is compatible with the UN Land Cover Classification System (LCCS).
This new product will update and complement other existing comparable global products, such as the global land cover map at 1 km resolution for the year 2000 (GLC2000) produced by the Joint Research Center (JRC). It is also improves such previous global products, in particular because of the finer spatial resolution (300m).

The GlobCover is an initiative of ESA that was launched in 2005. It is led by an international network of partners including EEA, FAO, GOFC-GOLD, IGBP, JRC, and UNEP.

• GlobCorine

GlobCorine aims to demonstrate an automatic service that can generate in a consistent way a land cover / land use map and a land change indicator, based on a CLC-compatible legend. The project will evaluate the automatic applicability of MERIS FR time series in the CLC update processing. MERIS FR time series could be used to detect hot spot changes, which could be further analyzed in more details with high resolution images. Such a method would allow the update on a regular basis of the land and ecosystem accounts prepared by the EEA on the basis of Corine land cover in Europe.

The GlobCorine project is based on the GlobCover software system but it concentrates on the pan-European area, although the system could be potentially extendable globally. Its main data source is the Envisat MERIS fine resolution (300m) mode data acquired between end 2004 and mid 2006. GlobCorine is an initiative of ESA with the partnership of EEA and is implemented by Université catholique de Louvain - UCL.