to be published in the Proceedings of ISPRS Commission VII Mid-Term Symposium: Remote Sensing: from Pixels to Processes, Enschede (NL), 8-11 May 2006 1 GLOBCOVER : A 300 M GLOBAL LAND COVER PRODUCT FOR 2005 USING ENVISAT MERIS TIME SERIES P. Defourny 1 , C. Vancutsem 1 , P. Bicheron 2 , C. Brockmann 3 , F. Nino 2 , L. Schouten 4, and M. Leroy 2 ,. 1 Research Laboratory in Environmetrics and Geomatics, UCL (Belgium) • 2 Medias-France • 3 Brockmann Consult (Germany) • 4 Synoptics (Netherlands) KEY WORDS: land cover, pre-processing, classification, global scale, mapping, ENVISAT MERIS, GLOBCOVER ABSTRACT: The objective of the ESA GLOBCOVER initiative is to develop a service which, in its first instance, will produce a 300 m global land-cover map, using as its main source ENVISAT-MERIS full resolution data acquired over the year 2005. The overall project includes the development of the pre-processing and classification chain, the production and the validation. Most of the global acquisition planning and the development of the entire processing chain are completed. The challenges of the geometric correction and of a regionally-tuned classification were successfully addressed. This unique global 300 m land cover product is expected to be made available to the international communities in the course of 2007. 1. INTRODUCTION In the context of increasing concern for the future functioning of the Earth system global land cover maps are considered as key source of information with regards to GCM improvement (Solomon et al, 1993), land cover modeling, biodiversity habitat and tropical forest monitoring and global Carbon cycle change. To face up the changes and the degradation of natural resources which currently work on the biosphere, accurate and detailed information on the state and the evolution of the land cover is required. In particular, such information is necessary for a better understanding of the global cycles, for the monitoring of land cover changes and the climatic change simulation. It also has an impact on the environmental security (Kyoto protocol) and the biodiversity. Finally, it would contribute to a sustainable management of forest, agricultural and natural resources at regional, continental and global levels. Early nineties the first global land cover map derived from remote sensing was produced by DeFries and Townshend (1994) using Advanced Very High Resolution Radiometer (AVHRR) (NDVI) data at 1° spatial resolution and then, at 8- km spatial resolution (DeFries et al., 1998). The international IGBP-DIS effort provided the first 1-km global map product derived still from AVHRR data (Loveland et al., 2000; Hansen et al., 2000). The main challenge was to acquire and put together the global data set. The legend and the accuracy were very much constrained by the poor AVHRR data quality (Loveland et al., 1999). However this definitively confirmed the needs for a consistent land cover map for the whole world and raised the validation issue of such 1-km global product. With the launch of TERRA the MODIS land cover product was expected to fulfill the user’s needs thanks to the advanced sensor capabilities. The MODIS land cover map (Friedl et al., 2002) has been generated with a supervised classification methodology that exploits a global database of training sites interpreted from high-resolution imagery in association with ancillary data. The procedure is automatic but supervised, what requires the definition of signatures for the final 17 classes. In this way, the results strongly depend on the quality of the identified signatures. While initially planned to be updated very regularly the global MODIS land cover product still faced major discrimination and confusion problem for various areas. More recently, the Global Land Cover 2000 project (GLC2000) project produced a new Global Land Cover database for the year 2000 thanks to an international partnership of about 30 research groups co-ordinated by the European Commission’s Joint Research Centre (Bartholomé et al., 2002). The project has adopted an ad hoc processing strategy for the different regions and relied on a multiple thematic resolution approach based on the FAO Land Cover Classification System (LCCS) (Di Gregorio & Jansen, 2000) to insure the consistency of the various outputs. This initiative took advantage of the great quality of the VEGETATION data to discriminate 22 different land cover classes at the global level. The GLC2000 product presents a greater level of detail than MODIS product (22 classes for GLC2000 vs 17 classes for MODIS product) and a better accuracy. The overall accuracy of 68.6% (Mayaux et al., 2005) for the GLC2000 global product confirmed the critical role of the geometric and radiometric performance to build the annual time series. This product seems to be recognized as the current reference worldwide. For instance it has been endorsed by FAO and UNEP and selected for the Year 2000 Millenium Assessment. Even if the production of a 1-km resolution global land cover map has been a challenge, most of the regions of the world would benefit from a finer space resolution, given the very large landscape heterogeneity. In that context the recent availability of the MERIS instrument, with its enhanced resolution (~ 300 m), offers a very interesting opportunity to improve the global land cover map product. MERIS has been launched in March 2002 by the European Space Agency onboard ENVISAT, an advanced polar-orbit Earth observation satellite that provides with measurements of the atmosphere, the ocean, and the land and ice surfaces. MERIS is one of the ten ENVISAT on-board instruments. It is a push-broom imaging spectrometer that measures the solar radiation reflected by Earth surfaces in 15 spectral bands in the visible and near infrared spectral domain. The MERIS swath is 1,150 km and the global coverage is obtained in 3 days. The primary mission of MERIS is the measurement of sea colour and the retrieval of atmospheric parameters over the marine domain. The MERIS instrument can also be programmed to acquire data over land but the nominal geometric specifications was defined according to its primary mission.