Abstract

Abstract The objective of the GLOBCOVER initiative is to develop a global land-cover map for the year 2005, using as main source of data the fine resolution (300m) mode data from MERIS sensor on-board ENVISAT satellite in course of acquisition over the full year 2005. The thematic legend of the final product is intended to be compatible with the FAO Land Cover Classification System (LCCS). The service will be developed in a way that any further update of the Global Land Cover map will be at recurrent cost to run the developed system. This new product is intended to update and to complement the other existing comparable global products, such as the global land cover map at 1 km resolution for the year 2000 (GLC-2000) produced by the JRC. It is also expected to improve such previous global products, in particular through a finer resolution (300m) and the corresponding geolocalisation required to use time series of images to capture the phenological cycle of some vegetation covers. The initiative includes a validation phase similar to the GLC-2000 validation exercise. The GLOBCOVER initiative is leaded by an international network of partners, in particular, JRC, F AO, UNEP, IGBP, GOFC and EEA are considered as the GLOBCOVER partners. The data acquisition and pre-processing is performed by ESA, the development of the GLOBCOVER V1 is developed by a consortium lead by MEDIAS France under contract with ESA and the GLOBCOVER V2 will be developed by JRC once the GLOBCOVER seasonal mosaics will be ready. Why Globcover? Building on the success of the GLC-2000 project (Global Land Cover map for the year 2000) coordinated by the JRC, ESA decided to launch the GLOBCOVER initiative in the framework of its Data User Element (DUE). The GLOBCOVER product is intended for European and international users, such as but not limited to, services of the European Commission, UNEP, FAO, GOFC-GOLD and IGBP. Recent efforts have been done to make European Environmental Agency (EEA) join the GLOBCOVER user group. By example, GOFC-GOLD, which is an international program to monitor forest cover and land dynamics globally, push strongly to improve global land cover assessment (GEOSS, GCOS IP, IGOL …) because existing land data though are not yet fully satisfying and IGBP has an interest in GLOBCOVER as a data source in the ongoing studies of land-cover and land-use dynamics particularly to improve our understanding of the variation of the nature-society dynamics of land management, thereby facilitating the regional and global modelling that is vital for climate impact and sustainability research. The GLOBCOVER concept of global land cover determination at fine spatial resolution with repeatability will contribute significantly to the body of data sources required for this effort. ESA made also considerable effort in order to improve MERIS data Geolocation accuracy. Accuracy better than 150 meters was requested by land community. Globcover can only use MERIS only if such requirement is satisfied. Today, the geolocation accuracy is limited by different factors: the satellite attitude accuracy, the relief deformation, the resampling processing. A final version of Amorgos tool developed by ACRI (France) an dintented to improve geolocation has been delivered to ESA, validation phase is still on going. The input to this software will be the L1B MERIS Full Resolution, restituted attitude files and operational or precise orbit files. DEM used for orthorectification is based on the ACE Digital Elevation Model and is called Getasse 30. BEAM3.2 MERIS orthorectification tool is also using this DEM. Globcover Input Data Status Figure 1. MERIS FRS density data acquisition The GLOBCOVER R&D Research and Development activities are both conducted in JRC and ESA: JRC Strategy: Building on the expertise acquired during the GLC-2000 project and based on the comparison of the GLC-2000 map with the other similar products (IGBP pathfinder, MODIS map, there is a clear demonstration that a regional tuned approach is needed to produce accurate global maps at a specific resolution. It is expected that, once a validated map will exist at a given resolution, it will be easily possible to update such a map at a five years interval through the reproduction of the processing steps, using the initial map as training dataset. Using the GLC2000 methodology, JRC will analyse the MERIS processed data set, assemble a global land cover map (GLOBCOVER V2) and validate the final map. MERIS imagery requires in-house processing for geometric correction. The ESA processing package (BEAM) allowed extracting control points but did not allow producing multi-channel images. ENVI was used in a second step for multichannel geometric corrections with triangulation and cubic convolution options. Existing softwares (such as BEAM, ENVI or GDAL) recognise the MERIS FR (Full Resolution) standard formats. The new MERIS FRS (Full Resolution Swath) format, which was specifically developed in the framework of the GLOBCOVER project, is not yet included in these softwares. A geometric processing chain based on IDL / ENVI has been specifically developed. The relative geolocation accuracy is not seem fully satisfying the GLOBCOVER requirements (<1/3 pixel size). In particular the Andaman Islands example (comparison of Channel-13 at dates 14 Dec and 27 December 2004) where differences of more than one pixel are visible. Even if the geolocation accuracy does not allow an accurate multitemporal analysis, a visual analysis leads to the following conclusion: for the regions covered by MERIS imagery (Part of Thailand coast including Phuket island, part of South India coast, Andaman Islands and Myanmar coast), damages of the Tsunami event on the vegetation or on the coastline are mostly not visible. Effects are only apparent in very limited areas (such as North of Phuket or along Indian Coast). Satellite imagery at medium spatial resolution (~250 m) can be used to detect the Tsunami effect if areas of more than ~500m width are affected and if (i) tree and vegetation cover is removed and (ii) areas remain flooded. The impact of the flooding is mostly visible the day of the event (26 December) and soon after; already one day later (27 December) the recognition of the temporary flooding can become difficult. Research in ESRIN: On the other hand, some research works in ESRIN are done in order to improve actual methodologies used for land cover determination. Research are focussing on the appropriateness of MERIS-FR data for land cover mapping in combination, with ASAR Wide-Swath data for cloudy areas and AATSR data especially with the short-wave infrared band (i.e. 1.58 -1.64 µm channel).