The GRFM Radar Mosaic of the Congo Basin

The GRFM (Global Rain Forest Mapping Project) approach calls for the coverage of a continental scale geographic area (a linear distance of some 6000 km in the case of the Africa mosaic) using a SAR sensor that acquires a scene within a swath of some 75 km and with a ground resolution of approximately 18 m. This is obtained by tiling together several acquisitions that are not taken instantaneously but still within a short time frame (2 months). Moreover due to the orbital characteristics of the JERS-1 spacecraft contiguous swaths are imaged within 2 consecutive days, which assures a smooth time evolution through the whole coverage.

GRFM mosaic on Central Africa and Madagascar. The colored image over Central Africa shows a combination radar data on the Congo River's low water season (January - March 1996) and high water season (October - November 1996).

Input Data

The JERS-1 SAR data acquisitions of the GRFM Africa project include two blanket coverages of the Central Africa tropical region between and . The first coverage was acquired during January-March 1996 and extends between to ; the second during October - November 1996 and extends between to . The two data sets will be dubbed in the following also the "low water" and "high water" data set, with reference to the perceived hydrological state of the main river network in the region, the Congo river and its tributaries. In addition, the low-water data set covers West Africa from to and the island of Madagascar (acquired on January 1997).

The SAR raw data corresponding to these acquisitions were correlated at NASDA Earth Observations Center EOC. The output of this processing phase is a product defined by NASDA as a standard geocoded image in a Universal Transverse Mercator (UTM) projection and dubbed Level 2.1 product. However the rows and columns of the Level 2.1 raster file are aligned with the slant range and azimuth directions, and not with the Mercator reference system.
The low water acquisition comprises 2173 scenes, the high water acquisition 1460 scenes in continental Africa and 263 scenes for the island of Madagascar. All the scenes were acquired on descending orbits.

Detail of the African Radar Mosaic at 200 m resolution (Congo River and swamp forest near the town of Mbandaka, D. R. Congo)

Data Flow

Proper management of the data flow, in terms of hardware resource allocation, operation synchronization and housekeeping information is an issue of primary importance in a complex processing system like the GRFM one.
Typically the following data flow takes place. The NASDA level 2.1 high resolution images are ingested in batches into the system. The data ingestion step (Flow chart, block 1) is the most time critical in the overall chain of operations; for the JRC computing resources the GRFM Africa data set ingestion took approximately two calendar months.
The high resolution data are then processed by the wavelet decomposition block (block 2 in Flow chart) which generates images and texture measures down-sampled at 100 m. Also a texture measure based on the normalized standard deviation is derived from the high resolution data (block 3 in Flow chart). The 100 m products are the basic building block for all successive operations that are needed to build the mosaic; products at this resolution will be called framelets.
The framelets are radiometrically calibrated (block 4 in Flow chart), geo-located and tiled together (block 5 and block 6 in Flow chart). This step produces the baseline bi-temporal mosaics at 100 m resolution.
Larger scale (lower resolution) maps and texture measures are generated again by wavelet decomposition from the baseline products (block 7 in Flow chart). The multi-scale radiometric and texture products are finally used to support further analysis, data distribution, visual interpretation and automatic thematic information extraction (blocks 9 and 10 in Flow chart). Classification in GRFM is performed using algorithms that assume a piece-wise constant image model with additive white Gaussian noise. To adapt the SAR imagery to this type of algorithm, a pre-processing step is adopted that generates edge preserving smooth approximations of the SAR imagery (block 8 in Flow chart). This technique is also based on a wavelet representation.

Flow chart of the GRFM Africa processing chain

For more information on this topic, please contact Gianfranco de Grandi

A CD-ROM holding the full Africa GRFM mosaic is available under the following conditions:

Send requests by email at:

Key Publications

The Global Rain Forest Mapping Project JERS-1 Radar Mosaic of Tropical Africa: Development and Product Characterization Aspects, G.F. De Grandi, P. Mayaux, Y. Rauste, A. Rosenqvist, M. Simard and S. Saatchi, 2000 , IEEE Trans. on Geoscience and Remote Sensing, vol. 38, no. 5.

Regional Scale Vegetation Maps Derived from the Combined L-band GRFM and C-band CAMP Wide Area Radar Mosaics of Central Africa, P. Mayaux, G.F. De Grandi, Y. Rauste, M. Simard, S. Saatchi, 2002, Int. J. Remote Sensing, (GRFM special issue), vol. 23, no. 7.

The ERS-1 Central Africa Mosaic: A New Perspective, G.F. De Grandi, J.P. Malingreau., M. Leysen, 1999, Radar Remote Sensing for the Global Monitoring of Vegetation, IEEE Trans. on Geosc. and Rem. Sens., vol. 37, no. 3, pp. 1730-1746.