P. Mayaux

Mapping tropical coastal vegetation using JERS-1 and ERS-1 radar data with a decision tree classifier

The objective of this paper is to investigate the complementarity of JERS-1 and ERS-1 data for mapping coastal tropical regions. We use a decision tree classifier to classify a coastal region of Gabon and describe the feature contribution using the decision tree diagram. The JERS-1 Global Rain Forest Mapping (GRFM) and ERS-1 Central Africa Mosaic Project (CAMP) datasets are used. The result is a land cover map of the west coast of Gabon. The analysis explicitly shows the complementary characteristics of the L- and C-band Synthetic Aperture Radar (SAR) instruments. We demonstrate the usefulness of combined use of L- and C-band data for large area mapping of coastal regions, especially in flooded areas for discrimination of high and low mangroves as well as grasses and tree swamps. The overall classification accuracy increases by 18% over single band classification.

The Global Rain Forest Mapping Project JERS-1 radar mosaic of tropical Africa: development and product characterization aspects

The Global Rain Forest Mapping Project (GRFM) is an international collaborative effort initiated and managed by the National Space Development Agency of Japan (NASDA). The main goal of the project is to produce a high resolution wall-to-wall map of the entire tropical rain forest domain in four continents using the L-band SAR onboard the JERS-1 spacecraft. The processing phase, which entails the generation of wide area radar mosaics from the raw SAR data, was split according to the geographic area. In this paper, the focus is on the part related to Africa. The GRFM projects goal calls for the coverage of a continental scale area of several million km 2 using a sensor with the resolution of tens of meters. In the case of the African continent, this entails the assemblage of some 3900 high resolution SAR scenes into a bitemporal mosaic at 100 m pixel spacing and with known geometric accuracy. While this fact opens up an entire new perspective for vegetation mapping in the tropics, it presents a number of technical challenges. In this paper, we report on the solutions adopted in the GRFM Africa mosaic development and discuss some quantitative and qualitative aspects related to the characterization and validation of the GRFM products. In particular, the mosaic geolocation and its validation are discussed in detail. Indeed, the internal geometric consistency (subpixel accuracy in the coregistration of the two dates), and the absolute geolocation (residual mean squared error of 240 m with respect to ground control points) are key features of the GRFM Africa mosaic. Other important aspects that are discussed are the multiresolution decomposition approach, which allows for tracking the evolution of natural phenomena with scale; the internal semi-automatic radiometric calibration, which minimizes artifacts in the mosaic; and the thematic information content for vegetation mapping, which is illustrated by a few examples elaborated by visual interpretation. Experience gained so far indicates that the GRFM products constitute an important source of information for global environmental studies.

Large-scale vegetation maps derived from the combined L-band GRFM and C-band CAMP wide area radar mosaics of Central Africa

Abstract A new dataset has been compiled by combining the wide area Synthetic Aperture Radar (SAR) mosaics over Central Africa generated in the context of the NASDA Global Rain Forest Mapping (GRFM) and the ESA/EC Central Africa Mosaic Projects (CAMP). The CAMP mosaic consists of more than 700 SAR scenes acquired over the Central Africa region (6° S-8° N and 5° E-26° E) by the ESA ERS satellites; the acquisitions were performed in 1994 (July, August) and in 1996 (January, February) in two different seasonal conditions. The GRFM Africa mosaic consists of some 3900 JERS-1 images acquired over the region (10° S-10° N, 14° W and 42° E) at two dates (January-March 1996 and October-November 1996). In this paper the methods used for combining the two wide area radar mosaics are at first presented. The GRFM Africa mosaic was processed using a block adjustment algorithm with the inclusion of external observations derived from high precision maps along the coastline, which assures an absolute geolocation residual mean squared error of 240 m with respect to ground control points. On the other hand, the CAMP mosaic was compiled taking into account only the internal relative geometric accuracy. Therefore the GRFM dataset was taken as the reference system and the C-band ERS layer composed by rectifying each ERS frame, after down-sampling at 100 m pixel spacing, to the reference mosaic. The rectification procedure uses a set of tie-points measured automatically between each ERS frame and the homologous subset in the JERS mosaic. Due to the different characteristics of the two sensors (microwave centre frequency, viewing geometry, polarization) and the different acquisition dates, each mosaic presents a different window over the same ecosystem. This fact suggests that a new dimension in terms of thematic information content can be added by the fusion of the two datasets. In support of this statement, the complementary characteristics of the two sensors are first discussed with respect to observations related to the vegetation cover in the Congo River floodplain. The potential of the combined dataset for vegetation mapping at the regional scale is further demonstrated by a classification pursuit of the main vegetation types in the central part of the Congo Basin. The main land-cover classes are: lowland rain forest, permanently flooded forest, periodically flooded forest, swamp grassland, and savannah. The classification map is validated using a compilation of national vegetation maps derived from other high resolution remote sensing data or by ground surveys. This first thematic result already confirms that the combined contributions from the L-band and the C-band sensors improve the information extraction capability. Indeed, the radar-derived vegetation map contains better spatial detail than any existing map, especially with respect to the extent of flooded formations.

New perspectives on global ecosystems from wide-area radar mosaics: flooded forest mapping in the tropics.

Large floodplains in the tropics, like the Congo river basin in Central Africa, are interesting ecosystems that function as water storage and faunistic and florensis habitat. Moreover, they host a series of bio-chemical processes, such as methane emission, which have a significance in global change issues. Characterization of these complex ecosystems can be tackled from different view points, such as bio-chemistry, geology, climatology, hydrology, geomorphology, floristics and forest structure. In this paper we focus on forest structure aspects and report about an approach for mapping two thematic classes - the swamp forest and lowland rain forest - by radar remote sensing at regional scale and high spatial resolution. The proposed solution hinges on the recent availability of a large radar mosaic acquired over Central Africa wall-to-wall by the Synthetic Aperture Radar instrument on board the ESA ERS-1 satellite. The focal points and main issues of this study are: the global mapping approach, using continuous spatial sampling over the region of interest; the signal processing techniques; the up-scaling to wide area of local area classification and (more critical) validation techniques. Results achieved so far already show that blanket radar coverage of the tropics can provide thematic information on the forest composition of a whole ecosystem at an unprecedented level of detail and accuracy.

Fusion of the L-band GRFM and C-band CAMP wide area Central Africa radar mosaics: a new data set with unprecedented potential for regional scale vegetation mapping

A new data set has been compiled by combining two wide area SAR mosaics over Central Africa: the L-band JERS-1 Africa mosaic, generated in the context of the NADSA Global Rain Forest Mapping project (GRFM); the C-band ESA ERS mosaic, developed by the JRC Central Africa Mosaic project (CAMP). The GRFM Africa mosaic was geolocated using a block adjustment algorithm which assures an internal geometric consistency at sub-pixel accuracy and an absolute geolocation residual mean squared error of 240 m with respect to ground control points. The projection used is a direct Mercator. The GRFM data set was therefore taken as the reference system and the C-band ERS layer composed by rectifying each ERS frame, after down sampling at 100 m pixel spacing, to the reference mosaic. Each mosaic consists of multiple features (amplitude and texture measures), and acquisitions at two dates; the resulting data set is therefore multi-temporal, multifeature, and multi-frequency. Clearly such a data set presents novel features, and an unprecedented potential for vegetation mapping at regional scale. As an indication, a test case is discussed related to the visual interpretation of the main vegetation types in the Congo river floodplain.

A vegetation map of the Central Congo basin derived from microwave and optical remote sensing data using a variable resolution classification approach

A vegetation map of the Central Congo basin was derived from observations performed by several imaging orbital instruments. These are the synthetic aperture radars on board the ESA ERS and the NASDA JERS-1 satellites (C-band and L-band), and the imaging spectrometer VEGETATION on board SPOT 4. The different properties of the composite microwave and optical observations are exploited in a complementary way to achieve the intended thematic goal. In particular the secondary forest formation that cannot be mapped consistently by the radar instruments, is captured by the optical observations. Information fusion is achieved at the level of the classification maps derived independently from the microwave and the optical instruments. The derived classification product is a data structure (dubbed VARMAP) composed of elementary variable size cells holding class labels. The VARMAP supports the generation of thematic products at different scales according to the end use. The paper touches upon some of the challenging issues that arise in the compilation of wide area multi-resolution thematic products, with emphasis on the classification methodology and in particular on novel non-contextual and contextual clustering techniques. The thematic products cover an area of approximately 576 million square kilometres, while keeping a resolution of 200 m for most of the thematic classes. It constitutes therefore an invaluable source of ecological information both for global change studies and for the sustainable management of local resources.

Cornerstones and epilogue of the GRFM Africa project: a gallery of regional scale vegetation maps

The Global Rain Forest Mapping project (GRFM) is an initiative started by the National Agency for Space Development of Japan (NASDA) in 1996 with the main goal of creating a wall to wall radar map of the tropical belt with homogeneous and consistent characteristics. GRFM Africa-the part of the project related to tropical Africa-has evolved through several years to the stage where significant thematic products have been generated. It is maintained that these products bear relevance to global change studies and to the sustainable management of local resources in the tropics. The objective of this paper is to lend support to this proposition by illustrating through a few examples the results achieved so far. In particular two land cover maps are presented covering respectively the Central Congo basin, and the Gabon country. Validation of these large-scale high-resolution products poses a challenging problem. The method adopted in GRFM Africa is outlined. It is based on comparison with independent thematic information with known error budget, derived from a combination of optical remote sensing observations, national forestry maps and ground surveys.

The ERS-1 Central Africa Mosaic: a new role for radar remote sensing in global studies of the tropical ecosystem

The Central Africa Mosaic project-CAMP-is an attempt to bring spaceborne SAR remote sensing into an entirely new perspective for global studies of the tropical ecosystem. The new approach hinges around the concept of multi-resolution information extraction, whereby using a high resolution radar sensor one can obtain information both at large geographical scale and at fine spatial detail; the access point to the data hierarchy-or the level of detail needed-is driven by the thematic application. CAMP consists of more than 450 ERS-1 SAR scenes, which were acquired on demand and in a short time frame (two months) over the entire Central African continent by the ESA Libreville ground station and correlated by the German PAF at DLR. The project is carried out by the European Commission Space Applications Institute MTV unit at Ispra and within the R/D activity of TREES (Tropical Ecosystem Environment Monitoring by Satellites). In this communication the basic concepts underlying the CAMP project are first summarized; aspects related to the thematic interpretation, the data processing and new initiatives for large scale radar maps are then discussed with emphasis on peculiarities of the CAMP approach.