ABSTRACT
Following the path laid by the 1994 CAMP project (see Chapter 1), the initiative for the compilation of a new and extended tropical vegetation radar mapping project was launched in 1996 by the National Agency for Space Development of Japan (NASDA). The scope of the project was to produce a pan-tropical map of the world’s forests from South and Central America to Africa, Asia and Australia using imagery acquired by the L-band radar aboard the JERS-1 spacecraft. Accordingly, the project was code-named Global Rain Forest Mapping (GRFM). The project called for truly international collaboration. In particular, the EC Joint Research Center and Caltech-NASA Jet Propulsion Laboratory (JPL) took the lead for the part related to tropical Africa. It is this part (GRFM Africa) that is presented in this chapter.
While the development of GRFM radar mosaic stemmed from the same underlying principles as the predecessor CAMP mosaic, it still offered several different and novel aspects:
Microwave observations at a longer wavelength (L-band, 22 cm), that in turn allowed for a different scattering scenario, including deeper penetration into the vegetation volume, higher sensitivity to the woody components and capability of sensing the ground condition (e.g. flooding).
Bi-temporal acquisitions (January–March 1996 and October–November 1996), corresponding to the perceived hydrological state (low water and high water) of the main river network in the region – the Congo River and its tributaries.
The development of a novel method for the geolocation and radiometric revision of the individual scenes that compose the mosaic. The method is based on a global error minimization algorithm, rooted in least-square estimation theory.
The implementation of a multiresolution decomposition of the mosaic based on the wavelet transform. This representation allows the image resolution to be adapted to the thematic context, the basic paradigm of mosaic spatial analysis (see Chapter 1).
This chapter deals in detail with the inner workings of the geolocation method, and presents results of an experimental geolocation validation procedure.Wavelets appear for the first time in this chapter in their role as acting for spatial analysis, and they will take center stage, though in different contexts, in later chapters.
The opportunity to compile a second edition of the GRFM Africa mosaic was offered in 2008 through the availability of the Shuttle Topographic Mission (SRTM) digital elevation model (DEM). The DEM made it possible to improve the GRFM’s mosaic accuracy, both in the spatial domain (orthorectification) and in the radiometric domain (backscattering coefficient dependence on the local incidence angle).
A neat solution was adopted that did not require the reassembling of the entire mosaic. Instead, a virtual backscatter image was obtained by illuminating the DEM in the radar slant range geometry, and warping the original mosaic to the virtual image by co-registration through homologous features generated by topographic effects.
