ABSTRACT
This chapter, and similarly the following ones in Part 1 of the book, recounts the unfolding of an ambitious radar remote sensing project, aimed at obtaining a snapshot of an entire ecosystem by pasting together many successive orbital radar acquisitions in a mosaic.
However, a radar mosaic is not just one BIG image. It is indeed posited that the radar mosaics of our story introduced a new paradigm in spatial analysis for radar remote sensing of the tropical forests. This point is illustrated by considering, by analogy, a mosaic composed of many colored glass tiles, such as the ones that can be admired in a Roman villa in Sicily.
Picking out a tile from the glass mosaic provides just a color, which cannot convey any particular meaning. The information is obtained through the spatial relationships of a tile with respect to the neighboring ones. Thus picking out a tile corresponds to the destruction of information. Whereas picking out a tile from the radar mosaic reveals plenty of information, which is embedded in the higher resolution of the tile. Information is not destroyed, it is augmented.
By zooming in and out at different resolutions on the mosaic, analysis can be fine-tuned to the thematic context. For example, high resolution for logging roads detection, and lower resolution for mapping the flooding extension in the swamp forests. Incidentally, multiresolution representations are the basic mechanism of human visual perception. It is also through this feature that a link is established with the main theme of the book: spatial analysis.
Chapter 1 describes a joint European Space Agency (ESA) and European Commission (EC) initiative named Central Africa Mosaic Project (CAMP). The project called for the acquisition and compilation of 477 ERS-1 scenes obtained in the period from July 15 to August 28, 1994. The mosaic canvas portrays the whole rainforest domain of Central Africa as well as the Northern and Southern transition zones. The land covered by the mosaic is over 3 million square kilometers (km 2).
The reader is first guided through a quick tour of the main thematic attractions offered by the radar mosaic, including views of the lowland forest, the savanna, and the swamp forests in the Congo Basin.
The compilation of large radar mosaics was not without its problems. The collection of the large mosaic called for the implementation of a bespoke processing chain. The main difficulties with the project are highlighted, and the way they were resolved in the data processing workflow is described.
In particular, the problem of arriving at a visually seamless compilation of the images within the mosaic canvas, while on the other hand preserving radiometric fidelity, is addressed. The solution turns out never to be a win–win strategy. Related issues comprise radiometric changes in time (because of the finite acquisition time of all images), within image radiometric changes in range (because of effective scattering area and radar cross-section dependence on incidence angle), and impact of quantization noise.
