Abstract :
[en] Every scientific studies should begin with the delimitation of the studied system. Ours aims at evaluating and quantifying the resistence of landscape classes and ecosystems to the urbanization process. This is done by studying a dozen cities in sub-Saharan Africa, conducting a diachronic (2000 - 2010) landscape evolution analysis from SPOT satellite imagery. Paradoxaly, when tackling this subject, one must recognize that no consensus exists about the definition and localization of the areas included in the urban-rural gradient. This prevents from comparing the results of different cities.
A bibliography analysis has been conducted in order to 1) identify the different areas in the urban-rural gradient, the characteristics and types of characteristics used to define the most cited ones (i.e. urban, suburbs, sprawl, exurban, rurban, periurban and rural) ; 2) Through citation frequency indexes, evaluate the relative importance of characteristics and types of characteristics for every area and then for the whole gradient; 3) Evaluate the principal characteristics according to a series of criteria (the best characteristic is supposed to be quantitative, integrative, marking a consensus, discriminative and easy to apply on the field); 4) On the basis of retained characteristics, propose single and simple definitions to the most cited areas. These new definitions aim at enable areas identification on the field and on satellite images. These new definitions have been applied to the field study of the city of Lubumbashi (D.R.C) and seem to be convenient.
Retained characteristics have then been translated into landscape composition indexes for the future study of the following cities on basis of satellite imagery, without field research. Indeed, such indexes are commonly used in landscape ecology because they allow the description of the urban landscape pattern or structure which, according to the central hypothesis of landscape ecology, i.e. “pattern-process paradigm”, influence landscape ecological processes.