| Summary: | To map vegetation and land cover changes on the West-African coastline using LANDSAT TM scenes taken at different dates, we applied a stacking of undirected classifications. The first step consists in interpreting the radiometric classes obtained via undirected classification so as to form the classes of the thematic nomenclature. The classification is subsequently improved by increasing radiometric contrasts within each of the classes. To do so, a Principal Component Analysis is applied to the LANDSAT channel. Then, a second classification within the classes enables to reallocate correctly those pixels that were wrongly classified previously. Cross-analysing the 2015 LANDSAT images and maps from earlier decades revealed that between the end of the years 1970 and the mid-2010s, wooded areas increased by 2% for the mangrove and 3.5% for dry land. This positive result is the outcome of significant regressions of woodlands compensated by even more significant progressions. It is in the mangrove that progression prevails over regression, especially in the last fifteen years. Finally, an opposition between particularly stable sectors and particularly changing sectors was observed, given that areas having gone through two changes or more are larger than those having gone through one change only in 36 years.
|
|---|
