Tropical forests mapping of bioko island using remote sensing techniques

Abstract EN

Forest sustainable management requires basically adequate vegetation mapping.

Remote sensing techniques delivers reliable classification scheme of medicinal species Prunus africana located in Bioko Island -Equatorial Guinea.

Prunus africana sustainable management relies principally on the population’s quantification of the sustainable trade volume.

Unsupervised and supervised image classifications techniques were implemented on Landsat OLI-8 (Operational Land Imager-8) to produce P.

africana thematic maps on Bioko.

Primarily, Support Vector Machine classification algorithm realized overall accuracy of 82.01%, with kappa coefficient of 0.79.

Forests misclassification was mainly confined between two interconnected classes of Guineo-Congolian/ Afromontane forest classes and lowland forest classes.

Therefore an extra rule of determent altitude (>1400 m) was added to the classification decision rule to improve the classification accuracies to be estimated as overall accuracy of 80.01% and a kappa coefficient of 0.81.

Regular ground truth data collection from nine transects found that both of P.

africana and Schefflera sp.

were dominantly the two arboreal species located in Bioko’s forests.

Thematic classification maps illustrated in the conducted research is an essential data for the sustainable management of P.

africana bark extraction.

These results may also be valuable for various future studies ranging from primate research to genetic variation of P.

africana on Bioko Island.