Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Master → Canada → Caractérisation spectrale de la dégradation des milieux naturels en régions semi-arides, à partir des données ASTER : cas du Moyen Atlas au Maroc

Université du Québec à Chicoutimi (2007)

Caractérisation spectrale de la dégradation des milieux naturels en régions semi-arides, à partir des données ASTER : cas du Moyen Atlas au Maroc

Bahri, El Mustapha

Titre : Caractérisation spectrale de la dégradation des milieux naturels en régions semi-arides, à partir des données ASTER : cas du Moyen Atlas au Maroc

Auteur : Bahri, El Mustapha

Université de soutenance : Université du Québec à Chicoutimi.

Grade : Maîtrise en ressources renouvelables (2007)

Résumé
The cedar forest of the Middle Atlas of Morocco is characterized by the heterogeneity of its stands and its fragmentation caused by the interaction among various factors, such as human activities, soil variability and climatic fluctuations. This results in a spatial and spectral heterogeneity that limits the reliability of the conventional methods used for classification of satellite imagery. To address this issue, the present study uses methods based on the spectral similarity to map major forest species of the cedar forest of Morocco : the linear spectral mixture analysis (LSMA) and the spectral angle mapper (SAM). The aim of the study was to compare : (i) methods used to extract spectral signature of pure pixels (endmembers) from the imagery, and (ii) the performances of LSMA and SAM in terms of appropriately mapping major forest species of the Middle Atlas. To achieve these goals, we used ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) images for the forest mapping. The results showed that SMA and SAM have led to similar patterns of the spatial distribution of studied forest species, but they have generated noticeable differences in the areas assigned to each mapped class. The classification results obtained by SMA and SAM were compared to those generated by the maximum likelihood classification method (our reference). This procedure showed that SMA has yielded a better classification of the dominant forest species than SAM ; this is illustrated by the value of Kappa Coefficient which was about 0.70 for SMA method and 0.66 for SAM approach. Keywords. Linear spectral mixture analysis ; spectral angle mapper ; pixel purity index ; iterative error analysis ; forest cover mapping ; ASTER.

Mots Clés : Télédétection—Maroc, Cartographie forestière—Maroc, THESE, CARTOGRAPHIE, FORESTIER, PEUPLEMENT, CEDRAIE, SIGNATURE, SPECTRAL, TELEDETECTION, ASTER, AMSL, SAM

Présentation

Version intégrale (2,74 Mb)

Page publiée le 30 août 2014, mise à jour le 7 septembre 2017