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

Accueil du site → Master → Pays Bas → 2004 → Land cover change detection in the Limpopo river basin, Mozambique

International Institute for Geo-Information Science and Earth Observation (ITC) 2004

Land cover change detection in the Limpopo river basin, Mozambique

de Sousa Rodrigues Pereira, M.C.

Titre : Land cover change detection in the Limpopo river basin, Mozambique

Auteur : de Sousa Rodrigues Pereira, M.C.

Etablissement de soutenance : University of Twente International Institute for Geo-Information Science and Earth Observation (ITC)

Grade : Master of Science in Geo-Information Science and Earth Observation 2004

Résumé
The study area of the current thesis is located in the Limpopo river basin, Mozambique. The work here developed addresses the problem of knowing if there were land cover changes in that area, in the period of 1999 – 2003, and if those changes are related with the exceptional floods that occurred in the year 2000. The general research questions considered to tackle this problem were : Can land cover change in the study area be assessed with the satellite imagery available ? What were those changes in land cover ? Is there a relation between (eventual) land cover changes and the 2000 floods ? The methodology for detecting land cover changes was based on the comparison between satellite images taken on different dates, using the postclassification operation. The relation between those changes and the floods was focused on the correlation between areas of change and no change and the maximum inundated area. From the imagery available, the Landsat ETM+ acquired on 22 August of 1999 was used to characterize the land cover in the same year ; the ASTER image taken on 14 June 2003 was used to characterize the current land cover status . The processing of the primary data col- lected during the fieldwork allowed the definition of land cover classes based on vegetation structure ; those classes were used for the classification of the images and generation of the correspondent land cover maps. To prepare the images for the change detection procedure, several image processing op- erations were applied, namely the radiometric and atmospheric normalization of the Landsat image to the ASTER one. The classification of the images was primarily based on the supervised method but other tools were also used, namely with ancillary data ; the evaluation of the supervised classification results was done through accuracy assessment and visual analysis (in this last case only for the AS- TER image). Throughout the work, numerous GIS operations were applied. In the change detection analysis two perspectives were considered : (1) areas of change versus areas where no changes occurred ; (2) transition between certain land cover classes. In terms of results, the changes observed occupied a bigger percentage of the study site than the areas of no change ; on the other hand, the most important directions of change were the transition from natural or semi-natural classes to crop fields and vice-versa. As a result of the work developed in this thesis, it was concluded that the land cover change observed in the study area between 1999 and 2003, as it was assessed, has no correlation with the maximum inundated area of the 2000 floods. Nevertheless, there may have been several factors affecting the change detection procedure, other than real change ; it is likely that those factors have introduced errors, thus affecting the results both of the change analysis and of its relation with the floods. From those factors, the following ones are suspected to be the most relevant : Seasonal differences related with plant phenology and crop calendars ; Factors related with the image classification procedures ; Difference in pixel size between the two sensors/platforms used. The difference in pixel size between the two sensors/platforms used in this study and the conse- quences it is likely to have had in terms of image processing and the comparison between the images in the change detection analysis, lead to the conclusion that it would have been strongly preferable to use the same or similar sensors

Version intégrale (ITC )

Page publiée le 5 avril 2018