Type d'édition : Papier
Add to wishlist

Article: Observation of impacts of transfering the forest management to local communities using multitemporal Landsat TM satellite images: the case of Didy in Alaotra-Mangoro region, Madagascar


1 Laboratoire de Géophysique de l’Environnement et Télédétection, Institut et Observatoire de Géophysique d’Antananarivo (IOGA), Université d’Antananarivo
2 Projet de Conservation et de Gestion des écosystèmes forestiers malgaches (COGESFOR-SAHAN’ALA), Madagascar


This study is a part of a project for the management of natural resources from regions   of biodiversity hotspot in Madagascar. It is a collaborative work between the Institute and Observatory of Geophysics of Antananarivo who did the image processing and the COGESFOR project. The aim of this work is to produce multidate land cover maps using remotely sensed images for evaluating the impact of transfering forest management to local population program. The site study is Didy, an area in Alaotra-Mangoro region, in the Center-North part of Madagascar. Seventeen operations of community base managements (CBMs) were carried out in the Didy region since in 2003 and 2004. Theses CBMs cover about 32 % of surface area of our site study. To assess the actions taken in the management during the three years of study, we analyzed in first time the temporal evolution of the land cover between the two periods: 1994-2001 and 2011-2009. Then the following three major ecological indicators were examined: forested area, burned area and cultivated area. We have at our disposal for this study three Landsat TM (ref. 158-073) scenes acquired respectively on November 21, 1994; November 16, 2001 and November 30, 2009. Images have a spatial resolution of 30 m. The interest area has a dimension of 289266 ha. After field works on June and October 2011 and discussion with forestry specialists, eight land cover types are retained: forest, forest re-growth (‘savoka’), burn and slash (‘tavy’), swamp sedge (‘zetra’), rice field, savanna, lake/river and bare soil. The multispectral images were first atmospherically corrected, and then classified with the non parametric classifier Support Vector Machine (SVM) method. Confusion matrixes of all classifications show global accuracies of more than 93 %. Forest land cover type was always well classified (omission errors inferior to 1 %). Some confusions are however noticed between field rice, savannah and bare soil due to similar spectral signature of theses classes. All of these land cover types were dry in November, just before wet season. Results show in general positive impacts of the transfering forest management to local communities in the Didy region: (i) diminution of 5.8 % of forested area is observed between 1994 and 2001 but significant  augmentation of 2.5 % of these area is noted between 2001 and 2009 when forest management was transferred to local population; (ii) superficies of burned area (‘tavy’) are decreasing  of about 40 % and 24.2 % respectively during the first and second period of the study ; (iii) for the forest re-growth area (‘savoka’), it was decreasing of about 1 % between 1994 and 2001 and was increasing of about 69 % between 2001 and 2009; (iv) for the cultivated area, particularly the ‘rice field’ land cover type, it was increasing of 434 % between 1994 and 2001 and of 14.6 % between 2001 and 2009. In perspective, for improving our classification results by reducing confusion between some classes and increasing classification accuracies, we propose the use of high spatial resolution images acquired in different period of the year (wet and dry seasons).

KEYWORDS : monitoring, remote sensing, multitemporal analysis, community base management

30 Items

16 other products in the same category:

This website uses cookies to ensure you get the best experience on our website