Skip to Main content Skip to Navigation
Conference papers

Spatial structure analysis of palm grove landscape in Congo basin using Landsat images

Abstract : Congo basin is the world’s tropical forest outside of the Amazon. With an important biodiversity, this forest plays a central role in global warming. These forest divide up on six countries, one of which is Cameroon. Tropical Congo basin forests are a vital element in the economic development of these countries and are prone to spoilage. Agricultural and urban expansion is among the causes of forest cover degradation in this area (Geist and Lambin, 2001). Agricultural development induces urbanization processes (Gourade and Bruneau, 1983). Our work aims to study cover forest structure evolution linked to oil palm cultivation in Cameroon. The studied oil palm cultivation is the SOCAPALM industrial palm grove, located in Kienkié, South of Cameroun. We developed a method to describe and characterize processes and transformations that induce covert forest degradation. For that purpose, we studied the link between forest, urban frame and oil palm areas evolution. The first step is about mapping of land use from Landsat5 TM and Landsat8 OLI images, from 1988 to 2017. For Land use mapping, because of mixed pixels (Komba Mayossa and Gadal, 2016), usual methods of automatic and supervised classification didn’t allow a good discrimination of land cover features. Thus we use a method based on supervised classification of Landsat segment images extract. These method consists of: (i) spatial segmentation using thematic masks; (ii) after extraction, we classify each segment using maximum likelihood algorithm; (iii) classified segments are merged to obtain a final map land use in 1988 and 2017; (iv) map accuracy evaluation using confusion matrix method; (v) we extract forest, urban frame and oil palm areas to study their evolution between 1988 and 2017. The last step is to extract landscape structure using contextual operator such as edge analysis. For that, we compute Principal Component Analysis (PCA) from spectral layers of Landsat sensors. First components PC1988 and PC2017 containing respectively 90% and 85% of spectral information were selected. From these components we make directional filters in 3x3 sliding window pixels. In this way we extract olive Landscape structure. The experimental results of this study shows that cartographic method proposed allows a good mapping of land use features. That claim is justified by the good values of Kappa , 0.98 in 1988 and 0.87 in 207. The proposed method involves a logic allowing understanding the systemic link between trends in urban frame, palm grove areas and forest structure. The study of areas evolution shows a decrease of forest cover (from 60% in 1988 to 40% in 2017), on the benefit of palm grove (from 25% to 35% in 2017) and urban frame (from 15% from 20% in 2017). At the same time the analysis of landscape structure shows the appearance of new urban structures such as roads and buildings inside and near agricultural surface, decreasing forest cover.
Complete list of metadatas
Contributor : Sébastien Gadal <>
Submitted on : Saturday, June 30, 2018 - 9:26:03 AM
Last modification on : Tuesday, October 13, 2020 - 3:10:45 AM


  • HAL Id : hal-01826890, version 1


Prune Christobelle Komba Mayossa, Sébastien Gadal, Jean-Marc Roda. Spatial structure analysis of palm grove landscape in Congo basin using Landsat images. GEOBIA 2018: From pixels to ecosystems and global sustainability, Jun 2018, Montpellier, France. ⟨hal-01826890⟩



Record views