Developing a method to map coconut agrosystems from high-resolution satellite images

Résumé En

Our study aims at developing a generalizable method to exploit high resolution satellite images(VHR) for mapping coconut-based agrosystems, differentiating them from oil palm agrosystems.We compared two methods of land use classification. The first one is similar to that described byTeina (2009), based on spectral analysis and watershed segmentation, which we simplified byusing the NDVI vegetation index. The second one is the semi-automatic classification based ontexture analysis (PAPRI method of Borne, 1990). These methods were tested in two differentenvironments: Vanua Lava (Vanuatu; heterogeneous landscape, very ancient plantations) andIvory Coast (Marc Delorme Research Station, monoculture, regular spacing, oil palm plantations);and their results were evaluated against manually digitized photo-interpretation maps.In both situations, the PAPRI method produced better results than that of Teina (global kappa of0.60 vs. 0.40). Spectral signatures do not allow a sufficiently accurate mapping of coconut and donot differentiate it from oil palm, despite their different NDVI signatures. The PAPRI methoddifferentiates productive coconut from mixed plantations and other vegetation, either high or low(70% accuracy). In both situations, Teina’s method allows counting 65% of the coconut treeswhen they are well spaced. To increase the method accuracy, we suggest (1) field surveys (forsmall scale studies) and/or finer image resolution, allowing a high precision in manual mappingwith a better discrimination between coconut and oil palm, thus limiting the proportion of mixedpixels. (2) A phenological monitoring could improve the distinction between coconut and oil palmagrosystems. (3) Hyper-spectral images should allow extracting more precisely the respectivesignatures of both species. Another possibility would be (4) an object-oriented analysis asproposed by the eCognition software. Finally, (5) coupling the Lidar system with watershedanalysis would allow a better characterization of coconut varietal types.