Characterization of the alternative to slash-and-burn benchmark research area representing the Congolese rainforests of Africa using near-real-time SPOT HRV data

P. S. Thenkabail

DOI: https://doi.org/10.1080/014311699212966

IF: 3.531

1999-01-01

International Journal of Remote Sensing

Abstract:This study used four near-real-time multispectral Systeme pour l'Observation de la Terre (SPOT) high-resolution visible (HRV ) images to establish land cover and forest classes of relevance to slash-and-burn agriculture. The study was conducted in 1.43 million ha of the Alternative to Slash-and-Burn (ASB) global benchmark research area for Africa selected as representative of the entire Congolese basin of Central Africa and located in Southern Cameroon. The land cover and forest classes mapped have different impacts on global phenomena as well as their own management challenges and include slash-and-burn dominant farmlands, C hromolenea orodata dominant short-fallows, Imperata cylindrica dominant weeds, long-fallows or regenerated forests, raphia palm-dominant lowlands, permanently flooded swamp forests, and primary and secondary forests. In order to map these distinct and complex classes the study proposes and implements a piecemeal approach to classification involving stratification of the image into several distinct discrete subsets of forest corridors, lowlands, and uplands. The approach involves the use of image texture indicators in conjunction with groundtruth data to divide the image into discrete subsets, performing unsupervised classification on these discrete subsets, masking problem classes and reclassifying them, editing certain spectrally inseparable areas, adopting post-classification strategies, and finally mosaicking the discrete subsets into one seamless image. This approach led to an overall mapping accuracy of 82.4% with individual classes mapped at accuracies above 72.9% (user's), and above 66.7% (producer's).

imaging science & photographic technology,remote sensing