In tropical rainforests, the sustainability of selective logging is closely linked to the extent of collateral stand damage. The capacity to measure the extent of such damage is essential for calculating carbon emissions due to forest degradation under the Reducing Emissions from Deforestation
and Forest Degradation (REDD+) process. The use of remote sensing to detect canopy gaps in tropical rainforests is an attractive alternative to ground surveys, which are laborious and imprecise. In French Guiana, the detection of logging-related gaps using very high spatial resolution optical
satellite images produced by the Système Pour l'Observation de la Terre (SPOT) 5 sensor is carried out by Office National des Forêts (ONF) (French National Forestry Agency). Gaps are detected using a segmentation method based on computer-assisted photointerpretation. Detection
has been automated to improve and accelerate the process. We developed an automatic method, which involves estimating segmentation thresholds using a statistical approach. The principle of the method presented in this article is to model the forest's spectral signature by using a Gaussian
distribution and calculate a divergence between that theoretical signature and the image histogram in order to detect gaps that constitute a reduction of forest cover. The segmentation threshold between gap and forest is thus no longer defined in the original radiometric area but as a discrepancy
between theoretical distribution and histogram. Computing the divergence to define the threshold made it possible to efficiently automate the detection of all gaps and skid trails with a surface area greater than 100 m2. The proportion of misclassified points measured during field
surveys is 12%, which is a high level of precision. The proportion of misclassified points obtained is 12%. This tool could be used to assess the quality of logging operations or biomass loss in other areas where the forest is undergoing deterioration while still remaining predominant in the
No Reference information available - sign in for access.
No Citation information available - sign in for access.
No Supplementary Data.
Document Type: Research Article
Office National des Forêts (ONF), Réserve de Montabo, 97307,Cayenne, French Guiana
NEVANTROPIC, SAS, 97300,Cayenne, French Guiana
CIRAD-ES, UPR 105 Forest Services, 34398,Montpellier Cedex, France
Publication date: 2013-01-20
More about this publication?