Search

The European Commission funded Tropical Ecosystem Environment Observation by Satellite (TREES) project aims to assess the forest cover and its changes within the pan- tropical belt. Statistics for the monitoring target periods 1990, 2000 and 2010 are derived from analyzing a systematic sample set of satellite imagery mainly from the Landsat TM and ETM+ sensors. To reach full coverage of all sample sites for the year 2010 epoch and to ensure data continuity for possible future epochs the usage of alternative satellite data is inevitable. In this study the replacement of Landsat data with ALOS AVNIR-2 data was tested. Adaptations of the existing processing chain are presented as well as the results of an accuracy assessment. The results show that AVNIR-2 data are generally suitable to substitute Landsat data in this context. [ABSTRACT FROM PUBLISHER]

Forest monitoring from earth observation is crucial over tropical regions to assess forest extent and provide up-to-date estimates of deforestation rates. Based on a systematic sample of 20x20 km size sites, a processing chain has been developed at the European Commission's Joint Research Centre (JRC) for producing deforestation estimates between years 1990, 2000 and 2005. Whereas this monitoring exercise was based on Landsat imagery, limitations in Landsat availability over Central Africa for year 2010 required alternative imagery such as the Disaster Monitoring Constellation (DMC). The classification module of the existing JRC processing chain is based on tasseled caps analysis (TCap-based). We adapted this module to DMC imagery by selecting the most suitable object-features through their assessments using a sub-sample of existing land-cover maps of years 1990 and 2000. The processing chain is adapted for the production of land-cover change maps between years 2000 and 2010. The accuracy of the land-cover maps produced for year 2010 with the two methods (original TCap-based and adapted Multi-Sensor) is assessed through a reference dataset. Overall accuracies are similar for both approaches (93% and 95% respectively), but the Multi-Sensor approach shows a significant improvement when considering only changed objects (83% overall accuracy versus 56% for TCap-based). Our results show that, even by using DMC imagery with lower radiometric quality (compared to Landsat) an automated classification can provide land-cover maps with similar accuracy thanks to an appropriate object-features selection. Similar adaptations need to be developed for other satellite imagery such as SPOT and RapidEye. [ABSTRACT FROM PUBLISHER]

The TREES-3 project of the European Commission's Joint Research Centre is producing estimates of tropical forest cover changes for two time periods: 1990–2000 and 2000–2010. This paper presents the method developed for the automatic change detection and classification of year 2010 imagery integrating the existing segmentation and classification results of the period 1990–2000. The year 2010 imagery is processed in three automatic steps: segmentation, change detection and object spectral classification. The validated maps of forest cover for the years 1990 and 2000 are integrated as thematic input layer into the image segmentation and classification process for the year 2010 data. An object-based change detection technique is applied using Tasseled Cap components and spectral Euclidean distances. Objects detected as changed are classified in two steps: parametric classification based on membership functions and change vector analysis for the remaining unclassified objects. All objects identified as ‘unchanged’ are used as training areas for parametric classification and spectral signatures are extracted from 2010 imagery. The change vectors are defined according to the validated land cover classification of the year 2000. The segmentation approach was tested on 568 sample units spread over Brazil. The segmentation results for year 2010 demonstrated consistency with the segmentation of imagery for the period 1990–2000. The resulting overall accuracy of the automatic classification was calculated for the 281 sample units of the Brazilian Amazon biome and for 201 sample units of three more complex Brazilian biomes, Caatinga, Cerrado and Pantanal, at 92% and 91% respectively. [ABSTRACT FROM PUBLISHER]

The Global Rain Forest Mapping (GRFM) radar mosaics, generated from L-band Japanese Earth Resources Satellite 1 imagery downsampled to 100-m pixel size, provide a two-season spatially continuous coverage of the humid tropical ecosystems of the world. This paper presents a novel classification approach suitable for regional-scale vegetation mapping using the GRFM datasets. The mapping system consists of: 1) an application-dependent wavelet-based edge-preserving smoothing algorithm and 2) a two-stage per-pixel hybrid learning nearest multiple-prototype (NMP) classifier, whose. unsupervised first stage is a per-pixel near-optimal vector quantizer, called enhanced Linde-Buzo-Gray (ELBG), recently proposed in pattern recognition literature. Identified as ENMP (NMP with ELBG), this novel classification approach is compared against two alternative systems in the classification of forest cover disturbances located across an area in the Amazon Basin. Surface classes of interest are primary forest, degraded forest, nonforest, and water bodies. Reference maps, derived from 30-m resolution Landsat Thematic Mapper imagery, are provided by the National Aeronautics and Space Administration and the Food and Agriculture Organization of the United Nations. Abundant quantitative and qualitative evidence shows that: 1) in a forest/nonforest data-mapping task, ENMP provides a testing accuracy of 87%, in line with training accuracies, i.e., the proposed method seems capable of generalizing well over the GRFM South America dataset and 2) among three competing approaches, ENMP provides the best compromise between ease of use, mapping accuracy, and computational time. Starting from these results, ENMJ is employed to generate a swamp forest map of the whole Amazon Basin from the two-season GRFM radar mosaic of South America, in the context of the Global Land Cover project (GLC 2000). [ABSTRACT FROM AUTHOR]

Reports the application of contextual clustering of image labeling to degraded forest assessment in Landsat Thematic Mapper images of Brazilian Amazon. Analysis of remote sensed images; Cause of intermediate states of forest alterations; Detection on the deforestation mapping program of the Food and Agriculture Organization.