Multiscale GEOBIA for the Quantification of Forest Cover Loss in Blue Nile Region, Sudan
El-Abbas, Mustafa; Csaplovics, Elmar

Natural forest cover loss in the region is threatening the sustainability of ecosystem functionality as vast forest areas are rapidly destroyed mainly as a result of mechanized rain-fed agriculture. Despite the importance of forest in fragile ecosystems, the spatiotemporal distribution of Gross Forest Cover Loss (GFCL) across the region is not well quantified. Therefore, efforts were carried out using GEO-Object-Based Image Analysis (GEOBIA) to provide a comprehensive knowledge about the GFCL, integrating rates and spatial distribution. Great attention has been given to its application in recent years, while the scale issue, as described here by segmentation-levels, seems to be widely ignored. Optical multi-spectral satellite imagery of 1990 and 2010 acquired from TERRA ASTER and LANDSAT sensors were used. The method adopted in this research consists in cross operation of classified images at multi-scale levels. A new objects domain was created representing the GFCL of each pair of classified maps as well as the overlapped areas. The quantitative result of the change detection reveals that the forest cover has been dramatically shrink and destroyed. GFCL is defined as the area of forest cover removed as a result of any disturbance, including both natural and human factors. The result achieved form GEOBIA reveals an intensive clearance of natural forest cover. GFCL was estimated to be 36585.11 ha representing 32.72% of the total area. The occupancy of the mechanized rain-fed agriculture expanded over the natural forest land was estimated to be 23636.19 ha, which represents 35.53% of the total gained area on that class during the period of study. While the recovered area represents only 7.88%, mainly from abandon agricultural field. It must be considered that most of the stable area observed is land managed under clear property rights. To conclude, the present study exhibits a great potential for accurate change detection, when utilizing GEOBIA technique with optical multispectral satellite imagery. The estimation of GFCL consequently requires larger observation meaningful segments, rather than discrete pixels. Furthermore, the proven capability of the adopted method for gaining knowledge of the change dynamics and it is driving forces is also shown.