Lake Maggiore - Italy, 3-6 April 2000
VEGETATION data for regional forest cover mapping of Southeast Asia
Stibig, R. Beuchle, V. Gond
Mapping of forest cover of Southeast Asia from satellite images is hampered by several factors characteristic of the region. Most significant is the persistent cloud cover over insular Southeast Asia where all satellite image acquisitions are contaminated by cloud and haze. For most of continental Southeast Asia good satellite images can be obtained during the dry season. However, in this part of the region the monsoon regime causes high variation in vegetation phenology. Land use patterns are small and intermingled (e.g. shifting cultivation) and particularly in the deciduous forest domain transitions from forest to woodland are gradual. In addition, in the mainly mountainous area the impact of shadow on the spectral reflectance is considerable.
The VEGETATION instrument of SPOT 4 can offer a valuable tool for forest cover mapping at regional scale. The improved spatial resolution and geometric quality of the new generation of coarse resolution satellite images make a more distinct mapping of forest cover feasible. The daily coverage of the whole continent allows an image selection according to best quality, lowest cloud cover and the optimal phenological stage of vegetation cover. Generating mosaics from series of images could provide a possibility for coping with persistent cloud cover.
VEGETATION 10-day standard composites (S10 products - with pixel extraction based on the maximum NDVI value) were tested for regional mapping of forest cover. For continental Southeast Asia twelve S10 mosaics were extracted during the dry season from December to March for the years 1999 and 2000. For the insular part of Southeast Asia 42 composites were acquired from May to November 1998 and 1999.
A further mosaicing procedure was applied for each sub-regional set of images, extracting pixels from the S10 composites by the minimum value in the short wave infrared band. Such selection would tend to give preference to cloud free or less cloud contaminated pixels and would reduce the artificial mosaicing effects in the original S10 products. Looking at vegetation phenology ‘green’ stages of vegetation would be chosen rather than leafless stages.
The mosaics and first classification results obtained open up some promising perspectives when based on an optimal selection of the seasonal and phenological ‘time window’.
VEGETATION 10-day-composites appear to offer a possibility for a rapid assessment and first level mapping of forest cover in Southeast Asia at continental scale. Such products are useful not only for producing a regional overview of forest cover, but also for stratification purposes and monitoring of large-scale forest conversion and trends of change.
Limitations to be expected are due to the nature of the data, the scale of the exercise and the characteristics of vegetation cover and land use in the region. Overestimation of forest cover can be expected due to shadow effects in mountain areas, the inclusion of woodland formations within deciduous forest cover and of small-scale plantations within evergreen forest cover. On the other hand underestimation may occur in areas of extensive shifting cultivation by missing small forest lots.