Leaf-Shedding Phenology in Lowland Tropical Seasonal Forests of Cambodia as Estimated From NOAA Satellite Images

The seasonality of forest leaf area index and photosynthetic activity is an important element of ecosystem function. The ecohydrological models representing water and carbon fluxes in tropical forests need regional-scale spatial information on forest leaf phenology. We aimed to evaluate the utility of the National Oceanic and Atmospheric Administration (NOAA)/Advanced Very high resolution radiometer (AVHRR) imagery for monitoring evergreen and deciduous forest phenology in lowland Cambodia. The temporal sequences of the normalized difference vegetation index (NDVI) data were constructed from ten-day composite imagery acquired from May 2001 to April 2002. A local maximum fitting (LMF) technique that combined time-series filtering and local curve-fitting techniques was used to extract cloud-free data. With this approach, different phenology between evergreen and deciduous forests was clearly observed during the dry season: Deciduous forests exhibited marked and spatially uniform loss and (later) gain in new leaf area, whereas evergreen forests showed lesser and spatially/temporally heterogeneous changes in leaf phenology. A significant finding was that about 30% of evergreen forests shed detectable quantities of leaves at two times during a dry season (i.e., in the early, and late, dry season). In summary, the capability of composite NOAA/AVHRR NDVI data sets processed with the LMF technique was established as a viable approach for monitoring heterogeneous forest phenology at regional scales during the tropical dry season, although it is apparent that further improvements in cloud clearing are required if the approach is to be useful also in the wet season.