General flowering only occurs in the dipterocarp forests in south-eastern Asia and for some species, these events are the main or even the only reproduction mechanism. These community-wide flowering events occur irregularly in both space and time. Main triggers are believed to be extreme weather events such as temperature drops, and droughts often associated with the El Niño Southern Oscillation (ENSO). For a good understanding of triggering mechanisms and ecological implications, research areas need to cover broad spatial extents, as facilitated by remote sensing. However, the detection of flowering is non-trivial and it is not known on which scale and with which spectral characteristics flowering canopies can be distinguished from non-flowering tree canopies. In this study, we aimed at detection of such events starting from phenocam and in-situ observations in Lambir Hill National Park, Malaysian Borneo. We studied several optical indices to find the spectral signature of flowering or the following mast fruiting of dipterocarp trees. Artificial up-scaling of segments on phenocam images provides a measure of distinguishability in spatial terms and information about the spatial scale at which the flowering can be observed. This is critical information for up-scaling the phenocam results to currently available satellite remote sensing imagery.
Results/Conclusions
On tree level and smaller, tree canopies with yellowish flowers, reddish seeds, or brownish fruits showed higher than usual Red Chromatic Coordinates, which means that the ratio of the red spectral band to the sum of red, green and blue increased. Especially the spatial variability over all segments within an image was significantly higher during and after the flowering period, indicating that the spatial heterogeneity within the canopy increased during flowering events. This effect became stronger throughout the general flowering period and peaked just after, i.e. in the fruiting period. The time lag depended on the intensity of the flowering period. With larger segments, the homogeneity within the segments, as well as the heterogeneity between the segments, decreased. This suggests that satellite data with high spatial resolution, i.e. sufficiently high for distinguishing single trees, is promising but that the signal is not strong enough for moderate resolution imagery. The step towards spatially determining flowering extents using high-resolution satellite remote sensing is feasible and valuable for ecosystem management purposes.