HR: 09:30h
AN: B31E-07 INVITED     [Abstracts]
TI: Influence of Rainfall Variability on Canopy Clustering in a Savanna Ecosystem
AU: * Scanlon, T M
EM: tms2v@virginia.edu
AF: University of Virginia, Department of Environmental Sciences Clark Hall, Charlottesville, VA 22904 United States
AB: Vegetation spatial pattern is indicative of dominant processes, and in water-limited ecosystems pattern formation could be strongly influenced by spatial interactions that are driven by fluctuating rainfall on interannual timescales. In the Kalahari region of southern Africa, vegetation pattern along a north-south aridity gradient of ~900-200 mm/yr of wet season rainfall suggests that water availability provides both global and local controls on the distribution of trees. Field surveys and satellite data indicate that tree cover density is linearly related to wet season rainfall, while analysis of high-resolution IKONOS data demonstrates that tree canopy clustering conforms to a power law distributions in the form of P( A≥a) α a^{- b}, where P( A≥a) is the cumulative probability of canopy cluster size greater than a, and b is a constant. The global and local controls on this apparent hydrologically-driven self-organization of the tree canopies are analogous to the controls on spin distributions in the Ising model of ferromagnetism. This cellular automata framework is adopted to evaluate the spatial patterns that emerge from such interactions. The ubiquity of the observed power-law clustering of the savanna tree canopies indicates that the Ising model parameters (i.e. local interaction strength and strength of the external magnet) should not require fine tuning, but rather be robust with respect to producing power-law distributions. Spatial interaction scales that consider only the immediate four neighbors (von Neumann neighborhood) were found to be ineffective in reproducing power-law clustering across the rainfall gradient, so local interactions in the Ising model were modified to be weighted according to a Pareto distribution as a function of distance. Simulations showed that fluctuating rainfall (external magnet strength) is effective in producing power-law distributions of tree canopy (spin) clusters along the rainfall gradient.
DE: 0466 Modeling
DE: 0480 Remote sensing
DE: 1813 Eco-hydrology
DE: 1854 Precipitation (3354)
SC: Biogeosciences [B]
MN: Fall Meeting 2005