Abstract
We modeled the distribution of the habitat of Lagerstroemia subcostata var. fauriei (LSF), a rare riparian species which is indigenous to Yakushima Island, southern Japan, by analysis of GIS-derived topographic factors. We measured the distribution of LSF at sample sites (ca 15 ha in total) located within two catchments (60 ha in total) on the western side of Yakushima Island. Based on this data, we modeled the distribution of tree density of LSF in 12.5m×12.5m grid cells covering the entire area of the two catchments (11280 cells in total) by using Poisson loglinear modeling of GIS-derived topographic factors. The modeling was performed by progressively adding four different levels of topographic factors as follows: Level-1, topographic attributes of the target cell, such as elevation or slope inclination; Level-2, cell position relative to the adjacent channel; Level-3, attributes of the adjacent channel; and Level-4, catchment-scale attributes. The AIC factor of the models obtained decreased with the addition of attribute data from the Level-1 to Level-4 factors. This indicated that the topographic attributes of the target cell (Level-1) alone are not sufficient to describe suitable habitat for LSF, and that channel characteristics (Level-2 and 3 factors) are also useful. Moreover, the models adopting Level-4 factors showed the best performance, indicating the importance of catchment characteristics, such as the existence of sediment sources or seed sources, in explaining the actual distribution of LSF.
