Geostatistical analysis of tree size distributions in the Southern Kalahari, obtained from remotely sensed data - art. no. 67420G

We propose using geostatistical methods for the spatial analysis of data pertaining to the size of trees (in terms of canopy surface area) obtained by means of remote sensing methods. Geostatistical methods are suitable because the locations of the trees are at the nodes of an unstructured grid. More specifically, we present a semivariogram analysis to detect correlations in the tree size spatial distribution, and we apply a novel method of anisotropy analysis to search for possible anisotropy in the size distribution. We use a combination of aerial photographs and satellite images in four snapshots covering 37 years to investigate the temporal behavior in addition to the spatial distribution at a single time. The aerial photographs were taken in 1964, 1984, 1993, and the IKONOS satellite image in 2001. We follow a study area covering over 139 ha and over 2,000 tree individuals. Our plots are located in the Southern Kalahari savanna near the city of Kimberley, South Africa.