Utilizing Local Variance of Simulated High Spatial Resolution Imagery To Predict Spatial Pattern of Forest Stands

Abstract

Spatial pattern, defined as the distribution of individuals in space, is an important characteristic of forest stands. It provides an insight into the allocation of above- and below-ground resources to a tree, as well as reflecting the stand history, microclimate, and competition between different species over time. The spatial arrangement of trees can be described as random, aggregated, or regular with a number of statistics existing that characterize the spatial pattern of a given tree population. High spatial resolution remote sensing is an obvious tool to facilitate the measuring and monitoring of spatial patterns in forest stands. Remote sensing imagery provides detailed information about forest structure while still allowing large areas of forest to be mapped and monitored. The increased availability of high resolution imagery coupled with improvements in scene processing and interpretation techniques allow additional information, such as texture, to be extracted from this type of imagery. In this paper the spatial pattern of trees within a forest stand is related to high spatial resolution imagery. This relationship is developed using a technique that relates scene texture variance to a statistic describing spatial pattern. The technique was tested on a number of simulated remote sensing scenes by systematically varying the size and spatial distribution of trees using a geometric-optical model. Results indicate that it is theoretically possible to derive the spatial pattern of trees within a high spatial resolution forested scene provided crown size is estimated a priori. It is also likely that the total projected foliage cover of the canopy will affect the ability to predict spatial distribution based on texture variance. It was concluded that the spatial pattern of trees within a scene can play a vital role in the amount and degree of variation existing within imagery. It is important to consider the implications of these type of relationships when developing variance-based models of forest structure.