Spatial Pattern Analysis Applied to Plant Ecology

Abstract

The spatial structure of a forest stand is an important signature of forest dynamics. The local environment determines competition among the trees, growth, death and regeneration. Therefore, the spatial configuration of the individual trees in a forest stand can provide information about the underlying ecological processes at the site.This thesis focuses on the determination and analysis of the local constellation of individual trees (characterized by species, age, size etc.) in a forest stand. A large number of spatial statistical methods can be applied to this spatial pattern analysis. Besides using some classical statistical methods, I propose two new methods for spatial analysis.In Chapter 1, I review some classical spatial statistical methods currently applied to plant ecology. These methods are suitable for the analysis of spatial configurations of individuals, such as plants or trees. Additionally, I present some explicit formulas for an area based edge effect correction method and compare this method with the Ripley edge correction factor.Chapter 2 shows an application of the methods presented in Chapter 1 to a data set obtained from two real mangrove forests located in the North of Brazil. The idea behind is to infer about the underlying ecological processes occurring in those sites from the spatial configuration of individual trees.In Chapter 3, I develop a new method for the spatial analysis of objects. The method approximates each individual tree as a circle, instead of a point, thus minimizing the bias of the classical methods.In Chapter 4, I finally propose a new method to obtain spatial scale information about the point processes occurring within a study site. This approach combines Multiresolution Decomposition Analysis through Wavelet Transform with the Kernel Density Estimation method. This methodology provides information not only about the type of spatial distributions but also about their particular locations. By this, this approach goes beyond classical point-pattern methods and is innovative.