Maximally selected two-sample statistics as a new tool for the identification and assessment of habitat factors with an application to breeding-bird communities in oak forests

Abstract

Near-natural forest management is one of the major principles of modern forestry in central Europe. A measure commonly used for the assessment of logging is the occurrence and density of character species typical for the local biotope. Near-natural forest structures can be designed only when sufficient knowledge about the relationship between species and habitat factors is available. We suggest the use of simple cutpoint regression models for the derivation of statistically valid rules of thumb. The methodology is applied to bird and structure data of ancient and present oak coppices with standards to demonstrate its practical application. The cutpoints are derived from estimates of break points by means of maximally selected two-sample statistics, and their validity is judged by multiple test procedures. Detailed results are presented for character species and their relationships to the environment. We show that, for example, the tree pipit (Anthus trivialis) is missing in stands with more than 40% canopy overstorey, whereas this species can be regularly observed in more open stands. For the tree pipit, the analysis is illustrated by means of a free software package.