High-resolution digital surface models (DSMs) for modelling fractional shrub/tree cover in a mire environment

This paper describes the development of a fractional shrub/tree cover in open mire land using logistic regression and airborne remote sensing data (the DSM is derived from colour infrared images). The present study was carried out in the framework of the Swiss Mire Protection Program, where shrub encroachment is a key issue. An example of the use of this modelling approach in a mire biotope in the Pre-alpine zone of central Switzerland is presented. As a first step, a DSM was automatically generated using an image matching approach from 12 colour infrared (CIR) aerial images. Two discrete forest masks with different levels of detail were then generated combining a canopy cover derived from the DSM with a multi-resolution segmentation and a fuzzy classification. Then, fractional shrub/tree covers for open mire land were calculated using six explanatory variables derived from the DSM only and the forest masks as the response variable. Validation with field samples revealed highest accuracies for the cover stratum 30-100%. An advantage of this approach for future mire protection work is that small shrubs and single trees of open mire land can be modelled with high accuracy and the estimated cover is not restricted to a simple forest/non-forest decision. The optimal, most reliable cover stratum can be achieved by individually tuning the probability threshold of the fractional shrub/tree covers according to the individual vegetation characteristics of a mire ecosystem. This study clearly revealed that 3-dimensional information, as obtained by digital photogrammetry, is indispensable for modelling tree and shrub occurrence.