Acquisition and Characterization of Canopy Gap Patterns of Beech Forests
by Robert S. Nuske
Date of Examination:2019-09-20
Date of issue:2019-11-22
Advisor:Prof. Dr. Joachim Saborowski
Referee:Prof. Dr. Joachim Saborowski
Referee:Prof. Dr. Kerstin Wiegand
Files in this item
EnglishCanopy gap research in European beech-dominated forests has experienced a remarkable upswing in the last decades. It contributes to answering both fundamental ecological questions and to designing silvicultural interventions according to the widely accepted concept of close-to-nature forestry. Although close-to-nature forestry aims at mimicking the dynamics of unmanaged forests, reference values of canopy gaps in old-growth or even semi-natural forests are still scarce. Old-growth forest remnants are rare, and the mapping of canopy gaps is extremely labor-intensive. From a forest structure perspective, canopy gap patterns are mostly characterized by the gap fraction and the gap size distribution and sometimes by gap shape and gap age. Although frequently demanded, an adequate description of the spatial distribution of canopy gaps is often lacking.
The present thesis represents a step towards gaining the needed reference values by automatically mapping and analyzing canopy gap patterns. It contributes to the methodology of automated delineation of canopy gaps based on remote sensing data and suggests a method to describe the spatial distribution of gaps respecting their finite size and irregular shape.
Three different approaches to canopy gap mapping based on remote sensing data are presented. Canopy gaps were mapped with (i) an adaptive median in a moving window exclusively using digital aerial photogrammetry (DAP) height models, (ii) a data fusion approach employing a support vector machine (SVM) combining color, texture, DAP height and height quality information from aerial images, and (iii) a two-part relative height threshold using a standard airborne laser scanning (ALS) data product of the Hessian land surveying office.
All three canopy gap mapping approaches are fully automated and thus not influenced by per gap or per stand subjective judgments. Mapping canopy gaps based solely on either the color or height information obtained from aerial images did not provide satisfactory results. The data fusion approach using a SVM allowed for better canopy gap maps than height or color separately. ALS showed the capacity to map canopy gaps of all sizes reliably over large forest areas.
In order to obtain status-quo canopy gap maps for large areas, ALS data are very well suited. However, if the focus is on the dynamics of the canopy gaps, time series of aerial images are still the best data source. The most promising approach for automating canopy gap delineation based on aerial images are data fusion techniques combining many information layers, such as DAP heights, color and texture.
An adaptation of the pair-correlation function is suggested for analyzing the spatial distribution of canopy gaps. In contrast to conventional point pattern analysis, the adapted pair-correlation function represents objects by their outer boundary. The method was first implemented using the geodatabase PostGIS and later as the R package "apcf" in C++ using the libraries GEOS and GDAL directly.
The adapted pair-correlation function applied to 35 forest sites was able to describe a large diversity of spatial distributions of canopy gaps. However, more samples are needed to study the relationship between the spatial distribution of canopy gaps and for instance forest structure, time since abandonment and local disturbance regime. The second implementation of the adapted pair-correlation function considerably increased usability and performance, rendering it possible to analyze larger and more complex patterns and generate confidence envelopes with a higher confidence level.
It was shown that the adapted pair-correlation function, in contrast to other approaches, avoids pseudo hard- and soft-core effects, is able to describe the real interactions at small scales and the size of the effects are not weighted by the size of the objects. The adapted pair-correlation function proved to be a useful analytical tool for analyzing the spatial distribution of canopy gaps.
Keywords: canopy gap; European beech; Fagus sylvatica; digital aerial photogrammetry; airborne laser scanning; LiDAR; spatial statistics; point pattern analysis; pair correlation function