|dc.description.abstracteng||A sustainable management of natural resources, in particular of forests, is of great importance to preserve the ecological, environmental and economic benefits of forests for future generations. An enhanced understanding of the current situation and ongoing trends of forests, e.g. through policy interventions, is crucial to managing the forest wisely. In this context, forest monitoring is essential for collecting the base data required and for observing trends. Despite the wide range of approved methods and techniques for both close-range and satellite-based remote sensing monitoring, ongoing forest monitoring research is still grappling with specific and unresolved questions: The data acquired must be more reliable, in particular over a long-term period; costs need to be reduced through advancements in both methods and technology that offer easier and more feasible ways of interpreting data.
This thesis comprises a number of focused studies, each with their individual and specific research questions, and aims to explore the benefits of innovative methods and technologies. The main emphasis of the studies presented is the integration of close-range and satellite-based remote sensing for enhancing the efficiency of forest monitoring.
Manuscript I discusses thermal canopy photography, a new field of application. This approach takes advantage of the large differences in temperature between sky and non-sky pixels and overcomes the inconsistencies of finding an optimal threshold. For an unambiguously separation of “sky” and “non-sky” pixels, a global threshold of 0 °C was defined. Currently, optical or hemispherical canopy photography is the most widely used method to extract crown-related variables. However, a number of aspects, such as exposure, illumination conditions, and threshold definition present a challenge in optical canopy photography and dramatically influence the result; consequently, a comparison of the results from optical canopy photography at a different point in time derived is not advisable. For forest monitoring, where repeated measurements of the canopy cover on the same plots were undertaken, it is therefore of utmost importance to devise a standard protocol to estimate changes in and compare the canopy covers. This paper offers such a protocol by introducing thermal canopy photography. A feasible and accurate method that examines the strong correlation (R2 = 0.96) of canopy closure values derived from thermal and optical image pairs.
Thermal photography, as a close-range remote sensing technique, also aids data collection and analysis in other contexts, for instance to expand our knowledge about bamboo tree species: Information about the maturity of bamboo culms is of utmost importance for managing bamboo stands because only then the process of lignification is finished and the culm is technically stronger and more resistant to insect and fungi attacks. The findings of a study (Manuscript III) conducted in Pereira, Colombia, show small differences in culm surface temperature between culms of different ages for the bamboo species Guadua angustifolia K., which may be a sign of maturity. The surface temperature of 12 culms was measured after sunrise using the thermal camera system FLIR 60Ebx. This study shows an innovative close-range remote sensing technique which may support researchers’ determination of the maturity of bamboo culms. This research is in its inception phase and our results are the first of this kind.
In the context of analyzing, in particular of thermal imagery time-series data, Manuscript (IV) offers a new methodology using advanced statistical methods. Otsu Thresholding, an automatic segmentation technique is used in a first processing step. O’Sullivan penalized splines estimated the temperature profile extracted from the canopy leaf temperature. A final comparison of the different profiles is done by constructing simultaneous confidence bands. The result shows an approximately significant difference in canopy leaf temperature. For this study, we successfully cooperated with the Center for Statistics at Göttingen University (Prof. Kneib).
The second close-range remote sensing technology employed in this thesis is terrestrial laser scanning which is used here to enhance our understanding about buttressed trees. Big trees with an irregular non-convex shape are important contributors to aboveground biomass in tropical forests, but an accurate estimation of their biomass is still a challenge and often remains biased. Allometric equations including tree diameter and height as predictors are currently used in tropical forests, but they are often not calibrated for such large and irregular trees where measuring the diameter is quite difficult. Against this background, Manuscript II shows the result of the 3D-analysis of 12 buttressed trees. This study was conducted in the Botanical Garden of Bogor, Indonesia, using a state-of-the-art terrestrial laser scanner. The findings allow for new insights into the irregular geometry of buttressed trees and the methodological approach employed in this paper will help to improve volume and biomass models for this kind of tree. The results suggest a strong relationship (R² = 0.87) between cross-sectional areas at diameter above buttress (DAB) height and the actual tree basal area measured at 1.3 m height. The accuracy of field biomass estimates is crucial if the data are used to calibrate models to predict the forest biomass on landscape level using remote sensing imagery.
The linkage between technology and methodology in the context of forest monitoring remote sensing enhance our knowledge in extracting more reliable information on tree cover estimation. The pre-processing of satellite images plays a crucial role in the processing workflow and particularly the illumination correction has a direct effect on the estimated tree cover. Manuscript IV evaluates four DEMs (Pleiades DSM, SRTM30, SRTM V4.1 and SRTM-X) that are available for the area of Shitai County (Anhui Province, Southeast China) for the purpose of an optimized illumination correction and tree cover estimation from optical RapidEye satellite images. The findings presented in this study suggest that the change in tree cover is contingent on the respective digital elevation models used for pre-processing the data. Imagery corrected with the freely available SRTM30 DEM with 30 m resolution leads to a higher accuracy in the estimation of tree cover based on the high-resolution and cost intensive Pleaides DEM.
These manuscripts eventually seek to resolve some of the issues and provide answers to some of the detailed questions that still persist at different steps of the forest monitoring process. In future, these new and innovate methods and technologies will maybe integrate into forest monitoring programs.||de