Reik Leiterer

A voxel-based approach for canopy structure characterization using full-waveform airborne laser scanning

Forests play a significant role in the global biogeochemical and -physical cycles and particularly the complex three-dimensional forest canopy structure influences the fluxes of energy and matter between the atmosphere and forests. Assessing this structure quantitatively using conventional fieldwork or traditional remote sensing methods is difficult, whereas airborne laser scanning (ALS) systems have proven to be suitable for providing explicit vertical information for large areas. However, most existing ALS based approaches include manual processing steps or need additional data about stand characteristics. To solve these issues, a robust and automatic multi-dimensional clustering method was developed to derive forest canopy structure types (CSTs) based on full-waveform ALS data. The results show that it is possible to develop an automatic, self-sustained and transferable method for: the extraction of CSTs without any previous knowledge about the forest stand; and the extraction of bio-physical parameters based on the resulting CSTs.

Operational forest structure monitoring using airborne laser scanning Flugzeuggestütztes Laserscanning für ein operationelles Waldstrukturmonitoring

Die Struktur des Waldes hat einen signifikanten Einfluss auf die globalen bio- geochemischen Stoffkreislaufe und kann daruber hinaus als Indikator dienen, um das Potential zum Erhalt der Biodiversitat abzuschatzen und die Widerstandsfahigkeit des Waldes gegen ausere Einflusse zu bestimmen. Flugzeuggestutztes Laserscanning (ALS) bietet hierbei die Moglichkeit einer raumlich hochaufgelosten Erfassung und Beschreibung sowohl der horizontalen als auch der vertikalen Waldstruktur. Wir stellen robuste Verfahren basierend auf flugzeuggestutzten Laserscanningdaten vor, um eine Extraktion von forstwirtschaftlich und -wissenschaftlich relevanten Strukturinformationen zu ermoglichen. Dies beinhaltet: i) die Einzelbaumextraktion, ii) die Bestimmung von Unterwuchs und Bodenbedeckung und iii) die Totholzerkennung. Die Datengrundlage bestand aus multi-temporalen, full-waveform Laserdaten in dichtem Laub- und Mischwald fur Testgebiete in der Schweiz (Lagern) und in Deutschland (Uckermark). Basierend auf der ALS-Punktwolke mit ihren geometrischen Attributen und den zugehorigen full-waveform Eigenschaften wurden folgende Methoden angewendet: i) hierarchisches, 3D-Clustering und die Ableitung von alpha shapes fur die Einzelbaumextraktion, ii) rasterbasierte, vertikale Stratifizierung fur die Charakterisierung von Unterwuchs, und iii) die Kombination aus map algebra und Vektorisierung fur die Totholzanalyse. Die erzielten Genauigkeiten der abgeleiteten Strukturvariablen entsprachen den Anforderungen der traditionellen Forstinventur. Vorbehaltlich der Verfugbarkeit einer entsprechenden Datengrundlage (multi-temporale ALS-Daten mit hohen Punktdichten) ist es mit den vorgestellten robusten Methoden moglich, ein grosflachiges und operationelles Waldstrukturmonitoring durchzufuhren.

Fusing imaging spectrometry and airborne laser scanning data for tree species discrimination

Accurate mapping of tree species composition within forest ecosystems is an important aspect of management planning and monitoring. Passive optical remote sensing in general and imaging spectroscopy (IS) in particular have played an important role in producing such maps, but are suffering from issues due to vegetation structure. On the other hand, the structural information provided by airborne laser scanning (ALS) was shown to be helpful for species discrimination, particularly in heterogeneous forests. In this paper, we investigate the potential of product-level fusion of IS and ALS to provide a better tree species differentiation based on their complementarity. Our results show that the fused tree species map does improve on the single system maps and more accurately provides the distribution and fraction of each tree species within the study area.

Optimal structural and spectral features for tree species classification using combined airborne laser scanning and hyperspectral data

In this paper, the performance of feature selection in tree species classification based on multi source earth observation data was studied. We applied a sequential forward floating feature selection on imaging spectroscopy (IS) and airborne laser scanning (ALS) data, as well as their combination. Qualitative comparison of the fused results shows that the selected spectral features are more distributed across the spectrum, in contrast to an accumulation of features in the near infrared region when using IS alone. A support vector machine (SVM) classifier was used for quantitative comparison of the different datasets. Assessing the classification accuracies confirmed the superiority of the selected subset of spectral and structural features compared to using all available features (improvement of > 7% in kappa accuracy).

Retrieval of canopy structure types for forest characterization using multi-temporal airborne laser scanning

We present a method to characterize forest canopy structure in space and time based on vertical echo distributions from airborne laser scanning (ALS). We developed a transferable, grid-based method using ALS data combined with an automatic determination of the best feasible spatial unit for canopy structure characterization. We derive canopy structure types (CSTs) using a hierarchical, multi-scale classification approach based on Bayesian robust mixture models (BRMMs), which satisfy structurally homogenous criteria without the use of in-situ calibration information. The validation shows promising results for the CSTs, particularly in terms of seasonal and horizontal variations in vertical canopy structure. We conclude that our method can improve the robustness and reliability of canopy structure characterization. Future work will include tests of transferability to a larger variety of forests and extensive testing using CSTs as a structural classification scheme.