This study aims to present and evaluate a new method for measuring the distribution of snow within built-up environments by differencing elevations collected by an Airborne Laser Scanner (ALS) before, and during peak snow accumulation. Few efforts have been made to study the distribution of snow within built-up environments due to the false assumption that high-intensity rainfall is the main contributor to peak yearly runoff rates. Traditional techniques for measuring snow are often difficult to replicate in built-up environments due to incompatibility of methods and barriers such as buildings, roads and private property. Light Detection and Ranging (LiDAR) technology, specifically ALSs, have previously been used to characterize the distribution of snow under forest canopy, and in remote mountain environments. This study investigates and assesses the utility of high resolution, non-intrusive ALS data for estimating the depth and distribution of snow within the town of Uxbridge, Ontario. ALS flights for this study were completed before the onset of snow accumulation, as well as near peak snow accumulation for the winters of 2010 and 2011. Pre and post snow accumulation ALS measured elevations were differenced to estimate the depth of the snowpack across the entire study area at a resolution of 0.5 m. Ground measurements of snow depth were also completed within 24 hours of each of the winter flights. The LiDAR-estimated and ground-measured snow depths were compared using Spearman's rank correlation coefficient as well as Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE). Results from this thesis show that: 1) Snow depths estimated by differencing elevations from two ALS flights show a MAE of 3 cm and an RMSE of 10 cm when compared to ground-measured snow depths. (2) There is a strong, statistically significant relationship (ρ = 0:82, p < 0:001) between LiDAR-estimated and ground-measured snow depths. (3) An average bias of -3 cm was found for the entire dataset showing an underestimation in the LiDAR-estimated snow depths most likely caused by the effects of low lying vegetation on the fall ALS measurements. The results presented in this study demonstrate that ALSs are capable of providing high spatial resolution snow depth estimates within built-up environments. Furthermore, snow depth measurements made using an ALS can be used to increase the current body of knowledge on the distribution and re-distribution of snow within built-up environments. Snow distributions measured by an ALS could also be used for future development and verification of urban hydrological models.

Submitted by Fernanda Rodrigues de Lima (fernanda.rlima@ufpe.br) on 2018-09-26T21:08:19Z No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) DISSERTAÇÃO Rafael Lopes Mendonça.pdf: 5746497 bytes, checksum: 6427f5355ad301a77ff31c0401d54da5 (MD5) / Approved for entry into archive by Alice Araujo (alice.caraujo@ufpe.br) on 2018-10-01T15:23:00Z (GMT) No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) DISSERTAÇÃO Rafael Lopes Mendonça.pdf: 5746497 bytes, checksum: 6427f5355ad301a77ff31c0401d54da5 (MD5) / Made available in DSpace on 2018-10-01T15:23:00Z (GMT). No. of bitstreams: 2 license_rdf: 811 bytes, checksum: e39d27027a6cc9cb039ad269a5db8e34 (MD5) DISSERTAÇÃO Rafael Lopes Mendonça.pdf: 5746497 bytes, checksum: 6427f5355ad301a77ff31c0401d54da5 (MD5) Previous issue date: 2017-08-31 / CNPq / Com o crescente uso do escaneamento a LASER aerotransportado para aquisição de Modelo Digital, tem-se observado um maior interesse em desenvolver técnicas de filtragens capazes de gerar um Modelo Digital do Terreno (MDT) de forma automática. Os métodos existentes na literatura consideram regiões mais uniformes, com características diferentes das encontradas nas cidades brasileiras, principalmente em áreas onde há uma maior variação do relevo e diversidade de elementos, tanto naturais quanto artificiais, sobre o solo exposto. Diante disto, o presente trabalho tem como objetivo analisar algumas técnicas existentes, e propor uma abordagem para a filtragem de dados Light Detection And Ranging (LIDAR) para geração de MDT. A área de estudo corresponde a uma região do bairro do Vasco da Gama, na cidade de Recife, capital Pernambucana. Esta região apresenta um relevo acidentado com grande concentração de imóveis, além da presença de vegetação arbórea e rasteira. Através de alguns métodos de filtragens foram analisados os resultados obtidos para essa área de estudo. A combinação dos métodos de identificação de bordas e da classificação supervisionada apresentou resultado mais satisfatório, em comparação aos outros utilizados nesta dissertação. / With the increasing use of the airborne LASER scanning to acquire Digital Model, it has been observed a greater interest in developing filtering techniques capable of generating a Digital Terrain Model (DTM) automatically. The existing methods in the literature consider more uniform regions, with characteristics different from those found in Brazilian cities, especially in areas where there is a greater variation of the relief and diversity of elements, both natural and artificial, above the ground. Therefore, the present work aims to analyze some existing techniques, and to propose an approach to Light Detection and Ranging (LIDAR) data filtering for MDT generation. The study area corresponds to a region of the neighborhood of Vasco da Gama, in the city of Recife, capital of Pernambuco. This region presents a rough relief with great urban agglomeration, besides the presence of arboreal and low vegetation. Through some filtration methods the results obtained for this area of study were analyzed. The combination of the methods of edge identification and supervised classification presented a more satisfactory result, compared to the others used in this dissertation.

Light detection is one of the crucial abilities of all animals. The light cues are important e.g. for maintaining of circadian rhythms, regulation of spawning cycles, changes of pigmentation and arguably most importantly for vision. Most animals detect light by opsins, members of the G protein coupled receptors superfamily. Amphioxus belongs to earliest branching chordate clade, cephalochordates. Thanks to their phylogenetic position, physiology and morphology, cephalochordates became the most relevant model organism for understanding the evolutionary origins of vertebrate specific traits. Amphioxus evince various reactions to light throughout its development. In the presented thesis light detecting systems of amphioxus were studied thoroughly. More specifically characterization of the opsin gene repertoire of two amphioxus species Branchiostoma floridae and Branchiostoma lanceolatum and their comparison with opsins from other animals is presented. In addition, remarkable similarity on the gene expression level between one of amphioxus visual organs, so called frontal eye, and neurons and retinal pigmented epithelium in vertebrate retina was shown. These data confirm the long time ago proposed homology between amphioxus frontal eye and vertebrate lateral eyes. Taken together all the presented data...

Weak optical signals have to be measured in different fields of sciences including chemistry and biology. For example, very low levels of fluorescence emission should be detected from the spots on a DNA microarray that correspond to weakly expressed genes. High sensitivity charge-coupled devices (CCDs) are used in these applications. CCDs require special fabrication and are difficult to integrate with other circuits. CMOS is the technology used for fabrication of CPUs and other widely used digital components. CMOS is not optimized for light detection. CMOS circuits are however cheap, low power and can integrate several components. Active pixel sensor (APS) is the most common pixel structure for CMOS photodetector arrays. In this work we provide an accurate analysis of the APS signal using new models for the capacitance of the photodiode. We also provide a complete noise analysis of the pixel to calculate the SNR of the pixel and provide optimum operation points. We propose a new mode of operation for APS that can achieve at least l 0 dB higher SNR, than conventional APS, at light levels of less than 1 μW/cm^2. We fabricated several APS pixels in CMOS 0.18 μm technology and measured them to confirm the proposed analyzes. There are applications like fluorescence lifetime imaging that require both sensitivity and fast response. Photomultiplier tubes (PMTs) are commonly used in these applications to detect single photons in pico- to nano-seconds regime. PMTs are bulky and require high voltage levels. Avalanche photodiodes (APDs) are the semiconductor equivalent of PMTs. We have fabricated different APDs along with different peripheral circuitries in CMOS 0.18 μm technology. Our APDs have a 5.5 percent peak probability of detection of a photon at an excess bias of 2 V, and a 30 ns dead time, which is less than the previously reported results. The low price of CMOS makes modem diagnosis devices more available. The low power of CMOS leads to battery-driven hand-held imaging solutions, and its high integration leads to miniaturized imaging and diagnosis systems. A low-light-level CMOS imager paves the way for the future generation of biomedical diagnosis solutions. / Thesis / Doctor of Philosophy (PhD)

Flood inundation modelling is highly dependent on an accurate representation of floodplain topography. These  remotely  sensed  accurate  data  are  often  not  available  or  expensive,  especially  in  developing countries. As an alternative, freely available Digital Elevation Models (DEMs), such as the near-global Shuttle Radar Topography Mission (SRTM) data, have come into the focus of flood modellers. To what extent  these  low-resolution  data  can  be  exploited  for  hydraulic  modelling  is  still  an  open  research question. This benchmarking study investigated the potentials and limitations of the SRTM data set for flood inundation  modelling  on  the  example  of  the  Papaloapan  River,  Mexico.  Furthermore  the  effects  of vegetation signal removal from the SRTM DEM as in Baugh et al. (2010) were tested. A reference model based on a light detection and ranging (LiDAR) DEM was set up with the model code LISFLOOD-FP and run for two flood events. Test models based on SRTM DEMs were run and output flood extents compared to the reference model by applying a measure of fit. This measure of fit, which was based on binary wet/dry maps of both model outputs, gave information on how well the test models simulated the flood inundation extents compared to the reference model by giving a percentage of the model performance from theoretically 0 to 100 %. SRTM-based models could not reproduce the promising results of previous studies. Flood extents were mostly underestimated and commonly flooded areas were almost exclusively made up out of the main channel surface. One of the reasons for this likely was the much steeper slope of the SRTM DEM as opposed to the LiDAR DEM where water probably was conducted much faster though the main channel. Too high bank cells as well as generally more pronounced elevation differences of the SRTM DEM throughout the whole floodplain were another problem of the SRTM DEM preventing accurate flood inundation simulations. Vegetation  signal  removal  was  successful  to  a  certain  degree  improving  the  fit  by  about  10 %. However, a realistic shape of flood extent could not be simulated due to too big pixel sizes of the used canopy  height  data  set. Also,  the  conditioned  models  overestimated  flooded  areas  with  increasing vegetation signal removal, rendering some of the models useless for comparison, as water leaving the model domain could not be accounted for in the measure of fit. This study showed the limitations of SRTM data for flood inundation modeling where an accurate approximation of the river slope as well as accurately captured bank cells and floodplain topography are crucial for the simulated outcome. Vegetation signal removal has been shown to be potentially useful but should rather be applied on more densely covered catchments. / Översvämningar skapar stora problem världen över och fler och fler människor lever i områden som är utsatta för risk för att svämmas över. Dessutom förväntas översvämningar förekomma mer frekvent i många delar av världen i framtiden på grund av klimatförändringar. Skada orsakad av översvämningar kan  överstiga  flera  miljarder  US$.  Men  översvämningar  orsakar  också  andra  problem,  förutom ekonomiska förluster. De senaste 10 åren har mer än 60 000 människor dött på grund av översvämningar. Ytterligare 900 000 000 människor har drabbats på något sätt. Därför är det viktigt att man vet vilka områden som är utsatta för hög risk. Ett av de verktyg för att avgöra  översvämningsrisker  är  hydrauliska  datormodeller  som  försöker  förutse  hur  en  bestämd översvämning breder ut sig. Modellerna är baserade på fysiska principer och topografisk information. Helst vill man ha topografisk information med hög kvalitet och upplösning. Ofta har man data från fjärranalyser, insamlade från flygplan. Ett exempel på det är LiDAR-data som är baserad på laser. Dock är det ofta dyrt eller inte tillgängligt med LiDAR i avlägsna områden och utvecklingsländer, där man behöver sådan data som mest. Därför har forskare försökt att använda globalt tillgängliga topografiska data av låg kvalitet för hydrauliska modeller. En sådan datauppsättning är det så kallade SRTM-datat från amerikanska NASA. SRTM samlas in med hjälp av radarstrålar från satelliter. I flera studier har man fått goda resultat inom översvämningsmodellering med SRTM. Dock måste man testa det vidare för fler avrinningsområden. I den här studien har man försökt att använda SRTM i en hydraulisk modell för den mexikanska floden  Papaloapan.  För  att  se  hur  bra  (eller  dålig)  SRTM-modellen  är  för  att  simulera  hur  en översvämning sprids har man jämfört den med en modell baserad på högkvalitativ LiDAR-data. Båda modellerna  simulerade  samma  översvämningar. Topografiska  information  från  SRTM-data  är  oftast inkorrekt där det finns väldigt tät och hög vegetation, eftersom radarsignalen då inte räcker till marken och den uppskattade höjden är därför för hög i sådana områden. Av denna anledning ville man därför i denna  studie  även  testa  hur  resultatet  av  SRTM-modellen  skulle  förbättras  om  man  tog  bort  viss vegetation. Dessvärre var den utformade SRTM-modellen inte så bra för det här fallstudieområdet och SRTM-modellen  förutspådde  mycket  mindre  översvämningar  än  den  förmodade  mer  korrekta  LiDAR-modellen. Då vegetation avlägsnandes kunde man förbättra SRTM-modellen till viss mån, men det var fortfarande  inte  tillräckligt  för  det  här  området.  Denna  studie  visar  att  det  är  viktigt  att  fortsätta undersöka hur passande och användbart SRTM är, eftersom det har visat sig att SRTM inte är lämpligt för att förutspå översvämningar i alla delar av världen.

The perception tasks in autonomous driving, namely 3D object detection and map segmentation, play a crucial role in enabling vehicles to perceive the surrounding environment. The traditional approach is to have a single network for each task and complete all tasks in a sequential manner. However, this method suffers from repeated feature extraction and error propagation, leading to inefficiency and reduced accuracy. While multi-task learning can eliminate redundant computations and facilitate information exchange among tasks, improving efficiency and overall system performance, it can also lead to a reduction in the performance of a particular task, compared to single-task training due to gradient dominance. To tackle this problem, this thesis aims to bridge the performance gap between multi-task and single-task learning. We first utilize the GradNorm method to dynamically readjust the loss weights while training. We further add individual encoders to allow fine-grained feature learning for each task. Based on the existing perception network, we adapt our dynamic loss strategy and new encoder architecture, which shows that our results match or even surpass the performance of each task in a multitask setting, compared to the single task. We also evaluate the computational efficiency of our method, further demonstrating the advantages of multi-task learning in the autonomous driving domain where real-time computing is non-negotiable. / Perception inom autonom körning, det vill säga 3D objektsdetektion och map segmentation, spelar en avgörande roll för att möjliggöra att fordon uppfattar den omgivande miljön. Traditionellt sett utförs detta av en grupp nätverk, en för varje uppgift, där uppgifterna utförs sekventiellt. Detta skapar problem i form av upprepad extrahering av egenskaper i datan och felfortplantning vilket leder till försämrad beräkningshastighet och resultat. Multi-task learning kan eliminera onödiga beräkningar och möjliggöra utbyte av information mellan uppgifter, vilket medför förbättringar inom effektivitet och systemets generella prestanda i relation till single-task learning. Däremot kan det leda till försämrade resultat i enskilda uppgifter på grund av gradient dominans. Denna avhandling syftar till att bemöta detta problem genom att överbrygga gapet mellan multi-task och single-task learning. Vi använder oss av GradNorm metoden för att dynamiskt justera gradienternas magnitud under träning. Enskilda encoders lades till i vardera nätverk för att möjliggöra fine-grained feature learning för varje uppgift. GradNorm och ytterligare encoders applicerades på det befintliga perception-nätverket, vilket gav resultat som är lika bra eller bättre vid multi-task learning som single-task learning för varje uppgift. Även effektivitet vid beräkning utvärderades, vilket ytterligare visade fördelarna av att använda multi-task learning i autonom körning där beräkning i realtid är av högsta prioritet.

No âmbito das mudanças climáticas globais, a quantificação do estoque de carbono em povoamentos florestais tem recebido mais atenção, principalmente pelo fato das florestas exercerem um papel fundamental no equilíbrio do estoque de carbono global. Com o objetivo de contribuir para esse processo, a parte investigativa deste trabalho foi desenvolvida em duas etapas. A primeira etapa objetivou o ajuste de modelos alométricos para a estimativa do estoque de carbono presente na biomassa total (Ctotal), no lenho comercial (Cleco) e parte residual (Crds) (casca, folhas e galhos) em plantações de Eucalyptus spp. em nível de árvores, através de uma amostragem destrutivas de árvores, análise elementar do carbono em laboratório e medidas convencionais de inventário. A medição do diâmetro à altura do peito (DAP) e altura total das árvores em parcelas amostradas instaladas nos talhões onde as árvores foram coletadas para determinação direta de carbono também foi realizada. A segunda parte, consistiu na avaliação do uso da tecnologia LiDAR (Light Detection and Ranging) aerotransportada (Airborne LASER scanning) como uma alternativa eficiente e versátil para a estimativa do estoque de carbono total (Ctotal), no lenho comercial de toras (Cleco) e no resíduo da árvore (Crds) em nível de parcelas em plantações de Eucalyptus spp usando como base o estoque de carbono estimado na primeira fase. Os resultados obtidos encontram-se resumidos em dois artigos científicos. O primeiro artigo mostra que os modelos baseados no logaritmo do diâmetro à altura do peito (DAP) e da altura total da árvore (Ht) oferecem boas precisão e exatidão para estimar o estoque de carbono em nível de árvore. O segundo artigo, permitiu a determinação das melhores métricas LiDAR para o cálculo do teor total de carbono, tanto no carbono total, lenho comercial e nas partes residuais da árvore em nível de plantação. Esses resultados, bem como os indicadores estatísticos utilizados para avaliar a qualidade dos ajustes, são o cerne desta dissertação. / In the context of global climate change, the quantification of the carbon content in forest plantations have received great attention. This is because vegetation play an important role in the global carbon budget. This master thesis was developed in two main parts. The first part was to adjust allometric equations for the estimation of the carbon content at a tree level. This was performed for the above ground section (Ctotal), in commercial logs (Cleco) and residuals ( Crds) (e.g. bark, leaves and branches/twigs). The experiment was based on the destructive model of individual trees harvested in commercial plantations of Eucaliptus spp. The experiment encompasses both forest inventory and laboratory analyses procedures. Additionally, in-situ measurements such as the diameter at the breast height (DBH) and the total tree height were also performed. These sample plots were located in homogeneous forest units and close to the areas were the trees have been harvested. The second part of this master thesis was the evaluation of the airborne LiDAR technology as a tool for the retrieval of the above ground biomass (Ctotal), the carbon present in the commercial logs (Cleco) and residuals. This procedure was performed at the sample plots level. This procedure was based on the information provided in the first part. The results are presented as two scientific manuscripts. The first manuscript shows that allometric equations based on the log of the variables diameter at the breast height (DBH) and total tree height (Ht) were good predictors for the retrieval of the total carbon content at a tree level. The second manuscript allow the selection of the best LiDAR derived metrics for the retrieval of the total carbon content, either at above ground level, commercial logs and residual parts of the tree at a sample plots level. These results, as well as the statistical indicators for the adjustment of several statistical models is the core of this master thesis.

No âmbito das mudanças climáticas globais, a quantificação do estoque de carbono em povoamentos florestais tem recebido mais atenção, principalmente pelo fato das florestas exercerem um papel fundamental no equilíbrio do estoque de carbono global. Com o objetivo de contribuir para esse processo, a parte investigativa deste trabalho foi desenvolvida em duas etapas. A primeira etapa objetivou o ajuste de modelos alométricos para a estimativa do estoque de carbono presente na biomassa total (Ctotal), no lenho comercial (Cleco) e parte residual (Crds) (casca, folhas e galhos) em plantações de Eucalyptus spp. em nível de árvores, através de uma amostragem destrutivas de árvores, análise elementar do carbono em laboratório e medidas convencionais de inventário. A medição do diâmetro à altura do peito (DAP) e altura total das árvores em parcelas amostradas instaladas nos talhões onde as árvores foram coletadas para determinação direta de carbono também foi realizada. A segunda parte, consistiu na avaliação do uso da tecnologia LiDAR (Light Detection and Ranging) aerotransportada (Airborne LASER scanning) como uma alternativa eficiente e versátil para a estimativa do estoque de carbono total (Ctotal), no lenho comercial de toras (Cleco) e no resíduo da árvore (Crds) em nível de parcelas em plantações de Eucalyptus spp usando como base o estoque de carbono estimado na primeira fase. Os resultados obtidos encontram-se resumidos em dois artigos científicos. O primeiro artigo mostra que os modelos baseados no logaritmo do diâmetro à altura do peito (DAP) e da altura total da árvore (Ht) oferecem boas precisão e exatidão para estimar o estoque de carbono em nível de árvore. O segundo artigo, permitiu a determinação das melhores métricas LiDAR para o cálculo do teor total de carbono, tanto no carbono total, lenho comercial e nas partes residuais da árvore em nível de plantação. Esses resultados, bem como os indicadores estatísticos utilizados para avaliar a qualidade dos ajustes, são o cerne desta dissertação. / In the context of global climate change, the quantification of the carbon content in forest plantations have received great attention. This is because vegetation play an important role in the global carbon budget. This master thesis was developed in two main parts. The first part was to adjust allometric equations for the estimation of the carbon content at a tree level. This was performed for the above ground section (Ctotal), in commercial logs (Cleco) and residuals ( Crds) (e.g. bark, leaves and branches/twigs). The experiment was based on the destructive model of individual trees harvested in commercial plantations of Eucaliptus spp. The experiment encompasses both forest inventory and laboratory analyses procedures. Additionally, in-situ measurements such as the diameter at the breast height (DBH) and the total tree height were also performed. These sample plots were located in homogeneous forest units and close to the areas were the trees have been harvested. The second part of this master thesis was the evaluation of the airborne LiDAR technology as a tool for the retrieval of the above ground biomass (Ctotal), the carbon present in the commercial logs (Cleco) and residuals. This procedure was performed at the sample plots level. This procedure was based on the information provided in the first part. The results are presented as two scientific manuscripts. The first manuscript shows that allometric equations based on the log of the variables diameter at the breast height (DBH) and total tree height (Ht) were good predictors for the retrieval of the total carbon content at a tree level. The second manuscript allow the selection of the best LiDAR derived metrics for the retrieval of the total carbon content, either at above ground level, commercial logs and residual parts of the tree at a sample plots level. These results, as well as the statistical indicators for the adjustment of several statistical models is the core of this master thesis.

Multi-spectral Satellite imagery, among remotely sensed data from airborne and spaceborne platforms, contained the NIR band information is the major source for the land- cover classification. The main purpose of aerial imagery is for thematic land-use/land-cover mapping which is rarely used for land cover classification. Recently, the newly developed digital aerial cameras containing NIR band with up to 10cm ultra high resolution makes the land-cover classification using aerial imagery possible. However, because the urban ground objects are so complex, multi-spectral imagery is still not sufficient for urban classification. Problems include the difficulty in discriminating between trees and grass, the misclassification of buildings due to diverse roof compositions and shadow effects, and the misclassification of cars on roads. Recently, aerial LiDAR (ULiUght UDUetection UAUnd URUanging) data have been integrated with remotely sensed data to obtain better classification results. The LiDAR-derived normalized digital surface models (nDSMs) calculated by subtracting digital elevation models (DEMs) from digital surface models (DSMs) becomes an important factor for urban classification. This study proposed an adaptive raw-data-based, surface-based LiDAR data-filtering algorithm to generate DEMs as the foundation of generating the nDSMs. According to the experiment results, the proposed adaptive LiDAR data-filtering algorithm not only successfully filters out ground objects in urban, forest, and mixed land cover areas but also derives DEMs within the LiDAR data measuring accuracy based on the absolute and relative accuracy evaluation experiments results. For the aerial imagery urban classification, this study first conducted maximum likelihood classification (MLC) experiments to identify features suitable for urban classification using LiDAR data and aerial imagery. The addition of LiDAR height data improved the overall accuracy by up to 28 and 18%, respectively, compared to cases with only red¡Vgreen¡Vblue (RGB) and multi-spectral imagery. It concludes that the urban classification is highly dependent on LiDAR height rather than on NIR imagery. To further improve classification, this study proposes a knowledge-based classification system (KBCS) that includes a three-level height, ¡§asphalt road, vegetation, and non-vegetation¡¨ (A¡VV¡VN) classification model, rule-based scheme and knowledge-based correction (KBC). The proposed KBCS improved overall accuracy by 12 and 7% compared to maximum likelihood and object-based classification, respectively. The classification results have superior visual interpretability compared to the MLC classified image. Moreover, the visual details in the KBCS are superior to those of the OBC without involving a selection procedure for optimal segmentation parameters.

Die letzten zwanzig Jahre haben gezeigt, dass die Integration luftgestützter Lasertechnologien (Light Detection and Ranging; LiDAR) in die Erfassung von Waldressourcen dazu beitragen kann, die Genauigkeit von Schätzungen zu erhöhen. Um diese zu ermöglichen, müssen Feldaten mit LiDAR-Daten kombiniert werden. Diverse Techniken der Modellierung bieten die Möglichkeit, diese Verbindung statistisch zu beschreiben. Während die Wahl der Methode in der Regel nur geringen Einfluss auf Punktschätzer hat, liefert sie unterschiedliche Schätzungen der Genauigkeit. In der vorliegenden Studie wurde der Einfluss verschiedener Modellierungstechniken und Variablenauswahl auf die Genauigkeit von Schätzungen untersucht. Der Schwerpunkt der Arbeit liegt hierbei auf LiDAR Anwendungen im Rahmen von Waldinventuren. Die Methoden der Variablenauswahl, welche in dieser Studie berücksichtigt wurden, waren das Akaike Informationskriterium (AIC), das korrigierte Akaike Informationskriterium (AICc), und das bayesianische (oder Schwarz) Informationskriterium. Zudem wurden Variablen anhand der Konditionsnummer und des Varianzinflationsfaktors ausgewählt. Weitere Methoden, die in dieser Studie Berücksichtigung fanden, umfassen Ridge Regression, der least absolute shrinkage and selection operator (Lasso), und der Random Forest Algorithmus. Die Methoden der schrittweisen Variablenauswahl wurden sowohl im Rahmen der Modell-assistierten als auch der Modell-basierten Inferenz untersucht. Die übrigen Methoden wurden nur im Rahmen der Modell-assistierten Inferenz untersucht. In einer umfangreichen Simulationsstudie wurden die Einflüsse der Art der Modellierungsmethode und Art der Variablenauswahl auf die Genauigkeit der Schätzung von Populationsparametern (oberirdische Biomasse in Megagramm pro Hektar) ermittelt. Hierzu wurden fünf unterschiedliche Populationen genutzt. Drei künstliche Populationen wurden simuliert, zwei weitere basierten auf in Kanada und Norwegen erhobenen Waldinveturdaten. Canonical vine copulas wurden genutzt um synthetische Populationen aus diesen Waldinventurdaten zu generieren. Aus den Populationen wurden wiederholt einfache Zufallsstichproben gezogen und für jede Stichprobe wurden der Mittelwert und die Genauigkeit der Mittelwertschätzung geschäzt. Während für das Modell-basierte Verfahren nur ein Varianzschätzer untersucht wurde, wurden für den Modell-assistierten Ansatz drei unterschiedliche Schätzer untersucht. Die Ergebnisse der Simulationsstudie zeigten, dass das einfache Anwenden von schrittweisen Methoden zur Variablenauswahl generell zur Überschätzung der Genauigkeiten in LiDAR unterstützten Waldinventuren führt. Die verzerrte Schätzung der Genauigkeiten war vor allem für kleine Stichproben (n = 40 und n = 50) von Bedeutung. Für Stichproben von größerem Umfang (n = 400), war die Überschätzung der Genauigkeit vernachlässigbar. Gute Ergebnisse, im Hinblick auf Deckungsraten und empirischem Standardfehler, zeigten Ridge Regression, Lasso und der Random Forest Algorithmus. Aus den Ergebnissen dieser Studie kann abgeleitet werden, dass die zuletzt genannten Methoden in zukünftige LiDAR unterstützten Waldinventuren Berücksichtigung finden sollten.