No description available.

No description available.

The PointNet architecture is a foundational deep learning model for 3D point clouds, solving classification and segmentation tasks. We hypothesize that the full potential of PointNet has not been reached and is greatly restrained by a single Max pooling layer. First, this thesis introduces new and more complex learnable aggregation functions. Secondly, a novel normalization technique, Context Normalization, is proposed for further feature extraction. Context Normalization is similar to Batch Normalization but independently normalizes each point cloud within a mini-batch and always uses dynamic statistics. The experiments show that replacing Max pooling with Principal Neighborhood Aggregation (PNA) increased classification accuracy from 73.3% to 78.7% on an SO(3) augmented version of the ModelNet40 dataset. Combining PNA with Context Normalization further increased accuracy to 84.6%.

Identifying the dynamical processes that are active in tropical cirrus clouds is important for understanding the role of cirrus in the tropical atmosphere. This study focuses on analyzing turbulent motions inside tropical anvil cirrus with the use of a Cloud Resolving Model. Convection in the transition from shallow to deep convection has been simulated with Colorado State University Large Eddy Simulator/Cloud Resolving Model System for Atmospheric Model (SAM 6.3) in a high resolution three-dimensional simulation and anvil cirrus formed in the end of this simulation has been analyzed. For model set up, data gathered during the Tropical Rainfall Measuring Mission Large-Scale Biosphere-Atmosphere (TRMM LBA) field experiment in Amazonas, Brazil have been used as large scale forcing. 31 anvil clouds have been localized from a single time step of the simulation, “a snapshot”, of the entire simulated cloud field consisting of convective clouds of different scales and subsequently divided into three categories that represent different stages of the anvil lifetime; growing, mature and dissipating anvil stages. The classification is based on in-cloud properties such as cloud condensate content and vertical velocities. The simulated anvils have been analyzed both individually and as groups to examine the transition from isotropic three-dimensional turbulence in the convective core of the thunderstorm to stratified two-dimensional turbulence in the anvil outflow. A dimensionless number F is derived and used as a measure of the “isotropic” behavior of the turbulence inside the cloud. F is expressed as the ratio between the horizontal part of TKE and the total (horizontal + vertical) Experiments show that SAM 6.3 clearly can resolve turbulent structures and that the transition from isotropic three-dimensional turbulence to stratified two-dimensional turbulence occurs in the middle layers of the mature and dissipating anvil stages. / Sammanfattning av ”Studier av turbulenta rörelser i städmoln med hjälp av numeriska simuleringar av tropisk konvektion”  Städmoln i tropikerna har stor inverkan på strålningsballansen på grund av de är så vanligt förekommande och att de ligger på hög höjd i atmosfären. Att förstå de drivande krafterna som är aktiva i skapandet och underhållandet av städmoln är viktiga för att få en bra bild av rollen städmoln spelar i den tropiska atmosfären. Den här uppsatsen fokuserar på att studera turbulenta rörelser inuti tropiska städmoln med hjälp av en molnmodell. Tropisk konvektion har simulerats med Colorado State University’s molnmodell SAM 6.3 i en högupplöst tredimensionell simulering. Data från en ”ögonblicksbild” av det simulerade molnfältet har analyserats och 31 städmoln har valts ut och studerats vidare. De simulerade städmolnen indelades i tre olika kategorier baserat på utvecklingsstadier; växande städmoln, moget städmoln och skingrade städmoln. Stadieklassificeringen bestämdes beroende på isvatteninnehåll och vertikalhastigheter i molnet. Städmolnen har därefter analyserats både individuellt och som grupper för att lokalisera och analysera övergången från tredimensionell isotropisk turbulens i kärnan av Cb-molnet till tvådimensionell stratifierad turbulens i städmolnet. För att initiera simuleringen användes mätdata insamlade under fältexperimentet TRMM LBA (Tropical Rainfall Measuring Mission Large-Scale Biosphere-Atmosphere) i Amazonas, Brasilien. För att beskriva turbulenta rörelser i molnen togs det dimensionslösa talet 𝐹 fram som ett mått på isotropin. 𝐹 uttrycks som kvoten mellan den horisontella delen av TKE och den totala (horisontell och vertikal). Den här studien visar att den undersökta molnmodellen SAM 6.3 klart kan simulera turbulenta i rörelser i övergången mellan isotropisk till horisontell turbulens i olika stadier av städmolnens livscykel. Mina analyser visar att övergången sker främst i de mellersta skikten av de mogna och skingrade stadierna av städmolnets utveckling.

O objetivo deste trabalho foi estudar o papel dos aerossóis em modificar o desenvolvimento das nuvens e da precipitação. Esta tem sido uma das mais intrigantes questões no estudo das mudanças climáticas. Medidas da concentração de núcleos de condensação de nuvens (CCN) e distribuições de gotículas de nuvem durante o Experimento de Grande Escala da Biosfera-Atmosfera na Amazônia (LBA) revelaram características distintas entre condições atmosféricas limpas e poluídas. As medidas foram conduzidas no Sudoeste da Região Amazônica durante os meses de setembro e outubro de 2002, focando a transição do final da estação seca para o início da estação chuvosa. Durante a transição, a análise da concentração de CCN dentro da camada limite revelou um decréscimo geral, de valores acima de 1200 cm-3 para menos de 300 cm-3. A comparação entre áreas limpas e poluídas mostrou concentrações de CCN cerca de 5 vezes maiores em áreas poluídas. As diferenças não foram tão grandes nos níveis acima da camada limite. As medidas também mostraram um ciclo diurno acompanhando a atividade de queima de biomassa. Distribuições de tamanho de gotículas medidas em duas regiões com concentrações de aerossóis extremamente diferentes foram analisadas. Em condições poluídas pela queima de biomassa foi encontrada alta concentração de gotículas, com diâmetro médio e conteúdo de água de nuvem aumentando muito pouco com a altura, em comparação com regiões limpas. A função gama foi usada para ajustar as distribuições de gotículas e o parâmetro de forma da função foi usado como critério para definir adequadamente a melhor representação das distribuições de gotículas. De acordo com os valores encontrados, distribuições gama estreitas (parâmetro de forma em torno de 5) são mais bem indicadas para representar condições poluídas enquanto que aquelas mais largas se ajustam melhor em condições limpas (parâmetro de forma em torno de 2). Com base nesses resultados, experimentos numéricos foram conduzidos com o Brazilian Regional Atmospheric Modeling System (BRAMS) para investigar os efeitos da concentração de CCN e do parâmetro de forma das distribuições de gotículas no desenvolvimento da precipitação em nuvens convectivas tropicais. Os resultados mostraram uma grande sensibilidade devido às mudanças nesses parâmetros. Altas concentrações de CCN e distribuições de gotículas estreitas (parâmetros de forma maiores), típicas de dias poluídos, produziram baixos valores médios para água líquida integrada na coluna e precipitação acumulada na superfície. Por outro lado, tendência oposta a este efeito foi encontrada em condições limpas (baixos valores para ambos, a concentração de CCN e o parâmetro de forma). O parâmetro de forma se mostrou ser mais importante que a concentração de CCN. Os efeitos da concentração de CCN e do parâmetro de forma também influenciaram a distribuição espacial dos campos de nuvem e precipitação. Embora o valor médio desses campos tenha diminuído em condições poluídas, o valor máximo aumentou. Como conseqüência da menor dispersão nas nuvens em condições poluídas, mais radiação solar esteve disponível na superfície. Isto é oposto aos resultados dos modelos globais que mostram redução na radiação solar como conseqüência do segundo efeito indireto dos aerossóis. Da mesma forma, este estudo encontrou que as diferenças são reduzidas quando é incluído o efeito direto dos aerossóis em absorver ou refletir a radiação solar. Sobretudo, os resultados sugerem que um maior número de modelos com tratamento explícito dos processos microfísicos de nuvem são necessários. Esses modelos permitem comparações, podendo mostrar o melhor tratamento numérico a ser usado na representação dos efeitos dos aerossóis no processo de precipitação como um todo. Estes resultados são importantes porque melhoram a compreensão de como o clima será afetado como conseqüência das mudanças futuras. / The objective of this work was to study the role of aerosols in modifying clouds and precipitation. This is one of the most difficult aspects in the study of climate changes. Field measurements of cloud condensation nuclei (CCN) and cloud size distributions performed during the Large Scale Biosphere-Atmosphere Experiment in Amazonia (LBA) campaign revealed distinct characteristics between clean and polluted atmospheric conditions. Measurements were conducted over the southwestern Amazon region during September-October 2002 focusing the transition from dry to wet seasons. During this period, analysis of CCN concentrations in the boundary layer revealed a general decreasing trend from mean values higher than 1200 cm-3 to values lower than 300 cm-3. The comparison between clean and polluted areas showed CCN concentrations 5 times higher than in polluted areas. These differences were not so strong above the boundary layer. Measurements also showed a diurnal cycle following the biomass burning activity. Cloud droplet size distributions at two regions with extremely different aerosols loading were also analyzed. During biomass-burning conditions, at high concentrations of cloud droplets, the mean diameter and liquid water content increased very little with altitude when compared with unpolluted conditions. A gamma distribution was used to fit the measured droplet spectra and the shape parameter was used as a criterion to define the best choice of spectra representation. According to the found values, narrow gamma distributions optimally fit polluted conditions (shape parameter around 5), while broad distributions are best fits for unpolluted conditions (shape parameter around 2). Based on these results, numerical experiments were carried out using the Brazilian Regional Atmospheric Modeling System (BRAMS) to investigate the effects of CCN concentrations and shape parameters of droplet spectra on the development of precipitation in tropical convective clouds. The results showed large sensitivity due to changes in these parameters. It was observed that high CCN concentrations and narrower cloud droplet distributions (high values for shape parameter), typical of the polluted days, produced low mean values of liquid water path and accumulated surface precipitation. On the other hand, an opposite trend to this effect was found under clean conditions (low CCN concentration and shape parameter values). Shape parameter showed to be much more important than CCN concentration. The effects of CCN concentration and shape parameter also influenced the spatial distribution of cloud and precipitation fields. Although mean values of these fields decreased under polluted conditions, maximum values were increased. Consequently, the less dispersion in clouds under polluted conditions, the more surface solar radiation was found. This is opposite to the results of global climate models, which predict reduction in solar radiation as a consequence of the second aerosol indirect effect. Also, it was found that the differences were reduced when the aerosols direct effect to absorb or reflect solar radiation is included. Moreover, the results suggest that additional models with explicit microphysics process treatment are necessary in order to allow further comparisons, which could show the best numerical treatment to be used in representing the aerosol effects on precipitation process. The importance of these results is to improve the understanding of future climate changes.

We present a novel new method for modelling magnetized molecular cloud cores using submillimetre linear polarization maps from thermal dust emission. Our PolCat modelling software builds a three-dimensional core model via the use of consecutive parametrized coordinate transformations, and produces simulated polarization maps to fit to observational datasets. We utilize the Ferret evolutionary optimizer to search the parameter space to simultaneously minimize chi-squared for the intensity and polarization position angle maps separately. We have applied PolCat to multiple test problems and several datasets from the SCUPOL Legacy Catalogue. We find that PolCat is able to distinguish between maps of twisted and non-twisted field geometries and identify twist symmetry. Preliminary fits to several datasets show that the best potential field geometries to our sample cores contain field twists. Further research using a larger number of maps is required to determine if twisted fields are commonplace in cores.

The middle atmosphere is a very important part of the Earth system. Until recently, we did not realize the importance of the structure of this vaporous shell and of the fundamental role it plays in both creating and sustaining life on the planet. Thanks to the development and improvement of new sounding methods and techniques, our knowledge of the composition of the atmosphere has become more detailed than ever before. We have also learned how to reveal complex interactions between different species and how they react to the incoming solar radiation. The middle part of the Earth’s atmosphere serves as a host for the Polar Mesospheric Clouds. These clouds consist of a thin layer of water-ice particles, only exsisting during the summer months and only close to the poles. There are indications that the occurrence of Polar Mesospheric Clouds may be linked to climate change. It has been pointed out that the first sightings coincide with the industrial revolution. Satellite observations have shown that Polar Mesospheric Clouds have become brighter and possibly more widely distributed during the 20th century. The clouds might therefore be suited as indicators of the variability of the climate - a good reason for studying this night-shimmering phenomena. The clouds can also be used as a proxy for middle atmospheric dynamics. In order to fully utilize Polar Mesospheric Clouds as tracers for atmospheric variability and dynamics, we need to better understand their local properties. The Optical Spectrograph and Infra-Red Imager System (OSIRIS) is one of two instruments installed on the Odin satellite. The optical spectrograph of this instrument observes sunlight scattered by the atmosphere and thus the Polar Mesospheric Clouds. This thesis deals with a tomographic technique that can reconstruct both horizontal and vertical structures of the clouds by using observations from various angles of the atmospheric region. From this information, microphysical properties such as particle sizes and number densities are obtained. The tomographic technique presented in this thesis also provides a basis for a new satellite concept - MATS. The idea behind the MATS satellite mission is to analyze wave activity in the atmosphere over a wide range of spatial and temporal scales, based on the scientific heritage from Odin/OSIRIS and the tomographic algorithms presented in this thesis. / <p>At the time of the doctoral defense, the following papers were unpublished and had a status as follows: Paper3: Submitted. Paper 4: Manuscript.</p>

The feedbacks among aerosols, clouds, and radiation are important components for understanding Earth's climate system and quantifying human-induced climate change, yet the magnitude of these feedbacks remain highly uncertain. Since every cloud droplet in the atmosphere begins with water condensing on a pre-existing aerosol particle, characterizing the ability of aerosols to uptake water vapor and form cloud condensation nuclei (CCN) are key to understanding the microphysics behind cloud formation, as well as assess the impact aerosols have on the Earth system. Through a combination of controlled laboratory experiments and field measurements, this thesis characterizes the ability of atmospheric aerosols to uptake water vapor and become CCN at controlled levels of water vapor supersaturation. The origin of the particle water uptake, termed hygroscopicity, is also explored, being from either the presence of deliquescent soluble material and/or adsorption onto insoluble surfaces. The data collected and presented is comprehensive and includes (1) ground samples of volcanic ash, collected from six recent eruptions re-suspended in the laboratory for analysis, (2) laboratory chamber and flow-tube studies on the oxidation and uptake of surface active organic compounds, and (3) in-situ aircraft measurements of aerosols from the Arctic background, Canadian boreal forests, fresh and aged biomass burning, anthropogenic industrial pollution, and from within tropical cyclones in the Atlantic basin. Having a more thorough understanding of aerosol water uptake will enable more accurate representation of cloud droplet number concentrations in global models, which can have important implications on reducing the uncertainty of aerosol-cloud-climate interactions, as well as additional uncertainties in aerosol transport, atmospheric lifetime, and impact on storm dynamics.

La consommation d’énergie est une préoccupation importante pour les systèmes de traitement Big Data à grande échelle, ce qui entraîne un coût monétaire énorme. En raison de l’hétérogénéité du matériel et des conflits entre les charges de travail simultanées, les stragglers (i.e., les tâches qui sont relativement plus lentes que les autres tâches) peuvent augmenter considérablement le temps d’exécution et la consommation d’énergie du travail. Par conséquent, l’atténuation des stragglers devient une technique importante pour améliorer les performances des systèmes de traitement Big Data à grande échelle. Typiquement, il se compose de deux phases: la détection de stragglers et la manipulation de stragglers. Dans la phase de détection, les tâches lentes (par exemple, les tâches avec une vitesse ou une progression inférieure à la moyenne) sont marquées en tant que stragglers. Ensuite, les stragglers sont traités en utilisant la technique d’exécution spéculative. Avec cette technique, une copie du straggler détecté est lancée en parallèle avec le straggler dans l’espoir qu’il puisse finir plus tôt, réduisant ainsi le temps d’exécution du straggler. Bien qu’un grand nombre d’études aient été proposées pour améliorer les performances des applications Big Data en utilisant la technique d’exécution spéculative, peu d’entre elles ont étudié l’efficacité énergétique de leurs solutions.Dans le cadre de cette thèse, nous commençons par caractériser l’impact de l’atténuation des stragglers sur la performance et la consommation d’énergie des systèmes de traitement de Big Data. Nous observons que l’efficacité énergétique des techniques actuelles d’atténuation des stragglers pourrait être considérablement améliorée. Cela motive une étude détaillée de ses deux phases: détection de straggler et traitement de straggler. En ce qui concerne la détection de straggler, nous introduisons un cadre novateur pour caractériser et évaluer de manière exhaustive les mécanismes de détection de straggler. En conséquence, nous proposons un nouveau mécanisme énergétique de détection de straggler. Ce mécanisme de détection est implémenté dans Hadoop et se révèle avoir une efficacité énergétique plus élevée par rapport aux mécanismes les plus récentes. En ce qui concerne le traitement de straggler, nous présentons une nouvelle méthode pour répartir des copies spéculatives, qui prend en compte l’impact de l’hétérogénéité des ressources sur la performance et la consommation d’énergie. Enfin, nous introduisons un nouveau mécanisme éconergétique pour gérer les stragglers. Ce mécanisme fournit plus de ressources disponibles pour lancer des copies spéculatives, en utilisant une approche de réservation dynamique de ressources. Il est démontré qu’elle améliore considérablement l’efficacité énergétique en utilisant une simulation. / Energy consumption is an important concern for large-scale Big Data processing systems, which results in huge monetary cost. Due to the hardware heterogeneity and contentions between concurrent workloads, stragglers (i.e., tasks performing relatively slower than other tasks) can severely increase the job’s execution time and energy consumption. Consequently, straggler mitigation becomes an important technique to improve the performance of large-scale Big Data processing systems. Typically, it consists of two phases: straggler detection and straggler handling. In the detection phase, slow tasks (e.g., tasks with speed or progress below the average) are marked as stragglers. Then, stragglers are handled using the speculative execution technique. With this technique, a copy of the detected straggler is launched in parallel with the straggler with the expectation that it can finish earlier, thus, reduce the straggler’s execution time. Although a large number of studies have been proposed to improve the performance of Big Data applications using speculative execution technique, few of them have studied the energy efficiency of their solutions. Addressing this lack, we conduct an experimental study to fully understand the impact of straggler mitigation techniques on the performance and the energy consumption of Big Data processing systems. We observe that current straggler mitigation techniques are not energy efficient. As a result, this promotes further studies aiming at higher energy efficiency for straggler mitigation. In terms of straggler detection, we introduce a novel framework for comprehensively characterizing and evaluating straggler detection mechanisms. Accordingly, we propose a new energy-driven straggler detection mechanism. This straggler detection mechanism is implemented into Hadoop and is demonstrated to have higher energy efficiency compared to the state-of-the-art mechanisms. In terms of straggler handling, we present a new speculative copy allocation method, which takes into consideration the impact of resource heterogeneity on performance and energy consumption. Finally, an energy efficient straggler handling mechanism is introduced. This mechanism provides more resource availability for launching speculative copies, by adopting a dynamic resource reservation approach. It is demonstrated, via a trace-driven simulation, to bring a high improvement in energy efficiency.

Auf der Azoreninsel Graciosa, etwa 1600 km westlich von Portugal, betreibt das U.S. Department of Energy eine American Radiation Measurement (ARM) Station, welche mit einer Vielzahl von Messgeräten ausgestattet ist. Durch ihre Lage im subtropischen Hochdruckgürtel eignet sich diese Station gut zur Erforschung von mariner Stratocumulusbewölkung. Für eine Klimatologie der Eigenschaften dieser Bewölkung wurden Messwerte von Radiosonden, einem Ceilometer, einem Distrometer und einem Mikrowellenradiometer aus dem Zeitraum vom 01. März 2014 bis 31. August 2018 verwendet. Daraus wurde der Tages- und Jahresgang der Wolkenhöhe und -dicke, des Flüssigwasserpfades, des Entkopplungs- und Niederschlagsverhaltens und der Inversionsstärke abgeleitet. Die Auswertung basiert auf einer Methode von Zhang et al. (2010) zur Detektion von Wolkenschichten aus Radiosondenmessungen, die für die Abgrenzung mariner Stratocumulusbewölkung von anderen Wolkenarten angepasst und anschließend mit den Messungen eines Ka-Band-Wolkenradars evaluiert wurde. Während die Parameter im Tagesgang nur geringe Änderungen zwischen 00 UTC und 12 UTC aufwiesen, konnten im Jahresgang deutliche Variationen bei Wolkenhöhe, Niederschlagsverhalten, Entkopplung und Inversionstärke gezeigt werden. Die beobachteten Variationen wurden mit den Jahresgängen der solaren Einstrahlung, der Temperatur sowie der synoptischen Variabilität über den Azoren verglichen. / The U.S. Department of Energy operates an American Radiation Measurement (ARM) station on Graciosa (Azores), an island located approximately 1600 km west of Portugal. Due to the wide variety of installed measuring devices and its location in the subtropical belt of high pressure the station is well suited for studying marine stratocumulus. To derive a climatology of the properties of these clouds, data from radiosondes, a ceilometer, a disdrometer and a microwave radiometer were used for the period from 1st March 2014 to 31st August 2018 to get the daily and annual cycle of cloud height and thickness, liquid water path, precipitation, inversion strength and the presence of decoupling. The evaluation is based on a method developed by Zhang et al. (2010) to detect cloud layers from moisture profiles in the sounding data. The method was adapted for the distinction between marine stratocumulus and other cloud types. For the evaluation of the method data from a Ka-band cloud radar were used. While only minor changes occurred in the daily variability of the properties, a clear annual cycle was observed for cloud height, precipitation, inversion strength and decoupling of the stratocumulus. The observed variations were compared with the annual cycles of solar irradiance, temperature and the synoptic patterns over the Azores.