One of the most important volatiles in magma is water. In magma, water is incorporated in both the melt and in crystals, including hydrous and nominally anhydrous minerals (NAMs). The purpose of this study is to determine the role of magmatic water content in controlling volcano explosivity especially at Merapi volcano, Indonesia, which is one of the most active volcanoes in the world. Here I apply clinopyroxene from two types of eruption products, lava and ash, to test the magmatic water content in the Merapi 2010 eruption. Both oriented and unoriented crystals were analysed using Fourier-Transformed Infrared (FTIR) spectroscopy to establish the clinopyroxene’s crystal water content. By combining the water content of clinopyroxene and geochemical data obtained from EPMA analysis, the magmatic water content is then estimated by inserting the data into the equation of O’Leary et al. (2010) [lnD= -4.2(60.2)16.5(60.5) VI[Al3+]-1.0(60.2) [Ca2+]]. I also investigated clinopyroxene crystals that were experimentally rehydrated, including four oriented crystals from a Merapi 2010 eruption ash sample annealed at a temperature of 700°C and pressure of 1 atm. Magmatic water contents of Merapi 2010 from rehydration experiments were then compared to the Merapi 2006 magmatic water content which were treated under the same conditions. The results revealed a wide variation between samples and among crystals taken from the same sample. However, oriented samples yielded a smaller variation between minimum and maximum clinopyroxene crystal water content and, overall, ash samples yielded higher water content in clinopyroxene than lava samples. The rehydration experiment significantly amplified clinopyroxene crystal water content by 3 to 7-fold compared to the result without the rehydration procedure. Comparison of data between the 2006 and 2010 Merapi eruptions demonstrates that there was a significant difference in magmatic water contents with the 2010 eruption displaying more than twice as much magmatic water than the 2006 eruption. The result implies that magmatic water plays potentially a crucial role in controlling explosivity of Merapi eruptions. / Indonesien är det land i Världen som har den största befolkningen som lever inom 100 km av en aktiv vulkan. Merapi är en av de mest farliga vulkanerna i Indonesien, som också är en av de mest aktiva vulkanerna i Världen och ligger på ön Java runt 15-25 km från Yogyakarta med en total befolkningstäthet på 1000 personer per kvadratkilometer. Utbrott från Merapi har dödat cirka 5000 personer under de senaste 500 åren. Således är det viktigt att studera hur vulkanen beter sig och orsaken till utbrotten för att kunna vita förebyggande åtgärder för att minska antalet dödsfall eller till och med undvika dödsfall helt. Vulkanutbrott kontrolleras av många faktorer, en av faktorerna är flyktiga ämnen i magman och en av de viktigaste flyktiga ämnen är vatten. Det här projektet syftar till att ta reda på vattnets roll på explosiviteten hos utbrotten genom att undersöka eruptiva produkter från utbrotten av Merapi under 2006 och 2010 med hjälp av Fourier-Transformed Infrared (FTIR) spektroskopi. Experimentet använder normalt vattenfria mineraler (NAMs) så som klinopyroxen men vatten kan införlivas i NAMs genom att binda väte i kristallgitteret under kristalltillväxten. Tidigare studier har visat att vattenhalten i klinopyroxenkristall skulle kunna användas för att beräkna magmatisk vattenhalt och därmed jämföra den magmatiska vattenhalt mellan olika utbrott med olika explosivitet i samma vulkan, vilket kan indikera hur vattenhalten påverkar vulkanutbrotten. Magmatiskt vatteninnehåll beräknat utifrån klinopyroxenvattenhalten visade en variation mellan de undersöka proverna från samma utbrott, varför ytterligare undersökning genom rehydreringsexperiment gjordes genom annealing av proverna vid en temperatur av 700 °C och tryck vid 1 atm. Även om det var variation i magmatisk vattenhalt mellan proverna, så har utbrotten av Merapi under 2010 högre magmatiskt vatteninnehåll än utbrottet under 2006 med 3-7 gånger. Sammantaget har Merapi 2010 utbrottet högre explosivitet och högre magmatisk vattenhalt än 2006 utbrottet, vilket innebär att högre vattenhalt ger högre explosivitet. På detta sätt kontrollerade vattnet vulkanens explosivitet för Merapivulkanen.Nyckelord: Merapi vulkan, NA

Neste trabalho estudam-se alguns novos métodos de seleção de variáveis no contexto da regressão linear que surgiram nos últimos 15 anos, especificamente o LARS - Least Angle Regression, o NAMS - Noise Addition Model Selection, a Razão de Falsa Seleção - RFS (FSR em inglês), o LASSO Bayesiano e o Spike-and-Slab LASSO. A metodologia foi a análise e comparação dos métodos estudados e aplicações. Após esse estudo, realizam-se aplicações em bases de dados reais e um estudo de simulação, em que todos os métodos se mostraram promissores, com os métodos Bayesianos apresentando os melhores resultados. / In this work, some new variable selection methods that have appeared in the last 15 years in the context of linear regression are studied, specifically the LARS - Least Angle Regression, the NAMS - Noise Addition Model Selection, the False Selection Rate - FSR, the Bayesian LASSO and the Spike-and-Slab LASSO. The methodology was the analysis and comparison of the studied methods. After this study, applications to real data bases are made, as well as a simulation study, in which all methods are shown to be promising, with the Bayesian methods showing the best results.

Neste trabalho estudam-se alguns novos métodos de seleção de variáveis no contexto da regressão linear que surgiram nos últimos 15 anos, especificamente o LARS - Least Angle Regression, o NAMS - Noise Addition Model Selection, a Razão de Falsa Seleção - RFS (FSR em inglês), o LASSO Bayesiano e o Spike-and-Slab LASSO. A metodologia foi a análise e comparação dos métodos estudados e aplicações. Após esse estudo, realizam-se aplicações em bases de dados reais e um estudo de simulação, em que todos os métodos se mostraram promissores, com os métodos Bayesianos apresentando os melhores resultados. / In this work, some new variable selection methods that have appeared in the last 15 years in the context of linear regression are studied, specifically the LARS - Least Angle Regression, the NAMS - Noise Addition Model Selection, the False Selection Rate - FSR, the Bayesian LASSO and the Spike-and-Slab LASSO. The methodology was the analysis and comparison of the studied methods. After this study, applications to real data bases are made, as well as a simulation study, in which all methods are shown to be promising, with the Bayesian methods showing the best results.

The aim of this thesis is to increase our knowledge and understanding of trace water concentrations in nominally anhydrous minerals (NAMs). Special focus is put on the de- and rehydration mechanisms of clinopyroxene crystals in volcanic systems, how these minerals can be used to investigate the volatile content of mantle rocks and melts on both Earth and other planetary bodies (e.g., Mars). Various analytical techniques for water concentration analysis were evaluated. The first part of the thesis focusses on rehydration experiments in hydrogen gas at 1 atm and under hydrothermal pressures from 0.5 to 3 kbar on volcanic clinopyroxene crystals in order to test hydrogen incorporation and loss from crystals and how their initial water content at crystallization prior to dehydration may be restored. The results show that extensive dehydration may occur during magma ascent and degassing but may be hindered by fast ascent rates with limited volatile loss. De- and rehydration processes are governed by the redox-reaction OH- + Fe2+ ↔ O2- + Fe3+ + ½ H2. Performing rehydration experiments at different pressures can restore the water contents of clinopyroxene at various levels in the volcanic systems. Subsequently water contents of magmas and mantle sources can be deduced based on crystal/melt partition coefficients. This thesis provides examples from the Canary Islands, Merapi volcano in Indonesia and the famous Nakhla meteorite. Using NAMs as a proxy for magmatic and mantle water contents may provide a very good method especially for planetary science where sample material is limited. The thesis’ second part focusses on analytical methods to measure the concentration of water in NAMs. Specifically the application of Raman spectroscopy and proton-proton scattering are tested. The hydrated mineral zoisite is thoroughly analyzed in order to be used as an external standard material. Polarized single crystal spectra helped to determine the orientation of the OH-dipole in zoisite. Further, Transmission Raman spectroscopy and a new method for the preparation of very thin samples for proton-proton scattering were developed and tested. The results provide new possibilities for the concentration analysis of water in NAMs such as three dimensional distribution and high spatial resolution.

Avec plus de 350 000 produits chimiques utilisés et de nouveaux arrivant sur le marché chaque année, des outils rapides et à coûts réduits sont nécessaires pour l'étude de ces produits. L’évaluation des risques pour ces produits est généralement faite à partir d’études animales, mais celles-ci présentent plusieurs limitations. Par exemple, évaluer le potentiel cancérogène d’une substance prendre jusqu'à trois ans et coûter six millions de dollars. En outre, il a été démontré que les modèles animaux n'ont qu'un faible pouvoir prédictif par rapport aux effets chez l’humain. Pour surmonter ces obstacles, on assiste actuellement à un mouvement mondial en faveur du développement et de l'acceptation de nouvelles approches méthodologiques (NAM) pour la priorisation des produits chimiques et l'évaluation des risques. Notre objectif était d’élaborer et d'évaluer une nouvelle approche d’extrapolation in vitro à in vivo (IVIVE) pour établir des niveaux d'exposition acceptables chez l'homme en combinant des des études in vitro et des outils de modélisation toxicologique. À cette fin, nous avons développé et évalué un outil informatique utilisé dans l'approche IVIVE et mené des études de cas sur deux produits chimiques pour lesquels étaient disponibles des données in vitro, des modèles d'exposition, et des études épidémiologiques associant l’exposition à des effets néfastes chez l’humain. Dans le premier article, nous avons développé et évalué un modèle de bilan de masse dynamique (IV-MBM DP v1.0) pour estimer les concentrations intracellulaires au cours d'expériences in vitro avec administration répétée, incluant une description du transport facilité. Pour évaluer la précision du modèle, nous avons paramétré et appliqué le modèle à des scénarios de dose unique et de doses répétées, et évalué les concentrations estimées aux données empiriques. En outre, nous avons simulé des scénarios de dosage répété pour des produits chimiques organiques représentant une diversité de caractéristiques physico-chimiques, et nous avons comparé leur dispersion dans le système au fil du temps. Dans l'ensemble, pour les scénarios de dosage unique et répété, la concordance entre les données simulées et expérimentales a illustré le pouvoir prédictif du modèle. i Dans les deuxième et troisième articles, nous nous sommes concentrés sur l'utilisation et l'évaluation de notre nouvelle approche IVIVE en faisant deux études de cas impliquant l'exposition placentaire et lactationnelle à des polluants organiques persistants. La première étape de notre méthodologie a été la sélection d'un point de départ à partir d'une étude in vitro utilisant des cellules humaines. Ensuite, nous avons appliqué la modélisation benchmark dose pour obtenir la concentration associée à un changement relatif de 5% de la réponse par rapport au contrôle. Nous avons ensuite appliqué la modélisation de de bilan de masse pour déterminer la concentration cellulaire pour un point de départ conduisant à un changement de réponse de 5%. Un modèle toxicocinétique pour le transfert placentaire et par l’allaitement a ensuite été utilisé pour calculer la dose équivalente administrée et la concentration plasmatique associée, et des facteurs d'incertitude (variabilité interindividuelle (10) et sous-chronique à chronique (10)) ont été appliqués pour calculer les apports quotidiens tolérables et les équivalents de biosurveillance. Les équivalents de biosurveillance ont été comparés aux concentrations dans le sang de la mère et du cordon ombilical mesurées dans les études épidémiologiques. Nos études de cas portaient sur la neurotoxicité développementale du 2,2',4,4'-tétrabromodiphényléther (BDE-47) et sur l'obésogénicité du dichlorodiphényldichloroéthylène (p,p'-DDE). Pour les deux études, les apports quotidiens tolérables calculés en tenant compte des facteurs d’incertitude étaient plus faibles que les valeurs toxicologiques de référence déterminées sur la base d’études animales. En outre, les deux études de cas ont produit des équivalents de biosurveillance se situant dans la gamme des concentrations maternelles et du cordon ombilical mesurées dans les études épidémiologiques. Dans l'ensemble, l'évaluation de notre modèle de bilan de masse, ainsi que les valeurs conservatrices générées par l'approche IVIVE dans nos études de cas, renforcent la confiance dans les NAM, ce qui est essentiel pour leur adoption future par les organismes de réglementation. / With over 350,000 chemicals in use and more entering the market every year, cost-effective and time-efficient tools are necessary for the investigation of these products. Whole animal models are traditionally used and accepted by regulatory agencies; however, animal models carry multiple limitations. Specifically, animal models may take up to three years and six million dollars to investigate the carcinogenicity of a compound. Additionally, animal models have been shown to have poor predictive power for human safety. To overcome these obstacles, a global movement toward the development and acceptance of new alternative methods (NAMs) for chemical prioritization and risk assessment is taking place. Our objective was to develop and evaluate a novel in vitro to in vivo (IVIVE) approach to establish acceptable exposure levels in humans by combining novel in vitro and biological/computational modeling technologies for chemical safety assessment. To this end, we tested and evaluated a computational tool utilized in the IVIVE approach, and conducted proof-of-concept studies on two case chemicals with publicly available in vitro data, exposure models, and epidemiological studies demonstrating adverse health effects. In the first paper, we aimed to develop and evaluate a dynamic partitioning mass-balance model (IV-MBM DP v1.0) to estimate intracellular concentrations during in vitro experiments of repeat dosing, and incorporate facilitated transport into the model. To evaluate the model accuracy, we parametrized and applied the model to single dose and repeat dosing scenarios and assessed the output against empirical data. In addition, we simulated repeat dosing scenarios for organic chemicals with different properties and compared their dispersion within the system over time. Overall, for single and repeat dosing scenarios, concordance between simulated and experimental data illustrated the predictive power of the model. In the second and third papers, we focused on the use and evaluation of our novel IVIVE approach through case studies involving placental and lactational exposure to persistent organic pollutants. The first step of our methodology was the selection of a point of departure from an in vitro study utilizing human cells. Next, we applied benchmark dose modeling to obtain the nominal iv concentration at a 5% relative change in response from control. We subsequently applied mass- balance modeling to determine the cellular concentration for the POD leading to a 5% change in response. A toxicokinetic model for placental transfer and lactation was then used to calculate the administered equivalent dose and associated maternal and cord plasma concentration, and uncertainty factors (interindividual variability (10) and subchronic to chronic (10)) were applied to calculate tolerable daily intakes and biomonitoring equivalents. Biomonitoring equivalents were compared to concentrations in maternal and cord blood measured in epidemiological studies. Our case studies were on 2,2',4,4'-tetrabromodiphenyl ether (BDE-47) developmental neurotoxicity and dichlorodiphenyldichloroethylene (p,p’-DDE) obesogenicity. For both studies, calculated tolerable daily intakes accounting for uncertainty factors were more conservative than the reference doses determined through the use of whole animal models. Moreover, both case studies produced biomonitoring equivalents within the range of maternal and cord levels measured in epidemiological studies. Overall, assessment of our IV-MBM DP v1.0 mass-balance model, as well as the demonstrated protective quality of the IVIVE approach in our case studies, enhances confidence in NAMs, which is essential for their future adoption by regulatory agencies.