Conventional sources of emissions have been a prime target for policymakers in designing pollution control strategies. However, the evolution of shale gas activities is a growing concern over the impact of unconventional sources on urban and regional air quality. Owing to the development of Barnett Shale production, the fast-growing Dallas-Fort Worth (DFW) metroplex has encountered both types of these emissions. Oil and gas activities result in emissions of ozone precursors, notably volatile organic compounds (VOC). The major objective of this study was to evaluate the spatio-temporal distribution of VOC in order to highlight the influence of unconventional emissions. The study utilized measurements from automated gas chromatography (AutoGC) monitors to analyze the patterns of the total non-methane organic compounds (TNMOC) and relative contributions from marker species of traffic versus oil and gas activities. In this study, data from 2001-2014 was obtained from the Texas Commission on Environmental Quality (TCEQ) for fifteen monitoring sites within the North Texas region. With over a thousand wells in a 10 mile radius, two of the rural sites measured twice as much TNMOC as compared to the urban site in Dallas. Source apportionment analysis was conducted using Positive Matrix Factorization (PMF) technique. The target site located in the urban zone resolved an eight factor model. Natural gas signature was the dominant source of emission with a 52% contribution followed by 31% from two separate traffic-related sources. Considering ethane to be the dominant species in oil and gas emissions, it was observed that the rising ethane/NOx ratio correlated with increasing annual average ozone post-2007. In this period, higher concentration of ozone was found to be associated with stronger winds from the Barnett Shale area – a region that did not seem to contribute to high ozone during 2001-2007. With traffic emissions having flattened over the years, the recent increase in oil- and gas-related emissions has a negative impact on the air quality in this area. Results indicate that the area has failed to observe a declining trend in ozone despite effective reductions in NOx and traffic-related VOC emissions. The findings of the study would be helpful in proper evaluation of the ozone-forming potential of unconventional VOC emissions. Although these emissions may not be strong enough to cause harm through direct exposure, underestimating their potential towards ozone formation could hinder the progress in ozone attainment in growing urban areas. After all, a major portion of the study area continues to be in nonattainment of the EPA designated ozone standards. The study therefore draws the attention of policymakers towards the new influx of emissions that have emerged as a powerful source within the DFW metropolitan area.

No description available.

No description available.

In this research, a proton exchange membrane fuel cell (PEMFC) sensor was investigated for specific detection of volatile organic compounds (VOCs) for point-of-care (POC) diagnosis of the physiological conditions of humans. A PEMFC is an electrochemical transducer that converts chemical energy into electrical energy. A Redox reaction takes place at its electrodes whereas the volatile biomolecules (e.g. ethanol) are oxidized at the anode and ambient oxygen is reduced at the cathode. The compounds which were the focus of this investigation were ethanol (C2H5OH) and isoflurane (C3H2ClF5O), but theoretically, the sensor is not limited to only those VOCs given proper calibration. Detection in biosensing, which needs to be carried out in a controlled system, becomes complex in a multivariate environment. Major limitations of all types of biosensors would include poor selectivity, drifting, overlapping, and degradation of signals. Specific detection of VOCs in multi-dimensional environments is also a challenge in fuel cell sensing. Humidity, temperature, and the presence of other analytes interfere with the functionality of the fuel cell and provide false readings. Hence, accurate and precise quantification of VOC(s) and calibration are the major challenges when using PEMFC biosensor. To resolve this problem, a statistical model was derived for the calibration of PEMFC employing multivariate analysis, such as the “Principal Component Regression (PCR)” method for the sensing of VOC(s). PCR can correlate larger data sets and provides an accurate fitting between a known and an unknown data set. PCR improves calibration for multivariate conditions as compared to the overlapping signals obtained when using linear (univariate) regression models. Results show that this biosensor investigated has a 75% accuracy improvement over the commercial alcohol breathalyzer used in this study when detecting ethanol. When detecting isoflurane, this sensor has an average deviation in the steady-state response of ~14.29% from the gold-standard infrared spectroscopy system used in hospital operating theaters. The significance of this research lies in its versatility in dealing with the existing challenge of the accuracy and precision of the calibration of the PEMFC sensor. Also, this research may improve the diagnosis of several diseases through the detection of concerned biomarkers.

Si la sueur fraîchement émise par le corps humain est inodore, la dégradation de celle-ci par la flore bactérienne cutanée produit des composés volatils malodorants, responsables des odeurs de transpiration. Les odeurs de transpiration apparaissent également sur les vêtements au cours de leur utilisation, particulièrement sur les textiles réalisés en fibres synthétiques. Ce travail a pour but d’améliorer la compréhension du phénomène d’émanation d’odeurs en étudiant l’effet du sujet testé, l’effet de la flore bactérienne et l’effet du textile sur les émissions de composés volatils malodorants.L’intérêt de ce travail réside dans l’approche globale de la problématique des odeurs de transpiration et dans la diversité des méthodes de mesure mises en place, tant dans l’étude de la flore microbiologique que dans les méthodes de mesures des composés odorants émis.Dans un premier temps, le dénombrement simultané de la flore bactérienne sur la peau et sur le vêtement a été réalisé sur un échantillon de 15 sujets à l’issue d’un exercice physique. Cette expérimentation a permis d’évaluer le taux de transfert bactérien moyen lors d’une activité sportive et d’étudier son rôle dans l’émission d’odeurs. Ensuite, afin d’affiner ces résultats, une méthode basée sur la biologie moléculaire a été mise en place pour réaliser le suivi qualitatif de la stabilité de la flore commensale axillaire d’un sujet pendant 3 mois. Le transfert bactérien spécifique entre la peau du testeur et le vêtement a été étudié pour 4 matières textiles sélectionnées (dont le coton et le PET). Ceci a permis de déterminer le rôle du transfert bactérien spécifique dans l’émission des odeurs à partir de textile.Enfin, le dernier chapitre est consacré à l’étude de l’émission de composés volatils et odorants à l’aide de mesures olfactives et d’un nez électronique au cours du temps par 8 composants textiles sélectionnés. Après traitement statistique par analyse en composante principale et étude détaillée des mesures, 9 composés chimiques ont été identifiés comme indicateurs d’un comportement textile malodorant. Ces derniers pourraient être utilisés dans la mise en place d’une méthode ciblée de mesure physico-chimique des mauvaises odeurs.Ce travail a permis de déterminer l’impact de chacun des facteurs sujet, flore bactérienne et textile dans l’émission d’odeurs. En outre, ce travail ouvre des perspectives sur l’étude des contaminations bactériennes par contact, mais également dans l’étude des odeurs, sur les phénomènes de désorption de molécules volatiles à partir de différentes matrices textiles et sur les solutions pouvant être envisagées pour limiter les émissions odorantes à partir de textiles. / Fresh human sweat is odorless. Odoriferous volatile compounds are produced by the metabolism of bacteria living on the skin, generating strong malodor. Sweaty body odors do also appear on clothes during use, and especially on synthetic fabrics. The aim of this document is to improve understanding of odor emission by investigating subject effect, microbiota effect and fabric effect on the emission of odoriferous volatile compounds.Odors of perspiration are hereby globally approached with a wide use of methods and experimental devices, for microbial flora study as well as for odoriferous volatile compounds emission study.First, microflora enumeration has been simultaneously processed on the skin and on the fabric after exercise for 15 subjects. This experiment allowed an evaluation of the average bacterial transfer yield during physical activity and the beginning of the investigation of its effect on odor emission.A molecular biology methodology has then been developed in order to refine these results. Monitoring of qualitative composition of the microbiota has been performed to study the stability of the armpit’s ecosystem on a subject during 3 months. Specific microbial transfer from subject’s skin to clothe has been performed for 4 textile fabrics (including cotton and PET). This leaded to characterize the effect of specific bacterial transfer on odor emission from fabric.The last chapter is dedicated to the study of the emission of odoriferous volatile compounds over time using olfactory measurements and electronic nose for 8 selected fabrics. Principal component analysis targeted 9 chemical compounds that have been selected as malodorous behavior indicators for a given fabric. Those 9 compounds could be used for setting up a fitted physicochemical method of malodor.To conclude, this study helped to understand the effect of 3 factors in odor perception from a fabric after sport : subject, microbial flora and fabric. Perspectives have been charted on contact microbial contamination, but also on odor, and especially on desorption of odoriferous volatile molecules from a textile or knitted matrix. The solutions that could be used to limit malodorous emission from fabrics have also been discussed.

Ces dernières années, le domaine de l’apprentissage profond a connu des progrès énormes dans des applications allant de la génération d’images, détection d’objets, modélisation du langage à la réponse aux questions visuelles. Les approches classiques telles que l’apprentissage supervisé nécessitent de grandes quantités de données étiquetées et spécifiques à la tâches. Cependant, celles-ci sont parfois coûteuses, peu pratiques, ou trop longues à collecter. La modélisation efficace en données, qui comprend des techniques comme l’apprentissage few-shot (à partir de peu d’exemples) et l’apprentissage self-supervised (auto-supervisé), tentent de remédier au manque de données spécifiques à la tâche en exploitant de grandes quantités de données plus “générales”. Les progrès de l’apprentissage profond, et en particulier de l’apprentissage few-shot, s’appuient sur les benchmarks (suites d’évaluation), les métriques d’évaluation et les jeux de données, car ceux-ci sont utilisés pour tester et départager différentes méthodes sur des tâches précises, et identifier l’état de l’art. Cependant, du fait qu’il s’agit de versions idéalisées de la tâche à résoudre, les benchmarks sont rarement équivalents à la tâche originelle, et peuvent avoir plusieurs limitations qui entravent leur rôle de sélection des directions de recherche les plus prometteuses. De plus, la définition de métriques d’évaluation pertinentes peut être difficile, en particulier dans le cas de sorties structurées et en haute dimension, telles que des images, de l’audio, de la parole ou encore du texte. Cette thèse discute des limites et des perspectives des benchmarks existants, des fonctions de coût (training losses) et des métriques d’évaluation (evaluation metrics), en mettant l’accent sur la modélisation générative - les Réseaux Antagonistes Génératifs (GANs) en particulier - et la modélisation efficace des données, qui comprend l’apprentissage few-shot et self-supervised. La première contribution est une discussion de la tâche de modélisation générative, suivie d’une exploration des propriétés théoriques et empiriques des fonctions de coût des GANs. La deuxième contribution est une discussion sur la limitation des few-shot classification benchmarks, certains ne nécessitant pas de généralisation à de nouvelles sémantiques de classe pour être résolus, et la proposition d’une méthode de base pour les résoudre sans étiquettes en phase de testing. La troisième contribution est une revue sur les méthodes few-shot et self-supervised de détection d’objets , qui souligne les limites et directions de recherche prometteuses. Enfin, la quatrième contribution est une méthode efficace en données pour la description de vidéo qui exploite des jeux de données texte et vidéo non supervisés. / In recent years, the field of deep learning has seen tremendous progress for applications ranging from image generation, object detection, language modeling, to visual question answering. Classic approaches such as supervised learning require large amounts of task-specific and labeled data, which may be too expensive, time-consuming, or impractical to collect. Data-efficient methods, such as few-shot and self-supervised learning, attempt to deal with the limited availability of task-specific data by leveraging large amounts of general data. Progress in deep learning, and in particular, few-shot learning, is largely driven by the relevant benchmarks, evaluation metrics, and datasets. They are used to test and compare different methods on a given task, and determine the state-of-the-art. However, due to being idealized versions of the task to solve, benchmarks are rarely equivalent to the original task, and can have several limitations which hinder their role of identifying the most promising research directions. Moreover, defining meaningful evaluation metrics can be challenging, especially in the case of high-dimensional and structured outputs, such as images, audio, speech, or text. This thesis discusses the limitations and perspectives of existing benchmarks, training losses, and evaluation metrics, with a focus on generative modeling—Generative Adversarial Networks (GANs) in particular—and data-efficient modeling, which includes few-shot and self-supervised learning. The first contribution is a discussion of the generative modeling task, followed by an exploration of theoretical and empirical properties of the GAN loss. The second contribution is a discussion of a limitation of few-shot classification benchmarks, which is that they may not require class semantic generalization to be solved, and the proposal of a baseline method for solving them without test-time labels. The third contribution is a survey of few-shot and self-supervised object detection, which points out the limitations and promising future research for the field. Finally, the fourth contribution is a data-efficient method for video captioning, which leverages unsupervised text and video datasets, and explores several multimodal pretraining strategies.