En raison de la nature cohérente du signal RADAR à synthèse d’ouverture (RSO), les images RSO polarimétriques (RSOPOL) sont affectées par le bruit de chatoiement. L’effet du chatoiement peut être sévère au point de rendre inutilisable la donnée RSOPOL. Ceci est particulièrement vrai pour les données à une vue qui souffrent d’un chatoiement très intense.Un filtrage du bruit est nécessaire pour améliorer l’estimation des paramètres polarimétriques pouvant être calculés à partir de ce type de données. Cette opération constitue une étape importante dans le traitement et l’analyse des images RSOPOL. Récemment une nouvelle approche est apparue en traitement de données visant la solution d’une multitude de problèmes dont le filtrage, la restauration d’images, la reconnaissance de la parole, la classification ou la segmentation d’images. Cette approche est l’apprentissage profond et les réseaux de neurones à convolution (RNC). Des travaux récents montrent que les RNC sont une alternative prometteuse pour le filtrages des images RSO. En effet par leur capacité d’apprendre un modèle optimal de filtrage, ils tendent à surpasser les approches classiques du filtrage sur les images RSO. L’objectif de cette présente étude est d’analyser et d’évaluer l’efficacité du filtrage par RNC sur des données RSOPOL simulées et sur des images satellitaires RSOPOL RADARSAT-2, ALOS/PalSAR et GaoFen-3 acquises sur la région urbaine de San Francisco (Californie). Des modèles inspirés de l’architecture d’un RNC utilisé notamment en Super-résolution ont été adaptés pour le filtrage de la matrice de cohérence polarimétrique. L’effet de différents paramètres structuraux de l’architecture des RNC sur le filtrage ont été analysés, parmi ceux-ci on retrouve entre autres la profondeur du réseau (le nombre de couches empilées), la largeur du réseau (le nombre de filtres par couches convolutives) et la taille des filtres de la première couche convolutive. L’apprentissage des modèles a été effectué par la rétropropagation du gradient de l’erreur en utilisant 3 ensembles de données qui simulent la polarimétrie une vue des diffuseurs selon les classes de Cloude-Pottier. Le premier ensemble ne comporte que des zones homogènes.Les deux derniers ensembles sont composés de simulations en patchwork dont l’intensité locale est simulée par des images de texture et de cibles ponctuelles ajoutées au patchwork dans le cas du dernier ensemble. Les performances des différents filtres par RNC ont été mesurées par des indicateurs comprenant l’erreur relative sur l’estimation de signatures polarimétriques et des paramètres de décomposition ainsi que des mesures de distorsion sur la récupération des détails importants et sur la conservation des cibles ponctuelles. Les résultats montrent que le filtrage par RNC des données polarimétriques est soit équivalent ou nettement supérieur aux filtres conventionnellement utilisées en polarimétrie.Les résultats des modèles les plus profonds obtiennent les meilleures performances pour tous les indicateurs sur l’ensemble des données homogènes simulées. Dans le cas des données en patchwork, les résultats pour la restauration des détails sont nettement favorables au filtrage par RNC les plus profonds.L’application du filtrage par RNC sur les images satellitaires RADARSAT-2,ALOS/PalSAR ainsi GaoFen-3 montre des résultats comparables ou supérieurs aux filtres conventionnels. Les meilleurs résultats ont été obtenus par le modèle à 5 couches cachées(si on ne compte pas la couche d’entrée et de sortie), avec 8 filtres 3×3 par couche convolutive, sauf pour la couche d’entrée où la taille des filtres étaient de 9×9. Par contre,les données d’apprentissage doivent être bien ajustées à l’étendue des statistiques des images polarimétriques réelles pour obtenir de bon résultats. Ceci est surtout vrai au niveau de la modélisation des cibles ponctuelles dont la restauration semblent plus difficiles. / Due to the coherent nature of the Synthetic Aperture Radar (SAR) signal, polarimetric SAR(POLSAR) images are affected by speckle noise. The effect of speckle can be so severe as to render the POLSAR data unusable. This is especially true for single-look data that suffer from very intense speckle. Noise filtering is necessary to improve the estimation of polarimetric parameters that can be computed from this type of data. This is an important step in the processing and analysis of POLSAR images. Recently, a new approach has emerged in data processing aimed at solving a multi-tude of problems including filtering, image restoration, speech recognition, classification orimage segmentation. This approach is deep learning and convolutional neural networks(CONVNET). Recent works show that CONVNET are a promising alternative for filtering SAR images. Indeed, by their ability to learn an optimal filtering model only from the data, they tend to outperform classical approaches to filtering on SAR images. The objective of this study is to analyze and evaluate the effectiveness of CONVNET filtering on simulated POLSAR data and on RADARSAT-2, ALOS/PalSAR and GaoFen-3 satellite images acquired over the San Francisco urban area (California). Models inspired by the architecture of a CONVNET used in particular in super-resolution have been adapted for the filtering of the polarimetric coherency matrix. The effect of different structural parameters of theCONVNET architecture on filtering were analyzed, among which are the depth of the neural network (the number of stacked layers), the width of the neural network (the number of filters per convoluted layer) and the size of the filters of the first convolution layer. The models were learned by backpropagation of the error gradient using 3 datasets that simulate single-look polarimetry of the scatterers according to Cloude-Pottier classes. The first dataset contains only homogeneous areas. The last two datasets consist of patchwork simulations where local intensity is simulated by texture images and point target are added to the patchwork in the case of the last dataset. The performance of the different filters by CONVNET was measured by indicators including relative error on the estimation of polarimetric signatures and decomposition parameters as well as distortion measurements on the recovery of major details and on the conservation of point targets.The results show that CONVNET filtering of polarimetric data is either equivalent or significantly superior to conventional polarimetric filters. The results of the deepest models obtain the best performance for all indicators over the simulated homogeneous dataset. Inthe case of patchwork dataset, the results for detail restoration are clearly favourable to the deepest CONVNET filtering. The application of CONVNET filtering on RADARSAT-2, ALOS/PalSAR andGaoFen-3 satellite images shows results comparable or superior to conventional filters. The best results were obtained by the 5 hidden layers model (not counting the input and outputlayers), with 8 filters 3×3 per convolutional layer, except for the input layer where the filtersize was 9×9. On the other hand, the training data must be well adjusted to the statistical range of the real polarimetric images to obtain good results. This is especially true when modeling point targets that appear to be more difficult to restore.

Epidemiological research around the world has shown that exposure of urban populations to fine microparticles (PM2.5) suspended in air from, among other things, car combustion, is responsible for many cases of lung and cardiovascular disease and even mortality. However, most of these studies examine urban centers as ensembles without considering that population exposure to microparticles is not homogeneous across an urban space. For example, individuals living near major arterial roads are much more exposed to microparticles than others living in low traffic neighborhoods. Unfortunately, ground stations measuring PM2.5 are few and far between to generate accurate microparticle concentration maps at fine scales. One way to spatialize information on microparticle concentrations is to introduce remotely sensed images that allows to calculate an optical parameter of aerosols, their optical depth. The use of medium-to-fine-resolution images is not common in this area. So, we wanted to look at their potential. Tests with hyperspectral and multispectral images at these resolutions have shown that optical depth can be estimated with enough accuracy. The AODFinder software developed for this purpose performs well. Unfortunately, the small sample of AOD values and PM2.5 concentration measurements did not allow us to conclude on the possibility of using AOD as a proxy for PM2.5 and thus on the possibility of refining microparticle monitoring at the local level. / Des recherches épidémiologiques à travers le monde ont mis en évidence que l’exposition des populations urbaines aux microparticules fines (PM2.5) en suspension dans l’air provenant, entre autres, de la combustion automobile, est à l’origine des nombreux cas des maladies pulmonaires et cardiovasculaires et même des cas de mortalité. Cependant, la plupart de ces études examinent les centres urbains comme des ensembles sans tenir compte que l’exposition des populations aux microparticules n’est pas homogène à travers un espace urbain. À titre d’exemple, les individus demeurant à proximité de grandes artères routières sont beaucoup plus exposés aux microparticules que d’autres demeurant dans des quartiers de faible circulation. Malheureusement, les stations terrestres de mesure des PM2.5 sont peu nombreuses pour permettre de générer des cartes de concentration des microparticules précises à des échelles fines. Un moyen pour spatialiser l’information sur les concentrations des microparticules est d’introduire l’imagerie de télédétection qui permet de calculer un paramètre optique des aérosols, leur profondeur optique. L’utilisation des images à résolution moyenne à fine n’est pas chose courante dans ce domaine. Ainsi nous avons voulu examiner leur potentiel. Les tests avec des images hyperspectrale et multispectrale à ces résolutions ont montré que la profondeur optique peut être estimer avec suffisamment de précision. Le logiciel AODFinder développé à cette fin se comporte bien. Malheureusement le faible échantillon des valeurs de AOD et des mesures des concentrations des PM2.5 ne nous a pas permis de se prononcer sur la possibilité d’utiliser le AOD comme proxy des PM2.5 et ainsi sur la possibilité de raffiner le monitoring des microparticules à l’échelle locale.

Indiana University-Purdue University Indianapolis (IUPUI)

Indiana University-Purdue University Indianapolis (IUPUI) / Phytoplankton size classes (pico-plankton, nano-plankton, and micro-plankton) provide information about pelagic ocean ecosystem structure, and their spatiotemporal variation is crucial in understanding ocean ecosystem structure and global carbon cycling. Remote sensing provides an efficient approach to estimate phytoplankton size compositions on global scale. In the first part of this thesis, a global sensitivity analysis method was used to determine factors mainly controlling the variations of remote sensing reflectance and inherent optical properties inverse algorithms. To achieve these purposes, average mass-specific coefficients of particles were first calculated through Mie theory, using particle size distributions and refractive indices as input; and then a synthesis remote sensing reflectance dataset was created using Hydrolight. Based on sensitivity analysis results, an algorithm for estimating phytoplankton size composition was proposed in the second part. This algorithm uses five types of spectral features: original and normalized remote sensing reflectance, two-band ratios, continuum removed spectra, and spectral curvatures. With the spectral features, phytoplankton size compositions were regressed using support vector machine. According to validation results, this algorithm performs well with simulated and satellite Sea-viewing Wide Field-of-view Sensor (SeaWiFS) and Moderate Resolution Imaging Spectroradiometer (MODIS), indicating that it is possible to estimate phytoplankton size compositions through satellite data on global scale.

Indiana University-Purdue University Indianapolis (IUPUI) / Heat is one of the most important factors in our atmosphere for precipitation (thunderstorm) formation. Thermal energy from local urban land-cover is also a common source of heat in the lower atmosphere. This phenomenon is known as the urban heat island effect (UHI) and is identified as a substantial cause to a changing climate in surface weather modification. The proceeding study investigates this connection between the UHI and surface weather using remote sensing platforms A ten-year analysis of the Indianapolis UHI and thunderstorms were studied from the summer months of May, June, July, August and September (MJJAS) from 2002 until 2011. LANDSAT space borne satellite technology and land-surface based weather radar technology was used in this analysis for thunderstorm investigation. Precipitation areas identified from land-based NEXRAD WSR-88D technology were used to identify convection from non-synoptic forcing and non-normal surface diurnal heating scenarios. Only convection appearing from the UHI were studied and analyzed. Results showed twenty-one events over eighteen days with the year 2005 and 2011 having the largest frequency of events. The month of August had the largest concentration with seven events during the late afternoon hours. UHI results showed July had the largest heat island magnitude with April and September having the lowest magnitude in UHI temperatures. Three events of the twenty-one storm paths did not had a significant mean temperature difference in the heat island below the storm reflectivity. The nineteen storm paths that were significant had a warmer area underneath storm path development by an average 6.2°C than surrounding areas. UHI initiation points showed twelve of the twenty-one events having statistically significant differences between 2 km initiation areas and the rest of the study areas. Land-cover results showed low intensity developed areas had the most land-cover type (48%) in the 2km initiation buffer regions.

Indiana University-Purdue University Indianapolis (IUPUI) / Over the past two decades, northern parts of the United States have experienced extreme heat conditions. Some of the notable heat wave impacts have occurred in Chicago in 1995 with over 600 reported deaths and in Philadelphia in 1993 with over 180 reported deaths. The distribution of extreme heat events in Indianapolis has varied since the year 2000. The Urban Heat Island effect has caused the temperatures to rise unusually high during the summer months. Although the number of reported deaths in Indianapolis is smaller when compared to Chicago and Philadelphia, the heat wave in the year 2010 affected primarily the vulnerable population comprised of the elderly and the lower socio-economic groups. Studying the spatial distribution of high temperatures in the vulnerable areas helps determine not only the extent of the heat affected areas, but also to devise strategies and methods to plan, mitigate, and tackle extreme heat. In addition, examining spatial patterns of vulnerability can aid in development of a heat warning system to alert the populations at risk during extreme heat events. This study focuses on the qualitative and quantitative methods used to measure extreme heat events. Land surface temperatures obtained from the Landsat TM images provide useful means by which the spatial distribution of temperatures can be studied in relation to the temporal changes and socioeconomic vulnerability. The percentile method used, helps to determine the vulnerable areas and their extents. The maximum temperatures measured using LST conversion of the original digital number values of the Landsat TM images is reliable in terms of identifying the heat-affected regions.