This study analyses the spatial and temporal variability of chlorophyll-a in Chesapeake Bay; assesses the performance of Ocean Color 3M (OC3M) algorithm; and develops a novel algorithm to estimate chlorophyll-a for coastal shallow water. The OC3M algorithm yields an accurate estimate of chlorophyll-a concentration for deep ocean water (RMSE=0.016), but it failed to perform well in the coastal water system (RMSE=23.17) of Chesapeake Bay. A novel algorithm was developed which utilizes green and red bands of the Moderate Resolution Imaging Spectroradiometer (MODIS) sensor. The novel algorithm derived the chlorophyll-a concentration more accurately in Chesapeake Bay (RMSE=4.20) than the OC3M algorithm. The study indicated that the algorithm that uses red bands could improve the satellite estimation of chlorophyll-a in the coastal water system by reducing the noise associated with bottom reflectance and colored dissolved organic matter (CDOM)

Les eaux superficielles sont nécessaires à toute forme de vie en tant que parties intégrantes de tout processus de vie sur Terre. Quantifier les eaux de surface et suivre leurs variations est primordial en raison du lien direct qui existe entre les variables hydrologiques et le changement climatique. La télédétection par satellite, de l’hydrologie continental offre l’opportunité unique d’étudier, depuis l’espace, les processus hydrologiques à différentes échelles (régionale et globale). Dans cette thèse, différentes techniques ont été développées afin d’étudier les variations des eaux superficielles ainsi que d’autres variables hydrologiques, au niveau du bassin inférieur du Mékong (entre le Vietnam et le Cambodge) et ce en utilisant plusieurs estimations satellitaires différentes. Cette thèse s’articule autour de quatre points principaux. Premièrement, l’utilisation d’observations satellitaires dans le visible et dans l’infrarouge (MODIS) est étudiée et comparée afin d’évaluer les eaux de surface au niveau du bassin inférieur du Mékong. Quatre méthodes de classification ont été utilisées afin de différencier les types de surface (inondés ou pas) dans le bassin. Les différentes méthodes ont donné des cartes d’eaux de surface aux résultats semblables en terme de dynamique saisonnière. La classification la plus adaptée aux régions tropicales a été ensuite choisie pour produire une carte des eaux de surface à la résolution de 500 m entre janvier 2001 et aujourd’hui. La comparaison des séries temporelles issues de cette carte et de celles issues du produit de référence MODIS donne une forte corrélation temporelle (> 95%) pour la période 2001-2007. Deuxièmement, l’utilisation des observations issues du satellite SAR Sentinel-1 est examinée à des fins identiques. L’imagerie satellitaire optique est ici remplacée i par les images SAR qui grâce aux longueurs d’ondes utilisées dans le micro-ondes, permettent de « voir » à travers les nuages. Un jeu d’images Landsat-8-sans-nuage est alors utilisé pour entraîner un Réseau de Neurones (RN) afin de restituer des cartes d’eaux de surface par l’utilisation d’un seuillage sur les sorties du modèle RN. Les cartes sont à la résolution spatiale de 30 m et disponibles depuis janvier 2015. Comparées aux cartes de référence Landsat-8-sans-nuage, les sorties de modèles RN montre une très grande corrélation (90%) ainsi qu’une détection "vraie" à 90%. Les cartes restituées d’eaux de surface utilisant la technologie SAR sont enfin comparées aux cartes d’inondation issues de données topographiques. Les résultats montrent une fois encore une très grande consistance entres les deux cartes avec 98% des pixels considérés comme inondés dans cartes SAR se trouvant dans les régions de très grande probabilité d’inondation selon la topographie (>60%). Troisièmement, la variation volumique des eaux de surface est calculée comme le produit de l’étendue de la surface avec la hauteur d’eau. Ces deux variables sont validées à l’aide d’autres produits hydrologiques et montrent de bons résultats. La hauteur d’eau superficielle est linéairement interpolée aux régions non inondées afin de produire des cartes mensuelles à la résolution spatiale de 500 m. La hauteur d’eau est ensuite analysée pour estimer les variations volumiques. Ces résultats montrent une très bonne corrélation avec la variation volumique induite par la mesure du contenu en eau du satellite GRACE (95%) ainsi qu’avec la variation des mesures in situ de débit des rivières. Finalement, deux produits globaux et multi-satellites d’eaux superficielles sont comparés à l’échelle régionale et globale sur la période 1993-2007: GIEMS et SWAMPS. Lorsqu’elles existent, les données auxiliaires sont utilisées afin de renforcer l’analyse. Les deux produits montrent une dynamique similaire, mais 50% des pixels inondés dans SWAMPS se trouvent le long des côtes. / Surface water is essential for all forms of life since it is involved in almost all processes of life on Earth. Quantifying and monitoring surface water and its variations are important because of the strong connections between surface water, other hydrological components (groundwater and soil moisture, for example), and the changing climate system. Satellite remote sensing of land surface hydrology has shown great potential in studying hydrology from space at regional and global scales. In this thesis, different techniques using several types of satellite estimates have been made to study the variation of surface water, as well as other hydrological components in the lower Mekong basin (located in Vietnam and Cambodia) over the last two decades. This thesis focuses on four aspects. First, the use of visible/infrared MODIS/Terra satellite observations to monitor surface water in the lower Mekong basin is investigated. Four different classification methods are applied, and their results of surface water maps show similar seasonality and dynamics. The most suitable classification method, that is specially designed for tropical regions, is chosen to produce regular surface water maps of the region at 500 m spatial resolution, from January 2001 to present time. Compared to reference data, the MODIS-derived surface water time series show the same amplitude, and very high temporal correlation for the 2001-2007 period (> 95%). Second, the use of SAR Sentinel-1 satellite observations for the same objective is studied. Optical satellite data are replaced by SAR satellite data to benefit the ability of their microwave wavelengths to pass through clouds. Free-cloud Landsat-8 satellite imagery are set as targets to train and optimize a Neural Network (NN). Predicted surface water maps (30 m spatial resolution) are built for the studied region from January 2015 to present time, by applying a threshold (0.85) to the output of the NN. Compared to reference free-cloud Landsat-8 surface water maps, results derived from the NN show high spatial correlation (_90%), as well as true positive detection of water pixels (_90%). Predicted SAR surface water maps are also compared to floodability maps derived from topography data, and results show high consistency between the two independent maps with 98% of SAR-derived water pixels located in areas with a high probability of inundation (>60%). Third, the surface water volume variation is calculated as the product of the surface water extent and the surface water height. The two components are validated with other hydrological products, and results show good consistencies. The surface water height are linearly interpolated over inundated areas to build monthly maps at 500 m spatial resolution, then are used to calculate changes in the surface water volume. Results show high correlations when compared to variation of the total land surface water volume derived from GRACE data (95%), and variation of the in situ discharge estimates (96%). Fourth, two monthly global multi-satellite surface water products (GIEMS & SWAMPS) are compared together over the 1993-2007 period at regional and global scales. Ancillary data are used to support the analyses when available. Similar temporal dynamics of global surface water are observed when compared GIEMS and SWAMPS, but _50% of the SWAMPS inundated surfaces are located along the coast line. Over the Amazon and Orinoco basins, GIEMS and SWAMPS have very high water surface time series correlations (95% and 99%, respectively), but SWAMPS maximum water extent is just a half of what observed from GIEMS and SAR estimates. SWAMPS fails to capture surface water dynamics over the Niger basin since its surface water seasonality is out of phase with both GIEMS- and MODIS-derived water extent estimates, as well as with in situ river discharge data.

Les données de télédétection disponibles jusqu'à présent ne permettaient pas d'envisager un suivi spatio-temporel détaillé de l'occupation et de l'utilisation des sols à l'échelle régionale dans des régions au paysage très fragmenté, en raison de leur résolution temporelle trop faible ou de leur couverture spatiale trop limitée. Les objectifs de cette thèse étaient d'une part d'évaluer des séries temporelles d'images de télédétection à moyenne résolution spatiale pour effectuer ce type de suivi et d'autre part d'identifier et de caractériser les changements d'usage des terres en région agricole intensive à travers l'évolution de deux indicateurs, le taux de couverture hivernale des sols et le ratio céréales-prairies. Pour cela, des séries temporelles MODIS ont été constituées sur la région Bretagne entre 2000 et 2008. Une chaîne de prétraitements et de traitements complète a été mise en œuvre. Elle comprend des méthodes déjà utilisées sur des images de télédétection, mais qui ont dû être adaptées aux séries temporelles MODIS, et des méthodes originales, notamment pour la phase de reconstruction des séries temporelles. Les résultats montrent des tendances d'évolution comme la diminution des prairies au détriment des céréales ou l'augmentation de l'implantation d'intercultures. Ils mettent aussi en évidence des changements ponctuels comme l'augmentation des céréales en 2006 et 2008 et du maïs en 2001. Les principaux facteurs de changements d'occupation et d'utilisation agricole du sol, qui sont des conditions climatiques particulières ou des changements de pratiques agricoles en fonction du contexte politique ou socio-économique, ont été identifiés

Cette thèse analyse (i) la variabilité pluviométrique contemporaine en Nouvelle-Calédonie et ses téléconnexions avec les températures de surface océanique (TSO) du Pacifique tropical et (ii) l'impact des anomalies atmosphériques sur l'activité des incendies estimés par satellites. La Nouvelle-Calédonie figure aujourd'hui parmi les 34 " points chauds " de la biodiversité à l'échelle planétaire. A ce titre, l'un de nos objectifs est de construire un modèle permettant de prévoir l'intensité de la saison des feux entre septembre et décembre, qui correspond au maximum annuel. Le croisement de trois bases de données de feux détectés par satellites (MODIS, ATSR et LANDSAT) avec le réseau des stations météorologiques de Météo-France a montré qu'aucune anomalie thermique significative n'est détectée avant un feu, tandis que de forts déficits pluviométriques par rapport à la normale sont enregistrés jusqu'à trois mois avant le départ des feux. Ces déficits pluviométriques sont partiellement liés aux phases chaudes du phénomène El Niño Southern Oscillation (ENSO) et plus particulièrement à celles durant lesquelles les anomalies thermiques se situent à proximité de la ligne de changement de date équatoriale lors du printemps austral. Ces anomalies renforcent la circulation moyenne de Hadley (via le dégagement de chaleur latente dans la moyenne et haute troposphère) et la subsidence au niveau des latitudes néo-calédoniennes, tandis que les épisodes les plus intenses du siècle dernier (i.e. 1982-83 et 1997-98) dont les anomalies thermiques les plus importantes sont localisées dans le Pacifique oriental, sont associés à des conditions pluviométriques proches de la normale dans le Pacifique sud-ouest (SW). La téléconnexion entre les TSO du Pacifique central et les précipitations du Pacifique SW s'affaiblit à partir du mois de décembre au moment où l'ENSO atteint, paradoxalement, son intensité maximale. Cette modulation saisonnière est le produit d'une interaction entre (i) le cycle saisonnier des TSO brutes dans le Pacifique central, (ii) le cycle de vie des anomalies thermiques des épisodes chauds et (iii) l'intensité du gradient zonal des TSO le long de l'équateur. Par ailleurs, la convection profonde semble particulièrement sensible à la propagation vers l'est des anomalies de TSO faibles-à-modérées au niveau de la ligne de changement de date équatoriale (soit légèrement à l'ouest de la boîte Niño 3.4), modifiant significativement la position et l'intensité de la courroie de transmission des téléconnexions. Une analyse en ondelettes montre que les pluies néo-calédoniennes sont également sensibles à des modes de variations plus lents (> 8 ans) du Pacifique central entre septembre et novembre. La synchronisation entre la saisonnalité des feux et la prévisibilité saisonnière liée à l'ENSO permet d'estimer la surface brûlée totale en septembre-décembre à partir des états thermiques de l'océan Pacifique en juin-août, dont la polarité est clairement établie plusieurs mois à l'avance. La corrélation entre l'observation et la simulation du logarithme du total des surfaces brûlées en Nouvelle-Calédonie est de 0.87 sur la période 2000-2010 selon un modèle linéaire en validation croisée.

TEMPORAL EVALUATION OF SNOW DEPLETION CURVES DERIVED FOR UPPER EUPHRATES BASIN AND APPLICATIONS OF SNOWMELT RUNOFF MODEL Marim, G&ouml / khan M.S., Department of Geodetic and Geographic Information Technologies Supervisor: Prof.Dr.A.&Uuml / nal Sorman September 2008, 112 pages Water is becoming very important issue day by day with descending usable water and energy resources. In the aspect of water resources management, especially for the optimum reservoir management, predicting runoff for large reservoirs by applying hydrologic model is a recent and crucial topic. The most important model input and predictor parameters to estimate runoff for the mountainous regions are to be distribution of rainfall / temperature and snow cover area, (SCA). It is seen that many predictor variables should be integrated with Geographic Information Systems (GIS) and Remote Sensing Techniques especially for hydrologic model variable preparation. Satellite products have the potential for obtaining those kinds of data in near real time. In this study, the changes of SDC are generated by the analysis of optical satellite and by using SDC as an input to hydrological models runoff is simulated for Upper Euphrates Basin (10215.7 km2) which is a sub basin of Euphrates Basin. Largest dams of Turkey / Keban, Karakaya and Atat&uuml / rk are located on Euphrates River. Optimum operations of these dams depend on forecasting incoming water in early summer season. Euphrates River is fed mainly from snowmelts in spring or early summer time.65-70 % of the annual flow is contributed from snowmelt in that region. Main objective of this study is to obtain the spatially and temporally distributed SCA percentages from optical satellite, which are required as one of the main input variables of the hydrological model used in the application. SCA percentages and SDC are obtained for snowmelt years 2004-2007 by using high temporal resolution optical remote sensing data: Terra Moderate Resolution Imaging Spectroradiometer (MODIS). In this study, Terra MODIS snow cover map product, MOD10A1 which has a spatial resolution of 500 m is used. As a hydrological model Snowmelt Runoff Model (SRM) was applied. SRM was built up on the well-known degree day approach. In this study SRM is simulated for two years 2006 and 2007.The simulation results are compared and resultant model parameters are obtained for future runoff forecast studies. In this study, beside recommendations, discussions on the variables and SRM parameters are also provided.

On April 20, 2010 a tragic explosion aboard the Deepwater Horizon (DWH) drilling rig marked the beginning of one of the worst environmental disasters in history. For 87 days oil and gas were released into the Gulf of Mexico. In August 2010, anomalous phytoplankton activity was identified in the Northeastern Gulf of Mexico, using the Fluorescence Line Height (FLH) ocean color product. The FLH anomaly was bound by approximately 30-28 degrees North and 90 and 86 degrees West and there was a suggestion that this anomaly may have occurred due to the presence of oil. This study was designed to examine alternative explanations and to determine what influence the Mississippi River and the freshwater diversions, employed in the response efforts, may have had on the development of the FLH anomaly. The combination of the anomalously high flow rate in the Mississippi River observed in June-August 2010, the use of freshwater diversions, and three severe storms increased the flow of water through the adjoining marshes. We propose that these conditions reduced the residence time of water and nutrients on the wetlands, and likely mobilized nutrients leading to increased fresh water and nutrients being discharge to the coasts around the Mississippi Delta. Salinity contour maps created from data collected by ships operating in the Northeastern Gulf of Mexico showed that the 31 isohaline was upwards of 250km east of the Mississippi River Birds Foot Delta in August 2010. The American Seas (AmSeas) numerical circulation model was used to examine the dispersal and distribution of water parcels from the Mississippi River and freshwater diversions. Two virtual particle seeding locations were used to trace particles to obtain a measure of the percentage of particles entering a Region of Interest (ROI) located in the center of the FLH anomaly, i.e. 150 km east of the Mississippi Delta. All environmental data examined suggest that the eastward dispersal of the Mississippi River water including that derived from freshwater diversions and storm activity contributed to the development of FLH anomaly in August 2010. Chapter two examines the spectral characteristics of water and oil collected by the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS). Several peaks in the spectral features of the total radiance of surface oil between 1907nm and 2400nm appear to be absent for water. An algorithm (Spectral Line Height) was created to measure the height of the peak at 2142nm relative to a baseline between 2013nm and 2390nm. A normalized difference technique developed by the USGS was used as a validation tool. Preliminary results of the SLH technique appear to compare favorably with the results derived using the USGS technique. The SLH technique worked in areas that did not show sunglint or shallow bottom features. Sunglint areas would require additional correction to remove the effect of specular reflection. The SLH technique shows promise but will require validation to develop into an operational remote sensing method.

No description available.

The sea ice extent change and variability of the Canadian Arctic Archipelago (CAA) are quite different compared to the Arctic as a whole due to its unique geographic settings. In this thesis, the sea ice retreat processes, the connection with other Arctic regions, and the linkages to the surface radiation flux in the CAA are examined. The sea ice retreat processes in the CAA follow a four-phase process: a slow ice melt phase that usually lasts until early June (phase 1); a quick melt phase with large daily sea ice extent change which lasts close to half-a-month (phase 2); a slow melt phase that looks like slow sea ice melt or even a small ice increase that lasts another half-a-month (phase 3); and a steady ice decrease phase (phase 4). With the help of Moderate-Resolution Imaging Spectroradiometer (MODIS) data, it is identified that the quick melt in phase 2 is actually melt ponding, with melt ponds being falsely identified as open water by passive microwave. A simplified data assimilation method is then developed to improve the passive microwave sea ice concentration estimation by fusion with MODIS ice surface temperature data. The ice concentration from the analysis is found to improve the original passive microwave sea ice concentration estimation, with the largest improvements during sea ice melt. The sea ice retreat patterns in the CAA region are correlated with the sea ice retreat patterns in other regions of the Arctic. A decision tree classifier is designed to segment the sea ice retreat patterns in the CAA into several classes and classification maps are generated. These maps are effective in identifying the geographic locations that have large changes in the sea ice retreat patterns through the years. The daily progressions of the surface radiation components are described in detail. Due to the lack of multiple reflection, the percentage of shortwave radiation at the top of atmosphere that reaches the surface is influenced by the form of melt ponds over ice surface. The roles that each surface radiation component plays in forcing sea ice retreat are different in different years.

Snow indicates the potential stored water volume that is an important source of water supply, which has been the most valuable and indispensable natural resource throughout the history of the world. Euphrates and Tigris, having the biggest dams of Turkey, are the two largest trans-boundary rivers that originate in Turkey and pass throughout the water deficit nations Syria, Iran, Iraq and Saudi Arabia bringing life as well as water all their way. Snowmelt runoff originating from the mountains of Eastern Turkey accounts for 60 to 70 % of total annual discharge observed in Euphrates and Tigris. For an optimum operation of the dams, maximizing energy production, mitigation of floods and satisfying water rights, hydrological models which can both simulate and forecast the river discharges of Euphrates and Tigris are needed. In this study a hydrological model, snowmelt runoff model (SRM), is used in conjunction with remote sensing and geographic information systems to forecast the river discharges in the headwaters of Euphrates River, Upper Euphrates Basin. NOAA and MODIS satellite images were used to derive the snow covered area (SCA) information required by SRM. Linear reduction methodologies based on accumulated air temperature, with constant or varying gradient, were developed to get the continuous daily SCA values from the discrete daily satellite images. Temperature and precipitation forecasts were gathered from two different numerical weather prediction models, namely European Center for Medium Range Weather Forecasts (ECMWF) and Mesoscale Model Version 5 (MM5) from Turkish State Meteorological Services. These data sets provided t+24 hour forecasts of both temperature and precipitation. Temperature, precipitation and SCA information are fed into SRM. Discharge forecasts obtained from the model outputs are compared with the observed values. The overall performance of the model was seen as promising. Possible reasons of the mismatches between the forecasted and observed values are searched. Experiences gained throughout the study are summarized and recommendations on further forecast studies are mentioned.

A vegetação tropical é uma das maiores fontes de evapotranspiração e desempenha um papel importante na dinâmica climática global e nos processos de produtividade primária de ecossistemas terrestres, uma vez que se encontra diretamente ligada aos processos de transferência de massa e energia. Objetivou-se nesta tese de doutorado avaliar a acurácia das estimativas de ET baseadas em modelos de sensoriamento remoto quando comparadas a medições pontuais obtidas a partir de torres de fluxo e a partir do fechamento do balanço hídrico a médio e longo prazo. Foram selecionados dois modelos: para escalas locais e regionais usou-se o algoritmo SEBAL, enquanto que para escalas continentais e globais utilizou-se o algoritmo MOD16. Em relação ao algoritmo SEBAL, a metodologia empregada apresentou bons resultados tanto nas estimativas pontuais dos fluxos de calor latente quanto em relação ao fechamento do balanço hídrico em grandes áreas. Em relação ao algoritmo MOD16, verificou-se que as estimativas de evapotranspiração apresentam melhores resultados a longo prazo e em grandes áreas. As análises aqui realizadas sugerem que os algoritmos SEBAL e MOD16 podem capturar razoavelmente bem as respostas da vegetação em relação à variabilidade espacial e temporal em larga escala. Considerando que esses algoritmos podem ser aplicados desde escalas regionais até escalas globais, os resultados estão em um intervalo aceitável de acurácia, indicando um potencial significativo no monitoramento espacial e temporal do processo de ET de forma contínua e sistemática a partir de informações obtidas por sensoriamento remoto. / Tropical vegetation is a major source of global land surface evapotranspiration (ET) and plays an important role in the global climate dynamics and in the processes of primary production of terrestrial ecosystems, since it is directly linked to the mass transfer and energy balance. The aim of this doctoral thesis was to evaluate the accuracy of ET estimates based on remote sensing models when compared to measurements obtained from flux towers and hydrological models. Two algorithms were selected: for local and regional scales SEBAL was used, while for continental and global scales, MOD16 was adopted. Regarding SEBAL algorithm, the applied methodology showed that significant results were achieved in both latent heat flux and water balance over large areas. Concerning MOD16 algorithm, the research suggested that evapotranspiration estimates showed better results in the long-term and in large areas. The analysis suggests that MOD16 and SEBAL algorithms can capture reasonably well the responses of vegetation in relation to spatial and temporal variability on a large scale. Taking into consideration that these algorithms can be applied from regional to global scales, the results are in an acceptable range of accuracy, indicating a significant potential for monitoring evapotranspiration using remote sensing data.