Multi-temporal remotely sensed data (MSS, TM and ETM+)were used for monitoring and mapping the desertification processes in North Kordofan State, Sudan.A liear mixture model (LMM) was adopted to analyse and the desertification proccesses by using the image endmembers. interpretation of ancillary data and field observation was adopted to verfiy the role of human impacts in the temporal changes in the study area. The findings of the study proved the powerfull of remotely sensed data in monitoring and mapping the desertification processes and come out with valuable recommendations which could contribute positively in reducing desert encroachment in the area.

The advance in remote sensing technology helps people more easily assess urban growth. In this study, the utility of multiple endmember spectral mixture analysis (MESMA) is examined in a sub-pixel analysis of Landsat Thematic Mapper (TM) imagery to map urban physical components in Tampa, FL. The three physical components of urban land cover (LC): impervious surface, vegetation and soil, were compared using the proposed MESMA with a traditional spectral mixture analysis (SMA). MESMA decomposes each pixel to address the heterogeneity of urban LC characteristic by allowing the number and types of endmembers to vary on a per pixel basis. This study generated 642 spectral mixture models of 2-, 3-, and 4-endmembers for each pixel to estimate the fractions of impervious surface, vegetation, soil, and shade in the study area with a constraint of lowest root mean square error (RMSE). A comparative analysis of the impervious surface areas (ISA) mapped with MESMA and SMA demonstrated that MESMA produced more accurate results of mapping urban physical components than those by SMA. With the multiyear Landsat TM data, we quantified sub-pixel %ISA and the %ISA changes to assess urban growth in the City of Tampa, Florida during the past twenty years. The experimental results demonstrate that the MESMA approach is effective in mapping and monitoring urban land use/land cover changes using moderate-resolution multispectral imagery at a sub-pixel level.

A Geobotânica por sensoriamento remoto é uma técnica de obtenção de informações geológicas indiretas em ambientes cobertos por vegetação e apresenta grandes perspectivas por sua capacidade de otimizar trabalhos de campo e gerar possíveis alvos a serem examinados. O objetivo deste estudo foi realizar a discriminação espectral, em escala de folha e de copa, de espécies arbóreas neotropicais associadas localmente a diferentes formações e fácies geológicas, bem como defini-las remotamente como indicadoras geológicas na Estação Ecológica de Mogi-Guaçu, São Paulo, Brasil. Dados obtidos em 70 unidades amostrais, como texturais de solos e sedimentos, químicos de solos, de nível do freático, altitudinais (modelo digital de terreno),fitossociológicos e fisiológicos/estruturais de vegetação (índices hiperespectrais de vegetação), foram minerados e analisados através da técnica de quantização vetorial Self-Organizing Maps (SOM). Inga veraWilld. subsp. affinis (DC.) T.D. Penn (INVE) e Calophyllum brasiliense Cambess. (CABR) mostraram-se associadas à planície de inundação, incluindo meandros abandonados (Depósitos Aluvionares), com amplo domínio das frações argila e silte nos sedimentos. Qualea grandifloraMart. (QUGR) e Tabebuia ochracea(Cham.) Standl. (TAOC) foram identificadas apenas nas colinas e platôs da Formação Aquidauana,com altas porcentagens das frações areia fina, média e grossa, e escasso silte. Cedrela fissilisVell. (CEFI) e Zeyheria tuberculosa(Vell.) Bur. (ZETU) demonstraram estar associadas a uma fácies aflorante da Formação Aquidauana, com distinta presença das frações de areia grossa e muito grossa, além da baixa porcentagem das frações silte e areiafina. As cinco primeiras espécies tiveram dados bioquímicos e espectrais (400-2.500 nm, FieldSpec 3 Hi-Res) coletados, em escala de folha, tanto no período chuvoso quanto noseco. Seus espectros foram classificados através da técnica Multiple Endmember Spectral Mixture Analysis(MESMA). As espécies foram bem discriminadas em ambos os momentos sazonais, nessa escala de trabalho. Dentre os melhores resultados por intervalo espectral, as exatidões do produtor e do usuário não foram inferiores a 87,5%. Esse sucesso mostrou estar intimamente ligado à alta variabilidade bioquímica observada em suas folhas. As variações intra e interespecíficas em compostos bioquímicos puderam ser correlacionadas às suas variações espectrais. A discriminação espectral em escala de copa foi realizada com dois membros-finais (MF) via MESMA para CEFI, INVE e QUGR. Os pixels das imagenspré-processadas do sistema de sensores aeroportados ProSpecTIR-VS (530-2.532 nm, 1 m de resolução espacial) foram modelados por três MF: MF da classe de espécie-alvo, MF de outras classes de vegetação e sombra fotométrica. A falta de comissão espectrale a relativa baixa omissão espectral atingidas por QUGR na modelagem com dois MF, que incluiu outras classes de vegetação, refletiu em um mapeamento satisfatório de sua fração espectral. As tendências em distribuição dessa espécie indicaram claramente as colinas e platôs da Formação Aquidauana na área estudada. / Geobotany via remote sensing is a technique for obtaining indirect geological information in vegetated areas and presents great perspectives by its capability for field work optimization and target generation to be evaluated afterwards. The aim of this research was to perform the spectral discrimination of Neotropical tree species (at leaf and crown levels) which are locally associated to geological facies and formations in the Mogi-Guaçu Ecological Station, in southeastern Brazil. Data from 70 sampling units, such as soils and sediments texture, soils chemistry, groundwater level, elevation (digital terrain model), plant sociology and vegetation physiology/structure (hyperspectral vegetation indices), were mined and analyzed through the vetorial quantization method called Self-Organizing Maps (SOM; Kohonen, 1982). Inga veraWilld. subsp. affinis(DC.) T.D. Penn (INVE) and Calophyllum brasiliense Cambess. (CABR) demonstrated to be associated to the floodplain, including paleochannels (Alluvial Deposits sequence), with clayey-silty sediments. Qualea grandifloraMart. (QUGR) and Tabebuia ochracea(Cham.) Standl. (TAOC) were sampled on hills and plateaus of the Aquidauna Formation, which stood out for higher fine, medium and coarse sand contents and lower silt content. Cedrela fissilisVell. (CEFI) and Zeyheria tuberculosa(Vell.) Bur. (ZETU) showed be associated to one outcrop facies of the Aquidauna Formation, with distinctive presence of coarse and very coarse sandas well as lower silt and very fine sand contents. Biochemical and spectral (400-2.500 nm, FieldSpec Hi-Res 3) data were collected from the leaves of the first five species, during both rainy and dry seasons. Their spectra were classified through Multiple Endmember Spectral Mixture Analysis(MESMA). All target species were well discriminated at leaf scale. Considering the best classification results per spectral range, user\'s and producer\'s accuracies were always higher than 87,5%. These results seem to be linked to the great biochemical variability of their leaves. The intra and interspecific variability of biochemical compounds were correlated with spectral variability. The spectral discrimination at crown scale was performed with two endmembers (EM) via MESMA for CEFI, INVE e QUGR. The 1-m pixels of the preprocessed ProSpecTIR-VS images (530-2.532 nm) were modeled by three EM: EM of the target species, EM of other vegetation classes, and photometric shade. The QUGRclass achieved a relatively lower spectral omission and had no spectra erroneously assigned to its class in the two EM classification, which included other vegetation classes. This classification result was reproduced in the three EM image unmixing. The distribution tendency of that species clearly indicated the hills and plateaus of the Aquidauana Formation in the study area.

Multi-temporal remotely sensed data (MSS, TM and ETM+)were used for monitoring and mapping the desertification processes in North Kordofan State, Sudan.A liear mixture model (LMM) was adopted to analyse and the desertification proccesses by using the image endmembers. interpretation of ancillary data and field observation was adopted to verfiy the role of human impacts in the temporal changes in the study area. The findings of the study proved the powerfull of remotely sensed data in monitoring and mapping the desertification processes and come out with valuable recommendations which could contribute positively in reducing desert encroachment in the area.

Heterogeneous data with different characteristics are ubiquitous in the modern digital world. For example, the observations collected from a process may change on its mean or variance. In numerous applications, data are often of mixed types including both discrete and continuous variables. Heterogeneity also commonly arises in data when underlying models vary across different segments. Besides, the underlying pattern of data may change in different dimensions, such as in time and space. The diversity of heterogeneous data structures makes statistical modeling and analysis challenging. Detection of change-points in heterogeneous data has attracted great attention from a variety of application areas, such as quality control in manufacturing, protest event detection in social science, purchase likelihood prediction in business analytics, and organ state change in the biomedical engineering. However, due to the extraordinary diversity of the heterogeneous data structures and complexity of the underlying dynamic patterns, the change-detection and analysis of such data is quite challenging. This dissertation aims to develop novel statistical modeling methodologies to analyze four types of heterogeneous data and to find change-points efficiently. The proposed approaches have been applied to solve real-world problems and can be potentially applied to a broad range of areas. / Ph. D.