Dados EASE-grid do Special Sensor Microwave-Imager (SSM/I) e imagens classificadas ASAR wideswath (WS), cobrindo a Península Antártica (PA), foram processadas e usadas em um Modelo Linear de Mistura Espectral (MLME), para a análise subpixel da Zona Superficial de Neve Úmida (ZSNU) em imagens SSM/I. As proporções dos componentes puros (imagens-fração) da área de estudo (ZSNU, Zona Superficial de Neve Seca e rochas) foram derivadas das imagens ASAR classificadas. As imagens-fração e imagens SSM/I co-registradas de mesma data (bandas 19H, 19V, 37H e 37V) foram usadas no MLME para estimar as assinaturas espectrais desconhecidas (i.e., temperatura de brilho em cada banda SSM/I). Essas assinaturas espectrais foram então usadas no MLME para estimar as imagens-fração da ZSNU, as quais foram comparadas com as imagens-fração ASAR correspondentes, por meio do cálculo do coeficiente de correlação. Foram identificadas as duas assinaturas espectrais que resultaram nos dados mais correlacionados, sendo também calculadas as correlações das imagens-fração da ZSNU resultantes do uso no MLME dos valores médio e mediano das assinaturas espectrais mais similares. Os valores medianos dessas assinaturas espectrais produziram as imagens-fração da ZSNU mais correlacionadas, que tiveram uma precisão global de classificação média (PGCM) de 95,6% e 97,3%, nas imagens de primavera e outono, respectivamente (amplitude de classes de 0,1), e uma PGCM de 72,6% nas imagens de verão (amplitude de classes de 0,2). Essas assinaturas espectrais medianas foram então usadas no MLME para estimar, com esses níveis de precisão global, a intensidade e extensão da ZSNU na PA, pelo uso de imagens calibradas SSM/I e SMMR (Scanning Multichannel Microwave Radiometer), possibilitando assim a análise diária e em nível subpixel dessa fácie superficial, de 1978 a 2008. Na análise espacial das imagens-fração da ZSNU estimadas, observou-se que o derretimento superficial médio começava no final de outubro e terminava no final de março, com auge em 7 de janeiro (cerca de 172.237 km2 ou 31,6% da área da PA). A área total mediana da ZSNU no verão foi de aproximadamente 105.100 km2. A análise de regressão com as imagens-fração dos verões entre 1978-1979 a 2007-2008 revelou a tendência de redução da área da ZSNU, totalizando 330,854 km2 nesse período. Todavia, essa tendência não é estatisticamente significante, devido à alta variabilidade interanual da área da ZSNU na PA. Forte derretimento superficial ocorreu nos verões de 1984-1985 (176.507,289 km2) e 1989-1990 (172.681,867 km2), enquanto fraco derretimento, nos verões de 1993-1994 (26.392,208 km2) e 1981-1982 (23.244,341 km2). O mais persistente e intenso derretimento superficial foi observado nas plataformas de gelo Larsen, Wilkins, George VI e Wordie e isto foi relacionado com os eventos de fragmentação e desintegração dessas massas de gelo, ocorridos nas últimas décadas. O derretimento superficial está intimamente relacionado com a estabilidade do sistema glacial antártico e com mudanças no nível médio dos mares. Esse poderia ser monitorado em toda a Antártica, por meio da análise subpixel de imagens SMMR e SSM/I proposta neste estudo. / Special Sensor Microwave-Imager (SSM/I) EASE-grid data and classified ASAR wideswath (WS) images, covering the Antarctic Peninsula (AP), were processed and used in a Spectral Linear Mixing Model (SLMM) for a subpixel analysis of the Wet Snow Zone (WSZ) in SSM/I images. The components’ proportions (fraction images) of the endmembers in the study area, namely WSZ, Dry Snow Zone and rock outcrops, were derived from classified ASAR images. These fraction images and co-registered SSM/I images (bands 19H, 19V, 37H and 37V), acquired on the same date, were used in the SLMM to estimate the unknown spectral signatures (i.e., brightness temperature on each SSM/I band). These spectral signatures were used to estimate WSZ fraction images, which were compared with the ASAR fraction images, by calculating the correlation coefficients. This work identified two spectral signatures that produced the most correlated data, and determined the WSZ fraction images correlations resulting from the use, in the SLMM, of the mean and median values of the most similar spectral signatures. The median values of these spectral signatures produced the most correlated WSZ fraction images, which had an average overall classification accuracy (AOCA) of 95.6% and 97.3% for spring and autumn fraction images, respectively (class range of 0.1), and an AOCA of 72.6% for summer fraction images (class range of 0.2). These median spectral signatures were then used in a SLMM to estimate accurately the WSZ intensity and its extension on the AP, by using calibrated SSM/I and SMMR (Scanning Multichannel Microwave Radiometer) imageries, allowing a daily subpixel analysis of this glacier facie on the AP from 1978 to 2008. Based on the spatial analysis of the WSZ fraction images, it was observed that melt primarily takes place in late October and ends in late March, with peak on January 7th (about 172,237 km2 or 31,6% of the AP area). The WSZ median total area in summer was about 105,100 km2. Regression analysis over the 1978-1979 to 2007-2008 summers, revealed a negative interanual trend in surface melt of 330.854 km2. Nevertheless, this trend inference is not statistically significant, due to the high WSZ interanual variability. Extremely high melt occurred in the 1984-1985 (176,507.289 km2) and 1989-1990 (172,681.867 km2) summers, while extremely weak melt occurred in the 1993-1994 (26,392.208 km2) and 1981-1982 (23,244.341 km2) summers. The most persistent and intensive melt was observed on Larsen, Wilkins, George VI and Wordie ice shelves and it was related to the break-up and disintegration events that occurred on these glaciers in the last decades. Surface melting is closely related to the stability of the Antarctic glacial system and global sea level changes. It could be monitored for the whole Antarctica, by using the WSZ subpixel analysis in SMMR and SSM/I imageries proposed by this study.

Dados EASE-grid do Special Sensor Microwave-Imager (SSM/I) e imagens classificadas ASAR wideswath (WS), cobrindo a Península Antártica (PA), foram processadas e usadas em um Modelo Linear de Mistura Espectral (MLME), para a análise subpixel da Zona Superficial de Neve Úmida (ZSNU) em imagens SSM/I. As proporções dos componentes puros (imagens-fração) da área de estudo (ZSNU, Zona Superficial de Neve Seca e rochas) foram derivadas das imagens ASAR classificadas. As imagens-fração e imagens SSM/I co-registradas de mesma data (bandas 19H, 19V, 37H e 37V) foram usadas no MLME para estimar as assinaturas espectrais desconhecidas (i.e., temperatura de brilho em cada banda SSM/I). Essas assinaturas espectrais foram então usadas no MLME para estimar as imagens-fração da ZSNU, as quais foram comparadas com as imagens-fração ASAR correspondentes, por meio do cálculo do coeficiente de correlação. Foram identificadas as duas assinaturas espectrais que resultaram nos dados mais correlacionados, sendo também calculadas as correlações das imagens-fração da ZSNU resultantes do uso no MLME dos valores médio e mediano das assinaturas espectrais mais similares. Os valores medianos dessas assinaturas espectrais produziram as imagens-fração da ZSNU mais correlacionadas, que tiveram uma precisão global de classificação média (PGCM) de 95,6% e 97,3%, nas imagens de primavera e outono, respectivamente (amplitude de classes de 0,1), e uma PGCM de 72,6% nas imagens de verão (amplitude de classes de 0,2). Essas assinaturas espectrais medianas foram então usadas no MLME para estimar, com esses níveis de precisão global, a intensidade e extensão da ZSNU na PA, pelo uso de imagens calibradas SSM/I e SMMR (Scanning Multichannel Microwave Radiometer), possibilitando assim a análise diária e em nível subpixel dessa fácie superficial, de 1978 a 2008. Na análise espacial das imagens-fração da ZSNU estimadas, observou-se que o derretimento superficial médio começava no final de outubro e terminava no final de março, com auge em 7 de janeiro (cerca de 172.237 km2 ou 31,6% da área da PA). A área total mediana da ZSNU no verão foi de aproximadamente 105.100 km2. A análise de regressão com as imagens-fração dos verões entre 1978-1979 a 2007-2008 revelou a tendência de redução da área da ZSNU, totalizando 330,854 km2 nesse período. Todavia, essa tendência não é estatisticamente significante, devido à alta variabilidade interanual da área da ZSNU na PA. Forte derretimento superficial ocorreu nos verões de 1984-1985 (176.507,289 km2) e 1989-1990 (172.681,867 km2), enquanto fraco derretimento, nos verões de 1993-1994 (26.392,208 km2) e 1981-1982 (23.244,341 km2). O mais persistente e intenso derretimento superficial foi observado nas plataformas de gelo Larsen, Wilkins, George VI e Wordie e isto foi relacionado com os eventos de fragmentação e desintegração dessas massas de gelo, ocorridos nas últimas décadas. O derretimento superficial está intimamente relacionado com a estabilidade do sistema glacial antártico e com mudanças no nível médio dos mares. Esse poderia ser monitorado em toda a Antártica, por meio da análise subpixel de imagens SMMR e SSM/I proposta neste estudo. / Special Sensor Microwave-Imager (SSM/I) EASE-grid data and classified ASAR wideswath (WS) images, covering the Antarctic Peninsula (AP), were processed and used in a Spectral Linear Mixing Model (SLMM) for a subpixel analysis of the Wet Snow Zone (WSZ) in SSM/I images. The components’ proportions (fraction images) of the endmembers in the study area, namely WSZ, Dry Snow Zone and rock outcrops, were derived from classified ASAR images. These fraction images and co-registered SSM/I images (bands 19H, 19V, 37H and 37V), acquired on the same date, were used in the SLMM to estimate the unknown spectral signatures (i.e., brightness temperature on each SSM/I band). These spectral signatures were used to estimate WSZ fraction images, which were compared with the ASAR fraction images, by calculating the correlation coefficients. This work identified two spectral signatures that produced the most correlated data, and determined the WSZ fraction images correlations resulting from the use, in the SLMM, of the mean and median values of the most similar spectral signatures. The median values of these spectral signatures produced the most correlated WSZ fraction images, which had an average overall classification accuracy (AOCA) of 95.6% and 97.3% for spring and autumn fraction images, respectively (class range of 0.1), and an AOCA of 72.6% for summer fraction images (class range of 0.2). These median spectral signatures were then used in a SLMM to estimate accurately the WSZ intensity and its extension on the AP, by using calibrated SSM/I and SMMR (Scanning Multichannel Microwave Radiometer) imageries, allowing a daily subpixel analysis of this glacier facie on the AP from 1978 to 2008. Based on the spatial analysis of the WSZ fraction images, it was observed that melt primarily takes place in late October and ends in late March, with peak on January 7th (about 172,237 km2 or 31,6% of the AP area). The WSZ median total area in summer was about 105,100 km2. Regression analysis over the 1978-1979 to 2007-2008 summers, revealed a negative interanual trend in surface melt of 330.854 km2. Nevertheless, this trend inference is not statistically significant, due to the high WSZ interanual variability. Extremely high melt occurred in the 1984-1985 (176,507.289 km2) and 1989-1990 (172,681.867 km2) summers, while extremely weak melt occurred in the 1993-1994 (26,392.208 km2) and 1981-1982 (23,244.341 km2) summers. The most persistent and intensive melt was observed on Larsen, Wilkins, George VI and Wordie ice shelves and it was related to the break-up and disintegration events that occurred on these glaciers in the last decades. Surface melting is closely related to the stability of the Antarctic glacial system and global sea level changes. It could be monitored for the whole Antarctica, by using the WSZ subpixel analysis in SMMR and SSM/I imageries proposed by this study.

Dados EASE-grid do Special Sensor Microwave-Imager (SSM/I) e imagens classificadas ASAR wideswath (WS), cobrindo a Península Antártica (PA), foram processadas e usadas em um Modelo Linear de Mistura Espectral (MLME), para a análise subpixel da Zona Superficial de Neve Úmida (ZSNU) em imagens SSM/I. As proporções dos componentes puros (imagens-fração) da área de estudo (ZSNU, Zona Superficial de Neve Seca e rochas) foram derivadas das imagens ASAR classificadas. As imagens-fração e imagens SSM/I co-registradas de mesma data (bandas 19H, 19V, 37H e 37V) foram usadas no MLME para estimar as assinaturas espectrais desconhecidas (i.e., temperatura de brilho em cada banda SSM/I). Essas assinaturas espectrais foram então usadas no MLME para estimar as imagens-fração da ZSNU, as quais foram comparadas com as imagens-fração ASAR correspondentes, por meio do cálculo do coeficiente de correlação. Foram identificadas as duas assinaturas espectrais que resultaram nos dados mais correlacionados, sendo também calculadas as correlações das imagens-fração da ZSNU resultantes do uso no MLME dos valores médio e mediano das assinaturas espectrais mais similares. Os valores medianos dessas assinaturas espectrais produziram as imagens-fração da ZSNU mais correlacionadas, que tiveram uma precisão global de classificação média (PGCM) de 95,6% e 97,3%, nas imagens de primavera e outono, respectivamente (amplitude de classes de 0,1), e uma PGCM de 72,6% nas imagens de verão (amplitude de classes de 0,2). Essas assinaturas espectrais medianas foram então usadas no MLME para estimar, com esses níveis de precisão global, a intensidade e extensão da ZSNU na PA, pelo uso de imagens calibradas SSM/I e SMMR (Scanning Multichannel Microwave Radiometer), possibilitando assim a análise diária e em nível subpixel dessa fácie superficial, de 1978 a 2008. Na análise espacial das imagens-fração da ZSNU estimadas, observou-se que o derretimento superficial médio começava no final de outubro e terminava no final de março, com auge em 7 de janeiro (cerca de 172.237 km2 ou 31,6% da área da PA). A área total mediana da ZSNU no verão foi de aproximadamente 105.100 km2. A análise de regressão com as imagens-fração dos verões entre 1978-1979 a 2007-2008 revelou a tendência de redução da área da ZSNU, totalizando 330,854 km2 nesse período. Todavia, essa tendência não é estatisticamente significante, devido à alta variabilidade interanual da área da ZSNU na PA. Forte derretimento superficial ocorreu nos verões de 1984-1985 (176.507,289 km2) e 1989-1990 (172.681,867 km2), enquanto fraco derretimento, nos verões de 1993-1994 (26.392,208 km2) e 1981-1982 (23.244,341 km2). O mais persistente e intenso derretimento superficial foi observado nas plataformas de gelo Larsen, Wilkins, George VI e Wordie e isto foi relacionado com os eventos de fragmentação e desintegração dessas massas de gelo, ocorridos nas últimas décadas. O derretimento superficial está intimamente relacionado com a estabilidade do sistema glacial antártico e com mudanças no nível médio dos mares. Esse poderia ser monitorado em toda a Antártica, por meio da análise subpixel de imagens SMMR e SSM/I proposta neste estudo. / Special Sensor Microwave-Imager (SSM/I) EASE-grid data and classified ASAR wideswath (WS) images, covering the Antarctic Peninsula (AP), were processed and used in a Spectral Linear Mixing Model (SLMM) for a subpixel analysis of the Wet Snow Zone (WSZ) in SSM/I images. The components’ proportions (fraction images) of the endmembers in the study area, namely WSZ, Dry Snow Zone and rock outcrops, were derived from classified ASAR images. These fraction images and co-registered SSM/I images (bands 19H, 19V, 37H and 37V), acquired on the same date, were used in the SLMM to estimate the unknown spectral signatures (i.e., brightness temperature on each SSM/I band). These spectral signatures were used to estimate WSZ fraction images, which were compared with the ASAR fraction images, by calculating the correlation coefficients. This work identified two spectral signatures that produced the most correlated data, and determined the WSZ fraction images correlations resulting from the use, in the SLMM, of the mean and median values of the most similar spectral signatures. The median values of these spectral signatures produced the most correlated WSZ fraction images, which had an average overall classification accuracy (AOCA) of 95.6% and 97.3% for spring and autumn fraction images, respectively (class range of 0.1), and an AOCA of 72.6% for summer fraction images (class range of 0.2). These median spectral signatures were then used in a SLMM to estimate accurately the WSZ intensity and its extension on the AP, by using calibrated SSM/I and SMMR (Scanning Multichannel Microwave Radiometer) imageries, allowing a daily subpixel analysis of this glacier facie on the AP from 1978 to 2008. Based on the spatial analysis of the WSZ fraction images, it was observed that melt primarily takes place in late October and ends in late March, with peak on January 7th (about 172,237 km2 or 31,6% of the AP area). The WSZ median total area in summer was about 105,100 km2. Regression analysis over the 1978-1979 to 2007-2008 summers, revealed a negative interanual trend in surface melt of 330.854 km2. Nevertheless, this trend inference is not statistically significant, due to the high WSZ interanual variability. Extremely high melt occurred in the 1984-1985 (176,507.289 km2) and 1989-1990 (172,681.867 km2) summers, while extremely weak melt occurred in the 1993-1994 (26,392.208 km2) and 1981-1982 (23,244.341 km2) summers. The most persistent and intensive melt was observed on Larsen, Wilkins, George VI and Wordie ice shelves and it was related to the break-up and disintegration events that occurred on these glaciers in the last decades. Surface melting is closely related to the stability of the Antarctic glacial system and global sea level changes. It could be monitored for the whole Antarctica, by using the WSZ subpixel analysis in SMMR and SSM/I imageries proposed by this study.

Nitrogen (N) is an essential element to plants for growth, maintenance and reproduction, however most N does not exist in a form that is biologically available to plants. In order to maximise the acquisition and retention of N, plants have evolved a variety of morphological and physiological adaptations and life history strategies, as well as the ability to respond plastically to changes in resource availability in ecological time. Determining the ecophysiological responses of plants to changes in root N availability is crucial to further understanding of the mechanisms underlying competitive interactions between plants, and between plants and other organisms, that ultimately contribute to community structure and ecosystem functioning. Carnivorous plants are ideal systems for investigating ecophysiological responses to N availability as:- (i) they share a unique adaptation for obtaining supplemental N from captured prey, therefore ecological stoichiometry and energetic cost/benefit models may be explored; (ii) the trait of botanical carnivory is widely considered to have independently co-evolved as a response to N-deficient, sunny and wet environments, therefore resource allocation trade-offs between plant investment in N and carbon (C) acquisition may be observed, and (iii) they are extremely sensitive to changes in root N availability in ecological time. In this research, the carnivorous plant Drosera rotundifolia (round-leaved sundew) was used to address several unanswered ecophysiological and evolutionary questions relating to patterns and processes of prey capture and the N nutrition of carnivorous plants. Furthermore, the potential for reducing uncertainty in the calculation of plant reliance on carnivory using a δ15N natural abundance multi-level linear mixing model was explored. A combined approach of in-situ and ex-situ studies was employed, using co-occurring non-carnivorous plants or carnivorous plant species with differing evolutionary lineages or prey capture mechanisms respectively to provide context. Results show that the adaptations of carnivory, high reproductive investment and a relatively short life span enable Drosera rotundifolia to survive and thrive in an extreme, N deficient environment. Phenotypically plastic responses by the plant to light and root N availability provide evidence of resource allocation trade-offs between investment in carnivory for N acquisition and in photosynthesis for C acquisition. Plants invested less heavily in prey capture (measured as the stickiness of leaf mucilage) as N availability increased or light availability decreased. These results show that the energetic costs associated with carnivory are avoided by the plant when less costly sources of N are available for uptake and that the production of carbon-rich mucilage is only made under nutrient-limited and well-lit conditions. Results obtained from the comparison of captured insect prey with background invertebrates of potential prey indicate that Drosera rotundifolia is a dietary generalist, where the quantity of prey captured per plant is positively correlated with leaf stickiness and total leaf area. Plant reliance on prey-derived N decreased with increasing root N availability, providing evidence that carnivory is only of net benefit to the plant in N-deficient and well-lit environments, as the photosynthetic costs of investment in the trait are not exceeded by the energetic gain from prey N uptake in shady or dry habitats. A more accurate and precise method for calculating plant reliance on botanical carnivory is presented which incorporates the insect diet of the plant. This method has wider significance for reducing uncertainty in the calculation of relative source contributions to a mixture for most natural abundance applications using a multi-level linear mixing model. To conclude, results from this research further understanding of the ecophysiological mechanisms underlying plant responses to changes in resource availability and the selective pressures driving the evolution of plant adaptations. These results therefore assist with predicting how plants and plant communities may respond to sustained N deposition inputs and future environmental scenarios.

Submitted by Rosivalda Pereira (mrs.pereira@ufma.br) on 2017-06-19T18:59:30Z No. of bitstreams: 1 EvanildoLima.pdf: 4481585 bytes, checksum: 8fa0d16613f524e7afcc180f1e362073 (MD5) / Made available in DSpace on 2017-06-19T18:59:30Z (GMT). No. of bitstreams: 1 EvanildoLima.pdf: 4481585 bytes, checksum: 8fa0d16613f524e7afcc180f1e362073 (MD5) Previous issue date: 2015-06-30 / Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) / In Maranhão (northeast of Brazil), the “quilombolas”, African slaves descendants, match a significant part of state traditional communities. Although there are studies about them, few papersconsider and relate the pressure of this communities on the ecosystems. This study, using satellite remote sensing tools and ecology landscape elements, aimed to realize the forest cover percentage in different center of distances of the community area in the twenty "quilombolas" sites of Maranhão in 2006. The central hypothesis of this study be progressive decrease in vegetation clearance from the center of the community (negative correlation between distance and percentage of forest cover), however, was not confirmed (p = 0.272). The high variation observed in the forest cover of the different communities is a significant indication of the great variability in factors involved in this type of approach, which would involve a large number of replicates the data to minimize these differences. Data obtained through a Power analysis indicated that better results would be obtained at a distance of 10km from the center of community. However, the average coverage observed (38.5%) is within the range expected from forest cover to rural areas, higher in urban areas. In addition, five communities had more than 50% of that class in the first kilometer (SJB, FER, PAL, SI and FAI), while in three it was observed under 20% in the same range (COC COT and CUP). Thus, further studies are needed, not only using other image analysis routines and other landscape indexes, but also investigating the history of communities and their social aspects, as well as their relationship with other social actors in the area. / No Maranhão, os "quilombolas", descendentes de escravos africanos, correspondem a uma parte significativa das comunidades tradicionais do estado. Embora existam estudos sobre eles, poucos trabalhos consideram e relacionam a pressão dessas comunidades sobre os ecossistemas. Este estudo, usando ferramentas de sensoriamento remoto e elementos da ecologia de paisagens, teve o objetivo de analisar o percentual de cobertura florestal a diferentes distâncias do centro da comunidade em vinte áreas quilombolas do Maranhão, no ano de 2006. A hipótese central desse trabalho de haver diminuição progressiva na vegetação com o afastamento do centro das comunidades (correlação negativa entre distância e porcentagem cobertura florestal), entretanto, não foi corroborada (p=0.272). A alta variação observada na cobertura florestal das diferentes comunidades é um indicativo expressivo da grande variabilidade nos fatores envolvidos neste tipo de abordagem, o que implicaria em um grande número de replicações nos dados para minimizar estas diferenças. Dados obtidos através de uma Power análise indicaram que resultados mais consistentes seriam obtidos a uma distância de 10km do centro da comunidade. Entretanto, a cobertura média observada (38,5%) está dentro dos valores esperados de cobertura florestal para áreas rurais, superior às zonas urbanas. Além disso, cinco comunidades apresentaram mais de 50% dessa classe no primeiro quilômetro (SJB, FER, PAL, SI e FAI), enquanto que em apenas três observou-se menos de 20% no mesmo intervalo (COC, COT e CUP). Dessa forma, são necessários mais estudos, não apenas utilizando outras rotinas de análise de imagem e outros índices de paisagem, mas também investigando a história das comunidades e os seus aspectos sociais, bem como, a sua relação com outros atores sociais na área.

Ces travaux de thèse concernent le développement de nouvelles méthodes de séparation aveugle de mélanges linéaires instantanés pour des applications à des données hyperspectrales en astrophysique. Nous avons proposé trois approches pour effectuer la séparation des données. Une première contribution est fondée sur l'hybridation de deux méthodes existantes de séparation aveugle de source (SAS) : la méthode SpaceCORR nécessitant une hypothèse de parcimonie et une méthode de factorisation en matrices non négatives (NMF). Nous montrons que l'utilisation des résultats de SpaceCORR pour initialiser la NMF permet d'améliorer les performances des méthodes utilisées seules. Nous avons ensuite proposé une première méthode originale permettant de relâcher la contrainte de parcimonie de SpaceCORR. La méthode MASS (pour \textit{Maximum Angle Source Separation}) est une méthode géométrique basée sur l'extraction de pixels mono-sources pour réaliser la séparation des données. Nous avons également étudié l'hybridation de MASS avec la NMF. Enfin, nous avons proposé une seconde approche permettant de relâcher la contrainte de parcimonie de SpaceCORR. La méthode originale SIBIS (pour \textit{Subspace-Intersection Blind Identification and Separation}) est une méthode géométrique basée sur l'identification de l'intersection de sous-espaces engendrés par des régions de l'image hyperspectrale. Ces intersections permettent, sous une hypothèse faible de parcimonie, de réaliser la séparation des données. L'ensemble des approches proposées dans ces travaux ont été validées par des tests sur données simulées puis appliquées sur données réelles. Les résultats obtenus sur ces données sont très encourageants et sont comparés à ceux obtenus par des méthodes de la littérature. / This thesis deals with the development of new blind separation methods for linear instantaneous mixtures applicable to astrophysical hyperspectral data sets. We propose three approaches to perform data separation. A first contribution is based on hybridization of two existing blind source separation (BSS) methods: the SpaceCORR method, requiring a sparsity assumption, and a non-negative matrix factorization (NMF) method. We show that using SpaceCORR results to initialize the NMF improves the performance of the methods used alone. We then proposed a first original method to relax the sparsity constraint of SpaceCORR. The method called MASS (Maximum Angle Source Separation) is a geometric method based on the extraction of single-source pixels to achieve the separation of data. We also studied the hybridization of MASS with the NMF. Finally, we proposed an approach to relax the sparsity constraint of SpaceCORR. The original method called SIBIS (Subspace-Intersection Blind Identification and Separation) is a geometric method based on the identification of intersections of subspaces generated by regions of the hyperspectral image. Under a sparsity assumption, these intersections allow one to achieve the separation of the data. The approaches proposed in this manuscript have been validated by experimentations on simulated data and then applied to real data. The results obtained on our data are very encouraging and are compared with those obtained by methods from the literature.