Submitted by ULISSES SILVA GUIMARÃES null (ulissesguimaraes21@yahoo.com.br) on 2017-04-23T14:46:14Z No. of bitstreams: 1 PPGCC_Tese_Guimaraes_mar2017.pdf: 11881192 bytes, checksum: b616aff4dd851b26e6151f6a753a3e62 (MD5) / Rejected by Luiz Galeffi (luizgaleffi@gmail.com), reason: Solicitamos que realize uma nova submissão seguindo a orientação abaixo: O arquivo submetido não contém o certificado de aprovação. Corrija esta informação e realize uma nova submissão com o arquivo correto. Agradecemos a compreensão. on 2017-04-25T19:47:14Z (GMT) / Submitted by ULISSES SILVA GUIMARÃES null (ulissesguimaraes21@yahoo.com.br) on 2017-04-27T00:35:55Z No. of bitstreams: 1 PPGCC_DR_UG.pdf: 11945196 bytes, checksum: 6a00a3b24776e561994802637366de42 (MD5) / Approved for entry into archive by Luiz Galeffi (luizgaleffi@gmail.com) on 2017-05-03T13:21:12Z (GMT) No. of bitstreams: 1 guimaraes_us_dr_prud.pdf: 11945196 bytes, checksum: 6a00a3b24776e561994802637366de42 (MD5) / Made available in DSpace on 2017-05-03T13:21:12Z (GMT). No. of bitstreams: 1 guimaraes_us_dr_prud.pdf: 11945196 bytes, checksum: 6a00a3b24776e561994802637366de42 (MD5) Previous issue date: 2017-03-06 / A Zona Costeira Amazônica (ZCA) é marcada por uma alta descarga de sedimentos e água doce sob a influência do rio Amazonas, possui ampla plataforma continental, extensas planícies de inundação e planaltos rebaixados. É uma região de clima tropical, caracterizando-se por chuvas e nebulosidade severas, além da influência de macromarés. Este estudo propõe-se a avaliar a precisão de Modelos Digitais de Superfícies (MDSs), elaborados a partir de dados de radar de abertura sintética (SAR) Cosmo-SkyMed (CSK) e TerraSAR-X (TSX), utilizando abordagens de reconstrução tridimensional por interferometria e radargrametria, para caracterizar esse relevo plano e dinâmico da costa amazônica. O estudo foi desenvolvido em quatro experimentos contemplando: i) as variações de linha de costa por meio de detecção de mudanças a partir de imagens ópticas; ii) mapeamento de ambientes costeiros; iii) elaboração e análise de MDSs interferométricos e iv) radargramétricos, por meio das suas respectivas cadeias de processamento SAR. A ZCA teve forte dinâmica nos últimos 15 anos, com acresção total de 5.582,18 km2 e sob a taxa de 372,15 km2.ano-1, erosão total de 5.475,90 km² e sob taxa de 365,06 km2.ano-1, resultando no balanço sedimentar de 106,27 km², com taxa de 7,08 km2.ano-1. O setor Insular Estuarino apresentou a maior dinâmica de linha de costa, com mudanças costeiras de 213,17±56,46 km2 e balanço sedimentar de 20,65±73,59 km2. Os ambientes costeiros amazônicos foram descritos pelo retroespalhamento e pela coerência, os quais compartilharam alta ambiguidade e dispersão elevada, sendo o pior caso de separabilidade e baixa coerência registrado para Planície Costeira. O mapeamento dos diferentes ambientes costeiros resultou em coeficiente Kappa entre 0,46 a 0,51, apontando os ângulos de incidência rasantes e o período seco como mais apropriados para o estudo. Os MDSs interferométricos e radargramétricos foram elaborados em passagens múltiplas de única revisita com compromissos entre ângulos de incidência, linha de base espacial e descorrelação temporal. A acurácia vertical foi realizada por testes estatísticos pareados com levantamentos de campo que resultou em discrepâncias, viés e precisão compatíveis com o Padrão de Exatidão Cartográfica Brasileiro para Produtos Cartográficos Digitais (PEC-PCD), em adição, os MDS foram comparados por meio dos diagramas de Taylor e Alvo. Os MDSs interferométricos alcançaram RMSE entre 9,57 e 25,18 m, com melhor desempenho para o MDS CSK, adquirido com 1 dia de revisita, ângulo de incidência íngreme, no período chuvoso e compatível a escala de 1:50.000, classe A. Entretanto, a abordagem interferométrica não foi capaz de solucionar a reconstrução tridimensional de ambientes que se mostraram incoerentes. Os modelos radargramétricos obtidos pelas abordagens do SARscape e Toutin alcançaram RMSE entre 4,34 e 7,76 m, com melhor desempenho para os modelos de Toutin, que foram compatíveis com a escala 1:50.000, classe A. A radargrametria permitiu a reconstrução tridimensional contínua, incluindo a Planície Costeira de comportamento incoerente. A comparação dos MDSs por meio dos diagramas de Taylor e Alvo, mostrou variações de precisão entre os sistemas CSK e TSX, e suas respectivas condições de aquisição e modelos, com destaque à menor variabilidade e ajuste da correlação para MDSs do sistema TSX, em incidências rasantes, no período seco e gerados pelo modelo de Toutin. O Tabuleiro Costeiro e Terraço Fluviomarinho apresentaram menor erro vertical entre 3,89 e 28,59 m, e entre 3,79 e 20,33 m, respectivamente, enquanto que a Planície Costeira teve maior RMSE entre 4,16 e 26,24 m. O Tabuleiro Costeiro foi o ambiente costeiro mais adequado para estimar altura, com posições plotadas próximas as referências de campo. Os dados CSK e TSX permitiram mapear o relevo plano e dinâmico da ZCA, por meio da banda X, alta resolução espacial e revisita, o que demostrou o suporte para cartografar em detalhe de escala espacial (1:50.000) e frequente atualização (semestral a anual). / The Amazon Coastal Zone (ACZ) is marked by a high discharge of sediments and fresh water under the influence of the Amazon River, which has a wide continental shelf, extensive flood plain and lowered plateaus. It is a region of tropical climate, rainfall, severe cloudiness and macrotidal influence. This study proposes to assess the performance of Digital Surface Models (DSM) based on Cosmo-SkyMed (CSK) and TerraSAR-X (TSX) dataset, using threedimensional reconstruction by interferometry and radargrammetry approaches on the flat and dynamic relief of Amazonian coast. The method focused on four experiments: i) shoreline variations through change detection of optical images, ii) mapping of coastal environments; iii) elaboration and analyses of interferometric DMSs and iv) radargrammetric through their respective Synthetic Aperture Radar (SAR) processing chains. The ACZ had high dynamic in the last 15 years with total sediment deposition of 5,582.18 km2 and under a rate of 372.15 km2 .yr -1 , and with total erosion of 5,475.90 km² and under a rate of 365.06 km2 .yr-1 . Besides, it was obtained a sedimentary balance of 106.27 km² and under a rate of 7.08 km2 .yr-1 . The Estuarine Insular sector presented the greater dynamics of shoreline, registering coastal changes of 213.17 ± 56.46 km2 and sedimentary balance of 20.65 ± 73.59 km2 . The Amazonian coastal environments were described by backscattering and coherence which shared ambiguity and high dispersion, with the lowest separability and coherence noted for Coastal Flat. The mapping of the coastal environments obtained Kappa coefficients between 0.46 and 0.51, indicating the shallow incidence angles during the dry season as more appropriated for the study. The interferometric and radargrammetric DSMs were elaborated in multi-pass and single revisit with commitment between incidence angles, spatial baseline and temporal decorrelation. A vertical accuracy assessment was performed with paired statistical tests at surveyed elevations in the field that resulted in discrepancies, bias and precision, in accordance to the Brazilian Cartographic Accuracy Standard for Digital Cartographic Products (PEC-PCD), in addition, the DSMs were compared throughout Taylor and Target diagrams. The interferometric DMSs achieved RMSE between 9.57 and 25.18 m, with better performance for the DMS CSK, acquired with 1 day of revisit, steeper incidence, in rainy season and compatible at a scale of 1: 50,000, class A. However, the interferometric approach was not able to solve the threedimensional reconstruction in incoherent environments. The radargrammetric models of SARscape and Toutin achieved a RMSE between 4.34 and 7.76, and the best performances were for the Toutin’s models compatible at a scale of 1: 50,000, class A. The advantage of radargrammetry was to provide continuous three-dimensional reconstruction, including the Coastal Flat of incoherent behavior. The comparison of the DMSs through the Taylor and Target diagrams showed fluctuations of precision between CSK and TSX systems and their respective acquisition conditions and models, but it is remarkable the stability of the lowest variability and the correlation well fitted for the DMSs given by TSX system, shallow incidences, dry season and Toutin model. The Coastal Plateau and Fluvial Marine Terrace had the lowest vertical error between 3.89 and 28.59 m, and between 3.79 and 20.33 m, respectively. On the contrary, the Coastal Flat had the highest RMSE between 4.16 and 26.24 m. The Coastal Plateau was the most suitable coastal environment to estimate the height following the Taylor and Target diagrams, with the plotted positions close to the field references. The CSK and TSX data allowed to map the ZCA precisely, based on X-band perspective, high spatial resolution and revisit, which has demonstrated the support for detailed cartography of spatial scale (1: 50,000) and frequent updating (semiannual up to annual).

Résumé : Le contexte récent d’accélération de la fonte des glaciers et calottes glaciaires (GCG) de l’archipel arctique canadien, jumelé aux difficultés de suivi des GCG de cette région, rendent essentiels le développement et l’utilisation de nouvelles approches innovatrices de suivi. Le potentiel de la stéréo radargrammétrie (SRG) RADARSAT-2 est ici caractérisé pour l’extraction d’élévations et le calcul de changements d’élévation et de bilans de masse (historiques et récents) sur les calottes glaciaires Barnes et Penny (Nunavut, Canada). Par la méthode semi-automatisée de recherche de corrélation à partir de couples stéréoscopiques RADARSAT-2 de 2013 (mode wide ultra-fin; résolution spatiale de 3 m; taille d’image de 50 km x 50 km), une précision verticale de ~7 m (LE68) est mesurée sur la terre ferme, et cette valeur de précision est possiblement légèrement supérieure sur la calotte Barnes, étant donné la variabilité de profondeur de pénétration. Par captage 3D, une précision altimétrique de ~3-4 m (LE68) est mesurée par différents photo-interprètes à partir de couples RADARSAT de 2012 en zone d’ablation de la calotte Penny. Sur la calotte Barnes, les changements d’élévation mesurés par rapport aux premiers modèles numériques de terrain disponibles permettent de mesurer un bilan de masse spécifique historique (1960-2013) de -0,49 ± 0,20 m w.e./année, pour un bilan de masse total de -2,9 Gt/année. Entre 2005 et 2013, le bilan de masse spécifique de cette calotte augmente significativement à -1,20 ± 0,86 m w.e./année, pour un bilan de masse total de -7 Gt/année. En zone d’ablation de la calotte Penny, un changement d’élévation annuel moyen de -0,59 m/année est mesuré entre 1958 et 2012. Parallèlement, plusieurs aspects méthodologiques et techniques sont discutés et analysés. Des profondeurs de pénétration nulles (bande C) sont mesurées à partir des images acquises sur la calotte Barnes à la toute fin de la saison d’ablation (fin septembre/début octobre), alors que cette profondeur augmente à ~2,5-3 m pour des images acquises à la fin octobre/début novembre (période de gel). Nos résultats suggèrent aussi que le modèle de fonction rationnelle, lorsqu’utilisé avec des images RADARSAT-2 en mode wide ultra-fin, permet d’obtenir des précisions plus constantes que le modèle hybride de Toutin. De par son indépendance des conditions météorologiques, son utilisation possible sans point de contrôle et sa simplicité de traitement, la SRG RADARSAT-2 s’avère donc être une excellente alternative aux technologies actuelles pour le suivi de GCG situés dans des régions affectées par des contraintes opérationnelles importantes. / Abstract : Given the recent melt acceleration of the Canadian arctic archipelago’s ice caps and the monitoring difficulties of this remote region, the development of new innovative monitoring tools has become essential. Here, the potential of the RADARSAT-2 stereo radargrammetry (SRG) is characterized for elevations extraction, as well as for elevation changes/mass balances calculations (historical and recent) on Barnes and Penny ice caps (Nunavut, Canada). Using the semi-automatic approach of correlation search from RADARSAT-2 stereoscopic couples of 2013 (wide ultra-fine mode; spatial resolution of 3 m; coverage of 50 km x 50 km), a vertical precision of ~7 m (LE68) is measured on ice-free terrain and this precision is possibly slighty worse on the ice cap because of the penetration depth’s variability. On the other hand, the 3D vision extraction approach reveals an altimetric precision of ~3-4 m (LE68) on the ablation area of the Penny Ice Cap. On the Barnes Ice Cap, elevation changes calculated relative to the oldest digital elevation models available allows to calculate an historical specific mass balance (1960-2013) of -0,49 ± 0,20 m w.e./year, resulting in a total annual mass balance of -2,9 Gt/year. Between 2005 and 2013, the specific mass balance of this ice cap increases to -1,20 ± 0,86 m w.e./year, which equals to a total annual mass balance f -7 Gt/year. On Penny Ice Cap’s ablation area, an average elevation change of -0,59 m/year is measured between 1958 and 2012. As also suggested in the literature, the recent melt acceleration is highly linked to warmer summer temperatures. Methodological and technical aspects are also presented and analyzed. No penetration depth (C band) is perceived on elevations derived from late ablation season images (late September/beginning of October), while a penetration of ~2,5-3 m is measured from images acquired in late October/beginning of November (freeze period). Our results also suggest the superiority and better consistency of the rational function model for geometrical correction of wide ultra-fine mode RADARSAT-2 images, compared to the hybrid Toutin’s model. Because of its all-weather functionality, its possible use without any ground control point and the simplicity and facility of its treatment, the RADARSAT-2 SRG represents a really good technology for glacier monitoring in regions affected by serious operational constraints.