Fechar
Metadados

@MastersThesis{Silva:2014:EsVaEs,
               author = "Silva, Ricardo Dal'Agnol da",
                title = "Estudo das varia{\c{c}}{\~o}es espectrais e texturais em 
                         florestas prim{\'a}rias e sucess{\~o}es secund{\'a}rias na 
                         Flona Tapaj{\'o}s usando dados ALI/EO-1",
               school = "Instituto Nacional de Pesquisas Espaciais (INPE)",
                 year = "2014",
              address = "S{\~a}o Jos{\'e} dos Campos",
                month = "2014-04-28",
             keywords = "sucess{\~o}es secund{\'a}rias, florestas tropicais, redes 
                         neurais artificiais, ALI/EO-1, textura, secondary successions, 
                         tropical forests, artificial neural networks, ALI/EO-1, texture.",
             abstract = "As sucess{\~o}es secund{\'a}rias s{\~a}o tipologias importantes 
                         no monitoramento e conserva{\c{c}}{\~a}o da biodiversidade, 
                         constituindo reservas de carbono para mitiga{\c{c}}{\~a}o de 
                         mudan{\c{c}}as clim{\'a}ticas. Com base na hip{\'o}tese de que 
                         h{\'a} um ganho no uso integrado de m{\'e}tricas de textura com 
                         dados espectrais no processo de mapeamento da cobertura da terra, 
                         especialmente das sucess{\~o}es secund{\'a}rias, os objetivos 
                         deste trabalho s{\~a}o: (1) mapear florestas prim{\'a}ria e 
                         secund{\'a}ria, al{\'e}m de outros componentes de cena na 
                         Floresta Nacional do Tapaj{\'o}s (PA) e arredores usando Redes 
                         Neurais Artificiais (RNA) aplicada a dados espectrais 
                         (reflect{\^a}ncia) e h{\'{\i}}bridos (reflect{\^a}ncia e 
                         m{\'e}tricas texturais) do sensor \emph{Advanced Land Imager} 
                         (ALI)/ \emph{Earth Observing One} (EO-1); (2) analisar as 
                         varia{\c{c}}{\~o}es espectrais, texturais e de {\'{\i}}ndices 
                         de vegeta{\c{c}}{\~a}o provenientes de diferentes est{\'a}dios 
                         de sucess{\~a}o secund{\'a}ria presentes na cena ALI/EO-1, 
                         comparando os resultados com o monitoramento de uma sucess{\~a}o 
                         secund{\'a}ria fixa com dados multi-temporais (1984-2010) do 
                         sensor \emph{Thematic Mapper} (TM)/Landsat-5; e (3) avaliar a 
                         rela{\c{c}}{\~a}o dos atributos espectrais, texturais e 
                         {\'{\i}}ndices de vegeta{\c{c}}{\~a}o, derivados dos dados 
                         ALI/EO-1, com par{\^a}metros biof{\'{\i}}sicos das tipologias 
                         florestais ({\'a}rea basal, altura m{\'e}dia, densidade de 
                         {\'a}rvores, biomassa, IAF e percentual de cobertura de dossel) 
                         obtidos de um levantamento flor{\'{\i}}stico e estrutural. Para 
                         o c{\'a}lculo dos atributos texturais, matrizes de 
                         co-ocorr{\^e}ncia de n{\'{\i}}veis de cinza (GLCM) foram 
                         utilizadas para determinar m{\'e}dia, vari{\^a}ncia, contraste, 
                         dissimilaridade, homogeneidade, correla{\c{c}}{\~a}o, segundo 
                         momento angular e entropia das diferentes bandas dos sensores ALI 
                         e TM. RNA foi utilizada tamb{\'e}m para a sele{\c{c}}{\~a}o de 
                         atributos de textura para fins de mapeamento. Os resultados 
                         obtidos mostraram uma exatid{\~a}o de mapeamento de 79\% para os 
                         dados espectrais (reflect{\^a}ncia) e 89\% para dados 
                         h{\'{\i}}bridos, compostos pelas m{\'e}tricas texturais 
                         \${''}\$m{\'e}dia\${''}\$ e 
                         \${''}\$dissimilaridade\${''}\$ e pelos dados de 
                         reflect{\^a}ncia espectral do ALI, para mapear as classes de 
                         floresta prim{\'a}ria (FP), sucess{\~o}es inicial (SS1), 
                         intermedi{\'a}ria (SS2) e avan{\c{c}}ada (SS3), pasto, culturas 
                         agr{\'{\i}}colas, solo, vegeta{\c{c}}{\~a}o 
                         n{\~a}o-fotossinteticamente ativa (NPV) e {\'a}gua. O 
                         padr{\~a}o espectral-textural observado na an{\'a}lise 
                         multi-temporal de uma {\'a}rea fixa de regenera{\c{c}}{\~a}o 
                         natural da vegeta{\c{c}}{\~a}o em cronossequ{\^e}ncia com dados 
                         TM e na an{\'a}lise de diferentes {\'a}reas de 
                         regenera{\c{c}}{\~a}o em uma data fixa com dados ALI foram 
                         consistentes entre si. Isso sugere que a influ{\^e}ncia de 
                         fatores locais sobre o desenvolvimento das sucess{\~o}es 
                         secund{\'a}rias, como por exemplo, o hist{\'o}rico do uso da 
                         terra e {\'{\i}}ndice de s{\'{\i}}tio, apesar de introduzirem 
                         variabilidade espectral-textural nos dados, n{\~a}o foram 
                         suficientes para alterar o padr{\~a}o geral observado nos dados 
                         ALI no desenvolvimento sucessional, quando comparado com os dados 
                         TM. Os atributos espectrais ALI apresentaram forte 
                         rela{\c{c}}{\~a}o com os par{\^a}metros biof{\'{\i}}sicos da 
                         floresta prim{\'a}ria e das sucess{\~o}es secund{\'a}rias, mas 
                         apenas o atributo \${''}\$textura m{\'e}dia\${''}\$ foi 
                         {\'u}til para tais estimativas. Entretanto, as 
                         correla{\c{c}}{\~o}es obtidas n{\~a}o representam 
                         necessariamente causa e efeito, pois refletem apenas a 
                         transi{\c{c}}{\~a}o das tipologias com estrutura menos 
                         desenvolvida e que apresentam alta reflect{\^a}ncia (SS1 e SS2) 
                         para as tipologias mais desenvolvidas e de menor reflect{\^a}ncia 
                         (SS3 e FP). ABSTRACT: Secondary successions are important 
                         typologies in the biodiversity monitoring and conservation, 
                         constituting carbon reserves to mitigate climate change. Based on 
                         the hypothesis that there is a gain in the integrated use of 
                         texture metrics with spectral data in the land-cover mapping, 
                         especially of secondary successions, the objectives of this study 
                         are: (1) to map primary and secondary forests as well as other 
                         land covers in the Tapajos National Forest (PA) and surroundings 
                         using Artificial Neural Networks (ANN) applied to spectral data 
                         (reflectance) and hybrid data (reflectance and textural metrics) 
                         of the Advanced Land Imager (ALI)/Earth Observing One (EO-1) 
                         sensor; (2) analyze the spectral, textural and vegetation indices 
                         variations for different secondary succession stages on ALI/EO-1 
                         data, comparing the results with the monitoring of a fixed 
                         secondary succession with multi-temporal data (1984-2010) from the 
                         Thematic Mapper (TM)/Landsat-5 sensor; and (3) assess the 
                         relationship of spectral, textural and vegetation indices 
                         attributes derived from ALI/EO-1 data with biophysical parameters 
                         of the forest typologies (basal area, average height, tree 
                         density, biomass, leaf area index and percentage of canopy cover) 
                         obtained from a floristic and structural survey. For the 
                         calculation of textural attributes, gray-level co-occurrence 
                         matrixes (GLCM) were used to determine mean, variance, contrast, 
                         dissimilarity, homogeneity, correlation, angular second moment and 
                         entropy of the different bands of ALI and TM. ANN was also used 
                         for texture attribute selection for mapping purposes. Results 
                         showed classification accuracy of 79\% using spectral data 
                         (reflectance) and of 89\% using hybrid data, which was composed 
                         by the texture metrics \${''}\$mean\${''}\$, 
                         \${''}\$dissimilarity\${''}\$ and by the ALI spectral 
                         reflectance, to map the land-cover classes of primary forest (PF), 
                         initial (SS1), intermediate (SS2) and advanced (SS3) successions, 
                         pasture, crops, soil, non-photosynthethically active vegetation 
                         (NPV) and water. The spectral and textural pattern observed in the 
                         multi-temporal analysis on a fixed area of regeneration 
                         (chronosequence) with TM data and the analysis of different areas 
                         of regeneration in a fixed date with ALI data were fully 
                         consistent to each other. This comparison suggests that the 
                         influence of other local factors on the development of secondary 
                         succession, such as the history of land use and the site index, 
                         may have introduced spectral-textural variability in the data. 
                         However, they were not sufficient to alter the general pattern 
                         observed in the ALI data in the successional development when 
                         compared to the TM images. The ALI spectral attributes showed a 
                         strong relationship with the biophysical parameters of primary 
                         forest and secondary successions, but only the metric 
                         \${''}\$mean\${''}\$ was the useful texture attribute for such 
                         estimations. However, the observed negative correlations did not 
                         necessarily mean cause and effect. They represented only the 
                         transition from typologies with less developed canopy structure 
                         and with high reflectance (SS1 and SS2) to the typologies with 
                         well-defined canopy structure and low reflectance (SS3 and PF).",
            committee = "Galv{\~a}o, L{\^e}nio Soares (presidente) and Santos, Jo{\~a}o 
                         Roberto dos (orientador) and Ponzoni, Fl{\'a}vio Jorge and 
                         Breunig, F{\'a}bio Marcelo and Liesenberg, Veraldo",
         englishtitle = "Study of the spectral and textural variations in primary forest 
                         and secondary successions at Flona Tapaj{\'o}s using ALI/EO-1 
                         data",
             language = "pt",
                pages = "133",
                  ibi = "8JMKD3MGP5W34M/3G4LUD8",
                  url = "http://urlib.net/rep/8JMKD3MGP5W34M/3G4LUD8",
           targetfile = "publicacao.pdf",
        urlaccessdate = "23 nov. 2020"
}


Fechar