Seasonal monitoring of atmospheric constituents using multi-angle MODIS data as support for atmospheric correction in the Amazon region

• Ano de defesa: 2017
• Autor(a) principal: Vitor Souza Martins
• Orientador(a): Cláudio Clemente Faria Barbosa, Lino Augusto Sander de Carvalho
• Banca de defesa: Yosio Edemir Shimabukuro, Mauro Antonio Homem Antunes
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Instituto Nacional de Pesquisas Espaciais (INPE)
• Programa de Pós-Graduação: Programa de Pós-Graduação do INPE em Sensoriamento Remoto
• País: BR
• Link de acesso: http://urlib.net/sid.inpe.br/mtc-m21b/2017/06.10.13.55

Satellite data provide the only viable means for systematic monitoring of remote and large ecosystem, such as Amazon. However, atmospheric attenuation distorts optical remote sensing measurements, and therefore, accurate atmospheric correction (A/C) is a key requirement for retrieving reliable surface reflectance (R$_{sup}$). In this sense, the knowledge of the seasonal patterns of cloud cover and atmospheric constituents is essential for remote sensing applications. Multi-angle Implementation of Atmospheric Correction (MAIAC) is a new Moderate Resolution Imaging Spectroradiometer (MODIS) algorithm that combines time series approach and image processing to derive surface reflectance and atmosphere products, such as aerosol optical depth (AOD), columnar water vapor, and cloud mask. In this research, three main analyses were performed: (i) validation of MAIAC AOD retrievals using ground-data from 19 AERONET sites in the South America; (ii) evaluation of seasonal pattern of cloud cover and key atmospheric constituents over the Amazon basin; and (iii) assessment of AC methods (6SV, ACOLITE and Sen2Cor) applied to MultiSpectral Instrument (MSI) Sentinel-2 image over Amazon floodplain lakes. In the first analysis (i), MAIAC AOD Terra/Aqua retrievals showed high agreement with ground-based AERONET measurements, with correlation coefficient (R) close to unity (R$_{terra}$: 0.956 and R$_{Aqua}$: 0.949). However, MAIAC accuracy varies with land cover type, and comparisons revealed a high fitness for cropland, forest, savanna and grassland covers, with more than 66\% of retrievals within the expected error ($\Delta$4AOD=$\pm$0.05$^{\ast}$AOD$\pm$0.05) and R exceeding 0.8 for both Terra and Aqua products. Over bright surfaces, however, MAIAC retrievals showed lower correlation than those of vegetated areas, and overestimated retrievals over shrubland and barren areas. In the second analysis (ii), the seasonal pattern of cloud cover and key atmospheric constituents presented clear distinction amongst four Amazon regions, with relative high (low) cloud cover and low (high) atmospheric loading during wet (dry) season, exception for water vapor content. The sub-basin analysis showed that Negro and Caqueta-Japurá sub-basins are under quasi-constant cloud cover (80-100\%) throughout the year, while High-Madeira and Tapajos present a cloudiness regime during dry season. For the temporal analysis, drought years present the most critical regimes of aerosol loading, with a peak in September. In the last analysis (iii), A/C results of the MSI visible bands illustrate the limitation of the methods over dark lakes (R$_{sup}$ < 1\%), and a better match of the Rsup shape compared with in-situ measurements over turbid lakes, although the accuracy varied depending on the spectral bands and methods. Particularly above 705 nm, R$_{sup}$ was highly affected by adjacent effects of forest, and the proposed adjacency effect correction minimized the spectral distortions in R$_{sup}$ (RMSE < 0.006). In conclusion, the availability of multi-angle MODIS products contributes with consistent information for both analyses of seasonal constituents and atmospheric correction, what opens a new endeavour for remote sensing studies over Amazon basin. Particularly for inland water, future studies should be focused on distinct surface-atmosphere conditions to assess the quality of these A/C methods.