Recently, remote sensing image time series analysis has being widely used to investigate the dynamics of environments over time. Many studies have combined image time series analysis with machine learning methods to improve land use and cover change mapping. In order to support image time series analysis, analysis-ready data (ARD) image collections have been modeled and organized as multidimensional data cubes. Data cubes can be defined as sets of time series associated with spatially aligned pixels. Based on lessons learned in the research project e-Sensing, related to national demands for land use and cover monitoring and related to state-of-the-art studies on relevant topics, we define the requirements to build Earth observation data cubes for Brazil. This paper presents the methodology to generate ARD and multidimensional data cubes from remote sensing images for Brazil. We describe the computational infrastructure that we are developing in the Brazil Data Cube project, composed of software applications and Web services to create, integrate, discover, access, and process the data sets. We also present how we are producing land use and cover maps from data cubes using image time series analysis and machine learning techniques.

Tropical forests regulate the global water and carbon cycles and also host most of the world’s biodiversity. Despite their importance, they are hard to survey due to their location, extent, and particularly, their cloud coverage. Clouds hinder the spatial and radiometric correction of satellite imagery and also diminishing the useful area on each image, making it difficult to monitor land change. For this reason, our purpose is to identify the cloud detection algorithm best suited for the Amazon rainforest on Sentinel–2 images. To achieve this, we tested four cloud detection algorithms on Sentinel–2 images spread in five areas of the Amazonia. Using more than eight thousand validation points, we compared four cloud detection methods: Fmask 4, MAJA, Sen2Cor, and s2cloudless. Our results point out that FMask 4 has the best overall accuracy on images of the Amazon region (90%), followed by Sen2Cor’s (79%), MAJA (69%), and S2cloudless (52%). We note the choice of method depends on the intended use. Since MAJA reduces the number of false positives by design, users that aim to improve the producer’s accuracy should consider its use.

The Brazilian National Institute for Space Research (INPE) produces official information about deforestation as well as land use and cover in the country, based on remote sensing images. The current open data policy adopted by many space agencies and governments worldwide provided access to petabytes of remote sensing images. To properly deal with this vast amount of images, novel technologies have been proposed and developed based on cloud computing and big data systems. This paper describes the INPE’s initiatives in using remote sensing images and cloud services of the Amazon Web Services (AWS) infrastructure to improve land use and cover monitoring.

Accurate and consistent Surface Reflectance estimation from optical remote sensor observations is directly dependant on the used atmospheric correction processor and the differences caused by it may have implications on further processes, e.g. classification. Brazil is a continental scale country with different biomes. Recently, new initiatives, as the Brazil Data Cube Project, are emerging and using free and open data policy data, more specifically medium spatial resolution sensor images, to create image data cubes and classify the Brazilian territory crops. For this reason, the purpose of this study is to verify, on Landsat-8 and Sentinel-2 images for the Brazilian territory, the suitability of the atmospheric correction processors maintained by their image providers, LaSRC from USGS and Sen2cor from ESA, respectively. To achieve this, we tested the surface reflectance products from Landsat-8 processed through LaSRC and Sentinel-2 processed through LaSRC and Sen2cor comparing to a reference dataset computed by ARCSI and AERONET. The obtained results point that Landsat-8/OLI images atmospherically corrected using the LaSRC corrector are consistent to the surface reflectance reference and other atmospheric correction processors studies, while for Sentinel-2/MSI images, Sen2cor performed best. Although corrections over Sentinel-2/MSI data weren’t as consistent as in Landsat-8/OLI corrections, in comparison to the surface reflectance references, most of the spectral bands achieved acceptable APU results.

Clouds and cloud shadows significantly impact optical remote sensing. Combining images from different sources can help to obtain more frequent time series of the Earth’s surface. Nevertheless, sensor differences must be accounted for and treated before combining images from multiple sensors. Even after geometric correction, inter-calibration, and bandpass, disparities in image measurements can persist. One potential factor contributing to this phenomenon is directional effects. Bidirectional reflectance distribution function (BRDF) corrections have emerged as an optional processing method to soften differences in surface reflectance (SR) measurements, where the c-factor is one of the available options for this task. The c-factor efficiency is well-proven for medium spatial resolution products. However, its use should be restricted to images from sensors with a narrow view since it causes subtle changes in the processed images. There are currently a limited number of open tools for users to independently process their images. Here, we implemented the required tools to generate a Nadir BRDF-Adjusted Surface Reflectance (NBAR) product through the c-factor approach, and we evaluated them for a study area using Landsat-8 and Sentinel-2 images. Several comparisons were conducted to verify the SR and NBAR differences. Initially, a single-sensor approach was adopted and later a multi-source approach. Notably, NBAR products exhibit fewer disparities compared to SR products (prior to BRDF corrections). The results reinforce that the c-factor can be used to improve time series compatibility and, most importantly, provide the tools to allow users to generate the NBAR products themselves.

Dados derivados de sensoriamento remoto aliados ao uso de técnicas de classificação digital de imagens fornecem uma visão sinóptica e informações sobre a dimensão temporal dos fenômenos espaciais, possibilitando gerar informações sobre a dinâmica e os padrões espaciais da paisagem em áreas de grandes extensões territoriais como a Amazônia. Nos mapeamentos extensivos, como os realizados com imagens de satélite pelo INPE para a Amazônia Legal, é comum a definição e descrição de classes de uso e cobertura da terra em função da resolução espacial e radiométrica dos sensores utilizados. Por essa razão, pequenas áreas que não atingem o tamanho mínimo da área de mapeamento e/ou com usos diversificados são, em geral, agregadas em uma única classe. Esse é o exemplo da classe denominada Mosaico de Ocupação presente na legenda do sistema de monitoramento do uso e cobertura da terra da Amazônia-TerraClass. Esta classe procura representar, em parte, a agricultura familiar, no entanto, como o mapeamento é feito pelo sensor TM ou OLI dos satélites da série Landsat, com resolução espacial de 30m e como o TerraClass define uma área mínima de mapeamento de 6,25 hectares (ha), a identificação de pequenas áreas agrícolas é comprometida, pois essas categorias são agregadas em classes mistas e não são mais distinguíveis. Para estudos que procuram dar visibilidade aos padrões espaciais de atividades de produção de pequena escala torna-se necessário realizar o refinamento dessas classes com dados de resolução espacial com melhor definição. Neste contexto, o objetivo desse trabalho é testar três algoritmos semiautomáticos de classificação, baseados em pixel e em regiões: os algoritmos MAXVER, Bhattacharya e K-Vizinho Mais Próximo (KNN), para avaliar a capacidade de refinamento da classe Mosaico de Ocupação do dado produzido pelo TerraClass-2014. A área de estudo compreende parte dos municípios de Cametá, Mocajuba e Baião, localizados na região Nordeste do Pará, onde a produção de mandioca, pimenta-do-reino, cacau e açaí têm grande importância econômica para população local. Para o mapeamento das categorias contidas na classe Mosaico de Ocupação, foram utilizadas imagens do RapidEye, sensor REIS, ortoimagens com 5m de resolução espacial. Foi estimada a exatidão global dos algoritmos testados obtendo-se índices de 26%, 38% e 78%, para os algoritmos MAXVER, Bhattacharya e K-Vizinho Mais Próximo, respectivamente. Além da maior exatidão Global (78%), o algoritmo K-Vizinho Mais Próximo apresentou melhores resultados relativo às classes de vegetação secundária, hidrografia, e pasto sujo, com mais de 90% de acerto. A classe agricultura anual de pequena escala apresentou acerto de 62%, enquanto o índice de acerto dos outros dois algoritmos testados não passou de 8%. A abordagem metodológica desenvolvida demonstrou a viabilidade do uso das imagens de alta resolução espacial e de métodos semiautomáticos para a classificação de classes de uso e cobertura da terra associadas à classe Mosaico de Ocupação do TerraClass. A metodologia pode ser utilizada para complementar às atuais bases de dados existentes para a Amazônia (TerraClass, MapBiomas e IBGE), explicitando as categorias agrícolas de produção em pequena escala, dando visibilidade a esses sistemas de produção, frequentemente negligenciados nos mapeamentos que abrangem grandes extensões territoriais.