Brazil has a monitoring system to track annual forest conversion in the Amazon and most recently to monitor the Cerrado biome. However, there is still a gap of annual land use and land cover (LULC) information in all Brazilian biomes in the country. Existing countrywide efforts to map land use and land cover lack regularly updates and high spatial resolution time-series data to better understand historical land use and land cover dynamics, and the subsequent impacts in the country biomes. In this study, we described a novel approach and the results achieved by a multi-disciplinary network called MapBiomas to reconstruct annual land use and land cover information between 1985 and 2017 for Brazil, based on random forest applied to Landsat archive using Google Earth Engine. We mapped five major classes: forest, non-forest natural formation, farming, non-vegetated areas, and water. These classes were broken into two sub-classification levels leading to the most comprehensive and detailed mapping for the country at a 30 m pixel resolution. The average overall accuracy of the land use and land cover time-series, based on a stratified random sample of 75,000 pixel locations, was 89% ranging from 73 to 95% in the biomes. The 33 years of LULC change data series revealed that Brazil lost 71 Mha of natural vegetation, mostly to cattle ranching and agriculture activities. Pasture expanded by 46% from 1985 to 2017, and agriculture by 172%, mostly replacing old pasture fields. We also identified that 86 Mha of the converted native vegetation was undergoing some level of regrowth. Several applications of the MapBiomas dataset are underway, suggesting that reconstructing historical land use and land cover change maps is useful for advancing the science and to guide social, economic and environmental policy decision-making processes in Brazil.

O objetivo deste trabalho foi mapear o uso da terra do Bioma Cerrado na escala de 1:250.000. As seguintes classes de uso da terra foram consideradas: culturas agrícolas, pastagens cultivadas, reflorestamentos, áreas urbanas e áreas de mineração. A metodologia envolveu a segmentação de imagens do satélite Landsat, a classificação visual dos segmentos e a análise da exatidão global do mapa final. Aproximadamente 39,5% do Cerrado apresentaram algum tipo de uso de terra. Pastagens cultivadas e culturas agrícolas foram as classes predominantes, com 26,5 e 10,5%, respectivamente.

In this study, the response of vegetation indices (VIs) to the seasonal patterns and spatial distribution of the major vegetation types encountered in the Brazilian Cerrado was investigated. The Cerrado represents the second largest biome in South America and is the most severely threatened biome as a result of rapid land conversions. Our goal was to assess the capability of VIs to effectively monitor the Cerrado and to discriminate among the major types of Cerrado vegetation. A full hydrologic year (1995) of composited AVHRR, local area coverage (LAC) data was converted to Normalized Difference Vegetation Index (NDVI) and Soil Adjusted Vegetation Index (SAVI) values. Temporal extracts were then made over the major Cerrado vegetation communities. Both the NDVI and SAVI temporal profiles corresponded well to the phenological patterns of the natural and converted vegetation formations and depicted three major categories encompassing the savanna formations and pasture sites, the forested areas, and the agricultural crops. Secondary differences in the NDVI and SAVI temporal responses were found to be related to their unique interactions with sun-sensor viewing geometries. An assessment of the functional behaviour of the VIs confirmed SAVI responds primarily to NIR variations, while the NDVI showed a strong dependence on the red reflectance. Based on these results, we expect operational use of the MODIS Enhanced Vegetation Index (EVI) to provide improved discrimination and monitoring capability of the significant Cerrado vegetation types.

Fires modify the structure of vegetation communities, the carbon and water cycles, the soil’s chemistry, and affect the climate system. Within this context, this work aimed to understand the distribution patterns of burned areas in Brazil, during the period of 2002 to 2010, taking into consideration each one of the six Brazilian biomes (Amazon, Caatinga, Cerrado, Atlantic Forest, Pampa and Pantanal) and the respective major land cover classes. Data from the MODIS MCD45A1 product (burned area), as well as thermal anomalies (MOD14 and MYD14) and precipitation (TRMM), were analyzed according to the 2002 Brazilian official land cover and land use map (PROBIO). The Brazilian savanna biome, known as Cerrado, presented the largest concentration of burned areas detected by MODIS (73%), followed by the Amazon (14%), Pantanal (6%), Atlantic Forest (4%), Caatinga (3%), and Pampa (0,06%) biomes. Indeed, in the years of 2007 and 2010, 90% and 92% of Brazil’s burned areas were concentrated in the Cerrado and Amazon biomes, respectively. TRMM data indicated that during these two years there was a significant influence of La Niña, causing low rainfall in the Amazon, Cerrado, Caatinga, and Atlantic Forest biomes. Regarding the land cover classes, approximately 81% of the burned areas occurred over remnant vegetation areas. Although no unequivocal correlation can be established between burned areas and new land conversions, the conspicuous concentration of fire scars, particularly in Amazon–Cerrado transition (i.e., the Arc of Deforestation) is certainly not a simple coincidence. Such patterns and trends corroborate the need of improved territorial governance, in addition to the implementation of systematic fire warning and preventive systems.

The rapid growth of satellites orbiting the planet is generating massive amounts of data for Earth science applications. Concurrently, state-of-the-art deep-learning-based algorithms and cloud computing infrastructure have become available with a great potential to revolutionize the image processing of satellite remote sensing. Within this context, this study evaluated, based on thousands of PlanetScope images obtained over a 12-month period, the performance of three machine learning approaches (random forest, long short-term memory-LSTM, and U-Net). We applied these approaches to mapped pasturelands in a Central Brazil region. The deep learning algorithms were implemented using TensorFlow, while the random forest utilized the Google Earth Engine platform. The accuracy assessment presented F1 scores for U-Net, LSTM, and random forest of, respectively, 96.94%, 98.83%, and 95.53% in the validation data, and 94.06%, 87.97%, and 82.57% in the test data, indicating a better classification efficiency using the deep learning approaches. Although the use of deep learning algorithms depends on a high investment in calibration samples and the generalization of these methods requires further investigations, our results suggest that the neural network architectures developed in this study can be used to map large geographic regions that consider a wide variety of satellite data (e.g., PlanetScope, Sentinel-2, Landsat-8).