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.

A comprehensive record of the land use history of South America reveals that human impact on land increased by 60% since 1985.

We address the task of mapping deforested areas in the Brazilian Amazon. Accurate maps are an important tool for informing effective deforestation containment policies. The main existing approaches to this task are largely manual, requiring significant effort by trained experts. To reduce this effort, we propose a fully automatic approach based on spatio-temporal deep convolutional neural networks. We introduce several domain-specific components, including approaches for: image preprocessing; handling image noise, such as clouds and shadow; and constructing the training data set. We show that our preprocessing protocol reduces the impact of noise in the training data set. Furthermore, we propose two spatio-temporal variations of the U-Net architecture, which make it possible to incorporate both spatial and temporal contexts. Using a large, real-world data set, we show that our method outperforms a traditional U-Net architecture, thus achieving approximately 95% accuracy.

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).

The paper describes the production and evaluation of global grassland extent mapped annually for 2000–2022 at 30 m spatial resolution. The dataset showing the spatiotemporal distribution of cultivated and natural/semi-natural grassland classes was produced by using GLAD Landsat ARD-2 image archive, accompanied by climatic, landform and proximity covariates, spatiotemporal machine learning (per-class Random Forest) and over 2.3 M reference samples (visually interpreted in Very High Resolution imagery). Custom probability thresholds (based on five-fold spatial cross-validation) were used to derive dominant class maps with balanced user’s and producer’s accuracy, resulting in f1 score of 0.64 and 0.75 for cultivated and natural/semi-natural grassland, respectively. The produced maps (about 4 TB in size) are available under an open data license as Cloud-Optimized GeoTIFFs and as Google Earth Engine assets. The suggested uses of data include (1) integration with other compatible land cover products and (2) tracking the intensity and drivers of conversion of land to cultivated grasslands and from natural / semi-natural grasslands into other land use systems.