Souza et al. Landsat sub-pixel land cover dynamics in the Brazilian Amazon

This study proposes a novel approach to characterize and measure land cover dynamics in the Amazon biome. First, we defined 10 fundamental land cover classes: forest, flooded forest, shrubland, natural grassland, pastureland, cropland, outcrop, bare and impervious, wetland, and water. Second, we mapped the land cover based on the compositional abundance of Landsat sub-pixel information that makes up these land cover classes: green vegetation (GV), non-photosynthetic vegetation, soil, and shade. Third, we processed all Landsat scenes with <50% cloud cover. Then, we applied a step-wise random forest machine learning algorithm and empirical decision rules to classify intra-annual and annual land cover classes between 1985 and 2022. Finally, we estimated the yearly land cover changes in forested and non-forested ecosystems and characterized the major change drivers.

Baeza et al. Two decades of land cover mapping in the Río de la Plata grassland region: The MapBiomas Pampa initiative

This work describes and analyzes the land cover changes in the entire Río de la Plata Grasslands (RPG) region for the first two decades of the 21st century, especially those related to grassland loss. In 20 years, RPG region lost, at least, 2.4 million ha of grassland (9% of the remaining grassland area in 2001). Most of these losses are concentrated in Brazil and Uruguay and are associated with new agricultural or forestry areas that increased by 5% and 100%, respectively.

Santos et al – Assessing the Wall-to-Wall Spatial and Qualitative Dynamics of the Brazilian Pasturelands 2010–2018, Based on the Analysis of the Landsat Data Archive

In this study, the spatial-temporal dynamics of pasture quality in Brazil between 2010 and 2018 were mapped and evaluated, considering three degradation classes: Absent (D0), Intermediate (D1), and Severe (D2. There was no variation in the total area occupied by pastures in the evaluated period, despite the accentuated spatial dynamics.

Cesar et al. – A Large-Scale Deep-Learning Approach for Multi-Temporal Aqua and Salt-Culture Mapping

Aquaculture and saliculture are relevant economic activities in the Brazilian Coastal Zone (BCZ). However, automatic discrimination of such activities from other water-related coverages/uses is not an easy task. In this sense, convolutional neural networks (CNN) have the advantage of predicting the class label of a given pixel, providing as input a local region (patches or named chips) around that pixel. Both the convolutional nature and the semantic segmentation capability provide the U-Net classifier with the ability to access the "context domain" instead of just isolated pixel values. Supported in the context domain, we present the results of the analyzes.