Tropical forests concentrate the largest diversity of species on the planet and play a key role in maintaining environmental processes. Due to the importance of those forests, there is growing interest in mapping their components and getting information at an individual tree level to conduct reliable satellite-based forest inventory for biomass and species distribution qualification. Individual tree crown information could be manually gathered from high resolution satellite images; however, to achieve this task at large-scale, an algorithm to identify and delineate each tree crown individually, with high accuracy, is a prerequisite. In this study, we propose the application of a convolutional neural network—Mask R-CNN algorithm—to perform the tree crown detection and delineation. The algorithm uses very high-resolution satellite images from tropical forests. The results obtained are promising—the $Recall$ , $Precision$ , and $F_1$ score values obtained were were $0.81$ , $0.91$ , and $0.86$ , respectively. In the study site, the total of tree crowns delineated was $\mathrm{59,062}$ . These results suggest that this algorithm can be used to assist the planning and conduction of forest inventories. As the algorithm is based on a Deep Learning approach, it can be systematically trained and used for other regions.

We present a new version of the Brazilian developments on the Regional Atmospheric Modeling System (BRAMS), in which different previous versions for weather, chemistry, and carbon cycle were unified in a single integrated modeling system software. This new version also has a new set of state-of-the-art physical parameterizations and greater computational parallel and memory usage efficiency. The description of the main model features includes several examples illustrating the quality of the transport scheme for scalars, radiative fluxes on surface, and model simulation of rainfall systems over South America at different spatial resolutions using a scale aware convective parameterization. Additionally, the simulation of the diurnal cycle of the convection and carbon dioxide concentration over the Amazon Basin, as well as carbon dioxide fluxes from biogenic processes over a large portion of South America, are shown. Atmospheric chemistry examples show the model performance in simulating near-surface carbon monoxide and ozone in the Amazon Basin and the megacity of Rio de Janeiro. For tracer transport and dispersion, the model capabilities to simulate the volcanic ash 3-D redistribution associated with the eruption of a Chilean volcano are demonstrated. The gain of computational efficiency is described in some detail. BRAMS has been applied for research and operational forecasting mainly in South America. Model results from the operational weather forecast of BRAMS on 5 km grid spacing in the Center for Weather Forecasting and Climate Studies, INPE/Brazil, since 2013 are used to quantify the model skill of near-surface variables and rainfall. The scores show the reliability of BRAMS for the tropical and subtropical areas of South America. Requirements for keeping this modeling system competitive regarding both its functionalities and skills are discussed. Finally, we highlight the relevant contribution of this work to building a South American community of model developers.

Many natural disasters in South America are linked to meteorological phenomena. Therefore, forecasting and monitoring climatic events are fundamental issues for society and various sectors of the economy. In the last decades, machine learning models have been developed to tackle different issues in society, but there is still a gap in applications to applied physics. Here, different machine learning models are evaluated for precipitation prediction over South America. Currently, numerical weather prediction models are unable to precisely reproduce the precipitation patterns in South America due to many factors such as the lack of region-specific parametrizations and data availability. The results are compared to the general circulation atmospheric model currently used operationally in the National Institute for Space Research (INPE: Instituto Nacional de Pesquisas Espaciais), Brazil. Machine learning models are able to produce predictions with errors under 2 mm in most of the continent in comparison to satellite-observed precipitation patterns for different climate seasons, and also outperform INPE’s model for some regions (e.g., reduction of errors from 8 to 2 mm in central South America in winter). Another advantage is the computational performance from machine learning models, running faster with much lower computer resources than models based on differential equations currently used in operational centers. Therefore, it is important to consider machine learning models for precipitation forecasts in operational centers as a way to improve forecast quality and to reduce computation costs.

Experimentally characterizing evapotranspiration (ET) in different biomes around the world is an issue of interest for different areas of science. ET in natural areas of the Brazilian Pampa biome has still not been assessed. In this study, the actual ET (ET_act) obtained from eddy covariance measurements over two sites of the Pampa biome was analyzed. The objective was to evaluate the energy partition and seasonal variability of the actual ET of the Pampa biome. Results showed that the latent heat flux was the dominant component in available energy in both the autumn–winter (AW) and spring–summer (SS) periods. Evapotranspiration of the Pampa biome showed strong seasonality, with highest ET rates in the SS period. During the study period, approximately 65% of the net radiation was used for the evapotranspiration process in the Pampa biome. The annual mean ET rate was 2.45 mm d⁻¹. ET did not show to vary significantly between sites, with daily values very similar in both sites. The water availability in the Pampa biome was not a limiting factor for ET, which resulted in a small difference between the reference ET and the actual ET. These results are helpful in achieving a better understanding of the temporal pattern of ET in relation to the landscape of the Pampa biome and its meteorological, soil, and vegetation characteristics.

Machine learning has experienced great success in many applications. Precipitation is a hard meteorological variable to predict, but it has a strong impact on society. Here, a machine-learning technique—a formulation of gradient-boosted trees—is applied to climate seasonal precipitation prediction over South America. The Optuna framework, based on Bayesian optimization, was employed to determine the optimal hyperparameters for the gradient-boosting scheme. A comparison between seasonal precipitation forecasting among the numerical atmospheric models used by the National Institute for Space Research (INPE, Brazil) as an operational procedure for weather/climate forecasting, gradient boosting, and deep-learning techniques is made regarding observation, with some showing better performance for the boosting scheme.