Alan James Peixoto Calheiros

Este artigo descreve o modelo de curadoria do DataMap/Amazon, desenvolvido para mitigar a dispersão e heterogeneidade de dados ambientais amazônicos. A solução integra quatro componentes funcionais (Dicionário, Arquivos, Datasets e Produtos) a um pipeline de qualificação progressiva (camadas Bronze, Prata e Ouro). Aplicada a quatro iniciativas de pesquisa (LBA, AmazonFACE, ATTO-Campina e CHUVA), a plataforma transformou acervos brutos em ativos científicos FAIR. Os resultados incluem a estruturação de 183 datasets, 1,58 TB de arquivos organizados e 165 DOIs gerados, fortalecendo a reprodutibilidade e o acesso aberto à ciência na região.

This study investigates the non-stationary evolution of Atlantic multidecadal variability and its influence on hydroclimate in Northeastern Brazil (NEB) over the last millennium. By integrating high-resolution paleoclimate proxies (speleothems and marine sediments) with isotope-enabled climate simulations from the GISS-E2-R model, we move beyond confirming the well-established AMV-ITCZ link and instead document how the tropical Atlantic-South American climate system responded to a superposition of drivers across distinct climatic intervals. During the Medieval Climate Anomaly (950-1300 CE), variability was characterized by a persistent 64-year cycle associated with a positive phase of Atlantic Multidecadal Variability (AMV) and a northward displacement of the Intertropical Convergence Zone (ITCZ), intensifying aridity over NEB. The transition period (1301-1499 CE) was characterized by a more irregular 8-year cycle, likely reflecting enhanced climatic instability driven by interactions among ENSO dynamics, volcanic forcing, and internal Atlantic variability. During the Little Ice Age (1500-1849 CE), a 32-year cycle emerged, associated with a negative AMV phase, a southward ITCZ shift, and wetter conditions over NEB. Our findings demonstrate that NEB hydroclimate reflects a non-stationary superposition of internal variability (AMV, ENSO) and external forcings (volcanism, solar variability), rather than a single dominant driver. This multi-scale framework provides a long-term perspective for understanding present-day climate variability and improving future drought risk projections in the region. Graphical Abstract The This graphical summary illustrates how Atlantic Multidecadal Variability (AMV), Pacific forcing, and external drivers influenced the position of the Intertropical Convergence Zone (ITCZ) and rainfall patterns in Northeastern Brazil (NEB) over the last millennium. During the Medieval Climate Anomaly (MCA, 950–1300 CE), a positive AMV phase contributed to a northward ITCZ shift and drier conditions. The transition period ( 1300–1500 CE) was marked by an 8-year mode reflecting interactions among ENSO, volcanic forcing, and Atlantic variability. During the Little Ice Age (LIA, 1500–1849 CE), a negative AMV phase favored a southward ITCZ migration and wetter conditions. These fluctuations reflect a superposition of internal variability and external forcings, rather than a single dominant driver, providing a long-term perspective for improving future climate adaptation in NEB.

This study evaluates the performance of two deep learning architectures, U-Net and convolutional long short-term memory (ConvLSTM), for predicting cloud-to-ground (CG) lightning probability in the central Amazon basin. Utilizing data from S-band weather radar and very low-frequency (VLF) lightning detection networks, we explored nowcasting capabilities at 12- and 6-min intervals. Various data configurations were tested, specifically examining the integration of volumetric radar reflectivity (dB Z ) and echo-top 35-dBZ (ET35). The evaluation encompassed objective skill metrics and spatial error analyses, accounting for sensor uncertainties. Results reveal a critical divergence in multimodal data assimilation. While the U-Net struggled with volumetric inputs, the ConvLSTM architecture successfully leveraged dB Z data, achieving the highest critical success index (CSI) of 0.546 and an effective spatial scale of 6 km, notably outperforming persistence baselines. Conversely, U-Net performed best using only quality-controlled lightning data or lightweight ET35 inputs. Furthermore, 6-min experiments demonstrated that extending the temporal lookback window notably enhances the skill of recurrent models. Spatial error analyses indicated well-collocated clusters with average centroid errors below 6 km at optimal probability thresholds. These findings suggest that while U-Net provides a robust baseline, ConvLSTM is superior for integrating high-dimensional radar variables in very short-term lightning nowcasting. Significance Statement: This exploratory study evaluates and compares emerging computational approaches for meteorological forecasting, specifically examining the effectiveness of U-Net and convolutional long short-term memory (ConvLSTM) architectures in predicting cloud-to-ground lightning probabilities in the central Amazon basin. By integrating data from weather radar and lightning detection networks, the research investigates methodological solutions for very short-term lightning prediction (6–12 min) in one of Earth’s most climatically complex regions. The comparative analysis reveals a critical distinction in data assimilation capabilities. While the ConvLSTM successfully leverages volumetric radar reflectivity to achieve effective spatial scales of 6 km, the U-Net performs optimally with simplified inputs like echo tops. This work identifies robust configurations that surpass traditional persistence baselines, laying the foundation for next-generation early warning systems capable of mitigating the socioeconomic impacts and risks of human fatalities caused by lightning.

Accurate short-term precipitation forecasting is critical for Latin America, but the region lacks a standardized framework to evaluate data-driven approaches due to the sparse coverage in the ground. This study introduces the Artificial Intelligence for Nowcasting Pilot Project (AINPP) Precipitation Benchmark (AINPP-PB-LATAM), providing curated datasets and a scalable, optimized training pipeline designed to accelerate deep learning development in high-performance computing environments. Beyond establishing a baseline using seven years of satellite-based data (2018–2024), the framework reduces engineering barriers, enabling researchers to focus on fine-tuning strategies to extract the full predictive capacity of models for regional specificities. As a demonstration use case, we trained and evaluated five deep learning architectures, AFNO, Inception-V4, ResNet-50, U-Net, and Xception, comparing them against Lagrangian Persistence and the operational AI Nowcast. The results reveal critical trade-offs: spectral methods like AFNO excel in continuous skill by capturing large-scale dependencies, while convolutional architectures offer robust categorical performance. However, pixel-wise optimization challenges persist, with systematic under-prediction of heavy rainfall. By providing open-access code and optimized baseline implementations for distributed computing, AINPP-PB-LATAM establishes a scalable foundation for collaborative research, facilitating the advancement of operational AI-based nowcasting and transferability assessments in data-scarce regions.

O crescente volume de dados demanda maior controle e monitoramento das informações geradas. Nesse contexto, ferramentas de gerenciamento de dados tornam-se essenciais para atender às necessidades de grandes instituições no Brasil. Órgãos como o ICMBio e o INMET já utilizam soluções desse tipo, mas o INPE ainda carece de um portal específico que atenda a essas demandas. Este artigo apresenta o desenvolvimento de uma ferramenta baseada nos princípios FAIR, voltada à inserção e indexação padronizada de dados em um banco de dados não relacional. Os resultados demonstram que a ferramenta atende adequadamente aos princípios FAIR.

Extreme precipitation events have become a central focus of the scientific community due to their increased occurrence in recent years. This study aims to analyze the variability and trends in aspects associated with the rainy seasons in the South American Monsoon System (SAMS) area from 1979 to 2022. The dates for the onset and demise of the rainy season (ONR and DER, respectively) were determined using antisymmetric outgoing longwave radiation (OLR) data relative to the equator (AOLR) for the clustered regions defined in a previous work. Based on these dates, the duration of the rainy seasons and the total precipitation for each rainy season were also calculated. The main advantage of this study is the analysis of trends within homogeneous regions derived from cluster analysis, which enables a more reliable assessment of precipitation patterns across the spatially heterogeneous SAMS domain. The non-parametric Mann–Kendall test and Sen’s slope estimator were applied to the ONR, DER, rainy season length, and total precipitation time series for each group over the 1979–2022 period. Quartile analysis was performed on the total precipitation time series to identify the most and least intense rainy seasons in the SAMS’s regions. These analyses revealed a trend of shortening of the SAMS rainy season over the 44 years of analysis, with a positive trend in the ONR dates and a negative trend in the DER dates, which is further confirmed by the decreasing trends in rainy season length and accumulated precipitation in most analyzed regions. The most (above the third quartile) and least (below the first quartile) intense rainy seasons were found to be concentrated at the beginning and end of the study period, respectively, for all monsoon regions. After removing the linear trend, the distribution of events appeared more uniform over time, yet the major droughts that occurred after 2010 remained clear. The results of this study contribute to a better understanding of the precipitation characteristics in the SAMS area, and these findings may assist climate forecasting and monitoring centers in improving regional precipitation assessments.

Precipitating systems (PSs) are key components of atmospheric dynamics, shaping both regional and global hydrological cycles. Accurate identification, tracking, and forecasting of these systems are essential for improving weather and climate predictions and mitigating the socio-economic impacts of extreme events. This study presents a global climatology of PSs based on the pyForTraCC (Python Forecasting and Tracking the Evolution of Configurable Clusters) algorithm, developed to identify and track PSs using satellite-based precipitation datasets. The algorithm applies multi-threshold segmentation, clustering, and vectorization techniques to extract features from GSMaP/MVK version 8 data, covering the period from 2015 to 2024 and spanning latitudes from 60°N to 60°S. This decade was selected specifically because it relies exclusively on data from the Global Precipitation Measurement (GPM) mission, ensuring the highest consistency and quality in satellite-derived precipitation estimates. The latest version of GSMaP includes significant improvements from the GPM era, such as enhanced inter-sensor calibration and the Microwave Rainfall Normalization module, which standardizes estimates across different sensors. In this study, pyForTraCC tracks PSs using four precipitation thresholds (0.1, 1, 5, and 10 mm/h), classifies clusters into six evolutionary types, and applies a 10% spatial overlap criterion to ensure continuity across time steps. To reduce uncertainties associated with displacement vectors, multiple correction methods were tested and validated using persistence-based forecasting and minimization of the False Alarm Rate (FAR). Only systems with a minimum duration of three hours and a spatial extent larger than 10 km² were included to ensure robustness. The resulting global climatology reveals spatial and seasonal distributions of PSs aligned with major atmospheric features such as convergence zones, monsoon regions, and storm tracks. Additionally, a high variability in the response to climate change can be observed. For example, in some regions, such as the Amazon Basin, precipitating systems are becoming more intense, and certain convergence zones are persisting for longer periods. Most importantly, this work provides the scientific community with both a flexible, open-source tool for PS tracking and a publicly available, curated database of all tracked systems over a full decade of high-quality GPM observations. This dataset includes detailed information on size, duration, intensity, and displacement, supporting future research on precipitation regimes, climate variability, and extreme weather. By making both the algorithm and database openly accessible, this study contributes to advancing satellite precipitation validation, nowcasting applications using AI, and data-driven meteorological research.

Extreme precipitation events have become a central focus of the scientific community due to their increased occurrence in recent years. This study aims to analyze the variability and trends of aspects associated with the rainy seasons in the South American Monsoon System (SAMS) area from 1979 to 2022. The Onset and Demise of the Rainy Season (ONR and DER, respectively) dates were determined using antisymmetric Outgoing Longwave Radiation (OLR) data relative to the equator (AOLR) for the clustered regions defined in a previous work. Based on these dates, the duration of the rainy seasons and the total precipitation for each rainy season were also calculated. The main advantage of this study is the analysis of trends within homogeneous regions derived from cluster analysis, which enables a more reliable assessment of precipitation patterns across the spatially heterogeneous SAMS domain. The non-parametric Mann-Kendall test and Sen’s slope estimator were applied to the ONR, DER, rainy season length, and total precipitation time series for each group over the 1979–2022 period. Quartile analysis was performed on the total precipitation time series to identify the most and least intense rainy seasons in the SAMS´s regions. These analyses revealed a trend of shortening of the SAMS rainy season over the 44 years of analysis, with a positive trend in the ONR dates and a negative trend in the DER dates, which is further confirmed by the decreasing trends in rainy season length and accumulated precipitation in most analyzed regions. The most (above the third quartile) and least (below the first quartile) intense rainy seasons were found to be concentrated at the beginning and end of the study period, respectively for all monsoon regions. When the linear trend was removed, the distribution of events became more homogeneous over time, but the main drought events after 2010 remained. The results of this study contribute to a better understanding of the precipitation characteristics in the SAMS area and these findings may assist climate forecasting and monitoring centers in improving regional precipitation assessments.

The National Institute for Space Research (INPE) has been a partner in significant projects that conduct atmospheric investigations impacting various sectors, such as the Amazon Tall Tower Observatory (ATTO) project. Since 2009, the project has conducted studies on the interactions between climate and the Amazon forest. ATTO has played an essential role in providing large volumes of data obtained by meteorological sensors, contributing to a deeper understanding of the atmospheric dynamics of the region. In a landscape where Artificial Intelligence-based rainfall forecast models gain prominence, this study explores the imbalance of data from the ATTO Campina field experiment and its influence on short-term rainfall forecasts using Artificial Neural Networks (ANNs). Metrics such as MAE, RMSE, and POD, as well as FAR indices, were applied in the assessment and revealed the connection between data balance and forecast results. More balanced data or data with greater weights for different rainfall ranges yield better results. The study emphasizes the importance of reliable data for training rain forecast models, aiming to improve the dexterity of these models. This approach is fundamental to increase the reliability of these models in real environments.