ResumoShort-Term Wind Speed Forecasting Using ERA5 Reanalysis: A Comparison of SARIMA and LSTM Models at the Santo Agostinho Wind Farm

Renewable energy is a central component of Brazil's electricity matrix, with wind power standing out as one of the fastest-growing technologies and reaching more than 33 GW of installed capacity in 2025. Despite this rapid expansion, the temporal variability of renewable energy generation can pose operational challenges for the power system, including grid instability and potential stress on transmission infrastructure. As wind capacity continues to expand, accurately understanding both the availability and short-term predictability of wind resources becomes essential for reliable grid operation and effective energy planning...

ResumoA Domain-Restricted Local LLM for Meteorology Education: Faster, More Accurate Responses with Reduced Energy and Water Footprint

The rapid expansion of Large Language Models (LLMs) has created new opportunities for climate science education and decision support. However, general-purpose LLMs that rely on large-scale internet data may introduce inaccuracies, hallucinations, or content from unreliable sources. To address these challenges, we developed a fully local, domain-restricted Retrieval-Augmented Generation (RAG) system designed to operate exclusively on curated meteorological textbooks and peer-reviewed academic materials...

ResumoAssessment of the Impact of Synoptic Events on Photovoltaic Efficiency in North-Central Minas Gerais Using Random Forest

The increasing frequency of extreme climate events drives the energy transition in Brazil towards renewable sources. In this scenario, solar photovoltaic energy consolidates itself as a fundamental mitigation strategy. However, the thermodynamic efficiency of the panels is inherently vulnerable to atmospheric variability and extremes. Given this context, this study evaluates the impact of synoptic systems on photovoltaic generation in the North-Central region of Minas Gerais, Brazil, one of the main national generation hubs characterized by high irradiation levels, with a transitional climate domain highly susceptible to thermal and rainfall anomalies...

ResumoMachine Learning Assessment of Synoptic and Hydroclimatic Influence on Persistent Wildfire Events Across Brazilian Biomes (2003-2024)

Brazil encompasses diverse climate regimes and biomes that are differently affected by climate change. Among the emerging challenges, fire activity has become a key environmental and socioeconomic concern due to its impact on carbon emissions, biodiversity integrity, and human health. Within this context, this study investigates the influence of different components on the persistence and propagation of wildfire events across Brazilian biomes from 2003 to 2024. Wildfire hotspots data were sourced from the satellite detection system of the Brazilian National Institute for Spatial Research (INPE)...

ResumoConstructing Homogeneous Water Quality Time Series Through Multivariate Modeling: A Case Study of the Paraíba do Sul River Basin

This study aims to create a homogeneous time series of water quality parameters in the Paraíba do Sul River Basin - Rio de Janeiro, Brazil. The basin has high socioeconomic and environmental relevance for the region, being intensively used for urban water supply and industrial and agricultural activities. The analyzed parameters were pH, dissolved oxygen, and turbidity. The time series of these variables present substantial temporal heterogeneity, varying according to the parameter and the monitoring station analyzed. Five gauging stations in the Paraíba do Sul River Basin were evaluated, using data obtained from Hidroweb of the National Water Agency (ANA), covering the period from 2000 to 2024...

ResumoIdentification of Environmental Factors Controlling Biogenic Secondary Organic Aerosol Formation in the Central Amazon: a machine learning approach

Biogenic secondary organic aerosols dominate submicron aerosol mass in the central Amazon during the wet season, yet their molecular composition and formation pathways remain difficult to resolve due to the complexity of precursor chemistry and atmospheric processing. Recent measurements using a chemical ionisation mass spectrometer (PTR-Qi-ToF-MS-CHARON) in Central Amazon, at the Atmospheric Tall Tower Observatory (ATTO) during the wet season of 2022 provided near-molecular-level characterization of SOA tracers associated with isoprene-, monoterpene-, and sesquiterpene-derived oxidation pathways...

ResumoArtificial Intelligence and Climate Action in Smart Cities: Catalyst, Tool, or Barrier?

Climate change is intensifying environmental, social, and infrastructural challenges in urban areas, heightening the need for innovative solutions. In this context, Artificial Intelligence (AI) has emerged as an increasingly present element in debates about climate adaptation and mitigation strategies in smart cities. However, it is still unclear how AI is being constructed and represented in the global public debate. This gap is particularly relevant given the growing influence of techno-solutions narratives and sociotechnical imaginaries that shape public expectations about digital technologies in sustainability transitions...

ResumoApplication of bias correction Artificial Intelligence techniques on various CMIP6 SST teleconnections

In recent decades, Brazil has experienced increasing impacts from extreme weather events associated with climate change. Previous studies have demonstrated robust links between large-scale teleconnection indices and synoptic-scale phenomena that trigger extreme events across the country. Among global teleconnections, those associated with sea surface temperature (SST) exhibit stronger correlations with synoptic indices, highlighting the oceans' role in modulating these events. Although climate model projections have limited horizontal resolution to directly represent regional extremes, they perform consistently in simulating large-scale features such as SST variability and associated teleconnection patterns...