Juliana Aparecida Anochi
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Juliana Aparecida Anochi has a total of 52 co-authors. Here are the top co-authors by number of shared publications:
- Haroldo Fraga de Campos Velho
National Institute for Space Research (INPE), São José dos Campos 12227-010, SP, Brazil
16 shared publications - Marília Harumi Shimizu
General Coordination of Earth Science, National Institute for Space Research, São José dos Campos 12227-010, SP, Brazil
5 shared publications - Vinícius Albuquerque de Almeida
Department of Meteorology, Federal University of Rio de Janeiro (UFRJ), Rio de Janeiro 21941-916, RJ, Brazil
5 shared publications - Matheus José Gomes
Universidade Federal de Itajubá image/svg+xml
3 shared publications - Vinicius Schmidt Monego
Coordination of Applied Research and Technological Development (COPDT), National Institute for Space Research (INPE), São José dos Campos, Brazil
3 shared publications - Afonso Augusto Magalhães de Araújo
Department of Water Resources and Environment, Federal University of Rio de Janeiro (UFRJ), Rio de Janeiro 21941-909, RJ, Brazil
2 shared publications
Latest Publications
- Open Access
- 25 September 2026
- Published by Universidade Federal de Santa Maria in Ciência e Natura
- Vol. 48, e90228
Use of artificial intelligence for seasonal precipitation forecasting in the Minas Gerais state, Brazil
Precipitation forecast is crucial for water resource management, agriculture, and climate disaster mitigation. This study investigates the use of artificial intelligence (AI) to predict seasonal precipitation in the state of Minas Gerais, Brazil. Utilizing neural networks trained on high temporal and spatial resolution datasets (GPCC, GPCP, MERGE-CPTEC, and CHIRPS), the research compares AI-based forecasts with numerical physics-based models. The results indicate that averaging multiple runs of each neural network model and using higher spatial resolution datasets, particularly CHIRPS, yield more accurate predictions, with lower Root Mean Square Errors (RMSE) and Relative Errors (RE). AI applications have shown promising results, offering forecasts with reduced computational processing time compared to traditional models. The study emphasizes the need to enhance computational infrastructure and explores the potential of advanced AI techniques to improve extreme climatic events forecasts, benefiting sectors such as agriculture and water resource management.
- 25 September 2026
- Published by Universidade Federal de Santa Maria in Ciência e Natura
- Vol. 48, e90228
Open AccessUse of artificial intelligence for seasonal precipitation forecasting in the Minas Gerais state, Brazil
Precipitation forecast is crucial for water resource management, agriculture, and climate disaster mitigation. This study investigates the use of artificial intelligence (AI) to predict seasonal precipitation in the state of Minas Gerais, Brazil. Utilizing neural networks trained on high temporal and spatial resolution datasets (GPCC, GPCP, MERGE-CPTEC, and CHIRPS), the research compares AI-based forecasts with numerical physics-based models. The results indicate that averaging multiple runs of each neural network model and using higher spatial resolution datasets, particularly CHIRPS, yield more accurate predictions, with lower Root Mean Square Errors (RMSE) and Relative Errors (RE). AI applications have shown promising results, offering forecasts with reduced computational processing time compared to traditional models. The study emphasizes the need to enhance computational infrastructure and explores the potential of advanced AI techniques to improve extreme climatic events forecasts, benefiting sectors such as agriculture and water resource management.
- Open Access
Hybrid Streamflow Forecasting with ERA5 and Machine Learning Across Daily and Monthly Time Scales
This study presents an updated Hybrid Hydrological Forecasting System (HHFS) for streamflow prediction at the Santa Branca outlet, located in the upper Paraíba do Sul River Basin in southeastern Brazil, aiming to support hydropower-oriented water resources management. This paper is explicitly framed as a companion paper which introduced the original HHFS framework and demonstrated the feasibility of combining deterministic and probabilistic machine-learning approaches for monthly streamflow forecasting. Building upon that foundation, the present study develops and validates a substantially enhanced and operationally oriented version of the system. The upgraded HHFS replaces the original BR-DWGD forcing strategy—a Brazilian gridded meteorological dataset useful for research applications but not routinely updated for sustained operations—with ERA5, the fifth-generation global atmospheric reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF), which provides temporally consistent and operationally updated meteorological fields. This transition renders the framework fully operational while preserving the original dual-stage architecture, composed of a deterministic forecasting module (GA1) and a hydro-adaptive uncertainty module (GA2). In addition, the study introduces a daily short-term forecasting extension using a single multi-output XGBoost 2.1.1 model to predict streamflow from D+1 to D+10. Predictive uncertainty is quantified using split conformal prediction, a distribution-free uncertainty method that provides valid prediction intervals with empirical coverage guarantees. Coverage represents the proportion of observed values falling within the prediction intervals and is used here as a reliability metric. For the monthly product, the ERA5-based methodology maintained and slightly improved deterministic skill relative to the original BR-DWGD benchmark, with independent-test NSE increasing to 0.798, KGE to 0.878, and RMSE decreasing to 18.778 m3/s. The probabilistic component preserved a high hit rate and similar relative width, although coverage declined modestly to 0.838, indicating slight undercoverage relative to the previous reliability target. For the daily forecasts, predictive skill decreased progressively with lead time, from NSE = 0.881 at D+1 to 0.394 at D+10, accompanied by coherent widening of the uncertainty intervals. Taken together, these results demonstrate that ERA5 is a robust and operationally practical forcing source for the HHFS, preserving monthly forecasting skill while enabling a promising multi-day extension for anticipatory streamflow prediction across multiple temporal scales.
Open AccessHybrid Streamflow Forecasting with ERA5 and Machine Learning Across Daily and Monthly Time Scales
This study presents an updated Hybrid Hydrological Forecasting System (HHFS) for streamflow prediction at the Santa Branca outlet, located in the upper Paraíba do Sul River Basin in southeastern Brazil, aiming to support hydropower-oriented water resources management. This paper is explicitly framed as a companion paper which introduced the original HHFS framework and demonstrated the feasibility of combining deterministic and probabilistic machine-learning approaches for monthly streamflow forecasting. Building upon that foundation, the present study develops and validates a substantially enhanced and operationally oriented version of the system. The upgraded HHFS replaces the original BR-DWGD forcing strategy—a Brazilian gridded meteorological dataset useful for research applications but not routinely updated for sustained operations—with ERA5, the fifth-generation global atmospheric reanalysis produced by the European Centre for Medium-Range Weather Forecasts (ECMWF), which provides temporally consistent and operationally updated meteorological fields. This transition renders the framework fully operational while preserving the original dual-stage architecture, composed of a deterministic forecasting module (GA1) and a hydro-adaptive uncertainty module (GA2). In addition, the study introduces a daily short-term forecasting extension using a single multi-output XGBoost 2.1.1 model to predict streamflow from D+1 to D+10. Predictive uncertainty is quantified using split conformal prediction, a distribution-free uncertainty method that provides valid prediction intervals with empirical coverage guarantees. Coverage represents the proportion of observed values falling within the prediction intervals and is used here as a reliability metric. For the monthly product, the ERA5-based methodology maintained and slightly improved deterministic skill relative to the original BR-DWGD benchmark, with independent-test NSE increasing to 0.798, KGE to 0.878, and RMSE decreasing to 18.778 m3/s. The probabilistic component preserved a high hit rate and similar relative width, although coverage declined modestly to 0.838, indicating slight undercoverage relative to the previous reliability target. For the daily forecasts, predictive skill decreased progressively with lead time, from NSE = 0.881 at D+1 to 0.394 at D+10, accompanied by coherent widening of the uncertainty intervals. Taken together, these results demonstrate that ERA5 is a robust and operationally practical forcing source for the HHFS, preserving monthly forecasting skill while enabling a promising multi-day extension for anticipatory streamflow prediction across multiple temporal scales.
- Open Access
- 5 May 2026
- Published by MDPI AG in Meteorology
- Vol. 5 (2) , 12
Use of Artificial Intelligence for Spatial Seasonal Precipitation Forecasting in Minas Gerais, Brazil
Seasonal precipitation forecasting remains challenging in regions with complex topography and high climatic variability, such as the state of Minas Gerais, Brazil. This study evaluates the performance of an Artificial Intelligence (AI)-based ensemble approach for seasonal precipitation prediction. The AI-based predictions are compared against outputs from multiple dynamical models, including those from the North American Multi-Model Ensemble (NMME) and the Copernicus Climate Data Store (CDS). The AI model was trained using high-resolution precipitation data from the Center for Weather Forecast and Climate Studies (CPTEC) dataset – MERGE-CPTEC – and subsequently applied to generate regional-scale seasonal forecasts. Model performance was assessed using Root Mean Square Error (RMSE), Mean Squared Error (MSE), and Pearson Correlation (r). The results indicate that the AI-based forecasts achieve competitive performance relative to dynamical models across all seasons, exhibiting lower error metrics and improved representation of spatial precipitation patterns. The highest forecast skill was observed during winter (June-July-August, JJA), when atmospheric conditions are more stable, and precipitation variability is low. During the wet seasons (December-January-February, DJF and September-October-November, SON), despite increased convective activity and spatial heterogeneity, the AI model maintained greater spatial coherence and closer agreement with observations than the dynamical forecasts. Overall, the findings demonstrate that AI-based approaches represent a promising and computationally efficient complementary tool for regional-scale seasonal precipitation forecasting, particularly in climatically heterogeneous regions.
- 5 May 2026
- Published by MDPI AG in Meteorology
- Vol. 5 (2) , 12
Open AccessUse of Artificial Intelligence for Spatial Seasonal Precipitation Forecasting in Minas Gerais, Brazil
Seasonal precipitation forecasting remains challenging in regions with complex topography and high climatic variability, such as the state of Minas Gerais, Brazil. This study evaluates the performance of an Artificial Intelligence (AI)-based ensemble approach for seasonal precipitation prediction. The AI-based predictions are compared against outputs from multiple dynamical models, including those from the North American Multi-Model Ensemble (NMME) and the Copernicus Climate Data Store (CDS). The AI model was trained using high-resolution precipitation data from the Center for Weather Forecast and Climate Studies (CPTEC) dataset – MERGE-CPTEC – and subsequently applied to generate regional-scale seasonal forecasts. Model performance was assessed using Root Mean Square Error (RMSE), Mean Squared Error (MSE), and Pearson Correlation (r). The results indicate that the AI-based forecasts achieve competitive performance relative to dynamical models across all seasons, exhibiting lower error metrics and improved representation of spatial precipitation patterns. The highest forecast skill was observed during winter (June-July-August, JJA), when atmospheric conditions are more stable, and precipitation variability is low. During the wet seasons (December-January-February, DJF and September-October-November, SON), despite increased convective activity and spatial heterogeneity, the AI model maintained greater spatial coherence and closer agreement with observations than the dynamical forecasts. Overall, the findings demonstrate that AI-based approaches represent a promising and computationally efficient complementary tool for regional-scale seasonal precipitation forecasting, particularly in climatically heterogeneous regions.
- Open Access
- 13 March 2026
- Published by Copernicus GmbH
PWV-GNSS JUMP as a tool for nowcasting in Brazil: an overview, the challenges, and opportunities
Ground-based GNSS (Global Navigation Satellite System) receivers have been used to estimate precipitable water vapor (PWV) with high temporal resolution. The quality in terms of precision and confidence has given the opportunity to explore this feature to predict the occurrence of thunderstorms. A sharp increase in the GNSS-PWV time series before the intense precipitation events has been found, which indicates the occurrence of this phenomenon and consequently demonstrates a good potential for application in nowcasting activities. This increasing pattern in the PWV-GNSS time series before strong precipitation has been termed GPS-PWV-jumps and occurs because of the water vapor convergence and the continued formation of cloud condensate and precipitation particles. This study presents an overview of the development of this technique in Brazil, presenting a summary of the latest results using the data collected in different campaigns in the last years over different regions of Brazilian territory. GNSS receivers and several instruments to observe the precipitation, such as disdrometers and X-band radar, were used. The long database has been explored, and extensive analyses of results were carried out using wavelet cross-correlation analysis, lag correlation method, and contingency table after defining a method to predict the precipitation using GNSS-PWV jump information. This approach is innovative because it uses only GNSS data and, consequently, the infrastructure used by geodesic applications, such as GNSS receiver networks present in big cities, can be explored for this purpose without additional investments. However, there are some challenges that need to be addressed yet, such as the PWV-GNSS-jump production in near real time, which involves the data reception and data processing in a suitable time to be evaluated and applied to the issuance of disaster warnings. Another challenge, just as important as the first, is ensuring that the performance of the GNNS-PWV jump is maintained when using near-real-time estimates. These challenges are treated in this work as an opportunity for researchers exploring artificial intelligence methods, which are discussed, and some possible strategies are presented. The future perspective of the GNSS receiver application as a humidity information data source used in the evaluation and data assimilation process in the community development of the MONAN (Model for Ocean-laNd-Atmosphere predictioN) model is also discussed.
- 13 March 2026
- Published by Copernicus GmbH
Open AccessPWV-GNSS JUMP as a tool for nowcasting in Brazil: an overview, the challenges, and opportunities
Ground-based GNSS (Global Navigation Satellite System) receivers have been used to estimate precipitable water vapor (PWV) with high temporal resolution. The quality in terms of precision and confidence has given the opportunity to explore this feature to predict the occurrence of thunderstorms. A sharp increase in the GNSS-PWV time series before the intense precipitation events has been found, which indicates the occurrence of this phenomenon and consequently demonstrates a good potential for application in nowcasting activities. This increasing pattern in the PWV-GNSS time series before strong precipitation has been termed GPS-PWV-jumps and occurs because of the water vapor convergence and the continued formation of cloud condensate and precipitation particles. This study presents an overview of the development of this technique in Brazil, presenting a summary of the latest results using the data collected in different campaigns in the last years over different regions of Brazilian territory. GNSS receivers and several instruments to observe the precipitation, such as disdrometers and X-band radar, were used. The long database has been explored, and extensive analyses of results were carried out using wavelet cross-correlation analysis, lag correlation method, and contingency table after defining a method to predict the precipitation using GNSS-PWV jump information. This approach is innovative because it uses only GNSS data and, consequently, the infrastructure used by geodesic applications, such as GNSS receiver networks present in big cities, can be explored for this purpose without additional investments. However, there are some challenges that need to be addressed yet, such as the PWV-GNSS-jump production in near real time, which involves the data reception and data processing in a suitable time to be evaluated and applied to the issuance of disaster warnings. Another challenge, just as important as the first, is ensuring that the performance of the GNNS-PWV jump is maintained when using near-real-time estimates. These challenges are treated in this work as an opportunity for researchers exploring artificial intelligence methods, which are discussed, and some possible strategies are presented. The future perspective of the GNSS receiver application as a humidity information data source used in the evaluation and data assimilation process in the community development of the MONAN (Model for Ocean-laNd-Atmosphere predictioN) model is also discussed.
- Open Access
- 23 February 2026
- Published by MDPI AG in Preprints.org
Use of Artificial Intelligence for Seasonal Precipitation Forecasting in the Minas Gerais State, Brazil
Seasonal precipitation forecasting remains challenging in regions with complex topography and high climatic variability, such as the state of Minas Gerais, Brazil. This study evaluates the performance of an Artificial Intelligence (AI)–based ensemble approach for seasonal precipitation prediction during 2024 and compares its results with those obtained from the NCEP Climate Forecast System version 2 (NCEP-CFSv2), a model from the North American Multi-Model Ensemble (NMME). The AI model was trained using high-resolution precipitation data from the MERGE-CPTEC dataset and applied to generate seasonal forecasts. Model performance was assessed using Root Mean Square Error (RMSE), Mean Square Error (MSE), and Relative Error (RE). Observed seasonal precipitation anomalies for 2024 were also examined to contextualize forecast skill under different climatic conditions. The results show that the AI-based forecasts consistently outperform the NCEP-CFSv2 from NMME across all seasons, exhibiting lower error metrics and improved representation of spatial precipitation patterns. The highest forecast skill was observed during winter (JJA), when atmospheric conditions are more stable and precipitation variability is low. During the wet seasons (DJF and SON), despite increased convective activity and spatial heterogeneity, the AI model maintained greater spatial coherence and closer agreement with observations than the dynamical forecasts. Overall, the findings demonstrate that AI-based approaches represent a promising and computationally efficient complementary tool for regional-scale seasonal precipitation forecasting, particularly in climatically heterogeneous regions.
- 23 February 2026
- Published by MDPI AG in Preprints.org
Open AccessUse of Artificial Intelligence for Seasonal Precipitation Forecasting in the Minas Gerais State, Brazil
Seasonal precipitation forecasting remains challenging in regions with complex topography and high climatic variability, such as the state of Minas Gerais, Brazil. This study evaluates the performance of an Artificial Intelligence (AI)–based ensemble approach for seasonal precipitation prediction during 2024 and compares its results with those obtained from the NCEP Climate Forecast System version 2 (NCEP-CFSv2), a model from the North American Multi-Model Ensemble (NMME). The AI model was trained using high-resolution precipitation data from the MERGE-CPTEC dataset and applied to generate seasonal forecasts. Model performance was assessed using Root Mean Square Error (RMSE), Mean Square Error (MSE), and Relative Error (RE). Observed seasonal precipitation anomalies for 2024 were also examined to contextualize forecast skill under different climatic conditions. The results show that the AI-based forecasts consistently outperform the NCEP-CFSv2 from NMME across all seasons, exhibiting lower error metrics and improved representation of spatial precipitation patterns. The highest forecast skill was observed during winter (JJA), when atmospheric conditions are more stable and precipitation variability is low. During the wet seasons (DJF and SON), despite increased convective activity and spatial heterogeneity, the AI model maintained greater spatial coherence and closer agreement with observations than the dynamical forecasts. Overall, the findings demonstrate that AI-based approaches represent a promising and computationally efficient complementary tool for regional-scale seasonal precipitation forecasting, particularly in climatically heterogeneous regions.
- Open Access
Integrating Regressive and Probabilistic Streamflow Forecasting via a Hybrid Hydrological Forecasting System: Application to the Paraíba do Sul River Basin
This study introduces the Hybrid Hydrological Forecast System (HHFS), a dual-stage, data-driven framework for monthly streamflow forecasting at the Santa Branca outlet in the upper Paraíba do Sul River Basin, Brazil. The system combines two nonlinear regressors, Multi-Layer Perceptron (MLP) and extreme Gradient Boosting (XGB), calibrated through a structured four-step evolutionary procedure in GA1 (hydrological weighting, dual-regime Ridge fusion, rolling bias correction, and monthly mean–variance adjustment) and a hydro-adaptive probabilistic optimization in GA2. SHAP-based analysis provides physical interpretability of the learned relations. The regressive stage (GA1) generates a bias-corrected and climatologically consistent central forecast. After the full four-step optimization, GA1 achieves robust generalization skill during the independent test period (2020–2023), yielding NSE = 0.77 ± 0.05, KGE = 0.85 ± 0.05, R2 = 0.77 ± 0.05, and RMSE = 20.2 ± 3.1 m3 s−1, representing a major improvement over raw MLP/XGB outputs (NSE ≈ 0.5). Time-series, scatter, and seasonal diagnostics confirm accurate reproduction of wet- and dry-season dynamics, absence of low-frequency drift, and preservation of seasonal variance. The probabilistic stage (GA2) constructs a hydro-adaptive prediction interval whose width (max-min streamflow) and asymmetry evolve with seasonal hydrological regimes. The optimized configuration achieves comparative coverage COV = 0.86 ± 0.00, hit rate p = 0.96 ± 0.04, and relative width r = 2.40 ± 0.15, correctly expanding uncertainty during wet-season peaks and contracting during dry-season recessions. SHAP analysis reveals a coherent predictor hierarchy dominated by streamflow persistence, precipitation structure, temperature extremes, and evapotranspiration, jointly explaining most of the predictive variance. By combining regressive precision, probabilistic realism, and interpretability within a unified evolutionary architecture, the HHFS provides a transparent, physically grounded, and operationally robust tool for reservoir management, drought monitoring, and hydro-climatic early-warning systems in data-limited regions.
Open AccessIntegrating Regressive and Probabilistic Streamflow Forecasting via a Hybrid Hydrological Forecasting System: Application to the Paraíba do Sul River Basin
This study introduces the Hybrid Hydrological Forecast System (HHFS), a dual-stage, data-driven framework for monthly streamflow forecasting at the Santa Branca outlet in the upper Paraíba do Sul River Basin, Brazil. The system combines two nonlinear regressors, Multi-Layer Perceptron (MLP) and extreme Gradient Boosting (XGB), calibrated through a structured four-step evolutionary procedure in GA1 (hydrological weighting, dual-regime Ridge fusion, rolling bias correction, and monthly mean–variance adjustment) and a hydro-adaptive probabilistic optimization in GA2. SHAP-based analysis provides physical interpretability of the learned relations. The regressive stage (GA1) generates a bias-corrected and climatologically consistent central forecast. After the full four-step optimization, GA1 achieves robust generalization skill during the independent test period (2020–2023), yielding NSE = 0.77 ± 0.05, KGE = 0.85 ± 0.05, R2 = 0.77 ± 0.05, and RMSE = 20.2 ± 3.1 m3 s−1, representing a major improvement over raw MLP/XGB outputs (NSE ≈ 0.5). Time-series, scatter, and seasonal diagnostics confirm accurate reproduction of wet- and dry-season dynamics, absence of low-frequency drift, and preservation of seasonal variance. The probabilistic stage (GA2) constructs a hydro-adaptive prediction interval whose width (max-min streamflow) and asymmetry evolve with seasonal hydrological regimes. The optimized configuration achieves comparative coverage COV = 0.86 ± 0.00, hit rate p = 0.96 ± 0.04, and relative width r = 2.40 ± 0.15, correctly expanding uncertainty during wet-season peaks and contracting during dry-season recessions. SHAP analysis reveals a coherent predictor hierarchy dominated by streamflow persistence, precipitation structure, temperature extremes, and evapotranspiration, jointly explaining most of the predictive variance. By combining regressive precision, probabilistic realism, and interpretability within a unified evolutionary architecture, the HHFS provides a transparent, physically grounded, and operationally robust tool for reservoir management, drought monitoring, and hydro-climatic early-warning systems in data-limited regions.
- 15 October 2025
- Published by Springer Nature in Computational and Applied Mathematics
- Vol. 45 (2) , 1-35
A machine learning algorithm for uncertainty quantification in climate precipitation prediction over South America
- 15 October 2025
- Published by Springer Nature in Computational and Applied Mathematics
- Vol. 45 (2) , 1-35
A machine learning algorithm for uncertainty quantification in climate precipitation prediction over South America
- Open Access
- 11 September 2025
- Published by MDPI AG in Meteorology
- Vol. 4 (3) , 25
Parameterization by Statistical Theory on Turbulence Applied to the BAM-INPE Global Meteorological Model
A parameterization for the planetary boundary layer (PBL) based on the statistical theory of turbulence formulated by Geoffrey Ingram Taylor is derived to be applied in the Brazilian Global Atmospheric Model (BAM). The BAM model is the operational system employed by the National Institute for Space Research (INPE), Brazil, to produce numerical weather and climate predictions. A comparison of the BAM model simulations using Taylor’s parameterization is carried out against other three turbulent representations. The forecasting from different parameterizations with BAM is evaluated with the ERA-5 reanalysis. Predictions were performed on different initial conditions, representing two types of climate seasons: dry and wet seasons, for the Southern Hemisphere. The comparison shows that Taylor’s approach is competitive with other turbulence parameterizations, especially for the dry season. It must be highlighted that the forecasting over the Amazon region—one of the regions on the planet with the most intense rainfall, where Taylor’s approach provided more effective precipitation forecasting, a particularly challenging meteorological variable to predict.
- 11 September 2025
- Published by MDPI AG in Meteorology
- Vol. 4 (3) , 25
Open AccessParameterization by Statistical Theory on Turbulence Applied to the BAM-INPE Global Meteorological Model
A parameterization for the planetary boundary layer (PBL) based on the statistical theory of turbulence formulated by Geoffrey Ingram Taylor is derived to be applied in the Brazilian Global Atmospheric Model (BAM). The BAM model is the operational system employed by the National Institute for Space Research (INPE), Brazil, to produce numerical weather and climate predictions. A comparison of the BAM model simulations using Taylor’s parameterization is carried out against other three turbulent representations. The forecasting from different parameterizations with BAM is evaluated with the ERA-5 reanalysis. Predictions were performed on different initial conditions, representing two types of climate seasons: dry and wet seasons, for the Southern Hemisphere. The comparison shows that Taylor’s approach is competitive with other turbulence parameterizations, especially for the dry season. It must be highlighted that the forecasting over the Amazon region—one of the regions on the planet with the most intense rainfall, where Taylor’s approach provided more effective precipitation forecasting, a particularly challenging meteorological variable to predict.
- Open Access
- 24 February 2025
- Published by MDPI AG in Meteorology
- Vol. 4 (1) , 6
Machine Learning with Voting Committee for Frost Prediction
A machine learning (ML)-based methodology for predicting frosts was applied to the southern and southeastern regions of Brazil, as well as to other countries including Uruguay, Paraguay, northern Argentina, and southeastern Bolivia. The machine learning model (using TensorFlow (TF)) was compared to the frost index (IG from the Portuguese: Índice de Geada) developed by the National Institute for Space Research (INPE, Brazil). The IG is estimated using meteorological variables from a regional weather numerical model (RWNM). After calculating the two indices using the ML model and the RWNM, a voting committee (VC) was trained to select between the computed outputs. The AdaBoostClassifier algorithm was employed to implement the voting committee. The study area was subdivided into three distinct subregions: R1 (outside Brazil), R2 (the south of Brazil), and R3 (southeastern Brazil). Two forecasting time scales were evaluated: 24 h and 72 h. The 24 h forecasts from both approaches (TF and RWNM) exhibited a similar performance in terms of the number of accurate predictions. However, in the region covering Uruguay and northern Argentina, the TensorFlow model demonstrated superior frost prediction accuracy. Additionally, the TensorFlow model outperformed the RWNM for the 72 h forecast horizon.
- 24 February 2025
- Published by MDPI AG in Meteorology
- Vol. 4 (1) , 6
Open AccessMachine Learning with Voting Committee for Frost Prediction
A machine learning (ML)-based methodology for predicting frosts was applied to the southern and southeastern regions of Brazil, as well as to other countries including Uruguay, Paraguay, northern Argentina, and southeastern Bolivia. The machine learning model (using TensorFlow (TF)) was compared to the frost index (IG from the Portuguese: Índice de Geada) developed by the National Institute for Space Research (INPE, Brazil). The IG is estimated using meteorological variables from a regional weather numerical model (RWNM). After calculating the two indices using the ML model and the RWNM, a voting committee (VC) was trained to select between the computed outputs. The AdaBoostClassifier algorithm was employed to implement the voting committee. The study area was subdivided into three distinct subregions: R1 (outside Brazil), R2 (the south of Brazil), and R3 (southeastern Brazil). Two forecasting time scales were evaluated: 24 h and 72 h. The 24 h forecasts from both approaches (TF and RWNM) exhibited a similar performance in terms of the number of accurate predictions. However, in the region covering Uruguay and northern Argentina, the TensorFlow model demonstrated superior frost prediction accuracy. Additionally, the TensorFlow model outperformed the RWNM for the 72 h forecast horizon.
- Open Access
Assessing NOAA/GFDL Models Performance for South American Seasonal Climate: Insights from CMIP6 Historical Runs and Future Projections
Climate prediction is of fundamental importance to various sectors of society and the economy, as it can predict the likelihood of droughts or excessive rainfall in vulnerable regions. Climate models are useful tools in producing reliable climate forecasts, which have become increasingly vital due to the rising impacts of climate change. As global temperatures rise, changes in precipitation patterns are expected, increasing the importance of reliable seasonal forecasts to support planning and adaptation efforts. In this study, we evaluated the performance of NOAA/GFDL models from CMIP6 simulations in representing the climate of South America under three configurations: atmosphere-only, coupled ocean-atmosphere, and Earth system. Our analysis revealed that all three configurations successfully captured key climatic features, such as the South Atlantic Convergence Zone (SACZ), the Bolivian High, and the Intertropical Convergence Zone (ITCZ). However, coupled models exhibited larger errors and lower correlation (below 0.6), particularly over the ocean and the South American Monsoon System, which indicates a poor representation of precipitation compared with atmospheric models. The coupled models also overestimated upward motion linked to the southern Hadley cell during austral summer and underestimated it during winter, whereas the atmosphere-only models more accurately simulated the Walker circulation, showing stronger vertical motion around the Amazon. In contrast, the coupled models simulated stronger upward motion over Northeast Brazil, which is inconsistent with reanalysis data. Moreover, we provided insights into how model biases may evolve under climate change scenarios. Future climate projections for the mid-century period (2030–2060) under the SSP2-4.5 and SSP5-8.5 scenarios indicate significant changes in the global energy balance, with an increase of up to 0.9 W/m2. Additionally, the projections reveal significant warming and drying in most of the continent, particularly during the austral spring, accompanied by increases in sensible heat flux and decreases in latent heat flux. These findings highlight the risk of severe and prolonged droughts in some regions and intensified rainfall in others. By identifying and quantifying the biases inherent in climate models, this study provides insights to enhance seasonal forecasts in South America, ultimately supporting strategic planning, impact assessments, and adaptation strategies in vulnerable regions.
Open AccessAssessing NOAA/GFDL Models Performance for South American Seasonal Climate: Insights from CMIP6 Historical Runs and Future Projections
Climate prediction is of fundamental importance to various sectors of society and the economy, as it can predict the likelihood of droughts or excessive rainfall in vulnerable regions. Climate models are useful tools in producing reliable climate forecasts, which have become increasingly vital due to the rising impacts of climate change. As global temperatures rise, changes in precipitation patterns are expected, increasing the importance of reliable seasonal forecasts to support planning and adaptation efforts. In this study, we evaluated the performance of NOAA/GFDL models from CMIP6 simulations in representing the climate of South America under three configurations: atmosphere-only, coupled ocean-atmosphere, and Earth system. Our analysis revealed that all three configurations successfully captured key climatic features, such as the South Atlantic Convergence Zone (SACZ), the Bolivian High, and the Intertropical Convergence Zone (ITCZ). However, coupled models exhibited larger errors and lower correlation (below 0.6), particularly over the ocean and the South American Monsoon System, which indicates a poor representation of precipitation compared with atmospheric models. The coupled models also overestimated upward motion linked to the southern Hadley cell during austral summer and underestimated it during winter, whereas the atmosphere-only models more accurately simulated the Walker circulation, showing stronger vertical motion around the Amazon. In contrast, the coupled models simulated stronger upward motion over Northeast Brazil, which is inconsistent with reanalysis data. Moreover, we provided insights into how model biases may evolve under climate change scenarios. Future climate projections for the mid-century period (2030–2060) under the SSP2-4.5 and SSP5-8.5 scenarios indicate significant changes in the global energy balance, with an increase of up to 0.9 W/m2. Additionally, the projections reveal significant warming and drying in most of the continent, particularly during the austral spring, accompanied by increases in sensible heat flux and decreases in latent heat flux. These findings highlight the risk of severe and prolonged droughts in some regions and intensified rainfall in others. By identifying and quantifying the biases inherent in climate models, this study provides insights to enhance seasonal forecasts in South America, ultimately supporting strategic planning, impact assessments, and adaptation strategies in vulnerable regions.