Dataset for Teleconnection-Based Modeling of Extreme Precipitation in South America Using Climate Indices (2002-2024)

Abstract

This dataset supports the analysis presented in the manuscript “Extreme Precipitation in South America: A Qualitative Linear Modeling Approach Using Climate Indices and Machine Learning Techniques”. It includes the filtered time series of extreme precipitation events (PM95 and pm5), predictor matrices of climate indices (CIs), regression model coefficients and intercepts, forecasts, error metrics, and variable importance rankings used throughout the study. The methodology combines hierarchical and K-means clustering, dimensionality reduction via NMF and PCA, and a multi-output regression framework using Linear, Ridge, Lasso, and ElasticNet algorithms. All files are organized and labeled following a structured convention explained in the accompanying README.txt. Embargo Notice: This dataset is currently under embargo and available only for peer review. Upon acceptance of the manuscript, it will be made fully public and assigned a DOI. For questions, contact the corresponding author.