QuakePredictEC: Plataforma de análisis y predicción de sismos en Ecuador asociados a variaciones climáticas extremas

Abstract

Ecuador is one of the countries with the highest seismic activity in the world, a condition that demands the exploration of new ways to manage seismic risk and anticipate potential hazards in order to protect and inform the population. This project, called QuakePredictEC, is based on a hypothesis that has been investigated over the past five years: “Although the origin of earthquakes is fundamentally tectonic, certain meteorological variations may alter the equilibrium of the Earth’s crust and favor their occurrence.” Based on this premise, we analyzed the statistical correlation between climatic variables and historical earthquake records in Ecuador, with the aim of training a predictive model for earthquake occurrence. To this end, historical seismic records from the Geophysical Institute of the National Polytechnic School were combined with meteorological data from the OpenMeteo platform. Climatic variations during the 30 days prior to each recorded earthquake were analyzed and processed. Using these data, three machine learning models: Random Forest, Multi-Layer Perceptron, and XGBoost were trained and compared. By analyzing climatic variations such as pressure, temperature, and precipitation over the previous thirty days for each canton in the country, the models aim to predict whether an earthquake will occur in the following days. The results indicated that the XGBoost model was the most effective, achieving a recall (sensitivity) of 88%, meaning it is capable of detecting the vast majority of earthquakes. Rather than providing a definitive prediction, this project offers a decision-support tool for risk management and planning, enabling authorities to move from generalized monitoring to focused surveillance in areas with a higher probability of seismic occurrence.