Advanced Geomagnetic Storm Forescasting

---

Geomagnetic storms pose significant risks to modern technologies, particularly GPS satellite systems and electrical power grids. These storms result from solar wind and storms from the Sun that affect Earth’s magnetic field. Accurate prediction of these storms is crucial but challenging due to the variable travel time of solar material to Earth. The Deep Space Climate Observatory (DSCOVR) provides critical data for forecasting geomagnetic storms by measuring solar wind parameters. However, with its mission extended beyond the initial five years and instruments experiencing sensitivity issues, there is an urgent need for improved prediction models. 

This project aims to develop an integrated database combining data from DSCOVR and CASSIOPE missions to predict the planetary K-index (Kp) with a lead time of 20 minutes to 1 hour. The integrated database will serve as input for advanced predictive models of geomagnetic activity. The methodology involves collecting data from 2016 to 2023, preprocessing to clean and normalize the data, temporal synchronization of timestamps, and data integration. A neural network model is implemented to classify Kp values using historical data, with the aim of providing early warnings of geomagnetic storms. 

The results include a detailed correlation analysis between various solar wind features and the Kp index, identifying key predictors for geomagnetic activity. The machine learn- ing model shows good overall performance with an accuracy of 82%, but specific areas need further optimization. This integrated approach enhances the predictive capabilities for geomagnetic storms, contributing to better preparedness and mitigation of their im- pacts on modern technology.