Solar Wind Speed Variations as Precursor Through X-Class Flare Intensity and Wind Density: A Radio Data Analysis

Objectives: This study aims to identify solar wind speed as a precursor to the giant class of solar flares, focusing on flare intensity and wind density. Methods: A three-tier analysis was conducted on radio data from X-class solar flares on January 9, 2023, and May 11, 2024. Forecasting involved analysing wind speed variations w.r.t flare intensity and wind density. Firstly the data processing included filtration, moving average regressions, Holt-Winters and ARIMA models to identify the best fit. Secondly, the forecasted data were scaled to solar parameter ranges, revealing binary time-scale dependencies. Boolean feature extraction was employed to determine wind parameter frequency. Final predictions were made using an ensemble Random Forest model. Temporal parameter variations of geomagnetic indices are also related to identifications of the perturbation. Findings: Machine learning models achieved R² scores of 0.503±0.0002 and 0.7125±0.0035 for the two dates. RMSE analysis showed a swing of |-4.94 ± 3.811| for May 11, 2024. Post-flare (for 1-18 hrs. passage) wind speed (kms-1) and density (particlesm-³) differences are |±1.882 ±0.2543| and |±0.257±0.01055| respectively. Boolean expressions revealed solar parameter time dependencies, with frequency analysis yielding a mean of 49.90±0.26 and standard deviation of 28.45±0.05. The Random Forest model yielded an R² of 0.8514±0.0012. A Dst change of 18 nT was observed with post-flare enhanced solar wind. Novelty: This study introduces a state-of-the-art approach through a three-tier forecasting methods, combining model fitness analysis, Boolean function feature extraction and prediction. This framework captures the unique dependencies between solar wind speed, flare intensity and density highlighting the retention of increased average wind speed following a flare. Keywords: Solar Flare, Solar Wind Speed, Regression, HW, ARIMA, Random Forest Model