Characterizing interplanetary magnetic field fluctuations at arctic using cosmic ray secondaries–An approach with machine learning

Abstract

This study explores the potential of ground-based cosmic ray measurements to quantify solar weather parameters, specifically the total interplanetary magnetic field (B_t). A compact Tissue Equivalent Proportional Counter (TEPC), meant for measuring radiation doses in human tissue, was set up at Ny-Ålesund near the North Pole, a region with zero geomagnetic cut-off, which allows for detailed measurements of the different components of cosmic rays. The TEPC continuously monitored low Linear Energy Transfer (LET) cosmic ray components, mainly electrons, and photons during two different seasons: January to March (winter) and September (summer) in 2024. Monte Carlo simulations using PHITS and EXPACS were carried out to understand the changes in cosmic ray flux related to solar weather. To model the relationship between cosmic ray flux and B_t, two machine learning algorithms were used: Gaussian Process Regression (GPR) and Artificial Neural Networks (ANN). Cosmic ray neutron data from the Oulu neutron monitor, which is part of the global neutron monitor network for studying solar weather, were included in the model. Adding the low LET data increased the R² value in the GPR model from 0.81 to 0.90 on the training data, and in the ANN model from 0.76 to 0.88 in comparison to only neutron data, showing a significant improvement in predictive ability. The results show a significant correlation between cosmic ray variations and B_t, suggesting that ground based cosmic ray data collected at low geomagnetic cut-offs—as captured by the TEPC in Ny-Ålesund, can be a reliable way to estimate B_t, especially when satellite data is unavailable. This approach offers a promising, cost-effective method for continuous solar weather monitoring, providing valuable insights into the effect of solar activity on cosmic rays, in turn, helping to make space-based technological systems more resilient.