The proposed work will improve on past studies, a majority of which focus on the near-surface (>20 MM) layer, by probing deeper layers of the convection zone (up to 60 MM of depth) using both…

The New Jersey Institute of Technology (NJIT) has been awarded a $515,973 research grant by a civilian federal agency to conduct a study on forecasting active region emergence using helioseismic and machine learning methods. The research aims to advance solar physics by examining subsurface layers of the sun's convection zone up to 60 millimeters in depth through both established and novel helioseismic techniques. The project will analyze approximately 100 active regions to evaluate the statistical reliability of pre-emergence signatures and will employ advanced computational methods, including convolutional neural networks (CNNs) and long short-term memory (LSTM) models, to identify patterns and improve prediction capabilities. The work builds upon previous studies by incorporating deeper subsurface analysis and integrating helioseismic data with machine learning approaches to better understand flux emergence processes.

The contract is scheduled for completion by May 22, 2028, with performance conducted at NJIT's Newark, New Jersey facility. This award reflects the institute's established expertise in solar physics research and artificial intelligence applications, capabilities demonstrated through prior federal funding from agencies including NASA and the National Science Foundation. The research has potential applications for space weather prediction and solar activity forecasting, which support broader national interests in understanding solar phenomena and their effects on Earth-based systems and infrastructure.

The proposed work will improve on past studies, a majority of which focus on the near-surface (>20 MM) layer, by probing deeper layers of the convection zone (up to 60 MM of depth) using both established and novel helioseismic techniques. This will study the flux emergence process using the previously described subsurface sound speed and flow maps generated from HMI dopplergrams. The team will generate maps up to 60 MM beneath the surface with time-distance methods in the deep-focusing regime, using both traditional and modified techniques, which preliminary testing has shown to have a greater signal-to-noise ratio. The team will also use the shallower sound speed and flow maps provided by jsoc, whose advantage lies in their extensive testing and validation. This will allow to study the flux emergence process from the upper convection zone to very near to the solar surface. We will use a sample of around 100 active regions to evaluate the statistical reliability of pre-emergence signatures, a figure suggested by past work as being sufficient for this task. For the machine learning portion of this investigation, the team will initially use convolutional neural networks (cnns) which excel at image analysis and feature recognition. The determination of parameters for the CNN will be a sub-task of this investigation, and other appropriate ML methods such as long short-term memory (lstm) models will be explored.