Interpolating Electron Diffusion Coefficients in Earth’s Radiation Belts Based on A Neural Network Model Inspired by Video Frame Interpolation

Many numerical models for studying radiation belt dynamics have been built to uncovering the physical mechanisms governing electron dynamics and enabling real-time forecasting. A critical input for these models is the diffusion coefficient obtained through linear interpolation from precomputed diffusion coefficient libraries to achieve real-time processing. However, linear interpolation unavoidably introduces overlap issues, compromising both the accuracy and realism of the simulation results. In this study, we propose a diffusion Coefficient Interpolation Neural nEt (CINE) model, inspired by video frame interpolation, to address overlap issues. The CINE model does not require any preexisting diffusion coefficients for training and successfully interpolates diffusion coefficients induced by various physical mechanisms. We also analyze optimal interpolation intervals for different diffusion coefficients (ΔL≤0.8 for hiss waves and ΔL≤0.2 for atmospheric collisions) based on a threshold of Structural Similarity Index Measure (SSIM)=0.98. The CINE model is easy to incorporate with current radiation belts models to obtain accurate and prompt simulation results for real-time forecasting.