Physics-Informed DeepONet for Sparse 4-Dimensional Ionospheric Specification Under Disturbed Conditions During May 2024

This record provides the input data and a reference implementation of the core PI-DeepONet-SAMI3 model associated with the manuscript “Physics-Informed DeepONet for Sparse 4-Dimensional Ionospheric Specification Under Disturbed Conditions During May 2024.” The dataset contains SAMI3 simulation inputs covering 1–31 May 2024. The simulations from 1–30 May are used for training, while the 31 May simulation is used for testing. The file initial_dene_e_20240501_20240531.zip contains the initial electron-density fields used as the Branch-network input. The file coor_lat_alt_20240501_20240531.zip contains the dipole spatial coordinates and time used as the Trunk-network input. The accompanying software package (version 1.0.0) provides a reference implementation of the PI-DeepONet-SAMI3 core model, including the convolutional Branch network, the fully connected Trunk network, Cartesian-product DeepONet output construction, full and sparse sampling losses, the SAMI3 electron-temperature-equation residual, and the optimizer and learning-rate schedule. The physical-constraint weight used in the reported experiments is 1.0 × 10⁻²¹. The same implementation can be configured as the data-driven DeepONet-SAMI3 baseline by disabling the physical-residual term. The dataset is released under the Creative Commons Attribution 4.0 International License (CC BY 4.0). The reference software implementation is released under the MIT License.