Geophysical reference-models test datasets and neural-network surrogate predictions

This dataset contains the independent test data used to evaluate the neural-network surrogates presented in “A Neural-Network Surrogate Workflow for Accelerating the Evaluation of Empirical Geophysical Models.” Separate datasets are provided for NRLMSISE-00, IRI 2020, an IGRF-14-based magnetic-field model, and an EGM-96-based geopotential model Each dataset contains the original model inputs, outputs generated by the corresponding private reference C++ implementation used in Juliette software (orbit processing software), and restored neural-network predictions in physical units. The datasets were used to calculate the mean, median, and 95th-percentile absolute percentage errors, total variation distance for IRI ion composition, and relative vector errors reported in the associated manuscript.