Neutron Monitor Data and EXPACS-simulated Cosmic Ray Dose Fields for Neural Operator Training

Dataset Description This dataset provides neutron sensor data and physics-based dose-rate simulations for training and validating neural operator models. The data span January 1, 2001 to December 31, 2023, combining sparse ground-based neutron monitor observations with dense atmospheric dose rate fields. Note on versions. This version contains only the measured-station data (the measured_*.zip archives, README_measured.md, MANIFEST.tsv and MD5SUMS). The neutron monitor data and the EXPACS dose fields described below (neutron_data_22yrs.npy, neutron_data_MXCO.npy, dose_array.npy, dose_2001_2023_0m.npy, grid_points.npy, grid_points_025.npy and Dose_City.zip) are in the previous version, 10.5281/zenodo.18406046. Download both versions to reproduce all results of the paper. Data Files Neutron Monitor Observations neutron_data_22yrs.npy: Ground-based neutron count rates from the NMDB network (2001-2023), gridded at 1° spatial resolution. Shape: [time_steps, n_stations]. Data preprocessed from raw NMDB measurements with quality control and normalization applied. neutron_data_MXCO.npy: High-temporal-resolution neutron measurements from Mexico City station for validation studies. Cosmic Ray Dose Rate Fields dose_2001_2023_0m.npy: Effective dose rate fields at ground level (0 m altitude), simulated using EXPACS at 0.25° spatial resolution. Shape: [time_steps, n_lat, n_lon]. Units: μSv/h. dose_array.npy: Effective dose rate fields at 1° spatial resolution, temporally aligned with neutron observations. Shape: [time_steps, n_lat, n_lon]. Units: μSv/h. Spatial Coordinates grid_points.npy: Geographic coordinate arrays (latitude, longitude) for 1° resolution grids. Shape: [n_points, 2]. grid_points_025.npy: Geographic coordinate arrays for 0.25° resolution grids. Shape: [n_points, 2]. Dose_City.zip YYY_dose_2023.npy: Effective dose rate field datasets in the 2023 period at cities (YYY). Data Generation Neutron Monitor Data: Obtained from the Neutron Monitor Database (NMDB, https://www.nmdb.eu) and processed to daily temporal resolution. Individual station operators are acknowledged in the associated publication. Dose Rate Simulations: Generated using the EXPACS software (Copyright 2006 Japan Atomic Energy Agency) with solar modulation parameters derived from contemporary cosmic-ray measurements. Simulations account for geographic variations in geomagnetic cutoff rigidity and atmospheric depth. Physical parameters: Altitude: Sea level (0 m) Atmospheric depth: 1033 g/cm² (representing sea-level atmospheric pressure) Ground environment: Groundwater fraction = 0.15 (accounting for average soil moisture effects) Geomagnetic field: Vertical cutoff rigidity automatically calculated using MAGNETOCOSMICS database at each location Solar modulation: W-index derived from neutron monitor data for each simulation date Data Format All files are NumPy binary format (.npy) readable with numpy.load(). Temporal coverage: 8,400 days (2001-01-01 to 2023-12-31). Measured station networks (added in this version) This version adds the data behind the measured-station-network experiments of the paper (main-text Fig. 8; Supplementary Section S11, Tables S10–S17): prepared GHCN-Daily and PeakWeather arrays with the input, training and held-out station sets, prepared arrays of the datasets reported as negative results (NOAA OI SST v2, SNOTEL, TAO/TRITON gate output), the ETOPO1 subsets used for the elevation maps, and the per-run metrics, checkpoints and held-out predictions. Archive contents, sources and terms are described in README_measured.md; MANIFEST.tsv lists every file with its SHA-256. The arrays are derived from GHCN-Daily (NOAA NCEI, CC0-1.0; modified, not original NOAA data), PeakWeather (CC-BY-4.0, source: MeteoSwiss), NOAA OI SST v2 (NOAA PSL), SNOTEL (USDA NRCS) and ETOPO1 (NOAA). We acknowledge the GTMBA Project Office of NOAA/PMEL for the TAO/TRITON data. Citations When using this dataset, please cite: The associated publication: Kobayashi, K., Tripura, T., Yoo, J. P., Abueidda, D., Koric, S., Chakraborty, S., & Alam, S. B. Spatiotemporal Reconstruction of Global Radiation from Sparse Sensor Sequences. Nature Communications (in press); arXiv:2506.12045. NMDB: Neutron Monitor Database, www.nmdb.eu, founded under EU FP7 programme, contract no. 213007 EXPACS: Sato, T. (2016). Analytical Model for Estimating the Zenith Angle Dependence of Terrestrial Cosmic Ray Fluxes. PLOS ONE, 11(8): e0160390. https://doi.org/10.1371/journal.pone.0160390 Sato, T. (2015). Analytical model for estimating terrestrial cosmic ray fluxes nearly anytime and anywhere in the world: Extension of PARMA/EXPACS. PLOS ONE, 10(12): e0144679. https://doi.org/10.1371/journal.pone.0144679