Five lunar oxide abundance maps for the polar regions (65°–90°N/S) derived from KAGUYA SP data

This dataset provides high-resolution abundance maps of five major lunar oxides (FeO, TiO₂, Al₂O₃, CaO, and MgO) for the lunar north and south polar regions, covering latitudes from 65° to 90° in both hemispheres, at a spatial resolution of approximately 1 km/pixel. The oxide abundance products were generated using a one-dimensional convolutional neural network (1D-CNN) inversion model applied to hyperspectral data acquired by the KAGUYA Spectral Profiler (SP). The model was trained and validated using oxide abundance results from Bian et al. (2025)(doi: 10.1109/JSTARS.2025.3549691) as sample data, achieving high inversion accuracy (R² > 0.94 for all oxides and RMSE < 0.3 wt.% for FeO, TiO₂, CaO, and MgO). The resulting oxide maps characterize the geochemical composition of the lunar polar regions and enable quantitative analysis of compositional variability in permanently shadowed and sunlit areas. These data are intended to support studies of lunar crustal evolution, polar resource assessment, and landing site evaluation, and provide geochemical reference information for future lunar exploration missions, including Artemis III and Chang’E-7.