Orogenic gold deposits host a substantial proportion of global gold resources, yet their internal alteration architecture and mineral assemblage variability are commonly reconstructed from random samples, limiting our ability to resolve along-profile zonation and its linkage to gold enrichment. Here we evaluate hyperspectral drill-core spectroscopy, integrated with surface spectroscopy and petrographic validation, as a rapid and spatially continuous approach to delineate alteration zoning in the Liba orogenic gold deposit (West Qinling Orogen, China). We acquired hyperspectral spectra (0.35–2.50 μm) from 255 evenly spaced surface points across two orebodies and from nine representative drill cores scanned at 1 m intervals, and organized the spectral dataset according to Au-grade domains (0.5 g/t). Spectra were quality controlled and interpreted using ENVI-based spectral library matching and The Spectral Geologist (TSG) processing workflows. Petrographic observations from 76 polished thin sections provide independent mineralogical constraints. The hyperspectral results resolve a systematic alteration progression from barren chlorite-dominated assemblages (Au-grade < 0.05 g/t) (locally with minor carbonate) through a weakly mineralized transition (Au-grade 0.05–0.5 g/t) characterized by diminishing chlorite and emergent kaolinite–carbonate with intensified sericitization, to an ore-grade (Au-grade > 0.5 g/t) assemblage dominated by kaolinite–sericite–carbonate–pyrite where biotite persists in distal and proximal zones but was progressively replaced by chlorite during subsequent hydrothermal stages; its modal abundance exhibits a clear inverse correlation with mineralization intensity. These results demonstrate that hyperspectral core scanning, coupled with targeted validation, can rapidly reconstruct a three-dimensional alteration architecture and provide practical mineralogical vectors for exploration targeting and process-oriented studies in orogenic gold systems.
Accurate landslide susceptibility mapping (LSM) is important for hazard prevention and land use planning in mountainous regions. Existing machine learning and deep learning methods mainly use raster-based conditioning factors. They often ignore landslide-related attribute information and spatial context. To address this issue, this study proposes an image–tabular joint deep learning framework for regional-scale LSM. The framework is based on a FiLM-conditioned U-Net. The model combines raster patches with an estimated soft attribute-prior vector and uses FiLM to guide condition-aware spatial feature learning. The proposed framework was tested in the Tacheng region, Xinjiang, China. The dataset includes a landslide inventory and conditioning factors related to terrain, hydrology, vegetation, geology, land cover, and human activities. Model performance was evaluated using stratified five-fold cross-validation, an independent test set, buffer-radius sensitivity tests, and spatial hold-out validation. FiLM-U-Net achieved the best performance among the tested models. It obtained an accuracy of 89.73%, an F1-score of 89.41%, and an AUC of 0.953 on the independent test set. In the spatial hold-out validation area, the model achieved an AUC of 0.921. Feature importance analysis showed that distance to roads, rainfall, NDVI, and terrain factors provided important predictive information. These results suggest that the proposed image–tabular joint framework can improve condition-aware feature learning and support regional landslide susceptibility assessment.