No Detail Found!

GeoHybridGNN for Next-Generation Porphyry Copper Targeting: Hybrid Remote Sensing, Graph Learning, and Geochemical Intelligence in the Western Chagai Belt, Pakistan

Author(s)
M. Atif Bilal College of Geoexploration Science and Technology, Jilin University,Changchun ,China
Kateryna Hlyniana School of Mathematics, Jilin University, Changchun , China
Muhammad Pervez Akhter Department, National University of Modern Languages, Faisalabad , Pakistan
Yongzhi Wang College of Geoexploration Science and Technology, Jilin University, Changchun , China; Institute of Integrated Information for Mineral Resources Prediction, Jilin University, Changchun , China
Abstract
Porphyry copper exploration increasingly requires intelligent spatial frameworks capable of integrating remote sensing, geological structure, lithology, and geochemical evidence within a unified prospectivity model. In the Western Chagai Belt of Pakistan, major copper-gold (Cu-Au) systems occur in a structurally complex arc environment where mineralization is controlled by intrusive activity, fault corridors, hydrothermal alteration, and geochemical enrichment. This creates a need for advanced prospectivity models capable of integrating multi-source geological evidence into spatially coherent exploration targets. This study presents GeoHybridGNN, a hybrid deep learning framework designed for next-generation porphyry copper prospectivity mapping. The approach combines a CNN-based raster-patch encoder with a GraphSAGE-based graph learning model. Remote-sensing-derived hydrothermal alteration predictors, Cu geochemical information, distance-to-fault data, and lithology-based node attributes were integrated on a common spatial grid. The CNN branch captures local raster-patch patterns related to alteration and surface expression, while the graph branch aggregates geological and raster-derived node information through spatial neighborhood learning. The results show that GeoHybridGNN generates prospectivity maps that are spatially coherent and consistent with the main mineralized corridor of the Western Chagai Belt. Compared with tested graph-based models, the hybrid framework improves target organization and produces more interpretable exploration patterns. Cu geochemical integration has limited effect on global discrimination but provides modest local sharpening in selected target zones, suggesting its value as a refinement layer rather than a dominant regional predictor. This work demonstrates how hybrid remote sensing, graph learning, and geochemical intelligence can support uncertainty aware mineral targeting in structurally complex and arid terrains. The framework provides a practical pathway toward AI-assisted mineral exploration screening and more transparent geological decision-making
Publication Details
Page(s) 1-93
DOI DOI not available
Published Journal: Book of Abstracts: 8th International Conference "Earth Sciences Pakistan 2026" (ESP-2026), June 12-14, 2026, Volume: 1, Issue: 1, Year: 2026

Request the full text from PASTIC

Keywords
GeoHybridGNN, Geochemistry, Remote Sensing, Porphyry Copper prospectivity, Mineral targeting, Graph neural networks
0

Citations

0

Downloads

12

Views