Explore the vast range of titles available.

Calcrete-type uranium prospectivity prediction is challenged by the strong heterogeneity of multi-source geoscientific raster datasets, weak anomaly responses, and the lack of explicit heterogeneous information organization in conventional deep learning models. In this study, the Yilgarn Craton of Western Australia was selected as the study area, and a geology-guided fixed-group fusion ResUNet model (GGF-ResUNet) was developed based on 12-channel multi-source geoscientific raster datasets. At the input stage, the evidence layers were divided into four fixed geoscientific proxy groups according to their data modality and geological interpretation, namely gravity, aeromagnetic, radiometric, and geochemical groups, and intra-group channel weighting together with inter-group gating was introduced to enhance the hierarchical representation and adaptive fusion of heterogeneous information. Ablation results showed that GGF-ResUNet achieved better performance than the baseline ResUNet, with AUC increasing from 0.9340 to 0.9740 and F1-score improving from 0.7264 to 0.8356. Further comparative experiments with Attention U-Net, U-Net, SegNet, and FCN showed that GGF-ResUNet achieved comparatively better quantitative performance and more spatially coherent prediction results under the current experimental setting. Without substantially increasing model complexity, the proposed method improves the representation and integration of heterogeneous geoscientific information and provides a feasible technical pathway for calcrete-type uranium prospectivity prediction under weak-anomaly conditions.

What are the main findings? An unsupervised deep learning framework, Geo-U-Mamba, was developed by integrating a Mamba-driven four-directional cross-scan mechanism into a U-Net architecture to model local and global geological features with linear computational efficiency. Evaluations indicate that the model achieves competitive performance in geological background fitting and anomaly separation compared to baseline methods like CAE, U-Net, and ViT. What are the implications of the main findings? The automatically delineated prospective areas align well with major ore-controlling faults and cover most known gold occurrences in the study area, demonstrating the framework’s potential in extracting relevant ore-controlling features. The proposed approach offers a viable quantitative paradigm for mineral prospectivity mapping using multi-source geoscientific data, helping to mitigate non-ore geological noise in complex metallogenic settings. Modern mineral exploration faces the pivotal challenge of detecting concealed mineral deposits in complex geology, as depleting outcropping ores have driven global exploration to depths where 1000 m deep mining is now commonplace. To address this, this study proposes Geo-U-Mamba, an unsupervised deep learning framework for gold mineral prospectivity mapping. The model integrates multi-source geoscientific data, encompassing geochemistry, remote sensing alteration indicators, topography, and structural distance fields. By incorporating a Mamba-driven four-directional cross-scan mechanism into a U-Net architecture, the framework effectively models the complex nonlinear mapping relationships between metallogenic elements and the geological environment. This approach recognizes gold geochemical anomalies with an 86.11% deposit capture rate, decoupling environmental noise by reconstructing the geochemical background field and extracting anomalies in combination with C-A fractal theory. When applied to China’s Hatu gold belt in Xinjiang, Geo-U-Mamba achieved an AUC of 0.83, consistently outperforming classical baselines such as CAE, U-Net, and ViT. Ultimately, the findings indicate that this framework provides a reliable and high-precision tool for modern mineral exploration, successfully separating mineralization signals from geological backgrounds in complex metallogenic belts to facilitate exploration targeting.