Application of advanced data mining methods to various types of geochemical data is able to fingerprint valid signatures of mineralization, thus unveiling ore genesis and discovering new minerals. But individual studies that apply data mining methods to both local- and regional-scale, both sediment and whole-rock multi-element geochemical data sets are relatively scarce. Here, we applied data mining methods, including multivariate statistical analysis (principal component analysis), spatial analysis (trend surface analysis), unsupervised machine learning algorithm (K-means clustering), supervised algorithms (random forest and deep neural network) to both regional sediment geochemical and local lithogeochemical data from the Duolun-Guyuan prospect, in order to determine the geochemical signatures of volcanic-type uranium mineralization through characterizing: (1) representative element associations; (2) axial zonation of primary haloes; (3) element distribution patterns; and (4) crustal structures (via deep learning-based predictive hafnium (Hf) isotopic mapping). Results of principal component analysis and random forest show that samples from known ore districts (e.g., Zhangmajing and Daguanchang) exhibit a distinct combination of major ore-forming elements (U and Mo), chalcophile elements (Ag, Hg, Pb, Sb and As), rare and rare earth elements (Be, Li, La, Nb and Y), tungsten (W), bismuth (Bi), and rock-forming elements (SiO2, K2O, Na2O and Al2O3), differing from samples of both the mineralized and barren areas. The axial zonation of primary haloes in Daguanchang is comprised of supra-ore haloes (rare earth elements, Th, Nb, Zr, Hf, Ga and Rb), near-ore haloes (U, Mo, Pb, Zn, Cd and Sb), and sub-ore haloes (Li, Be, Sc, V, Cu, Sr, Cs, Ba, W and Bi). Moreover, trend surface analysis shows that in the study area, the spatial distribution pattern of the supra-, near-, and sub-ore elements forms a northwesterly alignment, with the supra-ore elements concentrated in the southeast, the sub-ore elements in the northwest, and the near-ore elements in between. Finally, deep learning-based predictive hafnium (Hf) isotopic mapping reveals that the Duolun-Guyuan prospect is dominated by negative mean zircon εHf(t) values ranging from −17 to 0, except for some local areas in the west and southwest of Duolun and the north of Weichang. The above results may indicate critical signatures of volcanic-type U mineralization, consisting of meta- or pera-luminous, alkaline rhyolite resulted from crustal reworking, surrounding mantle-derived igneous rocks, proximal heat source, accompanying epithermal deposits (e.g., Ag, Au, etc.), and anomalous concentrations of U, Mo and relevant elements particularly Th, W, Bi, Ag and Sb etc. Our study will effectively provide new exploration geochemical indicators of volcanic-type U deposit.
Data mining for geochemical signatures of volcanic-type uranium mineralization, Duolun-Guyuan prospect, North China
Domenico Cicchella
2024-01-01
Abstract
Application of advanced data mining methods to various types of geochemical data is able to fingerprint valid signatures of mineralization, thus unveiling ore genesis and discovering new minerals. But individual studies that apply data mining methods to both local- and regional-scale, both sediment and whole-rock multi-element geochemical data sets are relatively scarce. Here, we applied data mining methods, including multivariate statistical analysis (principal component analysis), spatial analysis (trend surface analysis), unsupervised machine learning algorithm (K-means clustering), supervised algorithms (random forest and deep neural network) to both regional sediment geochemical and local lithogeochemical data from the Duolun-Guyuan prospect, in order to determine the geochemical signatures of volcanic-type uranium mineralization through characterizing: (1) representative element associations; (2) axial zonation of primary haloes; (3) element distribution patterns; and (4) crustal structures (via deep learning-based predictive hafnium (Hf) isotopic mapping). Results of principal component analysis and random forest show that samples from known ore districts (e.g., Zhangmajing and Daguanchang) exhibit a distinct combination of major ore-forming elements (U and Mo), chalcophile elements (Ag, Hg, Pb, Sb and As), rare and rare earth elements (Be, Li, La, Nb and Y), tungsten (W), bismuth (Bi), and rock-forming elements (SiO2, K2O, Na2O and Al2O3), differing from samples of both the mineralized and barren areas. The axial zonation of primary haloes in Daguanchang is comprised of supra-ore haloes (rare earth elements, Th, Nb, Zr, Hf, Ga and Rb), near-ore haloes (U, Mo, Pb, Zn, Cd and Sb), and sub-ore haloes (Li, Be, Sc, V, Cu, Sr, Cs, Ba, W and Bi). Moreover, trend surface analysis shows that in the study area, the spatial distribution pattern of the supra-, near-, and sub-ore elements forms a northwesterly alignment, with the supra-ore elements concentrated in the southeast, the sub-ore elements in the northwest, and the near-ore elements in between. Finally, deep learning-based predictive hafnium (Hf) isotopic mapping reveals that the Duolun-Guyuan prospect is dominated by negative mean zircon εHf(t) values ranging from −17 to 0, except for some local areas in the west and southwest of Duolun and the north of Weichang. The above results may indicate critical signatures of volcanic-type U mineralization, consisting of meta- or pera-luminous, alkaline rhyolite resulted from crustal reworking, surrounding mantle-derived igneous rocks, proximal heat source, accompanying epithermal deposits (e.g., Ag, Au, etc.), and anomalous concentrations of U, Mo and relevant elements particularly Th, W, Bi, Ag and Sb etc. Our study will effectively provide new exploration geochemical indicators of volcanic-type U deposit.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.