Quantitative Assessment Methods Of Geochemical Patterns And Its Application In Mineral Exploration

Geochemical data are a direct means of obtaining mineralization information and play an important role in mineral exploration. During long-term geochemical exploration work, a large number of high-quality, multi-scale, multi-medium geochemical data have been accumulated, which provide a rich source of data for implementing geochemical exploration research. This dissertation focuses on the study of new methods of information extraction, and provides support for mineral exploration in the study area.Spatial distributions of elemental concentration are controlled by a variety of geological processes, so the study on the spatial distribution pattern of geochemical variables is helpful to the interpretation of geochemical processes. Therefore, quantitative evaluation of distribution patterns of geochemical elements has great significance for mineral exploration. Existing quantitative analysis methods can be divided into two categories:frequency domain methods and spatial domain methods. Traditional analytical methods (including probability plots, multivariate analysis, trend surface analysis method, etc.) ignore the spatial characteristics of geochemical data. The development of spatial statistics (including gcostatistics, fractal/multifractal methods, spatial autocorrelation index, etc.) enables quantitative characterization of the spatial patterns of geochemical elements. Spatial statistical methods can be also divided into two categories:global spatial statistics and local spatial statistics.Our study area is a typical vegetation covered area located in Southwestern Fujian, which is one of demonstration areas for the project. As one of the most important Fe metallogenic belts in China, the study area contains almost 98% of the total Fe ore reserves of Fujian. A number of skarn-type Fe deposits (e.g., Makeng, Yangshan, Pantian, Luoyang, Zhongjia, and Panluo) were discovered in this area. These deposits situate primarily in the early Hercynian Yongan-Meixian fold belt, which is composed of a central anticline flanked by two synclinal basins, resting on a Caledonide basement. The dominant ore-hosting rocks are late Paleozoic lithologies, with numerous outcrops discontinuously distributed mainly along the NE-NNE basement faults. The intrusion of the Yanshanian granites, resulting in Fe enrichment, derived primarily from Carboniferous-Permian formations. The stream sediment dataset representing 1:20,0000 scale was chosen to investigate the mineralization features. Previous studies have demonstrated that the skarn-type Fe deposits are strongly correlated to element associations of Cu-Mn-Pb-Zn-Fe2O3. In this study, based on the theory of spatial autocorrelation, global and local spatial statistical methods (including spatial autocorrelation index, geographical weighting methods and their improved algorithms) are used to evaluate spatial patterns of Cu-Mn-Pb-Zn-Fe2O3. The main conclusions as follows:(1) Research on geochemical element behaviorGlobal forms of variogram, autocorrelation index and V-D fractal model were used to investigate elemental behaviors of ore-forming elements. The results show that Fe2O3, Cu, Pb, Mn and Zn have a strongly spatial autocorrelation, and the autocorrelation distance is similar. The fractal dimensions based on V-D fractal model are about 2.9 for Fe2O3, Cu, Pb, Mn and Zn, indicating that Fe2O3, Cu, Pb, Mn and Zn have similar spatial complexities. These statistical laws imply that spatial distribution of Fe2O3, Cu, Pb, Zn and Mn may be controlled by the same or similar geological process (e.g., diagenesis, geological structure and mineralization).(2) Research on spatial dependent between different analysis scales.Geographical weighted statistics was employed to explore the influence of analysis scale on geochemical univariate, bivariate and multivariate analysis. When the scale is too large or too small, it is not conducive to recognize spatial patterns of element. The result of cross-validation score shows that the optimal analysis scale of Fe2O3 is 10 km, which reveals not only spatial trends but also describes local features of Fe2O3. Spatial distribution of local spatial autocorrelation index has subtle variation with changing scale, so local spatial autocorrelation index is stable. For geographically-weighted principal components analysis (GWPCA), the highest loading on GWPC1 associated with Fe polymetallic mineralization was selected to investigate the spatial relationships between the known Fe deposits and main mineralized elements. It is observed that the known skarn-type Fe deposits are linked to different mineralized elements. In the northern and southern parts of study area, the known Fe deposits are associated with Fe2O3, Cu and Mn. The mineralized element for the same mineral deposit may change as the scale changes, indicating that the scale is an important factor in geochemical exploration. These results may indicate the complex geo-chemical background of Fe polymetallic mineralization. These studies provide a new idea for quantitative characterization of geochemical elements.(3) Research on the extraction of geochemical weak anomalies.The trend surface analysis method and S-A model are used to decompose the integrated geochemical pattern obtained from the robust principle component analysis (RPCA) and compared the results obtained from both methods. The obtained anomaly maps were similar, with the high anomaly areas showing a strong spatial relationship with intrusions that are related to Fe polymetallic mineralization, indicating that both the trend surface analysis and the S-A are useful tools to identify geochemical anomalies. The S-A model, based on distinct anisotropic scaling properties, was better in revealing local anomalies because it considered the spatial characteristics of the geochemical variables.(4) Research on interpolation combining multivariable.Geochemical elements often have strongly correlations with each other, and sometimes we can get a small amount of accurate data of element concentration. Therefore, we need to integrate these data during interpolation process. We developed an improved co-kriging method based on the Markov-Chain model, which is useful for reducing the computational complexity and avoiding the constraint of cross-covariance matrix positive definite, and is easier to implement. We give their applicable conditions through comparison with classic co-kriging, MM 1 and MM2 from theory and application results. In this study, the value of grade skarn polymetallic Fe deposit and Fe2O3 samples at 1:20,000 scale were used as primary data and auxiliary data, respectively. The results were compared with ordinary kriging, which show MM2 improved the prediction efficiency vs. ordinary kriging.