Application of the multivariate canonical trend surface method to the identification of geochemical combination anomalies

• Propose a new nonlinear model for geochemical combination anomaly identification
• The correlation, structure and singularity of spatial data are organically combined.
• Three-dimensional expression of combination anomaly in spatial distribution

In geochemical exploration, geochemically observed data are generally multidimensional. Due to the geochemical characteristics and the diversity of geological conditions, elements often show correlated, structural and singular characteristics in the spatial distribution. In a local mineralization block, these three characteristics are particularly strong. A group correlation between elements is represented by main mineralization factors. The spatial variation of elements under the large-scale condition mostly shows nonlinear characteristics. Based on the canonical correlation principle with spatial observation data, this paper presents a multivariate canonical trend surface method, which expresses a geochemical element multivariable by a regional canonical variable. This regional canonical variable is taken as the main mineralization factor. We combine the main mineralization factor and two-dimensional trend surface constituted in a plane right angle coordinate system with the maximum correlation, to obtain the canonical correlation coefficient. This method reflects not only the correlation between elements, but also the spatial structure and singularity. It must be stated that since this method is a statistical method established by treating spatial observation data, it is not suitable for the simulation of dynamic processes in ore-forming systems, but it can be somehow used to effectively treat spatially discrete observation data. Therefore, the Xichuan District in Baishan City, Jilin Province, northeastern China is chosen as a study area, in which we processed and analyzed 7675 geochemical samples using the proposed multivariate canonical trend surface method. The results using the multivariate canonical trend surface method show that the location of the identified multivariate geochemical anomaly is consistent with the location of the known mineral deposits in the study area. Since the proposed multivariate canonical trend surface method cannot be used to simulate the dynamic processes involved in an ore-forming system, the computational simulation method in the emerging computational geoscience discipline should be used as the first choice in future research, especially for the prediction of deep ore deposits.