Identification of geochemical anomalies via local RX anomaly detector

Multivariate geochemical anomalies are of great significance to the mineral exploration. The general method for multivariate geochemical anomalies is application of a hybrid method such as combining principal component analysis (PCA) and local singularity analysis (LSA). However, the unknown multivariate probability distribution of the geochemical data may could not meet the application condition of PCA for the detection of multivariate geochemical anomalies. In this study, the local RX anomaly detector based on double sliding windows was used to detect multivariate geochemical anomalies. Based on the idea of the local data of nonlinear manifold can be approximated linearly, the local RX anomaly detector converted the global nonlinear problem into a local linear problem in the multidimensional feature space of the geochemical data. The geochemical data from southwestern Fujian district (China) were carried out to validate the method. The anomaly map showed that majority of skarn Fe deposits are situated in areas with high value of RX(x), demonstrating that the detected anomalies may have a close spatial relationship with Fe mineralization. The comparing results with deep autoencoder network and the hybrid method combining with PCA and LSA, suggest that the local RX anomaly detector is a potential method to identify multivariate geochemical anomalies in a complex geological background.