Identification of geochemical anomalies through combined sequential Gaussian simulation and grid-based local singularity analysis

Local singularity analysis (LSA) has been proven to be an effective tool for identifying weak geochemical anomalies. The common practice of grid-based LSA is to firstly interpolate irregularly distributed observations onto a raster map by using either kriging or inverse distance weighting (IDW). The inherent nature of the weighted moving averaging of these methods typically subjects the interpolated map to a smoothing effect. Additionally, the traditional procedure did not allow for uncertainties on the values of geochemical attributes at unsampled locations. As such, these two aspects might affect LSA results. This paper presents a hybrid method, which combines sequential Gaussian simulation and grid-based LSA to identify geochemical anomalies. A case study of processing soil samples collected from the Jilinbaolige district, Inner Mongolia, China, further illustrates the hybrid method and helps compare the results with those from kriging-based LSA. The findings indicate that (1) the uncertainties of values at unsampled locations could affect the results of grid-based LSA, and (2) singularity exponents from kriging-based LSA roughly represent the trend (median) of singularity exponent distributions from simulation-based LSA, but the latter can also provide a measure of uncertainty of singularity exponent propagated from the uncertain values at unsampled locations, and (3) the procedure combining simulation-based LSA and analysis of distance is a feasible way for identifying geochemical anomalies with uncertainty being considered. The anomaly probability map obtained can provide a more generalized perspective than interpolation-based LSA to delineate anomalous areas.