Exploring Multiscale Non-stationary Influence of Ore-Controlling Factors on Mineralization in 3D Geological Space

The formation of mineral deposit is the coupled result of multiple ore-controlling geological factors in mineralization processes. Different ore-controlling factors affect the mineralization typically with different mechanisms at different scales. Geographically weighted regression (GWR) assumes the same bandwidth for all the ore-controlling factors, which is limited in handling multiscale issues simultaneously. Multiscale geographically weighted regression (MGWR) can provide optimal bandwidth for each independent variable. In this study, based on the program of GWR in 3D space, we implement the MGWR model in MATLAB language, and also verify the accuracy and stability of the GWR and MGWR models by comparing the predefined and estimated parameters of the two models based on designed simulation datasets. To detect the non-stationarity and multiple scales of the controls of geological bodies in natural deposits, with the Jinchuan Ni–Cu sulfide deposit as a case study, firstly, the multicollinearity of ore-controlling factors is excluded and the spatial non-stationarity of their impact on mineralization is detected; secondly, the results of two models are compared and high performance of both models are achieved; then, the non-stationary index and the influence scale for different ore-controlling factors are obtained; finally, the variations of parameter estimates of the two models are analyzed and the importance of the magma conduit to the mineralization is verified.