Multivariate regression analysis of lithogeochemical data to model subsurface mineralization: a case study from the Sari Gunay epithermal gold deposit, NW Iran

• Geochemical and drill core data at Sari Gunay are used to highlight true anomalies.
• Multiple regression modeling is applied to predict subsurface mineralization.
• Mineralization parameters are estimated by four multivariate regression functions.

In this contribution, multivariate regression was applied to surface channel rock and borehole geochemical data from the world-class Sari Gunay epithermal gold deposit, in northwest Iran, to model subsurface mineralization for further drilling. Multiple, factorial, polynomial and response surface regression models were applied to the geochemical data sets from a training mineralized area to evaluate the accuracy of these models using separate geochemical data from a test area. Geochemical data of 31 elements in surface channel rock samples were used as independent variables, and three parameters namely average grade, sum and productivity in individual 25 m by 25 m grid cells, obtained by kriging of borehole data, were used as dependent variables. All the multivariate regression models revealed high determination coefficients for three parameters, among which the response surface regression model yielded the highest values. The response surface regression yielded the best result, followed by the multiple regression, in modeling the geochemical data from the test area. Therefore, the result of the response surface regression was used to model subsurface gold mineralization at the Sari Gunay gold deposit in order to design additional drillings.