Abstract: In order to select the optimal spatial interpolation applied to radon measurement for hidden fire location at the local scale, it took the sample data from the fire area of Beiyan Coal Mine as the study object based on the Arc GIS geostatistic wizard, then quantitatively screened out the most accurate interpolation model among four interpolations by cross-checking, qualitatively discerned the spatial distribution characteristics of different models, and ascertained the optimal interpolation by drilling verification. The results show that the most accurate model among four interpolations: the Inverse Distance Weight selects the Power p=1 model, the Radial Basis Function uses the Thin Plate Spline model, and both the Universal Kriging and the Ordinary Kriging are a Gaussian model; with the comprehensive deliberation on the distribution characteristics both in interpolating and extrapolating surface, as well as drilling verification, the interpolations are ordered as Ordinary Kriging（Gaussian）＞Universal Kriging（Gaussian）＞Radial Basis Function（Thin Plate Spline）＞Inverse Distance Weighting（p=1）, the Ordinary Kriging（Gaussian） is preferred to obtaining surface radon distribution difference in coal spontaneous combustion fire areas.