Abstract: The study on the spatial variability of hydraulic conductivity is the basis for artificial recharge of groundwater reservoir. In order to study the spatial variability of the hydraulic conductivity of underground reservoirs in Fuping section of Shichuan River, Box-Cox transform and Johnson transform were introduced to preprocess the field double-loop percolation test and exploration hole data. Traditional statistical methods and geostatistical methods were applied to analyze the hydraulic conductivity of the reservoir area with the variation function as a tool. The results show that the hydraulic conductivity of the reservoir area varies from 0.02 m/d to 6.44 m/d and obeys both logarithmic normal distribution and Box-Cox transformation normal distribution, the spatial correlation of the hydraulic conductivity is medium, and the best fitting model is Gaussian model. The Kriging interpolation results based on the optimal model show that the hydraulic conductivity as a whole is larger in the northwest and smaller in the southeast. The hydraulic conductivity is the largest near Meijiaping Township and Nanshe Township, ranging from 2.84 m/d to 6.44 m/d, with small spatial variation scale. It varies significantly near Mizi Township and Zhuangli Township, with large spatial variation scale. It is the smallest in the south of Dongshangguan Township, all less than 0.2 m/d, with small variation scale. The spatial variation is influenced by the combination of topography, geomorphology, distribution of stratigraphic lithology, hydro-meteorological conditions, distribution of test sites and exploration sites, human activities and other factors. The location of recharge should be chosen in Meijiaping town and other locations with large hydraulic conductivity, small spatial variation scale and low influence by human disturbance. The research results can provide theoretical references for the construction of underground reservoirs.