Abstract: In geotechnical engineering design, accurately measuring the spatial variability of soil parameters determines the reliability and economy of the design structure, and the distance of auto -correlation of soil parameters is an important index to reflect its spatial variability. At present, the relationship between auto -correlation distance and structural reliability is not perfect. In this paper, the reliability design of pile foundation is taken as an example. The static cone penetration test (CPT) data of engineering site are chosen, and the random field theory is introduced to describe the distribution characteristics of side friction and tip resistances in the soil layers. The auto -correlation distance of side friction and tip resistances is respectively calculated by the spatial average method. The auto -correlation distance is considered to reflect the distribution of side friction and tip resistances, and the reliability index of pile foundation is calculated by the Monte Carlo method. The results show that the value of soil parameters, taking into account the influence of the auto -correlation distance, the dispersion degree of soil parameter distribution can be reduced, and then the pile foundation reliability index is higher.