On the basis of aircraft coordinates calculated by latitude and longitude conversion, this paper utilizes the BP neural network for predicting the spatial position of aircraft in accordance with the acceptable level of safety 5×10-9 put forward by the International Civil Aviation Organization. On the basis of observing the longitudinal deviation distribution law, it is assumed that it conforms to the Gumbel distribution and the K-S test method is used for verification. The verification result is then adopted to build a collision risk model for risk assessment of obstacles height in airport clearance. The paper provides a new way to evaluate clearance condition for selecting airport location and administrate obstacles in airport clearance.