Abstract: For satellite navigation systems, positioning errors are affected by the refractive index of the troposphere atmosphere. Improving the accuracy of predicting the refractive index of the troposphere atmosphere can reduce navigation positioning errors. The refractivity of tropospheric atmosphere is the main parameter for studying the influence of the troposphere on the propagation of electromagnetic waves, and the accuracy of its predictions is of great significance for radio systems. In this paper, a tropospheric refractivity prediction method based on BP neural network is proposed, which takes the year, month, day, time, surface refractivity, and altitude as the input of the BP neural network, and the corresponding refractivity at the input altitude as the output of the model. Similarly, by adjusting the input and output parameters, the BP neural network can also be used to predict the refractivity gradient of 1 km near the ground. Finally, the proposed algorithm is calculated and analyzed by using the historical aerial exploration data of Hongkong and Taiyuan, and compared with the methods in the existing papers. The results show that the proposed method has certain advantage in the calculation accuracy.