Abstract:To solve the problem that the number of nearest reference points or influence radius on each interpolation point cannot be dynamically selected in the spatial interpolation method of meteorological elements, and to solve the applicability of fitting function for spatial location to meteorological elements, a spatial interpolation method of meteorological elements based on dynamic parameters was proposed. In the method, shallow cascade back propagation neural network was used for nonlinear fitting of reference points, and the spatial interpolation of interpolation points was completed on three factors: distance, horizontal angle and height difference basis of gradient transformation, finally, the weights of three factors were synthesized into spatial weights, which were used as a pre-evaluation index for dynamic selection of the number of nearest reference points, thus the method could reduce the difficulty of parameter setting and improve the accuracy of interpolation method. The experimental results based on the Chinese land area of 2022 show that compared with Aspect and Inverse Distance Weighted(AIDW), Ordinary Kriging(OK) and Thin-Plate Smoothing Spline(TPSS), the method's average error in the interpolation of air temperature and absolute humidity are the smallest, where the MAE of air temperature decreased by 0.23℃ and the RMSE decreased by 0.44 ℃ compared with TPSS whose average errors are the second smallest; the MAE of absolute humidity decreased by 0.06 g·m-3 and the RMSE decreased by 0.12 g·m-3 compared with TPSS.