Abstract: With the continuous development of grid forecasting operations with higher resolution and more accuracy, the spatial and temporal accuracies and precision of grid forecast products need to be continuously improved. To maintain compatibility between the grid and site forecasts, FRI (fast refinement interpolation) method was explored, which considers the static real-terrain and dynamic atmospheric vertical variation process. In this study, 3–36-h forecast products for three models (ECMWF, CMA-GFS, and CMA-MESO) from 1 January 2022 to 31 March 2022, were selected as the experimental data. Further, the results of individual weather process tests, FRI parameter selection tests, and long-term interpolation tests were compared. The results of long-term interpolation tests and individual case trials showed that the FRI method has a clear advantage over the bilinear interpolation method. From a spatial perspective, the FRI method can improve the spatial interpolation accuracy of 2-m temperature near the ground, and the results are more consistent with topographic variations. From a temporal perspective, the FRI method significantly improves the accuracy of the 2-m temperature interpolation results compared with the bilinear interpolation method, especially in the western region with complex subsurface. Additionally, the loss parameter in the proposed method can be used as an indicator to check the three-dimensional atmospheric vertical variability of the model product. The FRI method is fast and efficient and offers a clear physical meaning, which provides important theoretical support for the more accurate reflection of near-surface meteorological forecast information.