Abstract: Long-term meteorological datasets with high spatio-temporal resolution are of great significance to meteorological operations and scientific research.However, challenges such as low quality and low spatial resolution emerge during application of existing datasets.With the development of artificial intelligence, machine learning and deep learning algorithms have been applied in the field of meteorology.Addressing the technical challenges in the development of meteorological datasets, we outlined four application scenarios for artificial intelligence technologies: quality control of observations, interpolation and reconstruction of missing data, fusion of multi-source data, and downscaling of low-resolution gridding data.We comprehensively reviewed on the advantages and disadvantages of various machine learning models in these application scenarios and conducted a quantitative analysis on research frontiers and development trends using bibliometric methods.The research showed that compared to traditional methods, artificial intelligence algorithms are more competitive in terms of computational efficiency, accuracy, and application flexibility.It is recommended to further promote the application of artificial intelligence in the development of meteorological data products by establishing high-quality training datasets, enhancing the use of multi-source and heterogeneous data, and exploring modeling based on the physical mechanisms of meteorological data.