Abstract: Data correction is one of the core processes in data assimilation, which aims to improve the assimilation of data by correcting and calibrating the data. The application of deep learning in meteorological data correction was comprehensively reviewed, including meteorological model correction, weather forecasting, and climate prediction. Firstly, the importance of meteorological data correction was introduced, and traditional methods such as statistics and traditional machine learning were reviewed, analyzing their advantages and limitations. Secondly, the application of deep learning in data correction was detailed in each scenario, including convolutional neural networks, recurrent neural networks, and Transformer transformers. By summarizing the current research progress, the strengths and weaknesses of deep learning methods and traditional methods in data correction were discussed. Finally, the limitations of deep learning in data correction were summarized, and the optimization methods and future development directions of deep learning in meteorological data correction were pointed out.

Key words: deep learning, data correction, data assimilation, climate prediction, weather forecast

摘要： 数据订正是资料同化的核心过程之一，通过对数据进行修正和校准来提高资料同化的效果。针对气象观测存在多种误差导致气象数据存在偏差的问题，对深度学习在气象数据订正中的应用进行了综述，应用场景包括气象模式订正、天气预报和气候预测。首先，本文介绍了气象数据订正的重要性，同时回顾了传统的气象数据订正方法，如统计学、传统机器学习等，并分析了它们的优点和局限性；其次，本文结合各个场景分别详细介绍了基于深度学习的数据订正在三个场景中的应用，深度学习方法主要包括卷积神经网络、循环神经网络和Transformer变换器，通过归纳总结当前的研究进展，讨论了数据订正中深度学习方法与传统方法的优劣；最后本文总结深度学习在数据订正中存在的局限性，同时指出了深度学习在气象数据订正中的优化方式和未来发展方向。

关键词: 深度学习, 数据订正, 资料同化, 气候预测, 天气预报