神经网络在气象领域中的应用

Application of Neural Network in Meteorology

神经网络一直是气象领域重要的研究方法之一,但是,由于数据的缺乏及相关理论不完善等因素,在实际应用中有诸多局限。近年来,随着数据资料质量的提升及深度学习研究的兴起,研究者们尝试在气象领域应用神经网络的相关方法,提高天气预报的准确性,更加深入地研究天气现象背后的物理规律。本文介绍了应用于气象领域的几种常见的神经网络模型,回顾了近些年来神经网络模型在数值模式后处理、图像特征提取及天气预报三个方面的应用,并提出在以后研究中的几点期望。以目前的研究成果而言,神经网络模型可以在一定程度上弥补数值预报模式的缺陷,但是不能完全取代数值预报模式。因此,需要更加深入地尝试将数值预报模式及神经网络模型相结合,发挥两种方法的优势,共同推动气象行业的发展。

Neutral network has always been one of the most important methods in the field of meteorology. However, many limitations will be found when it used in practical applications due to the lack of data and its imperfect theory. In recent years, the researchers have tried to apply the neutral network models in the field of meteorology in many ways with the improvement of data quality and the appearance of deep learning theory, to improve the weather forecast accuracy and find the physical laws behind the weather phenomena. This paper introduce several neutral network models that most used in the field of meteorology, review the application of neutral network models in the field of numerical forecast post-processing, image feature extraction and weather forecast, Then give some expectations in future research. According to the current research, the neutral network model can make up for some defects of weather numerical forecast to some extent instead, but cannot completely replace it. Therefore, it is necessary to combine them in a deeper degree and take advantage of them, to promote the development of meteorological industry.