太阳耀斑预报研究进展

Progress in solar flare forecasting

太阳耀斑是指发生在太阳表面局部区域中突然和大规模的能量释放过程.它是空间环境的主要扰动源,对地球空间环境造成很大影响.太阳耀斑预报是空间天气预报的重要组成部分,对其研究具有重要的实用价值和科学意义.现有的大部分太阳耀斑预报模型是从观测数据提取预报因子,利用各种统计和数据挖掘技术建立预报因子与耀斑发生之间的关系模型,利用建立的模型对未来时间的耀斑发生进行预报.在预报研究中,预报因子、预报方法和预报模型是3个主要研究领域.其中预报因子的选取和数据处理尤为重要,是建立预报模型的前期工作.预报因子主要采用太阳黑子、磁场参量和分形因子等.预报方法包括统计方法、机器学习方法和数据同化方法.统计方法在早期的耀斑预报建模中用的较多,随着数据挖掘技术的发展,越来越多的机器学习方法应用到预报模型中并取得了较好效果.而近期发展的数据同化方法有更好的模型修正能力.预报模型早期基本使用静态模型,后来发展起来的动态模型具有更强的优势;而自组织临界模型在物理方面给了耀斑发生更多的解释.本文分别从这3个方面总结了耀斑预报的研究进展,结合中国科学院国家天文台太阳活动预报中心的工作,评述了一些重要的研究成果.最后,对未来的研究方向进行了总结和展望.

Solar flare is an abrupt and scale of energy release process in solar surface local region. It is a main disturbance resource of space environment and has tremendous influence. Existed solar flare forecasting models include obtaining predictor from observed data, constructing relation model of predictors and flare occurrence by using statistic or data mining method, and predicting future flare occurrence with this model. In solar flare forecasting research, the predictor, the forecasting method and model are three main sides. As an important part, predictor and data process are preprocessing work. Predictors usually select solar sunspot parameter, magnetic parameters and fractal dimension and so on. Forecasting methods include statistic method, machine learning method and data assimilation method. Statistic method is mainly used in early model. With the development of data mining technique, more and more machine learning methods are concerned and receive satisfied result. Data assimilation method has good model correction ability. In forecasting model, most of them are static model. Recently, the time revolution dynamic model is developed and has better performance. Besides, self-organized model developed recently gives a physical description about burst mechanism of solar flare. This paper summarizes the research progress in these three sides of flare forecasting. Connected with the work in solar activity prediction center of National Astronomical Observatories, important research progresses are reviewed. Finally, future research trend is prospected.