文章基本信息

标题：Generalized Additive Modeling Combined With Multiple Collinear for ICME Velocity Forecasting
本地全文：下载
作者：J. Y. Lu ; C. Q. Jin ; M. Wang 等
期刊名称：Space Weather
印刷版ISSN：1542-7390
出版年度：2019
卷号：17
期号：4
页码：567-585
DOI：10.1029/2018SW002135
语种：English
出版社：American Geophysical Union
摘要：One of the main issues of space weather is the timely prediction of disturbed solar wind parameters at L1, especially caused by Coronal Mass Ejection (CME). Using the data from 170 front halo, flare-associated CMEs, and in-site solar wind data, an analysis of the Interplanetary Coronal Mass Ejection (ICME) peak velocity at L1 related to associated coronal parameters is performed. The statistical methods including the Generalized Additive Modeling (GAM) and Multiple Collinear (MC) have been applied to explain the underlying physical reasons and set up a new prediction model. Our results indicate that (1) X-flare integral flux, CME linear velocity, and Acceleration observed on corona play key roles in ICME velocity, while other coronal parameters only present a weak correlation, such as the CME Mass and Angular Width, (2) the relationship between ICME velocity and CME Acceleration, as well as CME linear velocity, is nonstationary, and the ICME velocity will increase with the increasing CME Acceleration or linear velocity until saturation, and (3) MC is an effective method to improve the forecast model performance. Compared with 0.52 for only GAM, the correlation coefficient using GAM + MC reaches to 0.71. To further testify the prediction ability, the GAM + MC model results are compared with the Back-Propagation network model and a typically empirical statistic relation proposed by Manoharan (2006, https://doi.org/10.1007/s11207-006-0100-y; their correlation coefficients and root mean squared errors are both roughly 0.6 and 100 km/s, respectively). It is found that the MC + GAM can upgrade the forecast at least over 10%.