| 基于广义回归神经网络的行星际/太阳风参数和地磁指数的紫外极光强度建模 |
| |
| 引用本文： | 胡泽骏,韩冰,连慧芳.基于广义回归神经网络的行星际/太阳风参数和地磁指数的紫外极光强度建模[J].地球物理学报,2020,63(5):1738-1750. |
| |
| 作者姓名： | 胡泽骏 韩冰 连慧芳 |
| |
| 作者单位： | 1. 中国极地研究中心自然资源部极地科学重点实验室, 上海 200136;2. 西安电子科技大学电子工程学院, 西安 710071 |
| |
| 基金项目： | 国家重点研发计划（2018YFC1407303），国家自然科学基金（41874195，41831072，61572384，61432014），中国科学院空间科学先导专项（XDA15350202）共同资助. |
| |
| 摘 要： | 极光卵极光强度的空间分布是太阳风-磁层-电离层能量耦合过程的重要表现,并且随着空间环境参数和地磁指数的变化而变化,是空间天气的重要指示器.建立合适的极光强度模型对亚暴的预测以及磁层动力学的研究具有重要意义.本文基于Polar卫星的紫外极光成像仪(Ultraviolet Imager,UVI)数据,采用两种不同的极光强度表征方法,即曲线拟合方法(从UVI图像数据中提取极光强度沿磁余纬方向上的曲线特征,Curve Feature along the Magnetic Co-latitude Direction of the Auroral Intensity,CFMCD_AI)和网格化方法(从UVI图像数据中提取极光强度的网格化特征,Gridding Feature of the Auroral Intensity,GF_AI),来构造极区极光强度特征数据库.然后,利用该数据库,采用广义回归神经网络(Generalized Regression Neural Network,GRNN)构建了以行星际/太阳风参数(行星际磁场三分量、太阳风速度和密度)和地磁指数(AE指数)为输入参数的两种极光强度预测模型(GRNN_CFMCD_AI模型和GRNN_GF_AI模型).利用图像质量评价指数结构相似度(structure similarity,SSIM)作为极光强度模型预测结果和对应的UVI图像的相似性评价标准(完全相似为1,不相似为0,一般认为SSIM大于0.5是具有较好的相似性),对两种极光强度模型进行了性能评价.结果显示,GRNN_GF_AI模型预测结果对应的SSIM值范围为0.36～0.77,均值为0.54,性能优于GRNN_CFMCD_AI模型的.
|
| 关 键 词： | 极光卵模型 空间天气 神经网络 紫外极光 |
| 收稿时间： | 2019-04-12 |
Modeling of ultraviolet auroral intensity based on Generalized Regression Neural Network associated with IMF/solar wind and geomagnetic parameters |
| |
| HU ZeJun,HAN Bing,LIAN HuiFang.Modeling of ultraviolet auroral intensity based on Generalized Regression Neural Network associated with IMF/solar wind and geomagnetic parameters[J].Chinese Journal of Geophysics,2020,63(5):1738-1750. |
| |
| Authors: | HU ZeJun HAN Bing LIAN HuiFang |
| |
| Institution: | 1. MNR Key Laboratory for Polar Science, Polar Research Institute of China, Shanghai 200136, China;2. School of Electronic Engineering, Xidian University, Xi'an 710071, China |
| |
| Abstract: | As an important indicator of space weather, the auroral intensity and distribution on the auroral oval is closely related with the solar wind-magnetosphere-ionosphere energy coupling, and change with the change of space and geomagnetic environment. It is important to establish a suitable model of auroral oval intenstiy for the prediction of substorms and the study of magnetospheric dynamics.The images acquired by Polar ultraviolet imager on Dec. 1996 and Jan. 1997 are used to model the intensity of auroral oval. Firstly, the auroral intensity characteristics of every UV image are extracted by two different methods, one is the curve fitting method which extracts the Curve Feature along the Magnetic Co-latitude Direction of the Auroral Intensity (CFMCD_AI) in the UVI images, and another is the gridding method which extracts the Gridding Feature of the Auroral Intensity (GF_AI) in the UVI images. Based on the two data sets, the generalized regression neural network (GRNN) is used to establish two prediction models of ultraviolet auroral oval intensity (GRNN_CFMCD_AI and GRNN_GF_AI), which have the input parameters with IMF/solar wind parameters (three components of IMF, solar wind velocity and density) and geomagnetic index (AE). In order to verify the validity of the models, the structure similarity (SSIM) of images is used to evaluate the similarity between the result the models and the UVI images (1: complete similarity; 0: completely dissimilarity; SSIM>0.5: good similarity). The evaluation results show that the SSIM of GRNN_GF_AI is between 0.36 and 0.77, and the average is 0.54, which is better than GRNN_CFMCD_AI. |
| |
| Keywords: | Auroral oval Space weather Generalized regression neural network Ultraviolet aurora |
| 本文献已被 CNKI 等数据库收录！ |
| 点击此处可从《地球物理学报》浏览原始摘要信息 |
| 点击此处可从《地球物理学报》下载免费的PDF全文 |
|
