【作者】 乐新安；

【导师】 万卫星；

【作者基本信息】 中国科学院研究生院（武汉物理与数学研究所） ， 空间物理学， 2008， 博士

【摘要】 地球电离层上连外层空间,下接中层大气,是日地环境系统中承上启下的重要环节和关键层次。对电离层的研究可以促进认识日地空间系统的整体行为、进而更好地为人类的空间活动服务,因此有重要的意义。近年来,随着人类空间活动和通讯系统的增多,对电离层的基本状态和电离层扰动的监测和预报的需求也越来越强烈。在此背景下,基于模式和观测的数据同化方法被引进到电离层领域,并且在电离层现报和预报上显示了较好的应用前景。本文围绕中低纬电离层模拟与数据同化展开研究,构建了中低纬电离层经验和理论模型,进行了一系列基本物理问题的模拟研究,并对电离层数据同化中若干关键技术进行了试验研究。本文的主要工作如下:1、利用观测数据进行了电离层建模研究,并分析了电离层长期变化趋势。武汉电离层观测站是世界上观测历史最悠久的台站之一,利用该站的观测数据,已经建立了系列的经验模型,如TEC、foF2、hmF2等经验模型,本文首先利用武汉站长达近40年的foE观测值建立了一个精度较高的经验模式,比较研究显示该模式精度要优于IRI等其它全球经验模式。这一方面是对以前武汉站系列参数经验模式一个有力的补充,同时还对中国电离层模式的建设有参考价值。此外,还利用BP神经网络方法构建了东亚/澳大利亚区域foF2高精度模型,并利用武汉单站数据对建立的模式进行了验证,结果显示神经网络模式具备相对较高的精度。同时利用该神经网络模型系统分析了东亚/澳大利亚扇区foF2的长期变化趋势。分析结果显示东亚/澳大利亚的foF2以平均0.05%/年的速度下降。单纯温室效应并不能解释这个现象,还可能与其它因素如太阳地磁活动、中性浓度、温度和速度等的长期变化有关。2、建立了一个电离层理论模式,对若干电离层现象进行了数值模拟研究。本小组在中性风场、电离层电场、中纬一维模式和低纬二维模式建模方面已经有了很多积累,本文在这些工作的基础上,通过求解等离子体连续性方程、动量方程和能量方程,构建了一个高精度、高时空分辨率、快速灵活的中低纬电离层理论模式(TIME-IGGCAS)。通过与众多经验模式和观测数据进行比较对模式的有效性进行了验证,比较结果表明,模式模拟的结果无论是数量级上还是各种变化形态上都与经验模式和观测符合得很好;模式能再现大部分电离层异常特征,如赤道异常、冬季异常、和半年异常;但模式在日出日落时段、在低高度模拟结果有相对较大的偏差,需要进一步加以改进,利用这个模式进行的模拟研究有:(1)模拟了赤道异常区的气候学特征。重点分析了南北驼峰和赤道槽的位置和对应处的TEC值、谷宽和峰谷比。模拟显示赤道异常区有显著的地方时、季节和太阳活动变化。赤道槽的位置一般随季节的变化在磁赤道的两侧附近变化;南北驼峰在至日时有显著的不对称;赤道异常在冬季正午发展的最好、在分点午后发展的最好、在夏季下午发展的最好;谷宽和峰谷比的季节变化比较显著。这些结果可以由向赤道风、跨赤道风和太阳直射点的季节变化来解释。(2)模拟了赤道异常区对电场扰动的响应。当扰动电场漂移速度是向上(下)时,电子浓度在赤道附近磁纬小于15度的范围内出现负(正)暴,而在较高的纬度(20度附近)出现正(负)暴,在正负暴之间有一个大约3度宽的过渡带;在地磁纬度大于30度的区域,电子浓度的扰动已经很弱;扰动电场的效应会在扰动电场消失后持续较长的时间;无论电场漂移是正扰动还是负扰动,EIA扰动都随电场漂移扰动幅度的增加而线性增加,这个在电场扰动发生时或者发生后都是满足的。(3)模拟了利用赤道台站hmF2的时间变化率来表示电场漂移速度的有效性。由于电离层电场观测的缺乏,许多研究者提出了一些间接的办法来获取电离层垂直漂移速度,这中间就包括利用赤道台站hmF2的时间变化率这种办法。本文利用TIME-IGGCAS模式模拟了这种方法表示电场漂移速度的有效性,及这种有效性随地方时、季节、和太阳活动的变化。同时对这种计算方法在扰动状态下的可靠性也作了探讨。模拟结果显示这种方法在日出、日落、及日落后几小时非常有效(0600-0730, 1700-2100 LT),但是在其它的地方时,用这种方法得到的速度严重偏小。在扰动时,hmF2的剧烈变化可以用来确定扰动电场的发生。(4)最后,首次模拟研究了地磁场构型变化对电离层长期趋势的影响。在利用测高仪观测分析电离层的长期变化趋势时,发现单纯的温室效应并不能造成这么大的变化,因此我们提出了地磁场的长期变化可能会造成电离层的长期变化。为了证实这个说法,利用TIME-IGGCAS模式首次模拟研究了地磁场构型变化对电离层长期趋势的影响。模拟结果说明:由于地磁构型的变化会导致中性风对电离层的作用发生变化,进而导致电离层的长期趋势发生变化;由于全球地磁构型变化的差异很大,导致这种由地磁构型变化引起的电离层长期趋势有显著的地区差异;模拟的长期趋势有显著的季节和地方时变化,且这种变化有地区差异;这可能与因为地磁构型变化导致的别的电离层驱动因素作用的相应变化有关,通过与已有的观测结果作对比,模拟结果可以在部分程度上解释为什么目前得到的电离层趋势的地方时和季节变化会有很大差异;地磁场构型变化导致的电离层趋势不能忽略,这点在地磁场变化比较剧烈的区域尤其重要。以上这些模拟工作一方面解决了相应的物理问题,另一方面进一步说明了模式的稳定性和可靠性,这为以后继续使用该模式进行模拟研究及在此基础上建立数据同化模式奠定了基础。3、利用建立的TIME-IGGCAS模式,结合实际电离层观测数据,进行了系列电离层数据同化试验研究。(1)基于最小二乘法,我们进行了一个电离层观测系统数据同化试验,试验结果一方面表明理论模式的稳定可靠,另一方面表明利用最小二乘法进行电离层数据同化和外驱动参量估算的可行性。基于同样的方法,我们利用东亚/澳大利亚扇区42个GPS台站的观测和TIME-IGGCAS模式同化模拟了2004年11月7-9日超级磁暴期间东亚/澳大利亚扇区电离层/热层的响应特征,同时估算电离层外驱动参量包括O/N2、子午风和垂直电场漂移速度,并把模拟的结果跟GPS、测高仪等观测作了对比,估算的外参量也尽量利用观测作了验证。结合模拟和观测结果,本次超级磁暴期间东亚/澳大利亚扇区的响应特征大致是:晚上向赤道风和东向扰动电场导致电子浓度的增强;白天北半球主要是正暴,南半球主要是负暴,这可能是半球间O/N2的不对称性、夏季到冬季的跨赤道风和扰动电场共同作用的结果。在这些因素共同作用下,使得电离层扰动的幅度和相位随地方时和纬度发生复杂的变化。(2)为了电离层数据同化中构建准确的误差协方差矩阵,利用观测统计分析了电离层的空间相关性。在电离层数据同化中,背景误差协方差对同化效果有很大的影响,为了准确的表述背景误差协方差,需要准确的知道电离层的相关性。因此,我们利用美国喷气动力实验室全球电离层地图在2000年和2005年的数据和MillStone Hill非相关散射雷达在2002年10月长达一个月的观测统计分析了电离层逐日变化的空间相关性,并分析了电离层在三个方向的相关距离的地方时、季节和纬度变化,还尝试性的对这些变化特性做出了解释,最后重点讨论了磁共轭点之间强相关性的可能原因、电离层相关距离的各种变化的可能因素及电离层暴对相关距离的影响。总的来说,电离层相关距离在垂直方向上随高度增加而增加;电离层的相关性在白天大于晚上、中纬大于低纬、太阳活动高年大于太阳活动低年,北半球中纬有显著的季节变化。我们的统计结果一方面从相关性的角度证实了影响电离层的因素存在不同的空间尺度,另外为进行电离层数据同化时构建误差协方差矩阵提供了参考。(3)结合一个一维中纬理论模式和MillStone Hill非相干散射雷达观测,首次尝试了高级资料同化方法——集合Kalman滤波法(EnKF),在电离层数据同化中的应用。同化结果表明:与三维变分(3D-Var)相比,EnKF能动态的模拟电子浓度的高度相关系数。EnKF同化的结果显示电离层电子浓度的垂直相关系数有显著的高度和地方时变化,非相干散射雷达观测也证明了这点。EnKF集合扩展与集合平均相对观测的偏差有几乎相同的时间和高度变化,这说明在同化时增加扰动的方式和考虑各种误差的方式是正确可行的,说明EnKF同化是成功的。与没有同化相比,3D-Var和EnKF方法都可以获得相对较小的RMSE。EnKF可以更好的把观测的影响从数据密集的地方传播到数据缺乏的地方。我们的试验研究表明,利用等效风方法对中性风场进行修正可以显著的提高模式的预报能力。在EnKF同化时,当集合数不是很大时,在两个相隔很远的点有时会存在一些虚假的强相关,这个问题可以通过增加集合数来解决。本文工作表明,电离层的模式化(经验模式与理论模式)不仅是对现有的电离层实验观测和理论研究成果的总结,同时又能促进对电离层结构与变化特性的深入认识以及对相应物理过程的研究。而电离层数据同化把电离层观测与电离层模式结合起来,开辟了电离层研究与预报的新思路。总之,本文的研究工作可以加深我们对电离层的认识,所构建的模式可以用来模拟研究电离层中的一些物理过程和现象,运用数据同化方法把模式和数据进行结合,有望提高我们进行空间天气尤其是电离层天气的现报和预报能力。更多还原

【Abstract】 The Earth’s ionosphere locates between the outer space and middle atmosphere and is an important part and key layer in the whole sun-earth environment system. The research on the ionosphere can enrich our knowledge of the sun-earth system and serve the human’s space activities. Therefore, it is significantly for us to continue the study on the ionosphere. In the recent years, with the increase of the human’s space activities and communication systems, there is a growing need to more accurately represent and forecast the ionospheric climate and disturbances. Several groups attempt to incorporate the observations into ionospheric models by using optimization schemes, which is known as data assimilation methods, to give specific representations of the ionosphere. This technique has manifested potential ability in ionospheric nowcast and forecast. In this paper, considering the geographic location of our country, we concentrate on the investigations of middle and low latitude ionospheric modeling and data assimilation. We constructed several empirical and theoretical ionospheric models and investigated several basic physical problems using these models. In addition, we also studied several important techniques of ionospheric data assimilation. The main contents are as follows:1, Constructing empirical models based on observations and analyzing ionospheric long term trend.Wuhan ionospheric observatory is one of the most long lasting ionosphere stations in the world. Based on its observations, several empirical ionospheric models have been developed, such as TEC, foF2 and hmF2 empirical models. In this paper, we first constructed an empirical foE model with high precision over Wuhan using fifty years’observations of foE. This model is an important complement to the local empirical ionospheric model over Wuhan. A comparison between our model and IRI and Titheridge’s model shows that our model has a better performance. By using Artifical Neural Network (ANN) method, we constructed empirical models of foF2 over Asia/Australia sector based on ionosonde observations. The model has been validated by comparison with Wuhan station’s observations and the results showed that the ANN model has high precision. Using this ANN model, we systematically analyzed the long term trends of foF2 over this area for the first time. Results illustrated that the foF2 in the Asia/Australia sector has an average decrease of 0.05% per year in the past fifty years. This trend can not be interpreted only by greenhouse effect. Many other factors which can influence the ionosphere, such as solar and geomagnetic activities and neutral background gas, might also contribute to the trend.2, Constructing theoretical ionospheric model and modeling several ionospheric phenomena.Our group has done many works on the ionospheric modes, including neutral wind, ionospheric electric field, middle latitude theoretical model and low latitude two dimensional theoretical model. On the basis of these works, we developed a middle and low latitude theoretical ionospheric model with high precision, high space and time resolution, speediness and flexibility, through solving plasma continuity, momentum and energy equations simultaneously. The model is named Theoretical Ionospheric Model of the Earth in Institute of Geology and Geophysics, Chinese Academy of Sciences (TIME-IGGCAS). TIME-IGGCAS was validated by comparisons with several other typical empirical ionospheric models and multi-observations. The results showed that the modeled electron density and plasma temperatures are quantitatively and qualitatively in good agreement with those of empirical models and observations. TIME-IGGCAS can reproduce most anomalistic features including equatorial anomaly, winter anomaly and semiannual anomaly. The model results have relatively large deviations near sunrise time and sunset time and at the low altitudes. There results give us a reference to improve the model and develop the data assimilation model in the future. We did several modeling studies based on the developed theoretical model.(1) Modeling the climate features of the equatorial ionization anomaly (EIA). Our analysis concentrated on the locations and the corresponding TEC values of northern crest, equatorial trough and southern crest, and also the width of the trough and the ratio of crest to trough. The modeling results showed that the EIA has typical local time, seasonal and solar variations. The equatorial trough usually lies on the two sides of the geomagnetic equator and varies with seasons. The two crests have obvious asymmetries in solstice. The EIA is fully developed around midday in winter, postnoon in equinoxes and late afternoon in summer. The width of the trough and the ratio of crest to trough have obvious seasonal variations. These seasonal dependences of EIA can generally be interpreted by the seasonal variations of the equator ward wind, the transequatorial neutral wind, and the subsolar point.(2) Modeling the response of EIA to the disturbed electric field. When the disturbance of the E×B vertical drift velocity is upward (downward), electron content has negative (positive) disturbance in the area with geomagnetic latitude less than 15 degree and positive (negative) in the area around geomagnetic latitude 20 degree. There is a transitional area with 3 degree’s width between positive and negative disturbance. In the area of geomagnetic latitude larger than 30 degree, there is no significant ionospheric disturbance. The effect of disturbed electric field would last for several hours after the disappearance of the disturbed electric field. The disturbance of EIA linearly increases with the increase of disturbed electric field during or after the disturbance of electric field.(3) Modeling the validation of the method of using hmF2 to derive electric field vertical drift velocity. Since the lack of direct observations of ionospheric electric field, many indirect methods have been put forward to obtain it. Several researchers use the time rate of change of hmF2 to represent the E×B vertical drift. In this paper, we used the TIME-IGGCAS model to confirm its validation from a view of theory and study the local time, seasonal and solar variations of its validation of this method. In addition, we also tested the validation of this method during disturbed conditions. Our modeling results indicated that this method is usable near sunrise, sunset and post sunset (0600-0730, 1700-2100 LT). The derived velocity by this method is smaller than observations in the rest local times. During disturbed conditions, the variations of hmF2 can be used to determine the occurrence of intense electric field disturbance.(4) Modeling the effects of the variations of geomagnetic field on the ionosphere. When analyzing the ionospheric long term trends from ionosonde observations, we found that just greenhouse effect is insufficient. We suggest that the long term variations of geomagnetic field may be another origin. To confirm this interpretation, we modeled the effects of long term variations of geomagnetic field on the ionospheric long term trend by TIME-IGGCAS for the first time. The modeling results indicate that the variations of geomagnetic field indeed can result in the long term variations of ionosphere because of the variations of the effects from the neutral wind on the ionosphere. Since the geomagnetic field variations differ from place to place, the ionospheric trends induced from geomagnetic field also have location dependence. The modeled ionospheric trends also have obvious seasonal and local time variations. Because the corresponding controlling factors have location dependence, these variations also show typical regional features. By comparison with existing results, we suggest that the changes of the global geomagnetic field may partly contribute to the inconsistent seasonal and local time variation patterns in ionospheric trends from different observations. We conclude that the effects of geomagnetic orientation on the ionospheric long term trend cannot be ruled out, especially in areas with large geomagnetic field variations. The above modeling studies not only investigate the corresponding physical problems, but also illustrate that our model is steady and credible. These results imply a good base for the using of the model and the development of the ionospheric data assimilation model in the future.3, Based on the developed model, we did several data assimilation experiments and studied several important techniques of data assimilation.(1) We carried out an observation system data assimilation experiment using TIME-IGGCAS model and non-linear least square fit method. The experiment showed that the theoretical model is steady and credible and the ionospheric external drivers can be estimated by non-linear least square fit method. Using the same method, we modeled the ionospheric and thermospheric response to the super storm during November 7-9, 2004 in Asia/Australia sector by assimilating the GPS observations from 42 stations. The electric field, neutral meridian wind, and the ratio of O to N2 (O/N2) are estimated through minimizing the differences between modelling results and observations by nonlinear least square fit method. The modeling results including electron density and the estimated drivers have been validated by empirical models and the corresponding observations. Combining observations and modeling results, this storm can generally be understood as follows. During night time of this storm the equator-ward wind and disturbed eastward electric field are the main factor that attribute to the observed ionospheric enhancement. During the daytime northern hemisphere shows mainly positive storm and southern hemisphere is predominated by negative storm. This is probably the combined effects of the asymmetry of O/N2 in two hemispheres, summer-to-winter wind, and disturbed electric field. Under the competitive influences of varied electric field, wind field and neutral compositions, the phase of storm shows complicated variations versus latitudes and local times.(2) We analyzed the spatial correlation of ionosphere to give better background error covariance in data assimilation. When doing the ionospheric data assimilation, the background error covariance has determined effect on the assimilated results. So an accurate correlation model of the ionosphere is very important. In this paper, we statistically investigated the spatial correlation of ionospheric day-to-day variability and the local time, seasonal, and latitudinal variations of correlation distances in three directions using JPL GIM during 2000 and 2005 and one month observations of Millstone Hill incoherent scatter radar observations during October, 2002. We also attempted to interpret these variabilities. At last we discussed the possible reasons of the relative larger correlations between geomagnetic conjugate points, the possible factors of the variations of ionospheric correlation distances, and the effects of ionospheric storm on the ionospheric correlation. Generally, the ionospheric correlation distance increases with the increase of altitude and is larger during daytime than at night, at middle latitude than low latitude, and during high solar activity than low solar activity. The correlations of northern middle latitudes have obvious seasonal variations. Our statistical results confirm that the factors which result in the ionospheric day-to-day variability have different spatial scales. In addition, our results are useful in the constructing of a background covariance matrix in ionospheric data assimilation.(3) Based on a middle latitude theoretical ionospheric model and Millstone Hill incoherent scatter radar observations, we explored the application of Ensemble Kalman filter (EnKF) known as an advanced data assimilation method in ionosphere data assimilation. It is found that the derived vertical correlation coefficients of electron density show obvious altitude dependence. These variations are consistent with those from ISR observations. Both the altitude and local time variations of the root mean square error (RMSE) of electron densities for the ensemble spread and ensemble mean from observation behaves similarly. It is shown that the spread of the ensemble members can represent the deviations of ensemble mean from observations. The EnKF technique has a better performance than the 3DVAR technique especially in the data-gap regions, which indicates that the EnKF technique can extend the influences of observations from data-rich regions to data-gap regions more effectively. To achieve a better prediction performance, the external driving forces should also be adjusted simultaneously to the real weather conditions. For example, the performance of prediction can be improved by adjusting neutral meridional wind using equivalent wind method. In the EnKF, there are often erroneous correlations over large distance because of the sampling error. This problem may be avoided by using a relative larger ensemble size.This paper indicates that ionospheric modeling not only is the summarization of ionospheric observations and theoretical research, but also can enhance our knowledge of ionosphere structure and variability and the corresponding physical problems. The ionospheric data assimilation can open us a new method to give better ionosphere research and forecast by combing the observations and models. In conclusion, the research results in this paper can advance our knowledge of the ionosphere. The constructed model can be used to model several physical processes and phenomena in the ionosphere. The data assimilation method has potential ability to give better nowcast and forecast of space weather especially ionospheric space through combining the model and observations.更多还原