PUBLICATIONS. Journal of Geophysical Research: Space Physics

Transcription

1 PUBLICATIONS Journal of Geophysical Research: Space Physics RESEARCH ARTICLE Key Points: September 2011 geomagnetic storm impact on the ionosphere was modeled Data assimilation model accuracy was assessed for the calm and disturbed period Ionosphere has shown significant anomalies which were captured by the data assimilation model Correspondence to: D. Solomentsev, Citation: Solomentsev, D., K. S. Jacobsen, B. Khattatov, V. Khattatov, Y. Cherniak, and A. Titov (2014), Ionosphere data assimilation capabilities for representing the high-latitude geomagnetic storm event in September 2011, J. Geophys. Res. Space Physics, 119, 10,581 10,594, doi:. Received 4 JUN 2014 Accepted 27 NOV 2014 Accepted article online 1 DEC 2014 Published online 20 DEC 2014 Ionosphere data assimilation capabilities for representing the high-latitude geomagnetic storm event in September 2011 Dmitry Solomentsev 1,2, Knut Stanley Jacobsen 3, Boris Khattatov 4,5, Vyacheslav Khattatov 1, Yakov Cherniak 1, and Anton Titov 1,2 1 Central Aerological Observatory, Dolgoprudny, Russia, 2 Moscow Institute of Physics and Technology, State University, Dolgoprudny, Russia, 3 Norwegian Mapping Authority, Hønefoss, Norway, 4 Fusion Numerics LLC, Boulder, Colorado, USA, 5 Laboratory of High Precision GNSS Positioning, National Research University of Information Technologies, Mechanics and Optics, St. Petersburg, Russia Abstract Severe geomagnetic storms have a strong impact on space communication and satellite navigation systems. Forecasting the appearance of geomagnetically induced disturbances in the ionosphere is one of the urgent goals of the space weather community. The challenge is that the processes governing the distribution of the crucial ionospheric parameters have a rather poor quantitative description, and the models, built using the empirical parameterizations, have limited capabilities for operational purposes. On the other hand, data assimilation techniques are becoming more and more popular for nowcasting the state of the large-scale geophysical systems. We present an example of an ionospheric data assimilation system performance assessment during a strong geomagnetic event, which took place on 26 September The first-principle model has assimilated slant total electron content measurements from a dense network of ground stations, provided by the Norwegian Mapping Authority. The results have shown satisfactory agreement with independent data and demonstrate that the assimilation model is accurate to about 2 4 total electron content units and can be used for operational purposes in high-latitude regions. The operational system performance assessment is the subject of future work. 1. Introduction Space weather conditions and the ionosphere state are of great importance for the Global Navigational Satellite Systems (GNSS) operations. Even without any disturbances caused by solar or geomagnetic events, regular ionospheric plasma densities cause Global Positioning System (GPS) signal delay in the range from 1 to 10 m. During magnetic disturbances, which occur at high latitudes on a regular basis, the density of the charged particles in the ionosphere can increase by more than 100%. Nowcasting and forecasting of the ionospheric state in the presence of strong geomagnetic or solar activity are therefore of high importance for GNSS-based applications and operations. Currently, the most commonly used tool for geophysical system behavior forecasting is an empirical or the first-principle model. The empirical models are based on the limited data sets, which often make them inaccurate in a number of specific cases. As for the first-principle models, the physical factors or drivers that control the ionospheric state at high latitudes lack a quantitative description, and their numerical description lacks sufficient accuracy. Data assimilation methods are gaining more attention in the space weather community. This set of techniques allows one to adjust model results to the given observations, thus making initial conditions for the next step increasingly more accurate with time. Several research groups have independently developed data assimilation ionosphere models, which utilize different underlying physics and have different data assimilation algorithms. Such works as Schunk et al. [2004], Wang et al. [2004], Millward et al. [2001], and Bust and Datta-Barua [2014] should be mentioned to name a few. One of the models, performing data assimilation technique for ionosphere nowcasting, was developed by Fusion Numerics LLC (Boulder, CO) and Central Aerological Observatory (Dolgoprudny, Russia). The detailed model outline is given by Khattatov et al. [2005]and Solomentsev et al. [2012]. One of the goals of the developed model is to provide the ionosphere corrections for single- and double-frequency GNSS receivers. Usage of the data assimilation model for the GNSS-based application requires rigorous case studies to assess the model s capabilities under different conditions. SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,581

2 The aim of the current paper is to introduce the sample results of the data assimilation model performance during severe geomagnetic storm conditions at high latitudes. The paper contains the description of the core first-principle model and the observations used for assimilation. The latter are GNSS-derived slant total electron content (TEC) values, obtained from the dense GNSS receiver network around Norway, provided by the Norwegian Mapping Authority. The event chosen for reproduction took place on 26 September The geomagnetic storm with the geomagnetic activity index, Kp, reaching the value of 7, was caused by a large coronal mass ejection that had a strong impact on the polar ionosphere and the magnetosphere. Modeling results, obtained using the ionosphere data assimilation system, reveal specific ionosphere behavior during the active phase of the geomagnetic storm. The comparison of modeling results with the independent observations show that the model-derived distributions are realistic and can be relied upon within the current case study and, thus, can be used to study the ionosphere behavior under the forced conditions of this current geomagnetic event. The results given below could be of interest for both the space weather community and GNSS-based service providers, aiming for stability and precision. 2. Assessment of Data Assimilation Capabilities for Ionospheric Nowcasting at High Latitudes During a Geomagnetic Event 2.1. The Ionosphere Model The data assimilation models often consist of two main units: the core physics-based model, describing the most important processes influencing the system of interest, and the data assimilation system, gathering and preparing the observations and adjusting the model results according to the modeling and observations error values. The data assimilation itself is a continuous loop process, consisting of physical model propagation one step forward, collecting the observations, and adjusting the model results toward the observed values, after which the physical model begins the next step from the initial conditions, containing the information about the true model state on the previous time step. The core physics-based model, used in the current research, takes into account the important physical processes that occur in the ionosphere. The model solves the continuity, momentum, and energy balance equations for electrons and seven species of ions, i.e., H +,He +,O +,O + 2,N +,N + 2, and NO +. The model s prognostic variables, therefore, are the ionic and electron densities, velocities, and temperatures. The results are obtained on the regular global grid with user-defined resolution. The list of the processes, described by model s equations, is provided below: 1. Photoionization and recombination. The ions and free electrons are produced and lost during the photochemical reactions. Concerning the solar activity, the F 10.7 proxy is used to reproduce radiation intensity at each ionizing wavelength. 2. Chemical reactions. Along with photodissociation, there are 21 chemical reactions involved in the model calculations, influencing the ion density distribution. 3. Geomagnetic field. The geomagnetic dipole coordinates are used as a reference frame for the physical model, making the equations essentially one dimensional. 4. Dynamic terms, e.g., ion and electron pressure and gravity. These terms are used to adjust the ion velocity parallel to the magnetic field in the momentum equation. 5. Neutral atmosphere-ionosphere interactions. The neutral atmosphere is presented within the equations with the external empirical models Mass Spectrometer Incoherent Scatter Atmosphere Model [Hedin, 1991] and horizontal wind model [Hedin et al., 1988]. In more detail, the physical model is described in Solomentsev et al. [2012, 2013]. It is worth mentioning that the high-latitude processes governing ionospheric parameter distribution are included in the model equations (e.g., Wiemer polar cap convection term, presented by Weimer [1995]). Nevertheless, some of the important drivers, such as auroral precipitation, are not included, since there were no robust and precise parameterizations for that kind of effects at the time of model development. Up to the moment, several models of auroral precipitation exist, and their implementation within the framework of the current data assimilation model is a subject of future work. The example given in Figure 1 shows the near-global map of TEC distribution according to the model results with no data assimilation. SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,582

3 Figure 1. Near-global map of vertical TEC distribution for 12 May 2014, 14:00, calculated using the physical model results without any data being assimilated. The color bar units are electrons per m Data Assimilation The operational ionosphere model, used in the current case study, was designed to assimilate GNSS-derived slant total electron content measurements. To extract this kind of information from the GNSS signal and then utilize it to correct the physical model predictions, the following information is used: 1. GNSS observation data. The information on the pseudorange and the carrier phase, received from each visible satellite, is contained in the observation files in RINEX (receiver-independent exchange) format. While operating on a global scale, the model downloads RINEX observation files from about 300 globally distributed stations, for which stable and regular updates are available. 2. GNSS navigational data. To determine the satellite position at the signal transmission time, the automatically generated orbit parameters are used. We assume that rather small errors (less than 1 km) cannot significantly impact the observationsaccuracy,giventhemodelgridresolutionof1 inlatitude per 1 in longitude per 20 km in altitude. 3. Differential code biases (DCB). The DCBs values can often be larger than ionosphere-induced delay in terms of signal propagation time. If ignored, DCBs values can lead to meaningless observations, e.g., negative TEC values. The data, gathered and cached by the preprocessing system, are then used to adjust the model-derived ionosphere state. The adjustment is calculated using the Kalman filter equations as given below. x f ði þ 1Þ ¼ Mx a ðþ i P f ði þ 1Þ ¼ MPM T þ QðÞ i KðÞ¼P i f ðþh i T ðþ i H ðþp i f ðþh i T ðþþr i ðþ i x a ðþ¼x i f ðþþk i ðþyi i ½ ðþ HðÞx i f ðþ i Š P a ðþ¼ i ½I KðÞH i ðþ i ŠPðÞ i 1 (1) Here x f is the forecast model state, i.e., a vector, containing electron densities in all the model grid points before the data assimilation. It results from the model state vector after the assimilation x a multiplied by the nonlinear physical model operator M. Also, i is the time step number, P is the covariance of x with the corresponding subscript, and H is the observational operator, mapping the 3-D electron density distribution to the slant total electron content values analogous to those gathered from the GNSS data. In the set of equation (1), y is the observations vector, K is the so-called gain matrix, Q is the modeling error matrix, and I is SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,583

4 Figure 2. Map of vertical TEC distribution for 12 May 2014, 14:00, calculated using the data assimilation model results. The system was working for 1 week and at this particular moment assimilated 450 slant TEC measurements. The color bar units are electrons per m 2. The round dots represent the locations of the stations used for assimilation. the identity matrix. The covariance matrix, obtained after the analysis, is used for the next time step of the model as the forecast covariance. The first time step covariance is initialized using a small ensemble of the ionosphere models starting with the slightly perturbed initial conditions. Because of the big system size (720,000 grid points in the current configuration), the equations of the Kalman filter cannot be applied directly and several simplifying assumptions are made to perform the calculations and minimize the storage. In fact, only diagonal elements of the covariance are computed directly, while the other elements are estimated using these values. A more detailed outline of the calculation methods and the assumptions can be found in Solomentsev et al. [2012] and Khattatov et al. [2005]. In Figure 2, an example of the data assimilation model results is given. The global TEC map represents the same moment in time, as shown in Figure 1. The qualitative comparison of the two figures shows that the current ionosphere state could significantly differ from the average ionosphere. On a regular basis, the model results are verified by comparing with independent observations. The example given below in Figure 3 shows model-derived electron density profiles compared against a sample observation by the Constellation Observing System for Meteorology, Ionosphere and Climate (COSMIC/FORMOSAT-3). The COSMIC/FORMOSAT-3 scientific mission aims to get more information about the Earth s upper atmosphere and ionosphere and is described thoroughly in the work of Rocken et al. [2000]. The sample electron density profile comparisons similar to those shown in the Figure 3 are made globally on a daily basis, and the statistics are calculated. The results of the statistical estimates of the comparisons are shown in Figure 4. The estimated observations error is electrons per m 3. The model used for this kind of comparison for Figures 3 and 4 did not assimilate any COSMIC/FORMOSAT-3 profiles. The only source of observations used is the set of ground-based GNSS receivers. The results of a regular validation show a reasonable error value for different kinds of comparison. For example, the global model with data assimilation is capable of predicting the slant TEC values obtained from nonassimilated GNSS receivers with about a 2 4 TECU accuracy (1 TECU is a TEC unit, equal to electrons per m 2 ). Nevertheless, given the situation of a high-latitude geomagnetic storm, both scientific and technological challenge for the space weather community, the capabilities of the data assimilation model should be assessed. The estimates of model error and definition of its sources can lead to possible improvements and usage of the modelforoperationalionospheric monitoring in high-latitude conditions. SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,584

5 Figure 3. Ionosphere data assimilation model results compared to the COSMIC/FORMOSAT-3 mission observations during the period of calm ionosphere conditions over the region of interest. (top) The electron density profiles derived from COSMIC data (blue circles) and from the data assimilation model results (green circles) for 24 September (bottom left) The electron density profiles derived from COSMIC data (blue circles) and from the model results without data assimilation (red circles) for the same moment in time. (bottom right) The signal tangent point trajectory over the Earth s surface (red line). To obtain this figure, the modeling results were interpolated to the tangent point locations from the global three-dimensional grid The Storm Event Modeling and the Assessment of Model Performance During Elevated Geomagnetic Activity The following section estimates the model performance during more challenging conditions of elevated geomagnetic activity, when we expect steeper gradients and more rapid temporal changes in electron density and TEC. To capture the steeper gradients, a denser array of GNSS stations is required, which are not available globally and in real time. This section of the validation during storm time conditions therefore Figure 4. Histogram of data assimilation model deviations from the electron density profiles obtained from the COSMIC/ FORMOSAT-3 mission for the period from 20 February 2012 to 23 February All global profiles were taken into account. SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,585

6 Journal of Geophysical Research: Space Physics Figure 5. The bar plot of the Kp index variations before, during, and after the chosen event. The data and table are supplied by the National Ocean and Atmospheric Administration, Space Weather Prediction Center, Boulder, CO, USA. focuses on a specific geographic region, where sufficient observations are available. The geomagnetic storm chosen as the subject for the current study took place in September The variation of Kp index before, during, and after the event is given in Figure 5. On a global scale, the data from the Scripps Orbit and Permanent Array Center archive ( edu/) are downloaded in RINEX format to calculate the total electron content values for each satellite-receiver link. For the geomagnetic storm period, the data were provided by the Norwegian Mapping Authority (Hønefoss, Norway). The network of the ground-based GNSS receivers is shown in Figure 6. To provide a better resolution, using the same computing power, a regional version of the model was used in the current study. Model domain was chosen to cover the region of interest and the side territories. Without any additional data, the physical model results essentially represent average ionosphere behavior. The observation RINEX files provided by the Norwegian Mapping Authority (NMA) covered a 1 week interval. In the case studies performed by the data assimilation model, it is often useful to cover a minimum of a week of data with the model results. This shows the evolution of the model solution stability and error development. The external information, containing the receiver antennas differential code biases were also provided by the Norwegian Mapping Authority. Figure 6. The network of ground-based GNSS receivers providing the slant total electron content observations for the data assimilation model. SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,586

7 Figure 7. Time evolution of the TEC during the 25 and 26 September 2011 obtained by the data assimilation model. The figure shows TEC distribution over the region of interest with 6 h time resolution (starting from 14:15 UTC on 25 September 2011 and going from left to right and from top to bottom). The active storm phase is represented by the bottom right, 20:15 UTC on 26 September SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,587

8 Figure 8. Time evolution of the TEC during 27 September 2011 obtained by the data assimilation model. The figure shows TEC distribution over the region of interest with 6 h time resolution (starting from 02:15 UTC and going from left to right and from top to bottom). Figure 7 illustrates the impact of the geomagnetic event on the high-latitude ionosphere captured by the data assimilation ionosphere model. The figure shows the TEC evolution with 6 h time resolution for a period of 3 day period on September Figures 7 and 8 show that the electron density increased significantly at the peak of the geomagnetic storm period at 20 UT on 26 September (check); thus, the case under study is a positive ionospheric storm. The physical processes driving the plasma density increase are likely some combination of increased and expanded magnetospheric convection and more intense auroral production. A simulation of the period using the thermosphere-ionosphere-electrodynamics general circulation model (TIEGCM) coupled thermosphere ionosphere model described in Wang et al. [1999] showed a somewhat weaker increase. The TIEGCM model used no data assimilation but was driven using the observed time history of the Kp index to describe the magnetospheric convection and auroral precipitation (Figure 9). The primary data assimilation model error assessment results are presented below. The validation used in the current study is a common method in the data assimilation model used, for example, in the paper [Araujo-Pradere et al., 2007]. Up to 10 random ground-based GNSS receivers are chosen among the entire ground-based station list and are not included into the assimilation procedure. Instead, their observed SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,588

9 Figure 9. Results of the TEC calculation by the TIEGCM model for the active storm phase (26 September 2011, 20:15 UTC). The input parameters for TIEGCM (indices Kp and F 10.7 ) correspond with the real values; however, no data assimilation has been performed. slant TECs are collected and compared against the analogous model results. Figure 10 gives an estimate of model errors and calculation stability (i.e., the modeling error does not rise during the electron density enhancement period). Figure 10 shows that the root-mean-square model error is about 3 TEC units, and the bias is 0.2 TECU. In terms of pseudorange error caused by such an ionospheric density, this RMS value would lead to approximately 45 cm of group delay of a signal with L1 GPS frequency. The stability of the solution is also observed in the figure, since no significant error increase occurred during the geomagnetic storm event, which is toward the end of the data interval depicted. The ionospheric model without data assimilation during this period shows significantly larger errors (10 TECU in RMS), which rise during the geomagnetic storm period and reach 15 TECU. Thus, provided with the estimate of model errors, the data assimilation model results were analyzed. Several examples of visualization and analysis are given in the next section. Figure 10 also shows that even in the period of high electron density increase, the model accuracy remains the same which shows that the data assimilation methodology is able to capture the ionosphere features during the active storm phase Comparison With Other Techniques One of the main ionosphere model products, which is of interest for the GNSS society, is the TEC map over the selected region. For this case study storm period three different sources of TEC were available and are compared; these include the physics-based data assimilation model results and the interpolation of the GNSS TEC observations without a physical model created by the Norwegian Mapping Authority (NMA) with different elevation cutoffs. Figure 8 compares the TEC maps for a single date and moment of time obtained from the three different sources. The vertical total electron content (VTEC) maps denoted as NMA model results are produced by Kriging interpolation of the VTEC values calculated for each ionospheric pierce point, using a single-layer ionosphere model. The single-layer model is the assumption that the entire TEC of the ionosphere is located in a thin spherical shell. For the calculations performed here, the altitude at which this shell is located is 350 km. The NMA TEC map can use two possible configurations for the 2-D thin-shell ionosphere model, which are described in Jacobsen and Schäfer [2012]. These two configurations concern elevation cutoff for slant observations selection: one option uses a 30 cutoff (which provides more data) and the other uses a 50 cutoff, which leaves only satellites with higher elevation angles. Figures 11 and 12 show the TEC maps obtained by different methods for two different times. The figures can be used to compare the ionospheric morphology obtained by the 3-D assimilative ionospheric model and SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,589

10 Figure 10. Regular validation algorithm results are shown. The data from a group of randomly chosen receivers were processed but not assimilated into the model. The results were compared to the model-derived analogues. (top) The root-mean-square error and bias time evolution. (middle) Relative error evolution. (bottom) A map of validation receivers. the empirical NMA model. The mentioned figures show several points of agreement and discrepancy. For example, in Figure 11, all the models, utilizing almost the same data sets (NMA model with different elevation angles and assimilative 3-D ionosphere model), agree over the region where there is good data coverage. All of them reveal a spot with rather large electron density values in the center of Scandinavia. On the other hand, the thin-shell model reveals another high-density spot between 70 N and 75 N in latitude, which may be a consequence of extrapolation in the Kriging mapping method. Since no independent data are available in that region, there can only be indirect conclusions concerning which of the models is closer to reality. One of these conclusions is that there are very few lines-of-sight passing the region of interest because of mutual positions of GPS satellites and the ground-based receivers in Norway. The first-principle model, however, implies parameterization of polar ionosphere physics to extrapolate the observations, which, theoretically, makes the 3-D assimilative results more probable. In Figure 12, the TEC maps, obtained by different models, are more diverse. While being in good agreement in the southeast region, the maps resulting from model calculations show different ionosphere behavior over central Norway. Both versions of the NMA model show a spot of relatively high TEC values, while the 3-D ionosphere models do not observe it. The sources of such discrepancies could be rather different from SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,590

11 Figure 11. TEC maps for 26 September 2011, 20:15 UMT. From left to right and from top to bottom, the panels represent the following: the TEC map obtained from 3-D data assimilation ionosphere model, the TEC map obtained from 2-D data-based NMA model of the thin-shell ionosphere at 350 km with elevation degree cutoff equal to 30, the locations of the ionospheric pierce points of the GPS signals at 350 km altitude, and the TEC map 2-D NMA ionosphere model with elevation cutoff equal to 50. misrepresented model physics to contradictory observations and errors in satellite and ground-based receiver biases estimation. Note that the NMA model can be locally sensitive to the values from a small number of satellite measurements, which means that the small spot can either be a valid measurement of a small plasma enhancement or it may be caused by an erroneous satellite DCB, for example. In any case, the maps from the 3-D model and NMA are in agreement at large scales. Unlike simple mapping of the TEC, the assimilative ionospheric physical model is able to provide the end user with the three-dimensional ionospheric structure.forexample,atanypointchosen,thevertical profile of electron density can be derived from the model results. Figure 13 shows the electron density profiles changing with day. Vertical distribution of the electron density exhibits expected behavior in the night hours during the calm days 24 and 25 of September 2011, i.e., 267 and 268 days of year 2011 correspondingly. But for the dates, when geomagnetic activity was significantly higher than average, the electron density profiles have shown significant increase in the peak F region plasma density at times when the TEC is elevated. In reality, such an increase could be caused by a number of processes, e.g., the neutral atmosphere ionization by the auroral particles precipitation, plasma transport, or stagnation of plasma in the presence of sunlight. The observations, of course, contain the response to every factor influencing the plasma density, whereas without data the physical model would only match the data if all the drivers are accounted for correctly. Given the uncertainty in the model drivers, the observations constrained by the model physics are likely to be the most prudent technique for mapping the three-dimensional structure of the plasma. One final validation example, in Figure 14, is to compare the physical model data assimilation results with independent data from the Tromsø GNSS dual-frequency receiver. The observations, derived from the SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,591

12 Figure 12. TEC maps for 27 September 2011, 11:15 UMT. From left to right and from top to bottom, the panels represent the following: the TEC map obtained from 3-D data assimilation ionosphere model, the TEC map obtained from 2-D data-based NMA model of the thin-shell ionosphere at 350 km with elevation degree cutoff equal to 30, the locations of the ionospheric pierce points of the GPS signals at 350 km altitude, and the TEC map obtained from 2-D NMA ionosphere model with elevation cutoff equal to 50. ground network slant TEC values, exhibit fast and strong oscillations, most likely caused by the geomagnetic event. The ionospheric data assimilation model does not capture the rapid oscillation in the data but tends to smooth through the variability. On of the possible explanations is that the discrepancies between the model and the data could be related to the relatively rare data assimilation frequency (10 min). Figure 13. Electron density profiles above Rognan (67 05 N, E). Different colored lines show profiles for different days of the year at 19:30 UMT. SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,592

13 Figure 14. Ionosphere data assimilation model results comparisons versus the NMA ionosphere model data. (top) The time series of the TEC over Tromsø obtained from the data assimilation and NMA models. (bottom) The histogram of models differences. Acknowledgments The authors are grateful to Timothy Fuller-Rowell for his helpful and insightful comments which helped to improve the current manuscript. The authors are grateful to Chris Hall (University of Tromsø) for providing and explaining the ionosonde data used in the current study and plotted in Figure 11. The authors would also like to thank all the organizations and the scientific missions mentioned for providing the freely accessible data for research purposes. The authors are grateful to the editor Michael Liemohn and two anonymous reviewers for their insightful comments which helped to improve the manuscript. Michael Liemohn thanks the reviewers for their assistance in evaluating this paper. 3. Conclusions The current study aims to assess data assimilation model capabilities in representing the geomagnetic storm impact on the ionosphere state in the high-latitude region. The observations, assimilated into the model, were provided by the NMA and contain information about slant TEC values observed from the dense network of ground GPS receivers. The challenge of model-based ionosphere nowcasting is that uncertainties in quantitative description of the physical processes and driving forces are likely to cause large differences between model results and the real system state. On the other hand, when forced by the data, the model can maintain a reasonable level of accuracy and provide the user with a detailed ionosphere description. Despite the model uncertainties, there are several advantages of the model-based ionosphere monitoring. These include the following: 1. Ability to get global estimates of the ionosphere state based on data and model combination. Because of the information spread, the model results over the regions with sparse data are more reliable compared to interpolation methods. 2. Ability for the short-term forecasting of the ionosphere state based on the model equations and provided with the adjusted initial conditions. In the current study, it is demonstrated that the physical-based model with the data assimilation capability can be used for ionosphere nowcasting in high-latitude regions during periods with high geomagnetic activity. The RMS error of the data assimilation model, estimated by the standard procedure, is estimated at the level of 3 TECU. The case study, outlined in this paper, was performed by our collaboration team in order to estimate the current data assimilation ionosphere model capabilities in representing high-latitude geomagnetic events. The subject of the future work includes operational model tests and checking the model forecasting capabilities under different conditions. References Araujo-Pradere, E. A., T. J. Fuller-Rowell, P. S. J. Spencer, and C. F. Minter (2007), Differential validation of the USTEC model, Radio Sci., 42, RS3016, doi: /2006rs Bust, G. S., and S. Datta-Barua (2014), Scientific investigations using IDA4D and EMPIRE, in Modeling the Ionosphere-Thermosphere System, edited by J. Huba, R. Schunk, and G. Khazanov, John Wiley, Chichester, U. K., doi: / ch23. Hedin, A. E. (1991), Extension of the MSIS thermosphere model into the middle and lower atmosphere, J. Geophys. Res., 96, , doi: /90ja Hedin, A. E., N. W. Spencer, and T. L. Killeen (1988), Empirical global model of upper thermosphere winds based on atmosphere and dynamics explorer satellite data, J. Geophys. Res., 93, , doi: /ja093ia09p SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,593

14 Jacobsen, K. S., and S. Schäfer (2012), Observed effects of a geomagnetic storm on an RTK positioning network at high latitudes, J. Space Weather Space Clim., doi: /swsc/ Khattatov, B., M. Murphy, M. Gnedin, J. Sheffel, J. Adams, B. Cruickshank, V. Yudin, T. Fuller-Rowell, and J. Retterer (2005), Ionospheric nowcasting via assimilation of GPS measurements of ionospheric electron content in a global physics-based time-dependent model, Q. J. R. Meteorol. Soc., 131, , doi: /qj Millward, G. H., I. C. F. Mueller-Wodarg, A. D. Aylward, T. J. Fuller-Rowell, A. D. Richmond, and R. J. Moffett (2001), An investigation into the influence of tidal forcing on F region equatorial vertical ion drift using a global ionosphere-thermosphere model with coupled electrodynamics, J. Geophys. Res., 106, 24,733 24,744, doi: /2000ja Rocken, C., Y. H. Kuo, W. Schreiner, D. Hunt, and S. Sokolovsky (2000), COSMIC system description, Spec. Iss. Terr. Atmos. Ocean. Sci., 11(1), Schunk, R. W., L. Scherliess, J. J. Sojka, and D. Thompson (2004), Global Assimilation of Ionospheric Measurements (GAIM), Radio Sci., 39, RS1S02, doi: /2002rs Solomentsev, D. V., B. Khattatov, M. Codrescu, A. Titov, V. Yudin, and V. U. Khattatov (2012), Ionosphere state and parameter estimation using the Ensemble Square Root Filter and the global three-dimensional first-principle model, Space Weather, 10, S07004, doi: /2012sw Solomentsev, D. V., B. V. Khattatov, and A. A. Titov (2013), Three-dimensional assimilative ionosphere model for European region, Geomagn. Aeron., 53(1), 78 90, doi: /s Wang, C., G. Hajj, X. Pi, L. Gary Rosen, and B. Wilson (2004), Development of the global assimilative ionospheric model, Radio Sci., 39, RS1S06, doi: /2002rs Wang, W., T. L. Killeen, A. G. Burns, and R. G. Roble (1999), A high-resolution, three-dimensional, time dependent, nested grid model of the coupled thermosphere-ionosphere, J. Atmos. Sol. Terr. Phys., 61, Weimer, D. R. (1995), Models of high-latitude electric potentials derived with a least error fit of spherical harmonic coefficients, J. Geophys. Res., 100(19), , doi: /95ja SOLOMENTSEV ET AL American Geophysical Union. All Rights Reserved. 10,594