Tomoko Matsuo

Annual and semiannual variations of thermospheric density: EOF analysis of CHAMP and GRACE data

[1] In this paper, observations from CHAMP and GRACE during 2002–2010 are used to study the seasonal variations of thermospheric density by characterizing the dominant modes of thermospheric density variability as empirical orthogonal functions (EOFs). Our results showed that the first three EOFs captured most of the density variability, which can be as large as 98% of total density variability. Subsequently, the obtained mean field, first three EOFs and the corresponding amplitudes of three EOFs are applied to construct a thermospheric density model at 400 km to study seasonal variations of thermospheric density under geomagnetically quiet conditions. Thermospheric density shows strong latitudinal dependence in seasonal variation, although it usually has maxima near the equinoxes and minimum in the local winter at middle and high latitudes. Semiannual variations imbedded in the annual variations are seen at all latitudes; annual variations however become dominant in the southern hemisphere. Specifically, the observations show that the annual amplitude can reach as large as 40–50% of the annual mean at high latitudes in the southern hemisphere and it decreases gradually from the southern to northern hemisphere. The semiannual component to the annual mean is about 15–20% without significant latitudinal dependence. Additionally, the relative amplitudes of annual and semiannual variations in the MSISE00 density agree fairly well with the observations, albeit the MSISE00 gives an opposite solar activity dependence for the annual and semiannual variations compared with the positive F107 dependence seen in the observations.

A real‐time run of the Coupled Thermosphere Ionosphere Plasmasphere Electrodynamics (CTIPe) model

[1] The availability of unprecedented amounts of real-time data from Global Navigation Satellite Systems and ionosondes coupled with new and more stringent requirements for specification and forecast of the neutral and electron densities in the thermosphere-ionosphere system are driving a new wave of development in data assimilation schemes for the thermosphere and ionosphere. However, such schemes require accurate knowledge of any biases affecting the state-propagating models, and characterizing such biases involves significant effort. A first step in the estimation of the model biases, a steady state neutral temperature comparison with the empirical Mass Spectrometer Incoherent Scatter model, was published in Space Weather in 2008. Here we present another step in the validation of the Coupled Thermosphere Ionosphere Plasmasphere Electrodynamics (CTIPe) general circulation model in preparation for its future inclusion in a data assimilation scheme. We describe an implementation of the model at the Space Weather Prediction Center (SWPC) and present real-time comparisons between CTIPe and GPS total electron content and F2 layer ionosonde measurements. The CTIPe results are generated automatically about 20 min ahead of real time. The model inputs are based on NASA’s Advanced Composition Explorer and F10.7 data available in the SWPC database. The results and the comparison with measurements for the current 2-week period are available at http://helios.swpc.noaa.gov/ctipe/. The results are quite encouraging and offer hope that physics-based models can compete with empirical models during quiet times and have tremendous potential to provide more reliable forecasts during periods of geomagnetic disturbance.

Comparison of magnetic perturbation data from LEO satellite constellations: Statistics of DMSP and AMPERE

During the past decade engineering-grade magnetic field measurements from the low Earth orbiting (LEO) Iridium constellation of communication satellites have been available to the geospace science community as a tool to map field-aligned currents. The Active Magnetosphere and Planetary Electrodynamics Response Experiment (AMPERE) applied to Iridium measurements markedly improved the temporal and spatial resolution of these data. We developed new methods to compare data from the latest improvement to AMPERE with those from a constellation of four LEO Defense Meteorological Satellite Program (DMSP) spacecraft that carry high-resolution magnetometers. To perform the comparisons, we transformed all data to a common coordinate frame and altitude (110 km) and developed a means of computing spacecraft magnetic conjunctions. These conjunctions yield discrepancies in the magnetic field perturbations measured at each proximate spacecraft. During the geomagnetic disturbance of 29–30 May 2010, the vector differences in the horizontal perturbations at closest approach (typically a few tens of kilometers) had mean, median, and standard deviation values of 132 nT, 112 nT, and 90 nT, respectively. The DMSP spacecraft tend to report larger perturbations in the northern polar cap and cusp regions, especially during active intervals. We attribute some of the differences to limitations of spacecraft-attitude knowledge that propagate into AMPERE data. Overall, for the magnetic storm, we provide clear evidence that AMPERE data can provide high-resolution auroral zone data in good agreement with DMSP data for use in data assimilation algorithms. Such dual-use commercial data can provide important global augmentation to the nations space weather monitoring capabilities.

Effects of inferring unobserved thermospheric and ionospheric state variables by using an Ensemble Kalman Filter on global ionospheric specification and forecasting

This paper demonstrates the significance of ion-neutral coupling to ionospheric data assimilation for ionospheric specification and forecast. Ensemble Kalman Filter (EnKF) is used to assimilate synthetic electron density profiles sampled according to the Formosa Satellite 3/Constellation Observing System for Meteorology, Ionosphere, and Climate into the Thermosphere-Ionosphere-Electrodynamics General Circulation Model (TIEGCM). The combination of the EnKF and first-principles TIEGCM allows a self-consistent treatment of thermosphere and ionosphere coupling in the data assimilation and forecast. Because thermospheric variables affect ionospheric electron densities, different combinations of an observed ionospheric state variable (electron density), and unobserved ionospheric and thermospheric state variables (atomic oxygen ion density, neutral temperature, winds, and composition) are included as part of the EnKF state vector in experiments. In the EnKF, the unobserved state variables are estimated and made dynamically and chemically consistent with the observed state variable, thus improving the performance of the data assimilation system. The impact on ensemble forecast is further examined by initializing the TIEGCM with the assimilation analysis. The main findings are the following: (1) by incorporating ion-neutral coupling into the EnKF, the ionospheric electron density analysis, and forecast can be considerably improved. (2) Thermospheric composition is the most significant state variable that affects ionospheric analysis and forecast. (3) Thermospheric variables have a much longer impact on ionospheric forecast (>24 h) than ionospheric variables (2 to 3 h). (4) In the TIEGCM, the effect of assimilating electron densities is not completely transmitted to the forecast step unless the densities of ion species are estimated.

Mapping high-latitude ionospheric electrodynamics with SuperDARN and AMPERE

An assimilative procedure for mapping high-latitude ionospheric electrodynamics is developed for use with plasma drift observations from the Super Dural Auroral Radar Network (SuperDARN) and magnetic perturbation observations from the Active Magnetosphere and Planetary Electrodynamics Response Experiment (AMPERE). This procedure incorporates the observations and their errors, as well as two background models and their error covariances (estimated through empirical orthogonal function analysis) to infer complete distributions of electrostatic potential and vector magnetic potential in the high-latitude ionosphere. The assimilative technique also enables objective error analysis of the results. Various methods of specifying height-integrated ionospheric conductivity, which is required by the procedure, are implemented and evaluated quantitatively. The benefits of using both SuperDARN and AMPERE data to solve for both electrostatic and vector magnetic potentials, rather than using the data sets independently or solving for just electrostatic potential, are demonstrated. Specifically, solving for vector magnetic potential improves the specification of field-aligned currents (FACs), and using both data sets together improves the specification of features in regions lacking one type of data (SuperDARN or AMPERE). Additionally, using the data sets together results in a better correspondence between large-scale features in the electrostatic potential distribution and those in the FAC distribution, as compared to using SuperDARN data alone to infer electrostatic potential and AMPERE data alone to infer FACs. Finally, the estimated uncertainty in the results decreases by typically ∼20% when both data sets rather than just one are included.

Ionospheric data assimilation with thermosphere-ionosphere-electrodynamics general circulation model and GPS-TEC during geomagnetic storm conditions

The main purpose of this paper is to investigate the effects of rapid assimilation-forecast cycling on the performance of ionospheric data assimilation during geomagnetic storm conditions. An ensemble Kalman filter software developed by the National Center for Atmospheric Research (NCAR), called Data Assimilation Research Testbed, is applied to assimilate ground-based GPS total electron content (TEC) observations into a theoretical numerical model of the thermosphere and ionosphere (NCAR thermosphere-ionosphere-electrodynamics general circulation model) during the 26 September 2011 geomagnetic storm period. Effects of various assimilation-forecast cycle lengths: 60, 30, and 10 min on the ionospheric forecast are examined by using the global root-mean-squared observation-minus-forecast (OmF) TEC residuals. Substantial reduction in the global OmF for the 10 min assimilation-forecast cycling suggests that a rapid cycling ionospheric data assimilation system can greatly improve the quality of the model forecast during geomagnetic storm conditions. Furthermore, updating the thermospheric state variables in the coupled thermosphere-ionosphere forecast model in the assimilation step is an important factor in improving the trajectory of model forecasting. The shorter assimilation-forecast cycling (10 min in this paper) helps to restrain unrealistic model error growth during the forecast step due to the imbalance among model state variables resulting from an inadequate state update, which in turn leads to a greater forecast accuracy.

Modes of high-latitude auroral conductance variability derived from DMSP energetic electron precipitation observations: Empirical orthogonal function analysis

We provide the first ever characterization of the primary modes of ionospheric Hall and Pedersen conductance variability as empirical orthogonal functions (EOFs). These are derived from six satellite years of Defense Meteorological Satellite Program (DMSP) particle data acquired during the rise of solar cycles 22 and 24. The 60 million DMSP spectra were each processed through the Global Airlglow Model. Ours is the first large-scale analysis of ionospheric conductances completely free of assumption of the incident electron energy spectra. We show that the mean patterns and first four EOFs capture ∼50.1 and 52.9% of the total Pedersen and Hall conductance variabilities, respectively. The mean patterns and first EOFs are consistent with typical diffuse auroral oval structures and quiet time strengthening/weakening of the mean pattern. The second and third EOFs show major disturbance features of magnetosphere-ionosphere (MI) interactions: geomagnetically induced auroral zone expansion in EOF2 and the auroral substorm current wedge in EOF3. The fourth EOFs suggest diminished conductance associated with ionospheric substorm recovery mode. We identify the most important modes of ionospheric conductance variability. Our results will allow improved modeling of the background error covariance needed for ionospheric assimilative procedures and improved understanding of MI coupling processes.

Dominant modes of variability in large‐scale Birkeland currents

Properties of variability in large-scale Birkeland currents are investigated through empirical orthogonal function (EOF) analysis of 1 week of data from the Active Magnetosphere and Planetary Electrodynamics Response Experiment (AMPERE). Mean distributions and dominant modes of variability are identified for both the Northern and Southern Hemispheres. Differences in the results from the two hemispheres are observed, which are attributed to seasonal differences in conductivity (the study period occurred near solstice). A universal mean and set of dominant modes of variability are obtained through combining the hemispheric results, and it is found that the mean and first three modes of variability (EOFs) account for 38% of the total observed squared magnetic perturbations (δB2) from both hemispheres. The mean distribution represents a standard Region 1/Region 2 (R1/R2) morphology of currents and EOF 1 captures the strengthening/weakening of the average distribution and is well correlated with the north-south component of the interplanetary magnetic field (IMF). EOF 2 captures a mixture of effects including the expansion/contraction and rotation of the (R1/R2) currents; this mode correlates only weakly with possible external driving parameters. EOF 3 captures changes in the morphology of the currents in the dayside cusp region and is well correlated with the dawn-dusk component of the IMF. The higher-order EOFs capture more complex, smaller-scale variations in the Birkeland currents and appear generally uncorrelated with external driving parameters. The results of the EOF analysis described here are used for describing error covariance in a data assimilation procedure utilizing AMPERE data, as described in a companion paper.

Inverse procedure for high-latitude ionospheric electrodynamics: Analysis of satellite-borne magnetometer data

This paper presents an analysis of data from the magnetometers on board the Defense Meteorological Satellite Program (DMSP) F-15, F-16, F-17, and F-18 satellites and the Iridium satellite constellation, using an inverse procedure for high-latitude ionospheric electrodynamics, during the period of 29–30 May 2010. The Iridium magnetometer data are made available through the Active Magnetosphere and Planetary Electrodynamics Response Experiment (AMPERE) program. The method presented here is built upon the assimilative mapping of ionospheric electrodynamics procedure but with a more complete treatment of the prior model uncertainty to facilitate an optimal inference of complete polar maps of electrodynamic variables from irregularly distributed observational data. The procedure can provide an objective measure of uncertainty associated with the analysis. The cross-validation analysis, in which the DMSP data are used as independent validation data sets, suggests that the procedure yields the spatial prediction of DMSP perturbation magnetic fields from AMPERE data alone with a median discrepancy of 30–50 nT. Discrepancies larger than 100 nT are seen in about 20% of total samples, whose location and magnitude are generally consistent with the previously identified discrepancy between DMSP and AMPERE data sets. Resulting field-aligned current (FAC) patterns exhibit more distinct spatial patterns without spurious high-frequency oscillatory features in comparison to the FAC products provided by AMPERE. Maps of the toroidal magnetic potential and FAC estimated from both AMPERE and DMSP data under four distinctive interplanetary magnetic field (IMF) conditions during a magnetic cloud event demonstrate the IMF control of high-latitude electrodynamics and the opportunity for future scientific investigation.

Nonstationary covariance modeling for incomplete data: Monte Carlo EM approach

A multi-resolution basis can provide a useful representation of nonstationary two-dimensional spatial processes that are typically encountered in the geosciences. The main advantages are its flexibility for representing departures from stationarity and importantly the scalability of algorithms to large numbers of spatial locations. The key ingredients of our approach are the availability of fast transforms for wavelet bases on regular grids and enforced sparsity in the covariance matrix among wavelet basis coefficients. In support of this approach we outline a theoretical proposition for decay properties of the multi-resolution covariance for mixtures of Matern covariances. A covariance estimator, built upon a regularized method of moment, is straightforward to compute for complete data on regular grids. For irregular spatial data the estimator is implemented by using a conditional simulation algorithm drawn from a Monte Carlo Expectation Maximization approach, to translate the problem to a regular grid in order to take advantage of efficient wavelet transforms. This method is illustrated with a Monte Carlo experiment and applied to surface ozone data from an environmental monitoring network. The computational efficiency makes it possible to provide bootstrap measures of uncertainty and these provide objective evidence of the nonstationarity of the surface ozone field.