Bust 2008

HISTORY, CURRENT STATE, AND FUTURE
DIRECTIONS OF IONOSPHERIC IMAGING

Gary S. Bust1,2 and Cathryn N. Mitchell3
Received 4 September 2006; revised 1 May 2007; accepted 6 August 2007; published 26 February 2008.
[1] With the current data availability from both ground- imaging is presented. The ability of imaging algorithms to
and space-based sources, the network of ground-based ingest multiple types of data and use advanced inverse
Global Positioning System (GPS) receivers, GPS occultation techniques borrowed from meteorological data assimilation
receivers, in situ electron density sensors, and dual- to produce four-dimensional images of electron density is
frequency beacon transmitters, the time is right for a discussed. Particular emphasis is given to the mathematical
comprehensive review of the history, current state, and basis for the different methods. The science that ionospheric
future directions of ionospheric imaging. A brief introduction imaging addresses is discussed, and the scientific
and history of ionospheric imaging is presented, beginning contributions that ionospheric imaging has made are
with computerized ionospheric tomography. Then, a described. Finally, future directions for this research area
comprehensive review of the current state of ionospheric are outlined.
Citation: Bust, G. S., and C. N. Mitchell (2008), History, current state, and future directions of ionospheric imaging, Rev. Geophys., 46,
RG1003, doi:10.1029/2006RG000212.
1. INTRODUCTION and thus open the ionosphere to the driving force of the
solar wind. Introductory texts on the ionosphere are given
[2] The overall aim of this review paper is to summarize by Hargreaves [1995] and Davies [1990].
the direction and challenges of an area at the forefront of [4] The physics of the Sun-Earth system poses a number
research: ionospheric imaging. This is the production of of scientific challenges. Above the Earth’s gaseous atmo-
two-, three-, and four-dimensional images of the free sphere is a fascinating region of complex physics: fluid
electron density in the Earth’s upper atmosphere. dynamics, magnetic fields, solar radiation, and particles.
[3] The ionosphere is the upper region of the Earth’s Solar storms ultimately have their major terrestrial impact
atmosphere where a small but significant number of the when they encounter the ionosphere through magneto-
neutral atoms are ionized, resulting in free electrons and sphere/ionosphere coupling. The physics of the Sun-Earth
ions (a plasma). The ionization levels in this near-Earth system is described in detail by Kivelson and Russell
space plasma are controlled by solar extreme ultraviolet [1995]. The ionosphere responds to magnetospheric inputs
(EUV) radiation and particle precipitation. The dynamics of in a number of different ways, with changes in electron and
the neutral atmosphere plays a significant role in causing ion temperature, electron and ion and neutral density and
movement of the ionized particles by collisions with neutral mixing of the neutral atmosphere resulting in changes to the
atoms and molecules from the surrounding thermosphere. ionic species. All of these are important physical parame-
The ionosphere is embedded within the Earth’s magnetic ters, but arguably, the most important is the electron density.
field and thus is constrained by interactions of the ionized Ionospheric imaging of the electron density provides snap-
particles with the magnetic field. At middle and low shots of the global plasma structure and its temporal
latitudes the ionosphere is contained within a region of evolution. The overall plasma structure and temporal evo-
closed field lines, whereas at high latitudes the geomagnetic lution in a large spatial region (such as the polar cap) can be
field can reconnect with the interplanetary magnetic field related to important energy and momentum drivers such as
variation in the solar magnetic field, changes in the total
1
Atmospheric and Space Technology Research Associates, San polar cap potential, hemispheric power, and the amount of
Antonio, Texas, USA. energy in the precipitation of electrons and ions from the
2
Formerly at Applied Research Laboratories, University of Texas at
Austin, Austin, Texas, USA. magnetosphere.
3
Department of Electronic and Electrical Engineering, University of
Bath, Bath, UK.
Copyright 2008 by the American Geophysical Union. Reviews of Geophysics, 46, RG1003 / 2008
1 of 23
8755-1209/08/2006RG000212 Paper number 2006RG000212
RG1003
RG1003 Bust and Mitchell: IONOSPHERIC IMAGING RG1003
[5] The electron density is certainly the most important and discussion of future directions of research and devel-
parameter from the applications perspective because it opment for 4-D imaging is presented in section 7.
governs all of the effects on radio signals. It is this fact
that makes ionospheric imaging such a useful technique for 2. THEORY: TWO-DIMENSIONAL IMAGING
radio systems applications; it allows the three-dimensional
(3-D) time-evolving (hence four-dimensional (4-D)) spatial 2.1. Introduction to Tomography and Ionospheric
field of electron density to be imaged, and this is exactly the Tomography
requirement for almost all radio systems applications. Mea- [8] The mathematical problem of reconstructing a func-
suring changes in the ionosphere, in particular the electron tion from its projections was originally solved by Radon
density, is central to understanding the solar-terrestrial [1917]. However, the first practical application was not
environment impact on communication, surveillance, and published until 1956, when the tomographic method was
navigation systems here on Earth. By having an organized applied to radio astronomy [Bracewell, 1956].
3-D near-global map of electron density updated regularly [9] The strict definition of tomography reflects its deri-
in time, the radio frequency system user can apply correc- vation from the Greek words ‘‘to cut or section’’ (tomos)
tions both where and when required. McNamara [1991] and ‘‘to write’’ (graphein), hence the use of the term for
discusses some of the ionospheric effects on radio systems. obtaining 2-D pictures of the interior of the body from
[6] Ionospheric imaging involves using integrated meas- integrated measurements. The use of tomography has now
urements of electron density, known as total electron broadened to cover a wide range of 2-D and 3-D inverse
content (TEC) measurements, to produce two-, three-, and problems that originate with multiangle measurements that
four-dimensional maps of electron density. Most of the relate to the interior of the object.
basic TEC measurements are obtained from networks of [10] Tomography is an example of an inverse problem,
geodetic Global Positioning System (GPS) receivers that where imaging of a particular quantity is performed from
measure the ionosphere in order to improve the accuracy of diverse measurements. The technique has many applications
their position calculations. These TEC measurements can be and is found in many different areas of science such as
used directly or organized into two-dimensional (2-D) TEC archaeology, biology, geology, and, the best known exam-
maps to infer information on the horizontal structuring of ple, in medicine. The recent interest in tomographic imaging
the electron density. However, information on how plasma began with the invention of the X-ray computerized tomog-
can be lifted to high altitudes and transported to other raphy scanner by Hounsfield [1972]. This original medical
regions, polar outflow, and other vertical dynamical changes application, the computer-aided tomography (CAT) scanner,
is lost with such simple mapping algorithms. In order to took measurements of the attenuation of X rays passing
obtain information on the vertical structure of the electron through a human body from many different angles. By
density, its temporal variation, and transport, 4-D imaging is converting these measurements directly into digital
necessary. Ionospheric imaging to obtain electron density is impulses and feeding them into a computer, a two-dimen-
complementary to using point profiles from ionosondes and sional, cross-sectional image of the body was obtained. In
incoherent scatter radars (ISRs). For the point profiles, there modern computed tomography a three-dimensional image
is no information about large-scale horizontal gradients and of a body and its internal organs is formed from a series of
convection of plasma that causes structuring at the profile two-dimensional X-ray images. More recent developments
position. The time-evolving nature of 4-D imaging is crucial in the medical field have seen the technique applied to
since for most raw satellite TEC data used in the imaging nuclear medicine, magnetic resonance imaging, and ultra-
the temporal and spatial sampling within a given region is sound and microwave imaging. Further information about
constantly changing. Moreover, the ionospheric conditions, tomographic imaging is given by Kak and Slaney [2001].
electron density in particular, can significantly change with [11] The application of tomography to imaging the elec-
space and time. With 4-D imaging, there is the capability to tron density in the ionosphere falls into the category of ray
understand how plasma structures evolve in time and how tomography. It has the advantage that the measurements can
they affect local smaller-scale plasma structuring. be approximated as straight-line integrals of the electron
[7] The paper is organized as follows: Section 2 describes density, and it is therefore not necessary to construct an
the theoretical underpinnings and historical developments algorithm to cope with ray bending, which only introduces a
of two-dimensional ionospheric tomographic imaging, in- few kilometer path difference for GPS, for horizontal
cluding a short tutorial on computerized tomography. Sec- occultations [Schreiner et al., 1999]. On the other hand,
tion 3 is focused on 3-D and 4-D ionospheric imaging. The the measurements are compromised by the incomplete
mathematical development of 3-D imaging methods is geometrical coverage and by assumptions of stationarity
presented for several different approaches, including those of the medium. While both of these limitations have been
that make use of data assimilation techniques. Early results addressed in the last few years and improvements have been
and experimental validation of 2-D and 3-D imaging is achieved, they remain important considerations in iono-
presented in section 4, while section 5 focuses on recent spheric imaging.
scientific results of 4-D imaging at high, middle, and low [12] The number of free electrons in a column (defined as
latitudes. Application results and developments of 4-D of unit cross-sectional area) from the satellite (S) to the
imaging are presented in section 6. Finally, a summary receiver (R) is the slant TEC along the path. This TEC can
2 of 23
be determined from the differential phase [see, e.g., Davies, [15] With LEO TEC data from NIMS or Cicada a rich set
1990] of dual-frequency satellite transmissions, but in that of data is available along the track of the satellite for rays
case it contains an unknown constant, c, because of the between the satellite and ground array of receivers. The
unknown number of differential phase cycles at the start of ionosphere can be considered static, and a 2-D tomographic
the measurement. This measured TEC is related to the true image can be produced along the satellite track. However,
slant TEC by the 2-D image is necessarily regional (10° in horizontal
distance), and the images are only produced when satellites
ZR pass over the array. For a midlatitude site, four NIMS
ys ¼ N ds þ c; ð1Þ satellites (the current number available) produce 10 passes
S per day to a receiver array. Thus traditional 2-D LEO
tomography is limited both spatially and temporally. GPS,
where ys is the measured TEC along the path, N is the on the other hand, continuously provides TEC data to a
electron density, ds is the differential distance along the path ground receiver from multiple satellites. However, the
between satellite and receiver, and c is an unknown satellites move slowly through the ionosphere (100 m/s
constant. TEC is the basic measurement used in ionospheric at F region altitudes). Thus, in order to develop global 3-D
imaging, and it is important to appreciate that the algorithms images of electron density, large numbers of ground GPS
always have to manage the TEC calibration (the determina- receivers are required, and some method of accounting for
tion of the constant) in some manner. For GPS imaging this temporal evolution of the ionosphere must be included in
can be assisted by using the differential code technique, the imaging algorithm. Modern 4-D imaging algorithms
where the code data [see, e.g., Mannucci et al., 1999] give have developed methods to account for both of these issues.
an absolute signal delay and hence absolute TEC but Finally, the radio frequencies used by GPS and LEO
include satellite- and receiver-specific interfrequency biases systems can have an effect on the fidelity of the imaging.
(IFB). Fortunately, the IFB values are relatively stable During severe magnetic storms in regions of strong scintil-
[Mannucci et al., 1999] and not dependent on the lation, receivers often lose lock at 150 MHz, causing data
ionosphere. This is not the case for the unknown number dropouts that can last for several hours. In addition, at
of differential phase cycles at the start of the measurement, lower-elevation angles, bending at 150 MHz can become an
which is ionosphere-dependent and hence unique for each important effect. Both of these effects reduce the effective-
measurement arc. Thus, for GPS, estimates of the IFBs can ness of LEO tomography, particularly during severe mag-
be obtained by taking models of the ionosphere (empirical netic storms.
or from 2-D to 4-D imaging) and computing average values [16] Recently, GPS receivers on LEO satellites have
of the IFB over several hours. However, any systematic provided new data sources that have been added to 4-D
errors in the ionosphere models will produce errors in the tomographic imaging algorithms. GPS occultation TEC
IFB value estimations [Juan et al., 1997]. comes from rising or setting occultations and provides the
2.2. Data Sources horizontal ray information that is not available from angle-
limited ground-based tomography, while dual-frequency
[13] Tomography algorithms that were developed for
navigation receivers provide upward looking TEC data
imaging the ionosphere with TEC data from polar-orbiting
and allow for improved 4-D imaging of the topside iono-
satellites recorded at multiple ground-based receivers are
sphere and plasmasphere.
described in section 2.3. The satellites are in either the
Russian Cicada system or the Navy Navigation Satellite 2.3. Developing Theory
System (NNSS) that later became known as the Navy [17] The first research work in ionospheric tomography
Ionosphere Monitoring System (NIMS). These satellites was done by Austen et al. [1988]. They set up a computer
are low-Earth-orbiting (LEO) satellites with the satellite simulation study in which the electron density lay within a
altitudes approximating 1100 km. They broadcast two piecewise-constant grid oriented along a latitude-altitude
coherent frequencies at nominal values of 150 and plane of constant longitude. TEC values were computed for
400 MHz. Owing to their LEO orbit the ionosphere can a realistic satellite-receiver configuration by integration
often (but not always) be considered temporally static along specified raypaths through their electron density
during a satellite pass. model. They then reconstructed the electron density distri-
[14] Later, the development of global, three-dimensional, bution using a finite series expansion reconstruction tech-
time-varying algorithms made use of the GPS constellation nique. The algorithm choice was based on work by Censor
of satellites as data sources. The GPS constellation is in a [1983], since such methods can be used with any raypath
55° inclination, with 24 or more active satellites. For a given geometry and are often the preferred option when the line
receiver location, at any time, between 8 and 12 satellites integral data have been collected over a limited range of
are in view. The GPS satellites are at an altitude of orientations. As an initial condition for the algorithm the
20,000 km and broadcast at frequencies of 1.2 and authors used a triangular vertical density profile with a peak
1.6 GHz. With GPS, true 3-D time-evolving tomographic density close to that of the model used in the simulation.
imaging becomes possible. The simultaneous iterative reconstruction technique (SIRT)
algorithm, a development of the least squares algebraic
3 of 23
reconstruction technique (ART), was used to reconstruct the features, and an iterative algorithm was then used to add in
image. ART is described in section 2.3.2. the detail [Raymund et al., 1993]. This two-stage method,
[18] Another iterative algorithm was introduced by involving the selection of the initial condition from models
Raymund et al. [1990]. They describe the application of followed by the MART algorithm, has been extensively
the multiplicative algebraic reconstruction technique used in experimental ionospheric tomography by research-
(MART) algorithm to ionospheric tomography. They used ers at the University of Wales, Aberystwyth. The two-stage
a simulation study similar to that of Austen et al. [1988]. technique is now described in more detail here. The first
MART is an entropy-optimization algorithm that was pro- stage, selection of initial conditions from models, is first
posed by Gordon et al. [1970]. It is considered a suitable described in section 2.3.1. Then the second-stage iterative
algorithm to use in situations where the data are incomplete. algorithms ART and MART are discussed.
To produce an initial starting condition for the algorithm, 2.3.1. Background Ionosphere
TEC along a satellite-to-receiver raypath is considered to be [21] In iterative algorithms the compensation for the
divided between the pixels that the raypath intersected, with incomplete information in the measurements is made
the proportion of TEC assigned to a pixel distributed through the use of a starting point for the algorithm, known
according to the length of the raypath within that particular as a background ionosphere. In this case it is chosen to be
pixel. An electron density value was obtained by dividing the best guess that can be made before the measurements are
the assigned TEC by the intersection length. This procedure, introduced. To make this guess, an empirical model of the
a form of back projection, was combined with a Chapman ionosphere such as the international reference ionosphere
profile (described by Hargreaves [1995]) weighting to (IRI) is chosen. Then, 24 different specifications of the
assign the initial electron densities for the algorithm. The electron density are made by running the model for a range
reconstructed images indicated that good agreement with of 24 discrete times around the time of measurements. Each
the original simulations could be achieved using this electron density model represents the ionosphere at the
method. resolution of the tomographic grid, at discrete time intervals
[19] Now that the feasibility of ionospheric tomography over a variable time period, centered on the time of the
had been demonstrated, the focus turned to limitations of satellite pass. It is assumed that the matrix of electron
the new technique. Yeh and Raymund [1991] investigated density in the background ionosphere, xo, can be expressed
some of the theoretical limitations of ionospheric tomogra- as a weighted sum of the 24 models, X ~ , where
phy. They noted that the sphericity of the Earth imposes a
limitation to the lowest elevation of a raypath, and therefore ~W
xo ¼ X ~ ð2Þ
near-horizontal TEC measurements through the ionosphere
are not available. This poses a restriction on the ability to and where W ~ is a vector of the 24 weighting coefficients
reconstruct the vertical profile in electron density. In addi- that are to be found. Each column of matrix X~ represents a
tion, refraction of the radio wave at low elevations limits the particular model. In each row an element represents an
accuracy of the TEC values obtained from such rays. A electron density in a specific ionospheric locality, defined
further limitation to the accuracy of the images occurs by a pixel of uniform electron density. Let H~ be a matrix
because of the discrete locations of the receivers. The describing the geometry of the raypaths with respect to the
temporal variation of the ionosphere during the data collec- grid, containing the path lengths of each ray in each pixel.
tion period was also noted, and for future work the imple- The TEC values along the raypaths through the background
mentation of a time-dependent algorithm was suggested. ionosphere yo are given by
This idea could be used with experimental measurements ~ o¼H
yo ¼ Hx ~X~ W:
~ ð3Þ
during ionospheric conditions where the assumption of an
unchanging ionosphere for the duration of the satellite pass [22] The matrix product H ~X~ can be represented by a
(10 – 20 min) breaks down. This has been adopted for later ~
matrix M, which contains calculated TEC values along
work on 3-D imaging with GPS data [Mitchell and Spencer, satellite-to-receiver raypaths through each of the model
2003]. ionospheres. The matrix M ~ consists of 24 column vectors,
[20] Raymund et al. [1994] ran simulation studies using one for each model. The number of column vectors was
the mathematical technique for least squares fitting known chosen initially to have a model from each hour in the day.
as singular value decomposition (SVD) to produce a tomo- Each row is for a satellite to receiver raypath. The measured
graphic image from a linear combination of model iono- TEC values, y, are assumed to be equal to a linear combi-
spheres. A number of ionospheres given by the fully nation of these model TEC values, so
analytic ionospheric model were considered, and SVD ~ W:
~
y¼M ð4Þ
was used to select a linear combination of these model
ionospheres. However, small-scale structures are not repre- ~ was then computed
[23] The least squares solution to W
sented in these ionospheric models, and hence no combi- using SVD to find the pseudoinverse, and
nation of the models could reproduce detailed images. A 1
development of the method was necessary where the ~ ¼ M
W ~ y: ð5Þ
selected combination of models was used as the initial
condition (background ionosphere) showing the large-scale
4 of 23
[24] The background ionosphere is then formed from will be maximized. The maximum entropy approach to
equation (2). Once a background ionosphere is formed from image reconstruction yields the image with the lowest
the linear combination of models, an iterative algorithm is information content that is consistent with the available
run to add the finer-scale features to the image. The iterative measurements [Menerbo, 1979]. This reduces the tendency
algorithms are described in sections 2.3.2 and 2.3.3. In the to introduce artifacts into the image region. Censor [1983,
description of iterative algorithms the notation changes to p. 414] notes that B. R. Frieden, writing about image
refer to individual pixels rather then matrices of the entire restoration, stated ‘‘the most likely object scene implied by
pixel grid. This is because iterative algorithms work recur- given image data is found to obey the principle of maximum
sively on individual pixels. entropy.’’ Censor suggests that this could also be applied to
2.3.2. Algebraic Reconstruction Technique image reconstruction.
[25] The ART algorithm was first published by Gordon et [28] The initial application of this algorithm to experi-
al. [1970]. Rather then trying to handle a given system of mental ionospheric tomography was by Raymund et al.
equations as a whole, the ART works by an iterative process [1990], although in an experimental situation with noise
and has been shown to converge to the least squares on the data the behavior of the MART algorithm is not well
solution. ART has been published with several modifica- understood. In particular, its response to inconsistent sets of
tions. The one described here is the ART algorithm with equations is unknown [Censor, 1983]. Nevertheless, the
relaxation from Censor [1983]. An initial guess, x0j , is MART algorithm has been used successfully for the recon-
obtained from the background ionosphere and for the struction of tomographic images of the ionosphere by
k + 1th iteration several authors [Kersley et al., 1993; Heaton et al., 1995;
Mitchell et al., 1995; Pryse et al., 1995; Vasicek and
P
n Kronschnabl, 1995].
yi Dim xkm
m¼1 [29] The MART algorithm, as in the case of ART,
xkþ1 ¼ xkj þ lk Dij ; ð6Þ
j P
n
requires an initial starting value for each pixel, x0j for
Dim Dim
m¼1 j = 1, . . ., n, and in the case of ionospheric tomography this
is the electron density in the background ionosphere. The
where lk, the relaxation parameter, is a number or series of algorithm then updates the electron density in the image
numbers used to control the convergence of the algorithm, region according to the formula
Dij is the length element of raypath i through pixel j, im is
the loop over m from 1 n for each i, and n is the total 0 1max
lk Dij
Dj
number of pixels in the grid. The current iterate, xkj , is B y C
refined to a new value, xk+1 B i C
j , by considering a single ray, i, xkþ1
j ¼ xkj BPn C : ð8Þ
@ kA
and changing the electron density value of the pixels, j, Dij xj
j¼1
intersected by the ray. The discrepancy between the
measured TEC and the TEC calculated through the current
image is then redistributed among the pixels along the ray in Thus for each TEC measurement a multiplicative correction
proportion to the length of intersection with each pixel Dij. is made to the electron density of every pixel. The
The algorithm then cycles through all raypaths several times relaxation parameter, lk, which controls the convergence
until convergence is achieved. of the algorithm, is bounded to lie between 0 and 1.
[26] The ART algorithm has been successfully imple- 2.4. Other Early Methods and Image Resolution
mented in experimental ionospheric tomography by [30] A number of noniterative algorithms have been
Andreeva [1990] and Kunitsyn et al. [1994a, 1994b, applied to ionospheric tomography. These are mathemati-
1995]. The SIRT algorithm is a development of the ART cally similar to the creation of the background ionosphere
algorithm, where the first update to the density is made after described previously, with some modifications to allow
all of the raypaths have been considered once, i.e., after one smaller features to be imaged.
iteration. The SIRT algorithm has been used in ionospheric [31] Na and Lee [1990] applied an orthogonal decompo-
tomography by Pryse et al. [1993]. sition technique in which they generated a set of Fourier
2.3.3. Multiplicative Algebraic Reconstruction orthonormal basis functions from a priori information. Each
Technique was derived from a different frequency component and
[27] The MART algorithm was originally proposed as a orientation. The tomographic image could then be formed
reconstruction algorithm by Gordon et al. [1970] and was from a combination of these basis functions. A further
later proven to converge to the maximum entropy solution publication [Sutton and Na, 1994] tested this approach to
by Lent [1997]. The maximum entropy solution means that the reconstruction using simulated data and produced some
the function promising results. However, Sutton and Na noted the
significant reliance of the vertical profile in the reconstruc-
1 x
f ð xÞ ¼ P
n ln P
n ð7Þ tion on a priori information. This problem was not only
lnðn2 Þ xj xj found with their method; in fact, it is a consequence of the
j¼1 j¼1
lack of complete data coverage [e.g., Cornely, 2003].
5 of 23
[32] Fremouw et al. [1992] used a set of vertical ortho- over Scandinavia, by Mitchell et al. [1997c]. They used a
normal vectors, created from ionospheric models, to image series of Chapman profiles to describe the vertical electron
the vertical profile and a power law spectrum to select the density profile and found that while the peak height could
horizontal structures from a Fourier basis. This approach often be determined to within about 30 km, the scale height
has proven to be highly successful and has been used and peak density relationship was ambiguous; a larger scale
subsequently in various modified forms by a number of height and lower peak density could often fit the original
researchers. A new algorithm, described by Raymund et al. TEC data as closely as a smaller scale height and higher
[1994], was similar to the technique used by Raymund peak density.
[1994], but the solution was no longer constrained to lie
in the subspace of ionospheric models, and it satisfied the 3. THEORY: THREE-DIMENSIONAL IMAGING
TEC values completely. However, the solution was not
unique, and forcing the solution to match an ionospheric [36] Hajj et al. [1994] first discussed the potential of
model in a least squares sense finally resolved the ambiguity using ground- and spaced-based GPS observations to image
in the solution. However, this initially promising algorithm the ionosphere. They suggested the satellite-to-satellite
never showed a significant improvement over other com- transmission of GPS to LEO dual-frequency L band signals
putationally less expensive methods. to provide the vertical structure, with ground-based
[33] Fougere [1995] pursued a novel approach in which a receivers recording from NNSS satellites providing the
range of Chapman profiles were used to form an orthonor- horizontal structure. They indicated that the GPS-LEO
mal set and were incorporated into the reconstruction in the measurements would be of particular help in defining the
form of horizontal rays. The solution to the modeled TEC E layer and that these additional measurements could
was then computed using a maximum entropy algorithm. provide information for either two- or three-dimensional
The method was not tested on any model representative of reconstructions. In a subsequent publication, Leitinger et al.
the ionospheric electron density, but several test simulations [1997] outlined the possibility of using data from the first
showed that excellent reconstructions of Gaussian enhance- radio-occultation satellite GPS/MET to complement
ments could be achieved, even with the geometrical con- ground-based NNSS tomography.
straints imposed by ionospheric tomography. [37] Rius et al. [1997] showed the first experimental
[34] Kuklinski [1997] set up a three-dimensional image results that used GPS data to image the ionosphere.
grid in longitude, latitude, and altitude providing voxels (a Although their vertical grid consisted of just four height
voxel is a 3-D pixel and can be in any coordinate frame) of voxels, this was a key step in the transition from 2-D
size 0.5° 0.5° 45 km. TEC measurements, recorded mapping [Wilson et al., 1995; Mannucci et al., 1999] and
during the Russian-American tomography experiment (see intermediate multiple shells [Juan et al., 1997] into 3-D
section 4), were then used to reconstruct a tomographic imaging. They used a Kalman filter approach to accommo-
image in the plane defined by the receiver chain and the date temporal changes in the images. Kalman filters have
satellite orbit. Kuklinski [1997] also presented simulation been used for numerous engineering applications where the
results, employing a network of some 50 receivers spread state of a dynamic system is to be estimated from incom-
over the United States, indicating the possibility of quasi- plete measurement data. The approach exploits knowledge
three-dimensional ionospheric images from a single NNSS of the evolution of the system to reduce the noise or to
satellite pass. In 1997, Mitchell et al. [1997a] showed interpolate missing data. In data assimilation [see Daley,
experimental results that realized a simplified version of 1991] the Kalman filter uses the map and error covariances
this idea, where two chains of NNSS receivers were from the previous time step as the background model and
deployed in Scandinavia. error covariances for the present time step. Howe et al.
[35] Na and Sutton [1994] investigated the resolution [1998] applied the Kalman filter approach to ionospheric
limits of ionospheric tomography. They separated their tomography. This combined and extended ideas from the
analysis into projections and considered the case of the vertical-slice ionospheric tomography work of Fremouw et
3-dB resolution of two structures within the ionosphere. al. [1992] and the 2-D horizontal mapping of GPS TEC
They found that the resolution improved as the density of [Wilson et al., 1995] into a full three-dimensional algorithm.
receivers increased but that a limit of around 50 km was An outline of the general approaches to three-dimensional
found. This is apparently consistent with experiments since imaging of the ionosphere is given in section 3.1.
it is smaller than the smallest verified features imaged in the
ionosphere using tomography [e.g., Mitchell, 1997b]. Sub- 3.1. Three-Dimensional Ionospheric Imaging
sequent work on the resolving capabilities of different [38] Imaging in three dimensions is essentially the same
geometries [Na and Sutton, 1994; Na et al., 1995] has as in two dimensions, but some complications have to be
allowed the theoretical resolutions of different ionospheric overcome because of the sparseness of the data coverage.
tomography experiments to be assessed. Heaton et al. For that reason, iterative pixel-based methods (such as ART
[1995] tested the incorporation of scaled ionograms into or MART) are not chosen, since they do not couple the
the imaging, and this resulted in some improvement in the changes applied to one voxel across to other voxels that
vertical profiles. The determination of vertical profiles was may have no intersections with measurements. Neverthe-
investigated in some detail, in the experimental environment less, the electron density field is discrete for at least part of
6 of 23
each imaging method. In some cases [e.g., Howe et al., physical models. These functions will then allow the recon-
1998] the field can be changed into functional form for part struction of the ionosphere. This empirical orthogonal
of the inversion procedure. In general, a three-dimensional function (EOF) approach was taken by Fremouw et al.
grid of voxels is defined, each bounded in latitude, longi- [1992]; the extension into 3-D GPS imaging with a Kalman
tude, and altitude. Several authors have chosen Sun-fixed filter was first shown by Howe et al. [1998]. A time-
coordinates [e.g., Juan et al., 1997] exploiting the solar dependent extension of this technique without a Kalman
EUV dominance of the ionization that can help to fill data filter, where the electron density is allowed to vary linearly
gaps. Others have preferred a general geographic frame in each voxel over a period of an hour, was shown by
[e.g., Mitchell, 2002] since they often aim to image the Mitchell and Spencer [2003].
disturbed ionosphere, which does not follow simple EUV- [43] A set of 3-D orthonormal basis functions is defined
dominated changes. in accordance with the geophysical conditions and data
[39] As for the 2-D case the length of each straight-line coverage. For example, horizontally, these could be pixels,
element of a satellite-to-receiver signal propagation path spherical harmonics, or wavelets. Vertically could be any
though each intersected voxel is computed from knowledge orthogonal basis set formed from empirical or mathematical
of the satellite and receiver locations. The unknown electron models. The time, location, and geophysical conditions can
concentration is defined to be constant within each voxel be used to tailor specific EOFs [e.g., Materassi and
and contained in the column vector xa. This can be Mitchell, 2005a, 2005b] and truncation of the spherical
expressed as harmonics to minimize aliasing in data-sparse regions
[Mitchell and Spencer, 2003]. A mapping matrix, X, ~ is
~ a;
y ¼ Hx ð9Þ used to transform the problem. This results in a situation
where the unknowns are coefficients of 3-D orthonormal
where the matrix H ~ transforms the electron density to the basis functions. The combination of these will give the final
form and location of the observations and y are the observed image of electron concentration. The horizontal functions
TECs. provide a flexible basis to determine the horizontal distri-
[40] There are several approaches to using the observed bution of ionization, which is generally well specified by the
TECs. The simplest conceptually is to calibrate the TEC data; the EOFs form a constraint to the underdetermined
observations before using them into the inversion. For GPS vertical profile.
data this can be achieved by using both the differential [44] This can now be expressed mathematically as
phase and the differential delay data and correcting for the
satellite and receiver interfrequency biases. While the orig- ~X
y¼H ~ W;
~ ð10Þ
inal estimates for the accuracy of interfrequency biases were
a few TEC units, recent work by Ciraolo et al. [2007] where the matrix X~ contains the basis functions. H ~X~ now
shows that much larger errors can be found. Consequently, represents the set of modeled TEC that are formed by
this approach should be undertaken with care, and consid- integration through the set of 3-D basis functions. Applying
eration for possible errors must be given, especially a matrix inversion, for example, using singular value
when using such data in assimilation algorithms where the decomposition, the inversion can be performed such that
estimate of measurement error is critical to the solution.
1
[41] An alternate approach is to use the relative changes ~ ¼ H
W ~X~ y; ð11Þ
of TEC along continuous satellite-receiver paths. If this
route is taken, then the inversion algorithm must calibrate and the solution to the inverse problem is then given by
the unknown constants, as was the case for the 2-D applying the matrix of ‘‘weights’’ back to the 3-D basis
algorithms of Andreeva [1990]. In this case the problem is functions,
analogous to that of using NNSS data: Differential phase
arcs are used, and the inversion uses relative differential ~ X:
xa ¼ W ~ ð12Þ
phase observations. Thus appropriate lines of, first, the
geometry matrix, H, ~ then the measurement matrix, y, are [45] The algorithm can be extended into a time-dependent
differenced, such that measurements along continuous inversion by incorporating a priori information about the
satellite-receiver arcs are taken relative to a certain reference evolution of the electron concentration during a specified
measurement within that arc. period of time. If the change electron concentration within a
[42] Equation (9) cannot be used directly to find the voxel with time is linear, then it is possible to write the same
electron density xa because the measurement geometry is system of equations to solve for the change in the relative
not ideal. The inversion system is underdetermined. Many contributions of each basis function [Mitchell and Spencer,
approaches exist to compensate for the missing information 2003].
to allow the reconstruction of a spatially continuous electron
density field. Mathematically based constraints can be 3.2. Data Assimilation Approaches
applied to obtain a physically reasonable solution. One [46] Another approach to three-dimensional, semicontin-
approach to applying the constraints is to form a basis set uous imaging of plasma density is data assimilation, as
of orthogonal functions that are formed from empirical or practiced in oceanography and meteorology. The similarity
7 of 23
between meteorology and space weather has led a number (IDA3D) algorithm. IDA3D produces an objective analysis
of researchers to discuss the idea of applying meteorological ‘‘map’’ of the 3-D electron density field that is projected
forecasting techniques to ionospheric forecasting. Daley forward in time through a Gauss-Markov Kalman filter. It
[1991] provides a comprehensive treatment of the theoret- represents an implementation of stages 2 and 4. Utah State
ical foundations of data assimilation for numerical weather University (USU) has developed the Global Assimilation of
forecasting. While the objective of full data assimilation is Ionospheric Measurements (GAIM) algorithm, which is a
to operationally specify and forecast the state of the iono- first-principles physics-based Kalman filter approach that
sphere, such methods can also be used to produce 4-D incorporates stages 2, 3, and 4. The Global Assimilative
images of ionospheric electron density, which is the main Ionospheric Model (GAIM) was developed by the Univer-
interest in this review. Below we present a brief discussion sity of Southern California and Jet Propulsion Laboratory
of ionospheric data assimilation and provide examples of (USC/JPL) and is an algorithm that also incorporates stages
the main techniques that have been developed for iono- 2, 3, and 4 but in a different manner. Both the USU and
spheric data assimilation. The goal here is present iono- USC/JPL GAIM models have Gauss-Markov implementa-
spheric data assimilation within the context of 4-D imaging. tions similar to IDA3D. The Gauss-Markov Kalman filter
Thus a full derivation and exposition of data assimilation approach is outlined below. IDA3D is used to demonstrate
methods is beyond the scope of this review. The interested the mathematical form of the solution. The two GAIM
reader can obtain a more complete discussion from the Gauss-Markov Kalman filters have similar mathematical
many references provided. forms.
[47] True data assimilation is fundamentally a model [50] IDA3D [Bust et al., 2000, 2004] organizes the data
specification and prediction technique that uses data to into spatial maps of electron density via an objective
improve the fidelity of the model. For the ionosphere the analysis technique. These spatial maps are projected for-
relationship between raw input data and resulting images of ward in time through a Gauss-Markov Kalman filter where
electron density is complex. It is based on first-principles they are used to initialize the next analysis. However, while
physics models that typically consist of multiple species of the background model can be a first-principles ionospheric
ions and neutrals, each with its own continuity, momentum, model, there is no feedback of the spatial maps into the first-
and energy equations and all coupled to Maxwell’s equa- principle model, and no correction or update to the model is
tions. The resulting electron density specification is depen- made. IDA3D is an objective analysis algorithm (step 2 in
dent upon the numerical model itself; that is, the resulting 4- the data assimilation cycle above), based upon three-
D electron density field is influenced by both the underlying dimensional variational (3DVAR) data assimilation [Daley
numerical model and the inputs that drive it. In contrast, and Barker, 2001; Daley, 1991]. The basic idea is to create
with 4-D imaging the relationship (or forward model) a procedure for combining a model output with actual
between the state variable to be imaged (ionospheric elec- measurement data. Obviously, sparsely measured quantities
tron density) and the observations is simple. For example, in could create discontinuities if they were used by directly
tomographic imaging the forward model is a path integral placing them into a model output. In addition, instruments
over the quantity to be imaged. Thus in tomography, there will measure a quantity over a given region of space, and
may be no physics model, though regularization through this measurement in itself will have a region over which it
mathematical techniques is often used to provide necessary can be considered. The 3DVAR is a statistical minimiza-
a priori information along with networks of data to create tion procedure that deals with these issues. It takes into
the image. account data, data error covariances, a background model,
[48] In data assimilation, there are four stages: (1) Collect and the background model error covariances.
the data to be assimilated and do quality control on the data. [51] Following the previous notation in this paper where
(2) Organize the data into 3-D (or 4-D) maps of the state appropriate, y is the vector of observations, R ~ is the error
variable being assimilated through an ‘‘objective analysis’’ covariance matrix for the observations, xf is the forecast (or
technique, using the current best model forecast of the state prediction) model, and P ~ f is the error covariance matrix for
variable. (3) Reinitialize the state variables in the first- the forecast model. For a given background forecast (or
principle model being used and (if desired) estimate the prediction) the method computes xa, which is the analysis
inputs to the model from the objective analysis. (4) Using electron density at a given time, and the analyzed error
the newly initialized model, forecast forward to the next covariance P ~ a. The 3DVAR (or Kalman gain) solution is
data assimilation time. given as
[49] Stage 1 is not discussed further in this review since it
is an essential step for all data analysis. The output of stage h i1
2 at a given time includes a 3-D estimation or image of xa ¼ xf þ P ~T R
~f H ~ þH
~P ~T
~f H ~ f
y Hx ð13Þ
electron density. Thus any data assimilation algorithm that
includes stage 2 results can be used for 4-D imaging. Stages
2 through 4 have been implemented in three separate
algorithms discussed in detail below. Applied Research
h i1
Laboratories, University of Texas at Austin has developed ~a ¼ P
P ~f P ~T R
~f H ~ þH
~P ~T H
~f H ~P
~f ; ð14Þ
the ionospheric data assimilation three-dimensional
8 of 23
where superscript T is the transpose of the matrix. The of the ionosphere. There are several versions of GAIM
analyzed density and error covariance are forecast to the under development at USU. One version, a Gauss-Markov
next time step by application of the Gauss-Markov Kalman Kalman filter [Schunk et al., 2004] is identical to IDA3D in
filter technique [Gelb, 1974]: the mathematical derivation and form of the estimation
solution. The Gauss-Markov version of GAIM uses the
Dt
xf ðtnþ1 Þ ¼ e t ½xa ðtn Þ xb ðtn Þ þ xb ðtnþ1 Þ ð15Þ Ionosphere Forecast Model [Schunk et al., 1997] as the
background model specification. The Gauss-Markov ver-
sion of GAIM is currently used operationally at the Air
~ f ðtnþ1 Þ ¼ 1 e2Dt
P t P~ b ðtn Þ þ e2Dt
t P~ a ðtn Þ; ð16Þ Force Weather Agency. The USU Gauss-Markov version of
GAIM has been validated against independent measure-
where tn is time associated with the analysis and model ments TEC and peak densities from ionosondes [Thompson
estimates and t is the decorrelation time. The background et al., 2006]. A second version of GAIM being developed
model state (empirical or first principle model) is given by by USU is a physics-based, reduced-state Kalman filter
xb, and its error covariance is given by P ~ b. assimilation algorithm [Scherliess et al., 2004]. There are
[52] The value of the resulting specifications is dependent two basic ideas. First, computing the full analyzed error
upon the amount, the distribution, and the quality of the covariance (number of grid points squared) is a computa-
observations. Any observation that can be predicted from tionally prohibitive task. So, a reduced-state error covari-
the background model can be ingested. Typically, ingestion ance is calculated. In practice, the number of grid points in
is limited to data types that can be predicted without major the global ionosphere is reduced to 1% of the original full
assumptions or complex calculations. To solve equations (13) grid, reducing the computational requirements by a factor of
and (14), IDA3D needs certain inputs, including back- 10,000. Second, the evolution in time of the model state and
ground climatology with a model grid, electron density error covariance is given by
specification, correlations, and a set of observations (direct
measurements or radiances from remote sensing) that is ~ þh
xf ¼ Mx ð17Þ
easily related to the electron density. The model grid is
an input that is chosen for its compatibility with the
specific scientific investigation that is planned. Typically, ~f ¼ M
P ~P ~ T þ Q:
~M ~ ð18Þ
the IRI [Bilitza, 2001], or a first-principle model such as
the thermosphere-ionosphere-mesosphere-electrodynamic [55] The operator M ~ is the full nonlinear operator that
general circulation model [Roble and Ridley, 1994], is represents the transition from one time step to the next of
used as the background model. The impact of the back- the first-principle physics model. The error on the model
ground model is significantly reduced by application of the prediction is given by h, and the model error covariance is
Gauss-Markov Kalman filter technique (equations (15) and given by Q. ~ To solve equation (18) exactly requires a
(16)). The background model error correlations are treated computationally prohibitive number of full nonlinear for-
as inputs that are independent of the background model. At ward model runs. To make the problem tractable, the full
present the correlations are treated as correlation lengths in physics model is linearized about some reference state.
latitude, longitude, and altitude. The background model Then M ~ becomes a linear matrix on the reduced-state grid.
correlations decrease exponentially as the ratio of the This reduced-state approach was tested under simulation
distance between the model points and the correlation [Scherliess et al., 2004], though for the simulation only the
length. The horizontal and vertical distances are treated predictions of electron density obtained from the first-
separately. In addition, a correlation time, t, is given as principle model (with driver adjustment) were used. The
input. These correlations allow the data to impact a larger full analyzed electron density and error covariance (such as
region of the specification than just of the grid points obtained from IDA3D, equations (12) and (13)) was not
affected by the observations and allow past observations used in the simulation. The results of the simulation are
to impact the present specification. Using these techniques, presented in Figure 1.
IDA3D is able to ingest any available observation of [56] The Global Assimilative Ionospheric Model (GAIM)
electron density or electron content. has been under development by the University of Southern
[53] IDA3D is primarily used as a 3-D electron density California and the Jet Propulsion Laboratory [Wang et al.,
imaging algorithm. It is a mature algorithm. It has been 2004; Hajj et al., 2004; Pi et al., 2003]. Both a Kalman
validated [Coker et al., 2001; Watermann et al., 2002] and filter and a four-dimensional variational (4DVAR) tech-
applied to several scientific investigations [Bust et al., 2000; nique are included in the USC/JPL GAIM algorithms. The
Bust and Crowley, 2007; Yin et al., 2006; Garner et al., Kalman filter is very similar to the USU approach described
2006; Crowley et al., 2006]. already and will not be discussed further. Here the focus is
[54] For the last several years, Utah State University has on the 4DVAR approach to data assimilation. In 4DVAR the
been developing the Global Assimilation of Ionospheric idea is to use measurements over some range of times to
Measurements (GAIM) model [Schunk et al., 2004, Schunk estimate both the initial state of the ionosphere and the input
et al., 2005a, 2005b; Scherliess et al., 2004]. GAIM is a drivers to a first-principle ionospheric model. The USC/JPL
data assimilation model that specifies and forecasts the state physics model is derived from the University of Sheffield
9 of 23
Figure 1. USU GAIM Image [from Scherliess et al., 2004].
model [Bailey et al., 1997]. However, it considers only a with k referring to time tk. The notation is the same as
single ion species O+. It solves the ion continuity and described in (13) – (16), with the exception that x0 now
momentum equations, but it ignores the time variation and refers to the electron density distribution at time 0. The
inertial terms in the momentum equation. The ion momentum input drivers are represented through the vector of unknown
equation is further broken into a field-parallel and field- parameters a. The parameters g n and g F are regularization
perpendicular component. The velocity component perpen- factors. The a priori designation refers to the initial guess or
dicular to the magnetic field is considered to be due entirely background values for the O+ density and the input drivers.
to E B and is an input driver. The parallel component of To minimize J, a nonlinear estimator such as the quasi-
velocity also has input drivers due to neutral winds and Newton method is used. However, a major challenge is the
electron and ion temperatures. Thus the only state variable calculation of the gradient of J for every parameter to be
solved for is the O+ density; the rest are input drivers to the solved for. The solution is to construct an adjoint equation
system. These driver inputs must be obtained from empir- and propagate the solution backward in time from the
ical models including the following: thermospheric densities measurement time to the initial time t0. The math is not
from the Mass Spectrometer Incoherent Scatter model repeated here. The interested reader can refer to Pi et al.
[Hedin, 1991], neutral winds from the horizontal wind [2003] and references therein. In principle, self-consistent
model [Hedin et al., 1996], solar EUV as described by estimation of the initial state and driving parameters will
Tobiska [1991], electric fields [e.g., Fejer, 1991; Heppner provide an accurate estimation of the O+. However, to date,
and Maynard, 1987; Scherliess and Fejer, 1999], and only simulation experiments have been reported in the
electron energy precipitation flux [Fuller-Rowell and Evans, literature. Simulation in this case means setting up a
1987]. A 4DVAR estimation algorithm for the O+ model computer model to test the feasibility of new technique.
described above was implemented by the USC/JPL group. One such experiment was conducted in 2003 [Pi et al.,
Following Pi et al. [2003], the cost function to be mini- 2003]. In the 2003 experiment, only vertical drift at the
mized is given as geomagnetic equator was simulated and estimated, while all
the other inputs were held at their empirical values. The
X
N vertical drift was parameterized by nine coefficients at
J ðx0 ; aÞ ¼ k yk Hk xk k þ g n k x0 xapriori k þ g F different local times. For such a small set of unknowns
k the simulated results were promising (Figure 2). This was
k a aapriori k; ð19Þ then followed by another simulation experiment in 2004 [Pi
et al., 2004] where both the vertical drift and neutral winds
10 of 23
Figure 2. USC/JPL GAIM Image [from Pi et al., 2003].
were solved for. Actual experiments that validate against These authors, from the Moscow State University, used
independent measurements of plasma density or driving TEC data collected at three receivers located at Murmansk
parameters have not yet been reported in the literature, (68.6°N, 31.8°E), Kem (65.0°N, 34.6°E), and Moscow
although Mandrake et al. [2005] show promising compar- (55.7°N, 37.6°E). Their images were reconstructed using
isons with other sources of TEC for the Kalman filter an iterative algorithm that operated on relative TEC data.
version of GAIM. Pryse and Kersley [1992] from the University of Wales,
Aberystwyth, used TEC measurements collected by
4. EARLY RESULTS AND EXPERIMENTAL receivers in Scandinavia at Kiruna (68.7°N, 20.4°E) and
VALIDATION OF IONOSPHERIC IMAGING Oulu (65.0°N, 25.5°E) to reconstruct the ionospheric elec-
tron density for a satellite pass recorded on 4 September
[57] Experiments in ionospheric tomography have been 1986. This was the first tomographic image of the iono-
conducted since the early 1990s. The receiver configura- sphere to be verified by another instrument. The European
tions depend on the choice of satellite system: In the case of Incoherent Scatter (EISCAT) radar made coordinated meas-
NIMS the satellites are in geographic polar orbit, and hence urements of electron density in the geographic meridian of
the receivers are placed close to lines of constant geographic Tromsø. The tomographic image showed a northward
longitude, whereas the for the Cicada satellites the config- gradient in electron density in broad agreement with that
uration is for the corresponding geomagnetic coordinates. measured by the radar. A special tomography campaign was
Both satellite systems are equally useful for ionospheric carried out in the United Kingdom during December 1990
imaging from geometrical standpoints. In either case it is [Pryse et al., 1993] with receivers placed along a meridional
necessary to find suitable receiver sites on the ground to chain at four sites: Lerwick (60.2°N, 1.1°W), Aberdeen
provide an approximate longitudinal line for the imaging. It (57.2°N, 2.2°W), Hawick (55.4°N, 2.8°W), and Aberyst-
should be noted, though, that an ideal latitude-altitude plane wyth (52.5°N, 4.1°W). Good agreement was found between
can never be achieved in practice because of the Earth foF2 measurements from the ionosonde and the corresponding
rotation under the satellite pass and because the approx- values obtained from the tomographic images. Tomographic
imations must be made to project the measurements into a images from eight consecutive satellite passes showed the
2-D image plane regardless of where that plane is. expected behavior of the midlatitude trough, demonstrating the
[58] Early experimental work was conducted through potential of tomography to image large-scale ionospheric
short-duration campaigns with manned receivers. By the structures.
mid to late 1990s these campaigns were largely replaced by [60] Kersley et al. [1993] presented results from the first
permanent deployments of networked receivers. Early ex- EISCAT radar observing program specifically designed to
perimental results, particularly those that included valida- provide an independent electron density comparison with
tion, are described below. tomographic images. The EISCAT radar was run in a
[59] The first experimental result showing a tomographic meridional scanning mode, and additional information
image of the ionosphere was published by Andreeva [1990]. about the height of the layer peak and the electron density
was available from ionosondes in Scandinavia. As the IRI
11 of 23
model allowed the direct incorporation of this ionosonde which was shown to compare reasonably well with an
data, it was used as a background ionosphere, and Kersley et electron density profile from the EISCAT radar. However,
al. [1993] concluded that the incorporation of ionosonde the ambiguity in the vertical profile was not solved, and the
data resulted in a significant improvement in the tomo- selection of the peak height remained an outstanding
graphic images. problem.
[61] Kunitsyn et al. [1994a] presented an overview of the [64] Several other experimental tomographic results were
various tomographic methods employed by Russian iono- published in a special issue of Annales Geophysicae.
spheric tomographers and presented results from their Kunitake et al. [1995] used a modified version of singular
various tomographic techniques, although no independent value decomposition to reconstruct the electron density over
measurements were available to verify their images. A Japan. This extension of SVD involves the truncation of the
further publication by Kunitsyn et al. [1994b] contrasted smaller singular values to remove small-scale perturbations
the merits of phase difference as opposed to calibrated and noise from the data. It also allows the direct incorpo-
phase values as input for the tomographic algorithm. Phase ration of other constraints into the algorithm and requires no
difference refers to the process of taking the difference background ionosphere. Other ionospheric information was
between the differential phases measured along adjacent provided by ionosondes and oblique sounders, which
raypaths. They described the problem of isolating the initial allowed verification of the tomographic images. The results
phase offset for each station. Their results suggested that the were encouraging, with good agreement found between the
determination of the initial phase offset, which enables the peak electron density in tomographic images and ionosonde
subsequent use of absolute TEC, was particularly difficult in foF2 values.
the case of strong horizontal gradients in electron density. [65] Mitchell et al. [1995] presented tomographic images
This confirmed the ideas of Leitinger et al. [1975]. They of field-aligned irregularities from TEC data collected at the
also concluded that the phase difference technique should chain of four receivers (Tromsø, Kiruna, Lycksele, and
be more sensitive to the reconstruction of localized electron Uppsala) on mainland Scandinavia and one additional
density disturbances in the ionosphere than methods using station at Ny Ålesund (78.9°N, 12.0°E). To reconstruct
calibrated TEC values. the images, the MART algorithm was used with a back-
[62] In late 1993 a joint Russian-American tomography ground ionosphere obtained from the IRI-90 model. The
experiment was held in the eastern United States and reconstructions showed a latitudinally narrow trough, with a
Canada [Foster et al., 1994]. Receivers were placed at four boundary blob on its poleward wall of latitudinal extent
sites located approximately along the 72°W meridian from around 50 km. The EISCAT radar provided verification of
Rhode Island, United States (41.4°N), to Roberval, Quebec, the electron density distribution in the images, confirming
Canada (48.8°N). Radio transmissions were monitored from the accuracy of the tomographic method (Figure 3). The
both the NNSS and the Russian Cicada satellites. The phase tomographic images also revealed the signature of auroral E
data, collected by both Russian and American receivers, in a tomographic image showing enhanced electron densi-
were analyzed independently by the respective research ties beneath the poleward wall of the trough. Further details
groups. The Millstone Hill incoherent scatter radar about experimental results from NIMS imaging can be
(42.6°N, 288.5°E) was operated in an elevation scan mode found in a review paper by Bernhardt et al. [1998].
in the plane of the receivers, and the resulting electron
density measurements were used as a comparison to the 5. SCIENTIFIC RESULTS
tomographic images. The Russian tomographic images were
produced using a horizontally stratified ionosphere as a 5.1. High Latitudes
background, whereas the American group used the method [66] The high-latitude ionosphere has been studied rather
described by Raymund et al. [1994] in conjunction with the extensively using ionospheric imaging. Tomography is a
parameterized ionospheric model [Daniell, 1991] to create useful technique for this remote region where sites for
the background ionosphere. The results contained measure- instrumentation are sparse and the combination of data from
ments recorded during a geomagnetic storm. In one exam- different sources can be useful. In addition, the physics is
ple, depleted ionization between 44° and 66°N was imaged complex, and scientific studies can benefit from the inter-
in the Cicada reconstruction and in the radar observations. pretation of results from multiple instruments. Ionospheric
[63] Markkanen et al. [1995] described a new reconstruc- imaging has proved a useful complement to other observa-
tion method based on a Bayesian approach. They presented tions in the auroral regions and the polar cap. Two techni-
simulated results in which several different peak heights ques that have been used with high-latitude ionospheric
were put into the reconstruction process as a priori infor- imaging are outlined below.
mation, and the effect of this information on the resulting [67] Incoherent scatter radars are large, sophisticated
tomographic images was investigated. The application of instruments for investigating processes in the ionosphere
this new method to an experimental situation was then [Beynon and Williams, 1978]. They operate by line-of-site
tested using data from the Cicada satellites. Measurements scatter from the plasma and thus can view an ionospheric
were recorded at a chain of four satellite receivers in region spanning several hundred kilometers from their
Scandinavia from Tromsø in Norway to Nurmijävi in Fin- location. Such facilities exist at several high-latitude loca-
land. Two tomographic images were produced, one of tions and have been used for both verification and for
12 of 23
Figure 3. Tomographic image and EISCAT verification. Reprinted from Mitchell et al. [1995], with
permission.
complementary physical studies with ionospheric imaging. [68] The Super Dual Auroral Radar Network is a network
High-latitude ISR facilities include Sondrestrom in Green- of HF backscatter radars that is used to form velocity maps
land, EISCAT in northern Scandinavia, the EISCAT Sval- of the plasma convection in the polar regions [Greenwald et
bard Radar (ESR) on Svalbard, and Irkutsk in Russia. A al., 1995]. This information is a useful complement to
new redeployable facility, Advanced Modular Incoherent ionospheric imaging because it can help to identify the
Scatter Radar, is currently active at Fairbanks, Alaska. ISRs large-scale movement of the plasma. In fact, it can also
have been very important in ionospheric physics because provide assistance in ‘‘static’’ imaging as it can identify
they provide a wealth of information about the ionosphere cases where the plasma is convecting at high velocity and
such as density, temperature, and velocity of the medium. may induce image blurring.
This can be important for the interpretation of ionospheric [69] The first attempts to use tomographic imaging to
images [Meggs et al., 2005] by helping to distinguish investigate the morphology of the ionosphere [Afraimovich
between processes that change the bulk electron density et al., 1992] were made soon after the technique first
within a region, such as precipitation and convection. appeared. Since many of the early experimental campaigns
13 of 23
Figure 4. Map of NmF2 revealing the formation of polar cap patches [from Bust and Crowley, 2007].
were located at high latitudes, both auroral and polar cap ward. Bust and Crowley [2007] combined a trajectory
processes have been investigated, particularly in the Euro- analysis with a model of the polar cap convection with
pean sector [Mitchell et al., 1995; Walker et al., 1996; Pryse their imaging to reveal patches being formed from solar
et al., 1997; Walker et al., 1998; Moen et al., 1998; Idenden et radiation building up the plasma density on the dayside of
al., 1998; Pryse et al., 1998a; Mitchell et al., 1998; both convection cells. Figure 4 presents an example of using
Watermann et al., 2002]. A variety of structures have been IDA3D imaging to investigate the transport of polar cap
imaged: auroral E layer, nighttime field-aligned irregulari- patches. Figure 4 reveals a tongue of ionization extending
ties formed from soft particle precipitation, dayside and from the noon sector over the polar cap. The convection
nightside auroral arcs, and polar cap patches. Both auroral E pattern was obtained from the assimilative mapping iono-
ionization and a field-aligned irregularity on the poleward spheric electrodynamics algorithm. ‘‘Q’’ in Figure 4 denotes
wall of the trough are evident in Figure 3. the location of Qaanaaq. In addition to the tongue of
[70] The origins of polar cap patches have been investi- ionization it also appears as if plasma is being convected
gated over the last decade by various researchers using from the morning sector toward noon where the solar
tomographic imaging. Early results used images to reveal radiation increases the plasma density as it convects past
high-density plasma structures over Svalbard [Pryse et al., noon and back toward the midnight sector. These studies
1997]. Walker et al. [1999] demonstrated that particle have shown that polar cap plasma can arise both from
precipitation in the cusp (dayside) enhanced the plasma convected precipitation-produced plasma and from solar-
density sufficiently to create a structure that subsequently produced plasma, which itself can originate at both polar
convected into the polar cap as a patch. Watermann et al. and subauroral latitudes.
[2002] revealed polar cap patches convecting across the [71] Auroral structures have been studied with tomo-
polar region and warned against the misinterpretation of graphic imaging since 1995, when Mitchell et al. [1995]
polar cap plasma in static tomographic imaging when the showed the ability of tomography to image the trough, a
structures are moving at high velocity. Sims et al. [2005] field-aligned boundary blob, and auroral E ionization.
show an example where multi-instrument studies reveal that Heaton et al. [1996] showed a trough in the Southern
the plasma can be drawn in from subauroral latitudes. Stolle Hemisphere, revealing the first tomographic images in
et al. [2005] show patch convection under storm conditions Antarctica. Kersley et al. [1997] showed a variety of
where the interplanetary magnetic field is strongly south- tomographic images of the trough at U.K. and Scandinavian
14 of 23
latitudes and suggested that tomographic imaging could be a nearly 50° geomagnetic latitude, while the ionosonde
new tool for investigating the trough and creating new showed dramatic variations in the virtual height of the
models. Mitchell et al. [1999] created a large database of ionosphere. Hernández-Pajares et al. [1998] initiated the
U.K. tomographic images and used them to show that use of GPS tomography for studying ionospheric storms.
previous models of the trough, dependent on geomagnetic Previous ionospheric storm studies had suffered from a lack
indices, could be improved upon by including a seasonal of real knowledge about the longitudinal variation of the
term. Pryse et al. [1998b] used tomographic images from a plasma, and GPS clearly offered a real possibility to observe
2-week campaign to reveal the dayside trough at the edge of the large-scale structure of the dynamic and disturbed
the convection cell. Mitchell et al. [1998] published a multi- ionosphere. A number of tomographic studies have been
instrument study of the dayside auroral ionosphere where an conducted using GPS data from the recent solar maximum.
auroral arc was observed in both ESR and tomography. The July 2000 storm revealed dramatic uplifts in the F layer
More recently, further trough studies have appeared; Yizen- height over the mainland United States [Yin et al., 2004] that
gaw and Moldwin [2005] have shown the colocation of the were confirmed by Millstone Hill ISR. Garcia-Fernandez et
trough seen in tomographic imaging with the plasmapause, al. [2005] show the use of the SAC-C satellite occultation
thus confirming its location on the boundary between open data in ionospheric imaging. Yin and Mitchell [2005]
and closed geomagnetic field lines. Voiculescu et al. [2006] demonstrated the use of the CHAMP satellite occultation
conducted a statistical study of the main trough and found data in imaging the April 2002 storm time ionosphere.
that the interplanetary magnetic field (IMF) plays a role in Yizengaw and Moldwin [2005] used tomography to show
the occurrence of the trough at different levels of geomag- the trough and thus to demonstrate with other data the
netic activity. They also confirmed the importance of the collocation of the plasmapause with the trough (Figure 5).
season in the trough location. Thus tomographic imaging Yizengaw et al. [2006] revealed a downward plasma drift for
has been useful in revealing ionization at low altitudes from the November 2003 storm. More recently, Yin et al. [2006]
auroral E, F layer field-aligned blobs, and the main trough showed a time dependence in the peak height elevations
and has been able to provide a new insight to identify during several storms and discussed the electric field or
important factors for modeling of the trough, namely, neutral wind role in this effect. Thus GPS imaging is being
season and IMF. used to create a new view of the midlatitude ionosphere,
and, in particular, it complements the 2-D GPS mapping
5.2. Midlatitudes [e.g., Wilson et al., 1995] to reveal the 3-D dynamics of the
[72] In 1995 a special issue of Annales Geophysicae plasma during storms.
contained several papers on ionospheric tomography. Three
of these papers presented experimental results showing 5.3. Low Latitudes
tomographic images of traveling ionospheric disturbances [74] At lower latitudes, images of the equatorial anomaly
(TIDs) [Cook and Close, 1995; Markkanen et al., 1995; have been produced from the Asian sector [Huang et al.,
Pryse et al., 1995]. These wave-like structures are the 1997; Franke et al., 2003; Xu et al., 2000a, 2000b; Tsai et
manifestation of internal atmospheric gravity waves in the al., 2000], from India [Thampi et al., 2004], and from
ionosphere. Pryse et al. presented results showing medium- southern Europe [Materassi et al., 2003]. Huang et al.
scale TIDs obtained during an extended experimental cam- [1999] undertook an interesting study into the response of
paign, lasting some 7 months, held in the United Kingdom the low-latitude ionosphere during a solar eclipse. Their
in 1992– 1993. Tomographic images of these features from tomographic results had implications for understanding the
successive satellite passes allowed the meridional compo- ionospheric dynamics in the equatorial anomaly region
nent of the wavelengths and southward velocities to be during the eclipse. Xu et al. [2000a] revealed the rapid
found. Cook and Close also presented several images of formation of a depletion in ionization at 10° geomagnetic
TIDs from data recorded during the Mid-America Comput- latitude, where the anomaly peak usually lies, after the
erized Ionospheric Tomography Experiment (MACE ’93) commencement of a storm. Tsai et al. [2000] have used
campaign. Subsequent work by Nygren et al. [1997] tomographic imaging to study the seasonal variations of the
showed the preferential visibility of the structures due to location, the time of occurrence, and the height and the
their field-aligned slope. electron density of the maximum peak of the anomaly.
[73] The response of the ionosphere to large changes in Andreeva et al. [2000] showed the anomaly under a variety
geomagnetic activity is called an ionospheric storm. Iono- of geomagnetic conditions and seasons. Xu et al. [2000b]
spheric storms have been observed with TEC, and an demonstrated how tomography can be used to study the
excellent review of the findings is given by Mendillo day-to-day variability of the anomaly crests and interpreted
[2006]. Buonsanto et al. [1997] recognized the need to their results in the light of changing equatorial electrody-
make a global specification of the ionosphere at storm time namics. Yeh et al. [2001] found from their extensive study
and recognized the use of both radio and optical tomo- that the movement of the anomaly crest was about 1° per
graphic techniques. The strength of tomographic imaging is hour poleward, slowing as it reaches its highest latitude
that the plasma dynamics can be observed over a large where it remained until weakening and receding in the
scale. Bust et al. [1997] studied a November 1993 storm to afternoon. They also found that the crest latitude correlates
reveal a deep equatorward surge of the midlatitude trough to with the fountain strength. Kunitsyn et al. [2003] used
15 of 23
communications. The electron density at a given time

defines the refractive paths of the signals, and hence
knowledge of this is of interest to communications planners.
Second, the smaller-scale electron density structures and
regions of strong refractive gradients can also be important
for communications because the signals can experience
diffraction and refraction causing fading known as scintil-
lations as they pass through such regions.
[79] The ionosphere affects satellite navigation systems
because the signals experience an unknown time delay as
they pass through it that maps into a navigation error. This is
a problem for single-frequency systems, such as the GPS L1
navigation signal. To bound and manage this problem, a
number of real-time mapping systems have been con-
structed (e.g., the wide area augmentation system) that
continually monitor the ionospheric delay and create and
broadcast maps and error estimates to make corrections for
the ionospheric errors. Ionospheric tomography has been
Figure 5. Tomographic image of the trough and the proposed as a next-generation ionospheric correction for
location of the plasmapause [from Yizengaw et al., 2006]. regions where the ionosphere is more complicated, such as
the equatorial anomaly regions over South America. Meggs
et al. [2004] and Meggs and Mitchell [2006] have investi-
successive images of the anomaly to reveal 2-D electron gated the gain between 2-D ionospheric mapping and 3-D
fluxes. imaging and found a threefold improvement in TEC map-
[75] Huang et al. [1995] demonstrated the use of the IRI ping using 3-D imaging.
model in reconstructing the equatorial ionosphere using [80] Bust et al. [1994] investigated the application of
tomography. Doherty et al. [1997] used tomography over ionospheric tomography to single-site location (SSL) range
the U.S. sector in conjunction with a low-latitude iono- estimation, the determination of the location of an unknown
sphere model to look at the effects of a storm. Huang et al. HF transmitter. An extended campaign known as MACE
[1997] used the low-latitude ionospheric tomography ’93 was held in 1993. Satellite receivers were deployed in
network at 121°E to view the diurnal behavior of the two configurations. The first involved nine receivers in a
ionosphere. meridional chain stretching from Pierre, South Dakota, to
[76] Ledvina et al. [2004] demonstrated the first attempts Pharr, Texas, and for the second configuration the receivers
to understand the morphology of the ionosphere in the were shared between two parallel meridional chains set
location of scintillation events using ionospheric imaging. apart by 5° longitude. The tomographic images were
They used a network of GPS TEC receivers in South reconstructed using the MART algorithm. Two target trans-
America and GPS scintillation receivers across Brazil to mitters were deployed in Pierre, South Dakota, and Nor-
show that the predominant regions of scintillation are man, Oklahoma, and the azimuth and elevation of these
coincident with strong TEC gradients on the edge of the signals were recorded in San Antonio, Texas. The classical
anomaly walls. method of SSL range estimation uses an ionosonde to
[77] Thus ionospheric imaging has been used to study the estimate the ionospheric electron density, but Bust et al.
morphology and dynamics of the equatorial ionosphere on found that ray tracing through tomographic images pro-
both short and long timescales. The large- and small-scale duced a statistically smaller error in the range estimates.
processes are being related. The results will be useful to This was a very encouraging result and showed the promise
improve ionospheric models. of tomographic imaging for applications as well as scientific
work.
6. APPLICATIONS [81] Rogers et al. [2001] set up an experiment to image
the ionosphere over the United Kingdom and to make
[78] Scientific studies of the ionosphere are fundamen- coincident measurements with an oblique sounder. Thus
tally investigations where the objective is to increase and the images could be ray traced through and subsequently
enhance our understanding of the physical world. However, checked against the oblique paths. The results showed that
in addition to understanding the ionosphere from a physics tomographic images could be used as an aid to HF com-
standpoint, specification of the electron density in the munications planning.
ionosphere is useful for a number of different applications: [82] Ruffini et al. [1998] demonstrated that ionospheric
communications, navigation, and surveillance. The iono- imaging could be used to provide an ionospheric calibration
sphere can have two very different effects on communica- for radar altimeters. Hernández-Pajares et al. [1999] im-
tions systems. First, at HF the ionosphere is the medium that aged the ionosphere at European midlatitudes using ground
refracts signals back to the Earth in over the horizon GPS and ionosondes. They demonstrated how a single
16 of 23
Figure 6. Slices of tomographic images using different instrument combinations. (a) Reference IRI
distribution, (b) inversion using semisynthetic GPS data and ionosondes, (c) only semisynthetic GPS, (d)
difference between Figures 6b and 6a, (e) result using real GPS and ionosonde data, and (f) only GPS
data. The five layer boundary heights are (bottom to top) 75, 125, 175, 250, 650, and 2000 km. Reprinted
from Hernandez – Pajares et al. [1999], with permission from Elsevier.
ionosonde when combined with GPS ground data could and double-differenced ionospheric corrections in real time
provide reliable 3-D images of electron density over 20° of and at very long distances between receivers (500– 3000 km).
latitude. Figure 6 shows horizontal images at five different It has been tested under difficult ionospheric conditions
heights. The five layer boundary heights are (starting at the during four consecutive weeks in March– April 2001 at solar
bottom) 75, 125, 175, 250, 650, and 2000 km. Figure 6a maximum and at latitudes ranging from -40° to +40° to
represents the reference IRI distribution, Figure 6b is the include the equatorial region. The results are very promising
inversion using semisynthetic GPS data and ionosondes, for application to geodetic problems.
Figure 6c is only the semisynthetic GPS, Figure 6d is the
difference between Figures 6b and 6a, Figure 6e is the result 7. FUTURE DIRECTIONS
using real GPS and ionosonde data, and Figure 6f uses only
GPS data. The inversion is for 2 November 1997. This type [84] Over the last 20 years, ionospheric tomography has
of quantification of the imaging error is essential for the developed from 2-D simulations into 4-D near-global spec-
assessment of ionospheric imaging for many different ification of the Earth’s ionosphere. New directions for the
applications. research are likely to be in several areas: (1) improved
[83] Hernández-Pajares [2000, 2002] has investigated electron density values that can be achieved through better
the application of ionospheric imaging to combine real-time algorithms and more data, (2) extension of the imaging to
geodetic and ionospheric techniques to achieve a significant variables other than the electron density, (3) combining
improvement in the reliability of carrier phase ambiguity tomographic images with models to identify the physical
resolution. The model is obtained from the GPS L1 and L2 driving parameters for a system, (4) extension of the
carrier phase data, and it is used to estimate undifferenced techniques into other regions of the Earth’s environment
17 of 23
and the solar system, and (5) self-consistently combining Vertical resolution can be achieved by GPS occultations for
imaging results at different spatial scales. These ideas are a constellation of satellites such as the current constellation
discussed in sections 7.1– 7.5. observing system for meteorology ionosphere and climate
(COSMIC) array. Finally, as more and more data sets become
7.1. Improved 4-D Imaging of Ionospheric Electron
available, the goal is for 4-D imaging algorithms to have
Density
less and less reliance on a priori information, with the
[85] While the current algorithms that provide 4-D im- imaging results only depending on the data.
aging of ionospheric density are mature, well-validated, and
accurate and have been applied to both science and appli- 7.2. Imaging Other Ionospheric and Thermospheric
cations, there is a lot of room for improvement. First, the State Variables
addition of more sensors of measurements than are currently [88] Any state variable that can be related to observations
used by 4-D imaging will increase the global data coverage, through some forward model, and with observations dis-
accuracy, and resolution. These measurements include TEC tributed over a large spatial region, can, in principle, be
from the European Space Agency Galileo satellites, in- amenable to 4-D imaging methods. For example, a thermo-
creased deployment of new ground-based, inexpensive spheric variable that could be imaged is neutral composition
GPS software receivers, deployments of Digisondes and using optical ground measurements and UV space-based
dynasondes, satellite-based TEC measurements from GPS measurements as data sources. In addition, multivariate
navigation receiver and GPS occultation measurements, imaging methods can be developed similar to multivariate
TEC measurements from the Doppler orbitography and objective analysis in meteorology [Daley, 1991]. In multi-
radiopositioning integrated by satellite (DORIS) network variate analysis, error cross-covariances between state vari-
of ground-based transmitters, cross-link TEC measurements ables exist, and the cross-covariances allow all observations,
between LEO satellites transmitting on 150 and 400 MHz both ionospheric and neutral, to influence the state variables.
and the new scintillation and tomography receiver in space Other possible state variables that could be imaged from
(CITRIS) recently launched on STPSat-1, and in situ observations include neutral winds, temperatures, and iono-
measurements of electron density from satellites. In addi- spheric currents.
tion, new data sets can be incorporated into 4-D imaging
algorithms. These data sets generally involve a more com- 7.3. Estimation of Physical Drivers From 4-D Imaging
plex relationship with the underlying electron density, [89] The idea of combining physical models with tomo-
including nonlinear relationships, and the need to specify graphic images is a very powerful tool for physical studies.
auxiliary information. Some of the new data sources being The goal is to extract as much information about the state of
considered for 4-D imaging include EUV observations at the ionosphere system from the imaging as is possible. The
1356 Å, the more basic time delay versus frequency 4-D imaging reveals the time-evolving electron density. The
observations obtained from ionograms [Friedman et al., next step is to obtain quantitative information on the
2006], time delay versus frequency from oblique HF instru- physical drivers that are producing the time-evolving elec-
ments, Riometer absorption data, optical data at 7774 and tron density. This is different from estimating inputs to
6300 Å, and time delay versus frequency from whistlers. sophisticated data assimilation algorithms [Pi et al., 2003,
Both the Riometer data and the optical 6300 data require 2004; Scherliess et al., 2004]. There, a highly complex,
specification of auxiliary parameters (collision frequencies complicated forward model exists between the desired
and neutral densities), but to the degree that empirical or first- inputs and the measured data, making accurate estimation
principle models can make good estimates of the auxiliary of the inputs an extremely challenging problem. In 4-D
parameters, these data sources can be used to improve imaging the relationship between the 3-D time-evolving
imaging in the E region and F1 layer of the ionosphere. images and the desired physical driver is kept as simple as
[86] Many of the measurements from these new data possible. To take just one example, at high latitudes if the
sources are nonlinearly related to the electron density. The variation in electron density is due primarily to horizontal
1356 EUV data and 7774 optical data are given by a line transport, the imaging can be used to obtain estimates on the
integral over the square of the electron density (assuming plasma drifts caused by the changing electric fields. This
that O+ is equal to electron density). Time delay versus could be done by identification of plasma structures in
frequency measurements from ionograms, HF oblique consecutive frames of the imaging and calculating the
measurements, and whistlers involve integrals over the necessary electric field strengths to convect the plasma
refractive index, which is strongly nonlinear and requires using information on the magnetic field strength from a
ray-tracing methods. model.
[87] In addition to adding additional data sources, iono- 7.4. Imaging of Other Geophysical Systems
spheric 4-D imaging can be improved in several ways. As
[90] Any physical variable that can be related through an
more data sets become available, higher-resolution spatial
integral equation to observables can, in principle, be applied
images can be made. A reasonable goal is to achieve
to 4-D imaging. However, sufficient observations are needed
horizontal resolution of 10 km and vertical resolution of
over sufficient regions to image accurately the quantity of
5 km. The horizontal resolution can be achieved by a local
interest. For other geophysical systems, new and novel ways
dense array of ground instruments such as GPS receivers.
to take measurements that can be related to underlying
18 of 23
physical variables need to be investigated so that the general PPARC. G.S.B. acknowledges support from NSF and ONR during
methodology of tomographic inverse imaging can be ap- the preparation of this manuscript. The authors are grateful to the
plied to the geophysical problem. Some of the geophysical late K. C. Yeh for originally suggesting the idea of doing iono-
problems that are open to exploration through 4-D imaging spheric tomography.
include unexplored Earth regions such as the deep ocean [95] The Editor responsible for this paper was Peter Riley. He
thanks Manuel Hernández-Pajares, an anonymous technical re-
and Antarctica, plasma density throughout the entire solar/
viewer, and an anonymous cross-disciplinary reviewer.
magnetosphere/ionosphere, and various geophysical prob-
lems on other planets. REFERENCES
7.5. Linking Multiple Scales and/or Different Regions
Afraimovich, E. L., O. M. Pirog, and A. I. Terekhov (1992), Diag-
[91] Imaging of different geophysical regions such as the nostics of large-scale structures of the high-latitude ionosphere
solar region, magnetosphere, and ionosphere leads to the based on tomographic treatment of navigation satellite signals
need to self-consistently couple these different imaged and of data from ionospheric stations, J. Atmos. Terr. Phys.,
regions into a single coherent picture. Self-consistently 54, 1265 – 1273.
combining imaging results at different spatial scales is also Andreeva, E. S. (1990), Radio tomographic reconstruction of io-
nization dip in the plasma near the Earth, J. Exp. Theor. Phys.
desired. For example, ionospheric diffraction tomography Lett., 52, 142 – 148.
acts on irregularities over scales of a few meters to few Andreeva, E. S., S. J. Franke, K. C. Yeh, and V. E. Kunitsyn
kilometers. Global 4-D imaging acts over (horizontal) scales (2000), Some features of the equatorial anomaly revealed by
ranging from a few tens of kilometers to the globe. A dense ionospheric tomography, Geophys. Res. Lett., 27, 2465 – 2468.
network system would image the scales in between global Austen, J. R., S. J. Franke, and C. H. Liu (1988), Ionospheric
imaging using computerized tomography, Radio Sci., 23, 299 –
and irregularity scales. These differently scaled imaging 307.
systems should all transition self-consistently from one to Bailey, G. J., N. Balan, and Y. Z. Su (1997), The Sheffield Uni-
the other, with large, global imaging providing the initial versity plasmasphere ionosphere model—A review, J. Atmos.
backgrounds for regional imaging, which then provides the Sol. Terr. Phys., 59, 1541 – 1552.
gradients that drives the irregularities. Bernhardt, P. A., et al. (1998), Two-dimensional mapping of the
plasma density in the upper atmosphere with computerized iono-
[92] One method that can be used to couple different spheric tomography (CIT), Phys. Plasmas, 5, 2010 – 2021.
spatial regions and scales is the full data assimilation Beynon, W. J. G., and P. J. S. Williams (1978), Incoherent scatter of
approach discussed in section 3.2. Data assimilation techni- radio waves from the ionosphere, Rep. Prog. Phys., 41, 910 – 956.
ques combine data (or imaging results) with first-principle Bilitza, D. (2001), International Reference Ionosphere 2000, Radio
Sci., 36, 261 – 275.
models. It is in theory possible to self-consistently couple
Bracewell, R. N. (1956), Strip integration in radio astronomy, Aust.
the models on different spatial scales and in differing J. Phys., 9, 198 – 217.
geophysical regions, and, in fact, such efforts are underway. Buonsanto, M. J., M. Codrescu, B. A. Emery, C. G. Fesen, T. J.
Then, these coupled, self-consistent systems can ingest data Fuller-Rowell, D. J. Melendez-Alvira, and D. P. Sipler (1997),
from a variety of data sources, spatial regions, and spatial Comparison of models and measurements at Millstone Hill dur-
ing the January 24 – 26, 1993, minor storm interval, J. Geophys.
scales, use them to self-consistently update the initial
Res., 102, 7267 – 7277.
conditions in the coupled system and the physical drivers Bust, G. S., and G. Crowley (2007), Tracking of polar cap patches
and then produce self-consistent images of the medium on using data assimilation, J. Geophys. Res., 112, A05307,
all scales of interest. Nevertheless, the complexity of this doi:10.1029/2005JA011597.
should not be underestimated. It is not simply a matter of Bust, G. S., J. A. Cook, G. R. Kronschnabl, C. J. Vasicek, and S. B.
Ward (1994), Application of ionospheric tomography to single-
increasing the resolution; scales are implicit in the formation site location range estimation, Int. J. Imaging Syst. Technol., 5,
of the existing ionospheric models, and in order to change 160 – 168.
scales and resolutions, models require more physics than Bust, G. S., T. L. Gaussiran, and D. S. Coco (1997), Ionospheric
they currently contain. It is also possible that our under- observations of the November 1993 storm, J. Geophys. Res.,
standing of the linkage between the physics of different 102, 14,293 – 14,304.
Bust, G. S., D. Coco, and J. J. Makela (2000), Combined iono-
scales could be deficient. spheric campaign: 1. Ionospheric tomography and GPS total
[93] At this point, data assimilation can be considered as electron count (TEC) depletions, Geophys. Res. Lett., 27,
a general technique to ingest data that itself may include 4- 2849 – 2852.
D imaging of any combination of parameters in the Sun- Bust, G. S., T. W. Garner, and T. L. Gaussiran II (2004), Iono-
Earth system. There are no longer simple integral forward spheric Data Assimilation Three-Dimensional (IDA3D): A glo-
bal, multisensor, electron density specification algorithm,
models relating state variables to observations: The full J. Geophys. Res., 109, A11312, doi:10.1029/2003JA010234.
power of self-consistent first-principle physics models is Censor, Y. (1983), Finite series expansion methods, Proc. IEEE,
required to couple all the imaging regions together. The 71, 409 – 419.
ultimate result of such a coupling is a single 4-D imaging Ciraolo, L., F. Azpilicueta, C. Brunini, A. Meza, and S. M.
result of the entire Sun-Earth system, involving all variables Radicella (2007), Calibration errors on experimental slant total
electron content (TEC) determined with GPS, J. Geod., 81(2),
of interest on all spatial and temporal scales of interest. 111 – 120, doi:10.1007/s00190-006-0093-1.
Coker, C., G. Kronschnabl, D. S. Coco, G. S. Bust, and T. L.
[94] ACKNOWLEDGMENTS. C.N.M. would like to ac- Gaussiran (2001), Verification of ionospheric sensors, Radio
knowledge support from the UK research councils EPSRC and Sci., 36, 1523 – 1529.
19 of 23
Cook, J. A., and S. Close (1995), An investigation of TID evolu- Atmosphere, Ionosphere, and Magnetosphere, Cambridge At-
tion observed in MACE ’93 data, Ann. Geophys., 13, 1320 – mos. Space Sci. Ser., vol. 5, Cambridge Univ. Press., New York.
1324. Heaton, J. A. T., S. E. Pryse, and L. Kersley (1995), Improved
Cornely, P. J. (2003), Flexible Prior Models: Three-dimensional background representation, ionosonde input and independent
ionospheric tomography, Radio Sci., 38(5), 1087, doi:10.1029/ verification in experimental ionospheric tomography, Ann. Geo-
2002RS002703. phys., 13, 1297 – 1302.
Crowley, G., et al. (2006), Global thermosphere-ionosphere re- Heaton, J. A. T., G. O. L. Jones, and L. Kersley (1996), Toward
sponse to onset of 20 November 2003 magnetic storm, J. Geo- ionospheric tomography in Antarctica: First steps and compari-
phys. Res., 111, A10S18, doi:10.1029/2005JA011518. son with dynasonde observations, Antarct. Sci., 8, 297 – 302.
Daley, R. (1991), Atmospheric Data Analysis, Cambridge Univ. Hedin, A. E. (1991), Extension of the MSIS thermosphere model
Press, Cambridge, U. K. into the middle and lower atmosphere, J. Geophys. Res., 96,
Daley, R., and E. Barker (2001), NAVDAS: Formulation and 1159 – 1172.
diagnostics, Mon. Weather Rev., 129(4), 869 – 883. Hedin, A. E., et al. (1996), Empirical wind model for the upper,
Daniell, R. E. (1991), Parameterized real-time ionospheric specifi- middle and lower atmosphere, J. Atmos. Terr. Phys., 58, 1421 –
cation model: PRISM version 1.0, Tech. Rep. PL-TR-91 – 2299, 1447.
Phillips Lab., Hanscom Air Force Base, Mass. Heppner, J. P., and N. C. Maynard (1987), Empirical high-latitude
Davies, K. (1990), Ionospheric Radio, IEE Electromagn. Waves electric field models, J. Geophys. Res., 92, 4467 – 4489.
Ser., vol. 31, P. Peregrinus, London. Hernández-Pajares, M., J. M. Juan, J. Sanz, and J. G. Sole (1998),
Doherty, P. H., D. N. Anderson, and J. A. Klobuchar (1997), Total Global observation of the ionospheric electronic response to solar
electron content over the Pan-American longitudes: March – events using ground and LEO GPS data, J. Geophys. Res., 103,
April 1994, Radio Sci., 32, 1597 – 1606. 20,789 – 20,796.
Fejer, B. G. (1991), Low latitude electrodynamic plasma drifts: A Hernández-Pajares, M., J. M. Juan, and J. Sanz (1999), New ap-
review, J. Atmos. Terr. Phys., 53, 677 – 693. proaches in global ionospheric determination using ground GPS
Foster, J. C., et al. (1994), Russian-American tomography experi- data, J. Atmos. Sol. Terr. Phys., 61, 1237 – 1247.
ment, Int. J. Imaging Syst. Technol., 5, 148 – 159. Hernández-Pajares, M., J. M. Juan, J. Sanz, and O. L. Colombo
Fougere, P. F. (1995), Ionospheric radio tomography using max- (2000), Application of ionospheric tomography to real-time GPS
imum entropy: 1. Theory and simulation studies, Radio Sci., 30, carrier-phase ambiguities resolution, at scales of 400 – 1000 km
429 – 444. and with high geomagnetic activity, Geophys. Res. Lett., 27,
Franke, S. J., K. C. Yeh, E. S. Andreeva, and V. E. Kunitsyn 2009 – 2012.
(2003), A study of the equatorial anomaly ionosphere using Hernández-Pajares, M., J. M. Juan, J. Sanz, and O. L. Colombo
tomographic images, Radio Sci., 38(1), 1011, doi:10.1029/ (2002), Improving the real-time ionospheric determination from
2002RS002657. GPS sites at very long distances over the equator, J. Geophys.
Fremouw, E. J., J. A. Secan, and B. M. Howe (1992), Application Res., 107(A10), 1296, doi:10.1029/2001JA009203.
of stochastic inverse-theory to ionospheric tomography, Radio Hounsfield, G. N. (1972), A method of and apparatus for exam-
Sci., 27, 721 – 732. ination of a body by radiation such as X-ray or gamma radiation,
Friedman, S. V., L. J. Nickisch, M. Aiello, and M. Hausman Patent 1283915, U.S. Patent and Trademark Off., Washington,
(2006), Real-time reconstruction of the three-dimensional iono- D. C.
sphere using data from a network of GPS receivers, Radio Sci., Howe, B. M., K. Runciman, and J. A. Secan (1998), Tomography
41, RS5S12, doi:10.1029/2005RS003341. of the ionosphere: Four-dimensional simulations, Radio Sci., 33,
Fuller-Rowell, T. J., and D. S. Evans (1987), Height-integrated 109 – 128.
Pedersen and Hall conductivity patterns inferred from the Huang, C. R., C. H. Liu, H. C. Yeh, W. H. Tsai, C. J. Wang, K. C.
TIROS-NOAA satellite data, J. Geophys. Res., 92, 7606 – 7618. Yeh, K. H. Lin, and H. L. Tsai (1995), IRI model application in
Garcia-Fernandez, M., A. Saito, J. M. Juan, and T. Tsuda (2005), low latitude ionospheric tomography, in Low and Equatorial
Three-dimensional estimation of electron density over Japan Latitudes in the International Reference Ionosphere (IRI), edited
using the GEONET GPS network combined with SAC-C data by K. Rawer et al., pp. 237 – 240, Elsevier, New York.
and ionosonde measurements, J. Geophys. Res., 110, A11304, Huang, C. R., C. H. Liu, H. C. Yeh, and W. K. Tsai (1997), The
doi:10.1029/2005JA011037. low-latitude ionospheric tomography network (LITN)—Initial
Garner, T. W., G. S. Bust, T. L. Gaussiran II, and P. R. Straus results, J. Atmos. Sol. Terr. Phys., 59, 1553 – 1567.
(2006), Variations in the midlatitude and equatorial ionosphere Huang, C. R., C. H. Liu, K. C. Yeh, K. H. Lin, W. H. Tsai, H. C.
during the October 2003 magnetic storm, Radio Sci., 41, Yeh, and J. Y. Liu (1999), A study of tomographically recon-
RS6S08, doi:10.1029/2005RS003399. structed ionospheric images during a solar eclipse, J. Geophys.
Gelb, A. (1974), Applied Optimal Estimation, MIT Press, Cam- Res., 104, 79 – 94.
bridge, Mass. Idenden, D. W., R. J. Moffett, M. J. Williams, P. S. J. Spencer, and
Gordon, R., R. Bender, and G. T. Herman (1970), Algebraic L. Kersley (1998), Imaging of structures in the high-latitude
Reconstruction Techniques (ART) for three dimensional electron ionosphere: Model comparisons, Ann. Geophys., 16, 969 – 973.
microscopy and X-ray photography, J. Theor. Biol., 29, 471 – Juan, J. M., A. Rius, M. Hernández-Pajares, and J. Sanz (1997), A
481. two-layer model of the ionosphere using global positioning sys-
Greenwald, R. A., et al. (1995), Darn/Superdarn: A global view of tem data, Geophys. Res. Lett., 24, 393 – 396.
the dynamics of high-latitude convection, Space Sci. Rev., 71, Kak, A. C., and M. Slaney (2001), Principles of Computerized
761 – 796. Tomographic Imaging, Soc. of Ind. and Appl. Math., Philadel-
Hajj, G. A., R. Ibanezmeier, E. R. Kursinski, and L. J. Romans phia, Pa.
(1994), Imaging the ionosphere with the global positioning sys- Kersley, L., J. A. T. Heaton, S. E. Pryse, and T. D. Raymund
tem, Int. J. Imaging Syst. Technol., 5, 174 – 184. (1993), Experimental ionospheric tomography with ionosonde
Hajj, G. A., B. D. Wilson, C. Wang, X. Pi, and I. G. Rosen (2004), input and EISCAT verification, Ann. Geophys., 11, 1064 – 1074.
Data assimilation of ground GPS total electron content into a Kersley, L., S. E. Pryse, I. K. Walker, J. A. T. Heaton, C. N.
physics-based ionospheric model by use of the Kalman filter, Mitchell, M. J. Williams, and C. A. Willson (1997), Imaging
Radio Sci., 39, RS1S05, doi:10.1029/2002RS002859. of electron density troughs by tomographic techniques, Radio
Hargreaves, J. K. (1995), The Solar-Terrestrial Environment: An Sci., 32, 1607 – 1621.
Introduction to Geospace—The Science of the Terrestrial Upper
20 of 23
Kivelson, M. G., and C. T. Russell (1995), Introduction to Space Meggs, R. W., C. N. Mitchell, and V. S. C. Howells (2005),
Physics, Cambridge Univ. Press, New York. Simultaneous observations of the main trough using GPS ima-
Kuklinski, W. S. (1997), Ionospheric tomography via iterative ging and the EISCAT radar, Ann. Geophys., 23, 753 – 757.
cross-entropy minimization, Radio Sci., 32, 1037 – 1049. Mendillo, M. (2006), Storms in the ionosphere: Patterns and pro-
Kunitake, M., K. Ohtaka, T. Maruyama, M. Tokumaru, A. Morioka, cesses for total electron content, Rev. Geophys., 44, RG4001,
and S. Watanabe (1995), Tomographic imaging of the ionosphere doi:10.1029/2005RG000193.
over Japan by the modified truncated SVD method, Ann. Geo- Menerbo, G. (1979), MENT: A maximum entropy algorithm for
phys., 13, 1305 – 1310. reconstructing a source from projection data, Comput. Graphics
Kunitsyn, V. E., E. S. Andreeva, O. G. Razinkov, and E. D. Image Process., 10, 48 – 68.
Tereshchenko (1994a), Phase and phase-difference iono- Mitchell, C. N. (2002), Imaging of near-Earth space plasma, Phi-
spheric radio tomography, Int. J. Imaging Syst. Technol., 5, los. Trans. R. Soc. London, Ser. A, 360, 2805 – 2818.
128 – 140. Mitchell, C. N., and P. S. J. Spencer (2003), A three-dimensional
Kunitsyn, V. E., E. S. Andreeva, E. D. Tereshchenko, B. Z. time-dependent algorithm for ionospheric imaging using GPS,
Khudukon, and T. Nygren (1994b), Investigations of the iono- Ann. Geophys., 46, 687 – 696.
sphere by satellite radiotomography, Int. J. Imaging Syst. Tech- Mitchell, C. N., D. G. Jones, L. Kersley, S. E. Pryse, and I. K.
nol., 5, 112 – 127. Walker (1995), Imaging of field-aligned structures in the auroral
Kunitsyn, V. E., E. S. Andreeva, A. Y. Popov, and O. G. Razinkov ionosphere, Ann. Geophys., 13, 1311 – 1319.
(1995), Methods and algorithms of ray radiotomography for io- Mitchell, C. N., S. E. Pryse, L. Kersley, and I. K. Walker (1997a),
nospheric research, Ann. Geophys., 13, 1263 – 1276. The correction for the satellite-receiver longitude difference in
Kunitsyn, V. E., E. S. Andreeva, S. J. Franke, and K. C. Yeh ionospheric tomography, J. Atmos. Sol. Terr. Phys., 59, 2077 –
(2003), Tomographic investigations of temporal variations of 2087.
the ionospheric electron density and the implied fluxes, Geophys. Mitchell, C. N., L. Kersley, and S. E. Pryse (1997b), The effects of
Res. Lett., 30(16), 1851, doi:10.1029/2003GL016908. receiver location in two-station experimental ionospheric tomo-
Leitinger, R., G. Schmidt, and A. Tauriainen (1975), An evaluation graphy, J. Atmos. Sol. Terr. Phys., 59, 1411 – 1415.
method combining the differential Doppler measurements from Mitchell, C. N., L. Kersley, J. A. T. Heaton, and S. E. Pryse
two stations that enables the calculation of the electron content of (1997c), Determination of the vertical electron-density profile
the ionosphere, J. Geophys., 41, 201 – 213. in ionospheric tomography: Experimental results, Ann. Geophys.,
Leitinger, R., H. P. Ladreiter, and G. Kirchengast (1997), Iono- 15, 747 – 752.
sphere tomography with data from satellite reception of Global Mitchell, C. N., I. K. Walker, S. E. Pryse, I. Kersley, I. W. McCrea,
Navigation Satellite System signals and ground reception of and T. B. Jones (1998), First complementary observations by
Navy Navigation Satellite System signals, Radio Sci., 32, ionospheric tomography, the EISCAT Svalbard radar and the
1657 – 1669. CUTLASS HF radar, Ann. Geophys., 16, 1519 – 1522.
Ledvina, B. M., P. M. Kintner, and J. J. Makela (2004), Temporal Mitchell, C. N., L. Kersley, S. E. Pryse, C. A. Willson, J. A. T.
properties of intense GPS L1 amplitude scintillations at midlati- Heaton, P. S. Cannon, and N. C. Rogers (1999), Imaging and
tudes, Radio Sci., 39, RS1S18, doi:10.1029/2002RS002832. modelling of the main ionospheric trough using radio tomogra-
Lent, A. (1997), A convergent algorithm for maximum entropy phy, in IEE National Conference on Antennas and Propagation,
image restoration, with a medical X-ray application, in SPSE 31 March – 1 April 1999, IEE Conf. Publ., 461, 192 – 195.
Conference Proceedings, edited by R. Shaw, pp. 249, Soc. of Moen, J., S. T. Berry, L. Kersley, and B. Lybekk (1998), Probing
Photogr. Sci. and Eng., Washington, D. C. discrete auroral arcs by ionospheric tomography, Ann. Geophys.,
Mandrake, L., B. Wilson, C. Wang, G. Hajj, A. Mannucci, and X. Pi 16, 574 – 582.
(2005), A performance evaluation of the operational Jet Propul- Na, H., and H. Lee (1990), Resolution analysis of tomographic
sion Laboratory/University of Southern California Global Assim- reconstruction of electron density profiles in the ionosphere,
ilation Ionospheric Model (JPL/USC GAIM), J. Geophys. Res., Int. J. Imaging Syst. Technol., 2, 209 – 218.
110, A12306, doi:10.1029/2005JA011170. Na, H., and E. Sutton (1994), Resolution analysis of ionospheric
Mannucci, A. J., B. Iijima, L. Sparks, X. Q. Pi, B. Wilson, and tomography systems, Int. J. Imaging Syst. Technol., 5, 169 – 173.
U. Lindqwister (1999), Assessment of global TEC mapping Na, H., J. Shen, and H. Lee (1995), A Fourier domain technique
using a three-dimensional electron density model, J. Atmos. for ionospheric tomography, Radio Sci., 30, 747 – 754.
Sol. Terr. Phys., 61, 1227 – 1236. Nygren, T., M. Markkanen, M. Lehtinen, E. D. Tereshchenko, and
Markkanen, M., M. Lehtinen, T. Nygren, J. Pirttila, P. Henelius, B. Z. Khudukon (1997), Stochastic inversion in ionospheric
E. Vilenius, E. D. Tereshchenko, and B. Z. Khudukon (1995), radiotomography, Radio Sci., 32, 2359 – 2372.
Bayesian approach to satellite radiotomography with applica- Pi, X., C. Wang, G. A. Hajj, G. Rosen, B. D. Wilson, and G. J.
tions in the Scandinavian sector, Ann. Geophys., 13, 1277 – Bailey (2003), Estimation of E B drift using a global assim-
1287. ilative ionospheric model: An observation system simulation ex-
Materassi, M., and C. N. Mitchell (2005a), A simulation study into periment, J. Geophys. Res., 108(A2), 1075, doi:10.1029/
constructing of the sample space for ionospheric imaging, 2001JA009235.
J. Atmos. Sol. Terr. Phys., 67, 1085 – 1091. Pi, X. Q., C. Wang, G. A. Hajj, G. Rosen, B. D. Wilson, and A. J.
Materassi, M., and C. N. Mitchell (2005b), Imaging of the equa- Mannucci (2004), Assimilative modeling of low-latitude
torial ionosphere, Ann. Geophys., 48(3), 477 – 482. ionosphere, paper presented at 2004 IEEE Position, Location
Materassi, M., C. N. Mitchell, and P. S. J. Spencer (2003), Iono- and Navigation Symposium—PLANS 2004, Monterey, Calif.,
spheric imaging of the northern crest of the equatorial anomaly, 26 – 29 April.
J. Atmos. Sol. Terr. Phys., 65, 1393 – 1400. Pryse, S. E., and L. Kersley (1992), A preliminary experimental
McNamara, L. P. (1991), Ionosphere: Communications, Surveil- test of ionospheric tomography, J. Atmos. Terr. Phys., 54, 1007 –
lance and Direction Finding, Krieger, Malabar, Fla. 1012.
Meggs, R. W., and C. N. Mitchell (2006), A study into the errors in Pryse, S. E., L. Kersley, D. L. Rice, C. D. Russell, and I. K. Walker
vertical total electron content mapping using GPS data, Radio (1993), Tomographic imaging of the ionospheric midlatitude
Sci., 41, RS1008, doi:10.1029/2005RS003308. trough, Ann. Geophys., 11, 144 – 149.
Meggs, R. W., C. N. Mitchell, and P. S. J. Spencer (2004), A Pryse, S. E., C. N. Mitchell, J. A. T. Heaton, and L. Kersley (1995),
comparison of techniques for mapping total electron content over Travelling ionospheric disturbances imaged by tomographic
Europe using GPS signals, Radio Sci., 39, RS1S10, doi:10.1029/ techniques, Ann. Geophys., 13, 1325 – 1330.
2002RS002846.
21 of 23
Pryse, S. E., L. Kersley, M. J. Williams, I. K. Walker, and C. A. the north polar ionosphere on 30 October 2003, in Plane-
Wilson (1997), Tomographic imaging of the polar-cap iono- tary Atmospheres, Ionospheres, and Magnetospheres, edited
sphere over Svalbard, J. Atmos. Sol. Terr. Phys., 59, 1953 – by O. Witasse and M. A. Shea, Adv. Space Res., 36, 2201 – 2206.
1959. Sutton, E., and H. Na (1994), Orthogonal decomposition frame-
Pryse, S. E., L. Kersley, and M. J. Williams (1998a), Electron work for ionospheric tomography algorithms, Int. J. Imaging
density structures in the polar cap imaged by ionospheric tomo- Syst. Technol., 5, 106 – 111.
graphy, in Polar Cap Thermosphere/Ionosphere and Its Role in Thampi, S. V., T. K. Pant, S. Ravindran, C. V. Devasia, and
Solar-Terrestrial Physics, edited by U. P. Lovhaug, Adv. Space R. Sridharan (2004), Simulation studies on the tomographic
Res., 22(9), 1385 – 1389. reconstruction of the equatorial and low-latitude ionosphere in
Pryse, S. E., L. Kersley, M. J. Williams, and I. K. Walker (1998b), the context of the Indian tomography experiment: CRABEX,
The spatial structure of the dayside ionospheric trough, Ann. Ann. Geophys., 22, 3445 – 3460.
Geophys., 16, 1169 – 1179. Thompson, D. C., L. Scherliess, J. J. Sojka, and R. W. Schunk
Radon, J. (1917), Über die Bestimmung von Funktionen durch ihre (2006), The Utah State University Gauss-Markov Kalman filter
Integralwerte längs gewisser mannigfaltigkeiten, Saechs. Ber. of the ionosphere: The effect of slant TEC and electron density
Akad. Wiss., 69, 262 – 277. profile data on model fidelity, J. Atmos. Sol. Terr. Phys., 68,
Raymund, T. D. (1994), Ionospheric tomography algorithms, Int. J. 947 – 958.
Imaging Syst. Technol., 5, 75 – 85. Tobiska, W. K. (1991), Revised solar extreme ultraviolet flux
Raymund, T. D., J. R. Austen, S. J. Franke, C. H. Liu, J. A. model, J. Atmos. Terr. Phys., 53, 1005 – 1018.
Klobuchar, and J. Stalker (1990), Application of computerized- Tsai, W. H., L. F. Huang, M. F. Chen, and C. H. Liu (2000),
tomography to the investigation of ionospheric structures, Radio A tomographic study of seasonal variations of the equatorial
Sci., 25, 771 – 789. anomaly in the Asian sector, Terr. Atmos. Oceanic Sci., 11,
Raymund, T. D., S. E. Pryse, L. Kersley, and J. A. T. Heaton 337 – 348.
(1993), Tomographic reconstruction of ionospheric electron den- Vasicek, C. J., and G. R. Kronschnabl (1995), Ionospheric
sity with European incoherent scatter radar verification, Radio tomography - An algorithm enhancement, J. Atmos. Terr. Phys.,
Sci., 28, 811 – 817. 57, 875 – 888.
Raymund, T. D., Y. Bresler, D. N. Anderson, and R. E. Daniell Voiculescu, M., I. Virtanen, and T. Nygren (2006), The F-region
(1994), Model-assisted ionospheric tomography: A new algo- trough: Seasonal morphology and relation to interplanetary mag-
rithm, Radio Sci., 29, 1493 – 1512. netic field, Ann. Geophys., 24, 173 – 185.
Rius, A., G. Ruffini, and L. Cucurull (1997), Improving the ver- Walker, I. K., J. A. T. Heaton, L. Kersley, C. N. Mitchell, S. E.
tical resolution of ionospheric tomography with GPS occulta- Pryse, and M. J. Williams (1996), EISCAT verification in the
tions, Geophys. Res. Lett., 24, 2291 – 2294. development of ionospheric tomography, Ann. Geophys., 14,
Roble, R. G., and E. C. Ridley (1994), A thermosphere-ionosphere- 1413 – 1421.
mesosphere electrodynamics general circulation model Walker, I. K., J. Moen, C. N. Mitchell, L. Kersley, and P. E.
(TIME-GCM): Equinox solar cycle minimum simulations Sandholt (1998), Ionospheric effects of magnetopause reconnec-
(30 – 500 km), Geophys. Res. Lett., 21, 417 – 420. tion observed using ionospheric tomography, Geophys. Res.
Rogers, N. C., C. N. Mitchell, J. A. T. Heaton, P. S. Cannon, and Lett., 25, 293 – 296.
L. Kersley (2001), Application of radio tomographic imaging to Walker, I. K., J. Moen, L. Kersley, and D. A. Lorentzen (1999), On
HF oblique incidence ray tracing, Radio Sci., 36, 1591 – 1598. the possible role of cusp/cleft precipitation in the formation of
Ruffini, G., E. Cardellach, A. Flores, L. Cucurull, and A. Rius polar-cap patches, Ann. Geophys., 17, 1298 – 1305.
(1998), Ionospheric calibration of radar altimeters using GPS Wang, C., G. Hajj, X. Pi, I. G. Rosen, and B. Wilson (2004),
tomography, Geophys. Res. Lett., 25, 3771 – 3774. Development of the Global Assimilative Ionospheric Model,
Scherliess, L., and B. G. Fejer (1999), Radar and satellite global Radio Sci., 39, RS1S06, doi:10.1029/2002RS002854.
equatorial F region vertical drift model, J. Geophys. Res., 104, Watermann, J., G. S. Bust, J. P. Thayer, T. Neubert, and C. Coker
6829 – 6842. (2002), Mapping plasma structures in the high-latitude iono-
Scherliess, L., R. W. Schunk, J. J. Sojka, and D. C. Thompson sphere using beacon satellite, incoherent scatter radar and
(2004), Development of a physics-based reduced state Kalman ground-based magnetometer observations, Ann. Geophys., 45,
filter for the ionosphere, Radio Sci., 39, RS1S04, doi:10.1029/ 177 – 189.
2002RS002797. Wilson, B. D., A. J. Mannucci, and C. D. Edwards (1995), Sub-
Schreiner, W. S., S. V. Sokolovskiy, C. Rocken, and D. C. Hunt daily Northern Hemisphere ionospheric maps using an extensive
(1999), Analysis and validation of GPS/MET radio occultation network of GPS receivers, Radio Sci., 30, 639 – 648.
data in the ionosphere, Radio Sci., 34, 949 – 966. Xu, J. S., S. Y. Ma, X. B. Wu, K. H. Lin, and K. C. Yeh (2000a),
Schunk, R. W., J. J. Sojka, and J. V. Eccles (1997), Expanded Low-latitudinal ionospheric effects during a moderate storm by
capabilities for the ionospheric forecast model, Rep. AFRLVS- tomographic imaging, Chin. J. Geophys., 43, 145 – 151.
HA-TR-98 – 0001, 1 – 142, Space Veh. Dir., Hanscom Air Force Xu, J. S., X. B. Wu, and S. Y. Ma (2000b), Tomographic
Base, Mass. imaging of ionospheric structures and disturbances in the re-
Schunk, R. W., et al. (2004), Global Assimilation of Ionospheric gion of East-Asian equatorial anomaly, Sci. China, Ser. E,
Measurements (GAIM), Radio Sci., 39, RS1S02, doi:10.1029/ 43(4), 395 – 404.
2002RS002794. Yeh, K. C., and T. D. Raymund (1991), Limitations of ionospheric
Schunk, R. W., L. Scherliess, J. J. Sojka, D. C. Thompson, and imaging by tomography, Radio Sci., 26, 1361 – 1380.
L. Zhu (2005a), An operational data assimilation model of the Yeh, K. C., S. J. Franke, E. S. Andreeva, and V. E. Kunitsyn
global ionosphere, pp. 512 – 581, Ionospheric Effects Sympo- (2001), An investigation of motions of the equatorial anomaly
sium Proceedings, Natl. Tech. Info. Serv., Springfield, Va. crest, Geophys. Res. Lett., 28, 4517 – 4520.
Schunk, R., L. Scherliess, J. J. Sojka, D. C. Thompson, and L. Zhu Yin, P., and C. N. Mitchell (2005), Use of radio-occultation data for
(2005b), Ionospheric weather forecasting on the horizon, Space ionospheric imaging during the April 2002 disturbances, GPS
Weather, 3, S08007, doi:10.1029/2004SW000138. Solutions, 9, 156 – 163.
Sims, R. W., S. E. Pryse, and W. F. Denig (2005), Spatial structure Yin, P., C. N. Mitchell, P. S. J. Spencer, and J. C. Foster (2004),
of summertime ionospheric plasma near magnetic noon, Ann. Ionospheric electron concentration imaging using GPS over the
Geophys., 23, 25 – 37. USA during the storm of July 2000, Geophys. Res. Lett., 31,
Stolle, C., S. Schluter, S. Heise, C. Jacobi, N. Jakowski, S. Friedel, L12806, doi:10.1029/2004GL019899.
D. Kürschner, and H. Lühr (2005), GPS ionospheric imaging of
22 of 23
Yin, P., C. N. Mitchell, and G. S. Bust (2006), Observations of the Yizengaw, E., M. B. Moldwin, A. Komjathy, and A. J. Mannucci
F region height redistribution in the storm-time ionosphere over (2006), Unusual topside ionospheric density response to the No-
Europe and the USA using GPS, Geophys. Res. Lett., 33, vember 2003 superstorm, J. Geophys. Res., 111, A02308,
L18803, doi:10.1029/2006GL027125. doi:10.1029/2005JA011433.
Yizengaw, E., and M. B. Moldwin (2005), The altitude extension
of the mid-latitude trough and its correlation with plasmapause

position, Geophys. Res. Lett., 32, L09105, doi:10.1029/ G. S. Bust, Atmospheric and Space Technology Research Associates,
2005GL022854. 11118 Quail Pass, San Antonio, TX 78249, USA. (gbust@astraspace.
net)
C. N. Mitchell, Department of Electronic and Electrical Engineering,
University of Bath, Bath BA2 7AY, UK.
23 of 23

Bust 2008

Uploaded by

Document Informationclick to expand document information

Copyright:

Available Formats

Bust 2008

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Bust 2008

Uploaded by

Copyright:

Available Formats

HISTORY, CURRENT STATE, AND FUTURE

DIRECTIONS OF IONOSPHERIC IMAGING

Figure 1. USU GAIM Image [from Scherliess et al., 2004].

Figure 2. USC/JPL GAIM Image [from Pi et al., 2003].

communications. The electron density at a given time

You might also like