GIGeoscientific Instrumentation, Methods and Data SystemsGIGeosci. Instrum. Method. Data Syst.2193-0864Copernicus PublicationsGöttingen, Germany10.5194/gi-5-263-2016Sodankylä ionospheric tomography data set 2003–2014NorbergJohannesjohannes.norberg@fmi.fiRoininenLassihttps://orcid.org/0000-0002-7014-6684KeroAnttiRaitaTeroUlichThomasMarkkanenMarkkuJuusolaLiisahttps://orcid.org/0000-0003-0864-5949KauristieKirstiFinnish Meteorological Institute, Helsinki, FinlandSodankylä Geophysical Observatory, University of Oulu, Sodankylä, FinlandDepartment of Mathematics, Tallinn University of Technology, Tallinn, EstoniaEigenor Corporation, Sodankylä, FinlandJohannes Norberg (johannes.norberg@fmi.fi)1July20165126327016November20154December201515April20162May2016This work is licensed under a Creative Commons Attribution 3.0 Unported License. To view a copy of this license, visit http://creativecommons.org/licenses/by/3.0/This article is available from https://gi.copernicus.org/articles/5/263/2016/gi-5-263-2016.htmlThe full text article is available as a PDF file from https://gi.copernicus.org/articles/5/263/2016/gi-5-263-2016.pdf
Sodankylä Geophysical Observatory has been operating a receiver network
for ionospheric tomography and collecting the produced data since 2003. The
collected data set consists of phase difference curves measured from COSMOS
navigation satellites from the Russian Parus network (Wood and Perry, 1980)
and tomographic electron density reconstructions obtained from these
measurements. In this study vertical total electron content (VTEC) values are
integrated from the reconstructed electron densities to make a qualitative
and quantitative analysis to validate the long-term performance of the
tomographic system. During the observation period, 2003–2014, there were
three to five operational stations at the Fennoscandia sector. Altogether the
analysis consists of around 66 000 overflights, but to ensure the quality of
the reconstructions, the examination is limited to cases with descending
(north to south) overflights and maximum elevation over 60∘. These
constraints limit the number of overflights to around 10 000. Based on this
data set, one solar cycle of ionospheric VTEC estimates is constructed. The
measurements are compared against the International Reference Ionosphere
(IRI)-2012 model, F10.7 solar flux index and sunspot number data.
Qualitatively the tomographic VTEC estimate corresponds to reference data
very well, but the IRI-2012 model results are on average 40 % higher than
that of the tomographic results.
Introduction
The use of tomographic methods for ionospheric research was
first suggested by . In ionospheric tomography with
low-Earth-orbit (LEO) satellites, the objective is to reconstruct the
ionospheric electron density in a two-, three- or four-dimensional domain
from ground-based measurements of beacon satellite radio signals. The
measured quantity is the phase shift of the transmitted radio signal. The
phase shift is proportional to the integrated number density of free
electrons along the signal path; hence, the measurements can be modelled as
line integrals of ionospheric electron density. As the measurement geometry
cannot provide horizontal ray directions, the information provided on
vertical structures is poor. This results in a limited angle tomography
inverse problem, which requires some regularisation scheme to stabilise the
problem. The method operated by the Sodankylä Geophysical Observatory
(SGO), is carried out within the framework of Bayesian statistical inverse
problems. The current method is reported by .
Stabilisation of the inverse problem is given with first-order difference
priors with a Chapman profile used in weighting the variances in the
altitude. In a more recent analysis development, a similar framework has been
used with Gaussian–Markov random field approximations for proper prior
covariance structures . provide a good
overview on other commonly used methods and on overall development of the
topic.
Current locations of SGO and FMI tomography chains.
SGO has been producing ionospheric tomography measurements operationally
since 2003. For the observation period 2003–2014, the data set consists of
around one solar cycle of measurements. SGO's measurements are based on
Russian polar orbiting COSMOS satellites equipped with
dual-frequency 150/400 MHz beacon transmitters. The geographical locations
of the SGO chain are plotted in black triangles in Fig. .
Recently, the Finnish Meteorological Institute (FMI), in collaboration with
SGO, has installed six additional stations in the region allowing the
observation area to expand from Tartu, Estonia to Longyearbyen, Svalbard,
Norway, shown as black circles in Fig. . These
TomoScand stations
are able to receive signals from any beacon satellite transmitting at
dual-frequency 150/400MHz. For example the
CASSIOPE/e-POP satellite mission has provided one new transmitter. A similar
chain, operated by Polar Geophysical Institute, is on the Kolan peninsula and
Karelia in north-west Russia . In this specific study,
the SGO receiver chain and COSMOS satellites are considered.
The lack of horizontal measurements in ionospheric tomography causes
well-known problems to reconstructions, especially when steep vertical
gradients are involved. However, it was reported by that
in the SGO algorithm, the local overestimations are usually compensated with
local underestimations elsewhere, and vice versa; e.g. overestimation in
layer thickness leads to reduced peak density. Hence, the solution for the
absolute level, i.e. the total electron content (TEC) of relative electron
density measurements is more stable than the actual profiles. To characterise
the long-term trends in the ionosphere the TEC measurements are integrated
vertically (vertical total electron content, VTEC) over each receiver
station. Compressing the data to an ionospheric VTEC value results in a more
robust statistic and allows for more straightforward visualisation.
Typical choices for VTEC validation are GPS measurements. These methods are
well-established, but due to the relatively low inclination of 55∘,
they do not measure directly the high-latitude ionosphere and provide
information essentially southwards from the receiver sites. Therefore, when
studying, e.g., travelling ionospheric disturbances at high latitudes, the
wave structures can be distinguished in instantaneous snapshots but following
their propagation is difficult . Moreover, the GPS
altitude is so high that the measurements include almost the entire
plasmasphere. Especially at night-time, the plasmaspheric contribution to
electron density can be significant. have performed similar
studies for the last two solar minimum periods with TOPEX and
JASON-1 satellites with orbital altitudes of 1337 km and
inclination of 66.038∘. From these satellites the VTEC can be solved
as a by-product of altimetry estimation, but only over areas above the ocean.
The specific objective of this paper is to investigate the solar cycle
variations in VTEC data obtained from the ionospheric tomography analysis. As reference material, the VTEC values derived from the IRI-2012 model (Bilitza et al. 2014), F10.7 solar flux index and sunspot number as extracted from NASA/GSFC's OMNI data set through OMNIWeb are used. By quantifying first the solar
cycle variations in the data, the opportunities to use similar data also in
studies on slower trends in high-latitude VTEC values can be considered.
Based on model predictions, increased levels of carbon dioxide and methane
are predicted to cool the thermosphere, and hence lower the so-called F2-peak
layer; see e.g. . Indirect measurements of the height (hmF2)
of the ionospheric F2 peak have been studied by a number of authors (see e.g.
). The
hmF2 values are derived empirically using routinely scaled ionograms. For
detailed information on the hmF2 estimation in the Sodankylä Geophysical
Observatory 1957–2014, see . In contrast to the hmF2
studies, the long-term TEC trends have not been widely studied. The first
study was carried out by , who considered GPS global and
regional trends (1995–2010) with the main finding of slow increasing trends
(0.6±0.3 total electron content unit (TECU) decade-1, TECU =1016 Ne m-2) in
the daily averaged global TEC values. One explanation for this trend could be
a reduction in the upper atmospheric recombination rates due to cooling in
the thermosphere. In the future, LEO tomography measurement could contribute
to refining the result of Lean et al. (2011) at high latitudes where GPS
measurements have accuracy problems due to oblique signal paths.
This paper is organised as follows: in Sect. the SGO
ionospheric tomographic data and the ionospheric VTEC estimation are
overviewed. Section includes a discussion of
the estimated VTEC and IRI-2012 model results. Section
concludes the study and provides some notes for future research.
Schematic plot of ionospheric tomography with LEO beacon
satellites.
Data and methodology
The ionospheric tomography reconstructions provided by SGO are solved in
two-dimensional latitude–altitude domain. The orbital altitude of COSMOS
satellites is approximately 1000 km and one such overflight takes
approximately 10 min. The measurement geometry then resembles the
schematic plot in Fig. . During the years 2003–2014, the number of
operational satellites varied between three and seven. Most of the time four
satellites have been providing transmissions. Figure shows an
example result sheet for one overflight from the SGO's tomography web archive
(http://www.sgo.fi/Data/Tomography/tomoArchive.php).
Tomographic reconstruction result from Sodankylä Geophysical
Observatory. The satellite trajectory, observed phase difference curves and
tomographic results with two different prior models. The receiver stations
are shown as coloured points. The origin of the kilometre axis is placed on Kokkola
station.
The number of satellite overflights over the whole observation
period with respect to minimum threshold elevation.
The inclination of COSMOS satellites is ∼ 83∘, i.e. compared to
a strictly polar orbit, the direction of the satellites is tilted slightly
eastwards. The geographical locations of the receiver stations are
Nurmijärvi (60.51∘ N, 24.65∘ E), Kokkola
(63.83∘ N, 23.16∘ E), Luleå (65.62∘ N,
22.14∘ E), Kiruna (67.85∘ N, 20.41∘ E) and
Kilpisjärvi (69.02∘ N, 20.86∘ E). At the beginning of the
observation period, a station in the European Incoherent Scatter Scientific Association (EISCAT) site in Tromsø, Norway,
was used, but this station was soon moved to Kilpisjärvi. The Nurmijärvi
measurements started in June 2004. The SGO receivers are installed along the
inclination angle so that the descending, i.e. southward, overflights are
somewhat parallel to the chain. Figure illustrates how in an
optimal case the descending satellite trajectory is aligned with the chain,
but also how the ascending, i.e. northward, overflights are almost
perpendicular to the chain. In Fig. the number of satellite
overflights are plotted against satellite elevation.
Magnetic local time mean values for tomographic VTEC, corresponding
IRI-2012 values and the difference between the two.
Magnetic local time mean values for summer time tomographic VTEC,
corresponding IRI-2012 values and the difference between the two.
Since in two-dimensional ionospheric tomography the longitudinal gradients
cannot be taken into account, the set of reconstructions is limited to
descending overflights with maximum elevation angles over 60∘, when
observed from the Kokkola station. To get an idea of the trajectories included in
analysis, two extremes of descending overflights with maximum elevation
angles close to 60∘ are shown in Fig. . Limiting the
original data of 66 000 overflights with these criteria results with a data
set of around 10 000 tomographic reconstructions, on average a little more than two
overflights per day. Instead of analysing complete two-dimensional
reconstructions the data are simplified to ionospheric VTEC measurements. The
VTEC is obtained by integrating the reconstructed electron densities above
each SGO receiver from ground level to satellite altitude.
The reference data sets of IRI-2012
(http://omniweb.gsfc.nasa.gov/vitmo/iri2012_vitmo.html), F10.7 and
sunspot (http://omniweb.gsfc.nasa.gov/form/dx1.html) number values are
all collected with the same time axis as the tomographic VTEC data. The
IRI-2012 VTEC values are integrated from the model results vertically between
the altitudes 0–1000 km at the receiver locations, similarly to the
tomographic VTEC.
Results and discussion
To characterise the data, they are first presented in
Figs. – as averaged VTEC values in magnetic
latitude and a magnetic local time (MLT) coordinate system. This is done
separately for the complete and seasonal data sets from summer, equinox and
winter. Winter is defined as one-third of a year centred around the winter
solstice. Summer starts one-third of a year after winter solstice and lasts
for one-third of a year. Everything else is defined as equinox. In
Figs. – first the data for tomographic then for
IRI-2012 VTEC values are shown. The third image illustrates the differences
between these two. In all Figs. – the relative
diurnal behaviour in VTEC values within different seasons are relatively
comparable between tomographic and IRI-2012 data. Both approaches show in
dayside VTEC values a dawn–dusk asymmetry with higher values on the dusk
side. This asymmetry is pronounced particularly during summer time in
Fig. , where according to IRI-2012 enhanced VTEC values extend to
pre-midnight hours. In the tomography results a similar trend is visible but
the extension of high VTEC to night-time hours beyond 18:00 MLT is missing. In
all seasons the electron densities are systematically higher in the IRI-2012
data, with the maximum difference close to 5 TECU. The difference plots show
that the differences in summer, in Fig. , are slightly smaller
than in equinox and winter seasons, in Figs. and . In
all Figs. –, at the magnetic local night-time, the
differences are in general somewhat smaller and both positive and negative.
Figures and indicate that in equinox and winter at
magnetic local night-time, the tomographic VTEC values at higher latitudes
are larger than the corresponding values from the IRI-2012 model.
Magnetic local time mean values for equinox time tomographic VTEC,
corresponding IRI-2012 values and the difference between the two.
Magnetic local time mean values for winter time tomographic VTEC,
corresponding IRI-2012 values and the difference between the
two.
In order to deduce whether the solar cycle can be observed from the data, in
Figs. and the data sets for the location of Kokkola
station are presented as time series for the complete period of 2003–2014.
Kokkola is chosen as the representative case as it is located close to the
centre of the tomographic domain and also provides good operational coverage
in the observation period. Furthermore, in illustrations of this kind, the
large-scale features are the same for all stations.
In Fig. the VTEC over Kokkola station is presented as monthly
means for each MLT hour. This is done for the whole observation period. In
the same figure the corresponding VTEC values from
the IRI-2012 model and the differences between them are presented again. First,
Fig. shows the nature of satellite availability. The period of
COSMOS satellites is 105 min, which produces a drift in daily times
of overflights. The images also show a maxima of VTEC in 2003 and in 2014.
Similarly to Figs. – the systematic differences
between tomographic and IRI-2012 VTEC values are visible. IRI-2012 VTEC
values are on average approximately 40 % higher than average tomographic
VTEC.
Monthly VTEC averages for each MLT hour over the Kokkola
station.
VTEC values over Kokkola averaged from 11:00 to 13:00 MLT vs.
corresponding IRI-2012 model values, sunspot number and solar flux index
F10.7.
The overestimation of high latitude Ne has been widely reported for different
versions of IRI model. reported that IRI-2001 overestimates
Ne at the peak altitude and above, especially in winter time compared to
incoherent scatter radar (ISR)
measurements. One of the main improvements for IRI-2007 was the topside Ne
modelling . compared the IRI-2007 model
results to orbital averages of CHAMP and
GRACE satellite measurements from 2000 to 2009, with the satellite height
range from 300 to 500 km. Especially during the solar minimum period
the overestimation was up to 60 %. The overestimation was concentrated on
the lower latitudes, but in a 20 % overestimation also
for a trough area was reported. utilised CHAMP and GRACE
satellite-based Ne measurements from 2005 to 2010.
These studies then suggested that despite the development, the modelling of
F peak and topside Ne still contains some problems. The improvements for
IRI-2012 were made for the thickness and the shape of the bottom-side F2
layer, as well as for the description of storm effects in the auroral
E region .
We have found it difficult to find a comprehensive account of the
different measurements used in the IRI model. In a
network of 27 ionosondes were used for the enhanced bottom-side modelling of
Ne. The closest ionosonde measurements to Fennoscandia in the network were
from Chilton, UK. In all, the network comprises two high latitude ionosondes,
both located in Greenland.
The Sodankylä tomographic set-up employs numerous measurements from the
high-latitude area. However, ionospheric tomographic inversion is well-known
to be an unstable inverse problem, and its performance, especially in
small-scale details in vertical structures, can be argued. The Bayesian
approach utilised here assumes zero electron density a
priori, and variations from zero background are then controlled with a
Chapman profile shaped standard deviation. The approach is hence more likely
to underestimate than overestimate the electron densities. Hence, if the
system has a bias, it would be most likely towards zero.
A discernible exception for the systematic difference between IRI-2012 and
tomographic VTEC is in the MLT hours around midnight of the year 2003 (blue
pixels in the lowest panel of Fig. ). It is known to be a year of
particularly strong space weather activity (see e.g. ).
Geomagnetic activity is strong particularly during 2–3 years after the solar
maxima . Both 2003 and 2014 are such years. Geomagnetic
activity is caused by processes in the night-side magnetosphere, which also
generate enhanced electron and proton precipitation into the ionosphere. The
impact of this precipitation is mostly visible in the E-layer densities. Our
results suggest that IRI-2012, as a statistical model, cannot describe these
special situations accurately, while the tomography inversion manages at
least partly to catch the altitude integrated impact from this precipitation.
In Fig. , in addition to IRI-2012, the tomographic data are
compared to the daily sunspot number and F10.7 solar flux index. Here only
the midday (11–13:00 MLT) VTEC from Kokkola station and corresponding
values from the IRI-2012 model were selected for the analysis. Due to a low
number of satellites, the measurement times are not uniformly represented;
i.e. some time slots are over-presented in the data. However, even despite
non-uniformity of the data, the solar cycle dependence is clear. The pattern
of tomographic VTEC corresponds essentially to reference data.
Conclusions
In this paper, the SGO's LEO-satellite
ionospheric tomography data set from the period of 2003–2014 is presented.
This data set covers approximately one solar cycle. The primary aim of this
paper is to see the solar cycle effect in the data. For this purpose,
the estimated VTEC values were used, which clearly exhibit similar solar
cycle-dependent features than in sunspot number and solar flux index F10.7
data.
The tomographic VTEC values also have a relative agreement with the
corresponding VTEC values obtained from the IRI-2012 model, but there is
a systematic difference between the two. The values based on the IRI-2012
model are on average 40 % higher than those of the tomographic results.
As an exception for the systematic difference, the results suggest that the
tomographic results capture geomagnetic night-time activity in increased VTEC
values.
Further studies are needed to resolve the reason for the significant
discrepancy between the tomographic and IRI-2012 VTEC values. However, as
mentioned in the Introduction, improved inversion methods for ionospheric
tomography are under development. The upgrading work includes better methods
to estimate the quality of inversion results. Therefore, we believe that
beacon-based tomography could be used more intensively in future research,
perhaps even in IRI validation and upgrading.
We suggest that the VTEC values from beacon-based tomographic inversion can
constitute a viable tool for studying long-term trends in the atmosphere. The
standard long-term trend is usually studied via the F2-layer peak. But as
this is a point value, the overall VTEC can import some extra information to
the analysis. However, as the data considered here consist of only one solar
cycle, it is practically impossible to say anything about the long-term
trends merely on the basis of the measured VTEC values. Hence, at least one
extra cycle for a proper long-term VTEC trend analysis is required, as well
as further studies to resolve the reason for the discrepancy with tomographic
and IRI-based VTEC values.
Data availability
The IRI-2012 electron density profiles are available from IRI-2012
(http://omniweb.gsfc.nasa.gov/vitmo/iri2012_vitmo.html). The F10.7
solar flux and sunspot number data are available from OMNI 2
(http:/omniweb.gsfc.nasa.gov/form/dx1.html). Ionospheric tomography
measurements and analysed data products used in this paper are available upon
request from the Sodankylä Geophysical Observatory. The quick-look plots
are available online
(http://www.sgo.fi/Data/Tomography/tomoArchive.php).
Acknowledgements
This work has been funded by European Regional Development Fund (Regional
Council of Lapland, decision number A70179), Academy of Finland (decision
number 287679) and European Research Council (ERC advanced grant 267700 –
Inverse problems).
Edited by: N. Partamies
ReferencesAltadill, D., Torta, J. M., and Blanch, E.: Proposal of new models of the
bottom-side B0 and B1 parameters for IRI, Adv. Space Res., 43,
1825–1834, 10.1016/j.asr.2008.08.014, 2008.
Austen, J. R., Franke, S. J., and Liu, C. H.: Ionospheric imaging using
computerized tomography, Radio Sci., 3, 299–307, 1988.Bilitza, D. and Reinisch, B.: The international reference ionosphere 2007:
Improvments and new parameters, Adv. Space Res., 42, 4, 599–609,
10.1016/j.asr.2007.07.048, 2008.Bilitza, D., Altadill, D., Zhang, Y., Mertens, C., Truhlik, V., Richards, P.,
McKinnell, L.-A., and Reinisch, B.: The international reference ionosphere
2012 – model of international collaboration, Journal of Space Weather and
Space Climate, 4, A07, 10.1051/swsc/2014004, 2014.
Bremer, J.: Ionospheric trends in mid-latitudes as a possible indicator of
the atmospheric greenhouse effect, J. Atmos. Terr. Phys., 54, 1505–1511,
1992.Bremer, J.: Trends in the ionospheric E and F regions over Europe, Ann.
Geophys., 16, 986–996, 10.1007/s00585-998-0986-9, 1998.Bust, G. S. and Mitchell, C. N.: History, current state, and future
directions of ionospheric imaging, Rev. Geophys., 46, RG1003,
10.1029/2006RG000212, 2008.
Cnossen, I. and Franzke, C.: The role of the Sun in long-term change in the
F2 peak ionosphere: new insights from EEMD and numerical modeling,
J. Geophys. Res.-Space, 119, 8610–8623, 2014.F10.7 and solar flux data: NASA/GSFC's Space Physics Data Facility's OMNIWeb
service, and OMNI data, Goddard Space Flight Center, NASA, available at:
http://omniweb.gsfc.nasa.gov/form/dx1.html, last access: 20 May 2016.Ionospheric tomographic Quicklook: Ionospheric tomography measurements and
analysed data products, Sodankylä Geophysical Observatory, Finland,
available at: http://www.sgo.fi/Data/Tomography/tomoArchive.php, last
access: 20 May 2016.IRI-2012: International Reference Ionosphere – IRI-2012 data, NASA,
available at: http://omniweb.gsfc.nasa.gov/vitmo/iri2012_vitmo.html,
last access: 20 May 2016.Jee, G., Lee, H.-B., Solomon, S. C., Cnossen, I., and Franzke, C.: Global
ionospheric total electron contents (TECs) during the last two solar minimum
periods, J. Geophys. Res.-Space, 119, 2090–2100, 10.1002/2013JA019407,
2014.Juusola, L., Viljanen, A., Van De Kamp, M., Tanskanen, E. I., Vanhamäki,
H., Partamies, N., and Kauristie, K.: High-latitude ionospheric equivalent
currents during strong space storms: regional perspective, Space Weather, 13,
49–60, 10.1002/2014SW001139, 2015.
Kunitsyn, V. E. and Tereshchenko, E. D.: Ionospheric Tomography,
Springer-Verlag, Berlin, Germany, 2003.Lean, J. L., Emmert, J. T., Picone, J. M., and Meier, R. R.: Global and
regional trends in ionospheric total electron content, J. Geophys. Res., 116,
A00H04, 10.1029/2010JA016378, 2011.Lühr, H. and Xiong, C.: IRI-2007 model overestimates electron density
during the 23/24 solar minimum, Geophys. Res. Lett., 37, L23101,
10.1029/2010GL045430, 2010.
Markkanen, M., Lehtinen, M., Nygrén, T., Pirttilä, J., Helenius, P.,
Vilenius, E., Tereshchenko, E. D., and Khudukon, B. Z.: Bayesian approach to
satellite radiotomography with applications in the Scandinavian sector, Ann.
Geophys., 13, 1277–1287, 1995.
Nevanlinna, H. and Pulkkinen, T. I..: Solar cycle correlations of substorm
and auroral frequency, Geophys. Res. Lett., 25, 3087–3090, 1998.Norberg, J., Roininen, L., Vierinen, J., Amm, O., McKay-Bukowski, D., and
Lehtinen, M.: Ionospheric tomography in Bayesian framework with Gaussian
Markov random field priors, Radio Sci., 50, 138–152,
10.1002/2014RS005434, 2015.
Nygrén, T., Markkanen, M., Lehtinen, M., Tereshchenko, E. D., Khudukon,
B. Z., Evstafiev, O. V., and Pollari, P.: Comparison of F-region electron
density observations by satellite radio tomography and incoherent scatter
methods, Ann. Geophys., 14, 1422–1428, 10.1007/s00585-996-1422-7,
1996.Roble, R. G. and Dickinson, R. E.: How will changes in carbon dioxide and
methane modify the mean structure of the mesosphere and thermosphere,
Geophys. Res. Lett., 16, 1441–1444, 10.1029/GL016i012p01441, 1989.Roininen, L., Laine, M., and Ulich, T.: Time-varying ionosonde trend: Case
study of Sodankylä hmF2 data 1957–2014, J. Geophys. Res. Space, 120,
6851–6859, 10.1002/2015JA021176, 2015.
Ulich, T. and Turunen, E.: Evidence for long-term cooling of the upper
atmosphere in ionosonde data, Geophys. Res. Lett., 24, 1103–1106, 1997.
Upadhyay, H. O. and Mahajan, K. K.: Atmospheric greenhouse effect and
ionospheric trends, Geophys. Res. Lett., 25, 3375–3378, 1998.van de Kamp, M., Pokhotelov, D., and Kauristie, K.: TID characterised using
joint effort of incoherent scatter radar and GPS, Ann. Geophys., 32,
1511–1532, 10.5194/angeo-32-1511-2014, 2014.Vierinen, J., Norberg, J., Lehtinen, M. S., Amm, O., Roininen, L.,
Väänänen, A., Erickson, P. J., and McKay-Bukowski, D.: Beacon satellite
receiver for ionospheric tomography, Radio Sci., 49, 1141–1152,
10.1002/2014RS005431, 2014.Wood, C. D. and Perry, G. E.: The Russian Satellite Navigation System,
Philos. T. Roy. Soc. A, 294, 307–315, 10.1098/rsta.1980.0037, 1980.Xiong, C., Lühr, C., and Ma, S. Y.: The subauroral electron density trough:
Comparison between satellite observations and IRI-2007 model estimates, Adv.
Space Res., 51, 536–544, 10.1016/j.asr.2011.09.021, 2011.Zhang, S-R., Holt, J. M., Bilitza, D. K., van Eyken, T., McCready, M.,
Amory-Mazaudier, C., Shoichiro, F., and Sulzer, M.: Multiple-site comparisons
between models of incoherent scatter radar and IRI, Adv. Space Res., 39,
910–917, 10.1016/j.asr.2006.05.027, 2007.