Available online at www.sciencedirect.com
Advances in Space Research 44 (2009) 1309–1313
www.elsevier.com/locate/asr
Use of varying shell heights derived from ionosonde data
in calculating vertical total electron content (TEC) using
GPS – New method
Sajan C. Mushini a,*, P.T. Jayachandran a, R.B. Langley b, J.W. MacDougall c
b
a
Department of Physics, University of New Brunswick, Fredericton, NB, Canada E3B5A3
Department of Geodesy and Geomatics Engineering, University of New Brunswick, Fredericton, NB, Canada E3B 5A3
c
Department of Electrical Engineering, University of Western Ontario, London, Ont., Canada N6A 5B9
Received 3 April 2009; received in revised form 15 July 2009; accepted 20 July 2009
Abstract
The dispersive nature of the ionosphere makes it possible to measure its total electron content (TEC). Thus Global Positioning System, which uses dual-frequency radio signals, is an ideal system to measure TEC. When data from an ionosonde situated in polar region
was observed, the height of an approximated thin shell of electrons (shell height) used in GPS studies was seen not to be fixed but rather
changing with time. Here we introduce a new method in which we included the varying shell heights derived from the ionosonde to map
the slant total electron content from GPS to obtain a more precise vertical total electron content of the ionosphere contrary to some
previous methods which used fixed shell heights. In this paper we also compared the ionosonde derived TEC with the GPS derived vertical TEC (vTEC) values. These GPS vTEC values were obtained from GPS slant TEC (sTEC) measurements using both fixed shell
height and varying shell heights (from ionosonde measurements). For the polar regions, the varying shell height approach produced better results than the fixed shell height and compared to exponential function, Chapman function seems to be a better function to model the
topside ionosphere.
Ó 2009 COSPAR. Published by Elsevier Ltd. All rights reserved.
Keywords: Polar ionosphere; GPS; Ionosonde; Total electron content; Chapman function
1. Introduction
The Global Positioning System’s availability both spatially and temporarily and also its usage of dual-frequency
signals (F1 = 1.575 GHz and F2 = 1.228 GHz) make it a
very suitable system for ionospheric research. The phase
advance and the group delay in the GPS signals produced
by the ionosphere depend on its total electron content
(TEC). Using these advances and delays one can calculate
the total electron content along the path the ray has traveled from the satellite to the receiver through the ionosphere (Komjathy, 1997).
*
Corresponding author.
E-mail addresses: m4s86@unb.ca (S.C. Mushini), jaya@unb.ca (P.T.
Jayachandran), lang@unb.ca (R.B. Langley), jmacdoug@uwo.ca (J.W.
MacDougall).
The point at which the ray path intersects the ionosphere is called the ionospheric pierce point (IPP) and the
electron content derived from the advances and delays
along that ray path is assumed to be present at that point.
Some previous studies (Komjathy, 1997) have assumed the
ionosphere to be a thin layer around the earth at a height
called the shell height.
The shell height is defined as the height at which the
electrons are distributed equally below and above it, and
this height is close to the peak height of the F layer in
the ionosphere (Komjathy, 1997; Horvath and Crozier,
2007). In many previous studies (e.g. Komjathy, 1997),
the shell height is approximated to be 350 km. However,
for the polar regions the fixed shell height assumption
may not be valid since the polar ionosphere is very dynamic
because of the solar wind–magnetosphere–ionosphere
0273-1177/$36.00 Ó 2009 COSPAR. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.asr.2009.07.015
1310
S.C. Mushini et al. / Advances in Space Research 44 (2009) 1309–1313
interaction and presence of ionospheric structures such as
polar patches (MacDougall and Jayachandran, 2007).
The electron content derived at the Ionospheric Pierce
Point (IPP) is called the slant total electron content (sTEC)
since the ray path makes an angle with the vertical. This
sTEC is then mapped to the vertical using a mapping function to obtain the vertical total electron content (vTEC) at
the same height of the IPP (shell height).
The ionosonde is another device which is generally used
to profile the ionosphere. It is a vertical sounder of the ionosphere which sweeps over a frequency range of 1–20 MHz
and works on the principle of reflection (Davies, 1990).
From ionosonde vertical sounding profiles, one can obtain
the peak frequency and peak heights which correspond to
the plasma frequency and height of the maximum electron
density at that epoch, respectively. These peak heights
obtained from the ionosonde were used as shell heights
for mapping the slant TECs to the vertical in our analysis
of the GPS data. We used sun-fixed solar geomagnetic
coordinates in our calculations since the ionosphere varies
much more slowly in these coordinates (Bregstrand and
Haas, 2004).
We compared the TEC values obtained from ionosonde
profiles (bottomside + modeled topside) with the GPS
derived vTEC values. The topside profiles from the ionosonde were modeled using a Chapman and exponential
functions (Chapman, 1931; Davies, 1990).
2. Data analysis
In this study, we have used GPS measurements from
Resolute Bay (74°410 5100 N, 94°490 5600 W geog), an International GNSS Service (IGS) site which is in the polar cap.
We have used data from three days in 2006 (11th November, 12th November, and 15th November 2006). The IGS
provides 30-s dual-frequency RINEX observational files.
These observational files contain information about range
measured in meters on P1, P2 (P-code pseudo-ranges on
L1, L2 frequencies, respectively) and C1 (C/A code
pseudo-range on L1). They also have phase information
(L1 and L2) measured in cycles on L1 and L2 frequencies.
We calculated the slant absolute total electron content
(SATEC) from pseudo-ranges and slant relative total electron content (SRTEC) from the carrier phases using Eq. (1)
and (2), respectively (Horvath and Crozier, 2007).
P2
P1
ð1Þ
2:852 10þ9 ;
SATEC ¼
c
c
60
ð2Þ
L1 ð2:3247Þ;
SRTEC ¼ L2
77
The relative slant total electron content is relative in the
sense that there is an ambiguity in the phase which does not
give an absolute value but it is very precise. The absolute
total electron content from the pseudo-range is absolute
but noisy. To obtain a better result we combined them
using a method called phase leveling (Komjathy, 1997;
Horvath and Crozier, 2007). To minimize the multipath
effect, we have used the phase and pseudo-range values
which correspond to ±15° from highest elevation angle
of the respective satellite pass to obtain the mean difference
between the phase and pseudo-range values. This bias was
then used to level the whole satellite pass. Since GPS hardware have their inherent delays (satellite bias and receiver
bias), they should also be taken into consideration (Komjathy, 1997; Horvath and Crozier, 2007).
We obtained these bias values from the University of
Bern
website
(http://www.aiub-download.unibe.ch/
CODE/) and combined them with our slant absolute total
electron content values before we did phase leveling (Horvath and Crozier, 2007). This phase leveled slant total electron content is called the slant total electron content. While
analyzing the data, cycle slips, where the phase is lost for a
certain period of time, were checked for and corrected
before phase leveling was done. After obtaining the slant
total electron content, we used the mapping function,
M(E) (Eq. (3)) (Komjathy, 1997), to obtain the vertical
slant total electron content at the ionospheric pierce point
at the shell heights obtained from ionosonde measurements. We then mapped the vertical TEC (vTEC) from
the pierce point to the point above the receiver using the
angle between them. Later, we averaged the vertical total
electron content from each satellite to obtain the averaged
vertical total electron content. This averaging would
smooth out variations caused by horizontal gradients in
the ionosphere. These horizontal gradients in the ionosphere are assumed to be small for a particular site and
epoch. The elevation angles (E) between the receiver and
satellites were obtained from the GPS Toolkit designed
by the University of Texas, Austin (Gaussiran et al, 2004).
MðEÞ ¼
1
h
i ;
e
cos arcsin cosðEÞ ReRþh
ð3Þ
where Re is the radius of earth and h is the shell height
(Komjathy, 1997).
In our study we also used the Canadian Advanced Digital Ionsonde (CADI) installed at Resolute Bay. For the
details of the CADI system please see MacDougall and
Jayachandran (2007). Given the dispersive nature of the
ionosphere, the heights we obtained from the ionosonde
are called the virtual heights. To obtain the real heights
from these virtual heights, one has to use numerical methods and we have used a commonly used method called the
Polynomial Analysis (POLAN) method (Titheridge, 1985).
CADI can provide peak height of F layer/shell height every
1 min. Fig. 1 shows the distribution of shell heights for the
three days we have used for this study. One can see that this
height varied between 240 km and 380 km during the time
period used in this study. So using a fixed shell height will
obviously introduce some errors in the vTEC calculations.
Instead of using a fixed shell height we have incorporated
the peak height obtained from ionosonde as the shell height
1311
S.C. Mushini et al. / Advances in Space Research 44 (2009) 1309–1313
0.3
2000
(a)
0.2
TECU(FIX-VARY)
TOTAL
1500
1000
0.1
0.0
-0.1
500
-0.2
200
250
0
300
350
400
450
SHELL HEIGHT(Km)
200 220 240 260 280 300 320 340 360 380 400
SHELL HEIGHTS (Km)
0.3
(b)
Fig. 1. Distribution of shell heights derived from ionosonde at Resolute
Bay on 11th, 12th, 15th November 2006. It was observed that the shell
heights varied from 250 km to 400 km. Most shell heights were observed in
the range 280–320 km. Bin value of 20 km was used in the figure.
TECU(FIX-VARY)
in Eq. (3) for the GPS vTEC calculations for all satellites as
previously mentioned.
Ionosondes can also be used to estimate the TEC in the
ionosphere. However, the ionosonde gives only bottomside
profiles i.e. profiles only up to the peak of the F layer. So, if
one wants to estimate the TEC of the ionosphere, a topside
profile of the ionosphere has to be included in the calculation. In this study, the topside profiles were modeled using
one of two mathematical functions (exponential and Chapman) (Chapman, 1931; Davies, 1990; Huang and Reinisch,
2001). The combined profiles (bottomside + topside) were
then integrated with height to get an estimate of the total
TEC of the ionosphere (Huang and Reinisch, 2001). We
then compared the TEC values obtained from these profiles
with the GPS derived vTEC values.
0.2
0.1
0.0
-0.1
-0.2
0
2
4
6
8
10
PEAK FREQUENCY(MHz)
Fig. 2. The difference between the total electron contents derived from
GPS using a fixed shell height analysis with shell height at 350 km and
using varying shell heights analysis is plotted against (a) shell height and
(b) peak frequency. Data were obtained at Resolute Bay on 11th, 12th,
15th November 2006. The varying shell heights were obtained from
ionosonde at Resolute Bay. The difference is observed to increase with the
electron content in the ionosphere (1 TECU = 1016 electrons/m2).
3. Results
Fig. 2(a) and (b) shows the scatter plots where the difference between the vTEC obtained using a fixed shell height
at 350 km and varying shell heights obtained from the ionosonde is plotted against the shell height and peak frequency, respectively. In Fig. 2(a), the maximum difference
between the vertical electron content obtained using a fixed
shell height and by using a varying shell height is around
0.3 TECU. It is also observed that a positive difference is
observed when we have shell heights less than 350 km
and a negative difference when the shell heights are more
than 350 km. It is to be noted in Fig. 2(b) that the difference increases proportionally with the peak frequency
which is directly proportional to the total electron content.
This indicates that on an active ionosphere day when the
ionosphere varies constantly, using a fixed shell height to
calculate the TEC values from GPS is not recommended.
As mentioned earlier, we modeled the topside ionosphere using two different functions, exponential function
and Chapman function. Then we combined the topside
with the bottomside profiles and integrated it to obtain
the TEC. Fig. 3(a) and (b) shows the scatter plots between
the ionosonde derived TEC (Chapman and exponential,
respectively) and the GPS vTEC (using ionosonde derived
shell heights) for the three days chosen. Agreement
between TEC estimates from GPS and Ionosonde (Chapman) is evident in the Fig. 3(a) (0.87 correlation coefficient
and 0.89 as slope of the best fit line). As from Fig. 3(b), we
saw that the exponential function derived TEC have much
1312
S.C. Mushini et al. / Advances in Space Research 44 (2009) 1309–1313
(a)
coefficients were slightly lower than ones obtained using
varying shell model. This indicates that the approximation
of using the shell height fixed at 350 km is not an accurate
but sufficient approximation for the polar region. As we are
slowly heading towards the solar maximum, the ionosphere
will become much more active and any high precision ionospheric studies should take account of the varying shell
height into their analysis.
4. Discussion and conclusions
(b)
Fig. 3. The correlation between the GPS derived TEC using varying shell
heights and the total electron content obtained using ionosonde for (a)
bottomside + Chapman topside model, (b) bottomside + exponential
topside model. The shift/bias of 0.28 TECU (exponential) and 0.36
TECU (Chapman) that were observed in the plots could be attributed to
receiver bias estimation, mapping functions used and also to the topside
profile model used. Chapman function (slope of the best fit line for
Chapman model derived TEC = 0.89) seems to be a better function to
model the polar topside ionosphere compared to exponential function
(slope of the best fit line for exponential derived TEC = 1.5). Data were
obtained at Resolute Bay on 11th, 12th, 15th November 2006 (1
TECU = 1016 electrons/m2).
smaller values than the GPS derived TEC (0.87 correlation
coefficient and 1.5 as slope of the best fit line). So Chapman
function (slope of the best fit line for Chapman model
derived TEC = 0.89) seems to be a better function to model
the polar topside ionosphere compared to exponential
function (slope of the best fit line for exponential derived
TEC = 1.5).We are in the process of doing a more elaborate study on the validity of different topside models for
the polar region. We have also correlated (figures not
shown) ionosonde TEC (Chapman and exponential) and
GPS vTEC using fixed shell height and the correlation
The ionosphere is a very dynamic system which changes
both spatially and temporarily. Most past studies assumed
that electron density of the ionosphere peaks at a constant
height (350 km) and some studies took varying shell
heights in to consideration but these shell heights were
obtained from ionosphere models (Komjathy and Langley,
1996). Our ionosonde data analysis show shell heights
which vary from 250 km to 400 km (Fig. 1). This we think
is a significant change in the ionosphere which should be
taken into account. It is seen in Fig. 2(a) that the difference
between the electron contents derived from using varying
shell heights derived from ionosonde data and a fixed shell
height at 350 km, increases as the electron content in the
ionosphere increases. This shows that if there is an increase
in electron content due to solar activities, the fixed shell
method is not accurate enough.
From the scatter plots (Fig. 3(a) and (b)) between the
total electron content derived from GPS and total electron
content from the ionosonde for all 3 days, we observed that
the electron content derived from the ionosonde using a
Chapman function modeled topside profiles had a shift/
bias of 0.36 TECU with the GPS derived electron content
while the electron content derived from the ionosonde
using exponential function modeled topside profiles had a
shift/bias of 0.28 TECU with the GPS derived electron
content. At midlatitudes (McKinnell et al., 2007) such a
shift or bias in the scatter plot is usually attributed to
plasmasphere electron content in ionospheric studies. But
since our station is in the polar region, the plasmasphere’s
contribution is negligible.
As less than 5% is the error that can be attributed to ionosonde derived bottomside TEC (Huang and Reinisch,
2001), we think that most of this shift or bias observed in
our results is mainly due to the following reasons. The first
reason we think is the receiver bias. It has to be noted that
receiver biases which were used were monthly averages.
Recent studies also show that for higher latitudes above
60°N, there is no proven method which would give us an
accurate receiver bias (Rideout and Coster, 2006). The second reason we think that may affect the results is the mapping function. The general mapping function we used is
designed best for mid and low latitudes where the highest
elevation angle of the GPS satellites can be 90° (Komjathy,
1997). For higher latitudes this is not true since the satellite
elevation angle is never 90°. This shows that more studies
are required to design a mapping function for higher polar
S.C. Mushini et al. / Advances in Space Research 44 (2009) 1309–1313
latitudes and as well as for receiver bias calculations for
GPS receivers at high latitudes given the activity in the
higher latitude ionosphere. Precise ionospheric studies
should also include varying shell heights since the ionosphere is not constant in time. The third reason we think
is due to the topside model used for ionosonde analysis.
Chapman function seems to be a better function compared
to exponential function to model the topside ionosphere
for polar latitudes. One still has to study different topside
models to find out a model which would give a better result
for polar ionosphere given its variability.
Acknowledgments
We thank the Natural Sciences and Engineering Research Council (NSERC). Operation of CADIs is done in
collaboration with the Canadian Space Agency (CSA).
We also thank IGS for providing us with GPS data.
References
Bregstrand, S., Haas, R. Comparison of ionospheric activity derived from
GPS and different VLBI networks. In: IVS General Meeting Proceedings, pp. 447–451, 2004.
Chapman, S. The absorption and dissociative or ionizing effect on
monochromatic radiation in an atmosphere on a rotating earth. Proc.
Phys. Soc. 43 (1), 26–45, 1931.
1313
Davies, K. Ionospheric Radio IEE Electromagnetic Waves Series 31. Peter
Peregrinus Ltd., 1990.
Gaussiran, T., Munton, D., Harris, B. et al. An open source toolkit for
GPS processing, total electron content effects, measurements and
modeling. In: Proceedings of the International Beacon Symposium,
Trieste, Italy, 2004.
Horvath, I., Crozier, S. Software developed for obtaining GPS-derived
total electron content values. Radio Sci. 42, doi:10.1029/
2006RS003452, 2007.
Huang, X., Reinisch, B.W. Vertical electron content from ionograms in
real time. Radio Sci. 36 (2), 335–342, 2001.
Komjathy, A., Langley, R.B. The effect of shell height on high precision
ionospheric modelling using GPS. In: Proceedings of the 1996 IGS
Workshop, pp. 193–203, 1996.
Komjathy, A. Global ionospheric total electron content mapping using
the Global Positioning System, Ph.D. Dissertation, Department of
Geodesy and Geomatics Engineering Technical Report No.188,
University of New Brunswick, Fredericton, New Brunswick, Canada,
1997.
MacDougall, J., Jayachandran, P.T. Polar patches: auroral zone precipitation effects. J. Geophys. Res. 112 (A5), A05312, doi:10.1029/
2006JA011930, 2007.
McKinnell, L.-A., Opperman, B., Cilliers, P.J. GPS TEC and ionosonde
TEC over Grahamstown, South Africa: first comparisons. Adv. Space
Res. 39, 816–820, 2007.
Rideout, W., Coster, A. Automated GPS processing for global total
electron content data. GPS Solutions 10 (3), doi:10.1007/s10291-0060029-5, 2006.
Titheridge, J.E. Ionogram analysis: least squares fitting of a Chapmanlayer peak. Radio Sci. 20 (2), 247–256, 1985.