IEEE - Aerospace and Electronic Systems - February 2022 - 12

Comparison of Atmospheric Mass Density Models Using a New Data Source: COSMIC Satellite Ephemerides
IRI model by providing the sunspot number or the F10.7
index as inputs (otherwise these are determined from historical
data). Similar to the AMD models discussed, thus,
far, the IRI is an empirical model, representing the syntheses
of data from a number of sources including ionosondes,
incoherent scatter radar, in situ space missions,
GNSS satellites, and others. The main physical parameters
that the IRI models output are the electron density and
temperature, ion temperature, and densities (including
compositions), the total electron content, and the speed of
the equatorial vertical ion drift [39].
Other output parameters of the IRI are based on applications
and how the ionosphere is measured. Ionospheric
measurement began with ionosondes and continues to be
important today. Therefore, frequencies and heights at
which radio waves reflect off the ionosphere are of interest.
The parameters that the IRI models include the heights
above the Earth's surface, reflective frequencies, and densities
of the maximum electron density for each ionospheric
layer (F2, F1, E, D), the highest frequency
(refracted in the ionosphere) that can be received at a distance
of 3000 km, the probability of spreading in the ionogram
signal, etc [38]. Since the introduction of IRI-2012,
space weather effects have also been modeled. Predictions
are now available on the change in the maximum reflection
frequency of the E and F2 layers due to ionospheric
storms (quiet versus storm time values) [39].
The IRI ionospheric model has undergone a number of
versions since its inception in the 1960s with the tables of
IRI-75 giving way to computer-based models from IRI-86
and eventually going online in IRI-95 [39]. The latest version
is the IRI-2016 model. However, even for a given version,
there are a number of updates that occur-for example as of
this work, the IRI-2016 model had implemented 24 updates
since its release. We have used the seventh update dated January
2, 2016 in the proposedmodel. Updates can be as small as
allowing the ability for users to manually choose parameters
whereas previously they were automatically determined
(such as the B0 and B1 ionospheric parameters introduced in
updates 8 and 13) or the introduction of new models for ionospheric
values (such as the different hmF2 models described
in this article). These can have small but noticeable differences
in the outputs, and therefore, the version and date oflatest
update are necessary ifIRI results are to be reproduced by the
readers.
IRI OPTIONS IN THE PROPOSED MODEL
A consideration within the IRI model is the height of the
ionospheric F2 layer where the electron density is a maximum
(hmF2). The two available models are described
in [40] and [41]. The former uses a large dataset of radio
occultation measurements from satellite missions CHAMP
(2001-2008), GRACE (2007-2011), and COSMIC (2006-
12
2012) as well as ionospheric sounding data from 62 Digisonde
ionosondes (1987-2012) while the later uses data
from 26 digisonde stations in the Global Ionospheric Radio
Observatory network (1998-2006) [42]. The empirical
measurements in both models were used as a basis for predicting
the values of hmF2 for different times and locations
in the ionosphere - Shubin [40] determined the
median value while Altadill et al. [41] determined the
mean [42]. The predictions for hmF2 in [40] were used in
the proposed model.
The IRI also has two models available for the critical
frequency of the ionosphere's F2 layer, one developed by
the Comite Consultatif International des Radiocommunications
[43] and the other by the International Union of
Radio Science (or URSI after the French acronym for
Union Internationale de Radio-Scientifique) Working
Group G.5 [44]. These are based upon averages taken
from ionosonde measurements over different times of day
and days of the year. The URSI version has been used in
the proposed model's implementation.
Therefore, ignoring the contribution of ions, the proposed
model would provide the same AMD predictions
as that of the MSISE90 model. Although the MSISE00
model includes " anomalous oxygen " as mentioned earlier,
other ionic species are not accounted for. The proposed
model on the other hand includes contributions
from all ions modeled under the IRI-2016 model,
including H+, He+, O+, O2+, NO+, and N+. This leads
to predictions different from both the MSISE90 and
MSISE00 models.
DATA RETRIEVAL AND SPACE WEATHER CONDITIONS
The aforementioned AMD models and the proposed
model are assessed via one-day OP of the COSMIC satellites.
The discrepancy between the COSMIC reference
ephemerides retrieved from CDAAC and the OP results
with each AMD model are calculated in terms of the three
dimensional (3-D) distance error and component along
each direction, especially the in-track direction.
Two datasets of forty day ephemerides were chosen
during solar maximum in 2014-2015 and solar minimum
in 2018-2019 for COSMIC 1 and COSMIC 6, respectively.
We present the data here in 10-day windows distributed
in each quarter of the year for each dataset (see
Figure 3). Note that the 10-day window may not be consecutive,
e.g., Day 275-287, due to unavailable ephemerides
or attitude profiles for COSMIC satellites in some days.
The average height ofeach day is also shown in the figure,
with values decreasing from 810 km to 806 km during
these periods. More specifically, the average height has a
decreasing tendency during 2014-2015 while it stays
more fluctuating during 2018-2019. Roughly speaking,
the COSMIC altitude decreases by 2km between two
IEEE A&E SYSTEMS MAGAZINE
FEBRUARY 2022

IEEE - Aerospace and Electronic Systems - February 2022

Table of Contents for the Digital Edition of IEEE - Aerospace and Electronic Systems - February 2022