Volume 123, Issue 8 p. 7011-7028
Research Article
Free Access

Global Longitudinal Behavior of IRI Bottomside Profile Parameters From FORMOSAT-3/COSMIC Ionospheric Occultations

Sampad Kumar Panda

Corresponding Author

Sampad Kumar Panda

Department of Electronics and Communication Engineering, Koneru Lakshmaiah Education Foundation, Guntur, India

Correspondence to: S. K. Panda,


Search for more papers by this author
Haris Haralambous

Haris Haralambous

Department of Electrical Engineering, Frederick University, Nicosia, Cyprus

Search for more papers by this author
Venkatesh Kavutarapu

Venkatesh Kavutarapu

Laboratório de Física e Astronomia, Universidade do Vale do Paraiba, Saõ José dos Campos, Brazil

Search for more papers by this author
First published: 25 June 2018
Citations: 21


The bottomside thickness and shape (B0 and B1) are the critical key elements for depicting a realistic electron density profile in the International Reference Ionosphere (IRI) model. We investigated their longitudinal variability using a large database of FORMOSAT-3/COSMIC (Constellation Observing System for Meteorology, Ionosphere, and Climate) electron density profiles covering the period 2006–2015 from which these parameters are derived using a simple analytic Chapman least squares fitting. The parameters are then compared with the IRI-2012 submodels, that is, Bil-2000, Gul-1987, and ABT-2009 during different seasons and solar activity. Besides apparent adherence of B0 to the equatorial region, the opposite occurrence pattern of maximum B0 to that of NmF2 is realized, manifesting semiannual distribution and hemispheric asymmetry. Unlike NmF2, the presence of single anomaly crest in B0 in all seasons except March equinox and their leading local time stamps are among the key observations. In spite of regular representation of IRI submodels during equinox, the performance of Gul-1987 seems to be inferior during solstice with respect to geomagnetic field and equatorial electrodynamics, thereby depicting an erroneous distribution of B0. The striking wave-like feature in B0 longitudinal variability is also revealed indicating its possible connection with the nonmigrating diurnal/semidiurnal tides. Moreover, the solar activity effect on B0 appears to be relatively higher over the Asian and Pacific longitudes (60–150°E). Although speculations are made over B1 variability, its complete longitudinal characteristics are indistinguishable in our results, except the subordinate magnitude as compared to IRI estimations. The important findings from this study may reinforce the understanding of bottomside variability and future improvements in IRI.

Key Points

  • Longitudinal variability of B0 and B1 is investigated from FORMOSAT-3/COSMIC occultation profiles
  • Longitudinal wave-like feature in B0 indicates nonmigrating tidal influences with latitudinal anticorrelation connection with NmF2
  • Discrepancy among the bottomside submodels in IRI is noticed with the relatively improved performance in ABT-2009

1 Introduction

The vertical distribution of electron density (Ne) in the ionosphere is important for understanding and modeling the plasma temperature, composition, and dynamics in connection with radio communication and positioning applications. An accurate representation of height profiles Ne(h), particularly for the bottomside and topside ionosphere, provide key information for estimation and correction of transionospheric propagation delays introduced in Global Navigation Satellite System signals, satellite and radio communication, etc. (Chuo et al., 2013; Lei et al., 2005; L. Liu et al., 2008; Tulasi Ram, Su, Liu, Reinisch, & McKinnell, 2009). Typically, the topside is the region above the F2 layer peak height (hmF2) connecting the plasmasphere, whereas the bottomside or the lower ionosphere is the region below hmF2. What is more, the topside and bottomside profiles are important due to their respective approximate contributions of 75% and 25% in determining total electron content (TEC), a characteristics that defines the effect on transionospheric communication and radio propagation signals (Astafyeva et al., 2015; Belehaki & Tsagouri, 2002; Bidaine & Warnant, 2010; Zhu et al., 2016). Although the bottomside TEC contribution is less, unlike absolute changes in the topside the relative changes in the bottomside TEC to the quiet time reference level is often more or even behaves reversely in the low and middle latitudes, particularly during geomagnetically disturbed days (Kuai et al., 2017; Zhu et al., 2016). The highlighting feature of bottomside profile for which the present study aimed is that it is one of the key elements in fitting suitable polynomials for deriving a true vertical profile in the modeling of electron density distribution while estimating TEC in the most widely accepted global empirical International Reference Ionosphere (IRI) model (Venkatesh et al., 2014). This suggests that an inappropriate representation of the bottomside may cause discrepancies in the full specification of the electron density profile. In other way, it is also possible to determine the topside ionospheric profile from the bottomside expression, but any discrepancy in the bottomside profiling will eventually affect the topside (Kutiev et al., 2009; L. Liu et al., 2006; Reinisch et al., 2004). Moreover, recent studies of Batista et al. (2017) indicate that B0 could be a good proxy for the F3 layer occurrence in the topside ionosphere. As the global bottomside profile data are significantly more abundant than the topside, extended research must be conducted for refinement of key parameters in the standard specification of the model.

In IRI model, the bottomside electron density profile is expressed with reference to F2 region peak electron density (NmF2), its height (hmF2), and the bottomside thickness (B0) as well as shape (B1) parameters. Typically, B0 is the altitude difference between hmF2 and h0.24, and the latter is the height where electron density profile decreases to 0.24*NmF2. B1 describes the shape of the bottomside profile. The analytical function in IRI to describe the ionospheric electron density profile Ne(h) in the bottomside is as follows (Ramakrishnan & Rawer, 1972):

The present version of IRI (IRI-2016) encompasses three different submodel options for estimating B0 and B1 parameters, that is, Bil-2000, Gul-1987, and ABT-2009 (default) (Bilitza et al., 2014). Bil-2000 consists of a set of tabular values (B0 and B1) involving an interpolation scheme. In Gul-1987, B0 is based on half-density height, but B1 is adopted from Bil-2000 database (Gulyaeva, 1987). ABT-2009 is the recently introduced model for B0 and B1 and is claimed to be the most improved model for bottomside specification (Altadill et al., 2009). However, the disadvantage of the above models is that their reliability depends strongly on the underlying database containing records from ground-based ionosondes whose spatial and temporal extent is sparse in some regions (Bilitza et al., 2014). Reportedly, Bil-2000 is based on average noon and midnight data from several ionosonde recordings across low and middle latitudes. With a coverage of four different seasons under two solar activity levels, it claims sensible results in the equatorial ionosphere. However, Gul-1987 is established mainly with data from midlatitude stations and the estimations are based on observed correlation between hmF2 and h0.5 in terms of solar zenith angle and season. Comparative analysis of bottomside profiles from existing IRI-2012 have been performed with the corresponding values deduced from ionosonde and incoherent scatter database to validate and improve its predictability (Altadill et al., 2009; Bilitza et al., 2000; Gulyaeva, 1987; L. Liu, Wan, et al., 2010; McKinnell et al., 2009; Sethi & Mahajan, 2002; Sethi et al., 2009). They found marked discrepancies in the seasonal and solar activity trends of B0; notably, Bil-2000 provides relatively improved results during daytime whereas Gul-1987 performed better during nighttime. Concerning B1 variation, it is reported that Bil-2000 performs relatively better than Gul-1987 though the solar cycle variation is not satisfactory (Bilitza et al., 2014; Lei et al., 2004). However, recently Altadill et al. (2009) claimed that ABT-2009 model provides comparatively improved results than the other two models. They have taken an extended database of 27 ionosondes distributed globally during 1998–2006 and used a spherical harmonics analysis to estimate the bottomside variations with modip (modified dip) latitude, local time, month, and sunspot number. Using the present model, they achieved a fair improvement of 32% and 40% over Bil-2000 and Gul-1987, respectively. Regarding B1, they claim that their model is better up to 20% over the Bil-2000 model. However, the ABT-2009 model may not fully represent the longitudinal effects on the profile parameters requiring further investigation (L. Liu, Wan, et al., 2010).

During past decades, several probing techniques, such as incoherent scatter radars (ISR), ground-based ionosonde/digisonde and in situ rocket measurements as well as modeling practices have been performed at different parts of the globe for exploring the bottomside ionosphere profile and studying the discrepancies between the IRI submodels as well as from other alternative estimation methods (Adeniyi et al., 2008; Alazo-Cuartas & Radicella, 2017; Altadill et al., 2009; Bello et al., 2017; Bilitza et al., 2014; Chen et al., 2006; Chuo et al., 2011; 2013; Gulyaeva, 1987; Jamjareegulgarn et al., 2017; Kalita & Bhuyan, 2017; Lazo et al., 2007; Lei et al., 2004; McKinnell et al., 2009; Panda & Haralambous, 2017; Sethi & Mahajan, 2002; Venkatesh et al., 2014; Venkatesh & Fagundes, 2016; Zhang et al., 2008). However, these techniques are limited both spatially and temporally; hence, less research has been conducted so far to assess the longitudinal variation of bottomside parameters and their corresponding estimation from models. Moreover, the difficulty in modeling the well-established seasonal as well as solar cycle morphology of profile parameters over the equatorial and adjoining low latitudes is due to the limited data spatial coverage on a global scale. Hence, a good global spatiotemporal investigation of profile representation over the aforementioned region is certainly beneficial to related research activities (Lei et al., 2004). Fortunately, with the introduction of spaceborne Global Navigation Satellite System on board low Earth-orbiting (LEO) satellites, it has now become a powerful tool to remotely sense the ionospheric profile with a huge database of accumulated ionospheric radio occultation (IRO) electron density profiles on a daily basis (L. Liu et al., 2009). Hence, such electron density profiles retrieved from Global Positioning System (GPS) on board LEO satellites, such as Challenging Mini satellite Payload for Geophysical Research and Application, FORMOSAT-3/COSMIC (Formosa Satellite 3/Constellation Observing System for Meteorology, Ionosphere, and Climate), and other missions is considered as an important database for studying the global ionospheric profile structure and behavior in contrast to other traditional ground-based measurement techniques (Hu et al., 2014; L. Liu et al., 2008).

FORMOSAT-3/COSMIC is the most successful mission consisting of six identical micro LEO satellites, each loaded with an advanced dual-frequency GPS-Radio Occultation (RO) receiver among other payloads on board and has accumulated a huge database of atmospheric data, including three-dimensional profiles of temperature, humidity, and pressure, as well as electron density profiles of ionosphere in a cost-effective way (L.-C. Lee et al., 2000). The retrieved electron density profiles are accessible from the Taiwan Analysis Center for COSMIC (http://tacc.cwb.gov.tw/) and the COSMIC Data Analysis and Archive Center (CDAAC, http://cdaac-www.cosmic.ucar.edu/) archives of University Corporation for Atmospheric Research. Broadly, these ionospheric occultation profiles are retrieved from TEC along COSMIC occultation pathway using the Abel inversion technique, a detailed procedure of which can be found in Yue, Schreiner, Lei, Sokolovskiy, et al. (2010, and references therein).

In the present paper, we used COSMIC IRO electron density profiles (.ionprf) for years 2006–2015 from CDAAC reprocessed (COSMIC2013) and postprocessed (COSMIC) database. With this huge database of IRO profiles, we conducted an analysis starting from diurnal and seasonal variation of bottomside parameters (B0 and B1) to their global longitudinal distributions and eventually centered our studies across the equatorial and low-latitude region. Corresponding estimations from IRI-2012 model are also gathered by independently choosing all available bottomside model options. In a broad sense, we tried to investigate the apparent longitudinal structure of bottomside profile in the equatorial regions across different longitude sectors. The appealing findings from our studies are the complicated global longitudinal seasonal morphology of bottomside parameters and their occurrence patterns in the low latitudes. Very few studies are conducted with global database of IRO profiles to speculate a ubiquitous picture of bottomside profile parameters through diurnal, seasonal, and solar activity levels, pertaining to their longitudinal variability across the global equatorial and low-latitude region.

2 Data and Methods

For the present study, we obtained level-2 COSMIC ionospheric electron density profiles for the period 2006–2014 from the reprocessed database (COSMIC2013 database) and the rest of the profiles till 2015 are accessed from the postprocessed database (COSMIC database) at CDAAC of the University Corporation for Atmospheric Research (http://cdaac-www.cosmic.ucar.edu) constituting approximately a decade of study period (2006–2015). Initially, we processed a worldwide database of occultation profiles to get a clear idea on the global morphology of bottomside profile parameters and then narrowed down to the global equatorial and low-latitude zone owing to its importance for communication, navigation, and ionospheric research. From the large database of ionospheric occultations, we chose the profiles for the global equatorial and low-latitude region, with geomagnetic latitude ranging within ±20° and discarding profiles for days corresponding to a sum of geomagnetic activity indices (Kp) larger than 24. The database of Kp index is available at World Data Center for Geomagnetism, Kyoto (http://wdc.kugi.kyoto-u.ac.jp). Additionally, we restricted the peak height (hmF2) in the range between 160 and 600 km and the peak density (NmF2) between 1∗103 and 1∗107 (cm3) ensuring elimination of spurious profiles and forwarding for further processing (Hu et al., 2014; Yue, Schreiner, Lei, Rocken, et al., 2010). Using an α-Chapman profile function, we constructed each individual profile by applying least squares fitting in terms of equation 1 and by limiting the aforementioned altitude limits with a number of data points of more than eight, of which at least three points should be present below the peak(Hu et al., 2014; Lei et al., 2004, 2005; L. Liu, Wan, et al., 2010; Kelley et al., 2009; Yue, Schreiner, Lei, Rocken, et al., 2010). From the coefficients of the fitting function, the B0 and B1 parameters are determined. Further, to ensure reliability of the results, a compromised criterion of maximal altitude gap between original profile and fitted one has been fixed to be less than 20 km. For visual understanding, few examples of good and bad fittings from our database are shown in Figure 1. At this point, it should be highlighted that the derived peak parameters NmF2, hmF2, and bottomside parameters (B0 and B1) typically look genuine with most of the fitted profile peaks being around the maximum position of electron density and conquer well with the data (Fox, 1994; Hu et al., 2014; Lei et al., 2004). To ascertain the reliability of COSMIC results, we collected ionosonde profiles from six different low-latitude locations (i.e., Ascension Island, Fortaleza, Jicamarca, Kwajalein, Ramey, and Saoluis) at which the derived B0 parameters are compared with the corresponding values from COSMIC profiles with the collected data points within 1 hr of ionosonde soundings and a stretch of 1°E from ionosonde location. The comparative analysis in Figure 2 shows correlation coefficients (R2>0.9) and root-mean-square error (RMSE ranging from 4 to 14) between COSMIC and Ionosonde derived B0 parameters within the strongly acceptable limits. This validation experiment supports conclusive remarks on consistency of occultation data sets for ionospheric investigations. Reportedly, many past studies have also been conducted on validation of F layer peak parameters retrieved from COSMIC profiles with the globally distributed ionosondes/ISRs measurements substantiating their good agreements and usability for understanding of ionospheric parameters and modeling practices (Angling, 2008; Chuo et al., 2011; Ely et al., 2012; Garcia-Fernandez et al., 2005; Jakowski et al., 2002; Krankowski et al., 2011; Lei et al., 2007; G. Liu, Immel, et al., 2010; Schreiner et al., 2007; Scherliess et al., 2008; Tsai & Tsai, 2004; Xu et al., 2013; Yue, Schreiner, Lei, Rocken, et al., 2010; Yue, Schreiner, Lei, Sokolovskiy, et al., 2010; Zhang et al., 2014).

Details are in the caption following the image
Examples of few bad and good electron density altitude profiles obtained through least squares fittings with α-Chapman function.
Details are in the caption following the image
Scatterplots of COSMIC-fitted B0 and ionosonde-derived B0 parameters at six different locations across the equatorial and low-latitude region. The correlation coefficients (R) and root-mean-square error (RMSE) between them are presented for each location.

The respective parameters from the current version of IRI (IRI-2012) corresponding to each occultation point and its local solar time are obtained using all three bottomside modeling options, that is, Bil-2000, Gul-1987, and ABT-2009 (Bilitza et al., 2014). The hourly median values of B0 and B1 parameters for each season are determined considering the local solar time over all longitudes. The whole study period is categorized into three seasons (equinox: March, April, September, and October; summer: May to August; winter: November to February) under two broad solar activity levels, low and high solar activity (LSA and HSA), to analyze the seasonal as well as solar activity dependence over the equatorial region. Here it should be noted that the above seasons refer to the seasons in the Northern Hemisphere. Hence, it is obvious that the vernal equinox and summer solstice in the Northern Hemisphere surmise the respective occurrences of autumnal equinox and winter solstice in the Southern Hemisphere. To assess the longitudinal behavior of diurnal peak B0 and B1 parameters in a particular season and solar activity level, we consider only those ionospheric occultations occurring during local peak hours of the day (11:00 to 14:00 LT). The bottomside parameters corresponding to occultations within each voxel for a particular season are accumulated separately to determine the seasonal median values of B0 and B1 under both solar activity conditions. The data points are again grouped into voxels of 5° in longitude; values inside each bin are then used to obtain the median value for the longitude range.

3 Global Behavior of B0 Parameter

We investigated the global variation pattern of B0 parameter extracted from COSMIC electron density profiles as well as those estimated from IRI-2012 model by selecting each of the three bottomside submodel options: Bil-2000, Gul-1987, and ABT-2009. Here we concentrate only on the peak hours (11:00–14:00 LT) interval of the day to visualize overall daytime global variation pattern under different seasons and solar activity conditions.

3.1 B0 from COSMIC observations

Figure 3, depicts the global daytime B0 profile parameter for three different seasons (equinox, summer, and winter) under low and high solar activity conditions by determining the average values within the voxels of 2° × 5° (latitude × longitude) during 11:00–14:00 LT and by constraining the occultation profiles with certain criteria as indicated in section 2. The black contours in the middle of the maps highlight the geographical location of the magnetic equator (middle) and the average crest latitudes at ±20° geomagnetic latitudes in both hemispheres. Figure 3 clearly demonstrates centrally positioned elevated B0 values (around 220–270 km) straddling along the geomagnetic equator that is shifted northward during northern summer and southward during southern summer. However, across midlatitudes, B0 is restricted within a range of 100–150 km. Although its average amplitude is minimal during equinox, a hemispheric asymmetry is being noticed with relatively higher B0 values over the summer hemisphere than the winter hemisphere. However, enhanced B0 with approximately 200-km thickness is distinctly visible in the Southern Hemisphere during northern winter, especially below the African and Asian territory. Concerning solar activity dependence, B0 plots during HSA show slightly higher values than the LSA period, though the difference is minutely noticeable from the figure. However, it is for sure that the irregularity is minimum during the LSA which is later inferred from Figure 8. To obtain a clear idea on existing IRI model B0 climatology, we further studied the B0 parameter by considering all submodel options.

Details are in the caption following the image
Global COSMIC occultation profile-derived longitudinal B0 parameters during equinox, summer, and winter seasons under (a) LSA (2006–2010) and (b) HSA (2011–2015) conditions. COSMIC = Constellation Observing System for Meteorology, Ionosphere, and Climate; LSA = low solar activity; HSA = high solar activity.

3.2 B0 from IRI model

From Figures 4-6, we show the global daytime B0 (averages of 11:00–14:00 LT) estimated from Bil-2000, Gul-1987, and ABT-2009 submodels of IRI-2012 during equinox, summer, and winter seasons under (a) LSA (2006–2010) and (b) HSA (2011–2015) conditions. Similar to COSMIC B0 plots in Figure 3, the geomagnetic equator and average anomaly crest (20° magnetic latitudes) on both hemispheres are highlighted as black color contours at the center of the maps. At a first glance, it can be marked from Figure 4 that during equinox, the significance of B0 apparently adheres to the equatorial and low-latitude zone, whereas the rest of the latitudes manifest almost similar values. However, there are perceptible discrepancies among the submodels; the spread of B0 magnitude by Bil-2000 is limited to the geomagnetic equator and its nearer latitudes, more prominent B0 by Gul-1987 with a wider spread across the geomagnetic equator and moderate values with a slight northward shifting in the case of the ABT-2009 model. What is more, ABT-2009 exhibits a relatively higher B0 around higher latitudes than the middle latitude region with a magnitude difference of about 30 km.

Details are in the caption following the image
Global B0 estimated from Bil-2000, Gul-1987, and ABT-2009 submodels of IRI during equinox under (a) LSA (2006–2010) and (b) HSA (2011–2015) conditions. IRI = International Reference Ionosphere; LSA = low solar activity; HSA = high solar activity.
Details are in the caption following the image
Global B0 estimated from Bil-2000, Gul-1987, and ABT-2009 submodels of IRI during summer under (a) LSA (2006–2010) and (b) HSA (2011–2015) conditions. IRI = International Reference Ionosphere; LSA = low solar activity; HSA = high solar activity.
Details are in the caption following the image
Global B0 estimated from Bil-2000, Gul-1987, and ABT-2009 submodels of IRI during winter under (a) LSA (2006–2010) and (b) HSA (2011–2015) conditions. IRI = International Reference Ionosphere; LSA = low solar activity; HSA = high solar activity.

With further investigation, we revealed that B0 from ABT-2009 is higher at northern high latitudes during vernal equinox in the Northern Hemisphere, whereas southern high latitudes depict fairly higher B0 value during the vernal equinox in the Southern Hemisphere (not shown in figure). However, the susceptibility of high-latitude B0 values to the episode of equinoctial occurrence in the same hemisphere seems to be relatively prominent in the Northern Hemisphere. Moreover, there is an obvious increase in the overall magnitude of B0 from low to high solar activity at all locations irrespective of season. The other interesting feature is the representation of B0 in all the above submodels in IRI-2012 during summer and winter solstice. In summer, the equatorial B0 by Bil-2000 model is shifted northward retaining the shape of magnetic equator across the globe and with relatively higher magnitude in the Northern Hemisphere than the Southern Hemisphere. However, from Figures 5 and 6 it seems that the Gul-1987-derived B0 do not follow the control of geomagnetic field; rather it appears to distribute B0 with the highest values within 0° to 20°N geographic latitudes during northern summer and 0° to 20°S geographic latitudes during southern summer. The overall B0 (Gul-1987) magnitude difference between the two hemispheres during summer and winter solstice months are realized. In the case of ABT-2009, B0 is represented realistically emphasizing the location of geomagnetic equator across the continents and highlighting solar activity differences. During summer solstice (winter solstice), the ABT-2009 estimated B0 is highest at equator and the surrounding region, moderate at northern (southern) higher latitudes, and followed by relatively lower values at the northern (southern) midlatitudes by exhibiting the lowest value over the Southern (Northern) Hemisphere. In brief, it can be summarized that all three models represent the regular pattern of B0 in a similar way only during equinox, particularly over the equatorial region though the magnitudes show some disparities. However, the performance of Gul-1987 seems to be inferior depicting erroneous distribution of B0 parameters during solstices not taking into account the role of the geomagnetic field reference frame. Also, Bil-2000 model produces a very narrow band of higher B0 values populated over the equator and does not extend to the highest variable crest latitude, while the rest of the regions in both hemispheres depict almost similar magnitudes. Hence, we narrowed down the region under investigation to equatorial and low-latitude zone extending up to 20° on both hemispheres, but with an extensive analysis on bottomside thickness (B0) and the shape (B1) parameters along with F layer electron density (NmF2) to derive a meaningful expression for their longitudinal variability.

4 Equatorial and Low-Latitude Features of Bottomside Profile Parameters

We investigated the longitudinal variation of seasonal B0 and B1 parameters over the equatorial and low-latitude region on both hemispheres under LSA and HSA conditions. The study focuses only on the local time interval of 11:00–14:00 LT to probe the average daytime behavior of the global low-latitude bottomside parameters. Figure 8 illustrates average B0 deduced from COSMIC electron density profiles through least squares fitting and the corresponding values at the occultation points estimated from the IRI-2012 model using all three bottomside modeling options, that is, Bil-2000, Gul-1987 and ABT-2009. Clearly, it can be noted that daytime B0 follows a semiannual pattern, with relatively higher values in solstice and lower values in equinox. Hence, the occurrence pattern of maximum B0 is exactly opposite to that of NmF2 as realized from the literature (C. C. Lee et al., 2008; L. Liu et al., 2009; L. Liu, Wan, et al., 2010), acknowledging a thicker profile in solstice and thinner profile in equinox. To emphasize a valid anticorrelation between seasonal B0 and NmF2 irrespective of longitude, we performed a comparative study by taking two dimensional surface plots of both parameters with respect to local time (10:00 to 18:00 LT) and geomagnetic latitude (−20 to +20) as shown in Figure 7. It is confirmed from the seasonal plots that although obvious equatorial ionization anomaly (EIA) crests are manifested in both hemispheres (around ±10° to ±15° magnetic latitude) during all seasons irrespective of solar activity, B0 distribution rather perceived a moderate double hump structure relatively toward the equatorward latitude (5–10° geomagnetic latitude) only during the low solar activity equinox season. The high solar active equinox season evidenced a distinctly visible unique crest in B0 over the geomagnetic equator which is pushed 5–10° northward and southward during summer and winter seasons, respectively. Additionally, we noticed the local time difference of anomaly crest developments in NmF2 (late afternoon) and B0 (local noon). Our implication on anticorrelation nature among NmF2 and B0 is consistent with earlier comments of L. Liu et al., (2009) and L. Liu, Wan, et al. (2010). Moreover, evidences of opposite behavior between instant changes of hmF2 and NmF2 have been well documented in the literature through observations and model simulations in the lower latitudes (Emery et al., 1996; Gulyaeva, 2012, and references therein). The highly positive correlation between hmF2 and B0 further strengthen the speculations on existence of anticorrelations between the parameters (C.-C. Lee & Reinisch, 2007; C. C. Lee et al., 2008; L. Liu et al., 2006).

Details are in the caption following the image
Comparison of local time (10:00 to 18:00 LT) and geomagnetic latitude (−20° to +20°) variations of COSMIC-derived NmF2 and B0 parameters during equinox, summer and winter seasons under (a) low solar activity (LSA; 2006–2010) and (b) high solar activity (HSA; 2011–2015) conditions. COSMIC = Constellation Observing System for Meteorology, Ionosphere, and Climate.

Figure 7 also demonstrates the appearance of single/double humped structure in NmF2 and B0. Like EIA, we hereby demonstrate the abnormality in B0 as equatorial bottomside thickness anomaly. The double crest structure in B0 is fairly visible around 5–10° from geomagnetic equator only during the LSA equinox, whereas the HSA equinox demonstrates a distinct crest over the geomagnetic equator; summer and winter solstices also reveal solitary crests but apparently shifted (toward magnetic equator) southward and northward, respectively. The fitting explanation for the shifting in crests can be given through amplitudes of effective meridional wind (EMW) component that is southward during summer, northward during winter, and modest during equinox (Abdu, 2001; Balan et al., 1998; Batista et al., 2017; Panda et al., 2015; Titheridge, 1995). Under normal E × B vertical drift and usual photochemical and dynamical processes, the northward EMW during winter pushes the plasma upward along the magnetic field lines at the Southern Hemisphere preventing effective diffusion process and retaining part of the ionization at higher altitudes to result in higher B0 closer to the magnetic equator in the Southern Hemisphere. Oppositely, the southward EMW during summer pushes the plasma upward along the magnetic field lines at the Northern Hemisphere to manifest higher B0 nearer to the magnetic equator in the Northern Hemisphere (Balan et al., 1998; Batista et al., 2017; Tulasi Ram, Su, & Liu, 2009). During equinox, the transequator EMW is incompetent and hence regular equatorial phenomena control the double crest structure in B0 that is barely noticed only during LSA equinox and disappears during HSA equinox in our study. Similar interpretations in case of hmF2 also have been reported by Luan et al. (2016) confirming evident daytime double crests around ±10° geomagnetic latitude with a trough over the magnetic equator during LSA and March equinox which they named as equatorial height anomaly of the ionospheric F2 layer. Their observations also realized less obvious equatorial height anomaly during September equinox and disappearance of the phenomena during solstices. Moreover, it is evident from the plots that the anomaly in NmF2 is quite different from B0 manifesting prominently visible double crest structure in all seasons. Nevertheless, the solar activity dependence of both the ionospheric parameters is clearly remarkable showing larger magnitudes under high solar activity. The difference in the seasonal median B0 under HSA with respect to LSA are found to be 10%, 13%, and 18% for equinox, summer, and winter seasons, respectively. It appears that the sensitivity of B0 is more (almost double that of equinox) to the level of solar activity than any other critical electron density parameters. Again, the departure between minimum and maximum values of daytime B0 is higher during winter solstice under HSA, whereas the value remains high in the summer solstice under both solar activity conditions.

The striking peculiarity in the longitudinal variability of B0 in Figure 8 is that it reveals a wave-like structure with four stable peaks (wavenumber-4) prominently appearing approximately at 170°W, 90°W, 0°E, and 100°E during the equinoctial seasons under both solar activity levels. Similarly, the northern winter and summer solstice seasons indicate wavenumber-3 ( 160°W to  170°W,  30°W to  0°E, and  100°E to  130°E) and wavenumber-2 (70°W to  100°W and  90° to  120°E) structures, respectively. The leading nonmigrating tidal components with zonal wavenumber (ZWN) responsible for such wavenumber-4, wavenumber-3, and wavenumber-2 longitudinal wave pattern is believed to be the dominating diurnal eastward (DE3; ZWN = 3), diurnal eastward (DE2; ZWN = 2), and semidiurnal westward (SW4; ZWN = 4), respectively, though the other trivial components do exist (Forbes et al., 2008; Pedatella et al., 2008). Hence, the present observations reinforce earlier observational and modeling studies on the appearance of longitudinal wave-like pattern in the equatorial and low-latitude ionospheric electron density profiles, peak height parameters (Brahmanandam et al., 2011; Kil et al., 2007; G. Liu, Immel, et al., 2010; L. Liu, Wan, et al., 2010; Lühr et al., 2007; Pancheva & Mukhtarov, 2010), drift velocity (Hartman & Heelis, 2007; Huang et al., 2010; Kil et al., 2008; Ren et al., 2009), thermospheric wind and temperature (Husler et al., 2007; Ren et al., 2008), TEC (Lin, Hsiao, et al., 2007; Scherliess et al., 2008; Wan et al., 2008; Wang et al., 2015; Zhong et al., 2017), and in the topside ionosphere (Bankov et al., 2009; Hawkins & Anderson, 2017; Kakinami et al., 2011). The daytime wavenumber-4 (equinox) and wavenumber-3 (winter solstice) longitudinal structure is also reported by Scherliess et al. (2008) from 13 years of TEC data retrieved from the TOPEX/Poseidon altimeter. The electron temperature and total ion density variation reports from a decade of Defense Meteorological Satellite Program F13 Satellite observations also pointed wavenumber-4, wavenumber-3, and wavenumber-2 longitudinal wave structures during equinox, summer, and winter solstice seasons, respectively. Of them, the last couple of structures are induced by both longitudinal variation of geomagnetic declination and diurnal eastward tides (Ren et al., 2008). The remarkable longitudinal wavenumber-4 patterns in the EIA crests with bulged amplitudes over South America, West Africa, Southeast Asia, and Central Pacific Ocean were also demonstrated earlier by Wan et al., (2008, 2012) through emphasizing their existence with the seasonal dependency in tidal wave propagation and electrodynamical source. In brief, such longitudinal anomalies have been credited to the dynamics of neutral wind, equatorial electric and geomagnetic fields, and their zonal variations along with upward propagating tides and other waves in the lower atmosphere (Immel et al., 2006; Lei et al., 2007; Lin, Wang, et al., 2007). Nevertheless, the offset between magnetic and geographic equator across different longitudinal sectors governs the severity of the above occurrences. The other noticeable feature in winter solstice in Figure 8 is frequent fluctuations in the longitudinal variability whose magnitude further increases with solar activity. Over and above, the solar activity effect on overall B0 is relatively higher over the Asian and Pacific region covering 60° to 150°E longitudes and manifesting its highest values during HSA.

Details are in the caption following the image
Global longitudinal variation of B0 profile parameter over the equatorial and low-latitude region from COSMIC observations and comparison with IRI estimations during (a) low solar activity (LSA; 2006–2010) and (b) high solar activity (HSA; 2011–2015) periods.

Concerning B1 longitudinal discrepancy, it seems difficult to define a meaningful variability pattern at least in winter solstice under both solar activity levels. However, a wavenumber-4 structure in B1 is merely noticeable in the LSA equinox seasons whereas B1 in HSA summer follows a wavenumber-2 structure apparently similar to that of B0 variability (see Figure 9). Whatever the pattern may be, the overall magnitude is noticeably increased from LSA (about 1.2) to HSA (about 1.3) almost through all seasons. In general, the complete longitudinal variability structure of B1 is not clearly understood in our results except that some speculations have been extracted to define the pattern for assisting further modeling practices and enrichment of literature.

Details are in the caption following the image
Global longitudinal variation of B1 profile parameter over the equatorial and low-latitude region from COSMIC observations and comparison with IRI estimations during (a) low solar activity (LSA; 2006–2010) and (b) high solar activity (HSA; 2011–2015) periods.

5 Comparisons of B0 and B1 Between COSMIC and IRI Over Equatorial and Low Latitudes

A comparative analysis of COSMIC-derived B0 with different IRI bottomside modeling options was performed to investigate the discrepancy in model estimations. In Figure 8, we show the COSMIC profile fitted B0 parameters and the corresponding estimates from IRI-2012 model using all three bottomside modeling options, that is, Bil-2000, Gul-1987, and ABT-2009. It can be seen from Figure 8 that in all plots COSMIC B0 is at a higher magnitude than all the IRI bottomside modeling options regardless of season and solar activity. The longitudinal pattern of IRI model plots are almost following COSMIC values, whereas the magnitude discrepancies are mismatched. Exceptionally, the Bil-2001 could not reflect the longitudinal wavenumber-4 structure distinctly during the equinoctial seasons under both solar activity levels as other models do. In most cases, Bil-2001 is at the highest departure from COSMIC-derived values and Gul-1987 performed better than the ABT-2009 model in deriving a profile close to reality.

In Figure 9, we present the comparative overview of B1 from COSMIC as well as IRI submodels. Needless to say the estimated daytime B1 in Bil-2000 produced a uniform value everywhere for all seasons and follows mean constant values during daytime (1.9) and nighttime (2.6) with smoothed transitions around local dusk and dawn (Bilitza et al., 2014). Unlike B0, the estimated B1 (Gul-1987) is replaced by B1 (Bil-2000) with the release of IRI-2012 version; hence, its plots are exempted in this study. Among those remaining two IRI bottomside models as shown in Figure 9, B1 from ABT-2009 is higher than Bil-2001 irrespective of seasons under LSA and underestimating Bil-2001 under HSA conditions. The exception to the above is HSA summer where the longitudinal B1 (ABT-2009) is straddling around the constant B1 (Bil-2001). Nevertheless, the IRI model-estimated B1 values are always larger than the extracted B1 from COSMIC least squares fitting.

6 Conclusions

We have extracted the bottomside electron density profile shape parameters from COSMIC RO measurements using α-Chapman function least squares fitting and compared them with those estimated by IRI 2012 submodels: Bil-2001, Gul-1987, and ABT-2009. Past evaluations have been conducted by various groups with the global distributed ionosondes and ISR, demonstrating the reliability of COSMIC RO electron density profiles and their feasibility in ionospheric physics studies. However, although sensible comparative analysis of COSMIC electron density, peak parameters, and related IRI simulations have been carried out, some of the crucial aspects related to the north-south asymmetry of bottomside parameters during solstice seasons and their longitudinal abnormality in the EIA as captured from the COSMIC observations are not clearly portrayed in the IRI model (Lei et al., 2007). The large database of COSMIC data provides vertical electron density profiles globally as opposed to a limited number of ionosonde, ISR, and other LEO occultation satellites. Also, the characteristics of ionospheric vertical profiles and their variations with varying geophysical conditions have their importance in understanding the ionospheric and thermospheric physics and chemical exchanges for space weather investigation and communication applications. Moreover, past records of COSMIC-derived ionospheric studies have proven their consistencies with other measuring techniques and model simulations. It must as well be noted that the longitudinal pattern of bottomside thickness parameters are not yet fully explored due to the scarcity of data coverage, and hence the present study is an attempt to cater finer knowledge on B0 and B1 longitudinal variability by exploiting COSMIC occultation profiles.

The important findings from the study are as follows:
  1. There is an apparent adherence of B0 to the equatorial and low-latitude zone, whereas for the rest of the latitudes almost similar amplitudes are manifested. However, a distinguishable hemispheric asymmetry must be noted in the profile thickness (B0) parameters.
  2. The occurrence pattern of maximum B0 is exactly opposite to that of NmF2 manifesting a thicker profile in solstice than equinox season to retain the semiannual pattern as evidenced in other ionospheric parameters.
  3. Unlike NmF2, the interesting feature in B0 latitudinal variation is appearance of double crests only in March equinox, whereas the rest of the seasons perceive a unique crest in the vicinity of geomagnetic equator that is pushed (toward magnetic equator) southward or northward during summer and winter seasons, respectively. The active participation of transequatorial EMW component during summer (southward) and winter (northward) is responsible for such abnormality, whereas during equinox its magnitude is suppressed from causing any significant northward/southward shifting in B0 (1998; 2017; Tulasi Ram, Su, & Liu, 2009).
  4. The longitudinal wave-like behavior in B0 with four bulged amplitudes (wavenumber-4), particularly in the equinoctial seasons, indicates its connection with the nonmigrating diurnal tides (DE3; ZWN = 3) propagating in the eastward direction. Likewise, the wavenumber-3 and wavenumber-2 longitudinal pattern in B0 apparently appearing during the northern winter and summer solstice seasons even under both the solar activity conditions agree with the corresponding leading role of eastward propagating diurnal (DE2; ZWN = 2) and westward propagating semidiurnal (SW4; ZWN = 4) nonmigrating tides. Hence, the present results strengthen the understanding from earlier observational and modeling studies on longitudinal pattern of other measuring parameters like electron density and peak heights, temperature, TEC, and drift velocity and topside ionospheric characteristics (Burns et al., 2008).
  5. The possible factors for such longitudinal behavior could be the local time variation in neutral wind, vertical drifts, and equatorial electric field associated with the offset between magnetic and geographic equator, zonal variations in the neutral wind, and geomagnetic field components and the upward propagating tides from the atmosphere underneath, though their individual controls vary with seasons (Immel et al., 2006; Lei et al., 2007; Lin, Wang, et al., 2007). Although the nighttime longitudinal variability structure of other electron density profile parameters has been discussed and compared with daytime variability indicating approximately 180° out of phase difference (Brahmanandam et al., 2011), our focus in this work is on the daytime component.
  6. The other notable observations in our study are frequent fluctuations in the longitudinal structure during winter solstice with magnitude increasing with solar activity and the impact of the latter being relatively more over the Asian and Pacific region covering 60° to 150°E longitudes.
  7. Although the three IRI submodels represent the regular pattern of B0 in a similar manner during equinox, Gul-1987 seems to be inferior in the solstice seasons not following the control of reference geomagnetic field and equatorial electrodynamics and thereby depicting an erroneous distribution of B0 in the transition region.
  8. ABT-2009 presents a realistic distribution of B0 over the seasons, emphasizing the control of electric and magnetic field alignment across the continents, and highlights the solar activity difference in B0 component similar to other electron density and peak parameters.
  9. While Bil-2000 projected a narrow band of enhanced B0 magnitude around the magnetic equator, a more prominent B0 with wider spread values is estimated by Gul-1987 leaving behind moderate magnitudes of B0 in ABT-2009 that is slightly shifted northward. The other notable feature in ABT-2009 plots is the northern (southern) high latitude depicting relatively higher values than the middle latitude during the vernal (autumnal) equinox evidencing a hemispheric high-latitude asymmetry.
  10. Our comparative findings indicate that irrespective of bottomside submodels options chosen in IRI 2012 model, the fitted B0 values from COSMIC electron density profiles exhibit relatively higher magnitudes than all the IRI bottomside modeling options regardless of season and solar activity.
  11. Though the longitudinal pattern (wavenumber-4) of IRI 2012-B0 is similar to the COSMIC-fitted B0 with exception to Bil-2000, discrepancy in magnitudes is apparent. In most cases, Bil-2000 is at highest departure from the COSMIC results, whereas there is a compromise between the Gul-1987 and ABT-2009 for improved performance in a particular season.
  12. The longitudinal profile of B1 parameter is not clearly understood from our results except the subordinate magnitudes in COSMIC least squares fitted profiles compared to IRI model estimations that is a point of attention for further investigation from this study.

In conclusion, more observational and modeling techniques are essential for better understanding on the peculiarity and morphology of various ionospheric parameters over equatorial and low latitudes. The bottomside F2 layer has the greatest contribution to TEC, which is a key parameter for quantifying the transionospheric radio propagation delays. But, besides NmF2 and hmF2, the F2 region profile specification in IRI model also critically depends on the B0 and B1. The B0 and B1 parameters are not yet fully explored in terms of diurnal as well as seasonal characteristics, and most importantly their complete longitudinal profile has not been well documented in literature. Hence, this study supports further improvement of the bottomside representation for enhancing IRI reliability, particularly across the global equatorial and low-latitude region.


The authors acknowledge GSFC, NASA for providing an online version of the IRI-2012 model (https://omniweb.gsfc.nasa.gov/vitmo/iri2012_vitmo.html). The authors are also grateful to COSMIC Data Analysis and Archive Center (CDAAC) for providing access to the FORMOSAT-3/COSMIC RO database (http://cosmic-io.cosmic.ucar.edu/cdaac/index.html). The database of planetary geomagnetic Kp index for discarding geomagnetically disturbed days is availed from the World Data Center for Geomagnetism, Kyoto archives (http://wdc.kugi.kyoto-u.ac.jp). The ionosonde sounding profiles used for validation study have been made available free for researchers at the DIDBase repository at the Lowell GIRO Data Center (http://giro.uml.edu/). The access to the ionosonde profiles and extraction of bottomside parameters through manual scaling has been performed with the SAO Explorer package available at (http://ulcar.uml.edu/SAO-X/SAO-X.html), approval for which has been granted with the kindness of Ivan Galkin at UMLCAR.