Longitudinal structure in the CHAMP electron densities and their implications for global ionospheric modeling



[1] Detailed examination of the electron densities in the equatorial ionosphere observed by the CHAMP satellite at ∼400 km and mapped for a fixed local time has revealed a 90°-wide longitudinal structure with a maximum to minimum ratio that can exceed two or more. To date, the role of these large density variations has not been considered in either the development or validation of current global assimilative models. The best defined structure has nodes near 60°, 150°, 240°, and 330°, and is variously called the 4-node or wave number 4 structure. Unless a global model includes the physics that leads to the 4-node structure, or has the structure imposed upon it by the assimilation of data that includes the structure, a very significant part of the physics of the equatorial ionosphere will be missing. The version of the Utah State University (USU) Global Assimilation of Ionospheric Measurements (GAIM) model currently used by the U.S. Air Force takes the Ionospheric Forecast Model (IFM) as its background ionosphere. The IFM does not produce a 4-node structure. However, USU-GAIM is driven mainly by the assimilation of slant GPS total electron content observations, which could impose a 4-node structure, given sufficient equatorial GPS sites. The UV radiances from the Special Sensor Ultraviolet Spectrographic Imager instrument on DMSP also have a 4-node structure, and this is maintained in the GAIM global specifications. In the absence of sufficient assimilated data to impose a 4-node structure, such a structure could be imposed on the IFM by using a model of the equatorial E × B drift that contains a 4-node structure.

1. Introduction

[2] This paper first addresses the 4-node structure found in equatorial UV radiances and satellite in situ electron density observations [see, e.g., Immel et al., 2006; Liu and Watanabe, 2008; Luhr et al., 2007] and described here in terms of the equatorial observations made by the polar-orbiting CHAMP (Challenging Mini-Satellite Payload) and DMSP (Defense Meteorological Satellites Program) satellites at altitudes of ∼400 and 840 km, respectively. The amplitude variations in the IMAGE (Imager for Magnetopause-to-Auroral Global Exploration) far ultraviolet radiances at the anomaly peaks were discussed and interpreted by Immel et al. [2006], who showed that when plotted as a world map at a fixed local time the amplitudes exhibited a wave pattern that had four nodes and four antinodes across 360° longitude. This pattern is variously described as 4-node or “wave number 4.” In fact, the 4-node pattern is just one of the more obvious “n-node” patterns that the CHAMP observations show to exist in the equatorial ionosphere. Note that the n-node patterns or structures exist in a fixed local time frame, not in the “real” snapshot or fixed UT frame.

[3] Immel et al. [2006] showed that ionospheric densities in the equatorial region vary with the strength of nonmigrating, diurnal atmospheric tides that are, in turn, driven mainly by weather in the tropics. They revealed a global 4-node longitudinal signature in the nighttime airglow intensity and in the position of the equatorial anomaly. Liu and Watanabe [2008] used the CHAMP observations to examine the longitudinal structure of the equatorial ionosphere in the noon and postsunset local time sectors in different seasons, and presented multiple illustrations of 4-node variations. Luhr et al. [2007] used simultaneous observations of the electron density and the zonal wind obtained by CHAMP to study systematic longitudinal variations for August–September 2004.

[4] Scherliess et al. [2008] found a 4-node pattern in the TOPEX/Poseidon vertical total electron content (TEC) observations that is created during the daytime hours at equinox and June solstice but is absent or washed out by other processes during the December solstice. During equinox the 4-node pattern is created around noon with well defined longitudinal enhancements in the low-latitude TEC over East Asia, over the Pacific, and over the Atlantic Ocean. These enhancements are symmetric about the geomagnetic equator during this season, last for many hours, and can be clearly seen past midnight. The variations between the well-defined longitudinal maxima and minima are of the order of 20%.

[5] Apart from investigating the n-node structure exhibited by the CHAMP and DMSP satellites (mainly the former), we have investigated the effects of this structure on assimilative global ionospheric models. The main type of data assimilated by such models is the ground-based slant TEC derived from observations of signals from GPS satellites. As discussed above, Scherliess et al. [2008] found that the vertical TEC observations exhibit an n-node structure, so the structure is clearly going to impact the performance of global ionospheric models.

[6] We address in particular the Utah State University (USU) Global Assimilation of Ionospheric Measurements (GAIM) model [Scherliess et al., 2004; Schunk et al., 2004; Thompson et al., 2006] although other global models such as EDAM [Angling and Khattatov, 2006; Angling et al., 2009] encounter the same issues. In fact, our global model discussions are restricted to the USU GAIM-GM model, in which the observations are assimilated using a Gauss-Markov scheme. This is the model that is used operationally by the Air Force Weather Agency, and which is currently being validated by AFRL. In the GAIM-GM model, a statistical Gauss-Markov process is used to describe the time evolution of the differences between the measurements and the background ionosphere provided by the Ionospheric Forecast Model (IFM [Schunk et al., 1997]). The primary output of the model is a time-dependent three-dimensional electron density distribution.

[7] If an ionospheric model does not contain a 4-node variation, and it does not assimilate data that does show the variation, there is no way for such a variation to appear in the model output. The IFM used in GAIM v2.4.3 (the version of GAIM-GM being validated by AFRL) relies on an empirical model of the vertical E × B drift in equatorial regions [Scherliess and Fejer, 1999]. This drift model does not include a 4-node variation, so neither does the IFM specification of the equatorial ionosphere. Retterer et al. [2008] have shown that combining an empirical model of the drifts based on ROCSAT (∼600 km) observations with the AFRL model of the equatorial ionosphere (PBMOD [see Retterer, 2005; Retterer et al., 2005]) successfully produces a 4-node variation in the electron density. (PBMOD concentrates on the equatorial ionosphere and is used for forecasting scintillations. It is not a global model.)

[8] Many of the same issues are expected to arise with empirical models developed by URSI and CCIR (ITU) that fit observations such as foF2 with spherical harmonic functions of the minimum possible order. The information missing from both the assimilative and empirical models of the equatorial ionosphere can be expected to have negative impacts on operational applications such as real-time HF propagation predictions [see, e.g., Angling et al., 2009; McNamara et al., 2009].

[9] Section 2 of the paper describes how we analyzed the CHAMP data and set up data files required for the various types of plot, and presents examples. Section 3 concentrates on high-resolution CHAMP plots for single months and different local times. The plots show 1- and 2-node variations as well as the more obvious 4-node variation. The plots are illustrative, not exhaustive. Section 4 presents observations of 4-node variations in the DMSP/SSIES (Special Sensors-Ions, Electrons, and Scintillation) electron densities. These are similar but not identical to the CHAMP variations. Section 5 presents some results of GAIM validations for the case in which the UV radiances from DMSP/SSUSI (Special Sensor Ultraviolet Spectrographic Imager) were assimilated. Section 6 shows that n-node variations also exist in some empirical global models. Section 7 summarizes and discusses our results.

2. Methodology and Examples

[10] The Plasma Langmuir Probe (PLP) used to determine the electron density at the CHAMP satellite location is a component of the AFRL Digital Ion Drift Meter (DIDM) instrument. The PLP basically consists of a ram-facing, electrically isolated gold rectangular sensor plate mounted on the lower front panel of the spacecraft. There are no grids in front of the sensor. The sensor plate of the PLP is alternately allowed to float for 14 s to track the spacecraft potential, and swept in voltage for 1 s to verify the floating potential and to determine the ion density and electron temperature. The density value is derived from the measurement of the current to the PLP plate during the negative portion of the 1 s voltage sweep, along with the plate surface area and spacecraft velocity. The PLP plasma frequencies have been validated by McNamara et al. [2007], through comparison with plasma frequencies derived from Digisonde ionograms at Jicamarca, Peru. That paper also describes the operation of the PLP in some detail.

[11] Most of the discussion of the CHAMP observations is in terms of plasma frequency, rather than electron density. Those who work with ionosondes work in terms of plasma frequency, while modelers tend to work in terms of electron density. (N = 1.24 × 1010 fn2, where fn is the plasma frequency in MHz, and N is in m−3.)

[12] The CHAMP plasma frequencies have been processed in groups of days that straddle the days of year for which the CHAMP passes are at a specified local time. Figure 1 shows for example the LT–day of year tracks of CHAMP for 2004, for dip latitudes less than 40°. CHAMP passes through 0000 LT for days 75 ± 10, 205 ± 10, and 340 ± 10 (and other groups of days for other LTs). It takes ∼4 years of data to build up a full annual/seasonal variation for a fixed LT.

Figure 1.

Local time versus day of year tracks for CHAMP passes, 2004.

[13] Our analysis program works with a specified year, LT and range of corresponding days. All CHAMP records with an LT within 30 min of the specified LT are distributed into latitude and longitude bins, and the median plasma frequency calculated for each bin. We have not specifically rejected any magnetically disturbed days.

[14] We have used two sets of coordinates, geographic latitude and longitude, and dip latitude and geographic longitude, each of which has advantages. For Figure 2, the dip latitude and geographic longitude bins were 2° and 5°, respectively. For Figure 4, the dip latitude and geographic longitude bins were 5° and 10°, respectively. For Figure 3, only two bins are considered, +10° to +20° dip latitude, and −10° to −20° dip latitude, to cover the anomaly peaks. The longitudinal variations are smoothed to eliminate some of the noise due to small sample sizes.

Figure 2.

Geographic longitude/dip latitude world map of the median CHAMP plasma frequency for 2000 LT, September 2004.

Figure 3.

Geographic longitudinal variation at 2000 LT for median CHAMP plasma frequencies in dip latitude ranges of 10–20 degrees, days 235–265, September 2004.

[15] Figure 2 shows the 2000 LT world map of the median CHAMP plasma frequency for September 2004, in a format generally similar to that used by Scherliess et al. [2008, Figure 5a] and Retterer et al. [2008].

[16] The axes are dip latitude and geographic longitude. Dip latitude was preferred to geographic latitude because the latter type of plot also shows the geographic variation of the dip equator, which tends to catch the eye instead of the variations of the plasma frequency with longitude. (The very dark southern hemisphere area near 300° longitude is associated with the behavior of the definition of dip latitude for this region of the world.) Figure 2 (top) shows the usual plasma frequency variation in the equatorial region, with the anomaly peaks at ±10 to 20° dip latitude, and a minimum at the magnetic equator. The longitudinal minima in plasma frequency occur at ∼60, 150, 240 and 330°, with maxima at ∼15, 105, 195 and 285° (central Africa, Vietnam, central Pacific, and Peru). Note that the days of year 235–265 used to make Figure 2 just provide a wide filter that is sure to catch all of the CHAMP data with the specified LT.

[17] Figure 2 (bottom) shows the longitudinal variation of the median CHAMP plasma frequency for the two 10–20° dip latitude bins. The blue curve is for the northern anomaly, and the red curve is for the southern. Figure 3 shows a plot of the same data in the format used in later plots that will be compared with Figure 3.

[18] As with Figure 2 (but with slightly different longitudinal smoothing), the minima in plasma frequency occur at ∼60, 150, 240 and 330°, with maxima at ∼15, 105, 195 and 285°. Note that the minima near 300° are not at the same geographic longitude. This is mainly because the geographic and dip equators subtend such a large angle at these longitudes (South America). The ratio of the maximum to minimum is useful for comparing different plots. For example, the ratio for the rise from 60° to 90° is ∼10/7 = 1.4. For the fall from 200° to 240°, the ratio is ∼10.2/7 = 1.4 again. This ratio corresponds to a factor of 1.42 = 2.0 change in electron density over 30–40 degrees in longitude.

[19] Figure 4 shows an alternative, more detailed, way of viewing the data, specifically for the northern anomaly. The dip latitude and geographic longitude bins were 5° and 10°, respectively.

Figure 4.

Northern hemisphere longitudinal variation at 2000 LT for median CHAMP plasma frequencies in dip latitude ranges of 5–10, 10–15, and 15–20 degrees, days 235–265, 2004.

[20] Note that the first and last points on each curve do not necessarily match, because of the smoothing. The red (2) curve in Figure 4 corresponds to the poleward side of the anomaly peak. Figure 4 shows clear minima at (approximately) 60°, 150°, 240° and 330°. The blue (3) curve shows clear minima at 60°, 150°, and 330°, but only a small dip at 240°. The black (4) curve corresponds to the equatorward side of the anomaly. For the red (2) curve, the range from maximum to minimum reaches a high value of ∼(9/5), which corresponds to an electron density range of ∼3. Figure 5 shows the corresponding plot for the southern anomaly.

Figure 5.

Southern hemisphere longitudinal variation at 2000 LT for median CHAMP plasma frequencies in dip latitude ranges of −5 to −10, −10 to −15, and −15 to −20 degrees, days 235–265, 2004.

[21] As with the northern anomaly plot, the red (9) curve shows minima at (approximately) 60°, 150°, 240° and 330°, although the 240° minimum appears on the shoulder of the curve as it drops down to a minimum of 4 MHz at ∼300°. The range of the plasma frequency variation near 300° is ∼2× (or ∼4× in electron density). The South Atlantic Anomaly (SAA) seems to have only a very small effect on the measurement of the CHAMP plasma frequencies, so the low values near 300° are not due to the SAA.

[22] As expected, the black (4 and 7) curves in Figures 4 and 5, which cover the 5–10 degree dip angle intervals, are very similar. Figure 4 (for the northern anomaly) shows that the red and black curves are antiphase, which is consistent with the E × B drifts moving ionization away from the magnetic equator (the black 4 curve) to the poleward side of the anomaly (the red 2 curve), and vice versa. It follows that the vertical drift velocity must therefore share in the 4-node variations, as shown by Kil et al. [2007] and Fejer et al. [2008]. In general, it would be expected that the variations of the CHAMP plasma frequencies at the magnetic equator would be antiphase with the drift velocity, since higher drift velocities move more ionization away from the equator.

3. CHAMP Plasma Frequency Variations at Different LTs

[23] The n-node variations such as seen in the CHAMP plasma frequencies at 2000 LT also exist at other local times. We have generated ∼100 plots for both day and night that show that the equatorial ionosphere as sampled by CHAMP (at ∼400 km) seems to be permeated at all times by very complex wavefields that coexist and interfere with each other, with particular waves such as the 4-node dominating under some conditions. The amplitudes of the waves can change by a factor of two in plasma frequency or four in electron density across a narrow range of longitude. We present here only a few representative examples that illustrate the variability of the n-node structure and the coexistence of nodes of different orders.

3.1. CHAMP Plasma Frequency Variations at 0800 LT

[24] Figure 6 shows the longitudinal variation of the CHAMP plasma frequency for 0800 LT, days 020–050, 2001. The minimum to maximum ratio near ∼180°E for each curve in Figure 6 is ∼1.5 (2.25 in electron density). The high values of plasma frequency occur basically over the Pacific Ocean (which may be related to the neutral atmosphere origin of the n-node variations). The minima near 150° and 330° may be related to the 4-node variation, which generally occur at these longitudes. The small dip near 210° is to the west of the 4-node minimum at ∼240°, and there is no minimum at ∼60°. In this particular case, it appears that the 4-node variation may be modulated by an asymmetric 2-node variation. Figure 7 shows the n-node structure for the southern anomaly.

Figure 6.

Northern hemisphere longitudinal variation at 0800 LT for median CHAMP plasma frequencies in dip latitude ranges of 5–10, 10–15, and 15–20 degrees, days 020–050, 2001.

Figure 7.

Southern hemisphere longitudinal variation at 0800 LT for median CHAMP plasma frequencies in dip latitude ranges of −5 to −10, −10 to −15, and −15 to −20 degrees, days 020–050, 2001.

[25] The southern anomaly shows the same basic structure as the northern anomaly, with a large peak over the Pacific Ocean. Partly because of the northward movement of the dip equator over Brazil, the red (9) and blue (8) curves extend further east. The black curves (4 and 7) both have a value of ∼5.5 MHz at 300°E.

[26] Figure 8 shows the world map (dip latitude and geographic longitude) of the CHAMP plasma frequencies at 0800 LT. Figure 8 can be compared with Figure 5c (top left) in the work of Scherliess et al. [2008]. It illustrates clearly the large maximum over the equatorial Pacific Ocean. Comparison of Figure 8 (0800 LT) with Figure 2 (2000 LT) shows much lower plasma frequencies for dip latitudes greater than ∼20° for the 0800 LT plot.

Figure 8.

Geographic longitude/dip latitude world map of the median CHAMP plasma frequency for 0800 LT, January/February 2001.

3.2. CHAMP Plasma Frequency Variations at 2200 LT

[27] Figures 9 and 10 show the longitudinal variation of the CHAMP plasma frequency for 2200 LT, days 320–350 (November), 2003. (The variations for 2000 LT were discussed in section 2.) Figure 9 (northern anomaly) shows the familiar 4-node variation, but with a 1-node modulation that has a very broad minimum near 180°. Figure 10 (southern anomaly) shows the same effects, especially for the black (7) curve, which has a range of over 3 MHz. 1-node variations can also be seen at other times, but with the minimum near 360°, rather than near 180° (such as 0800 LT, days 050–080; 1000 LT, days 255–285).

Figure 9.

Northern hemisphere longitudinal variation at 2200 LT for median CHAMP plasma frequencies in dip latitude ranges of 5–10, 10–15, and 15–20 degrees, days 320–350, 2003.

Figure 10.

Southern hemisphere longitudinal variation at 2200 LT for median CHAMP plasma frequencies in dip latitude ranges of −5 to −10, −10 to −15, and −15 to −20 degrees, days 320–350, 2003.

[28] Figure 11 shows the world map (dip latitude and geographic longitude) of CHAMP plasma frequencies at 2200 LT, days 320–350 (November), 2003. The white pixels in Figure 11 show the large peaks in both hemispheres near 300°, and the southern hemisphere peaks near 120° and 200° that are seen more clearly in Figure 10. The 1-node variation seen in Figures 9 and 10 is not as obvious in the more commonly used format of Figure 11. The low values at the magnetic equator are symptomatic of equatorial plasma bubbles, which have not been excised for this type of plot.

Figure 11.

Geographic longitude/dip latitude world map of the median CHAMP plasma frequency for 2200 LT, November 2003.

4. Observations at Other Altitudes

[29] USU-GAIM will also assimilate the in situ electron density observations made by the SSIES instrument onboard DMSP. We have analyzed these observations in a manner similar to our analysis of the CHAMP data, for March/April 2004 (F15, at ∼2100 LT) and September 2004 (F16, at ∼2000 LT). Figure 12 shows the September 2004 F16 results (in terms of plasma frequency).

Figure 12.

Geographic longitudinal variation at 2000 LT for median SSIES plasma frequencies in dip latitude ranges of 10–20 degrees, days 246–266, September 2004.

[30] Figure 12 can be compared with the CHAMP results given earlier in the same format (Figure 3). Although the values at 360° are suspicious (and not present in the following F15 results), it is obvious that the SSIES plasma frequencies also exhibit a 4-node variation, with maxima at ∼60, 150, 240 and 330°. The ratio of maximum to minimum plasma frequency is ∼1.4/1.2, or 1.36 in electron density. The CHAMP plasma frequencies shown in Figure 3, on the other hand, exhibit ratios of ∼2.0 in electron density.

[31] Figure 13 shows the SSIES March/April 2004 results. The March/April results in Figure 13 also show wave structure, but the patterns are not as simple as for September 2004. Both anomaly peaks show a minimum near 60°, but the northern anomaly shows an antinode at ∼300° while the southern anomaly shows a node. Apart from this longitude region, the ratio of maximum to minimum plasma frequency is ∼1.2/1.0, or 1.44 in electron density, which is very similar to the ratio for the September results. (There were no evening passes of CHAMP during March/April 2004.) In general, the CHAMP and SSIES n-node structures do not usually exhibit a node-by-node correspondence.

Figure 13.

Geographic longitudinal variation at 2100 LT for median SSIES plasma frequencies in dip latitude ranges of 10–20 degrees, days 080–104, March/April 2004.

[32] The normalized electron density ratio of (maximum − minimum)/(maximum + minimum)/2 is useful for comparing plots for different satellites/altitudes. For corresponding CHAMP and DMSP/SSIES passes for 2002 and 2004, for example, this ratio was ∼0.16 for CHAMP and ∼0.04 for SSIES. This means that the waves are ∼4× as “strong” at the CHAMP (∼400 km) altitude than at the DMSP (∼850 km) altitude. This is what would be expected, but a more careful analysis would be required to relate the waves at different altitudes.

[33] N-node variations would also be expected in the ROCSAT (Republic of China Satellite (ROCSAT-1)) observations of electron density at an altitude of ∼600 km, since the variations have been shown to exist in the observed vertical drifts and thence in the electron densities at the magnetic equator given by the physics-based model PBMOD [Retterer et al., 2008].

5. Assimilation of SSUSI Radiances by USU-GAIM

[34] One of the USU-GAIM validations being performed at AFRL is of the assimilation of DMSP/F16 SSUSI UV radiances for the evening pass at ∼2000 LT over the equator. The SSUSI instrument provides nadir (disk) scans and limb scans to the east (for the evening pass). AFRL has performed validations with both types of observations, but it is the near-nadir disk radiances that show a clear 4-node variation.

[35] Figure 14 shows the median September 2004 SSUSI disk radiances for latitude bins of +10 to +20 and −10 to −20 degrees dip latitude, versus geographic longitude (with some smoothing in longitude). The 4-node variation is clearly evident in these radiances, with a range of 2× or higher. The curve for the southern anomaly does not cover the longitude range 270° around to 30°, in order to excise spurious radiances that arise because of particle precipitation in the South Atlantic Anomaly region. Figure 14 may be directly compared with the earlier Figure 3, which shows the CHAMP plasma frequencies and the same 4-node variation.

Figure 14.

Longitude variation of the median SSUSI disk radiances for 2000 LT, September 2004, for the northern and southern anomaly peaks.

[36] Figure 15 shows in the same format the USU-GAIM plasma frequencies, for the case in which only the UV radiances are assimilated (no GPS TEC or SSIES observations). The sample sizes for each bin are rather low because GAIM was run for only 20 days, and the local time window was rather small, so the raw distributions are noisy with some zero counts (which cause the gaps in the plotted lines). The longitudinal variations have therefore been smoothed. It can be seen from Figure 15 that the GAIM plasma frequencies also exhibit the 4-node variation, with minima at 60°, 150°, and 240°. The expected minimum at ∼330° is not present. A 4-node variation is also evident when the GPS TEC and SSIES data are assimilated along with the UV data.

Figure 15.

Longitude variation of the median USU-GAIM plasma frequencies for 2000 LT, September 2004, for the northern and southern anomaly peaks, when SSUSI radiances are assimilated.

[37] Note that the maximum to minimum ratio near 60° longitude (for example) is ∼10/8 = 1.25, or 1.6× in electron density. This is in contrast to the radiances assimilated by GAIM that are plotted in Figure 14, which have a ratio of more than 2. The difference is possibly due to the relative weights that GAIM gives (dynamically) to the IFM and the assimilated data. GAIM tends to be conservative because of the ever present possibility of invalid assimilation data. For example, Decker and McNamara [2007] showed that GAIM does not reproduce the extreme values of foF2 observed during ionospheric storms.

[38] The AFRL validations of the assimilation by USU-GM of SSUSI radiances have confirmed that GAIM correctly assimilates the radiances and in so doing provides more accurate specifications of the equatorial ionosphere in those regions near the DMSP tracks. The assimilated radiances have been restricted to the evening pass (2000 LT for F16) through the equatorial ionosphere. The CHAMP plasma frequencies are the only useful source of ground truth since the equatorial ionograms at 2000 LT are almost invariably made unusable by the presence of spread F echoes. For the validations, the observer “rides on CHAMP,” and continually asks GAIM what it thinks the plasma frequency is. The comparisons are normally restricted to those for which there is a UV observation within 500 km of CHAMP. The validations are performed in retrospect, which means that the latency of the UV observations (the time taken for the observations to make their way from DMSP to GAIM, which is ∼90 min) is effectively set to zero. If a data latency exists, the latest observations can make no contribution to the real-time GAIM specifications of the ionosphere (although they can contribute to post facto analyses). The validations have also confirmed that GAIM correctly assimilates latent UV observations, to provide an improved but belated specification.

6. N-Node Structure in Empirical World Models

[39] Some global assimilative models such as EDAM [Angling and Khattatov, 2006] use the International Reference Ionosphere [Bilitza and Reinisch, 2008] as their starting point. It is therefore of some interest to determine if the IRI specifications show any evidence of an n-node structure. Figures 16 and 17 show the median plasma frequency at an altitude of 400 km when using the CCIR and URSI world maps of foF2 with the IRI, for 2000 LT, September 2004.

Figure 16.

Median CCIR plasma frequency at 400 km, September 2004, 2000 LT.

Figure 17.

Median URSI plasma frequency at 400 km, September 2004, 2000 LT.

[40] Figures 16 and 17 can be compared with the CHAMP data shown in Figure 3, which shows that the CCIR curves are in better agreement with CHAMP. The CCIR and URSI world maps are of foF2, which corresponds to an altitude hmF2 that is mostly below the altitude of CHAMP, so there is no guarantee that the same n-node structures will occur at hmF2 and the CHAMP altitude. The same n-node structure exists at all altitudes in the IRI profiles.

[41] In this particular case, the CCIR structure is similar to the CHAMP structure, while the nodes and antinodes of the URSI structure have less structure and wrong relative sizes of the northern and southern anomaly peaks. A limited comparison has shown that in general the nodal structure in the IRI/CCIR plots agrees with the CHAMP structure better than do the IRI/URSI plots. This might be because the CCIR maps are based on the years 1957 (IGY) and 1964 (IQSY), 2 years with large numbers of ionosondes. The URSI maps [Rush et al., 1989] are based on a reduced number of stations for which a solar cycle or more data were available.

7. Discussion

[42] The 4-node variation, or any other similar variation, will appear in the output specification of a global assimilative model such as USU-GAIM if the assimilated ionospheric data includes evidence of the variations, even if the model does not include the physics responsible for the variation. We presented evidence of this behavior when SSUSI UV data (which manifests the 4-node variation) is assimilated, specifically for September 2004. The early evening SSIES observations for the equatorial anomalies also exhibit 4-node variations. However, the SSIES data are usually overwhelmed by the much more voluminous GPS TEC data, and corresponds to a significantly higher altitude (DMSP, ∼850 km) than the altitude of the ground-truth observations (CHAMP, ∼400 km).

[43] The USU-GAIM v2.4.3 global specifications obtained when only the GPS TEC and SSIES observations are assimilated (but not the UV data) do show evidence of an n-node variation at a fixed LT, but the variation is not always clearly defined and consistent with the variations of the CHAMP plasma frequencies. A lot depends on the equatorial coverage provided by the assimilated GPS sites. In general, the more sites, the better. However, the number of equatorial GPS sites necessary to define the n-node structure adequately is not yet known.

[44] For models such as the AFRL low-latitude ionospheric model PBMOD [Retterer, 2005; Retterer et al., 2005] that do not assimilate real-time ionospheric observations, the variations can still appear in the output specification if one of the key drivers such as the vertical E × B drift includes the 4-node variation (or similar). Historically, PBMOD used the vertical ion drifts given by empirical drift models based on observations from Jicamarca and the AE-E satellite [Scherliess and Fejer, 1999]. However, Retterer et al. [2008] used PBMOD with vertical drifts derived from ROCSAT observations [Kil et al., 2007], which exhibit n-node variations, and confirmed that the n-node variations then appear in the model electron densities. The same holds for the physics-based IFM, which is the starting point for USU-GAIM. (In fact, PBMOD and IFM have a common heritage.)

[45] The bulk of the ionospheric physics in the Gauss-Markov version of GAIM (GAIM-GM) that is currently being used by AFWA is embedded in the physics-based IFM. The IFM does not include the neutral atmosphere, which is the seat of the nonmigrating, diurnal atmospheric tides that lead to the n-node structure of the equatorial ionosphere. The n-node structure must therefore be imposed on the IFM via an equatorial E × B drift model that does include the structure, such as the climatological model of the ROCSAT vertical drifts of Fejer et al. [2008].

[46] Scherliess et al. [2008] have shown that a 4-node variation appears in the TOPEX/Poseidon/Jason vertical TEC observations, with a ∼20% variation from maximum to minimum, which contrasts with the factors of two or more in electron density found for CHAMP. The cause of the lower ratios may be the averaging over multiple instances of interfering waves of different orders and amplitudes, which would tend to wash out some of the stronger variations. Another issue is the height dependence of the 4-node variation, as exemplified by CHAMP (∼400 km) and DMSP/SSIES (∼840 km). Both the vertical TEC and the GPS slant TEC represent integrals across height-dependent n-node variations. It has not yet been shown that the n-node structures that appear in the ROCSAT observations (∼600 km) are a sufficiently accurate surrogate for the structure that exists in the slant TEC.

[47] In the absence of observations to assimilate, modelers often assume that the ionosphere observed at a particular LT will be the same as it was 1 h (or more, depending on the correlation lengths) to the east at the same LT (the “Sun-fixed” assumption). The Sun-fixed assumption will also be affected by the n-node structure of the equatorial ionosphere. In terms of LT maps of the ionosphere, this assumption means that the plasma frequency contours would be aligned latitudinally. The existence of n-node structures in the equatorial ionosphere shows that this assumption has limited validity at these latitudes, since one to 1.5 h translates to one third to one half of the spacing between a node and an antinode of a 4-node structure.

[48] The n-node structure of the equatorial ionosphere would seem to imply fundamental limitations to the achievable accuracy of current global ionospheric models, for example, by limiting the utility of the Sun-fixed ionosphere assumption, and of latent data.


[49] The GAIM-GM code is provided to AFRL by USU to support a continuing AFRL validation effort on the part of the Air Force Weather Agency. The USU scientists continue to provide AFRL with valued support in testing a very complex system. The SSUSI UV radiances used by GAIM are provided by the Johns Hopkins University Applied Physics Laboratory. L.F.M. is supported in part by AFRL contract FA8178-08-C-0012.