Assimilation of global-scale and mesoscale electric fields from low-latitude satellites

Authors


Abstract

[1] Equatorial satellites can measure key drivers of the low-latitude ionosphere globally and frequently. The electric field and plasma drift instruments of the Communications and Navigation Outage System satellite will provide redundant observations of the most important driver of low-latitude ionosphere structure, that is, the zonal electric field and resulting E × B plasma drift. The electric field data from the San Marco D satellite have been reduced to E × B plasma drift equivalents and are used to, first, examine data ingestion into an ionosphere model for weather sensitive specifications of the ionosphere and, second, explore use of the mesoscale (<1000 km) features in the E × B plasma drift to forecast the spatial mapping of signal scintillation associated with equatorial spread F in the postsunset ionosphere. The ingestion of the global-scale, satellite-based plasma drift observations into an ionosphere model improves electron density specifications over climatology when results are compared with the satellite in situ electron density measurements. Examination of the presunset and postsunset passes of the San Marco D satellite over a specific longitude sector suggests that the mesoscale structure in the zonal electric field can be used to predict the position and spacing of large plasma plumes associated with equatorial spread F in the postsunset ionosphere. A presunset forecast of the postsunset spatial location of plasma plumes can be used to spatially refine forecast maps of signal scintillation.

1. Introduction

[2] The electric field is a primary driver of the low-latitude ionosphere structure caused by the vertical E × B plasma drift at the magnetic dip equator. Scherliess and Fejer [1999] presented a climatological model of the global-scale E × B vertical plasma drift in the low-latitude ionosphere. A particularly important feature in the low-latitude vertical drift is the evening prereversal enhancement (PRE). The PRE lifts the evening F region plasma to high altitudes, which destabilizes the F region ionosphere and allows the gravity-driven Rayleigh-Taylor (RT) instability to proceed. Mesoscale plasma bubbles or plumes grow from the unstable bottom side of the F region. Plumes of rarefied plasma beginning from the bottom side move rapidly up through the dense F region plasma to the topside ionosphere. One method of identifying plasma plumes in satellite data is to detect very large vertical plasma drifts in the data, though not all plasma plumes are drifting upward. On the steepened structures within a plasma plume, smaller-scale irregularities develop. The RT instability produces a cascade of irregularities through an incredible range of scale sizes (G. Haerendel, Theory of equatorial spread F, unpublished manuscript, 1973). This nighttime irregularity phenomenon is called equatorial spread F (ESF). The small-scale irregularities within the plasma plumes have a negative impact on communication and navigation systems that are dependent on transionospheric electromagnetic signal paths. Strong scintillation in L-band and UHF signal transmissions is caused by the irregularities within the large plasma plume structures of ESF. The one unfortunate characteristic of ESF is that the night-to-night occurrence and spatial development of plasma plumes with the associated signal scintillation is highly variable and presently unpredictable.

[3] Because of the small growth rate of the Rayleigh-Taylor instability in the F region, an initial perturbation in the plasma is required prior to the evening onset of ESF. Atmospheric gravity waves have been suggested as the driver of the initial perturbation in the F region plasma [Kelley et al., 1981]. The plasma perturbation might develop from chemical interaction of the atmospheric wave and plasma or from the modulated electric field induced by the atmospheric wave motion through the ionosphere [Huang and Kelley, 1996].

[4] In an attempt to improve low-latitude specification and forecast of communication and navigation outages associated with ESF, the Air Force is placing a well-instrumented satellite in an equatorial low-Earth orbit. The Communication-Navigation Outage Forecast Satellite (C/NOFS) will measure plasma and neutral parameters as well as the electric field. The 90-min orbit will capture conditions in each sector of the low-latitude ionosphere. A comparable equatorial satellite, the San Marco D, was flown in 1988 with similar instruments. The San Marco D electric field measurements are used in this study to examine methods of data ingestion into models of the low-latitude background ionosphere and postsunset irregularities. Improvements to background ionosphere specification are demonstrated. Additionally, evidence is given of correlation between mesoscale structure in the electric field before sunset and the location of large plasma plumes after sunset. This suggests that the gravity wave modulation of the electric field prior to sunset provides the seeding of the large plasma plumes in the postsunset ionosphere. Thus the presunset mesoscale structure of the electric field can be used to forecast refined spatial mapping of scintillation in the nighttime ionosphere.

2. San Marco D Satellite

[5] The electric field instrument on NASA's San Marco D satellite is similar to the future C/NOFS vector electric field instrument. C/NOFS will have continuous coverage of data, which will greatly enhance the predictive capabilities over this demonstration based on the sparse San Marco data set. San Marco D was launched on 25 March 1988. The inclination of 2.9° and slight elliptical orbit (614 km, 260 km) provided an optimal measurement platform for low-latitude ionosphere studies. However, the measurements were not continuous due to power constraints. Two pairs of 20-m wire antennas in the spin plane of the satellite were used to monitor electric fields using the double floating probe technique [Maynard et al., 1995]. The data have been calibrated and quality assessed for earlier studies [Eccles et al., 1999].

[6] The San Marco data interval from 27 August to 4 September 1988 is sufficiently dense near the sunset terminator to examine the ingestion of the vertical drift (zonal electric field) into global and mesospheric models of the ionosphere. These days (days 242–248 of 1988) were used in the San Marco mission to observe ionospheric parameters near sunset over Kwajalein in the southern Pacific. This provides measurements in one geographic longitude in the presunset ionosphere on one San Marco pass and postsunset ionosphere on the second pass 90 min later. Certainly, an equatorial satellite with continuous electric field measurements would provide proper statistical studies of the significance of the case studies presented herein. There are other days of data during 1988 that are used as well to find evenings when ESF plasma plumes were not observed in the data.

3. Analysis

[7] The electric field values were transferred from satellite coordinates to magnetic field coordinates (parallel, zonal, meridonal) using the International Geomagnetic Reference Field (IGRF) magnetic field model, then related to the electric field values at the apex of the magneticfield line (dip equator). It is assumed that the field lines are equipotentials and approximate dipoles. This allows a simple formula to relate the electric field observation at the satellite to the electric field at the apex of the field line passing through the satellite position [Haerendel et al., 1992]:

equation image
equation image

Eφ and EL are the zonal and vertical values, respectively, of electric field at the field line apex. Es and Eq are the zonal and meridional field components, respectively, at the satellite but perpendicular to the magnetic field. The ζ is the sine of the magnetic latitude. The plasma on a field line moves as a coherent flux tube drifting as E × B. The plasma drift at the apex is easily ingested into the ionospheric forecast model (IFM) [Schunk et al., 1997], which solves for the time history of plasma flow along a multitude of drifting flux tubes. However, because the global IFM model does not resolve mesoscale structures, the mesoscale structures of the vertical drift must be removed prior to ingesting into the IFM. This includes gravity wave modulations and postsunset plasma plume structure in the vertical drift.

[8] The global-scale structure in the electric field and mesoscale structure must be separated to be useful for assimilation into a global ionosphere model. A simple smoothing algorithm will not do, because the smoothing would average the strong upward plasma drifts within plasma plumes into the background drifts. It is necessary to remove drifts within the plasma plumes to obtain the background drift information. An algorithm was developed to determine plume location using the combined plasma density measured by the San Marco ion density probe and vertical plasma drift values.

[9] Figure 1 shows a composite of vertical plasma drift deduced from the electric field probe and the plasma density obtained from the San Marco retarding potential analyzer. The global-scale feature of the PRE can be seen. The vertical drift before and after the PRE is modulated with mesoscale structure. The positions of the wave crests are indicated by W, and the postsunset plumes are indicated by P. The ESF plasma plumes are indicated using the narrow density depletion features in association with the strong upward drifts. The presunset modulation of the vertical drift is present on most days as shown in Figure 2. Figure 2a is a composite of nights with no observed plasma plumes, and Figure 2b is a composite of nights with plasma plumes. We will term these non-ESF nights and ESF nights according to the detection of plasma plumes in the satellite data. The ESF nights are generally from the September 1988 data. The non-ESF nights are scattered throughout the 1988 flight. In Figure 2b (ESF nights), one can see the scattered large vertical drifts after 1930 LT. The background plasma drift is always negative after the PRE, except for large positive drifts that occur within ESF plasma plumes.

Figure 1.

Vertical plasma drifts and plasma densities. (a) Drifts are deduced from the electric field probe. Presunset wave crests are identified (W), and postsunset plasma plumes are identified in the next satellite pass (P). The question mark is a developing plume prior to the descent of the ionosphere. (b) The total ion density from the retarding potential analyzer on San Marco shows the rarefied plasma plumes.

Figure 2.

Vertical plasma drifts for (a) days without rising plasma plumes (non-ESF days) and (b) days with rising plasma plumes (ESF days). The presunset drifts have mesoscale (50–1000 km) structure capable of seeding plasma plumes on ESF days and non-ESF days.

[10] Prior to the PRE, smaller modulations exist in the vertical drift compared with those within plasma plumes after sunset. These smaller modulations appear on both the non-ESF nights and ESF nights. It would appear that there are large modulations on non-ESF nights when comparing the Figure 2 panels. One can see the mesoscale wave structure with about 300–800 km wavelengths on nearly all days plotted. The non-ESF days have continued wave structure after the PRE, but no plasma bubbles were present in these data. The mesoscale wave structure of the vertical plasma drift appears to be ubiquitous. Note that these are observed on magnetic field lines that have large field-line-integrated F region conductivity compared to the E region during the late afternoon and evening. The long wavelength and dominance of the F region conductivity in the late afternoon suggest that atmospheric gravity waves in the F region are responsible for the modulations in the vertical plasma drift in the magnetic equatorial plane, VL. The presence of large waves during the non-ESF nights suggests that the seeding mechanisms for plasma plumes are present, but the Rayleigh-Taylor instability is suppressed by the atmosphere and ionosphere conditions to limit nonlinear plasma plume development.

[11] To separate the global-scale background drifts and the mesoscale drifts, a several-step procedure was devised. First, the drifts within the plasma plumes were removed and, second, were filled linearly from the background vertical drifts just outside the plasma plume structure. This procedure did not remove gravity wave structure with smaller amplitudes in VL. The wave-like modulation of the vertical field did not have an accompanying structure in the total ion density. This suggests that the vertical drift modulations are produced by polarization electric fields created by gravity wave neutral wind modulations in the plasma. Once the plume plasma drifts were replaced with background drifts, the background VL was then smoothed using an fast Fourier transform (FFT) smoothing algorithm [Press et al., 1986] to produce a smooth background vertical drift (VLs). VLs was then subtracted from the original VL to obtain a residual mesoscale vertical drift, δVL. The δVL contains both the gravity wave modulations and the large drifts within plasma plume structures. Results are shown in Figure 3. There are three consecutive passes plotted in Figure 3, which allows two local time periods with double-valued vertical drifts. Though the plasma plume vertical drifts are positive in the data in this paper, the downward drifting plumes are also excluded by the plume identification software.

Figure 3.

The equivalent E × B vertical drift (solid curve in Figure 3a) observed by the San Marco D electric field instrument. The background vertical drift VLS (dashed curve in Figure 3a). The mesoscale vertical drift δVL (Figure 3b).

4. Assimilation Into Global Ionosphere

[12] The San Marco D observations of electric fields are reduced to equivalent vertical plasma drifts and are then used to drive a global ionosphere model. Assimilating the global-scale plasma drift into a global model is problematic because each longitude sees the satellite pass over once every 90 min. To accommodate for the data sparseness, a longitudinal range of the satellite position is allowed to determine the drift at a single longitude within the model ionosphere. Assuming the background vertical drift is defined by Sun-synchronous physics, we can translate the local time structure using

equation image

where VLobs is the observed vertical drift at local time t at the longitude φo within the model. The φS is the satellite longitude. Here it becomes transparent that the local mesoscale variation of the vertical drift must be removed from the data for assimilation into global models because the gravity wave induced variation is necessarily bound to longitude, i.e., is not Sun-synchronous physics. The translation of data to a Sun-synchronous coordinate must be limited due to differences in the climatological vertical drift [Scherliess and Fejer, 1999] and to the time variation of the conditions creating the electric fields. These limits of equation (3) are best determined by a larger data set such as that provided by the future C/NOFS satellite. An effective limit is applied to equation (3) through an increased uncertainty as the satellite moves away for the modeled sector (φS).

[13] A second concern for the assimilation of the vertical plasma drift is the single apex altitude of the satellite's observation. In the modeled sector, how does the vertical plasma drift vary with altitude? It has been shown that it is approximately constant with altitude [Pingree and Fejer, 1987] and this is assumed within the ionospheric model up to 2000 km apex altitude, and then the vertical drift is linearly reduced to zero at 3000 km. The vertical drift at the apex altitude is less effective in lifting F region plasma which is now beyond the low-latitude region for the dipole field line with 3000 km apex altitude.

[14] The observation of VLobs is blended with the Scherliess and Fejer [1999] empirical model of vertical plasma drift, VLM, for a specific longitude according to the uncertainty of the measurement. This allows for a continuous and reasonable vertical drift even if the data contain a large gap or has large uncertainty. In determining the best vertical drift for the ionospheric model, one must know the associated uncertainty of empirical model results as well as the measurements to be assimilated. The uncertainty of an empirical model is primarily associated with the inability to reproduce the day-to-day weather. The uncertainty in the vertical drift model, δVM, is approximately 8–10 m/s for all local times [see Scherliess and Fejer, 1999, Figure 6]. The error of the measurement of the vertical velocity must also incorporate the uncertainty in the appropriateness of the value as used in the model. Three uncertainties are estimated:

[15] 1. The first is δVE, the measurement error (5 m/s) based on the uncertainty of systematic error (satellite pointing error etc., N. Maynard, personal communication, 2000).

[16] 2. The second is δVW, local weather variation (∼2 m/s) in the measurement. This was determined from the average deviation of the measurement, VL, from the smooth value, VLs. San Marco data between 1500 and 1800 LT on days 240–248 are used in this determination. The local weather (gravity wave perturbations, etc.) is not appropriate for the global model ionosphere. This is called the representation error.

[17] 3. The third is δVI, the inappropriateness of the measurement for assimilation to the model ionosphere. This should increase as the satellite moves away in longitude from the model sector assimilating the value. There is also a component of satellite altitude above hmF2. The vertical drift model and the ionosphere model are based on the peak measurements of vertical drifts.

[18] The uncertainty due to the inappropriateness of the measurement for the model is given by the following formula:

equation image

Equation (4) is assigned without rigorous study. The observation uncertainty in a combination of independent uncertainties becomes

equation image

The assimilation vertical velocity combines the model and observed value through a weighting of uncertainties [Parratt, 1971]:

equation image
equation image

[19] This least squares combination of model and observation is ready to drive a low-latitude ionospheric model. The global model used in the study is the ionospheric forecast model (IFM) [Schunk et al., 1997]. The IFM is a physics-based ionosphere solving for O+, H+, NO+, and O2+. It includes empirical models for the atmosphere [Hedin, 1991; Hedin et al., 1991], and the ion and electron temperatures [Titheridge, 1998]. The vertical plasma drift is input by means of a data ingestion module using the methods described above.

[20] There are little data available for verification of the model results that ingest San Marco D data. The San Marco satellite had plasma density measurements that can be compared against model ionosphere results for an almost independent measure of the degree of improvement of the data-ingestion over climatology. Figure 4 shows a comparison of the empirically driven model densities (short-dashed curves), the data-driven model densities (long-dashed curves), and the measured densities (solid curves) for two San Marco orbits. The large difference between the climatological run and the data is caused by the difference in altitude of the F region between the runs. The satellite is observing the F region bottomside, and a small difference in the vertical velocity history can cause large differences in density observations. The results in Figure 4 show promise for the assimilation of C/NOFS plasma drift and electric field measurements. Improvement in the determination of the inappropriateness uncertainty (equation (4)) is needed through a future study of results from C/NOFS and other equatorial satellites. The longitudinal inappropriateness of satellite data determines how the satellite observations are combined with ground observations of the vertical drift as well as the vertical drift model.

Figure 4.

Electron densities measured by the San Marco D satellite (solid curve) for 1988, day 240; the electron densities obtained from the IFM model using Scherliess and Fejer [1999] vertical drift model (short-dashed line); and the electron densities obtained from the IFM model using the San Marco electric fields measurements to obtain vertical drift (long-dashed line).

5. Mesoscale Structure Forecast

[21] On magnetically quiet nights during advantageous seasons, the F region ionosphere rises and plasma plumes are launched. The steep density gradients on the plume walls generate smaller-scale irregularities, which interfere with satellite-to-ground electromagnetic signals. The strong signal scintillation of communications and GPS occur as the latitudinally extended plasma plumes pass between satellite and ground receiver. Predicting scintillation occurrence prior to sunset involves a prediction of the vertical plasma drift near and after local ground sunset. If the ionosphere is lifted high enough, then Rayleigh-Taylor instability will initiate ESF. However, even if this PRE prediction is made with accuracy and the postsunset of scintillation is certain, then the spatial map of the forecast scintillation strength is general in character. The true character of scintillation is patchy and associated with plasma plumes. Predicting plume placement and spacing could provide added definition to the forecast scintillation maps. Using the limited San Marco data set, we examine the possibility of plume position prediction.

[22] It has been suggested over the years that the Rayleigh-Taylor instability was not sufficient to create plasma plumes without initial variations or seed perturbations in the plasma at sunset. Gravity waves were presumed to be involved in seeding of ESF due to the regularly spaced occurrence of plasma plumes [Whitehead, 1971; Davis, 1973]. The gravity wave seeding was thought to occur through a resonant relationship between the gravity wave and the initial plasma wave that allowed chemical decay of the plasma in the modulated atmosphere. This required a spatial resonance of the gravity wave and growing plasma modulation [Kelley et al., 1981]. However, Huang and Kelley [1996] showed that a vertical plasma drift perturbation generated by a gravity wave neutral wind modulation could produce plumes without spatial resonance requirements. This wind-induced plasma drift would show up as mesoscale modulations seen in Figure 2. These wave-like plasma drift perturbation may have direct correspondence to large plasma plumes observed after sunset through causal relationships modeled by Huang and Kelley [1996].

[23] We examine the correlation of presunset and postsunset plasma plume perturbations in the plasma drift to see if there is a precursor signature of individual plumes. Figure 5 plots several days of vertical drift observations against local time of the observation. Each plot combines data from two successive passes; the black curve is the first pass. The left column plots both the vertical drifts and the smoothed vertical drifts against local time. Conveniently, the first pass covers presunset and the second pass covers postsunset. The middle column plots the data against the geographic longitude of the satellite. The smooth background vertical drift is removed from the drift value in the right column; this is the mesoscale structure only. Local time marks are placed for reference.

Figure 5.

The left column is the VL observed by the San Marco satellite during 1988 on successive orbits of Kwajalein Island longitudes. The smoothed VL value is also plotted in the left column. In the middle column the two passes from each day are plotted against longitude. The correspondence between presunset gravity waves (black curve) and postsunset plasma plumes (red curve) is difficult to determine. The right column plots the difference between the observed VL and the smoothed VL leaving the mesoscale structure of the two orbits.

[24] The first day plotted (day 243) shows no plasma plumes developing after sunset. The right column shows that the waves presunset (black curve line) and postsunset (red curve) are typically anticorrelated rather than correlated. This is true on two other days of data with no plasma plumes. The waves do not grow into plumes because, presumably, the ionosphere did not rise enough. The neutral gravity wave, however, has traveled to a new position so the mesoscale velocity structure is now anticorrelated.

[25] Regarding the days with ESF plume occurrence (244, 245, 246, and 248), some correlation could be supposed from the alignment of waves to plumes in geographic space. Plumes long after sunset (>2200 LT) have suffered westward drifting and may not be in phase with the presunset seed signatures. However, the presunset seeds between 1700 and 1830 LT have strong association in position with postsunset plasma plumes between 1900 and 2000 LT. In particular, the coherence of the alignment is very strong for day 248. A case study of five evenings does not provide the basis for predictive algorithms but does provide hints of a possible precursor forecast of plume positions and spacing after sunset during ESF season. Even though the presunset gravity wave are probably traveling, the rising ionosphere launches plumes at the position of the gravity wave influence and freeze the location in the F region plasma. After the plasma plume growth moves into the nonlinear region, the plume position and the gravity wave need not remain colocated.

6. Summary

[26] Assimilation of equatorial satellite data has been examined in a three-step analysis of (1) plume identification and removal, (2) smoothing of plasma drifts to obtain global-scale vertical drifts to assimilate, and (3) combination of observed drifts with model drifts to provide suitable vertical plasma drifts at all longitudes to drive global low-latitude ionosphere models. The mesoscale variations of vertical plasma drift are removed in the smoothing process. The mesoscale features are used separately to determine the uncertainty of the vertical plasma drift due to representation error associated with global models. The standard deviation of the mesoscale features is approximately 2 m/s. The observed and climatology drifts are combined through an uncertainty-weighted least squares method. Then resulting global-scale vertical plasma drift variation is ingested into a global ionosphere model. In this limited case study, the resulting model electron densities along the satellite track are significantly closer to the satellite observations than if the model ionosphere uses only climatology drifts.

[27] The mesoscale structure in the vertical plasma drift is then used to forecast plasma plume development in the presunset hours. In particular, the mesoscale wave structure is nearly frozen in longitude as the solar terminator races past the (supposed) gravity wave induced vertical drift structure. The gravity wave induced structure imposes perturbations in the F region plasma, which result in plasma plumes, if the PRE is large enough, where the maximum upward velocity occurs in the mesoscale wave. The correlation to the presunset mesoscale waves in the vertical drifts with the development of plasma plumes in the postsunset hours hints that a predictive algorithm of plasma plume position and spacing might be successfully used to increase the fidelity of forecast scintillation maps. Data from the C/NOFS satellite to be launched in 2004 will provide substantial insight into the validity of plume predictions and resulting scintillation prediction maps.

Acknowledgments

[28] This work was supported by the Air Force Research Laboratory under Air Force contract F19628-00-C-0026. I thank Nelson Maynard for his suggestions on this paper and also his help in understanding the electric field measurements of the NASA San Marco D satellite.

Ancillary