Structure of the Earth s lower ionosphere observed by GPS/MET radio occultation

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 107, NO. A5, 1057, /2001JA900158, 2002 Structure of the Earth s lower ionosphere observed by GPS/MET radio occultation K. Hocke and K. Igarashi Communications Research Laboratory, Tokyo, Japan Received 5 February 2001; revised 10 November 2001; accepted 10 November 2001; published 23 May [1] The application of the radio occultation technique to the lower ionosphere is described and analyzed by means of the international reference ionosphere (IRI) and the GPS/Meteorology (GPS/ MET) experiment. Electron density profiles are derived from total electron content (TEC) measurements along radio links between GPS satellites and a GPS receiver on board the low Earth orbit satellite Microlab-1 during radio occultations. The retrieval method is based on Abel inversion and requires only spherical symmetry for the lower ionosphere. The profiles are sorted for geographic latitude, noon, and midnight in June July 1995 and February By means of a sliding window average, meridional slices of electron density are calculated from the sorted profiles. GPS/MET slices of electron density agree better with corresponding IRI slices for noon than for midnight. This is mainly due to the variable nature of particle precipitation from the magnetosphere, which is a significant ionization source, in particular for the lower ionosphere at high latitudes around midnight. Depending on ionospheric conditions, the applied retrieval method seems to be limited to the recovery of absolute electron density values exceeding m 3. This is possibly due to incomplete separation of E and F region contributions to the refraction of the GPS signal. Around noon, GPS/MET and IRI observe/predict a similar base height of the E region as well as similar absolute values of electron density in the E region. The radio occultation data also permit retrieval of meridional slices of average plasma irregularities (vertical scales <7 km) of the lower ionosphere at noon, midnight, winter, and summer. INDEX TERMS: 2427 Ionosphere: Ionosphere/atmosphere interactions (0335); 2439 Ionosphere: Ionospheric irregularities; 6969 Radio Science: Remote sensing; 6929 Radio Science: Ionospheric physics (2409); 6994 Radio Science: Instruments and techniques; KEYWORDS: GPS ionosphere sounding, radio occultation, E region, ionospheric structure, plasma irregularities, international reference ionosphere 1. Introduction [2] The global structure of the lower ionosphere between 70 and 130 km is relatively unknown, probably because of two reasons: 1. The lower ionosphere is highly variable. It is a transition region between the D and E regions as well as between the mesosphere and lower thermosphere. In this height range the thermal structure, ionization rate, and ion and neutral composition of the atmosphere are rapidly changing, and nonlinear processes occur (thermosphere-ionosphere-magnetosphere interaction; breaking/dissipation of waves from the lower atmosphere at the mesopause; variations of electric fields, particle precipitation, and photoionization due to changes of solar wind and solar radiation) [Thomas, 1996]. 2. Satellite measurements of electron density in the lower ionosphere have not been possible in the past. Electron density profiles of the lower ionosphere are mainly obtained by groundbased incoherent scatter radars (h = km), partial reflection radars (h = km), and ionosondes (h = km). Global coverage of these radars (especially of the expensive incoherent scatter radars) is not sufficient, measurements of partial reflection radars are limited to small electron density values in the D region, and inversion of ionosonde ionograms into electron density profiles is difficult because of group retardation of radio signal in the ionosphere [Hunsucker, 1991]. In situ measurements of electron density by rockets are not appropriate for continuous monitoring of the lower ionosphere on a global scale. Copyright 2002 by the American Geophysical Union /02/2001JA [3] Sounding of the atmosphere, ionosphere, and Earth s surface by radio links between GPS satellites and GPS receivers on board low Earth orbit (LEO) satellites seems to be an economic and accurate method which can achieve high temporal and spatial coverage in future multisatellite missions such as COSMIC [Lee et al., 2001; Gorbunov, 1996]. The concept of bistatic radar sounding of planetary atmospheres, ionospheres, and surfaces was introduced by Fjeldbo [1964] and has been applied in many space missions to the planets of our solar system up to the present day. The U.S. GPS meteorology experiment ( ) is a proof of concept for application of this method to the Earth s atmosphere and consists of a small GPS flight receiver on board the Microlab-1 satellite in low Earth orbit at 730 km height, measuring phase path and amplitude variations of GPS L1 and L2 signals [Rocken et al., 1997]. Some of these variations are caused by atmospheric refractivity. In the case of a GPS radio occultation (setting of a GPS satellite at the horizon as observed from the position of the LEO satellite), the GPS-LEO radio link successively scans the atmospheric layers at the Earth s limb from ionospheric heights down to the surface. This allows a simple retrieval of the height profile of atmospheric refractivity using Abel inversion. It may be also called the onion-peeling technique: the refractivity gradient of the first layer is used to calculate the refractivity gradient of the second layer, the refractivity gradients of the first and second layers provide the refractivity gradient of the third layer, and so on, layer by layer. [4] Previous publications on ionosphere sounding with GPS/ Meteorology (GPS/MET) radio occultation have concentrated on validation and retrieval techniques [Hajj and Romans, 1998; Schreiner et al., 1999; Tsai et al., 2001; Vorob ev et al., 1999]. SIA 4-1

2 SIA 4-2 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE GPS/MET results differ from simultaneous ionosonde measurements in particular for occultations taking place in the dawn and dusk sector or at high and low latitudes [Tsai et al., 2001; D. N. Anderson, private communication, 1999]. These deviations are due to retrieval errors of the Abel inversion and to unstable ionospheric conditions during the morning and evening hours, when a sudden change of solar heating and photoionization rates disturbs the thermosphere and ionosphere around the solar terminator [Beer, 1978; Galushko et al., 1998]. In the latter case the deviations between ground- and space-based observations reflect the natural variability of the ionosphere on small scales during disturbed conditions, since it is impossible to have the same location, time, and sounding volume size for both observation techniques. By means of simulation of radio occultation and subsequent retrieval of electron density, Hajj et al. [2000] find errors related to the Abel inversion of <20% for recovery of the F 2 peak layer electron density in a nonspherical model ionosphere. [5] Profiles of high accuracy and with resolved sporadic E layers have been derived for the ionosphere at midlatitudes around noon or midnight from GPS/MET data, when the spherical symmetry assumption of the Abel inversion is fulfilled [Hajj and Romans, 1998; Schreiner et al., 1999; D. N. Anderson, private communication, 1999]. Vorob ev et al. [1999] investigated the validity of the spherical symmetry assumption for recovery of lower ionospheric electron density and found only one case (out of 50) of significant discrepancy of the GPS phase and amplitude profile, indicating the influence of nonspherical ionospheric refractivity. Analysis of GPS ionosphere sounding data by means of ionospheric models (data assimilation or model-assisted tomography) can solve the inversion problem for nonspherical ionospheric conditions [Lee et al., 2001]. [6] The present paper is an application study using a large amount of radio occultations of the GPS/MET database. Contrary to previous studies, we try to separate GPS signal propagation in the lower ionosphere from the signal propagation in the F region. For the first time, average states of the lower ionosphere are derived from GPS radio occultation data and compared to IRI. The daily (noon-midnight) and seasonal (summer-winter) variations of the electron density distributions in the lower ionosphere are analyzed and presented as a function of geographic latitude and height. These electron density fields are then compared to distributions of electron density irregularities, also obtained from the GPS/MET radio occultation data set. 2. Radio Occultation Experiment [7] Here we describe the radio occultation experiment and the Abel inversion. The Abel inversion is required for recovery of electron density from the integral measurement of GPS phase path excess along the radio link between the GPS and the low Earth orbit (LEO) satellite. The application of the radio occultation technique to the lower ionosphere is analyzed. [8] The scheme of ionospheric limb sounding or radio occultation is depicted in Figure 1 in a proportional scale. The GPS satellite is at a distance of around 20,000 kilometers at the left-hand side (not shown in the figure). The transmitted L1 and L2 radio signals have wavelengths of and cm and are received by a GPS antenna in low Earth orbit (h = 730 km for Microlab-1 satellite of the GPS/MET mission). Two horizontal radio links are drawn from the GPS to the LEO satellite, one with a ray perigee or tangent point height at 140 km and the other at 60 km. The radio links are close together and indicate the upper and lower borders of the lower ionosphere at the Earth s limb. During a radio occultation the radio beam moves with a geocentric radial velocity of around 2 3 km/s and scans the whole lower ionosphere, layer by layer, within a time of around 30 s. The height resolution is around 1 2 km, corresponding to the size of the first Fresnel zone. By means of Figure 1. Scheme of GPS limb sounding of the lower ionosphere in a proportional scale. The GPS signals are transmitted by a GPS satellite at x = 20,000 km (not depicted) and are received by a GPS receiver in low Earth orbit (LEO). Two horizontal radio links are drawn with tangent point heights at 60 and 140 km, indicating the lower and upper borders of the lower ionosphere at the Earth s limb. advanced data analysis methods, such as synthetic aperture or radio holography, the GPS phase and amplitude data recorded along the LEO trajectory can provide a vertical resolution of 100 m [Pavelyev, 1998; Vorob ev et al., 1999; Karayel and Hinson, 1997; Mortensen et al., 1999]. [9] In the present study we assume a straight line for the radio link in order to estimate the ray perigee height. This assumption may cause a height error of around 1 km for the averaged profiles of the TEC gradient. We will calculate this correctable error later. An exact data analysis under consideration of curved ray paths has been described by Hajj and Romans [1998] and Vorob ev et al. [1999]. However, formulas and illustration of the main problems of the retrieval are easier in the case of straight line assumption, which is selected here. The difference of the L1 and L2 phase path excess is used for derivation of total electron content as a function of ray perigee height. The data center of the GPS/MET mission at University Corporation for Atmospheric Research (UCAR) in Boulder provides the absolute phase path excesses derived by the double difference method using fiducial data of a GPS ground receiver. This double differencing removes clock errors of transmitter and receiver [Melbourne et al., 1994]. For the present study we actually do not need double-differenced data. Ephemeris errors, phase cycle slips, and clock instabilities would be removed by the difference of L1 and L2 phase path [Vorob ev et al., 1999]. For each occultation the phase path difference is calibrated to a value of zero at the top height of the ray perigee at around km. The excess of phase path difference is measured for all layers below with reference to this start value. The absolute value of the phase path difference is not known, but for atmosphere/ionosphere profiling we are interested only in the vertical gradient of phase path difference. [10] Advanced GPS receivers such as the space-qualified TurboRogue on board Microlab-1 measure the relative phase path with an error of 0.1 mm for a 1 s average [Melbourne et al., 1994, Figure 5.1]. This corresponds to an error of relative TEC of around el/m 2 for a 1 s average of an individual occultation.

3 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE SIA 4-3 According to Schreiner et al. [1999], the TEC profile, T(r o ), is determined by Tr ð o Þ ¼ f :3 S 1ðr o Þ þ k 1 ¼ f :3 S 2ðr o Þ þ k 2 ¼ f1 2f :3 f1 2 f 2 2 ½S 1 ðr o Þ S 2 ðr o ÞŠ þ k: [11] Units are [electrons/m 2 ] for T, [Hz] for GPS L1 and L2 frequencies, f 1 = GHz and f 2 = GHz, and [m] for phase path excess S 1 and S 2. The term r o denotes the radial distance of the ray perigee. The constants k 1, k 2, and k are not interesting since the data retrieval is based on the differences of TEC, e.g., the vertical TEC gradient. The constants are related to the signal phase ambiguity (e.g., clock biases, residual orbit errors). [12] T(r o ) is also given by the integrated electron density along the radio link. By means of Figure 1 we get following formulas: Z xtop Ty ð o Þ ¼ n e ðx; y o Þdx x Z top rtop rn e ðþ r Tr ð o Þ ¼ 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi dr: r o r 2 ro 2 [13] In the case where we have a complete profile of T(r o ), observed during a radio occultation, the Abel integral equation can be solved for the interesting parameter electron density n e (r) [Fjeldbo et al., 1971; Schreiner et al., 1999]: n e ðþ¼ r 1 p Z rtop r ð1þ ð2þ 1 dtðr o Þ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi : ð3þ ro 2 r2 [14] Calculation of the electron density profile n e (r) requires only the vertical gradient of TEC, dt(r o )/. In the Earth s ionosphere the vertical electron density gradient changes its sign at the F 2 peak. If a strong E layer is present, then the sign of the electron density gradient additionally changes in the valley between the E and F regions and at the E layer peak. The TEC gradient changes its sign in particular around the F 2 peak [Hajj and Romans, 1998]. In the case of integration along the radio ray through the F region (equation (3)), the lower F region cancels to some extent the dispersion of the GPS signal by the upper F region (as indicated by small values of observed TEC gradients at tangent point heights around km). This pleasant effect is valid for a nonspherical ionosphere as well as for high and low electron densities in the F region. The nonspherical nature of the ionosphere is often discussed as a crucial problem of ionospheric limb sounding and Abel inversion. However, recovery of electron density by Abel inversion is often better than expected. For example, linear horizontal gradients of electron density inside an ionospheric layer do not cause retrieval errors of the Abel inversion [Hajj et al., 2000]. [15] Now the situation for retrieval of electron density of the lower ionosphere is analyzed. Figure 1 indicates the piercing of the F 2 peak layer at h = 350 km by the radio rays with ray perigee heights at 60 and 140 km. The spherical symmetry assumption of the Abel inversion means that the electron density value of the F 2 peak, determined by a previous radio ray with tangent point at x = 0 km and y = r earth km, is used at the depicted piercing points. This is a bit questionable, since the piercing points are 14 apart from the place of determination of the F 2 peak electron density. Within 14 the F 2 layer may have a significant change of its characteristics, which would introduce a retrieval error of the small electron density values of the lower ionosphere. On the other hand, the horizontal distance of the F 2 peak piercing points of the radio rays at perigee heights h = 60 and 140 km is just 300 km, which is a rather small scale for the F region, where enhanced diffusion and thermal conduction suppress atmospheric variations of such scales. Thus the whole lower ionosphere is sounded while the radio rays traverse through almost always the same F layer. Neglect of spatial and temporal variations of the F layer during the sounding process of 30 s for the lower ionosphere should be justified. If there are variations, then these variations have possibly a stochastic nature, so that they disappear by averaging of a sufficient amount of radio occultation profiles. Our approach roughly assumes a constant contribution of the F region to the observed TEC gradient, and spherical symmetry assumption is used only within the sounded area of the lower ionosphere. The spherical symmetry assumption is often not fulfilled for the part of the lower ionosphere because of horizontal variations of the E region at scales of around 100 km. This obstacle diminishes by statistical analysis of occultation profiles. [16] The electron density profiles are calibrated at h = 60 km to a value of zero as described in the work of Vorob ev et al. [1999]. We additionally use the constraint that the electron density gradient between 60 and 70 km should be positive (dn e /dh 0). If a negative linear trend exists, then it is interpreted as bias due to the upper ionosphere and is removed. The TEC gradient observed below 140 km contains a lot of information about the local TEC trend of the upper ionosphere. This information can be used for automatic data processing and optimal removal of the TEC trend. [17] In the present study, information obtained by radio rays with tangent points beyond 140 km is not used. This is due to three reasons: (1) We do not have these GPS/MET data yet. (2) We like to show that the retrieval of the lower ionosphere does not depend on the spherical symmetry assumption for the F region. (3) Our data analysis method can be applied to occultation experiments which do not perform a continuous recording of the phase path from the upper F 2 region down to 60 km height and to LEO satellites which have orbit heights around km. 3. Concept of Data Analysis [18] We present a new concept of data analysis of occultation profiles observed at arbitrary places and times. The reader will see that this new concept is quite old. It is well-approved for the international reference ionosphere (IRI) and consists in sorting and archiving of ionospheric observations obtained at various places and times according to their latitude, longitude, local time, geomagnetic activity, and other parameters [Bilitza et al., 1993]. [19] Since a GPS receiver in low Earth orbit observes only a few hundreds of atmospheric profiles per day, the data are in general not sufficient for real-time studies and radio tomography. The situation will change when future multisatellite communication systems in low Earth orbit are equipped with GPS receivers. For understanding of the ionosphere and description of its average states, continuous data of a single satellite can be sufficient. Unfortunately, the GPS receiver on board the Microlab-1 satellite suffered a bit under the encryption of the GPS signals (antispoofing on mode). This encryption of the P code (precise code of GPS) is usually turned on and protects the P signals from being spoofed through the transmission of false GPS signals by an adversary. For the civilian user, who does not know the encryption, it results in reduced quality of the GPS phase data (P code of L2 signal). In GPS/MET data the L2 phase path noise is significantly increased during GPS antispoofing on. High-quality data (0.1 mm phase path noise for L1 and L2) exist only for three long time intervals of around 2 3 weeks in June July 1995, October 1995, and February 1997, when the U.S. Department of Defense switched off the antispoofing encryption of the GPS system. The new radio occultation experiment on the CHAMP satellite (launch in summer 2000) continuously records occultation profiles of high quality (around 230 per day). The enhanced noise of the L2 signal phase path during GPS antispoofing on is compensated by an advanced high-gain GPS antenna and receiver system with a high signal to noise ratio [Wickert et al., 2001]. Accurate occultation profiles are

4 SIA 4-4 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE Figure 2. GPS/MET observations of the average vertical gradient of total electron content (dt/dr) as function of tangent point height (Figures 2a and 2c) and retrieved, average electron density profiles (solid lines in Figures 2b and 2d): (a and b) 25 S, noon, June July 1995; (c and d) 60 N, midnight, June July The maxima of the TEC gradient are pointing to the bottom side of the E region, where the vertical change of electron density is maximal. now obtained for both GPS antispoofing on and off. Because of the CHAMP commissioning phase, these occultation data are not available for the present study. [20] Regular behavior and small details of the ionosphere can be extracted by means of the law of error and use of long data sets. Noise and stochastic effects disappear by averaging of a sufficient amount of observations, while the signal of the geophysical phenomenon remains. A famous example is the detection of the gravity effect of the Moon on the Earth s upper atmosphere by sorting and averaging a long time series of noisy ionospheric observations according to Moon time [Chapman, 1918]. In the present study the GPS/MET data are arranged as a function of local time of the Sun. Diurnal variations of the lower ionosphere according to daily change of photo ionization rate, particle precipitation, and tidal waves from the lower atmosphere are emphasized. This periodogram method supports the harmonic oscillations of 24 hours (24/n hours with n =1,2,3,...), while variations due to random atmospheric waves or plasma irregularities are suppressed. A previous study, using time series of electron density observed by the European Incoherent Scatter (EISCAT) radar in northern Scandinavia, showed a rather harmonic spectrum with spectral peaks at 24, 12, 8, 6,... hours for the lower ionospheric electron density [Hocke, 1996]. A time interval of around 30 days is commonly recommended for determination of mean characteristics of tidal variations in the mesosphere/lower thermosphere [e.g., Vial and Forbes, 1989]. The observation time intervals of the GPS/MET mission are near this recommendation, so that our data analysis approach is meaningful. [21] The future amount of available space-borne GPS data will allow detailed statistical studies. Global ionospheric changes due to solar variations, atmosphere dynamics, and anthropogenic effects can be described by GPS ionosphere sounding data. Regular effects of the ionosphere at the solar terminator (dusk and dawn side) and shape of the auroral ovals can be extracted from the GPS data by statistical methods. For example, GPS data can be sorted for geomagnetic activity and magnetic storms (J. Aarons, private communication, 2001). The inverse problem of radio tomography is solvable with a superior spatial resolution for a statistical ensemble of worldwide GPS data by renunciation of the chronology of measurements. (For example, an average meridional slice of the ionosphere can be calculated for magnetic midnight, December solstice, geomagnetic activity Kp > 5 by collection of appropriate GPS-LEO radio links over a sufficiently long time interval. The well-determined tomographic solution will describe the average behavior of the ionosphere in case of high geomagnetic activity.) An organized and open database of space-borne GPS data will be an invaluable tool for future ionospheric and atmospheric research as well as for space weather applications. Previous work related to the international reference ionosphere (IRI model) is exemplary for a GPS ionosphere database. 4. Data Analysis 4.1. Data Selection [22] The GPS/MET database contains three prime times of 2 3 weeks of observation time, when antispoofing mode of the

5 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE SIA 4-5 GPS system has been turned off in order to enable a higher accuracy of atmosphere profiling by GPS/MET. We select the two prime times that have the most occultation events and that are near solstices (19 June to 10 July 1995, 1900 occultations; 2 16 February 1997, 2690 occultations). The GPS receiver on Microlab-1 collected on average occultations per day. The geomagnetic activity was during both prime times low to moderate with Kp around 2 3. [23] The GPS/MET occultations around noon ( LT) and midnight ( LT) are selected for each prime time. These occultations are sorted corresponding to geographic latitude of occultation place. Geographic latitude has been favored against geomagnetic latitude since the lower ionosphere depends on neutral dynamics (e.g., tidal winds) and solar radiation. For the study of magnetospheric influences on the ionosphere, geomagnetic coordinates are better. In the following, ephemeris data of GPS and LEO satellites and L1 and L2 atmospheric phase path excess with a time step of 0.02 s (50 Hz sampling rate) are used (level 2 data files of the GPS/MET database at UCAR). The 50 Hz high rate occultation data usually start at around km height and go down to the surface Data Analysis of Electron Density Profile and Retrieval Test [24] All available occultation events are considered. For a set of arranged profiles of the vertical TEC gradient, dt(r)/dr, the average field is derived as a function of geographic latitude and height by means of a sliding window average of 20 in latitude and a step of 0.6. Figure 2 depicts two examples of the average TEC gradients (Figures 2a and 2c) and the retrieved electron density profiles (Figures 2b and 2d). The numbers of averaged profiles are around 80 in Figure 2a and 45 in Figure 2c for heights below 110 km. At heights beyond 110 km the numbers of available observations are significantly lower, so that the profile in Figure 2a becomes noisier. The profile in Figure 2a is derived from occultations at 25 S around noon in June July The clear maximum of the TEC gradient at around h = km indicates the sudden increase of electron density at the bottom side of the E region as shown by the solid line in Figure 2b. Later, we will compare this to the corresponding IRI profile, which agrees well. During nighttime the observation of the lower ionosphere is more difficult, since a dense F 2 layer still exists, but electron densities of the E region are often below 10 9 or m 3, which is near the errors introduced by ionospheric noise, radio occultation, and retrieval technique. Figure 2c shows the average TEC gradient at 60 N around midnight in June July The TEC gradient clearly indicates an increase of electron density at the base of the E region. Similar profiles are also observed at higher latitudes, so that we can be sure that these measurements are reliable. The increase of electron density is probably due to solar illumination during midsummer night at high latitudes and agrees well with the IRI prediction, which is shown later. The examples demonstrate that the lower ionosphere generates a clear signal in the vertical TEC gradient which is detectable by GPS radio occultation from space. [25] We use now the average TEC gradients of Figure 2 for calculation of the distortion of the curved GPS ray path from the straight line. According to Fjeldbo et al. [1971], the bending angle a is related to the gradient of the signal phase f by aðþ r ¼ l dfðþ r ¼ dsðþ r : ð4þ dr dr [26] Here the impact parameter p (product of refractive index and tangent point radius) has been substituted by the tangent point radius r, since the refractive index is close to 1 at E region heights. l is the radio wave length. By means of differentiation of equation (1), the bending angle a is aðþ r ¼ 40:3 f 2 dtðþ r ; ð5þ dr where f is the radio wave frequency in [Hz]. This relation depicts the importance of the TEC gradient. The angles of arrival of the L1 and L2 signal and also their difference are proportional to the TEC gradient. In Figure 2a the maximal TEC gradient is around m 3. Together with f = f 2, the maximal bending angle of the L2 signal, a 2, is around rad in Figure 2a. The distance between the tangent qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi point at h = 100 km and the LEO satellite position is d ¼ rleo 2 ð r e þ hþ 2 ¼ 2900 km. The vertical displacement r of the curved ray path with respect to the straight line is at the tangent point r d a: [27] For the maximal bending angle of the L2 signal we get a displacement r of around 600 m. In the case of Figure 2c the maximal displacement r is around 150 m for the L2 signal. The distortion of the L1 signal is even smaller, because of the higher frequency of L1. In the present study the displacement of the ray perigee from the straight line is neglected. However, by means of the last two equations one has the possibility to calculate a firstorder correction for the displacement using the information of the vertical TEC gradient. [28] The next steps are the description and test of our retrieval using simulated occultation data of IRI-95 at noon and midnight in July For the retrieval test we take IRI-95 electron density profiles at several geographic latitudes between 90 S and 90 N and calculate the corresponding TEC profile. The IRI model is an empirical description of the Earth s ionosphere. IRI has a long tradition and is based on ionospheric observations obtained by rockets, satellites, and ground-based radars [Bilitza et al., 1993, Bilitza, 2001]. For simplicity, horizontal gradients of the IRI ionosphere along the ray path are neglected in the case of integration of TEC according to formula (2). We already mentioned that our retrieval just removes the whole contribution of the F region by the assumption that the F region contribution is constant or a linear trend for rays with perigee heights between 60 and 140 km. In addition, the horizontal distance of the piercing points of the F 2 peak layer is just 300 km (Figure 1). From the TEC profile of IRI we use only the part below 140 km height for recovery of electron density, in order to have the same lack of data as in the case of GPS/MET high-rate data. The integral of equation (3) is split into two terms, B and C: A ¼ B þ C; 1 p Z rleo r eþ140 km n e ðþ¼ r 1 Z reþ140 km 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi p r ro 2 r2 1 dtðr pffiffiffiffiffiffiffiffiffiffiffiffiffiffi o Þ ; ro 2 r2 ð6þ dtðr o Þ ð7þ where r e is the Earth radius. According to previous considerations (Figure 1 and section 2), the integral of the TEC gradient along the ray path through ionospheric layers beyond 140 km is approximated by means of a constant k: C ¼ 1 p Z rleo r eþ140 km 1 p k Z rleo r eþ140 km 1 dtðr pffiffiffiffiffiffiffiffiffiffiffiffiffiffi o Þ ro 2 r2 1 pffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ C a : ð8þ ro 2 r2

6 SIA 4-6 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE Figure 3. Test of the retrieval. Dotted lines are true electron density profiles (IRI, July 1995, noon), and solid lines are the retrieved profiles. Retrieval errors are generally <50%. The retrieval recognizes the different vertical structures of electron density in the southern winter and northern summer ionospheres. Geographic latitude is shown as the title of each viewgraph. The dash-dotted lines show the contribution of the F region as given by integral C in equation (7). [29] The constant k is estimated at h = 60 km, where the electron density of equation (7) is assumed to be zero: k ¼ R reþ140 km 1 dtðr pffiffiffiffiffiffiffiffi o Þ r ro 2 r2 R rleo ffiffiffiffiffiffiffiffi 1 r eþ140 km ro 2 r2 p ; r ¼ r e þ 60 km: ð9þ [30] The integral in the denominator is analytically solved by using the relation R ðx 2 a 2 Þ 1=2 p dx ¼ lnðjxj þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffi x 2 a 2 Þ. By means of k and the approximated term C a (instead of C), the electron density profile n e (r) in equation (7) is determined. The dash-dotted electron density profiles in Figures 2b and 2d depict the results after performing these steps. The dotted lines just correspond to the integral B, which is in most cases very near the final solution. The dash-dotted profiles usually have a constant inclination to the left-hand side. This linear trend is estimated at heights between 60 and 70 km. If the electron density decreases from 60 to 70 km (dn e /dh < 0), then the linear trend is removed. So, the retrieval method estimates and removes the trend of the F region in an iterative manner. Finally, the profiles are shifted to n e (h)=0ath = 60 km. The last step is the most simple but efficient step. It has already been used in the data analysis by Vorob ev et al. [1999], Igarashi et al. [2000b], and Pavelyev et al. [2002]. For the IRI retrieval test, our retrieval method resulted in better agreement (by 10 50%) for the profiles at noon around the equator and in particular for the profiles at midnight. We may also mention that the approximated term C a is considered at all latitudes at midnight, and the retrieval is performed without a change of program parameter. In the case of the noon ionosphere, C a is weighted by C a (dt(h = 60 km)/dh/max) 2, where max denotes the maximum of the vertical TEC gradient measured at the equator at h =60km around noon. [31] The results of the retrieval test are shown in Figure 3 for noon in June The dotted line is the true IRI electron density profile, while the solid line is the profile retrieved by using solely TEC gradient data of rays with tangent points below 140 km. The dash-dotted line shows the trend or constancy of the TEC gradient caused by the F region (given by the integral C in equation (7)). In general, we find an excellent agreement between the retrieved profiles and the true IRI profiles. The test results for the midnight ionosphere in June 1995 are shown in Figure 4. Please note that the scale of the electron density axis has changed. In some cases the TEC gradient of the F region disturbs the retrieval, in particular at low and midlatitudes of the summer hemisphere, where the F region is upward lifted, forming a relative dense layer around h = km. At heights beyond 120 km the retrieved electron density values are often overestimated since it becomes more difficult to separate the E and F region contributions to the refraction of the GPS signal. However, we can see that the retrieved profiles show all main features of the midnight lower ionosphere of IRI along the meridian from the winter to the summer hemisphere. We conclude from this test that the retrieval method is appropriate for a preliminary analysis of the GPS/MET

7 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE SIA 4-7 Figure 4. Same as Figure 3, but for midnight. data of the lower ionosphere. The results of the statistical analysis of GPS/MET data in section 5 will give more insight about possible application of the radio occultation technique for sounding of the lower ionosphere at different seasons, local times, and latitudes Data Analysis of Electron Density Fluctuation Profile [32] The electron density fluctuation profile is directly obtained from the observed phase path fluctuations [Hocke and Tsuda, 2001a, 2001b]. The TEC fluctuation profile T(h) is estimated by the difference of T(h) and its average profile T (h), provided by a vertical 7 km sliding window average. The observed phase path variations are mainly due to electron density variations near the tangent point in the lower ionosphere. As discussed in section 2 and shown by Figure 1, the properties of the radio occultation experiment (aspect angle between layered structures and GPS-LEO ray, enhanced interaction length or time of the GPS signal in the sounding volume around the tangent point) and the natural scales of the ionosphere (result of a rapid enhancement of atmospheric diffusion and thermal conduction from the E to the F region) justify this assumption. In addition, ionospheric fluctuations of the F region would be randomly projected into limb-sounding TEC data observed at tangent point heights below, so that F region irregularities would mainly appear as noise (constant with height) in a statistical analysis of lower ionospheric irregularities. Thus the electron density fluctuation profile can be approximated by n e ðhþ ¼ Th p ð Þ 2 ffiffiffiffiffiffiffiffiffiffi ; ð10þ 2rr where r is the radial distance to the Earth center and r is the vertical scale of the fluctuation. In the present study we assume constant values r = 6472 km and r = 0.6 km. The choice of these values is not so important, since we discuss only the relative distribution of the electron density fluctuations as a function of latitude and height, and not the absolute values of the fluctuations. [33] A more serious error or better uncertainty is that the plasma irregularities are separated by means of high-pass filtering (subtraction of 7 km sliding window average). It is likely that the rapid vertical increase of electron density at the base of the E region is partly interpreted as fluctuation. This is also a principle problem of definition of background and perturbation. For example, sporadic E has a quite regular occurrence, often associated with tidal wind shears [Mathews, 1998]. Averages of many profiles obtained at the same location and at the same local time include this sporadic E. The question is then if sporadic E is acceptable as part of the background profile obtained by this average method. 5. Results and Discussion 5.1. Meridional Slices of Electron Density [34] The average meridional slices of the electron density in the lower atmosphere during February 1997 are depicted in Figure 5 for GPS/MET (Figures 5a and 5c) and IRI-95 (Figures 5b and 5d), where Figures 5a and 5b correspond to the ionosphere around noon and Figures 5c and 5d correspond to the ionosphere around midnight. The color scale is the same for GPS/MET and IRI. The number of averaged GPS/MET profiles within the moving 20 latitude window is depicted in the top panels of Figures 5a and 5c. These curves show where the statistics are relatively good or bad.

8 SIA 4-8 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE Figure 5. Meridional cross sections of electron density in the lower ionosphere in February 1997: (a) GPS/MET observations at noon, (b) IRI empirical model at noon, (c) GPS/MET at midnight, and (d) IRI at midnight. The top panels of Figures 5a and 5c show the number of averaged radio occultation events as a function of latitude. Color scales are the same in all plots for the sake of comparison. See color version of this figure at back of this issue. [35] On the first view the magnitude of electron density, the vertical electron density gradient around h = 95 km, and the latitudinal dependence of electron density agree for GPS/MET and IRI in Figures 5a and 5b at noon. In particular, the height dependence, with a sharp increase of ionization at km height, is similar for GPS/MET and IRI. At midnight, IRI predicts lower electron densities. IRI shows some enhancement of ionization at high latitudes (due to solar radiation at solstice and particle precipitation from the magnetosphere), but it is almost 10 times smaller than the GPS/MET observation. The curve in the top panel shows poor statistics of the GPS/MET average at high latitudes. The GPS/MET slice depicts a clear difference between the summer and winter hemisphere during nighttime. Ionization is found to be minimal in the lower ionosphere of the winter hemisphere at N. [36] The GPS/MET slice around noon in Figure 5a reveals an increase of ionization in the auroral zones, in particular in the northern winter auroral zone, where enhanced electron density is observed down to 75 km height. Collis et al. [1996] present and discuss EISCAT incoherent scatter observations of a long-lasting (>2 hours) energetic electron precipitation occurring around local noon in the northern auroral zone in March The electron density observed by EISCAT exceeds a value of el/m 3 at h = 65 km. High electron densities exceeding el/m 3 at D region heights are often observed in the high-latitude ionosphere by the EISCAT radar, which is located at geographic latitude 69 N [Kirkwood and Collis, 1987]. D region electron densities observed by a MF radar at Poker Flat (geographic latitude 65 N) in Alaska are higher by a factor of around 2 than IRI from December 1997 to March 1998 [Igarashi et al., 2000a]. Due to the variable nature of particle precipitation into the polar caps and auroral zones, it is of course a difficult task to derive or predict mean states of the ionosphere at high latitudes. The D region described by IRI is mainly based on a limited number of precise, in situ rocket measurements, so that the statistics is still a problem for IRI at D region heights [Bilitza, 2001]. We conclude from the GPS/MET slices that radio occultation data contain detailed and reasonable information on ionization structures. [37] Summer in the northern hemisphere (June July 1995) is depicted in Figure 6. GPS/MET has magnitudes and gradients of electron density similar to IRI at noon (Figures 6a and 6b). The ionization of the northern summer ionosphere is a bit stronger for the GPS/MET observation, and the ionization has no decrease northward of 30 N. It is remarkable that there is a discontinuity in the horizontal electron density distribution at 30 N, in particular at heights around 90 km. Associated with this discontinuity we find later a significant increase of plasma irregularities (sporadic E) in the GPS/MET observations and in simultaneous ionosonde measurements. [38] Comparing Figures 6c and 6d, the agreement between the ionization observed by GPS/MET and IRI in the northern summer polar region is obvious. This enhancement is probably due to solar

9 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE SIA 4-9 Figure 6. Meridional cross sections of electron density in the lower ionosphere in June July 1995: (a) GPS/MET observations at noon, (b) IRI empirical model at noon, (c) GPS/MET at midnight, AND (d) IRI at midnight. The top panels of Figures 6a and 6c show the number of averaged radio occultation events as a function of latitude. See color version of this figure at back of this issue. photoionization during midsummer night at high latitudes. At the south pole (winter) we find enhanced ionization down to 75 km height, which could be caused by particle precipitation over the polar cap [Winnigham and Heikkila, 1974; Meng and Kroehl, 1977]. However, the statistics of available GPS/MET occultations in the southern polar cap is poor. In agreement with Figure 5c, the ionization of the winter hemisphere is minimal from around 20 to 60 S Meridional Slices of Plasma Irregularities [39] The fluctuations with vertical scales <7 km are separately determined for each radio occultation. Then, the fluctuation profiles are arranged in the same manner as described for the electron density profiles in the previous section. [40] Figures 7a and 7b depict the average distributions of plasma irregularities around noon in February 1997 and June July 1995, respectively, observed by GPS/MET. In Figures 7c and 7d, decreased plasma irregularities at midnight are shown for both GPS/MET prime times. The irregularities around noon occur just below 100 km height. In particular, the irregularity distribution at noon in Figure 7a suggests a modulation of the plasma by tidal winds [Mathews, 1998], since the red layer of enhanced irregularities shows a sinusoidal shape. The Hedin wind model [Hedin et al., 1996] predicts a strong meridional wind shear with increase of poleward wind up to 50 m/s at southern latitudes and at noon in February The poleward wind is above 100 km height and could shift plasma via ion-neutral collision from the upper to the lower E region. In addition, the Hedin wind model predicts an eastward wind of around 25 m/s at 95 km height decreasing to 0 m/s at heights beyond 100 km. The eastward wind below 100 km height may cause an upward E B plasma drift [Mathews, 1998], so that a strong electron density gradient and plasma irregularities could be formed at around 100 km height. Comparison between distributions of electron density obtained by GPS radio occultation and neutral wind field measurements of the mesosphere/lower thermosphere (e.g., the future TIMED mission) can show the influence of neutral dynamics on the lower ionospheric plasma. With a deeper understanding of the effective processes, neutral and plasma dynamics of the lower ionosphere may be described by continuous monitoring of the electron density distribution by GPS radio occultation. [41] In the auroral ionosphere of the northern summer hemisphere (Figure 7d), two layers between 90 and 100 km and a layer in 120 km height appear (statistics are good). In the southern summer auroral ionosphere (Figure 7c), layers are present at 95 km and at 110 km (statistics are not so good). [42] In June July 1995, enhanced irregularities (Figure 7b) are centered around 30 N, at the same location where a possible horizontal discontinuity of the northern summer ionosphere occurs (Figure 6a). Also, around midnight some plasma irregularities are at this location (Figure 7d). The significant enhancement of plasma irregularities observed by GPS/MET has been simultaneously

10 SIA 4-10 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE Figure 7. Average distribution of plasma irregularities (vertical scales <7 km) observed by GPS/MET radio occultation: (a) around noon in February 1997, (b) around noon in June July 1995, (c) around midnight in February 1997, and (d) around midnight in June July See color version of this figure at back of this issue. measured by the Asia-Australia ionosonde chain. Figure 8 depicts a strong enhancement of the monthly average of plasma peak frequency foes at around 30 N (noon in June 1995, virtual height of Es (h 0 Es) is around km, confirming the GPS/MET observations in Figure 7b). The reason for this sharp increase of sporadic E activity is not known. For example, it could be related to the Earth s topography of the northern hemisphere, which possibly has a significant influence on the surface wind field and Figure 8. Geographic locations of the Asia-Australia ionosonde chain. Monthly average of plasma peak frequency, foes, observed by the ionosonde chain at noon in June 1995 is depicted as a function of geographic latitude.

11 HOCKE AND IGARASHI: STRUCTURE OF EARTH S LOWER IONOSPHERE SIA 4-11 on upward propagating gravity waves, tides, and planetary waves, interacting with the plasma of the lower ionosphere via ion-neutral collisions [Hocke and Tsuda, 2001a]. 6. Concluding Remarks [43] Our study has been focused on the application of radio occultation data for the study of the lower ionosphere. We find that the arbitrarily distributed radio occultation events can be analyzed by sorting the measurements according to local time, latitude, longitude, geomagnetic activity, or other parameters. This method is similar to the concept of the international reference ionosphere (IRI), which we extensively used for retrieval tests and comparisons. [44] We discussed the inversion of electron density and the relative new problem of ionospheric limb sounding in a detailed and illustrative manner, since GPS radio occultation is the first space-borne remote-sensing technique which is able to sound the lower ionospheric electron density through the dense F region. The retrieval method requires spherical symmetry assumption for the lower ionosphere but not for the F region. A test of our data analysis method which removes the vertical TEC gradient due to the F region indicates a retrieval error of electron density less than around 50% for electron density values greater than m 3 (depending on ionospheric state). Radio occultation and the retrieval method are in particular appropriate for detection and recovery of vertical electron density gradients at E layer heights. The retrieval test has been carried out by using electron density profiles at geographic latitudes from 90 S to 90 N of IRI at noon and midnight in July [45] Using the radio occultation data sets of the GPS/MET prime times in June July 1995 and February 1997, meridional cross sections of the lower ionospheric electron density are determined for noon and midnight. Since the prime times are near solstices, the change of the lower ionosphere from winter to summer hemisphere is described. The comparison to meridional slices of IRI generally shows a good agreement around noon. At high latitudes, GPS/MET observed enhanced electron densities down to h = 75 km. These enhancements are probably due to particle precipitations over the polar caps and auroral zones. [46] The high vertical resolution of GPS/MET radio occultation allows the study of discontinuities, small- and large-scale structures of the lower ionosphere. These structures can be interpreted as interaction of atmospheric winds and waves with the plasma but also as the influence of electric fields, particle precipitation, and solar radiation on the ionization distribution of the lower ionosphere. Meridional slices of plasma irregularities with vertical scales <7 km are retrieved and discussed together with the electron density slices. The irregularities have a clear diurnal and seasonal variation, and they mostly occur at the bottom side of the E region (just below 100 km height in the noon viewgraphs). The maximum of sporadic E activity appears in the northern, summer hemisphere at around 30 N. This enhancement of sporadic E is simultaneously observed by the Asia-Australia ionosonde chain. [47] GPS/MET radio occultation has demonstrated that it can significantly contribute to our knowledge on the structure of the Earth s lower ionosphere. Because of improved GPS receiver technology (for the purpose of atmosphere sounding) and because of future multisatellite missions with space-borne GPS receivers, a huge database of precise GPS radio occultations is expected for the near future. [48] Acknowledgments. We are grateful to Chris Rocken (UCAR, Boulder) for data provision. The GPS/MET teams at UCAR (Boulder) and JPL (Pasadena) are thanked for the GPS/MET experiment and raw data analysis of GPS/MET level 2 data. We are grateful to J. Wu and P. Wilkinson for Chinese and Australian ionosonde data. Professor Toshitaka Tsuda and the Radio Science Center for Space and Atmosphere (Kyoto University) are thanked for support of radio occultation data analysis of this study. One author (K.H.) thanks the Telecommunications Advancement Organization (TAO) of Japan for a research fellowship. [49] Hiroshi Matsumoto thanks D. Bilitza and another referee for their assistance in evaluating this paper. References Beer, T., On atmospheric wave generation by the terminator, Planet. Space Sci., 26, , Bilitza, D., International Reference Ionosphere 2000, Radio Sci., 36, , Bilitza, D., K. Rawer, L. Bossy, and T. Gulyaeva, International Reference Ionosphere Past, present, future, Adv. Space Res., 13(3), 3 23, Chapman, S., An example of the determination of a minute periodic variation as illustrative of the law of errors, Mon. Not. R. Astron. Soc., 78, , Collis, P. N., J. K. Hargreaves, and G. P. White, A localised co-rotating auroral absorption event observed near noon using imaging riometer and EISCAT, Ann. Geophys., 14, , Fjeldbo, G., Bistatic-radar methods for studying planetary ionospheres and surfaces, Ph.D. dissertation, Stanford Univ., Standford, Calif., Fjeldbo, G., A. J. Kliore, and V. R. Eshleman, The neutral atmosphere of Venus as studied with the Mariner V radio occultation experiments, Astron. J., 76, , Galushko, V. G., V. V. Paznukhov, Y. M. Yampolski, and J. C. Foster, Incoherent scatter observations of AGW/TID events generated by the moving solar terminator, Ann. Geophys., 16, , Gorbunov, M. E., Three-dimensional satellite refractive tomography of the atmosphere: A numerical simulation, Radio Sci., 31, , Hajj, G. A., and L. J. Romans, Ionospheric electron density profiles obtained with the Global Positioning System: Results from the GPS/MET experiment, Radio Sci., 33, , Hajj, G. A., L. C. Lee, X. Pi, L. J. Romans, W. S. Schreiner, P. R. Straus, and C. Wang, COSMIC GPS ionospheric sensing and space weather, Terr. Atmos. Ocean. Sci., 11, , Hedin, A. E., et al., Empirical wind model for the upper, middle amd lower atmosphere, J. Atmos. Sol. Terr. Phys., 58, , Hocke, K., Tidal variations in the high-latitude E- and F-region observed by EISCAT, Ann. Geophys., 14, , Hocke, K., and T. Tsuda, Using GPS satellites to study plasma irregularities, GPS World, 7, 34 36, 2001a. Hocke, K., and T. Tsuda, Gravity waves and ionospheric irregularities over tropical convection zones observed by GPS/MET radio occultation, Geophys. Res. Lett., 28, , 2001b. Hunsucker, R. D., Radio Techniques for Probing the Terrestrial Ionosphere, 293 pp., Springer-Verlag, New York, Igarashi, K., Y. Murayama, M. Nagayama, and S. Kawana, D-region electron density measurements by MF radar in the middle and high latitudes, Adv. Space Res., 25, 25 32, 2000a. Igarashi, K., A. Pavelyev, K. Hocke, and O. Yakovlev, Radio holographic principle for observing natural processes in the atmosphere and retrieving meteorological parameters from radio occultation data, Earth Planets Space, 52, , 2000b. Karayel, E. T., and D. P. Hinson, Sub-Fresnel-scale vertical resolution in atmospheric profiles from radio occultation, Radio Sci., 32, , Kirkwood, S., and P. N. Collis, The high-latitude lower ionosphere observed by EISCAT, Adv. Space Res., 7, 83 86, Lee, L.-C., C. Rocken, E. R. Kursinski (Eds.), Applications of Constellation Observing System for Meteorology, Ionosphere and Climate, 384 pp., Springer-Verlag, New York, Mathews, J. D., Sporadic E: Current views and recent progress, J. Atmos. Sol. Terr. Phys., 60, , Melbourne, W. G., E. S. Davis, C. B. Duncan, G. A. Hajj, K. R. Hardy, E. R. Kursinski, T. K. Meehan, and L. E. Young, The application of spaceborne GPS to atmospheric limb sounding and global change monitoring, JPL Publ.94 18, Jet Propul. Lab., Pasadena, Calif., Meng, C.-L., and H. W. Kroehl, Intense uniform precipitation of low-energy electrons over the polar cap, J. Geophys. Res., 82, , Mortensen, M. D., R. P. Linfield, and E. R. Kursinski, Vertical resolution approaching 100 m for GPS occultations of the Earth s atmosphere, Radio Sci., 34, , Pavelyev, A. G., On the possibility of radio holographic investigation on communication link satellite to satellite (in Russian), Radiotek. Elektron., 43, , Pavelyev, A. G., K. Igarashi, K. Hocke, S. S. Matugov, A. I. Kucherjavenkov, D. A. Pavelyev, O. I. Yakovlev, C. Reigber, J. Wickert, and G. Beyerle, First application of radioholographic method to wave obser-