Long-term analysis of ionospheric polar patches based on CHAMP TEC data

Transcription

1 RADIO SCIENCE, VOL. 48, 89 31, doi:1.1/rds.33, 13 Long-term analysis of ionospheric polar patches based on CHAMP TEC data M. Noja, 1 C. Stolle, J. Park, 1 and H. Lühr 1 Received October 1; revised 8 February 13; accepted 8 April 13; published 19 June 13. [1] Total electron content (TEC) from LEO satellites offers great possibility to sound the upper ionosphere and plasmasphere. This paper describes a method to derive absolute TEC observations aboard CHAMP considering multipath effects and receiver differential code bias. The long-term data set of 9 years GPS observations is used to investigate the climatological behavior of high-latitude plasma patches in both hemispheres. The occurrence of polar patches has a clear correlation with the solar cycle, which is less pronounced in the Southern Hemisphere (SH). Summed over all years, we observed a higher number of patches in the SH. The maximum occurrence rate of patches has been found at the dayside polar cusp during 1: 18: MLT (magnetic local time) supporting the mechanisms for patch creation by local particle precipitation and by intrusion of subauroral plasma into the polar cap through tongues of ionization (TOIs). The latter mechanism seems to be even more important in the SH. Investigating the patches in comparison with interplanetary magnetic field (IMF) conditions, we found that decreased IMF Bz and enhanced merging electric field preceded the patch observation; hence, patch creation follows a period of enhanced solar wind input into the magnetosphere/ionosphere. We further found an annual cycle in patch occurrence with maxima at equinox and December solstice and a June solstice minimum which reflects the global ionospheric seasonal asymmetry in electron density. We suggest that enhanced TEC at midlatitudes and low latitudes during December solstice provides a greater possibility to transport high-density plasma to the polar region through the buildup of TOIs. Citation: Noja, M., C. Stolle, J. Park, and H. Lühr (13), Long-term analysis of ionospheric polar patches based on CHAMP TEC data, Radio Sci., 48, 89 31, doi:1.1/rds Introduction [] Polar patches are prominent phenomena in the polar ionosphere. They represent islands of high plasma density in a low-density environment in the ionospheric F region. Their extent ranges from 1 to 1 km in horizontal size. Polar patches were first observed and studied by Buchau et al. [1983] and Weber et al. [1984] using optical (all-sky imaging photometer) and ionosonde measurements. Since then, patches have been studied using a number of different techniques, e.g., ground-based TEC measurements [Weber et al., 1986; Krankowski et al., 6], incoherent scatter radar [Pedersen et al., ;Carlson et al., ], and in situ measurements of satellites passing through the topside ionosphere [Coley and Heelis, 1995]. Seasonal and Universal 1 Helmholtz Centre Potsdam, German Research Centre for Geosciences, GFZ, Potsdam, Germany. National Space Institute, DTU Space, Technical University of Denmark, Copenhagen, Denmark. Corresponding author: M. Noja, Helmholtz Centre Potsdam, German Research Centre for Geosciences, GFZ, Potsdam, Germany. (maxx@ gfz-potsdam.de) 13. American Geophysical Union. All Rights Reserved /13/1.1/rds Time (UT) dependencies have been studied by Sojka et al. [1994] and Coley and Heelis [1998a], who classify polar patches primarily as a winter and equinox phenomenon, although patches are also found during local summer. Peak occurrence frequencies were found from 16: to 4: UT. [3] In the past decades, the sources, formation, propagation, and production mechanisms of patches have been studied and explained in different ways. Storm-enhanced density (SED) [Foster, 1993] and related tongue of ionization (TOI) phenomena are attributed to be major sources of polar patches as they cause strong enhancements of subauroral solar-euv produced plasma densities which get transported into the cusp [e.g., Hosokawa et al., 1; Carlson, 1]. Another potential plasma source is particle precipitation in the cusp [Rodger et al., 1994]. Polar patches form predominantly under southward interplanetary magnetic field (IMF) conditions which lead to the development of a strong two-cell convection pattern in the polar region. This convection pattern creates an antisunward flow that transports high-density plasma from the dayside cusp across the polar cap. The production mechanisms which separate the plasma into patches are subject to an ongoing discussion [see Carlson, 1]. Proposed mechanisms include rapid changes in IMF By and Bz components and changing solar wind conditions [e.g., Carlson, 1],

2 transient extension of the high-latitude convection pattern into the subauroral ionosphere [Anderson et al., 1988], transient bursts of magnetopause reconnection [Lockwood and Carlson Jr. 199], high-speed flow channel events separating cusp plasma [e.g., Pinnock et al., 1993], and IMF reversals [Valladares et al., 1998]. [4] Strong interest in studying polar patches lies in their important contribution to space weather effects in the polar region [Schunk and Sojka, 1996]. Due to steep plasma density gradients at the edges of patches, trans-ionospheric radio wave propagation can be heavily disturbed leading to radio wave scintillations which disturb navigation signals such as from the Global Positioning System (GPS) [e.g., Mitchell et al., 5]. Therefore, it is important to understand the occurrences and distributions of polar patches as complete as possible. [5] GPS-derived total electron content (TEC) from topside sounders aboard Low Earth Orbiting (LEO) satellites has offered great possibility to sound the upper ionosphere and plasmasphere [e.g., Heise et al., ; Pedatella and Larson, 1; Spencer and Mitchell, 11]. A good calibration of the satellite-received GPS signal to absolute TEC has been important for proper imaging of the near-earth space plasma for both occultation and topside sounders [e.g., Heise et al., 5;Stephens et al., 11]. In this study, we describe a method that is used to derive absolute TEC observations aboard CHAMP (Challenging Minisatellite Payload) considering multipath effects and receiver differential code bias. With the long-term and continuous CHAMP observations, we are able to investigate polar patches on a global and climatological scale which can be regarded as a great advantage compared to ground-based methods, especially for the Southern Hemisphere (SH) where ground-based instruments are difficult to maintain. We present new insights in longterm trends for both hemispheres with respect to seasonal and spatial distributions, dependency on solar cycle, and in comparison with solar wind conditions.. Instrumentation and Data [6] The CHAMP satellite is a German LEO mission with various objectives in scientific research and application. The satellite was launched on 15 July and reentered the atmosphere on 19 September 1. CHAMP was flying in a circular, near-polar orbit at altitudes between 45 km (1) and 7 km (1). As a basis for TEC determination, the data of the topside precise orbit determination (POD) antenna of the TRSR- Blackjack GPS receiver aboard CHAMP are used, which allow for a precise sounding of the topside ionosphere and plasmasphere. CHAMP POD GPS measurements are sampled at a rate of.1 Hz, which correspond to a spatial scale of 75 km; hence, TEC variations can be analyzed on short time and spatial scales. CHAMP GPS data from 1 to 9 are used in this analysis. Besides GPS observations and ephemerides from CHAMP, we used GPS satellite ephemerides from the International GNSS Service (IGS) and differential code bias data produced by the Center for Orbit Determination in Europe (CODE), which are required for TEC determination. Furthermore, we used IMF and solar wind data extracted from 1 min OMNI data available from NASA s Goddard Space Flight Center Absolute Slant TEC Retrieval [7] The basis of polar patch detection using CHAMP POD GPS observations is the determination of absolute slant TEC, which is described in detail in the following subsections. Slant TEC denotes the integrated number of electrons along the line-of-sight between receiver and transmitter, usually quantified in units of TECU (1 16 electrons/m ). TEC retrieval from GPS observations is based on the so-called ionospheric combinations P I and L I [Blewitt, 199]: P I P P 1 = ionp ionp1 DCB S DCB R + " P (1) P I 4.3m 3 s f 1 f f 1 f TEC DCB S DCB R + " P () where P and P 1 denote the precise (P-Code) code phase measurements and ionp and ionp1 denote the ionospheric propagation delays of the respective GPS signal. L I L 1 L = ionl ionl1 +N 1 1 N +DPB S +DPB R + " L (3) L I 4.3m 3 s f 1 f f 1 f TEC+N 1 1 N +DPB S +DPB R + " L (4) where L 1 and L denote the GPS carrier phase measurements and ionl1 and ionl denote the ionospheric propagation delays. N 1 and N denote the carrier phase ambiguities, and 1 and the carrier wave lengths. Information on slant TEC is contained in the difference of the ionospheric propagation delays. f 1 and f denote the GPS carrier frequencies. Besides slant TEC, the ionospheric combinations contain additional error terms which have to be corrected. For code phase TEC, these error terms are the differential code biases of the GPS satellite, DCB S, and of the GPS receiver, DCB R,andthe additional error term " P, which is mainly caused by multipath effects and thermal noise. For carrier phase TEC (L I ), the differential phase biases (DPB S and DPB R ) and the additional carrier phase error " I are by magnitudes smaller than their code phase equivalents. However, the carrier phase observations contain the ambiguities N 1 and N as an additional source of error. Additional terms which are part of the code phase and/or carrier phase observables are phase center variations, higher order ionospheric effects, and the carrier phase wind-up effect [Wu et al., 1993]. These terms have been omitted in equations (1) (4) as they have negligible amplitudes compared to those of the DCBs and the carrier phase ambiguities. Cycle-slip detection and correction to create continuous arcs of GPS observations and solving the carrier phase ambiguities by a common carrier-to-code leveling algorithm is based on the algorithms described in Blewitt [199] Multipath Effect [8] Multipath errors of GPS observations are primarily caused by the superposition of the direct signal with signals reflected in the vicinity of the antenna. A multipath error of 1 m is equivalent to approximately 1 TECU in slant TEC determination. Therefore, reducing multipath effects enhances the accuracy of the carrier phase leveling. Multipath calculation and correction is based on the

3 P-code multipath linear combinations [e.g., Byun et al., ; Montenbruck and Kroes, 3]: f M P1 + " P1 P f f f 1 f L 1 + f 1 f L B P1 (5) f M P + " P P 1 f 1 f L 1 + f 1 + f f 1 f L B P (6) [9] In these linear combinations, the receiver-to-satellite geometry and the ionospheric path delay both are removed. The multipath errors are denoted by the M Pi terms and the additional noise errors " Pi. The carrier phase ambiguities and the differential code biases are combined in the B Pi terms [Montenbruck and Kroes, 3]. These offsets can be assumed to be constant during one continuous satellite track of a certain PRN (pseudo random noise) number which takes up to 3 min. Assuming a zero mean multipath error (M Pi + " Pi =), equations (5) and (6) can be written as f B P1 P f f f 1 f L 1 + f 1 f L (7) f B P P 1 f 1 f L 1 + f 1 + f f 1 f L (8) where Qr = R Rec R Rec + h (1) and denotes the ray elevation. R Rec denotes CHAMP s radial distance from the Earth s center. The thickness of the ionosphere above the CHAMP satellite is given by h which usually takes an empirically determined value of around 4 km. The mapping function is based on a thick-layer model of the ionosphere. A thin-layer ionosphere model, which is commonly used to map ground-based TEC to the vertical, cannot be used properly for CHAMP GPS observations as CHAMP s altitude is generally higher than the ionospheric piercing point of the thin-layer model. Using the mapping function M(), the relation between absolute vertical TEC and absolute slant TEC is given by the following: VTEC abs = STEC abs M() (11) where VTEC abs denotes the absolute vertical TEC, STEC abs the absolute slant TEC, and the ray elevation. The relation of two paired VTEC observations can therefore be described by the following: STEC abs A M( A)=STEC abs B M( B) (1) [1] The offsets B Pi are determined by creating a weighted median of equations (7) and (8) from all observations of a continuous PRN track based on the signal-to-noise ratio (SNR) of the GPS observations. The weighting is created by only using observations whose SNR was at least the median SNR of the PRN track. To eliminate the additional noise errors " Pi, all multipath errors of 1 day are binned in a 1 ı grid of elevation and azimuth angle, and the median of each bin is calculated. 3.. Receiver Differential Code Bias [11] Besides multipath, the differential code bias (DCB) of a GPS receiver is one of the major error sources during TEC determination. The receiver DCB is mainly caused by path delay differences during the processing of P1 and P in the receiver hardware. Determining the receiver DCB is possible using a model of the ionosphere, [e.g., Heise et al., 5]. The approach which is used here to determine the DCB of the CHAMP GPS receiver is based on the method described in Syndergaard [7] which does not require a model. The basic assumption of this method is that vertical TEC is equal for paired simultaneous GPS observations. Mapping the relative slant TEC values to vertical TEC is achieved by a function M() depending on the GPS ray s elevation angle which is described by Foelsche and Kirchengast []: The absolute slant TEC can be rewritten as follows: where and STEC abs i STEC abs i = STEC rel i + DCB Si + DCB R (13) STEC rel i = P I 4.3m 3 s f 1 f f 1 f (14) denotes the absolute slant TEC. STEC rel i denotes the relative slant TEC which is derived from the ionospheric combination P I (see equation ()). The DCB of the GPS satellite, DCB Si, is usually given by an external provider such as CODE. The receiver DCB is denoted by DCB R. Rewriting equation (1) with respect to equation (13) results in the following: STEC rel A M( A)+DCB SA M( A )+DCB R M( A )= STEC rel B M( B)+DCB SB M( B )+DCB R M( B ) (15) M() = h (R Rec + h) cos(arcsin( Qr cos()) Qr sin()) (9) 91 [1] All paired observations of a day are used to create a system of linear equations in the form of y = Ax where x denotes the DCB R.TheDCB R is derived by applying the

4 method of least squares to the system of linear equations in the form of x =(A T A) 1 A T y and results in the following: P (M(B ) M( A ))( STEC d A M( A ) STEC d B M( B )) DCB R = P (16) (M(B ) M( A )) b TEC i denotes the sum of the relative slant TEC and the DCB S. [13] Besides the assumption that a GPS receiver DCB is constant over the course of a day, equation (1) also assumes a spherical ionosphere without significant horizontal gradients. This assumption only holds when the absolute vertical TEC is low, e.g., TEC 3 TECU. We therefore limited the TEC data that are used for the DCB R determination to midlatitudes and high latitudes and to ray elevation angles 45 ı GPS seconds 15 1 TECU 4. Polar Patch Identification [14] Identification of polar patches is based on analyzing continuous arcs of absolute CHAMP slant TEC data. The use of vertical TEC would be beneficial; however, the accuracy of the vertical mapping function in equation (9) is only suitable for determining the differential code bias under the above described conditions. Applying this mapping function for polar patch identification would produce unreliable results. Anyhow, the difference in TEC magnitude which is caused by a slant path is not as severe for CHAMP TEC as it is for ground-based TEC because the electron density in the topside ionosphere and plasmasphere is much smaller than in the F region. As an additional constraint, we impose a minimum elevation of 5 ı for all TEC data. Since polar patches are high-latitude phenomena, we also impose a minimum geomagnetic latitude of 55 ı for all TEC observations to prevent a consideration of possible low-latitude ionospheric irregularities during the detection. [15] The patch detection approach used here is similar to the one described in Coley and Heelis [1995]: Within a sliding window of s (15 km arc) length, there has to be a positive slope in the TEC data which is followed by a negative slope at some later point in the window. There can of cause be other positive and negative slopes in between. No minimum slope magnitude is required. Having found these two slopes, the patch peak is determined by the largest TEC magnitude between them. This peak magnitude has to be compared to a background level in order to validate the patch detection. The patch background level is determined by linearly interpolating between the TEC values at the boundaries of the patch range symmetrically to the peak position. The first boundary of the patch range is defined by the TEC value which is the greater one of the smallest TEC values to the left and right of the patch peak position. The second boundary is defined by the TEC value on the opposite side of the patch peak which is closest in magnitude to the TEC value of the first boundary. [16] The general criterion for detecting polar patches using ground-based techniques or in situ measurements is to have an electron density which is at minimum twice the background level magnitude [Crowley, 1996]. However, since CHAMP is expected to fly in the upper part of the patch ( 4 km) and sounds only the ionosphere above the patch, the signature of the polar patches is expected to 9 Figure 1. Polar patch identification: A peak in the absolute slant TEC with a minimum of 4 TECU over the background level has to be found within s (15 km). The background level of the patch is determined by linearly interpolating between the maximum of the smallest TEC value left and right of the peak and its closest counterpart in TEC magnitude on the opposite side of the peak. be weaker. The polar patch criterion that is applied in this detection approach is based on a mixture of absolute and relative patch magnitudes. The criterion for relative magnitude is described by the following: (TEC P TEC BG ) TEC BG 1. (17) where TEC P denotes the TEC value at the patch peak and TEC BG the background level TEC. We used the square term in the numerator to be able to detect smaller scale patches (< TECU) in high-density backgrounds (>1 TECU). The value 1. is an empirically determined threshold. The absolute magnitude criterion is described by the following: TEC P TEC start 4 TECU (18) and TEC P TEC end 4 TECU (19) where TEC start and TEC end denote the TEC magnitudes at the patch range boundaries. The criterion for absolute TEC magnitude is used to prevent small-scale peaks (<4 TECU) from being falsely detected as polar patches in a low background environment (<13 TECU). The value 4 TECU is also an empirically determined threshold which is used throughout the solar cycle. The chosen thresholds are kept throughout the solar cycle since the definition of patches is based on electron density gradients between background and patch. This gradient which is responsible, e.g., for radio wave scintillations, is regarded as independent from the absolute background electron density governed by solar flux. An example of polar patch identification from CHAMP TEC data is depicted in Figure Observations [17] For an in-depth analysis of the polar patches detected from CHAMP TEC data, we investigated seasonal and spa-

5 Number of Polar Patches per 5 days Northern Hemisphere Polar Patches Day of Year Number of Polar Patches per 5 days Southern Hemisphere Day of Year Polar Patches Figure. CHAMP TEC polar patch detections seasonal distribution 1 4: Number of polar patch detections bin-averaged over 5 days. tial distributions, solar wind conditions during patch events, and dependence on solar cycle. The analysis is generally separated by hemisphere. The occurrence distribution is considered in a quasi-dipole [see Richmond, 1995] geomagnetic latitude, magnetic local time frame if not stated otherwise. Observations resulting from the analysis are given in the following subsections Seasonal Distribution [18] The seasonal distribution of polar patch detections is analyzed in different ways, e.g., absolute number of occurrences, patch background level magnitude, and absolute and relative patch peak TEC values. Figure shows for both hemispheres the averaged annual variation. Events within 5 day bins have been summed up for the solar active years from 1 through 4. Both hemispheres indicate peak occurrence frequencies during the equinoxes and strong occurrence frequencies during December solstice months, as well as low occurrence frequencies during June solstice months. The SH shows generally more detections but not such a deep and long minimum during June solstice. [19] The seasonal occurrence distribution is quite similar to the patch background level distribution. Figure 3 depicts the absolute (TEC P TEC BG ) and relative ((TEC P TEC BG )/TEC BG ) patch peak TEC levels with respect to the TEC background level (TEC BG ). Absolute, relative, and background levels are given as medians of (TECU) 1 (days) bins. Both hemispheres experience low background levels during June season and peak background levels from autumn to spring, with the SH also dominating. Additionally, there are numerous gaps in the low background level regions 5 TECU in the NH from March to September and some smaller gaps in the SH from November 93 to February. The absolute patch peak levels (Figure 3 left column) are in good correlation to the background levels, with peaks occurring from September to March and only very low absolute levels in summer. The relative peak levels show similar results for both hemispheres with the NH having stronger relative levels from November to February. In the SH, there might be a slight increase during June solstice months which can only be suspected due to the absence of patches with larger background levels. 5.. Spatial Distribution [] Besides having an idea of the distribution in season and hemisphere, the spatial distribution in terms of q-dipole geomagnetic latitude and magnetic local time as well as geographic coordinates is of great interest. Figure 4 shows the spatial distribution of polar patch occurrence rates per satellite pass for the active years from 1 to 4, where occurrence rates below.1 are treated as. The upper right subfigure for geomagnetic SH exposes strongest occurrence frequencies in the cusp region at noon with a tail reaching to dusk at approximately 65 ı magnetic latitude (MLAT). Patches are present across the polar cap, and there is another smaller patch accumulation at the night side at around 6 ı at early morning. Between the 65 ı MLAT dusk tail and the polar cap is a dusk minimum region with smaller gaps from 7 ı to 8 ı. We also find larger gaps equatorward of the dayside cusp region and prenoon below 68 ı MLAT. [1] In the Northern Hemisphere, the patch distribution is similar; however, rates are much smaller. There are a lot of gaps in the dusk and dawn regions, and there is a large gap extending from dawn to noon below approximately 65 ı MLAT. Starting from the peak region in the dayside cusp, the noon-to-midnight polar region is more dominant on the

6 Figure 3. CHAMP TEC polar patch peak levels seasonal distribution 1 4: Patch background level median (abscissa) and days of the year (ordinate) in bins of (TECU) 1 (days). Corresponding median peak levels are given by color map. Absolute patch peak level SH (upper left), NH (lower left). Relative patch peak level SH (upper right), NH (lower right). dusk side. Additionally, the occurrence rates in the polar cap are not symmetrical about the pole but slightly shifted towards dusk. Looking at the nightside regions, higher occurrence rates are mostly confined to the premidnight (1: to : MLT) sector. However, we also find an accumulation of events around 64 ı MLAT in the postmidnight sector that is similar to the SH. [] Looking at the geographic distribution, the NH is clearly dominated by high rates in the North American area where patches can be detected down to approximately 5 ı. The peak occurrence region lies south of the geomagnetic pole (indicated by a white cross). In the European and Asian sector, patches are only present in the polar region. In the geographic SH, the dominant region is centered equatorward of the geomagnetic pole. In contrast to the NH, the dominant region is much more extended covering more than 9 ı in longitude and more than ı in latitude reaching out almost to Australia and New Zealand. In the polar region, moderate patch occurrence rates can be found across almost all longitudes with an exception of the sector around the Greenwich meridian. There are numerous gaps and very low occurrence rates from ı to 4 ı longitude at almost all latitudes. Similar to Europe and Asia in the NH, the areas opposite to the geomagnetic pole (South America, Atlantic Ocean) show patch detections only at high latitudes ( 7 ı ) Superposed Epoch Analysis [3] Polar patch creation responds to changing IMF and solar wind conditions [Sojka et al., 1994; Coley and Heelis, 1998b]. Most patches are expected during southward IMF 94 which is indicated by a negative IMF Bz. Furthermore, the merging electric field (Em) indicates the degree of the coupling between solar wind and the magnetosphere [Kan and Lee, 1979]: E m = v SW qb Y + B Z sin (/) () where v SW denotes the solar wind velocity, B Y and B Z denote the Y and Z components of the IMF, and denotes the clock angle of the interplanetary magnetic field tan() = BY B Z. Besides solar wind and IMF conditions, we investigated the influence of geomagnetic storm conditions by use of the SYM-H index which is an adequate high-resolution equivalent (1 min) to the hourly D ST index [Wanliss and Showalter, 6]. Figure 5 shows the superposed epoch analysis of the 1 min OMNI data based IMF By (separated by hemisphere), IMF Bz, Em, and SYM-H index 1 min before and after the polar patch detections (time of the peak, see Figure 1) for the years from 1 to 4. Solar wind data from OMNI is propagated to the bow shock nose. When considering both hemispheres, around 4, polar patches were detected. For each observable, the plot depicts the average value and standard deviation (and median for IMF By) of all the values in the 1 min bin. [4] Assessing IMF By in the SH, we find the average at around. nt and the median at.4 nt. Both average and median expose similar trends with smaller fluctuations. The standard deviation is around 5.5 nt with similar smallscale fluctuations. Looking at By in the NH, the average is quite stable around, and the median is always negative,

7 Figure 4. CHAMP TEC polar patch spatial distribution 1 4: Number of polar patches per GPS satellite pass separated by hemisphere (north left, south right) with respect to geomagnetic latitude and magnetic local time (top row) and geographic longitude and latitude (bottom row). Data are bin-averaged in a 1 ı (4 min MLT (magnetic local time)) grid. fluctuating a little more around.15. The standard deviation is slightly larger than in the SH with 5.9 nt. Although both hemispheres have a large standard deviation in comparison to the average and median, the uncertainty of the standard deviation average p n is below.1 nt. Looking at IMF Bz, strongest magnitudes are found 1 to 6 min before the patch detection where the average is around.95 nt. Afterwards, IMF Bz continuously increases (decreases in magnitude) until 1 min after the patch occurrence with minimal fluctuations. The standard deviation is around 5.1 nt and also fluctuates only slightly. The uncertainty of the Bz average is also below.1 nt. Similar to Bz, the Em average exposes strong magnitudes of 1.75 mv/m before the patch detection which continuously decreases after 6 min before the patch until it reaches a still considerable magnitude of approximately 1.57 mv/m at 1 min after the patch detection. The standard deviation of Em is around.1 mv/m, and it exposes a very similar trend to the Em average. The SYM- H index average shows a strongly negative gradient before the patch from 1.6 nt to 3. nt which is followed by 95 a slightly negative gradient that almost reaches 3.6 nt at 1 min after the patch detection. The standard deviation of the SYM-H index exposes an almost constant increase from 33.8 to 35.9 nt. The maximum difference in magnitude is similar for the average and standard deviation with SYM-H nt Solar Cycle Dependence [5] In order to investigate dependences on solar activity, the average number of polar patches per 5 days is compared to the P1.7 solar flux index [Richards et al., 1994] in Figure 6. The P1.7 is defined as follows: P1.7 = (F1.7 + F day avg )/ (1) where F1.7 denotes the solar flux index and F day avg denotes the 81 day average of the solar flux index centered on the day of interest. The CHAMP data ranging from 1 to 9 cover solar maximum conditions in 1/ and solar minimum from 7 to 9. Solar flux and patch detection trends are similar and indicate a close dependence

8 average [nt] CHAMP TEC Polar Patch Superposed Epoch Analysis From Day 1 1 to IMF BY SH Minutes BY avg. BY median BY std.-dev standard dev. [nt] average [nt] IMF BY NH BY avg. BY median BY std.-dev Minutes standard dev. [nt] average [nt] IMF BZ SH and NH Minutes BZ avg. BZ std.-dev standard dev. [nt] average [mv/m] average [nt] Merging Electric Field (Em) SH and NH Em avg. Em std.-dev Minutes SYM-H SH and NH SYM-H avg SYM-H std.-dev Minutes standard dev. [mv/m] standard dev. [nt] Figure 5. CHAMP TEC polar patch superposed epoch analysis 1 4: Average (and median for By) (left y axis) and standard deviation (right y axis) of IMF By, IMF Bz, merging electric field (Em), and SYM-H index 1 min before and after polar patch occurrences. in both hemispheres, despite the seasonal variation of patch occurrence rates. The polar patch detections also expose a strong day-to-day variability and a susceptibility to particular events in solar flux, e.g., in October/November 3 or July 4. [6] Looking at the long-term trends of polar patch occurrence frequencies, the detection rate decrease in the SH is much smoother with patches still detectable in 8 and 9. The NH is much more affected by the solar maximum in winter 1/; however, it also shows a very steep drop after 3. Figure already suggested a dominance of the SH in terms of detected events. However, it does not show the very prominent role of the years 1 and in the NH s seasonal distribution and the strong dominance of the SH in later years that can be seen in Figure 6. 96

9 Avg. Number of Polar Patches per 5 days Northern Hemisphere Polar Patches Polar Patches P1.7 6 Jan 1 1 Jan 1 Jan 1 3 Jan 1 4 Jan 1 5 Jan 1 6 Jan 1 7 Jan 1 8 Jan 1 9 Jan P1.7 [1 - Wm - Hz -1 ] Avg. Number of Polar Patches per 5 days Southern Hemisphere Polar Patches Polar Patches P1.7 6 Jan 1 1 Jan 1 Jan 1 3 Jan 1 4 Jan 1 5 Jan 1 6 Jan 1 7 Jan 1 8 Jan 1 9 Jan P1.7 [1 - Wm - Hz -1 ] Figure 6. CHAMP TEC polar patch detections and P1.7 index 1 9: Number of polar patch detections binned and averaged over 5 days are compared to the P1.7 index which is calculated as the average of the F1.7 index and the 81 day average of the F1.7 index. 6. Discussion [7] The observations described in the above sections show similarities to and differences from previous studies of polar patches which are discussed in the following. Starting with the spatial distribution, the majority of polar patch detections is found in the SH. Coley and Heelis [1998a] find occurrence frequencies more than twice as high in the SH. However, we only find on average 6% to 4% for the SH to NH from 1 to 9, which varies from year to year (cf. Figure 6). This imbalance between the hemispheres is mainly caused by the asymmetric offset of the geomagnetic poles from the Earth s rotation axis. In the SH, the magnetic pole is much more equatorward than in the NH. Hence, as the auroral oval expands deeper into the midlatitude sunlit high-density plasma regions, TOIs and SEDs may be drawn into the cusp region over a larger longitude sector. [8] The seasonal distribution of the NH is in good agreement with observations from Sojka et al. [1994] and Coley and Heelis [1998a]: high occurrence frequencies during the equinoxes and December solstice and very low occurrence frequencies during June solstice. Carlson [1] explains the low number of polar patches in June solstice (local summer in NH) with an already high level of density in the polar region which prevents the creation of large density differences. This is also reflected in the first row of Figure 7 where the polar region in the NH local summer shows a very homogeneous TEC distribution with TEC > 8 TECU. In contrast to NH June solstice, the SH December solstice (fourth row, local summer) shows a more heterogeneous TEC distribution which slightly decreases in magnitude from year to year. This heterogeneity in combination with a considerable level of TEC magnitude in the polar regions may explain the slower drop in patch occurrence rates and why patches are still detectable in SH summer until the end of 9. Due to the heterogeneity and numerous patch detections in the SH local summer, the explanation by Carlson [1] cannot be applied. The reason that patches are detectable in SH summer is probably also due to the larger geomagnetic pole offset causing the observed differences in local summer TEC distribution between the two hemispheres. [9] The most pronounced feature of our observations is the June/July minimum in patch detections in the Southern Hemisphere. The reason for this minimum may be derived from the second row of Figure 7. The TEC distribution is not very homogeneous, but it exhibits very small magnitudes in the polar region which rarely exceed 5 TECU; hence, the detection parameters (see section 4) applied here for identification ignore possibly too small scale TEC peaks. It could be suspected that CHAMP is too far above the occurrence of patches during July. However, Coley and Heelis [1998a] reported a maximum number of patches during June solstice in the SH using DMSP in situ plasma measurements at an altitude of 84 km. Hence, CHAMP s altitude appears not to be the reason for the missing patches during SH June solstice months. It has been shown by several works that patches are also well developed at altitudes of 4 km or higher [Mitchell et al., 5; Stolle et al., 6; Dahlgren et al., 1]. [3] Krankowski et al. [6] investigated the rate of change of ground-based TEC observations associated with polar plasma patches in the SH using data from Antarctic stations McMurdo-MCM4, Casey-CAS1, Mawson-MAW1, 97

10 Figure 7. CHAMP absolute slant TEC distribution 1 4: Median absolute slant TEC based on binned CHAMP orbit positions for a 1 day period centered in June and December solstices and separated by hemisphere. and Davis-DAV1. At CAS1, they found strong gradients of TEC before 1: UT over the entire year and preferably during May/June. However, after 1: UT, preferred months of strong TEC gradients have been March and November; no significant gradients have been detected between June and August. They pointed out the complex patch activity in dependence on season and UT. Rodger 98 and Graham [1996] investigated patch occurrence in the SH using HF radar data from Halley, Antarctica. They reported a midsummer and midwinter minimum of patch occurrence and highest occurrence rates during equinox and found their observations in good agreement with simulation results from Sojka et al. [1994] when applied to the Southern Hemisphere. Both publications support a minimum of patch

11 Figure 8. Comparison between CHAMP PLP in situ electron density measurements (red) and CHAMP absolute slant TEC (blue) for detected polar patch events on 6 March in the NH (top row) and SH (bottom row). occurrence in the SH during local winter, but they do not report on a striking difference between occurrence rates in summer and winter, as shown in our data. We might explain our observations with structured background TEC as shown in Figure 7. [31] Figure 7 depicts the spatial distribution of the median absolute slant TEC values for a 1 day period centered on the June and December solstices for the years from 1 to 4 separated by hemisphere. The first row depicts the NH June solstice (local summer), the second row the SH June solstice (local winter), the third row the NH December solstice (local winter), and the fourth row the SH December solstice (local summer). Especially during high solar activity years (1 and ), Figure 7 reveals higher low and midlatitude TEC values during December than June solstice months, especially in the noon-dusk sector (1: MLT 18: MLT). As revealed in Figure 4 and as suggested by Foster et al. [5], patch generation through intrusion of subauroral TOIs originate mainly from enhanced plasma at the noon-dusk sector. Thus, we suggest that higher density plasma can be inferred into the polar cap during December than during June, which leads to a more significant patch detection, irrespective of hemisphere. This further supports the yearly periodicity observed in Figure 6. [3] Assessing the spatial distribution in geomagnetic latitude and magnetic local time (cf. Figure 4, top), the patch observations of the SH have a lot of resemblance with Figure 4 of Foster et al. [5], which depicts the polar ground-based vertical TEC evolution during an SED event in the NH. According to Foster and Rideout [7], SED occurrences are magnetically conjugate; hence, a similar distribution pattern in the SH is very likely. The polar patch tail in the dusk auroral region of the Southern Hemisphere (Figure 4, upper right) also indicates the role of the sunward 99 subauroral polarization streams for transporting the highdensity plasma of SEDs from subauroral latitudes at dusk into the dayside cusp regions [Foster et al., 7]. Since the NH shows similar features, which are however less pronounced, we suggest that the TOI mechanism is even more pronounced in the SH than in the NH. [33] The geographic spatial distribution in Figure 4 (bottom row) is in good accordance with the general assumption for the NH that SEDs occur mostly in the North American region. Coster et al. [7] showed that SEDs also occur in Europe and Russia, however, at much higher latitudes. The stronger polar patch occurrence frequencies and the much larger dominant region in the 9 ı to 18 ı longitude sector of the SH may be again explained with the offset of the geomagnetic pole. Since the geomagnetic pole is located at a lower latitude, the auroral region reaches deeper into the sunlit area attracting not only more TOIs but also covering a wider region. For the SH, a comparison with other studies is difficult as they are coming mostly from ground-based observations with sparse spatial coverage. Nonetheless, the southern distribution of polar patches appears fairly reasonable with regard to the results of the geomagnetic distribution. [34] Considering in situ measurements from CHAMP as a means of verification for the TEC based polar patch detection, a comparison between in situ electron density measurements from the Planar Langmuir Probe (PLP) and absolute slant TEC measurements for selected polar patch events is given in Figure 8, where the top row displays events in the NH and the bottom row events in the SH. The plots expose similar trends for TEC and in situ electron density data. Especially peak densities, which form the basis for the applied detection method, are replicated in the PLP data. Additionally, Park et al. [1] did an analysis of ionospheric topside plasma irregularities for high-latitude

12 regions based on in situ electron density measurements from the Digital Ion Drift Meter (DIDM) mounted on CHAMP. Their results for the spatial distribution with respect to geomagnetic coordinates generally agree with the spatial distributions given in the top row of Figure 4. Furthermore, Park et al. [1] find an annual variation of the spatial distribution that is similar to the one described above. [35] The results of the superposed epoch analysis (Figure 5) for IMF Bz follow the reasoning of Sojka et al. [1994] and Coley and Heelis [1998b] that a majority of polar patch detections coincides with a southward IMF. However, the large IMF Bz standard deviation suggests the development of polar patches also during positive IMF Bz conditions which has been shown by Wood et al. [8] and Hosokawa et al. [9]. The results of the IMF By analysis indicate that IMF By is mainly positive in the SH and negative in the NH. This is in good agreement with the IMF By dependence of the empirical high-latitude electric field models of Heppner and Maynard [1987], which describe the two-cell antisunward convection pattern that is present under southward IMF conditions. This slight polarity difference of By for NH and SH events suggest the preferences of a banana-shaped dusk convection cell. For both hemispheres, the relatively high standard deviation in IMF By with respect to the low median and average indicates a strong fluctuation of IMF By in both positive and negative directions. This supports the assumption that polar patches coincide with changes in IMF By [see Carlson, 1]. The results of the Em verify the detected patches since the fairly large average magnitude suggests an enhanced coupling between solar wind and magnetosphere as a basis for polar patch development, especially with the stronger Em magnitudes about an hour before the patch detections. We may however note that our TEC patch detection method cannot provide any information on the drift direction of the patch; thus, the age of the patch cannot be determined. The superposed epoch analyses in Figure 5 thus includes patches at different age and at different locations. [36] An analysis of the solar cycle dependence of polar patches based on fof measurements for the NH is given by Dandekar []. This analysis showed maximum polar patch occurrence frequencies during solar maximum from 1989 to 199 and an almost complete absence of patches around solar minimum from 1996 to A similar behavior is presented in Figure 6 for the NH with regard to the solar maximum in 1 and and the minimum from 7 to 9. Results of the SH also indicate a solar cycle dependence. However, Figure 6 depicts a lower maximum in detection number for 1 and than in the NH, but considerable detection numbers are still present during 8 and 9. As described above, this is related to the lower latitude position of the SH geomagnetic pole which allows the development of TOIs and their convection into the cusp even under solar quiet conditions. The long-term trend of the SH shows that the SH is less dependent on solar cycle than the NH. Another interesting aspect of the solar cycle analysis using CHAMP TEC data is the satellite altitude, as polar patches are better detectable at altitudes around the F peak. Taking into consideration the decreasing CHAMP altitude from 1 (43 km) to 9 (3 km), the solar cycle dependence of the polar patches is even more pronounced. 7. Conclusions [37] In this paper, we presented a long-term analysis (9 years) of ionospheric polar patches based on CHAMP TEC data. Results of the analysis verified the applied patch identification method by showing many similarities with known polar plasma patch climatologies. For example, we found higher patch occurrence in the Southern Hemisphere than in the Northern Hemisphere which is explained by the more equatorward offset of the geomagnetic pole in the SH. We found that 6% of the total number of patches occur in the SH and 4% in the NH. We also revealed a decreasing patch occurrence rate with decreasing solar flux. The solar flux dependence is less pronounced in the SH, still showing a countable number of patches during very low solar flux (solar minimum 3/4); almost no patches have been detected in the NH at this time. Also, this observation has been explained by the larger offset of the geomagnetic pole in the SH. In contrast, during high solar flux years (1 and ), the number of patches in the NH exceeded the number of patches in the SH. [38] A novel and unusual finding of our study is an annual cycle of polar patch detections rather than a seasonal variation. In both the NH and the SH, highest patch occurrence has been detected during equinox and December solstice months. A clear patch occurrence minimum occurred during June solstice months. The annual cycle in polar patch detection results based on CHAMP TEC might be explained by the global ionospheric asymmetry which creates peak TEC magnitudes in December [see, e.g., Mendillo et al., 5], especially in the topside ionosphere [Liu et al., 7; Su et al., 1998]. We suggest that enhanced TEC at midlatitudes and low latitudes during December solstice provides a greater possibility to transport high-density plasma to the polar region through the buildup of TOIs. [39] In terms of spatial distributions, we found similarities for both hemispheres with SED/TOI occurrences confirming the importance of transported plasma into the polar cap through TOI formation and contributing to polar patch creation. A comparison of both hemispheres with regard to patch occurrence frequencies confirmed a prevalence of this patch formation mechanism in the SH. [4] It has been shown that TEC observations by LEO satellites are suitable for patch climatology in a global sense. There are questions that remain to be answered, e.g., the annual variation of patch occurrence similarities in the NH and SH, where LEO satellite observations can provide new insight. [41] TEC observations from LEO satellites provide an effective, global, and low-cost method for patch detection which are related to radio wave scintillations due to enhanced plasma structuring [Mitchell et al., 5; Dahlgren et al., 1]. [4] Especially through the upcoming Swarm mission [Friis-Christensen et al., 8], with three LEO satellites flying in formation and carrying a GPS receiver, we expect an effective and routine detection opportunity for polar plasma patches. [43] Acknowledgments. We thank Stig Syndergaard and Stefan Heise for fruitful discussions on TEC retrieval aboard LEO satellites. The CHAMP mission was sponsored by the Space Agency of the German Aerospace Center (DLR) through funds of the Federal Ministry of Economics and Technology, following a decision of the German Federal Parliament (grant code 5EE944). 3

13 References Anderson, D., J. Buchau, and R. Heelis (1988), Origin of density enhancements in the winter polar cap ionosphere, Radio Sci., 3(4), Blewitt, G. (199), An automated editing algorithm for GPS data, Geophys. Res. Lett., 17(3), 199. Buchau, J., B. Reinisch, E. Weber, and J. Moore (1983), Structure and dynamics of the winter polar cap F region, Radio Sci., 18(6), Byun, S. H., G. A. Hajj, and L. E. Young (), Assessment of GPS signal multipath interference, in Proceedings of the National Technical Meeting of The Institute of Navigation, pp , San Diego, CA. Available at: cfm?jp=p&idno=63. Carlson, H. (1), Sharpening our thinking about polar cap ionospheric patch morphology, research, and mitigation techniques, Radio Sci., 47, RSL1, doi:1.19/11rs4946. Carlson, H., K. Oksavik, J. Moen, A. van Eycken, and P. Guio (), ESR mapping of polar-cap patches in the dark cusp, Geophys. Res. Lett., 9(1), 1386, doi:1.19/1gl1487. Coley, W., and R. Heelis (1995), Adaptive identification and characterization of polar ionization patches, J. Geophys. Res., 1(A1), 3,819 3,87. Coley, W., and R. Heelis (1998a), Seasonal and universal time distribution of patches in the northern and southern polar caps, J. Geophys. Res., 13(A1), 9,9 9,37. Coley, W., and R. Heelis (1998b), Structure and occurrence of polar ionization patches, J. Geophys. Res., 13(A), 1 8. Coster, A., M. Colerico, J. Foster, W. Rideout, and F. Rich (7), Longitude sector comparisons of storm enhanced density, Geophys. Res. Lett., 34, L1815, doi:1.19/7gl368. Crowley, G. (1996), Critical review of ionospheric patches and blobs, in URSI Review of Radio Sci , edited by W. R. Stone, p. 619, Oxford Univ. Press, New York. Dahlgren, H., G. Perry, J. Semeter, J.-P. St.-Maurice, K. Hosokawa, M. Nicolls, M. Greffen, K. Shiokawa, and C. Heinselman (1), Spacetime variability of polar cap patches: Direct evidence for internal plasma structuring, J. Geophys. Res., 117, A931, doi:1.19/1ja Dandekar, B. (), Solar cycle dependence of polar cap patch activity, Radio Sci., 37(1), doi:1.19/rs56. Foelsche, U., and G. Kirchengast (), A simple geometric mapping function for the hydrostatic delay at radio frequencies and assessment of its performance, Geophys. Res. Lett., 9(1473), 4 pp. Foster, J. (1993), Storm time plasma transport at middle and high latitudes, J. Geophys. Res., 98(A), Foster, J., and W. Rideout (7), Storm enhanced density: Magnetic conjugacy effects, Ann. Geophys., 5, , doi:1.5194/angeo Foster, J., W. Rideout, B. Sandel, W. Forrester, and F. Rich (7), On the relationship of SAPS to storm-enhanced density, J. Atmos. Sol.-Terr. Phys., 69, doi:1.116/j.jastp Foster, J., et al. (5), Multiradar observations of the polar tongue of ionization, J. Geophys. Res., 11, A9S31, doi:1.19/4ja198. Friis-Christensen, E., H. Lühr, D. Knudsen, and R. Haagmans (8), Swarm - An Earth observation mission investigating Geospace, Adv. Space Res., 41(1), Heise, S., A. Wehrenpfennig, C. Reigber, and H. Lühr (), Sounding of the topside ionosphere/plasmasphere based on GPS measurements from CHAMP: Initial results, Geophys. Res. Lett., 9(1699), 4 pp. Heise, S., C. Stolle, S. Schlüter, and N. Jakowski (5), Differential code bias of GPS receivers in low Earth orbit: An assessment for CHAMP and SAC-C, in Earth Observation with CHAMP: Results from Three Years in Orbit, edited by C. Reigber, H. Lühr, P. Schwintzer, and J. Wickert, Springer-Verlag, Berlin. Heppner, J., and N. Maynard (1987), Empirical, high-latitude electric field models, J. Geophys. Res., 9(A5), Hosokawa, K., T. Kashimoto, S. Suzuki, K. Shiokawa, Y. Otsuka, and T. Ogawa (9), Motion of polar cap patches: A statistical study with all-sky airglow imager at Resolute Bay, Canada, J. Geophys. Res., 114, A4318, doi:1.19/8ja14. Hosokawa, K., K. Tsugawa, T. Shiokawa, Y. Otsuka, N. Nishitani, T. Ogawa, and M. Hairston (1), Dynamic temporal evolution of polar cap tongue of ionization during magnetic storm, J. Geophys. Res., 115, A1333, doi:1.19/1ja Kan, J., and L. Lee (1979), Energy coupling function and solar windmagnetosphere dynamo, Geophys. Res. Lett., 6(7), Krankowski, A., I. Shagimuratov, L. Baran, I. Epishov, and N. Tepenitzyna (6), The occurrence of polar cap patches in TEC fluctuations detected using GPS measurements in Southern Hemisphere, Adv. Space Res., 38, Liu, L., B. Zhao, W. Wan, S. Venkartramann, M.-L. Zhiang, and X. Yue (7), Yearly variations of global plasma densities in the topside ionosphere at middle and low latitudes, J. Geophys. Res., 11, A733, doi:1.19/7ja183. Lockwood, M., and H. Carlson Jr. (199), Production of polar cap electron density patches by transient magnetopause reconnection, Geophys. Res. Lett., 19(17), Mendillo, M., C.-L. Huang, X. Pi, H. Rishbeth, and R. Meier (5), The global ionospheric asymmetry in total electron content, J. Atmos. Sol.-Terr. Phys., 67, Mitchell, C., L. Alfonsi, G. De Franceschi, M. Lester, V. Romano, and A. Wernik (5), GPS TEC and scintillation measurements from the polar ionosphere during the October 3 storm, Geophys. Res. Lett., 3, L1S3, doi:1.19/4gl1644. Montenbruck, O., and R. Kroes (3), In-Flight Performance Analysis of the CHAMP BlackJack GPS Receiver, GPS Solutions, Springer-Verlag, Berlin. Park, J., R. Ehrlich, H. Lühr, and P. Ritter (1), Plasma irregularities in the high-latitude ionospheric F-region and their diamagnetic signatures as observed by CHAMP, J. Geophys. Res., 117, A13, doi:1.19/1ja Pedatella, N., and K. Larson (1), Routine determination of the plasmapause based on COSMIC GPS total electron content observations of the midlatitude trough, J. Geophys. Res., 115, A931, doi:1.19/1ja1565. Pedersen, T., B. Fejer, R. Doe, and E. Weber (), An incoherent scatter radar technique for determining two-dimensional horizontal ionization structure in polar cap F region patches, J. Geophys. Res., 15(A5), 1,637 1,655. Pinnock, M., A. Rodger, J. Dudeney, K. Baker, P. Newell, R. Greenwald, and M. Greenspan (1993), Observations of an enhanced convection channel in the cusp ionosphere, J. Geophys. Res., 98(A3), Richards, P., J. Fennelly, and D. Torr (1994), EUVAC: A solar EUV flux model for aeronomic calculations, J. Geophys. Res., 99(A5), Richmond, A. (1995), Ionospheric electrodynamics using Magnetic Apex Coordinates, J. Geomag. Geoelectr., 47, Rodger, A., and A. Graham (1996), Diurnal and seasonal occurrence of polar patches, Ann. Geophys., 14, Rodger, A., M. Pinnock, and J. Dudeney (1994), A new mechanism for polar patch formation, J. Geophys. Res., 99(A4), Schunk, R., and J. Sojka (1996), Ionosphere-thermosphere space weather issues, J. Atmos. Terr. Phys., 58(14), Sojka, J., M. Bowline, and R. Schunk (1994), Patches in the polar ionosphere: UT and seasonal dependence, J. Geophys. Res., 99(A8), 14,959 14,97. Spencer, P., and C. Mitchell (11), Imaging of 3-D plasmaspheric electron density using GPS to LEO satellite differential phase observations, Radio Sci., 46, RSD4, doi:1.19/1rs4565. Stephens, P., A. Komjathy, B. Wilson, and A. Manucci (11), New leveling and bias estimation algorithms for processing COSMIC/FORMOSAT- 3 data for slant total electron content measurements, Radio Sci., 46, RSD1, doi:1.19/1rs4588. Stolle, C., J. Lilensten, S. Schlüter, C. Jacobi, M. Rietveld, and H. Lühr (6), Observing the north polar ionosphere on 3 October 3 by GPS imaging and IS radars, Ann. Geophys., 4, Su, Y., G. J. Bailey, and K.-I. Oyama (1998), Annual and seasonal variations in the low-latitude topside ionosphere, Ann. Geophys., 16, Syndergaard, S. (7), FORMOSAT-3/COSMIC ionospheric data processing and availability for data assimilation systems, Presentation at the IUGG XXIV General Assembly, Perugia, Italy, July 13, 7. Valladares, C., D. Decker, R. Sheehan, D. Anderson, T. Bullett, and B. Reinisch (1998), Formation of polar cap patches associated with northto-south transitions of the interplanetary magnetic field, J. Geophys. Res., 13(A7), 14,657 14,67. Wanliss, J., and K. Showalter (6), High-resolution global storm index: Dst versus SYM-H, J. Geophys. Res., 111, A, doi:1.19/ 5JA1134. Weber, E., J. Buchau, J. Moore, J. Sharber, R. Livingston, J. Winningham, and B. Reinisch (1984), F layer ionization patches in the polar cap, J. Geophys. Res., 89(A3), Weber, E., J. Klobuchar, J. Buchau, H. Carlson Jr., R. Livingston, O. de la Beaujardiere, M. McCready, J. Moore, and G. Bishop (1986), Polar cap F layer patches: Structure and dynamics, J. Geophys. Res., 91(A11), 1,11 1,19. Wood, A., S. Pryse, H. Middleton, and V. Howells (8), Multi-instrument observations of nightside plasma patches under conditions of IMF Bz positive, Ann. Geophys., 3, Wu, J., S. Wu, G. Hajj, W. Bertiguer, and S. Lichten (1993), Effects of antenna orientation on GPS carrier phase measurements, Manuscr. Geodaet., 18,