Stationary planetary wave and nonmigrating tidal signatures in ionospheric wave 3 and wave 4 variations in FORMOSAT-3/COSMIC observations

Transcription

1 JOURNAL OF GEOPHYSICAL RESEARCH: SPACE PHYSICS, VOL. 118, , doi: /jgra.50583, 2013 Stationary planetary wave and nonmigrating tidal signatures in ionospheric wave 3 and wave 4 variations in FORMOSAT-3/COSMIC observations Loren C. Chang, 1 Chien-Hung Lin, 2 Jia Yue, 3 Jann-Yenq Liu, 1 and Jia-Ting Lin 2 Received 4 March 2013; revised 16 September 2013; accepted 18 September 2013; published 3 October [1] The wave 3 and wave 4 modulations of the Equatorial Ionization Anomalies are a robust feature of the low-latitude ionosphere, when viewed at constant local time. Although initially associated, respectively, with DE2 and DE3, nonmigrating diurnal tides in the mesosphere and lower thermosphere region, recent results have suggested that the wave 3 and wave 4 may also have significant contributions from other tidal and stationary planetary wave (SPW) signatures. We present observations of total electron content (TEC) variations associated with tidal and SPW signatures comprising the ionospheric wave 3 and wave 4 structures from FORMOSAT-3/COSMIC from 2007 to We find that the wave 3 (wave 4) feature is comprised predominately by DE2 (DE3) and SPW3 (SPW4) signatures in TEC throughout all 5 years, with contributions from SE1 (SE2) being less significant. The wave 3 component also has recurring contributions from DW4 during December/January. The absolute amplitudes of all the aforementioned tidal and SPW signatures are directly related to the level of solar activity and the semiannual variation in zonal mean TEC. After normalizing by the zonal mean, the relative amplitudes of the wave 4 signatures are inversely related to solar activity through 2010, which is not seen with the wave 3-related signatures. The seasonal variation and phases of the main constituents of wave 3 and wave 4 are consistent from year to year, as evidenced by the interannual recurrence in the peak and trough locations of wave 3 and wave 4. Citation: Chang, L. C., C.-H. Lin, J. Yue, J.-Y. Liu, and J.-T. Lin (2013), Stationary planetary wave and nonmigrating tidal signatures in ionospheric wave 3 and wave 4 variations in FORMOSAT-3/COSMIC observations, J. Geophys. Res. Space Physics, 118, , doi: /jgra Introduction [2] Since the initial discovery of wave 4 longitudinal variations in IMAGE (Imager for Magnetopause to Aurora Global Exploration) FUV airglow observations [Sagawa et al., 2005] and their connection to nonmigrating tides excited via tropospheric latent heat release [England et al., 2006; Immel et al., 2006; Hagan et al., 2007], it has become accepted that the nonmigrating tides propagating upward from the middle to the lower atmosphere can play an important role in producing such variability in the low-latitude ionosphere when observed at constant local times as commonly utilized in ionospheric studies. The wave 4 and wave 3 longitudinal variations of the low-latitude iono- 1 Institute of Space Science, National Central University, Jhongli, Taiwan. 2 Department of Earth Science, National Cheng Kung University, Tainan City, Taiwan. 3 Center for Atmospheric Science, Hampton University, Hampton, Virginia, USA. Corresponding author: L.C. Chang, Institute of Space Science, National Central University, No. 300 Jhongda Rd., Jhongli, Taoyuan County 320, Taiwan. (loren@ncu.edu.tw) American Geophysical Union. All Rights Reserved /13/ /jgra sphere have become associated primarily with DE3 and DE2, respectively. Rationales for this include the similarity of the DE3 and DE2 seasonal variation in the mesosphere and lower thermosphere (MLT) region to that of ionospheric wave 4 (wave 3) variability [Fang et al., 2009; Pancheva and Mukhtarov, 2010], as well as the long vertical wavelengths and low-latitude structure of DE3 and DE2 in the middle atmosphere, allowing them to penetrate well into the ionospheric E-region dynamo [Forbes et al., 2006, 2008]. These nonmigrating tides may then cause changes further aloft by a variety of mechanisms, including modification of E- and F-region dynamo electric fields, thermosphere neutral composition, and F-region meridional wind fields [England et al., 2010; Maute et al., 2012]. [3] However, DE3 and DE2 are not the only global-scale oscillations that may alias into wave 4 and wave 3 stationary planetary wave (SPW) patterns. As described by Forbes et al. [2006], any nonmigrating tide in a constant local time frame can alias onto an SPW with wave number that is the difference between the zonal wave number and the tidal harmonic. For example, in addition to DE3, the wave 4 variation in the ionosphere might also be driven by SE2, TE1, DW5, and SPW4, just to name a few. [4] Recent studies have begun to address this ambiguity. Oberheide et al. [2011] found through empirical modeling

2 and analysis of TIMED (Thermosphere Ionosphere Mesosphere Energetics and Dynamics) satellite data that although DE3 dominated in the low-latitude MLT, SE2 and SPW4 could also attain comparable amplitudes in that region, with the latter apparently generated via nonlinear interaction between DE3 and DW1. Similarly, Pancheva and Mukhtarov [2012] found significant contributions to the wave 4 structure (wave 3) from SPW4 and SE2 (SPW3, DW4, and SE1) in addition to DE3 (DE2) using COSMIC electron densities over a 1.5 year period. The contribution of SPW4 to the wave 4 ionosphere variation was also found in TIME-GCM (Thermosphere Ionosphere Mesosphere Electrodynamics General Circulation Model) numerical experiments by Pedatella et al. [2012], who also concluded that SPW4 was likely produced via nonlinear interaction between DE3 and DW1. In such an interaction, interacting parent waves produce child waves with frequencies and wave numbers that are the sum and difference of those of the parent waves [Teitelbaum and Vial, 1991]. As will be elaborated on later, this basic nonlinear interaction can also be used to describe changes to tidal or PW frequency or zonal wave number that results when such oscillations are mapped from the neutral MLT into ionospheric fields [Yue et al., 2013]. These phenomena were explored in the numerical experiments of Yue et al. [2013] concerning the mapping of the quasi-two day wave (QTDW) from the MLT into the ionosphere. Although the periodicity and zonal wave number of the MLT signal remained dominant in the ionosphere, the study also identified the occurrence of new QTDW components in the ionosphere with different zonal wave numbers from the original wave in the MLT, a phenomenon referred to as wave number broadening, and attributed to magnetic declination effects. [5] The variation of such tidal signatures in the thermosphere and ionosphere with the solar cycle is also a topic of interest. Pedatella and Forbes [2009] quantified interannual variations in the relative strength of the ionospheric wave 4 feature from 1960 to 1993 by using the ratio between ionosonde fyf 2 values measured at peak and trough locations of the wave 4 feature as a proxy. This f o F 2 ratio proxy was found to be anticorrelated to the dominant and 7.86 year periodicities of the solar cycle as manifested in the F 10.7 solar radio flux. However, after detrending to remove solar cycle effects, the f o F 2 ratio proxy was found to show a significant correlation with the Oceanic NiQno Index, suggesting that the ionospheric wave 4 can be sensitive to the El NiQno Southern Oscillation (ENSO) in the troposphere. [6] Another known source of interannual tidal variability in the neutral MLT region is the quasi-biennial oscillation (QBO). TIMED SABER (Sounding of the Atmosphere using Broadband Emission Radiometry) observations by Forbes et al. [2008] found prominent QBO-related interannual variability in the MLT temperature amplitudes of both the DE3 and DE2 nonmigrating tides, with the hemisphere of maximum amplitudes for DE2 showing potential QBOrelated changes in the mesosphere. Further expanding on this using SABER and TIDI (TIMED Doppler Interferometer) observations from the TIMED satellite, Oberheide et al. [2009] found that DE3 amplitudes in the neutral MLT from 2002 to 2008 showed signs of modulation by the QBO but no apparent signs of solar cycle dependence below 120 km. However, DE3 amplitudes in the neutral thermosphere 6652 above 120 km were found to be anticorrelated to the level of solar activity in both analysis by TIMED-derived Hough mode extensions, as well as in exosphere zonal wind and temperature observations from CHAMP. This was attributed to increased tidal dissipation in the thermosphere during solar maximum. In contrast to the neutral thermosphere, Luan et al. [2012] found that the absolute amplitudes of the TW3 migrating tidal response were positively correlated to the solar cycle in GPS total electron content (TEC) observations from 1999 to However, the TW3 relative amplitudes obtained by normalizing absolute amplitudes by the zonal mean values were found to show variable relation with the solar cycle depending upon latitude and year, demonstrating that tides in the ionosphere can show different dependencies relative to those in the neutral thermosphere. [7] In this study, we expand upon these past results by examining the tidal and SPW composition of the wave 3 and wave 4 structures in TECs using 5 years of FORMOSAT- 3/COSMIC observations. Our results agree with the numerical experiments of Pedatella et al. [2012], in finding that the wave 4 longitudinal variation is dominated by DE3 and SPW4. We also find that in contrast to the neutral thermosphere results of Oberheide et al. [2009], the constituent components of wave 3 and wave 4 in TEC show a positive relation to the solar cycle in absolute amplitude, while only the relative amplitudes of the wave 4 constituents are anticorrelated to the solar cycle. The 5 years covered by our study provides one of the first looks at interannual variability in the composition of ionospheric wave 3 and wave 4 features through the extended solar minimum of , into the subsequent upward cycle in solar activity. [8] We note that in the following study, we refer to any global-scale oscillation in the COSMIC TECs of appropriate frequency and zonal wave number using tidal or SPW component nomenclature, irrespective of whether that signature can be attributed to in situ excitation in the ionosphere or upward propagation from the neutral MLT. Although applying such middle atmospheric analysis methods is somewhat unconventional in the ionosphere, they have been employed to good effect on several recent studies focused on understanding tidal and precipitable water (PW) coupling in the ionosphere [e.g., Pancheva and Mukhtarov, 2012; Lin et al., 2007; Luan et al., 2012; Yue et al., 2013]. In this context, aliasing of forcing signatures from the neutral atmosphere with the local time and longitudinal dependencies in the ionosphere may also be viewed as having a similar effect to nonlinear interactions in the neutral atmosphere, producing child waves due to amplitude modulation [Teitelbaum and Vial, 1991; Yue et al., 2013], as mentioned previously. However, as an analytical global tidal solution to the ionospheric plasma equations does not yet exist, the use of tidal nomenclature in this study should only be considered to be a shorthand reference to components of the periodic analysis performed. 2. Methodology [9] FORMOSAT-3/COSMIC (hereafter referred to as COSMIC) is a constellation of six microsatellites distributed into a final configuration of six orbital planes at 800 km, 72 ı inclination [Lin et al., 2007]. Using GPS occultation, each spacecraft provides over 2000 pseudorandomly distributed

3 vertical profiles of electron density per day, up to an altitude of 800 km [Cheng et al., 2006]. [10] Our retrieval of tidal and SPW signatures in COS- MIC TECs is performed in a similar manner as that detailed in Lin et al. [2007]. Bad data points are removed, which are defined as points with negative values or plasma frequencies exceeding 20 MHz (corresponding to an electron density of cm 3 ). This cutoff is based on the longterm ionosonde observations by Liu et al. [2006], who found maximum values for NmF2 (ionospheric peak electron density) of about cm 3 (15.6 MHz) at F 10.7 levels of around 300. The vertical electron profile from each occultation is then vertically integrated above 200 km to compute the total electron content (TEC), which is then binned into an overlapping 5 ı magnetic apex latitude grid, over a 40 day averaging window, which is then stepped daily through the entire period between January 2007 and December This 40 day window provides full longitude/local time coverage in the low- to midlatitude region during most of this period, allowing for full tidal and SPW separation (see Appendix A for further details). The sampled TECs in each latitude bin and each day are then fitted to basis functions of the form F(, t) = NF+ 3X 4X 4X OF n,s cos(nt s O n,s )+ OF s cos(s O s ) (1) n=1 s= 4 [11] NF is the zonal mean value. n and s are the tidal harmonic and zonal wave number (westward negative). The perturbations account for all diurnal, semidiurnal, and terdiurnal tidal components with zonal wave numbers from westward 4 (s = 4) to eastward 4 (s =4), as well as SPWs with zonal wave number 1 to 4. OF and O are, respectively, the fitted amplitude and phase for each component. = 2 24 h 1, is the longitude, and t is universal time. The basis function amplitudes and phases were fitted using the linear least squares algorithm [Wu et al., 1995], while the amplitude and phase uncertainties were derived from the least squares fit covariances. [12] Figure 1a shows the daily (blue line) values of F 10.7 solar radio flux throughout the entire observational period between 2007 and the end of 2011, while the black line shows the 40 day averaged value, using the same sliding window as that used for the COSMIC least squares fitting. The extended solar minimum between 2008 and 2009 is apparent, as well as the subsequent increasing phase of solar activity thereafter. Values of F 10.7 range from around 65 at the end of 2008, to 40 day averaged values in excess of 145 in the latter portion of [13] The COSMIC zonal mean TECs are shown as a function of magnetic apex latitude and time in Figure 1b and show a single peak in the equatorial region. On an interannual scale, the zonal mean TECs show a similar trend as F 10.7, with equinox amplitudes of roughly 10 TECu during November 2008/April 2009, to values as high as 37 TECu in late October Seasonally, the zonal mean TECs are symmetric about the equator with larger amplitudes around the equinoxes and shift toward the summer hemisphere during the solstices with smaller amplitudes. The seasonal variation is consistent with that found previously by Mukhtarov and Pancheva [2011] using COSMIC f o F 2 observations from January 2008 to March s=1 2009, who associated the semiannual background amplitude variation to seasonal variations in neutral thermosphere composition [Rishbeth et al., 2000]. [14] Similar to some previous studies, we have also utilized a relative (to the zonal mean) amplitude in our analysis of ionospheric tidal and PW perturbations, in order to elucidate certain types of interannual variability relative to the solar cycle [Luan et al., 2012]. Our relative amplitudes are computed by normalizing absolute tidal and PW amplitudes by the maximum zonal mean TEC value in the low-latitude region (F 0 = OF ). This allows for the importance of each tidal or SPW component relative to the zonal F max mean to be identified, while still retaining information on the latitudinal structure by preventing artificial inflation of relative amplitudes in regions where the zonal mean TECs are small. The uncertainties for the relative amplitudes were then determined using error propagation of the perturbation and zonal mean amplitude uncertainties ("( OF) and "(F max ), respectively): "(F 0 0 OF "( 0 2 OF) + "(F max ) max 6653 [15] The resulting uncertainties for the dominant tidal and SPW signatures examined in TECs will be shown in the next section. Generally speaking, the uncertainties of both relative and absolute amplitudes are largest in the subtropics, likely due to the COSMIC sampling pattern, with the absolute amplitude uncertainties sharing a similar distribution and magnitude to those of the zonal mean TECs shown in Figure 1c. The absolute and relative amplitude uncertainties increase beyond 2010, which may be related to both the increased level of solar activity, as well as reduced coverage due to battery degradation aboard two of the COSMIC microsatellites during this time (Figure A1a). Figure 1d shows the zonal mean TEC uncertainties, expressed as percentage of the zonal mean TEC, as a function of time near their peak latitudes of 15 ı in both hemispheres. The relative zonal mean TEC uncertainties as quite small less than 1% through 2010, with peak values less than 4% thereafter. As will be shown later, the uncertainties alone are too small to account for the level of observed seasonal and interannual variability of the zonal mean, as well as several features of the dominant wave 3 and wave 4 tidal and PW signatures. [16] Figure 2a shows the latitude and time variation of the TEC DW1 (migrating diurnal) absolute amplitudes, which may be involved in the nonlinear generation of some of the tidal and SPW signatures associated with wave 3 and wave 4 that will be discussed in the following section. The latitudinal and seasonal variation of DW1 is quite similar to that of the zonal mean TECs showing a single peak near the equator, maximizing around the equinoxes. The DW1 phases (not shown) show that the local time of maximum is between 1400 and 1500 throughout the entire low/middle latitude region. This is again consistent with the simulation and observational results of Hagan et al. [2001] and Mukhtarov and Pancheva [2011], who associated these features with solar zenith angle variations manifesting in in situ photoionization of the neutral thermosphere by extreme ultraviolet absorption, as the upward propagating DW1 modes from the MLT are not capable of thermosphere propagation. Our multiyear results further show that this pattern is repeated across

4 Figure 1. (a) Daily (blue line) and 40 day averaged (black line) values of F 10.7 solar radio flux. (b) Zonal mean TECs as a function of magnetic apex latitude and time. (c) Zonal mean TEC uncertainty. (d) % uncertainty for zonal mean TEC at magnetic latitudes of 15 ı N (black line) and 15 ı S (blue line). multiple years, with absolute amplitudes increasing with solar activity. Figure 2b shows the DW1 relative amplitudes, which show a pronounced anticorrelation to the solar cycle, being large during 2008/2009, and decreasing thereafter. 3. Results 3.1. Wave 4 [17] Possible tidal and SPW signatures comprising wave 4 include DE3, DW5, SE2, SW6, TE1, TW7, and SPW Zonal wave numbers greater than four are not included in our fit, and these components are thus not described here. Additionally, TE1 was found to have negligibly small amplitudes compared to the other components and is therefore not shown. [18] The remaining constituents of wave 4 are shown in Figure 3, which shows the absolute amplitudes of DE3, SE2, and SPW4 fitted from COSMIC TECs as a function of magnetic apex latitude and time. The color scale is identical in all the plots to facilitate easy comparison. It can be seen that the

5 Figure 2. (a) Absolute and (b) relative amplitudes of DW1 in COSMIC TECs as a function of latitude and time. x axis is marked by month, with January marked by year. absolute amplitudes of DE3, SE2, and SPW4 manifest primarily in twin latitude bands between roughly 10 ı and 30 ı latitude in each hemisphere. Additionally, the amplitudes of DE3 and SPW4 are consistently larger than those of SE2, by almost a factor of 2 or more. [19] The absolute amplitudes of all the tidal and SPW signatures fitted from COSMIC TECs reflect the semiannual and solar cycle variation of the zonal mean TECs, a feature also observed by Mukhtarov and Pancheva [2011] in f o F 2 and attributed to changes in the background ionosphere. To suppress these two dependences, the daily absolute tidal and SPW amplitudes at all latitudes are normalized by the maximum daily equatorial zonal mean TEC, as described previously. These normalized amplitudes are shown in Figure 4. [20] From Figure 4a, it can be seen that DE3 has peaks in the northern and southern EIA regions between 15 ı and 20 ı magnetic latitude. The primary maxima for DE3 occur between July and October, with the southern peak consistently stronger than the northern peak, and roughly a month later. The secondary maxima occurring in March/April are also resolved from 2007 to 2010 and are also stronger in the Southern Hemisphere. The secondary maxima in 2011 occur around February/March and are stronger in the Northern Hemisphere. [21] In contrast to DE3, the relative amplitudes of SE2 (Figure 4b) are considerably smaller. SE2 also shows peaks at EIA latitudes, though slightly further poleward compared to DE3, with the southern EIA values again greater than those in the northern EIAs. Maxima in SE2 activity are resolved during September/October (corresponding to DE3 maxima), as well as December/February (corresponding to DE3 minima in all years except the beginning of 2007). The September/October SE2 maximum never exceeds more than about half the DE3 values. Over the long term, the SE2 amplitudes show a decreasing trend throughout the entire 5 year period. [22] SPW4 relative amplitudes are shown in Figure 4c. In contrast to SE2, SPW4 shows amplitudes comparable to, and in some cases exceeding those of DE3, particularly in the Northern Hemisphere. SPW4 also peaks in the EIA region at 15 ı magnetic latitude, amplifying between June and October. This is consistent with the July October maximum of DE3; however, SPW4 persists for a longer duration of time, and contrary to DE3, is consistently stronger in the Northern Hemisphere. A secondary enhancement during February April is also resolved, which is also stronger in the northern EIA region. [23] To better elucidate the interannual variation of the dominant wave 4 constituents, the variation of DE3 and SPW4 relative amplitudes, as well as their corresponding uncertainties, is plotted near the peak magnetic latitudes of 15 ı in both hemispheres, as shown in Figure 5. A 30 day boxcar average has been applied to the time series in order to highlight seasonal and interannual variation. The annual maximum relative amplitudes of DE3 during the boreal summer/fall are generally inversely related to solar activity, with the strongest relative amplitudes of 12% of the maximum zonal mean TECs occurring in late The exception to this is 2011, where the August peak in DE3 relative amplitudes is larger than that in 2010, although occurring over a shorter period of time. SPW4 shows similar interannual variation, although the difference between the boreal summer peaks in 2010 and 2011 is not as pronounced as those of DE3. From the range of uncertainties, it can be seen 6655

6 Figure 3. Absolute amplitudes of (a) DE3, (b) SE2, and (c) SPW4 in COSMIC TECs as a function of latitude and time. x axis is marked by month, with January marked by year. that seasonal variation cannot be attributed to measurement uncertainty. Similarly, the difference between the relative amplitudes of the annual boreal summer/austral winter peaks in the solar minimum year of 2008 and during the increasing solar activity of 2010 is as large as 5% in the Southern Hemisphere, which is larger than the relative amplitude uncertainty of about 1.5%. It can also be seen that SPW4 is considerably larger than DE3 at 15 ı N in 2008 and A similar relation holds at 15 ı N during most of the other years through late 2010, but the difference between SPW4 and DE3 amplitudes lies within the range of uncertainties. [24] A potential explanation for changes not directly related to the solar cycle could include interannual variability in the DE3 component in the MLT region due to lower and middle atmospheric sources such as ENSO [Pedatella and Forbes, 2009] and the QBO [Oberheide et al., 2009]. Unfortunately, a direct comparison to the detrended proxies in the former study cannot be performed since the 5 year COSMIC observational period used in this study is a little less than half a solar cycle. In the case of the latter, SABER and TIDI observations by Oberheide et al. [2009] found a prominent QBO modulation of DE3 amplitudes in the neutral MLT region, which resulted in stronger DE3 amplitudes during the boreal summer/fall of even numbered years. Although this could potentially explain the stronger DE3 relative amplitudes in 2008, it is difficult to discern its effects during the increasing phase of solar activity that follows. [25] In order to relate the fitted tidal and SPW signature amplitudes and phases to the constant local time frame commonly employed in ionospheric studies, we present Figure 6. The two figures in the top row show the TEC profile at 15 ı magnetic latitude (again normalized by the maximum daily TEC value) reconstructed from the amplitudes and phases of all of the fitted tidal and SPW component signatures (equation (1)), averaged during the period of largest wave 4 variation from August to November of each year, 6656

7 Figure 4. Relative amplitudes of (a) DE3, (b) SE2, and (c) SPW4 in COSMIC TECs as a function of latitude and time. Units % of maximum daily zonal mean TEC. x axis is marked by month, with January marked by year. at 16 h local time. The results are essentially the residual after removing the zonal mean. Uncertainties in the reconstructed TEC profiles are denoted by the dotted lines and were computed using the amplitude and phase uncertainties resulting from the least squares fit. It can be seen from Figures 6a and 6c that throughout all 5 years, the zonal variation of the EIAs show several consistent features, including the longitudes of the peaks and troughs, as well as the reduced values between 180 ı and 270 ı longitude in the southern EIA due to the magnetic declination angle tilt, and the corresponding increase in the northern EIA [Rishbeth, 1998; Liu et al., 2010]. This indicates a consistent relationship between the relative amplitudes and phases of the various constituent tidal and SPW signatures present in the EIA regions at this local time. [26] Figures 6b and 6d show the wave 4 variation in the northern and southern EIAs but reconstructed using only 6657 DE3, SE2, TW1, and SPW4 signatures, which alias into a wave 4 pattern when viewed at a constant local time. By comparing the locations of the peaks and troughs in these results to the combined values shown in Figures 6a and 6c, the contribution of the wave 4 pattern to the overall zonal EIA variation at this local time can be identified. Note the change in scale of the y axes between the top and bottom plots. In the northern and southern EIAs, the peaks and troughs of the overall zonal variation align relatively well with those of wave 4. The longitudes of the wave 4 peaks and troughs are also consistent throughout all 5 years. [27] The largest averaged values of wave 4 relative to the maximum zonal mean TEC are seen in 2007 and 2008, and the smallest in 2010 and 2011, consistent with the previous results shown in Figure 4. We must stress, however, that although interannual variability in the amplitudes of the

8 Figure 5. Relative amplitudes of DE3 (black) and SPW4 (blue) in COSMIC TECs as a function of time at (a) 15 ı N and (b) 15 ı S. Units of maximum daily zonal mean TEC are in %. Range of uncertainties are denoted by dotted lines. Figure 6. Profiles of the (a, b) northern and (c, d) southern EIAs at 16 LT reconstructed from all fitted tidal and SPW signatures (Figures 6a and 6c), and from DE3, SE2, TW1, and SPW4 only (Figures 6b and 6d). Results are taken from 15 ı magnetic latitude, averaged from August to November. Uncertainties are denoted by thin dotted lines. 6658

9 Figure 7. Same as Figure 3 but for (a) DE2, (b) SE1, (c) SPW3, and (d) DW4. wave 4 constituent signatures in TECs will be manifested here, the wave 4 amplitudes, as well as the longitudes of the peaks and troughs, also reflect the results of constructive and destructive interference between the different tidal and SPW signatures and will vary with local time. The relative amplitudes of the individual components shown in Figures 4 and 5 are therefore a better gage of interannual amplitude variation of specific components. However, it can be seen from the similar locations of the wave 4 peaks and troughs through all 5 years that the relative phases of the tidal and SPW signatures associated with wave 4 are interannually consistent. 6659

10 Figure 8. Same as Figure 4 but for (a) DE2, (b) SE1, (c) SPW3, and (d) DW Wave 3 [28] We now consider the constituent signatures of the wave 3 component, which can include DE2, DW4, SE1, SW5, T0, TW6, and SPW3. We again consider all tidal and SPW signatures with zonal wave number less than or equal to 4, of which the T0 component is found to be negligibly small and is not shown. The absolute amplitudes of the remaining DE2, SE1, SPW3, and DW4 components are shown in Figure 7 and again reflect both the semiannual seasonal variation and the positive correlation to solar activity. The aforementioned influences are again suppressed through normalization by the maximum daily zonal mean amplitude, with the resulting relative amplitudes shown in Figure 8 and relative amplitude line plots of the three 6660

11 Figure 9. Relative amplitudes of DE2 (black, 20 ı S), SPW3 (blue, 15 ı S), and DW4 (red, 20 ı S) in COSMIC TECs as a function of time. Units of maximum daily zonal mean TEC are in %. Range of uncertainties are denoted by dotted lines. largest components at selected magnetic latitudes shown in Figure 9. [29] From Figure 8a, it can be seen that DE2 is strongest in the Southern Hemisphere, where it maximizes from roughly October to March around roughly 30 ı S. Another distinct maxima occur between March and June, with the peak appearing to shift equatorward with time starting from about 15 ı S. The maximum relative amplitudes are smaller than those of DE3, attainting roughly 8% of the maximum daily zonal mean TEC. Correlation between DE2 and the solar cycle is not immediately apparent, and the line plots of relative amplitude in Figure 9 show that the level of interannual variability for the May peaks of DE2 at 20 ı S lies within the range of uncertainties, with the exception of that in [30] The SE1 relative amplitudes shown in Figure 8b show similar spatial and seasonal variation compared to SE2. However, the relative amplitudes of SE2 are at most, half those of DE2, and are smaller than those of the SPW3 and DW4 signatures. Again, there is no easily observable relation between SE2 and the solar cycle. [31] Figure 8c shows the SPW3 relative amplitudes. SPW3 occurs predominately in the southern EIA region, maximizing near 15 ı S from roughly April to August. This overlaps with the March June maximum of DE2, which also occurs around the same latitude. A secondary SPW3 maxima also occur from December to February, slightly further poleward around 20 ı S. This is concurrent in time with the DE2 October March peak, though the latter is located further Figure 10. Profiles of the southern EIAs at 16 LT during (a, b) March June and (c, d) November December reconstructed from all fitted tidal and SPW signatures (Figures 10a and 10c) and from DE2, SE1, T0, and SPW3 only (Figures 10b and 10d). Results are taken from 15 ı S magnetic latitude in March June and 20 ı S in November December. Uncertainties are denoted by thin dotted lines. 6661

12 south. Nonetheless, the similarities in spatial and seasonal variation between SPW3 and DE2 again point to the generation of SPW3 via a multiplicative interaction between DE2 and DW1. [32] The relative amplitudes of DW4 are shown in Figure 8d. DW4 maximizes during December February around 20 ı S, with a smaller peak resolved at similar latitudes and times in the Northern Hemisphere. As DW4 was not identified as being large in the SABER lower thermosphere temperature observations of Forbes et al. [2006], it is unlikely that this component is the result of upward coupling from the MLT. Rather, it is possible that DW4 in the COSMIC TECs is the result of aliasing of SPW3 in the neutral atmosphere with the migrating diurnal variability of the ionosphere. This is supported by the similar variability in time and space between the December February peaks of DW4 and SPW3, as can also be seen in Figure 9. It is not clear why DW4 is not also produced at similar amplitudes during the larger April August SPW3 maximum. [33] Figure 10 shows the zonal variation in the southern EIA region reconstructed from all fitted tidal and SPW components (top row) and from components corresponding to the wave 3 component (bottom row) during March June at 15 ı S (left column) and November December at 20 ı S (right column), as well as their associated uncertainties. The selected time periods and latitudes correspond to those where wave 3 constituent signatures were previously identified as being large (Figure 8). Similarly, the results for the northern EIA are not shown here since the wave 3 constituent signatures were found to be small in that region. From Figures 10a and 10c, it can be seen that the overall zonal variation from all the reconstructed components is again relatively consistent from year to year and is similar to the previous wave 4 results (Figure 6) in exhibiting a large trough between 180 ı and 270 ı longitude due to magnetic declination effects. [34] Figures 10b and 10d show wave 3 reconstructed from its constituent DE2, SE1, TS0, and SPW3 signatures. Compared to the previous wave 4 results, wave 3 has smaller relative amplitudes, and the contribution of wave 3 to the overall zonal variation is much less distinct. This may be related to the fact that during November December, the TEC peaks of the largest constituents of wave 3 (the DE2 and SPW3 signatures) showed a greater separation in latitude. This would limit the degree to which constructive interference could occur. The longitudes of the wave 3 peaks and troughs also show much less correspondence with those of the overall zonal variation, although the relative phases of the constant signatures of wave 3 are still interannually consistent, again indicating that the relative phases of the wave 3 constituents change little from year to year. 4. Discussion [35] It has been suggested that SPW4 can be generated as a byproduct of a DW1/DE3 nonlinear interaction in the neutral atmosphere [Hagan et al., 2009; Pedatella et al., 2012]. An additional source might also be the modulation of tides and SPW signatures in the neutral atmosphere by the local time variability that dominates the ionosphere. The numerical experiments of Maute et al. [2012] demonstrated that the importance of the wind dynamos in the E- and F-regions 6662 can vary depending on local time, due to the variations in conductivity with local time. The conductivities themselves determine the polarization electric fields formed by the wind dynamos, which then produce the E B drift driving the equatorial fountain forming the EIAs. We may infer from these results that the net effect on the tidal and SPW signatures in the ionosphere will be a modulation of the original MLT tidal signature by the local time dependence of the wind dynamos controlling the coupling. In an analogy to the neutral atmosphere, this aliasing produces the same result as a nonlinear interaction with the parent waves being the original MLT tide and a tidal or SPW component describing the time and longitudinal dependence of the ionospheric conductivities, which will produce sideband child waves as a byproduct [Teitelbaum and Vial, 1991]. [36] The largest component of local time dependence for both the Hall and Pederson conductivities in the ionosphere may be approximated as being roughly migrating diurnal (DW1) in nature large in the daytime, small at night. Hence, we can expect that in addition to frequency and zonal wave number of the original tidal signature in the MLT, a portion of the ionospheric response will correspond to the child waves produced by aliasing between the original MLT tide and the DW1 component of the ionospheric conductivities. For the cases where the MLT parent wave is DE3 (DE2), the child waves that would result would be SPW4 and SE2 (SPW3 and SE1). Similar child waves could also result from aliasing between DE3 (DE2) signatures in conductivity and the DW1 tide in the thermosphere neutral winds. Another possibility for child wave production in the ionosphere could be the zonal mean component of conductivity acting in conjunction with nonmigrating tidal winds to produce a polarization electric field and EB drift that varies with the tidal period and zonal wave number. The tidally modulated E B drift then aliases with the photoionizationdriven DW1 component of electron density via advection (transport). [37] Although the above hypotheses require further study, they do form a plausible explanation for the presence of the SPW and nonmigrating semidiurnal signatures in our observations and are supported in our results by the similarity in the seasonal variations of DE3 and SPW4, as well similarities in the interannual variation of both components: The relative amplitudes of SPW4 and DE3 are both largest in 2008 and smallest in However, the difference in hemispheric asymmetry between the two components is interesting and does not immediately agree with past examples of nonlinear interactions in the MLT, where parent and child waves showed similar spatial structure [Palo et al., 2007]. This may be explained by spatial differences between DE3 and SPW4 in the neutral MLT, hemispheric asymmetry in the ionospheric conductivities, and/or may also reflect the fact that the DW1 amplitudes in the ionosphere are significantly larger than those of the other tidal and SPW components [Mukhtarov and Pancheva, 2011; Chang et al., 2013]. Further numerical studies will also be required to understand why the SPWs are stronger than the nonmigrating semidiurnal components. [38] The reconstructed zonal variations of wave 4 (Figures 6b, 6d) and wave 3 illustrate that the relative phases of the tidal and SPW signatures comprising both types of ionospheric variation change little from year to year. As

13 these tidal signatures in the neutral MLT, particularly DE3 and DE2, have been attributed to forcing from tropospheric latent heat release, the recurrent phases are indicative not only of interannually consistent source distribution in the troposphere, but also long-term consistency in mechanisms affecting the vertical wavelength in the neutral middle atmosphere (such as zonal mean winds, as well as gravity wave and molecular dissipation), as well as any nonlinear coupling mechanisms in the ionosphere. The relative contributions of the physical mechanisms driving the latter (e.g., changes in conductivity, as well as details of the dynamo interactions) remain to be explored. [39] The relative amplitudes utilized in our study can be related to the f o F 2 ratio between peak and trough locations of wave 4, used by Pedatella and Forbes [2009] as a proxy for wave 4 amplitudes. Assuming a wave 4 amplitude of b fo F 2 and a much larger zonal mean value of f o F 2,thef o F 2 ratio can be expressed as (f o F 2 ) peak = f of 2 + fd o F 2 f 1+ d o F 2 (3) (f o F 2 ) trough f o F 2 fd o F 2 f o F 2 [40] This is essentially the same as the relative amplitudes shown here, plus a constant offset. We may therefore surmise that our relative amplitudes will show a similar long-term trend to the f o F 2 ratios utilized in the aforementioned study, with the difference being that our study explicitly separates the wave 4 (and wave 3) into its constituent tidal and SPW signatures. In the case of wave 4, Pedatella and Forbes [2009] found that the f o F 2 ratios were anticorrelated to F10.7. [41] Our results indicate that this inverse relation between relative amplitudes and F10.7 also holds for the two dominant constituent signatures of wave 4: namely DE3 and SPW4. Both signatures show reduced relative amplitudes during the increasing phase of solar activity in , compared to those during the time leading up to the extended solar minimum at the end of 2008, which may be related to increased thermosphere molecular dissipation during high solar activity [Oberheide et al., 2009]. Additionally, the zonal mean TEC shows features associated with the seasonal and solar cycle variation of photoionization and will therefore increase with the solar cycle, thus also contributing to the decrease in relative amplitudes. Nonetheless, there is still some degree of variability within the increasing and decreasing phases that cannot be directly related to the solar cycle, such as the stronger DE3 relative amplitudes in 2011 compared to [42] However, this inverse relation to the solar cycle is not apparent in the case of the dominant constituent signatures of wave 3, suggesting the influence of other mechanisms, such as differences in source or background middle atmospheric or ionospheric conditions. For example, the wave 3-related signatures tend to occur around the solstice months when zonal mean TECs are smaller, and the relative amplitudes would thus be reduced less by normalization compared to the wave 4 related components, which tend to occur around August November when zonal mean TECs are large. Variability in the tropospheric latent heat sources responsible for forcing the constituent tides of both wave 4 and wave 3 due to lower atmospheric climatological oscillations such as ENSO also cannot be ruled out, although observations of 6663 DE3 in the neutral MLT from 2002 to 2008 using SABER and TIDI by Oberheide et al. [2009] found QBO-related changes to be more prominent. 5. Conclusions [43] From the above COSMIC results, it can be seen that the eastward semidiurnal (SE) nonmigrating tidal signatures are consistently less important in producing the low-latitude ionospheric wave 3 and wave 4 variation compared to the diurnal eastward (DE) and SPW components, agreeing with the recent numerical experiments of Pedatella et al. [2012]. We have also identified recurrent tidal and SPW signatures that could be generated via aliasing between the DE components with the DW1 migrating diurnal ionospheric variability, resulting in the manifestation of the SPW components. In the case of SPW3, higher-order interactions with a DW1-related signal are also possible, generating DW4, which is of secondary importance in the composition of wave 3 variation. The overall zonal variation of the EIA regions has also been examined and is found to be relatively consistent from year to year, indicating that the relative phases of the constituent tidal and PW signatures is interannually recurrent. This is indicative of interannual consistency in both tidal and SPW sources, as well as the physical mechanisms responsible for mapping the signatures of the MLT oscillations into the ionosphere. However, the wave 4 component appears to provide a greater contribution to the overall zonal EIA variation compared to wave 3 during the time periods and local time examined. Absolute amplitudes of tidal and PW signatures are found to increase with the solar cycle for both wave 4- and wave 3-related signatures. While the relative amplitudes of the wave 4-related signatures tend to decrease with the solar cycle through 2010 as previously expected, a similar relation is not found in the case of the wave 3-related signatures. [44] The extended 5 year data coverage presented here provides additional confidence in the robustness of our results, as well as evidence that aliasing between globalscale oscillations in the neutral atmosphere with diurnal ionospheric variability may be a real and recurrent feature. However, the conditions that govern this mapping of global-scale oscillations from the neutral atmosphere into the ionosphere still remain to be resolved. The causes for the particularly broad response of SPW4 compared to DE3, as well as the difference in hemispheric asymmetry of SPW4 and DE3, require further examination through future modeling studies. Appendix A: COSMIC Sampling [45] The space-time sampling of the COSMIC data is quantified as follows: Data points at each latitude bin within the 40 day averaging window are binned into overlapping longitude bins spaced every 45 ı. If fully filled, this longitude grid is sufficient to resolve zonal wave numbers up to 4. To better illustrate changes in COSMIC sampling over time, we compute two UT binning cases in conjunction with the aforementioned longitude grid: a higher resolution version with 24 overlapping bins spaced by 1 h, and a lower resolution version with 6 overlapping bins spaced by 4 h. Fully filled, the first case with the 1 h grid is capable of

14 Figure A1. Percentage of missing longitude/ut bins at each magnetic latitude from COSMIC observations during the entire period from 2007 to 2011, using a 40 day averaging window and 45 ı longitude bins. UT bins of (a) 1 h, (b) 4 h. White cells denote full longitude/time sampling. resolving periods as short as 2 h, while the second case with the 4 h grid can resolve periods as short as 8 h. As the shortest period fitted in our study is 8 h, we note that the 1 h grid in the first case is oversampled in time, while the 4 h grid is the minimum sampling requirement that must be fulfilled for unambiguous separation of the tidal and SPW signatures examined in this study. [46] Figure A1a shows the percentage of longitude/ut bins at each magnetic latitude that are left unfilled during the entire period from 2007 to 2011, using the oversampled 1 h grid. It can be seen that from an equatorward of about 60 ı, full longitude/ut sampling is available during most of this time period, allowing for unambiguous tidal and SPW separation. Coverage is somewhat reduced during late 2010 and 2011, due to battery degradation aboard two of the COSMIC microsatellites during this time but still results in sampling of about 90% of the longitude/ut bins during most of this time frame. Comparing with the results of the 4 h grid in Figure A1b, it can be seen that despite the reduced number of observations, full sampling with regard to the basis functions utilized in this study is still possible in an equatorward of 60 ı magnetic latitude during the entire period between 2007 and [47] Considering the above sampling analysis, the simultaneous fitting of all the tidal and SPW components employed in our study, as well as the estimated uncertainties, we conclude that the selected 40 day window provides sufficient coverage for the purposes of this study. [48] Acknowledgments. This research was supported by grant NSC M MY2 from the National Science Council of Taiwan. Additional support for L.C. Chang was provided by startup funding from the NCU Institute of Space Science. We are also grateful to the Taiwan National Space Organization (NSPO) and the University Corporation for Atmospheric Research (UCAR) in the U.S. for provision of FORMOSAT- 3/COSMIC data. We thank the reviewers for their helpful suggestions. [49] Robert Lysak thanks the reviewers for their assistance in evaluating this paper. References Cheng, C.-Z. F., Y.-H. Kuo, R. A. Anthes, and L. Wu (2006), Satellite constellation monitors global and space weather, Eos Trans. AGU, 87(17), 166, doi: /2006eo Chang, L. C., C.-H. Lin, J.-Y. Liu, N. Balan, J. Yue, and J.-T. Lin (2013), Seasonal and local time variation of ionospheric migrating tides in FORMOSAT-3/COSMIC and TIE-GCM total electron content, J. Geophys. Res. Space Physics, 118, , doi: /jgra England, S. L., T. J. Immel, E. Sagawa, S. B. Henderson, M. E. Hagan, S. B. Mende, H. U. Frey, C. M. Swenson, and L. J. Paxton (2006), 6664

15 Effect of atmospheric tides on the morphology of the quiet time, postsunset equatorial ionospheric anomaly, J. Geophys. Res., 111, A10S19, doi: /2006ja England, S. L., T. J. Immel, J. D. Huba, M. E. Hagan, A. Maute, and R. DeMajistre (2010), Modeling of multiple effects of atmospheric tides on the ionosphere: An examination of possible coupling mechanisms responsible for the longitudinal structure of the equatorial ionosphere, J. Geophys. Res., 115, A05308, doi: /2009ja Fang, T.-W., H. Kil, G. Millward, A. D. Richmond, J.-Y. Liu, and S.-J. Oh (2009), Causal link of the wave-4 structures in plasma density and vertical plasma drift in the low-latitude ionosphere, J. Geophys. Res., 114, A10315, doi: /2009ja Forbes, J. M., J. Russell, S. Miyahara, X. Zhang, S. Palo, M. Mlynczak, C. J. Mertens, and M. E. Hagan (2006), Tropospherethermosphere tidal coupling as measured by the SABER instrument on TIMED during July September 2002, J. Geophys. Res., 111, A10S06, doi: /2005ja Forbes, J. M., X. Zhang, S. Palo, J. Russell, C. J. Mertens, and M. Mlynczak (2008), Tidal variability in the ionospheric dynamo region, J. Geophys. Res., 113, A02310, doi: /2007ja Hagan, M. E., R. G. Roble, and J. Hackney (2001), Migrating thermospheric tides, J. Geophys. Res., 106(A7), 12,739 12,752. Hagan, M. E., A. Maute, R. G. Roble, A. D. Richmond, T. J. Immel, and S. L. England (2007), Connections between deep tropical clouds and the Earth s ionosphere, Geophys. Res. Lett., 34, L20109, doi: / 2007GL Hagan, M. E., A. Maute, and R. G. Roble (2009), Tropospheric tidal effects on the middle and upper atmosphere, J. Geophys. Res., 114, A01302, doi: /2008ja Immel, T. J., E. Sagawa, S. L. England, S. B. Henderson, M. E. Hagan, S. B. Mende, H. U. Frey, C. M. Swenson, and L. J. Paxton (2006), Control of equatorial ionospheric morphology by atmospheric tides, Geophys. Res. Lett., 33, L15108, doi: /2006gl Lin,C.H.,J.Y.Liu,T.W.Fang,P.Y.Chang,H.F.Tsai,C.H.Chen,and C. C. Hsiao (2007), Motions of the equatorial ionization anomaly crests imaged by FORMOSAT-3/COSMIC, Geophys. Res. Lett., 34, L19101, doi: /2007gl Liu, L., W. Wan, B. Ning, O. M. Pirog, and V. I. Kurkin (2006), Solar activity variations of the ionospheric peak electron density, J. Geophys. Res., 111, A08304, doi: /2006ja Liu, H., S. V. Thampi, and M. Yamamoto (2010), Phase reversal of the diurnal cycle in the midlatitude ionosphere, J. Geophys. Res., 115, A01305, doi: /2009ja Luan, X., X. Dou, J. Lei, and G. Jiang (2012), Terdiurnal migrating-tide signature in ionospheric total electron content, J. Geophys. Res., 117, A11302, doi: /2012ja Maute, A., A. D. Richmond, and R. G. Roble (2012), Sources of lowlatitude ionospheric E B drifts and their variability, J. Geophys. Res., 117, A06312, doi: /2011ja Mukhtarov, P., and D. Pancheva (2011), Global ionospheric response to nonmigrating DE3 and DE2 tides forced from below, J. Geophys. Res., 116, A05323, doi: /2010ja Oberheide, J., J. M. Forbes, K. Hausler, Q. Wu, and S. L. Bruinsma (2009), Tropospheric tides from 80 to 400 km: Propagation, interannual variability, and solar cycle effects, J. Geophys. Res., 114, D00I05, doi: /2009jd Oberheide, J., J. M. Forbes, X. Zhang, and S. L. Bruinsma (2011), Wavedriven variability in the ionosphere-thermosphere-mesosphere system from TIMED observations: What contributes to the wave 4?, J. Geophys. Res., 116, A01306, doi: /2010ja Palo, S. E., J. M. Forbes, X. Zhang, J. M. Russell III, and M. G. Mlynczak (2007), An eastward propagating two-day wave: Evidence for nonlinear planetary wave and tidal coupling in the mesosphere and lower thermosphere, Geophys. Res. Lett., 34, L07807, doi: / 2006GL Pancheva, D., and P. Mukhtarov (2010), Strong evidence for the tidal control on the longitudinal structure of the ionospheric F-region, Geophys. Res. Lett., 37, L14105, doi: /2010gl Pancheva, D., and P. Mukhtarov (2012), Global response of the ionosphere to atmospheric tides forced from below: Recent progress based on satellite measurements, Space Sci. Rev., 161, 1 35, doi: / s Pedatella, N. M., and J. M. Forbes (2009), Interannual variability in the longitudinal structure of the low-latitude ionosphere due to the El NiQno Southern Oscillation, J. Geophys. Res., 114, A12316, doi: /2009ja Pedatella, N. M., M. E. Hagan, and A. I. Maute (2012), The comparative importance of DE3, SE2, and SPW4 on the generation of wavenumber- 4 longitude structures in the low-latitude ionosphere during September equinox, Geophys. Res. Lett., 39, L19108, doi: /2012gl Rishbeth, H. (1998), How the thermospheric circulation affects the ionospheric F2-layer, J. Atmos. Sol-Terr. Phys., 59, Rishbeth, H., I. C. F. Mller-Wodarg, L. Zou, T. J. Fuller-Rowell, G. H. Millward, R. J. Moffett, D. W. Idenden, and A. D. Aylward (2000), Annual and semiannual variations in the ionospheric F2-layer: II. Physical discussion, Ann. Geophys., 18, , doi: /s Sagawa, E., T. J. Immel, H. U. Frey, and S. B. Mende (2005), Longitudinal structure of the equatorial anomaly in the nighttime ionosphere observed by IMAGE/FUV, J. Geophys. Res., 110, A11302, doi: /2004ja Teitelbaum, H., and F. Vial (1991), On tidal variability induced by nonlinear interaction with planetary waves, J. Geophys. Res., 96, 14,169 14,178. Wu, D. L., P. B. Hays, and W. R. Skinner (1995), A least squares method for spectral analysis of space-time series, J. Atmos. Sci., 52, Yue, J., W. Wang, A. D. Richmond, H.-L. Liu, and L. C. Chang (2013), Wavenumber broadening of the quasi 2 day planetary wave in the ionosphere, J. Geophys. Res. Space Physics, 118, , doi: /jgra