Spectral Behavior of VHF Backscatter From Ionospheric Sporadic E Irregularities in the Equatorial Anomaly Crest Zone

In this article, the spectral behavior of the radar returns from iono­ spheric irregularities of the sporadic E layer occurring in equat.orial anomaly crest zone are investigated using 52 MHz Chung-Li \!THF radar. It is shown that the mean Doppler velocit)· and spectral width for the type 1 radar spectra are within 28-35 mis and 3-5 Hz, respectively. Compared with the observational results made "rith other VHF radars, the Doppler shifts of type 1 spectra obtained here are significant.I}· smaller than those in the equa­ torial and auroral zones. On the other hand, the mean Doppler velocities of the type 2 radar spectra \Vith the spectral width of 16-30 Hz are height­ dependent and have values betw·een -50-60 m/s for the present case. The correlation bet\iveen the Doppler spectral parameters of the type 2 irregu­ larities, namely, the echo intensity., mean Doppler shift, and the spectral width, are also analyzed in this paper. It is demonstrated that a high corre­ lation exists bet\\7een echo intensity and spectral width, while a consider­ ably poor correlation prevails bet\\1een mean Doppler frequency shift and spectral \\1idth. These results indicate that the primary mechanism causing the breadth of the Doppler spectrum of type 2 irregularities is the iono­ spheric electron density fluctuations, not the beam broadening effect. The a11alysis shows that there is a good correlation bet"1een mean Doppler shift of type 2 radar spectra and the slope of echo contour in the range-time­ intensit)1 plot, implying that the Es targets responsible for type 2 echoes are localized and discrete in the resolution volume. The comparison between the echo intensit)' of the type 2 radar spectra and f oEs and fbEs is also made in this article, \\rhere f oEs and fbEs scaled from the trace of a HF ionogram record are the critical frequency and blanket frequency of the sporadic E layer, respectively. This shows that the VHF peak power of the t)1pe 2 radar spectra i s proportional to the difference between f oEs and fbEs. A detailed discussion on this feature is given in the text.

type 1 spectra obtained here are significant. I}· smaller than those in the equa torial and auroral zones. On the other hand, the mean Doppler velocities of the type 2 radar spectra \Vith the spectral width of 16-30 Hz are height dependent and have values betw·een -50-60 m/s for the present case. The correlation bet\iveen the Doppler spectral parameters of the type 2 irregu larities, namely, the echo intensity., mean Doppler shift, and the spectral width, are also analyzed in this paper. It is demonstrated that a high corre lation exists bet\\7een echo intensity and spectral width, while a consider ably poor correlation prevails bet\\1een mean Doppler frequency shift and spectral \\1idth. These results indicate that the primary mechanism causing the breadth of the Doppler spectrum of type 2 irregularities is the iono spheric electron density fluctuations, not the beam broadening effect. The a11alysis shows that there is a good correlation bet"1een mean Doppler shift of type 2 radar spectra and the slope of echo contour in the range-time intensit)1 plot, implying that the Es targets responsible for type 2 echoes are localized and discrete in the resolution volume. The comparison between the echo intensit)' of the type 2 radar spectra and f oEs and fbEs is also made in this article, \\rhere f oEs and fbEs scaled from the trace of a HF ionogram record are the critical frequency and blanket frequency of the sporadic E layer, respectively. This shows that the VHF peak power of the t)1pe 2 radar spectra is proportional to the difference between f oEs and fbEs. A detailed discussion on this feature is given in the text.

INTRODUCTION
For many years, several scientific workers have used radars operating at 20 MHz or above to observ'e sporadic E irregularities (e.g., Miller and Smith, 1978;Ecklund et al., 1981�Tanaka and Venkateswaran, 1982a, 1982Riggin et al., 1986;Moorcraft and Schlegel, 1990;Fukao et al.� 1991;Yamamoto et lll., 1991Yamamoto et lll., , 1992. Based on these observations, adv'anced kno\vledge of the dynamics and morphology of sporadic E irregularities has been achiev'ed. However, on examination of the geomagnetic locations of the radars in operation used for the. obse. rvations ot· ionospheric irregularities, it is found that no radar is located in the region of 0° -28° geo magnetic latitude. It should be noted that this area is just the zone in \\lhich the well-known feature ot' equatorial anomal)' occurs. Therefore, in order to improve the understanding of sporadic E irregularities from the global point of vie\\t·, the observation made by the radar located in the equatorial anomalous zone, such as the Chung-Li VHF radar, seems to be required.
It has been known for )''ears that on the basis of the Doppler spe. ct1�a1 characteristics, the VHF coherent radar returns from the ionosphe. ric sporadic E irregularities in equatorial and mid-latitude areas can be categorized into two major groups, namely type 1 and type 2 echoes (e.g., Kelly, 1989). Type 1 echoes are generally associated with extremely narrow spectra (only a fe\\1 Hz) with a mean Doppler shit' t close to the phase velocity of the ion-acoustic plasma wave (around 360 mis). On the other hand, the breadths of the type 2 radar spectra are. much larger than those of the type 1 spectra, and their mean Doppler shifts are fairly smaller than the. ion-acoustic plasma wave speed. The mechanisms responsible for t)1pe 1 and type 2 irregularities occ.urring in the mid-latitude area are different. For a VHF radar at 50 MHz, type 1 irregularities are thought to be attributed to linearly unstable 3-meter gradient drift plasma waves generated b)' extremely sharp electron density gradients associated with sporadic. E layer. However, t)'pe 2 irregularities are considered to be due to the gradient drift plasma instability occurring on the bottom side of the sporadic E layer \Vith a sharp electron density gradient. This sort of plasma instability excites large scale primary plasma wave.s (at the wa\1e length of several tens of meters) first, and small scale secondary plasma wa\1es are. produced subsequently through the effect of non-linear interaction.
On the basis of the obser\1ational results of sporadic E (Es) irregularities by using the Chung-Li VHF radar (Chen, 1993), the VHF Es ec. hoes primarily appear in the height range of I 00-120 km apd center at the heig.ht of around 110 km. The duration of the. e -choes ranges from about a few minutes to more than 6 hours with a mean duration of about 45 minutes. Gener ally, the sporadic E i1·regularities occur in the dark hou1�s. Analysis shows that they appear more frequent1y in the period of pre-midnight than in the period of post-midnight, \V'hich is quite different from the observations made with MU radar. Occasionally, in the season of' June. and July, very few VHF Es echoes can be detected before sunset. The mean thickness of Es echoes presented in the height-time-intensity plot is about 20 km with a variance of 7 .5 km.
In this article, the observational results of ionospheric field-aligned sporadic E irregulari ties made by the Chung-Li VHF radar are presented and discussed. In Section 2, the character istics of the Chung-Li VHF radar for the ionospheric observation is introduced. In Section 3, the correlation analyses among echo pO\\i'er, mean Doppler shift and spectral \Vidth of the type Chu et al.

363
2 radar spectra is made. The comparison of the. echo intensity of type 2 irregularities with the dif' t' erence between the critical frequency and blanke. t frequency of the sporadic E lay'er scaled t' rom an ionogram trace is also given in this section. The conclusion is given in Section 4, ,;v·here it is sho\.\1n that the consecutive plume-like structures associated with positive radial Doppler velocitie . s (i.e., away from antenna) r·anging f' rom 30-55 m/s are obser\red. Moreo\1er, through the spectral analysis of' radar returns with fine. temporal resolution, two types of radar spectra, namely, type 1 and type ·2, are found� on some occasions, to appear simultaneousl)' in the same height range. Different f' rom pre,rious observational results (Fejer and Kelley, 1980; Kelley, l 989), the mean Doppler shifts of the type 1 radar spectra observed here are onl)' 32-38 Hz, somewhat smaller than those obtained in othe.r regions. The plausible mechanisms which cause such a small Doppler shift of type 1 spectra will be disct1ssed in the text ..

CHUNG-LI \7HF RADAR A�D ECHO SIGNAL ANALYSIS
The Chung-Li VHF radar is located on the campus ot' National Ce. ntral Uni\1ersity of Taiwan (R.O.C.) with geographic and geomagnetic latitudes ot' 24.9' . N and 14 CN , respec tiv-ely. The Chung-Li VHF radar was originally designed as an ST radar. Accordingly, the original antenna beam can only be ste. ered toward zenith and north, east, south and west at 17° off-zenith angle such that the 3-dimensional wind field and other important f' eatures� such as aspect sensitiv·it)7, tropopause, wind she . ar� precipitation, atmospheric stability·, stable layer thickness, lightning, grav·it)' w·ave activity and momentum flux in the troposphe. re and strato sphere can be obsenled. In vie\\/ of the t' act that the _ peak of the equatc)rial anomaly generall)1 appears in the range ot' the geomagnetic latitude of 15° to 20° (Rishbeth and Garriott, 1969), it is obvil)US that the Chung-Li VHF radar is just located �,ithin thhe . so-called equatorial anoma lous crest zt)ne. To the best of the authors' knowle. dge, this radar se. ems to be the only VHF radar in operation \vi thin the region of' ionospheric equatorial anomaly· in the \Vorld. The capa bility of ionospheric observation on the Chung-Li VHF radar has be . en implemented since the spring of 1992. A new antenna array composed of three rectangular subarrays with 4x8 Yagi elements each was constructed just adjacent to the old one. The half' power beam width of' each antenna subarray is 7.5° in elev·ation and 15° in azimuth direction. The radar beam is pointed toward the geographic northwest by 17° with a fixed zenith angle . of 41° such that it is able to be perpendicular to the geomagnetic field line at the altit· ude ot� 250 km. The. \7ery detailed characteristics of the Chung-Li VHF radar for ionosphe1·ic observation can be ret' erred to in Chen ( 1993) but is summarized in Table 1.
Because the Chung-Li VHF radar is adjacent to the Taipei International Airport, the radar returns coming from aircrafts may seriousl)1 contaminate the wanted radar echo signals on occasion. Furthermore, except for the type 1 and type 2 echoes from Es irregularities, the meteor trails may also produce . sufficiently strong VHF backscatter. In light of' the complexity of the observed Doppler spectrum, the identif� ication of the type I and type 2 radar spectra can onl), be done manual!)'. Once they are separated in the Doppler spectral domain, the echo po\ver, mean Doppler frequency shift and spectral width f' or these two spectra are estimated individually by the least square method, in which the Gaussian cur\te is employed to best fit the corresponding Dopple. r spectral component. , \,lol. 7, N(J. 3, September 1996  In general, the Chung-Li VHF radar paramete1�s emplo)1ed f' or the observation of the spo radic E irregularities are set as follo\\1s: a radar peak transmitted p(1wer of 55 kw, an inter-pulse pe. riod of 1.5 ms, a pulse width ot' 28 µ s with a 7-bits Barker cc)de (corresponding to range resolution of 600 m), a recei\'er bandwidth of 250 KHz so as to maximize the signal-to-noise ratio, a coherent integration of 2 or 4 times, a radar probing range gene. rall)' from 124 km to 160 km, and 60 range. gates are set for each experiment. It shpuld be noted that the . apex of the radar beam is steere. d toward 17° \�lest of the ge. ographic. north with a fixed zenith angle of 4 I 0 such that the radar beam can be normal to the geomagnetic field line at the height of 250 km. In view· of the field-aligned property (Ecklaund et ell., ] 98 I� Moorcraft and Schlegel, I 990), the backscatter of the sporadic E irregularities can only be generated from the direction per pendic. ular to the field line. at the height where irregularities are situated, i.e., around 110 km. Under these conditions� based on the calculation of the International Ge. omagnetic Field Model ' � (IGRF85), the zenith ang1e w·hich is responsible for the radar returns ot .. E region irregularities 0\1er the Chung-Li VHF radar site are around 38.5°, other than 41°. Theret . . ore, the height range that C()rresponds to the slant range from 124 km to 160 km CO\lers between 97. 7 km and 126.1 km. The 128-points FFT algorithm is usually emplo)1ed to compute the Doppler spectrum of Es radar e . cho signals.

OBSERVATIONS AND DISCUSSION
3.1 Type 2 Radar Spectra Figure 1 a presents the range-time-intensity plot of the radar backscatter from Es irregu larities during the period from 02:58: 13L T to 03: 11 :52L To n January 26, 1995. It should be pointed out that the <:1ltitude ()f the first range gate is about 90 km, w·hile the height increment corresponding to each range gate is 0.47 km. The time res(1lution f' or each data point is 1.536 sec(1nds. Because it takes time to transmit the received 1·adar echoes f1·om the master computer to the hard disk, the gaps separating the radar data exist in the plot at a constant rate. As indicated, the Es echoes primarily occur in the height 1·!1nge from about 95 km to 112 km t� or this case. It is note\\i'orth)' that several intense mete()1· ecl1c. 1es characterized by short duration (usually less than 0.5 sec) and random distribution i11 the height range t' rom 80 km to 120 km can be seen in the contour plot. These echoes ma)' complicate the observed Doppler spectra and should be remO\led in analyzing the normal Es echoes. Figure 1 b displays the height distri buti(lll of the Doppler spe. ctra f()r the radar returns in the period from 03:03:27LT to 03:06: 04L T (i.e., the time point from 200 to 305), as presented in Figure J a. As shown in the plot, the fairly broad Doppler spectra with breadth ranging from 5 Hz to 50 Hz distribute from range gates 12 to 53, corresponding to the height range from 95.67 km to 115.05 km. This kind of broad Dopp1er spectra with a height-varying mean Doppler shift is cate. gc)rized as type 2 radar spec t1·a. The echo intensity, Mean Doppler shift and spectral width of type 2 echoes can be ob tained by using the least square method, in which the Gaussian curve is employed to best fit to the observed type 2 spectrum. In the following, the statistical analyse. s among these three spectral moments is gi"·en so that the mechanism broadening the type 2 radar spectra can be examined. Figure 2 presents the scatter diagram of the spectral \:\/idth ve. rsus the mean Doppler shift at· the t:y·pe 2 spectra observed on July 15, 1994 from 00:12:47 LT to 05:10:44 LT. It is clear that no C()rrelation bet\veen these t"''O spectral parameters can be seen. This indicates that the beam broadening eff" ect, which arise. s as a result of the irregularities drifting across the rela tively· broad antenna beam, on the type 2 spectral width is so small that it can be negle. cted . It is \\10rth pointing out that basically the Es irregularities are field-aligned. This characte. ristics implies that the aspect sensiti-v ·ity of this kind of target is so high that the radar returns can only be observed in the direction of normal to the magnetic field line. This explains why the beam broadening effect is negligible in analyz .
ing the prope . rt:y· of the spectral width of the type 2 echoes. Figure 3 is the scatter diagram of signal power (in dB units) versus the spectral width of"' t),rpe � radar spectra. Ob\i'iously, a good positive correlation between the. se two spectral parameters is seen from this diagram. This feature arises due to the ±' act that, from the sc. attering theory, the ech() intensity ot' the ionospheric irregularities is proportional to the variance (lf electron density fluctuations. In addition, as mentioned above, the beam broad ening et' t'ect plays an insignit' icant role in contributi11g to the width of· the type 2 radar spectra due to the extreme narro\\iness of aspect sensiti\1it)1• These results suggest that the most impor tant t' actor in broadening the t)'pe 2 radar spectra is the random variations of the ionospheric elec. tron density· (lcc. urring in the sporadic E layer associated with plasma instability. If the 11otion as to the t" ormation of the type 2 spectral width is correct, a good correlation between the VHF radar backscatte11 from Es irregularities, which is closely related to the variance of electron density fluctuations, and the. difference between t' oEs and tbEs scaled from the trace of the sporadic E layer rec. orded in a conventional HF ionogram is anticipated. This is because, in a general sense� foEs-fbEs re. presents the large scale \,ariation of the electron density imbed ded in the sporadic E la)'er. Figure 4 presents the scatter diagram of the peak signal-to-noise r<:ltio ot' the Chung-Li VHF radar returns fro1n type 2 irregularities versus l\ Es and is defined as: A E == f oEs -tbEs ti s foEs (1) where foEs and tbEs are the critical frequency and ' blanket t' requency· of· the sporadic E laye. r, respective.1) 1, and where both are scale. d t' rom the trace of the sporadic E layer in the. ionogram obtained by using the Chung-Li HF ionosonde located about 5 km northwest from the Chung Li VHF radar station. The VHF and HF radar data employed in Figure 4 were taken in the period bet\veen April 5 and Jul)' 29 t)f 1992 . As shown in Figure 4, it is ob\rious that the VHF peak signal-to-n(lise ratio of ty·pe 2 irregularities is proportional to the difference between foEs and tbEs. It is noteworthy that the targets embedded in the sporadic E layer responsible f' or VHF and HF radar returns are quite different. The former are the 3-meter electron density i11regularitie. s, while the latter are the dense patches of e . lectron density in the sporadic E layer with the scale of a few hundreds meters. Ne. vertheless, the common property' of these t\\'O targets is the inh()ffi()geneity of� the electron density. On the basis of the definition of foEs and tbEs, it is e. vident that the. physical meaning of 8Es can be regarded as a measure of the degree ot' large scale inhomogeneity of the ele. ctron density in the sporadic E layer. From the theory ot� the echo mechanism of \'HF radar, it seems c. Jear that the VHF radar echo intensit)' from type 2 irregularities is pr()p{>rti()nal to the small scale rand()ffi fluctuations of electr()n density associated \\tith plasma turbL1lence. Since the v·ariations ()f electr()Il densit)' on the small scale can be generated from those on the large scale through the non-linear plasn1a turbulent cascade pr()Cess, it is expected that the �Es of' HF echoes are c]c)sel)' C(Jnnected to the signal-to-noise ratio at· VHF radar retur11s, as shown in Figure 4. The detailed examining of Figures 1 a and J b shows that the sign of=" the mean Doppler shit· t of type 2 echoes is strongly correlated with the direction of· the drift of type 2 irre. gL1lari ties. If the irregularities ffi()Ve upward \\1ith time, namely with a positive slope in the range time-intensity plot, the mean Doppler shift is negative, and vice versa. Figure 5 sho\�is the scatter diagram of the mean Doppler velocit)' of t)1pe 2 echoes versus the range slope ( ()f rate) 369 ()f the t)1pe 2 echoes scaled trom the range-time-intensity· pl(1t. It is necessary to point out that the radar echoes used t' or the determination ()f' the 1�ange slope are the selected echo patterns with striated structures in the range-time-intensit)' plot. One example ()f' such echo patterns is presented in Figure 1 a marked vvith 'A'. The straight line presented in Figure 5 is at the slope of 1. It is clear that the C()rrelation between the range rate and the mean Doppler shift of type 2 echoes is fairly good and the slope t)f the regression line is less than I, indicating that the t' o1,mer is slightly larger than the latter. This result implies that this kind of� target is responsible t' or the type 2 echoes localized and distributed inhomogeneously in the resolution volume, a finding \vhich is quite consistent \\ti th the field-aligned property' of the ionospheric irregulari-• ties.

Type 1 Radar Spectra
As 1nentioned earlier, the radar backscatter ft�()ffi type 1 irregularities are characterized b)7 an e . xtremely narro\v spectral width. Usually, type l echoes coexist with type 2 echoes in the receiv·ed radar returns. In fact� it is impossible to distinguish these two echoes in the. range time-intensity plot. However� it is easy· to identif' y types 1 and 2 echoes in the Doppler spectral dt)main. Figure 6a presents an example of the contour plot of Es radar returns in which type 1 echoes occur in combinati()fl with those of ty·pe 2. Figure 6b shows the corresponding Doppler spectral profile in the period t� rom 20:24:29L T to 20:25:46LT. It is clear that in the range gates from 9 to 19, pront1unced type 1 spectra with a constant mean Doppler shift of about 10 Hz and pectra with a mean Doppler shit' t of 10 Hz and a spectral width of 20 Hz are observed. It is to be Il()ted that according to . the previous 50 MHz radar C)bservations made in the geomagnetic equatorial and auroral z.ones (e.g., Kelle) ', 1989), the mean Doppler shift of type. 1 radar :00.00 -tJ.00 --t TAO, V(J/. 7, No. 3, September 1996. spectra for sporadic E irregularities is about 12() ± 2() Hz (i.e., corresponding to 360 ± 60 mis radial velocity) if the ion acoustic \Vave is responsible f" or that. By taking the phase speed of the ion acoustic wave and the relative large Nyquist frequency into account, it seems that the aliasing problem on the observed type 1 spectral peak cannot occur. Therefore, it seems that the type I radar spectra obtained here may not be explained by using the ion acoustic wave \Vhich is responsible f()f the type 1 spectra obser,7ed by 50 MHz radar. However, as reported by Riggin et al. ( 1986 ), the averaged Doppler velocity of the type 1 radar spect1·a observed with 50 MHz CUPRI radar located close to the Arecibo Observatory is about 150 mis, which is fairly smaller than the v a lue� reporte. d above. In order tc) explain the enormously small Doppler frequency shif' t <)f the type 1 spectra caused by the ion acoustic wave� the conditions of heavy ion mass and the presence of an extremely sharp electron density gradient of� the sporadic E layer sh()Uld be) th be taken into account. Under these considerations, the threshold velocityr of type 1 irregularities could be reached as low as one-fourth (even smaller) of the ion acoustic \\lave speed (Farley and Fejer, 1975;Riggin et al., 1986). Conse. quently, the possibil ity· that the type I radar spectra shown in Fig.6b is caused by ion acoustic waves cannot be r111.ed out.

Multiple Peaks of Es Radar Spectra
Based on the experience oh these authors, the feature of' multiple spectral peaks of Es radar returns oc. curs very frequently. The effects causing the bifurcation of observed Es Dop pler spectra can be attributed to the inte1·ference signals fro . m other radio sources and multi targets coexisting in the scattering ,rolume. The interference signals can be easily identified either in a range-time-intensity plc .

CONCLUSION
The capability of · the observati(1ns of sp()rc. 1dic E i� rregularitie.s \Vere implemented at the Chung-Li VHF radar in the spri11g (1f. 1992. Thi .s was succe. ssful in detecting type 1 and ty,pe 2 irregularities in tl1e sporadic E lay1er. The ()bservational evidence shows that the mean Doppler shift of' the type I radar spectra is enormously smaller than that ()bser\led in the equatorial and auroral regions, indic. ating that it is necessar)1 to develop a nl()fe appropriate plasma instability theory· t() illust1�ate the spectral behavi()f of' the ionospheric irregularities occurring in the. equa t()fial anomalous crest regi()n. The a\1erage Doppler mean \'elocit)1 and spectral \vidth ot· the t)lpe 2 i4adar spectra p14esented in this article are, in general, bet\\?Cen -50 mis -6() in/s �lnd 16 -30 Hz, respective. ly, \\: 'hich is CC)nsistent \Vi th the observatio11s made by the v·HF radars located in other geomagnetic latitudes. The good correlation between spec. tral width and echo p()\Ver of type 2 spectra al()ng \\1ith the lack of' C()r1·elation ot· the rnean Doppler shif' t and the spectral width suggest that the beam br()ctdening et· t-ect in contributing to the spectral width ()f' t)rpe 2 i4adar echoes is 11egligible.