Journal of Amospheric and Terrestrial Physm, Vol. 50, Nos 4/5, pp. 2X9-299, 1988. OOZI-9169/88 S3.lM-t .oO Printed in Great Britain. Pergamon Press plc

Incoherent scatter spectral measurements of the summertime high-latitude D-region with the EISCAT UHF radar

E. TURUNEN

Sodankylg Geophysical Observatory, SF-99600 Sodankyll Finland

P. N. COLLIS and T. TURUNEN

EISCAT Scientific Association, Box 812, S-981 28 Kiruna, Sweden

(Receivedfor publication 8 December 1987)

Abstract-Measurements of incoherent scatter spectra from the aurora1 D-region were obtained during the summer of 1985 using a sophisticated pulse-to-pulse correlation technique with the EISCAT UHF radar. The spectral width variations with altitude are interpreted in terms of ion-neutral collision frequency, neutral temperature, mean positive ion mass and negative ion number density. Close agreement with predictions of currently available atmospheric models is obtained, except for a narrow layer around 86 km altitude. This layer showed evidence of increased positive ion mass for most of the experiment, and for short intervals indicated a mean ion mass close to 200 a.m.u. It is suggested that the layer is composed of proton hydrates in the vicinity of a structured noctilucent cloud, and that the index of hydration is occasionally large

INTRODUCTION

The ability of the EISCAT radar to provide high resolution measurements of electron densities in the ionospheric D-region is now well established and a number of studies have already been reported. These include observations during aurora1 radio absorption events (RANTA et al., 1985 ; COLLIS et al., 1986a; DEVLIN et al., 1986; COLLIS and KIRKWOOD, 1987), during a polar cap absorption event (HARGREAVES et al., 1987) and a statistical study of both quiet and disturbed conditions (KIRKWOOD and COLLIS, 1987). Whilst electron density is the fundamental ionospheric parameter measured by incoherent scatter radar, the advantage of the technique is that it also allows the derivation of other quantities if measurements of the incoherent scatter spectrum are available. Until recently, however, such spectral measurements of the D-region with the EISCAT radar have been very limited.

Determination of the incoherent scatter spectrum at D-region heights is intrinsically more difficult than for higher altitudes because of the smaller target cross- section (i.e. lower densities). In addition, the higher atmospheric pressure increases the rate of collisions between ions and neutrals, leading to much longer correlation times of the medium (i.e. narrower spec- tra). Under such circumstances it is necessary to use the technique of pulse-to-pulse correlation, as has been employed for several years at Arecibo

(MATHEWS, 1976). An overview of the method, and

of some of the results obtained with the Arecibo radar, can be found in MATHEWS (1984a). At EISCAT, KOF- MAN et al. (1984) utilised the pulse-to-pulse correlation

technique for mesospheric experiments during 1982 and although it was possible to infer information on

winds, temperature and the presence of negative ions from the spectral observations, the height resolution of the measurements was rather coarse and the authors were cautious in interpreting their results.

Since those earlier experiments, however, improve- ments to the EISCAT radar system, in particular with respect to experiment design at the correlator level, now mean that experiments can be conducted in a more effective way. The design philosophy of the experiment algorithms is termed GEN-SYSTEM and

the program library specifically contains a code for D-region measurements using the pulse-to-pulse cor-

relation technique, GEN-11, and has been described

by TURUNEN (1986). Although intended mainly for VHF applications, several experiments have been car- ried out on the UHF system using this modulation.

In the present paper we report results from one of the first series of measurements using GEN-11 on the UHF system, from July 198.5. Properties of the background plasma, derived from spectral width measurements, are discussed, as is a localised feature which showed evidence of very heavy positive ions. In addition, the method of analysis is described in some

289

290 E. TURUNEN, P. N. COLLIS and T. TURUNEN

detail, since this is fundamental to the interpretation of the measurements.

EXPERlMENTAL ARRANGEMENT AND DATA

ANALYSIS TECHNIQUE

The detailed design of the pulse coding in GEN-11 is shown in fig. 4 of TURUNEN (1986), so we need only give a general description here. A double pulse scheme is employed, with the second pulse reversed in phase from the first, and with the gap varying between the two from pulse to pulse. These features are designed to reduce the effects of clutter from preceding pulses at higher altitudes and to eliminate low frequency and d.c. contamination of the wanted signals. The inter-pulse period is 2.222 ms, which is thus the lag increment of the autocorrelation functions (ACFs) computed from the returned signals. A ‘quasi-zero’ lag is included in the ACF, at a mean delay of 112 ps, and this may be used to give an estimation of electron densities. Good height resolution is achieved by a 13 bit Barker code of 7 ~LS bit length within the main 91 ~LS pulse, giving a range resolution of 1.05 km for each of the 42 gates. The first gate is centred at a range of 70 km and for the experiment described here the antenna was pointed along the local geomagnetic field direction (elevation 76.5”) resulting in a height res- olution of 1.02 km. The remote receivers at Kiruna and Sodankyla are not used in this monostatic exper- imental arrangement.

The experimental algorithm GEN-11 is designed so that the most important instrumental effects are automatically eliminated or compensated. This means that no corrections or calibration checks are needed for clutter, receiver offsets or gain variations. However, the algorithm is not able to compensate for the sidelobe effects of the Barker-coded element pulses and in critical applications this correction has to be done. This is a rather complicated process because different lags and different pulses have an individual sidelobe contribution structure. In this work the sidelobe correction was done for all the lags following the methods described by HUUSKONEK et ul. (1988).

At the very lowest altitudes the real part of the 112 ,US lag is a good approximation of the true power, but for other altitudes a correction must be made. If the spectrum can be assumed to be Lorentzian, one can use the extrapolation obtained by fitting an ex- ponential curve to the measured ACF points. This works remarkably well if the correlation time of the tar- get is at least a few times the lag increment (2.222 ms). In the upper D-region, however, this fails due to the

fact that the spectrum is not of Lorentzian form. The

remaining possibility is to use a model ACF in the zero-lag surrounding and to do the correction in this

way. This works well in the upper D-region and in the

lowest parts of the E-region. Thus in principle a ‘power profile’ can be obtained for all altitudes where the correlation time is essentially longer than 112 vs. The model may fail at the lowest altitudes covered, but there the correction is, in any case, very small and moderate inaccuracies in correction can be tolerated. The second method is used in this paper and the ionospheric model ACFs are taken from SCHLECEL (1979).

The correlation time of the target is obtained by

fitting an ACF of exponential form to the measured data points, i.e. by assuming a Lorentzian spectrum (MATHEWS, 1984b). The use of an analytical form for

the spectrum (or ACF) has the advantage that it allows direct estimation of values for the parameters affecting the shape. The Lorentzian form has been shown to be a good approximation on the basis of both theoretical work (e.g. FUKUYAMA and KOFMAN, 1980) and by measurements made at Arecibo (e.g. GANGULY, 1980; FUKUYAMA, 1981). An intrinsic limi- tation of this approach is that it neglects the effects of finite Debye length, which become important for electron densities below about 10’” mm3 (MATHEWS, 1978). A more rigorous method of analysis would

be to generate theoretical spectra for comparison with the measurements, as is normally done for measure- ments from higher altitudes. However, this is not yet implemented for the heights of interest here and we have based our analysis of these initial measurements on an analytical approach. A general expression for the incoherent scatter spectrum for a collision domi- nated plasma (appropriate to the n-region) has been given by TANENBAUM (1968)

where w is the frequency shift from centre frequency wO, N, is the electron density, re is the classical electron radius, $l is the normalised ion-neutral collision fre-

quency (= v,,,lJ2k Rv,, where vi,, is the ionneutral col-

lision frequency, k, is the radar wave number (= 47-c/l,, where I, is the radar wavelength) and v, is the mean ion thermal speed (= Jk,T/m,, where k, is Boltzmann’s constant, T is the ion and neutral tem- perature and mi is the positivejon mass) and 0, is the normahsed frequency (= w/,/2k,v,).

We measure the ACF of the plasma fluctuations in the experiment, which is the Fourier transform of the above spectrum and is given by

Spectral measurements of summertime D-region

A(A\t) = e-‘~/~~,

where t, = &k,viAt and At is the lag delay. The exponent in the expression for the ACF can be re- arranged into the form

where t,. = v,,/,~m,/32rr2k,T. Evaluating the constants and using i., appropriate to the EISCAT UHF radar (0.32 m). we obtain

t. = 2.37 x 10% m.lT s i in li ’

If negative ions are present in the plasma, the effect is to broaden the incoherent scatter spectrum (MATHEWS, 1978; FUKUYAMA and KOFMAN, 1980). Detailed treatment of this effect is complicated, especially for small electron densities, but an adequate approximation for the present study is to assume that the spectrum is widened by a factor (1 +A), where 1 is the ratio of negative ion to electron number density (MATHEWS and TA~E~BAUM, 1981). The general expression for the correlation time derived above then becomes

2.37 x 10’%,,m, [,. = --~cr-~~,--. (1)

Thus, by measuring the correlation time of the medium (the delay at which the ACF has fallen to l/e of the zero-lag value), it is possible to interpret the measurements in terms of ion-neutral collision fre-

quency, positive ion mass, neutral temperature and the ratio of number of negative ions to electrons. Since these are inter-related in the expression for correlation time, it is usual to work with assumed or model values for some of the parameters.

The Doppler velocities are so high that the Doppler shift is not insignificant when compared with the spec- tral width. This is taken into account in the analysis, and the coherence times shown in this paper are based on the Doppler corrected data. The Doppler velocity itself is measured by using the so-called ‘matched fil- ter’ method. This assumes that the Doppler shift is constant during the integration time. During the per- iod of observations reported here, a wave-like atmo- spheric motion was seen. Thus, for long integration times two separate estimates are formed. one based on direct integration and the other based on an arithmetic mean of shorter integration periods all having an indi- vidual Doppler correction. The accuracy estimates of the results are computed on the basis of the estimated variances of the individual dam points in the measured ACFs.

, J ,900 2ooQ ztw 2200 UT

Fig. 1. Geophysical activity during the experiment, illus- trated by the horizontal geomagnetic component measured at Tromso, and the radio absorption measured by riometer at Kilpisjlrvi (100 km south-east of Tromsa). Conditions were generally weakly disturbed, but the magnetic activity

increased towards the end of the experiment.

IONOSPHERIC CONDITIONS DURING THE

EXPERIMENT

The experiment was operated on the EISCAT UHF

radar at Ramfjordmoen (69.6”N, 19.2”E) on 1 July 1985 between 19.54 and 2134 UT. Geomagnetic con- ditions were still weakly disturbed following a strong disturbance the previous night. Figure 1 shows the local magnetogram for times containing the interval of the experiment, together with the riometer record

from Kilpisjlrvi (approximately 100 km south-east of Ramfjordmocn). Weak radio absorption was rec- orded by the riometer.

Bl ectron densities were estimated from the power

profiles deduced from the radar measurements as described above. The results are shown in the form of contours in Fig. 2. Two independent consistency checks were applied to the electron density results. The first entailed a comparison of the radio absorption expected from the electron density profiles with the independent measurements by riometer, using the method described by RANTA et al. (1985). The results of this comparison are shown in Fig. 3, where we also include the radio absorption measured by the riometer at Lavangsdalen, which is 20 km south of the EISCAT site and almost directly beneath the ionospheric vol- ume probed by the present experiment. The simi- larities between the two series of riometer measure- ments show that the horizontal extent of the region of absorption was rather uniform on a scale size of tens to a hundred or so kilometers, but the variations on time scales of the order of l-2 min in the absorption

292 E. TIJRUNEN, P. N. COLLIS and T. TURUNEN

calculated from the EISCAT densities may indicate more localised precipitation features.

Importantly, in the present context, the magnitudes of the absorption observed by riometer and calculated from EISCAT are in reasonable agreement, indicating that the derived densities are reliable, at least to within a few tens of per cent. In contrast to the spectral anatysis, the calculation of electron densities includes a correction factor for finite Debye length (MATHEWS

ef al., 1982). The importance of this correction increases for smaller densities: below 10” mm3 the corrected values can be up to a few tens of per cent greater than the raw values. The Debye correction also includes the electron to ion temperature ratio, which we have assumed to be unity ~KI~KW~D and COLLIS, 1987), but any real deviations from this assumed value will introduce uncertainties in the derived densities.

The second cross-check compared the derived den- sities with those obtained from power profile measure- ments using a single short (20 ps) pulse, during a different experiment at the same local time on the following day (2 July), when geomagnetic conditions were not disturbed. Good agreement was obtained for D-region heights, but above about 100 km the results from the present experiment were a little low and are not included in Fig. 2. This discrepancy is presumably due to the inapplicability of the model ACF with increasing altitude into the E-region.

Two distinctive periods of activity can be recognised in the electron densities in Fig. 2. The first (1954 2036 UT) was characterised by hard precipitation causing electron densities of more than 4 x 10’ mm3 occasionally at 75 km altitude. During the second, the precipitation was much softer and the electron densities do not allow reliable spectral measurements below the 80 km level. This period covers the time inter- val 2100-2134 IJT and contains an interesting narrow layer phenomenon around the mesopause altitudes. These two events with enhanced precipitation are sepa- rated by a period when the precipitation was too low to create electron densities of sufficient magnitude for reliable parameter estimation, and only an estimate of the power can be obtained. This period also con- tained a short interruption of the experiment.

DOPPLER SHIFTS, DOPPLER BROADENING AND

COHERENCE TIMES

Figure 4 shows the Doppler velocity variation along the beam during the interval 1958-2034 UT. The inte- gration time is 4 min per profile, and at this resolution clear indications of wave-like motions can be seen.

These features are not studied in more detail in this paper, but the results show that even by using the EISCAT UHF system, a mapping of the waves and winds in the D-region is possible under suitable con- ditions with a suitably arranged experiment. Here the Doppler velocities were analysed only for checking the possible effects of winds and waves in the coher- ence time determination.

Examples of the measured incoherent scatter spec- tra are shown in Fig. 5, which represent an average over the interval 1958-2034 UT. These were obtained by calculating individual 4 min averaged spectra, sub- tracting the Doppler shift from each of these and then taking the mean of the corrected spectra. Also included in Fig. 5 are the analytical tits to the data points, assuming a Lorentzian shape for the spectrum. The agreement between measured and fitted spectra can be seen to be very good, thus supporting the physical arguments which lead to the anticipation of this type of spectrum, and allowing us to interpret the measurements in a quantitative way within this formulation.

The necessity of removing the Doppler shift from the measured spectra can be judged from Fig. 4, where it can be seen that the velocity along the beam can change by several metres per second within a few minutes. We quantify this statement in terms of the effect on spectral width in Fig. 6. The heavy line shows the fitted spectral widths to the measurements in Fig. 5, calculated as described above, i.e. with the Doppler shift subtracted from each 4 min integration and then taking the mean. The second line in Fig. 6 is the spectral width obtained by averaging together the whole 36 min of data without prior Doppler correc- tion and then removing the resultant Doppler shift. The difference amounts to an overestimate of up to about 20 Hz between altitudes 75 and 87 km in this case, if the latter method is used. This becomes especially important at the lower altitudes, since the percentage error increases.

Apart from this illustration of Doppler broadening for longer integrations, all other results presented here are based on individual Doppler corrected 4 min aver- ages. These may also be broadened by the presence of more rapidly varying velocities, but we anticipate this effect to be small.

The coherence time profiles using 4 min integration times are shown in Fig. 7 for the two periods which allowed reliable analysis. Temporal variations within the earlier interval are rather small, thus justifying the use of 36 min averages of these data in Figs. 5 and 6. The later interval shows more variability, with patchy coverage due to small signals from the lower altitudes. For this second event, we shall concentrate on the

293

MB1 2-

l-

Spectral measurements of summertime D-region

Lavangsdalen

I I I I I I I I

20.30 21.00 UT

Kilpisjsrvi

I I I I I I I I 20.30 21.00 UT

EISCAT , calculated value

I I I I I I I I 20.00 20.30 21.00 UT

Fig. 3. Comparisons of radio absorption measured by the riometers at Lavangsdalen (20 km south of EISCAT) and Kilpisjirvi (100 km south-east of EISCAT), with that calculated from the electron densities measured by EISCAT. The differences in detail probably arise from small-scale structures, whilst the

general agreement indicates that the deduced electron densities are reliable (see text).

295

feature displaying a large coherence time at about

85 km altitude around 2110 UT.

DISCUSSION

(a) First interval, 1954-2036 UT

During the first period the coherence time (or spec-

tral width) was characterised by a rather smooth vari-

-20 0 20 bO 60 80 100

VELOCITY Ims-‘,

Fig. 4. Altitude profiles (73-90 km) of Doppler velocity between 1958 and 2034 UT, presented as consecutive 4 min averages, with the zero level of each shifted by + 10 m s-‘.

The velocity scale refers to the first profile.

W&J km

83.38 km

82.36 km

81 .X km

80.32 km

79.30 km

10.20 km

77.25 km

76.23 km

75.21 km

14.19 km

73.17 km

72.15 km

c 1

-300 -200 -100 100 200 300

FREQUENCY (Hz)

Fig. 5. Examples of measured incoherent scatter spectra (points) and Lorentzian fits to the spectra (full lines) for altitudes between 72.15 and 84.4 km, for the interval 1958% 2034 UT. The ‘ringing’ in the measured points is a result of

the method of transform from the measured ACF.

296 E. TURUNEN, P. N. COLIJS and T. TURUNEN

76 -

7L -

0 100 200 300 w)O 500

FREQUENCY (Hz)

Fig. 6. Comparison of altitude variation of fitted spectral width for the interval 1958-2034 UT, calculated with Dop- pler subtraction for each 4 min integration (thick line) and with Doppler subtraction after the whole 36 min integration (thin line). The broader spectra below 87 km for the latter case are due to wind or wave variations of velocity on time

scales shorter than 36 min (see text).

ation with altitude, and little temporal variation. For

this reason, we have based the study of this interval on the averages of the measurements over the whole 36 min (Figs. 5 and 6). In order to interpret the measurements, an equivalent collision frequency pro- file was derived from equation (l), with the assump- tion of a positive ion mass of 30.5, a model neutral temperature profile, and an absence of negative ions (expected to be dissociated in sunlit conditions). Two

models of the neutral atmosphere have been used in the calculations, the first being CIRA (1972), and the second the MSIS one (HEDIN, 1983), with extra- polation below the mesopause based on the work of

ALCAYD~ (1981). The resultant collision frequency profiles are plot-

ted in Fig. 8, together with the model temperature

profiles. For comparison, the collision frequency pro- files based on the assumed model atmospheres are also included in Fig. 8, calculated from the expression

v,, = 8.92 x lo9 p, where p is the atmospheric neutral density (BANKS and KOCKARTS, 1973). The agreement between the calculated and model collision frequency profiles is surprisingly good and lends confidence to our method of analysis. For the upper part of the profile the derived collision frequencies agree very closely with those of the MSIS atmosphere, whilst for

the lower part the results are almost identical with those predicted by the CIRA model. The only depar- ture from agreement is in a small altitude range around 86 km, where the measurements indicate larger values by a factor of about 50%. One possible explanation of this feature is that our assumption of

ion mass fails at these altitudes and that a realistic mean mass should be higher. A mass of about 40 at 86 km altitude is required to bring the model and derived collision frequencies into agreement. Since we are measuring the mean effective mass of a mixture of species, it is not possible to identify individual con- stituents, but we note that heavy positive ions of the type H+(H,O),, where n may be up to about 20, have been observed in narrow layers near the arctic mesopause by rocket experiments (BJ~~RN and ARNOLD, 198 1 ; KOPP et al., 1985). The present results suggest a mean value of n = 2 at 86 km altitude. Note that this is a mean value and that smaller proportions of heavier ions could be present. Metallic ions of meteoric origin are also commonly detected in the D-region, as are other complex clusters, but the low temperatures expected in the vicinity of the observed

increase of ion mass suggests the main component

to be proton hydrates, which form more easily with decreasing temperature (ARNOLD et al., 1980).

(al lb)

74, \,, ,,\\ , . ~ _,

0 20 40 60 80 100

9oll I

0 20 40 60 80 100

tc (msl tc Imsl

Fig. 7. Profiles of coherence time (4 min averages) over the altitude range 73-90 km for the two intervals (a) 1958-2034 UT and (b) 2058-2134 UT discussed in the text. Consecutive profiles are shifted by

+ 10 m SS’ in each panel. The coherence time scale refers to the first profile in each panel.

Spectral measurements of summertime D-region 291

T IK)

200

1

d lot

V,” (s-l) Fig. 8. Collision frequency profiles derived from the mea- sured coherence times (0) over the interval 1958-2034 UT, together with the model collision frequency ( x ) and the model neutral temperature (+), for the CIRA model (top

panel) and the extrapolated MSIS model (lower panel).

The heaviest proton hydrates are expected at the

transition between the molecular ions (NO+ and 0:) above, with lighter hydrates below. The good agreement between the CIRA model and the derived collision frequency below 86 km (and the reasonable agreement for the MSIS model), indicates that the

mean mass used in the calculation is probably correct. For proton hydrates, this would result from a mean hydration index of 1.5. Alternatively, if our sup- position of predominance of cluster ions during this interval is wrong, the results could indicate that the molecular regime covers all altitudes above about 75 km, which might be expected during the rather hard electron precipitation (COLLIS et al., 1986b). In

that case, it is likely that the layer near 86 km is com- posed of metallic ions.

It is clear from Fig. 8 that the accuracy of the measurements is of the order of the discrepancies

between currently available atmospheric models for this altitude range. Because of the assumptions required in the deduction of our equivalent collision frequency profile, the deviations from the model values can be accounted for in a number of ways, on which we can place limitations according to physical reasoning. For example, the discrepancy with the MSIS model below 83 km altitude can be explained by the presence of negative ions, with 1 = 0.2 at 82 km, increasing to 1 = 0.5 at 76 km. Alternatively, a positive ion mass smaller than the one of 30.5 used in the calculations may be more applicable, indicating hydrated protons with a mean hydration index slightly greater than one. The discrepancy with the CIRA model above about 88 km could be due to tem- peratures of the order of 3WO% less than the model. Both of these are realistic possibilities : the suggested i values are quite small, and low concentrations of negative ions may be present when the solar zenith angle is close to 90” (Z 85” for the present obser- vations). Very low temperatures may be expected around the summer mesopause at high latitudes

(ARNOLD and Joos, 1979) and 1OOK is not implaus- ible.

Further refinements in interpretation and evalu- ation of parameter values are possible, but usually at the expense of one or more assumptions. For example, TEPLEY and MATHEWS (1978) and TEPLEY et al. (1981) have demonstrated a technique for determining tem- perature, collision frequency and neutral number den- sity by subdividing the profiles into small enough height intervals for the assumptions of constant ion mass in an isothermal atmosphere to be valid, with further deductions in a self-consistent manner. For

the present purposes, we cannot confidently justify such assumptions in the aurora1 ionosphere, although as more data are gathered this approach could be explored. In the absence of further independent infor- mation, however, we are not able to define a unique collision frequency profile, nor to assess the detailed accuracy of the two models discussed here.

(6) Second interval, 2100-2134 UT

The second period allowed at least limited deter- mination of spectral characteristics. As seen from the contour plot (Fig. 2) the main ionisation decreases very rapidly below about 90 km. This indicates that the spectrum of the precipitating electrons is softer than in the first event, or that the loss mechanisms do

298 E. TURUNEN, P. N. COLLIS and T. TURUNEN

not behave in the same way. The clear boundary of the ionisation is at about the same altitude as seen by

rockets in the summertime high latitude D-region, at

the altitude where the hydrated ions disappear and the NO+ dominated regime begins (KOPP and HERR- MANN, 1984). In the low density region around 86 km altitude, a narrow layer with slightly enhanced elec- tron densities can be seen. This layer is about 2 km wide, changes with time and lasts for about 10 min. The main feature of this layer is that the coherence time is much longer than in the surroundings (Fig. 7). The altitude of the layer is the same as that in the first period where the local decrease in spectral width was observed. It is not reasonable to explain this type of local behaviour by temperature deviations (a local cold-spot of 40K would be required) and negative ions would decrease the coherence time. The most probable explanation is a layer with very heavy posi- tive ions, possibly hydrated protons with large hydration index. Wider discussion of this feature, and a more detailed analysis with finer time resolution, has been given by COLLIS et al. (1987). The longest measured coherence time using 4 min averaging (23 ms) requires an ion mass of almost 200 to explain the result. This corresponds to a mean hydration level of the order of 10. Such ions have been measured by rocket experiments in the arctic summer mesosphere in the presence of noctilucent clouds (KOPP et al.,

1985), so this explanation of our observations is not unrealistic, though we have no confirmation that

noctilucent clouds were present. The apparent time variation observed in this layer can be explained either by true temporal variations, or by horizontal drift of a structured layer. The latter seems more probable, although from the measurements themselves there are no ways to determine which alternative is the correct one.

CONCLUSIONS

We have described power profile and incoherent scatter spectral width measurements of the ionospheric

D-region by the EISCAT UHF radar during weakly disturbed summertime conditions. The results illus- trate that the radar is capable of measuring D-region parameters with good resolution, provided that the conditions are suitable. The accuracy obtained is com- parable with the uncertainty existing among current models of the neutral atmosphere for mesospheric heights, but a detailed interpretation of the obser- vations is limited by the lack of a single reliable model.

If we assume a description according to the CIRA model, then the results suggest a lower temperature around the mesopause and above, whilst if we assume the extrapolated MSIS model, it is necessary to allow for moderate concentrations of negative ions below the mesopause.

A clear observation which cannot be explained by an inaccurate atmospheric model is the localised mini- mum of the width of the incoherent scatter spectrum around 86 km altitude, which was at times up to an order of magnitude different from the surroundings. These results suggest a more or less persistent layer of ions of the type H+(H,O),, with a mean value of n of approximately 2 during the first part of the experi- ment, but increasing up to about 10 for short intervals during the latter part of the experiment. This conclusion, together with the conditions of the experi-

ment, are consistent with measurements in the vicinity of a structured noctilucent cloud.

Whilst the EISCAT VHF radar should eventually prove to be more suitable than the UHF radar for D-region studies in general, we have shown that Dop- pler velocities are clearly measurable with the UHF radar and future experiments could utilise the full steerability of the UHF system for studies of meso- spheric dynamics in a more effective way than with the VHF radar.

Acknowledgements-The EISCAT Scientific Association is supported by the Centre National de la Recherche Sci- entifique of France, &omen Akatemia of Finland, Max- Planck-Gesellschaft of F.R.G., Norges Almenvitenskaplige Forskningsrad of Norway, Naturvetenskapliga For- skningsridet of Sweden and the Science and Engineering Council of the U.K.

ALCAYDB D. ARNOLD F. and Joos W. ARNOLD F., KRANKOWSKY D., ZETTWITZ E.

and Joos W. BANKS P. and KOCKARTS G. BJ~RN L. G. and ARNOLD F. CIRA

1981 1979 1980

1973 1981 1972

Ann. Geophys. 37, 515. Geophys. Res. Lett. 6, 763. J. atmos. ferr. Phys. 42,249.

Aeronomy. Academic Press, NY. Geophys. Res. Lett. 8, 1167. COSPAR International Reference Atmosphere. Akad-

emie-Verlag, Berlin. COLLIS P. N. and KIRKW~~D S. 1987 Adv. Space Rex I, 349.

REFERENCES

Spectral measurements of summertime D-region 299

COLLIS P. N., KIRKWO~D S. and HALL C. M. COLLIS P. N., HARGREAVES J. K., BREKKE A.

1986a 1986b

and KORTH A. COLLIS P. N., TURUNEN T. and TURUNEN E. DEVLIN T., HARGREAVES J. K. and COLLIS P. N. FUKUYAMA K. FUKUYAMA K. and KOFMAN W. GANGULY S. HARGREAVES J. K., RANTA H., RANTA A., TURUNEN E.

1987 Geophys Res. Lett. in press. 1986 J. atmos. terr. Phys. 48, 795. 1981 J. geophys. Res. 86,9 152. 1980 J. Geomagn. Geoelect. 32,67. 1980 Geophys. Res. Lett. 7, 369. 1987 Planet. Space Sci. 35, 947.

and TURUNEN T. HEDIN A. HUUSKONEN A., POLLARI P., NYGREN T.

1983 J. geophys. Res. 88, 10170. 1988 J. atmos. terr. Phys. 50,265.

and LEHTINEN M. KIRKW~OU S. and COLLIS P. N. KOFMAN W., BERTIN F., RBTTGER, J., CREMIEUX A.

1987 Adv. Space Res. 7, 83. 1984 J. atmos terr. Phys. 46, 565.

and WILLIAMS P. J. S. KOPP E. and HERRMANN U. KOPP E., EBERHARDT P., HERRMANN U.

1984 Annls geophys. 2,83. 1985 J. geophys. Res. 90,304l.

and BJ~~RN L. G. MATHEWS J. D. MATHEWS J. D. MATHEWS J. D. and TANENBAUM B. S. MATHEWS J. D., BREAKALL J. K. and GANCULY S. RANTA A., RANTA H., TURUNEN T., SILEN J.

1976 J. geophys. Res. 81,467 1. 1984a J. atmos. terr. Phys. 46 975. 1981 Planet. Space Sci. 29, 335. 1982 J. atmos. terr. Phys. 44,441. 1985 Planet. Space Sci. 33, 583..

and STAUNING P. TANENBAUM B. S. TEPLEY C. A. and MATHEWS J. D. TEPLEY C. A., MATHEWS J. D. and GANGULY S. TURUNEN T.

1968 Phys. Rev. 171,215. 1978 J. geophys. Res. 83, 3299. 1981 J. geophys. Res. 86, 11330. 1986 J. atmos terr. Phys. 48, 777.

Reference is ulso made to the following unpublished material.

MATHEWS J. D. 1984b

SCHLEGEL K. 1979

J. atmos. terr. Phys. 48, 807. Annls gCophys. 4,211.

Handbook for MAP, p. 135, SCOSTEP Secretariat, University of Illinois, Urbana, IL, U.S.A.

Report MPAE-W-05-79-01. Max-Planck-Institut fiir Aeronomie, Katlenburg-Lindau, F.R.G.