Ionospheric variability from an incoherent scatter radar long-duration experiment at Millstone Hill



[1] An incoherent scatter radar experiment at Millstone Hill covering 30 consecutive days in September 2005 has enabled this study of day-to-day ionospheric variability. This was a period of low solar activity with few magnetically disturbed periods. Our discussion focuses on ionospheric variability during quiet magnetic activity in the 100–500 km height range, with emphasis on its height variation at noon. (1) Very large midday variability is present for the ion temperature Ti near 120 km, which is verified by two other 30-d experiments at Millstone Hill. This is not apparently associated with solar flux and magnetic activity. The percentage variability in the midday electron density Ne changes with height, being smaller between 150 and 250 km and larger in the topside. (2) With increasing solar flux, Ne decreases between 170 km and the F2 peak and increases elsewhere, being essentially unchanged near the F2 peak. With increasing magnetic activity, Ne decreases between 160 and 325 km. Ti increases with solar flux and magnetic activity, in particular in the F2 region. (3) There is a time lag of ionospheric responses, varying with height, to changes in solar-geophysical conditions: In the E region, the lag is almost zero; above the F2 peak, both Ne and Ti respond to F10.7 with a 2–3-d delay. The delay in response to 3-hourly ap index changes for Ne above the F2 peak can be 9–12 h and between 160 km and the F2 peak can be 0–3 h. The time delay for Ti is 6–9 h. (4) We estimate that the majority of the topside variability in Ti and Ne can be explained in terms of solar flux F107 and magnetic activity ap effects. Near the F2 peak, Ne variability seems to be complicated, and nearly one half (10%) of it cannot be ascribed directly to F10.7 and ap effects.

1. Introduction

[2] Ionospheric variability, defined as the deviation from climatological mean, is a pronounced and permanent feature of the upper atmosphere that occurs on hourly, daily, seasonally, and solar cycle timescales. In particular, the day-to-day variability and the variability over temporal scales from days to one month have been considered to be involved with, in addition to geomagnetic activity, lower atmosphere processes including acoustic, tidal, and planetary waves, as well as meteorological processes. Some recent publications are Forbes and Zhang [1997], Forbes et al. [2000], Rishbeth and Mendillo [2001], Altadill and Apostolov [2001], Mendillo et al. [2002], Laštovička [2006], Rishbeth [2006], and Moore et al. [2006]. Quantifying the variability is an important part of space weather and climate modeling with various science and engineering applications. A number of prior publications and technique reports have pursued the representation of the variability in foF2, hmF2, and MUF; see Radicella [2003], Bradley et al. [2004], Wilkinson [2004], and references therein for some recent major efforts.

[3] Almost all prior studies have typically focused on the F2 peak parameters (occasionally on the E peak electron density) provided by ionosondes and in some cases TEC. There exists an obvious gap in that the ionospheric variability in the broad E through F region has yet to be well characterized. While a major and significant part of the ionosphere, the F2 peak has a pronounced different nature from the photochemical dominated lower heights and the diffusion dominated topside. Electron density seems to be heavily studied compared to other ionospheric parameters. Plasma temperatures, however, carry important information on the thermal properties in the ionosphere and thermosphere. The thermospheric temperature can have fundamental influences on the ionospheric electron density by changing the neutral composition and winds and by altering the production-diffusion chemical loss equilibrium of the ionospheric ions. Exploring the association of the variability with solar-geophysical conditions and clarifying impacts from below is a topic of great importance to a full range of understanding of this important ionospheric phenomenon. Long-duration incoherent scatter radar (ISR) experiments covering as long as one full month, currently being conducted at Millstone Hill and elsewhere [Zhang et al., 2005] have enabled us to further study the variability, in particular, its height variation (rather than only near the F2 peak) for a set of important ionosphere/thermosphere parameters (in addition to electron density).

[4] This paper uses the September 2005 30-d ISR experiment at Millstone Hill to characterize ionospheric variability and to establish the correlation of the variability with solar-geophysical conditions under quiet to minor magnetic activity (kp ≤ 3). We focus on the height variation of the variability and of its correlation with geophysical indices. We will note, among many other features, a pronounced peak variability in ion temperature in the upper E region, and indicate that this peak is mostly uncorrelated with effects of solar flux and magnetic activity. Solar 10.7 cm flux F10.7, 3-hourly ap index, and hourly Dst indices are involved in the correlation analysis, in order to determine the extent of impacts of solar-geophysical variability. With kp ≤ 3, we are able to find a very sensitive correlation of sets of observed F2 region parameters with Dst. We can also see the time delay of ionospheric responses to solar flux and magnetic activity indices.

[5] In the following sections, we first describe the experiment setup and present an overview of the 30-d experiment in September 2005 with information on the solar-geophysical conditions (section 2). In the results section, a brief discussion on the local time variation of the variability and solar zenith angle effect is given in section 3.1. The midday variability is our focus and is shown for various parameters at different heights (section 3.2). This is followed by correlation analyses between the ionospheric parameters with solar-geophysical indices to indicate effects of dynamical and chemical processes (section 3.3). The next subsection addresses the time lag of ionospheric responses to solar and magnetic activity indices (section 3.4). We evaluate quantitatively the solar flux and magnetic activity induced variability in the last subsection of the results. This paper is concluded with a summary of our main findings.

2. Observation

[6] The Millstone Hill (42.6°N, 288.5°E, Invariant Lat. 53.4°) ISR operated for 30 d from 1 through 30 September 2005. The operation was mostly for 16 h between 0800 and 2400 UT on each day; there was one week (12–18 September) and 3 d (27–29 September) when the coverage was extended to 24 h d−1. This experiment was designed to provide rapid time resolution coverage of the E region and F region ionosphere, yielding good altitude resolution in both regions. A 68-m zenith antenna and a 46-m steerable antenna (MISA) were used during the whole campaign. The zenith antenna provides measurements directly overhead Millstone Hill and the steerable provides data for a number of other directions enabling the determination of the vector ion drift velocity as well as regional features of the ionosphere. This permits local neutral winds and electric fields to be determined.

[7] A zenith observation is made between each MISA position, so that the observational cycle was Zenith ⇒ MISA (north, 45° elevation) ⇒ Zenith ⇒ MISA (west, 45° elevation) ⇒ Zenith ⇒ MISA (west) ⇒ Zenith ⇒ MISA (north). The dwell time in each position is 8 min to provide the possibility of a long integration if needed, so the overall cycle time of the experiment is 65 min with a measurement triplet every 32 min. An interleaved single pulse of 480 μs and alternating code scheme was used for all pointing positions and antenna selections during the experiment cycle, offering range resolution of 72 km and 4.5 km, respectively.

[8] For our current study, we use only the zenith data for electron density Ne, electron temperature Te, ion temperature Ti, and vertical ion drift V0, thus we limit our discussions to features only local to Millstone Hill. Except for an overview immediately below of general features of the data and solar geophysical conditions over the entire campaign, our focus will be on the midday data with emphasis on their height variations. We will not use winds and electric field data in this study, both of which are derived on the basis of some assumptions, although these data are potentially helpful in addressing details of the complicated dynamical processes. In general, however, the dynamical processes such as winds and electric fields are affected by the solar-geophysical conditions, which serve as the ultimate origin of the ionospheric variability we are to discuss. Therefore our approach here is largely to tie the ionospheric manifestation with the ultimate origin of the variability, while skipping over the intermediate details.

[9] Our height variation results are discussed for two distinct height bounds: 100–270 km height for which alternating code data are used; 270–500 km for which single pulse data are used. Variations near 270 km from one bound to another should be physically continuous. As shown in the figures to be discussed, this continuity is reasonably well preserved in most cases where the percentage variability and correlation analysis are performed for multiple parameters. However, since (1) the range resolution for both modes of observations are significantly different, (2) the height range and number of data points in a bin near this boundary are not the same on each side of the bound, and (3) the level of measurement errors is different, a small discontinuity near 270 km is still present. It is more meaningful to use data from the same observational mode for quantitative comparisons of results for different heights. Therefore our discussions are given largely for two separated ranges: 100–270 km for the bottom of the ionosphere, including the F2 peak and 270–500 km for the topside.

3. Results and Discussion

3.1. Overview of Solar-Geophysical and Ionospheric Conditions

[10] The overall geophysical conditions and zenith Ne data are shown in Figure 1. The solar activity was rather low. F10.7 daily index was in the range of 72–119, with a median value of 86, mean value of 92, and a standard deviation of 16 with respect to the mean. F10.7 gradually increased from 3 to 4 September, reached high values near the middle of the month, and decreased toward the end of the month to complete its 27-d cycle. During this period, some solar flares occurred and produced sudden enhancements in the E-F region ionospheric electron density on 7, 9, and 10 September.

Figure 1.

Overview of 30-d experiment data and solar-geophysical conditions for September 2005.

[11] The hourly Dst index was low (negative) in the middle of the month when F10.7 was high. It reached minimum on 11 September. During the development of the main phase, an explicit feature of electron density enhancement, electron temperature abatement, ion temperature increase, and upward ion drift was observed on 10 September in the local afternoon sector, followed by an electron density decrease and ionospheric temperature increase on the next day (11 September). The electron density level recovered on 12 September and became low again on 13 September. The storm recovery phase, as indicated in the Dst index, lasted for about 10 d. Other days of appreciably low electron density include 4 and 5 September, following a small drop of Dst in the earlier hours and a brief kp increase on the prior day. It is noted that even in those intervals when Dst and F10.7 did not appear to change dramatically, the ionospheric day-to-day variability can be very large. Before we discuss in detail the midday variability behavior, it is worthwhile to first examine the time variation of the variability during the day.

[12] Day-do-day variability, as the central topic of this paper, is defined as the standard deviation of data with respect to the bin mean (average); a data bin consists of observations from different days for certain height and time intervals. It is sometimes represented as a percentage value of the standard deviation over the mean. The ionosphere during periods of high magnetic activity is very different from that during quiet periods. The variability we are studying, however, is not the same type as these sharp differences. It is the day-to-day fluctuation during relatively quiet periods that we are interested in. We select all data for kp ≤ 3 for most of our studies. Figure 2 shows the time variation of the percentage variability for below 270 km (Figure 2a) and for above 270 km (Figure 2b).

Figure 2.

Time variation of electron density variability for (a) <270 and (b) >270 km. Variability is in percent, defined as standard deviation over mean. Dots are the percentage variability in cosχ, where χ is solar zenith angle.

[13] Below 200 km, the percentage variability is large, over 30%, during sunrise and sunset periods, and about 10% near midday, although the absolute variability (standard deviation) is higher near noon than at sunrise and sunset. At these heights, much of the variability arises from the variability in the solar zenith angle over the 30-d period. This effect is illustrated in the percentage variability in cosχ (χ, solar zenith angle), which is presumably proportional to the photoionization rate. As shown by the dots, the variation pattern follows closely the Ne pattern. The effect of rapid changes in sunrise/sunset times on the variability is verified by the significantly reduced variability as a result of shifting the timescales for each morning and evening, to simulate a situation where sunrise/sunset times are approximately the same over the period. This solar zenith angle effect was also noted by Moore et al. [2006]. The cosχ variability, however, is quantitatively lower than the Ne variability, so other factors may also contribute to the Ne variability. As noted earlier, the solar flux represented by F10.7 shows some degree of variability. In fact, this is 12% for magnetically quiet days, and 19% for all the days during the period. Then

equation image

where std is standard deviation and the last term is (12%)2. Therefore it seems that the solar flux and solar zenith angle variability account for the majority of the variability in Ne during quiet conditions.

[14] Above 200 km, variability is generally larger than below. The percentage variability tends to be high near sunrise and sunset. The cosχ curve is mostly well below the Ne curve, except when it is far into the sunrise and sunset periods such that the cosχ assumption (or even Chapman function) fails. Slightly after noon, the percentage variability reaches its daytime maximum, which grows with increasing height. Examining time distribution of ap, F10.7 and their standard deviations, we note coincident peaks, which partly explains the timing of the daytime maximum variability in Ne. Given the same solar-geophysical conditions, however, we note that the variability changes with height. The next section addresses in detail this topic.

3.2. Variability at Different Heights

[15] Now we study in detail the midday variability under relatively quiet magnetic activity conditions. Data between 1700 and 1800 UT (∼1200–1300 LT) are binned according to height, with a bin size of ±20 km. The magnetic index kp has been less than 3 (kp ≤ 3) throughout the prior 9 h (or between 0800 and 1800 UT). We assume the 1-h time bin near noon is small enough that local time variation within this interval can be neglected. To efficiently remove outliers, we filter out data that depart from the median more than two standard deviation units, yielding a subset which is 95% of the total data for a Gaussian distribution. Results of the variability for Ne, Ti, and V0 are shown in Figure 3 where areas shadowed with horizontal lines represent the percentage variability, and gray areas are absolute variability (standard deviation). The absolute variability profile for Ne and Ti, respectively, is normalized to its maximum value on the profile so that only the shape of the height profile is meaningful.

Figure 3.

Height variations of variability in (left) Ne, (middle) Ti, and (right) V0 for (a) <270 and (b) >270 km. Variability in percent, defined as standard deviation over mean, is shown as the area marked by horizontal lines, and mean values are given as dashed lines. The absolute variability for Ne and Ti is normalized and is shown by the gray area, where a larger deviation from the box border line on the right indicates a larger absolute variability. For V0 the absolute variability uses the same horizontal axis as for the mean V0.

[16] Below 200 km (Figure 3a), the absolute variability in Ne increases with height; the corresponding percentage variability exhibits a small peak between 125 and 150 km where Ne starts to grow rapidly with height. The vertical ion drift (positive for upward) is downward above 150 km, as a result of poleward meridional winds at noon and of downward ion diffusion. The wind shear effect appears to favor ionization convergence at 150 km, although ionization would not actually build up substantially because of the large loss rate, the variability in Ne is slightly higher. This upward ion drift below 150 km, mostly contributed by the equatorial winds, can often be seen at noon in Millstone Hill observations. At this height, where the semidiurnal tide might be stronger than the diurnal tide which prevails in higher altitudes, the vertical drift reaches its maximum absolute variability.

[17] In Ti, the peak percentage variability at 120 km and the high absolute variability spanning the 120–150 km range are striking features. Such a pronounced variability occurs at the base height where mean Ti (or neutral temperature Tn) starts to grow rapidly (thermobase). The percentage variability can be as high as 40%, well above corresponding variability in Ne.

[18] Above 200 km (Figures 3a and 3b), we can see the variability is obviously higher in Ne but lower in Ti and V0. In Ne, the percentage variability can be nearly 35% and increases with height; the absolute variability, however, retains a single maximum slightly above the F2 peak which is at 250 km. The vertical drift changes to be upward at approximately 75 km above the F2 peak. At the peak the electron density height gradient is zero. Beyond this point toward the topside, ambipolar diffusion becomes dominant; in this region, variations at the F2 peak are amplified through scale height, or plasma temperature effects. They seem to be more important than neutral composition, winds and electric fields effects, although the ions are relatively long lived at high altitudes and any such effects would work for a longer time period.

3.3. Association With Solar-Geophysical Conditions

[19] In this subsection, we present rough estimates of contributions from solar and magnetic activity to the overall variability and identify the residual variability. It is well known that ionospheric behavior is strongly controlled by solar and magnetic activities. However, the correlation of some ionospheric parameters with daily solar flux is not as straightforward as one might have expected. The day-to-day variability in the F2 peak density was found not quite correlated with the daily flux [Rishbeth, 1993; Richards et al., 1994; Forbes et al., 2000; Rich et al., 2003]. We now examine correlation of parameters with solar-geophysical indices (Figure 4). In Figures 4a and 4c, corresponding to 100–270 km and 270–500 km ranges, respectively, correlation coefficients are calculated for a specific parameter and the F10.7 index. The F10.7 index is a daily value for the current day (denoted as fc). Below 170 km, there exists a positive Ne-fc correlation in E and lower F regions (100–170 km); that is, Ne increases with fc in the region where molecular ions dominate. The statistical significance of the correlation coefficient depends on the size of the sample. According to the standard t tables, for a sample of 220 data points (in a typical bin for >270 km), the 99% confidential level of significance is 0.17. This level is 0.14 for a sample of 350 points (in a typical bin for <270 km). It is more important to note the relative change of the correlation where high correlation coefficients, well above the 0.14 level, occur near the E peak and 160 km.

Figure 4.

Correlation of various parameters with solar-geophysical indices: (a) F10.7 index for the current day in the 100–270 km range, (b) −Dst index for the current hour in the 100–270 km range, (c) F10.7 index for the current day in the 270–500 km range, and (d) −Dst index for the current hour in the 270–500 km range.

[20] In the lower F region and the F2 peak (between 170 km and the F2 peak), the correlation is negative; near the F2 peak, Ne is almost uncorrelated with fc. The correlation in this region reflects the effects of competing processes: photoionization, which peaks between 150 and 200 km [Zhang et al., 1999], and chemical effects of molecular nitrogen and oxygen, responsible for ion recombination, which might be solar flux-dependent as well. It seems that the chemistry dominates photoionization such that an negative correlation can be present. Above the F2 peak (note for correlation coefficient calculations, each altitude bin has a ±9 km width for the <270 km plot and a ±18 km width for the >270 km plot), very clear positive correlation prevails. This may be related to the plasma scale height effect, since plasma temperatures have a positive correlation with fc, in particular, Ti is strongly positive correlated with fc.

[21] The correlation between Ti and fc is positive in the entire 100–500 km range, reaching nearly 0.8 around the F2 peak and approaching 0 at 150 km. The percentage variability peak for Ti near 120 km shown in Figure 3a does not correspond to any significant correlation between Ti and fc, indicating a weak solar flux effect.

[22] Te attains its largest correlation with fc near the F2 peak. At this height, where electron thermal conductivity is lower as compared to higher altitudes and the F2 peak density is uncorrelated with fc, the electron heat capacity and electron cooling rates are less dependent on fc such that the Ne-Te anticorrelation [see Schunk and Nagy, 1978; Zhang and Holt, 2004; Zhang et al., 2004] does not come to play; therefore it is the change of solar heating rate that results in the peak positive correlation with fc. Well above the F2 peak, however, Te is weakly correlated with F10.7 as a result of somewhat complicated Te-Ne coupling processes. On one hand, Te controls Ne through the scale height effect; on the other hand, Te is affected by Ne mainly through the Te-Ne anticorrelation due to electron cooling processes with rates proportional to Ne. At these heights, Ne is sensitive to fc, and the Te-Ne anticorrelation effect comes to play along with the solar heating effect.

[23] The vertical drift V0 is positive correlated with fc near the F2 peak, where the density gradient does not contribute to diffusion. This might indicate that the southward meridional wind, contributing to upward vertical drift, tends to increase as fc increases; that is, the regular noon time flow, which is poleward, is reduced because of the enhanced equatorward component of the wind as fc increases.

[24] As we are discussing variability during relatively quiet intervals, we have selected data for kp ≤ 3 (or ap ≤ 15) within the past 0–9 h in Figure 3. We can expect a very weak ionospheric response to such magnetic activity and low variability. Now we slightly modify our criteria to kp ≤ 4 (or ap ≤ 27) to allow for slightly stronger magnetic activity so better correlation evaluation can be obtained by adding more data. This latter criteria is used in the fc correlation evaluation discussed immediately above, and is also used here in magnetic activity correlation evaluation. Three-hourly kp and ap, and hourly Dst indices have been tested. The results for these indices are essentially similar, especially, for kp and ap; but ionospheric responses seem to be slightly more sensitive to the hourly Dst than to 3-hourly kp and ap. This seems to be due to two facts, (1) our Dst is hourly values with a higher time resolution, while kp/ap index is a 3-hourly value, and (2) since we are considering weak magnetic activity, the range of variation in ap, especially in kp, is very much smaller than that in Dst. Here we discuss Dst results as shown in Figures 4c and 4d; ap results are given in Figures 5b and 5d.

Figure 5.

Time delay of ionospheric responses to solar-geophysical indices. Correlation between Ne or Ti with (a) F10.7 for 0–3 d in the past in the height range 100–270 km, (b) ap for 0–9 h in the past in the height range 100–270 km, (c) F10.7 for 0–3 d in the past in the height range 270–500 km, and (d) ap for 0–9 h in the past for 270–500 km height range.

[25] We use dc, a Dst index for the current hour with the opposite sign, for characterizing the magnetic activity control, so a high positive dc means high activity of the ring current in the magnetosphere. We can see that Ne is negatively correlated with dc in the range 160–325 km; that is, Ne decreases with increasing magnetic activity. In this negative correlation area, the correlation between V0 and dc is positive and is somewhat stronger than in other regions. This positive correlation of V0 and dc cannot explain the Ne-dc negative correlation, because the increase in the upward vertical drift, as a result of the dc increase, causes the ions move upward to the region of low loss rate, favoring a higher Ne. It follows that the O/N2 composition effect may be responsible for the Ne decrease as dc increases (although magnetic activity does not reach a high level).

[26] At other heights, there is weakly positive correlation between Ne and dc. A small peak exists near the E peak. It is not clear whether this is due to some type of particle precipitation causing ionization at subauroral latitudes. In the topside, the scale height effect acts to give rise to an Ne increase when dc increases (therefore plasma temperatures increase), although the F2 peak density tends to decrease. Plasma temperatures are positively correlated with dc. The Te-dc correlation is much more stronger than Te-fc (Figure 4c). The heat source in the topside or the plasmasphere during high dc appears to impose its impact on the F2 region heights by thermal conductivity. Such a heat source can be energetic ring current particles that have been considered responsible for subauroral red arcs [Foster et al., 1994; Kozyra et al., 1997]. The dc seems to reasonably represent this heat source peculiarity, so the correlation between plasma temperatures and dc is positive and strong. At 120 km, the height of the percentage variability peak in Ti (Figure 3a), there is no indication that Ti is correlated with dc, although positive correlation does occur near 140 km.

[27] It is interesting to note that the fc correlation variations with height are very much similar to the dc correlation variations. This is likely due to the fact that a larger solar flux often coincided with stronger geomagnetic activity during this campaign (see Figure 1), so the daily F10.7 and the hourly dc for a given time of the day are correlated. This similarity introduces some ambiguity in qualitatively identifying effects of the solar flux and magnetic activity.

3.4. Time Delay

[28] Ionospheric responses to solar-geophysical variations take place when the ionosphere receives the signal and acts to reconfigure during which process the response can be maximized. We now investigate effects of the daily F10.7 indices for 0–3 d in the past and 3-hourly ap indices for 0–9 h in the past. Figure 5 shows comparisons of the correlation coefficients, with Ne shown in red and Ti shown in blue. In Figure 5a for <270 km and Figure 5c for >270 km, correlations with solar flux F10.7 for different days are examined. It can be seen that the correlation coefficient between Ne and a given F10.7 changes with height. In the E region, the current day index yields the highest correlation so the time delay appears to be insignificant. At higher altitudes, F10.7 for 3 d in the past yields correlation coefficients similar to those for 2 d in the past. Compared to the F10.7 for the past 2–3 d, the current day F10.7 tends to yield correlations shifted in the negative direction, i.e., stronger negative correlation in the 170–270 km range and weaker positive correlation for higher altitudes. These indicate that the time delay of the solar flux effect is insignificant between 170 and 270 km in the bottomside F2 region. In the topside, Ne is strongly positive correlated to F10.7 for 2–3 d in the past. Because of the significance of the plasma temperature scale height effect on Ne, one would expect that a time delay in Ti must also occur.

[29] This is indeed the case as shown in the Ti curves in Figures 5a and 5c. For Ti, the delay effect is pronounced above 160 km. With increasing F10.7, more energy is absorbed and accumulated in the ionosphere and thermosphere at midlatitudes because of neutral atmospheric absorption of solar heating and fast photoelectron heating. The ionospheric energy can be transported into the plasmasphere through thermal exchange, and the photoelectron energy can also be deposited in the plasmasphere. The result is a time delay in the Ti response to F10.7. A corresponding delay in the topside Ne is then possible because of the scale height effect.

[30] However, the response time in thermospheric parameters are different. Earlier results indicated that the thermospheric response time to the solar activity is nearly 1 d [Roemer, 1967; Buonsanto and Pohlman, 1998]. Jakowski et al. [1991] suggested, on the basis of a numerical simulation, a time lag of 2 d in the atomic oxygen concentration with respect to the solar radiation variation. This time shift in atomic oxygen is associated with photodissociation of molecular oxygen with a time lag of several days.

[31] We now examine the correlation between ionospheric parameters and magnetic activity indices with various time lags. We show 3-hourly ap index results for which we selected four time lags, 9, 6, 3, and 0 h. For 160–300 km, the highest correlation is in the current time 3-hourly ap index, and the lowest is in the 3-h time lag. This means that the noontime ionospheric F2 region Ne decreases at Millstone Hill tends to be better correlated with the “current time” ap generated within 1200–1000 LT (yielding the noontime ap index) than within 900–700 LT (denoted as “−3 hrs” in Figures 5b and 5d), 600–400 LT (denoted as “−6 hrs” in Figures 5b and 5d), or 300–100 LT (denoted as “−9 hrs” in Figures 5b and 5d). In the E region, the correlation between Ne and current time ap is not as high as between Ne and ap 3 h ahead. In the topside, the time lag effect in Ne is very obvious. A large correlation appears with a 6–12-h (in particular, 9–12-h) lag. For Ti, the correlation is the largest with 6-h lag from above approximately 210 km. The current time ap normally yields a small correlation, especially, in the topside.

3.5. Estimating the Variability Induced by Solar Flux and Magnetic Activity

[32] On the basis of our above results, we can now evaluate the variability associated with solar flux and magnetic activity. We assume that the midday ionospheric parameter p at a given height can be expressed as

equation image

where p0,…, p4 are coefficients to be determined using least squares fitting. In fact, the two quadratic terms are basically minor correction terms and insignificant. Fc and Ac are F10.7 and ap indices defined as Fc(d) = F10.7(d+τd) and Ac(h) = ap(h+τh), where d is the UT day and h is the UT hour at noon. According to our time delay results for various parameters, we use τd = −3 d for >270 km and τd = 0 for <270 km, and τh = −6 h. The model values evaluated using equation (1) are obtained for each data point with corresponding conditions used to generate Figure 3. They are then used to determine mean model value and standard deviation for each height bin. Figure 6 shows Ne and Ti results of the model percentage variability (the thick line and the area shadowed by horizontal lines) and the model absolute variability (gray areas).

Figure 6.

Solar flux and magnetic activity induced variability estimates based on a model equation (1) (see text) in (left) Ne and (right) Ti for (a) <270 and (b) >270 km. Variability in percent, defined as model standard deviation over model mean, is shown as the area marked by horizontal lines, and mean values are given as dashed lines. The absolute variability is normalized and is shown by the gray area, where a larger deviation from the box border line on the right indicates a larger variability.

[33] Below 200 km, it can be seen that both percentage and absolute variabilities are small. In Ne, the absolute variability increases with height while the percentage is below 10%. This percentage is very close to the observed variability after solar zenith angle variability (see section 3.1) is excluded from the 10–15% total variability presented in Figure 3a (as well as possible measurement errors are considered). This variability, arising from F10.7]− and ap-dependent forcings, is essentially insignificant in this range as compared to other ranges.

[34] Above 200 km, however, model Ne variability beyond the F2 peak (Figure 6b) is nearly equal to the overall variability shown in Figure 3b, in other words, the topside variability is essentially determined by the solar flux and ap indices. About 10% of the variability for the bottomside F2 region cannot be ascribed to the solar flux and ap indices, suggesting there are mechanisms not included in equation (1). The model Ti variability is slightly lower than but very close to the overall variability shown in Figure 3b.

[35] The Ti variability near 120 km evaluated from equation (1) appears to be much weaker than that observed (Figure 3a), with a percentage variability of 5% and a larger absolute variability near 150 km. We can conclude that the observed high variability of Ti near 120 km is unlikely to be due to solar and magnetic activity conditions or solar zenith angle changes. This high variability appeared also in other 30-d runs conducted at Millstone Hill. The 4 October to 4 November 2002 experiments were under relatively higher solar activity and were quiet for most of the days [see Zhang et al., 2005]. Analyzing the midday data below 200 km from alternating code measurements using the same method as for Figure 3 with kp ≤ 3, we can see (Figure 7, middle plots) that there is clearly a Ti percentage variability peak at 125 km, and an absolute variability peak at about 130 km. These features are very similar to those shown in Figure 3. The 6 March to 6 April 2006 was a period of slightly lower solar and magnetic activity as compared to the September 2005 case we have focused on. Again, our analysis of alternating code data for<200 km at noon with kp ≤ 3 indicates similar peaks of percentage and absolute variabilities for Ti. For other parameters, the variability patterns for corresponding V0 are very similar among all three 30-d runs; the variability in Ne is similar for the March/April 2006 and the September 2005 experiments not only in pattern but also in magnitude. The percentage variability in Ne for the October/November 2002 experiment indicates an obvious F1 layer variability centered between 150 and 175 km. Ne in the intermediate region between E and F2 layers starts to increase with height only from 150 km under such a relatively high solar activity condition, while it increases from 110 km for relatively low solar activity conditions. These differences among different levels of solar activity conditions involving interesting molecular ions processes can be a topic for our future studies.

Figure 7.

Height variations of variability in Ne, Ti, and V0 for two additional ISR 30-d runs: (a) October–November 2002 and (b) March–April 2006. Variability in percent, defined as standard deviation over mean, is shown as the area marked by horizontal lines, and mean values are given as dashed lines. The absolute variability for Ne and Ti is normalized and is shown by the gray area, where a larger deviation from the box border line on the right indicates a larger absolute variability. For V0 the absolute variability uses the same horizontal axis as for the mean V0.

[36] The electron temperature enhancement in the E region was also observed at subauroral latitudes [see St.-Maurice et al., 1989; Foster and Erickson, 2000, and references therein], and was considered to be associated with two-stream wave-induced E region heating. Oyama et al. [1983] noted the Te enhancement near 107 km associated with the Sq anomaly. It is not clear whether these enhancements in Te are related to our observations of Ti variability in the E region. It is also not clear whether this high Ti variability near the thermobase is associated with tidal components related heating activity. These may be interesting subjects to pursue but are beyond the scope of this current paper.

[37] Finally, it is worthwhile to note here instrumental errors and uncertainty of ISR measurements. At midday, the error in the measured electron density without being corrected for plasma temperature effects is normally <3% below 400 km, and grows to ∼10% at 500 km. Therefore below 400 km the true Ne error is actually well below 10%. The error for plasma temperatures is just a few percentages, smaller than that in Ne. We have used F10.7, Dst and ap indices to represent solar and magnetic activity controls. F10.7, for example, might not be good enough to be associated with Ne in the F2 peak region. Nevertheless, we find them convenient indices for current study which emphasizes more relative effects of solar and magnetic activities, and the relative change of the effects with height for a given index. It is also noted that the radar's range resolution and data binning in height may affect results in variability and correlation coefficients at heights where height gradients in them are significant.

4. Summary

[38] This study has used incoherent scatter radar measurements at Millstone Hill made during a 30-d campaign in September 2005, when the solar activity was low and there was little significant magnetic activity. We have focused our presentation on the ionospheric variability during quiet magnetic activity in the 100–500 km height range, with emphasis on its height variation at noon. We have investigated the variability associated with solar flux and with the weak magnetic activity variability, and discussed the time lag of ionospheric responses. We have also evaluated quantitatively the variability caused by solar flux and magnetic activity. Our main results can be summarized as follows:

[39] 1. Observations indicate clear effects of changes in the solar zenith angle on the day-to-day variability, especially near sunrise and sunset, which varies with height.

[40] 2. Near 120 km, there exists very large variability in the midday ion temperature, which is verified by observations during two other 30-d experiments at Millstone Hill. This is not apparently associated with solar flux and magnetic activity.

[41] 3. The percentage variability of midday Ne changes with height: It is relatively smaller from 150 to 250 km (from the lower F region through the F2 peak) and much larger in the topside. The absolute variability attains its maximum near or slightly above the F2 peak.

[42] 4. Some of the observed midday variability is due to the solar flux and magnetic activity variability. With increasing solar flux F10.7, Ne increases below 170 km with maxima at the E peak and 160 km, and it increases above the F2 peak as well. Between 170 km and the F2 peak, Ne decreases, with essentially no changes near the F2 peak. As Dst decreases (higher magnetic activity), Ne decreases between 160 and 325 km, while it increases below and above. Ti increases with F10.7 and -Dst, in particular, in the F2 region where the Ti correlation coefficients with these indices are above 0.5.

[43] 5. There exist time lags in the ionospheric responses, varying with height, to solar flux and magnetic activity indices: in the E region, there is almost no time delay; above the F2 peak, both Ne and Ti respond to F10.7 with a clear 2–3-d time delay. For magnetic activity, it is seen that, while no significant magnetic activity effects are observed below 160 km, above the F2 peak the time delay in the Ne response to the 3-hourly ap index can be 9–12 h, and between 160 km and the F2 peak the response appears more significant when ap in the last 3 h, compared to other hours, was large. The time lag for Ti responses to ap index changes is 6–9 h.

[44] 6. We estimate that the major part of the topside variability in Ti and Ne can be explained in terms of solar flux F10.7 and magnetic activity ap effects. Near the F2 peak, the cause of the Ne variability seems to be complicated and nearly one half of the variability (10%) cannot be ascribed to F10.7 and ap effects in a simple way.

[45] Our study leads to two findings that require further study; (1) Ti variability at 120 km and (2) a large portion of Ne variability near the F2 peak that cannot be simply ascribed to F10.7 and ap.


[46] We thank the members of the Haystack Observatory Atmospheric Sciences Group for assembling and maintaining the Madrigal database. Observations and research at Millstone Hill are supported by a cooperative agreement between the National Science Foundation and the Massachusetts Institute of Technology. This research was also supported by NSF Space Weather grant ATM-0207748.

[47] Zu Yin Pu thanks Henry Rishbeth and another reviewer for their assistance in evaluating this paper.