The complete seasonal variation patterns of the nocturnal mesospheric Na and Fe layers over Wuhan, China (30°N), have been established on the basis of several years of Na and Fe lidar measurements. Both the Na and Fe layer column abundances show strong annual variations as well as moderate semiannual variations with maxima in winter and double minima from late spring to midautumn (note that only one night of Fe data is presently available between mid-May and mid-July). The seasonal variation in the Fe abundance is evidently stronger than that of Na. The Na layer abundance has an annual mean of ∼2.5 × 109 cm−2, while this value for Fe is ∼7.5 × 109 cm−2. The Na and Fe centroid heights are dominated by semiannual oscillations with similar phases. The mean centroid heights are ∼91.4 km for Na and ∼88.7 km for Fe. The Na RMS width exhibits a strong semiannual oscillation with the layer slightly broader in winter, whereas the Fe width varies principally annually with a maximum in winter. The mean RMS widths of the Na and Fe layers are 4.5 and 4.1 km, respectively. The seasonal characteristics of the Na and Fe layers observed at 30°N have been compared with those currently available at other latitudes. The seasonal ratios of their abundances are smaller compared with 40°N and the South Pole. Their centroid heights and RMS widths also show less seasonal variations than the counterparts at all other latitudes. The annual mean Na and Fe abundances are about 60–77% of the counterparts at 40°N, 18°N, and the South Pole. This suggests that both the nocturnal Na and Fe layers have a low-abundance region around 30°N. On the basis of the results observed at the three latitudes in the Northern Hemisphere, the annual mean Fe layer width decreases with increasing latitude.
 Various free metal atom layers exist in the mesopause region from 80 to 110 km. They are believed to result from meteoric ablation. As meteoroids with velocities between 11 and 71 km s−1 enter the Earth atmosphere, the frictional heating causes them to vaporize, thus depositing gas-phase metal materials in the mesopause region. Five species of these metal atoms (Fe, Na, K, Ca, and Li) have been detected by ground-based resonance fluorescence lidars so far. Fe is the most abundant of the mesospheric metallic atoms, with an annual mean column abundance roughly twice that of Na [Kane and Gardner, 1993]. The mean column abundances of K and Ca are about 2 orders smaller [Eska et al., 1998; Gerding et al., 2000]. The large backscatter coefficients for the mesospheric Na and Fe layers, which are product of resonance backscatter cross section and number density, enable the strong lidar return signal to be obtained with a modest lidar system configuration (modest power- aperture product, altitude, and time resolutions). Therefore it is relatively easier to acquire the high-quality Na and Fe density data (i.e., high-accuracy atom density profiles with fine altitude and time resolutions) compared with K, Ca, and Li.
 The first complete monthly record for Fe layer seasonal variation came from Urbana (40°N, 88°W) [Kane and Gardner, 1993]. Above the 40°N site, the seasonal behavior of the Fe column abundance is dominated by an annual oscillation with a maximum in winter (December/January) and a minimum in summer (June/July), which is similar to that of the Na column abundance. The Fe seasonal ratio is ∼3.2, being slightly larger than that of the Na layer at the same site. The centroid height of the Fe layer has a strong semiannual oscillation, which resembles also the Na layer. The Fe layer width shows a strong annual oscillation similar to the Fe abundance. This differs from the seasonal behavior of the Na layer width at the same site (strong semiannual oscillation). The second report on the seasonal variation of Fe layer was from Arecibo (18°N, 66°W) [Raizada and Tepley, 2003]. It is noticed that the seasonal behavior of the Fe layer at the low-latitude site near 18°N differs clearly from that observed at 40°N [Kane and Gardner, 1993; Raizada and Tepley, 2003]. In particular, the centroid height and RMS width of the Fe layer are nearly out of phase with respect to its midlatitude counterparts [Raizada and Tepley, 2003], while the Fe column abundance lags that at midlatitude by about 3 months. The seasonal ratio at the low-latitude site has a value of 2.3, being smaller than at 40°N. Recently, the seasonal variation of the Fe layer over the South Pole has been studied by Gardner et al. . The Fe abundance is dominated by an extremely strong annual oscillation with a maximum in the austral winter (June). This is in accord with the Na abundance over the South Pole, representing the seasonal feature of the polar region. The seasonal ratio of the Fe layer over the South Pole is ∼4.2, being much smaller than the corresponding value for Na. The centroid height and RMS width of the Fe layer show both dominant annual oscillations and moderately strong semiannual oscillations. The dramatic seasonal variations of the Na and Fe abundances, centroid altitudes and RMS widths at the South Pole are partially ascribed to the Fe or Na depletion by PMC ice particles occurring in summer [Plane et al., 2004; Gardner et al., 2005].
 Since the seasonal characteristics of Fe layer have currently been obtained only at three different latitudes (18°N, 40°N and the South Pole), it is difficult to infer its latitudinal variation. Furthermore, the marked difference between the seasonal behaviors of the Fe layer observed at 18°N and 40°N [Kane and Gardner, 1993; Raizada and Tepley, 2003] makes the latitudinal variation more mysterious. In order to identify the seasonal characteristics and latitudinal variation of Fe layer, the Fe data with complete month coverage should be collected at more latitudes, particularly, at latitudes between 18°N and 40°N. In addition, although the latitudinal and seasonal variations of the global Na layer have been revealed by the satellite measurements, they represent the results observed merely at 0600 and 1800 LT [Fan et al., 2007]. Hence it is presently still needed to characterize the seasonal variation of Na layer based on ground-based lidar observations covering more local hours at more latitudes.
 The purpose of this paper is to reveal the seasonal characteristics of the nocturnal mesospheric Na and Fe layers at 30°N on the basis of extensive Na and Fe lidar measurements at Wuhan (30.5°N, 114.4°E) during the past several years. These characteristics represent the behaviors of the normal Na and Fe layers at the border between midlatitudes and low latitudes. They, together with those early reports [Kane and Gardner, 1993; Raizada and Tepley, 2003], may help to understand the puzzling latitudinal variation of the Fe layer in the Northern Hemisphere and help to validate the existing gas-phase chemical models of the mesospheric Fe and Na layers [Helmer et al., 1998; Plane et al., 1999; Plane, 2004].
 Our Na and Fe lidar systems were established respectively in March 2001 and December 2003. They both are installed in the lidar observatory building on the campus of Wuhan University. Their technical details have been described by Yi et al. [2002, 2007], and therefore only a brief presentation of the main operational parameters is given here. The transmitter of our Na lidar operates at a pulse repetition rate of 20 Hz and a pulse width of 6 ns, emitting a 589 nm beam with a pulse energy of ∼60 mJ. The beam is of a divergence of 0.5 mrad. The receiving telescope has an aperture of 0.52 m and a field of view of 1 mrad. Our Fe lidar utilizes a frequency mixing technique to obtain the Fe resonance beam at 372 nm. The 372 nm beam has a pulse energy of ∼40 mJ. Its divergence is 0.5 mrad. The pulse repetition rate and pulse width are the same as those of our Na transmitter. The receiving telescope of the Fe lidar is a 1-m Cassegrain system with a field of view of 1.0 mrad. A frequency stabilization device inside each transmitter effectively avoids the wavelength drift resulting from temperature variation. Though this frequency stabilization measure is taken, two pulsed wave meters are still employed to simultaneously monitor the wavelengths and line widths of two transmit beams. Long-term monitoring indicates that the two transmitters are highly stabilized in frequency (wavelength) and have an identical mean line width (1.8 GHz). The measured line width is used to calculate the effective Fe and Na backscatter cross sections respectively.
 Since installation, each lidar has been operated nearly routinely during nights when weather was good. For these routine measurements, both the Na and Fe density data were taken with a range resolution of 96 m and time resolution of 5 min. In this study we use the Na lidar data acquired from March 2001 through January 2002 and from June 2003 through November 2007, and the Fe lidar data from January 2004 through December 2007. The Na and Fe density profiles are derived from the corresponding lidar photon count profiles by using a standard inversion method [Gardner, 1989]. The normalization altitude is set to 30 km, and the number density value of air molecules at this altitude is taken from routine L-band radiosonde observations at Yichang Weather Station, China (111.3°E, 30.7°N). From more than 1 year (January 2006 to February 2007) of radiosonde data at30-km altitude (∼800 balloons reached this altitude), we obtain an annual mean of the atmospheric number density (3.875 × 1023 m−3) and a standard deviation (0.318 × 1023 m−3). This annual mean value is taken as the calibration density at the normalization altitude. The absolute accuracies of the nightly mean Na and Fe density profiles are generally limited to ±8.2% by the seasonal and diurnal variations in the calibration density. In addition, it is noticed that our calibration density (3.875 × 1023 m−3) is close to that value of 3.828 × 1023 m−3 at 30-km altitude from the U.S. Standard Atmosphere (1976). The sporadic Fe and Na layers are frequently observed at our lidar site near 30°N [Yi et al., 2002, 2007]. The sporadic layers are characterized by large density enhancements in a narrow altitude range and have no clear-cut relation with the normal metal layers (regular layers). For obtaining the normal Fe and Na layers without contamination from the sporadic events, we have excluded those profiles showing such sporadic layers from the Fe and Na data. The remaining data represent ∼970 h of Na measurements on 132 different nights in more than 5 years and ∼590 h of Fe measurements on 82 different nights in 4 years. The Na data duration per night ranges from ∼3 to ∼12 h with a mean of ∼7.3 h, while the corresponding quantity for Fe varies from ∼3 to ∼11 h with a mean of ∼7.2 h. This database allows us to characterize the seasonal variations of the Fe and Na layers at 30°N.
Figure 1 presents the weekly mean Na and Fe density profiles versus month, which have been smoothed vertically with a 1-km FWHM Hamming window. These weekly mean profiles were obtained by taking average of nightly means. The largest gap is 5 weeks (from mid-June to mid-July) for the Fe data and 2 weeks for the Na data. Following the approach used by Gardner et al. , we derive the contour plots of the Na and Fe densities versus month and altitude from these Fe and Na density profiles: The mean, annual and semiannual harmonic fits for each data set (Na or Fe) were first calculated at each altitude between 75 and 110 km. The weekly mean profiles plotted in Figure 1 were interpolated over gaps in the observations by the fits. The residuals were obtained by subtracting the harmonic fits from the weekly mean density profiles. After the residuals were smoothed by a Hamming window with a FWHM of 8 weeks and a resolution of 1 week, they were then added back to the mean plus annual and semiannual fits. The results are shown in the form of month/altitude Na and Fe density contours (see Figure 2). According to the two contour plots, the maximum and minimum peak Na densities are 2550 and 1560 cm−3, which occur respectively at 91.0 and 90.5 km in the 43th and 25th week of the year, while the corresponding values for Fe are 9050 and 4360 cm−3 occurring, respectively at 88.1 and 87.2 km in the 49th and 20th week. As a whole, the seasonal variations of the Na and Fe layers at 30°N exhibit some similarities. For instance, both the Na and Fe densities maximize in winter and minimize in summer. In addition, as seen from the Figure 2, both the Na and Fe layers generally show an evidently steeper density gradient on the underside than on the topside, and the borders of the Fe layer are clearly steeper than those of the Na layer. The two features are consistent with those shown in the individual Na and Fe profiles measured simultaneously [Yi et al., 2008].
 In order to further account for the seasonal variations of the Na and Fe layers, we calculated the nightly averages of their column abundance, centroid height, and RMS width as well as the mean plus annual and semiannual harmonic fits to the three layer parameters following Kane and Gardner . The three parameters are defined by the spatial moments of the layers [Gardner et al., 1986]. The calculated results are shown in Figures 3 and 4, where circles denote nightly mean values and the thick curves stand for the harmonic fits (mean+annual+semiannual) to the layer parameters with the dashed curves being 1 standard deviation above and below the fit. The primary characteristics of the harmonic fits to the Na and Fe layer parameters are summarized in Table 1. As seen in Figures 3a and 4a, the nightly averaged Na column abundance varies from 1.1 × 109 to 6.5 × 109 cm−2 with an annual mean of 2.5 × 109 cm−2, while the Fe column abundance ranges from 2.4 × 109 to 22.3 × 109 cm−2 with an annual mean of 7.5 × 109 cm−2. The average Fe/Na abundance ratio is ∼3. The Na and Fe layer column abundances are dominated by annual oscillations with similar phases (the phase difference is 7 days), whereas moderate semiannual oscillations with similar phases (the phase difference is 15 days) are also visible. Hence they both show a similar seasonal behavior with maxima in winter and double minima elongated from late spring to midautumn. Noting that there is only one night of Fe measurement between mid-May and mid-July, the fitted double minima for the Fe column abundance (from late spring to midautumn) should be verified by more midsummer Fe data. The seasonal ratios are respectively 1.6 for Na and 2.4 for Fe. The Na centroid height (Figure 3b) displays a variation ranging from 89.2 to 93.6 km and has a mean of 91.4 km, while the Fe centroid height (Figure 4b) is between 86.5 and 90.7 km with a mean of 88.7 km. The variations of the Na and Fe centroid heights are mainly semiannual, with lowest heights in early summer. The RMS width of the Na layer (Figure 3c) varies from 3.4 to 5.9 km with a mean of 4.5 km, while the Fe RMS width (Figure 4c) is between 2.9 and 5.7 km, and has a mean of 4.1 km. The Na RMS width exhibits a strong semiannual oscillation with the layer being slightly broader in winter, whereas the Fe width varies primarily annually with a maximum in winter. It should be mentioned that the RMS residuals from the fitted layer parameters for both Na and Fe are comparable to the corresponding annual and semiannual amplitudes. This results from large night-to-night variability in the observed layer parameters. A graphical comparison between the harmonic fits to the Na and Fe layer parameters is given in Figures 5a–5c. The Na and Fe abundances plotted in Figure 5a display alike seasonal variations because the Na and Fe oscillations associated with each harmonic have similar phases. But, the amplitudes for the Fe oscillations are apparently larger than those for the Na, which results in an uneven abundance difference between Fe and Na in different seasons. As seen in Figure 5b, the Na and Fe layer centroid heights vary nearly in the same way, whereas the Na layer is consistently ∼2.7 km higher than the Fe layer. Since the Na and Fe RMS widths are dominated respectively by semiannual and annual oscillations, the Na/Fe width ratio is different from month to month (see Figure 5c).
Table 1. Mean Plus Annual and Semiannual Fits to Na and Fe Layer Parameters Observed at 30°Na
Annual Mean A0
Annual Amplitude A1
Annual Phase d1 (days)
Semiannual Amplitude A2
Semiannual Phase d2 (days)
Function adopted is A = A0 + A1 cos [(2π/365) (d − d1)] + A2 cos [(4π/365)(d − d2)].
Na abundance (×109 cm−2)
Fe abundance (×109 cm−2)
Na centroid height (km)
Fe centroid height (km)
Na RMS width (km)
Fe RMS width (km)
3. Comparison With Observations at Other Latitudes
 In order to provide an insight into the latitudinal variation of these metal layers, here we compare the harmonic fits to the Na and Fe layer parameters observed at our site with those currently available at other latitudes. The harmonic fits to the Na layer parameters observed at 40°N and South Pole [Plane et al., 1999; Gardner et al., 2005] are reproduced in Figure 6 along with our own curves from Figure 3. As shown in Figure 6a, the Na layer column abundance at 30°N and 40°N is dominated by the annual oscillations with similar phases. Thus, as a whole, it displays a similar seasonal feature (with maxima in winter and minima in summer) at the two different latitudes. Interestingly, the fitted Na abundance presents double minima from late spring to midautumn at 30°N, in contrast to a single minimum in midsummer at 40°N. The calculated seasonal ratio is ∼1.6 at 30°N and ∼3.0 at 40°N. This is consistent with the latitudinal trend revealed by satellite measurements, that the seasonal variation of the Na layer becomes stronger with increasing latitude [Fan et al., 2007]. Note that the annual mean Na layer abundance at 30°N is ∼60% of that at 40°N. At the South Pole, the Na abundance shows a dominant annual oscillation with a phase being shifted by 6 months with respect to its counterparts at northern midlatitudes. The Na layer centroid height plotted in Figure 6b exhibits strong semiannual oscillations at 30°N and 40°N, but varies principally annually at the South Pole. The phase of the semiannual oscillation at 30°N leads that value at 40°N by about 2 months (52 days), while its amplitude is slightly smaller than that at 40°N. The seasonal variation of the Na centroid height at the South Pole is evidently stronger than the counterparts at northern midlatitudes. Although these curves shown in Figure 6b are unlike each other, the Na layer centroid height has similar annual mean values at the three different locations (91.6, 91.4, and 91.5 km respectively for 40°N, 30°N, and the South Pole). As seen in Figure 6c, semiannual oscillations with slightly different phases dominate the Na layer RMS width at 30°N and 40°N, while annual oscillation dominates at the South Pole. On average, the Na layer at 30°N is ∼0.1 km broader than at 40°N. It is narrowest at the South Pole. This result appears to coincide with that derived from satellite measurements (the Na layer tends to be broad at low latitudes) [Fan et al., 2007].
Figure 7 compares the harmonic fits to the Fe layer parameters measured at our site with those obtained at 18°N, 40°N and the South Pole [Kane and Gardner, 1993; Raizada and Tepley, 2003; Gardner et al., 2005]. As seen in Figure 7a, the seasonal variations in Fe column abundance are basically annual at all the four sites, while the semiannual oscillations are imperceptible except at 30°N, where there is a moderately strong semiannual oscillation. At 30°N and 40°N, the Fe abundance shows as a whole an identical seasonal variation with maxima in winter and minima in summer due to similar annual oscillation phases. At 18°N, a phase delay of ∼3.5 months is observed compared to 30°N, which approaches the lag value of ∼3.3 months compared to 40°N [Raizada and Tepley, 2003]. Thus the Fe abundance at 18°N shows a distinct seasonal variation (with a maximum in April/May and a minimum in September/October) in comparison with 30°N and 40°N. The noticeably large phase difference between the annual oscillations at 18°N and 30°N presents a contrast with the very small phase difference between 30°N and 40°N. Note that this result is based on the currently limited Fe lidar data at 18°N (only a total of 26 observation nights) [Raizada and Tepley, 2003]. Figure 7a shows that the seasonal variation of the Fe abundance has a tendency to strengthen with increasing latitude. The seasonal ratio is ∼2.3 at 18°N, ∼2.4 at 30°N, ∼3.2 at 40°N, and ∼4.2 at the South Pole. This tendency coincides with the latitudinal trend of the seasonal variation in Na abundance [Fan et al., 2007]. The annual mean Fe layer abundance at 30°N is smallest of the four sites, which is ∼71% of that value at 40°N, ∼77% of that at 18°N and ∼77% of that at the South Pole. According to an error analysis given by Tilgner and von Zahn , the maximum error for the derived absolute atom density comes from uncertainty in the atmospheric density at the normalizing altitude, which is generally limited to ±10% [Kane and Gardner, 1993; States and Gardner, 1999; Plane et al., 1999]. Hence, we suggest that the nighttime Fe layer has a low-abundance region around 30°N. The Fe layer centroid height varies principally semiannually at three latitudes in the Northern Hemisphere, but it shows a dominant annual oscillation at the South Pole (see Figure 7b). In addition to unlike phases, the seasonal variation in the Fe centroid height exhibits different strengths at different latitudes, being weakest at 30°N and strongest at the South Pole. The mean Fe centroid height at 30°N is respectively ∼0.7 and 0.2 km higher compared with 40°N and 18°N. As shown in Figure 7c, the RMS width of the Fe layer is dominated by annual oscillations at all the four sites. At 30°N, the seasonal variation of the Fe layer width behaves like that at 40°N, being broadest during winter and narrowest during summer. Interestingly, the seasonal variation of the Fe layer width at 30°N is nearly out of phase with respect to its counterpart at 18°N. On the basis of the results observed at the three latitudes in the Northern Hemisphere, the mean Fe layer width decreases with increasing latitude (∼5.5 km at 18°N, ∼4.1 km at 30°N, and ∼3.4 km at 40°N). This latitudinal trend is in accord with that of the Na layer width obtained by satellite measurements [Fan et al., 2007].
 As mentioned above, the Na and Fe abundances at 30°N have similar seasonal variations, and the Fe abundance variation amplitude (peak to peak) is larger than that of Na. This situation is also true at 40°N and the South Pole [Kane and Gardner, 1993; Gardner et al., 2005]. The relative strength of the Fe and Na abundance amplitudes appears to display a systematic change with latitude. On the basis of the data shown in Figures 6a and 7a, the Fe amplitude is a factor of 5.1 larger than that of Na at 30°N. The factor becomes 2.4 at 40°N and decreases to 1.8 at the South Pole. It is also noteworthy that the relation between the annual mean Na and Fe RMS widths is unlike at different locations. The Na layer is broader than the Fe layer at 30°N and 40°N, whereas a reversed situation occurs at the South Pole, where the Na layer is generally narrower than the Fe layer.
 It should be mentioned that all the Na and Fe data involved in the above comparisons represent nighttime-only observations except those at the South Pole, where lidar observations were conducted throughout a full diurnal cycle. In terms of the Na lidar observations at 40°N, States and Gardner  assessed the biases in nocturnally observed Na density and column abundance caused by diurnal variations. They found that the seasonal variations in both the Na density and column abundance had little nighttime bias. Hence, the seasonal variations of the Na and Fe layers from the full-diurnal-cycle observations at the South Pole have been adopted in the present comparisons.
4. Summary and Conclusion
 The complete seasonal variations of the nocturnal mesospheric Na and Fe layers at 30°N have been obtained from ∼5 years of Na lidar measurements and ∼4 years of Fe lidar measurements. The Na and Fe layer column abundances show strong annual oscillations as well as moderate semiannual oscillations. Since the Fe and Na oscillations associated with each harmonic have similar phases, the Na and Fe abundances display as a whole an identical seasonal variation with maxima in winter and double minima elongated from late spring to midautumn. Since only one night of Fe data is presently available between mid-May and mid-July, the authenticity of the fitted two minima for the Fe column abundance need to be verified by more midsummer Fe data. The seasonal variation in the Fe abundance is evidently stronger than that of Na. The calculated seasonal ratios are 2.4 for Fe and 1.6 for Na. The Na layer abundance has an annual mean of ∼2.5 × 109 cm−2, while the corresponding value for Fe is ∼7.5 × 109 cm−2. The two values yield an average Fe/Na abundance ratio of ∼3.0 at 30°N. The Na and Fe centroid heights are dominated by semiannual oscillations with a very small phase difference. Thus they show alike seasonal variations. The mean centroid heights are 91.4 km for Na and 88.7 km for Fe. The Na RMS width exhibits a strong semiannual oscillation with the layer being slightly broader in winter, whereas the Fe width varies principally annually with a maximum in winter. On average, the RMS widths of the Na and Fe layers are respectively 4.5 and 4.1 km.
 The seasonal variation characteristics of the Na and Fe layers observed at 30°N have been compared with those obtained at 40°N, 18°N and the South Pole [Kane and Gardner, 1993; Plane et al., 1999; Raizada and Tepley, 2003; Gardner et al., 2005]. Similar to the seasonal behavior at 40°N, both the Na and Fe layer abundances at 30°N show as a whole maxima in winter and minima in summer. The seasonal variation strength (seasonal ratio) for Fe layer is larger than for Na layer. Different from the single minimum at 40°N, both the Na and Fe abundances display double minima from late spring to midautumn at 30°N. This feature probably represents a unique seasonal behavior of Na and Fe layers in the transition region between midlatitudes and low latitudes. Both the Na and Fe seasonal ratios at 30°N are smaller than at 40°N and the South Pole. The annual mean Na and Fe abundances are about 60–77% of the counterparts measured at 40°N, 18°N and the South Pole. Noting that the absolute accuracies of the measured Na and Fe abundances are generally limited to ±10% by uncertainty in calibration constants, our results suggest that nocturnal Na and Fe layers have a low-abundance region around 30°N. Both the centroid heights and RMS widths of the Na and Fe layers at 30°N exhibit smaller seasonal variations than the counterparts at all other latitudes. Compared to the observations at other latitudes, the difference in the annual mean centroid height is only 0.1–0.2 km for Na and 0.2–0.7 km for Fe. On the basis of the results observed at the three latitudes in the Northern Hemisphere, the annual mean Fe layer width decreases with increasing latitude.
 The current lidar observations present the features of the Na and Fe layer seasonal variations at the border between midlatitudes and low latitudes. The further confirmation and explanation of these features, such as the double minima of the Na and Fe column abundances from late spring to midautumn, very small annual mean column abundances, and weaker seasonal variations, require further observational and modeling efforts. The comparisons with lidar measurements at 40°N, 18°N and the South Pole provide a very preliminary outline of the Na and Fe layer latitudinal variations. As pointed out by Kane and Gardner , for fully understanding the latitudinal nature of the mesospheric metal layers, a baseline study to establish their seasonal variations at various latitudes is required. In particular, at present more observations at low latitudes in the Northern Hemisphere are urgently necessary to clarify the puzzling transition of the metal layer seasonal variations in this region.
 This research is supported jointly by the National Natural Science Foundation of China through grants 40674085 and 40731055 and the Program for Changjiang Scholars and Innovative Research Team in University (PCSIRT0643). The authors thank the anonymous referees for their valuable suggestions and helpful annotations.