Statistical climatology of mid-latitude mesospheric summer echoes characterised by OSWIN radar observations

Mid-latitude mesospheric summer echoes (MSEs) appear in radar observations during summer months. The geophysical factors controlling the formation of MSEs include solar and energetic particle ionisation, neutral temperature, turbulence, and meridional wind transport. 12 years of summer month observations with OSWIN radar in Kühlungsborn, Germany have been analysed to detect MSE events and to analyse statistical connections to these controlling factors. A more sensitive and consistent method for deriving signal-to-noise ratio has been utilised. Daily and monthly composite analysis demonstrates 5 strong daytime preference and early summer seasonal preference for MSEs. The statistical results are not entirely conclusive due to the low occurrence rates of MSEs. Nevertheless, it is demonstrated that the meridional wind transport from colder highlatitude summer mesosphere is the important controlling factor, while no clear connection to geomagnetic and solar activity is found.

The MSE detection is based on a global sky noise estimate derived from all range gates between 50 to 120 km altitude where the truncated Gaussian fitting resulted in a SNR < -8 dB for each record. The global sky noise estimate allows to define a new SNR value for each record, which is then used further on to detect potential MSE events.
The available dataset has been processed to obtain hourly values of SNR, only for summer months (June-August). The detection of MSE is defined when the observed hourly value of SNR exceeds the SNR threshold. Note that the definition of 5 SNR used in this study differs from the definition of SNR used in Gerding et al. (2018), and in earlier MSE studies with OSWIN (Bremer et al., 2006;Zeller et al., 2009). In the analysis, once the hourly SNR value exceeds the threshold, the maximum SNR value in a vertical column is considered as the SNR of detected MSE and the altitude of the maximum SNR value is considered as a height of the detected echo.
OSWIN radar does not provide continuous wind measurements. Meridional wind data for this study has been obtained from 10 the VHF meteor radar located in Juliusruh, Germany (54 • N, 13 • E). While the meteor radar is not co-located with OSWIN, the zonal distance of 120 km is within a field of view of the meteor radar. The meteor radar provides hourly averaged wind values in the range of altitude between about 75 and 110 km using backscatter from meteor ionisation trails (e.g., Stober et al., 2017).
For the statistical study the meridional wind values are taken at the altitude closest to the height of detected MSE.
To analyse the effects of geomagnetic activity, the planetary 3-hour-range index (Kp) is used to characterise the global geo-15 magnetic activity and the auroral electrojet (AE) index is used to characterise the geomagnetic substorm-related activity, such as auroral particle precipitation and ionospheric Joule heating. The solar wind speed, measured in-situ on a spacecraft upstream from the Earth's magnetosphere (King and Papitashvili, 2005), is used as a proxy for the energetic electron precipitation from the Earth's outer radiation belts. The solar activity is characterised by the solar radio flux at 10.7 cm wavelength (F 10.7). For each hour of OSWIN observations the values of AE, Kp, V sw , and F 10.7 are assigned using the NASA's OMNIWeb database.

20
To avoid a contamination of the statistics by very small counts during extreme events, a threshold of ≥ 5 MSE detections is used here.
3 Statistical results and discussion

Daily and monthly occurrences of MSEs
Superposed epoch analysis of the entire OSWIN dataset for each day of summer months has been done with respect to 00 25 UTC time. The total occurrence of MSEs within each UTC hour, computed in percentage with respect to the total hours of observations, is shown in Panel A of Figure 1. Panels B, C, and D of Figure 1 show occurrences of MSEs during months of June, July, and August, respectively. As expected, there is clear day-time preference for MSEs, though there is noticeable occurrence of MSEs during night hours as well.
In terms of monthly distribution, the maximum MSE occurrence is observed during June, with the occurrence diminishing 30 in August. Such strong MSE preference for early summer, also noticed but unexplained by Bremer et al. (2006), is not seen in long-term observations of PMSEs at polar latitudes (Bremer et al., 2006;Latteck and Bremer, 2017). In addition to the summer solstice conditions, the early summer preference could be attributed, at least in part, to the meridional wind climatology. Figure  2 shows the 10-year climatology of mean meridional winds observed by Juliusruh meteor radar (extracted using the procedure described by Stober et al. (2017)), indicating that the southward (negative) winds at the relevant altitudes (∼ 85 km) are strongest in late June -early July (see further discussion is Section 3.3). The occurrence of MSEs, as a function of height and UTC, is shown in Figure 3. The predominant altitude of MSEs of ∼ 85 km is consistent with the earlier statistical analysis based on 3 years of OSWIN observations (Zecha et al., 2003). 5 Average occurrences of MSEs during each summer month of observations is shown in Figure 4 for all 12 years of OSWIN observations. While the MSE occurrence is always highest in June, there is noticeable variability from year to year, with some years (especially 2006) having smaller difference in occurrences between June and July. The reason for this variability is not clear, as it could be due to solar cycle effects, year-to-year variability in atmospheric circulation, and/or variable planetary wave activity. It has been suggested earlier that the increased planetary wave activity could play a role in the formation of MSEs by 10 affecting the meridional transport (e.g., Zeller et al., 2009). However, an investigation of this possible connection is beyond the scope of this paper, as it would require a rigorous analysis using global atmospheric circulation models.

Dependence on geomagnetic and solar activity
Dependencies of MSEs on geomagnetic and solar activity (AE, Kp, V sw , and F 10.7) for the entire OSWIN dataset are illustrated in Figure 5. In Panels A-D of Figure 5, the number of counts of MSE observations in a specific bin is shown by red 15 bars, as well as the number of total observation hours, divided by a factor of five (blue bars). It is clear that the distribution of counts with detected MSEs resembles the distribution of total counts, indicating weak or no dependence on geomagnetic and solar conditions. A distribution of MSE occurrences is shown in Panels E-H of Figure 5. No clear dependence on the solar radio flux is seen, though there is a positive correlation with Kp values (masked by low count threshold, as discussed below), which is generally consistent with earlier studies (Bremer et al., 2006;Zeller and Bremer, 2009).

20
The relevance of planetary Kp index for the formation of (P)MSE, suggested in earlier studies Bremer et al. (e.g., 2006), is not obvious as the ionisation of ionospheric D layer requires high energy of precipitating particles, e.g., electrons with energies above 50 keV would be needed to produce substantial ionisation at altitude of ∼ 85 km (Turunen et al., 2009). While strong geomagnetic storms could produce these energies of precipitation at polar-and sometimes at mid-latitudes, such events are rare and have spring/autumn seasonal preference (e.g., Gonzalez et al., 1994), and thus should have only marginal impact on 25 the MSE statistics. In this study the effects of strong storms are manifested by higher occurrences of MSEs at Kp > 7, but they do not appear in Figure 5 due to the low event count (< 5 events). Energetic precipitation (100 keV and above) from the Earth's outer radiation belts are known to be correlated with high speed solar wind intervals (Mathie and Mann, 2001;Meredith et al., 2011), which are frequent recurring phenomena and not necessarily manifested by high Kp (Tsurutani et al., 2006). Thus the solar wind speed (V sw ) is expected to be a better proxy for the high energy precipitation effects. In addition, of MSEs at V sw > 500 km/s, i.e. under the solar wind regime dominated by fast solar wind (Bothmer et al., 2007). A weak connection to the geomagnetic activity is likely to be related to: (a) low occurrence rates of MSEs leading to poor statistics, not well representing geomagnetically disturbed periods; and (b) mid-latitudes being too remote to be directly affected by geomagnetic activity, with an exception of energetic particle precipitation from the outer radiation belts (which may also explain relatively stronger correlation with V sw ). Taking (a) and (b) into account, it is advisable to conduct a statistical study 5 using radar observations of PMSEs at high latitudes, following the methodology described here, instead of using the Kp index alone as in earlier studies (e.g., Bremer et al., 2006;Latteck and Bremer, 2017).
It needs to be mentioned that the connections between radar-observed mesospheric echoes and D-region electron density enhancements during geomagnetic disturbances remain controversial. While it is generally accepted that (P)MSEs require  controlling the NLC occurrence at mid-latitudes. Thus it is reasonable to assume that the meridional wind advection plays the key role in the formation of MSEs, also contributing to the observed early summer preference for the MSEs.

Summary
12 years of summer-time observations of MSEs with the OSWIN VHF radar have been analysed to study the physical mechanisms responsible for the formation of mesospheric echoes. With the MSE detection procedure based on a global sky noise 5 estimate we were able to analyse the entire dataset with the same sensitivity threshold, though the OSWIN radar has been substantially reconfigured over the time period. The dataset is examined statistically to establish the main factors controlling the MSE occurrence, including time of a day, period of summer, geomagnetic and solar activity, and meridional wind regime.
The main conclusions can be summarised as follows.
-In agreement with earlier studies, the occurrence of MSEs is substantially lower relative to the occurrence of PMSEs 10 reported at polar latitudes. The occurrence of MSEs shows strong daylight preference, though a finite number of MSE events also appears under dark conditions. More sensitive detection method used in this study facilitates the detections of these night-time echoes. Monthly composite analysis indicates strong preferences of MSEs for early summer months, especially for June, which can be explained by a combination of the summer solstice conditions and the most favourable conditions for the southward meridional wind transport.

15
-No clear connection between the occurrence of MSEs, and solar and geomagnetic indices has been found. In addition to the dependencies on planetary Kp index, investigated earlier, we investigated the dependence on the auroral electrojet index and solar wind speed. The observed positive correlation with V sw may indicate the role of energetic electron precipitation from the outer radiation belts (associated with high speed solar streams) in the formation of MSEs. The connection to geomagnetic activity remains controversial and it needs to be investigated further using PMSE observations 20 with different radars at polar latitudes, where the occurrences of mesospheric echoes are higher.
-Clear MSE preferences for stronger southward meridional winds is indicated, especially pronounced when the meridional winds are averaged over the previous 12-24 hrs interval, i.e. when the contribution of terdiurnal and semidiurnal (dominant) tides averages out. The suggested explanation is that the colder summer mesospheric air from higher latitudes is effectively transported southwards, with MSEs forming in the process. This interpretation appears consistent 25 with satellite observations of NLC clouds and their environment, and with the previous results of the numerical modelling of NLC cloud transport.

Data availability
The OSWIN radar data and the Juliusruh meteor radar wind data are available upon request to Gunter Stober (stober@iapkborn.de). Geomagnetic indices, F 10.7 solar flux, and solar wind data were obtained from the NASA's OMNIWeb database (http://omniweb.gsfc.nasa.gov).
Author contributions. OSWIN radar raw data and meteor radar raw data were processed by GS. Statistical analysis of radar data and so-5 lar/geomagnetic activity data was conducted by DP in consultation with JLC and GS. DP prepared the manuscript with contributions from all co-authors.
Competing interests. The authors declare that they have no conflict of interest.