On the relationship of energetic particle precipitation and mesopause temperature

Energetic particle precipitation (EPP) has the potential to change the neutral atmospheric temperature at the mesopause region. However, recent results are inconsistent leaving the mechanism and the actual effect still unresolved. In this study we have searched for electron precipitation events and investigated a possible correlation between D region electron density enhancements and simultaneous neutral temperature changes. The rotational temperature of the excited hydroxyl (OH) molecules 5 is retrieved from the infrared spectrum of the OH airglow. The electron density is monitored by the European Incoherent Scatter Scientific Association (EISCAT) Svalbard Radar. We use all available experiments from the International Polar Year (IPY) in 2007–2008 until February 2019. Particle precipitation events are characterized by rapid increases in electron density by a factor of 4 at an altitude range of 80–95 km, which overlaps with the nominal altitude of the infrared OH airglow layer. The OH airglow measurements and the electron density 10 measurements are co-located. Six of the ten analysed electron precipitation events are associated with a temperature decrease of 10–20 K. Four events were related to temperature change less than 10 K. We interpret the results in terms of the change in the chemical composition in the mesosphere. Due to EPP ionisation the population of excited OH at the top of the airglow layer may decrease. As a consequence, the airglow peak height changes and the temperatures are probed at lower altitudes. The observed change in temperature thus depends on the behaviour of the 15 vertical temperature profile within the airglow layer. This is in agreement with conclusions of earlier studies, but is, for the first time, constructed from electron precipitation measurements as opposed to proxies. The EPP related temperature change recovers very fast, typically within less than 60 minutes. We therefore further conclude that this type of EPP event reaching the mesopause region would only have a significant impact on the longer-term heat balance in the mesosphere if the lifetime of the precipitation was much longer than that of an EPP event (30–60min) found in this study. 20 Copyright statement.

boundary of the neutral atmosphere, the behaviour of neutral gas and ionized particles differ, which is why complex interactions between dynamics, photochemistry, heating and transport mechanisms take place and the atmospheric energy budget can be altered. in October-November was found. The strongest temperature increase of 15-20 K in May-June at the altitudes of 110-115 km was reported due to high fluxes of 30-80 keV protons. As particle precipitation events change the Pedersen conductivity, Joule heating may contribute to the increase in temperature at these altitudes. Further down in the atmosphere, at 85-90 km, only a minor cooling of 3-4 K during periods of high Kp values was observed due to precipitation of 250-800 keV protons (an intense solar proton event). 35 The rotational hydroxyl (OH) airglow temperature was observed during six nights of the austral winter in 2008 by Suzuki et al. (2010). These nights were selected to include high auroral activity and clear weather. In their study, the temperature was derived from spectra recorded at Syowa Station in Antarctica by a high-sensitivity spectrometer. During only one of these nights (27-28 March 2008), an increase in the temperature of 10 K over a time period of 15 minutes was observed. Furthermore, a decrease in the relative intensity of the OH(8-4) Q branch of ∼23% was found by comparing the pre-EPP level to that half an 40 hour after the deepest magnetic deflection. No such coherent behaviour was seen during the other nights. Suzuki et al. (2010) suggested a relationship between EPP and the OH temperature based on measured disturbances in the horizontal magnetic field and variations in the cosmic radio noise absorption (CNA) over the course of several hours of activity. The average energy of the precipitating electrons during this night reached 10-20 keV. They discussed different mechanisms for causing the change in the temperature. Joule heating was concluded to not contribute much, as the estimated heating rate required to explain the 45 observed temperature increase at the mesopause height was three orders of magnitude higher than earlier particle precipitation observations suggested. Direct particle heating, on the other hand, is produced by precipitating particles colliding with the atmospheric neutrals. This heating process is dependent on the incident particle energy and their deposition altitude. Similarly to the Joule heating, the authors estimated the energy deposition rate of EPP at the mesopause height, and concluded that the high-energy particle flux required to explain the temperature change was unrealistically high. Atmospheric gravity waves were 50 also excluded, as the intensity of the OH airglow and the rotational temperature did not show a positive correlation, which is characteristic for dynamics driven conditions, as described and modelled by Cho and Shepherd (2006). Suzuki et al. (2010) further discussed the possibility of a change in the height distribution of the OH airglow emission during auroral events. The initial profile of OH volume emission rates retrieved by the SABER instrument onboard the TIMED satellite was compared to an example of the disturbed layer. The comparison showed a decrease in the upper part of the disturbed layer as the thickness 55 of the OH layer had decreased by 20%. If a change in the height distribution of the OH airglow emission occurred during an auroral event, an increase in temperature is not necessarily observed, as the outcome would also depend on the temperature gradient in the mesosphere.
The connection between the geomagnetic activity and the long-term temperature at the mesopause region during the solar cycle 23 and 24 was studied by Gavrilyeva and Ammosov (2018). The OH rotational temperature was measured by the ground-60 based infrared spectrograph at Maimaga station (63 • N,129.5 • E), and ascribed to an altitude of 87 km, which is commonly assumed to be the peak height of the OH layer. The seasonally averaged temperatures from 1999 to 2015 were included in the analysis. The results showed that the mesopause temperature from October to February is about 10 K higher during the years with high solar activity (Ap > 8) than during low activity years (Ap ≤ 8). A solar activity dependence of OH airglow temperatures was also reported by Holmen et al. (2014). They concluded on a temperature change of about 4 K per 100 solar flux units (SFU) of the F10.7 radio flux. The question on the detailed relationship of EPP and the mesopause temperature was still left open.
The purpose of this study is to characterize the effects of the EPP on the mesopause temperature in more detail, on a larger set of events, in higher temporal resolution and using co-located measurements of precipitating electrons and the mesospheric temperature. The instrumentation is further described in section 2. Section 3 outlines the data used in this study, as well as the 70 analysis of the EISCAT Svalbard Radar data. Finally, the results are shown and discussed in sections 4 and 5. The conclusions of the effects of the EPP on the mesopause temperature are summarized in section 6.

Instrumentation
Following Cresswell- Moorcock et al. (2013) we use the European Incoherent Scatter Scientific Association (EISCAT) radar data to identify electron precipitation events as electron density enhancements. Simultaneous and co-located neutral tempera-75 ture measurements are determined using the spectrometer measurements of the OH airglow. The derived rotational OH temperature is taken as the neutral temperature of the mesopause region, assuming that the excited OH molecules are in thermal equilibrium with the ambient atmosphere.

EISCAT Svalbard Radar
For this study the EISCAT Svalbard Radar (ESR, Wannberg et al. (1997)) in Longyearbyen, Norway (situated at a geographic is used. The radar operates at the 500 MHz band and has a 32 m steerable parabolic dish antenna and a 42 m fixed parabolic antenna aligned to the local geomagnetic field. For our purpose, we searched for experiments with good height resolution at D region altitudes. The manda experiment (Tjulin, 2017) resolves altitudes of 80 to 100 km with 1-2 km height resolution and was therefore chosen for the radar campaigns in January and February 2019 for this study. In addition to manda, all previously 85 run ipy experiments were analyzed, as that experiment also covers the mesopause region with a vertical resolution of 4-5 km.
The manda experiments utilize the 32 m dish and the ipy data are collected on the 42 m dish.

Ebert-Fastie airglow Spectrometer
The Ebert-Fastie spectrometer at the Kjell Henriksen Observatory (KHO) in Longyearbyen, Svalbard, is used to retrieve the winter temperature of the mesopause (Sigernes et al., 2003). The observatory is located only a kilometer away from the radar 90 site, so it is practically co-located. The spectrometer scans the near infrared wavelength region, from 824 to 871 nm, which includes the rotational OH(6-2) band of the OH airglow. The spectrometer points to the zenith with a field of view of 5 degrees. The spectral resolution of the OH(6-2) band is 0.4 nm. One scan of the wavelength range takes 25 seconds but to obtain a good signal-to-noise ratio several scans are averaged during post-processing of the data. Most earlier studies use 1hour averages. In this study, half-hour averaging is used for better temporal resolution. The spectrometer measures whenever 95 the Sun is more than 12 degrees below the horizon, which at Svalbard latitudes (78.2 • N) gives an optical observation season from the beginning of November until the end of February. The rotational OH temperatures are obtained by fitting a synthetic spectrum to the measured band of emission lines and using the intensities of four different emission lines of the best fit to determine the temperature. The details of this procedure are presented in Sigernes et al. (2003). The accuracy of the obtained rotational temperature is estimated to be ±2 K. In addition to the fitted temperature values, we use relative band brightness 100 (non-calibrated arbitrary units) of the P branch of the OH(6-2) transition. These intensity values are routinely calculated and saved together with the rotational temperatures.

Data description and event selection
A total of 10220 hours of ESR data were initially inspected. The ipy and manda experiment contribute 10144 hours and 76 hours of data respectively. These experiments provide a sufficient height resolution to detect enhanced electron densities at the 105 mesopause. Here the mesopause region is defined from an altitude of 80 to 100 km to include the hydroxyl (OH) peak height at 76-90 km (Mulligan et al., 2009). Experiments run in field aligned direction (42m: ipy, 32m: manda) or vertical pointing manda mode was run for a total of 48 hours in January (24 hours) and February 2019 (another 24 hours) to specifically collect data for this study. The experiment was run for six evenings between 16 to 22 UT (19 to 01 MLT).
The EISCAT raw data files (Auto-Correlation Functions (ACFs)) are analysed using the GUISDAP data package M. S. Lehti-115 nen (1996). This provides an iterative fitting of the ACFs and then, using a model ionosphere, produces the electron densities examined in this study. The post-integration time is 60 s. Data gaps are shown by white areas in the plots (Figures 1 and 2) and occur where GUISDAP has not managed to provide a fit. 10-minute averaged electron densities are calculated for the altitude ranges of 87-90 km and 91-94 km separately. Data from higher than 94 km are excluded. The error of the electron density is averaged to give a mean error for further analysis. The search for the EPP events in the radar data is based on an earlier study by Cresswell-Moorcock et al. (2013). The onset is found by a sudden increase of the electron density (median value) by a factor of 5 over 5 minutes. For this study the electron density is averaged to 10 min resolution with an altitude resolution of 4 km. This criterion was slightly adapted to better suit the data from the two different EISCAT experiments. The previous study's factor of 5 was attempted but since there were a number of events just below this acceptance threshold the factor was adjusted to 4. We searched for an increase of electron 125 density by a factor of 4 within 20 minutes from the precipitation onset. If onsets are found only 10 minutes apart, the latter onset is ignored.
The events are detected automatically but inspected and sorted visually based on the temporal evolution of the electron density and the OH temperature. Events with large electron density errors of more than 5 × 10 9 m −3 are excluded. Events which lack more than one temperature measurement within a ± 3-hour analysis window from the event onset are also excluded.
130 Events which show a particle precipitation signature of electron density enhancement through the whole ionospheric column (87-126 km) are sorted into the EPP category. Events which show an electron density enhancement limited to horizontal layer at the bottom of the E region (around 100 km) are categorized as sporadic E layers (Rapp et al., 2011). The remaining events not showing clear EPP or sporadic E layer behaviour are excluded from the analysis. The sporadic E layer events are not discussed further in this paper. The EPP events with a full coverage of the temperature data, and those only missing one temperature data 135 point are analyzed in more detail. This selection results in a set of eight EPP events, which where categorized into groups of increasing, decreasing and stable (no change) temperature evolution over the EPP onset.
The electron density during the radar run on 6 January from 16:00 UT to 22:00 UT is shown in Figure 1 (top panel). The onset of the EPP is at 19:50 UT and the event lasted until 20:20 UT. The electron density at the lower part of the ionosphere during this experiment was low (mainly below 10 10 m −3 ) but abruptly increased at the EPP onset time. The OH rotational 140 temperature (bottom panel) does not change at the event onset. However, it experiences a total decrease of 19 K half an hour after the EPP onset time.

Results
Our eight EPP events are listed in Table 1. The criterion for a decreasing or increasing mesopause temperature is that the change has to be larger than the standard deviation (STD) of the temperatures averaged over half an hour. An event is classified 145 as stable (no change), if the temperature change is within the STD of the averaged temperatures.     In the line plots the electron densities are from the lowest height range where the increase by a factor 4 was detected. Brief 150 descriptions of each event are given in the following paragraphs.
The first event commences at 23:30 UT on 29 December 2007 and is detected in the altitude range of 91-94 km. The electron density at the EPP event start time is 1.0×10 10 m −3 . The electron precipitation lasts for about 30 min. The mesopause temperature decreases by 55 K (from 234 to 179 K) over the EPP onset time, and recovers within half an hour. The relative OH(6-2) band intensity is 148 before the event and only slightly diminishes to 133 over the EPP onset, but recovers within 155 about 30 minutes similarly to the temperature.
The second event starts at 02:50 UT on 28 February 2008. The electron density enhancement is detected in the height range of 91-94 km with an onset time value of 1.2 × 10 10 m −3 . The precipitation lasts for about 40 min and is followed by another increase lasting for about 50 min (from 04:00 until 04:50 UT). An airglow temperature decrease of 20 K (from 227 K to 207 K) is observed at the EPP onset time. There is only one temperature data point available after the event onset, suggesting 160 an increase of 10 K an hour after the EPP onset. The relative OH(6-2) band intensity decreases gradually from 2 hours before the event until the event onset time. Thus, during this event the intensity behavior is much smoother than that of the temperature.
Little or no correlation is seen between the two parameters.
The third event is detected at 22:50 UT on 7 February 2011 at the altitude range of 91-94 km. The electron density at the EPP onset is 1.8 × 10 10 m −3 and the enhancement lasts for about 30 min. The mesopause temperature decreases from 205 K 165 prior to the EPP to 187 K at the onset time (18 • ). The minimum temperature is measured at the time of the electron density maximum. Again, the temperature change recovers within 30 minutes. There is a mild decrease in the relative OH(6-2) band intensity (of 80 arbitrary units) during this event, but the minimum of the emission intensity is reached about half an hour after the temperature minimum, when the temperature and the electron density changes have already recovered.
The fourth event starts at 21:50 UT on 6 December 2013. It is detected in the altitude range of 87-90 km with an electron 170 density value of 9.9 × 10 9 m −3 . The temperature decreases by 14 K (from 210 to 196 K) over the event onset time. Both temperature and electron density values recover within 30 minutes. The relative band intensity peaks at 1319 an hour before the onset, decreases to 1011 half an hour before the onset and minimizes at 864 at the electron density maximum. A positive correlation between the temperature and band intensity is found in this case.
The fifth event starts at 02:10 UT on 24 January 2014 at an altitude range of 87-90 km. The electron density at the pre-175 cipitation onset is 9.5 × 10 9 m −3 , and stays slightly elevated for about two hours after the initial enhancement, which only lasts for about 30 minutes. The mesopause temperature undergoes a decrease of about 12 K, from 210 K prior to the event to 198 K at the EPP onset. The temperature recovery follows that of the initial enhancement in electron density. The OH(6-2) band intensity decreases from 1890 to 1551 already an hour prior to the EPP onset. Similar to the temperature evolution, the emission intensity recovers within half an hour after the onset. There is a mild positive correlation between the temperature 180 and band intensity values.
The sixth event commences at 15:20 UT on 24 January 2014 in the altitude range of 87-90 km. The electron density enhancement is seen in a single profile only (10 min lifetime), with the value of 1.5 × 10 10 m −3 . The mesopause temperature undergoes mild fluctuations, but all changes are well within the errors and thus, this event is classified as stable. The relative OH band intensity strongly increases (from 460 to 1079) from one hour before to half an hour before the event. After the event onset 185 a more steady and gentle increase in the emission intensity is seen. A more persistent mesospheric electron density enhancement below 95 km may refer to an evolution of sporadic E layer which can affect the temperature evolution. Furthermore, the short lifetime of the electron density enhancement is a good candidate to explain the temperature stability.
The seventh event begins at 02:40 UT on 25 January 2014. It is detected at an altitude range of 87-90 km with an electron density value of 2.6 × 10 10 m −3 . The electron density enhancement is strong for about 20 min and remains at an elevated level 190 for the next couple of hours. The mesopause temperature undergoes a minor decreases of 10 K (from 221 K to 211 K) over the event onset. The OH band intensity values keep constant with respect to the intensities an hour before the EPP event.
The eighth event starts at 19:50 UT on 6 January 2019 and is detected in the altitude range of 91-94 km. The electron density at the EPP onset is 1.3 × 10 10 m −3 , and the enhancement lasts for about 30 minutes. The mesopause temperature does not initially decrease when comparing the level before the precipitation to that when the electron density is first enhanced.

195
Nevertheless, there is a decrease of 19 K (from 196 K to 177 K) at about half an hour past the event onset. The relative OH(6-2) band intensity at the event onset is 54 units higher than the level before the event onset time, but stays at a constant level throughout the analysis window.
An overall temperature response to the EPP onsets is shown in Figure 3 by a scatter plot (upper panel) and a superposed epoch analysis (lower panel). The scatter plot includes simultaneous temperature and electron density value pairs around the density and the airglow temperature with a large range of variation. The average temporal evolution of the temperature response is seen in the superposed epoch analysis from 3 hours before to 3 hours after the precipitation onset. The zero epoch time (EPP onset) shows the lowest temperature both in the median (blue) and the lower percentile (lower red) curves. This immediate temperature decrease is of the order of 10 degrees and recovers within half an hour after the EPP onset. The upper percentile 205 does not show a clear signature of a temperature decrease, which may be due to slightly different timing of the minima as well as the averaging over a time comparable to the temperature change (30 min).

Discussion
We have found and analyzed eight electron precipitation events which reached the mesopause and had a good coverage of co-located OH airglow temperature data. Table 1 shows that 7 out of 8 EPP events analyzed in this study were accompanied 210 by a decrease in the mesopause temperature by 10-50 K at the EPP onset. The pre-EPP temperature level (temperature values 1 hour before the event) varied between 196 and 234 K, and the temperature decreased to the range of 177-222 K.
At high latitudes the optical observation season is limited to the four winter months (Nov-Feb). The winter mesopause is characterized by a large temperature variability (of the order of 10 K, Suzuki et al. (2010)) due to gravity wave propagation.
It is thus understandable that there is no consensus of the temperature response to the energetic particle precipitation when 215 any minor change is likely to be lost in the highly variable background conditions. The temperature responses found in this study, however, are larger than 10 degrees. The short lifetime (∼30 minutes) of the observed temperature responses does not allow this signature to be detected in any analysis utilizing hourly or daily averaged temperature data. For instance, the fast temperature decrease and equally quick recovery shown by the superposed epoch analysis (bottom panel of Figure 3) would not be a significant change in hourly or daily averaged epoch.

220
The first event at 23:30 UT on 29 December 2007 (top panel of Figure 2) shows a strong decrease in the mesopause temperature at the time of the EPP onset. Similarly, in the third and fourth events (at 22:50 UT on 7 February 2011, and at 21:50 UT on 6 December 2013) a very quiet ionospheric conditions (low electron density) are seen prior to the EPP onset, and the connection between the electron density enhancements and the temperature changes is particularly clear. This is a typical behavior in our set of events. It is further demonstrated by the anti-correlation between the airglow temperature and the electron density 225 seen in the top panel of Figure 3.
The lifetime of the EPP events in this study is typically 20-50 minutes, and the corresponding neutral temperature response is equally short. In the second, fifth, and seventh event the short-lasting electron density enhancements are followed by longer lasting but gently elevated levels of electron densities. In all these cases the electron fluxes reaching the E region heights (above 100 km) are high for a couple of hours after the EPP event, but only at the very beginning of the precipitation event did they 230 reach the altitudes below 100 km. Consequently, the temperature behaviour is smooth and steady apart from the short-term change around the EPP onset.
Our sixth event (at 15:20 UT on 24 January 2014) was classified as stable in terms of the mesospheric temperature response.
The electron density increase in this event has the shortest lifetime of all our events, as it is seen in one 10-min profile only. response is seen.
A statistical study on high-latitude OH airglow temperatures and emission intensities by Shepherd et al. (2007) shows a strong positive correlation between the two parameters in the time scales from hours to seasons. This is explained by vertical motion of the airglow layer driven by atmospheric dynamics. For instance, as an airglow layer undergoes downward motion the adiabatic heating increases its temperature. The lower peak emission height coincides with higher mixing ratio of oxygen 240 and therefore, enhances the production of the excited OH. The temperature changes observed in our study typically take place in shorter time scales. The correlation between the OH(6-2) temperature and the relative band intensity of 30-minute averaged data can be visually inspected in Figure 3. While a positive correlation can be seen between the two parameters in case of the fourth and fifth event, no significant correlation across the entire event set was found (data not shown).
An increase in the mesospheric temperature during particle precipitation would agree with the Joule heating effect suggested 245 by earlier studies (Nesse Tyssøy et al., 2010). In fact, a temperature increase of about 10 K was observed in the study by Suzuki et al. (2010). In our study, however, it was found that the mesopause temperature typically responds to the particle impact with a decrease of about 20 K. A way to explain the temperature change Suzuki et al. (2010) discussed was that the EPP ionisation changes the mesospheric chemical composition by decreasing the population of excited OH at the top of the layer. As a consequence, the peak height of the airglow changes and the temperatures are probed at lower altitudes than before.

250
It was suggested that the energetic electron impact can dissociate oxygen and ozone molecules (Maeda, 1967). When less O 3 is available, less excited OH molecules are produced as ozone is a key ingredient in the production of excited hydroxyl: where v corresponds to the upper vibrational level of the OH molecule, which in our case is 6. The dissociation of molecular oxygen and ozone by energetic electrons can therefore lead to a decrease in the emission of the OH airglow. The fitted rotational 255 OH temperature corresponds to the average height of the airglow layer. The peak is assumed to be at about 87 km. If, however, the production of excited OH is temporarily prohibited at the top part of the airglow layer, the temperature will then represent the layer, which is now centered at lower heights. Depending on the local gradient in the mesospheric temperature profile, this may lead to increased or decreased temperature value. In this scenario, the relative OH(6-2) band intensity would decrease as the airglow layer becomes thinner, which is true for most of the events analyzed in this study. In particular at the top of the 260 mesosphere the temperature can vary on the order of 10 K over a height range of a few kilometers. Even if the temperature change produced in this scenario is of the right order of magnitude, it is interesting that all EPP events analyzed in this study, which did cause a temperature change, showed a decrease. Our results together with the previous results by Suzuki et al. (2010) present an inconsistent temperature response to EPP. Therefore, a larger number of events should be collected and examined to conclude if the OH airglow layer favors a height region with a positive temperature gradient upwards. Furthermore, 265 an immediate temperature response and its fast recovery suggests that the longer-term and larger-scale heat balance in the mesosphere is little affected by EPP, unless the actual precipitation has a substantial lifetime (hours to days).
A total of 10220 hours of electron density measurements were browsed in the search of enhancements due to energetic electron precipitation (EPP) events with simultaneous temperature calculations from OH airglow measurements. A total of eight 270 events of electron density enhancements were found and analyzed in this study. We investigated the existence of any coherent behaviour between the electron density enhancements at the D region heights and the mesopause temperature. The response of the mesopause temperature on the EPP energy deposition is predominately (7 out of 8 events) an immediate decrease of 10-50 K, which recovers within 30 to 60 minutes after the EPP onset. The temperature decrease found in this study, together with a temperature increase in a previous study is interpreted as an EPP ionisation decreasing the production of the excited 275 OH at the top of the airglow layer. As a consequence, the airglow layer becomes thinner, the peak height is reduced and the airglow temperatures correspond to lower altitudes. Investigating the change in the relative OH(6-2) band intensity shows a decrease during the majority of our EPP events and thus supports the thinning scenario as a valid mechanism for changing the measured temperature. Furthermore, the relative OH brightness values are only poorly correlated with the temperatures in the time scales of a few hours, which is not in agreement with purely dynamically driven temperature changes. Given the 280 short-lived characteristic of atmospheric temperature change, EPP may not have climate effects except for long-lasting events.
Data availability. The temperature data are available as quicklooks plots online at kho.unis.no. The Eiscat Svalbard Radar data have been downloaded from https://eiscat.se and are the intellectual property of the EISCAT Scientific Association. The EISCAT raw data files are analyzed by using the Grand Unified Incoherent Scatter Design and Analysis Package (GUISDAP) M. S. Lehtinen (1996) .

285
Author contributions. Florine Enengl carried out the analysis of the data and the writing of the paper. Noora Partamies proposed the idea of the study, fitted and pre-analyzed the airglow temperature data, shared her expertise and together with Nickolay Ivchenko took part in the discussions, interpretations, planning and structure of the work. Lisa Baddeley advised and helped describing and analysing the EISCAT data. All co-authors helped in the writing process with comments, suggestions and edits on the paper.
Competing interests. No competing interests are present.

290
Disclaimer. The data and figures have been used in the MSc thesis by F. Enengl, available through http://kth.diva-portal.org/