Winds and Tides of the Extended Uniﬁed Model in the Mesosphere and Lower Thermosphere Validated with Meteor Radar Observations

.

have the unfortunate effect of modifying the circulation in the model in an unrealistic fashion (Fomichev et al., 2002). 90 Dempsey et al. (2021) show that eCMAM generally reproduces observed diurnal tidal amplitudes in the polar regime well, and Davis et al. (2013) show that eCMAM is generally good in the equatorial regime with a trend of overestimating meridional amplitudes.
4. The Ground-to-topside model of the Atmosphere and Ionosphere for Aeronomy (GAIA; Fujiwara and Miyoshi, 2010;Jin et al., 2012, and references therein) combines three independent models: a whole atmosphere GCM, an ionosphere 95 model, and an electrodynamics model. GAIA also has a similar altitude range to WAM. Jin et al. (2012) show the ability of GAIA to model the impact of an SSW on migrating tides and the associated ionospheric response, with in general good agreement shown with Sounding of the Atmosphere using Broadband Emission and Constellation Observing System for Meteorology, Ionosphere, and Climate observations. 5. The Hamburg Model of the Neutral and Ionized Atmosphere (HAMMONIA; Schmidt et al., 2006;Meraner and Schmidt, 100 2016) is an extended version of MAECHAM5 Manzini et al., 2006), taking the upper boundary to approximately 250 km. The extended model includes important radiative and dynamical processes of the upper atmosphere and is coupled to a chemistry module containing 48 compounds.
6. The upper-atmosphere extension of ICON (Borchert et al., 2019) extends the standard ICON model so that model upper boundaries can be placed in the lower thermosphere. This includes a switch over to deep-atmosphere dynamics, as well 105 as an implementation of an upper-atmosphere physics package based on that implemented by Schmidt et al. (2006) in HAMMONIA.
In this study, we test the ability of the ExUM to model diurnal and semi-diurnal tides by comparing the seasonal variation of these tides in the model to observations of zonal and meridional winds made in the mesosphere and lower thermosphere by 155 two meteor radars. The two radars are at very different latitudes, one at the polar Antarctic site of Rothera (68 • S, 68 • W) and the other at Ascension Island (8 • S, 14 • W) in the equatorial Atlantic Ocean. The Rothera radar samples a latitude where the semi-diurnal tide is known to reach very large amplitudes but where the diurnal tide is small. In contrast, the Ascension Island radar samples a region where the diurnal tide is known to reach large amplitudes but the semi-diurnal tide is small. We use measurements of winds, tidal amplitudes, tidal phases and their seasonal variability as tests of the model's ability to accurately 160 represent these tides.
The current dynamical core (ENDGame; Wood et al., 2014) solves the non-hydrostatic, fully compressible deep-atmosphere equations of motion on a rotating sphere using a semi-implicit semi-Lagrangian formulation. The primary prognostic variables used are the three-dimensional wind components, virtual dry potential temperature 1 , Exner function of pressure 2 and dry 1 The potential temperature θ is the temperature that an unsaturated parcel of dry air would have if brought adiabatically and reversibly from its initial state to a standard pressure, p 0 , typically 1000 hPa. The virtual dry potential temperature is then the theoretical potential temperature of dry air that would have the same density as moist air. 2 The Exner function Π can be viewed as non-dimensionalized pressure and has the useful relationship that the absolute temperature T = θΠ.
6 density, whilst moisture prognostics are advected as free tracers. The discretised equations are solved using an iterative implicit 185 method -more details of which can also be found in Wood et al. (2014).
For the purposes of this case study, the horizontal resolution is fixed at 1.25 • N×1.875 • E -or the so called N96 resolution 3 .
The vertical resolution is extended from the 85-level, 85 km configuration of the standard UM using the model implementation of Griffith et al. (2020). This gives the aforementioned ExUM which builds on the standard model to extend the working height of the UM into the lower thermosphere. The work makes it possible for the Unified Model to run in a stable manner completely re-engineer the existing radiation scheme -a significant undertaking -an interim solution of relaxation :: or ::::::: nudging of the temperature field to climatological values was used. :::: This :::::: scheme :::: was ::::::::: engineered :: in ::::::::::::::::: Griffith et al. (2020) ::: and :::: more :::::: details ::: can :: be ::::: found ::::::: therein. With this addition, a stable ExUM implementation was successfully achieved with upper boundary heights of 100, 120 and 135 km. The 120 and 135 km implementations did however require additional stability modifications such as an increase in the value of the vertical damping coefficient, which damps vertical velocities as they approach the upper boundary.

210
Following this research, the radiation scheme was extended to include non-LTE effects and the model temperature now contains the appropriate realistic forcing up to around 90 km. This work is detailed by Jackson et al. (2020). The improvement to the summertime polar mesopause minimum and consequent improvement in the wind fields can be seen in Fig. 1.
Above around 90 km, the lack of appropriate high atmosphere chemistry and consequent heating via exothermic reactions means that the temperature profile cannot be assumed to be accurate. Given this lack of appropriate chemistry, the relaxation 215 Figure 1. Latitude-height zonal mean monthly mean :::::::: zonal-mean ::::::::::: monthly-mean : climatologies in December comparing (a) ExUM temperature before with (b) ExUM temperature after the non-LTE implementation. Also compared is (c) ExUM zonal (u) wind before with (d) ExUM zonal (u) wind after the non-LTE implementation. The more accurate modelling of the summertime polar mesopause minimum is evident upon introduction of the non-LTE radiation scheme with consequent effect on the modelled winds in the MLT. or nudging scheme is still :::: must ::: still :: be : used above 90 km ::: and :: is :: in :::: place ::: for ::: our ::::::::: simulation. This pushes the model temperature towards a globally uniform temperature field, which can be seen for this region in Fig. 2. 8 Figure 3. Vertical level sets for the 3 km maximal vertical spacing (red) and the 1.5 km maximal vertical spacing (blue).
In summary, this results in an ExUM which differs from the standard General Atmosphere (GA) 7.0 configuration of the UM (as described in Walters et al., 2017) in the following ways: 1. The model chemistry scheme is entirely switched off -the development of a chemistry scheme appropriate for the MLT 220 is currently a work in progress.
2. Atmospheric aerosols are switched off and ozone background files are switched on.
4. The forcing from the radiation scheme now includes the non-LTE effects which means it is physically realistic up to 90 km.

225
5. The temperature field above 90 km is nudged towards the prescribed climatological temperature profile -this accounts for the lack of the chemistry scheme.
With this, the model is now sufficiently mature to ask the question: are the wind fields produced by the new ExUM physically realistic in the mesosphere and lower thermosphere? In this research, we answer this question by performing an initial case study comparing ExUM wind fields and tides to corresponding fields from meteor radar observations.

230
To begin this case study, we use the work of Griffith et al. (2020)  The 94-level configuration requires no changes to the model's vertical damping coefficient (see Griffith et al. (2020) for 235 more details on the vertical damping coefficient), whereas the 113-level configuration requires a sixfold increase in the vertical damping coefficient used -which can have the undesirable effect of modifying the general circulation in an unrealistic manner (e.g., Fomichev et al., 2002). As well as this, over the two chosen radar locations, the vertical wavelength is typically around 20 km in the MLT (e.g. Davis et al. (2013) Table 1 for Rothera). The resolution of the radar observations is also 3 km in the vertical, and this is the resolution used in the MLT in other models such 240 as WACCM and eCMAM (described previously). Therefore, the 94-level configuration is chosen for this study, which has a 3 km vertical resolution in the MLT. It avoids the use of the larger value for the vertical damping coefficient, matches up well with the resolution of the meteor radar observations as well as appropriately resolving wave scales for both radar locations.
Given the above choice of resolution, a choice of start date is required. This is guided by the availability of radar observations 245 and this is discussed in Sect. 2.2 -with 2006 being the year chosen.
The model runs are then all initialised using the same operational analysis from 1 September 2005 at 00 UTC. This allows the model to settle after the initialisation -known as the spin-up period of the model. Following this, climatological data (rather than year dependent data) is used to force background fields such as atmospheric ozone. This choice was made primarily due to the unavailability of year dependent forcing for the recently developed ExUM . As well as this, the nudging of the temperature 250 field above 90 is also based on a climatology and it shall be the focus of upcoming research to improve these aspects of the extended model :::: (such :: as :::: that :::: used :: in ::::: more ::::::::: developed :::::: models :::: like :::::::::: WACCM-X ::::::::::::::::::: (e.g., Liu et al., 2018b)). The primary focus of this work is to provide a first-look at the atmospheric tides present in the model, and perform a first comparison of those tides with observations in order to justify that the core dynamics and physics of the model is sound. Differences seen here can then be used to educate future development.

255
The output attained from the model consists of hourly-sampled time profiles for both zonal and meridional wind fields for the whole of the model year considered -this high cadence is used so that diurnal and semi-diurnal frequencies can be accurately resolved. For simplicity, we only show results for a single simulation, but multiple simulations were performed to verify these results leading to the same conclusions. From these model fields, we compute monthly mean background wind fields and composite days for each month. Each composite day gives an average for each hour of the day over the course of the month 260 at each height in the 80-100 km range being considered. The atmospheric tidal amplitudes and phases are then calculated for each month by fitting a sinusoidal function to this composite day using a curve fitting algorithm.

The meteor radars
We will compare the ExUM model's winds and tides to those measured by meteor radars. Meteor radars are well suited for wind and tidal studies because they can make continuous, reliable, measurements of zonal and meridional winds at the heights 265 of 80-100 km where tidal amplitudes reach large values (e.g., Dempsey et al., 2021). In this particular case, we consider observations made by two commercially-produced all-sky "SKiYMET" radars. One such radar is sited at Rothera (68 • S, 68 • W) in the Antarctic, a latitude where we expect the semi-diurnal tide to dominate. The other is sited on Ascension Island (8 • S, 14 • W) in the equatorial Atlantic, a latitude where we expect the diurnal tide to dominate. The two radars both use the commercially-produced all-sky "SKiYMET" system making their measurements directly comparable. A description of the 270 SKiYMET radar can be found in Hocking et al. (2001). The availability of radar observations for both sites is shown in Fig. 4.
From this, it can be seen that the radars were simultaneously operational with the fewest interruptions throughout 2006, and so we use data from that year in our analysis. The time series of winds recorded by the radars were analysed to determine tidal amplitudes and phases for the diurnal and semi-diurnal tides. The method employed is essentially a standard least-squares fitting method common in tidal analysis. The 275 particular implementation used here is that described by Dempsey et al. (2021). In this, for each month a composite day of zonal and meridional hourly winds was constructed. A least-squares fit of sinusoidal oscillations with periods of 24, 12, 8 and 6 hour, corresponding to the tides, was then made for each month and each component at each height. The result of this analysis is a monthly vector mean estimate of the amplitude and phase of each tide at heights from 79-101 km in both the zonal and meridional components (we will not consider the 8 and 6 hour tides further in this study). These observed tides can then be 280 compared to those predicted by the ExUM for the two sites.

Results
In this section, we present the ExUM winds and tides for the latitudes of Rothera and Ascension Island and compare them to the observations made by the two radars. We begin by presenting, in Fig to upwardly-propagating tides. These tidal oscillations are superposed on background wind fields that themselves display variation in height and time. Before we consider the variability of the tides in more detail, we will thus consider the ExUM's zonal and meridional background winds at the two locations, examine how they vary throughout the year and compare them to 295 the radar observations. All months will be referred to by their 3 letter abbreviation in lists for brevity.

Mean winds
The monthly-mean zonal and meridional winds for Ascension Island are presented in Fig. 6 and those for Rothera in Fig. 7. In each case we also present the corresponding monthly-mean winds observed by the respective radar. The zero-wind line in the figures is indicated by a dashed black line.

300
Firstly, we consider the equatorial site of Ascension Island. The ExUM monthly-mean zonal winds clearly exhibit the wellknown mesospheric semi-annual oscillation, with wind maxima in January and June and minima in April and October. The amplitude of this semi-annual behaviour reduces at the upper heights in the figure and is largely absent at heights above about 95 km. The corresponding zonal winds observed by the radar also display a semi-annual cycle, but the height/time regions of westward winds (negative zonal wind) are rather more extensive than those of the ExUM, with an interval of westward 305 winds being observed to last from Jan-May which is not well reproduced by the model. Further, the ::::::::: maximum :::::::::::: monthly-mean observed wind speeds are also somewhat greater than :::: about :::::: double ::::: those : in the ExUMthan those observed, with the model's maximum monthly-mean , :::: with :::::::: observed wind speeds reaching about 40 ms −1 at heights near 90 km in June and -40 ms −1 at heights near 80 km in January. Nevertheless, the ExUM reproduces the general semi-annual pattern of zonal winds.
The corresponding monthly-mean meridional winds in the ExUM at heights below about 95 km display a seasonal pattern generally similar seasonal variation to that of the ExUM. However, the observed wind speeds are slightly larger throughout 315 most of the year and the region of strongest southward flow in Jun/Jul extends to lower heights than in the ExUM.
In summary, comparing the ExUM and observed winds for Ascension Island, we see that some essential features are well captured and that the semi-annual variation is reproduced. However, there remain some notable differences in detail. This is particularly notable in Feb/Mar when the observed strong westward winds are not well reproduced.
Secondly, we consider the monthly-mean wind fields at the location of Rothera. Here, the ExUM zonal wind is predominantly 320 eastward from Nov-May (i.e, through summer and into autumn) and reverses to be westward from Jul-Oct. The austral summer months exhibit a strong wind shear with velocities increasing from about -20 ms −1 at heights of 80 km to more than 75 ms −1 at heights of 100 km. The radar observations from Rothera reveal rather smaller absolute wind speeds in austral summer with values ranging from about -25 ms −1 at heights of 80 km to about 25 ms −1 at heights of 100 km -significantly less than predicted by the ExUM. The observed winds in winter are noticeably different from those of the ExUM. In particular, the 325 observed winds are eastwards at all heights from Mar-Oct and reach speeds of more than 20 ms −1 , whereas the ExUM yields westward winds at heights above about 85 km with speeds reaching -20 ms −1 for most of these months. This is probably the most notable difference between the winds of the ExUM and those observed by the radars. winds over Rothera reveal a broadly similar pattern of winds to those of the ExUM from Jan-Aug, although with rather stronger northward winds in Jan/Feb. However, in Aug-Dec the observed winds are rather different from those of the ExUM.
In particular, the observed winds are almost entirely northward at all heights and actually reach the largest values measured in December, whereas the ExUM winds are actually southwards in Nov/Dec at heights above about 90 km.

335
In summary, comparing the ExUM and observed winds for Rothera, we see that some aspects of the seasonal variation of the observed winds are reproduced well in the ExUM, particularly below 85 km. However, there is a notable difference in that the observed zonal winds are eastwards in austral winter at all heights whereas in the ExUM they are westwards except at the lowest heights. As well as this, the magnitude of the ExUM winds above 90 km in austral summer is also significantly larger than that observed. We will consider possible explanations for these differences in Sect. 4.

Diurnal tides
We now proceed to a more detailed comparison of the diurnal and semi-diurnal tidal amplitudes and phases in the ExUM at the two locations to those observed by the radars. As with the winds, we will consider monthly-mean properties because they provide a test of the model's ability to reproduce the seasonal variation of the atmosphere.
Monthly-mean tidal amplitudes and phases at heights of 80-100 km were calculated as described in Sect. 2 for both the 345 ExUM results and the radar observations. 14 3.

Amplitudes
For the location of Ascension Island, the zonal and meridional amplitude components are presented in Fig. 8.
In each panel of the figures, the amplitudes predicted by the ExUM are plotted alongside the meteor radar observations. The shaded regions denote the standard deviation from the curve fitting algorithm, and the black bars indicate the standard deviation 350 from the mean of the measured amplitudes across the month.
Considering the monthly-mean ExUM results, we see that the ExUM tidal amplitudes in most months increase from values of about 10-20 ms −1 at heights near 80 km to about 20-40 ms −1 at heights of 100 km. However, in Jan and Mar the amplitudes do not increase across this height range. The zonal and meridional amplitudes are generally similar, but not exactly the same.
For instance, in May and Nov the meridional amplitudes are notably larger than the zonal amplitudes. In fact, the largest In terms of agreement between ExUM and observed amplitudes, the agreement tends to be better for the zonal components Next, we will consider the equivalent monthly-mean diurnal tidal amplitudes at Rothera, which are shown in Fig. 9. Once

Phases
The tidal phases are defined as the Local Time at which the tidal wind first reaches a maximum value for a particular component.
Phases were calculated for zonal and meridional components for both the ExUM and observed winds at for :::: both Ascension and meteor radar (orange) are plotted.
Island and Rothera. As with the amplitudes, we present figures on which we plot both the ExUM tidal phases and the observed tidal phases.

385
The monthly diurnal tidal phases are presented for both the zonal and meridional wind components at Ascension Island in Next, the monthly diurnal tidal phases are presented for both the zonal and meridional wind components at Rothera in Fig.   11. It should be noted that the amplitudes for many months are small, and so caution must be taken in drawing conclusions from the corresponding phases. Nevertheless, we can look for qualitative features.

405
Again, we firstly consider the ExUM phases. The ExUM meridional phases are once more consistent in leading their zonal counterparts. In both components the phases remain roughly constant with increasing height for the majority of the year.
Secondly, we consider the observed phases. As with the modelled phases, the observed meridional phases consistently lead the observed zonal phases, with July being the only exception. In the zonal component, a general trend of decrease in phase with increasing height is seen in the majority of months, with the exceptions being at higher altitudes. In the meridional component, 410 for most months the phase is roughly constant with increasing height, with a weak decrease in phase with increasing height observed in some months. July is again the exception where an increase in phase with increasing height is observed for lower altitudes.
In terms of agreement between ExUM and observed phases, the agreement is better for the meridional component which is The differences observed in amplitude do not follow a clear trend, but often the accuracy of the amplitudes in comparison with observed values is in general better at lower altitudes, and more differences were seen towards the upper heights of the model. The modelled phases systematically lead the observed phases by around 4-10 h. Where differences in phase gradient are evident, at Ascension Island, the observed phase gradients are often steeper than that seen in the ExUM and at Rothera, the ExUM phase gradients are generally vertical, which is not always the case in observations. 430 20

Semi-diurnal tides
We now proceed to a detailed comparison of tidal amplitudes and phases for the semi-diurnal tide, from both the ExUM and meteor radar observations at both locations.
Monthly-mean tidal amplitudes and phases at heights of 80-100 km were calculated as described in Sect. 2 for both the ExUM results and the radar observations. In terms of agreement between ExUM and observed amplitudes, the agreement is excellent and is marginally better for 455 the zonal components in comparison with the meridional components and in general is best at lower altitudes. For the zonal components, excellent agreement is observed in the majority of months. October is the biggest exception, which differs from the observed amplitude by 20-30 ms −1 at 100 km. Otherwise, deviations from observed amplitudes are around 5-15 ms −1 .
Looking more closely at their relative magnitudes, the ExUM zonal amplitudes are similar to observed amplitudes in the majority of cases, but tend to be larger where the amplitudes do differ. For the meridional components, excellent agreement is 460 observed once more in the majority of months. Jan and Aug are the main exceptions, with deviations of around 20 ms −1 at higher altitudes, but still show excellent agreement below 90 km. Otherwise the difference is minimal at 5-10 ms −1 . Looking more closely at their relative magnitudes, the amplitudes are once more similar between the two, with the ExUM amplitudes again larger where they differ.
Next, the monthly semi-diurnal tidal amplitudes are presented for the zonal and meridional wind components at Rothera in 465 Fig. 13.
Once more, we first consider the ExUM amplitudes. The ExUM amplitudes are of very similar magnitude across both components -therefore we will summarise them both simultaneously. The growth of amplitude with height is evident across nearly all months, with March being the only exception. We observe the largest amplitudes of c. 40 ms −1 in Dec/Jan.
Secondly, we consider the observed amplitudes. The observed amplitudes are also of very similar magnitude across both In terms of agreement between ExUM and observed amplitudes, the agreement is mirrored for both the zonal and meridional components. Excellent agreement is observed in the majority of months with the best agreement in general at lower altitudes.

475
Dec-Mar show the largest deviations of around 10-20 ms −1 at higher altitudes. Otherwise, the agreement is excellent with deviations of around 5-10 ms −1 . Looking more closely at their relative magnitudes, in general the amplitudes are similar. In the few cases they do differ, no obvious trend is apparent -for some months the ExUM amplitudes are larger and for others they are smaller.
The monthly semi-diurnal tidal phases are presented for both the zonal and meridional wind components at Ascension 485 Island in Fig. 14. It should be noted that the amplitudes for many months is ::: are small, and so caution must be taken in drawing conclusions from the corresponding phases. Nevertheless, we can as before look for qualitative features.
Firstly, we consider the ExUM phases. The ExUM meridional phases in general lead their zonal counterparts. A decrease in phase with increasing height is observed for the majority of months indicative of upwardly propagating tides, however it is worth noting that the corresponding phase gradient is much shallower than that seen for the phases of the diurnal tide, and thus 490 indicative of a shorter vertical wavelength. In the zonal component in Mar, Nov and Dec, the phase becomes roughly constant with increasing height at high altitudes, and May shows an increase in phase with increasing height also above 90 km.
Secondly, we consider the observed phases. Contrary to the modelled phases, it is not at all obvious that there is a trend between the observed zonal and meridional phases. In general, the trend of decrease in phase with increasing height is apparent in the majority of months for both components. However, other trends are observed. In the zonal component, the months of In terms of agreement between ExUM and observed phases, the agreement is in general better for the meridional component 500 which is excellent for many months, such as Jan-Apr, Oct and Nov, but has larger differences in Jul-Aug. It is interesting to note that the model matches some of the less expected behaviour such as the increase in phase with increasing height above 90 km in September, however the observed amplitudes are fairly small here. For the zonal component, the agreement tends to be good at best, in months such as Jan, Sep and Oct. Again it is interesting that some more complex features are well captured in September. The roughly constant phase with increasing height is also captured in March, but is out of phase by around 2-3 h.

505
This characteristic is repeated in many other months, such as Jul-Aug; namely, the correct qualitative behaviour is seen, but they are out of phase by 2-6 h. However, in other months the ExUM phases do not match those observed, in particular in Nov and Dec. In general though, the agreement is good and the trend of a more shallow decrease in phase with increasing height is mirrored between the ExUM and observed phases.
Next, the monthly semi-diurnal tidal phases are presented for the zonal and meridional wind components at Rothera in Fig.   510 15.
Again, we firstly consider the ExUM phases. The ExUM meridional phases are once more consistent in leading their zonal counterparts, by around 2-6 h. Apart from this phase shift, the zonal and meridional components are practically identical across all months. The phases exhibit a general trend of decrease with increasing height. This decrease is steeper in some months than others, for example compare Feb/Mar (where it is shallow) with Sep/Nov (where it is steeper). This is indicative of varying 515 vertical wavelength throughout the year, but of consistently upwardly propagating tides.
Secondly, we consider the observed phases. As with the modelled phases, the observed meridional phases consistently lead the observed zonal phases by around 3-6 h. They also share the property that, apart from this phase shift, the zonal and meridional components are very similar across the majority of months. A general trend of decrease in the observed phase with increasing height is seen. The observed phases also exhibit a variety of phase gradients, with shallower gradients in Mar and 520 Oct, and steeper gradients in Jun-Sep, for example.
In terms of agreement between ExUM and observed phases, the agreement on the whole is very good, and is marginally better diurnal amplitudes at Rothera, a general increase in amplitude with height, a general decrease in phase with height (indicating 535 upward propagation), a similar magnitude for zonal and meridional components and meridional phases that lead the :::: their zonal counterparts.
In the particular case of the semi-diurnal amplitudes, notable differences between the ExUM and the radar observations are often more pronounced at the greater heights. Finally, the ExUM diurnal :::::::::: semi-diurnal : phases systematically lead the observed phases by around 2-6 h at Ascension Island.

540
Where differences in phase gradient are evident, the observed phase gradients are often slightly steeper than those seen in the ExUM at both locations, indicating that tidal vertical wavelengths in the ExUM are slightly shorter than observed.

4 Discussion
The results presented above reveal that there are many aspects of the background winds and the diurnal and semi-diurnal tides in the ExUM that agree well with observations made in the MLT by the meteor radars at the two sites. However, there are also a 545 number of notable differences, or biases. Here we will discuss the possible origins of these biases and consider how the ExUM might be developed in future to reduce them. Note that the focus of our discussion will : be : on the ExUM's representation of background winds and tides and how they compare to the observations. More complete investigations of the observed winds and tides themselves over these locations and discussions of how they compare to other observational studies can be found in Davis et al. (2013) for Ascension Island and Sandford et al. (2010) and Dempsey et al. (2021) for Rothera.

Monthly mean winds
The most striking difference between the ExUM's monthly-mean zonal and meridional winds in the MLT and those observed by the radars occur in two places, i) the Antarctic during austral summer, when the ExUM zonal winds at the upper heights are much stronger than observed over Rothera, and ii) the austral winter, when the observations reveal eastward winds at all heights from March to October, but when the ExUM predicts westward winds commencing in April, i.e, the observed winds 555 are actually in the opposite direction to those predicted by the ExUM.
The first of these differences most likely arises from the gravity-wave parametrizations used in the ExUM, which are not yet tuned for the high-latitude MLT and so may be resulting in ::: give :::: rise :: to : unrealistically high mean-flow accelerations.
However, the second difference is particularly striking because the existence of any eastwards winds in the polar winter MLT is unexpected since the strong eastwards winds of the underlying winter stratosphere will have removed (by critical-level filtering) 560 all ascending gravity waves :::: GWs with eastwards phase velocities and momentum flux -leaving no such waves to dissipate in the MLT where they could force eastwards winds.
Recently, an explanation for the existence of such eastwards winds in the polar winter MLT has been proposed in the modelling study of Becker and Vadas (2018). These authors suggest that non-primary gravity waves :::: GWs : are generated in situ over the Southern Andes in winter, either by nonlinear instabilities and/or by the local body forces from the temporally and 565 spatially localized wave drag resulting from the breaking of large-amplitude mountain (orographic) gravity waves :::: GWs. These non-primary gravity waves :::: GWs may include waves which have significant eastwards momentum fluxes and which are excited at heights above levels where they would otherwise be removed by the critical-level filtering of eastwards winds. When such eastward waves reach the MLT and themselves dissipate, their eastward momentum may then force eastward mean winds.
The results we have presented here suggest that their predicted eastward winds do indeed occur and so our observations are not in disagreement with the work of Becker and Vadas (2018) and suggest that non-primary gravity-waves may play a key role in 575 the circulation of the Antarctic MLT ::::::::::::::::::::::: (cf. Becker and Vadas, 2020).
The ExUM, in common with nearly all GCMs, does not include gravity-wave sources above the troposphere and so cannot produce an eastward forcing of the polar winter MLT since any eastward propagating waves in the model will be filtered out by critical levels before reaching the MLT. Therefore, the ExUM cannot produce the observed eastward winds. This limitation may well explain the lack of eastward polar winter winds also found in other GCMs which launch gravity-waves from the 580 surface only, including WACCM-X, eCMAM, MUAM and other high atmosphere models.
To further investigate and demonstrate the role of gravity waves :::: GWs in forcing the winds of the MLT in the ExUM, we examined the time series of monthly-mean zonal and meridional gravity-wave tendencies from the spectral scheme over the course of 2006, this ::::: 2006. :::: This is presented in Fig. 16.

Diurnal and semi-diurnal tides
The results presented above for the tides show that the ExUM captures many of the main features of both diurnal and semi-625 diurnal tides at Ascension Island and Rothera. However, the semi-diurnal tide at Ascension Island and the diurnal tide at Rothera reach only small amplitudes in both the ExUM and the observations and so the model biases may not be meaningful.
We will therefore restrict our discussion to the larger amplitude tides that dominate the motion field at each location -that is, the diurnal tide at Ascension Island and the semi-diurnal tide at Rothera.
In the case of the diurnal tide over Ascension Island, the ExUM tidal amplitudes are in most months in good agreement with 630 the observations and increase with height in a manner similar to that observed. However, there are differences in amplitude of greater than 20 ms −1 at some heights in some months in one or both components. This is particularly apparent in February, May and June in the meridional component.
In the case of the semi-diurnal tide over Rothera, the ExUM amplitudes are again generally in reasonable agreement with those observed, but there are some months where the ExUM amplitudes are rather larger than observed (January and December 635 at the upper heights) or smaller than observed (September and October at the lower heights).
At Ascension Island, the diurnal tidal phases have gradients (vertical wavelengths) that are in excellent agreement with the observations, although the absolute values of phase in the ExUM in most months lead the observed phases by about 3-4 h. This systematic difference may, in part, reflect the accumulated phase difference over several cycles of the (short vertical wavelength) tide as it propagates from its sources at lower heights if there is a mismatch between the model vertical wavelength and that of 640 the tide in the real atmosphere.
In the case of the semi-diurnal tide at Rothera, the phases are less well defined than is the case at Ascension Island. Indeed, in some months the vertical profile of tidal phase has a complicated structure without a uniform gradient across the height range considered. This is evident in both the ExUM results and the observations and is notable in, for example, the zonal phases in February and July. This behaviour may result from a superposition of different tidal modes across the height range considered.

645
However, there are also months where the ExUM and observed tidal phases are in good agreement (for example, the meridional phases in February or October).
The WACCM results presented by Dempsey et al. (2021) also included estimates of monthly mean tidal phase as a function of height and indicated a good agreement in the phase gradients (i.e., vertical wavelength) between WACCM and observations in some summer and winter months (particularly, January, February, May-August and December), but less good agreement around the equinoxes. Similar behaviour is apparent in our ExUM results, although again, in some months the agreement is 680 less good, e.g., meridional phases in May and July which suggest longer vertical wavelengths in the ExUM than observed. Davis et al. (2013) investigated both diurnal and semi-diurnal tides over Ascension Island using data from the same meteor radar used in our study. They also compared their observations to results from both WACCM and eCMAM. However, they presented their results as averages for the entire interval 2002-2011 and so, again, the results are not directly comparable with those we report here. We will thus again restrict our comments to consideration of the broad seasonal characteristics of the 685 large-amplitude diurnal tide at Ascension Island, since the amplitude of the semi-diurnal tide at this site is small in both models and observations.
In general, Davis et al. (2013) showed that eCMAM tended to overestimate the meridional amplitudes of the diurnal tide over Ascension Island, whereas WACCM tended to underestimate them. The differences were not so large in the case of the zonal component amplitudes. Both models predicted larger amplitudes at the upper heights considered. In contrast, the results

690
we have presented here show that the monthly mean ExUM diurnal tidal amplitudes are not systematically larger or smaller than those observed, but from month to month can vary and be either larger or smaller.
Estimates of monthly mean tidal phase as a function of height and corresponding vertical wavelengths were also presented by Davis et al. (2013) for WACCM and eCMAM. Both models predicted tidal phases and vertical wavelengths with good agreement to the radar observations around the equinoxes, but with less good agreement in the summer and winter months 695 (particularly eCMAM which predicted much shorter diurnal zonal vertical wavelengths than are observed in summer). The ExUM generally does well in predicting the diurnal tidal phases and phase gradients (i.e., vertical wavelength), but with some small differences in summer months.

Conclusions
We have presented the first study demonstrating the ability of the newly Extended Unified Model (ExUM) to capture the 700 background winds and the atmospheric tides of the MLT. We have detailed the changes made to the model which allowed these investigations, including i) the addition of a non-LTE radiation scheme and ii) the relaxation to a climatological temperature profile above 90 km. We tested the predicted winds and tides in the ExUM by comparing them to the tides observed by SKiYMET meteor radars at characteristic Antarctic and equatorial latitudes where we expect the diurnal and semi-diurnal tides, respectively, to dominate. We used data from 2006 and for each month determined monthly-mean tidal amplitudes and 705 phases.
Despite the simplified nature of this initial development of the ExUM, the model produces diurnal and semi-diurnal tides that display many characteristics of the observed tides. In particular, the monthly-mean amplitudes and vertical gradients of phase ::::: phase :::::::: gradients are in reasonably good agreement with the observations in most months and at most heights. It is still true that in some months and at some heights the predicted tidal amplitudes can differ significantly from those observed. Given 710 that the comparison of winds described above highlights limitations in the ExUM's gravity-wave parameterization, it may well be that this also impacts the model's tides and accounts for some of the differences.
1. The equatorial background MLT winds predicted by the ExUM capture some essential features well -the observed pattern of semi-annual variation is reproduced. However, there are several months where there are notable quantitative differences in the detail, e.g. Feb/Mar.

715
2. The polar background MLT winds predicted by the ExUM have some notable differences from those observed. Most striking are that i) the winds in the ExUM in austral summer are stronger than observed and ii) the observed eastward winds in austral winter are not reproduced in the model, which actually predicts westward winds.
3. We have proposed that these eastward winds in the real atmosphere are forced by the fluxes of non-primary gravity waves :::: GWs : generated when large-amplitude orographic gravity waves :::: GWs break in the upper stratosphere or mesosphere, as 720 suggested in the modelling study of Becker and Vadas (2018). These discrepancies between the model predictions and the observations highlight the limitations of gravity-wave parameterizations that only launch waves from :::: near the surface.
4. The equatorial tidal amplitudes predicted by the ExUM are generally in good agreement with observations. Key qualitative features are reproduced, including large diurnal amplitudes and small semi-diurnal amplitudes; a general increase in amplitude with height; and the meridional tide component exceeding the zonal component.

725
5. The polar tidal amplitudes are generally good and also reproduce many of the qualitative features mentioned above.
However, the ExUM noticeably overestimates the tidal amplitudes at the summer solstice. This is the height and time when the ExUM zonal winds are larger than those observed and we therefore propose the anomalous tidal amplitudes may be a consequence of these zonal winds.
6. The tidal phases of the larger tides have vertical phase gradients which are in very good agreement with observations.Key 730 features are replicated including a general decrease in phase with height and the meridional phases leading their zonal counterparts. A difference in phase is commonly seen but is expected given the ground-level source of parameterized gravity waves :::: GWs.
It is necessary for high top models to reproduce these key features which are critical for deep coupling models as we strive towards more accurate models in the MLT. Further, we have suggested details for future work and parts of the model for future 735 development. From this, we recommend two improvements to deal with the problems seen in the polar MLT, firstly the tuning of the spectral gravity wave ::: GW scheme to correct the wind direction in polar winter, and secondly reducing the magnitude of winds around 95 km in polar summer (which may in turn address the overly large tidal amplitudes observed in polar summer).
These improvements pave the way for the development of a whole atmosphere UM in the near future.
In summary, we have demonstrated that even with relaxation to a relatively simplified temperature field and the use of 740 monthly ozone background files, the ExUM can produce tides with many of the features observed, highlighting its usefulness for future tidal studies. Further, we have suggested that the ExUM's gravity-wave parameterization needs to be revised in light