Investigation of sources of gravity waves observed in the Brazilian Equatorial region on 08 April 2005

10 On 08 April 2005, a strong gravity wave activity (of more than 3 hours) was observed in São João do Cariri (7.4 S, 36.5 W). These waves propagated to the southeast and presented different spectral characteristics (wavelength, period and phase speed). Using the hydroxyl (OH) airglow images, the characteristics of the observed gravity waves were calculated; the wavelengths ranged between 90 and 150 km, the periods from ~26 to 67 min and the phase speeds ranged from 32 to 71 m/s. A reverse ray-tracing analysis was performed to research the possible sources of these detected waves. The ray-tracing 15 database was composed of temperature profiles from NRLMSISE-00 model and SABER measurements and wind profiles from HWM model and meteor radar data. According to the ray path, the likely source of these observed gravity waves was the Inter Tropical Convergence Zone with intense convective processes taking place in the northern part of the observatory. Also, the observed preferential propagation direction of the waves to the southeast could be explained using blocking diagrams, i.e. due to the wind filtering process. 20


Introduction
Since the publication of the pioneering works of Hines in the 1960s on the detection of irregular motions 'gravity waves' in the upper atmosphere, there have been numerous improvement work-outpours. Gravity waves are results of disturbances that occur in atmospheric fluids with the upper mesosphere and thermosphere region being largely impacted 25 (e.g., Fritts and Alexander, 2003). Potential sources of these waves are cold fonts (e.g Plougonven et al., 2017), troposphere convection (e.g., , wind shear (e.g Clemesha and Batista, 2008), topography and wave breaking (e.g., Sarkar and Scotti, 2017), solar eclipse (e.g., Marlton et al., 2016). These atmospheric structures have been identified as a key component in the transportation of energy in the mesosphere and lower thermosphere (MLT) region (e.g., Fritts, 1993;Medeiros et al., 2007;Campos et al., 2016). 30 Internal GWs are generated as adjustment radiations whenever a sudden change in forcing causes the atmosphere to depart from its large-scale balanced state. Such a forcing anomaly occurs during a solar eclipse (Campos et al., 2016;Marlton et al., 2016). The intrinsic properties of the gravity waves (observed horizontal phase speed, propagation direction, observed period, horizontal wavelength) can be calculated directly from the airglow images by spectral analyzing. Using the dispersion relation, the vertical wavelength can also be computed (Vargas et al., 2009). Gravity waves can be summarized as large-scale 35 waves, medium-scale waves, and small-scale waves. Small-scaled gravity waves are characterized by horizontal wavelengths in tens of kilometers (Medeiros et al., 2003), medium-scaled gravity waves propagate at ~ 170 km altitude, and large-scale waves have high phase speeds and travel farther horizontal distances compared to others ).
In the MLT region, there are several and continuous chemical reactions such as the OH airglow emissions (e.g., Sivjee et al., 1992;Taylor et al., 2009;Campos et al., 2016). These emissions among several others have been used by many 40 2 authors as a proxy for investigating gravity wave activities. Airglow emissions are faint luminescence that are produced as a result of the emission of solar radiations (ultraviolet and x-radiation) by ionized air molecules. These luminosity are usually captured by the all sky imagers (ASI) (e.g., Chamberlain, 1954;Krasovskij et al., 1965).
To identify source location of gravity waves, the reverse ray tracing method has been widely adopted. This algorithm allows the identification of the source locations of gravity waves. Several researchers have successfully implemented this 5 technique to identify points of generation of these waves under different atmospheric conditions using airglow images (e.g., Hecht et al., 1994;1997;Brown et al., 2004;Wrasse et al., 2006;Pramitha et al., 2014, Sivakandan et al., 2016. Wrasse et al., (2006) did a comprehensive study of gravity waves observed over Brazil and Indonesia and concluded that most of the studied waves have their sources in the troposphere. Similarly,  studied the propagation 10 of gravity waves observed during the SpreadFEx campaign in Brazil and found out that the likely sources of those waves were deep convection in Brazil. In addition, Pramitha et al., (2014) identified that 64 % of observed GWs over Gadanki, India originated from the upper troposphere while the remaining were seen to have been ducted in the mesosphere. Furthermore, in 2016, Sivakandan et al. (2016) studied Gws observed in the southern part of India and associated the sources to convection.
The objective of this current study is to extensively study a strong activity of GWs observed in São João do Cariri 15 (7.4 o S, 36.5 0 W) on 08 April 2005. More than three hours of GW activities were observed and the waves propagated exclusively to the southeast. An explanation for this uncommon pattern is presented in this work investigating the combining effect of the location of the source and the wind filtering process.

20
The GWs detected in this study were observed using the ASI installed at the observatory in São João do Cariri. The ASI is an optical instrument that provides monochromatic maps of aurora and atmospheric airglow emission of different wavelengths. It has been designed to keep track of the spatial and temporal variations of OH, OI5577, 6300 nm airglow emissions (e.g. Paulino et al., 2010).
This present study however utilized only the OH airglow images captured by the ASI at an altitude of 87 km. This 25 instrument comprises of a fish-eye (f/4) lens, a telecentric lens system, a field of view of 180 0 , a computer-controlled filter wheel with several slots for the observation of different emissions, and a charged coupled device (CCD) camera with a CCD device used as a photodetector for increased sensitivity. More technical and operational details about this particular imager at São João do Cariri can be found in Medeiros et al., 2007;Paulino et al. (2012).

30
The Sounding of the Atmosphere using Broadband Emission Radiometry (SABER) is one of the four instruments onboard on the Thermosphere Mesosphere Ionosphere Energetics and Dynamics (TIMED) satellite. This instrument measures the vertical temperature profile of the atmosphere. In this study, at 7.4 o S, 36.5 o W, the vertical temperature measurements from 20 to 108 km were obtained from SABER. Data from the Naval Research Laboratory Mass-Spectrometer-Incoherent-Scatter (NRLMSISE00) atmospheric model (Picone et al., 2002) was utilized for complementing 35 the measurements at unavailable heights of 0-19 km, and 109-400 km. These measurements were used to provide vertical profiles of kinetic temperature, pressure, geopotential height, and volume mixing ratios for the trace species (Mertens et al., 2001).

The SKiYMET Meteor Radar
The All-Sky Interferometric Meteor Radar (SKiYMET) system located at 7.4 o S, 36.5 o W provided measurements of the horizontal wind speeds and direction of the mesosphere (81-99 km). The radar which is composed of Yagi antennas-5 receiving antennas and one transmitting antenna operates at 35.24 MHz with a maximum power of 12 kW. This instrument detects the trail left behind by vaporized meteors, determines the angle-of-arrival by using the phase difference between the 5 receiving antennas, then measures the radial velocity by using the derivation of the speed and direction of the atmospheric winds carrying the meteor trail at a specified altitude. The phase delay between the transmitted and received signal is used to determine the position of the trail. Further detail about this radar has been published in (e.g., Hocking et al., 2001;Egito et al., 2018;Paulino et al., 2015).

Determination of gravity waves parameters
To obtain the characteristics of the detected gravity waves, a two-dimensional Fast Fourier Transform (FFT) was used in specified batches of OH airglow images. The pre-processing of the airglow images can be summarized in the following procedures: • Rotating the image to fit the top of the image with the north geographic region;

15
• Removing low frequency waves by applying the Butterworth high pass filter; • Contrast enhancement and unwarping the airglow images for FFT analysis.  Table 1 presents a   25 summary of all the wave events observed on this night. Additional information about the cross spectrum analysis used in this present study to obtain the characteristics of these waves can be found in (Wrasse et al., 2007).

The atmospheric profile
The SABER instrument provided temperature measurement from 23h46 to 23h51 UT on 08 April, and from 08h35 to 08h40 UT on 09 April for altitudes 20-108 km. Then, linear interpolation between 23h46 (08 April) and 08h40 (09 April) was done for the same altitude range. In addition, numerical values for temperature were obtained from https://omniweb.gsfc.nasa.gov/vitmo/msis_vitmo.html for time (12h to 22h) on 08 April 2005 and for missing heights of (0-5 19km, 109-400km) to complement the SABER measurements. Finally, a 13-profile temperature data with a temporal resolution of 2hrs was constructed for the whole period. However, some discrepancies were at the combining points, and so the data were smoothened so that the model can seamlessly match the measurements at these junctions. This methodology was discussed by Paulino et al. (2012). The temporal and spatial variation of the temperature (not shown) with a contour interval of 154 km shows an outbreak of cold air at heights lesser than 150 km (< 300K). The peak period between 17 h and 10 19 h has a temperature value 1000 K.
The SKYiMET radar provided the zonal and meridional wind measurements from 81-99 km every 3 km from 0h to  . Additional details about the reverse-ray tracing parameters can be found in (Vadas et al., 2007;, T is the temperature, g is the acceleration due to gravity, = 8314.5 is the gas constant, z is the altitude, and is the static pressure. The scale height = − ⁄ ⁄ is obtained from the ratio of the density to the derivative with respect to the altitude, while the potential temperature is = ( ⁄ ) ⁄ , Cp is the specific heat capacity at constant pressure, and the and Equation 48 of Vadas and Fritts, 2004, and cg is the group velocity. The derivatives of the group velocities 10 and the complete dispersion relation can be found in Vadas and Fritts (2005) and Vadas (2007).

30
The spectral result shown in Table 1 shows observed wavelengths with a standard deviation of 24 km and mean value of 130 km with a large amount of variability among the detected waves. This result agrees with the reports of Medeiros et al., (2007) for waves detected at this observatory. Thus, it can be confidently concluded that the observed wavelengths are good representatives of the gravity waves in this observatory.
6 Figure 2 shows the traveling direction of the 5 events. From the polar chart, it is seen that all the waves are propagating Southeast with an approximate azimuth of ~134 o from the north. This is in favorable agreement with the results obtained from previous studies for the same observatory (Essien et al. 2018). This anisotropicity will be discussed later on in this study.    Similarly, the ray-tracing results show backward propagation of the second wave event detected at 21:33 UT in Figure   5(a). This shows a similar analysis when compared to the first wave event. However, this wave suffered greater wind acceleration ~12 hours as did GW3 ~10 hours. However, just as in GW1, GW4 and GW5 were accelerated by the modeled winds just by ~ 2 hours and 1 hour respectively. 5 9 Figure 5: Same as Figure 4 but for Event #2.
It was observed that the events 2, 3, 4, and 5 with horizontal wavelengths of H~140 km propagated more than ~1300 km from their source region. All these 4 events are believed to have originated from the same source due to the resemblance in their characteristics. Events 1, although appear to have a different but a source closer to the observatory site as it traveled a shorter distance of ~700 km before detection-this wave is still source-linked to the convective processes in the ITCZ because the actual convective area is larger than the cloudy areas ).  All the 5 wave events were reverse traced to some convection activities occurring in the northern region of the OLAP observatory. These have been identified as possible generators of these waves. One question is evident, why is the propagation direction of all the waves southeastward, since the ITCZ extends horizontally and covers the northern-region of the observatory? These results are in close agreement with previous results of different studies obtained at this laboratory 5 (e.g., Medeiros et al., 2003 andEssien et al., 2018). It is expected that the gravity waves should propagate across all directions in accordance to the location of the ITCZ. Thus, further investigation was done to explain this anisotropy phenomenon. To better understand the physical mechanism that is producing the anisotropy, blocking diagrams have been used to investigate the role of the wind in the filtering process of these gravity waves.
Atmospheric winds are the default features of a real atmosphere. As these gravity waves propagate in the wind 10 direction into the upper atmosphere, they would be susceptible to the Doppler's effect and critical level dissipation (Bretherton, 1966). The critical level marks the region where the horizontal wind component annuls the wave's horizontal phase speed (Medeiros et al., 2003). This region is very important as it decides how and if a traveling wave would propagate further. To understand the anisotropy of these GWs, we apply the critical level theory of the atmospheric gravity waves 13 filtering (Fritts and Geller, 1976;Fritts, 1979). From Gossard and Hooke (1975) relation, the intrinsic frequency of the gravity wave under the influence of both horizontal wind components can be described by equation 4 written below.
where k is the magnitude of the horizontal wave vector, V represents the two horizontal wind components, and c is the horizontal phase speeds of the gravity waves. Equation (4) can also be re-expressed in terms of the zonal and meridional 5 components. Further details can be found in the works of Medeiros et al. (2003), Campos et al. (2016 and . According to = ∅ + ∅, where represents the phase speeds of these waves, is the zonal wind component and is the meridional wind component, we constructed blocking diagrams using the azimuthal angles (ø). With input winds from the Horizontal Wind Model, we aimed to understand the wind filtering effects on the gravity waves, investigate why all the waves have a preferential propagation direction, and also to detect regions where the phase speed of 10 the GW is ≤ the velocity of the winds. The results of the 3-D blocking diagrams are shown in the following figures. 14 Constructing a polar chart as a function of the azimuthal angles and phase speeds using the wind data from HWM and the SkiYMET radar, we show where the phase speeds of these gravity waves equaled the wind speed of the background.
The critical levels were projected into the blocking diagram showing for each horizontal wind speed and azimuths of the corresponding GWs. If the phase speed of the GW is trapped in the blocking lines, it represents that the wave is prohibited to propagate upwards.

5
The above blocking diagram allows the detection of regions where ≤ 0 on the night of 08 April 2005 for the OH emission layer. Every circle in this diagram shows the critical level of the vertical propagation of these waves. The red and black mesh signify the measured and modeled wind components respectively while the green arrow represents the magnitude and direction of the detected GW. The theory of filtering process of gravity waves disallows waves propagating into the shaded region due to the effect of the critical levels ).

10
Similarly, we observed that the phase velocity of all the GW events were indeed greater in magnitude than the blocking area. They had strong enough speed and momentum to escape and propagate through the critical levels easily. It is important to note the main contributions to the blocking area were due to the measured wind in the mesosphere and lower thermosphere, making this analysis strongly confident.
Thus, these detected waves avoided and escaped absorption in the forbidden regions by traveling at these interesting 15 angles. The anisotropy of these waves furthermore compels the source location of the wave to be in the Northwest because the location of this wave source played a key role in this preferential traveling (Fritts et al., 2008;. The sources of these waves were identified as the convective processes in the ITCZ zone.

Conclusions
Using OH airglow images captured by the ASI at São João do Cariri, we investigated the sources of some gravity 20 waves observed on the night of 08 April 2005 in the OH airglow layer. Using the spectral analyzing method, we obtained the characteristics of 5 major gravity events with horizontal wavelengths concentrated between 90 km and 149 km. The phase speeds were distributed between the range of 32 m/s and 71 m/s, and the observed periods extended from 26 min to 67 min.
These waves presented spectral characteristics that are very compatible with waves previously observed in the same site. In addition, southeast propagation suggested possible sources in the northwest of the observatory.

25
Focusing on the possible sources of these waves, we back-traced the trajectory of each of these waves from the OH layer (87 ) into the troposphere using the meteor radar wind data, the HWM model winds, and zero winds. We found out that the RRT put the source as active convective processes (in the ITCZ) in the northwest of the laboratory as shown in the back tracing results presented above. However, the ITCZ was extended by a long strip in the northern part of the observatory which suggested generation of other GWs that should have been observed in the south and southwest of the observatory.

30
Thus, the construction of blocking diagrams showed that only the spectrum of waves with propagation to the southeast were able to propagate vertically to altitudes of the OH layer. Therefore, the filtering effect of gravity waves was decisive for explaining the presence of observed waves propagating to the Southeast.
Data availability: All sky image data used in the course of this study can be requested from Aerolume (UFCG) or Lume (INPE) Groups by mailing igo.paulino@df.ufcg.edu.br.

35
Author contributions: OD-I has written the manuscript. IP has revised the manuscript and supervised the research. CAOBF has calculated the deep cloudy convection over the OLAP area. ARP has provided wind measurements from the meteor radar, and revised the manuscript. RAB and AFM have revised the full text. CMW has provided the spectral analysis for the observed gravity waves.