Frequency of Mountain Waves Over Kanto Area Revealed by Imaging Observations of OH Airglow


 Imaging observations of OH airglow were conducted at Meiji University, Japan (IN, mE), from May 2018 to December 2019. Mountainous areas, including Mt. Fuji, are located to the west of the imager, and westerly winds are dominant in the lower atmosphere throughout the year. Mountain waves (MWs) are generated on the leeward sides of mountains and occasionally propagate to the upper atmosphere. However, during the observation period (about 1 year and 8 months), only four possible MW events were identified. Based on previous reports, this incidence is considerably lower than expected. There are two possible reasons for the low incidence of MW events: (1) The frequency of MW excitation is small in the lower layers of the atmosphere, and/or (2) MWs do not propagate easily to the upper mesosphere due to background wind conditions. This study verified the likelihood of the former case. Under over-mountain airflow conditions, wavy clouds are often generated on the leeward side. Since over-mountain airflow is essential for the excitation of MWs, the frequency of wavy clouds in the lower atmosphere can be regarded as a measure of the occurrence of MWs. The frequency and spatial distribution of MWs around Japan were investigated by detecting the wavy clouds from color images taken by the Himawari-8 geostationary meteorological satellite (GSM-8) for one year in 2018. The wavy clouds were detected on more than 70 days a year around the Tohoku region, but just 20 days a year around Mt. Fuji. This suggests that few MWs are generated around Mt. Fuji. The differences between these two regions were examined focusing on the relationship between the local topography and dominant horizontal wind fields in the lower atmosphere. Specifically, the findings showed that the angle between the dominant horizontal wind direction and the orientation of the mountain ridge is a good proxy of the occurrence of wavy clouds, i.e., excitation of MWs in mountainous areas. We have also applied this proxy to topography in other areas of the world to investigate areas where MWs would be occurring frequently. Finally, we discuss the likelihood of "MW hotspots" at various spatial scales in the world.

considerably lower than expected. There are two possible reasons for the low incidence 23 of MW events: (1) The frequency of MW excitation is small in the lower layers of the 24 atmosphere, and/or (2) MWs do not propagate easily to the upper mesosphere due to 25 background wind conditions. This study verified the likelihood of the former case. 26 Under over-mountain airflow conditions, wavy clouds are often generated on the 27 leeward side. Since over-mountain airflow is essential for the excitation of MWs, the A cooled CCD camera (Clara, Andor Technology Ltd., UK) was used for all 125 observations. Table 1 and Figure 2 show the main specifications and components of the 126 camera, respectively. Specifically, the components of the camera consist of a fish-eye 127 lens (objective lens), a mechanical shutter, a collimator lens, an interference filter, a 128 focusing lens, and the detector. The bandpass interference filter, which is 15 nm width 129 and centered at 890 nm, is optimized to detect emissions from the Meinel OH (7-3) 130 band in the near-infrared wavelength. Figure 3 shows a transmittance profile of the 131 interference filter with locations of the major rotational lines belonging to the Meinel 132 OH (7-3) band. This wavelength region was selected because it is less contaminated by light pollution from surrounding urban areas. The collimator lens is set in front of the 134 filter to suppress the effect of a wavelength-shift in the bandpass filter which depends 135 on an angle of incident light to the filter. The imager was placed in a water-proofed box 136 for field observations; specifically, the box was equipped with a thermostatically 137 controlled heater and intake fan to keep the temperature and humidity inside the box 138 between 5°C and 25°C, and less than 65%, respectively. A mechanical shutter behind 139 the fish-eye lens prevents damage to the CCD due to direct sunlight. This shutter is 140 controlled by a trigger signal sent from the camera.  145 Figure 2 Optical components of the OH airglow imager. 147 148 Figure 3 Transparency characteristics of the interference filter. The wavelengths of Q-149 Branch and P-Branch in the OH (7-3) band are indicated by arrows above the image. Figure 4 shows the field of view (FOV) of the imager at a typical altitude of the OH 151 airglow layer (85 km). The effective FOV at this altitude has a radius of about 120 km 152 from the observation site, exposure time was 3 min, and the images were taken 153 continuously from sunset to sunrise. To obtain a higher signal to noise ratio (SNR) 154 airglow image, 2 × 2 binning was performed on the CCD. The final image acquired by 155 the observation had dimensions of 696 pixels × 520 pixels. The dark frame used for the 156 dark subtraction method described in section 2-3 was taken every night with the shutter 157 closed at the start of observation. The imager was controlled by a PC installed in the 158 box, and all the acquired image data was saved as TIFF format. 159 Airglow cannot be observed from the ground in cloudy weather. In addition, data with 166 the moon light and the associated scattered light were excluded from the analysis 167 because this light contamination is considerably stronger than the airglow emission. In 168 this study, image data that satisfied the following two conditions were defined as valid 169 data for further analysis: 170 1) Data was acquired in a clear sky. 2) Data without the moon, its scattered light, and following stray lights causing ghosting 172 and/or flaring. 173  The acquired image data was visually judged as being valid or not. When valid data 180 were obtained continuously for several hours, then the images were subjected to MW 181 analysis using the method described in Section 2-3. 182 183 2-3 Procedure for extracting MW signals from image data 184 The following analytical method was used for extracting MW signals from airglow 185 image data. Since the apparent horizontal phase speed of MW is zero, MW appears to 186 be almost stationary from the ground observer. In principle, the bright and dark patterns 187 in an airglow image generated by the MWs would become clear by simply integrating 188 the successive images. However, due to the van Rhijn effect, atmospheric extinction enhance the MW signatures with horizontal wavelengths that have same order as the 191 FOV (see Figure 4). The procedure employed to remove these effects from the 192 integrated airglow image data is described below. 193 First, the elevation and the azimuth angles corresponding to each pixel of the image are 194 determined using star images ( Figure 6). The scheme to fit the local horizontal which is the angle from true north, increases counterclockwise in an image. As seen in 197 The valid data acquired continuously for 2 hours which corresponds 40 images were 210 averaged (we call this averaged data ̅ ) (Figure 7(a)). This process smooths the wave 211 with non-zero horizontal phase speed. By this method, stationary AGWs with a 212 horizontal scale below the FOV (horizontal wavelength < 240 km) can be detected.
profile, ̅ ( ), which is a function of Φ.   Figure 8. If valid data were acquired over two hours, then 237 airglow images were analyzed using the method described in Section 2-3. Table 2  238 shows the total number of observation days and days with a clear sky in every month. 239  and counted the number of events with a stationary wavy pattern (i.e., possible MW 249 signatures). Figure 9 shows examples of the stationary wave events that continued for 250 more than 2 hours. The stationary structures were detected only on the four days shown 251 in the figure. This number of possible MW events is considered to be unexpectedly low 252 despite the observation site being located in a region that appears to be well suited for 253 observing MWs (i.e., on the leeward side of mountains that would be expected to 254 propagate MWs into upper atmosphere). 255  then there is a high possibility that wavy clouds will not appear, even if MWs are 293 induced. Therefore, the absence of wavy clouds does not mean that MWs are not being 294 induced. The formation of wavy clouds indicates that mountain-crossing airflow has clouds were formed on the leeward side of the mountains (for example, see Figure 10). 297 In other words, we used the formation of wavy clouds as a proxy for the induction of In this study, 'wavy clouds' are defined as a set of three or more clouds oriented in 305 parallel rows that are generated in the vicinity of mountainous areas. We examined the 306 generation frequency and the spatial distribution of these wavy clouds over Japan by analyzing color images acquired by GMS-8 in 2018. In addition, the direction of the 308 wave front and the distance between two adjacent clouds that make up a wavy cloud 309 were calculated using the following procedure: 310 311 i) Divide image data for Japan (3301 pixels × 2701 pixels) into smaller images (300 312 pixels × 300 pixels, i.e., 300 km × 300 km) and check if there are wavy clouds in this 313 area. When wavy clouds exist in a sub-image, further subdivide the image (100 pixels × 314 100 pixels, i.e., 100 km × 100 km) ( Figure 11) and move to the next step. 315 316 317 Figure 11 Subdividing the original GMS-8 image into sub-images to isolate wavy 318 clouds for deriving wavelength and wave front.
ii) Select two adjacent rows of clouds form a wavy cloud in a 100 pixels × 100 pixels 321 image. Mark two points on along the centerline of each cloud and acquire the position 322 of each point on the image (Figure 12(a)). The positive values of the ( , ) coordinates 323 in Figure 12 corresponds to the (east, north) directions, respectively. 324 325 iii) The equations for the two straight lines are derived from the position acquired in ii). 326 The gradients of each line are defined as 1 , and 2 , and the mean value, ̅, of 327 1 and 2 is defined as the mean gradient of the two lines (Figure 12(b)). be ( 1 , 1 ) and ( 2 , 2 ), respectively. The Y-axis intercept ′ 1 of an equation is 333 calculated by ′ 1 = 1 − ̅ 1 . Similarly, ′ 2 is calculated. The distance between two 334 cloud rows, i.e., the wavelength, , and the direction of the wave front, θ, are obtained 335 as shown in Figure 12(c). The difference between the Y-axis intercepts ′ 1 and ′ 2 of the two straight lines is defined as ∆ , and λ is calculated by ∆ cos . We assumed θ 337 to be 0 in the eastward direction and that it increases counterclockwise. 338 339 340 Figure 12. (a) and (b) show the calculation method of wavy cloud wavelength, λ, and Wavy clouds were typically stable and were continuously observed at the same location 343 for several hours. Although the GMS-8 color image is acquired over 2.5 min, we 344 analyzed the image data acquired every hour (taken at 9:00, 10:00, 11:00 …) to simplify 345 the analysis. When wavy clouds were detected in the same area in successive images, it 346 was considered that the image represented the same event. The number of days when 347 wavy clouds were observed was counted for each sub-image (100 km × 100 km). 348 349

4-3 Frequency of wavy clouds over Japan 350
We analyzed GMS-8 color images acquired from January 2018 to December 2018 to 351 detect wavy clouds over Japan. Figure 13(a) shows the occurrence frequency and spatial 352 distribution of wavy clouds over Japan in this period. The color scale represents the 353 number of days on which wavy clouds were observed. Wavy clouds were frequently 354 observed over Hokkaido and the Tohoku region. Focusing on Tohoku region, the area 355 with highest frequency of wavy clouds was 79 days/year (Figure 13(b)). In contrast, 356 even over the highest area of the Kanto region, which is also within the FOV of our areas around the Kanto region, wavy clouds were detected only on approximately 10 359 days/year (Figure 13(c)). The relatively low frequency of wavy clouds observed over 360 the Kanto region is consistent with the low frequency of MWs revealed by our airglow 361 observations. From these results, it is found that the occurrence frequency of wavy 362 clouds differs markedly among regions, even though there are numerous mountainous 363 areas throughout Japan. meteorological fields is 0.5° × 0.625°, and the time interval is 3 hours. We examined the angle between the wave front direction and the background horizontal wind direction 375 using this data set. Figure 14 shows the total occurrence frequency of the number of 376 events with wavy clouds that was enumerated using this angle. The findings showed 377 that more than 75% of wavy clouds had an angle ≥ 60. The direction of the phase line 378 of the wavy clouds was considered to reflect the orientation of the mountain ridgeline.

4-4 Relationship between topography and background wind in the lower 387 atmosphere 388
The difference in the number of MWs generated each region is considered to be the 389 result of the interaction between topography and the synoptic horizontal wind field in 390 the lower layers of the atmosphere. We examined the relationship between 391 topographical features and the horizontal wind field by using geographical elevation 392 data and a meteorological reanalysis data. The details of each data type are described in 393 this section. 394 First, we derived the elevation data by using PNG Elevation tile provided by Geospatial 395 Information Authority of Japan (Figure 15(a)) (Nishioka and Nagatsu, 2015). The 396 elevation value of each pixel is calculated from the pixel value using equations (1) as 397 (2) as follows: 398 where, R, G, and B are the color components of the pixel value of the elevation tile, ℎ (= 0.01 m) is the elevation resolution, and H is the elevation value. Figure 15(b) shows the elevation converted from Figure 15(a). Then, mountain ridgelines were 401 extracted from each tile using a median filter as described by Iwahashi (1994)  ridgeline is expressed as 0 ≤ '≤ 90 (Figure 15(d)). We used MERRA-2 reanalysis data to examine the horizontal wind directions in the 424 lower layer of the atmosphere. We adopted this wind field as a synoptic wind field. The 425 wind velocity at an altitude of 1000 m at the center of each elevation tile was calculated using this data. We defined the wind direction as 0 in the east and the angle increases 427

counterclockwise. 428
The angles between the horizontal wind directions and the directions of the mountain 429 ridgelines (θ') that were deduced using the above method were calculated for the entire 430 world; we defined this angle as α. As mentioned in Section 4-3, 60 ≤ ≤ 90 is 431 considered to be favorable for inducing wavy clouds, i.e., over-mountain airflow, which 432 is the primary source of MWs. Figure 16 shows the annual ratio, which satisfies the 433 condition, 60 ≤ ≤ 90 around Japan. Wind data on 9:00, 12:00 and 15:00 were 434 used on this calculation because we would like to compare with wave cloud occurrence 435 frequency which is derived from GMS-8 color images acquired during day. The isolated 436 data points surrounded by zero are omitted from this plot. These areas were highly 437 consistent with the observed wavy cloud counts shown in Figure 13. High values were 438 clearly observed in the Tohoku and Hokkaido regions, and low numbers were clearly 439 observed around Kanto region. shown. However, the findings showed that major mountain area are not 'permanent' hot 456 spots of wavy clouds. The annual ratio of 60° ≤ α ≤ 90° in those areas is about 25~50%. 457 In the high latitude and mid-latitude of the Northern Hemisphere, there are many spots 458 with the ratio over 25 %. The 25% means that wavy clouds (MWs) will occur in these 459 areas more than a season. In low latitude regions such as Mexico, the northern African, 460 southern India and Southeast Asia, there are some hotspots with the ratio over 50%. 461 Importantly, this plot merely shows the distribution of potential wavy cloud hot spots, 462 i.e., the locations at which the active induction of MWs by over-mountain airflow may 463 and/or dissipation occur (Eliassen and Palm, 1961). MWs in our observations (4 events/1 year and 8 months) was surprisingly low relative 502 to previous observations conducted near mountainous areas. To clarify this disparity, we 503 examined the relationship between topography and horizontal wind fields using 504 geographical elevation data and a model of meteorological fields. 505 First, we examined the occurrence frequency of wavy clouds that formed in the lower 506 atmosphere around and over Japan by analyzing color image data obtained by the GMS-507 8 satellite. The results showed that wavy clouds formed in the leeward sides of 508 mountain ridges and that the occurrence frequency of these clouds was high in northern 509 regions of Japan and low in southern regions (including our observation site). Since this 510 difference was considered to reflect the occurrence of over-mountain flow, which is a prerequisite for MW induction in mountainous areas, we then attempted to elucidate the 512 underlying reasons for this spatial disparity. As a result, we developed a simple index, 513 α, which is the angle between the orientation of mountain ridgeline and the background 514 wind. This measure can be used as a measure of the occurrence frequency of MW in 515 each region and it was confirmed to be a good proxy for explaining the occurrence 516 frequency of wavy clouds. The condition α > 60° was found to be favorable for 517 induction of the wavy clouds near mountainous areas, and we also mapped the global 518 distribution of α. The results showed that areas with α > 60° were highly consistent with 519 areas with high momentum flux to the upper atmosphere, as suggested by both previous 520 observations and the GCM. In addition, it was also found that there are many small 521 areas that satisfy the condition α > 60° for more than a season in high latitude and mid-522 latitude of the Northern Hemisphere. Even though each size of these areas was small, 523 integrated effect of waves generated in each of these small areas could possibly affect 524 circulation in the middle atmosphere. It is therefore considered necessary to verify 525 whether the MWs induced in these small areas can propagate to the middle atmosphere 526 or not.
As a next step, we are planning to expand the OH airglow imaging observations to the 528 Tohoku region and/or Hokkaido where the high occurrence rate of excitation of MWs is 529 expected to clarify the effects of small-scale topography on circulation in the middle 530 atmosphere. 531       Table 1 Specifications of the Clara CCD camera. 576 Table 2 Number of total observation days and days with clear sky (more than 2 hours) 577 in each month.