Our study focuses on EPB events in the period of March–April in 2003, 2004, and 2005. EPB activity was higher in 2003, moderate in 2004, and lower in 2005, in accordance with the decrease in solar activity. From the GAIA assimilation data, we first focused on day-to-day variability of temperature from March 1 to April 30 at different altitudes. Figure 2 shows the cross-correlation R between day-to-day variability of S4 and atmospheric temperature T on an equal-pressure plane (a) and (b) at 500 hPa (about 5 km in altitude), (c) and (d) at 2 hPa (about 40 km in altitude), and (e) and (f) at 4 × 10−4 hPa (about 100 km in altitude). We selected 0930 UT as the time for temperature data (a), (c), and (e), and 1530 UT as the time for temperature data (b), (d), and (f), which correspond to local times at Kototabang of 1630 LT and 2230 LT, respectively. Each panel in Fig. 2 covers an area of 28–168 E longitude and 37S–46 N latitude. Figure 2 is then similar to Fig. 1 in Ogawa et al. (2009) except that their OLR cloud-top temperature is substituted with the GAIA temperature at certain altitudes. Our study area is wider, and the area in Ogawa et al. (2009) is indicated by a red square in each panel.
In Fig. 2, at 1530 UT in panels (b), (d), and (f) near the western edge of the red square regions, we find a high positive R = ~0.4 at all altitudes (marked as “C” in each panel). This feature is close to those marked “A” in Fig. 1. This feature is at similar longitudes at altitudes of 5 and 40 km, but is slightly shifted toward the west at an altitude of 100 km. Another similar feature is that these areas are surrounded by high negative R areas such as the feature marked “B” in Fig. 1. These positive/negative high R structures are clearer at an altitude of 40 km than at altitudes of 5 or 100 km. At this altitude, we should note that the structures are elongated along the meridian. We can recognize the zonal train of such structures, marked as “D” in panel (d), which is distributed in the latitudinal region of 9 S–19 N. Here, a question may arise as to what the meaning of negative R is. As discussed later (Fig. 4), the temperature shows enhanced longitudinal fluctuations near the equator. A negative deviation of the temperature is then associated with the negative R. We understand that S4 enhancement is associated with a high correlation in both the positive and negative directions. In Fig. 2c, it shows the R distribution at 0930 UT, which is close to the distribution at 1530 UT. However, from a close look at the “D” feature at an altitude of 40 km, we can see a slight eastward motion of the pattern through time. We can see an approximately 900 km movement of the pattern eastward in 6 h, which corresponds to a propagation speed of 41 m/s eastward. The horizontal bracket in the panel covers three wavelengths of the pattern and is 6170 km long. The zonal wavelength is thus 2060 km. If this pattern is measured from a fixed location, the apparent periodicity should be about 13.9 h. This pattern of propagation is, however, not clear at different altitudes. At an altitude of 5 km, the R distributions at 0930 UT and 1530 UT are very close and do not show clear propagation. At an altitude of 100 km, patterns between 0930 UT and 1530 UT show an overall similarity but do not hold detailed similarities such as those at altitudes of 5 or 40 km, so that it is not easy to find the pattern propagation.
Figure 3 shows a similar analysis for data in 2004 and 2005, where we plot data at an altitude of 40 km only. Data in 2004 (Fig. 3a, b) show similar features as in 2003. A meridionally elongated pattern of high R is seen in the 50–85 E longitude region, and it shows a slight eastward propagation with a small change in the patterns (marked as “E” in the panels). This pattern propagated eastward for 996 km in 6 h. From this, we can see that the eastward propagation speed is 46 m/s, the zonal wavelength is 2010 km, and the apparent periodicity relative to the fixed location is 12.4 h. These numbers are close to those found in 2003. Data in 2005 (Fig. 3c, d), on the other hand, do not show clear patterns, and the R level is relatively low in both the positive and negative directions. This unclear R can be attributed to the fact that the EPB activity in 2005 had already decreased because of the low solar activity. We then focus on the data from 2003 and 2004 for later analysis.
What is the latitudinal distribution of the temperature perturbation? We selected an altitude of 40 km, and calculated the longitudinal deviation of the temperature from a 5-point running mean, and averaged them over 10 days when EPBs were the most active, during March–April in 2003 and 2004. Figure 4a, b, respectively, shows results in 2003 and 2004, and the analyses at latitudes of 32 N, 4 N, and 23 S are described in each panel. In both years, the perturbation is a maximum at the latitude 4 N, as compared with those at 32 N and 23 S. These results are consistent over 2 years, but a perturbation at 4 N is found at the longitudes 40–80 E in 2003 but at 40–90 E in 2004. This suggests that the longitudinal perturbation of the temperature is greater at low-latitude regions, with a possible enhancement on the EPB-active days.
We focus on the data at the latitude of 4 N, and compare atmospheric perturbations on EPB-active and EPB-inactive days. We selected the three most EPB-active and EPB-inactive days in each year and compared averaged GAIA data between them. As EPB activity is affected by the geomagnetic disturbance (Abdu 2012), contamination from this effect should be reduced. The “international Q-days and D-days” are the 10 monthly quiet and 5 monthly disturbed days determined from the Kp index, respectively, and the data are available online (http://wdc.kugi.kyoto-u.ac.jp/qddays/). We excluded these international D-days from the analysis, as the geomagnetic disturbance can affect the EPB activity in both positive and negative ways. Figure 5 shows a similar analysis to that in Fig. 4 for GAIA temperature (panels (a) and (b)), eastward wind (panels (c) and (d)), and northward wind (panels (e) and (f)). Data for the three most EPB-active and EPB-inactive days are shown as solid and dashed curves, respectively, in each panel. Each 3-day average of active and inactive days is further averaged over 6 h from 0630 UT until 1230 UT. Data are plotted with a standard deviation (vertical bars) around the average. For example, in panels (a) and (b), temperature perturbations are enhanced on EPB-active days compared to those on EPB-inactive days in both 2003 (panel (a)) and 2004 (panel (b)). The difference well exceeds the standard deviation. The longitudinal ranges of maximum temperature perturbation are 40–85 E and 60–90 E in 2003 and 2004, respectively. These ranges are well associated with the longitudes of the enhanced R areas that are marked as “D” and “E” in Figs. 2 and 3, respectively. In the wind velocity, data in panels (c)–(f), a similar EPB-active day enhancement is persistent over both the eastward and northward wind components in both 2003 and 2004. The amplitude of eastward wind perturbations is generally greater (2–3 times larger) than that of northward wind perturbations. In addition, the EPB-active day enhancement in 2003 is clearer than in 2004. This yearly variation may be associated with decreased solar activity from 2003 to 2004. However, further analyses are necessary to clarify this behavior.
Finally, we show the results of the same analysis at different altitudes. Figure 6 shows a comparison of temperature perturbations between EPB-active and EPB-inactive days at altitudes of 5, 20, and 100 km for the data in 2003. At an altitude of 5 km (panel (a)), amplitude of fluctuations on both EPB-active and EPB-inactive days are smaller compared with that at other altitudes. The differences between the solid and dashed curves are also small, but are enhanced more than the standard deviation in the 50–85 E longitude region. We should note that this appears at longitudes of high R at an altitude of 40 km (Fig. 5a). Considering the general behavior of the atmosphere, in which disturbances increase with altitude in accordance with an exponential decrease in density, this small but detectable solid and dashed line difference in the troposphere may be the seed of EPB activity. At an altitude of 20 km (panel (b)), the EPB-active day enhancement becomes more obvious compared to the results at an altitude of 5 km, but it is less clear compared with that at an altitude of 40 km. At an altitude of 100 km (panel (c)), the EPB-active day enhancement seems existing, but it is not very clear. This is because the standard deviation around the average is much larger than at lower altitudes and may reflect the natural behavior of the upper atmosphere. Another explanation is technical: an altitude of 100 km is much higher than the top altitude of the data assimilation (30 km), so that at higher altitudes the model is less forced by real data. Padatella et al. (2014), for example, compared data assimilation results across different whole-atmosphere models and found that the model-to-model difference increases above the stratopause region owing to variations in calculations, assumptions, and/or grid sizes across models.
Summarizing the results shown above, we suggest that differences in the atmospheric perturbations between EPB-active and EPB-inactive days are seeded in the troposphere, grow in the stratosphere, and reach the upper atmosphere.