After sunset, plasma density depletions, also called equatorial plasma bubbles (EPBs), sometime occur in the equatorial- and low-latitude ionosphere. A large number of studies have shown that EPBs generally start to develop shortly after sunset during geomagnetic quiet periods (e.g., Weber et al., 1980; Kelley et al., 1986; Xiong et al., 2010; Wu et al., 2018). It is generally believed that the Rayleigh–Taylor instability (RTI) is a plausible mechanism to trigger the EPBs (Kelley, 2009; Makela and Otsuka, 2012). The growth rate of RTI is influenced by a number of different factors, such as the zonal electric field, neutral wind, vertical gradient of plasma density at the bottom side of the F region, or ion-neutral collision frequency, as well as the strength of magnetic fields (Ott, 1978; Abdu, 2001; Burke et al, 2004). The pre-reversal enhancement (PRE) of the eastward electric field around sunset is a main reason for the development of EPBs (e.g., Fejer et al., 1999; Abdu, 2001; Kelley, 2009; Huang, 2018). Owning to the intensified eastward electric field, near the magnetic equator the ionosphere is rapidly elevated to higher altitudes via E×B drifts, which is favorable for the growth of RTI at the bottom side of the ionosphere.

The EPBs are thought to extend along magnetic field lines, and they can reach as high as magnetic latitudes of about ±20∘ (Kelley, 2009; Lühr et al., 2014). Xiong et al. (2016, 2018) suggest that EPBs have a typical zonal size of about 50 km, by using Swarm in situ electron density measurements as well as the ground-based airglow imager. Although the characteristics of EPBs have been widely studied, special events, especially those occurring during geomagnetic storms, are still one of the interesting issues to be fully addressed. Some of the results showed that geomagnetic storms can affect the development of EPBs (e.g., Abdu et al., 2003; Tulasi et al., 2008; Carter et al., 2016), and in some extreme cases, the EPBs can extend to middle latitudes during intense geomagnetic storms (e.g., Sahai et al., 2009; Patra et al., 2016; Katamzi-Joseph et al., 2017; Aa et al., 2018). Moreover, during storm time, EPBs near sunrise were occasionally observed by some instruments such as radar and satellite. Fukao et al. (2003) used observations from the Equatorial Atmosphere Radar to report EPBs near sunrise over the Indonesian region during a geomagnetic storm and suggested that the EPBs were likely associated with the geomagnetic storm. Huang et al. (2013) reported the observations of long-lasting daytime EPBs with the Communications/Navigation Outage Forecasting System (C/NOFS) satellite during a geomagnetic storm in which the EPBs were persistent from the post-midnight sector through the afternoon sector. Zhou et al. (2016) used observations from multiple low Earth-orbiting satellites, like the Swarm constellation, the Gravity Recovery and Climate Experiment (GRACE) satellite, and the C/NOFS satellite, to detect the EPBs around sunrise during the St. Patrick's Day storm. They suggested that the geomagnetic storm induced changes in ionospheric dynamics should be the reason for triggering the EPBs. But until now, there has been no research on the occurrence characters and evolution of EPBs around sunrise using optical remote sensing, which can provide different aspects of the EPBs near sunrise.

It is well known that the EPBs usually drift eastward as reported by many studies (e.g., Pimenta et al., 2001; Martinis et al., 2003; Park et al., 2007; Taylor et al., 2013; Wu et al., 2017). However, during storm periods westward-drifting EPBs have been also observed (Abdu et al., 2003; Basu et al., 2010; Santos et al., 2016). Abdu et al. (2003) reported some cases of EPBs that showed eastward drifts after sunset which later reversed to westward drifts. Basu et al. (2010) reported that the westward-drifting EPBs reached maximum velocities of about 80–120 m s-1. Santos et al. (2016) also showed some EPBs of zonal drifts reversal (eastward to westward) during a geomagnetic storm, and they suggested the reversal was caused by a vertical Hall electric field which was induced by a zonal prompt penetration electric field (PPEF) in the presence of enhanced conductivity in the E region during night.

From 6-year observations of airglow images located in southern China, we found only one case of EPBs starting to appear near sunrise during the storm recovery phase on 8 November 2015. The EPBs appeared before sunrise, kept developing, and vanished about 1 h after sunrise. Unlike the quiet-time eastward drifting EPBs, the EPBs drifted westward. In the rest, we provide a detailed analysis of this event. In Sect. 2, we give a general description of the instruments. Observational results are showed in Sect. 3. In Sect. 4, we provide comparisons with previous studies as well as discussions. Finally, the summary is given in Sect. 5.

Figure 2 shows the 3 h Kp index, the interplanetary magnetic field (IMF) Bz, auroral electrojet indexes (AE, AU, and AL), the symmetric disturbance field in H (SYM-H), and virtual height (h′F) at Fuke on 6–8 November 2015. To make the comparison easier with other observations, we converted the universal time to the local time (LT) at Qujing. A geomagnetic storm occurred during those days. In Fig. 2b, IMF Bz turned southward at ∼11:40 LT on 7 November 2015, and it reached to about -11 nT at ∼16:00 LT. During the storm main phase, the SYM-H had a rapid reduction from -40 to -100 nT. Meanwhile, the Kp index reached a value of 6; the AE and AL also reached ∼1500 and ∼-1500 nT, respectively. After 04:00 LT on 8 November 2015, IMF Bz began to turn to north. In the storm recovery phase, the value of SYM-H was back to -40 nT.

Figure 3 shows the time sequence of airglow images observed by the all-sky imager at Qujing from 05:15 to 06:21 LT on 8 November 2015. The time difference between the successive images is 6 min. For each image, we removed the effects of compression and curving of the all-sky imager lens by an unwarping process (Garcia et al., 1997). All images have been mapped into a geographic range from 97 to 111∘ E in longitude and from 18 to 32∘ N in latitude. The height of the airglow layer is assumed to be at 250 km. The top of each image is to the north, and the right is to the east. Two EPBs, marked as “b1” and “b2”, were observed by the all-sky imager during this period. They occurred during the geomagnetic storm recovery phase.

Around 05:21 LT, EPB b1 appeared in the FOV of the all-sky imager. EPB b1 was still developing, as it extended northward and reached close to 25∘ N around 06:21 LT. At 05:39 LT, the other EPB, b2, started to appear in the FOV of the airglow imager. EPB b2 was also developing and expanded to about 20∘ N at 06:21 LT. The two observed EPBs possibly continued to develop after 06:21 LT, as no hints of stopping can be seen in the last airglow image. However, there was no further image data after 06:21 LT because the all-sky imager had to be shut down after sunrise. We want to point out that the sunrise time at Qujing was around 06:15 LT at an altitude of 250 km on that day. The far north part of b1 reached about 24.5∘ N at 06:15 LT. After 6 min, the far north of b1 extended to about 25∘ N (as marked by the black horizontal line). In other words, the observational result from the all-sky imager suggested that the EPBs kept developing after sunrise.

Some interesting features can also be seen from Fig. 3. EPB b1 appeared at ∼105∘ E, and b2 appeared at ∼104∘ E at 05:39 LT. Based on the black vertical line at 106∘ E, we can clearly see that the two EPBs drifted from east to west. Besides this, bifurcation and merging processes of EPB b1 were also observed. After 05:45 LT, a bifurcation process occurred in b1. The lower latitude portion of b1 moved further westward. An obvious cleft occurred at ∼19∘ N of b1 near 06:03 LT. More interesting is the fact that a merging process occurred in the two bifurcation portions of b1 during its later development period. After ∼06:03 LT, the upper portion of b1 began to connect to the lower portion of b1, and they merged or combined together into one EPB after 06:15 LT. The bifurcation and merging processes are more obvious in the red rectangles of Fig. 3, which is indicated by the red arrow in each image.

Figure 4 shows a series of TEC residuals over 10–50∘ N and 80–130∘ E during 04:30–08:20 LT on 8 November 2015. The adjacent imaging is in 10 min intervals. At about 04:40 LT, some TEC depletions, which occurred to the south and west of the location of the all-sky imager, appeared at ∼115∘ E (∼24∘ N) and began to develop. About 05:30 LT, some additional EPBs appeared at ∼105∘ E (∼20∘ N), and they were also developing. EPBs in the two regions kept developing until they disappeared. Owning to the FOV of the all-sky imager, the EPBs outside the ∼115∘ E region were not observed.

In order to provide a much more detailed comparison between the all-sky imager and TEC measurements, we give the local-time variation of the absolute TEC after 05:15 LT (Fig. 5), which corresponds to the geographical area of airglow imaging. In Fig. 5, the TEC depletions at ∼105∘ E appeared near 05:30 LT, which correspond to EPB b1 and b2 observed by the all-sky imager. And after ∼07:45 LT, those TEC depletions disappeared. For a better representation, we showed ROTI variations which correspond to the geographical area and time of each airglow imaging of Fig. 3. In Fig. 6, the ROTI enhancement at ∼105∘ E also corresponds to EPB b1 and b2 observed by the all-sky imager near 05:30 LT. The ROTI enhancement moves away from 106∘ E with time (the black vertical line represents the 106∘ E in Fig. 6), which is consistent with the movement of EPBs observed by the airglow imager. Meanwhile, the northernmost part of the ROTI enhancement expanded to ∼25∘ N at 06:21 LT (the black horizontal line represents the 25∘ N in Fig. 6), which also agreed well with the observations of the all-sky imager. Interestingly, in Fig. 4, TEC residuals show that the northernmost EPBs at ∼105∘ E extended beyond 25∘ N after 06:20 LT. We can see that the northernmost of them reached about 28∘ N at 07:10 LT. In other words, TEC variations show that those depletions were still in existence after 06:21 LT, and they kept developing after sunrise but vanished near ∼08:00 LT. These observational results shown that the lifetime of those EPBs exceeds 3 h.

In this study we showed an special event of EPBs which was simultaneously observed by the all-sky imager and the ground GNSS network in southern China. One interesting feature is that the EPBs started to appear near sunrise. Afterward, they kept developing until they totally vanished. During their lifetime, the EPBs moved from east to west. Those EPBs occurred in the recovery phase of the geomagnetic storm, which indicates that the prompt penetration electric fields (PPEF) and disturbance dynamo, as well as disturbed neutral wind circulation, may play an import role in triggering the EPBs.

The drift velocities of EPBs were shown in Fig. 7. We used the cross sections (keogram) (Fig. 7a, c, and e) of the airglow images to separately calculate the meridian velocities (Fig. 7b) of b1 and zonal velocities of b1 at the geographical latitudes of ∼22∘ N (Fig. 7d) and ∼19∘ N (Fig. 7f). Figure 7a illustrates the N–S cross sections (between 104 and 105∘ E) of the airglow images shown in Fig. 3. Figure 7c illustrates the W–E cross sections (between 21.5 and 22∘ N) of the airglow images, and Fig. 7e illustrates the W–E cross sections (between 18.5 and 19∘ N).

We separately calculated poleward and zonal velocities of b1 based on the position of which it changed over time in Fig. 7a, c, and e. The initial poleward and zonal velocities of b1 were about 200 and 60 m s-1, respectively. The horizontal drift of an EPB is also an important issue, which is often related to the background zonal plasma drift (Fejer et al., 2005; Eccles, 1998). The westward motion of the F region should be caused by the ionospheric dynamo process in the early morning (Kil et al., 2000; Sheehan and Valladares, 2004). The drift direction of the background zonal plasma drift has a reversal (eastward to westward) near dawn (Fejer et al., 2005). Huang and Roddy (2015) also found that the drift velocity of EPBs was eastward at night and reverses to westward near dawn by using data from C/NOFS, and they showed enhanced geomagnetic activities caused a westward EPB drift in the nighttime through a disturbance dynamo process. In our case, all EPBs emerged after 05:00 LT. The background plasma should drift westward during the early morning hours. So, it could partly explain why the observed EPBs drifted westward. In addition, the disturbed westward neutral winds can also contribute to the westward drifting of EPBs. Xiong et al. (2015) found that the disturbance winds were mainly westward at low latitudes, most prominent during early morning hours. Abdu et al. (2003) found that the westward drift of an EPB was most likely caused by westward zonal winds during a geomagnetic storm. Makela et al. (2006) found that the eastern wall of EPBs can become unstable due to the westward and equatorward neutral winds associated with wind surges. When the wind blows westward, and thus the wind-induced Pedersen current flows downward, gradient drift instability can occur at the eastern wall of an EPB, where the plasma density gradient is eastward. So, secondary instabilities are more likely to occur at eastern wall of EPBs. In Fig. 3, a sub-branch of dark bands first occurred at the eastern wall of b1, which is indicated by secondary instabilities developed at the eastern edge, most likely due to the westward disturbance winds.

In Fig. 8, we used the Thermosphere–Ionosphere–Electrodynamics General Circulation Model (TIE-GCM) to simulate the horizontal winds on 8 November 2015 under magnetically active conditions, and the latitude-versus-longitude distribution of zonal wind velocities are shown at different times. The winds at 250 km are shown, and the spatial coverage has been confined to 0–40∘ N latitude and 90–120∘ E longitude. The dashed rectangles represent the location of b1 and b2 at different times. In Fig. 8, we can see that the horizontal winds at low latitudes are mainly westward, which is consistent with the motion of EPBs in this case. As already discussed above, the westward drift of those EPBs is possibly caused by the westward disturbance winds. Besides this, the zonal winds computed from TIE-GCM shown in Fig. 8 are smaller than the zonal drifts of EPBs shown in Fig. 7. This is because zonal drift value of EPBs was controlled by the background zonal winds and ionospheric electric field (Haerendel et al., 1992; Eccles, 1998). The value differences between the simulation and zonal drifts of EPBs should be influenced by the ionospheric electric field. Besides this, the difference between the model simulated background zonal winds and the derived zonal drifts of EPBs from airglow images is possibly due to the fact that the model simulation mainly reflects a general trend of the wind but not the exact wind velocity in reality.

As reported, most of the EPBs start to occur at pre-midnight hours. There were a very limited number of studies that used data from radar or satellite to report the occurrence of an EPB close to sunrise hours (e.g., Fukao et al., 2003; Huang et al., 2013; Zhou et al., 2016). However, until now, there has been no observation result of EPBs around sunrise using optical remote sensing. In fact, it is very difficult to observe an EPB near sunrise by an all-sky imager. Often, EPBs start to develop shortly after sunset and vanish before sunrise. Even though some EPBs occur around sunrise in their initial stage, they disappear when they drift eastward into the daytime. And almost no report shows that the EPBs still kept developing after sunrise. In our case, the developing EPB was first observed at about 05:30 LT (near dawn) by both the all-sky imager and the GNSS network. The local-time variation of absolute TEC showed that EPBs existed after sunrise and they disappeared after 07:45 LT. Our observational results show that they kept developing after sunrise, and they vanished about 1 h after sunrise. Those EPBs should have occurred near sunrise, which is different from post-sunset EPBs. Their development stages lasted for at least about 3 h.

In the rest, we try to explain why the EPBs occurred near sunrise. During storm time, disturbance winds can affect the low-latitude ionospheric electrodynamics as well as the zonal drift of an EPB. The ionospheric disturbance dynamo electric field (DDEF) caused by a storm will drive the plasma drift to move upward during nighttime (Blanc and Richmond, 1980). Meanwhile, a number of studies found that the high-latitude electric fields can penetrate into the middle- and low-latitude ionosphere as PPEF when IMF Bz turns southward or northward (Kelley et al., 1979; Scherliess and Fejer, 1997; Cherniak and Zakharenkova, 2016; Carter et al., 2016; Patra et al., 2016; Katamzi-Joseph et al, 2017). For the storm event, after IMF Bz turned southward at ∼12:00 LT on 7 November 2015, there was a long duration and high AE during storm time. A DDEF should be present at the recovery phase of storm time. And it reversed the ambient electric field from westward to eastward near sunrise, which enhanced the height of the bottom side of the ionosphere F region. Meanwhile, the northward turning of IMF Bz at ∼04:00 LT on 8 November 2015 caused an overshielding electric field, which produced an eastward PPEF into the low–middle-latitude ionosphere. The eastward electric field also moved the F-region ionosphere to higher altitudes via vertical E×B drifts. In Fig. 2e, the increased height of bottom side of the ionosphere F region can be seen at Fuke. In low-latitude regions, one of the necessary conditions for the generation of EPBs is that the F layer should be uplifted to a higher altitude, where the RTI becomes unstable and forms EPBs. The F-layer height is largely determined by the eastward field via the vertical E×B drift (Dabas et al., 2003).

In this study, EPBs were initially observed by the all-sky imager at about 05:15 LT. We think that only a portion of the EPBs were observed in our study, as an EPB usually extends along the whole magnetic flux tube. It also means that the EPBs should possibly occur before 05:15 LT at equatorial latitudes. But due to the lack of observations at the Equator, we cannot provide direct evidence about their generation. However, as shown in our Fig. 9, we also found that spread F began to appear in the ionograms from the digisonde at Fuke after 05:15 LT, which indicates that those EPBs occurred in the region of southeastern Qujing (note that Fuke is to the southeast of Qujing). The bottom side of the ionospheric F region at Fuke was rapidly elevated from ∼250 to ∼290 km near sunrise on 8 November 2015. The rapidly elevated height of the ionosphere can cause a stronger RTI at the bottom of the ionosphere F region, which is beneficial to the formation of an EPB. The initial occurring time of EPBs in this case should be during this time. Unfortunately, we do not have more observations from the southeast of Fuke. We used TIE-GCM to simulate the height of hmF2 at lower latitudes on 8 November 2015. Figure 10 shows the hmF2 as a function of longitude and latitude at different times. The model results plotted are in a geographic range from 0 to 40∘ N in latitude and from 90 to 120∘ E in longitude. In Fig. 10, we can see that hmF2 southeast of (the dashed rectangles) Qujing was rapidly elevated to higher altitudes near sunrise. In other words, when the IMF Bz turned northward at about 04:00 LT, the ionosphere in some regions southeast of Qujing could be rapidly elevated to higher altitudes at this time. Those EPBs occurred in the same time period as highlighted by the green rectangular area in Fig. 2. Previous studies have reported that the occurrence of the dawn enhancement in the equatorial ionospheric vertical plasma drift (Zhang et al., 2015, 2016). They found that the enhancement of the ionospheric vertical plasma drift occurs around dawn. They suggested that the vertical plasma drifts can be enhanced near sunrise in a way similar to the PRE near sunset. Fejer et al. (2008) found that the nighttime disturbance dynamo drifts are upward, and they have the largest values near sunrise. In our case, the model simulations and observations both show an increase of the height of the ionosphere around sunrise. The enhancement of low-latitude ionospheric vertical plasma drift caused by DDEF and PPEF associated with the geomagnetic storm should play a vital role in triggering those EPBs. Our results also provide evidence of the enhancement of low-latitude ionospheric vertical plasma drift around sunrise, which should be the main reason of the EPB generation near dawn.

In addition, some interesting features of EPBs are also shown in Fig. 3 in that the EPBs also showed bifurcation and merging processes. The merging phenomenon of EPBs has been studied by some researchers (Huang et al., 2012; Huba et al., 2015; Narayanan et al., 2016; Wu et al., 2017). However, there has been no study to report that bifurcation occurred first and merging occurred later in the evolution of an EPB. In Fig. 7f, at a latitude of 19∘ N, the zonal velocity of b1 was about 60–70 m s-1 between 05:20 and 06:15 LT. However, at the latitude of 22∘ N (Fig. 7d), the zonal velocity of b1 was decreased from about 70 m s-1 to about 50 m s-1 between 05:20 and 05:45 LT. After 05:45 LT, its velocity began to increase from ∼50 to ∼70 m s-1 from 05:45 to 06:00 LT. Then, it kept a velocity of ∼70 m s-1. Owning to the fact that the zonal velocity at higher latitudes was smaller than that at low latitudes before 05:45 LT, b1 had a bifurcation process of EPBs during this period. After 05:45 LT, the zonal velocity at higher latitudes was bigger than that at lower latitudes, and b1 exhibited a merging process of EPBs after 06:03 LT. The above results indicate that the bifurcation and merging processes of EPBs should be caused by the different drift velocities of the background plasma at different latitudes.

In this paper, a special EPB event was observed by an all-sky imager and the GNSS network in southern China. The evolution processes and characteristics of those EPBs were studied in detail. Our main findings are summarized below.

The observed EPBs on 8 November 2015 emerged before sunrise and kept developing. They dissipated at about 1 h after sunrise (∼ after 08:00 LT), and the development stage lasted for at least about 3 h. The evolution of EPBs developing around sunrise was observed for the first time by an all-sky imager and the GNSS network.