Magnetic Storm Effects on the Occurrence and Characteristics of Plasma Bubbles

The Communications/Navigation Outages Forecast System satellite mission was designed to investigate the ionospheric conditions that lead to the formation of irregularities. Here, we have studied the effect of magnetic storms on the formation and evolution of plasma bubbles during the satellite's lifetime (2008–2015). During this period encompassing solar minimum and maximum conditions, many magnetic storms of varying intensity developed, producing a unique and rich data set of 248 storms (14 intense, 69 moderate, and 165 weak) that occurred during the same timeframe to examine the role of external magnetospheric drivers in the production and dynamics of equatorial plasma bubbles. We have used the Planar Langmuir Probe and Ion Velocity Meter instruments to elucidate the role of magnetic storm intensity on the bubble's depth, internal speed, width, occurrence, and lifetime. The pre‐reversal enhancement (PRE) tends to increase during the main phase and when BZ is southward. New bubbles occur during large excursions of the PRE value. The bubble lifetime extends and remains active during the main and part of the recovery phase. The plasma velocity within the bubbles increases and typically becomes over 100 m/s during significant PRE and BZ negative times. The depth of bubbles reaches values close to 100% during intense storms. In general, the intensity of the storms seems to control and augment the plasma bubbles' depth, width, and internal velocity.


10.1029/2023JA031292
2 of 17 Equatorial spread-F corresponds to an unstable state of the low latitude F-region, discovered by Booker and Wells (1938).ESF comprises bubble-like structures that develop mainly after sunset in the nighttime ionosphere that, under some circumstances, don't decay until dawn.The occurrence of the bubble-like structures known as plasma bubbles during quiet periods (i.e., in the absence of geomagnetic storms) has been thoroughly studied over the past few decades (Hei et al., 2005;Kil & Heelis, 1998;Stolle et al., 2008).The initiation of the plasma bubbles follows the mechanism known as the Rayleigh-Taylor (RT) instability.The growth-rate of the RT instability varies with several factors, including the vertical drift of the plasma, the thermospheric neutral wind velocity, the height of the F layer, and the bottomside vertical density gradient.It was realized in the 70s that the development of plasma bubbles depends on seeding mechanisms like gravity waves, which are necessary to initiate the bubble development when environmental conditions are insufficient for producing the plasma bubbles (Krall et al., 2013;Röttger, 1973;Tsunoda, 2010aTsunoda, , 2010b)).
The growth of the plasma bubbles can sometimes be predicted by the change in the PRE velocity, that is, the peak velocity observed during sunset hours.If the PRE is high, the growth rate of the plasma bubbles increases.As mentioned before, during geomagnetic storm the eastward PPEF in the equatorial ionosphere increases the vertical plasma drift.Based on these arguments, it was expected that PPEFs during the storm main phase would enhance bubble production and increase bubble depth and vertical bubble velocity.The effect of the stormtime PPEF can be observed to enhance bubble occurrence in specific longitude sectors, as indicated by Basu et al. (2007Basu et al. ( , 2010)), Huang (2008, 2011), and Abdu (2012).
The variation of the growth rate of RT instability creates a seasonal-longitudinal and solar activity variation of plasma bubble occurrence during quiet periods.For most of the longitudinal sectors, the occurrence of plasma bubbles maximizes when the sunset terminator aligns with the magnetic field line (Burke et al., 2004;McClure et al., 1977;Tsunoda, 1980).The production and development of plasma bubbles varies largely with solar activity, with the frequency of bubble occurrence being greater during solar maxima than during solar minima.Studies by Smith and Heelis (2017, 2018a, 2018b) show the variation in the characteristic features of the plasma bubbles, that is, bubble depth, bubble width, and vertical bubble velocity, using data from the C/NOFS satellite, with varying solar activity levels over different longitudinal sectors.
This study aims to examine the occurrence and characteristics of plasma bubbles and find a relationship between bubble production and the development of the different magnetic storm phases (i.e., PPEFs).The methodology section (Section 2) of the paper describes the data and algorithms used to detect geomagnetic storm signatures from the variation of the SYM-H index every minute and to detect plasma bubbles from a 1 Hz variation of plasma density data measured by the C/NOFS satellite.The results section (Section 3) of the paper demonstrates the magnetic conditions and plasma bubble generation during two intense magnetic storms that occurred on 09 March 2012 and 01 June 2013.The discussion section (Section 4) of the paper deals with the results and conclusions from the study.

Methodology
This section describes the measurement techniques used to isolate storm signatures and identify storm phases using the SYM-H index, determine PRE peak velocities, and detect plasma bubbles using measurements from the IVM and PLP instruments of the C/NOFS satellite.The SYM-H index data at a resolution of 1 sample per minute were obtained from the H component of the magnetic field measured at various geomagnetic observatories located in mid to high latitudes.The 1 sample per minute north-south Interplanetary Magnetic Field (IMF) data, shifted to the location of the Earth's magnetopause, were obtained from the ACE satellite measurements.
The Communication/Navigation Outage Forecasting System (C/NOFS) satellite was designed to study the different processes that cause plasma instabilities to grow and how these instabilities affect the propagation of radio waves through the ionosphere.Launched on 16 April 2008, with an inclination of 13°, a perigee of 405 km, apogee of 853 km, and a period of approximately 93 min, C/NOFS measured electron density and drift velocities in altitudes ranging from 400 to 800 km.

Storm Signature and Phase Detection
During geomagnetic storms, the SYM-H index often first shows a slight rise, called the Storm Sudden Commencement (SSC), which lasts from a couple of minutes to an hour.The SSC is also called the initial phase of storms.
The key signature of a storm is the rapid decrease in SYM-H that follows the SSC.This is called the storm's main phase, which lasts from a few hours to a day.For the rarely occurring super intense geomagnetic storms, the minimum SYM-H index falls below −250 nT, but for most other storms, the minimum SYM-H index falls between −30 nT and −250 nT.We divided the storm strengths into intense, moderate, and weak storms, in which the SYM-H index minima fall between −100 nT to −250 nT, −50 nT to −100 nT, and −30 nT to −50 nT, respectively.The SYM-H minima also mark the end of the main phase.Following the main phase, the SYM-H index typically shows a slow recovery to the quiet time average value of −15 nT.The recovery phase can be divided into two parts.During the first part, the SYM-H index recovers to half of the original decrease.During the final part, the SYM-H index recovers fully to quiet time levels.For this study, we used the SYM-H index to identify magnetic storms and then select the different phases of the storm.Recurrent storms in which the SYM-H does not go above −30 nT were considered one storm.Resulting data sets of all storms that were detected using the method are added as Data Set S1.

PRE Peak Detection
To evaluate the PRE peak velocities observed during local sunset hours, data from the IVM instrument on C/ NOFS were used.The data contained plasma velocities in the vertical direction from the satellite and in the meridional direction.The resultant of these two velocities and the magnetic dip angle determined by the IGRF model, were used to determine velocity of the plasma perpendicular to the direction of the magnetic field lines at different locations of the satellite.Moving averages of the resulting velocity over 2 min were taken at first to remove noise and small-scale perturbations due to plasma irregularities.The resulting data approximately represented the vertical velocity of the background plasma in a direction that is perpendicular to the magnetic field lines.The C/ NOFS satellite crossed the sunset terminator once every 100 min during its lifetime.The maximum of the background plasma vertical velocity observed within the solar local times (SLT) from 1800 to 1900 hr during each satellite pass was considered the PRE peak velocity corresponding to the pass (Huang, 2018).Furthermore, only drift velocities observed within the magnetic latitude of less than 20° were considered for the analysis.

Bubble Detection
The electron density data from the PLP instrument on-board the C/NOFS satellite was used to identify and characterize plasma depletions observed along the satellite orbit.To differentiate plasma bubbles that are embedded in the background density, we developed a new bubble detection algorithm.First, the 32-512 Hz data was averaged to obtain 1 Hz data and remove small-scale variations inside the plasma bubbles.Electron density measurements greater than 10 8 cc −1 were then eliminated as instrumental noise.The resulting data were further processed to extract the universal times (UTs) corresponding to the bubble boundaries and characteristics, that is, depth, width, and vertical velocity.The background or undisturbed plasma density was obtained by performing six iterations.Each iteration consisted of averaging the electron density data and replacing the average with an electron density data that is greater than the corresponding average data.In the first iteration, the average was over 160 s (≈1,200 km).The average closely traced the background plasma density, but as the electron density data showed sudden variations within the plasma bubbles, the average was either greater or lesser than the electron density in the vicinity of the plasma bubbles.Figure 1a shows the result of the first iteration, here, the light blue trace indicates the averaged values.Any average data less than the electron density data were then replaced by the corresponding electron density data.Average densities greater than the measured electron density data were kept.The resulting average data were further averaged in the second iteration, which was taken over 80 s.Following the second iteration, the average data were replaced as described in the previous step.In the third, fourth, fifth, and sixth iterations, averages were taken over 40, 20, 10, and 5 s respectively, and the average data were modified as described in the previous iterations except for the final iteration.The average obtained in the final iteration closely represented the background plasma density (N bg ). Figure 1b displays the averaged density closely resembling a visual envelope or undisturbed density (N bg ).The background plasma density was then subtracted from the measured electron density data (N e ), and the difference was divided by the background plasma density to obtain the relative change in density throughout the data (see Figure 1c).As the background plasma density closely traced the electron density data when bubbles were absent, the relative change in density values was almost equal to zero, but within the plasma bubbles, the relative change in density values ranged from −100% to close to 0 with lower values representing deeper and higher values representing shallower depletions.As the electron density data ranges over several orders of magnitude, it was necessary to find the relative change in density throughout the data.Finally, to detect the bubble boundaries, we selected a relative change in density data less than the threshold value of −0.1.The UT corresponding to the data when the relative change in density values crossed the threshold with a negative slope were taken as the first wall of the bubble.The UT corresponding to the data when the relative change in density values crossed the threshold with a positive slope was taken as the second wall of the bubble.As the C/NOFS satellite traversed the Earth from west to east, the first wall to be detected was the west wall, and the second wall was the east wall of each of the detected bubbles. Figure 1d shows a series of plasma bubbles, highlighted in light blue, detected using the algorithm described above.
Following the detection of the bubbles, the depths, widths, and vertical velocities corresponding to the bubbles were determined (see Figure 1e).Bubble depths were calculated by taking the maximum percentage relative change in density within each bubble in %, that is, D bub = (| e /N e |) max × 100% where ΔN e denotes the absolute change in density and N e denotes the background density corresponding to the point of the maximum relative change in density within each bubble (Smith & Heelis, 2018a).Bubble widths were calculated by taking the distance between the two walls of each bubble in km, that is, w bub = (UT east − UT west ) × 7.5 km/s where the UT east and UT west denote the UTs corresponding to the east and west walls of each bubble with the time difference measured in seconds.The 7.5 km/s is the spacecraft velocity.Vertical bubble velocities were determined by taking the maximum vertical drift velocity observed between UT east and UT west in m/s), that is, The detected bubbles were further validated by using the bubble depth and a defined parameter called the "depth-to-width ratio," representing the ratio of the bubble depth and width.Bubbles with a depth of less than 25% and a depth-to-width ratio of less than 0.1 were discarded as background plasma density fluctuations.In addition, we constrained our study to bubbles with background density more than 10 5 cc −1 , to account for the large altitude variation of the satellite.

Results
This section shows the magnetic conditions and plasma bubbles characteristics observed with C/NOFS during the storms of 9 March 2012 and 1 June 2013.During these two intense storms, we observed enhancements in PRE peak velocities, changes in plasma bubble depths, and increases in vertical bubble velocities.Figures 2 and 5 describe the limits of the initial, main, and recovery phases, the magnetic conditions, and observed PRE peak values during these two storms.Figures 3 and 6 show orbit-by-orbit observations of plasma densities and velocities observed by C/NOFS during the primary stages of bubble development during the storms.Figures 4  and 7 present scatter plots of the depths, widths, vertical velocities, and SLT of observations of bubbles during the two storms.Finally, Figure 8 compares the enhancements of bubble depths, widths, and vertical bubble velocities during 14 intense storms observed during the lifetime of C/NOFS.

Intense Storm of 09 March 2012
Figure 2 shows the variation of the SYM-H indices (top panel), the IMF B Z (middle panel), and the PRE peaks (bottom panel) with respect to UT during the storm of 09 March 2012.The start of the storm's initial, main, and recovery phases are marked with vertical lines, respectively, in order of increasing UT.
The value of SYM-H (Figure 2a) remained close to −15 nT for about 12 hr before the beginning of the storm with onset on March 9 at 01:11 UT.The storm's main phase started at 2:11 UT on March 9 and lasted for about 6 hr when the SYM-H index showed a rapid reduction in value and decreased from the quiet time value of −15 nT to a minimum of 149 nT.The decline occurred in two steps.During the first step, SYM-H reduced to about −100 nT and then increased to approximately −50 nT, it decreased again in the next 2 hr and reached the minimum of −149 nT.The final reduction showed a rate of change in SYM-H of about −50 nT/hr.After the end of the main phase, SYM-H started to recover during the first half of the recovery phase and reached approximately −68 nT in about 13.5 hr.During the second half of the recovery phase, SYM-H returned to its quiet-time value of −15 nT slowly over more than 1.5 days.IMF B Z was significantly southward (−15 nT) during the storm's main phase (see Figure 2b).The green points in this panel present the 1 Hz value of B Z during the storm interval, and the black points represent the 15-s averaged values.Before the initial phase, B Z remained slightly northwards for about 12 hr.Prior to storm onset, B Z was northwards during the increase in SYM-H and before the SYM-H reduction.At the beginning of the main phase, B Z changed direction and turned significantly southward with a value of −15 nT.B Z remained southward during the rest of the main phase, about 6 hr, with a constant value of −15 nT.B Z continued to remain southward during the first half of the recovery phase but showed a gradual reduction in value until 15:00 UT on March 9.The variation of PRE peaks detected by the C/NOFS satellite is plotted in panel c.Each PRE peak value in the plot represents the maximum vertical drift velocity of the background plasma between 18 and 19 LT.During magnetically quiet times and before the storm's start, the PRE peaks typically show small day-to-day fluctuations in values.After the start of the main phase, when B Z became significantly southward, the PRE peaks showed a steady increase in value to a maximum of 68 m/s between 6 and 8 UT on March 9.The increment in the PRE peaks indicates the penetration of the PPEF from higher latitudes.The high PRE peaks and vertical velocity favor the development of plasma bubbles during main storm phases.The PRE peaks continue to decrease even when B Z remains southward during the first half of the recovery phase to less than 0 m/s, that is, the PRE peaks approached zero, and the vertical velocity of the background plasma became negative.The value of the peaks remained negative for the rest of the first half of the recovery phase except at a point when it showed a slight rise with a value of 25 m/s (06 UT on 10 March 2012).During the rest of the recovery period, the PRE peaks showed a slight increase in value.There is a gap in the data as the magnetic latitudes were beyond 20° during the sunset hours for several orbits near the end of the first half of the recovery phase.and 3c show yellow patches that demarcate the local time of 18-19 hr, that is, the sunset local times during the satellite orbits.For simplicity in descriptions, the observation regions in this orbit-by-orbit plots are divided into four longitude sectors: the African sector (0°-90° longitude), the Indian sector (90°-180° longitude), the Pacific sector (180°-270° longitude) and the South American sector (270°-360° longitude) which are labeled as "A," "I," "P" and "SA" in the figure .Panel (b) shows 13 orbits during the storm's initial, main, and recovery phases.During the first orbit, the increase in the SYM-H represents the sudden commencement during the initial phase.Orbits 21157 to 21159 correspond sunset hours became more significant, which implies the production of new bubbles on the next orbit.Although the bubbles in the Pacific sector decayed significantly, the bubble depths remained constant for the next 2 orbits, starting at 07:31 UT and 09:14 UT.The following orbit, starting at 07:31 UT, shows the development of new bubbles at the boundary of the Indian and Pacific sectors.Here, the bubble depths were relatively low during the orbit, but the bubble velocities were high and approximately 500 m/s.The bubbles in the Pacific sector showed almost no fluctuations in the vertical velocity.In the orbit that started at 09:14 UT, bubbles grew deeper in the Indian-Pacific sector boundary.The bubbles showed 3 orders of magnitude reduction in the plasma densities.The

10.1029/2023JA031292
11 of 17 corresponding vertical bubble velocities were very small or negative, indicating that the bubbles would decay afterward.During the orbit that started at 10:57 UT, bubbles in the Indian-Pacific sector boundary decayed, but the bubbles in the Pacific sector increased their depth.In the next orbit that started at 12:40 UT, the bubbles in the Indian-Pacific sector boundary showed enhancements in depth and in the following orbit that started at 14:24 UT, bubbles started spanning the longitudinal region and showed growth in depth.The corresponding bubble velocities were approximately 250 m/s.In the following orbit that started at 16:08 UT, bubble depths and widths increased while bubble velocities remained significant.In the next orbit that started at 17:52 UT, the series of bubbles have been partially replaced by an extended plasma decrease of ∼2,000 km.Although, some bubble-like structures were observed on one side of the plasma decrease.
Figure 4 shows a summary plot of the occurrence and characteristics of the bubbles observed during the storm of 09 March 2012.The beginning of the initial, main, and first and second halves of recovery phases are marked with vertical black lines Panel (a), (b), and (c) shows the bubble depths in percentage, widths in km, and vertical velocities in m/s with respect to the UT.The LT of the bubbles is indicated by the color of the circles as shown by the color scale on the right side of the figure.
Figure 4 reveals that the bubble occurrences during the main and the first half of the recovery phases were greater than bubble occurrences during the quiet time and the second half of the recovery phase.Panel (a) shows that during the quiet times, observed depths of the bubbles ranged up to 75%.As the main phase started and B Z turned southward, the average bubble depths increased, and deeper bubbles with depth >90% were frequently observed.At the beginning of the main phase, the bluish color of the bubble markers implies that there was a development of new bubbles, and the reddish colors of the bubbles denote that most bubbles didn't decay until midnight or dawn.Panel (b) shows that bubble widths did not vary much between the quiet period, the main, and the recovery phases.Panel (c) shows that the vertical bubble velocities were enhanced at the beginning of the main phase and several bubbles contained velocities higher than 230 m/s.Afterward, by the end of the main phase, the bubble velocities decreased.During the second half of the recovery phase, the bubble velocities increased again, but most of the bubble velocities remained under 370 m/s.This resurgence of the bubble velocity coincided with a period of B Z negative condition and when the PRE reached ∼50 m/s (see Figure 2c).

Intense Storm of 01 June 2013
Figure 5 shows the SYM-H index (nT), IMF B Z (nT), and PRE peaks (m/s) during the intense storm of 01 June 2013 from midnight of 31 May 2013 to 18 UT on 03 June 2013.The SYM-H index, at the top, shows that the main phase duration of this storm was shorter than the one on 09 March 2012.The main phase started at 2:04 UT and lasted for about 5 hr.The minimum SYM-H that was attained during the storm was −136 nT.The reduction of the SYM-H happened in two stages.At the beginning of the main phase, the SYM-H reduced very rapidly at a rate of −42 nT/hr from −15 nT to −113 nT in 2.3 hr.Finally, the SYM-H index decreased by 23 nT in the next 3.5 hr during the end of the main phase.Following the main phase, a slow recovery took place, which lasted until 03 June 2013.The middle panel reveals that the IMF B Z remained significantly southward with a nearly constant value of −15 nT during the initial and main phases of the storm.After the conclusion of the main phase, B Z rapidly approached zero and displayed a variation around zero throughout the recovery period.The bottom panel shows that the PRE was positive and large at the beginning of the storm's main phase, becoming negative about halfway through the main phase.During the 3-day interval, the PRE peaks showed large variations in magnitude.The highest magnitude of the peaks was 100 m/s, which occurred at the beginning of the main phase, probably due to the storm-time PPEF.During the second half of the recovery phase, PRE values remained below 35 m/s.The main difference with other intense storms is the short interval (6 hr) of B Z negative conditions.
Figure presents an orbit-by-orbit plot during the storm of 01 June 2013, similar to Figure 3. Panel (a) shows 13 orbits covering the quiet time before the storm, the initial, the main, and part of the recovery phase of the storm.The storm's initial phase started during orbit 27900.The main phase of the storm occurred during the next 4 orbits, and the storm recovery took place during the successive 6 orbits.The PRE peak velocity reached a maximum of 100 m/s during orbit 27901.The PRE peak velocities decreased from that point onwards.The area plot of IMF BZ shows that it was northward during orbits 27899 and 27900 and became significantly southward during orbits 27901 through 27904.During the remainder of the time interval shown, the magnitude of BZ remained close to zero.
Panel (b) shows an absence of bubbles during the first orbit that started at 19:56 UT.At the beginning of orbit 2 at 21:39 UT, plasma bubbles started appearing in the African sector.The bubble depths were significant that is, plasma densities inside the bubbles showed a significant decrease in value.The corresponding bubble velocities were relatively high, indicating that the bubbles were developing and growing.In the next two orbits that started at 23:21 UT and 01:04 UT, still under quiet conditions, the bubbles in the African sector grew significantly and became deeper and wider.The bubble velocities were also significant, with an average value of 250 m/s.In the next orbit that started at 02:46 UT, the bubbles in the African sector decayed, and the depths and widths of the bubbles decreased.During this orbit, the PRE peak velocity was the highest, as seen in panel (a).New bubbles started developing in the Pacific-South American sector boundary showing very high bubble velocities of more than 500 m/s.This region also coincides with the local sunset time.During the next 3 orbits that started at 04:28 UT, 06:10 UT, and 07:52 UT, the bubbles grew deeper and spanned most of the Pacific sector and a part of the South American sector.In addition, the bubble velocities remained very high during these orbits.In the following orbit that began at 09:34 UT, the bubbles started to decay and became shallow.Some bubbles still showed significant depth, but the bubble velocities decreased and became equal to the plasma background velocities.During the successive 3 orbits which started at 11:17 UT, 13:00 UT, and 14:42 UT, the bubbles showed decay but remained deep.During the last orbit of the interval shown, the bubbles decayed completely, and the bubble velocities remained near zero throughout the recovery phase.
Figure 7 shows the bubble depths, widths, and vertical velocities observed during the storm of 01 June 2012 in the same format as Figure 4. Panel (a) shows that the bubble depths observed during the disturbed period were slightly greater during the main phase when B Z was southward than bubble depths observed during the quiet or the recovery periods.The width of the bubbles was slightly larger during the main phase, although many wide bubbles were observed during the recovery period as well.The vertical bubble velocities showed significantly higher values during the main phase than the quiet time before and the recovery phase after.The overall count of observation of bubbles was higher during the main phase, and the blueish color of the markers denotes that most bubbles observed during the main phase were pre-midnight bubbles, that is, bubbles developed at the beginning of the main phase.The reddish color of the markers during the early recovery phase indicates that bubbles produced during the main phase did not decay until after midnight.Figure 8 also shows that the characteristics of the bubbles during quiet periods before the storms are positively correlated with the sunspot number, and bubble characteristics showed enhancement, that is, an increase in bubble depth, width, and vertical velocities for a majority of the storms.Especially, the vertical bubble velocities showed significant growth during several of the storms and became close to or greater than 600 m/s.

Discussions
This study aimed to improve our understanding of how the magnetic storm-induced E fields affect the production and evolution of EPBs during the main and recovery phases of intense, moderate, and weak storms that occurred during the lifetime of the C/NOFS satellite.We used density and velocity data collected by the C/ NOFS satellite that orbited the earth in the low latitude and equatorial regions for almost 7 years, covering all longitude sectors, all seasons, and solar activity levels allowing us to observe changes in the general quiet-time seasonal-longitudinal pattern of density depletions during magnetic storms.
Our study considered quiet magnetic conditions as the 24-hr period that occurred before the storm's commencement.Although these quiet periods do not reflect the seasonal-longitudinal or solar activity variations of plasma 10.1029/2023JA031292 15 of 17 bubble occurrences, these are used to highlight the changes in bubble occurrence pattern after the storm's commencement.As C/NOFS orbits were elliptical and had an apogee near 850 km altitude, the satellite's low inclination (13°) facilitated observing the same set of plasma bubbles during consecutive passes.These measurements were ideal for inspecting any sudden changes in the bubble occurrence patterns during the storms.
We also calculated the PRE peak ion drifts by analyzing the vertical C/NOFS velocities to quantify the role of the PPEF and other E fields during storms and correlate the PRE values with the bubble production.Huang (2015Huang ( , 2018Huang ( , 2019) ) demonstrated how PRE peak values change during the main phase of the storms.These studies pointed out that the PRE peak velocities showed a significant upward increase which lasted for ∼2.5 hr during the main phase of the storm of 14-16 December 2006 (Huang, 2019).Short-lived prompt-penetration electric fields caused the increment.Later, during the recovery phase, the vertical ion drifts started to decrease ∼4.7 hr after the storm onset as the disturbance dynamo process became dominant.Our study demonstrated similar changes in PRE peak velocities during the storm main and recovery phases.The changes in the drifts were also correlated with the direction of B Z during the main phase of storms.It was observed that PRE peaks were higher when the direction of B Z was southward and became significantly lower when it was northward.During the storm periods, we identified more bubbles with higher velocities (500 m/s) than during the preceding quiet times (see Figures 4 and 7).We observed that the PRE was proportional to the sunspot number (Figure 8).We suggest that the more significant number of plasma bubbles is related to the presence of additional electric fields and not tropospheric sources (e.g., gravity waves).Basu et al. (2010) described how the eastward prompt-penetration electric field influence the production of plasma bubbles in the low-latitude ionosphere during 30 intense magnetic storms within the solar cycle 23.Basu's study used ground measurements from the Scintillation Network Decision Aid (SCINDA) network and in-situ measurements from the DMSP satellite (orbiting altitude near 840 km) to observe scintillations and plasma bubble occurrences during the main phase of the storms.Their study revealed the development of plasma bubbles and scintillations within a specific longitude sector corresponding to dusk during the storm's main phase.Our analysis covers the first half of the solar cycle 24 and expands their study by examining 248 intense, moder ate, and weak geomagnetic storms (see Data Set S2) from the minimum to maximum solar activity levels during the solar cycle.The C/NOFS' PLP and IVM data follow bubbles' growth and decay patterns across all longitudes within the equatorial regions for a more significant altitude range of 400-800 km.This allowed us to investigate changes in the depth, width, and vertical velocities inside the bubbles to understand the role of E fields on bubbles across different phases of storms.We found that plasma bubbles initiate right after dusk, containing bubble velocities that reach 500 m/s and depths near 100%.However, during the storm of 09 March 2012, several short-lived bubbles developed after midnight but lasted during one orbit.We also studied bubble occurrence during the recovery phase of the storms and found that plasma bubbles can originate even during the second half of the recovery phase, but they are produced when B Z is directed southward and an enhanced PRE condition exists.
We observed that plasma bubbles initiate during the main and recovery phases of magnetic storms.The preferred location for the new bubble generation is near the local sunset; however, during storms, bubbles can occur at other local times.We also provided evidence of the sudden increase in bubble internal velocity for bubbles that started several hours before.These two facts may indicate a longitudinal or local time extension of the prompt penetration electric field.
Comparison of bubble parameters, that is, bubble depths, widths, and vertical bubble velocities, with results obtained by Smith and Heelis (2017, 2018a, 2018b) revealed a significant increase in bubble depths and vertical bubble drifts during the disturbed periods.Smith and Heelis (2018b) reported bubble depths in the Pacific and the Indian sectors to be between 20%-30% and 40%-50%, respectively.We found that the bubble depths were typically much higher during storm conditions, containing values above 80% for intense storms (see Figure 8).Two magnetic storms that occurred on 19 February 2014 and 27 February 2014 that initiated plasma bubbles in the Pacific and Indian sectors indicated bubble depths near 80% and approximately 100%.Smith and Heelis (2018a) provided evidence of the median vertical bubble velocities for all the longitude sectors and between 20 and 24 LT to be about 50 m/s during quiet periods.We found that during disturbed periods, the median vertical bubble velocities were greater than 150 m/s, sometimes reaching bubble drifts that exceeded 500 m/s.We also observed that the high bubble velocity could occur during a storm's main and recovery phases.During the recovery phase, diminishing bubble depths and vertical bubble velocities suggest the action of the disturbance dynamo electric fields.
In agreement with other investigations, we observed that during the storms, the bubbles lasted longer than those observed during the quiet period before the storm commencement and afterward during the recovery period.The long lifetime of the bubbles was likely due to the higher depth and faster vertical bubble velocity that brings the depletions to much higher altitudes.
Most of the bubbles observed during the quiet periods before the storms and afterward in the recovery periods developed in the post-sunset sector.During the magnetically disturbed periods, bubbles initially appeared in the post-sunset sector, and those bubbles grew deeper in each successive orbit.Very few new bubbles were produced during the post-midnight sectors during magnetically disturbed periods.A westward overshielding electric field may prevent new bubble formation during the main phase of the storms.

Conclusions
Analysis of 248 intense, moderate, and weak storms with available bubble and PRE peaks information between May 2008 and August 2014 have revealed that: 1. New bubbles occur during disturbed periods in the local post-midnight sectors.The primary time of observation of bubbles during storms was during the initial, main, and first half of the recovery phase when the SYM-H index remained below the half of the minimum SYM-H index.2. PRE peak values observed during the main phase when B Z was southward were observed to be higher than those observed when B Z was northward and during the recovery periods.Bubble production increased during higher PRE peak values.3. Bubbles observed during storms had a longer lifetime than quiet periods.Storm-time bubbles often did not decay until dawn.4. Bubble depths became significant during storms, and bubble depths increased to a median of 90% during the storm's initial, main, and first half of the recovery phases.5. Increments in vertical bubble velocities were observed during the storm's initial, main, and beginning recovery phases over time.During those periods, bubble velocities were observed being near 400 m/s during intense storms.

Figure 1 .
Figure 1.Graphical representation of the bubble detection algorithm on electron density measured by the C/NOFS satellite between 14:35 and 14:41 UT on 24 October 2011.Panels (a, b, and d) show the electron density in black from 14:35 to 14:41 UT.Panels (a and b) show the average background plasma densities in blue during the 1st and final iteration of the bubble detection algorithm.Panel (c) shows the relative density in black within the same time limits.Panel (d) shows the detected bubbles (shaded yellow) whenever the relative density values are reduced to less than −0.1.Panel (e) shows a large-scale electron density variation from 14:39 to 14:40 UT with the detected bubble depth and width.
value of B Z became close to zero and continued to show its quiet time fluctuations around −2 nT during the rest of the recovery period.
Figures 3b and 3c display the orbit-by-orbit measurements of plasma densities and vertical velocities observed by the C/NOFS satellite highlighting the growth and the decay of the plasma bubbles on 09 March 2012.Figure3a

Figure 2 .
Figure 2. Background magnetic conditions during the storm of 09 March 2012.Panels (a-c) show the variations of SYM-H index, B Z , and PRE peaks during the storm's initial, main, and recovery phases.Vertical lines in the panels indicate the duration of the quiet time before the storm (Q) and storm's initial (I), main (M), and two parts of the recovery phase (R1 and R2).Panel (c) shows the PRE peak velocities observed during 18-19hr local time corresponding to each C/NOFS orbit (blue dot).See text for the PRE detection algorithm.

Figure 3 .
Figure 3. Background magnetic conditions and corresponding orbit-by-orbit measurements of electron density and vertical plasma velocity by C/NOFS during the initial, main, and part of the recovery phase during the storm of 09 March 2012.Panel (a) shows the variations of the SYM-H index (red), B Z (pink area plot), and PRE peaks (blue) corresponding to the C/NOFS orbits.Panel (b) shows the electron density (black) and altitude (red) variations in UT.Horizontal blue dashed lines signify the magnitudes of 10 5 cc −1 electron density (600 kms altitude), and each tick along the Y-axis measures an order of magnitude change in electron density (200 kms altitude).Panel (c) shows plasma vertical velocity (black) and magnetic latitude (red) variations in UT.Horizontal blue dashed lines signify the values of 0 m/s plasma vertical velocity (0° magnetic latitude), and each tick mark along the Y-axis indicates 250 m/s.Panels (b and c) show the local sunset hours shaded in yellow corresponding to each orbit.Satellite orbit numbers are shown at the left of the panel (a).UTs corresponding to the start and end of the orbits at 0° longitudes are stated at the left and right ends of panels (band c).The longitude range is further divided into 4 sectors: African ("A": 0°-90°), Indian ("I": 90°-180°), Pacific ("P": 180°-270°) and South American ("SA": 270°-360°).

Figure 4 .
Figure 4. Bubble characteristics during different phases of the storm of 09 March 2012.Panels (a-c) show bubble depths, widths, and vertical bubble velocities, respectively.Vertical lines in all the panels divide the indicated duration in quiet time before the storm (Q) and storm's initial (I), main (M), and two parts of the recovery phase (R1 and R2).The color marker indicates the local time of observation of each bubble as specified in the color bar.

Figure 6 .
Figure 6.Same as Figure 3 during the storm of 01 June 2013.

Figure 7 .
Figure 7. Same as Figure 4 during the storm of 01 June 2013.

Figure 8 .
Figure 8.Comparison of disturbed and quiet condition bubble characteristics for 14 intense storms occurring from May 2008 to April 2014.Panel (a) shows the sunspot number and range of PRE peak velocities when IMF B Z was observed southward during the storms.Panels (b-d) show the depths, widths and vertical bubble velocities observed during the storms.The gray bars in each panel show the quiet time median of the characteristics.Bubble properties during storm periods with respect to their quiet time bubble depths were less than and greater than 75%, respectively are represented with continuous and dashed narrow bars.

Figure 8
Figure 8 shows distributions of PRE peak velocities observed during the intense storms and the corresponding sunspot number averaged over the month of occurrence of the storm (panel a).The medians of the bubble depths, widths, and vertical velocities observed during the quiet period before and the disturbed period during the storms are shown in panels (b), (c), and (d).The median bubble depths, widths, and vertical velocities during quiet periods (in gray bars) and disturbed periods (in red stem plots with circular markers) are shown in panels (b), (c), and (d).The plot in panel (a) shows that the solar activity levels were low during the storms on 06 August 2011, 09 March 2011, 24 April 2012, 15 July 2012, 01 October 2012, 09 October 2012, 14 November 2012, 17 March 2013, and 01 June 2012, as the sunspot numbers were below 100.The solar activity levels increased slightly during the storms on 26 September 2011, 25 October 2011, 19 February 2014, and 27 February 2014 when the sunspot numbers were greater than 100.The range of PRE peak velocities observed during the disturbed periods of the storms, when the IMF B Z was southward, is shown with box-and-whiskers plots.The 25th, 50th, and 75th percentile values corresponding to the distributions are shown with horizontal lines, and the minima and maxima are shown with whiskers at the bottom and the top of the boxes, respectively.The panel shows that during disturbed periods of the intense storms, PRE peak velocities greater than 50 m/s were observed when B Z was directed southwards.For some of the storms, for example, the storms that occurred on 25 October 2011, 15 July 2012, 09 October 2012, 14 November 2012, 17 March 2013, 01 June 2013, and 27 February 2014, the maximum PRE peak velocity observed during the phase reached close to or greater than 100 m/s.Panel (b) in the figure shows that the median bubble depths during storms on 06 August 2011, 09 March 2012, 15 July 2012, 13 October 2012, 01 June 2013, 19 February, and 27 February 2014 have increased significantly from the median bubble depths observed during the quiet periods before the storms.During the storm that occurred on 01 October 2012, the median bubble depth during the disturbed period was observed to be lower than what was observed during the quiet period before the storm.During storms that occurred on 26 September 2011, 25 October 2011, 24 April 2012, 09 October 2012, 14 November 2012, and 17 March 2013, the median of the bubble depths showed negligible change.Panel (c) shows that bubble widths remained similar during both quiet and disturbed periods for most of the storms, except the storms that occurred on 06 August 2011, 26 September 2011, 24 April 2012, and 01 June 2013, when bubble widths during disturbed periods showed enhancement.The overall bubble vertical velocities during quiet times were very low for all the intense storms.As shown in panel (d), during disturbed periods corresponding to the majority of the storms, the bubble vertical velocities are shown to have increased.Bubble vertical velocities during disturbed periods corresponding to the storms of 25 September 2011, 09 March 2012, 15 July 2012, 01 October 2012, 13 October 2012, 17 March 2013, and 01 June 2013 have increased significantly.