The blue luminous events (BLEs) recorded by ISUAL (Imager of Sprites and Upper Atmospheric Lightning) radiate unambiguous middle ultraviolet to blue emissions (230–450 nm) but contain dim red emissions (623–754 nm). The BLE appears to be dot‐like on one ISUAL image with an integration time of 29 ms. A few BLEs develop upward into blue jets/starters or type II gigantic jets (GJs). The associated sferics of the BLEs in the extremely low frequency to very low frequency band and in the low‐frequency band exhibit similar patterns to the narrow bipolar events (NBEs) identified in the very low frequency and low‐frequency band. The ISUAL BLEs are conjectured to be the accompanied light emissions of the NBEs. Both upward and downward propagating current obtained from the associated sferics of the BLEs have been found. The source heights of the six BLEs related to negative NBEs are estimated in the range of 16.2–17.8 km. These six events are suggested to occur between the upper positive charge layer and the negative screen charge layer on the top of the normally electrified thunderstorm. The six blue starters, one blue jet, and one type II GJ are inferred to be positive upward discharges from their associated sferics in the extremely low frequency to very low frequency band. Based on the simultaneous radio and optical observations, a NBE is conjectured to be the initiation discharge with rapidly flowing current within the thunderstorm, while a blue jet/starter or a type II GJ is suggested to be the ensuing discharge with slowly varying current propagating upward from the thunderstorm.

We report the ionization emission associated with N₂⁺ 1N band in a halo event without visible sprite streamers. To avoid the lightning contamination to the ionization emission, we find halos whose parent lightning light is blocked by the Earth's limb. Five halos in the 2004–2010 Imager of Sprites and Upper Atmospheric Lightning data set were identified as halos without visible sprite streamers. A halo with maximum N₂ 1P brightness had significant ionization emission of N₂⁺ 1N. The time‐integrated photon intensity of N₂ 1P, N₂ 2P, and N₂⁺ 1N emission is 2.2 × 10⁵, 2.1 × 10⁴, and 7.4 × 10² photons cm⁻², respectively at a distance of 4130 km. The total number of photons of N₂ 1P, N₂ 2P, and N₂⁺ 1N band emissions are 4.6 × 10²³, 4.3 × 10²², and 1.6 × 10²¹ photons, respectively. In the halo region, the electron density increased as 1–2 orders of magnitude higher than ambient electron density. From the emission ratio of N₂⁺ 1N to N₂ 2P, the reduced electric field is estimated to be 275–325 Td that is higher than the conventional breakdown electric field. The recorded electric field related to this halo event is produced by a lightning discharge that has a total charge moment change of −1450 C km. Based on the estimated electric field from optical emissions, it is found that the lightning‐induced electric field in the bright halo region is significantly relaxed with a rate faster than that estimated using ambient electron density, in agreement with previous modeling results showing that the electron density enhancement due to the ionization processes leads to a short dielectric relaxation time inside the halo region.

Atomic sodium (Na) is one of the major meteoric species existing as layers of atoms in the mesosphere/lower thermosphere (MLT) at altitudes of 80–105 km and atomic ions at higher altitudes. As the boundary between neutral atmosphere and ionosphere, the MLT region is of particular interest because it is influenced by the mesoscale convective weather and thunderstorms from below and exposed to solar photons from above. There has been great interest on the atmosphere‐ionosphere coupling, and numerous studies have been reported on the connection between thunderstorms and the ionosphere. However, the influence of thunderstorms on metallic species, which would significantly enhance our understanding of how Earth's atmosphere interacts with its ionosphere and studying the chemistry and physics of the MLT region, has rarely been studied. Here we present observational results on a statistical basis showing evidence that thunderstorm activities can affect the metal layer, by identifying a statistically significant enhancement of the neutral metal Na layer above thunderstorms at Haikou, China (20.0°N, 110.3°E). The thunderstorm‐generated gravity waves and electric field effects could be the mechanisms responsible for the lightning‐associated enhancement of Na layer.

We report the blue luminous events observed by the Imager of Sprite/Upper Atmospheric Lightning (ISUAL) payload on board the FORMOSAT‐2 satellite. The ISUAL 427.8 nm filtered Imager campaign was conducted near Australia in the summer of the Southern Hemisphere during February–March 2008. On 18 February 2008, the 427.8 nm filtered Imager recorded a series of blue luminous events. One blue starter and nine smaller blue starters were recorded in 2 min and 34 s in a localized region with the radius < 4 km over the cloud top. The average time interval between subsequent blue luminous events was ~ 17 s. The occurrence rate of blue luminous events was 3.5 events per minute, slightly lower than the occurrence rate of pixies (4.2 events per minute) but higher than the occurrence rate of gnomes and blue jets in the previous observations. The recorded first blue starter lasted up to 2–3 frame times (60–90 ms) and extended its altitude about 8 ± 0.3 km with a width of ~2–4 km over the cloud top. After the first blue starter, subsequent nine smaller blue starters had the decreased heights of ~2–4 km, and their optical duration was shorter and is down to 1 ms. But their major emissions were 2PN₂ and 1NN₂⁺, without lightning OI 777.4 nm emission. The ISUAL recorded blue smaller starters had the spatial average brightness of 130 kR for the 427.8 nm filtered Imager with exposure time (29 ms) and 1.2 MR for the spectrophotometer (337 nm). Using the spectrophotometer, the emission time of blue starters was 1 ms. It is estimated that the 1NN₂⁺ emission was ~ 22 MR and the 2PN₂ emission was ~ 132 MR. We can estimate the degree of ionization was 10⁻¹¹–10⁻¹² in these blue luminous events using the 427.8 nm filtered Imager measured 1NN₂⁺ (0,1) emission.

Thunderstorms are electric generators, which drive currents upwardly into the ionosphere. In this paper, we examine the effects of thunderstorm upward current on the ionosphere. We use a thunderstorm model to calculate the three‐dimensional current flows in the atmosphere and to simulate the upward current above the thunderstorm with the tripole‐charge structure. The upward current flows into the ionosphere, while the associated electric field causes the plasma E × B motion. The caused plasma motion redistributes the plasma density, leading to ionospheric density variations. In the nighttime ionosphere, the E × B motion may also cause the formation of plasma bubbles.

A short‐term (30 days before an earthquake) prediction of an earthquake is a big challenge in seismology. As a first step, we apply deep learning to the ionospheric total electron content (TEC) data between 2003 and 2014 to detect the seismo‐ionospheric precursors of M ≥ 6.0 earthquakes in Taiwan. The bidirectional Long Short‐Term Memory (Bi‐LSTM) network is employed to use observed input data (features) to obtain the sequential TEC variations. The five input features are sequential vectors of TEC, the geomagnetic index Dst, the solar activity index F10.7, sunspot number (SSN), and solar emission index Lyman‐α. The daily values of F10.7, SSN, and Lyman‐α are converted into hourly values, depending on the solar elevation angle. The calculated hourly TEC variations can be more precisely predicted with this data conversion. We calculate the normalized difference of errors between two 15‐day adjacent stages as the “relative error”. Three trained models with the best discrimination between the relative errors of earthquake and no‐earthquake cases are chosen as classifiers. These three classifiers are then used to have a majority vote to declare whether the 30‐day period is related to the preparation of an earthquake or not. The results show that all 22 positive cases (earthquakes) are successfully predicted, giving a true positive rate of 100%. Among the 19 negative cases (normal cases), 10 of them are true negative.” Overall, a high accuracy of 78.05% is obtained.