HO
 2 Generation Above Sprite‐Producing Thunderstorms Derived from Low‐Noise SMILES Observation Spectra

No direct observational evidence of sprite‐produced active radicals has been presented owing to the difficulty of observing a small event area in the nighttime mesosphere, whereas sprite chemical models have indicated that sprite discharge locally affects the atmospheric composition. We present the first observational evidence of a HO 2 production above sprite‐producing thunderstorms from the coincidence of temporal‐spatial observations of HO 2 spectra, sprite events, and thunderstorms by two space instruments, a submillimeter‐wave limb spectrometer and ultraviolet/visible Imager and a ground‐based very low frequency radiation lightning detection network. A total of three areas was identified with enhanced HO 2 levels of approximately 10 25 molecules. A chemical sprite model indicates an increase in HO 2 in the considered altitude region; however, the predicted production due to a single sprite event is smaller than the observed enhancement. Our observational results suggest that sprites potentially contribute 1% of nighttime background HO 2 generation at altitudes of 75–80 km globally.


Introduction
Current technical advancements in areas such as measurement sensitivity and multiple-satellite observations have allowed us to examine single sudden events, although the statistical approach has been universally used to understand atmospheric phenomena generally. The understanding of sprite phenomena has been improved by the global coverage provided by satellite observations (e.g., Blanc et al., 2007, Chern et al., 2003, Garipov et al., 2013, Jehl et al., 2013, Sato et al., 2015. The electric fields and electron energies of sprite streamers were inferred from high-quality morphological and spectral data of ultraviolet (UV)/visible imagers (e.g., Adachi et al., 2006;Kuo et al., 2005) and global distributions and occurrence frequency are estimated by long-term global observations (e.g., Chen et al., 2008). These observations were also

10.1029/2019GL085529
Key Points: • Clear HO 2 enhancements above sprite-producing thunderstorms were identified by coincidence of space and ground-based observations • Enhanced HO 2 amounts in a sprite event area were estimated to be 10 25 molecules • Sprites potentially contribute about 1% of nighttime background HO 2 generations at 75-80 km globally Supporting Information: • Supporting Information S1 • Figure S1 • Figure S2 • Figure S3 • Figure S4 Correspondence to: Y. Kasai, ykasai@nict.go.jp Sprites are one of the most familiar types of the various upper atmospheric lightning phenomena, which are characterized as transient luminous events (Neubert et al., 2008;Pasko et al., 2012). Since the discovery of the sprite in 1990 (Franz et al., 1990), it has been suggested that transient luminous events, particularly sprites, generate active radicals and ions by ion-neutral chemistry models (e.g., Arnone et al., 2014, Evtushenko et al., 2013, Gordillo-Vázquez, 2008, Parra-Rojas et al., 2015, Winkler & Notholt, 2014, although no conclusive observational evidence of chemical impact has been reported so far. Sprite discharges are induced by conventional air breakdown, caused by lightning-driven electric fields above thunderstorms (Hu et al., 2007;Pasko et al., 1995). They appear in an altitude range of 40-90 km, with a horizontal extent of several tens of kilometers. The duration of sprite luminescence ranges from a few to several tens of milliseconds (Barrington-Leigh et al., 2001). Estimates of the global rate of sprite occurrence range from one to three events per minute (Chen et al., 2008;Ignaccolo et al., 2006;Sato & Fukunishi, 2003); that is, there is approximately one sprite event for every thousand lightning flashes, as there are approximately 40-50 lightning flashes per second (Christian et al., 2003). It is recognized that once a sprite is produced, they appear repeatedly in the same region (Jing et al., 2008), although not every thunderstorm produce sprites.
Previously, two independent satellite instruments were used in an attempt to investigate sprite chemistry by observing lower mesospheric NO 2 trends with tropospheric lightning activity (Arnone et al., , 2009Rodger et al., 2008); however, no conclusive observational evidence has been reported thus far. Arnone et al. (2008) showed a statistical difference between the amount of NO 2 observed by the Michelson Interferometer for Passive Atmospheric Sounding mid-infrared emission spectrometer and thunderstorms activity within a field of view (FOV) of 500 × 30-km regions recorded by the Worldwide Lightning Location Network (WWLLN). Their statistical analysis and follow-up work (Arnone et al., 2009) indicates that NO 2 enhancements of about 10% at 52-km height and tens of percent at 60-km height immediately after thunderstorm activity. Rodger et al. (2008) compared regional variations of column amounts of NO 2 observed by the Global Ozone Monitoring by Occultation of Stars UV-visible spectrometer with respect to the lightning activity obtained from Optical Transient Detector observations (Christian et al., 2003) in the tropical and northern midlatitude regions (Rodger et al., 2008). The area between land and ocean regions was compared, because ∼85-90% of lightning activity occurs above land (Christian et al., 2003). They reported no evidence of any correlation within the detection levels of the Global Ozone Monitoring by Occultation of Stars measurements.
In this study, we present the first investigation of HO 2 generated by sprite events above sprite-producing thunderstorms from the coincidence of temporal-spatial observations of HO 2 spectra, sprite events, and thunderstorms by two space instruments, submillimeter-wave limb spectrometer and UV/visible Imager and a ground-based very low frequency (VLF) radiation network used for lightning detection. Single limb scan HO 2 spectra were observed using the Submillimeter-Wave Limb-Emission Sounder (SMILES) with full local time coverage. The SMILES instrument is suitable for single-scan HO 2 detection because the HO 2 single-scan detection limit is on the order of 1 ppbv in the nighttime mesosphere, which is an order of magnitude better than previous microwave/submillimeter limb instruments (Baron et al., 2011;Kasai et al., 2013;Kikuchi et al., 2010;Kreyling et al., 2013). A total of 127 emissions including sprites, halos, and gigantic jets was extracted from the Imager of Sprites and Upper Atmospheric Lightning (ISUAL) (Chern et al., 2003) observation data, and VLF radiation from lightning strokes was recorded by receiver sites of WWLLN (Lay et al., 2004) during the SMILES observation period from 12 October 2009 to 21 April 2010.
This paper is organized as follows. Section 2 describes observation data and analysis of the HO 2 spectra, sprite image, and lightning distribution at their spatiotemporal coincidence. In section 3, we estimate the amount of HO 2 enhancement in a sprite event area for the cases presented in section 2 and discuss a qualitative comparison with a sprite chemical model and a possible global impact of HO 2 generation by sprite events. Finally, in section 4, we conclude our main results.

HO 2 Spectral Enhancements Above Sprite-Producing Thunderstorms Using SMILES, ISUAL, and WWLLN
The SMILES instrument on board the International Space Station is a unique instrument that uses 4-K superconductive heterodyne receivers to obtain low-noise spectra in the submillimeter-wave region (Kikuchi et al., 2010;Ochiai et al., 2012). We use the SMILES observation spectra ranging from 649.12 to 650.32 GHz and detect the HO 2 transition at 649.701 GHz. Another SMILES frequency window ranging from 625. The ISUAL instrument on board the Taiwanese satellite FORMOSAT-2 takes a snapshot image of sprites and parent lightning using an intensified charge-coupled device camera while maintaining its FOV in the eastward direction of the Earth's limb (Chen et al., 2008;Chern et al., 2003). The camera has an FOV of 20 • (horizontal) × 5 • (vertical). We use imager data obtained with the 633-to 751-nm filter and an exposure time of 29 ms. During the SMILES observation period, we selected a total of 127 events including sprites, halos, and gigantic jets observed by ISUAL in the latitude range of 25 • S-45 • N. The direction of the ISUAL's FOV is maintained eastward of the Earth's limb with an inclination angle of 98.99 • . The location and vertical distributions of sprite emissions are estimated by the method reported in Adachi et al. (2008). The sprite was considered to appear above the point where the intensity of the parent lightning emission is the brightest. In this study, we selected sprite, halo, and gigantic jet events observed in front of the Earth's limb so that we can precisely estimate the location of parent lightning. The spatial uncertainty of estimating the sprite's location is derived from the 1-sigma decay of the parent lightning intensity. The altitude range of lightning emission is assumed to be 10 ± 5 km. Consequently, the longitudinal and latitudinal uncertainties in positioning the sprites from the ISUAL image are up to 3 • and 1 • , respectively.
The locations of lightning strokes obtained from the WWLLN data are used for indicating the area of active thunderstorm likely producing additional sprites around a sprite ISUAL detected. The WWLLN locates lightning by using the time of group arrival of the VLF (3-30 kHz) radiation from a lightning stroke. WWLLN had about 40 active VLF receivers during the SMILES observation period. The detection efficiency of cloud-to-ground flashes for peak currents stronger than ±35 kA is 10%, and the location errors are 4.03 km in the north-south and 4.98 km in the east-west directions (Abarca et al., 2010).
We excluded the following cases to find spatiotemporal coincidence between SMILES and ISUAL observations: (1) longitudinal and latitudinal differences between the SMILES observation area and the ISUAL sprite detection location are more than 3 • and 1.5 • because of the uncertainty in sprite location, SMILES  These local enhancements were most likely caused by sprite events, rather than from other events or variations. Although solar proton and energetic electron precipitation events are known to enhance the nighttime abundance of HO x (H, OH, and HO 2 ) in the high-latitude region owing to the reaction between electrons and hydrated cluster ions (Jackman et al., 2011(Jackman et al., , 2014Solomon et al., 1981), their activity was quite low at the low-latitude region and during the SMILES observations that took place from 12 October 2009 to 21 April 2010 (Andersson et al., 2014). The diurnal variation in mesospheric HO 2 is mostly controlled by the photolysis of H 2 O by sunlight , which only depends on the local time and lifetime of HO x species, making it incapable of producing the local enhancement. Horizontal transportation by zonal winds may impact the estimated HO 2 productions as demonstrated in Arnone et al. (2014) for nitrogen oxides. The zonal winds are between 0 and 40 m/s at November (Events A and B) and between −20 and 0 m/s on March (Event C) near 75-to 80-km altitudes . We estimated that the distances the air masses are moved away from the LOS are smaller than 300, 200, 300 km for Events A, B, and C, respectively. These ranges are of the same order of magnitude than the area of the thunderstorms. Hence, the transportation only induces a small underestimation of HO 2 abundances estimated in this analysis. Future studies will be performed with global chemical model to account for dynamical effects.

Estimation of HO 2 Amount Produced by Sprite Events
HO 2 number densities in the sprite event area are estimated with a least squares fitting by using the Atmospheric Radiative Transfer Simulator (Buehler et al., 2018;Eriksson et al., 2011). The simulation assumes that there is an HO 2 enhanced area where all the additional HO 2 is contained. This box is called the enhanced HO 2 box. The altitudes of the enhanced HO 2 boxes for estimating the number density of HO 2 are 75, 77, and 80 km, respectively. Line parameters include line position, intensity, pressure broadening, energy levels, and partition functions of HO 2 were taken from Perrin et al. (2005). The atmospheric profiles include the volume mixing ratio of HO 2 , temperature, and pressure. Vertical atmospheric profiles are obtained from the averaged values of the SMILES Level-2 research (L2r) product data Version 3.0.0, an updated version of Version 2.1.5 (Baron et al., 2011;Kasai et al., 2013;Kreyling et al., 2013), at adjacent observation scans. The calculated spectra are convoluted with frequency response functions of acousto-optic spectrometers (Mizobuchi et al., 2012).
The enhanced air column abundances from an initial value, Δ obs HO 2 , are also calculated as below: where l, V box , n fit , and n back are the length of the enhanced HO 2 box, the column volume of the enhanced HO 2 box, the best fit number density of HO 2 in the enhanced HO 2 box, and the averaged values of the SMILES L2r product data at adjacent observation scans, respectively. We assume that the SMILES FOV has an elliptical shape with a diameter of 3 km vertically and 6 km horizontally.
The best fit densities and one standard deviation errors of the fitting densities of HO 2 in the enhanced HO 2 boxes are 38 ± 10 × 10 6 , 38 ± 5 × 10 6 , and 160 ± 20 × 10 6 molecules/cm 3 for averaged densities of the adjacent air mass of 0.56 × 10 6 , 1.0 × 10 6 , and 0.85 × 10 6 molecules/cm 3 , respectively, in the case of the size of the enhanced HO 2 box, l, is same as the horizontal width of sprite emission ISUAL detected. The number densities of HO 2 in the enhanced HO 2 boxes are 68 ± 18, 38 ± 5, and 190 ± 24 times larger than the background values. The root-mean-square of the residual between the simulated and observed spectra is 0.23, 0.20, and 0.21 K, while the root-mean-square noise of the observed spectrum is 0.21, 0.21, and 0.22 K, respectively, as shown in Figures S2a-S2c in the supporting information.
The Δ obs HO 2 for each event is nearly constant relative to the length of the sprite box as shown in Figure  S3. They were 8.9 ± 2.5 ×, 16 ± 2.0 ×, and 17 ± 2.0 × 10 24 molecules for each event. We also investigated the sprite-induced chemical disturbances in the mesosphere on the timescale of a few hours after a sprite event by using a one-dimensional atmospheric chemistry and transport model (see supporting information Text S1 for the model description and simulation results). The model simulation showed that once HO 2 is formed, the enhanced concentrations remain basically constant for a few hours, although the enhancements are much smaller than the enhancements observed by SMILES as shown in Figure S4 in the supporting information. Multiple sprite events are likely necessary to explain the observed HO 2 enhancement. The long-lasting HO 2 increase in the model may allow an accumulation of HO 2 enhancements. There is a short-term decrease in ozone at 75 km with recovery after 1,500 s. The model indicates the HO 2 enhancements are caused by a conversion of water molecules into HO x through reactions and recombination of proton hydrates, which is similar to the situation in solar proton events. There are also possibilities of different electric field parameters than used in our model and missing chemical processes in the model. The preliminary model simulation only accounts for streamer tip electric fields. The effect of afterglow fields (Gordillo-Vázquez & Luque, 2010) has not been considered yet. They will be addressed in future simulations. Further model studies of a time range of several hours after sprite initiation should focus on factors related to the above reactions and take into account horizontal transport processes.
Furthermore, we estimated a global production of HO 2 molecules by sprite events during nighttime at altitudes of 75-80 km from the observational results. The total amount of HO 2 produced by the nighttime sprite events can be approximately 10 28 molecules if the occurrence frequency were assumed to be one to three events per minute, while the total amount of background HO 2 can be 10 30 molecules if background concentration of HO 2 were assumed to be 0.9 × 10 6 molecules/cm 3 at the altitude range. Thus, the production of 10 25 molecules of HO 2 by a sprite event area is potentially accounts for 1% of the global amount of nighttime HO 2 at 75-80 km. The background concentration of HO 2 is derived from the average of SMILES L2r product data in nighttime at 75-80 km during the SMILES observation period. This estimation simplified the sprite and atmospheric conditions such as the strength of electric field in sprite discharges and the background atmospheric composition. To be clear, the retrieved production is not necessarily showing the production of HO 2 from single sprites but rather the accumulation of production by the entire sprite-producing thunderstorm. As Arnone et al. (2014) demonstrated, it is important to investigate the climate chemistry sensitivity to sprite HO x production without the above simplification. The impact of sprite chemistry on the Earth's atmosphere may increase in the future because the source of HO 2 production and trigger of sprite occurrence are mesospheric water vapor and tropospheric lightning, which are increased by greenhouse gas emissions and global warming, respectively. Global warming is predicted to increase tropospheric lightning activity (5-10% for every 1 • of warming) (Krause et al., 2014;Michalon et al., 1999;Price & Rind, 1994;Romps et al., 2014;Trapp et al., 2007) although the prediction has a large uncertainty in the lightning parameterization (Finney et al., 2018). Long-term trends of increase in mesospheric carbon dioxide (CO 2 ) and water vapor (H 2 O) have been reported owing to recent improvement in the accuracy of global measurements (Chandra et al., 1997;Nath et al., 2018;Nedoluha et al., 2017;Qian et al., 2017). These trends, approximately +5.5% of CO 2 and <10% of H 2 O per decade, are related to anthropogenic CO 2 and methane (CH 4 ) emissions, respectively, and affect radiative forcing and mesospheric composition. Further observations and model studies regarding sprite and atmospheric discharge chemistry are needed to assess their potential impact on the atmosphere of the Earth.

Conclusions
In this study, we detected for the first time the local enhancements of HO 2 above sprite-producing thunderstorms at an altitude range of 75-80 km by using SMILES, ISUAL, and WWLLN data. All three areas showed that the enhancement of HO 2 amounts to be approximately 10 25 molecules. The chemical sprite model indicates an increase in HO 2 in the considered altitude region for several hours after a sprite discharge; however, the predicted production of HO 2 due to a single sprite event is smaller than the observed enhancement. The production of 10 25 molecules of HO 2 at a sprite event area potentially accounts for approximately 1% of the global amount of nighttime HO 2 at altitudes of 75-80 km according to the simplified estimation. The investigation of the impact of sprite discharge events on global atmospheric composition becomes more important in the scenario of increase in both lightning activity and H 2 O abundance in the upper atmosphere. Further observations and model studies regarding sprite and atmospheric discharge chemistry are needed to assess their potential impact on the Earth's atmosphere.