WACCM climate chemistry sensitivity to sprite perturbations

Abstract

Transient luminous events affect Earth's atmosphere between thunderstorm tops and the lower ionosphere through ion-neutral chemistry reactions. Particular emphasis has been given to sprites, with models and observations suggesting a capability of perturbing atmospheric nitrogen oxides at a local level, as it is known to occur for tropospheric lightning and laboratory air discharges. However, it is as yet unknown whether sprites can be a relevant source of nitrogen oxides for the upper atmosphere. In this paper, we study the sensitivity of the Whole Atmosphere Community Climate Model (WACCM) to sprite-like nitrogen oxide perturbations. We take a top-down approach to estimate what magnitude sprite perturbations should have to become significant as compared to other relevant atmospheric processes and study the sensitivity of the model response within the given uncertainties. We show that, based on current predictions by sprite streamer chemistry models, sprites can perturb Tropical NO_x at 70 km altitude between 0.015 ppbv (buried in the background variability) and 0.15 ppbv (about 20%), adopting a local NO_x production per sprite of 1.5·10²³ and 1.5·10²⁴ molecules respectively at this altitude. Below the lowest of the adopted values, sprites are irrelevant at global scales. Sprite NO_x may build up to significantly larger amounts locally above active thunderstorms, further aided by other transient luminous events and possibly terrestrial gamma ray flashes. We also use model results to interpret the available observational studies and give recommendations for future campaigns.

Key Points

• First sprite-NO_x parameterization in GCM (WACCM) and sensitivity study
• Results show sprite-NO_x up to 20% at 70 km altitude depending on input adopted
• Inclusion of sprite-NO_x source essential to link predictions to observations

1 Introduction

Just over two decades ago, the first sprite, a huge luminous emission extending for tens of kilometers above a thunderstorm, was accidentally recorded by Franz et al. [1990]. It was the first observation of a whole family of upper atmosphere electrical luminous processes collectively known as transient luminous events (TLEs) [e.g., Rodger, 1999; Füllekrug et al., 2006; Neubert et al., 2008; Pasko et al., 2012]. Beside TLEs, unexpected gamma ray energy photons coming from the Earth's atmosphere above thunderstorms were captured by astrophysical satellites and later called terrestrial gamma ray flashes (TGFs) [e.g., Fishman et al., 1994; Smith et al., 2005]. These discoveries confirmed early predictions by Wilson [1925] and raised a strong scientific interest in the impact of thunderstorm processes on the atmosphere above.

Transient luminous events occur in the upper troposphere-lower thermosphere region between the top of thunderclouds and the lower ionosphere. Their occurrence, appearance, and development depend on charge processes within the parent thunderstorm and on the properties of the atmosphere above. Above a thundercloud, weakly ionized plasma channels called streamers can form up to roughly 70 km altitude. Above this altitude, the dielectric relaxation timescale becomes comparable with that of dissociative attachment leading to the occurrence of diffuse emissions [Pasko et al., 1998]. TLEs include blue jets [Wescott et al., 1995] and gigantic jets [Pasko et al., 2002], which are streamers directly injected from the thundercloud top toward the ionosphere, the latter creating a direct thunderstorm-ionosphere connection. Red sprites [Sentman and Wescott, 1993; Lyons, 1994; Sentman et al., 1995] are luminous flashes that initiate at about 70–80 km altitude, can extend downward to 40 km as streamers and upward to 90 km altitude as diffuse emission, and be tens of kilometers wide [Stenbaek-Nielsen et al., 2000]. Sprite halos occur as diffuse emission around 70–80 km altitude [Barrington Leigh et al., 2001; Wescott et al., 2001]. Elves appear as horizontally expanding diffuse emission rings at the lower edge of the ionosphere [Inan et al., 1996; Fukunishi et al., 1996]. Depending on the relaxation timescales at the altitude of occurrence, TLEs last from a few hundreds of milliseconds (jets) down to milliseconds (elves) and are therefore transients as compared to the much longer chemical and dynamical timescales typical of the upper atmosphere.

Similarly to other air plasma processes such as tropospheric lightning and laboratory discharges (see review by Ebert et al. [2006]), TLEs ionize, dissociate and excite neutral air constituents (mainly N₂ and O₂) as evident from the consequent well recognizable optical emissions, therefore perturbing the chemistry of the atmosphere [see, e.g., Pasko et al., 2012]. The altitude of occurrence of TLEs covers the stratospheric ozone layer which is regulated by the NO_x and HO_x families [see, e.g., Brasseur and Solomon, 2005]. Above, in the middle mesosphere, very low densities of NO_x are found because of a fading contribution of N₂O oxidation and inefficient downward transport of upper mesospheric NO_x (except in the winter polar vortex). Therefore, a sprite contribution may easily dominate over other sources in this region. Since sprites are tens of kilometers in size and occur at a global rate of 1–3 sprites per minute, as much as half of middle atmospheric air may be processed by sprites over 1 year [Ignaccolo et al., 2006; Chen et al., 2008]. Sprite-NO_x production should be therefore constrained and compared to known NO_x sources due to oxidation of tropospheric N₂O and energetic particle precipitation [e.g., Callis et al., 2002; Funke et al., 2005; Arnone and Hauchecorne, 2012].

In 2008, the first model studies of the ion-neutral chemical effects of sprites were published, evaluating their possible chemical impact on the neutral atmosphere. Enell et al. [2008] and Sentman et al. [2008] estimated sprite-induced NO_x perturbations within sprite volumes to be from a few percent to a few tens of percent between 50 and 80 km altitude in representative scenarios, up to hundreds of percent at 70 km altitude in extreme cases. Gordillo-Vázquez [2008] and Hiraki et al. [2008] estimated orders of magnitude increases in NO_x and HO_x. Furthermore, Hiraki et al. [2008] studied ozone changes and found these to be significant as well. These models show that key reactions dominating the NO production in the altitude range 60–70 km are electron impact on N₂ with atomic N being either in its ground (⁴S) state or in its (²D) state, and consequent N oxidation, with atomic O either in the ground or (¹D) state, together with minor contributions from ion species in the first 1–10s [Sentman et al., 2008]. The production of NO depends on the branching ratio into N(²D), whereas NO destruction depends on the branching into N(⁴S). The main reaction that would lead to reconversion of NO into N₂ (N+NO→ N₂ +O) is slower than diffusion timescales so that the NO produced can be diffused away before being destroyed. On timescales of 10²−10³ s, NO is completely converted into NO₂ through reactions with O₃, O, and HO₂, with the final production step NO₂ lasting several hours. Large enhancements of seed electrons were found also by the plasma model by Evtushenko et al. [2013], although with no estimates of NO production. Winkler and Notholt [2014] investigated the impact of possible daytime sprites, rarely observed at twilight: they found a similar NO_x production, which, persisting during daytime as active NO, has an enhanced impact on ozone. At global scales, sprites may be therefore significant NO_x sources whereas HO_x and O_x perturbations remain local. An attempt to estimate the local and global relevance of sprite-NO_x was performed by Enell et al. [2008] with simple order-of-magnitude calculations, suggesting sprites could lead to large NO_x changes above active thunderstorms. Arnone et al. [2008a] and Neubert et al. [2008] set up simplified simulations on global circulation models to investigate the dynamical response to possible sprite-induced ozone changes in the Tropical upper stratosphere-lower mesosphere (i.e., the overall global envelope of sprite perturbations), finding negligible temperature changes.

To date, no direct measurement of TLE-induced chemistry perturbations exists. Peterson et al. [2009] produced a first laboratory experiment simulating NO_x by TLEs in a pressure controlled laboratory chamber. Their results point toward a significant local impact but negligible contribution to the global budget of nitrogen oxides, although the representativity of their experiment for real sprites was severely questioned (see Nijdam et al. [2010], de Urquijo and Gordillo-Vázquez [2010], and replies by Peterson et al.). Two independent observational studies investigated possible sprite-induced chemical changes in middle atmosphere satellite measurements, using tropospheric lightning as a proxy of sprite activity. Arnone et al. [2008b] and their follow-up work [Arnone et al., 2009] found an anomalous NO₂ increase of 10% at 52 km altitude and of tens of percent at 60 km altitude in coincidence with active thunderstorms, possibly the result of sprite activity. Using a climatological approach, the multiyear study by Rodger et al. [2008] found no global sprite-signature in partial columns of NO_x in the middle atmosphere concluding that sprites and other TLEs occurring below 70 km altitude cannot exert a significant global impact on neutral chemistry. Even though Arnone et al. [2008b] made use of first-order backward trajectories to locally account for transport by winds, neither of these studies modeled the effect of overall transport in their global approaches so that a TLE contribution to global NO_x may have remained buried in the background due to zonal transport.

The WACCM model includes relevant atmospheric processes from the Earth's surface to the thermosphere and is therefore ideal for comparing TLE and TGF chemical sources to other sources at global scale. In the present study, we focus on sprites, introducing a sprite-NO_x parametrization of increasing magnitude in WACCM in order to study its chemical and dynamical response. The WACCM model and its components relevant to this study are described in section 2. The sprite parametrization and the criteria adopted for the model simulations are presented in section 3. The results are presented in section 4 and discussed in section 5. Finally, conclusions and recommendations for future observational campaigns are given in section 6.

2 The WACCM Model

The Whole Atmosphere Community Climate Model, WACCM, is a high top atmospheric model that is part of the NCAR Community Earth System Model (CESM). WACCM is an optional extension of the Community Atmosphere Model (CAM), one of the models that can be used for the atmospheric component of the CESM. WACCM has been designed to simulate the interactions of chemistry, radiation, and dynamics in the lower, middle, and upper atmosphere from the Earth's surface to 5.96×10⁻⁶ hPa. Chemical interactions are modeled using the Model of OZone And Related Tracers (MOZART3) [Kinnison et al., 2007]. The chemistry calculation includes all species and reactions that are expected to play a role in the middle atmosphere. Marsh et al. [2013] describe the current model version, WACCM 4, and show its success in simulating many aspects of the middle atmosphere.

In this study we use WACCM in the specified dynamics mode, abbreviated as SD-WACCM. Lamarque et al. [2012] describe the SD option. Horizontal wind and temperature fields in the troposphere and stratosphere are nudged at each model time step using meteorological analysis. In the current simulations, the analyses are from Goddard Earth Observing System 5 (GEOS-5). GEOS-5 analyses are available with a time resolution of 6h; the fields are interpolated to the 30min nudging intervals. The nudging is applied below 50 km, it is then tapered off between 50 and 60 km and removed above 60 km. SD-WACCM is therefore free running above 60 km. With the specified dynamics option of WACCM, the meteorology is constrained to be similar in separate model runs. This allows us to focus on the composition response to imposed changes. The present integrations have a horizontal resolution of 2.5° in longitude and 1.9° in latitude and have 88 levels in the vertical. Control and perturbed simulations for most of this study were run under summer conditions, starting on 1 July 2011 and lasting 40days. For comparison, we also produced a winter simulation starting on 1 January 2011 and lasting 40days.

3 Simulating Sprite Chemical Perturbations

A parametrization of sprite-induced NO_x was introduced in WACCM in terms of NO_x production by an individual sprite event and distributed following the expected sprite occurrence around the globe. While the global distribution of sprite occurrence is relatively well constrained by observations and knowledge of meteorology, the latter NO_x production per sprite event is uncertain by orders of magnitude. For both distribution and production, only nighttime conditions will be applied here as reference scenario. We therefore adopted a fixed distribution of sprite occurrence and maintained the NO_x production by an individual sprite as the main free parameter we wish to investigate in order to reach a significant chemistry climate response in WACCM.

3.1 Sprite Distribution

Sprites are related to lightning activity in thunderstorms and, in particular, to the occurrence of positive cloud-to-ground flashes [see, e.g., Pasko et al., 2012]. At the 1.9°/2.5° and 30 min resolution of the model, all sprite processes are instantaneous and subgrid. Clusters of sprites can be localized by their parent thunderstorms which generally extend for a few hundred kilometers and last for a few hours. The sprite NO_x perturbation is therefore introduced as an atmospheric volume affected by a certain number of sprites during the 30 min model time step.

At global scale, the distribution of sprite occurrence was shown to resemble that of lightning activity in observations from the ISUAL satellite [Chen et al., 2008], with about sprites observed over continental and coastal areas and over ocean. This can be supported by the large number of sprite observations recorded from ground-based facilities [e.g., Chanrion et al., 2007], which typically show a majority of sprites over continental areas, whereas a large gap exists in the distribution of these ground-based facilities over oceans and other not easily accessible regions. Estimates of the global rate of sprite occurrence range between 1/min [Chen et al., 2008] and 3/min [Ignaccolo et al., 2006], i.e., about 1 sprite every 1000 lightning flashes (which rates at about 40 per second in Christian et al. [2003]). In order to have a reference sprite distribution, we adopted a summer lightning scenario by Christian et al. [2003] as proxy for thunderstorms and sprites and considered the occurrence of one sprite every 1000 lightning flashes. We used the June, July, August (JJA) seasonal average data provided within LIS/OTD 0.5° High Resolution Monthly Climatology (HRMC). The resulting distribution is shown in Figure 1 and was applied to simulations under summer conditions. In order to study the effects of seasonal changes, we also produced a simulation under winter conditions (December, January, February - DJF). The adopted scenarios for the sprite distribution lead to a global average sprite activity of 3.9 sprites/min in July and 2.3 sprites/min in January.

Figure 1

Adopted distribution of sprite occurrence in summer (JJA) built from the summer average lightning distribution by Christian et al. [2003] and the global sprite occurrence rate of 3 sprites/min by Ignaccolo et al. [2006]. Data were regridded for the model simulations and are shown as sprite per hour per model input grid (5°×5°).

Sprite occurrence was applied to the model based on the above externally imposed lightning-sprite distribution and was kept constant in time. The choice of not taking into account the sprite daily cycle is due to the as yet unknown sprite occurrence during daytime, beside rare observations at twilight. The sprite occurrence will therefore not be consistent with the internally modeled properties of deep convection and midlatitude thunderstorm activity; however, it will be consistent with the cumulative impact of sprites because of the long NO_x lifetime in the mesosphere as discussed below.

3.2 Chemical Perturbation by a Sprite Event

We defined a reference sprite-NO_x production profile per event based on model streamer NO_x estimates diluted into a sprite-air volume. Because of the large (orders of magnitude) uncertainties affecting such estimates, the reference magnitude was then kept as a free parameter in the study.

Estimates of the chemical impact of individual sprite streamers at nighttime were produced by detailed ion-neutral chemistry models [Enell et al., 2008; Sentman et al., 2008; Gordillo-Vázquez, 2008; Hiraki et al., 2008, and follow-up works]. Ionization within sprite streamer heads lead to production of electrons that persist within the streamer trails for about 1s. Losses occur for dissociative attachment to ozone and for dissociative recombination with the positive ion cluster . Large populations of metastable species are produced with fractional density of ∼10⁻⁹−10⁻⁸, e.g., O(¹D), O(¹S), N(²D), leading to consequent reactions that fade within the 1s timescale, while their related active neutral species can persist for 100s due to the lack of dissipation channels. Persistent changes in NO_x are predicted to occur by all the models, with NO production being promptly converted into NO₂ at night and then persisting for longer than 1 h. Sentman et al. [2008] estimated that 5×10¹⁹ NO molecules are produced at 70 km altitude per sprite streamer. Enell et al. [2008] estimated changes in NO_x by 30% at 83 km, 50% at 73 km, 4% at 63 km, and 2% at 53 km under typical conditions, with 200% at 81 km and 500% at 71 km above extraordinary sprite-producing thunderstorms. Gordillo-Vázquez [2008] found 1 order of magnitude change for NO and NO₂ at 68 km altitude and 3 orders of magnitude increase for NO₃. Hiraki et al. [2008] found up to 6 orders of magnitude increase in NO_x, although they point out that improved rates would lead to 3 orders of magnitude change. All these estimates are subject to orders of magnitude uncertainties due to the adopted electric fields and streamer parameters.

Refinements to the above estimates were discussed by follow-up works of the same authors. Gordillo-Vázquez and Donkó [2009] considered the impact of humidity and gas temperature on the electron energy distribution and found a small impact for the conditions at the sprite streamer tips (although they can be relevant for blue jets and gigantic jets). Sentman and Stenbaek-Nielsen [2009] investigated chemical effects in the trailing columns of sprite streamers under the weaker electric fields and found modest enhancement of metastable species compared to field-free tails. In their study, they estimate a conduction current moment of less than 45A km for a trailing column of 10 km, while estimates derived from ELF measurements are of the order of 200kA km. This large difference is due to both the adopted approximations and to the large number of streamers occurring within one sprite event and leading to the observed ELF signature. If we attempt to constrain the streamer-to-sprite NO_x production relationship, this would require a scaling factor of 4500 which appears as a reasonable estimates of the number of streamers involved in a sprite event. The 5×10¹⁹ NO molecules at 70 km altitude per sprite streamer by Sentman et al. [2008] multiplied by 4500 streamers, i.e., 2×10²³ NO molecules, is then in fair agreement with the prediction by Enell et al. [2008, Table 1] for their typical case scenario at 70–80 km altitude (3×10²² – 3×10²³).

Since we are interested in a reference perturbation profile and in a maximum perturbation simulation, we adopted the maximum case scenario from Enell et al. [2008]. This choice takes in consideration also the larger estimates obtained by the two other sprite chemistry model studies. We assume a horizontal section of 50×50 km², and an altitude extent of 50–90 km in which the perturbation is introduced. The adopted perturbation profile per sprite is shown in Figure 2. Winkler and Notholt [2014] found very small differences in the NO_x produced by daytime and nighttime sprites, although the partitioning into NO during daytime leads to a more efficient action on ozone. Since we include in the model sprite NO molecules, the daily change in the partitioning and effect on chemistry is already dealt with by WACCM chemistry scheme. On the contrary, the possible change in altitude of occurrence of daytime sprites as compared to the adopted typical altitude range was not considered to allow an easier interpretation of the results. Results obtained with the perturbation profile at this reference magnitude will be then discussed in terms of current knowledge on sprite chemistry in the following sections.

Figure 2

Sprite-NO_x perturbation profile adopted from the maximum case scenario of Enell et al. [2008]. (left) Total number of molecules produced by one sprite over a 50 km×50 km region at various altitude levels (slices of 10 km thickness) filled by streamers at a fraction 0.01 of the volume. (right) As in Figure 2 (left) but in terms of percent change compared to the atmospheric background. See details in Enell et al. [2008].

4 Results

As a first test, a sprite-NO_x perturbed lower mesosphere scenario was applied to WACCM by assuming sprite perturbations with magnitudes significantly higher than the reference value, adopting factors of 10 and 100 (hereafter labeled 10x and 100x). These model simulations allowed to obtain a significant model response under extraordinary sprite activity and helped to understand the main characteristics of the response prior to reducing the magnitude to realistic values.

Results for the high-magnitude simulation 10x in summer are presented in Figures 3-5. The 10x perturbed simulations show the formation of a secondary peak in NO_x at about 70 km altitude. Similar but much more perturbed results are obtained in the 100x case. Figure 3 shows the NO_x distribution on the first and second day of the simulation at the altitude of the 70 km peak. Local sources are clearly identifiable above main continental lightning chimneys, resembling the large sprite activity confined in the summer Northern Hemisphere. As sprite-NO_x builds up above the main continental sources, transport competes to spread it around the globe. The simulations show that it takes about 1 day to exceed 2 ppbv of NO_x both in central Africa and in Central America, with a slower buildup elsewhere. Southern midlatitudes are affected also by a contribution from high-latitude downward transport of mesospheric NO_x during the first days, and to a lesser extent by a buildup of weaker sprite-NO_x sources under longer night time hours.

Figure 3

Distribution of WACCM NO_x (ppbv) from the sprite-NO_x simulation 10x at 70 km altitude over the Tropics. Results are shown for (left) day 1 and (right) day 2 of the simulation at (top to bottom) 0, 4, 8, 12, 16, and 20 UT.

Figure 4

Cross sections of WACCM NO_x at (a–c) 25° longitude and (d–f) 70 km altitude at 0UT on the last day of the simulations. Results are shown for the 1x sprite-NO_x simulation (Figures 4a and 4d), unperturbed control simulation (Figures 4b and 4e), and difference between sprite-NO_x and control (Figures 4c and 4f, in percent). Refer to color bars on the right hand side. Horizontal white dashed lines in Figures 4a–4c indicate the altitude of the cross sections reported in Figures 4d–4f, and vice versa. White patches indicate data below color scale.

Figure 5

Evolution in time of WACCM NO_x at (a) 70 km and (b) 58 km altitude for the sprite-NO_x simulation 10x (bold lines) and control simulation (dashed lines). Results are for target regions central North Africa (red), central North America (yellow), Southeast Asia (green), and the Tropics (black).

Already within a few tens of hours, high levels of NO_x are seen outside the continental source regions, following zonal transport. This is particularly evident across the Atlantic, the Northern Pacific, and the equatorial Indian Ocean. Within a few weeks, transport leads to spread the NO_x sources across most of the Tropics. The NO_x distribution in the sprite-NO_x10x simulation is compared with the unperturbed control simulation in Figure 4 on the last day of the simulation. The formation of the secondary NO_x peak around 70 km altitude is visible comparing model cross sections at 25° longitude (Figures 4a–4c): the sprite-NO_x maximum occurs where background NO_x in the unperturbed atmosphere is negligible. The difference exceeds 400% at its maximum. The altitude affected by the perturbation is consistent with the vertical profile of the applied perturbation (i.e., 55 to 90 km altitude), with changes limited to North of −40° latitude. In the Southern high latitude, changes are seen due to both the buildup of the weak sprite-NO_x source present at Southern midlatitudes in July and to the natural variability of the Southern winter polar region. Geographically (Figures 4d–4f), NO_x builds up more easily in the Tropical summer (Northern) hemisphere, with differences with respect of the unperturbed control simulation exceeding 100% in most of the Tropics.

Figure 5 shows the average NO_x at 70 and 58 km altitude above the Tropics and above the three most active regions central North America (0° to 30°N and −107° to −70°E), central North Africa (0° to 15°N and 5° to 35°E) and Southeast Asia (0° to 30° and 90° to 120° E). This allows a quantitative evaluation of the response comparing sprite-NO_x simulation (bold lines) with the control simulation (dashed lines). At 70 km altitude, transport produces a zonally averaged NO_x enhancement spread around the Tropics (Figure 5, black line), with regional levels (colored lines) slowly reaching an equilibrium with the zonally averaged Tropical levels. This implies that first, despite having strongly localized sources (above the continents), after a few weeks the local source is less competitive as compared to the zonally averaged contribution of the previous weeks; second, both the regional levels and the background zonally averaged levels of NO_x tend to saturate to a maximum value, when the sprite-NO_x source gets in equilibrium with its sinks (whereby its components due to photolysis are proportional to the available amount of NO_x). After about 20 days, the Tropical average at 70 km altitude has reached its maximum of about 2.1 ppbv and persists at this level for the rest of the simulation. Regional NOx keep oscillating but tend to converge around higher levels. This is due to the stronger sprite activity in the summer (Northern) hemisphere. As compared to the background 0.8 ppbv of unperturbed NO_x, this implies more than 150% increase in the zonally averaged Tropical NO_x and a 2–300% increase in the active regions. At 58 km altitude, just above the altitude of the weakest applied perturbation, the zonally averaged NO_x at the end of the simulation shows a much smaller increase of about 0.2 ppbv (i.e., few percent), although persistently above the level of the unperturbed simulation. These changes are reflected also above the active regions with maxima reaching about 10% increase, although they oscillate loosing significance at random times. Clearly, changes at 58 km altitude are strongly dependent on the period and region considered.

4.1 Dependence on Magnitude

Since a clear model NO_x response was obtained applying the large 10x magnitude simulations, we performed further simulations of reduced magnitude until the response was lost. Results at 70 km altitude for magnitudes 1x, 0.2x, and 0.1x are shown in Figure 6, focusing on the zonal average over the Tropics. The results are reported in terms of difference to the control simulation. The 1x perturbation shows a saturation response of about +0.15 ppbv after day 20, and peaks at +0.2 ppbv, down from the +1.3 ppbv of the 10x case; The 0.2x perturbation shows a saturation response of about +0.03 ppbv, peaking at +0.07 ppbv, whereas the 0.1x perturbation leads to a response which barely deviates from the control simulation. The 0.2x magnitude sets therefore the threshold of minimum magnitude for sprite-NO_x perturbations we looked for. The magnitude of the response therefore scales almost linearly close to about 0.15 ppbv per unit of perturbation. Scaling down to the 0.1x case, the expected response is 0.015 ppbv which then falls below the variability visible in the trends. For very large perturbations, the linearity is lost and the response saturates at levels which are proportionally lower: the change in the 10x case was about 10 times larger than the change in the 1x case, whereas the 100x perturbation led to about only 6 times the increase of the 10x perturbation simulation.

Figure 6

Evolution in time of WACCM NO_x at 70 km altitude over the Tropics for the sprite-NO_x simulations (a) 1x, (b) 0.2x, and (c) 0.1x. Results are shown as deviation from the unperturbed control simulation. Figure 6a reports also results above individual regions color coded as in Figure 5.

Regional changes above active regions are shown in Figure 6a only for the 1x case because they are too noisy at lower magnitude. In the 1x perturbation case, NO_x above active regions is persistently higher than in the control simulation and randomly increase by as much as 50% above Africa (compare colored bold lines in Figure 6a to dashed ones in Figure 5a). In contrast, at 0.2x and 0.1x, NO_x above individual regions of the perturbed atmosphere is often below that in the control simulation, making the local response highly dependent on dynamical conditions.

4.2 Dependence on Variability

In order to evaluate the significance of the NO_x changes found in the simulations as compared to atmospheric variability, we studied the model response for two further 1x simulations having slightly different atmospheric conditions. All 1x simulations were driven by the same meteorological fields in the troposphere and stratosphere but differ in the mesosphere due to the small differences in the initial conditions. The standard deviation of NO_x at 70 km altitude among the three 1x simulations is shown in Figure 7. Considering zonal averaged NO_x over the Tropics, the variability is typically around 0.01–0.02 ppbv, with the largest values rached in the second half of the simulation. The model response to NO_x perturbations down to the 0.2x magnitude are therefore larger than variability and the above considerations can be considered robust. Consistently, the 0.1x perturbation case with expected response of 0.015 ppbv falls at the edge of the variability and may therefore be considered to be buried in the atmospheric background. Above individual regions, the variability among different simulations affected by the very same perturbation is strongly oscillating and can exceed 0.05 ppbv because of the large dependence on transport. The response above a certain region will therefore be strong only on timescales smaller than the local timescale of transport, above which dynamical variability dominates.

Figure 7

Evolution in time of the variability (standard deviation) of WACCM NO_x at 70 km altitude over the Tropics among three sprite-NO_x simulations of magnitude 1x having slightly different starting conditions. Individual regions are color coded as in Figure 5.

4.3 Time Dependence

The above simulations were performed applying a perturbation which was kept constant in time, therefore neglecting both the randomness and daily cycle in the occurrence and characteristics of sprites. The meaningfulness of such simulations depends on the timescale of relaxation of the atmospheric NO_x back to unperturbed conditions after a sprite perturbation has halted. Figure 8a shows the results of a 1x simulation whereby the sprite perturbation was halted after 20 days, i.e., around the time when saturation is reached. The figure compares these results (dashed line) with those obtained in the previous 1x case (bold line) in terms of difference with the control simulation. Halting the perturbation causes the atmospheric NO_x to relax back to background conditions in about 15 to 20 days, i.e., roughly the same period that it takes to reach saturation from the start of the simulation.

Figure 8

Evolution in time of WACCM NO_x at 70 km altitude over the Tropics for the sprite-NO_x simulation 1x (bold lines) (a) in summer as in previous figures (JJA) and (b) in winter (DJF). Figure 8a reports also a simulation 1x in summer with perturbation halted after day 20 (dashed lines).

The relaxation is more evident in the Tropical zonal average, whereas NO_x above individual target regions tends to oscillate around the Tropical values making it more difficult to define a timescale (not shown). Relaxation timescales of about 15 days imply that sprite NO_x perturbations which pulse on shorter periods will act with a smoothed integrated effect: this supports the results obtained in the simulations previously discussed, including the neglect of a daily cycle. Clearly, at a local level and on timescales close to the individual pulse of a burst of sprite activity above a thunderstorm, the much larger sprite NO_x production confined in such a short period will lead to a temporary stronger change.

On longer timescales, seasonal changes lead to a shift in lightning activity, which mainly occurs in the summer hemisphere. In order to investigate the impact of the atmospheric background under the opposite season conditions, Figure 8b reports the results for the Tropical average NO_x at 70 km altitude for simulations starting on 1 January in terms of difference with the respective control simulation. The 1x simulation saturates at smaller values, about 0.1–0.13 ppbv larger than the unperturbed case in the January simulation, in fair agreement with the reduced global sprite activity as compared to July (about 2.3 sprite/min in January versus 3.9 sprite/min July). Seasonal changes can therefore reduce the Tropical NO_x change by about .

5 Discussion

The model simulations we performed allowed an evaluation of several aspects of the chemical impact of sprites and other thunderstorm-related processes onto the atmosphere.

The simulations aid describing how the sprite-NO_x is transported, gets zonally homogeneous and saturates. The obtained results show that, without consideration of the competition of transport and sinks, previous attempts to estimate the relevance of sprite-NO_x to the atmosphere at regional or global level [see, e.g., Enell et al., 2008; Neubert et al., 2008] are unreliable since the atmosphere needs to reach a new equilibrium with the sprite source: In the absence of an equilibrium with transport and sinks, sprite-NO_x may be added over the years until the sprite-NO_x-dominated scenario would occur. With the 10x perturbation, the Tropical atmospheric NO_x at 70 km altitude increases its equilibrium value from 0.8 ppbv to 2.1 ppbv (i.e., a +1.3 ppbv increase) and then persists at this level. This is a value of NO_x in contrast with NO_x climatological background minima reported by observations at this altitude; we can therefore conclude that such magnitude for sprite perturbation is not realistic. The nonlinearity of the response suggests that weaker perturbations will require more time for reaching an equilibrium with the sink, if an equilibrium at all can be reached (see, e.g., model results at 58 km in Figure 5).

Scaling down the magnitude to more realistic values, the increase in the Tropical lower mesosphere is of the order of +0.2 ppbv or smaller, which is not unexpected from observations of climatological NO_x at this altitude. The magnitude of the perturbation was decreased until the response was lost below 0.2 times our reference scenario. Our simulations point to a 0.15 ppbv contribution to Tropical NO_x at 70 km altitude by sprites assuming Enell et al. [2008] maximum case, down to 0.015 ppbv buried within background variability assuming the largest of their typical case values. This is of the order of the variability we found for Tropical NO_x at 70 km altitude (about 0.015 ppbv). Below these values the sprite contribution to Tropical NO_x becomes irrelevant. We further show that relaxation timescales for Tropical NO_x at 70 km altitude are of the order of 15 days, so that the global effects of sprite NO_x will be integrated over such time periods even though they occur impulsively. Seasonal changes can reduce these estimates by about in winter, consistently with a reduced global sprite activity.

We also used the model results to interpret available observational studies. Considering periods of maximum buildup under slow transport, we can expect peaks of more than 10% NO_x change at 60–85 km altitude. At this magnitude, sprite-NO_x have detectable levels, especially under favorable conditions such as above intense sprite-producing thunderstorms and with low background winds. This confirms what was suggested by the analysis of satellite observations of NO_x by Arnone et al. [2008b], Rodger et al. [2008], and Arnone et al. [2009], with significant impact at local scale and negligible at global scale. On the other hand, we performed ad hoc calculations to resemble the method applied by Rodger et al. [2008]: We found that the use of partial columns of NO_x extending down to 50 km altitude as they adopted is not sensitive enough to variations of sprite perturbations, since the lower atmospheric layers having the major weight in the partial columns have almost no response to sprite-NO_x. The lack of sensitivity was found even for our 10x case scenario. We therefore recommend to repeat these studies including only partial columns at the peak altitude of the sprite perturbation.

Orders of magnitude uncertainties affect the available estimates for TLE chemistry at various stages of their calculation. At all scales, in fact, the capability of correctly resolving relevant processes is showing to be fundamental in correctly understanding the evolution of a TLE event, e.g., as shown by the first simulation with self-adaptive grid of an individual sprite streamer channel emerging from the ionosphere [Luque and Ebert, 2009]. To start overcoming these shortages, studies such as Enell et al. [2008] and Hiraki et al. [2008] have diluted their modeled chemical products due to an individual streamer tip into the volume of air of an idealized event and further into the whole atmosphere by reasonable dilution factors, taking into account the expected size of individual streamers, a likely number of streamer per sprite event, the volume of one event, the number of events per thunderstorm, and finally the global occurrence of these events. These diluted calculations showed that sprites-NO_x, perturbations may (a) reach hundreds of percent increase of NO_x over very active thunderstorms, therefore reaching satellite detection levels and (b) give a negligible contribution at a global scale. However, depending on whether the orders of magnitude uncertainties are acting on the low or high side of these estimates, sprite can as yet range from being irrelevant also at a local level to be one of the main NO_x sources in the atmosphere. The thresholds we evaluated can now be used to constrain these sprite chemistry models.

Some comments should be spent on the adopted sprite distribution and occurrence. The vast majority of sprites are triggered by positive CG flashes which are approximately 10% of all CGs. This would require a refinement of the zero-order distribution, in terms of temporal and spatial distribution. In particular, the distribution of −CG activity is observed to peak around 3–4P.M. local time, increasing from almost no lightning during the whole morning and fading at late night. Positive CGs tend to occur preferably during late stages of thunderstorm development, so that their peak activity is shifted forward in time-triggering sprite activity deep into nighttime. Since WACCM does not discern the polarity of CGs, the main error affecting the approximation of our lightning distribution is a lag in time of the sprite perturbation. A simplified correction may be applied by shifting forward in time the sprite perturbation by 6 h, i.e., by the time lag between −CG activity and a typical peak in sprite activity [see, e.g., Arnone et al., 2008]. At global scales, the meaningfulness of such a time lag is however diminished by consideration of the 15 days relaxation timescale and would therefore apply only to the study of the local production.

Some refinements would also be required in the adopted geographical distribution. In midlatitude coastal areas and above seas, and possibly especially during the winter season, the fraction of +CG to −CGs can reach unity and TLE development can be more favorable, as observed, e.g., in small thunderstorms over the Mediterranean sea [van der Velde et al., 2010], or over North America coastal areas [Ely and Orville, 2005] and Japan seas [Takahashi et al., 2003]. The distribution of sprite occurrence based on lightning occurrence should be then refined in space and time considering a much higher fraction of occurrence over seas and in winter, and seasonal changes. Our simulations adopting idealized seasonal sprite distributions showed that the average estimates over the Tropics are only marginally affected, whereas the regional response will clearly scale increasing the activity in a certain region. These changes in the +CG/−CG ratio are hard to investigate in the Tropics because of the paucity of direct measurements, although the +CG/−CG ratio appears to have larger changes at midlatitudes. A more refined correction covering both the second and third issues discussed would need consideration of the triggering mechanisms of +CGs related to parameters available in the model to correctly identify sprite-triggering lightning activity.

Although we focused here on sprites, early modeling by Mishin [1997] showed that a single blue jet led to 10% change in NO_x and 0.5% change in ozone at 30 km altitude. Lehtinen and Inan [2007] found substantial ionization below 50 km altitude that persists for more than 10 min after the occurrence of a gigantic jet in their simplified five constituents model of the stratosphere-lower ionosphere chemistry. Neubert et al. [2011] discussed the persistent influence of a gigantic jet in the formation of a subsequent sprite. Hiraki et al. [2004] and Parra-Rojas et al. [2013] developed chemical models and applied them to the case of halos. These early results point toward a chemical impact from blue and gigantic jet streamers onto the stratosphere and further support the results for sprite streamer simulations. Because of the much larger number density, it is very unlikely that a global effect can be reached by these processes at lower altitude. However, local effects under favorable transport conditions may not be excluded. On the other hand, the results of our study at sprite altitude are valid for any other kind of thunderstorm-induced NO_x perturbation at sprite altitude, which include, e.g., TGFs, accelerated electrons [see, e.g., Fullekrug et al., 2013], and other thunderstorm-related processes in the lower mesosphere.

6 Summary and Conclusions

An assessment of sprite-NO_x perturbations in the lower mesosphere was carried out using the WACCM whole atmosphere model with inclusion of specified NO_x perturbations of various magnitudes. The perturbed simulations showed the formation of a secondary peak in NO_x at about 70 km altitude and allowed a clear interpretation of the effects of transport and saturation of the source. Transport creates a zonally averaged NO_x enhancement spread around the Tropics, whereas it allows a strong buildup above source regions at random favorable times.

Our simulation points to a 0.15 ppbv contribution to Tropical NO_x at 70 km altitude by sprites (about 20% of the background) assuming Enell et al. [2008] maximum case prediction, down to 0.015 ppbv (buried within background variability) assuming the larger of their typical case values (respectively 1.5·10²³ and 1.5·10²⁴ molecules per sprite slice in the 70–80 km layer). Winter conditions lead to estimates reduced by about . This values are above the variability we found for Tropical NO_x at 70 km altitude, about 0.015 ppbv. The possible difference in daytime occurrence of sprites may affect our estimates. Below these values, the sprite contribution to Tropical NO_x becomes irrelevant, so that predictions by sprite streamer chemistry models should be constrained by these reference thresholds. We further showed that relaxation timescales for Tropical NO_x at 70 km altitude are of the order of 15 days, so that the global effects of sprite NO_x will be integrated over such time periods even though they occur impulsively.

Due to background variability and transport, enhancements above a particular region tends to oscillate and show up only under specific days. Scaling the magnitude of the applied perturbations, we can expect peaks of more than tens of percent change in NO_x at 60–85 km altitude and with detectable levels, confirming suggestions by previous observational studies. We point out to the importance of the effects of transport on the detectability of such perturbations and recommend the use of vertical resolved measurements or partial columns targeting the 65–85 altitude regions in order to be sensitive to sprite perturbations. The vast family of thunderstorm-induced processes that are being discovered to occur above thundercloud tops suggest that the overall impact of thunderstorms on the upper atmospheric chemistry may be larger than previously predicted.

Acknowledgments

E. A. acknowledges support by the European Space Agency for the project CHIMTEA within the framework of the Changing Earth Science Network Initiative. The CESM project is supported by the National Science Foundation (NSF) and the Office of Science (BER) of the U.S. Department of Energy. Computing resources were provided by the Climate Simulation Laboratory at NCAR's Computational and Information Systems Laboratory (CISL), sponsored by the NSF and other agencies. NCAR is sponsored by the NSF. The LIS/OTD 0.5 Degree High Resolution Monthly Climatology lightning data were obtained from the NASA EOSDIS Global Hydrology Resource Center (GHRC) DAAC, Huntsville, AL. http://thunder.nsstc.nasa.gov/.