Linking solar minimum, space weather, and night sky brightness

This paper presents time-series observations and analysis of broadband night sky airglow intensity 4 September 2018 through 30 April 2020. Data were obtained at 5 sites spanning more than 8500 km during the historically deep minimum of Solar Cycle 24 into the beginning of Solar Cycle 25. New time-series observations indicate previously unrecognized significant sources of broadband night sky brightness variations, not involving corresponding changes in the Sun's 10.7 cm solar flux, occur during deep solar minimum. New data show; (1) Even during a deep solar minimum the natural night sky is rarely, if ever, constant in brightness. Changes with time-scales of minutes, hours, days, and months are observed. (2) Semi-annual night sky brightness variations are coincident with changes in the orientation of Earth's magnetic field relative to the interplanetary magnetic field. (3) Solar wind plasma streams from solar coronal holes arriving at Earth’s bow shock nose are coincident with major night sky brightness increase events. (4) Sites more than 8500 km along the Earth's surface experience nights in common with either very bright or very faint night sky airglow emissions. The reason for this observational fact remains an open question. (5) It is plausible, terrestrial night airglow and geomagnetic indices have similar responses to the solar energy input into Earth's magnetosphere. Our empirical results contribute to a quantitative basis for understanding and predicting broadband night sky brightness variations. They are applicable in astronomical, planetary science, space weather, light pollution, biological, and recreational studies.


Data collection and analysis
In this paper, photometers are employed to obtain a time-series of differential photometric brightness measurements of the same place on the celestial sphere along the zenith declination relative to celestial sources. This procedure minimizes errors encountered using data from several different instruments and individual instrument drift in sensitivity if it exists.
Our research is enabled by accurate, low cost, scientific quality SQM-LU-DL 31 and TESS-W 32 photometers. They provide continuous measures of zenith night sky brightness dusk to dawn every night. These two photometers use the same detector, have slightly different fields of view, and different red responses 31 . The SQM-LU-DL uses a color filter. The TESS-W employs a dichroic filter. These filters and TSL237 photodiode they both employ sets the spectral response of each instrument. The SQM-LU-DL has a spectral response which spans most of the Johnson B and V filters while the TESS-W has a substantially greater red response spanning more than the Johnson-Cousins B,V, and R pass bands 9 . The SQM-LU-DL and TESS-W report instrumental M(t) values in mag/arcsec 2 . These two instruments are slightly different measures of broadband night sky brightness. However, their differential photometric measurements produce similar results. In the differential photometry mode we employ, our photometers are more accurate (error < 0.03 mag/arcsec 2 ) compared to when used as calibrated absolute photometers (error ~ 0.1 mag/arcsec 2 ) 31 . The irradiance-to-frequency semiconductor detector employed by both the SQM-LU-DL and TESS-W instruments is calibrated to report the measurements in mag/arcsec 2 . 33 .
An approximate conversion of mag/arcsec 2 to cd/m 2 is: L 0 is 1.475 × 10 5 cd/m 2 in the AB System and 1.2216 × 10 5 cd/m 2 in the Vega system 34 .
On clear, astronomically dark nights, these single channel photometers, pointed at zenith, measure light accumulated from terrestrial airglow, stars, planets, scattered star light, zodiacal light, nebulae, galaxies, other faint astronomical sources, and anthropogenic skyglow, if present. SQM and TESS-W photometers sum all emissions in a broad cone to the edge of space over a relatively wide area of the sky. These characteristics make www.nature.com/scientificreports/ it difficult to identify the physical processes creating the emissions. These instruments produce time-series data to identify broadband airglow brightness events for further study. Time-series data are collected at Cosmic Campground International Dark Sky Sanctuary (CCIDSS) and Catalina Sky Survey Mt. Lemmon Station (CSSMLS). Data from TESS-W photometers located at Spain Observatorio Astrofísico de Javalambre-Arcos de las Salinas/Teruel (Stars 18), Centre d'Observació del'Universe, Àger, Lleida, Spain (Stars 62), and Observatorio del Teide, Izaña, Tenerife, Spain. (Stars 211) were downloaded from the TESS Data Monthly data files using IAU-IDA format 35 .
The artificial levels in Table 1 are estimates from satellite data adjusted by ground-based SQM observations 36 . CCIDSS is a unique standard more than 60 km away from any significant source of artificial light. Analysis of all-sky images and the satellite estimate, of less than 1/2% artificial light, indicates anthropogenic skyglow is unmeasurable at zenith at CCIDSS 36,37 .
We have developed techniques and software to process the time-series data. Our software selects individual instrumental time-series measurements, M(t), taken at time t, when the Sun was more than 18° below the horizon, the Moon was more than 10° below the horizon, and the sky was clear. The 10° Moon limit is necessitated by large changes in lunar brightness as the Moon passes through its phases every month. On photometric nights, at the CCIDSS, measurements show that if the Moon is less than 74% illuminated it provides negligible light at zenith when it is 10° or more below the horizon. The situation changes rapidly when the Moon approaches the horizon. For example, a 19% illuminated Moon, 5° below the horizon, increases the zenith sky brightness by 0.02 mag arcsec 2 . The Moon's zenith illumination, when it is still below the horizon, could be effected by moisture and/or dust in the atmosphere thus it is better to be conservative in selecting a value for this parameter 31 .
Sky clearness is measured by computing Chi Squared from a straight line fit to the data extending for 45 min on either side of the point in question. A Chi Squared of less than 0.009 for at least 1.5 h rejects suspect data but not the rising Milky Way 31 . This metric is employed to exclude data from marginal non-photometric nights. As an additional check, the TESS-W near IR sensor is employed to estimate cloud cover 32   www.nature.com/scientificreports/ R.A., is due to changes in terrestrial airglow. Since M(t) are measured in mag/arcsec 2 brighter values are smaller numerically. Plots similar to Fig. 1, for the other sites, observationally establish a quiescent value of airglow at each location on the celestial sphere. The existence of a minimum intensity of broadband airglow at each RA on the celestial sphere is a unique observational fact for each site. Alternatively, one could, also, establish a value for the quiescent airglow level by adding up the minimum values for all known sources of diffuse night sky brightness 16 .
Along the zenith declination on the celestial sphere, 1/2 h bins in R.A., are used as 48 standard candles. Each such standard candle is the average of the 10% faintest M(t) measurements at its location on the celestial sphere. A set of joined polynomials are fitted to the 48 standard candles to produce a continuous light curve of quiescent broadband airglow brightness. This fitted M(t) versus RA light curve is the background contribution from celestial sources and the broadband night airglow when it is at minimum. These concepts using data from CCIDSS are shown in Fig. 1. Each site has a unique light curve of quiescent broadband airglow brightness which depends on its latitude and the degree to which it is influenced by anthropogenic light. These light curves are an observational characteristic of each site. Each measured point's brightness above the quiescent airglow, ΔM(t), is obtained by subtracting the continuous light curve of quiescent broadband airglow from the data, point by point. This procedure removes light from the stars, planets, Milky Way, zodiacal light, other celestial sources, and constant anthropogenic light if present. Thus, each ΔM(t), is a differential photometric night sky brightness at time t, relative to the brightness of the same point on the celestial sphere when the airglow is at minimum. The same procedure is used to produce, ΔM(t), a time series of airglow brightness above its quiescent level for each site.
To evaluate conditions during the night at each site, we averaged the ΔM(t) data into 1/2 h time intervals relative to local midnight. The results are plotted in Fig. 2. The error bars, produced by real airglow variations, are ± 1 standard deviation for the 1/2 h bin averages. At CCIDSS, the quiescent airglow light curve is relatively flat (average 0.18 mag/arcsec 2 ). The standard deviation of 0.136 mag/ arcsec 2 is produced by real changes in airglow. On long winter nights, before the onset of astronomical twilight, there does seem to be an increase of approximately 0.07 mag/arcsec 2 . The origin of this increase is unclear. There appear to be other reports of this phenomena in the literature 38 . It is plausible natural night sky broadband airglow, in certain situations, has a UT dependence similar to the geomagnetic indices 29 .
At CSSMLS, ΔM(t) is correlated with automobile driving patterns and scheduled outdoor lighting changes in and around Tucson, AZ. At Stars 18, Stars 62, and Stars 211 there appears to be prolonged morning and evening twilight when the Sun is more than 18° below the horizon. This result could be due to the extended red response of the TESS-W photometers and/or the European pattern of artificial lighting in nearby cities.
For each site the ΔM(t) data versus the time relative to local midnight were fit to a quadratic function. Corrections using these functions were used to flatten the curves of Fig. 2, tighten the agreement between sites, and are small. In mag/arcsec 2 , the site (median, stdev) correction values are CCIDSS (0.005, 0.008), CSSMLS (− 0.018, 0.056), Stars 18 (0.017, 0.048), Stars 62 (0.016, 0.040), and Stars 211 (− 0.033, 0.029). Each data point, ΔMC(t), is the corrected broadband airglow brightness above the continuous light curve of quiescent airglow brightness (Fig. 1). The resulting time-series of data points, ΔMC(t), for each site are used to track the natural zenith broadband night sky airglow uncontaminated by celestial and other sources above its quiescent level.
We compare the zenith natural broadband airglow above its quiescent level, ΔMC(t),during a deep solar minimum, with conditions in the near Earth environment. We employ solar wind data as compiled and presented on the NASA Omni Plus Browser 39 , sunspot counts, 10.7 cm(t) (2.8 GHz) solar flux, and Geomagnetic indices [40][41][42][43] . We

Results
Time-series photometric observations were taken at 5 sites spanning more than 8500 km during the minimum of Solar Cycle 24 into the beginning of Solar Cycle 25. During this time one might have expected nightly airglow variations to be at a minimum. These data verify the natural night sky is rarely, if ever, constant in intensity. At a natural night sky location like CCIDSS during deep solar minimum the 5 min cadence time-series broadband airglow had a range in ΔMC(t) of 0.729 mag/arcsec 2 . This corresponds to a maximum/minimum flux ratio of 1.957. The nightly airglow average ΔMCN(t) over a total of 241 nights at CCIDSS ranged from a very active night (2,458,792.979282JD) at − 0.536 mag/arcsec 2 to a quiescent one (2,458,522.840410 JD) at 0.028 mag/arcsec 2 . This nightly range of − 0.564 mag/arcsec 2 corresponds to a flux ratio maximum/minimum intensity ratio of 1.68.
At CCIDSS, September 2018 through April 2020, 10 nights were recorded to have an average minimum broadband SQM brightnesses of 22.07 mag/arcsec 2 (stdev 0.03 mag/arcsec 2 ). These data were accumulated in the RA range 10.5-12.5 h.
Despite our instrument's dusk to dawn coverage every night, the observations were unavoidably interrupted by Sun, Moon, clouds, and instrument down time. For the total elapsed time during this research, 4 September 2018 through 30 April 2020, the % of time logged during clear astronomical dark conditions was for CCIDSS (7.4%), CSSMLS (5.6%), Stars 18 (1.6%), Stars 62 (3.3%), and Stars 211 (5.1%). Our data must be regarded as a small sample of sky brightness during this time. Thus, unless one has a worldwide network of monitoring stations, at dark sky locations, many important airglow events will be missed.  Fig. 4. The celestial and anthropogenic sources have been removed as outlined in "Data collection and analysis" of this paper. The x axis is the time in Julian Date and the vertical axis is the nightly average of broadband airglow brightness, ΔMCN(t), above its quiescent level. In this paper ΔMCN(t) is defined to be the nightly average broadband airglow intensity above its quiescent level with the quadratic correction of Fig. 2. Since ΔMCN(t) are measured in mag/arcsec 2 brighter values are smaller numerically. The error bars are ± 1 standard deviation. These standard deviations represent changes in broadband airglow during the night. Figure 4 shows during a deep solar minimum natural night broadband airglow brightness varies by more than 0.5 mag/arcsec 2 (intensity ratio 1.58). Local night sky airglow brightness events span a distance of a few hundred km. Others can extend 8500 or more km along the Earth's surface (see   www.nature.com/scientificreports/ Event A is one of several broadband airglow increase events during an active period that spanned several solar rotations. Event A began at 2,458,398.895833 JD when a pulse in the dimensionless solar wind proton kinetic energy, NKE(t), was observed at Earth's bow shock nose. At this time our planet encountered an energetic stream in the solar wind. Before Event A the space weather conditions were relatively steady. The upper panel of Fig. 6 plots some space weather parameters before and after Event A. The dimensionless solar wind kinetic energy, NKE(t), increased to more than four times its median level. NAp(t), the daily dimensionless Potsdam geomagnetic Ap index increased to more than five times its median value. The peak Ap on day 2,458,399 JD was 56. It increased to more than five times its median level, indicating a significant geomagnetic disturbance. Meanwhile, the solar EUV as measured by the 10.7 cm(t) radio flux remained near its median value of 70 SFU.
In the lower panel of Fig. 6 the x axis is the time in Julian Date-2,458,000 (days) and the vertical axis is the nightly average of broadband airglow brightness, ΔMCN(t), at CCIDSS, above its quiescent level (mag/arcsec 2 ). The error bars are ± 1 standard deviations about the mean. They are real changes during the night. The time-series data for individual nights are plotted in Fig. 8.
At 2,458,399.15646 JD an astronaut on the International Space Station took an image showing Earth engulfed in a bright orange airglow (Fig. 7).   Fig. 10 shows the image's temporal relationship to the other data. In Fig. 9, Green (558 nm) oxygen and orange (589 nm) sodium airglow are visible over the entire sky. The R,G, and B channels in the digital camera data provide estimates of the strength and spatial structure of the oxygen and sodium lines 37 Fig. 10. The upper panel is a plot of normalized solar wind and geomagnetic parameters versus the time in JD (days). The lower panel is a plot of the 5 min cadence broadband time-series airglow observations above the quiescent level, ΔMC(t), for each site in mag/ arcsec 2 versus the time in JD-2,458,000 (days). There was a slow rise and fall in the airglow brightness over a two week interval. This period of time was characterized by predominately negative [B(t)z]GSM and a NKE(t) which varied significantly above its median value. Event B began near 2,458,780.8333 JD (24 October 2019) when a high energy stream in the solar wind produced a shock wave at Earth's bow shock nose. A pulse in NKE(t), nearly 5 times its median level, deposited energy into the Earth's magnetosphere. This produced a dramatic increase in geomagnetic activity, NAp(t). Meanwhile, the solar EUV as indicated by the N10.7 cm(t) flux was low and constant and is uncorrelated with broadband airglow brightness changes. The period of negative [B(t) z]GSM which followed Event B allowed energetic charged particles to penetrate deep into Earth's ionosphere. The shock wave triggered large variations in airglow brightness during the night reaching peak brightnesses near local midnight (Fig. 11). Of interest is the broadband airglow increase before the shock wave arrived. Perhaps, the shock wave triggered a release of stored magnetospheric energy in the near-Earth tail as has been demonstrated for geomagnetic events 29 . This plot defies a simple explanation.
The nightly evolution of Event B, initiated by a shock wave at 2,458,780.8333 JD, is shown in Fig. 11. The broadband 5 min cadence, time-series airglow measurements, ΔMC(t), above their quiescent levels are plotted versus the time in hours relative to local midnight for 6 nights in Fig. 11. Before Event B the space weather Figure 8. The 5 min cadence, broadband time-series airglow observations, ΔMC(t), at CCIDSS are plotted versus the time in hours relative to local midnight. Each ΔMC(t) point is the differential photometric broadband airglow brightness relative to the continuous light curve of quiescent airglow brightness (Fig. 1). This plot shows the shock wave Event A produced a broadband airglow intensity disturbance which lasted for several days. www.nature.com/scientificreports/  www.nature.com/scientificreports/ conditions were not steady (see Fig. 10). The Event B shock wave triggered a geomagnetic disturbance whose peak daily Ap(t) was more than 6 times it's median value. The Ap    Fig. 13. On the vertical axis, each point is the corrected nightly average of airglow brightness, ΔMCN(t) above its quiescent level. The celestial and anthropogenic sources have been removed as outlined in "Data collection and analysis" of this paper. The horizontal axis is the time in fractions of a year (F). The data were sorted into 36 bins with an F width of 0.0278. The smoothed data curve was obtained from the 36 data bins using a 5 point triangular weighting function. The smoothed, binned, data curve of Fig. 13 shows a semi-annual variation in airglow brightness with broad peaks near 0.273 F and 0.837 F. The peak near 0.837 F has an amplitude and location strongly influenced by high speed streams in the solar wind from coronal holes on the face of the Sun which impact the Earth's magnetosphere (Event A and Event B). The amplitude of the unperturbed broadband airglow semiannual variation during solar minimum is ~ 0.2 mag/arcsec 2 .   The (CCIDSS,CSSMLS) points in Fig. 15 track together more tightly than do the (CCIDSS, Stars 211) points. On a number of nights the broadband airglow at Stars 211 was significantly brighter than it was on the same night at CCIDSS. However, some of the very brightest night sky airglow nights at CCIDSS correspond to some of the very brightest airglow nights at both CSSMLS and Stars 211.

Discussion
There is no simple cause and effect relationship between solar activity, space weather, and changes in broadband night terrestrial airglow. However, there are coincidental relations between events on the sun, conditions in the near Earth environment, and the brightness of the night sky which need to be explored. This paper is based on detailed analysis of individual astronomically dark nights during a deep solar minimum. Taking a completely different approach, Alarcon et al. used statistical methods to analyze data from dozens of TESS-W photometers during solar minimum 9 . Many of their instruments are in locations with more than 0.5 mag/arcsec2 of human caused artificial illumination. They report short time-scale variations in the night sky airglow which they attribute to events in the mesosphere.
Variations in broadband night sky airglow are not always accompanied by changes in 10.7 cm solar flux (Fig. 5). To put this fact into prospective, relationships between the brightness of the natural night sky and solar activity as measured by the 10.7 cm radio flux are discussed by many authors. Krisciunas et al. obtained V-band sky brightness from Cerro Tololo Inter-American Observatory CCD images over the course of a solar cycle 12   www.nature.com/scientificreports/ Cycle 23. Patat reports a clear seasonal variation in the broadband VRI passbands with two broad maxima (April-May and October) and two broad minima (July-August and December-January). In the aeronomy literature, observations obtained over many years show there are semiannual oscillations in OI 558 nm and OH 730 nm 18,19 . TIMED/SABER observations show a global distribution of oscillations in OH night glow emission with semiannual, annual, and quasi-biennial time scales 50 . Our data show during a deep solar minimum there are semiannual broadband night airglow variations with maxima near the equinoxes (Fig. 13). Our observations cover only a 1.65 year long period. As a result the location of the maxima and minima in Fig. 13 are shifted by space weather events such as Event A and Event B. It should be emphasized; in every case the semiannual modulation in night sky airglow is derived from noisy data. The precise details of the semiannual variation likely require a long base line in time.
Is it plausible the semiannual variations in night sky air glow are driven by similar processes to those explained by the Russell and McPherron model for the statistical behavior of geomagnetic events? 26  6. The time development takes place over several nights. The peak brightness occurred several days after the triggering event. (Figs. 8 and 11). 7. There are substantial variations in broadband airglow during the night (Figs. 8, 10, and 11). 8. The large variations in broadband airglow during the night at various sites are apparently correlated (Fig. 10). 9. There are large amplitude geomagnetic events (Figs. 6 and 10). Fig. 8 shows a airglow brightness change of approximately 0.40 mag/arcsec 2 (1.44 times in intensity units). Figure 10 for Event B the airglow brightness also changes by more than 0.35 mag/arcsec 2 (1.38 times in intensity units). Alarcon et al. used statistical methods to analyze data from dozens of TESS-W photometers 9 . They report observing short time-scale variations on most nights which they attribute to airglow events in the mesosphere. They do not report on the details of any specific events.
The orange tint of the images presented in Figs. 7 and 9 suggest Na airglow may have been excited during Event A and Event B 37 .
For broadband night sky increase Event C: 1. There are relatively small broadband night airglow brightness increases over a wide geographic area (Fig. 12).  9 . They find a correlation between the physical separation of photometers and the standard deviation of the differences in their measurements. The data plotted in Fig. 15 show some very bright broadband airglow nights at CCIDSS are also very bright on the same night at both CSSMLS and Stars 211. Similarly some of the nights at CCIDSS during which the night sky broadband airglow was faintest were also, on the same night, among darkest nights at both CSSMLS and Stars 211. The observations plotted in Fig. 15 suggest the presence of broadband night sky airglow events of various dimensions or, perhaps, they demonstrate semiannual variations for stations at similar latitudes. The amplitude of semiannual broadband night sky airglow variations shown in Fig. 13 as well as those published in the literature are on the order of 0.2 mag/arcsec 2 . 9,15 The range of variations observed in Fig. 15 is much larger than these measured semiannual amplitudes of broadband night sky airglow. There are a number of nights which have bright broadband night airglow at Stars 211 and are simultaneously dim at CCIDSS. These nights do not seem to support the semiannual hypotheses. Even so, the reason for the trend displayed in Fig. 15 remains an open question.
Data presented in this paper show variations in broadband airglow with time-scales on the order of minutes, hours, days, and months. There are, undoubtedly, a number of interacting physical processes which cause what is observed. Localized aurorae outside the auroral oval have been observed and have an origin which is not completely understood 52 . We show during solar minimum the variations in broadband airglow are unrelated to changes in the 10.7 cm solar radio flux. Instead, they are likely caused by charged particles from the solar wind www.nature.com/scientificreports/ which enter Earth's magnetosphere. Observations by NASA's IMAGE spacecraft and the joint NASA/European Space Agency Cluster satellites show that huge cracks develop in the Earth's magnetosphere for hours allowing charged particles from the solar wind to enter the ionosphere 51 . Some of these cracks appear on a seasonal basis and others present themselves in a more random fashion. They are associated with magnetic reconnection of Earth's magnetic field lines with those in the interplanetary magnetic field. It is postulated that magnetic fields from the Sun and Earth reconnect on Earth's day side. From there the solar wind transports the reconnected magnetic flux to Earth's night side where it is stored in the magnetospheric tail. This stored energy can be released by a triggering event [53][54][55] . The data from Events A and B, qualitatively, match this scenario (see Figs. 6,8,10,and 11). Event C implies broadband night time airglow around the vernal equinox follows the Russell-McPherron prediction for geomagnetic activity (please see Figs. 12, 13, and 14). Broadband night sky airglow and geomagnetic activity are both likely responses to changes in space weather. Sounding of the atmosphere by the SABER instrument aboard the NASA TIMED satellite over the course of a solar cycle relate changes in thermospheric cooling with variations in solar ultraviolet irradiance and geomagnetic activity 56 . Solar ultraviolet irradiance and geomagnetic processes are, also, related to cooling of the thermosphere by infrared radiation from nitric oxide over the duration of a solar cycle 57 . Thirteen years of data from the SABER instrument find the intensities of four night glow emissions are strongly coupled to solar radiation 58 . In this paper, Events A and B show broadband night sky airglow increase events coincident with Earth interacting with an energetic stream in the solar wind. Event C documents observed night sky broadband airglow brightness variations coincident with the alignment of the z component of Earth's magnetic filed and the interplanetary magnetic field. Our results contribute to the continuing effort to unravel how solar-terrestrial interactions modulate night sky airglow emissions.

Conclusions
New observational data reveal changes in night time airglow which are significant in broadband astronomical and artificial light at night studies. SQM and TESS-W photometers sum all emissions in a broad cone to the edge of space and over a wide area of the sky. These characteristics make it difficult to identify the physical processes creating the emissions. However, these instruments produce time-series data which can be used to identify broadband airglow brightness events for further study.
During deep solar minimum the broadband night sky airglow is never constant in intensity. For our data set, there were nights when the SQM broadband airglow intensity at the natural night sky location CCIDSS, became as faint as 22.07 mag/arcsec 2 . On other nights the SQM broadband airglow was brighter than 21.57 mag/arcsec 2 .
We report, during solar minimum, significant episodes of increased night sky airglow are not produced by changes in 10.7 cm solar flux. We find these night sky brightening events are coincident with: 1. Changing orientation of the interplanetary magnetic field relative to Earth's magnetic field and 2. Earth entering streams of energetic solar wind.
It is plausible episodes of increased broadband night sky airglow we observe could be amplified by a release of energy stored in Earth's magnetospheric tail triggered by a shock wave in the solar wind.
Sites more than 8500 km along the Earth's surface experience nights in common with either very bright or very faint night sky airglow emissions. The reason for this observational fact remains an open question.
Our data suggests the terrestrial night airglow responds to the energy input into the Earth's magnetosphere in a fashion similar to the geomagnetic indices.
We strongly advocate the establishment of a global network of photometers located in places where anthropogenic skyglow is at a minimum. These instruments would be used to track brightness variations of the natural night sky. Established astronomical observatories are the places to start. These measurements will have a significant impact on the studies of astronomy, space weather, light pollution, biology, and recreation.