Earliest datable records of aurora-like phenomena in the astronomical diaries from Babylonia

The astronomical diaries from Babylonia (ADB) are excellent sources of information of natural phenomena, including astronomical ones, in pre-Christ era because it contains the record of highly continuous and systematic observations. In this article, we present results of a survey of aurora-like phenomena in ADB, spanning from BCE 652 to BCE 61. We have found nine records of aurora-like phenomena. Philological and scientific examinations suggest that five of them can be considered as likely candidate for aurora observations. They provide unique information about the solar and aurora activities in the first millennium BCE. Graphical abstract .


Introduction
The solar activity has been monitored by telescopic observations of sunspots for more than 400 years (Zolotova and Ponyavin 2015; Vaquero and Vázquez 2009;Owens 2013). Reconstructing the earlier solar activity has been of great interest from the viewpoints of the longterm variation of solar magnetism and its effect on the Earth climate (Haigh 2007;Miyahara et al. 2008;Hathaway 2010;Usoskin 2013). One common way to reconstruct the past solar activity is the analysis of cosmogenic radioisotopes such as carbon-14 content in tree rings and beryllium-10 in ice cores (Steinhilber et al. 2009).
Records in historical documents provide another way to investigate the solar activity in the pre-telescopic era (Willis et al. 1996;Vaquero and Vázquez 2009;Usoskin et al. 2007;Usoskin 2013;Hayakawa et al. 2016b, c, submitted). Eddy (1977bEddy ( , 1980 and Vaquero and Vázquez (2009) even claim that historical documents can offer us data with more precise date than radioactive isotopes.
Chinese official histories (Yau and Stephenson 1988;Yau et al. 1995;Xu et al. 2000;Hayakawa et al. 2015Hayakawa et al. , 2016dKawamura et al. 2016) and Korean official histories (Lee et al. 2004) are especially suitable source documents for the purpose of studying the long-term variation because these sources provide wellformatted records based on continuous observations by professional astronomers (Keimatsu 1976;Hayakawa et al. 2015).
Recently, the past solar activity is also attracting the interest from the viewpoint of extreme space weather (e.g. Tsurutani et al. 2003;Schrijver et al. 2012) particularly after the discoveries of "superflares" in solar-type stars (Maehara et al. 2012;Shibayama et al. 2013;Maehara et al. 2015) as well as the sharp increase in cosmogenic radioisotopes in tree rings around CE775 and CE994 (Miyake et al. , 2013Mekhaldi et al. 2015), for which one of possible causes are suggested to be large solar flares. The discoveries of "superflares" in solar-type stars posed the question whether such extremely intense solar flares and space weather events can occur in our Sun (e.g. Shibata et al. 2013;Aulanier et al. 2013).
The most intense solar flare throughout the history of telescopic observation is believed to be so-called Carrington event (Carrington 1859; Tsurutani et al. 2003;Cliver and Dietrich 2013). It was associated with worldwide aurora observations by amateur observers, even at low geomagnetic latitudes (Loomis 1859(Loomis , 1860a(Loomis , c, d, e, 1861a(Loomis , b, 1865Kimball 1960;Nakazawa et al. 2004;Green and Boardsen 2006;Hayakawa et al. 2016b). A sharp spike of nitrate in polar ice cores has also been found associated with this event (McCracken et al. 2001), although the usage of nitrate as an index of solar flares is debated (Wolff et al. 2012). Searches for the historical records corresponding to the cosmic ray event in CE775 and CE994 (Miyake et al. , 2013 have been carried out by several groups Chapman et al. 2015;Stephenson 2015;Hayakawa et al. 2016a, e, submitted).
From the viewpoint of the history of science, it is also interesting to investigate how far back we can trace the solar and space weather events using historical records. Previously, the earliest record was thought to be a Chinese record from BCE 193 (Yau et al. 1995). However, Stephenson and Willis (2002) as well as Stephenson et al. (2004) found a much earlier record from BCE 567 within the astronomical diaries from Babylonia (ADB) to be the oldest observation of aurora in the world and the sole aurora observation in ADB.
In this article, we made further careful examinations of ADB in order to search for the potential records of an aurora and hence to obtain insights into the aurora observations and records in pre-Christ era that were not previously well known due to the shortage of historical documents.

Astronomical diaries from Babylonia
ADB is a series of Akkadian cuneiform texts inscribed on clay tablets. Most of the tablets were rediscovered in excavations at the site of Babylon in the late nineteenth century and now preserved in the British Museum, London. Contemporarily, each text of these tablets was titled "regular observation [ (EN.NUN) šá gi-né-e]". These records were mostly written and compiled by the families of astronomer-astrologers ( Enūma Anu Enlil) sponsored by the assembly of Esangil, the temple of god Marduk in Babylon (N32°33′, E44°26′) (Mitsuma 2012(Mitsuma , 2015. The scholars compiled the diaries from generation to generation, at least from the mid-seventh to the mid-first centuries BC. Towards the mid-third century BC, they fixed the format of the diaries, especially of the so-called Longer Diaries or Standard Diaries. Each tablet of the standard diaries covers half a year, i.e. 6 or 7 months if an intercalary month is included, or a third of a year, i.e. 4 or 5 months including an intercalary month. Horizontal rulings divide each tablet into four, five, six, or seven sections, according to its coverage. Each section covers a month, and its entries are arranged into subjects in the following order: daily report of the sky, price list of commodities, summary of the positions of the visible planets, level of the Euphrates, and unusual historical event(s).
In this article, we mainly examine the daily sky reports of ADB. They use consistent terminology and have a consistent set of criteria for the choice of what should be recorded. Their observations were carried out continuously as the original title of the diary tablet suggests. The astronomer-astrologers inserted a passage "I did not watch (NU PAP)", when they could not make their observations (Sachs and Hunger 1988;Mitsuma 2009).
Importantly, we have only fragments of the original ADB series, so our analysis is limited. Although the earliest known tablet in the ADB dates from BCE 652 and the latest up to BCE 61, we have not unearthed or reconstructed all tablets for every month. We have at most 5-10% of the estimated original complete ADB series (Stephenson et al. 2004; see also Sachs 1974). Future reconstruction and deciphering of undated clay tablets could improve this situation at least partially.
We first surveyed the critical editions of ADB published by Sachs and Hunger (1988, 1989, 1996 and by Hunger and van der Spek (2006). They edited the transliterations of Akkadian texts from all the ADB tablets whose dates became clear by 2006. We then examined the colour photographs recently taken by one of the authors (Y Mitsuma) at the British Museum, and copies made by EF Weidner and TG Pinches.

Date conversion
The Babylonian calendar was a lunisolar calendar. One year consists of 12 lunar months ( ) or 13 lunar months including an intercalary month. Each day starts at sunset, as is the case in the Bible (Parker and Dubberstein 1956, 26;Sachs and Hunger 1988, 15), each month starts when a new crescent moon is observed, and each year starts at the month around the vernal equinox. The Julian dates of the beginnings of Babylonian months covered by the diaries were calculated by Parker and Dubberstein (1956) and by Sachs and Hunger (1988, 1989, 1996, except for those in BCE651, the oldest diary known to us. Note that these conversion tables only cover up to BCE626. We follow the procedure by Sachs and Hunger to convert Babylonian dates in this article. Based on the converted dates, we computed normalized lunar ages according to the algorithm of Kawamura et al. (2016) that employs the interactive data language (IDL) programs in the astronomy user IDL library of NASA/Goddard Space Flight Center (Landsman 1993) and our program developed for numerically determining the minimum of lunar luminosity.

Result
In total, we found nine candidates of aurora observations recorded in the ADB. Those we found are shown with their ID numbers, keywords, references, dates in Julian calendar, normalized lunar ages, dates in Babylonian calendar, transliterations of Akkadian texts, and English translations. The notation "−n" such as −651 in every entry is the text number of the diaries published by Sachs and Hunger (1988, 1989, 1996 and Hunger and van der Spek (2006). In the section of Babylonian date, "n/n−1 BC" is used to show the Julian equivalent to a Babylonian year. Babylonian months are shown with roman numerals. For the convention used to indicate a part of a diary, see Sachs and Hunger (1988, 36-38).

Red glow: akukūtu
The record #2 is the well-known record from Stephenson et al. (2004), where the term was introduced and examined philologically and scientifically. The term akukūtu means "flame, blaze" or "red glow in the sky (as a rare meteorological phenomenon)" (CAD: I-1, p. 285). It is the second meaning that appears in this sentence. CAD (The Assyrian Dictionary of the Oriental Institute of the University of Chicago) relates this term to aurora as well.
Stephenson et al. (2004) made a positive interpretation of #2 because its last signs-2 DANNA-"2 double hours" indicate the duration of the phenomenon, and there is no other light source that can supply red light for as long as 4 h. Their conclusion can be reinforced with the flat topography in Babylonia. Babylon is located in the alluvial plain caused by the rivers of Euphrates and Tigris, in an area with no mountains until the Zagros Mountains about 180 km away in the north-east. Its weather is very dry, allowing few trees to grow. This means there is nowhere that a mountain fire or long-lasting light can be caused on the ground.

Unusual rainbow: TIR.AN (manzât)
(d) TIR.AN(.NA) (manzât) of #3 and #8 usually means as "rainbow" or "a name of star" or sometimes as "halo" with a subsequent description such as "surrounding the sun/moon" (CAD, X-1, pp. 230-232). #1 shows TIR. It seems to be an abbreviated form of TIR.AN and is followed by the words ma-diš SA 5 "very red", as is suggested by the comment of Sachs and Hunger (1988, p. 46) to #1. manzât was used with the verb GIB (parāku) "stretch" in #1 and #3, and the same verb may be restored in #8. These "rainbows" appear "stretched" in the sky with a red colour (#1, #8) or at night (#3, #8). Since these appearances do not match the nature of normal rainbows, manzât in #1, #3, and #8 seem to have been something different.
The unusual "rainbow" or TIR.AN in #3 was observed "before sunrise". The lunar age of this event is about 0.3 with the moon at about the waxing quarter, and the moon had already set at this time "before sunrise". Therefore, moonbow or any other moon-related atmospheric optical phenomenon cannot explain this event. Considering that this event was seen in the northern direction, which is usual for low-latitude auroras, it is reasonable to leave this record as the candidate for an aurora observation.
The term d TIR.AN.NA (manzât) in #8 is recorded after the "setting of the moon". The exact observational date is broken in the original tablet. However, this record is located between the record for BCE 123/122. II. 5 and that for BCE 123/122. II. 8 and thus can be located on II. 6/7, namely on 28/29 or 29/30 April BCE 123. Its moon phase is 0.227 or 0.261, and thus, the moon was approximately in its waxing quarter. This means "after moonset" was equivalent to just after midnight. This red "rainbow" may be regarded as an auroral arc elongated "from north to south" across the sky.
The unusual "rainbow" or TIR in #1 was also red in colour. Unfortunately, the Julian date of this record (BCE 652/651. XII. 28) was not calculated by Sachs and Hunger (1988) nor by Parker and Dubberstein (1956). Despite its abnormality, this record is very unlikely to represent aurora due to the observation time being "in the afternoon (KIN.SIG)". Red rainbows can be observed especially at the time of twilight when red-coloured light from the sun gets refracted by waterdrops. The "rain" immediately before this phenomenon might have supplied the waterdrops to cause rainbows in the quite dry weather in Babylonia.
Nevertheless, we should note two historical reports for luminescence phenomena observed in the day time in a huge magnetic storm, namely in the Carrington event. Loomis (1860b) cited a letter from Lieut. N. Home to describe aurora observation at Halifax on 28 August 1859. At 17:00, "a long narrow belt of cloud from E. to W. having a peculiar orange-white appearance" was seen, and at 20:00 "this cloud suddenly became luminous at its eastern extremity". This cloud is unlikely to have been aurora in the day time, but we cannot rule out the possibility that it was aurora.
Meanwhile, "unusual rainbow" can sometimes be related to bow-shaped auroras worldwide. A Norwegian drawing for aurora on 26 November 1710 was drawn like a rainbow in Ramus (1745), and Chinese historians reported some auroras as "white/unusual rainbow" in their official histories as highlighted by Hayakawa et al. (2016a).

Redness: sūmu
The term sūmu used in #5 and #6 means "redness, red glow" or "red spot". While the term in the latter meaning is used for red spots on the body, the term in the former meaning is used in astronomical contexts (CAD: XV, pp. 381-383). The sūmu of #5 and #6 was probably celestial events, although they appear not in reports of celestial events but in reports of historical events. Those observations of "redness" might have been considered as omens for historical events and recorded with them. Another celestial event, for example, lightning strike (miqitti išāti), is often inserted into the historical parts of the diaries and seems to have an ominous significance (Pirngruber 2013).
These terms of "redness" are used almost in the same expression with the signs GAR.GAR-an, which represents the Akkadian verb ittaškan, Ntn, i.e. passive habitative iterative (Caplice and Snell 2002: pp. 51-52), preterite of šakānu "to place". This indicates that the "redness" of #5 and #6 appeared repeatedly.
These philological analyses show us that "redness" appeared repeatedly in the same month in both cases. This phrase may describe low-latitude auroras whose movement is quite slow and which thus seem to remain in the same place for a long time. During the Halloween storm (29-31 October 2003), aurora observations were made for 3 successive nights (Shiokawa et al., 2005). Alternatively, this phrase is likely to describe a stable red auroral (SAR) arc that can be seen as red, faint lights. Data from optical imaging on board Dynamics Explorer 1 show a SAR arc lasting for 28 h (Craven et al., 1982). In fact, auroras can appear repeatedly for several days when complex active regions on the sun continue to launch multiple coronal mass ejections. At the Carrington event, the strong magnetic storms brought a cluster of auroras to the earth from 1859/08/28 to 09/04 (Loomis 1859(Loomis , 1860a(Loomis , c, d, e, 1861a(Loomis , b, 1865Kimball 1960;Green and Boardsen 2006;Hayakawa et al. 2016b). In September 1770, a series of aurora observation were made in East Asia for at least three days straight (Willis et al. 1996;Kawamura et al. 2016). These records (both #5 and #6) suggest the possibility of strong solar activity in BCE 145-144. In summary, although it is slightly mysterious that they appeared in the east and west, these records are not contradictory with aurora records.
Phenomenon of record #9 has an extension from east to west. The problem is to identify if this is with motion. Its verb "SUR ( )" means "to light up", "to flare up", or "to display a sudden luminosity" (CAD: XVI, p. 100). This verb itself does not provide an answer to our question whether this recorded phenomenon appeared without motion. There is a possibility that this is aurora. Bright aurora looks white or greenish white because it consists of emissions from O (green colour) and from N 2 and N 2+ (red and blue colours). After auroral breakups, bright aurora expands towards the north and west. The leading edge of the bright aurora is called a westward travelling surge. The westward travelling surge moves westward, but it is unclear whether such an elongated structure is expressed as a torch. Thus, this reminds us of a fireball.
As for record #7, its subject is lost. This makes any detailed discussion difficult. However, something like "a torch" appeared with "UD.DA ( , bright light)". The word frequently refers the light of the sun or the moon (CAD: XVI,. Therefore, strong brightness, at least, seems to be associated with this recorded phenomenon. According to the date written in the clay tablet, its date is 10/11 or 11/12 November BCE 137. In case if it is observed at daytime, we can hardly regard this event as aurora. In case if it is observed at night on 11/12, we have still a little possibility to regard this event as aurora. We have examples for low-latitude aurora, which becomes partially brighter (Shiokawa et al. 2005). Thus, we cannot exclude the possibility of aurora for this record.

Likeliness of every record
Based on the analysis above, we can rate these auroralike records as follows (Table 1). #2 has no reason not to be related to aurora as Stephenson et al. (2004) stated. The same can be stated for the "redness" in #5 and #6. They appeared repeatedly and remind us of clustering aurora caused by huge geomagnetic storms like those at the Carrington event. The unusual "rainbows" in #3 and #8 are not unlike aurora, since these rainbows are observed at night. Despite the same terminology, #1 is unlikely to be aurora due to its daytime observation. #4 and #9 seem like fireballs or meteors from the descriptions, and we do not have any positive intention to relate them with aurora. #7 could be aurora as the term dipāru can be used for description of , a variant of akukūtu, which can mean aurora, though its subject is lost and we cannot make a clear conclusion on this.
Korte and Stolze (2014) estimated the location of the aurora zone for past 10,000 years. During the period spanning from BCE 652 to BCE 61, Babylon was situated near the centre of the two boundaries where aurora was visible on the horizon at the geomagnetic activity levels Kp = 4 and Kp = 9. According to the statistical study by Remick and Love (2006), the mean wait times between successive events are 7.12, 16.55, 42.22, and 121.40 days for Kp ≥ 5, 6, 7, and 8, respectively. Thus, it is quite likely that aurora was frequently visible at Babylon somewhere between the horizon and zenith.

Comparison with long-term solar activity
In order to map the aurora-like records shown above with long-term solar activity, we compared the dates of these records with proxy-based solar activity levels. Figure 4 shows a comparison of the aurora-like records with the solar activity level anomaly reconstructed using multiple proxies of radioactive isotopes by Steinhilber et al. (2009) at a resolution of 20-30 years. However, it should also be noted that this comparison between characteristics of the solar activity at different timescales is hard and complex, as on one hand auroras occur at daily scale, and on the other hand, the reconstruction of solar activity by Steinhilber et al. (2009) has a decadal time resolution (see Vaquero et al. 2002 for the similar problems in case of naked-eye observations of sunspots). This figure shows us that most of the aurora-like records, except for #3 in BC 385 in solar minimum (Eddy 1977a, b;Usoskin et al. 2007;Nagaya et al. 2012), are located in phases of high solar activity. The cases #5-#7 are clustered in BCE 140-130 s, and #8-#9 are around BCE 120. These peaks are comparable with contemporary aurora records from other regions. For the former peak, red vapour was recorded in BCE 139 in Chinese official history Hànshū (Yau et al. 1995), and reports of sky fire in BCE 147 and night sun in BCE 134 in Western classics (Stothers 1979). For the latter peak, we have reports of milk rain in BCE 124 and BCE 118 in Western classics (Stothers 1979). Although we do not have exact simultaneous observation of aurora, these clustering records support the peaks at both BCE 140 s-130 s and around BCE 120 shown in the ADB.

Conclusion
We surveyed aurora-like records in ADB spanning from BCE 652 to BCE 61. We found 9 records of aurora-like phenomena including one mentioned in Stephenson et al. (2004) and examined them to evaluate 5 likely, 3 unlikely, and 1 possible aurora event. The main characteristics of these events are provided in this article. Our result is quite consistent with the long-term solar activity level, although comparison between characteristics of the solar activity at different timescales has difficulties as stated above. This article rewrites the history of aurora observation in this early period with dates to provide clear insights for the era before Christ. This article also examines proxies at a high resolution in this early period in order to consider long-term solar activity and rare extreme space weather events. A lot of unpublished clay tablets of the ADB are preserved in the British Museum. Although most of them are fragmentary, study of those tablets will provide us further astronomical information from the mid-seventh to the mid-first centuries BCE, possibly including further records of aurora-like phenomena.
Abbreviations BM: tablets in the collections of the British Museum; VAT: tablets in the collections of the Staatliche Museen, Berlin.

Fig. 4
Aurora-like records compared with total solar irradiance anomaly. Black curve total solar irradiance anomaly from Steinhilber et al. (2009). Black vertical lines likely aurora records from the ADB. Gray vertical lines unlikely aurora records from the ADB. Black dashed vertical lines possible aurora records from the ADB