Sporadic aurorae observed in East Asia

Abstract. All the accessible auroral observations recorded in Chinese and Japanese histories during the interval AD 1840–1911 are investigated in detail. Most of these auroral records have never been translated into a Western language before. The East Asian auroral reports provide information on the date and approximate location of each auroral observation, together with limited scientific information on the characteristics of the auroral luminosity such as colour, duration, extent, position in the sky and approximate time of occurrence. The full translations of the original Chinese and Japanese auroral records are presented in an appendix, which contains bibliographic details of the various historical sources. (There are no known reliable Korean observations during this interval.) A second appendix discusses a few implausible "auroral" records, which have been rejected. The salient scientific properties of all exactly dated and reliable East Asian auroral observations in the interval AD 1840–1911 are summarised succinctly. By comparing the relevant scientific information on exactly dated auroral observations with the lists of great geomagnetic storms compiled by the Royal Greenwich Observatory, and also the tabulated values of the Ak (Helsinki) and aa (Greenwich and Melbourne) magnetic indices, it is found that 5 of the great geomagnetic storms (aa>150 or Ak>50) during either the second half of the nineteenth century or the first decade of the twentieth century are clearly identified by extensive auroral displays observed in China or Japan. Indeed, two of these great storms produced auroral displays observed in both countries on the same night. Conversely, at least 29 (69%) of the 42 Chinese and Japanese auroral observations occurred at times of weak-to-moderate geomagnetic activity (aa or Ak≤50). It is shown that these latter auroral displays are very similar to the more numerous (about 50) examples of sporadic aurorae observed in the United States during the interval AD 1880–1940. The localised nature and spatial structure of some sporadic aurorae observed in East Asia is indicated by the use of descriptive terms such as "lightning", "rainbow", "streak" and "grid".


Introduction
The occurrence of the aurora borealis or aurora australis at relatively low geomagnetic latitudes during periods of weakto-moderate geomagnetic activity has been reported in the literature on rare occasions.A salient feature of these reports is that the aurora was observed at geomagnetic latitudes that would conventionally imply the concurrent existence of a great geomagnetic storm.Botley (1963) drew attention to this phenomenon and used the term "sporadic aurora", which she adopted from the earlier paper by Abbe (1895).In particular, Botley (1963) defined sporadic aurorae as "comprising such instances as a single ray in a sky otherwise seemingly clear of auroral light, or isolated patches well to the equatorial side of a great display".Moreover, she noted a casual mention by Fritz (1881) of observations in low geomagnetic latitudes when there was apparently little incidence of the aurora in high latitudes.Within the context of auroral physics, the word "sporadic" often has the connotation of a "spatially isolated" region of the night sky.Therefore, the term "sporadic aurorae" may be used to describe localised auroral displays that occur at low geomagnetic latitudes (approximately, −45 •  ≤ geomagnetic latitude ≤+45 • ) during intervals of weak-to-moderate geomagnetic activity (aa≤50).The converse situation ("non-sporadic aurorae") is the more familiar one in which extensive auroral displays at low geomagnetic latitudes occur at the times of great geomagnetic D. M. Willis et al.: Sporadic aurorae observed in East Asia storms and hence during intervals of strong geomagnetic activity (aa>150).Silverman (2003) stated that he had originally discounted the early reports of sporadic aurorae but had been forced to re-evaluate the situation after he found a number of similar records in a compilation of data from the United States.Indeed, Silverman (2003) provided the first proper scientific study of the existence of sporadic (or localised) auroral observations at relatively low latitudes, during periods of weak-to-moderate geomagnetic activity, by tabulating some 54 cases of such sporadic auroral occurrence.Apart from one observation from Grahamstown, South Africa, all his examples of sporadic aurorae are from the United States and occurred during the interval AD 1880-1940.The sporadic auroral observations presented by Silverman (2003) were taken primarily from a database he compiled that is based largely on the series entitled "Climatological Data of the United States".Some additional observations were taken from other sources, principally the "Monthly Weather Review".
As noted by Silverman (2003), the only two previous papers that specifically discuss sporadic aurorae have reported a sufficiently small number of cases (Abbe, 1895, cites two and Botley, 1963, cites nine) to justify initial doubt about the reality of the phenomenon.From the historical viewpoint, the limited early evidence for sporadic aurorae may be summarised briefly as follows.Eddie (1894) presented an isolated example of such an event observed in the Southern Hemisphere, from Grahamstown, South Africa (geographic co-ordinates 33 • 19 S, 26 • 32 E), on 1894 July 9 during a period of moderate geomagnetic activity (half-daily values of the aa index during the interval July 7-10: 6, 9, 32, 37, 25, 26, 8, 9).Abbe (1895) published a description of an aurora on 1895 August 26, which was observed from both Charleston, South Carolina (geographic co-ordinates 32 • 48 N, 79 • 58 W), and Greendale, Kentucky (geographic co-ordinates 38 • 06 N, 84 • 32 W), during a period of weak geomagnetic activity (half-daily values of the aa index during the interval August 24-27: 5, 8, 3, 4, 15, 9, 4, 5).Silverman (2003) quotes part of the text of the record from Grahamstown and the full text of the record from Charleston (which are both discussed in Sect.3.2).Although Botley (1963) purportedly described nine such cases of sporadic aurorae, several of which may be classified as "doubtful" by the criteria presented by Silverman (2003), most of these cases actually occurred at times of strong geomagnetic activity.
The main purpose of this paper is to point out that some of the auroral observations recorded in Chinese and Japanese histories provide evidence for the occurrence of sporadic aurorae in another region of the world.To achieve this goal, every accessible and exactly dated (year, month and day all known precisely) oriental auroral observation that lies within the interval AD 1840-1911 has been extracted from the database of East Asian auroral observations discussed by Willis et al. (2005).However, three minor changes have been made to this database for the purposes of the present investi-gation; these changes are defined in Sect. 2. The first exactly dated and reliable East Asian auroral observation in this time interval was recorded in China on 1844 April 21 and the last was recorded in Japan on 1909 September 25 and 26 (the former day being the date of a great geomagnetic storm).The present study is deliberately restricted to the interval 1840-1911, so that direct comparisons can be made with the lists of great and small geomagnetic storms , compiled by the Royal Greenwich Observatory (1955), as well as with the magnetic indices Ak (1844-1867) and aa (1868 -present), discussed by Nevanlinna and Kataja (1993) and Mayaud (1980), respectively.Prior to 1840, there are no proper quantitative measurements of geomagnetic activity that can be utilised in direct comparisons with the historical East Asian auroral observations, compilations of which do not extend beyond 1911.After 1911, lists of East Asian auroral sightings do not begin again until about the time of the International Geophysical Year (1957).
Therefore, the present investigation augments the study of sporadic auroral occurrence in the United States (Silverman, 2003) by identifying the occurrence of the same phenomenon in China and Japan.(There appear to be no reliable Korean auroral observations during the selected interval.)The approach adopted here differs slightly from that employed by Silverman (2003), at least to the extent that East Asian auroral observations coincident with great geomagnetic storms are not automatically eliminated from the investigation.Such observations are retained because they confirm the general reliability of the East Asian auroral observations , which might otherwise be considered to be intrinsically less reliable than the more recent auroral observations from the United States .Hence the approach adopted here is similar to that of Botley (1963), who presented auroral observations acquired at times of both strong and weak geomagnetic activity.

The East Asian auroral observations
The 1198 entries originally included in the auroral database compiled by Willis et al. (2005) are largely derived from the catalogue of auroral observations from China, Korea and Japan (193 BC-AD 1770) published by Yau et al. (1995).The material from this catalogue has been supplemented with additional data from several sources, notably: (1) Chinese auroral records included in the catalogue published (in Chinese) by the Beijing Observatory (1988), which extends the Chinese auroral observations back to at least 210 BC and up to AD 1911; (2) the list of Japanese auroral records since AD 1600 included in the catalogue published (in Japanese) by Osaki (1994);and (3) a research paper by Matsushita (1956) on ancient aurorae seen in Japan.Three minor amendments have been made to the original auroral database compiled by Willis et al. (2005).First, the supposed Japanese "auroral" observation on AD 1850 August 15 has now been rejected because the quoted double hour shen (15:00-17:00 LT) corresponds to late afternoon, which is long before sunset in August.Second, the supposed Chinese "auroral" observation on AD 1885 December 26 has now been rejected because the description of the phenomenon is obscure and does not necessarily correspond to an auroral display (despite the fact that it is included in the list compiled by the Beijing Observatory, 1988).Third, it has subsequently become clear that the extensive Japanese auroral observations on AD 1872 February 4 were followed by similar observations (also in Japan) on February 6.All the East Asian auroral observations within the interval AD 1840-1911 have been carefully translated into English, with appropriate help from experts in Classical Chinese and Japanese.The translations of the relevant records are presented in Appendix A. The reasons for rejecting the "auroral" observations on AD 1850 August 15 and AD 1885 December 26 are discussed in slightly more detail in Appendix B, which also includes a detailed discussion of the reasons for rejecting a potential Korean "auroral" observation on AD 1843 March 12.
It should be noted that the auroral catalogue published by the Beijing Observatory ( 1988) is divided into two parts: the second part is entitled "Not Definite Category".Although not strictly germane to this paper, a similar statement applies to the sunspot catalogue published by the Beijing Observatory (1988).It appears that the compilers of the Beijing catalogues have put entries in the "not definite category" if: (i) the date is very general; (ii) the content of the record is unclear; or (iii) the record is cited indirectly from some unavailable or unidentifiable source.Moreover, in the specific case of auroral observations (ji guang), entries seem to be included in the "not definite category" if there is any uncertainty whatsoever about the identification of an auroral display; for example, reference to a "black vapour like a rainbow", "glittering lights -like lightning" or a "dim light in the sky like a rainbow".In this paper, all Chinese auroral records extracted from the "not definite category" are identified by placing an asterisk after a chronological reference number (e.g.Ref. No. 01*), which is introduced later in this section.
The combined auroral catalogue that results from all these sources currently contains 1154 entries in the interval 210 BC-AD 1911 (although there are no exactly dated auroral observations in 1910 and 1911).However, the associated auroral database currently contains 1197 entries.The discrepancy between the number of entries in the catalogue and database arises partly from the fact that independent auroral observations (e.g. from different countries) are always included as separate entries in the auroral database and partly from the avoidance of auroral date ranges.In particular, auroral observations on consecutive days, which are sometimes included as a single entry in an auroral catalogue, are always included as separate entries in the auroral database.The paper by Willis et al. (2005) provides further details on the auroral database.As in this earlier paper, the present study is restricted to auroral records from East Asia for which the year, month and day are all specified exactly.This restriction is essential if meaningful comparisons are to be made with daily values of the Ak or aa index and also the dates on which great and small geomagnetic storms occurred.Therefore, imprecise records, with inexact dates for which no more than the lunar month is specified, are automatically excluded from this investigation (cf.Willis et al., 2005).
Table 1 lists all the accessible and exactly dated Chinese and Japanese auroral observations that occurred during the interval 1840-1911 (apart from the two rejected records -see Appendix B).To facilitate cross-referencing between the tables and the text, a chronological reference number (Ref.No.) is provided in the first column of each entry in Table 1.Subsequent columns give the exact Gregorian date (year, month, day) of each auroral observation, as well as details of the place at which each observation was made, namely the country, the city, district or province/prefecture (with the present-day capital in parentheses), and the associated geographic and magnetic coordinates.All of the Chinese auroral observations presented in Table 1 and Appendix A have been gleaned from local histories or gazetteers.In most cases, only the provinces in which the observations were made have been recorded (see Appendix A).However, as investigation of reports of other celestial events -such as eclipses and meteors -clearly reveals, local histories are concerned with specific provinces.Duplication between histories of separate provinces is extremely rare.Hence it is appropriate to cite the geographic and magnetic co-ordinates of the relevant provincial capital -see Table 1.Although the provincial Chinese auroral observations may have been made at unknown locations outside the present-day capitals, the differences should amount to no more than a very few degrees in both latitude and longitude.Such differences are insignificant in this study.For the Japanese auroral observations, both specific locations and prefectures are quoted; the geographic and magnetic coordinates listed in Table 1 refer to the present-day capitals of the prefectures.
If an auroral display was observed from two or more locations in the same country on the same night (e.g.Shanghai Municipality & Zhejiang Province (China) in 1866; Hamada Town & Shimane Prefecture (Japan) in 1872), the co-ordinates listed apply to the first location specified in column 4 of Table 1.However, in all cases the co-ordinates of the other locations are cited either elsewhere in Table 1 or in the accompanying footnotes.
The geographic co-ordinates of all locations have been extracted from the Index in "The Times Comprehensive Atlas of the World" (1999); moreover, the spelling of place names follows that used in this index (apart from Seoul).It is clear from the entries in Table 1 (including the footnotes) that the geographic latitudes of the locations at which aurorae were observed in East Asia (i.e. the locations of the observers) during the interval    1 are presented in Appendix A, which provides a complete historical reference for all but one of these records.Most of these auroral records (some of which are rather terse and stereotyped by modern scientific standards) have not previously been translated into English or any other Western language.However, the importance of the descriptions of the oriental displays is enhanced by the fact that the great majority of the records of sporadic auroral events listed by Silverman (2003) merely report an observation on a particular date at a known location.The records included in the series entitled "Climatological Data of the United States" do not provide descriptions, although some descriptions may still exist in various national data archives in the United States.In Sect.3.3, the descriptions of sporadic aurorae observed in East Asia are compared with the two descriptions of sporadic aurorae presented by Silverman (2003); namely the one from Grahamstown, South Africa, and the one from Charleston, South Carolina.In general, however, the information recorded in East Asian histories is inadequate to determine the height of the auroral display above the horizon (or even whether or not it was overhead).

Scientific interpretation of the observations
Table 2 contains a succinct summary of the scientific content of each of the East Asian auroral records listed in Table 1.For completeness, the date of the oriental observation is repeated in Table 2 but the country of origin is now denoted simply by a capital letter in parentheses (C = China, J = Japan).The time of occurrence, colour(s), compass position(s) and spatial extent(s) of each auroral observation have been extracted from the translations of the East Asian records presented in Appendix A. Table 2 also indicates whether or not ("Yes", "No") a great geomagnetic storm was in progress at the time of each auroral observation, according to the lists of geomagnetic storms (both great and small) published by the Royal Greenwich Observatory (1955).In addition, daily values of the Ak (Helsinki) magnetic index (before 1868) and the aa magnetic index (from 1868) are presented for a sevenday interval centred on the day (date) of the East Asian auroral observation (Mayaud, 1973(Mayaud, , 1980;;Nevanlinna and Ketola, 1993;Nevanlinna et al., 1993;Nevanlinna and Kataja, 1993).To achieve consistency between the numerical accuracies of the tabulated Ak and aa indices, daily values of the aa index have been rounded to the nearest integer.The final column of Table 2 (entitled "Comments") serves to emphasise the dates of both great and small geomagnetic storms and also those occasions when there was clearly defined spatial structure in the auroral display, as inferred from the description in the auroral record (see Appendix A).For example, several records suggest structure by the use of terms such as "lightning" (dian), "rainbow" (hong), "streak" (dao) and "grid" (jing).(In one instance, the term "rainbow" is abbreviated to R in Table 2.) Detailed lists of both great and small geomagnetic storms (based on Greenwich data up to the end of 1925 and Abinger data since the beginning of 1926) are available for the interval 1874-1954 and a list of great geomagnetic storms (based on Greenwich data) is available for the earlier interval 1840-1874(Royal Greenwich Observatory, 1955).The lower limit of range for a great geomagnetic storm at Greenwich (or Abinger) was adopted as 60 in declination (D) or 300 nT (γ ) in either the horizontal magnetic field (H ) or the vertical magnetic field (Z).For small geomagnetic storms, the lower limit for inclusion is 30 in D or 150 nT in H or in Z.
In the published tables, the ranges are given to the nearest minute of arc in D and the nearest 5 nT in H and Z. (For completeness, it should be noted that the same limits apply to the period from 1916 to 1924 inclusive, when the north component of the geomagnetic field was recorded instead of the horizontal component.) It is clear from Table 2 that about 25% of the East Asian auroral records refer to observations that occurred when a great geomagnetic storm was in progress (and hence invariably occurred under conditions of strong geomagnetic activity), whereas about 75% of the records refer to auroral observations that occurred when no great geomagnetic storm was in progress (and hence almost always occurred under conditions of weak-to-moderate geomagnetic activity).As noted in the Introduction, auroral observations coincident with great geomagnetic storms are retained in this study because they confirm the general reliability of the East Asian auroral observations in the interval 1840-1911.The oriental auroral observations are discussed separately in the two following subsections (3.1 and 3.2) for conditions of strong and weak-to-moderate geomagnetic activity.The two subsequent subsections (3.3 and 3.4) discuss the evidence for spatial structure in the descriptions of sporadic aurorae and the reliability of the East Asian auroral observations.1.For brevity, the country of origin of each observation is now denoted simply by a capital letter in parentheses (C = China, J = Japan).The scientific content of each oriental auroral observation is summarised briefly, with specific reference (whenever possible) to the time of occurrence, colour(s), compass position(s) and spatial extent (see Appendix A).The presence ("Yes"), or absence ("No"), of a great storm (RGO) implies inclusion in, or exclusion from, the lists of great geomagnetic storms compiled by the Royal Greenwich Observatory (1955).Daily values of the Ak index (before AD 1868) or the aa index (from AD 1868) are presented for a 7-day interval centred on the day of the oriental auroral observation (values of the aa index have been rounded to the nearest integer).In the "Comments" column, explicit reference is made to both great (Gt) and small (Sm) geomagnetic storms and also the existence of spatial structure in the auroral display, as inferred from the description recorded in the oriental history or gazetteer (see Appendix A).Definitions: All quoted times and time intervals (e.g.Yin [04:00-05:00]) are expressed in local time (LT); N, NE, E, SE, S, SW, W, NW denote points of the compass; Z denotes the zenith; ↑ indicates "upwards" and → indicates "extension" (usually from one compass point to another); the abbreviations Bl = Blue, Gr = Green, Pu = Purple, Re = Red, Wh = White, Ye = Yellow are used to specify the colours of auroral displays; a Ref.No. 31 also states "within the cloud there was one star like a bowl; its colour was blue"; ND signifies that no data (Ak) exist for that particular day; R = Rainbow (which necessarily implies "Structure" when used alone for brevity, as in the case of the Chinese auroral observation on AD 1908 July 12).

East Asian auroral observations during great geomagnetic storms
The entries in the final column of  and 42).More extensive low-latitude auroral observations associated with this great geomagnetic storm have been compiled and discussed by Silverman (1995).The daily values of the aa index for the seven-day interval centred on September 25 are as follows: 15, 10, 6, 329, 23, 16, 26.Conversely, the small geomagnetic storm on 1908 July 15.6-17 is most unlikely to be associated with the Chinese auroral observation on July 12 (Ref.No. 40) because in this case the auroral observation preceded the geomagnetic storm by three days.
Of course, there is no exact one-to-one correspondence between the great geomagnetic storms included in the lists compiled by the Royal Greenwich Observatory (1955) and auroral observations recorded in East Asia.Many potential auroral observations must have been missed as a result of extensive cloud cover (or other adverse viewing conditions), whereas the Royal Greenwich Observatory (RGO) lists of geomagnetic storms are based on magnetic measurements that are largely unaffected by the concurrent meteorological conditions.In particular, the RGO lists include 101 great geomagnetic storms in the interval 1840-1911 but only 5 of these storms appear in Table 2.The likely effects of cloud cover over East Asia can be estimated approximately by examining the global cloud cover maps that have been constructed using data acquired during the International Satellite Cloud Climatology Project (ISCCP).Since 1983, as part of ISCCP, visible and infrared images from polar orbiting and geostationary meteorological satellites have been processed to produce monthly data sets of global cloud cover and radiative properties (see, for example, http://www.gewex.org/isccp.html).The maps of mean annual cloud amount indicate that the frequency of occurrence of cloudy conditions over East Asia in the interval 1983 (July) -2005 (June) lies in the approximate range 50-80%.Therefore, cloud cover can probably account for the fact that some of the great geomagnetic storms recorded at Greenwich in the interval 1840-1911 were not accompanied by auroral observations in East Asia.However, this provisional estimate of cloud cover over East Asia requires considerable further refinement within the context of auroral physics.For example, extensive auroral displays can be seen in the gaps between clouds (which may be less than the pixel size of the satellite images) and it also seems most unlikely that such displays would be completely obscured by thin cirrus cloud (which implies that cloud type is also an important factor).Despite these "meteorological" complications, the oriental auroral observations that are available correctly identify some of the really great geomagnetic storms that occurred during the second half of the nine-teenth century and the first decade of the twentieth century.
An extremely important feature of the East Asian auroral observations is that brief descriptions of the auroral displays have been recorded in the various histories.Although these descriptions are usually tantalisingly terse and somewhat stereotyped if judged by modern scientific standards, as noted in Sect. 2 they often provide some basic physical information on the structure of auroral displays observed under conditions of both strong and weak-to-moderate geomagnetic activity.Many of the auroral observations listed in Table 2 merely state that "the sky was red", or refer to a "red light" or a "red vapour" in the sky.Sometimes the colour "red" is qualified by such commonplace terms as "blood" or "fire".As is also clear from Table 2, the reference to a red sky is used for auroral displays that occur under conditions of both strong and weak-to-moderate geomagnetic activity.However, in the case of great geomagnetic storms, there is usually some mention of the large extent of the sky covered by the "red light" or "red vapour", often by reference to more than one point of the compass (Ref.Nos. 13,14,25,[29][30][31][32]. More detailed information is available for the great geomagnetic storm on 1853 July 12-13 (Ref.No. 06*), which is associated with one of the Chinese auroral observations somewhat surprisingly included in the "not definite category" (Beijing Observatory, 1988).For this particular storm, the description (see Appendix A) refers to "two streaks of white vapour"; the centre being "blue-green" while the lower part was "purple".One streak was "long" and the other was "short".The description also states that the streaks had "the form of a roll of silk gauze reaching straight to the zenith" and notes further that the aurora extended "from the S to the N, gradually twisting to the W and disappearing".The level of detail provided in this particular auroral description is essentially comparable with that available from many contemporaneous European and North American auroral observations (e.g.Fritz, 1873;Loomis, 1859Loomis, -1861;;Silverman, 2006).
Although the descriptions of East Asian auroral observations during great geomagnetic storms usually indicate that the auroral display covered a large region of the sky, the descriptions of several of the East Asian sporadic auroral observations during weak-to-moderate geomagnetic activity suggest auroral structure over a more localised region of the sky, as discussed in the following subsection.

East Asian observations of sporadic aurorae
The detailed entries in Table 2 indicate that many of the East Asian auroral observations in the interval 1840-1911 occurred under conditions of weak-to-moderate geomagnetic activity.This conclusion is confirmed by relatively small values of the magnetic indices, Ak or aa, defined by Nevanlinna and Kataja (1993) and Mayaud (1980), respectively, or by the absence of a great geomagnetic storm ("No" in column 4 of Table 2) according to the lists of great geomagnetic storms compiled by the Royal Greenwich Observatory (1955).No values of the magnetic index Ak are available (ND) for three of the 42 auroral observations listed in Table 2 (i.e.Ref. Nos. 01*, 09 and 10).It then follows from column 4 of this table that 29 (=74%) of the 39 East Asian auroral observations for which magnetic indices are available occurred at times when no great geomagnetic storm was in progress.Moreover, the daily magnetic index, Ak or aa, was ≤50 throughout a five-day interval centred on the date of each of these 29 observations (including the auroral observation that occurred three days before the small geomagnetic storm on 1908 July 15.6-17).For a centred seven-day interval (as used in Sect.3.1), the daily value of Ak or aa slightly exceeded 50 for just two of the 203 daily observations (Ak=56 for Ref.No. 20* and Ak=55 for Ref.No. 21).On both occasions, however, this excess occurred three days before the day (night) of the auroral observation (i.e. on Day -3) and represented the nearest local maximum in the value of Ak immediately before the date of the auroral observation.For 12 (=31%) of the 39 auroral observations for which daily magnetic indices are available, Ak or aa≤25 throughout a sevenday interval centred on the date of the auroral observation (Ref.Nos.03*, 07*, 08, 12, 16, 17, 18, 19*, 23, 28, 37, 38).For comparison, 10 (=26%) of the 39 auroral observations (for which magnetic indices are available) occurred at times when a great geomagnetic storm was in progress, although these particular auroral observations correspond to just 5 distinct geomagnetic storms.
For the Chinese auroral observation on 1859 August 4 (Ref.No. 12), Ak≤5 throughout the seven-day interval centred on this date.Although it would be natural to question the reliability of any low-latitude auroral observation under such extremely weak, or quiet, geomagnetic conditions, it should be noted that the auroral observation on 1859 August 4 occurred 29 days before the commencement of the great geomagnetic storm on 1859 September 2-5.Actually, dual storms commenced on August 28 and September 2, both of which were accompanied by brilliant auroral displays seen from many parts of the world (Royal Greenwich Observatory, 1955;Silverman, 2006).Therefore, the auroral observation on August 4 occurred about one (mean) synodic-solarrotation period (27 days) before the dual storms on August 28 and September 2. The implication of this time sequence of events is that recurrent solar activity probably existed at this time, despite the apparently quiet geomagnetic conditions.In this same context, it should be noted that the It should be emphasised that some caution must be exercised when making comparisons between the magnetic index Ak (Nevanlinna and Kataja, 1993), which is based entirely on magnetic measurements at Helsinki, and unaidedeye auroral observations recorded in Chinese and Japanese histories.The difference in local time (LT) between Northern Europe and East Asia (ranging from about 5-9 h) may be significant for all but the most intense geomagnetic storms.Indeed, owing to the local-time variation of magnetic disturbances at Helsinki, the daily values of the Ak index cannot be compared with the daily values of the aa index (Nevanlinna and Kataja, 1993).In addition, as noted in Sect.3.1, the Ak index probably underestimates the true strength of even the great geomagnetic storm on 1859 September 2. Obviously, a similar cautionary remark applies to comparisons between magnetic measurements at Greenwich and auroral observations in East Asia.The magnetic index aa (Mayaud, 1980), which is based on magnetic measurements at one observatory in England and one in Australia, is perhaps marginally better, if only because the magnetic observatory in Australia (Melbourne; geographic co-ordinates 37 • 49 S, 144 • 59 E) is at approximately the same geographic longitude (and hence LT) as the locations of the East Asian auroral observations (see Table 1).Nevertheless, the magnetic indices Ak and aa are not necessarily ideal for identifying magnetic disturbances that might be associated with sporadic aurorae in East Asia.In a similar vein, Kamide ( 2006) has noted that it is misleading to use the D st magnetic index to define how much the entire solar-terrestrial system is disturbed.

Evidence for spatial structure in descriptions of sporadic aurorae
Several of the descriptions of auroral observations included in Table 2 , 12, 17, 19*, 27, 28, 33, 35, 36, 37), as in the case of great geomagnetic storms (Sect.3.1).In this context, it should perhaps just be mentioned that the Chinese auroral observation on 1862 July 29 (Ref.No. 17) occurred six days before the great geomagnetic storm of August 4-5.Since different types of aurorae appear to occur under weak-to-moderate geomagnetic activity, it is difficult to offer a unique physical explanation.However, the use of the terms "lightning", "rainbow" or "streak" in some East Asian auroral records may signify a single ray of light in a sky otherwise seemingly clear of auroral luminosity, in accord with the definition of sporadic aurorae propounded by Botley (1963) and quoted by Silverman (2003).Moreover, the term "rainbow" probably implies both shape and colour and, as intimated previously, there is some evidence that the term "lightning" is used to emphasise the flickering nature of the light (Ref.No. 20*).
The detailed descriptions of the East Asian auroral observations presented in Appendix A, which are summarised succinctly in Table 2 (column 3), show some similarities to the descriptions of sporadic aurorae quoted by Silverman (2003).For example, the description of the aurora australis observed on 1894 July 9 (Eddie, 1894), from Grahamstown, South Africa, states that the display lasted from 21:15 LT to 23:15 LT (i.e. 2 h).Despite bright moonlight (Moon at first quarter), this auroral display "manifested itself in a pale greenish white light extending around the south-eastern and southern horizon, and by broad streamers of a lovely roseate hue, which fitfully ebbed and flowed as they extended upwards from the south-east, at times right up to the zenith . . ." Likewise, the aurora borealis observed on 1895 August 26 (Abbe, 1895), from Charleston, South Carolina, lasted from 22:47 LT to 22:58 LT (i.e. about 10 min) and extended approximately from the NW to the N.As noted in the description, "a dark segment or arch rose to 17 • , with a crown of light to 24 • , one large streamer lasting from 22:51 LT to 22:53 LT (i.e. 2 min); azimuth 116 • to 122 • (measured in the sense S→W→N→E), and altitude 30 • ."This record also states "the display was accompanied by a thin veil of vapour, through which the stars could be seen."No detailed description is available for the display of the aurora borealis observed on 1876 May 20 (Fritz, 1873;Botley, 1963), from Canea, Crete (Chania, Kriti; geographic co-ordinates 35 • 30 N, 24 • 02 E).
It is evident from the preceding descriptions that sporadic aurorae can last from several minutes to several hours.Although none of the East Asian auroral observations unambiguously implies a time interval as short as a few minutes, some of these observations suggest that the auroral display lasts for about one or two hours (Ref.Nos.01*, 04*, 05,25,26,36,40).The descriptions of the auroral displays observed from Charleston and Grahamstown both refer to "streamers".Although the observation from Grahamstown actually refers to "broad streamers", there is a clear intimation of elongated rays (or streamers) in both cases.More specifically, the observation from Charleston suggests that one large streamer extended 6 • in azimuth and about 30 • in altitude, whereas the observation from Grahamstown suggests that the broad streamers extended over several multiples of 10 • in altitude (since at times they reached up to the zenith).This is in general agreement with the use of the terms "lightning", "rainbow" and "streak" in the East Asian auroral observations.Two of the East Asian auroral observations mention explicitly that the displays were either several tens of zhang (i.e.very many degrees) long, or several tens of zhang long and several chi (degrees) wide (see Ref. Nos. 07* and 16 in Appendix A).
None of the dates of the auroral observations from East Asia  in the absence of a great geomagnetic storm ("No" in column 4 of Table 2) coincides with the date of a sporadic aurora observed in the United States , as tabulated by Silverman (2003; see his Tables 1 and  2).A similar statement applies if the relevant East Asian auroral observations are compared with the sporadic aurorae observed from Canea, Crete, on AD 1876 May 20 (Botley, 1963) and from Grahamstown, South Africa, on 1894 July 9 (Eddie, 1894).The lack of any coincidences between East Asian auroral observations in the absence of a great geomagnetic storm and sporadic aurorae observed elsewhere in the world provides some, albeit limited, evidence for the truly localised nature of such auroral displays, although it should be emphasised that the East Asian observations were by no means systematic.

Reliability of the East Asian auroral observations
In the preceding discussion, it has been tacitly assumed that the East Asian auroral records listed in Tables 1 and 2 and presented in Appendix A are all equally reliable observations of the aurora borealis.Clearly, this assumption is not necessarily true.For example, the Japanese auroral observation on 1845 March 22 (Ref.No. 02) refers to a "white vapour" without stating explicitly that the observation was made during the night.In principle, it is at least possible that such a "white vapour" could have been seen during the daytime.Similarly, the detailed record of the Chinese auroral observation on 1862 January 1 (Ref.No. 16) does not state explicitly that the phenomenon was observed during the night, although judging from the description such an assumption seems entirely reasonable in this particular case.Likewise, the record of the Chinese auroral observation on 1863 February 18 (Ref.No. 19*) does not state explicitly that the phenomenon occurred at night, although the reference to a "red vapour" strongly supports the interpretation of the phenomenon as an auroral display.
Conversely, the record of the Chinese auroral observation on 1853 April 23 (Ref.No. 05) states: "At the third watch of the night [23:00-01:00 LT], the sky was red like blood."In this instance, there is specific reference to the auroral display occurring at a particular time during the night, while the colour "red" is the one most commonly associated with low-latitude aurorae.It is known from modern observations that low-latitude aurorae are predominantly "red" in colour because they are typically characterised by: (i) N + 2 1N (427.8 nm) emission at high vibrational levels and (ii) a high (>10) ratio of red (630.0nm) to green (557.7 nm) atomic oxygen lines.An aurora of the second type sometimes produces intense visible emissions, which result in a spectacular red aurora in the night sky (Tinsley et al., 1984;Davis, 1992;Shiokawa et al., 2005).Other historical auroral records from East Asia are slightly more ambiguous.The record of the Chinese auroral observation on 1853 August 12 (Ref.No. 07*) indicates that the phenomenon was observed at night but states enigmatically: "there was a black (hei) vapour like a rainbow; it extended from the west to the east, several tens of zhang (i.e.very many degrees) in length."Regrettably, this Chinese record contains inadequate information for a completely reliable scientific interpretation, although it is tempting to speculate that this might be a genuine historical observation of a "black" aurora.A "black" aurora is a lack of auroral emission in a small, well-defined region (e.g. a beanshaped pod or a thin arc-like feature) within an otherwise uniform, diffuse background or within an aurora exhibiting a type of behaviour intermediate between that of "diffuse" aurora and "discrete" aurora (Davis, 1978(Davis, , 1992;;Newell et al., 2001).
In this context, however, it should be noted that all of the auroral observations which are clearly associated with great geomagnetic storms (Ref. Nos. 06*,13,14,25,(29)(30)(31)(32)41 and 42) state explicitly that the observation was made during the "night" and all but two of these observations (Ref.Nos.41 and 42) refer specifically either to the colour "red" or to a colour that is neither "white" nor "black" (see Ref. No. 06*).The Chinese record of the auroral observation on 1908 July 12 (Ref.No. 40), which occurred three days before a small (probably independent) geomagnetic storm (Royal Greenwich Observatory, 1955), notes that at the first watch of the night (19:45-21:30 LT) a rainbow extended from east to west and that after a while it began to disappear.This description is quite compatible with the occurrence of an auroral display and the use of the term "rainbow" suggests that the spatial extent of the auroral display is limited, which is consistent with the occurrence of a sporadic auroral observation.
The Chinese record for 1863 January 4 (Ref.No. 18) states that "multicoloured clouds" appeared in the middle of the night and there seems little doubt that this is a reference to the aurora borealis.The Chinese record for 1869 July 25 (Ref. No. 23) refers to a "blue-green rainbow" extending across half the sky (E to W) in the evening, which could also be a genuine observation of the aurora borealis.The Chinese record for 1908 July 12 (Ref. No. 40) notes that at the first watch (19:45-21:30 LT) a "rainbow" extended from east to west, which could again be a genuine observation of the aurora borealis (three days before the small geomagnetic storm on July 15.6-17).If these three additional Chinese records are accepted as being authentic observations of the aurora borealis not associated with a great geomagnetic storm, then 21 (=50%) of the 42 auroral records from East Asia in the interval 1840-1911 refer to sporadic auroral observations.Conversely, only 10 (=24%) of the 42 auroral observations from East Asia in this interval are definitely associated with great geomagnetic storms.
The six Chinese observations on 1844 April 21, 1853 August 12, 1862July 29, 1866August 27, 1872November 30 and 1900 January 30 (Ref. Nos. 01*, 07*, 17, 20*, 34* and 38) and the Japanese observation on 1855 August 23 (Ref.No. 08) either refer explicitly to the "dusk", "evening", "night" or else imply a night-time observation by use of the phrase "shone like daylight" (Ref. No. 38).Moreover, the wording used to describe each of these seven observations is largely compatible with the occurrence of an auroral display.The Japanese record for 1868 July 6 (Ref.No. 22) indicates that at midnight two arcs extended across the sky from east to west (Matsushita, 1956), which could be a genuine observation of the aurora borealis, despite the fact that Matsushita classified this observation as being "unlikely".Future identification of the original text may eventually clarify this matter (see Appendix A).If these eight additional observations are regarded as being authentic observations of the aurora borealis that are not associated with a great geomagnetic storm, then 29 (=69%) of the 42 auroral records from East Asia in the interval 1840-1911 refer to sporadic auroral observations.
It is now necessary to consider the three remaining observations.The Japanese observation on 1845 March 22 (Ref.No. 02) merely refers to a white vapour seen in the zenith (i.e. in mid-sky) at an unspecified time.The Chinese observation on 1862 January 1 (Ref.No. 16) refers to a split in the sky several tens of zhang (i.e.very many deg) in length and several chi (deg) in width and also glittering lights that eventually conjoined.This observation is not specifically stated to have occurred during the night, as noted previously, although in this particular case it is a reasonable inference.The Chinese observation on 1863 February 18 (Ref.No. 19*) simply states that a red vapour appeared at the SE.As also noted previously, there is again no mention of night but reference to the colour "red" suggests an auroral display.If these three further observations are provisionally accepted as genuine observations of the aurora borealis that are not associated with a great geomagnetic storm, then 32 (=76%) of the 42 auroral records from East Asia in the interval 1840-1911 apparently refer to sporadic auroral observations.Finally, a comment should perhaps be made on the relative merits of the East Asian auroral observations discussed in this paper and the observations from (mainly) the United States discussed in the paper by Silverman (2003).The available Chinese records of atmospheric phenomena during the period in question are virtually all from local histories.Unlike the reports in the astronomical treatises of the official histories, they are thus not based on systematic observation and their inclusion in local histories will to some extent reflect the interests of the compilers.However, it should be emphasised that the goal of this paper is to investigate aurorae that were observed in East Asia.This is a unique region with an unparalleled history of auroral (and other) observations extending back over two millennia.Moreover, as noted previously, although the auroral observations from East Asia  might initially be considered to be intrinsically less reliable than the more recent auroral observations from the United States (1880-1940), they have the added advantage that descriptions of the auroral displays are immediately available (see Appendix A), contrary to the situation for the great majority of the American observations of sporadic aurorae (Silverman, 2003).

Discussion and conclusions
The main purpose of this paper is to point out that a significant number of the auroral observations recorded in Chinese and Japanese histories during the interval AD 1840-1911 provide evidence for the occurrence of so-called "sporadic aurorae" in East Asia.This investigation therefore extends the work of Silverman (2003), who tabulated about 54 cases of such sporadic auroral occurrence from elsewhere in the world.Apart from one observation (Eddie, 1894) from Grahamstown, South Africa, all his examples of sporadic auroral occurrence are from the United States.The salient features of "sporadic" auroral observations are that they occur at rela-tively low latitudes (approximately, −45 • ≤ geomagnetic latitude ≤+45 • ) during periods of weak-to-moderate geomagnetic activity (aa or Ak≤50).Moreover, as observed from the surface of the Earth, they sometimes appear to be spatially localised in the night sky.
Table 1 lists all the accessible, exactly dated and reliable East Asian auroral observations in the interval AD 1840-1911, which have been extracted from the database of oriental auroral observations discussed by Willis et al. (2005).This table provides the Gregorian date of each auroral observation, the country, city, district or province/prefecture of the place at which the observation was made, as well as the geographic and magnetic co-ordinates of the place of observation.The translations of the oriental auroral records are presented in Appendix A and the reasons for rejecting three implausible "auroral" records are discussed in Appendix B. Table 2 summarises the scientific attributes of the exactly dated East Asian auroral observations, as inferred from the translated records, and indicates whether or not ("Yes", "No") a great geomagnetic storm was recorded at Greenwich on the same date.In addition, daily values of the Ak index (before AD 1868) or the aa index (from AD 1868) are presented in Table 2 for a 7-day interval centred on the day of the oriental auroral observation.
Using the information presented in Table 2, it is possible to separate the exactly dated and reliable East Asian auroral observations in the interval AD 1840-1911 into two categories.The first category comprises those auroral observations that occurred in the presence of great geomagnetic storms ("Yes" in column 4) and hence occurred at times of strong geomagnetic activity (aa>150 or Ak>50 for at least one of the seven entries in column 5).The second category comprises those auroral observations that occurred in the absence of great geomagnetic storms ("No" in column 4) and at times of weakto-moderate geomagnetic activity (aa or Ak≤50 for at least the central five entries in column 5).Observations in the second category are examples of sporadic aurorae recorded in East Asian histories.Moreover, the spatially localised nature of some of the oriental observations of sporadic aurorae is indicated by the use of a variety of descriptive terms (e.g."lightning", "rainbow", "streak" or "grid").
There are two tentative reasons for believing that sporadic aurorae may be associated with solar activity, despite the fact that they occur under weak-to-moderate levels of geomagnetic activity.First, as noted in Sect.3.2, there is some evidence within the interval AD 1840-1911 for recurrent, though possibly intermittent, auroral activity on a timescale almost exactly equal to the mean synodic-solar-rotation period.There is similar evidence for recurrent auroral activity at many other times during the past two millennia.For example, there is strong evidence for recurrent (but intermittent) auroral activity between about the middle of AD 1127 and the middle of AD 1129 (Willis and Stephenson, 2001).Second, Willis et al. (2006) have drawn attention to the presence of large sunspots near the central solar meridian at the times of modern Japanese auroral observations.Although sunspot observations are not readily available throughout the interval 1840-1911, a provisional study of the recorded sunspot measurements from Kew and Greenwich suggests that there was at least one sunspot that was both large enough to be seen with the unaided eye and reasonably close to the central solar meridian at the times of some of the East Asian auroral observations in the interval 1840-1911 (Tables 1 and 2).Therefore, the level of solar activity may possibly have been high at these particular times, even if the auroral observations are apparently associated with weak-to-moderate levels of geomagnetic activity.A detailed study of the complex association between solar, geomagnetic and auroral activity, particularly at such early times when the available information is quite primitive by modern standards, is beyond the intended scope of the present paper.
However, it is hoped that similar investigations of the auroral observations recorded by members of the Aurora Section of the British Astronomical Association will reveal more recent examples of sporadic aurorae (albeit at higher geomagnetic latitudes) that have occurred at times when detailed geomagnetic, interplanetary and solar data are available.Moreover, such auroral observations have the added advantage that they are directly comparable with the East Asian auroral observations, at least in the sense that both sets of observations were made without the aid of specialised instrumentation.
In conclusion, evidence has been found in Chinese and Japanese histories for the occurrence of sporadic aurorae in East Asia during the interval 1840-1911.Special attention has been paid to the careful translation of the oriental historical records into English, in order to extract the relevant scientific information.In particular, the recorded descriptions of the auroral displays suggest that several of the East Asian examples of sporadic aurorae are characterised by spatial structure.Silverman (2003) discussed the occurrence of sporadic auroral observations at relatively low magnetic latitudes in the United States at various times in the interval 1880-1940.However, there is a notable difference between his results and those obtained in the present investigation.Because of the configuration of the geomagnetic field during the nineteenth and twentieth centuries, magnetic latitudes are less than geographic latitudes in East Asia (Table 1), whereas magnetic latitudes are greater than geographic latitudes in the United States (Silverman, 2003).Consequently, the lowest corrected geomagnetic latitude at which the aurora was seen in East Asia, namely about 16 • (Guangdong Province, China,  in 1863, 1871, 1872, 1897 and 1899), is very much closer to the estimated threshold magnetic latitude of approximately 15 • for a visual auroral event (Silverman, 1995) than are the corrected geomagnetic latitudes of the stations recording sporadic auroral observations in the United States (Silverman, 2003).Further research is required to explain how sporadic aurorae can occur at such low magnetic latitudes at times that apparently correspond to weak-to-moderate levels of geomagnetic activity.a Japanese auroral observation recorded in a scientific journal, the title of the journal (in italics) and the specific volume number and year are given in parentheses (e.g.(Tenmon geppo, vol. 26, 1933)).Following the customary style of writing oriental book and journal titles, a capital letter is used only in the first word of the title (apart from place names).
In a general context, it should be noted that two of the Japanese records on 1872 February 4 (Ref.No. 30) use the terms "polar light" (kyoku hikari) and "northern light" (kita hikari) to describe the auroral display.Therefore, it seems likely that by this date some Japanese observers had an understanding of the aurora, as described in the contemporary literature from the Western world, and used what are now regarded as fairly standard terms.[Japan, Tōkyō] Meiji reign period, 1st year, 5th month, 17th day.The authors have not yet been able to obtain the original text for this particular observation, despite several searches by helpful Japanese colleagues.However, Matsushita (1956) includes this observation in his list of ancient aurorae seen in Japan.From the information presented in his (ii) [Japan, Shimane Prefecture] Author's note to the previous record: "It seemed that the NE sky was as red as fire not only for one night.Moreover, although the districts [subsequently] affected by the earthquake disaster seemed to be on fire, the people living there did not see the phenomenon.Thus the people around Tsuwano in Kinoashi County thought there was a fire in the direction of Takatsu in Mino County, whereas people in Masuda thought it was in the direction of Hamada and people in Ōmori thought it was in the direction of Kiduki.Around Taisha [the Grand Shrine], it was thought that there was a great fire on Oki [islands]."(Kochibe Tadyoshi: The earthquake disaster at Iwami in the 5th year of the Meiji reign period) (iii) [Japan, Shimane Prefecture] "Last year, the 12th month, 28th day, in the N sky, a fire pillar (or fire pillars) was seen.This was the festival at the end of the year.The farmers from Imohara village who were gathered in my house were shocked and returned home in haste.They later told me that flames were visible in the N direction from their place."(Report by Shimada and his colleagues) (iv) [Japan, Shizuoka Prefecture] "Tanaka Masao . . . in the 4th year of the Meiji reign period, 28th day . . .travelled from Kōbe on the special steamer service 'Keier'.When he reached the coast at Enshu at midnight the N sky was bright and red."(Imamura Akitune: The Hamada earthquake in the 5th year of the Meiji reign period) [N.B.
Enshu is equivalent to the present-day western part of the Shizuoka Prefecture.In this particular case, the location of the auroral observation is taken to be Hamamatsu (34 Chinese auroral observations on 1853 February 27 (Ref.No. 04*) and 1853 April 23 (Ref.No. 05) occurred 135 (=5×27) days and 80 (=3×27-1) days, respectively, before the Chinese auroral observation on 1853 July 12 (Ref.No. 06*), which was the date of another great geomagnetic storm (1853 July 12-13).Likewise, it should be noted that the Chinese auroral observation on 1853 August 12 (Ref.No. 07*) occurred 30/31 days after this great geomagnetic storm.On all three days (apart from 1853 July 12), Ak≤31 throughout a seven-day interval cen-tred on each of these dates.These Chinese auroral observations provide evidence for recurrent, albeit intermittent, auroral activity, on a timescale almost exactly equal to the mean synodic-solar-rotation period, before the great geomagnetic storm on 1853 July 12-13.Extensive cloud cover may have prevented a complete sequence of auroral observations separated by approximately 27 days.Finally, the Chinese auroral observations on 1872 January 10 and 11 (Ref.Nos.27 and 28) occurred 25 days before the Chinese and Japanese auroral observations (Ref.Nos.29-32) associated with the great geomagnetic storm on 1872 February 4-5.

Table 1 .
Exactly dated Chinese and Japanese auroral observations recorded during the interval AD 1840-1911.After a chronological reference number (Ref.No.), introduced to facilitate cross-referencing, the exact Gregorian date of each auroral observation is given, followed by the country and then the city, district or province/prefecture of the place of observation (see Appendix A for further details).

Table 2 .
Scientific attributes of exactly dated East Asian auroral observations in the interval AD 1840-1911.The chronological reference number (Ref.No.) and associated date is exactly the same as in Table

Table 2
follows: 51, 19, 19, 21, 14, 5, 3; the corresponding values on the two previous days -February 11 and February 12 -are 50 and 158, respectively.Similarly, one of the great geomagnetic storms of the first decade of the twentieth century, namely that on 1909 September 25.4, is identified by auroral observations from Japan on the nights of both September 25 and 26(Ref.Nos.41 wich Observatory, 1955).Moreover, the second of these two storms is associated with auroral observations in China on the nights of February 4 and 5(Ref.Nos.29 and 31)and auroral observations in Japan on the nights of February 4 and 6 (Ref.Nos. 30 and 32).The daily values of the Ak (Helsinki) The values of the Ak index for the seven-day interval centred on July 12 are as follows: 6, 12, 33, 51, 53, 11, 12.It is also just possible that the great geomagnetic storm of 1871 February 11-13 is associated with an auroral observation from China on February 16 (Ref.No. 25).The values of the aa index for the seven-day interval centred on February www.ann-geophys.net/25/417/2007/Ann.Geophys., 25, 417-436, 2007 16 are as

Table A1 .
The Chinese sexagenary cycle (with the day names written in hanyu pinyin).

The auroral observations on 1859 September 2 (Ref. Nos. 13 and 14)
There was a split in the sky several tens of zhang (i.e.very many degrees) in length and several chi (degrees) in width; there were glittering lights, which eventually conjoined."(Xiushanxian zhi, 3)[N.B.Not specifically stated to have occurred at night, although probably obvious in this case.] The auroral observation on 1862 January 1 (Ref.No. 16) [China, Sichuan Province] Xianfeng reign period, 11th year, 12th month, day yimao (52)."

Table 1 ,
it is www.ann-geophys.net/25/417/2007/Ann.Geophys., 25, 417-436, 2007possible to construct the following pseudo-translation: "At midnight, two arcs [extended across the sky] from E to W." Moreover, this auroral display was observed from Tōkyōthe Japanese capital at the time.Although Matsushita classifies this observation as being "unlikely" on a fivefold classification system ranging from "certain" to "unlikely", it is retained in this study because the description suggests an auroral display rather than some other astronomical or meteorological phenomenon.Japan, Hamada (Shimane Prefecture)] Meiji reign period, 4th year, 12 month, 26th day."Atnight in the N direction it was like fire.In the N direction from above the sea to the sky a fire pillar stood, according to eyewitness(es).Night by night it became dimmer and moved slowly to and fro and finally disappeared.I presume that this was the polar light."(Annals of Hamada town) The well-known geologist Kochibe Tadyoshi obtained two separate accounts from different people: Fujii Muneo (dated 12th month, 26th day) and Masuda Reizo (dated 12th month, 28th day).These, and his own comments, Kochibe inserted in his book entitled: The earthquake disaster at Iwami in the 5th year of the Meiji reign period.These accounts are cited in (iv) and (v).(iv) [Japan, Shimane Prefecture] "Meiji reign period, 4th year, 12 month, 26th day (author's note: below written -i.e. in a further source -as the 28th day).At night in the N direction it was red like fire . .." (Abstract of written note by the old man Fujii Muneo: quoted by Kochibe Tadyoshi: The earthquake disaster at Iwami in the 5th year of the Meiji reign period) [N.B.The auroral description ends after the word "fire" and continues on a separate subject.Kochibe Tadyoshi also cited further accounts from the 12th month, 28th day -see 1872 February 6.] (v) [Japan, Shimane Prefecture?] Meiji reign period, 4th year, 12 month, 26th day."In the year before the earthquake, 12th month, 26th day, at night in the NW sky there appeared a red coloured light, so that I could count the number of trees in the mountain forest.I presume that it was the northern light."(ImamuraAkitune: The Hamada earthquake in the 5th year of the Meiji reign period) [N.B.The specific term "northern light", not "polar light", is used here.asthe SW; light shone like fire.Within the cloud (yun) there was one star like a bowl; its colour was blue.After a while it began to disappear."(Jing xian zhi, 14) Japan, Shimane Prefecture] Meiji reign period, 4th year, 12th month, 28th day."At night I was in the office on night duty.At around the 2nd hour of the morning, someone shouted that there was a fire.When I got up and looked around, I saw that due E the sky was bright red like a distant fire.It did not disappear until the sky started to brighten towards the dawn."(Report by Masuda Reizo; quoted by Kochibe Tadyoshi: The earthquake disaster at Iwami in the 5th year of the Meiji reign period) [N.B.The way the time of night is expressed is precisely the way in which "2 a.m." is expressed today.] "At night in the fourth watch [01:20-03:50 LT], red light filled the sky, as if the Sun had risen.[It lasted] from the [double] hour chou [01:00-03:00 LT] to the [double] hour yin [03:00-05:00 LT].After this time it gradually dispersed."(Huairen xian xinzhi, 1) ] The auroral observation on 1872 February 5 (Ref.No. 31) [China, Hubei Province] Tongzhi reign period, 10th year, 12th month, 27th day."At the [double] hour chou [01:00-03:00 LT], red vapour filled the sky, rising in the NE as far Ann.Geophys., 25, 417-436, 2007 www.ann-geophys.net/25/417/2007/