Challenges of Using Historical Aurora Observations for the Reconstruction of Solar Activity before the 19th Century, Especially during and near the Maunder Minimum

Abstract

In order to complement gaps in the surveillance of solar activity in historical times, various proxies are used to reconstruct past solar cycles and long-term maxima and minima of solar activity, the most famous being the Maunder Minimum (MM), which is usually defined to span the period between the years 1645 and 1715. We explain the problems within existing data bases and call upon trying to find the original sources of Schröder, since his aurorae catalog spans the whole MM and contradicts what has been deduced from more used compilations. We take a critical look at the proposed source-critical scheme introduced by Neuhäuser and Neuhäuser and show it to be counterproductive because it largely ignores the source situation, i.e., the scientific understanding of the reporters of times long past and their intentions. While historical sunspot and aurora reports can be useful to fine-tune our knowledge of solar activity in times before the onset of systematical surveillances, they should not be used as an index of solar activity, since they cannot be quantitatively expressed due to the non-scientific manner of the reports and ambiguous wording. Reconstructions based on cosmogenic isotopes are significantly preferable for establishing the level of solar activity in the past. The conclusions reached by this review should be regarded as a caution against expecting important conclusions to emerge from low quality data.

1. Introduction

Today’s state of research assumes that few aurorae occurred during a period that is known as the Maunder Minimum (MM) of solar activity. Amongst various proxies that are used to reconstruct solar activity in the past, aurorae play a central role. Schröder [1] compiled a catalog that spans the whole MM, while Neuhäuser and Neuhäuser [2] proposed a source-critical scheme that puts into question the value of aurora catalogs in general, which leads to the conclusion that sunspot records and reconstructions based on cosmogenic isotopes are to be preferred for establishing the level of solar activity in the past. In addition to direct measures of solar activity, there are those indirect indicators, notably auroral frequency, but also geomagnetic measurements and cosmic ray variability as deduced from cosmogenic isotopes recorded in natural archives like tree rings, sediments or glacier ice, since the modulation of galactic cosmic rays in the heliosphere depends on solar activity. During low solar activity, more galactic cosmic rays reach the Earth’s atmosphere, thereby leading to a higher production rate of cosmogenic isotopes, like Carbon-14 and Beryllium-10, which are most often used in this type of analyses. Carbon-14 and Beryllium-10 are excellent proxies to identify lower average solar activity, because they correlate with low production of those isotopes in the atmosphere. (e.g., Usoskin et al. [3], Beer et al. [4]).

The frequency and emission intensity of auroral displays has been used in many studies with the scope of reconstructing the strength of solar cycles long past, starting with Schove’s 1955 benchmark article [5]. Reports of aurorae from earlier times are, however, just as fragmentary as sightings of naked eye sunspots. Contrary to what has been suggested by different authors for years (e.g., Wang and Siscoe [6], Usoskin [7], Vaquero and Vazquez [8]), there was clearly no systematic surveillance of either sunspots or aurorae before the 18th and 19th centuries (e.g., Stangl and Foelsche [9], Usoskin et al. [10]). While in Eastern Asia, notably in China and Korea, where celestial phenomena have been monitored more carefully than in Europe, even there we cannot expect anything comparable to modern standards in order to compare auroral frequencies with actual values or those of the 19th and 20th centuries. However, it has to be noted that with regard to single events, East Asian reports sometimes proved useful for supporting or contradicting claims of important solar events deduced either by direct observation (e.g., Willis and Stephenson [11]) or by proxies (e.g., Hayakawa et al. [12]). Naked eye sunspot reports are of special interest for the MM period, actually even more than for aurorae, because the latter attracted attention at all times, while sunspots rarely do, due to the difficulties to discern anything at all due to the Sun’s glare. Naked eye sunspot reports are like lucky hits. However, within a Grand Minimum, lucky hits are far less probable. Imagine a very inactive sun for seven decades and then by chance see a sunspot large enough to be clearly visible to the naked eye. This is much more probable in periods of high solar activity.

The most extensive collection of historical aurora sightings is that of Yau, Stephenson and Willis [13], which covers the period from 193 BC to 1770 AD. A brief listing of additional Japanese sightings is given by Nakazawa et al. [14]. Aurorae from Korea between the years 992 to 1756 AD were collected by Lee et al. [15].

For the Arabian Peninsula, Basurah [16,17] has collected aurorae in two short articles and found 25 sightings for the period 816–1570 AD. Due to the southern location of the oriental cultural space, on first sight, they seemed to be of particular interest, but since the contemporary magnetic latitude for the time span in question (Korte et al. [18]) indicates almost similar conditions as in East Asia, their significance should at least not be overestimated.

Fritz [19] gives a worldwide collection of historical auroral phenomena in his classic study. The occurrence of aurorae in the temperate or even tropical zone is linked to geo-effective coronal mass ejections (CMEs), so phenomena from these latitudes are much more meaningful than those from high latitudes. For the 11th to 19th centuries, Krivsky and Pejml [20] compiled a revised list based on Fritz’s catalogue, as well as several other lists, and singled out those phenomena that occurred south of the 55th parallel. The aim of the two Czech authors was to collect northern light observations from their home country, the historical regions of Bohemia and Moravia, and also to publish them verbatim in English translation. Schröder [1] gives a more extensive, but also more time-limited, list of Central European sightings for the period 1545 to 1724 AD. Another Central European collection that is frequently cited is that of the Hungarian historians Réthly and Berkes [21], who are best known for their vast collection of historical sources of relevance for the studies of past terrestrial climate (e.g., Racz [22]). As a new contribution, Stangl and Foelsche [23] extracted from historical records aurorae sightings from only one part of Hungary, namely the province of Transylvania and also quoted or paraphrased them in verbal form. Some of these phenomena also appear in the Réthly/Berkes collection, but a rather surprising number of new documents could be found.

It was only in the 18th century that auroral phenomena were studied and collected systematically by contemporary scientists (see Usoskin et al. [10], Stangl and Foelsche [23]). Reports in former centuries have not been made in any systematic manner and might have been interpreted erroneously by historical chroniclers and/or modern researchers, see for example Hayakawa et al. [24]. The problem is analogous to the one concerning early sunspot observations (telescopic as well as naked-eye).

Hoyt and Schatten [25] tried to circumvent the problem of sparse early sunspot observations by constructing long time series of group sunspot numbers (GSN), which, according to them, should be less affected by errors than the counting of individual sunspots. Recent studies, however, clearly point to the direction that the early part of this time series (which starts in 1610) should not be used as a measure of solar activity, as it is based on too few actual observations (e.g., Stangl and Foelsche [9], Vaquero et al. [26], Hayakawa et al. [27]). Auroral statistics are less affected by such kind of misinterpretation, because they were never claimed to originate from scientific surveillances, but this has other drawbacks. Not surprisingly, the conclusions drawn from the data have proven to be very contradictory, as we will show in the following sections.

2. General Problems with Historical Aurora Catalogs

Eddy [28] had defined the MM quite precisely as 1645–1715 AD, confirmed today by isotope analyses (e.g., Usoskin et al. [10]). Nevertheless, in the recent past attempts have been made to fine-tune the duration of the MM. The approach by Vaquero and Trigo [29] on redefining the epoch is based on the one hand on reliable isotope analyses, but on the other hand on the auroral compilation from Hungary by Réthly and Berkes [21] and also on the GSN values of Hoyt and Schatten [25], which proved unreliable for the whole time span in question, although they are, of course, a very useful measure for later observations, starting around the middle of the 19th century, when data density became high enough to calculate these values in a reliable way, as shown by Clette et al. [30]. While the situation of data density and quality of aurorae sightings during the MM is already more than unsatisfactory to reliably deduce the strength of solar activity during the epoch, a restriction to the Réthly–Berkes catalog cannot be supported either, especially because it is based on the assumption that “the series of auroras observed in Hungary were built using non-scientific literature and appear to insure better uniformity and a more straightforward link to solar activity” [29] than the Krivsky and Pejml [20] compilation. The numbers derived from Réthly and Berkes agree quite well with the GSN by Hoyt and Schatten, but rather by coincidence, because both sources (although Hoyt and Schatten more so) proved unreliable (see Stangl and Foelsche [9,23]). Other aurorae catalogs, such as the one by Schröder [1], which are clearly contradictory, have been left aside by Vaquero and Trigo [29] and others.

Landsberg [31] not only denied Eddy’s assumption of the connection between the Little Ice Age (LIA) and the MM, but also the whole idea of the latter, by drawing attention to 52 sunspot reports from the period 1645–1715 AD. Out of these 52 sunspots, Eddy [32] claimed only 20% of them to have been previously “unknown” and further remarked that five of them fall into years, not “representative” for the MM, namely the years 1681, 1688, 1703 and 1708 AD. After considering those sunspots as not conflicting the concept of the MM, Eddy had to deal with Landsberg’s list of auroral reports. He argued that 54 out of 90 of these aurorae (which were first mentioned by Schröder [33]) fall into years when also the sunspot activity within the MM was supposed to be higher than during the rest of the grand minimum. A total of 30 more reports fell into the years 1704 to 1715, which might represent a transition phase from a “deep” MM into the normal state. Conversely, Eddy admits that Schröder has still found 36 aurorae, mostly from German speaking regions in Central Europe, in years in which the average sunspot number was supposed to be zero.

Eddy also tried to relativize the importance of aurorae as proxies for high solar activity and claimed them to be only loosely related to sunspots or flares, a somewhat problematic position. In addition to Schröder’s work, the aurora catalog compiled by Link [34] also points to a higher auroral activity in Central Europe than might be expected within a grand minimum of solar activity. Of course, today we have more relevant proxies due to isotope analysis, which started with radiocarbon investigations on tree rings, which have been highly refined recently in order to reconstruct the solar cycle throughout the last millennium, see for example Brehm et al. [35]. However, historical reports for sunspot and aurora would be useful to fine-tune our knowledge of solar activity during the MM and old visual aurorae observations have shown significant implications for the magnitudes of associated geomagnetic storms on the basis of their geographical extensions (e.g., Hattori et al. [36]). Furthermore, it has been supposed, that auroral activity at mid-northern latitudes, as it is the case in the Central European sector, can be triggered by rather moderate geomagnetic storms caused not only by CMEs but also by CIRs (Corotating Interaction Regions), following the empirical correlations of the auroral oval extension and geomagnetic storm intensity Yokoyama et al. [37], Richardson et al. [38]. Multiple CIR generated magnetic storms have been reported even near the recent deep solar minimum in 2008/09, see Jian et al. [39]. Provided that coronal holes appear occasionally, the auroral frequency in the European sector might not be reduced as drastically as has been supposed before. We also have to note that there are some robust candidate aurorae supported by naked-eye sunspots reported from other sources and associated with each other only recently (e.g., Willis and Stephenson [11,40], Willis et al. [41]), as well as promising identification of solar storms via geomagnetic measurements (e.g., Vaquero and Trigo [42]).

Direct observation would be the only way to disprove proxy results, and could count as benchmark, but as no systematic efforts have been made to monitor the Sun during the MM (see Stangl and Foelsche [9]), the best semi-direct indicator would be the manifestation of solar activity via auroral displays. The searches for sunspots by great observers like Hevelius (Carrasco et al. [43]), Cassini (Carrasco and Vaquero [44]) or the astronomical logbook entries of less well-known observers like several members of the Eimmart Family in Nürnberg (Hayakawa et al. [45]) were not made in a sufficiently systematic manner, apparently due to the lack of interesting phenomena to be seen on the solar disc.

The end of the MM seems to have been marked by spectacular celestial fireworks. On 17th March 1716 AD (converted to Gregorian date), strong Northern Lights flashed across continental Europe and the British Isles, one of their scientific observers being Edmond Halley [46], who made the following, very interesting statement regarding the aurora phenomenon: “I then should have contemplated propriis oculis, all the several Sorts of Meteors I remember to have hitherto heard or read of. This was the only one I had not as yet seen, and of which I began to despair, since it is certain it hath not happen’d to any remarkable Degree in this Part of England since I was born [i.e., 1656 AD]; nor is the like recorded in the English Annals since the Year of our Lord 1574…”. Halley could not find any “credible” reports of northern lights between the years 1621 and 1709, whereas there were some aurorae reported during the time of the MM with reduced frequency (e.g., Stangl and Foelsche [23], Hayakawa et al. [24], Riley et al. [47]). Even from London itself they have been reported, as can be seen from a sighting made there in 1661, as shown by Usoskin et al. [10].

While Mairan [48] claimed that Northern Lights over France and the neighboring countries had been very rare for at least seven decades before their return in 1716, nevertheless he notes that within the period 1686–1690 AD several have been seen from Europe, including one especially vivid display observed in Germany’s Rheingau Province. The statement by Usoskin et al. [10] that “Mairan’s original survey reported 60 occurrences of aurorae in the interval 1645–1698” seems to be some kind of translation error, at least it cannot be found in Mairan’s (1731) book (in French) quoted by the authors. Mairan [48] instead refers to 60 years, in which he could not find any aurora reports: “Depuis 1621 jusqu’en 1686, c‘est-à-dire, dans l‘intervalle de plus de 60 anées, je ne trouve aucune Observation bien marquée de l‘Aurore Boréale”. (Reprise XX., p. 172).

Contrary to the assumption of a decrease in auroral activity during the MM, several authors have repeatedly pointed out that they occurred at least so often, that even the solar cycle can be identified based on those reports, i.e., that there were more aurorae occurring at an interval of about 11 years (e.g., Schlamminger [49]). This dilemma is usually explained by the assumption that aurorae always have been present at certain latitudes, but weaker than usual and that only the frequency of phenomena at low latitudes was weak. However, this contradicts the Northern Lights over Central Europe mentioned by Schröder [1].

Cosmogenic isotope data indicate the existence solar cycles within the MM (see Usoskin et al. [3]), which seems to be supported by re-evaluation of sunspot observations [50], even though the so-called “active-day statistics” still suffers from a lot of misinterpretation and spurious data, as can be seen for example in the recent re-examination of the Eimmart observations by Hayakawa et al. [45]. Nearly all recent studies point to rather pronounced solar cycles, so the original assumption by Eddy [28] and others, that during the MM the cycles all but stopped, is regarded as highly unlikely today. For a long time, the scientific community was misled by a vast number of alleged spotless days owing to contaminations mainly from astrometric measurements confused with physical examinations and from general comments taken as proof for observations on a regular basis. This grave mistake was introduced by Hoyt and Schatten [25], leading to spurious suppressions of solar cycles (e.g., Beer et al. [51], Hayakawa et al. [52]). Given this situation, it is evident that auroral records supporting solar cycles within the MM, would not conflict with isotopic results for the solar activity during that period. Sunspot data, cosmogenic isotope data and auroral data can well complement each other. While sunspots directly reflect magnetic activity on the solar surface, cosmogenic isotopes show variable injection amounts of cosmic rays which anti-correlate with the solar wind, while midlatitude auroral records reflect the frequency of geomagnetic storms, and subsequently indicate the frequency of solar eruptions including CMEs (see Usoskin et al. [53]). Therefore, even for the long-term variability, it would be desirable to combine records for sunspots, aurorae, and cosmogenic isotopes.

How does the situation in the “deep” MM compare with that in the “shallow” Dalton minimum of 1790–1830 AD? According to Silverman and Hayakawa [54], aurorae have occurred more sparsely than normal. Few northern lights have been seen below a geomagnetic latitude of 56° in both periods. Although the number in the Dalton minimum is significantly higher, one must keep in mind that at the beginning of the 19th century the reporting situation was undoubtedly far better than ever before. Interestingly, several aurorae were seen in Korea during the MM, which were reported as red rays in all directions, including the south [13]. Further conflicts of the Neuhäuser scheme, which we will discuss in Section 3, to the observational evidence were discussed by Stephenson et al. [55] with a case study, comparing a doubtful event from 776 AD with a confirmed aurora from 1882 AD. Using different catalogs, Usoskin et al. [10] conclude that auroral activity within the MM must have been much lower than during the short Dalton minimum. While this may well have been so, we point out that the situation of historical sources is not comparable and that aurorae even before the MM are mentioned disproportionately rare.

A study of aurorae within the territories of today’s Czech Republic and Slovakia was presented by Krivsky and Pejml [20]. However, this catalog, as well as the Réthly collection, which was consulted in detail by Stangl and Foelsche [23], is probably less representative due to the more limited geographic scope than Schröder’s [1] compilation. In any case, we should keep in mind that of nearly 4000 auroral sightings known before the year 1900, little more than 300 fall into the first half of the second millennium, as underlined from the very beginning by Eddy [28].

Potential auroral sightings from the 11th to the 18th century, reported from Korea were collected by Lee et. al. [15]. Contrary to what is generally accepted, the authors think that their historical data clearly show solar cycles and, in contrast to Schröder’s collection from Central Europe, the Spörer Minimum (SM, about 1410 to 1540 AD, according to [3]) and the MM also seem to show up as periods of aurora poverty. Lee et al. [15] state: “On the other hand, Schröder (1994) reported that auroral activity during the period 1450–1550 was normal, using auroral data from Central Europe (…). [It] contradicts what we can see in Figure 1. It seems that he could not detect the real behavior of solar activity, because of insufficient auroral data from Central Europe”. It has to be noted that Schröder has identified more northern lights from the period in question, so “insufficient data” can hardly be claimed, but on the contrary: it might be argued that he used too much data, i.e., possibly included doubtful observations. Looking at Lee et al.’s Figure 1 [15], showing the distribution of the Korean Northern Lights, one can see that there are few promising sightings. During the MM not much can be seen, but this is also true for the adjacent times before and after. Conspicuous peaks show up around the year 1625 and during the 16th century, as well as in the late 14th century and small peaks around the years 1200, 1600 and 1725. The SM shows as a clear gap with only the year 1467 standing out as a lonely peak of auroral activity. However, one has to keep in mind that the total amount is so small, that even the slightest source of error such as inhomogeneity in historical tradition (which is inevitable) would completely change the graph, as the mentioned strong peak of 1467 means no more than three aurorae. The problem is similar to that of early sunspot observations: with so few reports, it is impossible to calculate reliable activity levels. For whatever reason, the fifteenth century has the least entries of all within the second millennium in the catalog of Lee et al. [15] and so would fit the SM very well.

To verify the reliability of their sources, Lee et al. [15] checked, whether known celestial phenomena of the time, such as eclipses and conjunctions show up in them. A “reliability” graph they give in a Figure 2, shows also a minimum of reliability within the time of the SM, suggesting that it might actually be nothing more than a gap within their data. Of course, this does not mean that aurorae might not have been rare during the SM, but it means that for the MM the density of reliable sources is much better, although still no better than for the 12th to 14th centuries, according to their Figure 2.

Neuhäuser and Neuhäuser [2] have defined five criteria to verify aurorae reports. First, of course, an aurora has to take place at night, but also becomes somewhat questionable if it was reported around the full moon, i.e., not in a dark sky. Secondly, northern lights should be visible in the North, although cases are known where they appear in the South because the observer is northward of the aurora oval, but this is extremely rare at mid to low latitudes we are dealing with. Thirdly, aurorae at lower latitudes are generally characterized by a reddish hue. Fourthly, Neuhäuser and Neuhäuser [2] allow the word “fire” to express the movements of the Northern Lights only for “European reports”, because, according to them, but without giving any convincing philological evidence for this statement, “East Asian reports” would use the word “fire” only in the context of red color, and a change in some form should be mentioned, otherwise a halo, comet, or other static celestial phenomenon will be suspected. Last, an aurora should not remain visible for more than a few days.

While most of Neuhäuser’s criteria could be applied with some confidence to modern observations, they are asking way too much of an early chronicler (and “early” can be said for anything well into the 18th century at least) to fulfill all these criteria at once, and above all, the five criteria mentioned have a very different weight, especially regarding the accuracy of observation and tradition, in order that the annalist should have given all that detail allowing us to check all five points. With the exception of the account of an auroral display from 10 September 1580, which was described by an obscure Italian scientist and unearthed by Kazmer and Timar [56], only very rarely (for descriptions made by laymen, many examples are cited full length by Stangl and Foelsche [23]) enough details are reported to apply Neuhäuser’s criteria with any confidence. We plead for more tolerance in screening and to consider the circumstances of the reports, i.e., for more historical understanding, because it can never be decided safely what was really seen at that time, due to a lack of accurate information anyway. One must always keep in mind that old reports are subjected to high uncertainty, i.e., one must beware of the statistics of small numbers and should avoid drawing any far-reaching conclusions from them.

It is strange enough, that in reports from earlier centuries no clear distinction between the “Wild Hunt” and aurorae can be made, and that people saw entire armies marching in the sky, when nothing more than an auroral curtain showed up and brought some movement into the statics of a (post)medieval firmament (Schroeder [1], Krivsky and Pejml [20]). To the five criteria introduced by Neuhäuser and Neuhäuser [2] it might be added that aurorae in reality never emit sounds, while there are many old reports mentioning war cries in the air, accompanying the “heavenly armies” reported—several examples are listed for example by Krivsky and Pejml [20]. In any case, it is hard to understand today how the imagination of our ancestors created such claims. However, since it is an incontrovertible fact that they did, and not rarely, but frequently, we must take note of it and should therefore be more forgiving in the acceptance of strange attributes in old reports and accept that they are severely distorted and not observations in the modern sense. Furthermore, this does not apply for the remote past only. Take, for example, the meteorite fall of Agen in France on 5 September 1814. No one can doubt it because the fallen stones are preserved in museums up to this day and additional contemporary scientific documentation does exist, most notably the study by Saint-Amans [57]. However, in their catalog of unexplained flying objects within historical times, Vallee and Aubeck [58] mention a number of fancy attributes (a hovering of a flying object in the air for one whole hour, sudden movements up and down and a spinning around its axis) by eyewitnesses that are in clear contradiction to a natural object. This means that the phenomenon has been reported in a very distorted way, something the historian is aware of, while natural scientists often fall into the trap, believing that modern understanding of observational standards can be applied for epochs long past.

Another problem with Neuhäuser’s strong filtering is that a lot of convincing auroral reports exist, that would have to be excluded, because they contradict one or more of the criteria and not only because data about the circumstances in question are missing. For example, Stephenson et al. [55] have shown multiple cases of aurorae, which were seen in brightly moonlit nights, at other azimuths than the north and in colors other than red. While it is true that aurorae cannot been seen during the daytime, twilight observations of strong displays must also be considered (see also Hayakawa et al. [12]). It would go beyond the purpose of this article to discuss in detail exceptions to each of Neuhäuser’s criteria. For the purposes of illustration, let us mention the following examples: Rethly and Berkes [21] list five sightings out of 46 for the period 1891–1956 AD, that were made in brightly moonlit nights near the full phase. In addition to the dominant red color in auroral displays, all kind of colors including white, yellow, orange, green and even blue have been reported in more than 10% of all catalog entries between the years 1523 and 1960 AD. It must not be forgotten, however, that a violation of Neuhäuser’s criteria does not mean at all, that an aurora could not have been genuine. So many fantastic additions crept into old reports, that the mention of cries in the air like for the aurora of January (exact date not available) 1593 AD in the Krivsky/Pejml collection [20] should not hinder the historian or physicist to suspect the reality of the auroral occurrence itself.

While the Neuhäuser criteria indeed do not apply to every single auroral display, they are still useful for probability examinations because they reflect a typical expected display. It does not lessen the idea, that derivations, especially in more than one point, are not the common case and could be used as criteria, provided there is enough data available, but exactly this is not the case for many old reports. Proving that low latitude aurorae also appear in the south, show colors different from red or that cases are known that were recognized in full moon nights, does not contradict the fact that these cases are by far fewer than the others. However, given the source situation during and near the MM, Neuhäuser’s criteria are not very helpful and should be left aside.

In summary: the five-point criteria by Neuhäuser and Neuhäuser [2] are useful as a reminder for critical examination of old auroral reports and as a call to look more closely and be more selective when counting historical aurora appearances. However, it must be noted, that too strict a selection would be clearly counterproductive and also lacks historical understanding. It is true that many catalogs like the one by Lee and al. [15] list anything found, that just might have been an aurora, and sometimes it is evident that the reports do not refer to aurorae as for example entry no.65 in the catalog of Yau, Stephenson and Willis [13], where it reads for January 584 AD: “At night, the sky opened from NW to SE. Within it there was a blue-yellow colour. There was a sound like thunder”, which clearly points to a bright meteor seen, not an auroral display. Xu, Pankenier and Jiang [59] list only reports of phenomena that have explicitly occurred at night, but by doing so, they have undoubtedly also eliminated several true aurorae, where the chronicler just did not mention the time of the sighting. Conversely, a large number of wrong entries were undoubtedly also eliminated this way, so that on the whole this filtering might be regarded more useful than harmful, would it not be for the fact that they have contradicted quite a serious number of confirmed auroral displays that did not follow Neuhäuser’s strict criteria, as has been demonstrated above. So, hyper-critique, as introduced with Neuhäuser’s scheme on the other hand, seems to be rather counterproductive. There are many examples of robust and critical studies analyzing visual reports of conspicuous auroral events (e.g., Hattori, Hayakawa and Ebihara [36], Hayakawa et al. [60], Carrasco et al. [61], Hattori and Hayakawa [62]), which do very well without applying any kind of dogmatic filtering. In some cases, Neuhäuser’s criteria directly contradict the observational evidence during known space weather events.

3. Results

In a study of aurorae frequency during the MM, Schlamminger [49] claimed no reduced values for Central Europe, questioning the reality of the MM. He quotes 404 reports on aurorae in Central Europe during the 16th and 17th centuries, especially from the Holy Roman Empire, i.e., what today amongst others is Germany, Austria, Czechia, Hungary and parts of other neighboring states. According to Schlamminger, there was no visible decrease during the MM (118 northern lights over the sky of Central Europe between 1645 and 1712 AD). Furthermore, he claims, that the usual cyclicality of roughly eleven-years shows up, as far as the sparse data allow such a statement at all.

According to Schröder [1] northern lights, which occurred between 1645 and 1715 AD over Central Europe at latitudes from 48° to 53° should represent an undisturbed continuation of the solar cycles: “In almost every year—before and during the Maunder Minimum—aurorae have been observed in Middle Europe. Because they are indicators of solar geomagnetic activity, it follows that no unusual variability in solar or geomagnetic activity is indicated”.

How can these claims be reconciled with modern results of isotope analysis? Schröder drew his reports apparently from various sources and combined them in a table which seems to be the most complete and best list for the period in question, especially because he claimed source-critical methods. Unfortunately, the source catalog itself seems not to have been published, so that it could be verified. Efforts should be made to find it.

The imaginative descriptions of the epoch certainly reflect more the spirit of the age, the experiences and fears of the people of the time, than the physical parameters of the appearances to us today, many of the old reports seem quite strange (see also Stangl and Foelsche [63]). What interests us more than details, is their number and, if possible, that the data collected originate all from the same geographic area. “Generally speaking, from the statistics of auroras, the period 1545–1715 represents a distribution mainly similar to that of recent years”, is Schröder’s conclusion.

Riley et al. [47] and others, on the other hand, concluded that few aurorae have been recorded throughout the MM, using the well-known catalog of Réthly and Berkes [21], which lists very few northern lights for the years in question. However, comparable auroral poverty in their catalog equally show up for other periods, too. Strangely, Riley et al. [47], when claiming near ceasing auroral activity during the MM, declare, that their findings would not change if they used Schröder’s catalog instead of the one by Réthly and Berkes [21]. They probably refer to the fact, that in Schröder’s compilation, more sightings occur before and after than during the MM, but ignore that Schröder claimed more than a hundred appearances within the MM, while Réthly and Berkes knew only a handful.

Figure 1 shows the frequencies of auroral occurrences for the time span 1547–1721 AD, according to Rethly and Berkes [21] and Schröder [1] respectively, given as five-year moving averages. The numbers of Rethly and Berkes represent every entry in their catalog considered by them as genuine aurorae, even though likely “contaminations” due to atmospheric phenomena and meteors have been admitted by the authors themselves, so the actual number is some kind of an “upper estimate” of the numbers found. Schröder, who has much higher numbers, but might be supposed to suffer from “contamination” as well, gives for several years a range of numbers, depending on what he felt the “sureness” of his findings. In order to better compare with the Rethly and Berkes (1963) [21] data, we chose to give the lower values and to put them to zero whenever Schroeder was in doubt if the number was higher than that.

While the data from Rethly and Berkes [21] show a sharp decline of auroral activity starting in the third decennium of the 17th century, lasting almost exactly one hundred years, the data from Schröder [1] show not very dramatically reduced values, and actually a nicer fit to the established classic definition of the MM as 1645 to 1717 AD. This is also in good accordance to newer studies that support a higher solar activity near the end of the MM (see Landsberg [30]), which is very well supported by the very strong peak for the year 1707 AD, that even equals the famous “aurora year” of 1716 AD. Note also the high auroral activity for the 17th century before the onset of the MM in the Schröder data, while the Rethly and Berkes graph suggests after 1618 AD low activity comparable to the MM. It is also notable that the year 1618 marks the beginning of the Thirty Years War in Central Europe, thus changing the situation for historic tradition drastically. No homogeneity within the data of either Rethly and Berkes [21] or Schröder [1] can be expected. The paradigm that Rethly and Berkes’ compilation is more “homogenous” than others seems largely to be a fallacy, because they seem to be in very good accordance to the SM and MM established by isotope analysis. However, this does not take into account 1. the very lack of any Hungarian sources prior to 1540 AD with the single exception of one in 1523—so the SM is impossible to show up, and 2. that the nature of sources is not homogeneous at all, neither from the historical, nor from the geographical viewpoints (there is a strange concentration of sources from the province of Transylvania for the time in question—see [23]), but rather reflects historical and political circumstances. After the battle of Mohács, which took place in 1526 AD, after which most of the country fell under the dominancy of the Ottoman Empire, the source situation changed completely. The source situation within the kingdom of Hungary from the fading of the SM to the late phase of the MM, also has much to do with the ever changing fortunes and misfortunes of the Transylvanian voivods. While several of them established harsh regimes within the country, others, most notably Bethlen Gábor, created a climate of tolerance and offered a breeding ground for culture and science. However, it was not under his rule, that the Italian scientist Squarcialupi [56] has written his excellent and first scientific treatise of an auroral display, but under the domination of Báthory Kristóf, who is not regarded as a particular important ruler (cf. Sienerth and Wittstock [64], Teutsch [65]). The sources used by Rethly and Berkes [21] are of a quite mixed nature. Neither completeness, nor homogeneity are strengths of Rethly and Berkes, but the limited geographic space and the time span over half of a millennium. However, even the geographical distribution of the sightings reflect the political and cultural circumstances due to available sources rather than the real distribution of the aurora visibility zone. Furthermore, Vaquero and Trigo [29] are mistaken by stating that “the series of auroras observed in Hungary [21] were built using non-scientific literature” only. This is just not the case, especially not for later periods when far more observations by professionals (for example the astronomer Miklós Konkoly) overwhelm the others. While we fully agree with Vaquero and Trigo [29], that “civilization factors” as tried by Krivsky and Pejml [20] are a “rather artificial fix to the problem”, there seems to be little supporting evidence to suggest that the Rethly/Berkes catalogue should “insure better uniformity and a more straightforward link to solar activity”, as the authors claim.

However, the main issue, which would make a reevaluation of Schröder’s findings so eminently important, is the sheer number of phenomena given by him, which would suggest that, despite a Grand Minimum, auroral activity would still have been at quite respectable levels in Central Europe, contrary to what is generally believed to have been the case, i.e., dramatically reduced frequency of the aurora phenomenon during the MM. This can be further suggested by the parallel example of the much shallower Dalton Minimum, where John Dalton’s auroral observations show a significant decrease of auroral activity, as shown by Silverman and Hayakawa [54]. A partial explanation could be the magnetic situation, which according to the reconstruction by Korte et al. [18] was somewhat more favorable during the 17th century for the sightings of aurorae than in later times. As shown by Hayakawa et al. [24], Central Europe was ~4 degree closer to the magnetic pole during the MM than it is now, so the regions studied by Rethly and Berkes, Schröder and others, would be expected to have witnessed greater auroral activity than it might be expected otherwise. In the whole, the activity fluctuations following Schröder’s data coincide better with the accepted solar activity situation during the 17th century than Rethly and Berkes, but show a much higher rate than is currently expected to have been the case, especially during the decades 1650 to 1690 AD where the reconstructed decadal sunspot number in the study of Wu et al. [66] drops below SSN 10, while the aurora numbers according to Schröder remained at about the same levels as in the decades before and after.

While a drop of activity during the MM shows up in the Schröder graph as well, i.e., a drop in numbers compared to before and after, the decline is far less dramatic than in the Rethly/Berkes graph. The MM period itself, i.e., 1645–1715 AD, is actually more clearly visible in Schröder than in Rethly and Berkes, because there a drop around the year 1620 AD continues to well after the grand Minimum ended, i.e., around 1715 AD. The actual situation is that Rethly and Berkes found so small a number of aurora sightings at all (between the years 1623 and 1724 only 14 events compared to 194 listed by Schroeder), that the use of Rethly and Berkes must necessarily be very difficult due to their small numbers that are more vulnerable to errors and coincidences than Schröder’ s rich material. Therefore, Rethly and Berkes should be regarded as highly unlikely to reflect any real fluctuations in auroral activity during a time span of more than a century.

4. Discussion and Summary

Aurorae were studied systematically starting from the year 1716 AD, so reliable series of records only became available from then. Both aurorae and sunspots, especially from pre-telescopic times, but also predating the 18th century, should be used as supplementary information rather than considered fully trustworthy on their own.

Regarding the Maunder Minimum (MM), different aurora catalogs give very contradictory information. While Réthly and Berkes [21] knew only a handful auroral appearances within the MM, Schröder [1] claimed more than one hundred within Central European latitudes. The Krivsky and Pejml [20] catalog, as well as the Réthly collection [23], are probably less representative due to the more limited geographic target areas than Schröder’s compilation. This severe contradiction needs reevaluation, so efforts should be made to gain access to Schröder’s original data, which are a very crucial source for the question of the actual auroral activity during the grand minimum of the MM epoch within the 17th and 18th centuries. The compilation of a revised catalogue of aurora sightings from Middle Europe on the basis of this yet lost catalog would be a very important contribution.

Aurorae as well as sunspot observations before the 18th century are too rare to construct long-term trends, and they do not live up to today’s standards. Data bases like the GSN (Group Sunspot Number) values deduced by Hoyt and Schatten [25], mainly due to methodological shortcomings, but also due to many translation mistakes, proved not to give any new insights over what has been found out by isotope analyses. Both sunspot and auroral data from pre-telescopic time should therefore only be used as supplementary information, but never as fully trustworthy on their own. Isotope analysis should be the preferred proxy method in order to reconstruct past solar activity levels.

However, even sparse observations can be valuable, when they are a positive prove for solar activity, in particular during times of (apparent) low solar activity. In order to enlarge our knowledge of solar activity in early times, i.e., predating systematic observational coverage, observations of sunspots and aurorae (e.g., Hayakawa et al. [24]) should be used rather than lack of them, because the latter can have causes other than the absence of the phenomenon in question, i.e., lack of interest, possibilities, tradition, etc. Rather than trying to reconstruct long-term solar activity fluctuations based on aurora sightings at mid and low latitudes, the positive sightings should be used in order to infer possible specific solar events like coronal mass ejections. As an example, observations of aurorae at mid latitudes during the MM (as for example unearthed by Stangl and Foelsche [23]) are more significant than their absence in the records. As no systematic efforts have been made to monitor sunspots and aurorae during the MM, positive and negative sightings come more as a coincidence and do not allow reliable statistics for long term solar activity levels within that period. Every new look into one of the sunspot data sets given by Hoyt and Schatten [25] has since proved to result in major revisions, often dramatically reducing the number of days with physical solar observations during the MM, see for example the observations of the Eimmart clan, which were reevaluated in an exemplary manner by Hayakawa et al. [45]. Until all datasets used by Hoyt and Schatten [25] are not re-evaluated, no reliable picture can be drawn and in the meantime observation numbers during the MM continue to be in a free fall, with unrealistically high percentages of active observations still cited in the literature (e.g., Vaquero et al. [26]). However, even today, when data calibration has been perfected to a high degree [30], data scarcity is a serious problem and may be impossible to overcome. It should not be forgotten that little stimulus must have existed for an observer during the MM period to keep a daily surveillance for sunspots. While the paucity of known “active days” between 1645 and 1715 AD is a good indicator for genuine low sunspot activity during the MM, many groups and single spots might have easily been missed.

The criteria used by Neuhäuser and Neuhäuser [2] and Neuhäuser, Neuhäuser and Chapman [67] to distinguish genuine from false aurorae reports cannot be supported for the time before the 19th century, as they contradict the scientific understanding of earlier epochs and neglect the intentions of early chroniclers, who in most of the cases utilized heavenly signs to transmit religious messages, as shown very clearly by Wozniak [68].

As the imaginative descriptions of the epoch certainly reflect more the spirit of the age, i.e., the experiences and fears of the people of the time, than the physical parameters of the phenomena, we plead for more tolerance and historical understanding, when interpreting early observation reports. One must always keep in mind that old reports are subjected to high uncertainty, and should avoid to draw any far-reaching conclusions from statistics of small numbers anyway (see Stangl and Foelsche [63]).

While it is true that many suspected aurorae may well have been other phenomena, it is in most cases impossible to surely discern between true and false. Take for example the “fire sign” reported from Lübeck in 1680, which was meticulously analyzed by Hayakawa et al. [24]. Although in this case conclusions sound extremely convincing, a meteor still cannot be ruled out because in case of a really bright bolide, a “fiery cloud”, distorted by winds, could remain for hours in the sky. This would also explain the sounds mentioned in some of the reports.

5. Conclusions

The criteria used by Neuhäuser, Neuhäuser and Chapman [67] are counterproductive because they eliminate true aurorae along with the false because chroniclers of former times did not apply today’s scientific standards when describing phenomena in the sky. Furthermore, recent studies (e.g., Stephenson et al. [55]) have shown that the source criteria in Neuhäuser and Neuhäuser [2] and Neuhäuser, Neuhäuser and Chapman [67] contradict several confirmed auroral observations in the modern scientific literature. However, reconstructions based on cosmogenic isotopes are by far to be preferred for establishing the level of solar activity in the past. The importance of our review lies within showing the clear limitations of the usefulness of historical aurora reports. While this does not mean they should be disregarded, analysis must be made in a historically critical manner and no far-reaching conclusions should be drawn from them. Further research is welcome, and we especially call upon trying to find the original sources of Schröder [1], since his aurorae catalog spans the whole MM and contradicts what has been deduced from more used compilations. A more complete picture of auroral activity, especially during the MM, can then be compared to the latest results from isotopic studies, as well as to different models of the solar dynamo (e.g., Zharkova et al. [69]).

The major conclusions from this review can be summed up as follows:

-

Neither sunspots nor aurorae have been observed in a systematic and scientific way before long after the end of the MM (1645–1715 AD)

-

Almost all of the aurorae reports before the 18th century AD were made by laymen and have furthermore been reported in a distorted manner and often even deliberately been edited in order to transport religious and moral messages.

-

The scheme of Neuhäuser and Neuhäuser [2] and Neuhäuser, Neuhäuser and Chapman [67] cannot be applied with confidence to pre-19th century aurorae reports.

-

The poor data material of auroral observations generally supports low solar activity during the MM. However, depending on the catalogs used, quite different pictures about auroral activity emerge.

-

Thus, even further studies are not likely to surpass our knowledge obtained via isotope analysis.