The geomagnetic data of the Clementinum observatory in Prague since 1839

The historical magnetic observatory Clementinum operated in Prague from 1839 to 1926. The data from the yearbooks that recorded the observations at Clementinum have recently been digitized and were subsequently converted in this work into the physical units of the International System of Units (SI). Introducing a database of geomagnetic data from this historical source is a part of our paper. Some controversial data are also analysed here. In the original historical sources, we identified an error in using the physical units. It was probably introduced by the then observers with the determination of the 5 temperature coefficient of the bifilar apparatus. By recalculating the values in the records, some missing values are added; for instance, the temperature coefficients for the bifilar magnetometer, the baselines, and the annual averages for the horizontal intensity in the first years of observations were re-determined. The values of absolute measurements of the declination in 1852, which could not be found in the original sources, were also estimated. The main contribution of this article rests in critically reviewed information about the magnetic observations in Prague, which is as complete as no other so far. The work also con10 tributes to the space weather topic by revealing a record of the now almost forgotten magnetic disturbance of 3 September 1839.

possible. Historical geomagnetic records can be useful to deal with this task, however they can be heterogeneous, incomplete or consist of corrupted data. Their processing should involve several steps like digitization, conversion from scale units to proper physical units, dealing with inaccuracies or missing data, searching for accompanying information like auroral records, etc. To obtain complete datasets, compiling archive records from several magnetic observatories is needed. In previous works by e.g. 25 Jackson et al. (2000); Jonkers et al. (2003); Arneitz et al. (2017), historical geomagnetic data were collected, processed and analysed with the aim of preparing datasets usable for broader geoscience community.
Historical geomagnetic observations are thus understood to be an important source of information about the structure and time behaviour of the geomagnetic field. The magnetic needle has already been commonly used in naval navigation since the 13th century, however at first it was supposed that the needle was oriented towards the true north. The fact that the orientation of 30 magnetic needle depends on the position was documented in the 16th century by explorers sailed in Atlantic and Indian Oceans.
On the other hand, the first sustained series of measurements at a single site in Greenwich showed that the geomagnetic field was subject to time-dependent change. The magnetic needle moved of about 7 degrees westerly over period 1580-1634. The first magnetic inclinometer (a magnetized needle rotating on a horizontal axis in the vertical plane of the magnetic meridian) was constructed by London compass maker Robert Norman at about 1580. 35 Relative magnetic intensity data began to be compiled from 1790s by comparing the time it took a magnetic needle displaced from its preferred orientation to return to it, or by the duration of a given number of such oscillations.

The method of absolute determination of magnetic intensity was developed by Carl Friedrich Gauss in cooperation with
Wilhelm Weber in 1832. The method combines vibration and deflection experiments in order to separate intensity of the magnetic field and magnetic moment of the magnet used in the experiment (Gauss, 1833). In 1833, Gauss and Weber finished the 40 construction of the Magnetic Observatory of Göttingen and developed or improved instruments to measure the magnetic field, such as the unifilar and bifilar magnetometers. The Göttingen Observatory became the prototype for many other observatories worldwide. Improvement of observatory practice was not a goal of Gauss's work, but just a tool for understanding the nature of the Earth magnetic field. Gauss and Weber therefore joined the activity of Alexander von Humboldt in establishing a worldwide network The magnetic rods in bifilar magnetometers constructed by the famous instrument makers M. Meyerstein in Göttingen and advertised in the Resultate aus den Beobachtungen des magnetischen Vereins im Jahre 1837 weighed even up to 25 pounds (nearly 12 kg).
The absolute measurements were carried out in the Imperial gardens near the Prague castle, in a place sufficiently distant from any buildings. As there was no hut or shelter at the beginning, the measurements were carried out in shine windless days 70 in order to eliminate the influence of wind on the rod oscillations.
Regular magnetic observations were started in July 1839. In the first year the observations were performed 17 times per day, however, their frequency soon decreased to 10 measurements and later to 5 and 3. Simultaneous measurements at specific intervals (term days) in the frame of the Göttingen Magnetic Union were performed up to 1849. In the first decade also measurements with the frequency of 2 minutes were carried out during periods of magnetic storms. 75 It was important that from the very beginning all measurements were published in the yearbooks called Magnetische und meteorologische Beobachtungen zu Prag. It was the best way how to save all data without a danger of lost by fire, flooding or incompetency and disregard of the future staff. The yearbooks contain tables of variation observations (magnetic and meteorological), reports on absolute magnetic measurements and discussion to their conversion to physical units. In the first volumes also observations of vegetation were included (development of sprouts, leafs, flowers, etc.). Most observatories operating within the Göttingen Magnetic Union were closed already in the 1840's or 1850's. Just a few observatories established before 1850 were in operation up to the year 1900 or later. According to the information about be usable for data digitization. The manual digitization was thus carried out by means of spreadsheets with pre-programmed templates that allowed also for preliminary data check and repair of rough errors: computed monthly means were compared with monthly means published in the yearbooks. All declination and horizontal intensity data of regular observations have been already digitized. The digitization of data from disturbed periods will follow.
As the above-mentioned calibration of sunspot numbers by Wolf and Wolfer shows, geomagnetic observatory data can serve 100 as a proxy of space weather parameters. Old geomagnetic records can thus provide invaluable information for space weather studies long time before the launch of artificial satellites. It is known that the most severe geomagnetic disturbances, including the well-known Carrington event in 1859, occurred during the historical solar cycles. Searching for geomagnetic disturbances in the past and their analysis is important for understanding the causers of extreme geomagnetic activity. At present, this kind of research can be beneficial for setting up reliable space weather forecast models. It is known that besides of the the ring-current 105 storms there are even more violent geomagnetic field variations called auroral substorms. In (Valach et al., 2019), the data from Clementinum observatory were used to study strong geomagnetic disturbances which were interpreted as substorms.
Structure of the paper is as follows. In Section 2, we present the database consisting of the Clementinum data. In Section 3, we deal with a-few-year period of corrupted declination record. The key part of the paper rests in Section 4 which is devoted to the study of horizontal intensity registered at Clementinum using a bifilar magnetometer. An example of interesting magnetic 110 disturbance is presented in Section 5.

Database of the Clementinum magnetic data
The magnetic data published in the yearbooks Magnetische und Meteorologische Beobachtungen zu Prag include absolute observations, regular variation observations, more frequent observations during magnetically disturbed conditions and simultaneous observations on term-days agreed in the frame of Göttingen Magnetic Union. The first four volumes cover the first four 115 subsequent periods: from July 1839 to July 1840, from August 1840 to July 1841, from August 1841 to July 1842, and from August 1842 to December 1843. All other volumes coincide with the calendar years.

Absolute observations
As the Clementinum building, where the daily variation observations were carried out, was not situated in a completely ironfree environment, the absolute observations were performed in the Emperor Garden near the Prague Castle. At the beginning, 120 the observations were held in the open air, later a wooden hut was built. Declination, inclination and horizontal intensity were measured. The latter one followed the procedure developed by Gauss (1833). The absolute measurements started as late as estimate the base value and size of the scale unit.

Regular variation observations
The variation magnetometers were installed in a 4.5 m wide corridor below the Astronomical Tower of Clementinum. The instruments including telescopes for scale readings were spaced out in such a way that one observer was able to carry out measurements of declination, horizontal intensity and inclination within 2 minutes. Regular variation observations started 130 in July 1839 by measurements of declination and horizontal intensity. Two months later, measurements of inclination and oscillation period of inclination needle were added, the latter providing some information about the total field. In the first year, 19 observations were carried out per day and later the numbers of observations were gradually reduced.
Time stamps in the yearbooks show Göttingen astronomical time. Compared to Prague astronomical time, the difference is 18 minutes. According to astronomical convention, 0 hour means noon and 12 hours midnight. More precisely, declination 135 is measured on the hour, horizontal intensity 2 minutes later and inclination again 2 minutes later. The oscillation period of inclination needle is observed before and after the above mentioned measurements. The summary of magnetic variations is given in Table 1. The time of measurements in Table 1 corresponds to the Göttingen "civic time", i.e. 12:00 corresponds roughly to 11:20 UT.
As the magnetization of the needle is temperature dependent, the temperature inside the case with needle was also recorded, 140 at the beginning twice a day, later during all measurements. Kreil was aware that the temperature dependence of the needle as well as the issues of the geomagnetic observations in general were not yet sufficiently understood and that is why the raw data in scale units were published. The data in physical units were published from Vol. 33 and publication of scale unit data was stopped in Vol. 45. We have decided to digitize first the declination and horizontal intensity data. Transformation of the data available only in scale units to physical units will be reported in the next sections of this paper. The complete data set of 145 declination from 1839 to 1917 and horizontal intensity from 1839 to 1904 is provided in Supplement.

Records of magnetic disturbances
If the observers on duty noticed exceptionally fast change of declination or horizontal intensity, they started observations of these components with higher frequency. The interval between measurements was in range from 24 seconds to 10 minutes with a typical period of 2 minutes. About 150 events were recorded from September 1839 to December 1843, which represent 200 150 full pages in the yearbooks. In the preface to volume 5, Kreil assessed hitherto known practice and came to the conclusion that only sufficiently large disturbances are worth to be included into the reports in the future. These high-frequency measurements were stopped in 1851 after Kreil left Prague to establish the Central Institute for Magnetic and Meteorological Observations in Vienna. However, the information about disturbed days was not fully omitted. Some brief reports on magnetic storms were published in monthly or yearly summaries. The yearbooks include also observations of northern lights. They are scattered 155 between above mentioned lists of disturbed days and observations of atmospheric phenomena. About 20 observations in Prague between 1839 and 1905 were recorded. Thanks to the dark night sky in the 19th century, the polar light was observable even in the city downtown. Information on the polar lights appearance across Europe and North America was also published.

Term-days observations
As mentioned in the Introduction, Prague observatory shortly joined the simultaneous observations of the magnetic field on  3 Corrupted data of magnetic declination around the year 1852 Figure 1 with the time series of magnetic declination clearly shows the secular variation (a systematic trend in which the orientation of the horizontal projection of the geomagnetic field vector shifts to the east) as well as a distinct seasonal variation (deviations resembling sinusoids). However, in the period that began sometime in early 1852 and lasted for about two years, the course of the declination had a different character than during the rest of the period. This special period begins with a sudden 175 decrease in the size of the declination by about 20 angular minutes; subsequently the time series continues, as if without secular variation. Such a course of the magnetic element is unexpected and highly probable is also incorrect.
Indeed, the yearbook for 1852 (Böhm and Kuneš, 1855, p. IV) also paid attention to this peculiarity. It presents comparisons of the absolute value determined by absolute measurements with the values determined from the variation apparatus using the formula from the previous year (i.e. from 1851). These comparisons show that the difference between these values has been 180 steadily increasing almost systematically; in June 1852 the difference ranged from 4 to 5 angular minutes (with a negative sign), in December it was already more than 12 angular minutes (again with the minus sign). Böhm and Kuneš (1855, p. IV) wrote: "Over the next year, this difference widened without being able to determine its cause, and only later did it become apparent that a spider, which was in the box where the small magnet of the variation apparatus was hung, caused these differences." The variations in the yearbook in 1852 and in the first part of 1853 were thus contaminated, likely by a spider web which in 185 some way mechanically affected the position of the magnet in the device. There is probably no way to remedy these variations, because the spider could have rebuilt its web in a way we have no way to determine. Therefore, we cannot reliably determine the annual averages for the years 1852 and 1853 from those variations. After examining the table comparing the absolute measurements made in June and October 1853 with the absolute data determined from the variations (Böhm and Karlinski, 1856a, p. III), it seems that for June and October 1853 there are no more systematic deviations between absolute measurements 190 and variations; although there are some differences, but they do not seem to be systematic but rather random, with different signs. Therefore, we assume that the variations in the second half of 1853 (including June) are already correct.
However, despite the contaminated variation data, we could still obtain some substitute values for the annual averages of absolute declination for 1852 and 1853, which could contain valuable information; more valuable than mere interpolation from existing surrounding annual averages.

195
Although the absolute measurements of 1852 were not found in contemporary sources, there are available the records of differences between the absolute values from variations and from direct measurements (Böhm and Kuneš, 1855, p. IV), those which we have already mentioned above. These records were captured on the following days: 22 and 23 June, 6 and 7 July, 21 and 22 October, and 24 December of 1852. Using the formula for converting declination variations to absolute declination and the coefficients which are valid for 1851 together with the records of those calibration measurements, we may easily reconstruct 200 the results of the absolute measurements.
The absolute measurements of declination in June and early July can be interpreted as measured during the summer solstice, the measurements in October are closest to the autumn equinox, and the December measurements are almost exactly at the time of the winter solstice. Figure 2 compares the absolute values in these three time points to the monthly averages in the surrounding periods. We admit that the comparison is not entirely perfect, because we have compared the averages of absolute 205 measurements with the monthly averages at 10.00 pm (they should be the least affected by daily variation), but the absolute measurements were made between 8.00 am and 10.00 am (local solar time). Even so, we can conclude that the values fit well into the time series. This, to some extent, justifies us to declare the average of the declination calculated from these three instances to be a substitute for the annual average of the declination in 1852. It gives 14 • 17.29'. Note that declination at that time was western in Prague. In accordance with the convention in which the positive direction for the declination is eastern, a 210 negative sign should be added to the then value of declination in Prague.
A substitute value for the annual average of 1853 can be calculated from the monthly averages from June to December 1853, which can be found in the yearbook (Böhm and Karlinski, 1856a). The formula obtained from the measurements in the following year (Böhm and Karlinski, 1856b, p. III) can be used for these calculations (Böhm and Karlinski, 1856b, p. III). The resulting absolute value of the western declination for this year is thus 14 • 10.49'.

215
The database on the website of the Institute of Geophysics of the Czech Academy of Sciences also preserves the original variation data of magnetic declination affected by the spider net. Being aware of the limitations of these data, they can still be considered useful for studying rapid variations such as magnetic storms, at least for the purpose of qualitative analysis.
the Prague Clementinum.The methodology of these measurements was invented only a few years before, the absolute method in 1832 and the method for the variations observations in 1837 (e.g., Garland, 1979). Probably the novelty of this issue caused some ambiguities or imperfections in the first records of these measurements in the Clementinum yearbooks. In this section we describe some of the problems we encountered while studying the oldest Prague records during the preparation of this article.
Solving these problems requires a basic knowledge of the operation of a bifilar magnetometer, which is clarified below.

The equation of the bifilar magnetometer
The bifilar magnetometer, the type of an instrument that was used for to perform the observations of the horizontal intensity variations in the mid 19th century (Gauss, 1838), has recently been reminded to the geomagnetic community by Garland (1979) and Nevanlinna (1997).
The main part of the instrument was a large magnetic needle that was suspended by two long fibres (a typical length of the 230 fibres was more than a metre). The fibres maintained the horizontal position of the needle and at the same time allowed the needle to turn in the horizontal plane; as they were very thin and hanged close to each other (a typical distance between them was a few of centimetres), they put only a small resistance against the rotary horizontal movement. The length of the needle was about one meter and its typical weight was of the order of two kilograms. The console from which the fibres hanged could be revolved around the vertical axis (Fig. 3).

235
During the initial adjustment of the instrument, the console was revolved so that the magnetic needle came to a position perpendicular to the magnetic meridian. The console was locked in this position.
When there was a small change of the magnitude of the magnetic force, which balanced the torque caused by the torsion of the pair of fibres, the magnetic needle slightly deviated from the perpendicular direction in the horizontal plane. In a small mirror, placed on a magnetic needle at the axis of rotation, the reflection of a scale was observed with a telescope with an index 240 line in its objective. It was a scale that was fixed and was placed, for example, on the wall of the room. This made it possible to determine by which angle the needle deviated (Fig. 4).
The change in magnetic force could be due to either a change in horizontal intensity or a change in needle magnetization due to a temperature variation. It can be shown (e.g. Nevanlinna, 1997) that the balance between the torsion of a pair of fibres and the rotational effect of a magnetic force yields the equation However, in order to be able to use such a variation device, it is necessary to know the scaling coefficient c 1 and the temperature coefficient c 2 .
In the yearbooks from the Clementinum Observatory in Prague, the values of constants c 1 and c 2 have been explicitly given replaced by another apparatus. Therefore, two partial problems we had to solve before the equation (1) could be used:
In the following section (Sec. 4.2) we will report on the solution of these two tasks. in the yearbooks and we consider its values to be correct.

270
A similar task, which is to find the temperature coefficient from values of the quantities h and t, was solved for observations at the Helsinki Observatory by Nevanlinna (1997). However, in the case of Helsinki, it was typical that the temperature variations during the day were very large. It was probably due to that their observation room was apparently unheated and insufficiently insulated from the outside environment. This allowed Nevanlinna (1997)  Because we know that the variations in h caused by the change in needle magnetization were much greater than the variations due to the actual change in magnetic field, we might use a similar approximation as Nevanlinna (1997) on a time scale of several months to years. Also in our approximation, we considered changes in h only as a consequence of changes in temperature t, but unlike (Nevanlinna, 1997) they were not the daily variations of the values of h and t, but the seasonal variations. This allowed 285 us to use linear regression in the same way to estimate the temperature coefficient.
We also used another simplification, namely we considered only the temperature dependence for the magnetization of the needle. Like Nevanlinna (1997), we did not consider that the magnetization of the rod at any fixed specific temperature could change over time during the studied period. Although the yearbook (Böhm and Karlinski, 1857, p. XV) stated that the magnetism of the rod (needle) "weakened" over time, we think it was an interpretation based on incorrect data. According to our 290 reasoning, the authors of the yearbook confused the physical units in the data on which they based their interpretation. In our opinion, the correct data on the temperature coefficient should be as stated in The values in Table 2 do not indicate any systematic increase or systematic decrease in the c 2 coefficient; we think that the reason why the listed values differ from each other is rather the inaccuracy with which those values were determined.
Therefore, we prefer to assume in our study that the c 2 coefficient, which stems from the temperature dependence of the needle magnetization, was still approximately the same value for the same needle.
However, from information in the yearbooks, it can be concluded that two needles or two bifilar devices were used. The first 300 bifilar apparatus was in operation from the beginning of the magnetic measurements in Clementinum until 31 December 1845.
On 1 January 1846, another apparatus, which was working with another magnetic needle, started to be employed. It means that two values of the c 2 coefficient have to be determined: the first of them being valid until the end of 1845 and the second one being valid from the beginning of 1846.
Regarding the first case (i.e., for years 1840-1845), we found no specific mention about the value of coefficient c 2 in the 305 yearbooks. Therefore, in Section 4.2.1, we calculate it as a new and hitherto unknown numerical value. In the second case, for the period from the beginning of January 1846, three values of the coefficient were given in the yearbooks (we have listed them in We estimated the value of the temperature coefficient for the period from 1 January 1840 to 31 December 1845 for six isolated time periods, between which, according to notes in the yearbooks, the setting of the bifilar apparatus was changed. The readjustment of the apparatus was probably performed for two reasons: either the fibres on which the needle has been suspended have ruptured, 1 In some places, there are errors in the yearbook when the value in parts of the whole horizontal intensity is given instead of the value in absolute units.
The error can be eliminated by multiplying by 1.9, which is the value of the then horizontal intensity in absolute units of the system mm-mg-s introduced by C. F. Gauss. needle systematically increased to the limit of the measuring range of the instrument.
Using coefficient c 1 , the values of which we read from the yearbooks, we converted the variations given in the scale divisions to the numerical values given in the parts of the whole horizontal intensity.
Subsequently, we calculated temperature coefficients for individual periods using the method of least squares. In doing so, we determined the parameters of the function In this linear equation, c 2 is the temperature coefficient sought and c 3 is the parameter, which determines the value of the fraction ∆H/H at the temperature t = 0 • R (the specific value of c 3 is of no interest for this study). The obtained values of the temperature coefficient for the period 1840-1845 are given in Table 3. According to these data, the mean of the temperature coefficient is Since we do not have another estimate of the temperature coefficient for the very first years of observations in Clementinum, we will use this numerical value in further calculations (Sect. 4.3).

Verification of the temperature coefficient of the bifilar apparatus from 1846 to 1854/55
Even from 1 January 1846, when a new bifilar apparatus began to be used for observations of horizontal intensity, until the 330 end of 1854, the values of the temperature coefficient were not given in the yearbooks. In the 1855 yearbook, the temperature coefficients were back-calculated, but due to uncertainties about the physical units in which their results were given, we considered it necessary to verify them with our own calculations. We proceeded in the same way as in Section 4.2.1.
For the period from the beginning of 1846 to 8 March 1855 we had five continuous time periods at our disposal, for which we determined the temperature coefficient satisfying equation (2) Table 4.
The mean value for the temperature coefficient calculated from the data in Table 4 is The value that we obtained is in good agreement with the corrected data stored in Table 2. After this verification, the numerical values in Table 2, or rather their mean value, may be taken as the temperature coefficient for the whole period from We will use this value in Section 4.3 in our calculations for the period up to the time when the values of the temperature coefficient starts to be provided in the yearbooks (in the headers of tables that contain the variations given in scale divisions).

A baseline for the horizontal intensity in years 1840-1854 and the annual means
At the beginning of the geomagnetic observations at Clementinum, two types of measurements were performed: occasional measurements of the whole horizontal intensity by means of the Gauss absolute method, regular observations of the variations of the horizontal intensity (expressed in parts of the whole horizontal intensity).
These two types of measurements were linked only by the fact that observers needed a value of the total horizontal intensity to express variations in horizontal intensity in absolute units. However, in contrast to nowadays practice, the baselines of the 350 variation apparatus were not calculated afterwards. Therefore, in this section we will calculate the baseline for the period from 1840 to 1854. Since 1855, the yearbooks provide the baseline values (under the designation "Constante").
To calculate the values of the baseline, we followed the procedure that became the standard method today, i.e. we compared the trends in the time series of registered variations of horizontal intensity (only the pure variations with no baseline added) with the values of horizontal intensity determined by absolute measurements. The time series of the registered variations with 355 no baseline added were calculated for each of the continual periods separately (Tables 3 and 4). Using the tabulated variations, which are provided by the yearbooks in scale divisions, together with scaling coefficient c 1 and temperature coefficient c 2 , we obtained the values for the variations of the horizontal intensity expressed in the absolute units. Other necessary data are the results of absolute measurements of the horizontal intensity. Figure 6 displays the data on the absolute measurements that we found in the yearbooks. A comparison with the course of the data of the Munich Observatory, which is only 300 km away from Prague, shows at first sight that the absolute measurements before 1844 deviate from what we would expect. A confrontation with the IGRF model also shows that the absolute measurements in the first years of geomagnetic 370 observations in Prague were probably characterized by a disproportionately large error, likely systematic observational error.
Therefore, for this period, we preferred to use the value of horizontal intensity, which we obtained by extrapolation based on a linear regression from data between 1844 and 1853.
The final baseline that we then obtained is presented in panel (c) of Fig. 5. We also need to mention that to determine the baseline we used the variation data that were observed at 04.00 pm, because absolute measurements in the first years of the observatory's operation were usually performed around 04.00 pm, which is information found in the yearbooks.
By adding the baseline values to the time series of the variations, we obtained time series of the absolute values of the horizontal intensity (Fig. 5, panel (d)). From these data, we then calculated annual means of the horizontal intensity for Prague (see data in Table 5). The comparison of the annual means that we obtained (Fig. 6) with the annual averages from Munich indicates a good agreement in the trend of the time series at these two not-too-far-away observatories. We believe that this good 380 agreement can be considered a confirmation of the correctness of the calculations we made in Sections 4.2 and 4.3.
5 An example of recorded strong magnetic disturbances: the event on 3 September 1839 In addition to obtaining the time series that capture the secular variations of the geomagnetic field, the study of historical records from old observatories has another, at least equally important, benefit. This benefit is the records of intense magnetic storms that occurred in the 19th century. As an example of such a record, we show here the course of the horizontal intensity 385 during an interesting magnetic storm, which commenced in the evening of 3 September 1839. Subsequently, after midnight and in the early morning time, two sudden, relatively deep, and short-lasting drops in horizontal intensity appeared in the Prague records. Probably these two swift variations were local phenomena, probably substorms or some other variations that were caused by electric currents in the auroral oval. Here we assume that the auroral oval was reaching as far as Central Europe at that time. Our assumption that the auroral oval was located in relatively low geographical (or magnetic) latitudes around  Table 6). The difference between local times of these two places is thus about one hour. Although this particular geomagnetic disturbance and the accompanying aurora borealis are not well known to the scientific community today, Snow (1842) states that the press at the time informed about this aurora.
The aurora borealis accompanying the magnetic storm of 3 and 4 September 1839 was also observed in Prague. However,

400
Kreil (1840b) writes that they could not pay full attention to the observation of this phenomenon because they measured magnetic field, namely declination and horizontal intensity, all the time.
Other examples of interesting geomagnetic disturbances observed in the Clementinum are the extreme variations of the magnetic field that occurred on 17 November 1848 and 4 February 1872 (Valach et al., 2019). We believe that the digitized records from the Clementinum observatory of Prague will provide an opportunity to study other interesting events.
The presented study can be primarily conceived as a presentation of the unique database collected from geomagnetic registration at the Clementinum Observatory in Prague during the 19th century. From another point of view, this study can serve as an example of how to deal with historical geomagnetic data in context with current research trends in geomagnetism. It is also documented here, that putting the records in proper physical units and coping with data inconsistencies is an essential part of 410 the research.
In dealing with proper scaling of the historical geomagnetic records, this study partly builds up on earlier work by Nevanlinna (1997). The major outcome of these efforts is the proper adjustment of thermal coefficient in the equation for bifilar magnetometer. Having this constant determined, it is possible to re-scale the data in geomagnetic records and to reconstruct the time series for horizontal intensity. As such, the adopted geomagnetic records can be cast in consistent forms to serve as a 415 reliable source of information for research purposes. As detailed in the main part of this paper, the procedure of computing the temperature dependence here in some sense can be considered as complementary to that in (Nevanlinna, 1997).
Closer study of the early Clementinum data reveals an interesting magnetic disturbance which is nearly unknown and almost forgotten among the present-day geomagnetic community. There is an indication that this disturbance can be interpreted as caused by currents related to the auroral oval transiently extending to mid-latitudes. We proposed that this disturbance could 420 be identified as two subsequent magnetic substorms.
Our database has a potential of adding valuable information to hitherto known global geomagnetic data resources. Recently, for instance, the HISTMAG database was created by Arneitz et al. (2017), consisting of large amounts of paleomagnetic and archeomagnetic data as well as a rich collection of instrumental historical observations. Compiling the historical data from several observatories can be helpful in dealing with inconsistencies or missing data. The study of multiple records is also 425 necessary for a more global view of a considered geophysical phenomenon, in contrast to the study of data from a single location on the Earth's surface. For example, in (Valach et al., 2019), the Clementinum data together with the geomagnetic records from other observatories have been used to analyse selected strong geomagnetic disturbances.
In a broader context, the scientific benefit of the presented work can be twofold. First, using the properly scaled series of geomagnetic records, the long term behaviour of the geomagnetic field at the Clementinum observatory can be reconstructed.

430
These data along with historical data from other observatories combined with some supplementary (archeomagnetic, paleomagnetic) data can be useful for modelling the global secular variation. Second, the consistent historical magnetic record supported with some other sources of data (e.g. records of aurorae) can be searched for extreme geomagnetic disturbances (magnetic storms and substorms). This kind of approach can make contribution to the study of extreme space weather events.     These moments correspond to magnetic local times 01:39 MLT and 04:59 MLT, respectively. Judging from their shapes and from the fact that they happened at night, these two swift disturbances might probably have been substorms. The figure also demonstrates the procedure with which the magnetic storms were commonly registered that time: i.e. shortening the time interval between readings of the values from the variation instruments immediately after an interesting magnetic variation commenced, as described in Section 2.3.  I 5, 6, 7, 8, 9, 10, 10:30, 11:30, 12:30, 13, 13:30, scale units 14:30, 15:30, 16:30, 18, 19, 20, 21, 22 1 Jan-Jul 1840 D, H, I 0, 2, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19 increasing urban noise ∆D = difference D(t) -D(t -5 min), and similarly for ∆H