Correction of the Tokyo geomagnetic data in Meiji era

Technical Report of the Kakioka Magnetic Observatory, Vol. 2, No. 2, pp. 1-49, December, 2004

Takeshi Toya, Takashi Koide and Akio Yoshida

Abstract

Before the Kakioka Magnetic Observatory commenced the activities in January 1913 in Kakioka, Ibaraki Prefecture, continuous observation of the geomagnetic field had been carried out from 1897 until 1912 in the premises of the Central Meteorological Observatory in Tokyo. Although magnetograms and observation notes that had been kept in the Central Meteorological Observatory were burned up when the city was severely damaged by the 1923 Great Kanto Earthquake, printed yearbooks were fortunately left in the libraries of the Kakioka Magnetic Observatory and the Japan Meteorological Agency. Recently, the Kakioka Magnetic Observatory digitized the hourly data over the whole 16 years and released them on its own homepage. While doing this, we were looking to see whether or not there might be any errors that could have occurred in transferring the data from the observation records to the yearbooks or any errors that occurred during typesetting. We realized in that process that there was a large number of apparently obviously mistaken data items mixed in when the original data were printed in the yearbooks a. At the first release of the digitized data, the original data as listed in the yearbooks was given more importance and was released uncorrected, just as the data had been. However, we considered that obvious mistakes in these valuable geomagnetic data of the Meiji era should be corrected for easy use in various research works, and so we decided to revise them entirely this time. In principle, we restricted corrections to only errors judged to be obvious mistakes. Whether to revise each datum or not relies on a judgment made in the following way:

An immediate judgment that a figure in the 100s digit of an hourly value is a typing error is acceptable when, for example, only one datum with a value in the 900s range is contained mixed in with other values in a day when all the other values are in the 600s range. In this case a judgment is made that a "6" was mistakenly changed to a "9." The validity of this judgment was confirmed, checking that the mean value calculated using the revised value agreed with the mean value written in the yearbook.
In order to find mistakes in the figures, we referred to the above procedure used in (1) to confirm the validity of the revision. (Checking whether the mean value written in the yearbook agrees with the mean value re-calculated using the digitized data.) If a disagreement is found in both the mean values of 24 hourly data in a day (lateral mean values) and the monthly mean values of the data at the identical time of each day (vertical mean values), it is suggested that the magnetic field value at the day and the hour where the lateral row and the vertical column cross might be mistaken. In that case, we checked whether the disagreement is resolved by changing the hourly value at the cross section. Only one figure, not two, was changed within either the 100s, or 10s, or units digits to change the values. Further, alteration of the units digit value was restricted by the criteria described in the following.
When the difference in two kinds of mean values is less than about 0.3, we considered the data should not be revised. The data was not revised also even when the difference was 0.4 or 0.5, unless it was estimated that obvious mistakes were almost certainly made, judging from variations in the magnetic field on the day.
In such cases that there are multiple locations in both the vertical and lateral lines where the two kinds of mean values described above are different, so that it could not be specified which column data should be revised, data were not revised (were not able to be revised) even if the difference in the mean value exceeded 0.5.
Even when the above disagreements of two kinds of mean values could be resolved, the data were not revised in the following cases: when the data were unnatural from the viewpoint of the magnetic field variation trend on that day (that is, the maximum or minimum value is estimated to occur on that day at the nearest hour, but the revised value deviated greatly from those values, or that day was recorded as "calm," but the revision, if applied, would cause a large variation), or when any revision would cause a deviation from the maximum or minimum ranges on that day. We did not revise data in those cases because we judged that the hourly value itself written in the yearbook was correct, but the mean value in the yearbook was mistaken (mistakes during checking or reading figures at mean value calculation).
When the 10s or 100s figures are lacking due to unclear printing, we inserted a reasonable figure in the spot, judging from the values before and after the hourly value. In most cases, we confirmed that this was effective to have the two kinds of mean values agree with each other.
After revision, we drew a magnetic change map for each element every year to find some locations showing abnormal changes, such as a pulse wave or a rectangular wave. Investigation of this abnormal change reveals that the mean value was not calculated, because of a lack of measurements, resulting in missing mistakes in the 100s figures. Such data were corrected by estimating the correct value from magnetic field values before and after the abnormal changes.

   Although the revision was limited to the hourly value data that was judged to be obviously mistaken, the number of revisions exceeded 900 locations. We show the entire list of the places revised and their contents. Besides that, we also attached the list of places where we decided not to make a revision, because it was impossible to specify the date and hour to be corrected or we judged it better not to revise them due to some reasons. For these locations, we think, there remains a possibility to find out in the future that our judgment not to revise was in fact not correct. In addition to the above-described two cases, there were many places where we just looked at the plotted magnetic change map and had a doubt if the change was actually observed. They show changes over several tens of nT in a calm day or changes over 100 nT during a few days. This is not understood as a natural change in the magnetic field, but is left uncorrected as it is, because we did not find any mistakes which were taken up to be revised in this work. We listed those places as well, hoping for progress of study in the future.
   Comparison of figures drawn using the original geomagnetic hourly values with those made from the revised data indicates that this revision was effective to clean up a lot of data noise due to artificial mistakes. The H and Z components, however, have significant large drift variations at some areas. A few hints to consider the abnormal changes are noted, but clarification of the cause is left to future studies. We can take in geomagnetic variations such as magnetic storms from the figures. In fact, a total of 307 bromide analog copies of magnetic storms were inserted in the yearbook. The bromide records are very valuable, but they are of such a large volume, difficult to contain in the paper, that only a list of the records is given.
   Assuming that the local magnetic difference between Tokyo and Kakioka that was measured in 1916 is constant, the data were converted to Kakioka's value to plot the yearly mean values with Kakioka's data since 1924. Although the Z component is outstanding from 1909 through 1911 because of drift changes that cannot be considered natural, the data continuity of each component is rather good. In particular, it is excellent in the D component. Therefore, it is no exaggeration to say that this revision has made it possible to discuss long-term geomagnetic changes near Japan based on continuous observation data over one century. We hope that the published data will be useful in research for long-term geomagnetic changes, and annual variations of magnetic storms in Japan.

Received 13 October 2004; Received in revised form 1 December 2004; Accepted 6 December 2004