583 The 27-Day Period in Magnetic Phenomena .
By C. Chree , Sc. D. , LL. D. , F.R.S. ( Received June 6 , \#151 ; Read June 25 , 1914 .
) S 1 .
In two earlier papers , termed here for brevity SMi* and SM2 , f I have discussed data supporting the existence of a period of about 27 days in magnetic phenomena .
The results have , I think , put it beyond question that there has been\#151 ; at all events since 1890\#151 ; a decided tendency for the magnetic character of a day , whether disturbed or quiet , to resemble that of a day either 27 days earlier or 27 days later more closely than it does that of the average day .
There is no period\#151 ; or , if there is , its amplitude is very small\#151 ; in the usual sense of the word , the phenomena do not go on repeating themselves in an indefinite sequence like the phenomena of day and night .
Various ways suggest themselves of prosecuting the enquiry further , with a view to ascertaining its connection with other natural occurrences , especially with sunspots or other solar phenomena .
Having at hand a long series of magnetic records , the method that appeals to me as a practical man is the utilisation of these records .
The next question is how to use them to best advantage .
If the 27-day period , as we may term it for brevity , is directly associated with sunspots , we should expect it to be influenced by the difference between the solar latitude of sunspots in years of many and few sunspots .
My previous work has shown that the period , though ordinarily far from insignificant , may sometimes be little in evidence .
There is much that in the present state of our knowledge must be regarded as " accidental " in magnetic phenomena , and a few " accidents " conspiring together may render the 27-day period unapparent in an individual year .
Thus , for an enquiry into the nature of the relationship , if any , between the period and sunspots , or any other solar or lunar phenomenon , data are wanted from a period so long that accidental phenomena will be largely eliminated from the groups of years representative of any one given set of conditions such as sunspot maximum or minimum .
As a preliminary to any such general enquiry , the measure of magnetic disturbance to be adopted claims consideration .
The measure adopted in SM2 , and also mainly in SMi , was the " character " of the day on the international scale 0 ( quiet ) , 1 ( moderately disturbed ) , 2 ( highly disturbed ) .
Since 1906 mean " character " figures taken out to 01 , based on * 1 Phil. Trans. , ' A , vol. 212 , p. 75 .
t ' Phil. Trans. , ' A , vol. 213 , p. 245 .
Dr. C. Chree .
the data from a large number of stations , have been published at De Bilt .
These enable the days of each month to be arranged easily , and probably with adequate accuracy , in the order of their disturbance .
It is , however , impossible to say how the day whose character in the international lists is , say , 05 compares with the day whose international figure is , say , 1*5 .
If one takes the " activity " of the day as defined by Prof. Bidlingmaier , which represents practically the integral of the squares of instantaneous departures from a normal value , the day whose international figure is 1*8 is much more than twice as disturbed as the day whose international figure is 0*9 .
There is a good deal to be said for " activity " as a measure of disturbance , but activity data are practically non-existent , and deriving them for any considerable period of years would represent an enormous deal of labour .
A second objection is that there are a few days whose " activity " is so enormously greater than that of the average day that an " accidental " occurrence of two of these might introduce a false period into the records of even a number of years .
International " character " figures , if available for years prior to 1906 , would have much to recommend them .
Their unprejudiced nature is above suspicion , and their aid enables one to separate out at once the five most and the five least disturbed days of the month , five being a number which is generally recognised as at least convenient .
But international data do not exist prior to 1906 , so one must use something else .
In SMi , in the main , I took as a measure of disturbance " character " figures 0 , 1 , 2 , which I allotted myself from consideration of the Kew curves alone .
One can allot " character " figures on this scheme to the days of a whole year in a couple of days or less , and the summation of columns containing only 0 's , l 's , and 2 's is simplicity itself .
Naturally , however , one has more than five 0 's and fewer than five 2 's a month .
Thus characterisation from a single station does not lend itself to the selection of a fixed number of days of a particular class .
In SMi , dealing with the years 1890-1900 , I considered only the five most disturbed days of the month , and took the amplitude of the horizontal force range as the criterion of disturbance in their selection .
Treated in this way the years 1890-1900 , as a whole , showed the 27-day period as unmistakably as did the years 1906-1911 dealt with in SM2 .
I felt , however , at the time that the hybrid nature of the method\#151 ; i.e. the selection of the days by reference to the horizontal force range , but the measurement of the disturbance by reference to the " character " figure\#151 ; might arouse criticism , and partly for that reason , partly for my own satisfaction , I considered for two years , 1894 and 1895 , the result of taking the range of H as itself the measure of the disturbance .
I found that in 1895 the 27-day period was The 27-Day Period in Magnetic Phenomena .
" as unmistakable in the range as in the character figures/ ' but " in 1894 the character figures and the range figures both show not a peak at day w + 27 or n -f 28 ( ?
* , being the representative selected disturbed day ) , but a high plateau from about day n-\- 25 to day w + 34 .
The range figures , in fact , would seem to favour a Schmidt or 30-day period."* The substitution of range figures for " character " figures added largely to the arithmetical labour , and I did not prosecute the enquiry further .
It seemed , however , desirable to undertake this labour now , to see whether the method was likely to prove a satisfactory one for dealing with years prior to 1890 .
S 2 .
Daily ranges at Kew for D ( declination ) , H ( horizontal force ) , and Y ( vertical force ) being now at my disposal , it seemed desirable to obtain results from all three elements , so as to compare their relative advantages for the object in view .
The five days of largest daily range were taken as the five most disturbed days , the D , H , or Y range serving as the criterion of selection according to the element which was being used as the measure of magnetic disturbance .
The five days a month which the Astronomer Eoyal had selected as representative quiet days were accepted as the five quietest days .
The latter choice , I dare say , might be improved on for purposes such as the present .
The desirability of securing that the mean time represented by the selected quiet days should fall near the middle of the month exerted , I believe , some influence on the Greenwich choice .
In a good many cases the selected days are obviously not the quietest of the month , and if one cared to spend a lot of time , a choice somewhat superior for the present purpose could doubtless be made .
But the Astronomer Royal 's days have at least the great advantage of being free from any suspicion of bias .
A choice of quiet days based on the diurnal range alone has several disadvantages .
The differences amongst the ranges of the 10 or 15 quietest days of the month are usually not large .
In many cases there would be several ties amongst the days competing for the choice , and the difference in range between days selected and days not selected would frequently be of the same order as the probable error of measurement .
This is especially true of H and Y , where temperature corrections are usually necessary to the measured curve ordinates .
No continuous record of temperature in the magnetograph room had been obtained at Kew until 1894 .
Temperature corrections to the ranges were thus impossible for the earlier years , and under these circumstances it appeared best to omit such corrections entirely .
The uncertainty arising from the absence of a temperature correction to Kew H ranges is trifling , but this is unfortunately by no means true of Y. A choice of the five quietest days of the month based on * Loc .
cit. , p. 104 .
Dr. C. Chree .
the ranges , especially Y ranges , would be an uncertain one , especially at seasons when the magnetograph room temperature is altering most rapidly .
The absence of a temperature correction is seldom serious in the choice of the five most disturbed days of the month , because the ranges on the most disturbed days usually stand conspicuous above the others .
Still the fact remains that a few even of the Astronomer Royal 's quiet days figured amongst the five days of largest Y range , and their inclusion was more likely due to large temperature range than to real disturbance .
It is only proper to mention this , but it should at the same time be pointed out that irregular variations are very rare in Y as compared with H or D curves , and a quiet day has often a large regular diurnal variation .
S 3 .
Owing to the necessity of reducing arithmetical computation to a minimum , attention was confined to the ranges on days n \#151 ; 2 , n \#151 ; 1 , n , n-f 1 , n + 2 , and w + 25 to w-f-30 inclusive , n being a representative selected day , disturbed or quiet .
The five days n\#151 ; 2 to n-\- 2 served to show the nature of the preliminary " pulse " \#151 ; whether representing excess above the normal on disturbed days or deficiency on quiet days\#151 ; while the six days n-\- 25 to n -f 30 represented the secondary pulse constituting the 27-day period .
No days were selected for December , 1900 , as the corresponding secondary pulse would have fallen in 1901 .
This left 131 months , and so 655 selected days of either type , or 1310 in all .
Consequently , the total entries of daily ranges in the 11 columns numbered 14.410 for each element .
The D ranges were measured directly in minutes of arc to 0T ' , but the H and Y ranges in millimetres , going in each case to 0T mm. In the majority of cases , in the H and Y curves , 1 mm. represented 5 7 throughout the entire year , but in some cases changes of the order of 2 per cent , or more occurred in the course of 12 months .
For the purposes of the present enquiry , it was unnecessary to take account of small variations , incapable of affecting sensibly the relative values of the means derived from the successive columns of figures , thus the readings taken in millimetres were usually not converted individually into force , but were treated as if the equivalent of 1 mm. were always the same .
In the first instance , each year was considered separately , and the selected disturbed and quiet days independently .
The sum of the ranges from the 60 selected days ( 55 in 1900 ) appearing in each column was formed , and expressed as a percentage of the mean range from all days of the year multiplied by 60 ( in 1900 by 55 ) .
Tables I , II , and III give the percentage figures thus obtained for D , H , and V separately .
Table IY shows the arithmetic mean of the D , H , and V percentages .
The percentages were all calculated to 0*1 , but the last figure has been omitted , except in the final means .
The 27 -DayPeriod in Magnetic Phenomena .
Table I.\#151 ; D Ranges .
Percentage of Mean Absolute Daily Range .
Year .
Five days of largest D range and associated days .
n\#151 ; 2 .
n\#151 ; 1 .
n. n + 1 .
n + 2 .
n + 25 .
n + 2Q .
n + 27 .
n + 28 . !
n + 29 .
n + 30 .
1890 ... 103 116 153 116 102 99 106 106 104 107 Ill 1891 ... 105 124 172 127 115 302 115 114 123 115 106 1892 ... 91 136 196 129 104 93 93 108 108 103 98 1893 ... 99 116 155 118 104 99 101 104 109 104 102 1894 ... 94 123 193 132 108 92 100 100 103 107 116 1895 ... 100 115 165 123 110 97 106 122 117 120 113 1896 ... 103 127 180 141 118 108 116 121 122 116 99 1897 ... 89 122 176 118 95 98 110 114 111 98 108 1898 ... 102 138 187 129 103 105 106 121 122 122 106 1899 ... 97 124 175 123 108 105 121 120 113 101 99 1900 ... 99 115 155 106 99 97 102 96 95 93 103 Mean ... 98 *3 123 -2 173 -5 123*9 105 *9 99 *5 106 *9 111 *4 111 6 107 9 105 *6 .
Five selected quiet days and associated days .
Year .
n\#151 ; 2 .
n\#151 ; 1 .
n. n + 1 .
n + 2 .
n + 25 .
w + 26 .
n + 27 .
n + 28 .
n + 29 .
n + 30 .
1890 ... 104 91 78 94 104 92 93 94 95 97 87 1891 ... 94 82 73 95 100 99 90 92 90 104 113 1892 ... 95 79 70 93 98 94 104 105 92 91 87 1893 ... 93 86 76 92 100 91 91 95 105 102 102 1894 ... 81 81 69 94 120 103 92 91 97 93 87 1895 ... 100 91 68 90 108 96 90 89 105 102 108 1896 ... 86 76 66 90 105 87 79 73 78 86 94 1897 ... 91 84 67 90 94 90 78 95 94 107 101 1898 ... 91 82 67 81 97 95 88 80 82 90 96 1899 ... 84 80 69 88 108 87 87 90 91 97 105 1900 ... 100 89 81 96 112 96 95 100 95 93 102 Mean ... 92 -6 83 -5 71 *3 91 *2 104*3 93 *6 89 *8 91 *4 93 *2 96 *6 98-2 Dr. C. Chree .
Table II.\#151 ; H Ranges .
Percentage of Mean Absolute Daily Range .
Year .
Five days of largest H range and associated days .
n\#151 ; 2 .
n\#151 ; 1 .
n. n + 1 .
n + 2 .
n + 25 .
n+ 26 .
n + 27 .
n + 28 .
n + 29 .
n + 30 .
1890 ... 102 118 163 119 107 107 107 116 117 105 109 1891 ... 104 128 181 130 114 95 104 104 111 107 110 1892 ... 112 166 228 153 100 93 103 122 105 100 100 1893 ... 98 123 170 128 103 101 101 116 114 110 104 1894 ... 115 163 215 144 119 108 107 104 105 112 124 1895 ... 110 133 171 124 109 109 116 129 120 115 105 1896 ... 113 133 188 145 120 111 115 124 133 128 106 1897 ... 97 122 172 125 103 96 105 117 109 101 102 1898 ... 118 156 199 131 117 104 129 124 129 129 114 1899 ... 118 133 180 129 118 105 119 132 126 113 103 1900 ... 99 116 172 111 103 96 102 108 101 98 104 Mean ... 107-9 135 *6 185 *4 130-7 110 *2 102*2 109 *8 117 -7 115*4 no -7 107-4 Five selected quiet days and associated days .
Year .
n\#151 ; 2 .
n \#151 ; 1 .
n. n+1 .
n + 2 .
n 4- 25 .
n + 26 .
n + 27 .
n + 28 .
n + 22 .
n + 30 .
1890 ... 99 88 74 92 105 96 89 87 89 102 94 1891 ... 95 84 72 89 91 97 98 96 85 104 110 1892 ... 92 74 61 86 85 94 90 104 90 89 85 1893 ... 98 82 72 92 99 90 88 95 108 99 97 1894 ... 76 75 64 92 119 102 93 89 95 91 82 1895 ... 100 87 68 86 110 89 87 83 99 102 106 1896 ... 87 73 62 85 96 83 83 73 77 88 96 1897 ... 94 85 67 88 92 93 81 85 88 98 103 1898 ... 89 71 60 73 102 96 87 75 78 83 98 1899 ... 87 76 62 84 101 89 86 87 90 94 106 1900 ... 95 83 79 97 111 97 92 100 94 97 108 Mean ... 91 *9 79 -8 67 -2 87 *9 100 *8 93*2 88 -4 88 *5 90 *3 95 -1 98 *5 The 27 -DayPeriod in Magnetic Phenomena .
Table III.\#151 ; V Ranges .
Percentage of Mean Absolute Daily Range .
Year .
Five days of largest Y range and associated days .
n \#151 ; 2 .
n \#151 ; 1 .
n. n +1 .
n + 2 .
n + 25 .
n + 26 .
n + 27 .
n + 28 .
n + 29 .
n + 30 .
1890 ... 93 102 161 Ill 100 94 100 103 106 105 99 1891 ... 111 147 210 143 118 98 112 121 126 107 111 1892 ... 101 205 284 177 115 84 87 116 107 92 88 1893 ... 99 124 202 129 100 94 101 118 111 102 102 1894 ... 107 185 284 161 128 83 113 102 111 106 155 1895 ... 112 132 191 129 99 105 115 140 130 108 108 1896 ... 113 146 213 146 124 119 131 126 128 121 110 1897 ... 90 12 L 192 121 97 96 111 104 107 107 102 1898 ... 121 166 219 123 92 95 104 142 119 125 108 1899 ... 112 140 185 139 113 103 119 116 109 99 90 1900 ... 91 119 173 108 94 122 93 94 97 90 92 Mean ... 104 *5 144*3 210 4 135 *1 107-1 99 *5 107 -8 116 *6 113 *8 105 *6 105 *8 Five selected quiet days and associated days .
' Year .
n \#151 ; 2 .
n \#151 ; 1 .
n. n + 1 .
n + 2 .
n + 25 .
n + 26 .
n + 27 .
n + 2 8 .
n + 29 .
n + 30 .
1890 ... 99 91 90 92 108 99 99 89 95 98 97 \#171 ; 1891 ... 97 72 66 89 98 104 103 87 86 113 117 1892 ... 86 70 56 79 90 91 88 113 89 85 82 \gt ; 1893 ... 93 81 73 100 103 85 81 96 115 102 93 , 1894 ... 64 61 51 106 133 108 83 88 97 90 73 !
1895 ... 97 87 68 88 118 90 87 82 98 101 111 i 1896 ... 87 75 65 85 99 91 88 79 80 85 89 i 1897 ... 97 84 74 92 92 91 82 88 86 97 95 i 1898 ... 81 77 68 77 107 98 85 73 75 83 106 | 1899 ... 87 79 75 94 109 85 82 87 95 91 104 ; 1900 ... 93 86 90 93 112 95 93 105 102 99 102 Mean ... 89 1 78-3 70-5 90 *4 106 *3 94*2 88 *3 89 *9 92 *4 94 *9 97 *2 Dr. C. Chree .
Table IV.\#151 ; D , H and V Ranges .
Means of Percentages from Tables I , II and III .
Year .
Five days of largest range and associated day s. n\#151 ; 2 .
n\#151 ; 1 .
n. n + 1 .
n + 2 .
n + 25 .
n + 26 .
ft+ 27 .
n + 28 .
ft + 29 .
ft + 30 .
1890 99 112 159 115 103 100 104 108 109 106 106 1891 107 133 187 133 115 99 110 113 120 110 109 1892 102 169 236 153 106 90 94 116 106 98 95 1893 99 121 176 125 102 98 101 113 111 105 103 1894 105 157 230 146 118 94 107 102 106 108 132 1895 107 127 176 12\#163 ; ) 106 104 112 130 122 115 109 1896 110 135 194 144 121 112 121 123 128 122 105 1897 92 122 180 122 98 97 109 112 109 102 104 1898 113 153 202 127 104 102 113 129 124 125 109 1899 109 132 180 130 113 104 120 123 116 104 97 1900 96 117 167 108 99 105 99 99 98 94 100 Mean 103 *6 134 *4 189-7 129 *9 107-7 100*4 108 *1 115 *2 113 *6 108 *1 106 -3 " Character " 1890-1900 110 *9 150-t ) 216 *5 158 *9 123-9 101 *5 118*5 134 *8 132 *0 120 *9 113 -7 1906-1911 101 *5 144-4 201 -1 154 *6 116-7 98 *4 112 *8 126 *5 121 *6 112 *5 106-2 Five selected quiet days and associated days .
jl 68\gt ; r\#187 ; n\#151 ; 2 .
n \#151 ; 1 .
n. n + 1 .
n + 2 .
n + 25 .
ft+ 26 .
n + 27 .
^+ 28 .
j ft + 29 .
ft+ 30 .
1890 100 90 81 93 106 95 94 90 93 99 93 1891 95 79 70 91 96 100 97 91 87 107 113 1892 91 74 62 86 91 93 94 107 90 89 85 1893 95 83 74 95 101 89 86 95 109 101 97 1894 74 72 61 97 124 104 90 89 96 92 81 1895 99 88 68 88 112 92 88 85 101 102 108 1896 87 74 64 87 100 87 83 75 78 86 93 1897 94 84 69 90 92 91 80 89 89 101 100 1898 87 77 65 77 102 96 87 76 78 85 100 1899 86 78 69 89 106 87 85 88 92 94 105 1900 96 86 83 96 112 96 93 102 97 96 104 Mean 91*2 80 *5 69 *7 89 *8 103 *8 93 -7 j 88*9 89 *9 92 *0 95 *5 98 *0 " Character " 1890-1900 \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; \#151 ; 1906-1911 82-7 52 -8 20 *5 62 *3 101 *1 89 *1 73 -9 73 -3 80 *7 92 *9 99-7 The last two lines in Table IV give , for purposes of comparison , corresponding percentages based respectively on the " character " figures assigned at Kew for the years 1890-1900 and the international character figures The 27-Day Period in Magnetic Phenomena .
591 calculated at De Bilt for the years 1906-1911 .
No quiet and associated day data existed for the earlier period .
Considering first the range results from the years 1890-1900 , it is seen that the general nature of the conclusions derived from the D , H , and V ranges is the same .
All show , the 27-day period distinctly , in the case alike of the disturbed and the quiet days .
It is not quite clear what is a fair basis for comparing the three elements .
On days % + 27 and n-f-28 , n being the representative day of largest range , the percentage figures are decidedly higher for Y than for D , but the same is true of the representative days n themselves .
If we take the excess over 100 in the percentage figure in column n as a measure of the primary pulse , and the corresponding excess in column n + 27 or n + 28\#151 ; whichever is the larger\#151 ; as a measure of the secondary pulse , the ratio of the secondary to the primary pulse has the values 0*158 for D , 0*207 for H , and 0*150 for Y. The corresponding values of the ratios from the " character " figures were decidedly larger , being 0*299 for the years 1890-1900 and 0*262 for the years 1906-1911 .
S 4 .
In Table IV , 1900 is the only year in which the percentages for days 27 and 28 subsequent to the representative disturbed day fall below 100 .
This year was very quiet , and the difference between the average selected disturbed and the normal day unusually small .
Also a marked decline took place in disturbance throughout the year .
Thus the range of the second of two days separated by 27 days tended to be slightly the smaller , irrespective of whether the first was a selected disturbed day or not .
In Table IY , except in 1900 and 1894 , the largest percentage in the days included in the secondary pulse following the selected disturbed days occurs either in column n + 27 or column n + 28 .
But in 1894 much the largest percentage appears in column n + 30 .
This is true of all the elements , especially Y , the average range for that element on day w-f-30 being actually 55 per cent , above that of the normal day .
That the average range from 60 days selected on a principle which takes account only of what happened 30 days earlier should exceed the normal by 55 per cent , appears , at least at first sight , a phenomenon impossible to attribute to chance .
It seemed desirable to look into this more closely , as it suggested that in 1894\#151 ; a highly disturbed year\#151 ; the period of 30 days which Prof. Ad .
Schmidt has observed in specially large storms was abnormally developed .
The first question was whether the specially large disturbances of Y followed one another at 30 days ' interval .
In the whole 11 years there were only 18 days on which the Y range exceeded 300 7 , and 8 of these occurred in ^894 .
Entering the ranges from these eight days in the order of their magnitude in the column headed n , Dr. C. Chree .
and distinguishing by heavy type ranges from selected disturbed days , the facts are as presented in Table Y. Table Y.\#151 ; Apparent 30-day Period in 1894 .
V Ranges ( Unit 1 7 ) .
n-31 .
w\#151 ; 30 .
n. n + 30 .
n + 31 .
67 25 638 41 508 638 229 508 132 124 16 51 429 40 26 34 19 410 26 27 226 82 362 52 81 52 48 306 29 57 59 52 304 51 29 34 226 302 45 52 The largest range of all , 638 7 , was followed 31 days later by the second largest range , 508 7 , but this was the only case in which two of these specially large disturbances showed a 30- or 31-day sequence .
The phenomenon does not seem to be explained by Schmidt 's period , but to be due to the " accident " that the day 30 days earlier than a day of specially large V range was in six cases out of eight a selected disturbed day .
In four of these six cases , the first disturbance was quite trifling .
Even in 1894 the normal V range was only 46 7 , so that the incidence in column w + 30 of Table III of the six days specified above , having an aggregate range of 22117 , produced an enormous effect .
It may be added that column w-f27 of Table III included only one of the above eight days of largest Y range .
The fact that the lowest percentage in 1894 in the secondary pulse associated with the selected quiet days occurs in column n-f 30 of Table III is also remarkable .
Another outstanding result is the occurrence , in 1892 , of 113 and 107 in column %+27 of Tables III and IY respectively , n being in this case the representative quiet day .
In 1892 there were 28 occasions when the Y range exceeded 100 7 , and of these no less than 9 occurred 27 days after one of the Astronomer Royal 's quiet days .
If the data under review had been confined to 1892 and 1894\#151 ; the two most disturbed of the 11 years\#151 ; it is difficult to say what conclusions might not have been drawn .
Without investigating the records from similarly disturbed earlier years , we do not know what significance to attach to this .
S 5 .
It has been already remarked that a gradual decline in disturbance , such as distinguished 1900 , naturally reduces the amplitude of the secondary The TJ-Day Period in Magnetic Phenomena .
593 pulse of disturbance .
An analogous effect follows from the ordinary seasonal change in the daily range .
If we express all ranges in terms of the mean daily range from the whole year , the consequence is to reduce the apparent amplitude of the secondary disturbance pulse in the beginning of winter , when the daily range is falling , and to correspondingly enhance it in the end of winter , when the range is rising .
This defect could be reduced , though not wholly removed , by expressing all ranges as percentages of the normal range for the particular month in which they occur .
This may be necessary for an exhaustive enquiry as to how the 27-day period varies throughout the year .
It would entail- , of course , much extra labour .
The same object may , however , be attained fairly satisfactorily in a simpler way , by taking the difference between the pulses associated respectively with disturbed and with quiet days .
Any common tendency , whether to enhanced or diminished daily range , is then largely eliminated .
Table VI gives percentage differences obtained in this way for the several ranges and " character " figures .
It is confined in each case to the whole period dealt with , but corresponding results for the ranges in individual years can be deduced from Tables I-IV .
Table VI.\#151 ; Excess of Percentages from 5 Highly Disturbed and Associated Days over Percentages from 5 Selected Quiet and Associated Days .
n \#151 ; 2 .
n \#151 ; 1 .
n. n + 1 .
n + 2 .
n + 25 .
n + 26 .
n + 27 .
n + 28 .
n + 2 9 .
ra + 30 .
1890-1900\#151 ; 17 -i D ranges 67 39 *1 102 -2 32 *7 1-6 5-9 20 -0 18*4 11 3 7*4 H " 16 *0 55-8 118 *2 42 -8 9*4 9 *0 21 *4 29 *2 25 1 15 '6 8*9 V " 15 *4 66 -0 139 *9 44 *7 0*8 5*3 19 *5 26 -7 21 *4 10 *7 8*6 Mean D , H , and V ranges 12 *4 53 *9 120 *0 40*1 3*9 6*7 19 *2 25 *3 21 *6 12 *6 8*3 1906-1911\#151 ; International " character " 18-8 91 *6 180 *6 92 *3 15 *6 9*3 38 -9 53 -2 40-9 19 *6 6*5 In Table VI the figure in column % + 27 is always substantially greater than that in column 71 + 28 , and the figure in column n+ 28 always exceeds that in column w-f-26 .
This suggests , of course , as dominant a period in excess of 27 days , but considerably nearer 27 days than 28 .
The data , however , in Tables I-IV for the days associated with the selected quiet days suggest a period somewhat nearer 26 than 27 days , while the 11-year mean data from days associated with disturbed days in Table I suggest a period slightly in excess of 27 J days .
This is distinctly suggestive of a possibility , pointed to by some results in SM2 , that the period tends to lengthen as the standard of disturbance represented in the primary pulse is increased .
S 6 .
If five days were selected by lot from each month , the expectation that Dr. C. Chree .
the 27th day before ( or after ) one of these days would be itself a selected day would be roughly 60/ 365 , i.e. the event would happen in about 16J per cent , of the days .
Taking the 655 selected days of one type from 10 years 11 months , one would thus expect such a sequence on about 108 occasions .
Table YII shows how often selected days of specified types showed a 27-day sequence .
Table VII.\#151 ; 27-Day Sequences of Selected Days .
First day .
Disturbed .
Quiet .
Quiet .
Disturbed .
Second day ... Disturbed .
Quiet .
Disturbed .
Quiet .
Element D. H. Y. D , H , and Y , mean .
\#151 ; H. H. Year .
1890 10 16 12 13 12 3 9 1891 15 13 16 15 14 7 7 1892 11 12 12 12 11 12 7 1893 11 20 15 15 15 10 6 1894 14 16 13 14 13 4 10 1895 16 19 22 19 14 3 7 1896 25 22 22 23 21 3 3 1897 16 14 14 15 13 2 4 1898 19 19 20 19 17 1 3 1899 15 19 16 17 14 5 7 1900 12 16 7 12 9 7 9 Total 164 186 169 173 153 57 72 The sequence of a disturbed day by a disturbed day , it will be seen , was decidedly more frequent for H than for D or V. Also , whichever of the three elements one takes , the sequence of a disturbed day by a disturbed day was more frequent than the sequence of a quiet day by a quiet day .
This is what we should expect on any hypothesis , because the selected disturbed days were more outstanding as representatives of disturbance than were the Astronomer Royal 's days as representatives of quiet conditions .
Even for the quiet days the sequences were more numerous by 40 per cent , than they would have been according to pure chance ; 1896 showed an outstandingly large number of sequences , whether of disturbed days by disturbed days , or quiet days by quiet days .
The cases in which a selected quiet day was followed by a selected disturbed\#187 ; day were , on the whole , decidedly fewer than those in which the converse phenomenon occurred .
The remarkable prevalence , however , of the former phenomenon in 1892 has been already remarked on .
The 27 -Day Period in Magnetic Phenomena .
The number of 28-day sequences of disturbed days was also investigated for the H ranges .
Taking the years in chronological order the numbers were respectively 17 , 12 , 8 , 11 , 16 , 16 , 20 , 11 , 22 , 14 , 15 .
The total 162 is less by 24 than for the corresponding 27-day sequence .
1898 showed the largest number of 28-day sequences , and 1892 the fewest .
S 7 .
Taking the H disturbances , it was investigated how far sequences of 27 days could be traced .
The results were as follows:\#151 ; Sequence of 27 days ... ^ 2 days or more .
3 days or more .
4 days or more .
5 days or more .
6 days or more .
7 days or more .
8 days .
Number of cases 186 j 32 17 7 2 1 1 The 32 cases are included in the 186 , the 17 in the 32 , and so on .
Thus in the 32 cases where there were three or more sequences the second of the two days was at once the second day of one sequence of two , and the first day of another sequence of two .
The longest sequence , covering 7 x 27 or 189 days , began on December 10 , 1895 , and ended on June 16,1896 .
The next longest extended from December 20 , 1897 , to May 4 , 1898 .
Several things have to be borne in mind when dealing with apparent sequences .
If a disturbance were a sharply marked incident of short duration and if its recurrence period , if it recurred at all , were 27 + / days , where / is a definite fraction of 24 hours , then the number of apparent recurrences obtained as above could not exceed r = ( 24\#151 ; N)/ / , where FT is the hour of the first occurrence .
There might be more than r recurrences , but the ( r-f l)th would fall on the 28th day not the 27th after the rth .
The number of apparent 27-day recurrences could not exceed 24/ /and a limit might thus be found for / .
For instance , if we accepted 7 as the maximum possible number of recurrences , the limiting value of / would be 3-f hours .
A disturbance , however , is not a shaply marked incident , and disturbed conditions are often shared by two or more consecutive days .
Also the chances against an apparent sequence being a real one would seem to be not inconsiderable , in view of the fact that about 9 per cent , even of the Astronomer Royal 's quiet days had for the 27th day thereafter a selected disturbed day .
This , of course , tends to emphasize the desirability of finding some criterion for discriminating between true and apparent sequences .
S 8 .
The first thing one naturally looks to when in quest of a criterion is the relative size of disturbances forming an apparent sequence .
A sunspot , for instance , often develops very rapidly , but is seldom large for more than one VOL. XC.\#151 ; \#166 ; \#166 ; \#166 ; A. 2 y Dr. C. Chree .
solar revolution .
An investigation was thus made into the amplitudes of the 186 H ranges which showed a 27-day sequence .
Table VIII shows the average size of the ranges on the first and second occurrences , and for comparison the average size of the range from all the selected disturbed days of the year .
It also gives the number of times when the first range of the sequence exceeded the second or fell short of it .
It should be remembered that the 186 cases included 32 in which the sequence extended to three or more days , and in these cases at least the second member of the sequence figured as the second day of one sequence and the first of another .
The ranges are given as measured in millimetres .
The value of 1 mm. was usually 5*0 7 , but in some years it departed from this by 1 or 2 per cent. Table VIII.\#151 ; H Ranges forming Sequences of Two .
Ranges in mm. ( 1 mm. = 5 y approx. ) .
Number of occasions when Year .
Average disturbed day .
First day of sequence .
Second day of sequence .
First range the larger .
Second range the larger .
1890 14 *5 16 *5 16 *4 9 7 1891 22 *0 18 *4 18-4 7 6 1892 38 -3 47 *0 47 *3 7 5 1893 23 *4 22 *9 22 *9 9 10 1894 34 3 29 *5 25 -9 10 6 1895 22 *8 25 *7 24 T 10 9 1896 23 *8 24 T 22 *9 12 10 1897 17 *5 18 *5 18 *7 8 6 1898 21 *9 21 *3 21 -5 10 9 1899 17*5 18 *7 17 *7 9 9 1900 13 *2 14 *6 12 *9 8 8 Totals or !
means J 22-7 23*4 22 *6 99 85 ' On two occasions , the first in 1893 , the second in 1898 , the ranges were equal .
It will be seen that in most individual years the range on the first day of a sequence was more often than not the larger , and on the average from the 11 years the first day 's range was some 4 per cent , in excess of the second day 's .
But in 1890 , 1891 , 1892 , 1893 , 1897 , and 1898 the difference between the mean ranges from the first and second days of a sequence was almost microscopic .
S 9 .
The cases of three or more sequences led to the results given in Table IX .
The ranges in this case are given in millimetres , and in forming the means the value of 1 mm. was treated as invariable , which was not always strictly the case .
The 27-Day Period in Magnetic Phenomena .
597 Table IX.\#151 ; H Eanges ( in mm. ) in Days forming Sequences of Disturbances .
First day .
Second day .
Third day .
Fourth day .
Fifth day .
Sixth day .
Seventh day .
Eighth day .
32 sequences of 3 days 22 *1 20 *5 24 -1 31 " 3 ji 22 *3 20 *5 20 *3 17 " 4 \gt ; \gt ; 22 *9 21 -6 27 *7 22 -0 16 " 4 \gt ; \gt ; 23 *4 21 *6 20 *5 19 *7 7 " 5 j ) 25 *2 22 *4 21 -0 22 *5 21 *9 6 " 5 \gt ; \gt ; 18 -9 21 *5 20 -2 22 *5 20 *7 1 " 6 jj 30 *9 17*0 19 -1 23 *9 20 *9 22 -3 1 " 8 9\gt ; 16 *1 21 *1 27 *6 22 -2 25 -0 17 -4 26 -6 22 *2 Two sets of results are given for the sequences of three , four , and five days , because the means derived from the full number of cases , whether 32 , 17 , or 7 , were largely influenced by one outstanding day .
February 13 , 1892 , the most disturbed day of the whole 11 years , figured as the third day of a sequence of four , the ranges of which in millimetres ran as follows:\#151 ; 15*4 20-7 144*0 59*5 The run of the ranges suggests that the connection of the 144 mm. range with those preceding it may well have been accidental , and in any case the omission of this outstanding occasion from the three- and four-day sequences seemed likely to lead to a more representative result than its inclusion .
Omitting this day , there does seem a tendency to reduction in the range of the later members of the three- and four-day sequences , but it is very slight .
The five-day sequence omitted in forming the third last line of Table IX began with a range of 63*3 mm. , which was much the largest found on the first day of any of the sequences , the next in size being only 42*9 mm. These two ranges occurred on successive days , forming two of a remarkable succession of highly disturbed days , February 21-25 , 1894 .
The ranges in millimetres on these five days and on successive groups of five days separated by 27-day intervals were as follows :\#151 ; Monthly mean .
February 21 to 25 35 7 30 *6 63 *3 42 *9 67 *7 19 *3 March 20 " 24 11*1 14 -5 27 *7 25 0 13 *4 15*4 April 16 " 20 11 0 33 6 26 -2 20 *4 15 *5 15 *5 May 13 " 17 17 -0 33 *8 22 *2 20 9 14 *2 16 *2 June 9 " 13 27 *8 45 *1 29*2 16 *9 16*6 18 *6 July 6 " 10 14 *0 13 *2 17 *6 20 *8 13 *4 20 *9 2 u 2 Dr. C. Chree .
The mean range from all days of each month is shown for comparison .
Heavy type indicates a day which qualified as a disturbed day .
February 22 would have qualified in any but an exceptionally disturbed month .
If March had been as disturbed a month as February there would have been no apparent sequence at all .
If , as seems on the whole most probable , the above represents four recurrences of a single disturbance , the changes in duration and intensity that took place between the first and second occurrences were on a different scale from those that took place between the subsequent occurrences .
A parallel , however , to the changes between the first and second occurrences is afforded by the completeness of the disappearance after the fifth occurrence .
Another complex case is that presented by the long apparent sequence of eight disturbed days in 1895 to 1896 , and the eight days immediately previous to them .
The ranges , in millimetres , on these eight pairs of days were as follows , heavy type indicating that the day in question qualified as a disturbed day:\#151 ; Dates Dec. 9 10 Jan. 5 6 Feb. 1 2 Feb. 28 29 Mar. 26 27 April 22 23 May 19 20 June 15 16 Ranges in mm. 19 -2 16*1 16 9 21*1 16 *5 27 6 38 9 22 2 23 8 25*0 27 0 17 *4 21 -4 26'6 11 0 22 2 All the 16 ranges in the two lines were decidedly above the normal for their respective months , except 1T0 mm. on June 15 .
Thus the fact that the days in the second line show an unbroken sequence of eight disturbed days , while the longest sequence of disturbed days in the first line is only three , may he purely an accidental consequence of our restricting the selected disturbed days to five a month .
Three ranges in the first line which did not qualify all exceed the first range in the second line which did .
The successive pairs of days may well have represented repetitions of one and the same disturbance .
If they did , then it probably presented at least 10 sequences , for November 12 , 1895\#151 ; the 27th day before December 9\#151 ; qualified with a range of 251 , and July 12 , 1896\#151 ; the 27th day after June 15\#151 ; also qualified with a range of 23-4 .
On this view the climax would seem to have occurred on February 28 , after three if not four previous appearances .
An analogous , though somewhat different , case is represented by the following pairs of days , extending from October 27 , 1897 , to May 5 , 1898 The 27-Day Period in Magnetic Phenomena .
599 Dates 1 Oct. 27 28 Nov. 23 24 Dec. 20 21 Jan. 16 17 Feb. 12 13 Mar. 11 12 April 7 8 May 4 .
5 Ranges in mm. 7-0 16 3 35 14*5 30 *9 17 *3 17 0 20 *3 19 *1 9-0 23 9 16 O 20 *9 10 *1 22 *3 16 7 As before , heavy type indicates that the day in question was a selected disturbed day .
The last six entries in the first line represent a sequence of six disturbed days , commencing on December 20 .
November 23 , the 27th day prior to this commencement , was a selected quiet day , and had the smallest range of any of the quiet days of the month .
Thus regarding the first line of figures alone , we would accept December 20 unhesitatingly as the first of the sequence .
But while November 23 and October 27 did not qualify as disturbed days , the following did:\#151 ; October 29 with 18*7 , October 2 with 19*6 , and September 5 with 17*8 ; also November 25 had a range 10T , considerably above the average .
Thus October 28 and 29 , November 24 and 25 , and December 20 and 21 may represent sequences of a disturbance which originated at least as far back as September 5 , but which had a recurrence period averaging a little under 27 days .
On this view there may have been a repetition of the disturbance later than those shown , for May 30 qualified with 34*2 .
It is thus by no means improbable that there was in reality a sequence of 11 , extending from September 5,1897 , to May 30 , 1898 .
The above examples will show the difficulty of saying which disturbances form sequences and which do not .
If instead of 5 we had taken 10 disturbed days a month , or had adopted some criterion which allowed a variable number averaging more than five , we should have found the difficulty greater .
Unless we can recognise some common trait in disturbances arising from what we may regard as a common ancestor , or obtain guidance from some allied branch of science , we shall probably find general statistical methods such as those adopted in the first part of the paper the safest to follow .
The expense of measuring the daily ranges employed in this investigation was defrayed by a grant from the Government Grant Committee .

