NATURE OF THE SHOCK.
Little has yet been made known with regard to the nature of the shock, and the published records of the accompanying sound are so rare that it seems as a rule to have passed unheard. The seismographs at Gifu and Nagoya registered the first half-dozen vibrations, and were then buried beneath the fallen buildings. In the following table, the data from these two stations are therefore incomplete:—
PRINCIPAL MEASUREMENTS OBTAINED FROM SEISMOGRAPHIC RECORDS.
- - Tokio Gifu. Nagoya. Osaka. (Imp. Univ.). - - Maximum horizontal motion > 18 mm. > 26 mm. 30 mm. > 35 mm. Period of ditto 2.0 secs. 1.3 sec. 1.0 sec. 2.0 secs. Maximum vertical motion > 11.3 mm. 6.2 mm. 8 mm. 9.5 mm. Period of ditto 0.9 sec. 1.5 sec. 1.0 sec. 2.4 secs. - -
If the period of the principal vibrations were known, the observations of Professor Omori on the overturning of bodies would enable us to determine the range of motion at different places. For instance, the maximum acceleration at Nagoya was found by these observations to be 2,600 millimetres per second per second, and if we take the period of the greatest horizontal motion to be the same as that of the initial vibrations—namely, 1.3 second, the total range (or double amplitude) would be 223 millimetres, or 8.8 inches. With the same period, and the maximum acceleration observed (at Iwakura and Konaki) of more than 4,300 millimetres per second per second, the total range would be greater than 14.5 inches.
In the meizoseismal area, many persons saw waves crossing the surface of the ground. At Akasaka, according to one witness, the waves came down the streets in lines, their height being perhaps one foot, and their length between ten and thirty feet. To the north of the same area, we are told that "the shoreline rose and fell, and with this rising and falling the waters receded and advanced." Even at Tokio, which is about 175 miles from the epicentre, the tilting of the ground was very noticeable. After watching his seismographs for about two minutes, Professor Milne next observed the water in an adjoining tank, 80 feet long and 28 feet wide, with nearly vertical sides. "At the time it was holding about 17 feet of water, which was running across its breadth, rising first on one side and then on the other to a height of about two feet." Still clearer is the evidence of the seismographs in the same city. Instead of a number of irregular waves, all the records show a series of clean-cut curves. The heavy masses in the horizontal pendulums were tilted instead of remaining as steady points. They were not simply swinging, for the period of the undulations differed from that of the seismograph when set swinging, and also varied in successive undulations. It was ascertained afterwards, by measurement with a level, that to produce these deflections, the seismograph must have been tilted through an angle of about one-third of a degree.
Direction of the Shock.—Shortly after the earthquake, Professor Omori travelled over the meizoseismal area and made a large number of observations on the directions in which bodies were overturned, taking care to include only those in which the direction of falling would not be influenced by the form of the base, such as the cylindrical stone lanterns so frequently found in Japanese gardens. At some places these bodies fell in various directions, at others with considerable uniformity in one direction. For instance, at Nagoya, out of 200 stone lanterns with cylindrical stems, 119 fell between west and south, and 36 between east and north; the numbers falling within successive angles of 15 being represented in Fig. 43. The mean direction of fall is W. 30 S., coinciding with that in which the majority of the lanterns were overturned. Similar observations were made at forty-two other places within and near the meizoseismal area, and the resulting mean direction for each such place in the Mino-Owari district is shown by short lines in Fig. 44, the arrow indicating the direction towards which the majority of bodies at a given place were overturned. It will be seen from this map that the direction of the earthquake motion was generally at right angles, or nearly so, to that of the neighbouring part of the meizoseismal zone, and that on both sides of it, the majority of overturned bodies at each place fell towards this zone.
VELOCITY OF THE EARTH-WAVES.
The times of the great earthquake and of sixteen minor shocks on October 28th and 29th and November 6th were determined at the Central Meteorological Observatory at Tokio, and at either two or three of the observatories of Gifu, Nagoya, and Osaka, each of which is provided with a seismograph and chronometer. The after-shocks referred to originated near a point about 6 miles west of Gifu, and the difference between the distances of Tokio and Osaka from this point is 89-1/2 miles, of Tokio and Nagoya 147 miles, and of Tokio and Gifu 165 miles. The mean time-intervals between these three pairs of places were 67, 111, and 128 seconds respectively; and these give for the mean velocity for each interval 2.1 kilometres (or 1.3 mile) per second. Thus there appears in these cases to be no sensible variation in the velocity with the distance from the origin.
As might be expected, an earthquake of such severity was recorded by magnetometers at several distant observatories. Disturbances on the registers of Zikawei (China), Mauritius, Utrecht, and Greenwich have been attributed to the Japanese earthquake, but the times at which they commenced are too indefinite to allow of any determination of the surface-velocity of the earth-waves to great distances from the origin.
THE GREAT FAULT-SCARP.
As in all disastrous earthquakes, the surface of the ground was scarred and rent by the shock. From the hillsides great landslips descended, filling the valleys with dbris; and slopes which were formerly green with forest, after the earthquake looked as if they had been painted yellowish-white. Innumerable fissures cut up the plains, the general appearance of the ground, according to Professor Milne, being "as if gigantic ploughs, each cutting a trench from 3 to 12 feet deep, had been dragged up and down the river-banks." But by far the most remarkable feature of the earthquake was a great rent or fault, which, unlike the fissures just referred to, pursued its course regardless of valley, plain, or mountain. Although at first sight quite insignificant in many places, and some time hardly visible to the untrained eye, Professor Koto has succeeded in tracing this fault along the surface for a distance of forty miles, and he gives good reasons for believing that its total length must be not less than seventy miles.
The general character of the fault-scarp changes with the surface features. On flat ground, where the throw is small, it cuts up the soft earth into enormous clods, or makes a rounded ridge from one to two feet high, so that it resembles, more than anything else, the pathway of a gigantic mole (Fig. 46). When the throw is considerable—and in one place it reaches from 18 to 20 feet—the fault-scarp forms a terrace, which from a distance has the appearance of a railway embankment (Fig. 47). Or, again, where the rent traverses a mountain ridge or a spur of hills, "it caused extensive landslips, one side of it descending considerably in level, carrying the forest with it, but with the trees complicatedly interlocked or prostrate on the ground."
At its southern end, the fault was seen for the first time crossing a field near the village of Katabira. The field was broken into clods of earth, and swollen up to a height of 5-1/2 yards, while a great landslip had descended into it from an adjoining hill. A little farther to the north-west, the ground was sharply cut by the fault, the north-east side having slightly subsided and at the same time been shifted horizontally through a distance of 3-1/4 to 4 feet to the north-west Adjoining fields were formerly separated by straight mounds or ridges running north and south and east and west, and these mounds were cut through by the fault and displaced, as shown in Fig. 48. From this point the fault runs in a general north-westerly direction, the north-east side being always slightly lowered with respect to the other and shifted to the north-west. Near Seki it takes a more westerly direction, and continues so to a short distance east of Takatomi, where the north side is lowered by five feet, and moved about 1-1/4 feet to the west. At the north end of Takatomi, a village in which every house was levelled with the ground, the fault is double, and the continuous lowering towards the north has converted a once level field into sloping ground. At this point, the small river Toba, flowing south, is partially blocked by the fault-scarp, and an area of about three-quarters of a square mile, on which two villages stand, was converted into a deep swamp (Fig. 49), so that, as the earthquake occurred at the time of the rice-harvest, the farmers were obliged to cut the grain from boats. After passing Takatomi, the fault again turns to the west-north-west, but, the throw being small, it resembles here the track of an enormous mole. At Umhara it crosses a garden between two persimmon trees, appearing on the hard face of the ground as a mere line; but the trees, which were before in an east-and-west line, now stand in one running north and south, without being in the least affected by the movement (Fig. 50). From here to Kimbara, where the fault enters the Neo valley, the north side is always depressed and shifted westwards by about 6-1/2 feet.
It was in the Neo valley that the supreme efforts of the earthquake were manifested. Landslips were so numerous that the greater part of the mountain slopes had descended into the valley, the whole appearance of which had changed. "Unfamiliar obstacles," remarks Professor Koto, "made themselves apparent, and small hills covered with forest had come into sight which had not been seen before." But the ground was not only lowered and shifted by the fault; it was permanently compressed, plots originally 48 feet in length afterwards measuring only 30 feet. In fact, "it appears," in the words of Professor Milne, "as if the whole Neo valley had become narrower."
A few miles after entering the Neo valley, the throw of the fault reaches its maximum at Midori. But instead of the relative depression of the east side, which prevails throughout the rest of the line, that side is here about 20 feet higher than the other. It is, however, shifted as usual towards the north, by about 13 feet; and this displacement is rendered especially evident by the abrupt break in the line of a new road to Gifu (Fig. 47). That the east side has really risen is clear, for, a little higher up, the river has changed from a shallow rapid stream 30 yards wide into a small lake of more than twice the width, and so deep that a boatman's pole could not reach the bottom. At Itasho, about a mile north of Midori, both sides are nearly on the same level, the fault appearing like a mole's track; and seven miles farther, at Nagoshima, the east side is relatively depressed by more than a yard, and at the same time shifted about 6-1/2 feet to the north.
At Nogo, the main Neo valley turns off at right angles to the east, and the fault continues its course up a side valley, the east side, with respect to the other, being continually depressed and shifted towards the north. It was traced by Professor Koto through Fujitani (Fig. 46), where there were many unmistakable evidences of the violence of the shock, as far as the eastern shoulder of Haku-san; and here, after following the fault for 40 miles, the lateness of the season compelled him to return. There can be no doubt, however, that it runs as far as Minomata; and it is probable, from the linear extension of the meizoseismal area, that it does not entirely die out before reaching the city of Fukui, 70 miles from its starting-point at Katabira.
For some hours after the earthquake, shocks were so frequent in the meizoseismal area that the ground in places hardly ever ceased from trembling. Without instrumental aid, detailed record was of course impossible; but fortunately the buried seismographs at Gifu and Nagoya were uninjured, and in about seven hours both were once more in working order. To the energy by which this result was accomplished, we owe our most valuable registers of the after-shocks of a great earthquake.
Until the end of 1893—that is, in little more than two years—the total number of shocks recorded at Gifu was 3,365, and at Nagoya 1,298. None of these approached the principal earthquake in severity. Nevertheless, of the Gifu series, 10 were described as violent and 97 strong; while of the remainder, 1,808 were weak, 1,041 feeble, and 409 were sounds alone without any accompanying shock. The slight intensity of most of the shocks is also evident from the inequality in the numbers recorded at Gifu and Nagoya, from which it appears that nearly two-thirds were imperceptible more than about 25 miles from the chief origin of the shocks. Only 70 of the after-shocks during the first two years were registered at Osaka, and not more than 30 at Tokio.
Distribution of After-shocks in Time.—The decline in frequency of the after-shocks was at first extremely rapid, the numbers recorded at Gifu during the six days after the earthquake being 303, 147, 116, 99, 92, and 81, and at Nagoya 185, 93, 79, 56, 30, and 31; in fact, half of the shocks up to the end of 1893 occurred by November 23rd at Gifu, and by November 6th at Nagoya. The daily numbers at these two places are represented in Fig. 51, in which the crosses correspond to the numbers at Gifu, and the dots to those at Nagoya; and the curves drawn through or near the marks represent the average daily number of shocks from October 29th to November 20th. It will be seen that these curves are hyperbolic in form, the change from very rapid to very gradual decline in frequency taking place from five to ten days after the great earthquake. Fig. 52 illustrates the distribution in time of the after-shocks at Gifu to the end of 1893, the ordinates in these cases representing the number of shocks during successive months.
A similar rapid and then gradual decline in frequency characterises the strong and weak shocks recorded at Gifu. Of the ten violent shocks, only one occurred after the beginning of January 1892; and of the 97 strong shocks, only three after April 1892. But at the commencement of the series, feeble shocks (i.e., shocks that could just be felt) and earth-sounds without any accompanying movement were comparatively rare, and did not become really prominent until two months had elapsed. Of the 308 after-shocks recorded in 1893, none could be described as strong, only 10 were weak, while 263 were feeble shocks and 35 merely earth-sounds.
The last two diagrams show at a glance that the decline in frequency of after-shocks is very far from being uniform. Some of the fluctuations are due to the occurrence of exceptionally strong shocks, each of which is followed by its own minor train of after-shocks. Others seem to be periodic, and possibly owe their origin to external causes unconnected with the earthquake.
Method of representing the Distribution of After-shocks in Space.—The maps in Figs. 54-57 show the distribution of the after-shocks in space during four successive intervals of two months each. They are founded on Professor Milne's great catalogue of Japanese earthquakes, which give, among other data, the time of occurrence and the position of the epicentre for every shock until the end of 1892. For the latter purpose, the whole country is divided by north-south and east-west lines into numbered rectangles, each one-sixth of a degree in length and breadth; and the position of an epicentre is denoted by the number of the rectangle in which it occurs. The area included within the maps is bounded by the parallels 34 40' and 36 20' lat. N., and by the meridians 2 10' and 3 50' long. W. of Tokio, so that ten rectangles adjoin each side of the map. The number of epicentres lying within each rectangle having been counted, curves are then drawn through the centres of all rectangles containing the same number of epicentres, or through points which divide the line joining the centres of two rectangles in the proper proportion. Taking, for example, the curve marked 5, if the numbers in two consecutive rectangles are 3 and 7, the curve bisects the line joining their centres; if the numbers are 1 and 6, the line joining their centres is divided into five equal parts, and the curve passes through the first point of division reckoned from the centre of the rectangle in which six epicentres are found. Thus the meaning of the curve marked, say, 5 may be stated as follows:—If any point in the curve be imagined as the centre of a rectangle whose sides are directed north-south and east-west, and are respectively one-sixth of a degree of latitude and longitude in length; then the number of epicentres within this rectangle is at the rate of 5 for the time considered.
Preparation for the Great Earthquake.—At first sight, there appears to have been but little direct preparation for the great earthquake. Except for a rather strong shock on October 25th, at 9.14 P.M., it occurred without the warning of any preliminary tremors. But a closer examination of the evidence shows, as we should indeed expect, that there was a distinct increase in activity for many months beforehand. The region had become "seismically sensitive." Of the hundred rectangles included in the maps in Figs. 53-57, there are thirteen lying along the meizoseismal area of the earthquake of 1891, in which nearly all the after-shocks originated. During the five years 1885-89, 53 out of 125 earthquakes (or 42 per cent.) had their epicentres lying within the thirteen rectangles; or, in other words, the average frequency in one of the rectangles of the meizoseismal area was five times as great as in one of those outside it. In 1890 and 1891 (until October 27th), the percentage in the thirteen rectangles rose to 61, and the average frequency in one of them to ten times that of one of the exterior rectangles.
The curves in Fig. 53 illustrate the distribution of epicentres during the latter interval. It will be seen that they follow roughly the course of the meizoseismal area southwards to the Sea of Is, and that to the south-east they continue for several miles the short branch of the meizoseismal area which surrounds the southern end of the fault-scarp.
Thus, the preparation for the great earthquake is shown, first, by the increased frequency of earthquakes originating within its meizoseismal area; and, secondly, by the uniformity in the distribution of epicentres throughout the same region, the marked concentration of effort which characterises the after-shocks being hardly perceptible during the years 1890-91.
Distribution of After-shocks in Space.—We have seen that the after-shocks were subject to a fluctuating decline in frequency, rapid at first, and more gradual afterwards. It is evident, from Figs. 54-57, that a similar law governs the area within which the after-shocks originated. During the first two months, epicentres occur over nearly the whole of the meizoseismal area, but afterwards they are confined to a smaller district, which slowly, though not continually, decreases in size.
The most important feature in the distribution of the epicentres is the central region of extraordinary activity; but there are also districts of minor and more short-lived activity near the three extremities of the meizoseismal band. The seat of chief seismic action shifts slightly from one part to another of the epicentral region, especially about the end of 1891, as will be seen by comparing the innermost curves of Figs. 54 and 55. Thus, with the decline in frequency of the after-shocks and the decrease in their sphere of action, there took place concurrently a gradual but oscillating withdrawal of that action to a more or less central region of the fault.
Sound Phenomena of After-shocks.—While comparatively few observers seem to have noticed any noise with the principal earthquake, many of the after-shocks were accompanied by sounds. Professor Omori describes them as belonging to two types. They were either rushing feeble noises like that of wind, or loud rumbling noises like those of thunder, the discharge of a gun, or the fall of a heavy body. In the Neo valley, sounds of the second type were most frequent and distinct, but they either occurred without any shock at all, or the attendant tremor was very feeble; while, on the other hand, severe sharp shocks were generally unaccompanied by distinctly audible sounds.
It is remarkable, also, that sounds were less frequently heard with the early than with the later after-shocks. In November 1891, the percentage of audible shocks was 17, and from December to the following April always lay between 10 and 12. In May the percentage suddenly rose to 39, and until the end of 1892 was always greater than 32, while in November 1892, it rose as high as 49. This, of course, agrees with Professor Omori's observation that sounds attended feeble shocks more often than strong ones.
The distribution of the audible after-shocks in space is shown in Fig. 58. These curves are drawn in the same way as those in Figs. 53-57, but they represent the percentages, not the actual numbers, of shocks accompanied by sound. It will be noticed that all three groups of curves lie along the meizoseismal area, or the continuation of the south-east branch; while the axis of the principal group of curves lies to the west of the central regions in which most after-shocks originated.
The explanation of these peculiarities is no doubt connected with the comparative inability of the Japanese people to perceive the deep sounds which in Europe are always heard with earthquake shocks. The sounds are rarely heard by them more than a few miles from the epicentre. We may therefore conclude that slight after-shocks originated nearer the surface than strong ones, that the mean depth of the foci decreased with the lapse of time, and that the axes of the systems of curves in Fig. 58 mark out approximately the lines of the growing faults. The separation of the two westerly groups of curves appears to show that the main branch of the meizoseismal area is connected with a fault roughly parallel to that traced by Professor Koto, but of which no scarp (if it existed) could be readily distinguished among the superficial fissures produced by the great shock.
EFFECT OF THE EARTHQUAKE ON THE SEISMIC ACTIVITY OF THE ADJOINING DISTRICTS.
So great and sudden a displacement as occurred along the fault-scarp could hardly take place without affecting the stability of adjoining regions of the earth's crust, and we should naturally expect to find a distinct change in their seismic activity shortly after October 28th. In Fig. 59 two such regions are shown, bounded by the straight dotted lines. The district in which the principal earthquake and its after-shocks originated is enclosed within the undulating dotted lines. The continuous lines inside all three districts are the curves corresponding to 10 and 5 epicentres for the years 1885-92. Not far from the axes of the outer groups of curves there are probably transverse faults, approximately parallel to the great fault-scarp and the main branch of the meizoseismal band, and distant from them about 45 and 55 miles respectively.
In the district represented in the north-east corner of Fig. 59, 29 earthquakes originated between January 1st, 1885, and October 27th, 1891, and 30 between October 28th, 1891, and December 31st, 1892, 7 of the latter number occurring in November 1891. In the south-west district, the corresponding figures before and after the earthquake are 20 and 36, 8 of the latter occurring in November 1891. Thus, in the north-east district, for every shock in the interval before the earthquake there were six in an equal time afterwards, and at the rate of 10 during November 1891; and in the south-west district, for every shock before the earthquake there were 10 afterwards, and at the rate of 16 during November 1891.
Now, it is unlikely that the gradual increase of stress should be so nearly proportioned everywhere to the prevailing conditions of resistance as to give rise to a marked and practically simultaneous change in seismic activity over a large area; whereas the paroxysmal occurrence of a strong earthquake might alter the surrounding conditions with comparative rapidity, and so induce a state of seismic excitement in the neighbourhood. It therefore seems very probable that the increased activity in the two districts here described was a direct consequence of the occurrence of the great earthquake.
ORIGIN OF THE EARTHQUAKE.
The preponderance of preliminary earthquakes within the meizoseismal area and the outlining of the fault-system by the frequency curves of 1890-91 (Fig. 53) point to the previous existence of the originating fault or faults, and to the earthquake being due, not to the formation of a new fracture, as has been suggested, but to the growth of an old fault.
The last severe earthquake in the Mino-Owari plain occurred in 1859, so that for more than thirty years there had been but little relief to the gradually increasing stresses. Now, the distribution of stress must have been far from uniform throughout the fault-system, and also the resistance to displacement far from proportional to the stresses at different places. At certain points, therefore, the effective stress would be greater than elsewhere, and it would be at these points that fault-slips would first occur. Such slips tend to remove the inequalities in effective stress. Thus, the function of the slight shocks of 1890 and 1891 was, briefly, to equalise the effective stress over the whole fault-system, and so to clear the way for one or more great slips throughout its entire length.
As to which side of the fault moved during the great displacement, or whether both sides moved at once, we have no direct evidence but as regards the neighbourhood of Midori, and there the conditions were exceptional. Professor Koto thinks that it was probably the rock on the north-east side that was generally depressed and always shifted to the north-west But the disturbance in reality seems to have been more complicated. That this was the case, that displacement occurred along more than one fault, is probable from the branching of the meizoseismal area, the isolation of the audibility curves of the after-shocks (Fig. 58), and the sudden increase in seismic activity both to the north-east and south-west of the epicentre. The detached portion of the meizoseismal area near Lake Biwa may also point to a separate focus. The whole region, indeed, was evidently subjected to intense stresses, and the depression on the north-east side of the fault-scarp can hardly fail to have been accompanied by other movements, especially along a fault running near the western margin of the main branch of the meizoseismal area.
The later stages of the movements are somewhat clearer. From a study of the after-shocks, we learn that the disturbed masses began at once to settle back towards the position of equilibrium. At first the slips were numerous and took place over the whole fault-system, but chiefly at a considerable depth, where no doubt the initial displacement was greatest. After a few months, stability was nearly restored along the extremities of the faults; slips were confined almost entirely to the central regions, while a much larger proportion of them took place within the superficial portions of the faults.
The official records bring down the history to the end of 1893. Since that time more than one strong shock has been felt in the Mino-Owari plain; but the stage of recovery from the disturbances of 1891 is probably near its end, and we seem rather to be entering on a period in which the forces are once more silently gathering that sooner or later will result in another great catastrophe.
1. CONDER, J.—"An Architect's Notes on the Great Earthquake of October 1891." Japan Seismol. Journ., vol. ii., 1893, pp. 1-91.
2. DAVISON, C.—"On the Distribution in Space of the Accessory Shocks of the Great Japanese Earthquake of 1891." Quart. Journ. Geol. Soc., vol. liii., 1897, pp. 1-15.
3. —— "On the Effect of the Great Japanese Earthquake of 1891 on the Seismic Activity of the Adjoining Districts." Geol. Mag., vol. iv., 1897, pp. 23-27.
4. —— "On the Diurnal Periodicity of Earthquakes." Phil. Mag., vol. xiii., 1896, pp. 463-476, especially pp. 466-468.
5. —— "On Earthquake-Sounds." Phil. Mag., vol. xlix., 1900, pp. 31-70—especially pp. 49-53, 60-61.
6. KOTO, B.—"The Cause of the Great Earthquake in Central Japan, 1891." Journ. Coll. Sci. Imp. Univ. Japan, vol. v., 1893, pp. 295-353.
7. MASATO, H.—"Report on Earthquake Observations in Japan." Cent. Meteor. Obs. of Japan (Tokio, 1892), pp. 16-18, 41, and map 30.
8. MILNE, J.—"A Note on the Great Earthquake of October 28th, 1891." Japan Seismol. Journ., vol. i., 1893, pp. 127-151; Brit. Assoc. Rep., 1892, pp. 114-128.
9. —— "A Catalogue of 8,331 Earthquakes recorded in Japan between 1885 and 1892." Japan Seismol. Journ., vol. iv., 1895, pp. 1-367—especially pp. 134-234, 303-353.
10. —— "On Certain Disturbances in the Records of Magnetometers and the Occurrence of Earthquakes." Brit. Assoc. Rep., 1898, pp. 226-251—especially pp. 227, 232, 234, 241, and 245.
11. MILNE, J., and W.K. BURTON.—"The Great Earthquake in Japan." Journ. Coll. Sci. Imp. Univ. Japan, vol. v., 1893, pp. 295-352.
12. OMORI, F.—"On the After-shocks of Earthquakes." Journ. Coll. Sci. Imp. Univ. Japan, vol. vii., 1894, pp. 111-200; abstract in Japan Seismol. Journ., vol. iii., 1894, pp. 71-80.
13. —— "A Note on the Great Mino-Owari Earthquake of October 28th, 1891." Pub. Earthquakes Inves. Com. in Foreign Languages, No. 4, Tokio, 1900, pp. 13-24.
14. —— "Sulla velocit di propagazione e sulla lunghezza delle onde sismiche." Ital. Soc. Sismol. Boll., vol. i., 1895, pp. 52-60—especially pp. 52-57.
15. —— "Sull' intensit e sull' ampiezza del movimento nel gran terremoto giapponese del 28 ottobre 1891." Ital. Soc. Sismol. Boll., vol. ii., 1896, pp. 189-200.
16. —— "Note on the After-shocks of the Mino-Owari Earthquake of Oct. 28th, 1891." Pub. Earthquakes Inves. Com. in Foreign Languages, No. 7, Tokio, 1902, pp. 27-32.
17. —— "Note on the relation between Earthquake Frequency and the Atmospheric Pressure." Tokyo Phys.-Math. Soc. Reports, vol. ii., 1904, No. 8.
18. TANAKADATE, A., and H. NAGAOKA. "The Disturbance of Isomagnetics attending the Mino-Owari Earthquake of 1891." Journ. Coll. Sci. Imp. Univ. Japan, vol. v., 1893, pp. 149-192.
 I have not referred to the results of this survey, for, though changes in all the magnetic elements (especially in horizontal intensity) have taken place between 1887 and 1891-92, these changes cannot be ascribed with confidence to the earthquake in the absence of a thorough knowledge of the secular variation.
 From the formula a=x*g/y, where a is the maximum horizontal acceleration, g the acceleration due to gravity, y the height of the centre of gravity, and x its horizontal distance from the edge about which the body was overturned.
 These estimates are made, on the supposition of simple harmonic motion, from the formula 2*a=alpha*t^2/(2*pi^2), where 2a is the total range or double amplitude, a the maximum acceleration, and t the period of the vibration.
 Professor Omori finds that the mean daily number of earthquakes y during the month x (reckoned from November 1891) may be approximately represented by the equation—
y = 16.9 / (x + 0.397);
or, taking the semi-daily earthquake numbers during the five days between October 29th and November 2nd, 1891, by the equation—
y = 440.7 / (x + 2.314),
where y denotes the number of earthquakes observed during the twelve hours denoted by x, the time being measured from the first half of October 29th. It is interesting to notice that, taking account of the mean annual frequency of earthquakes in ordinary years, the number of shocks observed at Gifu during the two years 1898-99 should, according to the latter formula, be 163; the actual number recorded was 160.
 The last violent shock before the end of 1893 occurred on September 7th, 1892, and its effects on the frequency of after-shocks is shown by the daily numbers recorded at Gifu during the first fortnight in September. These are—2, 2, 2, 3, 5, 5, 28 (on September 7th), 8, 8, 5, 4, 3, 2, 4, 3.
 The periodicity of after-shocks is discussed in the papers numbered 4, 12, 16, and 17 at the end of this chapter. In these, the existence of diurnal and other periods is clearly established. Professor Omori also shows that the mean daily barometric pressure is subject to fluctuations with maxima occurring on an average every 5-1/2 days, and that earthquakes are least frequent on the days of the barometric maxima and minima, and more frequent in the days immediately preceding and following them.
 Of the Japanese earthquakes of 1885-92 originating beneath the land, twenty-six per cent. were accompanied by a recorded sound; but less than one per cent. of those originating beneath the sea and not more than ten miles from the coast.
THE HEREFORD EARTHQUAKE OF DECEMBER 17TH, 1896, AND THE INVERNESS EARTHQUAKE OF SEPTEMBER 18TH, 1901.
Among the earthquakes described in this volume, the Hereford and Inverness earthquakes hold but a minor place. The damage to buildings, though unusual for this country, was slight when compared with that caused by the preceding shocks; there was no loss of life, not a single person was injured by falling masonry. The interest of the earthquakes lies entirely in the detailed study rendered possible by numerous observations of the shock and sound, and in the bearing of this evidence on the general theory of the origin of earthquakes.
THE HEREFORD EARTHQUAKE OF DECEMBER 17TH, 1896.
The principal earthquake of this series occurred at 5.32 A.M. on December 17th, and was preceded by at least nine minor shocks (the first of which was felt at about 11 or 11.30 P.M. on December 16th), and followed by two others on the same day, and by a third and last on July 19th, 1897. The accounts of these preliminary movements will be found on a later page, as their bearing will be more fully apparent after the discussion of the principal shock.
ISOSEISMAL LINES AND DISTURBED AREA.
On the map in Fig. 60, the continuous curves represent isoseismal lines corresponding to the degrees 8, 7, 6, 5, and 4 of the Rossi-Forel scale. The isoseismal 8, which is the most accurately drawn of the series, is an elongated oval, 40 miles long, 23 miles broad, and containing an area of 724 square miles. The longer axis is directed W. 44 N. and E. 44 S. Within this curve, there are 73 places where buildings are known to have been damaged, 55 places being in Herefordshire, 17 in Gloucestershire, and one in Worcestershire.
The most important damage occurred in the city of Hereford, which, in 1901, contained 4,565 inhabited houses. Here, no fewer than 218 chimneys had to be repaired or rebuilt. The Cathedral was slightly injured. The finial of a pinnacle of the Lady Chapel was thrown down, a fragment of a stone fell from one of the arches in the south transept, and the three pinnacles of the western front were fractured. Several churches suffered to a similar extent, while, at the Midland Railway Station, all the seven chimney-stacks were shattered. At Dinedor, Fownhope, Dormington, Withington, and a few other villages, the damage was also relatively greater than elsewhere, these places all lying within a small oval about 8-1/2 miles long, which surrounds, not the centre, but rather the north-west focus, of the isoseismal 8.
The isoseismal 7, which includes places where the shock was strong enough to overthrow ornaments, vases, etc., is also very nearly an ellipse, whose axes are 80 and 56 miles in length, and whose area is 3,580 square miles. Its longer axis, running from W. 42 N. to E. 42 S., is practically parallel to that of the inner curve. Next in succession comes the isoseismal 6, surrounding those places where the shock was strong enough to make chandeliers, pictures, etc., swing; but, as most of the observers seem to have slept in darkened rooms, the number of determining points for this curve is less than usual, and its course is therefore laid down with a somewhat inferior degree of accuracy. The error, however, is probably small, and we may therefore regard the isoseismal 6 as another ellipse, 141 miles long, 116 miles broad, and containing an area of 13,000 square miles. Its longer axis is again nearly parallel to those of the preceding isoseismals.
The next two isoseismals are nearly circular in form. It will be noticed that large portions of them, and especially of the isoseismal 4, traverse the sea. In these parts, the paths of the curves are to some extent conjectural. In drawing them, the chief guides are their trend before leaving the land and the known intensity along the neighbouring coastlines. The isoseismal 5 bounds the area within which the shock was perceptible as a sensible displacement and not merely a quiver. Its dimensions are 233 miles from north-west to south-east, and 229 miles from south-west to north-east, and its area 41,160 square miles. The isoseismal 4, which includes places where the shock was strong enough to make doors, windows, etc., rattle, is 356 miles from north-west to south-east, and 357 miles from south-west to north-east, and 98,000 square miles in area; its centre coincides nearly with that of the small oval area in the neighbourhood of Hereford, where the damage to buildings was relatively greater than elsewhere.
Outside the isoseismal 4, the earthquake was observed at several places. The shock was certainly felt at Middlesbrough, 12-1/2 miles from the curve, and probably at Killeshandra (in Ireland), 65 miles distant. Thus, if we consider the boundary of the disturbed area to coincide with the isoseismal 4, its area would be 98,000 square miles, or 1-2/3 that of England and Wales; if it were a circle concentric with the isoseismal 4, and passing through Middlesbrough, its area would be 115,000 square miles, or nearly twice that of England and Wales; while, if it passed through Killeshandra, its area would be 185,000 square miles, or more than three times the area of England and Wales.
Position of the Originating Fault.—The form, directions, and relative positions of the isoseismal lines furnish important evidence with regard to the originating fault. We conclude in the first place that its mean direction is parallel to the longer axes of the three innermost isoseismal lines—that is, north-west and south-east, or, more accurately, W. 43 N. and E. 43 S. In this case, the elongated forms of the isoseismal lines cannot be attributed to variations in the nature of the surface rocks. The district embraced contains about 13,000 square miles, and it is improbable that the axes of the three isoseismals should retain their parallelism over so large an area, if these variations had any considerable effect. Moreover, in the same district, an earthquake occurred in 1863, whose meizoseismal area was elongated from north-east to south-west, or almost exactly perpendicular to the direction in 1896.
Secondly, it will be noticed (Fig. 60) that the isoseismal lines are not equidistant from one another. On the north-east side, they are separated by distances of 20, 34, 55, and 51 miles; and on the south-west side by distances of 13-1/4, 25, 60, and 77 miles. It follows from this that the fault-surface must hade or slope towards the north-east; for, near the epicentre, the intensity is greatest and dies out more slowly on the side towards which the fault hades.
If we could ascertain any one place through which the fault passed, its position would thus be completely determined. Unfortunately, there is no decisive evidence on this point. There are, however, several places to the south-west of Hereford where the intensity of the shock was distinctly less than in the surrounding district, and it is possible that this was due to their neighbourhood to the fault-line (see p. 135). If so, the originating fault must have extended from a point about a mile and a half west of Hereford for a distance of about 16 miles to the south-east; and a fault in this position would certainly satisfy all the details of the seismic evidence.
NATURE OF THE SHOCK.
Throughout the disturbed area, considerable variations were observed in the nature of the shock. These changes were due to the mere size of the focus, to its elongated form and, as will be seen, to its discontinuity, and also to the distance of the place of observation from the epicentre.
At places near the epicentre, rapid changes in the direction of the shock were observed owing to the large angle subtended by the focus; while, at considerable distances, this angle being small, the changes of direction were imperceptible. A further variation with the distance was an increase in the period of the vibrations. Close to the epicentre, the general impression was that of crossing the wake of a steamer in a very short rowing-boat, or of riding in a carriage without springs. At distances of a hundred miles or more, the movement is described as being of a pleasant, gentle, undulating character, like that felt during the rocking of a ship at anchor or in a carriage with well-appointed springs.
The most remarkable feature of the shock, however, was its division into two distinct parts or series of vibrations, separated by an interval, lasting two or three seconds, of absolute rest and quiet. And this was no mere local phenomenon. With the exception of a narrow band that will be referred to presently, records of the double shock come from nearly all parts of the disturbed area, even from districts so remote as the Isle of Man and the east of Ireland. The two parts differed in intensity, in duration, and in the period of their constituent vibrations. For instance, at Oaklands (near Chard), a shivering motion was first felt, and then, after about three or four seconds, a distinct rocking from side to side. At Exeter, there was a sudden tremor lasting about two seconds, followed, after two or three seconds, by another and more severe shaking lasting four or five seconds. Again, at West Cross (near Swansea), an undulatory movement for about four seconds was followed soon after by a tremulous shock. At Liverpool, the durations of the first part, interval, and second part were respectively estimated at about six, two, and four seconds.
As a first result of the observations, then, it appears that in the south-east half of the disturbed area, the second part of the shock was the stronger, of greater duration and consisted of longer-period vibrations (as at a, Fig. 61); while, in the north-west half, the same features characterised the first part of the shock (b, Fig. 61). A closer examination of the records shows, however, that the boundary between the two portions of the disturbed area was not a straight line, but slightly curved, the concavity facing the south-east. The broken line on the map (Fig. 60), which is hyperbolic in form, represents roughly the position of this curved boundary.
Along this hyperbolic boundary-line, or rather within a narrow band of which it is the central line, the shock lost its double character, and was manifested as a single series of vibrations gradually increasing in intensity and then dying away. Close to the edges of this band, careful observers were able to distinguish two maxima of intensity connected by a continuous series of tremors (c, Fig. 61). Thus, within the band, the two series of vibrations, which elsewhere were isolated, must have been superposed on one another; while, near the edges of the band, the concluding tremors of the first series overlapped the initial tremors of the second.
Origin of the Double Series of Vibrations.—The Hereford earthquake thus belongs to the same class as the Neapolitan, Andalusian, Charleston, and Riviera earthquakes. As in these cases, the hypothesis of a single focus is inadmissible. The division of the disturbed area into two regions of opposite relative intensity, duration, etc., is sufficient proof that a single series of vibrations was not duplicated by reflection or refraction, or by separation into longitudinal and transverse waves. It is equally conclusive against a repetition of the impulse within the same focus. We must therefore infer that the focus consisted of two nearly or quite detached portions arranged along a north-west and south-east line, and that the impulse at the north-west focus was the stronger of the two. The only question that remains to be decided is whether the impulses at the two foci were simultaneous or not.
Now, if the impulses occurred at the same instant, the waves from the two foci would travel with the same velocity, and would therefore coalesce along a straight band which would bisect at right angles the line joining the two epicentres. But we have already seen that this band is curved, and it thus follows that the two impulses were not simultaneous. Again, since the concavity of the hyperbolic band faces the south-east, the waves from the north-west focus must have travelled farther than those from the south-east focus before the two met along the hyperbolic band; in other words, the impulse at the north-west focus must have occurred two or three seconds before the impulse at the other.
Position and Dimensions of the Two Foci.—There can be little doubt that the impulse at the north-west focus was responsible for the greater damage to buildings at Hereford, Dinedor, Fownhope, etc. The centre of its epicentral area must therefore lie about three miles south-east of Hereford. It is probable, also, that the corresponding centre of the other focus is similarly placed with respect to the south-east portion of the isoseismal 8—that is, about two or three miles north-east of Ross. These two points are eight or nine miles apart. Now, since, as we shall see, the mean surface-velocity of the earth-waves was about 3000 feet per second, and the mean duration of the quiet interval between the two series was 3-1/2 seconds, the nearest ends of the two foci must have been separated by a distance of not less than two miles. Moreover, since the series of vibrations from the north-west or Hereford focus lasted a few seconds longer than that from the south-east or Ross focus, the former must have been about two miles longer than the latter, and we may therefore estimate their lengths at about eight and six miles respectively. Including the undisturbed intermediate portion, this would give a total length of focus of about 16 miles, a result we have already inferred from the dimensions of the isoseismal 8.
DIRECTION OF THE SHOCK.
Although no question was asked with regard to the direction of the shock, no fewer than 469 observers made notes on this point. As a general rule, their determinations are extremely rough, few referring to more than the eight principal points of the compass. Moreover, in any one place, the directions assigned to the shock are very varied. For instance, in the city and suburbs of Birmingham, eight observers give the direction along a north and south line, eight east and west, eleven north-west and south-east, and five north-east and south-west, while there are five other intermediate estimates. But, when these directions are plotted on a map of the district, it is seen at once that they are either nearly parallel or perpendicular to the roads in which the observers were living; that is, the apparent direction of the shock was at right angles to one of the principal walls of the house. This, of course, is a result to be anticipated, for, whatever be the direction of the earthquake-motion, a house tends to oscillate in a plane perpendicular to one or other of its walls.
It is extraordinary to how great a distance the direction of the shock is perceptible. Records come from Brighton (137 miles from the epicentre), Maldon in Essex (144 miles), Harrogate (147 miles), Douglas in the Isle of Man (167 miles), Dublin (176 miles), and Baltinglass in Co. Wicklow (180 miles).
Nevertheless, whatever the distance may be, the sense of direction must be most perceptible in those houses whose principal walls are at right angles to the true direction of the earthquake-motion, and we should therefore expect to find the observations of direction most frequently made in such houses, or in others which approximate to this situation. Thus, the average of all the observations within a fairly small area should give a result not very far from the true direction of the shock; and, the smaller the area and the farther from the epicentre, the more reliable should be the result. Now, in Birmingham the mean direction of the shock is E. 39 N., which differs only by 2 from the line joining the city to the epicentre; in London it is E. 21 S., the difference being again 2. In other cases, the observations from different counties are grouped together, and the mean direction is taken to correspond to the centre of the county. Yet, even then, there is often a close agreement between the mean direction of the shock and the direction of the county-centre from the epicentre; the difference being not more than two or three degrees in the counties of Buckingham, Devon, Stafford, Warwick, and York. In other cases, where the deviation exceeds this amount, either the number of observations is small or the county is near the epicentre and so subtends a large angle.
Two results of some importance follow from this analysis: (1) that while, with a few isolated observations, the "method of directions" is almost sure to fail, with a large number of observations closely grouped, the position of the epicentre may be determined with a fair approach to accuracy; and (2) that, at any rate outside a radius of forty miles, the earth-waves travelled in approximately straight lines outwards from the epicentre.
COSEISMAL LINES AND VELOCITY OF EARTH-WAVES.
Coseismal lines were defined by Mallet as long ago as 1849, but, owing to the difficulty of ascertaining the correct time, they have so far been of little service in the investigation of earthquakes. In the case of the Hereford earthquake, the distances traversed by the earth-waves are small; but, on the other hand, the time-records are numerous and frequently trustworthy to the nearest minute. Rejecting all estimates earlier than 5.32 A.M., and later than 5.36, as well as a number at 5.35, there remain fairly good observations from 381 places, and exceptionally accurate ones from 33 places. The latter were obtained from signalmen and other careful observers who were in possession of Greenwich time, or who compared their watches shortly afterwards with well-regulated watches.
With evidence so abundant, a new method of drawing coseismal lines becomes possible. According to this method, each place of observation is indicated on the map by a mark corresponding to the particular minute recorded. If the records were quite correct, there would be a central area occupied by the marks corresponding to 5.32 A.M., surrounded by a series of zones in which the times were respectively 5.33, 5.34, and 5.35. The curves separating these zones would be coseismal lines corresponding to the times 5.32-1/2, 5.33-1/2, and 5.34-1/2.
Owing, however, to the inevitable inaccuracy of all the time-records, these different zones intrude on one another, and the coseismal lines have therefore to be drawn about half-way through the overlapping regions, special weight being attributed to the apparently more accurate observations.
The coseismal lines obtained in this manner are represented by the continuous curves in Fig. 62. The isoseismals, which are added for the sake of comparison, are indicated by the dotted lines. It will be seen that the coseismal lines are elongated in the same direction as the isoseismals, but to a less extent, and this no doubt is due to the fact that the epoch selected by the majority of observers was one not far from, and slightly preceding, that of the maximum intensity of the shock.
Now, the average distance between the two inner coseismals is 32-3/4 miles, between the two outer ones (so far as drawn) 35-1/6 miles, and between the first and third 67-1/6 miles. The mean surface-velocity between the two inner coseismals is therefore 2,882 feet per second, and between the two outer ones 3,095 feet per second. There is thus an apparent increase in the velocity with the distance, but the accuracy of the coseismal lines is unequal to establishing this as a fact. The mean surface-velocity of 2,955 feet per second between the first and third coseismals is probably, however, the most accurate estimate of the surface-velocity yet made in a slight earthquake.
Nature of the Sound.—The sound which accompanied the shock was of the same character as that heard during all great earthquakes. It is often described in such terms as a deep booming noise, a dull heavy rumble, a grating roaring noise, or a deep groan or moan; more rarely as a rustling or a loud hissing rushing sound. As a rule, it began faintly, increased gradually in strength, and then as gradually died away; and this no doubt is the reason why it sometimes appeared as if an underground train or waggon were approaching quickly, rushing beneath the observer, and then receding in the opposite direction. Occasionally, the sound was very loud, being compared to the noise of many traction-engines heavily laden passing close at hand, or to a heavy crash or peal of thunder. But its chief characteristic was its extraordinary depth, as if it were almost too low to be heard. According to one observer, it was a low rumbling sound, much lower than the lowest thunder; and another compared it to the pedal notes of a great organ, only of a deeper pitch than can be taken in by the human ear, a noise more felt than heard. It will be seen presently how the sound, from its very depth, was inaudible to many persons.
A few observers described the sound in terms like those quoted above, but by far the larger number compared it to some more or less well-known type, and in many cases the resemblance was so close that the observer at first attributed it to the object of comparison. The descriptions, which present great varieties in detail, may be classified as follows: (1) One or several traction-engines passing, either alone or heavily laden, sometimes driven furiously past; a steam-roller passing over frozen ground or at a quicker pace than usual; heavy waggons driven over stone paving, on a hard or frosty road, in a covered way or narrow street, or over hollow ground or a bridge; express or heavy goods trains rushing through a tunnel or deep cutting, crossing a wooden bridge or iron viaduct, or a heavy train running on snow; the grating of a vessel over rocks, or the rolling of a lawn by an extremely heavy roller; (2) a loud clap or heavy peal of thunder, sometimes dull, muffled or subdued, but most often distant thunder; (3) a moaning, roaring, or rough, strong wind; the rising of the wind, a heavy wind pressing against the house; the howling of wind in a chimney, a chimney or oil-factory on fire; (4) the tipping of a load of coal, stones, or bricks, a wall or roof falling, or the crash of a chimney through the roof; (5) the fall of a heavy weight or tree, the banging of a door, only more muffled, and the blow of a wave on the sea-shore; (6) the explosion of a boiler or cartridge of dynamite, a distant colliery explosion, distant heavy rock-blasting and the boom of a distant cannon; (7) sounds of a miscellaneous character, such as the trampling of many men or animals, an immense covey of partridges on the wing, the roar of a waterfall, the passage of a party of skaters, and the rending and settling together of huge masses of rock.
The total number of comparisons made was 1,264. Of these, 45.4 per cent. refer to passing waggons, etc., 15.0 per cent. to thunder, 15.5 to wind, 3.9 to loads of stones falling, 2.7 to the fall of a heavy body, 7.2 to explosions, and 10.3 per cent. to miscellaneous sounds.
Generally, the sound adhered throughout to one of the types mentioned above, and, if it varied at all, varied only in intensity. At some places, however, the character of the sound was observed to change. For instance, one person described it as like the rumbling of a train going over a bridge, with a terrific crash, such as is heard in a thunderstorm at the instant when the shock was strongest, the rumbling dying away afterwards for some seconds.
Inaudibility of the Sound to some Observers.—The total number of observers who give a detailed account of the earthquake is 2,681, and, of these, 59 per cent. state that they heard the sound, 23 per cent. give no information, while 18 per cent. distinctly say that they heard no sound; that is, roughly, out of every five observers, three heard the sound, one made no reference to it, and one failed to hear the sound.
In a few cases, no doubt, this failure was due to the distance of the observer, but this is far from being a complete explanation; for, in Herefordshire, six out of 179, and in Gloucestershire 17 out of 227, observers heard no sound. Nor is the peculiarity a local one, for at Clifton two out of five observers who were awake did not hear the sound, at Birmingham four out of 23, and in London, eight out of 18. Even in the same house, it would happen that one observer would hear a sound as of a heavily-laden traction-engine passing, while to another it was quite inaudible.
Again, a large number of observers who heard the sound expressly state that they were unconscious of any while the shock lasted. The noise at first resembled the approach of a steam-roller or traction-engine up the street, it became gradually louder, and then ceased more or less suddenly as the shock began; while, to others in the same places, the sound continued to grow in loudness until the strongest vibrations were felt.
Even when observers in the same place agreed in hearing the sound, it presented itself to them under different aspects. Thus, at Hereford, a crash or bomb-like explosion was noticed by some, but not by all, observers; at Ledbury, the sound according to one began like a rushing wind and culminated in a loud explosive report, another heard a noise like distant thunder, which ended when the shock began, while a third heard no sound at all. At places more distant from the epicentre, the same diversity, both in character and intensity, is manifested. Thus, at Birmingham, the accounts refer on the one hand to the distant approach of a train and the rising of the wind, on the other to the reports of large cannons and to a noise as if tons of dbris had been hurled against the wall of the house; at Bangor, to muffled thunder, wind through trees, and a loud rumbling sound.
The first explanation of these apparent anomalies which presents itself is inattention on the part of the observers; but it is one that will not bear examination, though it may apply in some cases. The sound is too loud, at any rate near the epicentre, to escape notice, and it is generally heard before the shock begins to be felt. Moreover, as described in the last chapter, three out of every four earthquakes in Japan are unaccompanied by recorded sound, and the Japanese as a race cannot be accused of such constant inattention. The defect, it can hardly be doubted, is inherent to the observer, and not dependent on the conditions in which he is placed.
That the higher limit of audibility varies with different persons has long been known; and there can be no reason for doubting that there is a similar variability in the lower limit. Thus, to some observers, the sound remains inaudible throughout, however intently they may be listening. Again, it is found that, the deeper the sound, the greater must be the strength of the vibrations required to render them audible. As the vibrations which reach an observer increase in period, it may therefore happen that, sooner or later, the strength of some does not attain or exceed that limiting value, and, at that moment, the sound will cease to be heard. Moreover, for vibrations of a given period, this limiting value varies for different persons. Thus, to one observer, the sound may become inaudible, while another may continue to hear it. Lastly, the vibrations which affect an observer at any moment are of various strength and period. One may hear all perhaps, while a second may be able to hear some and not others. Thus, to one observer, the sound may be like a rising wind, to another like a heavy traction-engine passing; one may hear the crashes which accompanied the strongest part of the shock, while a second may be deaf to the same vibrations; to one the sound may become continually louder and cease abruptly, to another it may increase to a maximum and then die away.
Sound-Area.—While the sound was a very prominent feature of the earthquake in and near the epicentral area, records at a great distance are naturally difficult to obtain, and, on this account, the number of stations for determining the boundary of the sound-area is too small to allow of it being accurately drawn. As a rule, however, it must lie between the isoseismals 5 and 4, but it is less nearly circular than either of these lines. Its length, from north-west to south-east, is 320 miles, its breadth 284 miles, and the area contained by it about 70,000 square miles, or roughly two-thirds that of the disturbed area.
Isacoustic Lines.—The dotted lines in Fig. 60 represent isacoustic lines—that is, lines which pass through all places where the percentage of observers who recorded their perception of the sound is the same. For instance, if we take any point in the line marked 80 and describe a small circle with that point as centre, then 80 per cent. of the observers within that circle would hear the earthquake-sound. The isacoustic lines thus show how the audibility of the sound varies throughout the sound area. To draw the curves with a close approach to accuracy, the unit of area should be small and of constant dimensions; but, in the present case, owing to the comparative paucity of the observations, a smaller unit than the county would give unreliable results. At the centre of each county, the sound audibility may be regarded as proportional to the percentage of the total number of observers within the county who distinctly heard the sound. To draw the curve marked 50, the centre of every county in which the average percentage is less than 50 is joined to the centres of those adjoining counties in which it is above 50, and these lines are then divided in the proper ratio so as to give a point where the percentage would be exactly 50. A number of points at which the percentage is 50 is thus obtained, and the curve drawn through them is the required isacoustic line. The percentage of audibility varies from 87 in Herefordshire to 23 in Essex and the east of Ireland, but the only isacoustic lines which can be completely drawn are those that correspond to the percentages between 80 and 50 inclusive.
The peculiar form of the isacoustic lines will be evident at a glance. They bear little relation to the isoseismal lines. Their greatest extensions are not along the axes of those lines, but in two directions which are a little east of north-east and south of south-west. They lie indeed along a hyberbolic line which, towards the south-west, agrees closely with the curvilinear axis of the hyperbolic band represented by the broken line in Fig. 60. Towards the north-east, the coincidence is not so close, but this is chiefly owing to the magnitude of the northern counties, which causes a deflection of the isacoustic lines towards the north.
It will be remembered that the hyperbolic band is the area within which the vibrations from the two foci were superposed. Now, the sound accompanied each part of the shock, and ceased entirely during the interval between them. Also, the stronger series of vibrations was accompanied by the louder sound; but, while the difference in strength was considerable between the two parts of the shock, it was very slight between the two sounds. There is therefore no marked distortion of the isoseismal lines when crossing the hyperbolic band, while the isacoustic lines are completely diverted from their normal course.
Thus, the study of the isacoustic lines strongly confirms the conclusions at which we have arrived above (p. 223)—namely, that there were two distinct foci arranged in a north-west and south-east line, and that the impulse at the former focus occurred a few seconds earlier than that at the latter.
Variations in the Nature of the Sound throughout the Sound-area.—In one respect, the sound exhibited a marked uniformity all over the sound-area—namely, in its great depth; the word "heavy" being used in one out of every four accounts of the sound, whether close to the epicentre or near the boundary of the sound-area.
The type of comparison employed varies in different parts of the sound-area. As we recede from the origin, the sound becomes on the average less like thunder or explosions and more like wind. The references to passing waggons, etc., are so numerous that it is possible to draw curves, in the same way as isacoustic lines, which represent equal percentages of comparison to this type out of the total number of comparisons. The curves are somewhat incomplete, but it is noteworthy that those corresponding to the higher percentages cling to the extremities of the hyperbolic band, probably because the uninterrupted duration of the sound is greater there than elsewhere.
The effect of distance from the epicentre, however, is most noticeable in connection with changes in the character of the sound. It is only on the immediate neighbourhood of the origin that the explosive reports or crashes were heard in the midst of the rumbling sound. At a moderate distance, the sound before and after the shock became smoother, while the sound which accompanied the shock retained to a certain extent its rougher and more rumbling or grating character. Close to the boundary of the sound-area, the irregularities were still further smoothed away, and the only sound heard was like the low roll of distant thunder.
The explanation of these changes depends on the fact that, as we recede from the epicentre, the vibrations of every period tend to become inaudible. The limiting vibrations of the whole series will be the first to be lost, especially those of the longest period. Thus, near the epicentre, sound-vibrations of many different periods will be heard, and the sound will be more complex than it is elsewhere. The greater the distance, the narrower are the limits with regard to period between which the audible vibrations lie, until, near the boundary of the sound-area, the sound becomes an almost monotonous deep growl of nearly uniform intensity.
Time-relations of the Sound and Shock.—The principal epochs to be compared are the beginning, the epoch of maximum intensity, and the end. The beginning of the sound preceded that of the shock in 82 per cent. of the observations on this epoch, coincided with it in 12, and followed it in 6 per cent.; the epoch of maximum intensity preceded that of the shock in 21 per cent. of the records, coincided with it in 73, and followed it in 6 per cent.; while the end of the sound preceded that of the shock in 22-1/2 per cent., coincided with it in 27-1/2, and followed it in 50 per cent. Thus, as a general rule, the beginning of the sound preceded that of the shock, the sound was loudest when the shock was strongest, and the end of the sound followed that of the shock. In other words, the duration of the sound was in most cases greater than that of the shock.
Of the twelve undoubted minor earthquakes, nine occurred before, and three after, the principal shock, the times of the first eleven lying between limits about seven hours apart. With three exceptions, the records are insufficient to determine the positions of the epicentre with any approach to exactness.
The first occurred at about 11 or 11.30 P.M. on December 16th. The boundary of the disturbed area, which coincides nearly with that of the fifth shock (E, Fig. 63), is 97 miles long from north-west to south-east, 83 miles wide, and contains about 6,300 square miles. The focus was apparently situated between the two foci of the principal earthquake and partly coincided with them.
Then came three slight shocks (at about 1 A.M. on December 17th, 1.30 or 1.45 A.M., and 2 A.M.), about which little is known except that they probably originated somewhere near the Ross focus.
The fifth shock (E, Fig. 63) occurred at about 3 A.M., and disturbed an area 104 miles in length, 79 miles in width, and about 6,400 square miles in area. Its boundary occupies approximately the position that would be taken by an isoseismal of intensity between 7 and 6 of the principal earthquake. We may therefore infer that this shock and the principal earthquake were caused by slips along the same fault and in about the same region of the fault. Also, as there is no evidence of discontinuity in the vibrations of the minor shock, it is probable that the focus was continuous, and occupied the space between the two foci of the principal earthquake, as well as part or the whole of both these foci.
The next four shocks occurred at about 3.30, 4, 5, and 5.20 A.M., and were more closely associated with the Ross than with the Hereford focus, and then followed the principal earthquake at 5.32 A.M.
A few minutes later, at 5.40 or 5.45 A.M., a very slight shock was felt, the focus of which was possibly situated in the central region between the two foci. The next, at about 6.15 A.M. (K, Fig. 63), disturbed an area 41 miles long, 27 miles broad, and containing about 870 square miles. Its focus must have coincided approximately with the Ross focus of the principal earthquake, and this was also the case probably with the last shock of all, which occurred on July 19th, 1897, at 3.49 A.M.
ORIGIN OF THE EARTHQUAKES.
The greater part of the epicentral district is covered by a sheet of Old Red Sandstone (Fig. 64), but, just to the north-east of the position laid down for the originating fault (indicated by the straight broken line), is the well-known Woolhope anticlinal, by which Silurian beds are brought to the surface. The anticlinal axis runs approximately north-west and south-east, and is thus roughly parallel to the earthquake-fault. Moreover, the thinning-out and occasional disappearance of some of the Silurian beds on the south-west side of the anticlinal (as compared with those on the north-east side) is suggestive of a north-west and south-east fault or rapid flexure at or near the south-west junction of the Old Red Sandstone and the Silurian strata. If it be a fault, it must hade to the north-east, and would therefore satisfy two of the conditions determined by the seismic evidence. It would lie, however, about two miles too far to the north-east, being in fact to the north-east of the villages which suffered most from the earthquake.
But only a few miles to the south-east of the Woolhope anticlinal, and almost in the same line with it, there is a second anticlinal, that of May Hill. This is a triangular area, and is known to be bounded on all three sides by faults. The fault on the north-east side has an average north-west and south-east direction, and, if it were continued through the Old Red Sandstone towards the north-west, but bending at first a few degrees more to the west, it would pass through a point about 1-1/2 miles west of Hereford. It is worthy of notice that both this fault and another nearly parallel to it, about half-a-mile farther north-east, stop, according to the Geological Survey map, at the points where they enter the Old Red Sandstone. The latter is an area which has never been investigated with thoroughness by modern stratigraphical methods, and in which it is difficult to trace faults. It therefore appears not improbable that the earthquakes were due to slips along a continuation of this fault.
Whether this be the case or not, however, it is clear that the earthquake-fault must pass between the anticlinal areas of Woolhope and May Hill, the former being on the north-east, and the latter on the south-west, side of the fault. At the Hereford focus, the fault must hade to the north-east; and, at the Ross focus, it is probable, from the distribution of places where damage occurred to buildings, that it hades to the south-west If this be the case, the fault must change in hade between the two foci.
How long a time had elapsed since the last sign of growth in the earthquake-fault took place, it is impossible to say; but it must be many years in length. During this interval, the stresses tending to produce movement along the fault-service had been gradually increasing, until they were sufficient to overcome the resistance opposed to them. It is worthy of notice that the earliest perceptible movements were slight. Their function seems to have been to prepare the way for the great slips by equalising the difference between stress and resistance over a large area of the fault-surface. We cannot trace with accuracy the transference of the seat of movement from one part of the fault-surface to another. The first slip seems to have taken place chiefly in the region between the two foci of the principal earthquake; possibly it overlapped both of them partly. The next three slips were apparently in the neighbourhood of the Ross focus, and were followed by a fifth in the same area as the first. Then came a series of small movements that we cannot locate further than by saying that they were more closely connected with the Ross focus than the other.
In consequence of the preliminary slips within and near the Ross focus, the effective stress in that portion of the fault was diminished; and this may be the reason why the first great slip took place at the Hereford focus. The immediate result of such a movement would naturally be an increase of stress in and beyond the terminal regions, and the next slip might have been expected in an area partly overlapping the Hereford focus, and either to the north-west or south-east of it. Instead of this, for a distance of two miles in the latter direction, there was not the least perceptible movement during the principal earthquake, and the second great slip occurred in the region beyond occupied by the Ross focus. This second slip, moreover, occurred within two or three seconds after the other; that is, before the earth-waves had time to travel from the Hereford to the Ross focus. In other words, the slip at the Ross focus was not a consequence of the slip at the Hereford focus; but both were due to a single generative effort.
Now, a section drawn parallel to the earthquake-fault and on the north-east side of it, would show an anticline near the Hereford focus and a corresponding syncline near the Ross focus, with an undisplaced portion in the intermediate region; while a parallel section on the other side of the fault would show a syncline near the Hereford focus, an anticline near the Ross focus, and again an undisplaced portion in the intermediate region. If further movements tending to accentuate such a structure were to occur (that is, if the anticlinals were to be made more anticlinal and the synclines more synclinal), there would therefore be two slips, one in each focus; while, along the fault-surface between, there would be practically no displacement. At any rate, the earlier stresses in that region may have been fully relieved by two slight preliminary slips (those causing the first and fifth minor earthquakes), and those resulting from the great displacements by the first after-slip which followed in about ten minutes.
Half-an-hour later, another slip took place at the Ross focus, and by this the equilibrium of the rock-masses was almost completely restored; for we have no certain evidence of any further movements until seven months have elapsed (July 19th, 1897), when there was a final slip in the same region of the fault.
THE INVERNESS EARTHQUAKE OF SEPTEMBER 18TH, 1901.
Between the north-east end of Loch Ness and the Moray Firth at Inverness, there lies a tract of land not more than seven miles in length, which is notable as one of those most frequently shaken by earthquakes in the British Islands. In the intensity of its shocks it is inferior to the south-east of Essex and the centre of Herefordshire, and, in mere number, to the celebrated village of Comrie in Perthshire. But, in the interest of its seismic phenomena, in the light which they cast on the development of the earth's crust, the neighbourhood of Inverness has no equal in Great Britain, and not many superiors in any part of the world.
For this importance from a seismological point of view, the district is indebted to the great fault which traverses Scotland along the line of the Caledonian Canal, and to the fact that this fault, although it dates from Old Red Sandstone times, has not yet finished growing. As results of its formation, we have the almost straight cliff along the south-east coast of Rossshire, and the long chain of lakes, beginning with Loch Dochfour and Loch Ness, and ending with Loch Oich, Loch Lochy, and Loch Linnhe. As evidences of its persistent though intermittent growth, we have the slight tremors and earth-sounds occasionally observed at and near Fort William, and the much stronger shocks felt in the neighbourhood of Inverness.
During the nineteenth century there were three strong earthquake shocks in this district. The first and most severe occurred on August 13th, 1816, and was felt over the greater part of Scotland; the second on February 2nd, 1888; and the third and weakest on November 15th, 1890. This last shock was followed by several slighter ones, the series ending with a rather smart shock on December 14th. Between this date and the summer of 1901 no earthquakes seem to have been felt at or anywhere near Inverness.
The date of the first shock of 1901 is not quite certain. One is said to have been felt at Aldourie (see Fig. 66) some time in June, and a second at Dochgarroch in July. These may have been succeeded by others too slight to attract much notice, but the first to be generally observed occurred on September 16th at 6.4 P.M. A weak tremor, accompanied by a faint sound, was perceived over a nearly circular area about 12 miles in diameter, and with its centre about 1-1/2 miles south of Dochgarroch. On the next day, at 11 P.M., a quivering lasting two seconds was felt at Inverness, and a weak tremor, accompanied by sound, at Dochgarroch at 1.15 A.M. on September 18th. Nine minutes later, at 1.24 A.M., occurred the principal earthquake, the shock of which would be called a strong one, even in Italy and Japan.
EFFECTS OF THE SHOCK.
In Inverness, the damage to buildings, though seldom serious, was by no means inconsiderable. One brick building used as a smithy was destroyed, several chimneys or parts of them fell, and many chimney-cans were displaced or overthrown. At Dochgarroch and other places within the meizoseismal area, walls were cracked, chimneys thrown down, and lintels loosened.
But, for this country, an unusual effect of the earthquake was a long crack made in the north bank of the Caledonian Canal near Dochgarroch Lochs. It occurred in the middle of the towing-path, and could be traced at intervals for a distance of 200 yards to the east of the Lochs, and 400 yards to the west, being often a mere thread, and in no place more than half-an-inch wide. Soon after its formation, however, the fissure was obliterated by heavy showers of rain.
ISOSEISMAL LINES AND DISTURBED AREA.
The map (Fig. 65) shows the area over which the earthquake was perceptible. The isoseismal lines are drawn partly continuous and partly dotted—continuous where some confidence can be placed in their accuracy, and dotted where their course must be regarded as doubtful, owing to the rarity or absence of observations.
The innermost isoseismal (shown on a larger scale in Fig. 66) corresponds to the intensity 8 of the Rossi-Forel scale, and includes the places where the shock was strong enough to cause slight structural damage to buildings. It is elliptical in form, 12 miles long, 7 miles broad, and 67 square mile in area, with its centre at a point about 1-1/2 mile east-north-east of Dochgarroch, and its longer axis running N. 33 E. and S. 33 W.
The remaining isoseismals are less accurately drawn, owing to the scarcity of observations made in the west of Scotland. Except towards the west, however, the course laid down for the isoseismal 7 may be trusted. Its length is 53-1/2 miles, width 35 miles, and area 1,500 square miles. Its longer axis is almost exactly parallel to that of the preceding isoseismal, but the distance between the two curves is 9 miles on the north-west, and 14 miles on the south-east, side. The isoseismal 6 is 105 miles long, 87 miles wide, and contains 7,300 square miles; and the isoseismal 5, 157 miles long, 143 miles wide, and about 17,000 square miles in area.
The isoseismal 4 may be regarded as the boundary of the disturbed area of the earthquake, for, so far as known, the shock was not noticed at any point outside it. Towards the north, it was felt at Wick, Castletown, and other intermediate places; towards the west at Tobermory in the island of Mull; and, towards the south, at Skelmorlie (in Ayrshire), Paisley, Belsyde (near Linlithgow), Gullane (near North Berwick), and Dunbar. Along the east coast of Scotland, between Wick and Dunbar, there are few places of any size where the shock was not felt. The disturbed area of the earthquake is thus 215 miles long from north-east to south-west, 198 miles wide, and contains about 33,000 square miles.
Position of the Originating Fault.—The only isoseismals which are drawn accurately enough to determine the earthquake-fault are the two inner ones, those marked 8 and 7; but these are sufficient for the purpose. It is clear, from the direction of their longer axes, that the average direction of the fault must be N. 33 E. and S. 33 W. Again, the isoseismals are farther apart towards the south-east than towards the north-west, implying that the fault hades to the south-east. Lastly, as the intensity of the shock is greater on the side towards which the fault hades, it follows that the fault-line must lie a short distance (about a mile or so) on the north-west side of the centre of the isoseismal 8.
Now, the great fault alluded to above occupies almost exactly the position indicated by the seismic evidence. Its mean direction from Tarbat Ness to Loch Linnhe is N. 35 E. and S. 35 W., it hades to the south-east, and the fault-line passes through a point about three-quarters of a mile to the north-west of the centre of the isoseismal 8 (Fig. 66). There can be little doubt, therefore, that the earthquake was caused by a slip of this fault; and the evidence of the after-shocks, as will be seen, offers additional support to this conclusion.
The region in which the slip took place may be determined roughly from the position and form of the innermost isoseismal. Its centre must have been close to the point marked A in Fig. 66, which corresponds to a point about 1-1/2 mile east-north-east of Dochgarroch. In a horizontal direction, its length must have been at least five or six miles; otherwise, the isoseismal 8 would have been less elongated. It must therefore have reached from about half-a-mile north-east of Loch Ness to about half-a-mile south-west of Inverness. Its width, measured along the dip of the fault-surface is unknown; but the small distance between the centre of the isoseismal and the fault-line shows that the principal movement took place at a depth which was probably under, rather than over, one mile.
NATURE OF THE SHOCK.
We come now to the evidence afforded by the nature of the shock, in which there was but little variation throughout the disturbed area. At Inverness, a gentle movement was first felt, followed by an extraordinary quivering, which increased in force for two or three seconds, and then decreased for two or three seconds; just as the quivering was about to cease, there was a distinct lurch or heave, after which the vibration was much more severe than before and lasted several seconds longer than the first part of the shock. Dalarossie lies about fourteen miles south-east of Inverness, and here the first indication was a loud sound, as of an express train, coming from the east, rushing close to, and then under, the house; this lasted for a few seconds, and towards the end of it the house vibrated. Then succeeded an interval of quietness for about a second, followed by a terrific burst or crash, not unlike the crash of a loud thunder peal, of about two seconds' duration, during which the house distinctly heaved up once and then sank back. After another brief interval of quietness, there was a low rumble, like the sound of a dying peal of thunder.
It will be noticed, in this account, that the two parts of the shock were no longer consecutive. There was a short interval of rest between them, the intermediate vibrations observed at Inverness being too weak to be felt at Dalarossie. Still farther away, the extinction became more marked. At Aberdeen, for instance, the shock consisted of two parts, the first a tremble, followed, after an interval of a few seconds, by a swinging movement of longer duration than the tremble.
In all parts of the disturbed area, the shock maintained the same character of division into two parts, the second of which was of greater duration and intensity than the first and consisted of vibrations of longer period. A phenomenon of such wide occurrence was clearly not due to local influences. It must have been caused by two separate initial impulses, the stronger succeeding the other after an interval of a few seconds and taking place in nearly the same region of the fault.
Outside the isoseismal 5, there are but few records of the earthquake-sound; but it was heard faintly at Skelmorlie (in Ayrshire), Belsyde (near Linlithgow), and Gullane (near North Berwick). Towards the north, it was not observed beyond Wick and Wathen (in Caithness). The boundary of the sound-area cannot be laid down with any approach to accuracy, but it must have included a district containing about 27,000 square miles.
Throughout the whole disturbed area, 84 per cent. of the observers heard the sound. The percentage varies in different counties, from 93 in Inverness-shire to 77 in the counties of Perth and Aberdeen; but the records in the more distant regions are too few to allow of the construction of isacoustic lines.
In its character, the sound resembled that usually heard with strong earthquakes, 39 per cent. of the observers having compared it to passing waggons, traction-engines, etc., 25 per cent. to thunder, 14 to wind, 8 to loads of stones falling, 3 to the fall of heavy bodies, 4 to explosions or the firing of heavy guns, and 7 per cent. to miscellaneous sounds. The intensity of the sound gradually diminished outwards from the epicentre, and most rapidly near the isoseismal 7, which abounds approximately the area in which the sound was very loud from that in which it was distinctly fainter, and also includes nearly all the places at which loud explosive crashes were heard with the strongest vibrations.
In the time-relations of the sound and shock, the Inverness earthquake resembles the Hereford earthquake of 1896. The beginning of the sound preceded that of the shock in 72 per cent. of the records, coincided with it in 20, and followed it in 8 per cent.; the epoch of maximum intensity of the sound preceded that of the shock in 20 per cent. of the records, coincided with it in 73, and followed it in 7 per cent.; while the end of the sound preceded that of the shock in 15 per cent. of the records, coincided with it in 34, and followed it in 52 per cent.
Somewhat similar proportions hold over the greater part of the disturbed area, the percentages being nearly the same in the counties of Inverness, Ross, Nairn, Elgin, Banff, and the most distant counties. But in Aberdeenshire an exception occurs, the three epochs of sound and shock in most cases coinciding with one another. The majority of the observations in this county come from the southern part, and the line joining this district to the epicentre is nearly perpendicular to the line of the earthquake-fault. This result has an important bearing on the origin of the sound-vibrations. For, if the general precedence of the sound with respect to the shock were due to its superior velocity, the percentage of records in which the beginning of the sound preceded that of the shock would vary only with the distance, and not with the direction from the origin. Indeed, with increasing distance from the origin, this percentage should continually approach 100; while that in which the end of the sound followed that of the shock should diminish to zero. There is, however, no trace of either tendency, the sound being heard after the shock at places close to the boundary of the sound-area. On the other hand, it the sound-vibrations were to start simultaneously, or nearly so, from all parts of the focus, but especially from its marginal regions, then, in the greater part of the disturbed area, the sound would be heard both before and after the shock; for the lateral margins of the focus would be the portions nearest to, and farther from, most observers; while, at places near the line through the epicentre at right angles to the earthquake-fault, the three principal epochs of the sound and shock should approximately coincide.
The inference that the sound-vibrations heard before and after the shock come from the margins of the focus is also supported by the observations on the relative duration of the sound and shock. If we take only those records which are free from doubt, in 78 per cent. of the total number, the duration of the sound was greater than that of the shock; while, in Aberdeenshire, according to 93 per cent. of the observers, the durations of sound and shock were equal.