FIG. 66.—Diagrams to show how the compression of a district and its consequent contraction during an earthquake may close up the joint spaces within the rock basement and concentrate the contraction of the overlying mantle where this is partially cut through and so weakened in the valley sections.] Water pipes laid in the ground at a depth of some feet may be bowed up into an arch which appears above the surface; lines of curbing are raised into broken arches, and the tracks of railways are thrown into local loops and kinks which imply a very considerable local contraction of the surface (Fig. 64). With unvarying regularity railway or other bridges which cross rivers or ravines, if the structures are seriously damaged, indicate that the river banks have drawn nearer together at the time of the disturbance. In such cases, whenever the bridge girder has remained in place upon its abutments, these have either been broken or back-tilted as a whole in such a manner as to indicate an approach of the foundations which was prevented at the top by the stiffness of the girder (Fig. 65). [Illustration: FIG. 67.—Map of the Chedrang fault which made its appearance during the Assam earthquake of 1897. The figures give the amounts of the local vertical displacement measured in feet (after R. D. Oldham).] The simplest explanation of such an approach of the banks at the sides of the valleys cut in loose surface material is to be found in a general closing up of the joint spaces within the underlying rock, and an adjustment of the mantle upon the floor mainly in the valley sections (Fig. 66). [Illustration: FIG. 68.—Map giving the displacements in feet measured along an earthquake fault formed in Alaska in 1899 (after Tarr and Martin).] =The plan of an earthquake fault.=—In our consideration of earthquake faults we have thus far given our attention to the displacement as viewed at a single locality only. Such displacements are, however, continued for many miles, and sometimes for hundreds of miles; and when now we examine a map or plan of such a line of faulting, new facts of large significance make their appearance. This may be well illustrated by a study of the plan of the Chedrang fault which appeared at the time of the Assam earthquake of 1897 (Fig. 67). From this map it will be noticed that the upward or downward displacement upon the perpendicular plane of the fault is not uniform, but is subject to large and _sudden_ changes. Thus in order the measurements in feet are 32, 0, 18, 35, 0, 8, 25, 12, 8, 2, 0. The fault formed in 1899 upon the shores of Russell Fjord in Alaska (Fig. 68) reveals similar sudden changes of throw, only that here the direction of the movement is often reversed; or, otherwise expressed, the upthrow is suddenly transferred from one side of the fault to the other. Such abrupt changes in the direction of the displacement have been observed upon many earthquake faults, and a particularly striking one is represented in Fig. 69.