That it is by means of the gravel and other detritus pushed along the bottom by the current, rather than by the mere friction of the water on its bed, that a river excavates its channel, is most strikingly shown immediately below a lake. In traversing a lake, the tributary streams are filtered.

stream.

Depositing their sediment on the floor of the lake, they unite in the clear transparent river which escapes at the lower end. The Rhone, for instance, flows into the Lake of Geneva as a turbid river; it issues from that great reservoir at Geneva as a rushing current of the bluest, most translucent water which, though it sweeps over ledges of rock, has not yet been able to grind them down into a deep gorge. The Niagara, also, filtered by Lake Erie, has not acquired sediment enough to enable it to cut deeply into the rocks over which it foams in its rapids before throwing itself over the great Falls.

One of the most characteristic features of streams is the singularly sinuous courses they follow. As a rule, too, the flatter the ground over which they flow, the more do they wind. Not uncommonly they form loops, the nearest bends of which in the end unite, and as the current passes along the now straightened channel, the old one is left to become by degrees a lake or pond of stagnant water, then a marsh, and lastly, dry ground. We might suppose that in flowing off the land, water would take the shortest and most direct road to the sea. But this is far from being the case. The slightest inequalities of level have originally determined sinuosities of the channels, and trifling differences in the hardness of the banks, in the accumulation of sediment, and in the direction of the currents and eddies have been enough to turn a stream now to one side now to another, until it has assumed its present meandering course. How easily this may be done can be instructively observed on a roadway or other bare surface of ground. Seen when quite dry and smooth, hardly any depressions in which water would flow might be detected on such a surface. But after a heavy shower of rain, runnels of muddy water will be seen coursing down the slope in serpentine channels that at once recall the winding rivers of a great drainage-system. The slightest differences of level have been enough to turn the water from side to side. A mere pebble or projecting heap of earth or tuft of grass has sufficed to cause a bend. The water, though always descending, has only been able to reach the bottom by keeping the lowest levels, and turning from right to left as these guided it.

course.

When a river has once taken its course and has begun to excavate its channel, only some great disturbance, such as an earthquake or volcanic eruption, can turn it out of that If its original pathway has been a winding one, it goes on digging out its bed which, with all its bends, gradually sinks below the level of the surrounding country. The deep and picturesque gorge in which the Moselle winds from Trèves to Coblenz has in this way been slowly eroded out of the undulating tableland across which the river originally flowed.

In another and most characteristic way, the shape of the ground and the nature and arrangement of the rocks over which they flow, materially influence rivers in the forms into which they carve their channels. The Rhone and the Niagara, for instance, though filtered by the lakes through which they flow, do not run far before plunging into deep ravines. Obviously such ravines cannot have been dug out by the same process of mechanical attrition whereby riverchannels in general are eroded. Yet the frequency of gorges in river scenery shows that they cannot be due to any exceptional operation. They may generally be accounted for by some arrangement of rocks wherein a bed of harder material is underlain by one more easily removable. Where a stream, after flowing over the upper bed, encounters the decomposable bed below, it eats away the latter more rapidly. The overlying hard rock is thus undermined, and, as its support is destroyed, slice after slice is cut away from it. The waterfall which this kind of structure produces continues to eat its way backward or up the course of the stream, so long as the necessary conditions are maintained of hard rocks lying upon soft. Any change of structure which would bring the hard rocks down to the bed of the channel, and remove the soft rocks from the action of the current and the dash of the spray would gradually destroy the waterfall. It is obvious that, by cutting its way backward, a waterfall excavates a ravine.

The renowned Falls of Niagara supply a striking illustration of the process now described. The vast body of water which issues from Lake Erie, after flowing through a level country for a few miles, rushes down its rapids and then plunges over a precipice of solid limestone. Beneath this hard rock lies a band of comparatively easily eroded shale. As the water loosens and removes the lower rock, large portions of the face of the precipice behind the Falls are from time to time precipitated into the boiling flood below. The waterfall is thus slowly prolonging the ravine below the Falls. The magnificent gorge in which the Niagara, after its tumultuous descent, flows sullenly to Lake Ontario is not less than 7 miles long, from 200 to 400 yards wide, and from 200 to 300 feet deep. There is no reason to doubt that this chasm has been entirely dug out by the gradual recession of the Falls from the cliffs at Queenstown, over which the river at first poured. We may form some conception of the amount of rock thus removed from the estimate that it would make a rampart about 12 feet high and 6 feet thick, extending right round the whole globe at the equator. Still more gigantic are the gorges or cañons of the Colorado and its tributaries in Western America. The Grand Cañon of the Colorado is 300 miles long, and in some places more than 6000 feet deep (Fig. 9). The country traversed by it is a network of profound ravines, at the bottom of which the streams flow that have eroded them out of the table-land.