I take this occasion to acknowledge myself indebted to Mardigny's interesting memoir just quoted for all the statements I make respecting the floods of the Ardeche, except the comparison of the volume of its water with that of the Nile.] It is easy to see that the damage occasioned by such floods as I have described must be almost incalculable, and it is by no means confined to the effects produced by overflow and the mechanical force of the superficial currents. In treating of the devastations of torrents, I have hitherto confined myself principally to the erosion of surface and the transportation of mineral matter to lower grounds by them. The general action of torrents, as thus fur shown, tends to the ultimate elevation of their beds by the deposit of the earth, gravel, and stone conveyed by them; but until they have thus raised their outlets so as sensibly to diminish the inclination of their channels--and sometimes when extraordinary floods give the torrents momentum enough to sweep away the accumulations which they have themselves heaped up--the swift flow of their currents, aided by the abrasion of the rolling rocks and gravel, scoops their beds constantly deeper, and they consequently not only undermine their banks, but frequently sap the most solid foundations which the art of man can build for the support of bridges and hydraulic structures. [Footnote: In some cases where the bed of rapid Alpine streams is composed of very hard rock--as is the case in many of the valleys once filled by ancient glaciers--and especially where they are fed by glaciers not overhung by crumbling cliffs, the channel may remain almost unchanged for centuries. This is observable in many of the tributaries of the Dora Baltea, which drains the valley of Aosta. Several of these small rivers are spanned by more or less perfect Roman bridges--one of which, that over the Lys at Pont St. Martin, is still in good repair and in constant use. An examination of the rocks on which the abutments of this and some other similar structures are founded, and of the channels of the rivers they cross, shows that the beds of the streams cannot have been much elevated or depressed since the bridges were built. In other cases, as at the outlet of the Val Tournanche at Chatillon, where a single rib of a Roman bridge still remains, there is nothing to forbid the supposition that the deep excavation of the channel may have been partly effected at much later period. The Roman aqueduct known as the Pont du Gard, near Nismes, was built, in all probability, nineteen centuries ago. The bed of the river Gardon, a rather swift stream, which flows beneath it, can have suffered but slight depression since the piers of the aqueduct were founded.]