In addition to the natural increase in the eroding power of the Somme brought about by the increase in its volume, on account of the greater precipitation in the Glacial age, there would also be, as Prestwich has well shown, a great increase in rate through the action of ground-ice, which plays a very important part in the river erosion of arctic countries, and in all probability did so during the Glacial period in the valley of the Somme. "When the water is reduced to and below 32° Fahr., although the rapid motion may prevent freezing on the surface for a time, any pointed surfaces at the bottom of the river, such as stones and boulders, will determine (as is the case with a saturated saline solution) a sort of crystallisation, needles of ice being formed, which gradually extend from stone to stone and envelop the bodies with which they are in contact. By this means the whole surface of a gravelly river-bed may become coated with ice, which, on a change of temperature, or of atmospheric pressure, or on acquiring certain dimensions and buoyancy, rises to the surface, bringing with it the loose materials to which it adhered. Colonel Jackson remarks, in speaking of this bottom-ice, that 'it frequently happens that these pieces, in rising from the bottom, bring up with them sand and stones, which are thus transported by the current.... When the thaw sets in the ice, becoming rotten, lets fall the gravel and stones in places far distant from those whence they came.' "Again, Baron Wrangell remarks that, 'in all the more rapid and rocky streams of this district [northern Siberia] the formation of ice takes place in two different manners; a thin crust spreads itself along the banks and over the smaller bays where the current is least rapid; but the greater part is formed in the bed of the river, in the hollows among the stones, where the weeds give it the appearance of a greenish mud. As soon as a piece of ice of this kind attains a certain size, it is detached from the ground and raised to the surface by the greater specific gravity of the water; these masses, containing a quantity of gravel and weeds, unite and consolidate, and in a few hours the river becomes passable in sledges instead of in boats.' Similar observations have been made in America; but instances need not be multiplied, as it is a common phenomenon in all arctic countries, and is not uncommon on a small scale even in our latitudes. "The two causes combined--torrential river-floods and rafts of ground-ice, together with the rapid wear of the river cliffs by frost--constituted elements of destruction and erosion of which our present rivers can give a very inadequate conception; and the excavations of the valleys must have proceeded with a rapidity with which the present rate of erosion cannot be compared; and estimates of time founded on this, like those before mentioned on surface denudation, are therefore not to be relied upon."[EQ]