198 EXPLORATIONS IN TURKESTAN. 
back from whose front it rises sharply lOO feet or more. It is of the same material 
as all the others, but is decidedly fresher in form. The hills and hollows are very 
clearly defined, though all are well covered with ^rass, and some of the kettles 
contain water. Where the Taka Su traverses the moraine it has cut a flood-plain 
about 100 yards wide, on the sides of which are steep, ungraded walls, disclosing 
the structure of the underlying deposits. Here it is seen that the moraine lies upon 
a consideral)le deposit of gravel. The .same feature is seen in a number of other 
cases, where water-laid gravel underlies moraines of various ages. This sliows that 
previous to the formation of .such moraines there was a time of aggradation, 
probably due to the increasing load of the stream which heralded the advance of 
the ice. As soon as the stream leaves the fourth moraine and enters the third there 
is a sudden and ver^- marked change in the character of the flood-plain, which 
expands abruptly to five or six times its fonner width. This is not due to change in 
material, for the moraines are alike in composition ; nor is it due to the less tliickness 
of the deposit which is cut through in the older moraine, for if this were tlie case 
the broadening would be gradual and funnel-shaped instead of sudden and at right 
angles. Moreover, in the older moraine the flood-plain is not limited by steep 
stream-cut walls as it is above, but by gently sloping, sinuous shores rather than 
banks ; for the moraine is half-drowned in flood-plain gravel, so that the kettles fonn 
deej) ba)S and the hills form islands and promontories. The gravel which drowns the 
third moraine seems to belong to the .same epoch as that which underlies the 
fourth, although lack of time made it impossible to trace one into tlie other. 
Apparently at the end of the third glacial epoch the ice retreated above the upper 
limit of the terraces, and normal stream erosion proceeded far enough to ctit into 
the underhing rock along the terraced portion of the valley, and to cut a broad 
swath through the latest moraine. Then, as the interglacial epoch drew to a close, 
there seems to have been an increase in the load of the streams. As a result, the 
whole valley was aggraded, and in the region of the moraine the aggradation was 
so great that it not only filled the valley eroded in the moraine, but caused the 
gravel to overflow and cover the adjacent parts of the moraine itself Meanwhile the 
glacier was advancing. In its upper cour.se it doubtless widened and deepened the 
valley, but near the front it ceased to erode and flowed on top of the gra\el and the 
earlier glacial deposits, there forming a new moraine, the fourth of our series. 
The youngest of the old moraines is a small one far up the valley, close to that 
which is now being formed. There is nothing to prove that it does not mark 
merely a stage in the retreat of the fourth glacier. The other moraines, however, 
seem to represent epochs and not stages, so that, judging from analog}' and even 
without the evidence of the Kan Su moraine (p. 193) the fifth moraine ought to repre- 
sent a glacial advance following a retreat. It is noticeable that each moraine is smaller 
than its predecessor and, except for the older ones, of which the ends are cut off", each 
moraine extends to a less distance downstream than its predecessor. The amount 
of erosion in the successsive interglacial epochs also grows less and less. 
