SUBGLACIAL AND ENGLACIAL CHANNELS. 311 



englacial channels under long reaches of ice imbroken at the surface, and 

 it is not admissible that they were formed along longitudinal or any other 

 crevasses. The problem is solved so far as the much-crevassed ice is 

 concerned. It now remains to inquire how we can account for the exist- 

 ence of longitiidinal channels within or under parts of the ice solid on the 

 surface, or if broken, not longitudinally, but transversely into long prismoids 

 attached at the ends to unbroken ice, so that apparently they ought to dam 

 the subglacial waters. 



At the outset we are confronted by another query: How nearly do 

 the surface crevasses represent those of the bottom ! Except at precipices, 

 crevasses form at right angles to the tension, or nearly perpendicular to the 

 ice surface. Usually they are not planes, but more or less curved and 

 irregular; but even when approximately plane when first made, they 

 become greatly distorted by the unequal flow of the ice. Also, since the 

 upper ice moves much faster than the lower, the successive fractures divide 

 the ice into blocks that are much wider, lengthwise of the glacier, at the 

 top than at the bottom. When the ice has great depth and rapid motion 

 and crevasses form at short surface intervals, the bases of the prismoidal 

 slabs must often be narrow, and it may even happen that the crevasses 

 meet at the ground or above it. We must admit, therefore, that the basal 

 ice is broken by crevasses nearer together than at the surface, and also that 

 by the intersection of curved and irregular crevasses it may not seldom 

 happen that transverse slabs of ice that appear on the surface to be capable 

 of acting as dams to the subglacial waters are broken through in the 

 depths. After making' a most liberal allowance for cases where there are no 

 apparent longitudinal crevasses but yet the transverse crevasses connect, we 

 still have a residue of apparently unbroken ice penetrated by longitudinal 

 subglacial streams. 



Into all crevasses the siu'face waters pass. Part of the crevasses do 

 not reach to the ground; part open into subglacial channels and part do not. 

 Those crevasses down which the waters succeed in forcing a passage are 

 soon enlarged into the shaft or well of a moulin. This enlargement is 

 significant and must be accounted foi*. At the instant of melting, the 

 surface water has the temperature of 32°, but under sunlight it absorbs 

 heat and rises in temperature. The water in contact with the ice then gives 

 up its surplus heat to melt a portion of the adjacent ice. This is a slow 



