FOEMATION OF BEOAD OSAR CHANNELS. 447 



process of time tlie middle of the arch might rest on the previously depos- 

 ited osar, where there was one, but in any case there would often be more 

 enlargement of the tunnel laterally than in height. Where the course of 

 the glacial river was approximately parallel to the ice flow, the slow set- 

 tling of the roof of the tunnel would continue to modify the same mass 

 of ice in its prog'ress for a term of years and cause a somewhat continuous 

 depression of the surface. In this depression or valley surface waters 

 would collect and melt more or less ice before reaching a crevasse. Many 

 conditions, such as the extension of the n(^ve line northward, might cause 

 an increased supply of waters with flooding of the subglacial tunnels. 

 Collapse of the roof or clogging of the channel would cause the water 

 to rise into englacial or superficial channels, and the latter would follow 

 the depression caused by the settling of the roof and often cause the forma- 

 tion of temporary surface lakes. Where the waters rose in crevasses or 

 went down again into them after passing an obstruction, deep pools would 

 form if the overflow was long continued or often repeated. When one or 

 more pools were formed or openings were made through the roofs, the 

 heat of the sun would be absorbed in increasing amount by the subglacial 

 waters, the separate pools would gradually become confluent in a contin- 

 uous channel open above to the sun, and this channel would then rapidly 

 broaden till it sometimes came to extend across a whole valley. Many of 

 the conditions for oversupply of water as compared with tunnel capacity 

 would depend on purely glacial conditions, such as rate of melting, rate 

 of ice flow, etc. When the falling of a single block of ice into a tunnel 

 may have changed thg course of a glacial river overflowing on the ice 

 into a new valley in the ice surface^ it will not be expected that we shall 

 be able to trace all the accidents of broad-channel formation. North of 

 hills crossed by the osar rivers this process was probablj^ often, perhaps 

 always, assisted by the pool of slac^i water there collected, and here the 

 enlargement may have often proceeded as the extension of a fringing or 

 marginal lake formed north of the hill. 



This hypothesis, postulating the change in the development of an 

 osar system from a narrow to a broad osar and again to the narrow 

 type, demands that we shall not regard the broadening of the channel as 

 extending recessively northward. Rather it took place locally, leaving 

 reaches of narrow osar in the course of the same system. We can admit a 



