FOEMATION OF BROAD OSAR OHAT^TNELS. 445 



The breadth of these broad channels (one-eighth of a mile to one-half mile 

 or more) is such that it seems inadmissible to postulate ice arches of such 

 dimensions without their roofs collapsing-. Russell reports the roof of a 

 stream of the Lucia glacier collapsing. This stream is about 150 feet wide.-"- 

 Now, although the subsidence of the roof in this case appears on the ice 

 surface only so long as the roof is not very thick, it by no means follows that 

 there is not also an inward flow of the roof and walls with increasing depth. 

 But the roofs of the broad osar channels would be from ten to twenty times 

 as broad as this stream of the Lucia glacier. To postulate self-supporting- 

 roofs is an enormous demand. I do not see even one feature of the gravels 

 or any property of ice that warrants the assumption. Locally we can con- 

 ceive of such arches floating on the slack water north of hills crossed by 

 the osar rivers, but the broad osars are also found on southern slopes where 

 there could be no slack Avater. 



Perhaps the principal question involved in the problem is this: Where 

 are we to find the supply of heat necessary to melt and enlarge such great 

 channels 1 For I assume that melting is a greater cause of enlargement 

 than erosion. The channels in which were deposited the broad osars, the 

 osar bowlder clay, the narrow marine deltas, also the lake-like enlarge- 

 ments in which were deposited the peculiar formation elsewhere named 

 "lacustrine massives," are all a connected series of phenomena. Any com- 

 plete theory must account not only for these very broad channels but also 

 for the narrow ones. If we assume that the broad osar channels were 

 formed subglacially, we may as well assume that lacustrine massives 5 to 

 10 miles long and 1 to 2 miles wide were also formed subglacially. But if 

 we assume that these very broad channels were subglacial, how are we to 

 account for the narrowness of the osars proper? Ordinary subglacial 

 streams depend for the heat with which they enlarge their tunnels chiefly 

 on waters of superficial melting, slightly warmed before the pkmge down 

 the crevasses. This supply of heat is small and only moderately enlarges 

 the tunnels. ■ This accounts for the narrowness of the earlier tunnels. We 



'Nat. Geog. Mag., vol. 3, p. 107, May, 1891. "The course of the stream below the movith of 

 the tunnel may he traced for some distance by scarps in the ice above, formed by the settling of the 

 roof. Some of these may be traced in the illustrations. When the roof of the tunnel collapses so 

 completely as to obstruct the passage, a lake is formed above the tunnel, and when the obstruction 

 is removed the streams draining the glacier are flooded." 



This description refers ro the tunnel by which the stream descends beneath the ice after having 

 risen to the surface and flowed a mile and a half on the ice. 



