THE NEW ZEALAND SOUND BASINS 53 



pelled to establish channel grades along which to do efficient work. 

 This follows immediately from gravitative considerations, and is 

 illustrated by the grades of all stream channels as known to the author. 

 The ultimate result of this continual approximation to baselevel is a 

 complete flattening of channel grades. Flattening or lowering of 

 channel grades is directly proportional to the strength of the eroding 

 agent, and is a common fact of observation. A normal stream may 

 possess a certain grade ; a flood finds this so steep that it is enabled to 

 reduce the grade locally, and yet maintain great efficiency as regards 

 transportation. Thus at certain points, notably those of stream con- 

 vergence, floods excavate holes below local or even main baselevel. 

 These holes show undercutting of stream banks and reversal of channel 

 grade down-stream. Stream studies show the amount of this excava- 

 tion below baselevel to be directly proportional to the stage of stream 

 development (i. e., steepness of grade) and the strength of flood. Thus 

 a stream 5,000 feet deep working along a young channel would alto- 

 gether overshadow the work accomplished by a stream 50 feet deep 

 flowing along an excessively broad (i. e., a very flattened grade) valley. 



Now, the glacial period was an ice--flood, the streams of ice, in the 

 canyons (young channels) of southwestern New Zealand being more 

 than 5,000 feet thick. 



The action was evidently that of a viscous stream, there being a 

 continuous slope in the mass from summit to baselevel. Therefore, 

 from gravitative considerations, the ice-floods would find the channel 

 grades of the insignificant preglacial streams so steep that they could 

 induce in them excessive local flattening and yet maintain great effi- 

 ciency of transportation;' i. e., their general velocity, as shown by their 

 still steep surface curves above baselevel, would still be considerable. 

 The disparity in volume between ice- and stream-floods, even allowing 

 for loss in velocity of ice, would induce, at certain marked canyon 

 convergences, such flattening of grade as would (by analogy with 

 stream studies in New England, New South Wales) be commensurate 

 even with the depth of the ice-stream. Thus would here arise basins 

 thousands of feet deep, showing reversed grades lower down stream; 

 also the undercutting of canyon sides, and alignment of cliff bases. 



Upon the retreat of the ice-flood or floods, one would expect—as 

 with ordinary streams — to find the flood grade altogether too flat for 



