The data presented in part 1 of this paper (Meier et al., this issue) are here used to assess the role of water input/output, water storage, and basal water pressure in the rapid movement of Columbia Glacier, Alaska. Consistently high basal water pressures, mostly in the range from 300 kPa below to 100 kPa above the ice overburden pressure, are responsible in an overall way for the high glacier flow velocities (3.5-9 m d-1), which are due mainly to rapid basal sliding caused by the high water pressure. Diurnal fluctuation in basal water pressure is accompanied by fluctuation in sliding velocity in what appears to be a direct causal relation at the upglacier observation site. The water pressure fluctuation tracks the time-integrated water input (less a steady withdrawal), as expected for the diurnally fluctuating storage of water in the glacier far from the terminus. At the downglacier site, the situation is more complex. Diurnal peaks in water level, which are directly related to intraglacial water storage as well as to basal water pressure, are shifted forward in time by 4 hours, probably as a result of the effect of diurnal fluctuation in water output from the glacier, which affects the local water storage fluctuations near the terminus. Because of the forward shift in the basal water pressure peaks, which at the downglacier site lead the velocity peaks by 6 hours, a mechanical connection between water pressure and sliding there would have to involve a 6-hour (quarter period) delay. However, the nearly identical nature of the diurnal fluctuations in velocity at the two sites argues for a single, consistent control mechanism at both sites. The velocity variations in nondiurnal ''speed-up events'' caused by extra input of water on the longer timescale of several days are only obscurely if at all correlated with variations in basal water pressure but correlate well with water storage in the glacier.

It appears that small variations in water pressure (≤100 kPa) sufficient to produce the observed velocity variations (15--30%) are mostly masked by pressure fluctuations caused by spontaneous local reorganizations of the basal water conduit system on a spatial scale much smaller than the longitudinal coupling length over which basal water pressure is effectively averaged in determining the sliding velocity. At the achieved level of observation the clearest (though not complication free) control variable for the sliding velocity variations is basal water storage by cavitation at the glacier bed. ¿ American Geophysical Union 1994