So as if an elephant were pulling you down beneath the water and you only had your strength to grip against it. Can’t be done.
<s>I looked but was not able to find CFM/CMM for the input and output</s>; but that would give a reasonable idea of what the volume of the canyon must be because what goes in must come out. So the volume of the flow at the outlet and inlet gives a very good idea of just how crazy it can be.
Edit: Input flows: https://nrfa.ceh.ac.uk/data/station/meanflow/27096 Output flows: https://nrfa.ceh.ac.uk/data/station/meanflow/27043
https://en.wikipedia.org/wiki/Slot_canyon https://en.wikipedia.org/wiki/Bernoulli%27s_principle
In an incompressible flow (e.g. water under anything resembling normal circumstances), there is no net flow into or out of any given volume, i.e. a 3-D box or other shape. But the surface is 2-D! So you can look at a little box with its top at the surface and its bottom slightly below the surface. If there is downward flow out the bottom (and zero flow out the top, of course), there must be net inward flow in the sides. Nothing wrong with that.
This could happen if the river has constant speed and constant cross-sectional area everywhere but gets progressively narrower and deeper. The water moves inward because it has to get away from the encroaching banks and moves down to fill the deepening channel. If this is abrupt enough even in the absence of turbulence, the downward current a foot or so below the surface will be substantial.
Loads of fatalities in turbulent water occur because the victim rapidly becomes unconscious in the water.
It's an underground tunnel with a small exposed strip at the top.
the water can be circulating around the tunnel, flowing back and forth under the rocks. It can be pulling sideways or down at the exposed strip with the water circulating back up in the underground areas out at the sides.