It's interesting that in the specialist literature we still see maps showing long straight ice edges -- like the one above, in a paper by Etienne et al in 2006. Given that this is to some extent a scale issue, with minor crenellations in the ice edge "evened out" for ease of interpretation, it is also unfortunate in that some readers might get a very simplistic view of how ice edges behave. In general, an ice edge like that shown in the map would only exist where very special conditions exist on the ground -- for example on an plateau surface or on a low-level plain virtually devoid of prominent surface features. Here is one example from the Greenland Ice Sheet near Cape Alexander:
Much more commonly we find this sort of thing:
The upper two photos are oblique aerial photos and the lower one is a satellite image from Google Earth -- all three images are from Greenland. What we see are ice edges that are very complex indeed, with ice tongues pushing into all low-lying areas and with higher ground / hill masses standing proud of the ice as enclaves or nunataks or as elongated "peninsulas". Lakes are impounded against rising ground, but the picture portrayed in the map at the top of this photo would have been almost impossible in the real world.
In the real world of West Wales, as I have said before, the Devensian ice edge of the Irish Sea Glacier must have been somewhat on the lines of the maps below:
Bearing in mind that the coast was not in its present position at the time, and that most of Cardigan Bay and St George's Channel was dry land, the key determinants of ice edge position (apart from the glaciological ones) would have been the uplands of Preseli and Carningli - Dinas Mountain and -- at a lower level -- the old cliffline of North Pembrokeshire, presenting a rampart which was in some places more than 100m high. The ice managed to surmount this old cliffline along the full length of the North Pembrokeshire coast, but in St Bride's Bay and around the Dale Peninsula the ice was not sufficiently thick or powerful to surmount the obstacle and press far inland. So from Newgale southwards all of the Devensian glacial and fluvioglacial deposits are plastered along the coastline, or at the most a kilometre or so inland. The ice must have pushed some way into Milford Haven, for there are ice contact deposits at West Angle Bay.
South of Milford Haven the question of whether Devensian ice actually passed across the limestone plateau of Castlemartin is still to be resolved. But I'm pretty sure that the ice pressed eastwards along this coastline, maybe with the glacier flank pressed against the limestone cliffs, at least as far east as Caldey Island. The fresh till in Ballum's Bay, which is uncemented and which contains ORS cobbles and larger erratics, must have been laid down by ice travelling from west towards the east.
The only glaciological scenario which makes sense here is one involving ice travelling ACROSS St George's Channel, and not along it, as shown on the (hypothetical) map of the Celtic Sea Piedmont Glacier. I am also happy with the idea that Devensian ice reached Lundy Island, the west-facing coasts of Devon and Cornwall, and the Scilly Islands.
So the map published by Clark et al (2010) in Quaternary Science Reviews just doesn't make sense to me, since ice does not flow in long thin lobes with virtually zero gradient across wide open coastal plains....... I'm not up to speed with everything going on in British glacial geomorphology at the moment, but I trust that by now this map has been substantially revised!
So what we need in the Bristol Channel, if we are to make sense of the field evidence, is a highly crenellated ice edge, entirely in tune with what we see in analogous situations in Greenland today
Pattern and timing of retreat of the last British-Irish Ice Sheet
Chris D. Clark, Anna L.C. Hughes, Sarah L. Greenwood , Colm Jordan , Hans Petter Sejrup
Quaternary Science Reviews(2010) pp 1-35