In the week that this site has been online, my map has been downloaded more than a thousand times (thanks!), so here is how I made it, and what the hex-symbols mean.
Tracing over a map As you can see on the sources page, I used hexographer, with a public-domain topographic map of the British Isles as an underlay. If you start hexographer, you can open both this map (using the 'load map' option at the bottom of the menu) and the hxm file I made. Then, in the main menu on the left, stretch the underlay to 350% in the x and y dimensions and adjust the opacity to see both maps, and begin tracing.
Mountains first I find that it is generally a good idea to create or recreate maps from the top on down; as mountains anchor hillsides, hills determine rivers, rivers make valleys, leaving everything else that isn't water as plains.
Thus, I began by putting in snow-capped mountains or peaks wherever the contents of a cell where mostly white, and regular mountains where they showed red-brown. Next, I put in rugged hills where the map showed ochre. Mountains, on maps as well as in how we imagine them, are pretty unambiguous terrain, and the map reflects them accurately - see the legend for elevation categories. In places like Wales or the western Scottish coastline however, many cells contain everything from mountain peaks to narrow valleys near sea-level. While the six-mile scale is a fairly detailed scale, one hex still covers about 31 square miles of terrain, so a certain loss of detail is inevitable. In these cases, I tried to preserve the large-scale contours and flow of mountain-chains around the peaks.
Rivers and Lakes Next, I put in rivers, which are helpful for orientation when placing towns, and in deciding whether a given hex should be rugged hills, grassy hills or plain. In general, wherever a major river bisects a hex with high elevation variance, I tended to use a lower elevation symbol, to make river valleys more apparent.
I only put in the major lakes and kind of ran out of steam by the time I got to Scotland. If you want to set a game in the north, keep in mind that the highlands are riddled with lochs and glens not shown at this resolution.
Plains Last, I turned every unambiguously plain hex into green grassy plains, then filled the remaining hexes with plains or grassy hills, to further shape the contours of hilly areas until all the land area was filled in. Again, there is some loss of resolution, but I tried to have clusters of hills, and mountain chains buffered by grassy hills for easier orientation and narration in play.
Coastlines While hexographer lets you trace exact coastlines, I decided against this as it is a lot of work for relatively little gain. If you want to use the map to simulate movement speed or land yield, uniform hexes are easier to use than ambiguous ones that are half sea and half land. In most cases, I ignored narrow inlets and small islands, and moved larger islands like the Isle of Wight a bit further out to sea to fit a full sea hex into the coastal gap.
Cities and Forts I got the information for settlements from this Map of the Roman World, and often consulted this site for more information. While the major towns like London, York, or the tribal capitals are unambiguous, many Roman forts morphed into small villages over time. Thus, every fort or fortress hex should be read as a predominantly military installation, while most villages also had a fort nearby, but are mostly civilian in nature. Towns and cities were walled, while villages (with some exceptions in the tribal borderlands) were not.
Note that the actual footprint of cities in this age was quite small - even London's population was only around 10.000 people, and the built-up area of any given settlement hex actually only takes up some 5% of the land area inside. Similarly, the forts along Hadrian's Wall form a solid line on this map, but were actually separated by five or more miles of mostly empty terrain.
Roads With the settlements in place, putting in roads was easy. The main Roman roads were laid out for speed and connected towns and forts in straight lines where possible. The main roads like the Fosse Way or Watling Street were of excellent quality, paved with stone and lined with ditches and mile-markers, while smaller roads connecting outlying villages may have been little more than footpaths. However, the historical record is too uncertain to draw a clear distinction, so any connection shows as a road on the map.
Terrain Cover: Forests and Farms The last step, and the one that involved the most guesswork, was the placement of vegetation. Generally, irrigated grassland adjacent to settlements would have been used as farmland. I also interpreted clusters of Roman villas (of which there were hundreds, especially in the south and east of England) as farmland, but only highlighted the big, palatial ones with a villa symbol. Plains and green hills without woodland cover would have been used for grazing sheep and cattle.
Arable land further away from settlements or in the vicinity of major known forest was mostly represented as light woodland. This doesn't mean that there weren't any people living there, or that it was all evenly wooded. Rather, it should be narrated as plains or hills, dotted with clusters of trees and the occasional roundhouse farmstead.
True dense forests, with blanket-like tree cover (like that of Germania across the Rhine) was rare in Roman Britain: Most of the great forests that must have once covered the island were thinned out or cleared by the onset of the Iron Age, centuries before the coming of the Romans. The main forests mentioned in the historical sources were the Weald in the south, the forests of Dean, Nottingham and the Yorkshire area, and the primordial evergreens of Caledonia, but even here we cannot be sure to what degree they were inhabited, or even managed for the production of timber. Roman forts were all first built in wood, and surrounded by miles-long palisades that must have consumed all trees for miles around.
However, in a fantasy game context, discrete clusters of forest are easier to narrate than a random jumble of terrain, so when in doubt, I put in dense, contiguous forests where the absence of settlements and farms left sufficient space to justify them.
Swamps and Moors Before the drainage projects of the middle ages, many low-lying plains and riverbanks in Britain were swamps; most notable are the fens of East Anglia and the Somerset downs. These were hard to traverse except witj boats, but supported low densities of population, engaged in fishing and the production of salt.
Waterlogged terrain at higher elevations was characterised by moorland - soggy peat soil, dotted with flowers, isolated trees and rock outcroppings covered in moss. The moors of Cornwall were rich in tin, which was extracted for the production of bronze.
Both swamps and moors were often shrouded in mist, giving them an eerie, otherworldly feel. The Romans, unfamiliar with this kind of terrain, avoided the moors, making them perfect for placing elves, fairies or the more fantastical kinds of monsters.
Wilderness From a Roman point of view, Ireland, most of Scotland and some parts of Wales were considered wilderness, and the historical and archaeological record for these areas is thin to non-existent. The terrain didn't produce enough of a surplus to support towns or paved roads; so cattle and sheep raising were the main complement to subsistence-level agriculture, centred on hillforts, wooden roundhouses or lakeside huts on stilts. I didn't spend much time on detailing the terrain, so the GM should feel free to change empty hexes into farmland or forest as fits the story.
So, did the British Isles really look like this? Nobody knows. At this resolution, we can be fairly sure about the coastlines and mountain-ranges, as well as the location and nature of the major settlements. Everything in between is educated guesswork at best. However, what is unfortunate for the historian is actually liberating for wargamers and GMs - feel free to make the land and the creatures of Britannia fit your vision for the game, knowing that no one can conclusively prove you wrong.