Tonypandy

Most places can receive several transmitters — a median Output Area hears twelve — and only one of them can be the best, which is why the two columns are so far apart. Reach counts an area once the model clears the reception threshold there at all, and it inherits the sweep's screen: only the strongest twenty candidates per area are stored, so a mast lying twenty-first somewhere does not count as reaching it. Both are a share of areas, deliberately, and not a number of homes: checked against Ofcom's own coverage the model ranks transmitters correctly (Spearman 0.861) but over-states how far each one gets by a median factor of 2.5. A share is a ranking statement and survives that; a household count would not. How this is worked out.

It is an envelope, not a boundary: terrain can still block reception well inside it, and it is an edge, not a fill — the colour says how far the signal gets, not that every spot under it works. Zoom in and the shading gives way to the per-area prediction, which is the same question answered Output Area by Output Area. The television ring was contoured from the stored predictions, which keep only the strongest candidates for each area — the true edge can extend beyond what is drawn [I].