Designing the UK domain: where the edges go
The first decision in any WRF setup is where the model sees air. Here is how the Convek UK domain is laid out and why the boundaries sit where they do.
Before you pick a PBL scheme, before you argue about microphysics, before any of the physics that actually matters for soaring, you pick a domain. The domain is the rectangle of air WRF integrates over. Everything inside the edges is simulated. Everything outside is borrowed from the global model driving you, nudged in at the boundaries every three hours.
The geometry matters more than most tutorials let on. WRF is accurate in the middle of your domain and progressively less so near the edges, where the boundary conditions from GFS are imprinted onto the solution. A rough rule is that the outer 8 to 10 grid cells on each side are contaminated by boundary adjustment. At 4 km spacing that is 30 to 40 km of unusable rim. Any soaring area sitting inside that rim is going to forecast badly, and there is nothing you can do in the physics to save it.
So the first design question is: where are the pilots, and how much buffer do we need around them? For the UK, the furthest-flown areas in summer are the Scottish border ridges in the north, the Long Mynd and Mid Wales in the west, the South Downs and Kent in the south, and East Anglia. The domain has to comfortably cover all of them and then add the rim on top.
The current shipping UK domain is deliberately tight: 6 W to 2 E, 50 N to 56.5 N. That covers England, Wales, and a useful slice of southern Scotland, but it does cut off the north of Scotland and most of Ireland. We made that trade on purpose. On the upgraded benchmark worker, the trimmed domain ran in roughly 13 minutes of WRF wall-clock instead of nearly an hour for the full-UK box. On the live `raspuk` worker, the same trimmed domain is slower in absolute terms but still fits the four-daily cycle comfortably. The full Scottish Highlands and Northern Ireland are on the roadmap as a separate domain when the cost makes sense.
We run a single domain at 4 km rather than a nested parent-and-nest setup. Nests are useful when you want high resolution over a small area and do not care about the rest, but at this domain size a flat 4 km grid is simpler to reason about. No nested-boundary weirdness, no ratio-of-three grid jumps, no fiddly feedback options. 1 km nests will come later over specific sites such as the Lakes or the Welsh ridges, but the base layer stays flat.
The downside of a tight domain is that the western and southern edges sit closer to UK soaring areas than is ideal. Cornwall and Pembrokeshire are within the contaminated boundary rim on south-westerly flow days. That is the single strongest argument for widening the domain westward when compute permits, and it is one of the things the validation work in the queue is set up to measure.
The map itself will get its own post once the coverage page gets the interactive version. Next post: vertical levels, and why the bottom 3 km of atmosphere gets most of the detail.