The 2021 thesis code
The original branch holds the code exactly as submitted for the BSc thesis at the Faculty of Physics, University of Bucharest, in June 2021, together with the thesis itself (BSc_thesis_2021.pdf, in Romanian). main shares no history with it.
How far the 2021 results move
Every correction recorded in the validation page and in the changelog is real, but only some of them bite on the CANDU case the thesis assessed. Running that configuration both ways:
| Correction | Effect on the reference case |
|---|---|
| Wind-direction convention | Most-exposed sector S instead of N — a half turn — and χ/Q there higher by a factor of 1.51 at every distance |
| Plume-rise constants | none: the stack is buoyancy-dominated, 55.7 m of buoyant rise against 10.6 m of momentum, so the corrected constants are never reached |
| Building-wake coefficient | none: no buildings in the reference configuration |
| Roughness coefficients | none: the corrected rows are z₀ = 0.01 and 0.04 m, and the case is pasture at 0.1 m |
| Mixing layer, which the 2021 code did not have | none within 2 km; in the most exposed sector +2 % at 5 km, +11 % at 10 km and +22 % at 20 km, nearly all of it class F, whose plume is held at its 100 m lid |
So the answer is short: the published dose maps are rotated by a half turn, and the most-exposed sector carries 1.51 times what was reported — more beyond a few kilometres, where the lid adds up to 22 %. Everything else corrected here would change a different configuration — a momentum-dominated stack, a site with buildings, grassland or arable roughness — and leaves this one alone.
That is worth saying plainly because the reverse would have been easy to assume. A dozen corrections do not compound into a dozen shifts; most of them are inactive in any one configuration, and the one that dominated was the one that needed no numerics to see.