Lough Neagh seiche — historical model replay

– m/s from – –

–surface elevation–

5 km

–

η is surface displacement about still water, in centimetres. The basin is closed — no Lower Bann outflow, no tributaries — so its mean stays zero to round-off and η carries no lake-level trend. Arrows are depth-averaged current at cell centres — not surface drift, not wave motion. Row 0 is north; the map is drawn at true aspect in Irish Grid metres, one cell per model cell. The satellite backdrop is a locator only — reprojected onto the model grid offline, never sampled by the model, and not the definition of the shoreline. –

What this is, and what it is not

It is a replay of a solved model

Each storm was integrated offline with the reference solver at 300 m resolution and a 5 s timestep, from rest, with 84 hours of spin-up before the displayed window. The browser only decodes and draws those results. It does not solve anything, and it cannot be made to.

The periods are right; the amplitude is a lower bound

The model's free modes (1.63 h, 1.21 h, 1.12 h) match independent spectral peaks in the gauge record (1.59 h, 1.15 h), and it reproduces the observed north–south antiphase of the fundamental — with nothing fitted anywhere. Its amplitude is several times below the observed basin-coherent amplitude, for the reason in the next panel. Read the shapes, the timing and the mechanism — not the centimetres.

Hourly wind cannot excite a 1.6 h resonance

Interpolating hourly reanalysis wind to the timestep applies a sinc² response to the stress spectrum: only about 24% of the amplitude survives at the basin's own 1.63 h period. Adding sub-hourly wind variance bounded by the ERA5 gust field lifts the modelled response to and past the observed level, which puts the deficit on the forcing rather than on the basin dynamics. This replay uses the deterministic hourly forcing, so what you see is a lower bound.

Not all of what a gauge records is the seiche

Across eleven storms the two ends of the long axis — Toome and Oxford Island — anticorrelate at −0.44 in the 1–6 h band, which is the fundamental mode's signature. But a substantial share of each gauge's variance in that band is local to its shoreline site and not basin-wide, so a raw station-by-station comparison overstates how much the model is missing.

Currents are limited by quadratic bottom drag

Bottom stress is C_d|u|u with C_d = 2.5×10⁻³, a declared standard value rather than a fitted one. It matters: a linear drag law cannot limit a steady wind-driven gyre and a simple linear closure lets currents run to metres per second. Depth-averaged speeds shown here settle at a few centimetres per second, with the largest values in the shallowest cells where u = M/H is least trustworthy.

Barotropic and depth-averaged

No short wind waves, no vertical shear, no internal seiche, no sediment or water quality. The bathymetry's vertical datum and provenance are unresolved, which admits a systematic period bias of a few percent.

A storm name is a label, not a result

Names and impact dates come from Met Office storm-season records and identify the weather event whose archived wind field was used. They do not validate the model, and they do not assert that the named storm caused any particular observed water level.

Everything here is traceable

Each run publishes a manifest with the forcing request and its hash, the bathymetry and grid hashes, every physical and numerical parameter, the solver module hashes and a hash of every published object. Download it and check.

–