# Pre-registration: does EGMS show a Praid-specific signal? (13 Nov 2026)

Written before any number below was computed for this project.

## Disclosure: what was seen first
- The sister project `strada-care-se-lasa` ran a first Praid test (T4) on 29 Oct 2026. Its summary was read before writing this file: inside 1 km vs ring 4–6 km, same EGMS release; the share of cells at ≤ −5 mm/yr after the flood rose to about 11% in both zones. So a Praid-only effect is already doubtful.
- This test changes three things to remove that artefact: a same-months seasonal control, a difference between inside and ring, and a national placebo (random discs elsewhere, which see the same radar track windows and weather).

## Data (fixed)
- EGMS Ortho vertical (U), CALIBRATED/ORTHO Level 3, "2025 release", 100 m cells, 2020-12-09 → 2025-11-25, 303 epochs (open viewer endpoints, as in strada-care-se-lasa). Release coverage was checked on the portal's dataset list: every dataset is "2020-2025"; no later release exists in it.
- Praid point: OSM "Salina Praid" 46.55069 N, 25.12052 E. INSIDE = cells within 1.0 km. RING = cells 4.0–6.0 km. Not moved after seeing results.
- Event: the freshwater inrush into the mine, open 5–29 May 2025 (dates to be fixed from official sources; any date inside May 2025 gives the same windows below).

## Windows (fixed)
- BEFORE: 2020-12-09 → 2025-04-30. AFTER: 2025-06-01 → 2025-11-25.
- Season windows for the control: JUN–NOV of 2021, 2022, 2023, 2024, 2025; reference months JAN–APR of the same year.

## Statistics (fixed)
- For each cell with ≥ 20 BEFORE epochs: vB = linear velocity over BEFORE (mm/yr).
- G1 (signal before): dB = median(vB inside) − median(vB ring). Placebo: 30 discs elsewhere in Romania (same radii, same procedure). Praid passes G1 if |dB| ≥ 1.0 mm/yr AND |dB| is above the 95th percentile of the placebo |dB|.
- For each cell and year y: S_y = mean displacement JUN–NOV of y − mean displacement JAN–APR of y (mm). D_y = median(S_y inside) − median(S_y ring).
- G2 (signal after): X = D_2025 − mean(D_2021..2024). Praid passes G2 if |X| ≥ 1.0 mm AND |X| > 2 × sd(D_2021..2024) AND |X| is above the 95th percentile of the same X computed for the placebo discs.
- Gate: PASS only if G1 and G2 both pass ("significant signal before and after"). If only one passes, it is reported as that and the gate is FAIL.

## Placebo discs (fixed)
- Seed 5. Candidate centres: Romanian 100 m cells with GHSL built surface ≥ 1000 m². Keep a candidate if it has ≥ 60 cells inside 1 km and ≥ 150 cells in 4–6 km, lies ≥ 15 km from Praid and ≥ 15 km from any of the ten salt sites listed in strada-care-se-lasa BRIEF. 30 discs are drawn in order of a seeded shuffle; the first 30 that qualify are used. Radii are not tuned. For cost, a placebo disc uses at most 150 of its ring cells (seeded random subset, seed 5) and at most 150 inside cells; Praid uses all of its cells. Salt sites (lat, lon): Turda 46.58770,23.78733; Praid 46.55069,25.12052; Cacica 47.63502,25.89802; Ocna Dej 47.11287,23.86121; Ocna Mureș 46.39,23.86; Târgu Ocna 46.28,26.62; Ocnele Mari 45.08556,24.30472; Slănic 45.23333,25.93917; Ocna Sibiului 45.88167,24.06139; Ocna Șugatag 47.78111,23.93944.

## If the gate fails
The page leads with the national ground-motion map and an address lookup and says in words that the radar sees no Praid-specific motion; the cut through the dome and the flood chronology are drawn from published sources and labelled as drawn.