TSU · Physics
Tsunami
PTHA18 offshore probabilistic hazard reuse, propagated to 90 m with coastal amplification and overland attenuation; regional source branches (NEAMTHM18, GAR15) plus non-seismic sources · v1.1 · 2026-09
Sendai / Sanriku — the Tohoku 2011 validation coast
Coastal tsunami screening: offshore probabilistic hazard reused from PTHA18 and propagated to 90 m. A screening climatology for portfolio triage — explicitly not evacuation planning, life-safety zoning, or engineering design.
What ships
Surfaces — P(inundation ≥ 0.5 m) at 50 / 250 / 500-year horizons (500-year is the default planning basis), plus 500-year inundation depth in metres, a five-class zoning cut, uncertainty and coverage. Scored on ocean-connected land below 50 m elevation and within 30 km of the shoreline — the inland edge follows terrain rather than a fixed line, and land below sea level (polders, subsided plains) is scored rather than excluded.
Every release: 0–1 score · 5-class band · per-pixel uncertainty · input-coverage mask · STAC item · methodology doc.
Inputs
PTHA18 + GAR15 (Geoscience Australia), NEAMTHM18 (INGV), NGI, GEBCO 2024 bathymetry, SRTM15+, Slab2, NOAA NCEI historical run-up database
Version history
What changed between releases — the same notes that ship with the data. Newest first.
v1.1
2026-09
Correction release. Fixes an inland-attenuation defect on below-datum coastal plains and replaces the fixed 15 km scored band with a physical domain. Offshore hazard, sources and the Tōhoku calibration are unchanged.
- Inland decay measured from the shoreline. v1.0 measured the overland energy-line distance from any pixel below sea level, so subsided plains, polders and deltas (about 1.2 % of scored land, affecting attenuation over roughly 8 %) restarted the decay from zero. On the Ishinomaki plain that placed a 500-yr inundation depth of 8 m some 13 km inland, where the 2011 event never reached. Distance is now taken from the GSHHG shoreline, with the per-tile seam removed at the same time.
- Below-datum land is scored. Land is defined by the shoreline mask, not the sign of the elevation, so Dutch polders, New Orleans and post-2011 subsided Japanese plains return values rather than nodata.
- Physical scored domain. Ocean-connected land below 50 m within 30 km of the coast replaces the fixed 15 km band; the inland edge now follows terrain rather than a line. Inland water bodies are excluded.
- Bare-earth elevation. GEDTM30 v1.2 (global) with DeltaDTM v1.1 on coasts, over Copernicus. Building and canopy height no longer read as ground; a 2–7 m residual remains in dense city cores and is documented.
- New audit gates at the sites where v1.0 failed, so this class of defect cannot ship again.
- Coastal-fringe elevations corrected. The shoreline blend that joins land to sea floor was smearing bathymetry onto the first two pixels of land wherever the sea bed falls away steeply, inventing ground below sea level and therefore inventing inundation depth on top of it. Cádiz fell from 15.9 m to 3.0 m and Puerto Rico from 14.8 m to 1.0 m. This was the single cause behind three separate symptoms.
- Offshore source selection now follows water, not straight lines. Each coastal cell draws its offshore hazard from the nearest source reachable across water. Banda Aceh previously drew from a shadowed point on the Andaman side of Sumatra rather than from the Indian Ocean the 2004 wave crossed. Sheltered water keeps its sheltering: an inner bay is still reached through its mouth.
- Known limitations: tsunami propagation up river channels (the Kitakami reached 49 km in 2011) is not modelled. Where the model disagrees with the historical record it says so rather than hiding it — at Banda Aceh the offshore rate model places a 2004-scale event beyond the 500-year level, so the city reads near-zero depth at this return period; see the validation report. A small number of cells on volcanic and pond-embanked coasts sit below datum in the bare-earth source and may carry a metre or two of inundation that is really standing water.
- Source selection is over-water, not straight-line. Each location now uses the nearest offshore point reachable across water (geodesic Voronoi over the ocean), so a point shadowed by a headland is no longer chosen. A straight-line rule understated 15% of scored coastal locations by >1.25x and 6.2% by >1.5x, worst at peninsula tips and strait mouths. Real sheltering is preserved — reaching an inner bay still means travelling through its mouth. Documented in methodology.md -> *Which source point a location uses*.
v1.0
2026-08
Initial vintage. Coastal band only; see README for the scored domain.
- Three-basin offshore hazard. Published, commercially-clean reference models used where each is authoritative: PTHA18 (Indo-Pacific), NEAMTHM18 (NE Atlantic + Mediterranean), GAR15 Global Tsunami Model (Caribbean, US East Coast, S Atlantic). The Atlantic is genuinely modelled, not reported as zero.
- Non-seismic sources. Volcanic sector collapse, submarine eruption and shallow-seamount explosion contribute their own rates through a displacement kernel calibrated on ten published back-analyses — so the Sunda Strait carries Anak-Krakatau-type hazard that a seismic-only model reports as nothing.
- Coastal amplification at 90 m. Green's-law shoaling to a common reference depth plus a nearshore-slope-dependent run-up factor, over a seamless GEBCO + Copernicus bathy-topo surface.
- Inundation with an observed-calibrated inland decay. Reproduces the 2011 Tōhoku Sendai-plain inundation limit (3 km wet, 5 km marginal, 8 km dry) and its peak run-up (39.9 m modelled vs 39.26 m observed).
- Three horizons (50 / 250 / 500 yr) plus the 500-year inundation depth.
- Honest nulls. Deep inland and open ocean are nodata by design; land beyond wave reach scores zero. Where the model reads lower than the historical record (Cádiz 1755), the validation report says so rather than tuning the number.
Other hazards
LandslideAvalancheWildfireEarthquakeWindstormExtreme heatExtreme coldFrost / freezeLightningSoil erosionHurricaneDroughtVolcanic EruptionStorm SurgeHail