Data formats¶

Everything the engine consumes and produces is plain CSV or JSON. The bundled demo set lives in the plugin's Data/ folder; point your own files at the same loaders.

Portfolio CSV¶

One row per insured location:

LocationId,Latitude,Longitude,ElevationM,TIV,Occupancy,Deductible,Limit,Share
LOC-0001,26.24938,-79.99590,7.3,193000,ResidentialWood,9600,0,1.0
LOC-0002,26.20410,-80.08993,4.6,1284000,ResidentialMasonry,12800,1054000,0.25

Column	Meaning
`LocationId`	Unique site identifier
`Latitude` / `Longitude`	WGS-84 degrees; geo-projected into the level around the configured origin
`ElevationM`	Site elevation in metres (drives flood/surge exposure)
`TIV`	Total insured value
`Occupancy`	Construction/occupancy class — must match a class in the vulnerability curves
`Deductible`	Site deductible (absolute)
`Limit`	Site limit; `0` = unlimited
`Share`	Insurer's share, 0–1

An optional tenth column Insured (1/0) marks whether the location carries a policy; omitted, every location counts as insured.

To load: create a SceloPortfolioAsset Data Asset (Content Browser → right-click → Miscellaneous → Data Asset → SceloPortfolioAsset) and call its Blueprint-callable LoadLocationsFromCSV / LoadVulnCurvesFromCSV (absolute file paths) from any editor Blueprint — e.g. an Editor Utility Widget. Or skip authoring entirely: the bundled sample DA_Ekhaya2 (SceloAtlas Content/SceloAtlas/Sample/) ships with locations and curves already baked in. OED exposure files import directly with Scelo.ImportOED <file>.

Vulnerability curves CSV¶

Per-(peril, occupancy) damage curves, linear in intensity between anchor points:

Peril,Occupancy,Intensity,MeanDamageRatio,DamageCV
TropicalCyclone,ResidentialWood,15,0.00,0.40
TropicalCyclone,ResidentialWood,25,0.02,0.40
TropicalCyclone,ResidentialWood,35,0.08,0.40

Intensity units are per-peril (wind: m/s peak 3-second gust at 10 m). MeanDamageRatio is the expected damage fraction of TIV at that intensity; DamageCV parameterizes Beta-distributed secondary uncertainty around the mean (toggle and tune at runtime with Scelo.SU). Inspect the loaded table with Scelo.CurveViewer on — it logs the table and writes curves.csv with the mean ± SU band.

Event catalog JSON¶

A stochastic catalog of parametric events with annual rates:

{
  "catalog_name": "Atlantic Demo Catalog v0.1",
  "events": [
    {
      "event_id": "HU-ANDREA",
      "peril": "TropicalCyclone",
      "annual_rate": 0.12,
      "lat": 24.4, "lon": -79.2,
      "delta_p_hpa": 35, "rmax_km": 30, "holland_b": 1.3,
      "heading_deg": 300, "forward_speed_ms": 6.5
    }
  ]
}

Cyclone events carry the Holland (1980) parameters directly: central pressure deficit delta_p_hpa, radius of maximum winds rmax_km, shape parameter holland_b, plus track heading and forward speed. Frequency is Poisson over the summed rates by default; switch to negative binomial with Scelo.Freq nb <lambda> <k>. Real historical tracks replay from HURDAT2 files with Scelo.Replay.

Earthquake strong-motion records¶

Data/QuakeRecords/ bundles eight processed records from real events (South Napa 2014, Ridgecrest, Anchorage, Ferndale) — response-removed, high-pass filtered at 0.08 Hz, baseline-corrected, acceleration + displacement in E/N/Z, with a manifest.json carrying magnitude, PGA/PGD, significant duration and sampling metadata per record. These drive the recorded-motion playback in the quake directors.

The export pack¶

Scelo.Export All produces a complete, versioned actuarial results pack (see a real example: a 40,000-year run on a 14-location book):

File	Contents
`summary_metrics.json`	Schema-versioned run summary: AAL (with standard error), annual σ, TVaR at 90/95/99/99.5/99.6, OEP & AEP by return period — all in three perspectives (GU / Gross / Net) — plus convergence diagnostics and units
`ept.csv`	EP table: `Perspective,EPType,ReturnPeriod,Loss,CI_Low,CI_High` — confidence bands included
`ylt.csv` / `melt.csv` / `splt.csv`	Year-loss, mean event-loss and sample period-loss tables
`attribution.csv`	Loss attribution
`locations*.csv`	Per-location detail
`program_economics.csv`	Program economics (via `Scelo.Econ`)
`footprints/`	Per-event hazard footprints
`manifest.json`	Pack manifest with run id, seed and file inventory

Every number in the pack is recomputable by an independent validator — the reference validation suite passes 108/108 checks against engine output.