SCS-TCWave: an event-based hindcast dataset of typhoon waves and reconstructed wind fields for the South China Sea (1979–2025)

Quality stratification: method_tier (tier1 = observed vortex structure, 2002+; tier2 = empirical fallback, 1979–2001, method-switch bias +1.0% [−1.7, +4.2] with 7.8% per-event scatter) and qc_flag (wind–pressure-inconsistent best-track phases, vortex amplitude clamped).Data generation. Best tracks: China Meteorological Administration (CMA) tropical cyclone database, with vortex-structure parameters (radius of maximum wind, radius of 34-kt wind) from IBTrACS v04r01 where available.

Wind forcing: Holland (1980) profile with Georgiou (1985) motion asymmetry and translation deduction, blended into ERA5 hourly 10-m winds (e-fold band 300/150 km), passing a reconstruction gate (parametric vortex peak ≤ best-track Vmax over sea). Wave model: SWAN 41.51 (ST6 source terms, Madsen bottom friction, BSBT propagation), GEBCO 2024 bathymetry, 48 h spin-up before each event window (approximately −48 h / +24 h around the basin residence of the storm).

Temporal resolution 1 h; spatial resolution 0.25° (outer grid, 99–135°E, 0–30°N), 0.05° (inner grid, 108–121°E, 15–24°N) and 0.1° (wind forcing).Recommended layer by distance to the storm centre: within 200 km (eyewall / near core) use SCS-TCWave, adding the correction annex over open water; at 200–400 km use SCS-TCWave with the annex; beyond 400 km (far-field swell) ERA5-based wave products are recommended — the parametric increment adds no skill there and far-field swell is under-predicted by 40–60% at Taiwan-side reference buoys.

The annex corrects the systematic radial bias only and must not be applied at depth-limited nearshore locations.Files. 47 per-year zip archives scs_typhoon_waves_v1.0_<YYYY>.zip (1979–2025; 62.75 GB in total; 0.41–2.28 GB each) contain one directory per event named by the CMA storm id <YYYY>_<NNNN>, each with four CF-1.8 NetCDF-4 files: <id>_waves_inner.nc (hourly hs, tp, tm01, dir, depth, u10, v10; 0.05° grid), <id>_waves_outer.nc (same fields, 0.25° grid), <id>_wind.nc (reconstructed 10-m wind, 0.1°, hourly), <id>_catalog.nc (event-maximum Hs and wind per cell, peak hour, distance of the footprint maximum to the storm centre).

Missing values: land and dry cells are _FillValue. Thirteen documentation files accompany the archives: README_v1.0.md (quick start), DATA_DICTIONARY_v1.0.md (files, variables, units, attributes, CSV columns), PRODUCT_DOCUMENTATION_v1.0.md (configuration, residual tables, reconstruction gate, quality flags, correction annex, recommended layer, limitations), CALIBRATION_RECORD_v1.0.md (calibration decision record), scs_tcwave_usage_example.ipynb (xarray usage examples), event_catalog_index.csv (one row per event: window, peak Hs and location/time, maximum wind, method_tier, qc_flag, peak distance and recommended correction factor), CORRECTION_ANNEX.csv (distance-binned correction factors with 95% CI), MANIFEST_v1.0.csv (SHA-256 of all 1,920 product files), matrix_by_intensity.csv and per_storm_peaks_final.csv (validation tables), envelope_table_final.csv (buoy and literature anchors), frozen_v1.0-calib.yaml and frozen_wind_v1.0-calib.yaml (frozen configuration snapshots).

Files open with any NetCDF-4 library (e.g. xarray, netCDF4-python, MATLAB, Panoply).Known limitations. Core (0–100 km) sea states at severe-typhoon intensity and above are biased high before correction (0–100 km bias +1.7 m over the full dataset; median per-event peak error +9.5% raw, −1.4% with the annex, 78% of events within ±25%). Altimeter winds saturate above ~25 m/s; peak period has no systematic verification; tier2 events (1979–2001) have no direct altimeter overlap and are bounded by a cross-era experiment; the 0.05° grid does not resolve harbour sheltering.How to cite.

Please cite both the data descriptor and the dataset: (1) Fang, F., Yang, Y., Jin, P., and Zhou, B.: SCS-TCWave: an event-based hindcast dataset of typhoon waves and reconstructed wind fields for the South China Sea (1979–2025), Earth Syst. Sci. Data (under review), DOI: [TO BE ADDED UPON ACCEPTANCE]; (2) the dataset via its Science Data Bank DOI (this record).

For the up-to-date citation including the final article DOI, see this dataset's Science Data Bank page.