| 24 Sep 2026
SCS-TCWave: an event-based hindcast dataset of typhoon waves and reconstructed wind fields for the South China Sea (1979–2025)
Abstract. Extreme-wave estimates for the South China Sea (SCS) depend on the peak sea states of tropical cyclones, yet reanalysis winds smooth the storm core and no openly archived, event-resolved typhoon wave record exists for the basin. We present SCS-TCWave, an event-based hindcast of the 480 tropical cyclones that reached tropical storm intensity inside the SCS during 1979–2025. Each event is forced by 10-m winds reconstructed from best-track records (a Holland-type vortex with motion asymmetry and far-field gale-radius anchoring, blended into ERA5) and simulated with SWAN on one-way nested 0.25°/0.05° grids under a frozen, fully audited configuration. Each event ships three CF-1.8 NetCDF layers: hourly wave fields, the reconstructed wind forcing, and an extreme-value catalogue of per-cell event maxima. Validation against 64,529 calibrated altimeter super-observations from 187 events shows near-zero bias in background sea states and a single-signed core overestimation that grows with intensity (+1.15 to +2.99 m within 100 km of the centre). A distance-binned multiplicative correction annex with bootstrap confidence intervals centres the per-event peak distribution: the median relative error of per-event maxima moves from +9.5 % raw to −1.4 % corrected, with 78 % of events within ±25 %. At the highest observed peak (9.4 m, Pearl River estuary, Typhoon Mangkhut) the product stays within ±25 % where an identically configured ERA5-forced run does not; beyond 400 km ERA5-based products remain preferable, and this distance-based usage rule ships with the data. Method tiers, quality flags, the frozen configuration, and the full calibration record accompany the archive (https://doi.org/10.57760/sciencedb.011ul; Fang et al., 2026).
The manuscript presents a valuable dataset of wave conditions driven by tropical cyclones in the South China Sea. This dataset would be useful in regional assessments, its value is clear. However, I have several major comments that need to be addressed before publication, in order to ensure that the manuscript provides sufficient context for the future users of this dataset.
Comment 1. Readability and scientific style
The text would benefit from careful copy-editing to improve readability and consistency of scientific style. In several sections the writing is highly condensed and occasionally adopts an overly assertive or informal tone for a data descriptor manuscript. It also contains some difficult to understand sentences. Improving clarity and adopting a more neutral presentation would strengthen the manuscript. Some examples are:
L37: "Each group that needs one builds one"
L38: "The community's collective effort accumulates in inaccessible form"
L49, L131: "The dataset ships uncorrected", “Each event ships…”
L29: “wave extremes forced by those winds are biased low in proportion”
L171: “Extreme-value applications consume one number per event”
L151: “Across the 187 matchup storms, …”
Comment 2. Terminology and methodological clarity
I am unfamiliar with the term "wave anchor" (e.g. in Fig 1 and Fig 2, Section 3.5). If this refers to wave buoy observations or fixed validation stations, I recommend using standard terminology throughout the manuscript.
The manuscript often relies on project-specific terminology (e.g., "reconstruction gate", "degraded pathway", "amplitude clamping", "method tiers") in situations where the underlying methodology could be described more directly. While these terms are eventually defined, the resulting text is unnecessarily difficult to follow for a reader. I encourage the authors to replace terminology-based descriptions with more explicit explanations of the underlying procedures.
The manuscript repeatedly uses “the annex” as shorthand for the bias-correction method, for example “the annex operating in its design zone” and “the annex encodes an open-water radial bias”. An annex is a document section rather than a method, so this wording is difficult to interpret. Please refer directly to the “distance-dependent bias correction” or “radial correction factors” and state more explicitly that these factors were derived from open-water altimeter comparisons and may not be applicable at depth-limited coastal stations.
The statement “The two verification levels draw a consistent map, and the dividing line sits in the extreme tail” is difficult to interpret. Please state explicitly which results are consistent, what is being separated by the dividing line, and which variable or distribution is meant by the “extreme tail”.
Comment 3. Representativeness and independence of the validation strategy
The multi-level validation framework is useful, particularly the combination of basin-wide altimeter matches, fixed-station observations and an ERA5-forced control. However, I am not convinced that the presented validation fully supports the intended use of the event-maximum catalogue. The “per-event peak” comparison appears to use the maximum over the available altimeter matches, whereas the released product contains the temporal maximum at every grid point. Because altimeters sample only narrow tracks at discrete times, the largest observed matchup is not necessarily representative of the true event maximum, its location or its timing. The authors should clarify precisely which quantity is compared and quantify how satellite sampling affects its representativeness.
In addition, the direct near-core validation is based on relatively few station–storm combinations, several of which involve storms used during model calibration. The Tier-2 period has no direct altimeter validation and is assessed mainly by applying the Tier-2 reconstruction method to 27 completed Tier-1 cases. These limitations should be reflected more clearly in the strength and scope of the conclusions.
Comment 4. Limitations and applicability
The limitations section identifies several important constraints, including the omission of tides, storm surge and wave-current interaction, limited representation of complex coastal geometry, the absence of systematic peak-period validation, and reduced suitability for background sea states. However, the section currently combines model limitations, input-data uncertainty, validation gaps and usage recommendations without clearly distinguishing these categories. I recommend reorganising it into model-physics and resolution limitations, forcing and reconstruction uncertainty, observational and validation limitations, and implications for use.
The statement that nearshore engineering applications should use the dataset as offshore boundary forcing (L231) rather than as direct design values is an important limitation and should be introduced earlier and more prominently. The Introduction currently frames the dataset with reference to offshore wind farms, platforms, ports and coastal defences and describes it as serving engineering applications (L57). This framing could lead users to interpret the gridded event maxima as locally applicable design conditions. However, the simulations exclude tides, storm surge and wave–current interaction, and do not resolve smaller-scale coastal sheltering and nearshore transformation. The Pearl River estuary example further demonstrates that both the raw fields and the open-water bias correction may behave differently at depth-limited locations. Please clarify the supported applicability domain in the Abstract and Introduction and provide some rough guidance for identifying locations where the product should only be used as boundary forcing.
Minor comments
Figure 5: not discussed in the text. Panel c) shows a substantial difference with altimetry observations – this needs to be explained or interpreted in the text.