CycloneNet: A Reproducible Pipeline and Leakage-Safe Two-Basin Dataset for Tropical-Cyclone Rapid-Intensification Analysis

This record contains the CycloneNet v3 preprint and the CycloneNet v2 dataset package. The repository snapshot corresponding to this release (tag v3.0.2) is available through the GitHub release linked below and under Related works. SUPERSEDES v1.0.0, v1.0.1 AND v2.0.0 — WITHDRAWN METRICSSpecific metrics from the earlier releases are withdrawn and should not be cited: v1.0.0 / v1.0.1 — ROC-AUC 0.97, recall 0.92, "sub-pixel spatial accuracy", "~26 km mean spatial error", "18 hurricanes (1989–2024)": withdrawn for scope, not provenance. These numbers are reproducible — they come from validation_report.txt, released in both v1 records — but the benchmark comprises 58 track points from 19 named Atlantic hurricanes at 44.8% RI prevalence, reported without confidence intervals or a declared train/test separation. It is not comparable to any measurement on the full archive (16,780 events, 4.8% prevalence). v2.0.0 — ROC-AUC 0.83, recall 0.905 on 2,193 test samples: not reproducible from the public artifacts. The v2 release covered only 2020–2023 (972 events; test split 155 samples); the reported metrics were computed on data that release did not contain. The reproducible replacements, from the full 1980–2023 archive: ROC-AUC 0.796 [0.753–0.837], PR-AUC 0.251 [0.179–0.331], recall 0.852 on a frozen test split of 2,679 events — metrics of a model now retired, standing as historical record only. Also withdrawn: the FuelMap spatial-attribution claim (unsupported — did not distinguish from a storm-centre baseline); the "physics-guided" framing (the released v2 code left every training.physics.* weight at 0.0, so the model trained as a plain CNN); the 3D-CNN architecture (retired by a pre-registered verdict). The primary contribution of this release is the reproducible dataset and the auditable pipeline. The complete correction record is Section 9 of the enclosed preprint. ABOUT CYCLONENETCycloneNet is an open, reproducible dataset and processing pipeline for retrospective (hindcast) analysis of tropical-cyclone rapid intensification (RI). The dataset is derived from ERA5 reanalysis and IBTrACS best-track observations and is intended for reproducible research on tropical-cyclone rapid intensification. The repository contains the complete pipeline required to reconstruct the released dataset from the original public data sources. DATASET SUMMARY16,780 valid events from 992 tropical cyclones (578 East Pacific and 414 North Atlantic by genesis basin), 1980–2023, June–November hurricane season. Each event consists of one IBTrACS track point, its preceding 24-hour ERA5 history, and a 40×40×5×14 float32 tensor (40×40 grid window ~10°×10° centred on the storm; five 6-hourly time levels: t0, −6h, −12h, −18h, −24h; 14 channels), with metadata and provenance. LABELS (strict-temporal, tri-state) 799 RI-positive events (ΔV24 ≥ 30 kt, Kaplan & DeMaria 2003) 15,962 RI-negative events 19 NULL NULL means UNDEFINED — no exact 24-hour temporal partner exists (storm dissipation or reporting gap). It must not be coerced to 0, and NULL events must be excluded from RI-classification tasks. SPLITSTrain/validation/test assigned at the storm (SID) level using deterministic SHA-256 hashing with a frozen override map; no storm crosses splits. Normalization statistics are computed on the training split only. The frozen test split (2,679 events; 112 RI-positive; 6 NULL) is distributed for reproducibility. The design controls train/test non-independence through storm identity; it does not control the temporal axis (storms from all decades appear in all splits). Per-event SID and timestamp are included, so year-based or leave-one-year-out evaluation protocols can be constructed directly. PROVENANCE AND REPRODUCIBILITYPer-window SHA-256 manifests, a complete label-correction audit (label_diff_v1_v2.csv, 32,989 rows), and a basin-repair record. The RI-label chain is byte-reproducible from the raw IBTrACS archive (32,989/32,989 rows) via the replication-gate script in the repository. CRITICAL READER NOTEThe string "NA" in the basin field means North Atlantic — it is a categorical value, not a missing value. When reading any distributed CSV with pandas, use keep_default_na=False, na_values=[""]. Without this, "NA" is silently coerced to NaN and basin assignments are corrupted. This bug class occurred six times during this project's own development. PACKAGE CONTENTSThe preprint (PDF) and the data package — 46 files, ~6.03 GiB: 44 per-year cubic shards, 1 metadata archive, 1 checksums file. See DATA_DICTIONARY.md and TECHNICAL_VALIDATION.md inside the metadata archive. The repository snapshot is available through the GitHub release linked below and under Related works. LICENSEDataset: CC BY 4.0. Code: MIT (see LICENSE in the repository snapshot). Redistribution of the dataset must preserve the third-party attributions below (see the NOTICE file). ATTRIBUTIONS (mandatory) ERA5: Contains modified Copernicus Climate Change Service information (1980–2023). Neither the European Commission nor ECMWF is responsible for any use that may be made of the Copernicus information or data it contains. Hersbach, H., Bell, B., Berrisford, P., et al. (2018), ERA5 hourly data on single levels (doi:10.24381/cds.adbb2d47) and on pressure levels (doi:10.24381/cds.bd0915c6), Copernicus Climate Data Store. IBTrACS: International Best Track Archive for Climate Stewardship (IBTrACS), Project Version v04r00, NOAA National Centers for Environmental Information, doi:10.25921/82ty-9e16 (Knapp, K. R., et al., 2010, Bull. Amer. Meteor. Soc., 91, 363–376). IBTrACS follows the World Data Center for Meteorology policy of full and open access; WMO Resolution 40 guides commercial use. NOAA/NCEI provide the data without warranty. ADT (auxiliary, 761 events, 2020–2023): Generated using E.U. Copernicus Marine Service information. CODEhttps://github.com/estefano-ferreira/cyclone-net (MIT) Contact: [email protected]