Data and Code for "An Optimal-Transport Diagnostic of Inner-Radius Contrast in Tropical-Cyclone CloudSat Reflectivity Profiles"

This repository contains the data, code, figures, tables, and supporting outputs required to reproduce the principal results presented in the manuscript, “An Optimal-Transport Diagnostic of Inner-Radius Contrast in Tropical-Cyclone CloudSat Reflectivity Profiles.” The analysis uses archived CloudSat 2B-GEOPROF observations and IBTrACS tropical-cyclone best-track data. The complete end-to-end workflow reconstructs storm proximity, generates adjacent-profile metrics, performs the statistical analyses, and reproduces the manuscript figures and tables. Repository contents cloudsat_hdf/ — CloudSat 2B-GEOPROF granules analysed in the study. Ibtracs data/ — IBTrACS tropical-cyclone best-track dataset used for storm verification. Data/ — Reconstructed analysis datasets, adjacent-profile metrics, cloud-mask-aware fixed-support outputs, and supporting figure data. Tables/ — Primary statistical summaries, unique-storm and leave-one-storm-out results, conventional-metric benchmarks, common-bin-overlap diagnostics, and sensitivity-analysis outputs. Figures/ — Reproduced manuscript figures and supporting visual outputs. full_reproduce_from_raw_data.py — End-to-end workflow that reconstructs the analysis from the archived CloudSat and IBTrACS inputs. reproduce_results.py — Independent verification script that reproduces the manuscript’s principal statistical summaries from the archived derived datasets. README.md — Documentation describing the repository structure, software dependencies, methods, and reproduction workflow. Analysis summary Candidate CloudSat tropical-cyclone overpasses are verified using IBTrACS storm positions interpolated to individual CloudSat profile times. Adjacent-profile 1-Wasserstein distances are then calculated to quantify changes in the vertical distribution of observed radar reflectivity as a function of true distance from the interpolated storm centre. The primary analysis compares a fixed 0–25 km storm-centred zone with a 25–110 km comparison region. Each adjacent pair is evaluated on its pairwise common-valid vertical support using both shifted-dBZ and linear-reflectivity weighting. Statistical inference is performed at the overpass level rather than by treating individual adjacent-profile pairs as independent observations. The archived analyses include: overpass-level paired tests; aggregation by unique tropical cyclone; leave-one-storm-out sensitivity analysis; minimum common-bin-overlap sensitivity tests; cloud-mask-aware fixed-support validation; comparison with reflectivity-weighted centroid displacement, L1 distance, L2 distance, and correlation distance. The secondary cloud-mask-aware analysis uses fixed vertical support, excludes bad, missing, or clutter-contaminated profiles, and evaluates R05 CPR cloud-mask thresholds of 20 and 30. This analysis tests whether the reported radial contrast persists after explicit hydrometeor-confidence screening and removal of pair-specific support differences. The corrected end-to-end workflow was rerun from the archived raw inputs and reproduced the reported candidate counts, verified overpasses, adjacent-profile datasets, statistical results, sensitivity analyses, and manuscript figures. This repository supports transparent reproduction and independent verification of the archive-specific methodological and observational results reported in the manuscript.