Title: Growth–Recovery Decoupling in SYM-H Disturbance Timing, 1981–2026: Two-Channel Burst Catalogs, Fit Tables, and Robustness Results

Growth–Recovery Decoupling in SYM-H Disturbance Timing, 1981–2026: Two-Channel Burst Catalogs, Fit Tables, and Robustness Results Data archive accompanying: Wanliss, J. A. (2026). Growth–recovery decoupling in the timing of geomagnetic disturbances across five solar cycles. Geophysical Research Letters (submitted). Author: James A. Wanliss, Department of General Engineering, Anderson University, Anderson, South Carolina, USA ([e-mail]) License: Creative Commons Attribution 4.0 International (CC BY 4.0) DOI: https://doi.org/10.5281/zenodo.21435907 Overview Every negative excursion of the 1-min SYM-H index below a threshold u is decomposed at the first attainment of its minimum into a growth lifetime τ_g (downward threshold crossing → minimum) and a recovery lifetime τ_r (minimum → upward crossing). The record spans 1981-01-01 to 2026-05-31 (solar cycles 21–25). This archive contains the derived two-channel burst catalogs, the windowed and pooled fit tables, and the robustness-test results underlying all figures of the paper, together with a MATLAB script that redraws the figures from these tables alone. The underlying geomagnetic index is the property of the World Data Center for Geomagnetism, Kyoto, and is not redistributed here (see Data Sources below); the catalogs and tables in this archive are derived products. Contents catalogs/ — two-channel burst catalogs One CSV per detection threshold: burst_catalog_p10.csv … burst_catalog_p90.csv, where pNN is the percentile of the full 1981–2026 record at which the threshold is placed (p10 → −36 nT, p25 → −21 nT, p50 → −10 nT, p75 → −1 nT, p90 → +5 nT). The p50 (median, −10 nT) catalog is the primary ensemble of the paper (78,095 complete events). Column Meaning thresholdCrossingTime UT time of the downward threshold crossing (event start, t₀) tauGrowthMin growth lifetime τ_g in minutes (t_m − t₀) minimumTime UT time of the first attainment of the excursion minimum (t_m) tauRecoveryMin recovery lifetime τ_r in minutes (t₁ − t_m) minSYMH_nT SYM-H value at the excursion minimum Durations are raw catalog values. The fits in the paper discard durations ≤ 1.9 min (the 1-min Nyquist bin) before estimation; apply the same filter to reproduce fitted sample sizes (e.g., 33,829 growth and 53,130 recovery samples at p50). fits/ — windowed and pooled fit tables distributions_p50.csv — log-binned pooled densities of both channels at the median threshold with the maximum-likelihood parameters of the finite-range power law f(τ) = A τ^(−μ) exp(−τ/T_c) (data for Figure 1b). channel_window_exponents.csv — per-window MLE fits (μ, T_c, standard errors, sample sizes) for both channels in the 89 one-year 50%-overlap windows at all five percentile thresholds, with window-center and window-mean sunspot number (data for Figures 2a, 2b, S3). decoupling_stats.csv — paired windowed separation Δμ = μ_r − μ_g per threshold set: Wilcoxon signed-rank p, Cohen's d_z. solar_correlations.csv — windowed exponent–sunspot correlations (Pearson/Spearman) and solar-min/max means per channel. delta_mu_ladder.csv — pooled-fit Δμ across the absolute (−30 … −300 nT) and percentile threshold ladders (data for Figure 2c). lsq_crosscheck.csv — least-squares (binned) cross-check estimator vs. MLE; the LSQ estimator reads 0.09–0.17 higher by construction. nonoverlap_comparison.csv, nonoverlap_decoupling_stats.csv — the same conclusions re-derived on the 45 non-overlapping (independent) windows. robustness/ — robustness-test results robustness_R1_nonoverlap.csv — independence check on non-overlapping windows. robustness_R2_neff.csv — effective-sample-size (Bretherton) corrected correlations. robustness_R3_gof.csv — KS bootstrap goodness of fit and Vuong model comparison against exponential, lognormal, and truncated Weibull (Text S1, Figure S1). robustness_R4_covariance.csv — μ–ln T_c likelihood correlation and Hessian conditioning (Text S1). robustness_R5_windowed_tc.csv — solar-cycle dependence of the windowed cutoffs (Text S2, Figure S3). robustness_R6_taumin.csv — lower-cutoff sensitivity scan (Table S1, Figure S2). robustness_R7_blockboot.csv — moving-block bootstrap standard errors (Text S2). robustness_R8_mixture.csv, robustness_R8_summary.csv — pooled-tail mixture diagnostics. tiebreak_comparison.csv, tiebreak_tie_stats.csv — sensitivity of the decomposition to the treatment of tied excursion minima (Text S3, Table S2). figure_data/ fig1a_superstorm_symh_1989.csv — SYM-H excerpt (16 September – 23 September 1989, 1-min) around the 19 September 1989 superstorm used in Figure 1a. For this event at u = −10 nT: t₀ = 1989-09-18 19:12 UT, t_m = 1989-09-19 05:23 UT, t₁ = 1989-09-21 20:45 UT (τ_g = 611 min, τ_r = 3801 min, minimum −292 nT). code/ plotPaper0Figures.m — MATLAB script (base MATLAB only) that redraws Figures 1a, 1b, 2a–2c, S2, and S3 from the CSV tables in this archive. Run it from the archive root. The script evaluates the fitted density of Equation 1 from the archived (μ, T_c) parameters; it performs no event detection or parameter estimation. The full analysis pipeline (event detection, maximum-likelihood estimation, bootstrap machinery) supports an ongoing paper series and will be released with a later paper in the series; this archive contains everything needed to verify and reuse the results of the present letter. Method summary (for standalone use of the catalogs) Events are negative level-crossing excursions of SYM-H below threshold u, with crossings evaluated at u + 0.001 nT (the offset clears the point mass of the 1-nT-quantized index at threshold). Gaps of at most five samples are bridged within events; no other declustering is applied. The excursion minimum t_m is the first sample attaining the minimum (the standard storm-timing convention); because of 1-nT quantization about 70% of median-threshold events have tied minima, and the sensitivity of all conclusions to this convention is quantified in Text S3 / Table S2 of the paper's Supporting Information and in robustness/tiebreak_comparison.csv. Fits use maximum likelihood for f(τ) = A τ^(−μ) exp(−τ/T_c) on raw durations > 1.9 min, with at least 50 events per fit. Data sources SYM-H: World Data Center for Geomagnetism, Kyoto (https://wdc.kugi.kyoto-u.ac.jp/); Nosé, M., Iyemori, T., Sugiura, M., & Kamei, T. (2015). Geomagnetic ASY/SYM indices. WDC for Geomagnetism, Kyoto. https://doi.org/10.17593/15031-54800. The assembled record combines definitive values through 2019, finalized revisions for 2020–2021, and provisional/quicklook values from approximately 2025 onward. Sunspot number: SILSO World Data Center, Royal Observatory of Belgium, Brussels (https://www.sidc.be/SILSO/), monthly total sunspot number, version 2.0. Citation If you use these data, please cite both the paper and this archive: Wanliss, J. A. (2026). Growth–recovery decoupling in the timing of geomagnetic disturbances across five solar cycles. Geophysical Research Letters. Wanliss, J. A. (2026). Growth–recovery decoupling in SYM-H disturbance timing, 1981–2026: Two-channel burst catalogs, fit tables, and robustness results [Dataset]. Zenodo. https://doi.org/10.5281/zenodo.21435907