Data and code for: Hidden Markov model identifies four phytoplankton regimes associated with seasonal forcing and monsoon disturbance at a coastal station in Jinhae Bay, Korea

Complete data and code supporting the manuscript submitted to Limnology and Oceanography (LO-26-0267). Contents: raw and processed daily HPLC pigment and environmental data; the frozen-state R pipeline (Gaussian HMM fitted from a single prespecified 15-seed multi-start set; biomass emission variable = log10[TChl-a + fucoxanthin + peridinin]; CZM zero replacement via zCompositions; CLR-PCA); fitted model objects; the decoded regime sequence and posterior probabilities; state-number selection (AIC/BIC/ICL, k=1-6) under both a single continuous-sequence and a gap-aware (seven-segment) specification, with k=4 vs k=5 cross-tabulation, fifth-state profile, and k=5 robustness of transitions/dwell; k=3-6 order-robustness correspondence (Table S9); per-state within-state conditional-independence matrices; rank-based effect sizes (Cliff's delta) with moving-block bootstrap confidence intervals for external environmental variables (Table S5); CLR transformation-sensitivity; parametric bootstrap; below-detection (censored) sensitivity; per-pigment non-detection counts; figure code; Supplementary Tables S1-S9. Reproducibility (v5.0): state-number-selection outputs (Table S1, Table S8, Fig. S6) are generated from one common 15-start continuous-sequence execution and cross-checked by consistency_audit.R (which also verifies the k=3-6 order robustness, the NH4/DIN regime statistics, and that Table S5 reports effect sizes, not p-values); a single master runner (run_all.R) works from the repository root or the code/ directory; an automated value-check (verify_values.R) with a clean-run log (CLEAN_RUN_LOG.txt); a file manifest with checksums (MANIFEST.sha256); and a machine-restorable environment (renv.lock, R 4.5.2). One command regenerates all principal tables, figures, and numerical values and passes the automated checks. Changes from v4.0: added the k=3-6 order-robustness analysis and Table S9; replaced the Table S5 pairwise Wilcoxon p-values with autocorrelation-aware Cliff's delta effect sizes and moving-block bootstrap CIs (external variables only; state-defining pigments excluded as circular); reconciled Table S1 with Table S8 (k=1-6, ICL column, identical rounding and state order); corrected the k=6 selection fit to the full 15-seed set; extended the automated audit; and added the file manifest. Funding: National Research Foundation of Korea (NRF) grants RS-2026-25468730 and RS-2026-25550603.