Replication data and code for: Global Food Security in a Climate-Oscillating World: Spectral Evidence and Early Warning Implications for Sustainable Food Systems

This archive contains the data, analysis code, and generated figures required to reproduce all results reported in the manuscript "Global Food Security in a Climate-Oscillating World: Spectral Evidence and Early Warning Implications for Sustainable Food Systems" (submitted to Sustainability, MDPI; manuscript ID sustainability-4486005). The data comprise the FAO Food Price Index (headline and five sub-indices) together with seven climate oscillation indices (Niño 3.4, SOI, IOD, NAO, PDO, AMO, MEI v2), assembled at monthly resolution for 1990–2025 (432 observations) and at annual resolution for 1961–2025, plus a supporting country–year panel of food-import shares, undernourishment, and climate indices. Niño 3.4 is supplied as the sea surface temperature anomaly relative to the 1991–2020 climatology; the original absolute-SST series is retained alongside it for reference. The code implements the complete analysis pipeline: Welch cross-spectral coherence and phase estimation (segment length L = 128 months, 50% overlap, K = 5 segments), Monte Carlo AR(1) and IAAFT surrogate significance testing (500 surrogate pairs), lagged cross-correlation, sub-period comparison, and 15-year and 10-year rolling-window coherence. It also includes the first-round revision computations (10-year window sensitivity, IOD excluding-2025 robustness, Table 4 correlations with p-values, and per-cell significance counts for the sub-index heatmap) and the scripts that generate the updated Figure 5 (phase spectra with significant frequencies marked), Figure 7 (coherence heatmaps with significant cells marked), and Figure F1 (rolling-window comparison, Appendix F). A second-round revision script reproduces the additional analyses introduced during peer review: a seasonality diagnostic and deseasonalised coherence estimates; Morlet central-frequency sensitivity at ω₀ = 4, 6, and 8; Augmented Dickey–Fuller tests with the trend term inside the regression together with Phillips–Perron tests, both applied to the raw series; a Bonferroni-adjusted bound and a block-bootstrap distribution of the argmax lag for the IOD lead time; a block-bootstrap test of the sub-period coherence differences, which supersedes the Fisher z-test where the effective number of independent spectral estimates is small; and the frequency-band restriction reported in Appendix G. All stochastic components are seeded, so the values quoted in the manuscript are exactly reproducible. Climate indices are sourced from NOAA PSL/CPC and the food-price data from FAO; the assembled analysis-ready extracts are redistributed here for reproducibility. Code is released under the MIT Licence and data/figures under CC BY 4.0. See CHANGELOG.md for the differences between versions.