Global Rice Flowering-Stage Heat Exposure, 1961–2021: Derived Data and Reproducible Workflows
Abstract
This repository provides the derived datasets, quality-control records, data dictionaries, analysis code, software environment specifications, and automated validation procedures supporting a global assessment of rice flowering-stage heat exposure from 1961 to 2021. The reproducibility package integrates rice-distribution information, crop-calendar data, daily temperature records, and time-varying harvested-area weights. It includes the main flowering-stage exposure estimates, sensitivity analyses using relative flowering centres of 0.65, 0.70, and 0.75, and nonlinear trend diagnostics. The nonlinear workflow includes LOWESS, cubic-spline modelling, segmented regression, temporal block cross-validation, and a 1,000-replicate moving-block bootstrap for breakpoint uncertainty. The package contains the data and workflows required to reproduce the reported derived results, including 424 ERA5-Land climate points, 530,862 rice-growing grid cells, 750,395 grid-season phenology records, and 77,775 cluster-year-centre sensitivity estimates. Newey–West heteroskedasticity- and autocorrelation-consistent inference uses a Bartlett kernel with a four-year lag. The flowering-centre sensitivity results are robust across the three tested settings. The nonlinear analysis identifies an estimated acceleration in warm-night exposure around 2001, with a bootstrap 95% interval of 1986–2010; this interval does not support interpreting 2001 as a uniquely determined breakpoint. Original ERA5-Land, GloRice, RiceAtlas, and FAOSTAT source data remain subject to their respective access conditions and licences. Users should consult the README, data dictionary, third-party data notice, and file manifest before reusing the materials.
