Supplementary code and derived data
Seasonal memory in Hudson Bay sea ice: autumn freeze-up predicts spring break-up, 1990-2026

Contents
--------
code/
  build_region.py          Hudson Bay domain mask (Natural Earth 1:10m, 0.05 deg)
  download_seaice_ext.py   NSIDC Sea Ice Index G02135 v4.0 daily windows, 1989-2026
  download_power_ext.py    NASA POWER / MERRA-2 daily fields, 110-point grid, 1990-2026
  build_panel.py           per-cell ice phenology on the 1 Sep - 31 Aug ice season
  build_forcing.py         bay-wide and land-ring daily atmospheric means
  build_dataset.py         merge of panel metrics and seasonal atmospheric aggregates
  analysis_final.py        Sen-Theil / Mann-Kendall trends, memory model, IV, bootstrap
  analysis_ml.py           leave-one-year-out benchmark, gradient boosting, SHAP
  anal_panel.py            per-cell estimation of the memory coefficient
  paper_figures.py         Figures 1-5
  build_manuscript.py      builds the .docx from the Convergence template
  manuscript_text.py       manuscript text

data/
  season_dataset.csv       37 ice seasons x seasonal metrics and atmospheric predictors
  cell_coef.csv            1,428 cells: memory coefficient, p value, local trends
  annual_table.csv         annual series used for the trend table
  results.json             all headline statistics reported in the text
  ml_results.json          cross-validated model comparison and predictions
  dml.json                 double machine-learning estimates

Reproduction order
------------------
1. build_region.py
2. download_seaice_ext.py ; download_power_ext.py
3. build_panel.py ; build_forcing.py ; build_dataset.py
4. analysis_final.py ; analysis_ml.py ; anal_panel.py
5. paper_figures.py ; build_manuscript.py

Data sources (all public, no registration)
------------------------------------------
NSIDC Sea Ice Index G02135 v4.0   https://nsidc.org/data/g02135
NASA POWER (MERRA-2)             https://power.larc.nasa.gov
NASA GISTEMP v4                  https://data.giss.nasa.gov/gistemp/
NASA Worldview / GIBS            https://worldview.earthdata.nasa.gov

Note: raw daily sea-ice fields (~250 MB compressed) and the full atmospheric grid
(~60 MB) are not included; the retrieval scripts regenerate them deterministically.
