Hidden spatial compensation sustains apparent stability in global maize cultivation

The code was developed and tested using Python v3.10.13. The repository is organized into Code_core and Data_core.Code_core/Code_for_main_analysis contains memory-safe Python scripts used for the main analyses, including input inspection, trajectory compaction, construction of the 0.015° cell-event table, covariate attachment, maize mobility and recovery-replacement accounting, water-risk and hydroclimatic overlays, basin/country accounting, uncertainty analyses, and production/trade exposure analyses.Code_core/Code_for_SI contains supplementary scripts for preparing the GGCP10 maize production-importance baseline, summarizing annual maize fractions to 0.015° grids, exporting basin/country mechanism vectors, and running sensitivity analyses for intensity tolerance and denominator choices.Data_core contains the derived tables and aligned layers required by the scripts and used to support the main figures and supplementary analyses.

Core files include the compact 0.015° trajectory table (trajectory_summary_by_0p015_cell_compact.csv) and the covariate-enriched cell-event table (cell_event_table_0p015_with_covariates2.csv). Additional files provide validation and confusion-matrix inputs (accuracy_confusion_rates_by_region_season_year.csv, annual_confusion_for_error_null_GLOBAL.csv), basin and country reporting-unit layers (basin_idx.tif, basin_map.csv, country_idx_aligned_to_maize0p015.tif, country_map.csv), GGCP10 production-importance layers (GGCP10_yield_layer/), recovery–replacement and water-risk summaries, aggregation-blindness diagnostics, event-year hydroclimatic overlays, and FAOSTAT production/trade tables.