2026-10-06T14:38:43Z https://api.figshare.com/v2/oai

oai:figshare.com:article/32314854 2026-10-02T13:13:00Z category_25795 item_type_3 month_year_10_2026

Data and code supporting “Beyond binary canopy recovery: Continuous margins retain information about signed late GPP state” haojie zhang (23925654) Hechun Quan (25156108) Geoscience data visualisation Canopy recovery Spatial cross-validation ecological indicator gross primary production NDVI leaf area index compound drought inverse probability weighting negative control remote sensing reproducible research <p dir="ltr">This dataset and code archive supports the manuscript “Beyond binary canopy recovery: Continuous margins retain information about signed late GPP state”. It contains the verified 2001–2024 archived tropical and subtropical panel, event input, numbered Python analysis scripts, fixed spatial outer-fold predictions, inverse-probability weighting diagnostics, absolute-state construct comparisons, calibrated non-event negative-control outputs, result tables, figures, checksums, and a machine-readable manifest.</p><p><br></p><p dir="ltr">The archived panel comprises 208,224 cell-years from 8,676 land cells. The primary event risk set contains 689 impact-confirmed, stable-land-cover records, of which 418 have an observed late GPP outcome without a repeat qualifying event in years +4:+5. The standardization baseline is 2001–2012, eligible event years are 2013–2019, recovery windows extend through 2022, and late GPP follow-up extends through 2024.</p><p><br></p><p dir="ltr">The data form a designed validation sample. Original cell-selection probabilities and some preprocessing provenance were not preserved, so the archive does not support area-weighted prevalence inference. The primary endpoint is a signed late GPP state rather than functional closure. The calibrated non-event control did not support compound-event-specific enhancement. Recovery margins outperformed absolute +1:+3 canopy state, but did not show resolved superiority over the four-variable pre-event-plus-recovery-state representation. PML and MOD17 share inputs with the canopy predictors and are not independent physiological validation.</p><p><br></p><p dir="ltr">Detailed run order, software requirements, construct definitions, interpretation constraints, and SHA-256 checksums are provided inside the archive.</p> 2026-10-02T13:13:00Z Dataset Dataset 10.6084/m9.figshare.32314854.v7 https://figshare.com/articles/dataset/_b_Land-cover_transitions_modulate_photosynthetic_activation_in_temperate_land_systems_b_/32314854 CC BY 4.0