Code and data for: Sampling design governs satellite resilience estimates in seasonally frozen forests: structural–functional divergence is concentrated in evergreen needleleaf systems in Northeast Asia

This record contains the code and relevant derived data supporting the study “Sampling design governs satellite resilience estimates in seasonally frozen forests: structural–functional divergence is concentrated in evergreen needleleaf systems in Northeast Asia.” The study evaluates satellite-derived forest resilience across Northeast Asia during 2001–2024. It compares conventional full-year temporal-coverage requirements with a May–September growing-season design and examines whether functional resilience, represented by gross primary productivity and solar-induced fluorescence, is changing differently from structural resilience represented by kNDVI, EVI and NIRv. A conventional full-year completeness rule retained approximately 20% of the vegetated domain and 2.7% of Russia, whereas the growing-season design retained approximately 85% and 87%, respectively. Regional functional-minus-structural divergence was not distinguishable from a shared year-block permutation null (+0.0003 yr−1; p = 0.87). Divergence was instead concentrated in evergreen needleleaf forest relative to deciduous needleleaf forest (+0.0108 yr−1; p = 0.001). Compound hot–dry event recurrence was not an important spatial predictor, and divergence did not anticipate subsequent tree-cover loss. The archive contains 23 ordered Jupyter notebooks, numerical validation tests, a portable configuration file, reference layers, analysis-ready growing-season residuals, final GeoTIFF rasters, Zarr datasets, statistical tables, provenance documentation, a data dictionary, file manifests and SHA-256 checksums. Provider-scale raw archives, temporary files, caches, logs, obsolete analysis branches and duplicate figure renderings are excluded. Source-acquisition notebooks and provenance documentation identify the original Earth-observation products. The primary analysis branch is v2_per_month_detrend. Per-pixel permutation probabilities are provided as uncorrected consistency diagnostics; inferential conclusions are based on pooled zonal permutation tests and spatial-block bootstrap intervals.