Data and code for: Can satellites see shade-grown coffee? Detectability limits for monitoring agroforestry ecosystem services

Data and analysis code for the manuscript "Can satellites see shade-grown coffee? Detectability limits for monitoring agroforestry ecosystem services" (Agroforestry Systems, revision 1). Study area: Chiang Rai, northern Thailand. Files: training_samples_3490.csv (1,207 shade-coffee and 2,283 non-coffee training samples with WGS84 coordinates and the 40 predictors derived from PlanetScope SuperDove, Sentinel-1, ALOS-2 PALSAR-2 and SRTM; non-coffee samples from ESA WorldCover 2021 grassland, cropland and bare/sparse classes); field_reference_148_uncertain.csv (148 field-verified locations of previously uncertain status, visited 11-13 May 2026, with coffee presence within 10 m, map class, plant count, height and canopy width); README.txt (column descriptions); r1m1_spatial_cv.py, r1_ablation.py, r1_figures.py, r1_si_figures.py (spatial cross-validation, ablation and figure scripts for the revision). The v1.0 zip contains the original model-training repository. Coordinates of the 300 map-guided verification visits were not recorded.