Thai sugarcane green-harvest transition: mill panel, burning calendars, and a harvest-scheduling model (2019/20–2025/26)
Description
Between the 2019/20 and 2025/26 crushing seasons, the share of sugarcane delivered burned to the mills of Northeast Thailand fell from 50.17% to 2.84%. This dataset assembles the industry, satellite, and air-quality records documenting that transition, together with the code of a harvest-scheduling model calibrated on them. Contents. An audited mill panel of 401 mill-seasons covering 66 Thai sugar mills over seven seasons, compiled from the closed-book production reports of the Office of the Cane and Sugar Board, with a Northeast subset of 157 mill-seasons across 27 mills; a within-season burning calendar derived from OCSB biweekly weighbridge reports; 378,889 NASA FIRMS VIIRS active-fire detections for Thailand, 2022 to 2024; daily PM2.5 records for five Northeast monitoring stations of the Pollution Control Department, 2021 to 2025; monthly Sentinel-2 NDVI features for 97 quality-screened sugarcane parcels in Udon Thani; and the full outputs of the modelling campaign as CSV and JSON. Code. A multi-objective harvest scheduler in which a regulatory cap on burned deliveries acts as a weekly throughput ceiling, together with the calibration, capacity-sweep, robustness, and benchmark scripts, an independent verification suite, and the scripts that rebuild the mill panel from the source reports. Reuse notes. The per-field yield layer of the base instance is synthetic, calibrated to the accuracy band of published Thai satellite studies, and should not be treated as observed yield. The OCSB source workbooks and the GISTDA parcel geometry are cited by URL rather than redistributed, because their publishers state no redistribution terms; scripts to rebuild the derived tables from them are included. Burned shares are computed by the author from published tonnages, and the regulatory caps varied in legal form by season, so a season share above a nominal cap is not equivalent to a regulatory violation. Data and results are released under CC BY 4.0; code under MIT.