Causal evidence that the Taklamakan green belt expanded cropland and grain output(Data)

Authors. Zhaojie Chen¹*, Jinhui Chang²*, Rongjuan Tan², Yong Jin¹Affiliations. ¹ Faculty of Business, ² Department of Food Science and Nutrition, The Hong Kong Polytechnic UniversityCorresponding. [e-mail] · [e-mail] manuscript. Chen Z., Chang J., Tan R., Jin Y. Causal evidence that the Taklamakan green belt expanded cropland and grain output. Submitted to Nature Food (2026).---What this archive containsThe cleaned 2010–2024 county-year analytical panel and supporting intermediate datasets used to estimate the causal effect of the 3,046-km Taklamakan green belt on cropland and grain output. Reproducing the manuscript's headline numbers requires only the contents of `analytical_panel/` together with the code archive on Zenodo

---

CoverageGeography. 105 counties: 27 Tarim-periphery treated counties + 78 arid controls (Inner Mongolia, Gansu, Ningxia, latitudes 35–45° N, outside the Tarim hydrological system).Time. Annual, 2010–2024 (15 years).Treatment. Four staggered cohorts of green-belt segment opening: 2014 (Hotan southern, n = 6), 2017 (Kashgar western and Minfeng, n = 6), 2020 (Aksu northern and Akto, n = 8), 2023 (Bayingol eastern arc and ring closure, n = 7).Outcomes. Cropland area (ha), grain output (tonnes), grain yield (t ha⁻¹) in levels and natural logs.Covariates. Aridity index, elevation, baseline shelterbelt density, baseline sandstorm days, water-network indicator, latitude, longitude.Provenance (summary — see PROVENANCE.md for URLs)Cropland & grain. Xinjiang Statistical Yearbook and China Rural Statistical Yearbook county tables (NBS), supplemented by prefectural yearbooks where county tables were missing.Cropland validation. MODIS MCD12Q1 v6.1 land-cover product (NASA/USGS LP DAAC).Aridity. UNEP World Atlas of Desertification, 3rd edition (2018).Elevation. SRTM 90 m DEM, county-mean.Sandstorm days. China Meteorological Administration county-level station records (2010 baseline).Shelterbelt density. State Forestry and Grassland Administration (2010 baseline).Treatment cohorts. State Forestry and Grassland Administration tendering/commissioning records, cross-validated against People's Daily and Xinhua coverage.Compatible codeThe analysis pipeline that produces every figure and table in the manuscript from these files is deposited in a companion Zenodo record:> Chen Z., Chang J., Tan R., Jin Y. *Code: Causal evidence that the Taklamakan green belt expanded cropland and grain output.* Zenodo (2026). doi:10.5281/zenodo.XXXXXXX *(link to be inserted on publication)*Reproduction is a single command after extracting both archives side-by-side:```bashpython analysis/00_build_panel.pypython analysis/01_did_analysis.pypython analysis/02_figures.py```A full reproduction takes ~10 minutes on a modern laptop (8 GB RAM is sufficient; the placebo step is the binding constraint at 1,000 permutations).