Spatial Distribution Dataset of Afforestation in the "Two Rivers and Four Tributaries" Basin (2014-2022)

The Spatial Distribution Dataset of Afforestation in the "Two Rivers and Four Tributaries" (Liangjiang Sihe) Basin (2014–2022) covers the major watersheds of the Yarlung Zangbo, Nujiang, Lhasa, Nianchu, Yalong, and Shiquan rivers within the Tibet Autonomous Region (approximately E 79°–97°, N 27°–32°), spanning 2014 to 2022. This dataset aims to reveal the spatial distribution patterns and area dynamics of afforestation under the greening project, providing spatial data support for ecological effectiveness assessment.The dataset was produced through two technical approaches. The first employed deep learning-based remote sensing identification using Sentinel-2 Level-2A multispectral imagery (2019–2021, cloud cover <20%) and Sentinel-1 dual-polarization SAR imagery (2020) obtained via Google Earth Engine (GEE), both at 10 m spatial resolution. Preprocessing involved extracting the blue, green, red, red-edge, and near-infrared bands (with the red-edge band resampled to 10 m) to compute annual mean RVI, ReNDVI, and EVI products, along with VV/VH polarization annual means aligned with the optical data. Training samples were constructed by delineating forest extent using ESA 10 m global forest cover data, extracting areas with canopy height below 15 m from a global canopy height dataset, and overlaying with a global 30 m plantation distribution dataset to select high-confidence plantation pixels. A Temporal-Spectral-Spatial Vision Transformer (TSSViT) model was developed for pixel-level afforestation extraction, achieving a micro-average accuracy of 81.78%, training accuracy of 85.25%, and micro-average IoU of 69.18%. The second approach used forest age change detection based on the China 30 m Annual Forest Age dataset (CAFA V2.0), comparing adjacent-year forest age data annually to extract newly afforested areas transitioning from non-forest to forest, with ArcGIS zonal statistics employed for area calculation by watershed unit (pixel size: 900 m²). Results show that afforestation area increased from 118,300 mu in 2015 to approximately 390,000 mu in 2022, with the Yarlung Zangbo and Nujiang basins as core afforestation regions. Gaps between monitored and planned areas exist due to 30 m resolution limitations and forest age mapping errors.The dataset is stored in Shapefile format (.shp/.dbf/.prj/.shx) and can be processed using ArcGIS, QGIS, or Python (geopandas).