A dataset of surface water in the upper Yellow River from 1999 to 2023

In this study, we constructed a 15 m–resolution and 25-year (1999–2023) dataset of the main stream of the upper Yellow River using Landsat series imagery on the Google Earth Engine (GEE) platform. To address the challenges posed by fragmented and narrow waterbodies, pan-sharpening techniques were applied to enhance spatial detail, and waterbody extraction was performed through an OTSU adaptive thresholding approach combined with manual correction. Based on the extracted water body boundaries, the river centerlines were further delineated using ArcGIS software. Finally, six morphological indicators including average width, sinuosity, channel stability, and others were quantified. The results demonstrated that the pan-sharpening method significantly enhanced the extraction capability for small water bodies in the Upper Yellow River, while the Automated Water Extraction Index (AWEIsh) performed robustly across the study region. The final waterbody extraction achieved an overall accuracy of 0.94 and a Kappa coefficient of 0.84. This dataset comprises the main stream water bodies of the upper reaches of the Yellow River in the Qinghai section from 1999 to 2023, with data collected at 4-5 year intervals across six periods: 1999, 2003, 2008, 2013, 2018, and 2023. The dataset has a spatial resolution of 15 m and includes vector data and raster data in *.shp and *.tif formats for 16 river segments and the entire river section, as well as parameter statistics in *.xlsx format.