Global Dam Image Dataset V1.0

The Global Dam Image Dataset Version 1.0 (GDID V1.0) is a publicly available global dataset developed to support research on environmental monitoring, change detection, assessment of environmental changes associated with dam construction, and decision-making related to dam development. The dataset provides temporally paired multispectral observations acquired before and after dam construction, together with spatial guidance information describing dam crests and reservoir areas and a Digital Elevation Model (DEM) for each sample. GDID V1.0 was developed by integrating information from multiple data sources, including the Global Dam Watch database, Landsat Collection 2 Level-2 Tier-1 imagery from Landsat 4–9, Google Earth Engine, OpenStreetMap, the GLC_FCS30D and Dynamic World land-cover datasets, and the Detection of Initial Reservoir Impoundment (DIRI) algorithm developed in this study. DIRI was used to estimate the year of first reservoir impoundment based on long-term surface-water dynamics within the spatial extent of each reservoir. Each GDID V1.0 sample consists of four components:(1) a six-band pre-construction multispectral Landsat image;(2) a corresponding six-band post-construction multispectral Landsat image acquired after the first reservoir impoundment;(3) a guidance mask delineating the dam crest, reservoir area, and background; and(4) a DEM corresponding to the same spatial extent as the multispectral images. The multispectral images have a spatial dimension of 512 × 512 pixels, corresponding to a ground coverage of 15.36 × 15.36 km². Each image contains six spectral bands: Red, Green, Blue, Near-Infrared (NIR), Shortwave Infrared 1 (SWIR1), and Shortwave Infrared 2 (SWIR2). The guidance masks were generated using NDWI-based surface-water extraction and land-cover information and were subsequently reviewed and manually refined to improve the spatial correspondence of reservoir boundaries and dam crests with the Landsat imagery. The DEM data were spatially aligned with the corresponding multispectral images. Following image quality assessment, visual inspection, and data refinement, GDID V1.0 contains 5,685 samples. The dataset is publicly available and provides a comprehensive data resource for studies involving dam-related environmental change analysis, remote sensing, reservoir monitoring, and Earth observation applications.