UzEO: A Boundary-Validated Multimodal Earth-Observation Corpus for Uzbekistan (Sentinel-2, Sentinel-1, DEM)

UzEO is a multimodal Earth-observation corpus covering three contrasting landscapes of Uzbekistan — the irrigated Fergana Valley, the Tashkent urban region, and Karakalpakstan (Aral Sea basin) — delineated by national administrative boundaries rather than a bounding box. The corpus contains 709 grid cells within the borders of Uzbekistan, each covering approximately 10.68 x 10.68 km at 512 x 512 pixels, with three co-registered modalities: Sentinel-2 L2A optical (709 cells), Sentinel-1 RTC radar (648 cells), and Copernicus DEM elevation (709 cells). Cells were selected by a point-in-polygon test against GADM v4.1 level-1 boundaries; 651 cells falling outside Uzbekistan were excluded, ensuring that per-region statistics correspond to the stated territory. Sentinel-2 and DEM are complete for all 709 cells; Sentinel-1 is available for 648 cells (91.4%), with 61 cells lacking radar owing to gaps in the source acquisition coverage. A manifest.json lists every cell with its grid identifier, region, coordinates, and modality paths. The corpus is derived from the open Major TOM Core archive and is intended for self-supervised pretraining of regional geospatial foundation models, multimodal fusion, soil-salinity and land-degradation studies, crop and land-cover analysis, and urban change analysis over Central Asia. Source imagery: Copernicus Sentinel data via the Major TOM Core archive. Administrative boundaries: GADM v4.1.