The multisource rasters for the detection of ancient irrigation traces - central Iraq - 48 bands

The dataset comprises a composite created in Google Earth Engine and reference data, which are used to detect levees in Mesopotamia with the Swin UNETR deep learning model and to evaluate its accuracy. The data used in the study were assembled from multitemporal Landsat 5 imagery, Sentinel-1 SAR imagery, Sentinel-2 multispectral imagery, and the Copernicus GLO-30 DSM and were combined with vegetation and moisture indices, PCA reductions of seasonal variability, and Multi-Scale Relief Model (MSRM) outputs.