This dataset supports first-stage site screening for Direct Air Capture and CO₂ Storage (DACCS) in the North German Basin (NGB) and the German North Sea. It comprises (1) multi-criteria decision-making (MCDM) suitability layers and scores for onshore and offshore areas, derived from geospatial criteria (elevation, soil type, population, infrastructure, storage traps, renewable energy potential, etc.); (2) aligned Sentinel-2 satellite imagery with MCDM-derived suitability labels; (3) DINOv3 (ViT-L/16) patch features extracted from that imagery; (4) a trained deep neural network (RegressionMLP) that predicts suitability from Sentinel-2 images alone (`suitability_regression_model.pth`); and (5) evaluation outputs (metrics, calibration, confusion matrix, spatial and ranking indices). The MCDM maps provide reference suitability in 0–1 for rational site selection; the trained model enables suitability prediction from only Sentinel-2 imagery in data-sparse regions where full MCDM layers are unavailable. All workflows are scripted (Python) and reproducible with documented dependencies and paths.