Water retention in granular soils is a key mechanism for understanding transport processes in the vadose zone for various applications from agronomy to hydrological and environmental sciences. The macroscopic pattern of water entrapment is mainly driven by the pore-scale morphology and capillary and gravity forces. In the present study, the drainage water retention curve (WRC) was measured for three different granular materials using a miniaturised hanging column apparatus. The samples were scanned using X-ray micro-computed tomography during the experiment. A segmentation procedure was applied to identify air, water and solid phases in 3D at the pore-scale. A representative elementary volume analysis based on volume and surface properties validated the experimental setup size. A morphological approach, the voxel percolation method (VPM) was used to model the drainage experiment under the assumption of capillary-dominated quasi-static flow. At the macro-scale, the VPM showed a good capability to predict the WRC when compared with direct experimental measurements. An in-depth comparison with image data also revealed a satisfactory agreement concerning both the average volumetric distributions and the pore-scale local topology. Image voxelisation and the quasi-static assumption of VPM are likely to explain minor discrepancies observed at low suctions and for coarser materials.