Terahertz scattering-type scanning near-field optical microscopy (THz-s-SNOM) has emerged as a powerful technique for high-resolution imaging. However, most previous studies have focused on simplified smooth surface models, overlooking the realistic surface roughness induced by contamination during sample preparation. In this work, we present a novel 3D model, to the best of our knowledge, that combines the point dipole model with the finite element method to investigate the influence of sample morphology on scattered signals. We explore surfaces with a protrusion, a depression, and random roughness, characterizing the variations in scattered signals and highlighting the role of higher-order scattering in mitigating surface roughness effects. Our findings provide valuable insights into the impact of sample morphology on THz-s-SNOM imaging.