Measuring the interactions between engineered nanoparticles and natural substrates (e.g. soils and sediments) has been very challenging due to highly heterogeneous and rough natural surfaces. In this study, three-dimensional nanostructured slanted columnar thin films (SCTFs), with well-defined roughness height and spacing, have been used to mimic surface roughness. Interactions between titanium dioxide nanoparticles (TiONP), the most extensively manufactured engineered nanomaterials, and SCTF coated surfaces were measured using a quartz crystal microbalance with dissipation monitoring (QCM-D). In parallel, in-situ generalized ellipsometry (GE) was coupled with QCM-D to simultaneously measure the amount of TiONP deposited on the surface of SCTF. While GE is insensitive to effects of mechanical water entrapment variations in roughness spaces, we found that the viscoelastic model, a typical QCM-D model analysis approach, overestimates the mass of deposited TiONP. This overestimation arises from overlaid frequency changes caused by particle deposition as well as additional water entrapment and partial water displacement upon nanoparticle adsorption. Here, we demonstrate a new approach to model QCM-D data, accounting for both viscoelastic effects and the effects of roughness-retained water. Finally, the porosity of attached TiONP layer was determined by coupling the areal mass density determined by QCM-D and independent GE measurements.