The quantitative description of the soil surface roughness is necessary for effective monitoring of wind, water and tillage erosion, hydrological processes or greenhouse gas emissions. The aim of this work was to build soil roughness predictive models based on the type of tillage tool, the roughness indices and soil properties. The roughness formed by five tillage tools was determined. Two surface roughness indices: Height Standard Deviation (HSD) and T3D (Tortuosity index) were calculated from Digital Elevation Model. The both roughness indices demonstrated a significant correlation, however, they provided different information about soil roughness. The HSD describes roughness on the “macro,” while T3D refers to the “micro” scale. Hence, our findings show that a single index is not sufficient to describe the roughness of post‐treatment surface. The linear and random forest models were built to describe the relationships between the roughness indices, type of tillage tool and soil properties. The HSD analysis indicated that the type of tillage tools had the greatest impact on post‐treatment roughness. In contrast, T3D analysis found soil texture to have a significant effect, together with tillage tools. In all modeling scenarios, T3D was more accurately predicted than HSD by both the linear (R2 = 0.62 vs. R2 = 0.60) and random forest models (R2 = 0.58 vs. R2 = 0.55). Predictive soil surface roughness models can be applied effectively for estimating water retention in soil, the intensity and speed of surface water flow, soil erosion, the level of reflected shortwave solar radiation or soil properties by remote sensing techniques.