The application of the seismic reflection method is often limited in areas of complex terrain. The problem is the incorrect correction of time shifts caused by topography. To apply normal moveout (NMO) correction to reflection data correctly, static corrections are necessary to be applied in advance for the compensation of the time distortions of topography and the time delays from near-surface weathered layers. For environment and engineering investigation, weathered layers are our targets, so that the static correction mainly serves the adjustment of time shifts due to an undulating surface. In practice, seismic reflected raypaths are assumed to be almost vertical through the near-surface layers because they have much lower velocities than layers below. This assumption is acceptable in most cases since it results in little residual error for small elevation changes and small offsets in reflection events. Although static algorithms based on choosing a floating datum related to common midpoint gathers or residual surface-consistent functions are available and effective, errors caused by the assumption of vertical raypaths often generate pseudo-indications of structures. This paper presents the comparison of applying corrections based on the vertical raypaths and bias (non-vertical) raypaths. It also provides an approach of combining elevation and NMO corrections. The advantages of the approach are demonstrated by synthetic and real-world examples of multi-coverage seismic reflection surveys on rough topography. ?? The Royal Society of New Zealand 2005.