Seismic monitoring in desert environments has many challenges and changes in surface sand topography in one of them. We present a numerical study of the impact of surface elevation changes on repeatability of seismic data recorded with buried receivers. We focus on the early arrivals since they are affected only by the near-surface structure. We define changes in the surface topography as a homogeneous Gaussian random field. We show that for a homogeneous near surface layer, NRMS and predictability depend only on changes in the surface topography but not on its slope. For a heterogeneous near surface we observe worse repeatability for the zones with a thin sand layer (< 5 m), whereas areas of thick sand (> 10m) behave similarly to a homogeneous model may cause significant non-repeatability up to 60% of NRMS error and predictability down to 75%. These numbers are similar to the NRMS measured on field data in Saudi Arabia, suggesting that such factors may be significant for land 4D seismic in a desert. In addition, sand topography variations can accumulate thus explaining experimentally observed trends showing that land seismic repeatability degrades over time from days to months to years.