We conducted an restoration experiment on a dike at River Danube near Deggendorf. For the experiment we seeded 288 plots (2 x 3 m²) in 2018. The vegetation was surveyed until 2021. We established a full-factorial design on both dike expositions (north, south). The full set of treatments was replicated within six blocks. We used two different seed mix types (lowland hay meadow vs. dry grassland), two seed densities (4 vs. 8 g/m²), two soil depths (15 vs. 30 cm), and three sand admixtures (0, 25, 50%). The seed mixtures had two different species pools and each single mixture had an individual subset of 20 species out of the two species pools.