The study area is located in an eastern beech stand on a fertile and mesic upland site in the western Black Sea region of Turkey. During natural regeneration of the beech stand, rhododendron was differentially controlled. The experiment used a randomized block design with four replicates and two different site preparation practices. The manual treatment was hand grubbing. The whole plant was uprooted and left on the site. The mechanical treatment scarified the surface soil to 5–10 cm depth with a bulldozer equipped with a brush rake. After 17 years following site preparation, the number of trees on the grubbing site (GS) was 60% more than that of mechanical treatment sites (MTS). Total tree biomass per hectare was 39% greater on GS than that of MTS. Beech trees on the grubbing sites had 18, 115, 29, 17 and 78% more N, P, K, Ca and Mg than MTS, respectively. Forest floor N content on GS was 50% more than that of MTS. On the GS, Ca content also was three times more than that of MTS. At 0–20 cm depth, cation exchange capacity on GS was 48% higher than that of MTS. At the same depth, GS had 50, 100, 78, 23, 8, 10 and 49% more C, N, P, Ca, K, Mg and S concentrations than those of MTS, respectively. Removing of organic layer and upper soil with mechanical site preparation depletes ecosystem nutrient pools compared with manual grubbing. This difference is sustained in the long‐term. Grubbing is both effective and by leaving organic layer on the sites, grubbing preserves the long‐term productivity of the ecosystem.