At Klosterhede, Denmark experimental manipulations of water and nutrient inputs to a forest stand were carried out under a 1200 m² roof simulating i) summer drought, ii) removal of N and S input, and iii) optimal nutrition and water availability. In practise, manipulation of element fluxes in a complex system like a forest may cause unintended changes such as climatic effects from the roof which may interfere with the results of the intended manipulation. This paper illustrates the extent and effects of such unintended changes. The most important climatic change under the roof was a 50% reduction of photosynthetic light which caused a decrease in moss cover on the ground floor. Soil temperatures under the roof were decreased 0.3°C during summer and increased 0.2°C during winter. Air humidity was not changed. Throughfall water amount decreased close to the stems whereas element concentrations in throughfall increased close to the stems and with tree size on the ambient plots. This distinct spatial pattern was changed under the roof to a more variable water distribution from the sprinkling system and to constant element concentrations in the sprinkling water. This loss/change of spatial variability in the input was reflected in the soil solution. The concentration patterns found in soil water (increasing close to the stem and with tree size) in the ambient plots were no longer observed under the roof. Exclusion of throughfall by the roof disturbed the internal cycle of nutrients leached from the canopy. The removal of canopy leached Ca and K by the roof caused a decline in soil water concentrations, even on the drought plot where natural throughfall was only cut off during two summer months. The internal cycle of Ca and K had to be restored by addition of Ca and K under the roof. Further, leachable elements and soluble compounds may be washed out from litterfall during the collection period on the roof, and thus lost from the internal cycle by by exclusion of throughfall. It is important to consider these problems in the design of new manipulation experiments and in the interpretation of the results. Some recommendations are given.