Abstract
Most of the currently experienced global environmental changes (rising atmospheric CO2 concentrations, warming, altered amount and pattern of precipitation, and increased nutrient load) directly or indirectly affect ecosystem surface energy balance and plant transpiration. As a consequence, the relative humidity of the air surrounding the vegetation changes, thus creating a feedback loop whose net effect on transpiration and finally productivity is not trivial to quantify. Forcedly, in any global change experiment with the above drivers, we can only treat small plots, or 'islands', of vegetation. This means that the treated plots will likely experience the ambient humidity conditions influenced by the surrounding, non-treated vegetation. Experimental assessments of global change effects will thus systematically lack modifications originating from these potentially important feedback mechanisms, introducing a bias of unknown magnitude in all measurements of processes directly or indirectly depending on plant transpiration. We call this potential bias the 'island effect'. Here, we discuss its implications in various global change experiments with plants. We also suggest ways to complement experiments using modelling approaches and observational studies. Ultimately, there is no obvious solution to deal with the island effect in field experiments and only models can provide an estimate of modification of responses by these feedbacks. However, we suggest that increasing the awareness of the island effect among both experimental researchers and modellers will greatly improve the interpretation of vegetation responses to global change.