The F. excelsior samaras originated from different trees growing in forest seed stands in Sachsen. The samaras of both species were put in 16 glass beakers (10 cm width, 12 cm height, 600 ml volume) and

filled with about 300 ml distilled water ( Horsch, 2001 and van den Broek et al., 2005), corresponding to eight replications with 50 samaras per species. The beakers were placed on a flaskshaker, which moved gently at a frequency of 100 /min and at an amplitude of about 1 cm. The water movement prevented the adherence of the samaras to the glass beakers. The proportion of samaras floating per tree species was captured at progressive time intervals (after 5 min, after 1, 2, 4, 6, 9, 24 and 72 h and after 1 week). The experiment ended check details after 1 week, when nearly all of the samaras had sunk to the bottom of the beakers. The data were analysed in Origin 8G (OriginLab Corporation, USA). The dependence of buoyancy on time was described using the χ2 minimisation fitting routine. For the fitting routine, 200 iterations were performed. The best fitting model was selected by evaluating the goodness-of-fit criteria

(R2 and χ2/df values). R2 is the adjusted coefficient of determination and χ2/df represents the magnitude 17-AAG of scattering (χ2) of observed data and a theoretical curve normalised by a degree of freedom (df). The time-dependent buoyancy (number of samaras y(x)) was Tangeritin described using a four-parameter logistic growth function (dose–response function): equation(1) y(x)=A2+A1-A21+(xx0)pParameter A1 describes the minimum value of the asymptote, A2 the final value and the parameter p indicates the power of the function. x0 is the inflexion point of the function and corresponds to the species-specific half-value period when 50% of the samaras have sunk. Accordingly, results were obtained for the two most important parameters: the maximum floating time and the proportion of seeds floating for a certain time period. The data for the wind dispersal distances of both ash species derived from simulations using the programme PAPPUS (Tackenberg,

2003), which is a process-based seed dispersal model. In the model the plant species are characterised by the terminal velocity of their diaspores (Vterm: F. excelsior = 1.58 m/s; F. pennsylvanica = 1.62 m/s) and the height of the infructescence, the ‘release height’ (Hrel: F. excelsior = 25 m; F. pennsylvanica = 20 m). The wind-vector and turbulence data were measured over a period of 1 year for a low-cut grassland situated within a flat landscape in central Europe ( Tackenberg et al., 2003). The wind kernels were first computed for a limited number of combinations of Vterm and Hrel, and the species’ kernels were subsequently drawn from these data by means of bilinear interpolation according to the mean terminal velocity and release height.