What shapes giant hogweed invasion? Answers from a spatio-temporal model integrating multiscale monitoring data

In this study we simulated the invasion of Heracleum mantegazzianum with a spatiotemporal model that combined a life-cycle matrix model with mechanistic local and corridor dispersal and a stochastic long-distance dispersal in a cellular automaton. The model was applied to the habitat configuration and invader distribution of eight 1 kmA(2) study areas. Comparing the simulations with monitoring data collected over 7 years (2002-2009) yielded a modelling efficiency of 0.94. We tested the significance of different mechanisms of invasion by omitting or modifying single model components one at a time. Thus we found that the extent of H. mantegazzianum invasion at landscape level depends on both landscape-scale processes and local processes which control recruitment success and population density. Limiting recruitment success (100 -> 30 %) and successionally decreasing the carrying capacity of habitats (max -> 0) over 30 years significantly improved the projections of the invasion at the landscape level. Local dispersal reached farther than 10 m, i.e. farther than previously assumed, but appeared to be unaffected by wind directions. Long-distance dispersal together with local dispersal dominated the invasion quantitatively. Dispersal through corridors accounted for less invasive spread. Its importance, with respect to invasion speed (number of colonised model grid cells) is probably limited over short periods of time (7 years). Only dispersal along rivers made a significant quantitative contribution to invasion of H. mantegazzianum. We suggest that biotic heterogeneity of suitable habitats is responsible for varying invasion success and that successionally increasing competition leads to declining population densities of H. mantegazzianum over several decades slowing down the spread on the landscape scale.