A vegetation change from open tundra to dense taiga will fuel the global warming by positive feedback caused by albedo decreases. Yet, it is unclear how fast the arctic treeline, formed of Larix species, will advance north in the next decades. The most determinant factor of tree migration is the ability to disperse seeds (and pollen). Hence, to realistically forecast the migration of tree species in a dynamic vegetation model, it is crucial to incorporate reliable estimates of dispersal. Classical methods, for example counting seeds in seed traps, have been used to describe local dispersal abilities but are not applicable to give precise estimates on rare long-distance dispersal events. In this study we overcome this with the help of modern molecular techniques. By using a set of 16 nuclear microsatellites we inferred the cryptic signal of heritage among larch individuals to study the migration history among well-established tree stands and for different time-cohorts. We analyzed the genetic structure of larch populations for several latitudinal transects spanning north-to-south from tundra to open taiga forests in Siberia and additionally of several age cohorts which established throughout the last century in prevailing cold and warm periods. Finally, we present the results of simulations with our individual-based model LAVESI which was developed by us originally to study population dynamics of larch forest stands. Using downscaled global climate models and 'representative carbon pathway' (RCP) scenarios it is feasible to project the future treeline in Siberia.