The Arctic is experiencing particularly high temperature increases, which will potentially cause major ecosystem changes, such as a northward expansion of boreal forests. About half of the global circum-arctic boreal treeline is located in Siberia, where it is formed by one of three species of larch, Larix sibirica, L. gmelinii and L. cajanderi, distributed from West to East, respectively. They hybridize in their boundary areas, but show ecological separation, especially with regard to survival on permafrost, and they differ in a number of other characteristics, such as growth rates and dispersal distances. Previously published models suggest an overall distributional shift of these species to the Northeast, as they track their climatic envelopes, but these projections do not consider biogeographical constraints or interspecific competition. Empirical data on past larch forest dynamics can aid here, as significant fluctuations in forest extent during the Holocene and Pleistocene are documented. DNA stored in sediments offers the possibility to obtain such data. Most analyses of environmental DNA employ metabarcoding techniques that provide an overview of biotic communities. As a consequence of targeting large groups, the genetic markers employed are limited in their taxonomic resolution, but within the sedimentary DNA we can also target much more variable markers to track dynamics of single species. We are using sedimentary ancient DNA to analyze past vegetation changes with DNA metabarcoding and the past distribution of single-nucleotide polymorphisms (SNPs) in mitochondrial and chloroplast DNA of Larix. On the southern Taymyr peninsula, where the ranges of L. sibirica and L. gmelinii come together, we trace temporal changes of mitochondrial haplotypes through most of the Holocene. A comparison of these results with simulations using the larch population dynamics model LAVESI indicate that projections including both the bioclimatic envelope and interspecific competition predicht climate change outcomes for larch forests more accurately.