The Arctic treeline ecotone is characterised by a steep vegetation gradient from arctic tundra to northern taiga forests, which is thought to influence the water chemistry of thermokarst lakes in this region. Environmentally sensitive diatoms respond to such ecological changes in terms of variation in diatom diversity and richness, which so far has only been documented by microscopic surveys. We applied next-generation sequencing to analyse the diatom composition of lake sediment DNA extracted from 32 lakes across the treeline in the Khatanga region, Siberia, using a short fragment of the rbcL chloroplast gene as a genetic barcode. We compared diatom richness and diversity obtained from the genetic approach with diatom counts from traditional microscopic analysis. Both datasets were employed to investigate diversity and relationships with environmental variables, using ordination methods. After effective filtering of the raw data, the two methods gave similar results for diatom richness and composition at the genus level (DNA 12 taxa; morphology 19 taxa), even though there was a much higher absolute number of sequences obtained per genetic sample (median 50,278), compared with microscopic counts (median 426). Dissolved organic carbon explained the highest percentage of variance in both datasets (14.2 % DNA; 18.7 % morphology), reflecting the compositional turnover of diatom assemblages along the tundra-taiga transition. Differences between the two approaches are mostly a consequence of the filtering process of genetic data and limitations of genetic references in the database, which restricted the determination of genetically identified sequence types to the genus level. The morphological approach, however, allowed identifications mostly to species level, which permits better ecological interpretation of the diatom data. Nevertheless, because of a rapidly increasing reference database, the genetic approach with sediment DNA will, in the future, enable reliable investigations of diatom composition from lake sediments that will have potential applications in both paleoecology and environmental monitoring.