Biodiversity is declining globally. Yet, the biological and genetic processes associated with these declines on a longer timescale are still poorly understood. Ancient DNA is a powerful tool to study evolution in real-time. Despite advances in the field, there is further need for refinement of laboratory and computational techniques. In this thesis, I used mitochondrial and nuclear genomes, as well as radiocarbon data, to study the evolutionary history and extinction dynamics of the woolly mammoth (Mammuthus primigenius). In Chapter I, I developed and optimized a silica column-based extraction protocol for ancient DNA. Based on systematic tests, I advise against routine use of pretreatment methods, like bleach wash and/or predigestion, for well-preserved permafrost samples. Furthermore, I suggest that USER enzyme, which removes uracil from damaged DNA molecules, is effective at half the concentration compared to an established control protocol. Finally, I did not find a significant difference between different silica columns for the clean-up steps, or concentrator columns with different DNA retention sizes. In Chapter II, I used five high coverage Siberian mammoth genomes to develop a method based on differences in read depth to identify indels, insertions and deletions, in the mammoth genome. The results show that indels are enriched in intergenic regions, suggesting strong selection against structural variants affecting gene function. Nevertheless, 87 genes were identified that were severely affected. These genes are related to various functions like body-fat distribution, fur growth and hair shape, body temperature, and body size, and most likely represent important adaptations to the cold steppe-tundra. In Chapter III, I studied the population and extinction dynamics of the woolly mammoths in Siberia by combining Bayesian age models from radiocarbon data with inferences from complete mitogenomes. The results show that the woolly mammoth’s extinction was a complex process with consecutive extirpations, but also partial recolonizations, occurring in different Siberian localities. I hypothesize that Wrangel Island, one of the last refugia of the woolly mammoth, was colonized by mammoths from, or closely related to, a population from central or western Siberia. Mammoths reappeared on the island around 10 ky ago, where they became isolated due to rising sea levels, and went extinct around 4 ky ago. To investigate the genetic consequences of the founder bottleneck event of Wrangel Island and long-term survival at small population size, I analyzed time-series data of 21 woolly mammoth genomes in Chapter IV. Changes in heterozygosity and inbreeding show that while the initial bottleneck was dramatic, the population recovered fast and remained remarkably stable. Analysis of mutation load shows that mammoths purged highly deleterious mutations over time, but also accumulated slightly deleterious mutations, indicating reduced efficacy of selection. Nevertheless, there were no clear signs of a mutational meltdown, and it remains a mystery why mammoths went extinct on the island. In conclusion, this thesis presents methodological advances for DNA extraction and detecting structural variants (i.e., deletions) in ancient genomes. Furthermore, I demonstrate that extinctions are a complex process. Finally, I show that ancient DNA is a powerful tool to study evolutionary processes over long timescales.