Context for and purpose: Retrieval of modern and ancient environmental DNA (eDNA) from sediments has revolutionized our ability to reconstruct present and past ecosystems. Little emphasis has been placed, however, on the fundamentals of the DNA-sediment associations and, consequently, our understanding of taphonomy and provenance of eDNA in sediments remains extremely limited. If we are to be able to accurately infer community dynamics across time and space from eDNA data, we need to understand how depositional processes and sedimentary associations of DNA molecules in different settings influence our interpretation. Approach and methods: Here, we introduce interfacial geochemical principles to the field of eDNA and discuss current interpretational biases. We outline a way to increase the scope and resolution of ecological interpretations from eDNA by combining mineralogic composition with experimental adsorption data. We apply distribution coefficients to assess the relationship between the DNA fraction in water columns and DNA fraction sequestered by suspended sediment particles. We further evaluate the challenges with drawing ecological inference using eDNA from sedimentary systems that receive input from different ecosystem types as a consequence of sedimentary processes. Main results: We show that: The retention of DNA in aqueous environments depends on the mineralogy of sediment particles and on the number of particles loaded in the water column. DNA attached to sediment particles from distal systems can be deposited in proximal systems and skew the interpretation of the proximal sediment samples. High particle loading in the water column can deplete suspended DNA and cause inaccurate interpretation of aqueous DNA samples. High particle loading in surface sediment pore waters enhances sequestration of DNA from benthic communities relative to that of water column communities, resulting in skewed estimates of species richness and abundance from sedimentary DNA. We discuss how to integrate taphonomy and provenance knowledge into the reconstruction of modern and past ecosystems, and ecosystem monitoring from eDNA data. Conclusions and the wider implications: Our findings demonstrate that integrating information about eDNA taphonomy and provenance into modern and past ecosystem reconstruction from eDNA data can enhance the scope, resolution and accuracy of our interpretations.