Method used to produce the sediment DNA datasets: DNA sequences were generated using g (5'-GGGCAATCCTGAGCCAA-3') and h (5'-CCATTGAGTCTCTGCACCTATC-3') primers for plants and MamP007F (5'-CGAGAAGACCCTATGGAGCT-3') and MamP007R (5'-CCGAGGTCRCCCCAACC-3') primers and the human-blocking primer (5'-GGAGCTTTAATTTATTAATGCAAACAGTACCC-3') for mammals. The sequences have been produced by the Illumina technology (HiSeq instrument). The bioinformatic treatment of the sequences was operated applying the Obitool program (Boyer, F. et al. OBITOOLS:aUNIX-inspired software package for DNAmetabarcoding. Mol. Ecol. Resour. 16, 176–182 (2016)) and following the procedure described in Giguet-Covex et al. New insights on lake sediment DNA from the catchment: importance of taphonomic and analytical issues on the record quality. Scientific Reports. 9, 14676 (2019) (DOI https://doi.org/10.1038/s41598-019-50339-1). Further filtering treatments were then applied, as also described in Giguet-Covex et al. 2019. These second treatments allow the removal of 1) taxa assigned to a reference taxon at <95 % of sequence identity, 2) potential contaminations (based on sampling, extraction and PCR controls, and recognition of exotic taxa) and 3) stochastic detections. Several analytical replicates by samples were performed (8 for both, the plants and mammals). Data are presented in number of positive replicates by sample. A replicate is considered as positive if more than 5 sequences are detected.