Predicting provenance of forensic soil samples: soil DNA predicts habitat and environmental properties

Environmental DNA is increasingly applied in ecological studies, including forensic ecology where eDNA from soil can be used to pair samples or reveal sample provenance. We collected soil eDNA samples as part of a large national biodiversity research project across 130 sites in Denmark. We investigated the potential for soil eDNA in predicting provenance in terms of environmental conditions, habitat characteristics and geographic regions. We used linear regression for predicting environmental gradients of light, moisture, soil pH and nutrients (represented by Ellenberg Indicator Values, EIVs) and quadratic discriminant analysis (QDA) to predict habitat class and geographic region. We found high predictive power for environmental gradients (R2 > 0.73). The discriminatory power of QDA in predicting habitat characteristics varied from high accuracy in predicting certain forest types, less accurate prediction of heathland and poor accuracy for geographic region. We demonstrate the application of provenance prediction in forensic science by evaluating and discussing two mock crime scenes. Here, we supplement with plant species lists from annotated sequences. Where predictions of environmental gradients and habitat classes give an overall accurate description of a crime scene, care should be taken when interpreting annotated sequences, e.g. due to erroneous assignments in GenBank. The outlined approach clearly demonstrates that basic ecological information that can be extracted from soil eDNA, contributing to the range of potential applications of eDNA in forensic ecology.