Mammalian–Environmental–DNA-Derived Probabilistic Modeling Reveals Re-Established Trophic and Co-Habitation Network Structure in Fire-Affected Terrestrial Communities

Climate change threatens natural ecosystems with more frequent and intense extreme weather events, including forest fires. This study evaluated mammalian communities, a primary bio-indicator of forest health, in Uljin Forest, using both traditional camera traps and advanced environmental DNA (eDNA) metabarcoding. Fire severity at the study sites was categorized based on field observations and satellite images using the differenced Normalized Burn Ratio. Water basins were installed for 7 days, after which the water was collected for eDNA extraction. A multi-primer approach using mitochondrial cytochrome c oxidase subunit 1 and 12S rRNA genes was employed. Comparative analysis of the mammalian relative abundance indicated that while camera traps remain the optimal tool for monitoring large mammals, eDNA methods were significantly more effective for quantifying small, cryptic species. Although long-term effects remain observable in the biodiversity and abundance of mammalian communities, Bayesian network analysis at the order level inferred positive ecological signals three years post-fire, with the food-web-related and cohabitation dynamics re-established. eDNA metabarcoding improved mammal detection relative to traditional methods and enabled probabilistic modelling to reveal ecological structure that would not be possible solely with biodiversity analysis. Overall, this multi-primer/multi-analysis approach highlights the potential of eDNA for ecosystem health monitoring.