Comparing the efficacy of conventional survey methods to eDNA in detecting vertebrates in a tropical freshwater ecosystem in Madagascar

Madagascar is globally recognized for its exceptional biodiversity and endemism yet remains severely under-sampled for many taxa. We evaluate the effectiveness of environmental DNA (eDNA) metabarcoding as a rapid, non-invasive method for surveying vertebrate biodiversity in the Makira Forest Protected Area (MFPA), comparing its performance to conventional surveys across a range of vertebrate taxa. eDNA sampling detected 158 OTUs across 17 sites, outperforming conventional surveys in birds ( p = 0.000107), mammals ( p = 0.000718) amphibians ( p = 0.000725), reptiles ( p = 0.000877) and ray-finned fish ( p = 0.00788). We further examined the result for ray-finned fish by modelling species-accumulation curve asymptotes which supported this result. However, significant primer bias and amplification inconsistencies were observed, particularly in amphibian detections, and a high proportion of OTUs could not be resolved to species level due to taxonomic gaps in reference databases. We demonstrate, using species accumulation curves of our eDNA data and inventoried taxa lists of the MFPA, that the widespread practice of considering only resolved OTUs, rather than all unresolved OTUs, significantly underestimates biodiversity. This study underscores both the promise and current limitations of eDNA as a tool for tropical biodiversity assessment, specifically when compared to conventional sampling techniques. Combining eDNA and conventional survey methods increased species detection rates by 66.4%, highlighting their complementary strengths and reinforcing the value of integrated monitoring strategies for biodiversity-rich, data-deficient regions.