Abstract
DNA from the environment (eDNA) has been increasingly used as a new tool to conduct biodiversity assessment. Because of its noninvasive and less time-consuming nature, many studies of recent years solely rely on this information to establish a species inventory. eDNA metabarcoding has been shown to be an efficient method in aquatic ecosystems, especially for fish. However, detection efficiency is not clear for mammals. Using the existing literature, we conducted a meta-analysis to investigate if eDNA metabarcoding allows greater detection success compared to conventional surveys (such as field surveys, camera traps, etc.). Although only 28 articles were retrieved, showing the lack of comparative studies, still representing more than 900 taxa detected, we found that detection success was method dependent, but most importantly varies on the taxonomy of the targeted taxa. eDNA metabarcoding performed poorly for bats compared to the traditional mist nests. However, strong detection overlaps were found between conventional surveys and eDNA for large-bodied mammals such as ungulates, primates, and carnivores. Overall, we argue that using both molecular and field approaches can complement each other and can maximize the most accurate biodiversity assessment and there is much room for metabarcoding optimization to reach their full potential compared to traditional surveys.