Estimation accuracy and eDNA production source
Since the first study targeting common carp (Cyprinus carpio ) (Takahara et al., 2012), most research investigating the correlation between eDNA concentration and species abundance has targeted fish species (e.g., Klymus et al., 2015; Jo et al., 2017; Capo et al., 2021; Table 1). According to these studies, fish eDNA concentration predicts relative species abundance with relatively high accuracy regardless of environmental conditions, which was also confirmed in our meta-analysis. However, the meta-analysis showed that such tendencies might not necessarily be similar among target taxa, and the abundance of some taxa may not be accurately estimated via current eDNA applications.
We found that eDNA-based estimation accuracy of species abundance (R2) was significantly lower for crustaceans and mussels than fish. A previous study hypothesized that species morphology and/or behavior might affect eDNA production, reporting lower eDNA shedding rates in grass shrimp than in fish and jellyfish (Andruszkiewicz et al., 2021). Fish and jellyfish are likely to constantly produce eDNA as epidermis and/or muco-substances (Merkes et al., 2014; Sassoubre et al., 2016). In contrast, crustaceans are characterized by their hard exoskeletons and segmented bodies plans (Hadley, 1986) and are thus unlikely to shed large amounts of eDNA from their body surfaces unless they are molting. Similarly, mussel soft tissue is covered with a hard, calcified shell that is less likely to shed eDNA (Sansom & Sassoubre, 2017). Consequently, crustaceans and mussels infrequently and irregularly shed eDNA, which may impede sufficient eDNA detection in the field and prevent accurate abundance estimation via eDNA analysis (Dougherty et al., 2016; Mächler et al., 2016; Johnsen et al., 2020).
The mean correlation between amphibian eDNA concentration and abundance was similar to that of fish, which is reasonable given amphibians likely shed eDNA constantly from their epidermis and/or mucus. However, the variation (95 % CI) was much larger for amphibians relative to fish. The discrepancy could simply be explained by biases derived from the smaller number of corresponding studies or different experimental conditions. Among the studies collected for our meta-analysis, most of the correlations between amphibian eDNA concentration and abundance were studied in natural environments (e.g., Thomsen et al., 2012; Pilliod et al., 2013). This is likely because of the difficulty in conducting controlled laboratory experiments using amphibians due to its rarity, unless some invasive species such as American bullfrog (Lithobates catesbeianus ). In addition, possibly for the same reason, many of the collected studies depended on visual counts for abundance estimation, which could ambiguate correlations between amphibian eDNA concentration and abundance. On the other hand, Everts et al. (2021) assessed the correlation between eDNA concentration and abundance of American bullfrog tadpoles and juveniles using mesocosm experiments, and reported relatively high R2 values (0.64 to 0.99). Thus, amphibians are potentially suitable for accurate estimation of species abundance via eDNA analysis. Similarly, given that the study targeting mussels was not conducted under laboratory conditions (Currier et al., 2018), the relatively low R2 values for mussels may partly include such biases. Accumulating studies targeting various taxa in laboratory conditions and natural environments could help us understand the effects of ecological characteristics (morphology, physiology, and ethology) on the process of eDNA production, and may provide us with keys for improved approach of eDNA-based abundance estimation.