Estimation accuracy and environments
Higher R2 values were reported for laboratory conditions than natural environments, which is consistent with previous findings (Yates et al., 2019) and supports the methodological validity of our statistical analyses. As discussed in Yates et al. (2019), myriad and complex dynamics of eDNA could hamper the prediction of species abundance based on eDNA concentration in natural than controlled environments. However, we observed little difference in estimation accuracy among natural environments. Relative to lentic freshwater environments, such as ponds and lakes, eDNA diffusion and degradation would be substantial in riverine, coastal, and marine environments due to flow and tidal effects. Such factors can transport eDNA very long distances (Deiner & Altermatt, 2014; Andruszkiewicz et al., 2019) while also resulting in rapid dilution (Baker et al., 2018). In contrast, although eDNA may diffuse less in lentic environments, residual eDNA may cause false-positive inferences of species presence and inaccurate estimation of species abundance. It is noted that our results must be cautiously interpreted given the bias in the number of studies among natural environment types, but eDNA-based abundance estimation accuracy might not necessarily be worsened for lotic freshwater and marine environments relative to lentic environments.
Additionally, the spatial location of eDNA sampling within the study environments might also influence the results. Marine eDNA studies generally collected water samples from both near-shore and offshore sites, whereas some lentic freshwater studies relied solely on water sampling from shore sites (e.g., Kamoroff & Goldberg 2018; Kakuda et al., 2019). Even if eDNA detection rates were higher at shore sites (Hänfling et al., 2016; Jo et al., 2021), the lowered diffusion in lentic environments and highly heterogeneous distribution (Hänfling et al., 2016; Li et al., 2019) may not accurately reflect the overall species abundance in studied sites, especially when large lakes and reservoirs were targeted. Further studies are required to clarify how the heterogeneity of eDNA distribution influences the detection sensitivity, yield, and relationship to species abundance.