Literature search and data extraction
Using Google Scholar (https://scholar.google.co.jp/), we conducted a literature search relating to eDNA production and species abundance published during 2008 to 2021. the literature search was 1 June, 2021). The terms “eDNA” or “environmental DNA”, included in the title and/or text, were targeted for the literature search. We then filtered and selected papers that (i) targeted eDNA from macro-organisms (not from microbes such as bacteria and fungi, or viruses), (ii) were published in international journals, (iii) were peer-reviewed (not preprints), and (iv) described the relationships between eDNA concentration, quantified by real-time PCR or digital PCR, and species abundance (biomass and/or density) by fitting linear or linear-mixed models. For criterion (iv), most studies reported positive relationships with statistical significances, while some of results were not significant.
We then extracted R2 values from linear fitting, indices of species abundance estimation accuracy based on eDNA concentration, target taxa, filter pore sizes used for water filtration (μm), and PCR amplicon sizes (base pair; bp) from the selected studies. Taxa were classified as fish, amphibians, other vertebrates, crustaceans, mussels, and other invertebrates. In studies involving aqueous eDNA collection via precipitation or centrifugation, the filter pore size was regarded as 0 μm. We did not include Pearson’s correlation coefficients in our meta-analyses because the index was not available in all collected literatures. If the manuscripts only reported Pearson’s correlation coefficients, we squared the coefficients and substituted them for R2 values. Different R2 values based on different experimental conditions within the same study (e.g., species, filter type, and amplicon size) were treated separately. Abundance metrics (biomass/density) were pooled here because of its mere effect on correlation between eDNA concentration and species abundance (Yates et al., 2019). In addition, we extracted information on the study environments and classified them as laboratory, lentic freshwater, lotic freshwater, and marine environments. Moreover, we calculated the sample size (the number of water samples or sampling sites) required for fitting each linear model based on figures and/or text in the corresponding literature.