Evaluating Sampling Designs to Survey Fish Diversity in Lakes From Northern Temperate Zones

Long-term biological monitoring and management depend on efficient protocols and methodology to characterize and precisely describe species distributions and diversity. In recent years, environmental DNA has progressively become a tool of choice in survey programs. However, the effect of variables such as sampling effort and sampling design still requires consideration. Simple random, grid, and transect-based sampling methods are widely used in ecological surveys to obtain an unbiased estimation of species richness and community structure. However, under certain conditions where spatial information is available, sampling design and sequencing depth can be optimized to reduce effort and cost. Here, we evaluate different subsampling approaches to identify sampling strategies that are both easily implemented in the field and provide optimal recovery of species diversity for a given sampling effort. With a homogeneous grid-based sampling (25–50 samples by lake) of 12 freshwater lakes in southeastern Québec, and using the 12S MiFish metabarcoding primer set, we demonstrate that random and stratified designs perform similarly to detect 90% and 95% of species. However, we found that, under certain circumstances, stratified sampling outperformed random sampling, requiring lower numbers of samples to detect the same species diversity. We also demonstrate that for the minimum sequence threshold and sample replication used in our study, a sequencing depth of 50K reads per sample is adequate to obtain a reliable portrayal of species richness. In this study, we contribute to the effort of eDNA sampling standardization by providing data for selecting the best sampling design, sequence depth, and sample size to detect 90%–95% of fish species found in temperate lakes.