Multispecies site occupancy modelling and study design for spatially replicated environmental DNA metabarcoding

Abstract

1. Environmental DNA (eDNA) metabarcoding has become widely applied to gauge biodiversity in a non-invasive and cost-efficient manner. The detection of species using eDNA metabarcoding is, however, imperfect owing to various factors that can cause false negatives in the inherent multistage workflow. Imperfect detection in the multistage workflow of eDNA metabarcoding also raises an issue of study design, namely, how available resources should be allocated among the different stages to optimize survey efficiency.2. Here, we propose a variant of the multispecies site occupancy model for eDNA metabarcoding studies where samples are collected at multiple sites within a region of interest. Unlike traditional site occupancy models that deal with binary detection data, this model describes the variation in sequence reads, the output of the high-throughput sequencers. It explicitly accounts for the hierarchical workflow of eDNA metabarcoding and interspecific heterogeneity and allows the analysis of the sources of variation in the detectability of species throughout the different stages of the workflow. We also introduced a Bayesian decision analysis framework to identify the study design that optimizes the effectiveness of species detection with a limited budget.3. An application of the model to freshwater fish communities in the Lake Kasumigaura watershed, in Japan, highlighted a remarkable inhomogeneity in the detectability of species, indicating a potential risk of the biased detection of specific species.Species with lower site occupancy probabilities tended to be difficult to detect as they had lower capture probabilities and fewer sequence reads. The expected abundance of sequence reads was predicted to vary by up to 23.5 times among species. An analysis of the study design suggested that ensuring multiple withinsite replications of the environmental samples is preferred to achieve higher species detection effectiveness, provided that tens of thousands of sequence reads were secured per replicate.4. The proposed framework makes the application of eDNA metabarcoding more error-tolerant, allowing ecologists to monitor ecological communities more efficiently.

Evaluating the sampling effort for the metabarcoding‐based detection of fish environmental DNA in the open ocean

“…Therefore, accumulating basic knowledge of the ecology of eDNA in various environments in the open ocean is required to develop the practical use of eDNA. Using a statistical model, such as a multispecies site occupancy model that estimates occurrence/detection probabilities of species based on eDNA detection, will help to determine an efficient sampling strategy and optimal sampling effort (Doi et al, 2019; Fukaya et al, 2022). Further studies attempting to integrate field experiments and statistical approaches will contribute to establishing a basis for developing an optimal eDNA survey for the open ocean.…”

“…It is also expected that analyzing spatial and temporal replicates of eDNA samples to account for the heterogeneity of eDNA distribution will result in a more integrative and consistent estimation of fish community composition, though this possibility was not explored in this study. However, more extensive sampling requires more time and manpower, which could lead to higher analytical costs (Fukaya et al, 2022). Therefore, we suggest that the sampling strategy of eDNA for investigating fish diversity and distribution in the open ocean should be optimized for practical use by maximizing the robustness and reproducibility of metabarcoding results with allowable sampling effort, rather than aiming for complete detection of local fish fauna.…”

Evaluating environmental DNA metabarcoding as a survey tool for unionid mussel assessments

“…Therefore, the use of additional gene regions may help reduce the potential for mis-assignment of MOTUs (Klymus et al, 2021). The ecology of eDNA and its transport within the environment remains a need for future research, and further improvement of eDNA interpretations may be possible with the use of eDNA occupancy models (Fukaya et al, 2022). The unique mussel rescue survey within this study provided a rare opportunity for an in-depth analysis of eDNA detection across both small sampling cells (152 m) and a large sampling region (c. 1.5 km).…”

Integrated Population Models: Achieving Their Potential

“…In particular, the form of available data is continually evolving, particularly with advances in technology [104]. For example, citizen science data collection continues to grow in popularity [105,106] and eDNA data collection is increasingly being used due to its capability to detect multiple species from water or air samples [107]. Remote sensing technology such as drones are providing finer-scale aerial survey data of animals [108], satellite earth observation data over fine scales of 30-50cm are becoming available across larger geographical areas and acoustic recording technology is enabling monitoring of elusive marine species [109], from which machine learning techniques can provide population count estimates [110].…”