Statistical Models for the Design and Analysis of Environmental DNA (eDNA) Surveys of Invasive and Imperiled Species
Science Center Objects
Detecting invasive species at low densities or prior to population establishment is critical for successful control and eradication. For example, Burmese pythons occupy thousands of square kilometers of mostly inaccessible habitats.
The Science Issue and Relevance: Detecting invasive species at low densities or prior to population establishment is critical for successful control and eradication. For example, Burmese pythons occupy thousands of square kilometers of mostly inaccessible habitats. Previous tools aimed at detecting and controlling Burmese pythons have resulted in a < 1% detection rate. These include traditional tools such as visual searching or trapping as well as innovative tools for detection and control (e.g., detector dogs, remote sensing, attractant traps, “Judas snakes”, etc.). As an alternative, application of environmental DNA (eDNA) tools have effectively detected Burmese pythons, increasing detection rates from 1% to 90% (Hunter et al., 2015).
Environmental DNA methods can assist with early detection of novel invasive species and range delimitation or expansion of established species. These methods use abiotic samples (e.g. water, soil) to detect the eDNA of imperiled and invasive species that are rare or cryptic. Environmental DNA originates from cellular material (via skin, saliva, excrement, etc.) that is left as the animal travels through the area. Non-invasive monitoring of aquatic habitats can assist in identifying colonized areas, as well as movement corridors and pathways of dispersal. Environmental DNA tools can also assist with short- or long-term monitoring to determine whether invasive species control or eradication efforts have been successful.
While the body of eDNA literature is rapidly growing, many of the basic assumptions and data interpretation tools have not adequately been addressed for the field. To account for the effects of imperfect detection manifested as false-negative observations, multi-level occupancy models were developed to correct for imperfect detection and false negative data (Hunter et al., 2015). A new analysis method and model was also developed for all types of analysis using state-of-the-art digital PCR (dPCR; Dorazio and Hunter 2015). Digital PCR allows for more sensitive and accurate eDNA detections than quantitative PCR (qPCR).
Methodology for Addressing the Issue: The high sensitivity of eDNA molecular assays has been shown to produce false-positive eDNA detections. False-positive detections incorrectly increase occupancy-based estimates of species occurrence and may be influencing conservation and management decisions in cases where this important source of bias is not considered (e.g. Asian Carp in the Great Lakes). To reduce false positive eDNA detections, we propose the development of a limit of detection (LOD) model using two-levels of external standards. Models will be developed for both qPCR and dPCR platforms.
Additionally, basic study design and sampling strategies are also lacking in eDNA surveys. An important aspect is deciding on sample number and density within various habitats and species densities. A statistical sampling model will be developed to more accurately detect eDNA when the species is at low densities and to help distinguish between inhabited vs. newly colonized areas and reduce false negative data.
The WARC’s droplet digital PCR platform can detect a single molecule of DNA from an environmental sample and enhances accuracy and precision compared to traditional methods. Use of this tool has been applied successfully to the detection of Burmese and African rock pythons from southern Florida water samples and will be used for this study. To detect individual species, three species-specific markers (two primers and a fluorescently labelled probe) were developed. Filtered water samples are then split into ~20,000 PCR droplets, each containing the markers and, if present, a copy of the target species’ DNA. The droplets illuminate fluorescently if DNA of the targeted species is detected, with the number of illuminated droplets corresponding to the number of DNA molecules in the sample. This information can be used to accurately estimate the likelihood of a species being present or absent in the environment.
Future Steps: Future research will focus on testing the models using field data in areas with unknown occurrence and density. Models incorporating additional environmental covariates or eDNA parameters could be developed to further optimize the use of eDNA to detect invasive and imperiled species.
Location of Study Area: Everglades, south Florida