Accounting for observation processes across multiple levels of uncertainty improves inference of species distributions and guides adaptive sampling of environmental DNA

Authors

Amy J. Davis, USDA National Wildlife Research CenterFollow
Kelly E. Williams, University of Washington
Nathan P. Snow, USDA APHIS Wildlife ServicesFollow
Kim M. Pepin, USDA APHIS Wildlife ServicesFollow
Antoinette J. Piaggio, USDA APHIS Wildlife ServicesFollow

ORCID IDs

0000-0002-4962-9753

0000-0002-4701-0746

Date of this Version

2018

Citation

Ecology and Evolution. 2018;8:10879–10892.

Comments

Open access

This article is a U.S. Government work and is in the public domain in the USA

DOI: 10.1002/ece3.4552

Abstract

Understanding factors that influence observation processes is critical for accurate assessment of underlying ecological processes. When indirect methods of detection, such as environmental DNA, are used to determine species presence, additional levels of uncertainty from observation processes need to be accounted for. We conducted a field trial to evaluate observation processes of a terrestrial invasive species (wild pigs‐ Sus scrofa) from DNA in water bodies. We used a multi‐scale occupancy analysis to estimate different levels of observation processes (detection, p): the probability DNA is available per sample (θ), the probability of capturing DNA per extraction (γ), and the probability of amplification per qPCR run (δ). We selected four sites for each of three water body types and collected 10 samples per water body during two months (September and October 2016) in central Texas. Our methodology can be used to guide sampling adaptively to minimize costs while improving inference of species distributions. Using a removal sampling approach was more efficient than pooling samples and was unbiased. Availability of DNA varied by month, was considerably higher when water pH was near neutral, and was higher in ephemeral streams relative to wildlife guzzlers and ponds. To achieve a cumulative detection probability >90% (including availability, capture, and amplification), future studies should collect 20 water samples per site, conduct at least two extractions per sample, and conduct five qPCR replicates per extraction. Accounting for multiple levels of uncertainty of observation processes improved estimation of the ecological processes and provided guidance for future sampling designs.