River otter population monitoring in northeastern Pennsylvania using non-invasive genetic sampling and spatial capture-recapture models

Open Access
Forman, Nicholas Sean
Graduate Program:
Wildlife and Fisheries Science
Master of Science
Document Type:
Master Thesis
Date of Defense:
October 13, 2015
Committee Members:
  • William David Walter, Thesis Advisor
  • David Miller, Thesis Advisor
  • Cassandra Michaela Miller Butterworth, Thesis Advisor
  • River otter
  • non-invasive genetic sampling
  • spatial capture-recapture
  • population size
  • population density
  • Pennsylvania
River otter (Lontra canadensis) populations in Pennsylvania experienced a range reduction and subsequent expansion of the remnant population, as well as re-colonization of parts of the state through reintroduction efforts and expansion of neighboring populations. There are currently no estimates of population size or densities for river otter populations in Pennsylvania, and large-scale monitoring efforts are hampered by the elusive behavior of river otter. Non-invasive genetic sampling has been used to survey river otter populations, but given the river otter’s unique distribution across the landscape, estimation of population size and densities has been limited to linear habitats in river systems or along coastlines. Spatial capture-recapture models incorporate spatial information from captures into the estimation process, and estimates are more explicitly linked to the area in which observations occur. I analyzed the efficacy of non-invasive genetic sampling to identify individual river otter and I used spatial capture-recapture models to estimate river otter population size and density, and in northeastern Pennsylvania. I surveyed nine counties in northeastern Pennsylvania, opportunistically collecting samples from latrine sites on public and private land. Latrines were visited on three to four occasions at 6–14 day intervals, clearing latrines after each visit, in a capture-recapture framework. I amplified DNA extracted from the samples at ten microsatellite markers, to generate a genotype for each sample. I matched genotypes using program CERVUS to identify individuals. My first analysis compared amplification success rates and error rates for samples of different type and time of environmental exposure or freshness, and compared my amplification success rates to other studies. Previous studies on river otter had lower genotyping success rates than those for other otter species, and did not follow a common sampling protocol despite laboratory studies for the river otter and recommendations from field studies on other otter species. My amplification success rates were most comparable to those from studies on otter species conducted in the winter with samples collected in a storage buffer. I observed similar patterns of success rates as other studies for different sample types and samples classified for different categories related to lengths of environmental exposure, but had higher success rates for every category. Amplification error rates for the different sample types and environmental exposure categories were not reported in the literature, but I included them in the study as another measure of sample quality and to better inform future studies. The importance of comparing success rates and error rates is to better inform future studies on the preferred sampling protocol, and give measures for the amount of effort necessary for studies looking to use non-invasive genetic sampling to identify individual river otter for population analyses. To estimate population size and density in spatial capture-recapture models, I compiled spatial encounter histories given the location and occasion of collection of each sample assigned to an individual. I also used full likelihood models in program MARK to test for differences in capture and recapture probabilities. I reported the first density estimates for a river otter population in northeastern Pennsylvania (2.1 otter/100 km2, 1.4–5.0 otter/100 km2 95% Asymptotic Wald-type CI). The estimates of capture and recapture probabilities in the MARK model with those parameters estimated separately indicated that capture and recapture probabilities were not different, but that the probability of capturing an individual did vary by occasion. I observed a difference in density estimates for my SCR and MARK models. I would recommend using SCR models because of the spatial justification for density estimates, and the ability to include landscape covariates to build more informed models, which may prove to be useful for river otter given their unique space use. Future studies conducting non-invasive genetic sampling for river otter should conduct their studies in winter and use a storage buffer for samples. Sample type and length of environmental exposure should be considered when considering the amount of sampling effort to derive a genotype for identification of individual otter. NGS and SCR can be used to generate reliable population or density estimates, but as I documented from my MARK estimates of capture probability, numerous sampling occasions are desirable because of the variation in capture probability between occasions. Spatial capture-recapture models are preferable for river otter in Pennsylvania because the area for which density is being estimated is directly tied into the model, which is ideal given the diversity of linear and non-linear habitat types in northeastern Pennsylvania.