Investigating distribution and abundance of mesocarnivores in western Maryland, USA: a case study of coyote and fisher
Date
2021
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
University of Delaware
Abstract
Mesocarnivore species can influence the structure and dynamics of ecosystems with their presence and distribution. These species commonly occur in the same environment and share similar resources that lead to intraguild interactions. These types of interactions and variables that have an impact on their occurrence and abundance have not been sufficiently studied for most mesocarnivores. Therefore, I focused on two native mesocarnivore species, coyote (Canis latrans) and fisher (Pekania pennanti), occupying much of western Maryland. ☐ I planned to investigate the distribution and abundance of these mesocarnivore species in western Maryland to aid in understanding the current population status and potential interactions. Moreover, I expected that my findings will inform wildlife conservation and management practices conducted by the Maryland Department of Natural Resources (MDNR). In this context, my objectives were 1) to model the occupancy of coyote and fisher using environmental and landscape variables, 2) to determine environmental and landscape variables that correlate with the co-occurrence of fisher and bobcat, and 3) to estimate the abundance and density of coyote and fisher using two non-invasive survey methods in western Maryland. ☐ I used remote cameras combined with baited hair snares to survey 120 grid cells in total on 3 study areas, Potomac-Garrett State Forest, Savage River State Forest, and Green Ridge State Forest, in western Maryland during January-March 2019 and 2020. ☐ First, I built single-species occupancy models by using detection/non-detection data obtained by camera traps and habitat variables. No variables predicted coyote occupancy due to insufficient coyote detections. Fisher detection was predicted by the model including the combination of lure type and week-year interaction (AICwt= 0.87). The model including ruggedness index was the top model for probability of occupancy (AICwt= 0.27), indicating that fisher occurred in less rugged areas. This negative association may have been due to the area’s inaccessibility, more energy requirement to catch prey, the presence of other competitors, and the high risk of predation in rugged areas. The predicted fisher occupancy map derived from model-averaged outputs indicated fisher could occupy most of Garrett and Allegany counties. ☐ Second, I constructed multi-species occupancy models and informed models by incorporating detection and occupancy covariates derived from single-species occupancy models. I used detection/non-detection data recorded by camera traps and habitat variables calculated within a 1.17 km radius of each camera site. I counted both fisher and bobcat in 22 grid cells in 2019 and 10 grid cells in 2020. Given the multi-species occupancy model results, no models ranked above the null model (AICwt= 0.212). Therefore, I could not find any variables to predict the probability of fisher and bobcat coexistence, most likely due to insufficient pair detections, low covariate variation, or altered temporal behavior as a function of intraguild interaction to reduce mortality risk. ☐ Third, I estimated abundance of coyote and fisher by using Royle-Nichols models that estimate variation in abundance with the effect of environmental and landscape variables. I used detection/non-detection data obtained by camera traps to build abundance models. No variables predicted coyote abundance due to the sparseness of coyote detections. The ruggedness index model was the top model for fisher abundance (AICwt= 0.45), suggesting that fisher had lower abundance in more rugged areas. Competition, predation, or ability to catch prey might lead to this association. Further, I derived coyote detections from the results of DNA analysis on scat samples collected during the 2019 summer season to estimate density in the state using spatially explicit capture-recapture (SECR) models. The null model was the top model (AICwt= 0.91), indicating that coyote density was constant across study areas. Julian date seemed to affect detection probability of coyote at its activity center (AICwt= 0.39). ☐ Overall, my results provided an insight for assessing the population status of species and an important step for future investigations. Based on my findings, I suggested that more comprehensive and long-term studies should be conducted to investigate these elusive species by increasing sampling units and survey season length.
Description
Keywords
Abundance, Co-occurrence, Coyote, Density, Fisher, Occupancy, Maryland