A joint Kaplan-Meier known-fate and Brownie tag-recovery model to estimate harvest and survival rates
Open Access
- Author:
- Buderman, Frances Elizabeth
- Graduate Program:
- Wildlife and Fisheries Science
- Degree:
- Master of Science
- Document Type:
- Master Thesis
- Date of Defense:
- August 21, 2012
- Committee Members:
- Duane R Diefenbach, Thesis Advisor/Co-Advisor
- Keywords:
- white-tailed deer
deer harvest
deer survival
deer abundance
tag-recovery
known-fate
Kaplan-Meier
wildlife
estimation - Abstract:
- Incorporating multiple data sources into mark-recapture estimation models may eliminate the need for some model assumptions and can improve precision of parameter estimates. For some estimators, incorporating multiple data types into the modeling procedure can help correct for violations of model assumptions. The Brownie tag-recovery model assumes that all individuals tagged in each cohort survive to the start of the first hunting season after tagging. However, it may not be efficient or possible to tag individuals just prior to the hunting season. For example, white-tailed deer (Odocoileus virginianus) in Pennsylvania are trapped and tagged in spring and not harvested until the following fall. If individuals die between tagging and the hunting season, fewer individuals are actually available to be harvested than the model assumes and harvest rates will be underestimated. I developed a joint model that uses auxiliary known-fate data, provided by radio-transmitters, to inform a Brownie tag-recovery model to estimate harvest rates for game species; using auxiliary information accounts for tagging-to-harvest mortality. I performed simulations using 3 model species with varying tagging-to-harvest survival rates, annual survival rates, and harvest rates. I compared accuracy and precision of model species’ harvest estimates between the joint model and a standard Brownie tag-recovery estimator. The level of precision that the joint model obtained depended upon the biology of the study species. With some species, the acceptable level of precision (CV = 10-15%) for harvest estimates was reached given certain thresholds for allocations of reward tags and radio-transmitters, but in other species, this level of precision was nearly impossible to obtain. In all cases, releasing 20 radio-transmitters and 100-150 reward-tags allowed the joint model to outperform, based on a metric combining accuracy and precision, the standard Brownie tag-recovery estimator. I then applied the joint model to Pennsylvania white-tailed deer and estimated harvest and annual survival rates for deer in 4 Wildlife Management Units (WMUs). I used a multi-step approach, in which I first selected the most parsimonious model structure for tagging-to-harvest survival rates. I then incorporated this model structure into the Brownie tag-recovery estimator. This model allowed me to test whether radio-transmittered deer are differentially selected by hunters for harvest compared to non-radio-transmittered deer. Radio-transmittered individuals often are used to infer the population level harvest rate, however, if fates of these individuals differ from unmarked individual, harvest rates may be not reflect the population level harvest rate. I failed to detect a transmitter effect in either male or female white-tailed deer. Harvest rate precision was comparable to what I estimated based on my computer simulations. Also I developed a model that allowed me to estimate harvest rates for the discrete periods of hunting that occur during white-tailed deer season in Pennsylvania. Although there were insufficient data to do this for each of the WMUs, I was able to obtain an estimate for each season by pooling across WMUs. Finally, I used an Integrated Population Model (IPM) to investigate the repercussions of shortening the antlerless firearm season to the last 7-days of a 12-day season. Although I had hoped to test this empirically, the Board of Commissioners for Pennsylvania Game Commission did not enact regulations to allow me to implement the study design. Consequently I used IPMs to utilize multiple sources of data and a biologically reasonable population model to calculate pre- and post-hunt abundance. It also allowed us to determine the harvest rate that would have been required to stabilize the population at 2007 levels, while implementing a 7-day antlerless season over 4 hunting seasons. I used hunter efficiency to back calculate from the necessary antlerless harvest to the number of licenses that the PGC would have needed to sell each year. In all cases, there needed to be a substantial, but not unreasonable, increase in the number of harvested antlerless deer and licenses sold.