Exploration of methods for analyses of resource selection using location-based data

Open Access
Author:
Carrollo, Emily Megan
Graduate Program:
Wildlife and Fisheries Science
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
March 03, 2016
Committee Members:
  • William David Walter, Thesis Advisor
Keywords:
  • Captive cervid facility
  • chronic wasting disease
  • logistic regression
  • mule deer
  • negative binomial
  • resource selection
  • sunflower
Abstract:
Resource selection functions (RSFs) are commonly used by wildlife researchers to identify resources necessary for the presence or absence of a species in a given area. Resource selection functions can assist decision-makers in protecting necessary resources that can mean the difference between increasing populations or extinction. Hence, it is important that researchers use appropriate RSFs for differing types of data and study designs. The misuse of RSFs is not uncommon, however, and can have negative effects on experiment wise error (i.e., increases in Type 1 error) or give misleading results that lead to poor management plans. I identified three commonly used RSFs (logistic regression, negative binomial, and discrete choice), and more closely examined two (logistic regression and negative binomial) via the use of Global Positioning System (GPS) and count data. I used logistic regression to assess the selection of agriculture by mule deer in southwest Colorado, and in particular their selection of sunflowers. I then used negative binomial to identify the location characteristics of Pennsylvania captive cervid facilities to address the potential risk that the spread of chronic wasting disease (CWD) poses to wild deer via captive cervid facilities and vice versa. Resource selection can be used to identify human-related resources that are selected by wildlife and potentially damaged. In southwest Colorado, mule deer (Odocoileus hemionus) are a popular game animal, and are economically and environmentally important to the local area. Identification of specific crops and conditions that may increase crop damage can provide cost savings for farmers and wildlife managers by protecting crops most likely to be damaged by mule deer. I used GPS data and logistic regression to identify selection of specific crops during 8 seasonal/diel periods. Forest, alfalfa and sunflower had the greatest influence on mule deer resource selection. Identification of these specific crops will help wildlife managers implement crop depredation techniques in areas where they will be most beneficial. Resource selection functions can also be used to identify human selection of resources associated with practices such as placement of captive cervid facilities. Captive cervid facilities in Pennsylvania have a large economic value. They also pose a risk to wild cervid populations via the spread of CWD. I used negative binomial regression to assess the resource selection of captive cervid facility owners in Pennsylvania, and to identify areas where wild deer may come in contact with captive deer and increase the potential for CWD transmission. Agriculture had the most support for areas selected by captive facility owners, and several other covariates had varying effects. Resources that were identified as influential factors to selection by captive facility owners were then used to create a predictive surface of areas most as risk for CWD transmission between captive and wild cervids. Resource selection functions are a versatile tool in wildlife management. When used appropriately they can provide results that help wildlife managers make the most effective and efficient management plans for resources and animals. There are many places where a researcher can make inappropriate decisions regarding study design and RSF selection and thus produce poor results. In my thesis I discuss the many considerations a researcher must account for when choosing a RSF, and I also give two examples of appropriate RSF analysis using logistic regression and negative binomial.