Graduation Year

2022

Document Type

Dissertation

Degree

Ph.D.

Degree Name

Doctor of Philosophy (Ph.D.)

Degree Granting Department

Geography, Environment and Planning

Major Professor

Joni Firat, Ph.D.

Committee Member

Lori Collins, Ph.D.

Committee Member

Steven Reader, Ph.D.

Committee Member

Philip Van Beynen, Ph.D.

Keywords

Agent-based Model, Dynamic Interaction, GIScience, Mallard, Space-time

Abstract

Dynamic interaction (DI) describes the synchronous interaction of one or more individual animals over time. Current methods to quantify DI from tracking data are lacking. They typically require synchronized tracking data which limits their applicability to most studies. They also often do not have built-in tests of significance to evaluate interaction, and if so, rely on random walk models that lack biological realism. This research aimed to develop and evaluate a new way of measuring animal interaction that overcomes these limitations. The goal was to create a statistical method for quantifying and evaluating interaction that is relevant across computational, behavioural, ecological, and geographical information sciences and apply that method to a selected species, the Mallard (Anas platyrhynchos). A detailed activity budget on Mallards was also generated to bridge a gap in urban Mallard studies.

The new method presented in this dissertation, referred to as Intersection of Probabilistic Space-time Prisms with Agent-based Models or I-P-STPABM, combines probabilistic space-time prisms (P-STPs) with agent-based models (ABMs) to quantify DI and evaluate if interactions occur more, less, or as expected at random. The I-P-STPABM approach uses a three-part procedure. First, P-STPs are used to estimate the observed number of times the animals interacted over a tracking duration. Second, a species-specific ABM is used as a null model for simulating random animal movements in the same environment, with repeated runs yielding the expected numbers of interactions for the tracking duration. Third, the observed interactions are compared to the confidence interval for the expected interactions to determine if DI occurred more, less, or as expected at random.

The I-P-STPABM method was developed and demonstrated using an existing Muscovy Duck (Cairina moschata) ABM and generating 1-meter resolution P-STPs with tracking data on 4 dyads sampled every 2 minutes for 1 hour at University of South Florida ponds in Tampa, Florida, with each individual tracked 1-minute apart. The model operated using habitat-transition frequencies, distances traveled, and habitat-use data to inform agent-decisions in a simulated Campus environment. Start locations, tracking durations, and the sampling interval of observed behaviours were incorporated. Simulated interaction was shown as statistically random with 1,000 runs and predicted differing levels of interaction significance. For the method’s application, 201 Mallards were sampled every 15 seconds for 20 minutes using an ethogram and were analyzed. Both sexes showed significant variation in behaviour by time of day. Thirteen Mallard dyads were tracked with same sampling scheme as Muscovies to generate I-P-STPs. A new ABM was creating using the collected behavioural data to evaluate observed data. Combining methods predicted less than expected and as expected levels of interaction for respective Mallard dyads.

This study concludes that the new I-P-STPABM method is a viable way to quantify and evaluate animal interaction and overcomes former tracking data restrictions. This was also one of the first case studies on urban Mallard behaviours and detailed analysis of their interaction. Our ABM offers a basic framework using transition frequency, distance, and habitat-use data that could be used for other species. However, creating an ABM takes a lot of effort due to programming, testing, focal observation, habitat classification, and choosing frameworks to best simulate movement. Data collection was somewhat impeded by pandemic restrictions. An increased sample size of behaviours could have increased robusticity. A specified velocity input and cell size for prisms takes careful consideration with currently no established way to pinpoint the best values. Prisms are processing intensive and require more storage space. The methods used in this dissertation requires a great deal of planning, dedication, and data management. Overall, I-P-STPABM provides more insightful look into the interaction of animals and is a step forward into exploring new ways to quantify dynamic interaction. Future research might test its application on different species, sampling schemes, locations, and environments.

Download

Share

COinS