Methodological Development for Identifying Foraging Behaviors from GPS Data Among Artisanal Fishers in the Caribbean

Abstract

This project addresses questions about human foraging behavior in the ethnographic context of small-scale fishing-foraging in the Commonwealth of Dominica, an island in the Eastern Caribbean. The first goal of this project is to further develop a method of inferring foraging behaviors from GPS data by testing a recent partial sum approach—the CUMSUM method. The principle underpinning this method of research is that remotely gathered movement data can be accurately translated into meaningful data on foraging activities. GPS data produces movement tracks that are used to parse out changes in behavior, but segmentation of GPS tracks into different bouts of foraging activities is not straightforward. Previous research demonstrates that the CUMSUM method has benefits for detecting behavioral shifts and identifying patches of resources behaviorally, but it has seen limited testing across different foraging contexts. Developing this method has broad application across a range of disciplines, and one relevant utility is using CUMSUM segments to test foraging models. A second goal of this project is to demonstrate by testing a prediction of the marginal value theorem. The MVT explores generalized decision-making rules on patch residence time and was primarily developed in experimental settings with non-human animals. There are few tests of the MVT among human populations in naturalistic settings.

Research activities took place across three field sessions in the rural village of Desa Ikan, Dominica, among artisanal fisher-foragers. I tested the CUMSUM method with fishing data and found the method correctly identifies about 90% of patches with relatively small error rates. The strength of this approach is using both directly observed behavioral data to ground-truth simultaneously collected GPS data. I tested an aspect of the MVT using patch data from both observational data and CUMSUM segment data. Observational data supports the theoretical prediction that fishers spend more time in patches with higher travel costs, while support from CUMSUM model-generated data is equivocal.