1. How to use random walks for modeling the movement of wild animals
- Author
-
Berthelot G, Saïd S, and Bansaye
- Subjects
0106 biological sciences ,0303 health sciences ,biology ,business.industry ,Movement (music) ,Computer science ,Foraging ,Perspective (graphical) ,Machine learning ,computer.software_genre ,Random walk ,010603 evolutionary biology ,01 natural sciences ,Roe deer ,03 medical and health sciences ,biology.animal ,Key (cryptography) ,Global Positioning System ,Feature (machine learning) ,Artificial intelligence ,business ,computer ,030304 developmental biology - Abstract
Animal movement has been identified as a key feature in understanding animal behavior, distribution and habitat use and foraging strategies among others. At the same time, technological improvements now allow for generating large datasets of high sampled GPS data over a long period of time. However, such datasets often remain unused or used only in part due to the lack of practical models that can directly infer the desired features from raw GPS locations and the complexity of existing approaches. Some of them being disputed for their lack of rational or biological justifications in their design. We propose a simple model of individual movement with explicit parameters based on essential features of animal behavior. The main thrust was to stick to empirical observations, rather than using black-box models that could possibly perform very well while providing little insight from an ecological perspective. We used a simple model, based on a two-dimensional biased and correlated random walk with three forces related to advection, attraction and immobility of the animal. These forces can be directly estimated using individual raw GPS data. The performance of the model is assessed through 5 statistics that describe the spatial features of animal movement. We demonstrate the approach by using GPS data of 5 roe deer with high frequency sampling. We show that combining the three forces significantly improves the model performance. We also found that the model’s parameters are not affected by the sampling rate of the GPS, suggesting that our model could also be used with low frequency sampling GPS devices. Additionally, the practical design of the model was verified for detecting spatial feature abnormalities (such as voids) and for providing estimates of density and abundance of wild animals. Our results show that a simple and practicable random walk template can account for the spatial complexity of wild animals. Integrating even more additional features of animal movement, such as individuals’ interactions or environmental repellents, could help to better understand the spatial behavior of wild animals.
- Published
- 2020
- Full Text
- View/download PDF