Complexity USRG – EPSRC Summer Bursaries 2014
June 24, 2014
by Peter Prince
Peter Prince: ‘An agent-based model to explore time allocation towards foraging in response to rainfall and temperature conditions in large African herbivores’
I am a computer science student moving from my third year to my final Masters year in September. I have an existing interest in ecological modelling, previous creating an agent-based model of brown bears, attempting to minimise fatal interactions between humans and bears in American national parks. My interest in this field comes from wanting to take computer science knowledge I’ve gained and use it to help solve problems in other fields, such as ecological problems.
The plan for the research project is based on GPS tracking data on a number of large herbivores in Africa. Over a number of years a large dataset has been amassed, documenting the movement of various species such as antelope, wildebeest and buffalo across areas of Africa. By exploring the work done in previous papers on similar data, the idea of exploring the behaviour of certain herbivores in response to drought and extreme temperatures was decided upon.
For some species, effects such as high rainfall have been shown to cause population declines due to increased predation when foraging at night. However, certain species experience declines after periods of high rainfall, with long grass growing and providing more cover for predators during the day. It is effects such as these which our project aims to explore, optimising time allocation throughout the day and night.
The datasets do not record individual activities the tracked animals are performing, however research exists looking into translating movement rates into likely activity states. Taking movement rates, locations and previous behaviours, estimates can be made about the actions of each animal at a specific time step. Using these techniques our project team plans to build a model based on the activity states in response to recorded conditions.
The model itself is planned to be an agent-based model, taking into account the previous behaviours of each individual agent as well as its herd to govern future behaviour. Such models are extremely difficult to mathematically simulate, and make a lot of assumptions to minimise the complexity of the model. Agent-based models allow for more complex individual behaviour, which is a perfect fit for the problem being investigated.
In my fourth and final year of my degree I will be taking part in a year-long group design project, based on modelling and artificial intelligence. I feel that my work done as a part of this research project will help to equip me for the challenges of this style of research using artificial intelligence techniques.