David Hunter
- @davidhunter
- | he/him
University of Colorado Boulder
Hello All,
I'm wondering if anyone has any advice on creating Individual or Agent-Based Modelling simulations for large animals like deer? What is the typical approach with this kind of animal?
Are there established rules for their behaviour you program into the system/software? Do you use recorded data and have that determine rules? Is it a case of the animal agent evaluating the different land-use types it could move to next, or is it weighing up multiple factors like topography, temperature, fear of certain things (roads, people). Do you give the animal agent attributes like food needs or memory? Do you use machine learning instead?
I have looked at a small number of papers and spoken to a couple of people and heard different approaches. I'm from a human-computer interaction background, not ecology, so I don't have a good sense of what would be the "right" approach.
For context, I have several years of data of tracked deer that include latitude and longitude, timestamp, external temperature, with measurements taken a few times per day.
Many thanks, David
19 May 2025 12:18am
Hi David,
In short, there is no "right" answer. That's why you have heard different ideas from different people. Basically, you need to start with the research question, and then figure out how best to answer it using a model calibrated and validated with your data. If you have high-level movement data for the deer, you may not need to understand the driving forces and decision making methods the deer used in order for you to build a useful model. But whether or not this is the case can only be determined by reference to the question(s) you wish to use your simulation to answer!
Kinds regards, Alan
Vivian Hawkinson
University of Washington
22 May 2025 2:31pm
Hi David,
I would reiterate what Alan has said here - there is no concrete "right" or "wrong" way! It can be really helpful if you have animal movement data, particularly if it's from the area you'd like the model to emulate, but not a requirement.
Based on the question you're asking, you could analyze the deer data you have to determine how different variables influence their movement, which you could then use to set some of the rules for your model. I would note that deer and elk can move across a landscape quite differently, so if you have a species mismatch this might not be the best approach. I have taken this approach for some models, but I have also built others where the rules were entirely based on information pulled from an in-depth literature review. All of my models deal with large mammals (deer, elk, cattle, wolves, etc.).
One of the biggest things you'll have to decide is which/how many rules and parameters you want to include in the model, because you can end up with a long list rather quickly, making the model complex and heavy. How you set up your rules and parameters will also depend on the scale, of your model, of course. Behaviors and movement decisions appear differently at an hourly, daily, weekly scale across meters, kilometers, etc. so you'll want to account for that.
Good luck!
Vivian
Alan Dorin