DEER MIGRATION STUDY
GEOG 6160 - Spatial Modeling and Geocomp
Identifying Utah’s mule deer migration routes has gained importance as urban areas expand further into their ranges. The routes traveled that link summer and winter ranges serve as a critical movement corridor for them. While the corridors themselves facilitate movement, the stop over sites within these corridors help animals accumulate enough energy reserves necessary to complete migration. The stop over sites should be prioritized and identified to more holistically conserve mule deer populations.
Spring migration will be simulated in a Net logo model using an agent-based model (ABM) approach, which allows for a simple rule structure to determine mule deer interactions with the environment. The ABM will simulate behaviors of an individual’s decision to stay and leave a stopover site based on the resources available and energy expenditure. The emergent behavior from the model will provide insight into how mule deer explore stop over sites in times of good vegetation and poor.
Background
Our model is set up for mule deer migrating from their winter range, along I-15 in Panguitch, to their summer range over Price Mountain and into the Markagunt plateau. Datasets of home ranges, elevation, vegetation, and roads were cleaned up in ArcGIS or ran through R for the initial data analysis. Spring migration and home ranges were derived from the GPS collar locations of deer and then computed through a Brownian Bridge Model in R Studio. Brownian Bridge models movement data from the start and end location and the elapsed time in between those points. It calculates the probability of an animal being in an area based on these locations and classifies areas as low, medium or high use.
Vegetation Suitability Table
Elevation is also an important influence on mule deer migration to summer ranges. Mule deer will follow high-quality foraging to get to their summer range, also known as spring “green-up” (Merkle et al. 2016). This tends to follow an increase in elevation on the landscape. The sampled mule deer collars from Panguitch also followed this trend in migration. Thus, higher elevations are reflected in the net logo model with higher suitability scores. The 30 m resolution DEM used for this model was downloaded from Utah AGRC. Agents in this model will find the most suitable path to their summer range by selecting the highest patch score of vegetation and elevation at each turn. Memory is also an important aspect of migration as deer rely on this knowledge to get to certain ranges (Merkle, et al. 2019). The summer and winter ranges from the Brownian Bridge model were used as a starting and finishing point for the agents.
Figure 1: Vegetation Suitability Index
NetLogo Model Parameters
Specific to our model, we have the to forage, to forage-move, and the to migrate-move parameters. The to forage parameter simply tells the deer that they gain energy from foraging, and that the amount of energy the gain is controlled by the deer-gain-from-food slider in the model window. As stated before, the to forage-move parameter is activated if the deer need to forage but the patch they are on does not have suitable resources. It tells the deer to move to a neighboring patch that has a suitability of greater than 0, where they can then gain some energy and continue on to the summer range. The to migrate-move parameter tells the deer to face the destination patch of the summer range and then creates a cone of patches that it can move to next. The cone restricts the deer’s movement to be in the general direction of the destination patch, but allows it some freedom to forage at the same time. This parameter takes into account the vegetation suitability and the elevation suitability when telling the deer where to move, and the deer will prefer the patch with better vegetation and higher elevation.
Our model contains some code to change the way the model looks as the deer move over time. The to paint-resource and to recolor-patches exist to give the user a little more context about the model after it is set up and is running. To paint-resource colors the vegetation patches with their level of suitability for foraging by the deer; green being the most suitable. While the model is running, to recolor-patches changes this color based on whether or not a patch has been foraged by the deer. If they have, then the green color that the patch may have had is erased and it is no longer suitable for foraging.
Model Interface & Experiments
In the first experiment, we start by putting the deer-gain-from-food slider at 8, simulating a good year of vegetation. Then the model is run a few times and we can look at the average foraging and migration times that we get. The time it takes for the deer to move from its winter range to its summer range in this scenario is looking to be around 10 days, with 121 patches used for migration and only 7 patches used for foraging. The deer is able to gain enough food whenever it stops to travel a great distance. Next, we will turn the slider down to 5 to simulate a mediocre year of vegetation. This creates quite a bit more variance between runs as it can take anywhere between 10-15 days to reach its summer range. The deer also has to forage significantly more, which would explain the difference in time required to migrate. However, the patches used for migration doesn’t increase very much, showing that it is simply spending more time foraging. In the last scenario with low vegetation quality, the model does not perform as well. The deer has to spend almost all of its time foraging, and the model shows it going around in circles more often than heading towards the destination. That being said, it does often get at least within the vicinity of its summer range after a month of migrating and has often foraged at least 200 patches.
In the second experiment, we run the model with all 42 of the tagged deer. The model itself doesn’t perform quite as consistently with so many more agents, but well enough that we can still get results from it. We also are not limiting the starting and stopping points in this model like we were in the first experiment, as every winter and summer range is visible here. With the deer-gain-from-food slider set to 8 simulating that strong vegetation year, it takes around 20 days for the most of the deer to reach their destination. Around 250 patches are used for foraging here too. Turning the slider down to 5 once again keeps a relatively similar migration time of 20 days, but the patches used for foraging more than triples. The deer in general have to spend a substantially larger amount of time foraging. However, when the slider is turned down to 3 to simulate a bad year of vegetation, the behavior of the deer changes significantly. Only about a quarter of them ever reach their intended summer range, while most of them are just searching everywhere for any food source they can find. Some even end up crossing the interstate highway in model, which doesn’t typically happen under the other conditions. This shows that in a bad year of vegetation, the deer would migrate far more unpredictably. Although in the real world, they likely would still attempt to head toward their final destination due to their natural tendencies.
Figure 2: NetLogo Model Interface
Conclusion
This strategy of modeling deer migration isn’t perfect, as there are tons of variables for the deer to consider on their journey that we simply can’t implement. That being said, I think we did a good job of capturing some of the major factors that contribute towards the way they migrate such as vegetation quality and elevation. The first experiment is good because it allows us to focus on one specific deer and focus on what might affect its route without interference from other agents. That being said, the second experiment is more realistic in some ways by incorporating all of the tagged deer into the model. In general, I think I would choose it over the first experiment unless I was specifically studying a single deer as I think it is better at capturing overall behavior.
Skills Learned / Used
Communication Skills
GIS Analysis
Model Building
Programming & Scripting