An Agent-Based Model (ABM) is a type of simulation that can employ spatial data to model complex behaviors of individual "agents" (organisms, people, vehicles, etc.). ABMs can account for the adaptive behavior of agents, allowing for the study of emergent behavior that could arise in a dynamic system. For example, an ABM could be employed to model worker behavior in a warehouse. The results of the model could inform the placement of products to minimize traffic or time to fill an order and improve operations.

For this project, an ABM was designed in the NetLogo software to simulate the propagation of wildfire at the edge of an agricultural zone in Central Washington, United States. This area was selected due to dry drought conditions at the time. An existing model simulating wildfire in a hypothetical forest was heavily modified to accomplish this. The area of interest was selected from satellite imagery in ArcGIS Pro and exported to NetLogo to establish a new spatial context.

Through trial and error, the wildfire model was modified to select one or more pixels categorized as "dry fuel" to initiate a fire. As the fire spread, it would check adjacent pixels for their land cover type: dry fuel, wet fuel, built, or water. A probability of fire spread was assigned to each of these land cover types, allowing for every iteration of the model to behave differently, highlighting the unpredictability of wildfires. Further, an option was added to simulate rainfall, which gave each dry fuel cell a probability of being converted to wet fuel, slightly inhibiting wildfire propagation.

While this project was relatively rudimentary, it provided an exciting challenge in moving data between unrelated software environments and in carefully considering how an agent could interact with and respond to its environment. The map above shows the model in progress with two wildfires. Note how the left-hand fire does not cross over the body of water.