Networks can also be a useful way of modeling the spread of a wildfire. A landscape can be divided into cells and the midpoints of the cells form the nodes of the network. Arcs connect nodes to their neighbors and the weight on an arc represents the time a fire takes to travel between nodes, based on landscape and weather characteristics. Thus, we have applied network interdiction to the problem of allocating resources to a landscape prior to a fire in order to prevent fire outbreak.
Wildfire managers use initial attack (IA) to control wildfires before they grow large and become difficult to suppress. Although the majority of wildfire incidents are contained by IA, the small percentage of fires that escape IA causes most of the damage. Therefore, planning a successful IA is very important. In this article, we study the vulnerability of IA in wildfire suppression using an attacker-defender Stackelberg model. The attacker’s objective is to coordinate the simultaneous ignition of fires at various points in a landscape to maximize the number of fires that cannot be contained by IA. The defender’s objective is to optimally dispatch suppression resources from multiple fire stations located across the landscape to minimize the number of wildfires not contained by IA. We use a decomposition algorithm to solve the model and apply the model on a test case landscape. We also investigate the impact of delay in the response, the fire growth rate, the amount of suppression resources, and the locations of fire stations on the success of IA.
We study a security problem in which an adversary seeks to attack a landscape by setting a wildfire in a strategic location, whereas wildfire managers wish to mitigate the damage of the attack by implementing a fuel treatment in the landscape. We model the problem as a min–max Stackelberg game with the goal of identifying an optimal fuel treatment plan that minimizes the impact of a pyro-terror attack. As the adversary’s problem is discrete, we use a decomposition algorithm suitable for integer bi-level programs. We test our model on three test landscape cases located in the Western United States. The results indicate that fuel treatment can effectively mitigate the effects of an attack: implementing fuel treatment on 2, 5, and 10% of the landscape, on average, reduces the damage caused by a pyro-terror attack by 14, 27, and 43%, respectively. The resulting fuel treatment plan is also effective in mitigating natural wildfires with randomly placed ignition points. The pyro-terrorism mitigation problem studied in this article is equivalent to the b-interdiction-covering problem where the intermediate nodes are subject to interdiction. It can also be interpreted as the problem of identifying the b-most-vital nodes in a one-to-all shortest path problem.
In this research, we study the vulnerability of landscapes to wildfires based on the impact of the worst-case scenario ignition locations. Using this scenario, we model wildfires that cause the largest damage to a landscape over a given time horizon. The landscape is modeled as a grid network, and the spread of wildfire is modeled using the minimum travel time model. To assess the impact of a wildfire in the worst-case scenario, we develop a mathematical programming model to optimally locate the ignition points so that the resulting wildfire results in the maximum damage. We compare the impacts of the worst-case wildfires (with optimally located ignition points) with the impacts of wildfires with randomly located ignition points on three landscape test cases clipped out from three national forests located in the western U.S. Our results indicate that the worst-case wildfires, on average, have more than twice the impact on landscapes than wildfires with randomly located ignition points.