A Robust Optimisation Approach for the Placement of Wildland Fire Suppression Resources

This research devises an initial attack plan to combat wildfires that break out in wildland-urban interfaces. To be eligible for such a plan, the landscape must have been previously mapped and modelled concerning spatial and topographic data and fuel levels. Thus, when ignition occurs, one can predict the expected fire behaviour in terms of spread direction and rate of spread. With such information, decisions can be taken on where and when to position the suppression resources. This paper extends a recent contribution to this subject, generalising each node’s resource requirement and allowing a more precise modelling of non-homogeneous landscapes. Moreover, we treat the cases where the estimated number of resources may not be sufficient to deal with the fire intensity, which only becomes revealed at the fire scene. In such cases, additional resources may be needed to contain the fire effectively. This worst-case approach is modelled with the support of the robust optimisation paradigm. We propose a deterministic mathematical programming model, a robust optimisation counterpart, a preprocessing routine to reduce the network size, and a robust tabu search algorithm. We adapt instances from the literature, which are solved by a commercial optimiser and the robust tabu search. The proposed algorithm could optimally solve 94 out of 96 instances. Finally, we conducted a Monte Carlo simulation as part of a risk analysis assessment of the generated solutions