Methods for integrating models of wildfire progression with disaster response.

摘要

Challenges associated with the allocation of limited resources to mitigate the impact of natural disasters inspire fundamentally new theoretical questions for decision making in coupled human and natural systems. Two of these questions are: (1) how can complex, disparate sources of information be combined to make optimal response decisions, and (2) what are the issues that arise when a disaster is dynamically evolving with the response effort? We use wildfires in California as a case study to address these questions in a series of scenarios.;To examine the synthesis of disparate data sources, we develop a modeling framework that projects long-term costs based on the combination of a dynamic fire spread model, an economic cost model, and population data. We are motivated by an analysis of California wildfire data, which exhibits high variability (power laws) in event size statistics and demonstrates that fire size is not an accurate proxy for wildfire damage. Our study uses model generated fire catalogs to estimate the effect of suppression strategies on fire size, and our cost function incorporates both suppression costs and loss of assets. This yields statistical estimates of the long-term economic impact of fire response policies. Tradeoffs between resource costs and assets at risk determine the optimal response for an individual fire. We also compare the costs of different policies for division of limited resources between multiple fires using scenarios motivated by the 2003/2007 California wildfire seasons.;In the second model, we address the role of dynamic coevolution of a natural disaster with the response effort. Wildfires are one of several types of disaster phenomena, including oil spills and disease epidemics, in which the disaster evolves on the same timescale as the response effort, and delays in response can lead to increased disaster severity and greater demand for resources. We introduce a minimal stochastic process to represent wildfire progression that nonetheless accurately captures the power law statistical distribution of fire sizes observed in nature. We then couple this model for fire spread to a series of response models that isolate fundamental tradeoffs in strength and timing of response, as well as in division of limited resources across multiple competing suppression efforts. Using this framework, we compute optimal strategies for decision making scenarios that arise in fire response policy.