Search

Progressing climate changes have effects on forests. As a consequence, more and more woods will be endangered by forest fires. This is the motivation to get the needed information about forest fire propagation in order to enhance prevention concerning forest protection and firefighting. To model forest fire spreading we apply a physical model which considers the chemical and physical processes like combustion and heat and mass transfer mechanisms. Following this approach we are led to a time-dependent non-linear convection-diffusion-reaction-problem. Based on this framework, we present a numerical solution by a collocation method and a time-stepping scheme. Afterwards, we give an approach for stabilization which is needed in the numerical treatment of the underlying equations. Finally, we present some numerical simulations of forest fire spreading.

Markovian Agent Models are a rather new modeling technique to deal with complex systems composed by a multitude of interacting entities, whose spatial location is also relevant in determining their interaction. An example of application to the study of outdoor fire propagation is provided. The dynamic of this phenomenon strongly depends both on the type of materials being incinerated and on the wind direction and intensity. Therefore, the ability of the model to correctly reproduce the fire propagation is closely related to the spatial dependent interactions among agents. Moreover, the model is flexible enough to analyse scenarios in variable environmental conditions, such as wind direction and intensity, and in the presence of fire-barriers which prevent flames from propagating.

Every year forest fires provoke enormous looses from the ecological, economical and social point of view, by burning millions of hectares and killing several people. Nowadays, several forest fire simulators are used for helping in forest fire prevention and mitigation. Most of these simulators are based on Rothermel model. However, in most cases it is not possible to know the exact real-time values of model input parameters. This input data uncertainty causes predictions that are far from the real fire propagation. For this reason, we propose a fire propagation prediction system based on two stages: a calibration stage and a prediction stage. The calibration stage bases on the observed real fire evolution, so that the behavior of the fire is assimilated by the system at run-time matching the Dynamic Data Driven Application Systems (DDDAS) paradigm.

Discovery of evolving regions in large graphs is an important issue because it is the basis of many applications such as spam websites detection in the Web, community lifecycle exploration in social networks, and so forth. In this paper, we aim to study a new problem, which explores the evolution process between two historic snapshots of an evolving graph. A formal definition of this problem is presented. The evolution process is simulated as a fire propagation scenario based on the Forest Fire Model (FFM) [17]. We propose two efficient solutions to tackle the issue which are grounded on the probabilistic guarantee. The experimental results show that our solutions are efficient with regard to the performance and effective on the well fitness of the major characteristics of evolving graphs.

Human activities namely forest management interact with natural conditions to determine fire occurrence and fire impact in a complex form. Climate change tends to facilitate even larger and more dangerous fires. The wildland urban interface is an emerging problem derived from the expansion of urban areas associated to high risk fire conditions. Extreme fire behaviour in the form of eruptive fires, crown fires and spot fires are associated to great damage to the environment and to loss of human lives. Some cases of fatal accidents that occurred recently in Europe are presented to illustrate the proposed concepts.

There is a need to understand how people and environment react in a fire building emergency. Sometimes in the wayfinding process decision errors may occur mainly based on topological errors of the signage. A situation is critical if a decision about which path to take cannot be made with certainty, especially in a crisis situation. An agent-based simulation of human’s behavior in escaping from the fire with due attention to the building’s signage and dynamic nature of fire propagation affecting the wayfinding task is outlined in this paper. The hypothesis of the paper is that successful navigation is possible if the agent is able to make the correct decision through well-defined cues in critical cases, so the design of the building signage is evaluated through the agent-based simulation. Construction of mental representations of spatial environment and exploring models in the agent-based simulation have been proposed and a computational model successfully tested in an indoor complex hospital environment in different situations and the evacuation time from the building is computed. The most appropriate signage design resulted in the shortest evacuation time in various situations.

A Dynamic Data Driven Application System (DDDAS) was created to study interaction between fire and agent models during a fire evacuation. The analysis from that research can be used to validate proposed ideas in evacuation and building designs to ensure safety of buildings given various agent behaviors. Two separate models were used to simulate the components of the emergency situation: fire and agent. The independent models were able to run using data computed by the other interacting models, allowing careful examination of real-time interactions in a situation. Through study of the interactions, a better understanding is gained of how individual variables such as exit position and width affect the evacuation process and escape rate in the given scenario. Exits can be relocated and changed to quickly assess the effect on the model. The results can be used for improving building design and regulations as well as training first responders.

Classical prediction fire schemes do not match the real fire propagation, basically, because of the complexity of the physical models involved, the need for a great amount of computation and the difficulties of providing accurate input parameters. We describe an enhanced prediction scheme, which uses recent fire history and optimization techniques to predict near future propagation. The proposed method takes advantage of the computational power offered by distributed systems to accelerate the optimization process at real time.

Fire propagation simulation is seen as a challenging problem in the area of simulation, due to the complexity of the physical models involved, the need for a great amount of computation and the difficulties of providing accurate input parameters. Input parameters appear as one of the major sources of deviation between predicted results and real-fire propagation. Evolutionary algorithms have been used to optimize the input parameters. However, such optimization techniques must be carried out during real-time operation and, therefore, certain methods must be applied to accelerate the optimization process. These methods take advantage of the computational power offered by distributed systems.

Forest fire propagation modeling has typically been included within the category of grand challenging problems due to its complexity and to the range of disciplines that it involves. The high degree of uncertainty in the input parameters required by the fire models/simulators can be approached by applying optimization techniques, which, typically involve a large number of simulation executions, all of which usually require considerable time. Distributed computing systems (or metacomputers) suggest themselves as a perfect platform to addressing this problem. We focus on the tuning process for the ISStest fire simulator input parameters on a distributed computer environment managed by Condor.

This paper summaries experiences designing and developing a parallel/distributed simulator for existing models of forest fire propagation using PVM. The methodology used to parallelise the application is described. Results obtained for the parallelisation on different platforms are presented: clusters of SUN workstations and PCs, and one parallel machine (Parsytec CC) with different internal node interconnections. The studies present several measurements of performance & scalability of the proposed solution for this forest fire application.