Contenido
Al igual que todos los modelos en ciencia computacional, los modelos de fuego necesitan encontrar un equilibrio entre fidelidad, disponibilidad de datos y ejecución rápida. Los modelos de incendios forestales abarcan una amplia gama de complejidad, desde principios simples de causa y efecto hasta los más complejos físicamente, que presentan un desafío de supercomputación difícil que no puede ser resuelto más rápido que en tiempo real.
Los modelos de incendio forestal se han desarrollado a partir de 1940, pero todavía quedan por resolver muchas cuestiones químicas y termodinámicas relacionadas con el comportamiento del fuego. Los científicos y sus modelos de incendios forestales desde 1940 hasta 2003 se enumeran en el artículo.[3] Los modelos se pueden clasificar en tres grupos: Empíricos, Semi-empíricos, y basados en principios de la Física.
Empirical models
Conceptual models based on experience and intuition you develop from past fires can be used to anticipate the future. Many semi-empirical fire spread equations, such as those published by the USDA Forest Service,[4] Forestry Canada,[5] Nobel, Bary, and Gill,[6] and Cheney, Gould, & Catchpole[7] for Australasian fuel complexes have been developed for rapid estimation of fundamental parameters of interest, such as fire spread speed, flame length, and fire line intensity of surface fires at a point for specific fuel complexes, assuming a wind representative of the site and slope of the terrain. Based on the work of Fons in 1946,[8] and Emmons in 1963,[9] the quasi-steady equilibrium spread rate calculated for a surface fire on flat terrain under windless conditions was calibrated using data from stick piles burned in a flame chamber/wind tunnel to represent other wind and slope conditions for the fuel complexes tested.
Two-dimensional fire growth models such as FARSITE")[10] and Prometheus,[11] the Canadian wildfire growth model designed to work on Canadian fuel complexes, have been developed that apply such semi-empirical and other relationships regarding land-crown transitions to estimate fire spread and other parameters along the surface. Certain assumptions must be made in models such as FARSITE and Prometheus to shape fire growth. For example, Prometheus and FARSITE use the Huygens principle of wave propagation. Richards in 1990 developed a set of equations that can be used to propagate (shape and direction) a fire front using an elliptical shape.[12] Although more sophisticated applications use a three-dimensional numerical weather prediction system to provide inputs such as wind speed to one of the fire growth models listed above, the input was passive and does not take into account fire feedback on wind. atmospheric and humidity.
Models based on physical principles and coupling with the atmosphere
Simplified two-dimensional fire spread models based on conservation laws that use radiation as the dominant mechanism of heat transfer and convection, accounting for the effect of wind and slope, lead to reaction-diffusion systems with partial differential equations.[13][14].
More complex physical models couple computational fluid dynamics models with a wildfire component and allow the fire to feed back into the atmosphere. These models include NCAR's Coupled Atmosphere-Fire-Wildland Environment (CAWFE) model developed in 2005,[15] NCAR's WRF-Fire and the University of Colorado Denver),[16] which combines the Weather Research and Forecasting Model with a group-level extended model, the University of Utah's Coupled Atmosphere Wildfire Large Eddy Simulation model developed in 2009,[17] FIRETEC") developed by the Los Alamos National Laboratory developed in,[18] the Fire Dynamics Simulator (WFDS) WUI (Wildland Urban Interface) in 2007,[19] and, to some extent, the two-dimensional FIRESTAR model.[20][21][22] These tools have different emphases and have been applied to better understand fundamental aspects of fire behavior, such as fuel inhomogeneities in fire behavior,[18] feedbacks between fire and the atmospheric environment as the basis of the universal form of fire,[23][24] and are beginning to be applied to the spread of fires in the forest-urban zone interface in house-to-house dispersion at a community scale.
The cost of added physical complexity is a corresponding increase in computational cost, so much so that a full three-dimensional explicit treatment of combustion in forest fuels by direct numerical simulation (DNS) at scales relevant to atmospheric modeling does not exist, is beyond the capability of existing supercomputers, and does not currently make sense to do so due to the limited ability of climate models with spatial resolution less than 1 km. Consequently, even these more complex models parameterize fire in some way, for example, Clark's work[25][26] uses equations developed by Rothermel for the USDA Forest Service[4] to calculate local fire spread rates using fire-modified local winds. And, although FIRETEC and WFDS have conservation equations for the reacting fuel and oxygen concentrations, the computational grid cannot be fine enough to resolve the reaction rate-limiting fuel-oxygen mixture, so approximations must be made regarding the subgrid-scale temperature distribution or the combustion reaction rates themselves. These models are also too small a scale to interact with a weather model, so the fluid movements use a computational fluid dynamics model confined in a much smaller box than the typical wildfire.