IMPLEMENTATION OF WILDLAND FIRE MODEL COMPONENT

IN THE WEATHER RESEARCH AND FORECASTING (WRF) MODEL

Janice Coen (MMM/RAP) and Ned Patton (MMM)

1. Summary

The goal of this work is to implement NCAR's wildland fire model, currently an extension of the Clark-Hall atmospheric model, as an application within the Weather Research and Forecasting (WRF) Model, and, when completed, to offer this capability to the WRF community of users. The benefits would be 1) the availability of a coupled atmosphere-fire model in a well-supported community model for eventual community research, 2) a stable framework into which other scientific components related to the Wildland Fire Research and Development Collaboratorycan be built, including an end-to-end fuels - combustion chemistry - emissions module, greatly increasing the opportunity for university colleagues investigating regional haze, photochemical production of ozone, and the effects of smoke on cloud precipitation efficiency to build upon it, 3) the eventual operational capability required by operational applications, such as those for smoke management, prescribed burns, and wildfire nowcasting and mitigation, 4) a test of WRF with strong forcing at small scales, 5) reduction in the redundancy of effort developing needed capabilities within the Clark-Hall model (land surface schemes, portability across computing platforms) that would come with the evolution of a community model.

2. Background

2.1 NCAR fire model

NCAR's coupled atmosphere-wildland fire simulation model has been developed to represent the complex interactions between fires and local winds. This system is composed of two parts: the Clark-Hall atmospheric numerical model to which a wildfire simulation model has been added. In previous studies, (Clark et al. 1996a, b; Coen and Clark, 2001) this system was used to demonstrate the dynamic causes of the bowing of fire lines, the formation of vortices within the fireline, and sudden outbursts of small fingers of flame from the fireline.

The innermost, finest-resolution atmospheric model domain (if nesting is being used to refine the resolution) is the one that interacts directly with the fire model. The atmosphere and fire are fully coupled in that evolving modeled atmospheric information is used to drive the propagation of the fire line, and the sensible and latent heat from the fire model is released into the modeled atmosphere, greatly changing the atmospheric motions, creating strong convective updrafts, convergence near the surface, and strong near surface winds that, in turn, determine the spread rate and direction of the fire. More precisely, at each time step of the inner model domain, atmospheric wind velocities from a specified height in the atmospheric model (often the fuel or canopy height, 1-10 m) are passed into the fire model, where they are used to calculate fire spread rate at many points along the fire line and advance the fire line to a new position. During this time, fuel is burned both in the fine fuels that carry the flaming front and in slower burning fuels behind the front; the heat and moisture fluxes from this combustion enter the fluid dynamics model as heat and moisture distributed near the surface, decreasing with height over a specified depth, usually 50 m.

Each rectangular atmospheric fluid dynamic grid cell is subdivided into one or more rectangular fire fuel cells. The model carries information for each fuel cell, including the remaining ground and (tree) canopy fuel mass, and the position of four tracers assigned to each grid cell that define the burning region within each cell and, when viewed as a whole with their neighbors, define the fire boundary. A local contour advection scheme keeps the line coherent, but fire spread rates at tracer points along the fire line are calculated using fire spread rate formulas dependant on local fuel characteristics and local wind speeds. This allows fine-scale dynamic feedbacks between the spreading fire and the atmosphere, producing vertically-oriented fire vortices ("firenadoes") and forward bursts. Although these simulations were performed at fine horizontal and vertical grid spacings (10s of meters), it is not yet clear how coarsely these simulations may be done and still capture essential fire-atmosphere interactions.

However, the fire component is entirely modular, and can be coupled to another atmospheric model. This would provide good results provided that the atmospheric model has suitable numerical schemes to handle large releases of buoyancy and accurately represents relatively fine-scale motions, particularly at small scales and in complex terrain. We must, of course, examine how some processes may be treated differently in the two atmospheric models, such as whether heat transport from the lower boundary layer is explicitly modeled (as in the Clark-Hall model) or parameterized (WRF, currently).

The Clark-Hall meteorological model is a three-dimensional non-hydrostatic numerical model based on the Navier-Stokes momentum, thermodynamic, and conservation of mass equations using the anelastic approximation. Vertically-stretched terrain-following coordinates allow users to simulate in detail the airflow over mountainous topography. It can ingest large-scale gridded data to incorporate a changing mesoscale atmospheric environment, although it does not have the sophisticated tools that MM5 has for obtaining and preparing this initialization data and significant user effort is required for each case. Its two-way interactive nested grids capture the outer forcing domain scale of the environmental mesoscale winds while allowing us to telescope down to the meter-sized fine dynamic scales of vortices in the fireline through horizontal and vertical grid refinement. While this has been a valuable research tool, the code structure impairs readability, limits the ability to add new capabilities as "plug-in" modules, hinders porting and parallelization without significant modification, and is extremely time consuming for users to learn. It is also not supported as a community model.

2.2. The Weather Research and Forecasting Model (WRF)

The WRF model (Michalakes et al, 2000) is a joint development effort between NCAR, the National Oceanic and Atmospheric Administration (NOAA) Forecast Systems Laboratory and National Centers for Environmental Prediction, and the Center for the Analysis and Prediction of Storms (CAPS) at the University of Oklahoma, with collaboration with scientists at a number of universities. The model provides a common framework for both research and operational weather prediction, and is designed to be a modular, flexible, maintainable, and extensible code that performs well on diverse computing architectures.

WRF is targeted for 1-10 km grid-scale, and is intended for operational forecasting, regional climate prediction, air quality simulations, and idealized dynamic studies. Although fire modeling at first appears to be outside the scales upon which WRF is focused, fine-scale processes such as at the 10's of meters scale will be within the range of motions that WRF is expected to accommodate, once advanced features such as grid refinement have been incorporated. The atmospheric and fire community of users will benefit greatly from NCAR's commitment to WRF as a community research and operational tool, as WRF developers plan to offer extensive user support through workshops, tutorials, email lists, web pages, and ftp sites for code access and updates.

3. Statement of Work

Because of the interactions between a wildland fire and the atmosphere, a key component of many research and development components of the Collaboratory is a coupled atmosphere-fire model. For example:

· Released atmospheric particulate concentration and size (and consequent impacts on cloud microphysical and radiative properties) depend on whether the fire is flaming or smoldering, which in turn depends upon local atmospheric conditions, fuel conditions, and the flow in complex terrain.

· Fire emissions have a large impact on the air quality of western and southern states. The gases released by vegetation due to the heat stress of an approaching fire depend similarly on the chemical pathways of combustion, the flaming or smoldering nature of the fire, as well as the atmospheric motions near the fire. Circulation can bring the gases back into the fire where the volatile ones participate in further chemical combustion processes, while if they are vented to the atmosphere, many participate in photochemistry to produce pollution in remote areas or regional haze. Installation in WRF would allow us to tie into the atmospheric chemistry module WRF Chem.

· Land Surface Models are being developed for WRF to assess the sensitivity of hydrological cycles to land surface properties. Fires not only impact the hydrologic cycle through being a heat and water vapor source, but also post-fire through their impact on surface properties, some of which can be quite devastating through floods, mudslides, and the destruction of watersheds. Land surface properties such as surface moisture and its dynamic response to weather are in turn a large factor in fire behavior. A dynamic LSM co-existing in the same model with a fire behavior component would allow many exciting research topics to be explored.

· Decision support tools we hope to develop for wildland fire managers as part of the broader Wildland Fire Research and Development Initiative include an operational model, which would require a smooth, efficient, portable model framework.

· Educational and societal impacts tools would be most effective if they were interactive and allowed the user to see the effects of their decision; this requires a portable, well-supported atmosphere-fire model framework.

Thus, we propose to implement the unique component of the model, the wildfire component, into a more versatile, evolving community research and forecasting model. By incorporating NCAR's fire model in the Weather Research and Forecasting Model (WRF), we can establish a stable foundation from which we and the other partners in the Wildland Fire Research and Development Collaboraty can pursue research goals, practical technology transfer applications, and tools for education and stakeholder discussions.

WRF is currently undergoing testing as new options and packages are added. Developers regularly conduct workshops and users tutorials. To see its use in daily operational forecasts, see http://rain.mmm.ucar.edu/mm5/ .

REFERENCES

Coen, J. L, T. L. Clark, and D. Latham, 2001: Coupled Atmosphere-Fire model simulations in various fuel types in complex terrain. 4th Symp. Fire and Forest Meteor. Amer. Meteor. Soc., Reno, Nov 13-15.

Clark, T. L., M. A. Jenkins, J. Coen and David Packham, 1996a: A Coupled Atmospheric-Fire Model: Convective Feedback on Fire Line Dynamics. J. Appl. Meteor., 35, 875-901.

Clark, T. L., M. A. Jenkins, J. Coen and David Packham, 1996b: A Coupled Atmospheric-Fire Model: Convective Froude number and Dynamic Fingering. Intl. Journal of Wildland Fire. 6:177-190.

Michalakes, J., S. Chen, J. Dudhia, L. Hart, J. Klemp, J. Middlecoff, W. Skamarock, 2000: Development of a next-generation regional weather research and forecast model. Proceedings 9th ECMWF Workshop on the use of Parallel Processors in Meteorology. Reading, U.K., November 13-16. Argonne National Laboratory Preprint ANL/MCS-P868-0101.