Computational Modeling of Large Wildfires: A Roadmap
-
Upload
galvin-hatfield -
Category
Documents
-
view
21 -
download
1
description
Transcript of Computational Modeling of Large Wildfires: A Roadmap
![Page 1: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/1.jpg)
Computational Modeling of Large Wildfires: A Roadmap
Computational Modeling of Large Wildfires: A Roadmap
Janice Coen1 and Craig Douglas2
1 National Center for Atmospheric Research, Boulder, Colorado, USA2 University of Wyoming, Wyoming, USA
![Page 2: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/2.jpg)
Wildland Fire - Weather Basics
![Page 3: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/3.jpg)
3 Environmental Factors that affect Wildland Fire Behavior
Fuel Moisture, mass/area, size, hardwood vs. conifer, spatial continuity, vertical arrangement
Weather wind, temperature, relative humidity, precipitation Weather CHANGES: fronts, downslope winds, storm downdrafts, sea/land breezes, diurnal slope winds
TopographySlope, aspect towards sun, features like narrow canyons, barriers (creeks, roads, rockslides, unburnable fuel)
duff Surface litter, grass, shrubs, twigs, branches, logs
Tree crowns
These are not independent
![Page 4: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/4.jpg)
Weather conditions over wide time and space scales influence:
• where a fire occurs (lightning)• ignition efficiency• how fast/where it spreads• combustion rates• whether or not the fire produces extreme behavior
Weather is one of the most significant factors influencing wildland fire and (compared to terrain and fuel
characteristics) the most rapidly changing.
Weather
![Page 5: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/5.jpg)
Flow in the vicinity of a fire• Updraft• Flow into the base • Compensating
downdrafts
Notes:
- Air is not ‘nothing’.
- It has weight & exerts a force.
- There are rules about how it can behave.
- We calculate these forces with mathematical equations of fluid dynamics, mass conservation, and thermodynamics
![Page 6: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/6.jpg)
Fluid dyanmics basics
![Page 7: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/7.jpg)
Fire creates its own weather & can change winds over several miles
• Wildland fires don’t just respond to weather, they interact with the atmosphere surrounding them:– The fire releases heat, water vapor into the atmosphere.– This alters winds, pressure, humidity, etc. in the fire, in the fire
plume, & in the fire environment– This in turn feeds back on the fire behavior.
• This is a basic component of ALL fire behavior, not just in “plume” vs. “wind” driven, or high vs. low intensity, crown vs. grass fires.
Photo courtesy of Jeff Zimmerman
![Page 8: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/8.jpg)
Is this understanding currently being used?
ACCEPTED:Changes in wind and slope cause variations in fire behavior.
HOWEVER:Current operational tools & training modules
•Do not include interplay between fire & atmosphere•Are diagnostic, not predictive.•Example:
They cannot predict dynamical effects (i.e. changes in fire behavior or blowups) that lead to severe fire behavior or firefighter safety problems.
The next generation of fire models for operations and planning & training materials should include realistic fire, wind, and terrain interactions.
![Page 9: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/9.jpg)
How does fire interact with its environment?
Fire whirls
Photo courtesy of Josh Harville
Photo courtesy of Canadian Forest Service
![Page 10: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/10.jpg)
Photo courtesy of Josh Harville
![Page 11: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/11.jpg)
How does fire interact with its environment?
Convective fingers
Photo courtesy Charles George, USDA Forest Service
![Page 12: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/12.jpg)
How does fire interact with its environment?
Turbulent bursts
![Page 13: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/13.jpg)
Photograph courtesy of Jeff Zimmerman
![Page 14: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/14.jpg)
Wildland Fire Modeling at Fire Behavior Scales
![Page 15: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/15.jpg)
Wildland Fire Modeling
A Hierarchy of Models•Experience, Intuition
•Point-based Empirical, semi-empirical equations (i.e. Rothermel 1972)
•2-D in horizontal, non time dependent (FARSITE)
•3-D, time dependent, weather model:semi-empirical fire, 10 m – 1 km grids), NCAR/USFS CAWFE, WRF-Fire
•3-D, time dependent, air flow model:parameterized combustion, 1 m grids - Los Alamos FIRETEC, NIST FDS
•Full combustion treatment (1 cm grids) (does not exist)
Increasing complexity
- Computational science problem! The ‘best’ one: depends on the constraints of what you are trying to do.
![Page 16: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/16.jpg)
Possible usage scenarios for a model in a wildfire incident
Time Horizon Specific Decision Information
Will there be extreme behavior on this fire this afternoon? (safety)
Where will the fire be at the end of today ?What are the Best/Worst/Most Likely scenarios for
where this fire can be stopped? Could this become one of the “biggies”?
What would a fire do in this location? How would various fuel treatments affect this?
These pose different challenges, require different development tracks, and may be suited best for different models.
“What If?”:
Few hours:
1 day :
~ 1 Week:
Few hours:
1 day :
~ 1 Week:
“What If?”:
![Page 17: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/17.jpg)
An Atmospheric (“Weather”) Model
• While modeling flow over U.S., can telescope to model flow over single mountain valley with high resolution
• Models atmospheric structures like fronts, windstorms, formation of clouds, rain, and hail in “pyrocumulus” clouds over fires
Numerical weather prediction models solve predictive equations of fluid motion to forecast air velocity, temperature, water vapor, cloud water, rain, and ice on a grid in a 3-dimensional box.
![Page 18: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/18.jpg)
Wildland Fire Modeling
• Important output quantities: Rate of spread (ROS) of the leading edge of the flaming front, intensity, flame length
• In dynamic models (where forces are calculated), spatial and temporal variability of consumption is also very important
• Concept of “fire line” – interface between burned and fuel that has not yet ignited
![Page 19: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/19.jpg)
FIRE
Fire Behavior Model
Atmospheric Dynamics
ATMOSPHERE
(Weather model)
Heat, water vapor, smoke
Fire Propagation
FIRE ENVIRONMENT
Dead Fuel moisture
Coupled Weather – Wildland Fire models2-way coupling
![Page 20: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/20.jpg)
Fire Model
Contains components representing:1. Surface fire
– Spread of “flaming front” depends on wind, fuel, and slope. Based on Rothermel (1972) semi-empirical equations.
– Post-frontal heat/water vapor release– Subgrid-scale tracer points define the
interface between burning and unburning regions (computational science problem!)
– Fire line shape evolves naturally– Fuel can be heterogenous, can change any
property
2. Crown fire– If the surface fire produces enough heat, it
heats, dries, and ignites the tree canopy.
3. Heat, water vapor, and smoke fluxes released by fire into atmosphere
Surface fire
Crown fire
![Page 21: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/21.jpg)
Fireline self-organization
![Page 22: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/22.jpg)
Defining Burning Regions within each atmospheric cell
At the surface, within each atmospheric grid cell…
• Fuel is divided into rectangular grids (4x4, 10x10, etc.). When a grid of fuel is ignited, 4 tracers are created, defining what part of a fuel cell is burning.
• Fire tracers can move with the wind, back, or move normal to it.
Common Computational science problem: Keeping track of an interface within grid cells
![Page 23: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/23.jpg)
Fire Line Propagation
Fire spreads perpendicular to the (local) fire line
Spread rate specified by empirical formulas that calculate the rate of spread as a function of local winds, terrain slope, and fuel
Contour advection scheme keeps fireline coherently shaped.
![Page 24: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/24.jpg)
Mass loss rate
An approximation to the BURNUP (Albini and Reinhardt) algorithm treats the rate of mass loss due to burning for fuel of different types and sizes.
![Page 25: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/25.jpg)
Fraction fuel remaining with time Heat flux (kW/m2)
Grass
W=30
Woody fuel mixture
W=500
![Page 26: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/26.jpg)
The “Universal” Fire shape
The head (the flaming front), flanks, and the backing region.
Photo courtesy Ian Knight, CSIRO
![Page 27: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/27.jpg)
ANIMATION (fire1.avi)
Field: smoke Arrows: strength and direction of winds near ground level
The “universal” fire shape and fire whirls evolve from fire-atmosphere interactions.
![Page 28: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/28.jpg)
Landscale-scale case studies
![Page 29: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/29.jpg)
1. Big Elk Fire, CO-ARF-238 July 17–23, 2002
“Simple” Case study• “Simple” weather & airflow regime• “Simple” fuel distribution
![Page 30: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/30.jpg)
Big Elk (cont.)
Photo courtesy of Kelly Close
![Page 31: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/31.jpg)
Inputs
AtmosphereInitialize atmosphere & provide later
boundary conditions with large-scale weather forecast
Topography US 0.33 - 3 second topography
Fuel - Surface and canopy fuels.Mass/area Physical characteristics
Dead (live) Fuel moisture
6 nested domains ( 106-107 gridpoints):• 10 km, 3.3 km, 1.1 km, 367 m, 122 m, 41 m horizontal atmospheric grid
spacing. (Fuel grids much finer.). Time step in finest domain < 1 sec.
Configuration for a research problem
Domain 6
6.7 km
6.7 km
![Page 32: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/32.jpg)
4 hr simulation
Grid spacing ~50 m.
Red: Isosurface 10 oC (18 oF)buoyancy
White: smoke concentration
Arrows: strength & direction of wind
Frame each 30 sec.
W
N
ANIMATION: Big Elk Firebigelk_cinepak.avi
(4400 acres) Pinewood Springs, CO 17 July 2002
![Page 33: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/33.jpg)
• Although this influence is most dramatic near the fire, model simulations show this influence can change winds by several miles per hour even miles from the fire.
![Page 34: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/34.jpg)
[With fire] – [Without fire]
![Page 35: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/35.jpg)
2. Esperanza Wildfire
• Collaborators: Phil Riggan, Francis Fujioka, David Weise (USDA Forest Service, Riverside Fire Lab), Charles Jones (Univ. of California Santa Barbara)
![Page 36: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/36.jpg)
• Complex meteorology• Complex fuels• Complex terrain
Santa Ana Winds & Wildfires that occur during them
![Page 37: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/37.jpg)
Santa Ana Winds
![Page 38: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/38.jpg)
Esperanza Fire Overview• Ignition: arson-caused wildfire on 10-26-06 at 0112 in a river wash outside Cabazon,
CA (Riverside Co.)• Burned uphill influenced by strong winds and steep slopes. Spread WSW.• Rapid fire spread due to Santa Ana winds (E 6-10 mph, 20-25 mph gusts), 6% RH,
flammable brush
![Page 39: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/39.jpg)
Photograph courtesy of Jeff Zimmerman
![Page 40: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/40.jpg)
![Page 41: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/41.jpg)
Typical study addresses the meteorological factors, fire behavior, and interactions leading to the observed phenomena in the 1st day of the fire
• Experiment:– Start with 10-26-06 0 Z (10-25-06 1700 local time) 48-hour
MM5 simulation (C. Jones, UCSB) 10 km domain– Interpolate to Clark-Hall weather model grid – Use to Initialize and update CAWFE boundary conditions
• Use CAWFE to model weather and fire– Nest grids, refining from 10 km -> 3.3 km -> 1.1 km -> 0.37
km– Simulate 17 hrs of weather + fire beginning at 10-25-06 at
2100.
![Page 42: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/42.jpg)
FuelFirst approximation: LANDFIREDeveloping improved fuel information from pre-fire
aerial imagery and hand-collected samples
Grass
Grass: (1) short grass
Shrub: (5) mixed brush
Forest: (10) heavy litter
Brush
Forest
![Page 43: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/43.jpg)
ANIMATION (seq2_repaired.avi)Fire sensible & latent heat flux and smoke. Simulation covers 1:00 am – 9:43 am
![Page 44: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/44.jpg)
FireMapper Data USDA Forest Service, Riverside Fire Laboratory Laboratory
10-26-06 1117
![Page 45: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/45.jpg)
Landsat 5
![Page 46: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/46.jpg)
LandSat 5 image
False color composite
![Page 47: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/47.jpg)
Wave motions in air caused by two upwind ranges of mountains are bringing high momentum air down on the fire, causing it to spread quickly.
Fire site
Wave motions excited in air by 2 ranges upsteam
![Page 48: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/48.jpg)
WRF-FireA module to enable simulations of wildland fire behavior and interactions with weather
• Purpose: Develop, maintain & test a public domain wildland fire module for use with WRF (Weather Research and Forecasting model) that simulates wildland fire behavior over prescribed and landscape-scale fires. Why?
– Understand 2-way interactions between 4-D weather & wildland fires that influence fire behavior and growth
• Status: WRF-Fire module and documentation released in WRF 3.2, April 2010.
J. L. Coen, J. D. Beezley, M. Cameron, J. Mandel, J.Michalakes, E.G. Patton, K.Yedinak, WRF-Fire: Coupled Weather-Wildland Fire Modeling with the Weather Research and Forecasting Model. in preparation.
J. Mandel, J. Beezley, J. L. Coen, and M. Kim , 2009: Data Assimilation for Wildland Fires Ensemble Kalman filters in coupled atmosphere-surface models, IEEE Magazine.
Image: Ronan Paugham
![Page 49: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/49.jpg)
Sensitivity studies
• Study the sensitivity of simulated fire characteristics such as perimeter shape, intensity, and rate of spread to external factors known to influence fires such as fuel characteristics, wind, and terrain slope. Using theoretical environmental vertical profiles, experiments vary these external variables
• Results:– simulated fires evolve into the observed bowed shape as a result of fire-atmosphere
feedbacks that establish the flow in and near fires.
– Coupled model reproduces recognized differences in fire shapes and head intensity between flashy grass fuel, shrub surface fuel, and forest litter types.
– Coupled model reproduces the observed tendency of heavy, dry fuels in strong winds on steeper slopes to lead to faster moving fires.
J. L. Coen, J. D. Beezley, M. Cameron, J. Mandel, J.Michalakes, E.G. Patton, K.Yedinak, in preparation.
![Page 50: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/50.jpg)
LINGERING ISSUES
• Tough problem. Challenges include all those of weather modeling + fire behavior – Modeling/Forecasting limitations to predictability – physical processes that span a vast range of scales
(combustion: mm or cm, weather: m - 10s of km)– processes such as spotting (where burning embers are
lofted and tossed ahead of the fire, starting new fires) that should perhaps be modeled stochastically instead of deterministically
– estimating the consequences of uncertainty, – difficulty of gathering pertinent data for verification and
initialization in a dangerous environment
![Page 51: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/51.jpg)
Physical Processes That Span a Vast Range of Scales
• Combustion processes– [mm, computational domains of (10 cm)3]
• Meteorology– Boundary layer scales of 10’s of m to synoptic scale 100’s of
km
The gap between computationally feasible and functionally relevant scales has not been bridged (even with grid refinement and parallel computing)
![Page 52: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/52.jpg)
Fire Modelingat different scales
![Page 53: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/53.jpg)
Forecasting must be done faster than “real-time”
• Progress in faster processor speeds has given everyone a supercomputer on their desk, often with several processors for OpenMP or MPI parallel capabilities.
• Numerical weather prediction models can take advantage of many processors but efficiency flattens out at hundreds of processors.
• Are GPUs the answer?
![Page 54: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/54.jpg)
Ignition of new fires by “Spotting”, the lofting of burning embers that start new fires
Big Elk Wildfire (Pinewood Springs, CO)
A necessary component of models, but…
Can estimate combined probability - size distribution of embers, probability of a burning ember staying lit to the ground, distribution of landing spots, and probability of ignition – but cannot model deterministically.
![Page 55: Computational Modeling of Large Wildfires: A Roadmap](https://reader035.fdocuments.net/reader035/viewer/2022062717/56812add550346895d8ec621/html5/thumbnails/55.jpg)
Thank you
Photos: http://zimmermanmedia.com
NCAR is sponsored by the National Science Foundation.
Portions of this work sponsored by NSF ITR, NSF CDI, and NSF MRI.