main collaborators: janice coen, ncar morwenna griffiths,monash mary ann jenkins,york don latham,...

Main Collaborators: Janice Coen, NCARMorwenna Griffiths,MonashMary Ann Jenkins,YorkDon Latham, USFSDon Middleton,NCARDavid Packham, BoMLarry Radke,NCAR/RIMichael Reeder,MonashRoland Stull, UBC

Coupled Fire-Atmosphere ResearchObservations and Modeling

by Terry L. Clark/UBC

OUTLINEObservations using IR Imagery

-overview of IR camera and analysis techniques -some prescribed and wild fire field experiments

Model-Description of dynamic code-Description of early NCAR fire code

-some results-Description of current WFIS fire code

-some results

OBSERVATIONSAND InfraRed Image ANALYSIS

Onion sage brush fire in Owensvalley, Ca 1985 courtesy of C. George

Street Patterns observed in Fires

Photo courtesy Brenner -observed in Florida

OBSERVATIONS

A. Infrared Camera

- Inframetrics PM380

- 3 to 5 m

- 256 by 256 array

- Sterling cooler

- 16 deg lens gives 1.4 rad resolution per pixel

Image Flow Analysis Applications• Understand fire behavior

• Calculate combustion zone winds

and their statistics

• Use derived data to validate

numerical models

Image Flow Analysis

Assumptions

– IR camera sees incandescent soot particles– Motion is on a distorted two-dimensional surface – Local features can be followed for short periods– We can fit data to simple types of motions, i.e.

translation, rotation, dilation and shear

. .

Image Registration1. Reduce image resolution (e.g. 7:1 in x and 5:1 in y)

2. Align image using IR intensity center of mass

3. Refine alignment using correlation analysis

n+1x +xy +yn(x,y))2

Minimize to estimate xandy

4. Extract linear trends in x t) andy(t)

5. Registered IR images:

- used in image flow analysis to estimate winds within the combustion zone

Image Flow AnalysisGradient Approach (Helmholtz theorem): Two-dimensional motions can be represented as the sum of six components. Translation

. .

)( 210 fdtd

)()( 3022 yfxfdtd

Rotation

)(40 yfdtdyfxf

dtdxf

Uniform expansion

And two shear components which most researchers ignore.

Robust StatisticsUsing the two components of translation,

0 fdtd

ztfxtf

w

u

zzfxzfxzfxxf

we obtain the matrix equation

which we solve at each pixel for u and w. If |u2 + w2| > S2 thenwe flag that pixel as an outlier and avoid considering it in future calculations. Typically, S= 20 to 40 m/s.

Least Squares Minimization

ztf

xtf

w

u

zzfxzf

xzfxxf

After identifying outliers we sum over a patch of data as

We typically use 7 by 7 pixels.Outlier points are not included in any of the summations.

This simplest approach requires the inversion of a second order matrix to estimate u and w..

International Crown Fire Modelling Exp

Cameras on 50fttower

•Canadian and US Forest Services •Near Fort Providence NWT Canada• Prescribed crown fires•150 by 150 m plots•June – July 1997•Tower based IR measurements•Plot 6–9 July 1997

Workshop at the Site

Plot 5 Fire Whirl

Plot 6 Ignition

Plot 6 at 2:08

Plot 6 at 2:09

Plot 6 at 2:10

Plot 6 at 2:10 plus

Plot 6 at 2:11

Turbulent Burst Sequence

Video: Plot 6 Visual

Derived Winds for file=7004

Fig 10a Clark et al. 1999, JAM

Wild Fire Experiment

• NCAR• Sept 1998, Montana, Colorado and California • Infrared Camera mounted on NSF/NCAR C-130• Wildfires were the target of opportunity• First case was in Glacier National Park

-Challenge Fire Complex 4 Sept 98

-100 m long hairpin vortex observed with IR

-fire about 2 km away from camera

IR Imagery from C-130FOD - Part II

Finger shot outabout 100 m in 1-2 sec

Indications of burning fuel after finger retreats

Hairpin or Turbulent Burst

Video: IR Observations over Glacier National Park - 4 Sept 1998

Northern Territory Grass Fire Experiment- Australia

• Spear grass burns 40 km South of Darwin

• Kerosene grass burns near Batchelor

• Used 19 m high cherry-picker as platform

• Platform motion requires apparent motion treatment

Cherry Picker

Fuel Type - Australian Spear Grass (Sorgum_Intrans)

Litchfield Kerosene Grass (19 m up)

Video: IR Data Hughes plot 3

Camera 200 m from fire

giving 30 cm pixels

Video: Unprocessed Images Plot-3a Hughes Airfield

Video: Fire Winds using least squareswith registration

Velocity Statistics – Least Squares

MODELING

Numerical Model• 3D Non-hydrostatic 2nd order finite-differences

• Terrain following - geo-spherical coordinates

• Vertical and horizontal grid refinement

– 2-way interaction

• Vertically stretched grids with grid refinement• - Clark (1977,JCP), Clark-Farley(1984,JAS),

• Clark-Hall(1991,JCP; 1996,JAM)

• Boundary-Initial conditions from NWP

• Bulk parameterizations of rain/ice processes-Kessler (rain), Murray-Koenig (ice) and much more

recent approaches under development.

Coordinates

.

),(

)()(

),()/),()((

gridsconstantontocastisModel

orography.theofheighttheis

and

00

where

1

Δζ

yxh

HHFF

yxhHyxhFz

incrementslatitudeconstantrepresents

incrementslongitudeconstantrepresents

yΔ

xΔ

Horizontal Coordinates

Vertically Stretched Coordinates

Multi-Processing Approach

Message Passing Interface (MPI) software is usedfor multi-processing.

Model Configuration for 3 LayersNVRT=0 (no tiling)

SingleProcessorframework

N=1

N=3

N=2

Multi-Processing Configuration for 3 Layers

NVRT=1 (with tiling) Layer 1 details

Multi-processorFramework

Four sub-domainsper layerMCPU=4

N=3

N=1

N=4

N=2

Green=lmx1,lmx2lmy1,lmy2

Blue=mi2mo

Grid Refinement Using 5 Domains

Example fromClark et al. 2000, JAS

Grid size ranges from26 km to 200 m(4:1 4:1 4:1 2:1)

CO

DOMAIN 5 OROGRAPHY

Example fromClark et al. 2000, JAS

Wildfire Modeling

Rational for Wildfire Modeling

• Wildfire propagation physics is poorly understood• FS spread models use empirical fits from

– Low intensity small fires– Laboratory fire tunnels– Neither can hope to represent the vast parameter space of intense fires

• Understanding fire behavior involves– Combustion winds interacting with the fire and ambient flow– Fire-atmosphere heat exchange– Fire-fuel heat exchange – Chemical release and transport by the convection

• Some Applications of Coupled Fire-Atmosphere Models – Study burn paradigms – Understand fire related sources/sinks to atmospheric budgets– Develop suppression techniques– Like the NWP problem, fires are too non-linear to predict using empirically

derived rule based techniques, i.e. there is no empirical fit to a severe nonlinear event.

NCAR FIRE CODE

-Physical treatment of fire at a very ‘first cut’ level, i.e. useful for preliminary evaluation.

Fire Atmosphere Coupling

The sensible and latent heat fluxes were added to the vertical diffusion terms as:

)5()),((

)4()),((

txFvqvqKqvvqK

txsFKK

Where Fs and Fare the heat and moisture inputs from the fire. Fs can reach values up to 1-3 MW m-2. .

)/exp(,,),,(

)/exp(),,(),,(

lzhyxlFzyxlFszhyxsFzyxsF

Where s and l are extinction lengths for the sensible andlatent heat fluxes.

Spread Rate Treatment

01 RwspS )(

)( nvww

s is the slope coefficient

is the wind coefficient

nv is the wind normal to the fire front

R0 depends on fuel type and moisture content.A BURNUP type curve is used to describethe rate of mass loss for each fuel cell.

BEHAVE formulation

• Strong need for improved spread rate parametrizations appropriate for coupled fire-atmosphere models

Fire Line Propagation Scheme

•Contour advection scheme•Avoids assuming shape such as ellipse•We want the physics to determine shape

Tracer Code

• Area inside tracers designates ignited fuel•Thex and y of fire model range from 10 to 300 m • z ranges from 5 to 10 m• Fuel cellxf and yf range from 5 to 60 m

•Grid point method used to track fuel and fire• Four particles (or tracers) for each fuel cell•Local contour advection scheme used to move tracers

Fig3 Normal Vectors

Fig 4 Spot Fire- Zero Wind

Fig 5 Spot fire- West Wind of 3m/s

Fig 6 Spot Fires at t=200 s

Kinematical test usingBEHAVE with fuel class-3 (tall grass)U= 3 m/s

Fig 7. Line fire in oscillatory wind

Fig. 8 Windmill time seqence

Fig 9 Windmill -Converging fire lines

Video: Animation of six tests

Line fire propagating over a hill

• Slope about 60 percent

• Nh/U about 1 giving deep evanescent modes

• Without fire critical level amplifies wind speeds by about a factor of 3

• Fire winds dominate ‘ambient flow’

• Line spreads up hill with parabolic shape

• Bifurcation occurs at hill crest

• Vortex production along front

Video: Buoyancy and Enstrophy

Red=2 deg Buoy

Purple= .1/s Enstropy

Video – 4 deg K buoyancywith volume rendering of smoke

Video: Isosurfaces of .07/s vertical vorticity

Video: Plan View of Vertical Vorticity

Weather Fire Integrated System(WFIS) FIRE CODE

-Physical parameterization of fire at the level ofsophistication where we can begin to compare with the natural event.

Following results fromClark, Griffiths,Reeder and Latham, 2003:Numerical Simulations of Grassland Fires in the Northern Territory, Australia: A New Sub-Grid-Scale Fire Parameterization, J. Geophysical Research, (in Press)

Assumptions1. Assume a time scale for combustion b, e.g. 1-2 s.

2. Assume combusting material and air are two interacting fluids

3. The mixing ratio of the combustible material is Mf.

4. The burning portion of Mf has a volume mixing ratio f.

5. The temperature of the combusting material is Tb.

6. The effective grid cell area of radiation is determined so that Tb (maximum) = 1200k. This assumption provides us with an

effective volume/area (= lf) ratio.7. Exchange of heat between the combusting fluid and the air is

assumed to occur from:• Interfacial molecular diffusion using an ls length scale

(without/with cases treated)• H20 and CO2 absorption. (not considered as yet)

8. Above assumptions designed for models using horizontal and vertical grid sizes of about 1 to 3 m.

Governing Equations

)( fMKbSfSzfM

dt

d

Sb Mf / b

Sf Soexp( z / s )

d

dtMb Sb Sv (KMb)

Sv 0.56Sb 0.56Mf / b

Conservation equation for Mb (smoke)

Conservation equation for the mixing ratio of combusting material, Mf

Governing Equations

)(

)(42

*

bTKpc

bTfs

pcbSbCf

bTbT

dt

dpc

d

dt

sf(Tb )(K)

Thermodynamic equation for temperature of combusting material,Tb.

Conservation equation for showing the fire-atmosphere heat exchange terms.

Effective buoyancy that drives updrafts is now defined as

bTfffef )1(

Figure 8.Horizontal and vertical cross sections from experiment CFA. (a) is at the surface, (b) is at x = 150 m, (c) is a section through

y = 150 m, and (d) is taken at a height of 1 m.

Figure 4.Maximum and minimum u, v, w and maximum

versus time for experiment CFA.

Figure 6.Plan view of fire line at t = 1.4 min showing the wind vectors

at z = 2m above ground and plotted every 2 m for experiment CFA. The bold line marks the position of the fire front. Only a section of

the domain is shown to highlight the variations along the fire.

Figure 7.(a) Maximum combustion temperature, (K), (b) spread rate (ms-1), (c) heat flux (MWm-2) and (d) total sensible heat flux (GW) from

the fire versus time for experiment CFA.

Figure 9.Horizontal cross section of heat flux at z = 3 m above the surface for experiment CFA

Figure 2. at t = 24 s from a series of Australian grass fire experiments. The values of are 1., .5, .25 and .125 cm, respectively. The contour

interval of is 8 K.

ls =1 cm ls =.5 cm

ls =.25 cm ls =.125 cm

AVI showing Isosurface of W (1 m/s)(yellow) with Tb (600K) in blue shown near surface

Conclusions

1. Provided research funding is forthcoming we have a modeling framework that can be extended to the self-determining level by treating the fire-fuel heat exchange.

2. We can easily add new field variables to study the transport of various chemical species.

3. We can also treat additional physical effects through the addition of new field variables, e.g. break Mf into two variables to consider flashover effects.