Environmental Monitoring & Modelling Research Group

Parameterization of Fire Injection Height Ronan Paugam Environmental Monitoring and Modelling Research Group, Dept. of Geography, King’s College London, UK. Martin Wooster, Weidong Xu, Maria Val Martin, Martin Schultz, Saulo Freitas 1

Outline

    

Introduction literature review: data? 1D Plume Rise Model (PRM) PRM Validation: Optimization & Sensitivity Studies Conclusion and Perspective

2

Introduction Data?

Fire characteristic Atmosphere profile

Model?

Model Briggs 1984

empirical

Injection Height statistical

physical

x

Saulo Freitas et al. 2007, 2010

x

Rio et al 2010

x

Sofiev et al 2012

Available Data  Individual scale: MISR stereoheight

 Large scale: CALIOP, MOPITT

x

Large scale study  Pfister 2008 : CAM-no PRM, Alaska 2004  Session 2010 : WRF-Chem + (Freitas,PBL,35km), Alsaka 2008  Grell 2011 : WRF-Chem + Freitas, Alaska, 2004  Pfister 2011 : WRF-Chem + Freitas, California 2008  Raffuse 2012 : CMAQ+Briggs (BlueSky),US 2006-2008,

3

Direct Relationship: FRP, CALIOP, MISR Level1

It doesn’t work! - atmospheric effect are not resolved - matching between FRP and htop ?

[Raffuse 2012] Level2

Aerosol layer product

[Amidiris 2010]

unstable FT

[Val Martin 2010] 4

Data – FRP,MISR stereoheight htop = Model(FRP, parameters)

FRP

Fire dynamics

FRP time variation (≠fire emission)

≈11h Terra

≈14h Aqua

Local time

Plume dynamics Time delay stability ∝ Atmospheric fire

5

Data – FRP,MISR stereoheight ? Terra alone is unlikely to capture fire variability

Terra - MISR

MOD021KM emissive band

Terra - 18:30 UTC

?

Aqua - 20:20 UTC

2h FRP = 1907 MW Af area = 10 ha BT = 469 K

FRP = 6961 MW Af area = 134 ha BT = 223 K

6

Ideal Parameterization Quick overview:  No Direct Relationship to htop  Empirical equation (eg Briggs) have been developed on stack plume experiment (≠Radiative and convective fraction)  Statistical Approach: Sofiev et al 2011 equation is highly affected by not correlated data.  Physical Model: Freitas Model is the most used. No proper validation at fire scale event. assumption are based on few parameters rely strongly on AfArea More the model is physically constraint, less it can be skewed by data Better selection of [FRP,htop] Data Objective: Improvement and Validation of a new PRM based on the Freitas Model Improve Physics model Mass conservation New entrainment scheme

[Raffuse 2012 ]

Validation of the Physics on well characterised fire event

Introduce more time dependence with an Implementation on-line

MISR + MODIS + Atmospheric Profile

CALIOP product for validation of large regional scale

7

PRM – Main Physical Processes • Convective zone Atmospheric interaction

Pyroconvection

Wind Drag

PBL

entrainment R

• Combustion Zone •Vegetation

Convective Flux Convective Flux

Plume Rise Model Freitas et al 2007 Fire Radiative Power Active Fire Area 8

PRM: time dependent Equations Original equation, Freitas 2007

Modified equation Prognostic variable

Shallow convection [pergaud et al 2009]

mass conservation

with

Entrainment/detrainement scheme

Closure [Morton Turner and Taylor 1956]

MTT 1956 Rio 2010 New approach based on the Brightness Temperature

PRM Validation: Data Base For each MISR Observation: All Clusters within 10km to the reference Point of the MISR Plume contour: •MODIS FRP •Active Fire Area (Af Area) •Brightness Temperature = f(FRP,Af Area) Dozier Algorithm MISR top height MISR Median height terrain variation 10km2 (Gtopo30) Land Cover from the dominant type defined in GEFD3 ECMWF Analysis profile at the time of Terra Overpass PBL Height derived from the 1st local max of RH Problems: Validity of the Dozier Algo when Applied to MODIS? Impact of the Atmospheric profile reanalysis? plume interference? Impact of the Cluster Algo? 10

PRM Validation: Cluster Algo?

Impact of fire regime (smoldering/flaming)

Af Area (ha)

Same MODIS14 but run different algo for clusterization

weidong maria

maria weidong

FRP (MW)

FRP

AFarea

fire

BT

fire

fire

PRM Validation: selection process nbreObs Selection process: Start from the whole MISR data set over Northern America ......................... 2114 1. Remove fires with plume interference and no convergence of the Dozier Algo ........................... 896 2. Keep 1 Vegetation type (Extratropical Forest) ........................... 775 3. Keep only Stable fire : effect of diurnal cycle, fire fighter high latitude: Yukon and Alaska, lat > 60 ...........................412 4. Ensure that Intensity (W/m2) and AfArea peak happen before or during Terra Overpass .......... 29 FRPm2 = 106 MW/ha AfArea = 66 ha

select multiple overpass of Aqua and Terra

21h25 UTC

21h25 UTC

FRPm2 = 64 MW/ha AfArea = 59 ha

21h40 UTC

12

PRM: Optimization (undergoing...) en/detrainment scheme

closure scheme

Buoyancy flux F = f(E,AfArea) Convective Heat Flux E=f(FRP) = β FRP [Freeborn et al 2008; Riggan et al 2004]

y=0.8x+0.4 r2=0.8

α = 0.04637888 Cε,Cδ, Cεdyn,Cδdyn = , 2.54068068, -20.26901446, 4.02378368, 2.5951726 β = 0.96146643

13

PRM: Result Single fire :

29 fires :

y=0.7x + 0.7 r2 = 0.7

detrainment zone and time dependence?

14

PRM: Sensitivity Studies MISR data: 29 fires 4

number

800

FRPm2 (MW/ha) AfArea (ha)

160

MISR data : all fire

PRM result for one atmosphere

800

AfArea (ha)

160

FRPm2 (MW/ha)

altitude (km)

FRPm2 (MW/ha)

800

AfArea (ha)

160

15

Effect of Atmosphere profile

FRPm2 (MW/ha)

PRM: Sensitivity Studies

16 AfArea (ha)

Conclusion and Perspective Conclusion:  Literature shows a need for more robust parameterization  So far, the plume rise model approach seems the most adequate  Current Work:  comprehensive data set  new assumption in the model  Optimization and Validation are still undergoing

Perspective:  Use of more MISR Observation: other vegetation type and geographical location  Possible collection of more detailed data (SAMBBA) usable for validation  Use of high resolution model? [Trentmann 2002]  Validation at both fire event and regional scale

17

18

PRM Validation: Data? example of weird plume contour MINX user?

19

PRM – Different approach  Rio et al 2010 – mass flux formulation

 f = ρw   ∂f  ∂z = e − d

no microphysics implementation in LMDZ no proper validation on fire scale event

 Freitas et al 2007, 2010 – 1D time dependent cloud model it can take into account pyro-convection

20

PRM Comparison Initial Conditions - size of the fire R0 - buoyancy flux F=f(CHF,R)

Dozzier(1981) algorithm CHF defined by Type of Vegetation or FRP

equations

initialisation

Saulo

original

R0 and vegetation type

Maria

original

R0 and FRP

Ronan

modified

R0 and FRP 21