a`1
Transcription
a`1
Workshop Mathematical Modelling and Numerical Simulation of Forest Fire Propagation Predicting Eruptive Fire Behaviour by Domingos Xavier Viegas Department of Mechanical Engineering, University of Coimbra, Portugal xavier.viegas@dem.uc.pt Vigo, 29/30 November 2007 Contents • Introduction • Modeling Fire Behaviour • Eruptive Fire Behaviour • Case Studies • Conclusion WMMNSFFP November 2007 2 Introduction • For many years we have observed that forest fires can behave in an extreme way with a sudden acceleration and reach very high rates of spread. • This sudden change in fire behaviour has caught by surprise many fire fighters and has been the cause of many fatal accidents. WMMNSFFP November 2007 3 • Scientists and managers have tried to explain and model this phenomenon in the past. • Current fire behaviour models do not deal easily with ROS values that change in orders of magnitude. • Several arguments have been proposed to explain it, but in my opinion most of them are not convincing. WMMNSFFP November 2007 4 • This phenomenon is known in the literature as “blow-up”. • We prefer to call it a fire eruption. • In this lecture we will deal with the problem of sudden fire acceleration in forest fires and propose an explanation for it and present a model to predict fire behaviour. • I hope that you will understand why this phenomenon can not be called unpredictable. WMMNSFFP November 2007 5 • This lecture is based on the research that is carried out by my Group at the University of Coimbra for the past 20 years in forest fire propagation. • I acknowledge the contribution given by my co-workers and the support given to our research program by several institutions. WMMNSFFP November 2007 6 • Methodology: – Analytical and numerical work – Field and laboratory experiments – Analysis of real cases. WMMNSFFP November 2007 7 Forest Fire Research Laboratory 2004 Seoul September 2007 8 Combustion wind tunnel Túnel de-Combustão Lousã Portugal Seoul September 2007 9 Joint effect of slope and wind Seoul September 2007 10 Fire spread in canyons – Large Table DE 4 Seoul September 2007 11 Vertical Combustion Tunnel Seoul September 2007 12 Experimental field of Gestosa Lousã (Portugal) Seoul September 2007 13 Gestosa 2002 31st May 2002 Seoul September 2007 14 Observation of real fires Fundão, 30 Jul 03 Seoul September 2007 15 Analysis of accidents Seoul September 2007 16 Modelling Fire Behaviour • Fire spread regimes: – Low intensity fire (Ground fire) – Average intensity fire (Surface fire) – High intensity fire • Eruptive fire • Crown fire • Spot fire WMMNSFFP November 2007 17 • Classical fire behaviour modelling is based on various assumptions: RB SA RB A UA B WMMNSFFP November 2007 18 Fu ro l teo Me el Triangle of Fire Factors y og Topography WMMNSFFP November 2007 19 • Concept of rate of spread Present fire behaviour modeling is based on the concept of rate of spread ROS R = φ (U, α, mf, Mc, σ, β, …) WMMNSFFP November 2007 20 • Concept of rate of spread ROS: – The movement of the fire front is characterized by a ROS vector that is well defined at each point of the fire front. • Local independence: – The ROS vector depends only on the local properties at each point of the fire front. WMMNSFFP November 2007 21 • Reference or Basic ROS U=0 and α=0º R=Ro Ro = φ (mf, Mc, σ, β, …) WMMNSFFP November 2007 22 Combustibility Table Determination of Ro Combustibility Table Seoul September 2007 23 Determination of Ro Ro (cm/s) 0,5 0,4 P. Pin. 0,3 0,2 0,1 0,0 0 10 20 30 40 50 60 70 80 mf (%) Seoul September 2007 24 • Effect of wind or slope R R' = = φ' (U, α) Ro • Concept of elliptical growth RB SA RB A UA B WMMNSFFP November 2007 25 3 - Eruptive Fire Behaviour • Fires in steep slopes and in canyons have a peculiar behaviour. • Their rate of spread increases continuously and in the limit they produce a “fire eruption” WMMNSFFP November 2007 26 Eruptive processes in Nature Forest fires Volcanoes Monte de Santa Helena Seoul Incêndio de Thirtymile, 10 Julho 2001 September 2007 27 Monte de Santa Helena Loop Fire 1966 Seoul September 2007 Staff Ride 28 WMMNSFFP November 2007 29 DE 3 Canyon shaped Table WMMNSFFP November 2007 Oct.2002 30 Point ignition in a canyon Fuelbed of pinus pinaster 16 14 R'in R'med 12 R'fin δ=40º R/Ro 10 8 6 4 2 0 -45 WMMNSFFP -30 -15 November 2007 0 α º 15 30 45 31 Fire spread with wind R1 U’o L’ Uo R R’ D D’ WMMNSFFP November 2007 32 The Square of Fire: Time Vegetation Meteorology Topography WMMNSFFP November 2007 33 Jan. 2002 DE 2 WMMNSFFP November 2007 34 Δt=20” WMMNSFFP November 2007 35 WMMNSFFP November 2007 36 250 s1 200 20º 150 a=0º a=10º 100 a=20º a=30º 50 a=40º 0 0 WMMNSFFP 100 200 300 400 November 2007 500 600 700 37 16 ds1/dt 14 Rate of spread cm/s Velocidade de propagação 12 10 8 a=0º a=10º 6 a=20º a=30º 20º a=40º 4 2 0 0 WMMNSFFP 100 200 300 November 2007 400 500 600 t (s) 38 700 WMMNSFFP November 2007 39 WMMNSFFP November 2007 40 A m2 1600 25000 1400 Área Area(m2) Perimeter(m) Perímetro 20000 1200 1000 15000 800 600 10000 400 5000 200 0 00:00 WMMNSFFP 05:00 10:00 15:00 20:00 25:00 November 2007 30:00 35:00 0 40:00 41 P m 1800 30000 Mathematical model • Based on the laboratory experiments we developed a mathematical model for this phenomena. • This is a semi-empirical model based on two assumptions that takes into account the feedback between the fire and the surrounding flow. Viegas, 2005 WMMNSFFP November 2007 42 First assumption: • There is a univocal relationship between the surrounding flow velocity and the rate of spread (ROS): • R=f(U) R b1 R' = = 1 + a1U Ro WMMNSFFP November 2007 43 Combustion tunnel Lousã - Portugal WMMNSFFP November 2007 44 R/Ro 25 Wind effect 20 Exp. 15 2R’=1+1.10 U 2.02 10 5 Pine needles 0 0 WMMNSFFP 1 2 November 2007 3 4 U m/s 5 45 Second assumption: • The instantaneous wind velocity is modified by the flow induced by the fire: U=Ua+ΔUf • This induced wind velocity change is univocally related to the Rate of Spread: ΔUf= g(R). dt WMMNSFFP November 2007 46 • The fire induced flow velocity increase during time step dt is given by the following function of ROS: dU = a 2R' .dt b2 WMMNSFFP November 2007 47 Combining the two previous equations we can eliminate the flow velocity variable U between them to obtain a differential of the rate of spread (ROS) as an explicit function of time: df df dR dR = .dU = .g.dt = .dt dU dU dt WMMNSFFP November 2007 48 In the present case the development is the following: b1 R ' = 1 + a1.U ⎡ (R'−1) ⎤ U=⎢ ⎥ ⎣ a1 ⎦ dU = 1 1 1 1b1 ba WMMNSFFP 1 b1 (R'−1) 1 −1 b1 .dR' November 2007 b2 dU = a 2 .R' .dt 49 Differential equation of the model 1 1− b1 1 dR' = a1b1 a 2b1(R'−1) dt R' b2 dR R − Ri = ∫ .dt t i dt t WMMNSFFP November 2007 50 Non-dimensional form • Reference parameters – Ro – to – Uo Basic rate of spread Residence or reaction time Reference wind velocity (Uo=1 m/s) WMMNSFFP November 2007 51 a’1=a1Uob =a1 1 a’2=a2.to R’=R/Ro t’=t/to WMMNSFFP November 2007 52 Non-dimensional equations R’=1+a’1U’b1 dU’=a’2R’b2.dt’ WMMNSFFP November 2007 53 • Combining both equations we obtain: 1 1− b1 1 dR' = a'1b1 a'2 b1(R'−1) dt' 1 dR' = a1b1 a 2b1(R'−1) dt 1− WMMNSFFP November 2007 1 b1 R' R' b2 b2 54 Determination of a’2 and b2 1 −1 b1 dR' (R'−1) Y = a' 2 R ' = . 1 dt' b a' b1 b2 1 WMMNSFFP November 2007 1 55 δ=11º δ=20º δ=32º δ=40º α=0º α=10º α=20º α=30º α=40º Configurations tested in the basic program Viegas and Pita, 2004 WMMNSFFP November 2007 56 Dead litter of Pinus pinaster a1=1.10 b1=2.02 to=80s 0.22 cm/s<Ro<0.34 cm/s WMMNSFFP November 2007 57 Ref. α δ Ref. α δ 501 10º 40º 512 20º 20º 502 20º 40º 513 30º 20º 503 30º 40º 514 40º 20º 506 10º 32º 516 10º 11º 507 20º 32º 517 20º 11º 508 30º 32º 518 30º 11º 509 40º 32º 531 40º 0º 511 10º 20º 532 30º 0º Test cases WMMNSFFP November 2007 Validation 58 0,20 0,15 Outliers 533 518 531 512 514 508 506 502 Y 516 0,10 0,05 0,00 0 5 10 15 R' WMMNSFFP November 2007 59 For pine needles a’2=0. 496 and b2=1.16 Using R’i=1.1 we integrate the differental equation. WMMNSFFP November 2007 60 15 Model 517 519 511 10 513 507 R' 509 501 503 5 0 0 50 100 150 200 250 300 t (s) WMMNSFFP November 2007 61 • Particular case of b1=b2=1 R’=1+a1U dU=a2R’dt dR' = a1a 2 .dt R' R' a1a 2 .t =e R'i WMMNSFFP dR’=a1.dU [ln R'] R' R 'i = a1a 2 .t Exponential growth of ROS November 2007 62 Model parameters Ro Basic rate of spread to Residence time of the fire in the fuelbed a1 Wind law coefficient b1 Wind law exponent a2 Induced velocity coefficient b2 Induced velocity exponent R1 Initial value of the ROS WMMNSFFP November 2007 63 Extension to other fuels • Generalization of the model to other fuel types besides the ones considered in the experimental program. • Field experiments and real cases. WMMNSFFP November 2007 64 Common fuel types • • • • WMMNSFFP Herbaceous Litter Shrubs Slash November 2007 65 Range of variation of the model parameters Herbaceous Litter Shrubs Slash Min Max Min Max Min Max Min Max Ro 0,001 0,01 0,001 0,01 0,001 0,01 0,001 0,01 a1 3 1 1,5 0,5 1 0,2 0,5 0,1 b1 2 3 2 2,5 1,5 2,5 1 1,5 a2 0,001 0,01 0,001 0,01 0,0001 0,001 0,00005 0,0005 b2 1,5 2 1 1,5 1 1,5 0,5 1 t0 10 50 30 100 100 2000 1000 10000 a'1 3 1 1,5 0,5 1 0,2 0,5 0,1 a'2 0,05 0,1 0,1 0,3 0,2 0,1 0,5 0,5 WMMNSFFP November 2007 66 Typical values of the parameters Ro HB 0,005 LT 0,002 SR 0,005 SL 0,001 a1 1,4 1 0,8 0,6 b1 2,3 2,2 2 1,13 a2 0,01 0,0062 0,0005 0,0001 b2 1,5 1,2 1,1 0,8 t0 30 80 1000 5000 a'1 1,4 1 0,8 0,6 a'2 0,3 0,496 0,5 0,5 WMMNSFFP November 2007 67 10,000 HB LT SR SL R (m/s) 1,000 0,100 0,010 0,001 1 10 100 1000 10000 100000 t (sec) Viegas, 2006 WMMNSFFP November 2007 68 1000,00 X (m) 100,00 10,00 1,00 HB LT SR 0,10 SL 0,01 10 100 1000 10000 100000 t (sec) WMMNSFFP November 2007 69 Same fuel with different moisture content values 10,000 R (m/s) 1,000 0,001 0,002 0,005 0,100 0,010 0,001 1 10 100 1000 t (sec) WMMNSFFP November 2007 70 10,000 X (m) 1,000 0,001 0,002 0,005 0,100 0,010 0,001 1 10 100 1000 t (sec) WMMNSFFP November 2007 71 Other models and explanations • Complete mathematical models • Empirical explanations – – – – Topography Fuels Meteorology Miscelaneous WMMNSFFP November 2007 72 • Fuels – Fuel dessication – Thermal belt – Fuel transition WMMNSFFP November 2007 73 • Meteorology – – – – – – – – Passage of a cold front (wind shift) Venturi effect Vertical instability of the atmosphere Low altitude jet stream Daily air temperature variations Air turbulence Air vorticity Colapse of convection column WMMNSFFP November 2007 74 • Miscelaneous – Spot fires – Radiation from one slope to another – Air buble with volatile gases WMMNSFFP November 2007 75 4 - Case Studies • The eruptive fire behaviour is related to many accidents with fatalities that have occurred in forest fires. • As the ROS is initially very low the fire fighters can be mislead by the behaviour of the fire when attacking the fire. WMMNSFFP November 2007 76 • If the suppression of the fire is not achieved in time the fire may erupt and reach their position with great intensity. • Unfortunately this has happened – and continues to happen – too many times in the past. • Brief mention to some case studies. WMMNSFFP November 2007 77 Freixo de Espada-a-Cinta • The year 2003 was particularly bad in terms of forest fires in Portugal. There were 21 persons killed in 18 fire related accidents. • This case occurred in the North of Portugal on the 5th of August of 2003. • Two persons (a couple 50 and 40 years old) were killed in this accident. WMMNSFFP November 2007 78 WMMNSFFP November 2007 79 Meteo station Accident Start of Fire Start of blow-up WMMNSFFP November 2007 80 5-8-03 2 Victims Freixo de Espada à Cinta WMMNSFFP November 2007 81 WMMNSFFP November 2007 82 WMMNSFFP November 2007 83 60 Freixo E C 50 5 Aug 03 Air temperature Temperature 40 30 20 10 0 0:00 WMMNSFFP 3:00 6:00 9:00 12:00 November 2007 15:00 18:00 21:00 0:00 84 70 60 Relative humidity 50 40 30 20 10 0 0:00 3:00 WMMNSFFP 6:00 9:00 12:00 November 2007 15:00 18:00 21:00 0:00 85 360 315 Rumo do vento Wind direction 270 225 180 135 90 45 0 0:00 WMMNSFFP 3:00 6:00 9:00 12:00 November 2007 15:00 18:00 21:00 0:00 86 100 90 80 Average and maximum Wind velocity Wind velocity 70 60 50 40 30 20 10 0 0:00 WMMNSFFP 3:00 6:00 9:00 12:00 November 2007 15:00 18:00 21:00 0:00 87 120 Model prediction 100 Modelo Umax U km/h 80 Uaver 60 40 20 0 0 WMMNSFFP 20 40 60 November 2007 80 100 120 Minutos 88 • This case shows that the interaction between the fire and the surrounding air can modify dramatically both fire behaviour properties and the meteorological conditions. WMMNSFFP November 2007 89 Mann Gulch Fire • USA 1949 • 13 Victims WMMNSFFP November 2007 90 A Race that could not be Won Rothermel, 1990 WMMNSFFP November 2007 91 Loop Fire 1st November 1966 12 Victims WMMNSFFP November 2007 92 Loop Fire under the influence of the Santa Ana wind. Staff Ride WMMNSFFP November 2007 93 South Canyon Fire USA 1994 14 Victims WMMNSFFP November 2007 94 WMMNSFFP November 2007 95 1200 South Canyon R' 1000 800 Observações 600 400 Modelo presente 200 Ro=0.166 cm/s 0 0 WMMNSFFP 300 600 900 November 2007 1200 1500 1800 96 t (s) WMMNSFFP November 2007 97 WMMNSFFP November 2007 98 Thirtymile Fire 10 July 2001 4 victims WMMNSFFP November 2007 99 160 140 Thirty Mile Accident 120 R' 100 80 Model 60 Observ 40 20 0 0 1 2 3 4 5 6 7 8 t (h) WMMNSFFP November 2007 100 Guadalajara • This accident occurred on the 17th July 2005. • 11 Firefighters were killed by a double fire eruption. • This accident raised great concerns in Spain and is promoting many changes. WMMNSFFP November 2007 101 Guadalajara July 2005 WMMNSFFP November 2007 102 Guadalajara 17 de Julho de 2005 WMMNSFFP November 2007 103 WMMNSFFP November 2007 104 WMMNSFFP November 2007 105 Famalicão • This accident occurred in Famalicão da Serra (Guarda- Portugal) on the 9th July 2006. • Six fire fighters were killed in this accident. Five of them were Chilean citizens working as professionals in fire suppression in Portugal. WMMNSFFP November 2007 106 Famalicão da Serra Julho de 2006 WMMNSFFP November 2007 107 WMMNSFFP November 2007 108 Lição 7 Casos de Estudo 2 109 6 - Conclusion • Eruptive fire behaviour is relatively common in complex terrain and it is associated to many fatal accidents in forest fires. • The sudden acceleration of the fire front can be explained by the convective flow induced by the fire itself. WMMNSFFP November 2007 110 • A simple mathematical model was proposed to explain and predict eruptive fire behaviour. • The parameters of the model can be determined from experiments or from field data. • The model can be applied to a wide range of situations. WMMNSFFP November 2007 111 • I have the feeling that this is a piece of fire behaviour science that everyone dealing with forest fires should know. • Yet we still find persons that know very much about fires but are not aware of this. WMMNSFFP November 2007 112 WMMNSFFP November 2007 113
Similar documents
Shock Absorbers
PM(XT) Technical Data, Accessories and Sizing Curves . . . . . . . . . . . . . . . . . . . . . . 46-60 PRO Technical Data, Accessories and Sizing Curves . . . . . . . . . . . . . . . . . . . . . . ...
More information