Upload
others
View
0
Download
0
Embed Size (px)
Citation preview
824548
I t lfi-l 13 1 0 5 rlt U
--
-
ThG unedited intcncl note rccords the author B ViLJ for p~UpOSGB of
xcyi ~ v (ud CCtalllGllt This note d)06 net clIr r y the uthoX-ity of the Scicntific
bull
R HlJSTPN H J DAR~~--S
JUNpound 197amp ) I
bull
bull
bull
Ti2 noto d~8crib()13 n proposed modol lhich Hy be uGed to 1epl csent the
rc1ntionBhip bcten the G1odth of n fought fir Hnd tlw nnler of jeto hich
0 flpplied rho model can be used to predict the effect et applyine jcto at
()i~fcrent times and hence can be u6cd in the 6tlldy of thc nlOnber of Dpplanc~G
rcqlired at caell fire atation
ll8 model Io dermiddotivcd b~ conGidcrine 3impe E0thcmaticul form of the artgt
gimiddotth curve hich can be used to d~8cribe the observed bc)ioUl of 1(81
fi18( and which arc comptiblt Iith tn6 observed charnctel5tics of exp-imenO
firos Tllo different forms of 8 oyarhatic g1olth model al0 derived cnd Game
eS~iLlates of the coefficientn are given
~h- model desclibed her s deteri1l)n16tic cnd it lemaillG to investigate tha
propertics of DJl 1-quivalent stochastic model and to find SOn0 means of vcl idutirmiddots
01 displpving this proposed model bull
bull
A PROPOSM) MODEL FOH IIIE GH(J OF FUUCHT FIRE
Contentsshy
1 In troduction
2 Obsclvlltione from the DDto
4 Thooretical Hodol of Firo Grmmiddotth
6 EsUmates of the Parllmotcrs of the Conctullt ~ioratic Hodd
7 Alternativo Forms of the Qaadratic Il ooel
8 The Uso of the Quadratic Ilodcl
9 ]imitations of the Data
10 Alternative Methods of Estimating th~ ParCJTleters
11 Further Workmiddot
IHE GP-OH1Il OF A FOUGHT Fl[
1 II1rnOllUCrION
In 01 study of the numbJx of Plllf per GcCtion 0) a otudy of the filst attendance
rules it io eGsentiJl to have uu UlldcrGicndinG of the Hay in which fires amprolt
after the arival of the first pump ThLo Horldne noto iD a deoc6ptiol1 of a
proposed modol
Thin model is ~pplicle to buEdinC fir ~ c IIhich lulte Bprend beyond ch room
of ori-gin
I 2 OBSEllVATIONS FHOH THE DAfA
I I
AI1Y model which rcplaquosents the [oOltth of a fought filo must be copable of
exploining the foJloHb effects 11hich arc ob6cnuc in the 5AF2 data
i ~he middotaverace amount of Bpread for fir8 IIhich amp10 of tll~ snIDe nile Rt
arriVD1 middotincreDsell with th(l number of jets UGeurod This is due to the
differing fiercenccs of the fires An examination of the number of jets
UBed ugDinst the iinrl oren elso shown that spjcndinl3 fires use molS jeto
ii The proportion of fires which spread after Dlli val increases middotd th the
size of fire at 8rrival
iii But the average Mount of Gpread (for opreading and non-spreadJng
fires) is approxima tely constant for all sizES of fire
iv ii and Hi imply that for fires gthich spread tho average spread
(AfArgt ) io leGS for larGe fires
v Fires for which all the pumps errive srnuJutnncolsly are still likely
to spread nfter arrival
vi The conttol time (as recorcied on the K433 f orms) increuses ith
increaGing final urea
3 RFl3RESSION ANALYSIS
An onalJsis of the rclRtionship neblcen the ollnt f d d ~h t lth spree en le lm~s il fo
1
~h the jotc Hero vppllcd HOulcl PlOVC the oimplcst )KmiddotCi blo model of
or the follming eGGumptions
(j) Th rate of cromiddotth of firec is cxponcntiel
(ii) Each pump oupplicd 001) jet
(iii) Thc arrival timeG of the pumps are the times a c hich hc jete
f le
rlI model UGccl to reprctlcnt the grolth of a j jet fire as A ~ Ao e
rot~ jtpound1 -0 V j- ~OOCl~ e J
Hh0n Ap- is the oregtt dllmagccl Dtext in ction
Ao is the alea dam~ecl at arril8l
f is the exp(luentiRl growth rata
ti~1 le the i-1 th interarrival time ie the interval beteen the arrival
of tho itll and (i_1)th pUOlp k e represents finy croHth not ficcounted for by intc)$rrivlll groth
This regresGion analysiB RS tried on multi-storey non-chJellil1g firen l-Ihicb hUo
spraad bGyond the room of ori(in
1 jGt Ures Af 111f Hean Ar- 111
(n =598) AQ Ao
2 jet Ures Ap ooltt1 Hean Af ~ 116 116(n 366) A 0 Aa
All the R v81u(~ Hera very low and the regression coampfficiGnts woro -hot sienifcn
3 jet firos (n 11+6)
4 jet fires (n 95)
5 jot fires (n 42)
or only sigliificaut at the 10 level Host of the spread is cccounted fo by
dclfy in lntcrnrrival timer l1j)U providEl8 on order o ~l lnW t itn rj- ecirlto of
dOOG not provide u reasonable rD oil3 for D mod(Il of fh0 ijr-o-rcli(o
4 A THlWUhTICAL HCDFJ OF FIRE GliOI-lTH
If the firc size is roduced by applying Iter to tll p imetcr of the Hre
the tillle tden to extinguish the fire gtlDl be P)~OPGlltmiddot io middot i to At hel A is
the size of the fir( (Se Bsldlir 11olinok ThoCilflO Fire Re~car(h Note No 8811)
ell - at ob (42)dt
A _ AD - ~ 2 + bt (43) 2
The only inforruation El have about a fire iB tho urea at alrival (Ao) the
arell damcged (A) and the extinction tile (t ) The information defiueD twoE
points nnd a tangent on the graph of fire sizI] against time The quadratic
equation 11 42 43 is the simplest flUlction which ill fit a curve defiMd
by threo points
t -gt
EqlUltion 42 6h0l8 that the maximum a)ea i8 reached at time t ba a[tol
arrival This maxirutlIa fire size ill b thlt ni t damDged at extinction
3
lUU tbo ebsolut u DOlOunt ()f
rhe extinctioi1 ti(o js the tilT) c1 lhich 0
ba
a detGrmineB hOil qui0kly the fie si7o iD roduced 8 wUl c~epend on ~middothc
char[cteristic[J of the firo (ie Bnd filtrceu~Gs ) and lIil nlso reflec t the
nIgtJnba of j ~ tB uGed to figt 1ho firc As more j~t5 are apoJied the fire GheI
Hill be rGduccd or e quickly and the pa1rgtJueter a llill iucrcace
The parameter b represents the colltinucd grogtlth of the firo UFo it it being
fouGht If tho fire is not being fought a must hamp zero 81ulto Md the fir
groVlth will be dAdt b 0lthotlSh b ItlPy not be constant in gtIhich CO6 b is a tanGent to the grogtth curve at tho point A h2dAdt
Considet first tile 5impleot GittoCion ~hcre 1J1l thu pUlps middothich ay bo needed
arc available at tho time of the fhst alltrivnl and [lS5ume the parameters lgt and b
are constantt
The firEls HHl be charactcrised by their inItial 5izG Ao and their fierceness
The brigade in apply the appropiintc number of jets accolding to their
lisspoundsment of the fire size Dnd severity Fires of tho Sml0 initial size
Ao Hill folloll different groHth curvet according to their severity
As the severity increascs the ~ount of spread and control tiGe also incr~eso
Yo- a ghcn Ao [0 tmiddotJle fJc1rc6 ty incrcaceo the paXarncteZmiddot Cl will accrclt)cc
MCor the parumetcL b middotill incrcace
It 5s u66umed thampt all th~ dHfcrcncf)s n the cromiddot th of th6 fhcc ll 0 due
to the differences bchloen fires He a60UC1( thnt the brlgad Dpp1y tIle right
number of jctll and do net allovl for a fercc effoct by 6uGcel3ting that th
fhmiddotc may be put out more guielly by usinf more jet
5 A middotLfHOD OF ESTIHATING TIlE PAllAJmiddot~EW~S 1011 AN AVERAGE FIRE
~ ue still asouraLng that the puametels a und b a rc COllGtents and conBiceriq
onJy those fireG for Hhieh all the pumpG 10)C availo1gtle Ilimul tallcouGly If
take o eroup of fil06 of the BMa initial Gize Ao fought by the GBruG num0T
of jets He can observe avcraee values for the sprpad and control timo ftom
thr SAF2 data If Wc flUther aSGume tbat thil group of fires IJaG the same
(or at 16ilGt simila) Il and b 1uGO then Ie can use thu hlo obs epved avcr6r~B
to eShD1ilto the two porameters
2-
A rrrangcmcllt of the cqttation6 AI - Ao b 2a nnd tE ba +Ga giVeuro6
a ~ z[fiL~JAF J2
b J 2ak
2where l~ il3 the absolute amow1t of spreild) b 2a
If e can make soma assumption about the ay in vhich u changes RG edditioncl
jets are brought into we then can generalise the growth equation to include
thoGe fircl3 in lIhich there wae a delay in the arrival of the second and subsequent c-t-b
jets In a study of wood ~ firos (0 Donghcrty and Young Fire Research Note e
No 603) it HUS fOWld thut the maxiGm] amount of fire spreld IWS propottional lo
) -121 ( Bate of Hater Application bull If this result is used it Iltould irnplJ th~t
5
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
bull
bull
bull
Ti2 noto d~8crib()13 n proposed modol lhich Hy be uGed to 1epl csent the
rc1ntionBhip bcten the G1odth of n fought fir Hnd tlw nnler of jeto hich
0 flpplied rho model can be used to predict the effect et applyine jcto at
()i~fcrent times and hence can be u6cd in the 6tlldy of thc nlOnber of Dpplanc~G
rcqlired at caell fire atation
ll8 model Io dermiddotivcd b~ conGidcrine 3impe E0thcmaticul form of the artgt
gimiddotth curve hich can be used to d~8cribe the observed bc)ioUl of 1(81
fi18( and which arc comptiblt Iith tn6 observed charnctel5tics of exp-imenO
firos Tllo different forms of 8 oyarhatic g1olth model al0 derived cnd Game
eS~iLlates of the coefficientn are given
~h- model desclibed her s deteri1l)n16tic cnd it lemaillG to investigate tha
propertics of DJl 1-quivalent stochastic model and to find SOn0 means of vcl idutirmiddots
01 displpving this proposed model bull
bull
A PROPOSM) MODEL FOH IIIE GH(J OF FUUCHT FIRE
Contentsshy
1 In troduction
2 Obsclvlltione from the DDto
4 Thooretical Hodol of Firo Grmmiddotth
6 EsUmates of the Parllmotcrs of the Conctullt ~ioratic Hodd
7 Alternativo Forms of the Qaadratic Il ooel
8 The Uso of the Quadratic Ilodcl
9 ]imitations of the Data
10 Alternative Methods of Estimating th~ ParCJTleters
11 Further Workmiddot
IHE GP-OH1Il OF A FOUGHT Fl[
1 II1rnOllUCrION
In 01 study of the numbJx of Plllf per GcCtion 0) a otudy of the filst attendance
rules it io eGsentiJl to have uu UlldcrGicndinG of the Hay in which fires amprolt
after the arival of the first pump ThLo Horldne noto iD a deoc6ptiol1 of a
proposed modol
Thin model is ~pplicle to buEdinC fir ~ c IIhich lulte Bprend beyond ch room
of ori-gin
I 2 OBSEllVATIONS FHOH THE DAfA
I I
AI1Y model which rcplaquosents the [oOltth of a fought filo must be copable of
exploining the foJloHb effects 11hich arc ob6cnuc in the 5AF2 data
i ~he middotaverace amount of Bpread for fir8 IIhich amp10 of tll~ snIDe nile Rt
arriVD1 middotincreDsell with th(l number of jets UGeurod This is due to the
differing fiercenccs of the fires An examination of the number of jets
UBed ugDinst the iinrl oren elso shown that spjcndinl3 fires use molS jeto
ii The proportion of fires which spread after Dlli val increases middotd th the
size of fire at 8rrival
iii But the average Mount of Gpread (for opreading and non-spreadJng
fires) is approxima tely constant for all sizES of fire
iv ii and Hi imply that for fires gthich spread tho average spread
(AfArgt ) io leGS for larGe fires
v Fires for which all the pumps errive srnuJutnncolsly are still likely
to spread nfter arrival
vi The conttol time (as recorcied on the K433 f orms) increuses ith
increaGing final urea
3 RFl3RESSION ANALYSIS
An onalJsis of the rclRtionship neblcen the ollnt f d d ~h t lth spree en le lm~s il fo
1
~h the jotc Hero vppllcd HOulcl PlOVC the oimplcst )KmiddotCi blo model of
or the follming eGGumptions
(j) Th rate of cromiddotth of firec is cxponcntiel
(ii) Each pump oupplicd 001) jet
(iii) Thc arrival timeG of the pumps are the times a c hich hc jete
f le
rlI model UGccl to reprctlcnt the grolth of a j jet fire as A ~ Ao e
rot~ jtpound1 -0 V j- ~OOCl~ e J
Hh0n Ap- is the oregtt dllmagccl Dtext in ction
Ao is the alea dam~ecl at arril8l
f is the exp(luentiRl growth rata
ti~1 le the i-1 th interarrival time ie the interval beteen the arrival
of tho itll and (i_1)th pUOlp k e represents finy croHth not ficcounted for by intc)$rrivlll groth
This regresGion analysiB RS tried on multi-storey non-chJellil1g firen l-Ihicb hUo
spraad bGyond the room of ori(in
1 jGt Ures Af 111f Hean Ar- 111
(n =598) AQ Ao
2 jet Ures Ap ooltt1 Hean Af ~ 116 116(n 366) A 0 Aa
All the R v81u(~ Hera very low and the regression coampfficiGnts woro -hot sienifcn
3 jet firos (n 11+6)
4 jet fires (n 95)
5 jot fires (n 42)
or only sigliificaut at the 10 level Host of the spread is cccounted fo by
dclfy in lntcrnrrival timer l1j)U providEl8 on order o ~l lnW t itn rj- ecirlto of
dOOG not provide u reasonable rD oil3 for D mod(Il of fh0 ijr-o-rcli(o
4 A THlWUhTICAL HCDFJ OF FIRE GliOI-lTH
If the firc size is roduced by applying Iter to tll p imetcr of the Hre
the tillle tden to extinguish the fire gtlDl be P)~OPGlltmiddot io middot i to At hel A is
the size of the fir( (Se Bsldlir 11olinok ThoCilflO Fire Re~car(h Note No 8811)
ell - at ob (42)dt
A _ AD - ~ 2 + bt (43) 2
The only inforruation El have about a fire iB tho urea at alrival (Ao) the
arell damcged (A) and the extinction tile (t ) The information defiueD twoE
points nnd a tangent on the graph of fire sizI] against time The quadratic
equation 11 42 43 is the simplest flUlction which ill fit a curve defiMd
by threo points
t -gt
EqlUltion 42 6h0l8 that the maximum a)ea i8 reached at time t ba a[tol
arrival This maxirutlIa fire size ill b thlt ni t damDged at extinction
3
lUU tbo ebsolut u DOlOunt ()f
rhe extinctioi1 ti(o js the tilT) c1 lhich 0
ba
a detGrmineB hOil qui0kly the fie si7o iD roduced 8 wUl c~epend on ~middothc
char[cteristic[J of the firo (ie Bnd filtrceu~Gs ) and lIil nlso reflec t the
nIgtJnba of j ~ tB uGed to figt 1ho firc As more j~t5 are apoJied the fire GheI
Hill be rGduccd or e quickly and the pa1rgtJueter a llill iucrcace
The parameter b represents the colltinucd grogtlth of the firo UFo it it being
fouGht If tho fire is not being fought a must hamp zero 81ulto Md the fir
groVlth will be dAdt b 0lthotlSh b ItlPy not be constant in gtIhich CO6 b is a tanGent to the grogtth curve at tho point A h2dAdt
Considet first tile 5impleot GittoCion ~hcre 1J1l thu pUlps middothich ay bo needed
arc available at tho time of the fhst alltrivnl and [lS5ume the parameters lgt and b
are constantt
The firEls HHl be charactcrised by their inItial 5izG Ao and their fierceness
The brigade in apply the appropiintc number of jets accolding to their
lisspoundsment of the fire size Dnd severity Fires of tho Sml0 initial size
Ao Hill folloll different groHth curvet according to their severity
As the severity increascs the ~ount of spread and control tiGe also incr~eso
Yo- a ghcn Ao [0 tmiddotJle fJc1rc6 ty incrcaceo the paXarncteZmiddot Cl will accrclt)cc
MCor the parumetcL b middotill incrcace
It 5s u66umed thampt all th~ dHfcrcncf)s n the cromiddot th of th6 fhcc ll 0 due
to the differences bchloen fires He a60UC1( thnt the brlgad Dpp1y tIle right
number of jctll and do net allovl for a fercc effoct by 6uGcel3ting that th
fhmiddotc may be put out more guielly by usinf more jet
5 A middotLfHOD OF ESTIHATING TIlE PAllAJmiddot~EW~S 1011 AN AVERAGE FIRE
~ ue still asouraLng that the puametels a und b a rc COllGtents and conBiceriq
onJy those fireG for Hhieh all the pumpG 10)C availo1gtle Ilimul tallcouGly If
take o eroup of fil06 of the BMa initial Gize Ao fought by the GBruG num0T
of jets He can observe avcraee values for the sprpad and control timo ftom
thr SAF2 data If Wc flUther aSGume tbat thil group of fires IJaG the same
(or at 16ilGt simila) Il and b 1uGO then Ie can use thu hlo obs epved avcr6r~B
to eShD1ilto the two porameters
2-
A rrrangcmcllt of the cqttation6 AI - Ao b 2a nnd tE ba +Ga giVeuro6
a ~ z[fiL~JAF J2
b J 2ak
2where l~ il3 the absolute amow1t of spreild) b 2a
If e can make soma assumption about the ay in vhich u changes RG edditioncl
jets are brought into we then can generalise the growth equation to include
thoGe fircl3 in lIhich there wae a delay in the arrival of the second and subsequent c-t-b
jets In a study of wood ~ firos (0 Donghcrty and Young Fire Research Note e
No 603) it HUS fOWld thut the maxiGm] amount of fire spreld IWS propottional lo
) -121 ( Bate of Hater Application bull If this result is used it Iltould irnplJ th~t
5
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
Ti2 noto d~8crib()13 n proposed modol lhich Hy be uGed to 1epl csent the
rc1ntionBhip bcten the G1odth of n fought fir Hnd tlw nnler of jeto hich
0 flpplied rho model can be used to predict the effect et applyine jcto at
()i~fcrent times and hence can be u6cd in the 6tlldy of thc nlOnber of Dpplanc~G
rcqlired at caell fire atation
ll8 model Io dermiddotivcd b~ conGidcrine 3impe E0thcmaticul form of the artgt
gimiddotth curve hich can be used to d~8cribe the observed bc)ioUl of 1(81
fi18( and which arc comptiblt Iith tn6 observed charnctel5tics of exp-imenO
firos Tllo different forms of 8 oyarhatic g1olth model al0 derived cnd Game
eS~iLlates of the coefficientn are given
~h- model desclibed her s deteri1l)n16tic cnd it lemaillG to investigate tha
propertics of DJl 1-quivalent stochastic model and to find SOn0 means of vcl idutirmiddots
01 displpving this proposed model bull
bull
A PROPOSM) MODEL FOH IIIE GH(J OF FUUCHT FIRE
Contentsshy
1 In troduction
2 Obsclvlltione from the DDto
4 Thooretical Hodol of Firo Grmmiddotth
6 EsUmates of the Parllmotcrs of the Conctullt ~ioratic Hodd
7 Alternativo Forms of the Qaadratic Il ooel
8 The Uso of the Quadratic Ilodcl
9 ]imitations of the Data
10 Alternative Methods of Estimating th~ ParCJTleters
11 Further Workmiddot
IHE GP-OH1Il OF A FOUGHT Fl[
1 II1rnOllUCrION
In 01 study of the numbJx of Plllf per GcCtion 0) a otudy of the filst attendance
rules it io eGsentiJl to have uu UlldcrGicndinG of the Hay in which fires amprolt
after the arival of the first pump ThLo Horldne noto iD a deoc6ptiol1 of a
proposed modol
Thin model is ~pplicle to buEdinC fir ~ c IIhich lulte Bprend beyond ch room
of ori-gin
I 2 OBSEllVATIONS FHOH THE DAfA
I I
AI1Y model which rcplaquosents the [oOltth of a fought filo must be copable of
exploining the foJloHb effects 11hich arc ob6cnuc in the 5AF2 data
i ~he middotaverace amount of Bpread for fir8 IIhich amp10 of tll~ snIDe nile Rt
arriVD1 middotincreDsell with th(l number of jets UGeurod This is due to the
differing fiercenccs of the fires An examination of the number of jets
UBed ugDinst the iinrl oren elso shown that spjcndinl3 fires use molS jeto
ii The proportion of fires which spread after Dlli val increases middotd th the
size of fire at 8rrival
iii But the average Mount of Gpread (for opreading and non-spreadJng
fires) is approxima tely constant for all sizES of fire
iv ii and Hi imply that for fires gthich spread tho average spread
(AfArgt ) io leGS for larGe fires
v Fires for which all the pumps errive srnuJutnncolsly are still likely
to spread nfter arrival
vi The conttol time (as recorcied on the K433 f orms) increuses ith
increaGing final urea
3 RFl3RESSION ANALYSIS
An onalJsis of the rclRtionship neblcen the ollnt f d d ~h t lth spree en le lm~s il fo
1
~h the jotc Hero vppllcd HOulcl PlOVC the oimplcst )KmiddotCi blo model of
or the follming eGGumptions
(j) Th rate of cromiddotth of firec is cxponcntiel
(ii) Each pump oupplicd 001) jet
(iii) Thc arrival timeG of the pumps are the times a c hich hc jete
f le
rlI model UGccl to reprctlcnt the grolth of a j jet fire as A ~ Ao e
rot~ jtpound1 -0 V j- ~OOCl~ e J
Hh0n Ap- is the oregtt dllmagccl Dtext in ction
Ao is the alea dam~ecl at arril8l
f is the exp(luentiRl growth rata
ti~1 le the i-1 th interarrival time ie the interval beteen the arrival
of tho itll and (i_1)th pUOlp k e represents finy croHth not ficcounted for by intc)$rrivlll groth
This regresGion analysiB RS tried on multi-storey non-chJellil1g firen l-Ihicb hUo
spraad bGyond the room of ori(in
1 jGt Ures Af 111f Hean Ar- 111
(n =598) AQ Ao
2 jet Ures Ap ooltt1 Hean Af ~ 116 116(n 366) A 0 Aa
All the R v81u(~ Hera very low and the regression coampfficiGnts woro -hot sienifcn
3 jet firos (n 11+6)
4 jet fires (n 95)
5 jot fires (n 42)
or only sigliificaut at the 10 level Host of the spread is cccounted fo by
dclfy in lntcrnrrival timer l1j)U providEl8 on order o ~l lnW t itn rj- ecirlto of
dOOG not provide u reasonable rD oil3 for D mod(Il of fh0 ijr-o-rcli(o
4 A THlWUhTICAL HCDFJ OF FIRE GliOI-lTH
If the firc size is roduced by applying Iter to tll p imetcr of the Hre
the tillle tden to extinguish the fire gtlDl be P)~OPGlltmiddot io middot i to At hel A is
the size of the fir( (Se Bsldlir 11olinok ThoCilflO Fire Re~car(h Note No 8811)
ell - at ob (42)dt
A _ AD - ~ 2 + bt (43) 2
The only inforruation El have about a fire iB tho urea at alrival (Ao) the
arell damcged (A) and the extinction tile (t ) The information defiueD twoE
points nnd a tangent on the graph of fire sizI] against time The quadratic
equation 11 42 43 is the simplest flUlction which ill fit a curve defiMd
by threo points
t -gt
EqlUltion 42 6h0l8 that the maximum a)ea i8 reached at time t ba a[tol
arrival This maxirutlIa fire size ill b thlt ni t damDged at extinction
3
lUU tbo ebsolut u DOlOunt ()f
rhe extinctioi1 ti(o js the tilT) c1 lhich 0
ba
a detGrmineB hOil qui0kly the fie si7o iD roduced 8 wUl c~epend on ~middothc
char[cteristic[J of the firo (ie Bnd filtrceu~Gs ) and lIil nlso reflec t the
nIgtJnba of j ~ tB uGed to figt 1ho firc As more j~t5 are apoJied the fire GheI
Hill be rGduccd or e quickly and the pa1rgtJueter a llill iucrcace
The parameter b represents the colltinucd grogtlth of the firo UFo it it being
fouGht If tho fire is not being fought a must hamp zero 81ulto Md the fir
groVlth will be dAdt b 0lthotlSh b ItlPy not be constant in gtIhich CO6 b is a tanGent to the grogtth curve at tho point A h2dAdt
Considet first tile 5impleot GittoCion ~hcre 1J1l thu pUlps middothich ay bo needed
arc available at tho time of the fhst alltrivnl and [lS5ume the parameters lgt and b
are constantt
The firEls HHl be charactcrised by their inItial 5izG Ao and their fierceness
The brigade in apply the appropiintc number of jets accolding to their
lisspoundsment of the fire size Dnd severity Fires of tho Sml0 initial size
Ao Hill folloll different groHth curvet according to their severity
As the severity increascs the ~ount of spread and control tiGe also incr~eso
Yo- a ghcn Ao [0 tmiddotJle fJc1rc6 ty incrcaceo the paXarncteZmiddot Cl will accrclt)cc
MCor the parumetcL b middotill incrcace
It 5s u66umed thampt all th~ dHfcrcncf)s n the cromiddot th of th6 fhcc ll 0 due
to the differences bchloen fires He a60UC1( thnt the brlgad Dpp1y tIle right
number of jctll and do net allovl for a fercc effoct by 6uGcel3ting that th
fhmiddotc may be put out more guielly by usinf more jet
5 A middotLfHOD OF ESTIHATING TIlE PAllAJmiddot~EW~S 1011 AN AVERAGE FIRE
~ ue still asouraLng that the puametels a und b a rc COllGtents and conBiceriq
onJy those fireG for Hhieh all the pumpG 10)C availo1gtle Ilimul tallcouGly If
take o eroup of fil06 of the BMa initial Gize Ao fought by the GBruG num0T
of jets He can observe avcraee values for the sprpad and control timo ftom
thr SAF2 data If Wc flUther aSGume tbat thil group of fires IJaG the same
(or at 16ilGt simila) Il and b 1uGO then Ie can use thu hlo obs epved avcr6r~B
to eShD1ilto the two porameters
2-
A rrrangcmcllt of the cqttation6 AI - Ao b 2a nnd tE ba +Ga giVeuro6
a ~ z[fiL~JAF J2
b J 2ak
2where l~ il3 the absolute amow1t of spreild) b 2a
If e can make soma assumption about the ay in vhich u changes RG edditioncl
jets are brought into we then can generalise the growth equation to include
thoGe fircl3 in lIhich there wae a delay in the arrival of the second and subsequent c-t-b
jets In a study of wood ~ firos (0 Donghcrty and Young Fire Research Note e
No 603) it HUS fOWld thut the maxiGm] amount of fire spreld IWS propottional lo
) -121 ( Bate of Hater Application bull If this result is used it Iltould irnplJ th~t
5
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
A PROPOSM) MODEL FOH IIIE GH(J OF FUUCHT FIRE
Contentsshy
1 In troduction
2 Obsclvlltione from the DDto
4 Thooretical Hodol of Firo Grmmiddotth
6 EsUmates of the Parllmotcrs of the Conctullt ~ioratic Hodd
7 Alternativo Forms of the Qaadratic Il ooel
8 The Uso of the Quadratic Ilodcl
9 ]imitations of the Data
10 Alternative Methods of Estimating th~ ParCJTleters
11 Further Workmiddot
IHE GP-OH1Il OF A FOUGHT Fl[
1 II1rnOllUCrION
In 01 study of the numbJx of Plllf per GcCtion 0) a otudy of the filst attendance
rules it io eGsentiJl to have uu UlldcrGicndinG of the Hay in which fires amprolt
after the arival of the first pump ThLo Horldne noto iD a deoc6ptiol1 of a
proposed modol
Thin model is ~pplicle to buEdinC fir ~ c IIhich lulte Bprend beyond ch room
of ori-gin
I 2 OBSEllVATIONS FHOH THE DAfA
I I
AI1Y model which rcplaquosents the [oOltth of a fought filo must be copable of
exploining the foJloHb effects 11hich arc ob6cnuc in the 5AF2 data
i ~he middotaverace amount of Bpread for fir8 IIhich amp10 of tll~ snIDe nile Rt
arriVD1 middotincreDsell with th(l number of jets UGeurod This is due to the
differing fiercenccs of the fires An examination of the number of jets
UBed ugDinst the iinrl oren elso shown that spjcndinl3 fires use molS jeto
ii The proportion of fires which spread after Dlli val increases middotd th the
size of fire at 8rrival
iii But the average Mount of Gpread (for opreading and non-spreadJng
fires) is approxima tely constant for all sizES of fire
iv ii and Hi imply that for fires gthich spread tho average spread
(AfArgt ) io leGS for larGe fires
v Fires for which all the pumps errive srnuJutnncolsly are still likely
to spread nfter arrival
vi The conttol time (as recorcied on the K433 f orms) increuses ith
increaGing final urea
3 RFl3RESSION ANALYSIS
An onalJsis of the rclRtionship neblcen the ollnt f d d ~h t lth spree en le lm~s il fo
1
~h the jotc Hero vppllcd HOulcl PlOVC the oimplcst )KmiddotCi blo model of
or the follming eGGumptions
(j) Th rate of cromiddotth of firec is cxponcntiel
(ii) Each pump oupplicd 001) jet
(iii) Thc arrival timeG of the pumps are the times a c hich hc jete
f le
rlI model UGccl to reprctlcnt the grolth of a j jet fire as A ~ Ao e
rot~ jtpound1 -0 V j- ~OOCl~ e J
Hh0n Ap- is the oregtt dllmagccl Dtext in ction
Ao is the alea dam~ecl at arril8l
f is the exp(luentiRl growth rata
ti~1 le the i-1 th interarrival time ie the interval beteen the arrival
of tho itll and (i_1)th pUOlp k e represents finy croHth not ficcounted for by intc)$rrivlll groth
This regresGion analysiB RS tried on multi-storey non-chJellil1g firen l-Ihicb hUo
spraad bGyond the room of ori(in
1 jGt Ures Af 111f Hean Ar- 111
(n =598) AQ Ao
2 jet Ures Ap ooltt1 Hean Af ~ 116 116(n 366) A 0 Aa
All the R v81u(~ Hera very low and the regression coampfficiGnts woro -hot sienifcn
3 jet firos (n 11+6)
4 jet fires (n 95)
5 jot fires (n 42)
or only sigliificaut at the 10 level Host of the spread is cccounted fo by
dclfy in lntcrnrrival timer l1j)U providEl8 on order o ~l lnW t itn rj- ecirlto of
dOOG not provide u reasonable rD oil3 for D mod(Il of fh0 ijr-o-rcli(o
4 A THlWUhTICAL HCDFJ OF FIRE GliOI-lTH
If the firc size is roduced by applying Iter to tll p imetcr of the Hre
the tillle tden to extinguish the fire gtlDl be P)~OPGlltmiddot io middot i to At hel A is
the size of the fir( (Se Bsldlir 11olinok ThoCilflO Fire Re~car(h Note No 8811)
ell - at ob (42)dt
A _ AD - ~ 2 + bt (43) 2
The only inforruation El have about a fire iB tho urea at alrival (Ao) the
arell damcged (A) and the extinction tile (t ) The information defiueD twoE
points nnd a tangent on the graph of fire sizI] against time The quadratic
equation 11 42 43 is the simplest flUlction which ill fit a curve defiMd
by threo points
t -gt
EqlUltion 42 6h0l8 that the maximum a)ea i8 reached at time t ba a[tol
arrival This maxirutlIa fire size ill b thlt ni t damDged at extinction
3
lUU tbo ebsolut u DOlOunt ()f
rhe extinctioi1 ti(o js the tilT) c1 lhich 0
ba
a detGrmineB hOil qui0kly the fie si7o iD roduced 8 wUl c~epend on ~middothc
char[cteristic[J of the firo (ie Bnd filtrceu~Gs ) and lIil nlso reflec t the
nIgtJnba of j ~ tB uGed to figt 1ho firc As more j~t5 are apoJied the fire GheI
Hill be rGduccd or e quickly and the pa1rgtJueter a llill iucrcace
The parameter b represents the colltinucd grogtlth of the firo UFo it it being
fouGht If tho fire is not being fought a must hamp zero 81ulto Md the fir
groVlth will be dAdt b 0lthotlSh b ItlPy not be constant in gtIhich CO6 b is a tanGent to the grogtth curve at tho point A h2dAdt
Considet first tile 5impleot GittoCion ~hcre 1J1l thu pUlps middothich ay bo needed
arc available at tho time of the fhst alltrivnl and [lS5ume the parameters lgt and b
are constantt
The firEls HHl be charactcrised by their inItial 5izG Ao and their fierceness
The brigade in apply the appropiintc number of jets accolding to their
lisspoundsment of the fire size Dnd severity Fires of tho Sml0 initial size
Ao Hill folloll different groHth curvet according to their severity
As the severity increascs the ~ount of spread and control tiGe also incr~eso
Yo- a ghcn Ao [0 tmiddotJle fJc1rc6 ty incrcaceo the paXarncteZmiddot Cl will accrclt)cc
MCor the parumetcL b middotill incrcace
It 5s u66umed thampt all th~ dHfcrcncf)s n the cromiddot th of th6 fhcc ll 0 due
to the differences bchloen fires He a60UC1( thnt the brlgad Dpp1y tIle right
number of jctll and do net allovl for a fercc effoct by 6uGcel3ting that th
fhmiddotc may be put out more guielly by usinf more jet
5 A middotLfHOD OF ESTIHATING TIlE PAllAJmiddot~EW~S 1011 AN AVERAGE FIRE
~ ue still asouraLng that the puametels a und b a rc COllGtents and conBiceriq
onJy those fireG for Hhieh all the pumpG 10)C availo1gtle Ilimul tallcouGly If
take o eroup of fil06 of the BMa initial Gize Ao fought by the GBruG num0T
of jets He can observe avcraee values for the sprpad and control timo ftom
thr SAF2 data If Wc flUther aSGume tbat thil group of fires IJaG the same
(or at 16ilGt simila) Il and b 1uGO then Ie can use thu hlo obs epved avcr6r~B
to eShD1ilto the two porameters
2-
A rrrangcmcllt of the cqttation6 AI - Ao b 2a nnd tE ba +Ga giVeuro6
a ~ z[fiL~JAF J2
b J 2ak
2where l~ il3 the absolute amow1t of spreild) b 2a
If e can make soma assumption about the ay in vhich u changes RG edditioncl
jets are brought into we then can generalise the growth equation to include
thoGe fircl3 in lIhich there wae a delay in the arrival of the second and subsequent c-t-b
jets In a study of wood ~ firos (0 Donghcrty and Young Fire Research Note e
No 603) it HUS fOWld thut the maxiGm] amount of fire spreld IWS propottional lo
) -121 ( Bate of Hater Application bull If this result is used it Iltould irnplJ th~t
5
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
IHE GP-OH1Il OF A FOUGHT Fl[
1 II1rnOllUCrION
In 01 study of the numbJx of Plllf per GcCtion 0) a otudy of the filst attendance
rules it io eGsentiJl to have uu UlldcrGicndinG of the Hay in which fires amprolt
after the arival of the first pump ThLo Horldne noto iD a deoc6ptiol1 of a
proposed modol
Thin model is ~pplicle to buEdinC fir ~ c IIhich lulte Bprend beyond ch room
of ori-gin
I 2 OBSEllVATIONS FHOH THE DAfA
I I
AI1Y model which rcplaquosents the [oOltth of a fought filo must be copable of
exploining the foJloHb effects 11hich arc ob6cnuc in the 5AF2 data
i ~he middotaverace amount of Bpread for fir8 IIhich amp10 of tll~ snIDe nile Rt
arriVD1 middotincreDsell with th(l number of jets UGeurod This is due to the
differing fiercenccs of the fires An examination of the number of jets
UBed ugDinst the iinrl oren elso shown that spjcndinl3 fires use molS jeto
ii The proportion of fires which spread after Dlli val increases middotd th the
size of fire at 8rrival
iii But the average Mount of Gpread (for opreading and non-spreadJng
fires) is approxima tely constant for all sizES of fire
iv ii and Hi imply that for fires gthich spread tho average spread
(AfArgt ) io leGS for larGe fires
v Fires for which all the pumps errive srnuJutnncolsly are still likely
to spread nfter arrival
vi The conttol time (as recorcied on the K433 f orms) increuses ith
increaGing final urea
3 RFl3RESSION ANALYSIS
An onalJsis of the rclRtionship neblcen the ollnt f d d ~h t lth spree en le lm~s il fo
1
~h the jotc Hero vppllcd HOulcl PlOVC the oimplcst )KmiddotCi blo model of
or the follming eGGumptions
(j) Th rate of cromiddotth of firec is cxponcntiel
(ii) Each pump oupplicd 001) jet
(iii) Thc arrival timeG of the pumps are the times a c hich hc jete
f le
rlI model UGccl to reprctlcnt the grolth of a j jet fire as A ~ Ao e
rot~ jtpound1 -0 V j- ~OOCl~ e J
Hh0n Ap- is the oregtt dllmagccl Dtext in ction
Ao is the alea dam~ecl at arril8l
f is the exp(luentiRl growth rata
ti~1 le the i-1 th interarrival time ie the interval beteen the arrival
of tho itll and (i_1)th pUOlp k e represents finy croHth not ficcounted for by intc)$rrivlll groth
This regresGion analysiB RS tried on multi-storey non-chJellil1g firen l-Ihicb hUo
spraad bGyond the room of ori(in
1 jGt Ures Af 111f Hean Ar- 111
(n =598) AQ Ao
2 jet Ures Ap ooltt1 Hean Af ~ 116 116(n 366) A 0 Aa
All the R v81u(~ Hera very low and the regression coampfficiGnts woro -hot sienifcn
3 jet firos (n 11+6)
4 jet fires (n 95)
5 jot fires (n 42)
or only sigliificaut at the 10 level Host of the spread is cccounted fo by
dclfy in lntcrnrrival timer l1j)U providEl8 on order o ~l lnW t itn rj- ecirlto of
dOOG not provide u reasonable rD oil3 for D mod(Il of fh0 ijr-o-rcli(o
4 A THlWUhTICAL HCDFJ OF FIRE GliOI-lTH
If the firc size is roduced by applying Iter to tll p imetcr of the Hre
the tillle tden to extinguish the fire gtlDl be P)~OPGlltmiddot io middot i to At hel A is
the size of the fir( (Se Bsldlir 11olinok ThoCilflO Fire Re~car(h Note No 8811)
ell - at ob (42)dt
A _ AD - ~ 2 + bt (43) 2
The only inforruation El have about a fire iB tho urea at alrival (Ao) the
arell damcged (A) and the extinction tile (t ) The information defiueD twoE
points nnd a tangent on the graph of fire sizI] against time The quadratic
equation 11 42 43 is the simplest flUlction which ill fit a curve defiMd
by threo points
t -gt
EqlUltion 42 6h0l8 that the maximum a)ea i8 reached at time t ba a[tol
arrival This maxirutlIa fire size ill b thlt ni t damDged at extinction
3
lUU tbo ebsolut u DOlOunt ()f
rhe extinctioi1 ti(o js the tilT) c1 lhich 0
ba
a detGrmineB hOil qui0kly the fie si7o iD roduced 8 wUl c~epend on ~middothc
char[cteristic[J of the firo (ie Bnd filtrceu~Gs ) and lIil nlso reflec t the
nIgtJnba of j ~ tB uGed to figt 1ho firc As more j~t5 are apoJied the fire GheI
Hill be rGduccd or e quickly and the pa1rgtJueter a llill iucrcace
The parameter b represents the colltinucd grogtlth of the firo UFo it it being
fouGht If tho fire is not being fought a must hamp zero 81ulto Md the fir
groVlth will be dAdt b 0lthotlSh b ItlPy not be constant in gtIhich CO6 b is a tanGent to the grogtth curve at tho point A h2dAdt
Considet first tile 5impleot GittoCion ~hcre 1J1l thu pUlps middothich ay bo needed
arc available at tho time of the fhst alltrivnl and [lS5ume the parameters lgt and b
are constantt
The firEls HHl be charactcrised by their inItial 5izG Ao and their fierceness
The brigade in apply the appropiintc number of jets accolding to their
lisspoundsment of the fire size Dnd severity Fires of tho Sml0 initial size
Ao Hill folloll different groHth curvet according to their severity
As the severity increascs the ~ount of spread and control tiGe also incr~eso
Yo- a ghcn Ao [0 tmiddotJle fJc1rc6 ty incrcaceo the paXarncteZmiddot Cl will accrclt)cc
MCor the parumetcL b middotill incrcace
It 5s u66umed thampt all th~ dHfcrcncf)s n the cromiddot th of th6 fhcc ll 0 due
to the differences bchloen fires He a60UC1( thnt the brlgad Dpp1y tIle right
number of jctll and do net allovl for a fercc effoct by 6uGcel3ting that th
fhmiddotc may be put out more guielly by usinf more jet
5 A middotLfHOD OF ESTIHATING TIlE PAllAJmiddot~EW~S 1011 AN AVERAGE FIRE
~ ue still asouraLng that the puametels a und b a rc COllGtents and conBiceriq
onJy those fireG for Hhieh all the pumpG 10)C availo1gtle Ilimul tallcouGly If
take o eroup of fil06 of the BMa initial Gize Ao fought by the GBruG num0T
of jets He can observe avcraee values for the sprpad and control timo ftom
thr SAF2 data If Wc flUther aSGume tbat thil group of fires IJaG the same
(or at 16ilGt simila) Il and b 1uGO then Ie can use thu hlo obs epved avcr6r~B
to eShD1ilto the two porameters
2-
A rrrangcmcllt of the cqttation6 AI - Ao b 2a nnd tE ba +Ga giVeuro6
a ~ z[fiL~JAF J2
b J 2ak
2where l~ il3 the absolute amow1t of spreild) b 2a
If e can make soma assumption about the ay in vhich u changes RG edditioncl
jets are brought into we then can generalise the growth equation to include
thoGe fircl3 in lIhich there wae a delay in the arrival of the second and subsequent c-t-b
jets In a study of wood ~ firos (0 Donghcrty and Young Fire Research Note e
No 603) it HUS fOWld thut the maxiGm] amount of fire spreld IWS propottional lo
) -121 ( Bate of Hater Application bull If this result is used it Iltould irnplJ th~t
5
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
~h the jotc Hero vppllcd HOulcl PlOVC the oimplcst )KmiddotCi blo model of
or the follming eGGumptions
(j) Th rate of cromiddotth of firec is cxponcntiel
(ii) Each pump oupplicd 001) jet
(iii) Thc arrival timeG of the pumps are the times a c hich hc jete
f le
rlI model UGccl to reprctlcnt the grolth of a j jet fire as A ~ Ao e
rot~ jtpound1 -0 V j- ~OOCl~ e J
Hh0n Ap- is the oregtt dllmagccl Dtext in ction
Ao is the alea dam~ecl at arril8l
f is the exp(luentiRl growth rata
ti~1 le the i-1 th interarrival time ie the interval beteen the arrival
of tho itll and (i_1)th pUOlp k e represents finy croHth not ficcounted for by intc)$rrivlll groth
This regresGion analysiB RS tried on multi-storey non-chJellil1g firen l-Ihicb hUo
spraad bGyond the room of ori(in
1 jGt Ures Af 111f Hean Ar- 111
(n =598) AQ Ao
2 jet Ures Ap ooltt1 Hean Af ~ 116 116(n 366) A 0 Aa
All the R v81u(~ Hera very low and the regression coampfficiGnts woro -hot sienifcn
3 jet firos (n 11+6)
4 jet fires (n 95)
5 jot fires (n 42)
or only sigliificaut at the 10 level Host of the spread is cccounted fo by
dclfy in lntcrnrrival timer l1j)U providEl8 on order o ~l lnW t itn rj- ecirlto of
dOOG not provide u reasonable rD oil3 for D mod(Il of fh0 ijr-o-rcli(o
4 A THlWUhTICAL HCDFJ OF FIRE GliOI-lTH
If the firc size is roduced by applying Iter to tll p imetcr of the Hre
the tillle tden to extinguish the fire gtlDl be P)~OPGlltmiddot io middot i to At hel A is
the size of the fir( (Se Bsldlir 11olinok ThoCilflO Fire Re~car(h Note No 8811)
ell - at ob (42)dt
A _ AD - ~ 2 + bt (43) 2
The only inforruation El have about a fire iB tho urea at alrival (Ao) the
arell damcged (A) and the extinction tile (t ) The information defiueD twoE
points nnd a tangent on the graph of fire sizI] against time The quadratic
equation 11 42 43 is the simplest flUlction which ill fit a curve defiMd
by threo points
t -gt
EqlUltion 42 6h0l8 that the maximum a)ea i8 reached at time t ba a[tol
arrival This maxirutlIa fire size ill b thlt ni t damDged at extinction
3
lUU tbo ebsolut u DOlOunt ()f
rhe extinctioi1 ti(o js the tilT) c1 lhich 0
ba
a detGrmineB hOil qui0kly the fie si7o iD roduced 8 wUl c~epend on ~middothc
char[cteristic[J of the firo (ie Bnd filtrceu~Gs ) and lIil nlso reflec t the
nIgtJnba of j ~ tB uGed to figt 1ho firc As more j~t5 are apoJied the fire GheI
Hill be rGduccd or e quickly and the pa1rgtJueter a llill iucrcace
The parameter b represents the colltinucd grogtlth of the firo UFo it it being
fouGht If tho fire is not being fought a must hamp zero 81ulto Md the fir
groVlth will be dAdt b 0lthotlSh b ItlPy not be constant in gtIhich CO6 b is a tanGent to the grogtth curve at tho point A h2dAdt
Considet first tile 5impleot GittoCion ~hcre 1J1l thu pUlps middothich ay bo needed
arc available at tho time of the fhst alltrivnl and [lS5ume the parameters lgt and b
are constantt
The firEls HHl be charactcrised by their inItial 5izG Ao and their fierceness
The brigade in apply the appropiintc number of jets accolding to their
lisspoundsment of the fire size Dnd severity Fires of tho Sml0 initial size
Ao Hill folloll different groHth curvet according to their severity
As the severity increascs the ~ount of spread and control tiGe also incr~eso
Yo- a ghcn Ao [0 tmiddotJle fJc1rc6 ty incrcaceo the paXarncteZmiddot Cl will accrclt)cc
MCor the parumetcL b middotill incrcace
It 5s u66umed thampt all th~ dHfcrcncf)s n the cromiddot th of th6 fhcc ll 0 due
to the differences bchloen fires He a60UC1( thnt the brlgad Dpp1y tIle right
number of jctll and do net allovl for a fercc effoct by 6uGcel3ting that th
fhmiddotc may be put out more guielly by usinf more jet
5 A middotLfHOD OF ESTIHATING TIlE PAllAJmiddot~EW~S 1011 AN AVERAGE FIRE
~ ue still asouraLng that the puametels a und b a rc COllGtents and conBiceriq
onJy those fireG for Hhieh all the pumpG 10)C availo1gtle Ilimul tallcouGly If
take o eroup of fil06 of the BMa initial Gize Ao fought by the GBruG num0T
of jets He can observe avcraee values for the sprpad and control timo ftom
thr SAF2 data If Wc flUther aSGume tbat thil group of fires IJaG the same
(or at 16ilGt simila) Il and b 1uGO then Ie can use thu hlo obs epved avcr6r~B
to eShD1ilto the two porameters
2-
A rrrangcmcllt of the cqttation6 AI - Ao b 2a nnd tE ba +Ga giVeuro6
a ~ z[fiL~JAF J2
b J 2ak
2where l~ il3 the absolute amow1t of spreild) b 2a
If e can make soma assumption about the ay in vhich u changes RG edditioncl
jets are brought into we then can generalise the growth equation to include
thoGe fircl3 in lIhich there wae a delay in the arrival of the second and subsequent c-t-b
jets In a study of wood ~ firos (0 Donghcrty and Young Fire Research Note e
No 603) it HUS fOWld thut the maxiGm] amount of fire spreld IWS propottional lo
) -121 ( Bate of Hater Application bull If this result is used it Iltould irnplJ th~t
5
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
dclfy in lntcrnrrival timer l1j)U providEl8 on order o ~l lnW t itn rj- ecirlto of
dOOG not provide u reasonable rD oil3 for D mod(Il of fh0 ijr-o-rcli(o
4 A THlWUhTICAL HCDFJ OF FIRE GliOI-lTH
If the firc size is roduced by applying Iter to tll p imetcr of the Hre
the tillle tden to extinguish the fire gtlDl be P)~OPGlltmiddot io middot i to At hel A is
the size of the fir( (Se Bsldlir 11olinok ThoCilflO Fire Re~car(h Note No 8811)
ell - at ob (42)dt
A _ AD - ~ 2 + bt (43) 2
The only inforruation El have about a fire iB tho urea at alrival (Ao) the
arell damcged (A) and the extinction tile (t ) The information defiueD twoE
points nnd a tangent on the graph of fire sizI] against time The quadratic
equation 11 42 43 is the simplest flUlction which ill fit a curve defiMd
by threo points
t -gt
EqlUltion 42 6h0l8 that the maximum a)ea i8 reached at time t ba a[tol
arrival This maxirutlIa fire size ill b thlt ni t damDged at extinction
3
lUU tbo ebsolut u DOlOunt ()f
rhe extinctioi1 ti(o js the tilT) c1 lhich 0
ba
a detGrmineB hOil qui0kly the fie si7o iD roduced 8 wUl c~epend on ~middothc
char[cteristic[J of the firo (ie Bnd filtrceu~Gs ) and lIil nlso reflec t the
nIgtJnba of j ~ tB uGed to figt 1ho firc As more j~t5 are apoJied the fire GheI
Hill be rGduccd or e quickly and the pa1rgtJueter a llill iucrcace
The parameter b represents the colltinucd grogtlth of the firo UFo it it being
fouGht If tho fire is not being fought a must hamp zero 81ulto Md the fir
groVlth will be dAdt b 0lthotlSh b ItlPy not be constant in gtIhich CO6 b is a tanGent to the grogtth curve at tho point A h2dAdt
Considet first tile 5impleot GittoCion ~hcre 1J1l thu pUlps middothich ay bo needed
arc available at tho time of the fhst alltrivnl and [lS5ume the parameters lgt and b
are constantt
The firEls HHl be charactcrised by their inItial 5izG Ao and their fierceness
The brigade in apply the appropiintc number of jets accolding to their
lisspoundsment of the fire size Dnd severity Fires of tho Sml0 initial size
Ao Hill folloll different groHth curvet according to their severity
As the severity increascs the ~ount of spread and control tiGe also incr~eso
Yo- a ghcn Ao [0 tmiddotJle fJc1rc6 ty incrcaceo the paXarncteZmiddot Cl will accrclt)cc
MCor the parumetcL b middotill incrcace
It 5s u66umed thampt all th~ dHfcrcncf)s n the cromiddot th of th6 fhcc ll 0 due
to the differences bchloen fires He a60UC1( thnt the brlgad Dpp1y tIle right
number of jctll and do net allovl for a fercc effoct by 6uGcel3ting that th
fhmiddotc may be put out more guielly by usinf more jet
5 A middotLfHOD OF ESTIHATING TIlE PAllAJmiddot~EW~S 1011 AN AVERAGE FIRE
~ ue still asouraLng that the puametels a und b a rc COllGtents and conBiceriq
onJy those fireG for Hhieh all the pumpG 10)C availo1gtle Ilimul tallcouGly If
take o eroup of fil06 of the BMa initial Gize Ao fought by the GBruG num0T
of jets He can observe avcraee values for the sprpad and control timo ftom
thr SAF2 data If Wc flUther aSGume tbat thil group of fires IJaG the same
(or at 16ilGt simila) Il and b 1uGO then Ie can use thu hlo obs epved avcr6r~B
to eShD1ilto the two porameters
2-
A rrrangcmcllt of the cqttation6 AI - Ao b 2a nnd tE ba +Ga giVeuro6
a ~ z[fiL~JAF J2
b J 2ak
2where l~ il3 the absolute amow1t of spreild) b 2a
If e can make soma assumption about the ay in vhich u changes RG edditioncl
jets are brought into we then can generalise the growth equation to include
thoGe fircl3 in lIhich there wae a delay in the arrival of the second and subsequent c-t-b
jets In a study of wood ~ firos (0 Donghcrty and Young Fire Research Note e
No 603) it HUS fOWld thut the maxiGm] amount of fire spreld IWS propottional lo
) -121 ( Bate of Hater Application bull If this result is used it Iltould irnplJ th~t
5
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
lUU tbo ebsolut u DOlOunt ()f
rhe extinctioi1 ti(o js the tilT) c1 lhich 0
ba
a detGrmineB hOil qui0kly the fie si7o iD roduced 8 wUl c~epend on ~middothc
char[cteristic[J of the firo (ie Bnd filtrceu~Gs ) and lIil nlso reflec t the
nIgtJnba of j ~ tB uGed to figt 1ho firc As more j~t5 are apoJied the fire GheI
Hill be rGduccd or e quickly and the pa1rgtJueter a llill iucrcace
The parameter b represents the colltinucd grogtlth of the firo UFo it it being
fouGht If tho fire is not being fought a must hamp zero 81ulto Md the fir
groVlth will be dAdt b 0lthotlSh b ItlPy not be constant in gtIhich CO6 b is a tanGent to the grogtth curve at tho point A h2dAdt
Considet first tile 5impleot GittoCion ~hcre 1J1l thu pUlps middothich ay bo needed
arc available at tho time of the fhst alltrivnl and [lS5ume the parameters lgt and b
are constantt
The firEls HHl be charactcrised by their inItial 5izG Ao and their fierceness
The brigade in apply the appropiintc number of jets accolding to their
lisspoundsment of the fire size Dnd severity Fires of tho Sml0 initial size
Ao Hill folloll different groHth curvet according to their severity
As the severity increascs the ~ount of spread and control tiGe also incr~eso
Yo- a ghcn Ao [0 tmiddotJle fJc1rc6 ty incrcaceo the paXarncteZmiddot Cl will accrclt)cc
MCor the parumetcL b middotill incrcace
It 5s u66umed thampt all th~ dHfcrcncf)s n the cromiddot th of th6 fhcc ll 0 due
to the differences bchloen fires He a60UC1( thnt the brlgad Dpp1y tIle right
number of jctll and do net allovl for a fercc effoct by 6uGcel3ting that th
fhmiddotc may be put out more guielly by usinf more jet
5 A middotLfHOD OF ESTIHATING TIlE PAllAJmiddot~EW~S 1011 AN AVERAGE FIRE
~ ue still asouraLng that the puametels a und b a rc COllGtents and conBiceriq
onJy those fireG for Hhieh all the pumpG 10)C availo1gtle Ilimul tallcouGly If
take o eroup of fil06 of the BMa initial Gize Ao fought by the GBruG num0T
of jets He can observe avcraee values for the sprpad and control timo ftom
thr SAF2 data If Wc flUther aSGume tbat thil group of fires IJaG the same
(or at 16ilGt simila) Il and b 1uGO then Ie can use thu hlo obs epved avcr6r~B
to eShD1ilto the two porameters
2-
A rrrangcmcllt of the cqttation6 AI - Ao b 2a nnd tE ba +Ga giVeuro6
a ~ z[fiL~JAF J2
b J 2ak
2where l~ il3 the absolute amow1t of spreild) b 2a
If e can make soma assumption about the ay in vhich u changes RG edditioncl
jets are brought into we then can generalise the growth equation to include
thoGe fircl3 in lIhich there wae a delay in the arrival of the second and subsequent c-t-b
jets In a study of wood ~ firos (0 Donghcrty and Young Fire Research Note e
No 603) it HUS fOWld thut the maxiGm] amount of fire spreld IWS propottional lo
) -121 ( Bate of Hater Application bull If this result is used it Iltould irnplJ th~t
5
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
Yo- a ghcn Ao [0 tmiddotJle fJc1rc6 ty incrcaceo the paXarncteZmiddot Cl will accrclt)cc
MCor the parumetcL b middotill incrcace
It 5s u66umed thampt all th~ dHfcrcncf)s n the cromiddot th of th6 fhcc ll 0 due
to the differences bchloen fires He a60UC1( thnt the brlgad Dpp1y tIle right
number of jctll and do net allovl for a fercc effoct by 6uGcel3ting that th
fhmiddotc may be put out more guielly by usinf more jet
5 A middotLfHOD OF ESTIHATING TIlE PAllAJmiddot~EW~S 1011 AN AVERAGE FIRE
~ ue still asouraLng that the puametels a und b a rc COllGtents and conBiceriq
onJy those fireG for Hhieh all the pumpG 10)C availo1gtle Ilimul tallcouGly If
take o eroup of fil06 of the BMa initial Gize Ao fought by the GBruG num0T
of jets He can observe avcraee values for the sprpad and control timo ftom
thr SAF2 data If Wc flUther aSGume tbat thil group of fires IJaG the same
(or at 16ilGt simila) Il and b 1uGO then Ie can use thu hlo obs epved avcr6r~B
to eShD1ilto the two porameters
2-
A rrrangcmcllt of the cqttation6 AI - Ao b 2a nnd tE ba +Ga giVeuro6
a ~ z[fiL~JAF J2
b J 2ak
2where l~ il3 the absolute amow1t of spreild) b 2a
If e can make soma assumption about the ay in vhich u changes RG edditioncl
jets are brought into we then can generalise the growth equation to include
thoGe fircl3 in lIhich there wae a delay in the arrival of the second and subsequent c-t-b
jets In a study of wood ~ firos (0 Donghcrty and Young Fire Research Note e
No 603) it HUS fOWld thut the maxiGm] amount of fire spreld IWS propottional lo
) -121 ( Bate of Hater Application bull If this result is used it Iltould irnplJ th~t
5
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
being fouht by J jetl3 In tlliii proUminnry anlys1gtj huv rcatlOned
that in reDl fireo th additional jets ill havo le e cUed thn they oula
huVG in experimental firco and hovc used the cinlplcr re1laquo t iomiddotl ip aJ Jl1
If D fhe ia etinguishcd by J jets hich arrivG Ht 11 t 2 bullbull t _ afterJ 1
middotthe firot j0t tho e10llth curve lIill be of th9 fOlon chmn ueoo
JI i~--- ~I I
I
I
Thlt3 equation of the groVlth in the filml soction 111 be
A Aa + 0 I (b - 1t) dt +jt2 (b - D t) dt + bullbullbullbull2f t1
A bull Ao -l- bt - e ) +1
If baJ gt t J~1 the maxirJuu Hill occur in the final section ond the area
damaged nt extinction F will be
2 (t( - Cl ) AI = Ao b +~ ll2a i 1J
Tlw extinction time te i s obtainod from (5 bull3) b) solving fOl t Ilhen
A =o J-1
o = Aa - bt + C (0 - u )E 4+1 li=1
If we use the reI tioaship 1I = it a then [1 nrd b can be estilLO ted trolni 1
the cilllultlillcoUS oquntions
[
6
I
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
room of oriGin gtCl t3ed to test the es timatioll methods The fires Here gnngtpcu
according to the size at Glrivols the nur~bor of jets used ono the intenlrriv~l
timeD if more than o)a jet an -Gcti the inte-ulrival timeD ~Jcre calculated
by wl3UJlline that ith jet D9 applied llt the time the i th pump llldvcd For each
g~OlP of firoB p-n averagc (xtil~tion time and avcage GpreHd has c81c tloted and
the flVe~agc8 used tc caticoato v and b CG de6cribed in section 50 Tbe results1
are GhoHn id toble 1 I
oIithin tlie limitations of the lteta (sec ocction 8) Tho leonlt6 OhOlI
(i) Jar n giell Aa pound1 derrcnses and b incNases slightly with inereaning1
munbcgt of 1eta
(ii) for the two jet fires of n given Aa the values of a and bare
npployir1ate ly constant for IllrioUB valueo of t 1
0
(ii) a and b both increall6 I th incre~Bine Aa1
Figures 1 and 2 sholtl the l~elatioBhip bet-een deg 1
b and Aa The Glopes of
the log alog Aa and log blog AD elationships are about the 6Bllle and both
a 8m b are proportional to Amiddot 75 1
7 ALTERNATIVE FORlS OF THE QUADRATIC middotODEL
Hodel lsing P[oportional P=Dmet~2
The parameters b nnd a represent the absoluto rates of increaso nnd decreasD
of the fire It Ifotlld seem reanonllble to suggest that a and b might be
functions of the firo siz~ If a and b are directly proportional to the fire
size then the nodel beco01eo
bull2log A ~ log Aa - o( + pt (72) 2
In the more general ccse where there are delays bobcen the arrival of Gubneqtwnt
jets the sproad will be
7
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
bull
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
~shy
= 2r
rtud the control timo Clt1ll be estimated fl~om the cquntion J -I ~ 2 2o = log A + j3tE + Ii~n t (C(i+1 - 0( ) - tr1 oC (7~)i=1 12 J
2
11 this model the parameter 3 ia the exponential eruVJh fnetol The cctimatcs
of~J 1 and p [10 derive d ill on equivalent Hoy to that oec clibcd in section 6
Tho esti0l9to arc shOlm ill table 2
lhc ootimntcs oJlaquo 1 and ~ 010 quito doble with )7(Spcct to chanGinG Ao
decrea60s Ii th incI(ocing ritJllo C- of jets but there iD nstlfficiont eidc~c
to decide whether the grmlth a te f3 vories
~ehie model jmplies that the relationGhip between spread and arrival timo is
of the form
+ pound( 0( - c( ) t 2 i+1 t ~
2
This lelation8hip call be tested by applying this form of the recsroG6ion model
to the SAr2 dc-tu In the limited llUJober of regressions Imiddothich VleC t imed it
US found that this relationship gp-ve a bet t er fit or at least no OlSe a fit
than the intolarrivul time r1IodGl fbst tried ( section 3)
An Intermedi a te Hod2l
The relationship bob-lcen the constant palametOls )1 band Ao suggest a lliQdel
in which a and b aro propoltional to 75
1dA - 8 dt
A more genelaquol model and DJ] estimation procedure can be del iYei in a similar
fOlill to the o t her two models considered
8 (
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
8 THE sE OF QUADRtrIC 1lonEL
of the brieade dc-ponds on the charucteribticG of the Ere nuel a130 all the
action tken by the brilgtde A modd Ihich iu to be uB~d to In ~dic~_ tho
fire ard the brirgt2pound fhe (lqu~tions discussed in thiG note represent the
fire grovth for a given l3equence of arrivlll times t2 tJ Tile - Bequenco of arrival times will depend on the nlUnber of jotH Ihich arc applied
initially (from the first attendance or the aS6ist~nce Hl1ich ie roquested) and
0160 on the suboquent decisions to brlrlg on morc jets as tilt fire g1ol-l8
The relationship betlleen the area dRr~aged at oxtillction ilnd tIle totol number
of jets used has Blready been stndied Tbe relatiollship Ciltn bc explensed u
the form shmm beloH
C~~lo -K
p 0 ro
~ v~~ ~_r b 4 1egt
lt I rG T ~
o
A fire of a given siZQ can be charactelised by ito lelat~e sever-Ht - ie
a number between Omiddot and 100 which dctermilles its p0sHion on the cumulativ~
scale of utLnbors of jets required
s~~(-The followiug genenl principle could be used to FuiiCgt the growth of an
individual fire which is of size Ao at arrival
i Generate a random nUJlber to lepre6cnt the relative 6everity of the f51Clt
9
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
H Apply tho Il ~r of iClgt nc terrnincu by tho pocitiOh of middottile fire
on the jet aroD gl ph Tu () j e ts QuId be provideu by the firc t
attenuance or if necessary fJI3Glstanc() would ba roqvcf ted 1113 GOOa llll
the firt pump arrives
Hi ))termine ~ll8 a and h pawmoters of the fire hich dll u[o in be
a func tion of a)e aml rcl~t) v e severHy
iv PlGoict the (loHth of the fire and check wheth cl the f i re lI its
mrolmvm sie anu Hith the ~me soverity index would requiro Rdditiol19l
jets I v Reeucst additional jets if roquired and continue ~xtrapoliJtlng the
I grol-lth of the firo Hi th th~ chaneing a value
1
90 LDlITNIWNS OF TilE DATA
the pumps ani we hllt e lkd to ass ume Goroe sirJpl rule about the number cmiddotf jetbull (cAt- L tCr
provided by ~ach pump For any illliividurmiddotl fi) e He do not know hOgtl many jets
came fIOIl each pump or eveu hether the jets Cle appljed immedia tely the
pump arrived
he Extinction Time He have taken the extinction time to be the control
time as recorded on the Klf33 fOl-inS The control time is the tirJe bcteen tho
arrival of the fhst pWllP and the stop lIlessoge
For the smaller firos the time of the stop message 11 approxiroate to the
time the firo is out but for the larger fires th) stop message may be
wnt Holl before the fire is Ol1t
The AmOlU1t et S)re~ _ The runOWlt of spread is taken from the area at alriva)
and final area damaee RS recorded on the SAF2 forms The area at arrival Hill
usuRlly not be knoa with any accuracy and is ofteu recorded as beillg the GarJe
as the aleo at extinction The aroas are recod~d to two significnnt fieureG
10
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
an~ thir roulldjlC error introducej cltdd i jonal illHCCUlDCY
~Y~~i~t~ Ill uictributions of cplefld Incl control tillle ar hiGhly
Gk~cd Dncl if one V0lY llu(o value iG incluurccl in - OBlTIplo thio can conpletcly
outlmiddoteiCh 011 the other obnervationso
le ALTBIIHArIVE tlETlIODS OF ESTII1ApoundlIW rifE PJH111jETERS
1 EstiMlto the pllr ~ tcrs for- o11ch homOeCne01J8 eronp of flres g~ch C)OP
C011Biflt5 of fins IIhich hLwa the some ar~e at In )ival nutrbOl of jota and
irtcrarrival time6 Obtain an array of param2telS for eclt jet-area group
by interp012tior~ and smoothing of the eroup cnt)~ tes
2 Estim~te the parameter velues for each individual fire possibly iatroducir
II -ando01 clement desi~n~cl to put back the ronnciing erroro Dotermine the
parIJlJletels foX (middot8ch jet-area group by taking the aveNGe of the indivmiddot i(~ual vaJucc
in the group ThB re(son for puttilllS back the rounding error is to get a
continuouo distlibution of spread Rnd hence a continuous dictribution of estiulDtes
of a ond b
3 Obtain a r~gre6sion r~latiomgthip bebeen sYead and interarrival times anu
extinction tirue and interarXiva1 times Use the legression to estiOlatlt the spred
and extinction time for fires of a givltm area and number of jets Estimate the
a b parameters for this group using the) regrcssion estiMates of Gprcad and
control time
4 Use the regression ro1a tionLihip bettecn splcad contlol time number
of jets and area at arrival to estimate a b directly by considering the
correspondence between the coefficieJlts of the rlresGion model and the quadlat)c
modal
11 ~Eng2~iORK
The RelaJj2poundEJip betHeen ~ini5tic and ~chastic Bode
tIe have l[sumed that the pa rDJIleters a Iind b arc constant for each grol1p of
fires Ihelcas in Ienlity there Hill be a distribution of a and b values lhe
11
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
c_ b ~p viour of 90 fire Hith avor u b valueo mDy not bo the l3ame aB the
oTilEC of u11 fheEl dth individual 0 b Vaht0130 Tho properbol3 of the
GtnGhostic mOc1 Gl need to bE p-yllmincu anu in partiGuler it moy bo Horthhilu
consiuering a mpdcl in Hhich GOlDe fircn hccve zero b VllUClO (no growth) Dnd
tho remo inlng fheB luwe pos itivo b lues bull
Yilidcc ti~2 It should be possible to validate (or to diGprove) tllio Pl~opce J
mcdol by a more thorough study of the lorr6sion rclatioTJl3hips obtainable
flDm the SAF2 data
reotrictionc on groNth In practico lDany fires will bl) confined by the buHoillg
or ut lcal3t some fire compartment Hith i n the buildjng The effect of builuine
sie must be Qxamined
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
I
i r -1
shy bullI
bullbull
~
i
I I
I 1
-
--i
I I
I
i
I
1-
I i
bull I
bull l
i I
bull I
1
11
I
bull
__
_
I
I I
9
Ishy
v
------l
shy
--
-
_--
_
shy
-
-1
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
~cmiddot I ~ ~c ~r- ~(_ Cr - l 1
0 gt ox (kv(Ji (It )
(b~ 3f~( )
2 7
12
7 1
middot3
173
Yl7
54
~29
bull ~ 110
1I7
37$0
-rn r RC
10 I
b I
0- b0 2
I middotz 2
5 ( 4( 4 bull gt
I I
-77 s middotq I 3 ~s
I I
J 32 smiddott 1I b 8middot2 I
( e7 I 12shy
I 2middot81 -~ I I 1 I
I middot 71 11middot S 2middot1 JYmiddotS
44 34 middot j$ b 27y
28 middot gt l-t ~ middoto
fJ I 2 )middot7I Dmiddot
9middot ~7 76 93middot5
7- 7Er y ~( ~
~~I
b
pound 11middot0
07 N
9middot~5 951
7
1CAI
bull IS 5
middot7 5~
fi middot SS~
zmiddote 27
t -
0 1
CI If
2 -) t
7 (2
middot7
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
) vt 1
I I
I middot0
I
CA
coS
b
AeQ fe- F I(~C ~~
0 b o
L 2middot I
[0 0
10 0
Cc b
c 00 I shy I
Lo t 1 lt
0 c
L 0
c shy
o
S
S
3 SmiddotS
f 82
lJ_ middot1 7-shy lb
j
i I 0
I I
8
I (
2middot7
2qmiddot~
2$
34
34 (
I
1middot5
L-~ I
2 middot lt)
IgQ
2
369
I )_
21
1
5
ID
0
q
1middot7
middotHq 2o-Q 7
10
z t
zS r
Q ~ 0 I I o 2q
7( 13gt shy Lmiddoto 14 8
I
-
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
-n Cl ( 2 C~ W0 ( t t1 ( dc ) ~ ~ ~ ) jd )
shyi-J J~ J 0 r~ (middotVA( lLt ~middotlc ( G f Ctt- ~ -l( -Q e~ ~
dJ
I Kn d 1 (
1 ) Cl H ~ ~ 01
2 1 middot oS lj
147- -02(
71 -Da
middot O~713 _
17 - IYS7
-JD ~ O (I 0
0 )4 SC 3
gLc bull 03 7_-
Il (-) bullO$
bull Cgt 21911
gt~~ 00 17
y Cmiddot du-Q
2 ~ ~ loi middotmiddot J - 3 dd -
I I ()
d fj of pr- -1( 9 O~ - I f)
-1 3 ) 0 0 0
-o S ) 0( 2 middotc(
- 11 9 0 4 - 0
o8 C) 011 -10
0 6 1 0 1 middot08
0 1 ~ 02) 0 2 bull 0 1( IS
0( t -024 -08 000-1 O~
0 11 o~
0pound( -012 ()DqL
07 or et
I
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t
If I I I
i bull
I
~
~_ J~
ll
I 1
fmiddot
4-- -i -shy
bullbull bull r
- ~-
--
~
-
-
I
11 I -
middotmiddot 1
- -shy
L
I middot f _
_bull
-shy----shy
_ _shy
-1
bullbull
shy --shyjl _ 1
bull bull j shy
-
_
----rshy- ---- - 1
-r ~) t