>
Chapter 4 Challenges and Needs in FireManagement: A Landscape Simulation
Modeling Perspective
Robert E. Keane', Geoffrey J. Cary and lvlike D. Flannigan
AbstractFire management will face many challenges in the future from globalclimate change to protecting people, communities, and values at risk.Simulation modeling will be a vital tool for addressing these challengesbut the next generation of simulation models must be spatially explicit toaddress critical landscape ecology relationships and they must use mechanistic approaches to model novel climates. This chapter summarizes important issues that will be critical for wildland fire management in thefuture and then identifies the role that simulation modeling can have intackling these issues. The challenges of simulation modeling include: (i)spatial representation, (ii) uncertainty, (iii) complexity, (iv) parameterization, (v) initialization, (vi) testing and validation. The LANDFlREproject is presented as an example on how simulation modeling is usedto support current fire management issues. Research and managementneeds for successful wildland fire-related simulation modeling projectswill need (i) exteusive mechanistic research programs, (ii) comprehensive databases, (iii) statistical validation methods and protocols, (iv)software and hardware research, (v) modeling science explorations, and(vi) extensive training. Models will continue to play an integral role infire management but only if the science keeps pace and managers arepoised to take advantage of advances in modeling.
• Robert E. Keane: USDA Forest Service, Rocky Mountain Research Station, Fire Sciences~boralory. 5775 Hwy 10 West r...tissoula, Montana 59808 USA. E-mail: rkeaneOf.s.fed.us
e l!. S..Government's right to retain a non·exclusive, royalty-rree licence in and to anyooPYTIght IS acknowledged.
.....
d~mul~:ionf i~ an important tool to predict fire behavior and effects for a widedlver~~y::r ue management applications. However. the lenninology used to
en e e models can be confusing and inconsistent.
Tltree types of variabl· Il .the cent a1 t'f b .es al.e genera y used 1Il models: state variables describechange t~e :;~:el~~ri e~g slm~~ted. flux .variabl:s represent the processes that\"3!iables B a €s. an 'l.1ltermedzate vanables are used to compute f1tLX
variable ;vOUolrdebxamIPle. trbee leaf carbon might be a state variable and a fluxe t ,e car all lost ea h If' -are involved in s. I t' . c year to ea talL In general. four tasks
. UTIli a lOll modelma- }",·t· t·· t· . I ..mg values' tl . O' ~ W lza tOn Il1VO ves asslgmng start-'a 'e state vanabIes p. t· .model parameters that. . . marne .e1'lzatzon concerns quantifying those
in10nnediate . bl ale used 111 algonthms that compute state flux andvana es Vahdat' t '1 . ' .rl~ull are realistic and . l~~ en a1 5 testmg the model to ensure that the
lanily (Rykiel 1996). L,;uantlf} 1I1g the accuracy of results to estimate uncerttT::o and alga .thru stly, senszt1,1Jlty analysls 111\'0Ives modifying !Jarame-
n s to determ· tl·· flef al. 2007). me lell 111 uence on model results (Cariboni
4.2 Simulation modeling in fire management 7;
els ca~l also be ~Ised to update broad-scale digital maps and dcsi n futursampltng strategies for assessing change Fire behavior alld e" t . gil e
b. t . t d' . . . nec S moc e scan
e Ul egra e to simulate WIldfire s!Jread and resllitan' " .. .f .11" ' oJre seventy to deterIUlne I WI ("res are providing ecological b " (I( -
I h. . ene"IS eane and Karau 2010)
:\ ec alllstIc landscape models can be used t . 1 . t.:.' . •., . 0 exp Ole ule. clImate and vegetatlon mteractIons and to quantify fire regimes in S Jace . . . et a!. 2003. i\eilson el a!. 2005). :'IIost importantly, ~echa~~~ti~'~i~l1~~~el~~~model,s can help precilct potential fire dynamics in future el·llllates t . I6 e m 1 g . ." I' . . 0 proVIC C
r UI a emelll cutIca lI1formatlOn to mitigate adverse ef'eets (~Icl( . ta1. 2011). " " enzle e
This chapter discusses the challenges of using simulnlioll III d I· ."g A
. d . < a e mg 111 "reDIana ement. n llltro uctlOn to simulation n ad I· .1 ( 1 e mg IS presented that willlay t Ie groundwork for understanding most material 'In tl· I Th
h11 f... . < lIS C lapter. en the
c a enges 0 usmg sImulation in fire manalTement ap J..• t· .'
d_ I . ( 0 pled lOns are dIscussed
an an examp c of the lise of slIllUlation i r.:. .
L In III e management IS presented
ast, researc 1 needs. future direction. and possible sol t· . .1
_ . u Ions are StUllmanzedto ensure t Iat sllllulatlon modeling becomes an a t', c. .. I f ell€C 1\ e lire management toolIII t le uture because traditional fire management appl·oacl ·U b
t d. ' les WI e severely
tes e III a warmer clImate (Flannigan et a1. 2008). .
76 Chapter 4 Landscape Simulation in Fire Management
KeywordsEcological modeling, spatial dynamics, landscape ecology, mechanistic
simulation, parameterization.
4.1 Introduction
Fire management in the United States faces a number of challenges in thenext century. Seventy years of fire exclusion policies implemented under successful fire suppression programs have resulted in areas with increased canopyand surface fuels, especially those that historically experienced frequent fire,that now require extensive fuel treatments to reduce fire hazard and restoreecosystems (Ferry et a1. 1995). In some areas, extractive land managementpractices, such as grazing and timber harvest, along with exotic species invasions have created novel ecosystem and fuels characteristics that may also require innovative pro-active fuel and ecosystem treatments. ?"1eanwhile l hwnandevelopment is expanding into the nation's wildland extending the wildlandurban interface thereby making fire fighting difficult, heightening the risk ofloss of property or life from wildfire, and increasing the need for intensive fueltreatments (Radeloff et a1. 2005). Fire suppression costs are spiraling upwards,along with the ecooomic and social costs of treating fuels to reduce high fireseverity. New fuel treatment technologies, such as mastication, are findingfavor in fire management because they are less risky, easier, and cheaper toimplement I but their impacts on ecosystems remain unknown (Agee and Skinner 2005). Above all, future climates are predicted to be wanner and drierresulting in substantial increases in fire size l severity, intensity, and frequency(Cary 2002, Running 2006, Westerling et a1. 2006). Curiously, these same fireprone forests are being proposed for storage of carbon from the atmosphereeven though they will probably burn long before they can be effective carbon sinks (Sampson and Clark 1995, Tilman et a1. 2000). Fire managementwill need to develop new policies, strategies, and tools to meet these futurechallenges and ensure the sustained health of US landscapes (GAO 2007).
Simulation modeling will be one of the most important tools for fire management in the challenging future by providing an effective, standard, and objective context to evaluate management actions and ecological change (Lauenroth et a1. 1998). Models can be used to simulate effects of alternative treatments to determine the most effective fuel reduction or ecosystem restorationstrategy (Miller 2000). 'ovel treatments can be simulated to determine resultant short- and long-term effects on a diverse array of ecosystem elements(Ryu et al. 2006). Fire hazard and risk can be simulated to prioritize areasfor treatment and to design the most effective treatment prescriptions (Keaneet a!. 2008). Simulation can also be used to approximate historical landscapeconditions that can then be used as reference for ecologically based landscapeprioritization and planning (Wimberly et a1. 2000). Predictive landscape lUod-
4.2
4.21
Simulation modeling in fire management
A simulation modeling primer
78 Chapter 4 Landscape Simulation in F'ire Management
i\lodel approaches can be described as empirical. mechanistic. stochastic,and deterministic. Empirical models are created from extensiye data oftenusing statistical modeling techniques. Examples include t.he Australian firebehavior model (~IcArthur 1967), the CONSUA1E fire effects model (Oltmaret al. 1993), and the growth and yield model FVS-FFE coupled to a fireand fuels extension (Beukema et al. 1997). Empirical models are accmate butlimited in application to the conditions represented by the data. Mechanisticmodels simulate biophysical processes using universal physical and chemicalrelationships} therefore mechanistic models are applicable to a wide rangeof domains, but they are often complex} which often results in instability:inaccw·acy. and difficulty in parameterization. Stochastic models contain numerical relationships that use probability distributions. which often requirerepeated runs to quantify the variability in results. Deterministic models contain mathematical equations that represent important processes resulting inoutput that often does not vary for a particular set of inputs. In reality. mostmodels contain a diverse mixture of these four approaches, especially longterm landscape models. For example l a landscape fire succession model cansimulate fire using a mechanistic function) seed dispersal using stochastic al·gorithm l tree growth using mechanistic biophysical equations. and fire effectsusing deterministic decision trees (Keane et al. 2004).
Two groups of simulation models are used in fire management. Fi1"e behavior models simulate the physical combustion processes of wildland fire suchas spatial growth, rate of spread} firelil1e intensity} and flame length. Theseare strategic models for real-time. operational usc under wildfire conditionsor planning applications to describe fire hazard. Examples of these models include the mechanistic fiTe model of Rothermel (1972) that is implemented intothe BEHAVE software for point evaluations and into FARS1TE for spatial applications, and the empirical McArthur (1967) Australian fire spread model.Fi1"C effects models simulate direct and indirect effects of fire on ecosystems(Reinhardt et al. 2001). Direct or first order fire effects include fuel consumption, tree mortaility, and smoke production (Ottmar et al. 1993, Reinhardtet al. 1997), willie second order or indirect effects include vegetation development, erosion, and fire regimes (He et al. 2008). Fire effects models includeall those ecological simulation models that contain any representation of wildland fire from the stand-level gap models that simulate individual tree growth,mortality} and regeneration to the landscape fire succession models that simulate ecological processes. such as fire regime, in a spatial domain (Keane etal. 2004)" This chapter mainly covers landscape level fire effects simulatiollmodeling which nearly always contains embedded fire behavior models.
4.3 Technical challcllges in fire management modeling 79
4.3 Technical challenges in fire management modeling
There are assorted difficulties and dilemmas encountered b" 1110d I tl. d . .' . Jeers as leybUll ,·aJ lOllS computer programs for fire management tllat f I. span rom lOW todesign a model to how to usc the model correctly.
4.3.1 Model design
The chief challenge facing modelers is to build fire sim"lal'lon d I .. . 1110 e s usmg a
mechalllstic approach such that causal processes are "lllke,1 t. 0 ecosystem re-sponses so that new. unforeseen results can be generated (P I I T"Iaca a aile I man1994. Rastetter et al. 2003). Properlv designed mechanl'stl'c Inod I "t
• L e s are qUi erobust 111 terms of scope and application so that thel'I' SI'lllulated( consequencesslIch as responses to climate change, become emergent properties of the modeirather than predetermllled results generated as a conseq'le"c f ". . - eo parametenza-t'OIl (Peng 2000). The major challenge is to design l'ob'ISt fi" did. . Ie 1110 e s aroundetlUled algonthms that use nux variables to represellt '1111 t I' I.. . .. pol' ant p lYSlcarelationships and Il1teractlons between dynamic, readily quantifiable inputs.Unfortunately. research has quantied a fraction of the " I' I I. . . - major p lYSlca re a-tlOnslllps In a handful of ecosystems) so simulation design compromises arealways made to account for the limited state of knmvledge (Keane et al. 2010).F\lrthermore l those processes With IlltrmslC uncertainty. such as fire ignition.may. always need to be modeled stochastically because of their inherent compleXIty lUld cross-scale influences. It is important that modelers identifY thepiau Ible extent .o~ mechanis~ic design using available literature and existingmodels and exphcltly reeogmze these bounds in the results.
Future fire models must also be designed to be spatially explicit to ad:rcss complex scale issues (Peters et al. 2004). This means that the mod. lees must ~plicItly Itlcorporate spatial relationships in model design and(~Plementatron. One-dimensional (lD) or point models such as BEHAVElin~:ws and Bevins 1999) and FOFEM (Reinhardt et ~l. 1997), may have
aI use III the future because they can represent fire behavior at only one1.., e. Fu
l. tu:e fir~ behavior models, especially research-oriented mOdels must,
ot mu tl-dnnensl I' d " ,- ona III space an tnne to ensure that those processes thatI)(~ur at one seal d I' .!o< I . e an ocatlon arc affectmg processes that OCcur at other
a"" and 10catlOus (Gard . t I 1991) TI'for' I nel ea. . liS approach has man'\! obstaclesImp ementation' I d' 1 k f .",. J
hil'h d me U lllg ac 0 SluuClent data across appropriate scales, emand for comp t . . 'd'fi' 'for s' I . U CJ Iesources: I entl catIOn of the appropriate scales
IInll atlon (e I t" I' . <hn' activity (:Ma .g., se ec IIlg t le nght ~lxel ~ize). spatial autocorrelation inHoweve -I gnussen 2008), and speCIficatIon of the proper spatial extent.
r, exp oratIOns of t' I . .h-p}), ,,' I spa la lllteractions are the only wau to eomprehen-" .Imu ate Iir b h' d J
e e aVlor an effects across landscapes (Keane et aJ. 2010):
bO Chapter -t Lalldscape Simulation in Fire ~Ianagelllent
It is important that the spatial scales represented in the model are reCOllciled to the ecosystem processes that they represent and to the lime scales at\yhich these proce:sses operale (\rarillg and Running 1998). Tree regenerationd~·namics. for example. may require a spatiall,v explicit. anllual seed dispersalmodel and a simulation of reproduction phenology at a daily time step toproperly reflect climate interactions of tree life history.
Another challenge is lhat future simulation lIlodels must be able to simulate the complex interactions of state and flux "ariables across scales (Rastetter et a1. 2003 1 Drban 2005). Dynamic feedbacks and cross-scale interactionswill allow the prediction of no\"{~l ecosystem responses and interactions betweell climate. vegetation. and disturbance which are likely to lead to nonlinear model bellm-ior and cause important phase transitions that are criticalfor landscape management ('.IcKenzie et al. 201l[in press)). The trend andmagnitude of these interactions are Illostl~- unknown for man~' ecosystems andthev are difficult to ~tudy outside of a sinmlatioll approach. One importantint;raction is the role th~t humans play in past (e.g.. "\"ative American burning). present (fire exclusion era). and future (enlightened fire management) onlandscape dynamics (Kay 2007). [Ilieractions can dictate important thre:;holdsand phase transitions of landscapes in changing climates so that managementcan anticipate these changes and respond (Allen 2007). AIso important arehow multiple factor interactions. such as multiple disturbances. create novellandscape conditions that may accelerate landscapes toward important thresholds and phase transitions, Ecosystem science has only scratched the surfaceill determining the sign and amplitude of most ecological interactions largelybecause of the complexity in the nested scales of time and space involved(Allen and Starr 1982. King and Pimm 1983).
One last challenge is balancing complexity with utility in model design.In general. simulations are more difficult to conduct as model complexity increases because with complexity come additional parameterization, detailedinitializatioDs. higher computing demands. and complicated model behavior.Developing a parsimonious list of important variables to model is criticalto efficiently simulating ecological processes. otherwise a model can becomeO\-erly complex and difficult to parameterize because of lack of information.It is also easy to oversimplify model design such that simulation results aremeaningless. Conversely. if too much detail is included. the intrinsic uncertainties associated with each modeled process may compound to produce equallymeaningless results (Rastetter et a!. 1991, illcKenzie ct al. 1996). Moreover,manager:; may not have the time. expertise. and resources to operate ~nd
interpret highly complex fire models. Therefore. it is critical that simulatIOndesign is properly matched to the level of information required by managersand this is effectiYel.\' accomplished by plainly stating the objeeth"cs of thesimulation efrort.
1.3 Tl.'Chnical challenges ill fire management modeling 81
4.3.2 7Ilodcl use
One of the greatest challenges in modeling is to dearh' articlliatn s· Ib ' . . \,; .lmU8-
tion 0 Jcctlves to inform the simulatioll project. ""hi Ie seemingk ob .. .." '1 I .J \ lOllS.
thiS IS easl y t Ie ~~ast undcrstood concept in the design and implementHtionof fire 1ll0~e1s, \\ Ilhout an explicit statemeut of simulation objectives. it isprob~cmatlc .. and perhaps impossible. to build a comprchcnsh-e model thatprondes eaSily understandable results to address research aud maU8 crementC~IICCl·~IS. A c1e~r mode~ing objective allows the modeler to easily identify the(I) \'anablc~ to II1clude 111 the model structure. (ii) sequeuce of 5imulation forselected vanables. (iii) input and output file structures. (h·) critical ecosystem~rocesses to Slllllll~.te. (\-) ill1~ortal1t interactions 10 iuclude. (vi) timc St~ps tolIn~l~menl.Rnd (\"11) appropnate spatial and temporal resolutions aBd extellt.ft IS Important to state this objectivc so that the nlost appropriate modelsare ~elected or built. the right pararneters arc selected or quantified. the simulatIons arc successfully completed in an acceptable time. and the results areeasily ulldersto?d. \Vhile. in general. additional objectives can be explored~ the cO~~lplexlty of mechanistic models increases" it is unlikely that thereWill be a uber-model that addresses all objecth'es because there will llever besufficient science to support its development or computing reSOurces to conduct the simulation. Therefore~ it will always be imperative to focus modeldevelopment with a clear simulation objective.
. It seems logical t.hal .fire simulation models will be much more complexm tI~e f~ture: and thIS will demand increased computer resources. higher expertls~ ~n model lise. and more extensive parameterization. Complex modelsarc Cl:,tlcally needed because it is nearly intractable to design wildland fireexp~rtments that explore the dynamic relationships of fire and ecosystem behaVior over. multiple time and space scales. Crown fires. for example. areextremely difficult to study using empirical approaches because it is difficult~nd costly to measure heat flux across the large spatial scales involved llSmg contemporary experimental cquipment (Albini 1999). The long temporalscales II1vol\'oo In exploring dynamic fire regimes may preclude short-term anwers from intensive field surveys. which are undoubtedly invaluable in the
longer term. However. it is unlikch' that the fire manaucment would '~dOI>ttbese ne I· d I .". 0 ,
w comp lcate moe els 01 uecessanly afford the computers needed to:0 these models in an operational application. Therefore~ the challenge willh to develop complex spatially explicit fire models for research purposes and
tben synthesize them and their results to create manalTcmcnt-oricnted modelstatma tb . . 0
. ~ no e as 10bust as expected but will perform well in operationalapphcatlons becaus tl . .(K e ICY al e easy to parametenze. execute. and understand
eane and Finney 2003).The input of cr t· . I· .
lI11a e lIlto Slmu atlOn models IS an increasing challenge fac-ing modele d d .tit t rs an mo el users III the futnre (Keane et al. 2010). It is important
a models have a I· ·t . . .n exp ICI lepreSentatlOn of clImate across multiple scales
82 Chapter 4 Landscape Simulation ill Fire i\lanagelllent
to enSlU'C a realistic response of ecosystem dyuamics to climate. Phenology.for example. may need a daily climate time step at 100-111 rcsolution. whereasdecomposition may need monthly time steps at i-kIll resolution (Edmonds1991. White et al. 1997). Identification of the appropriate hierarchical scaleswill be difficult. but the task of matching the ecosystem processes of disturbance and plant dynamics to the appropriate climate scales may be evenmore challenging. Identifying the most parsimonious climate data stream toinput into models is also crucial given that too much weather data couldpotentially complicate and slow fire model simulations. This is inherently difficult because each weather variable has an intrinsic scale (e.g.. micl'osite,regional) and resolution (e.g.. vertical layers abo,·e ground. grid size). ~Iul
tiple weather streams representing past. prescnt. and future projections ofclimate are needed to determine potential climate effects on disturbance andvegetation. And. multiple climate scenarios are needed to bracket the range ofpotential effects and to identify important thresholds of ecosystem change. Anexplicit simulation of atmospheric transport is also desirable for fire models.\Vind speed and direction at various heights can feed disturbance proce es(e.g.. windthrow. fire spotting, insect epidemics) and plant dynamics (e.g. l
seed dispersal) (Greene and Johnson 1995). Atmospheric transport can alsobe llsed to simulate importallt feedbacks such as smoke dispersal. atmosphericdeposition. and radiation budgets.
A last challenge is quantifying the uncertainty im'olved in fire simulationsso that fire managers and researchers fully ullelerstand the impact and significance of the predictions and results (Bunnell 1989. Araujo et al. 2005).This includes developing methods to present simulation results that containan estimate of error or degree of uncertainty (Bart 1995). This assessmentof uncertainty should account for the error in parameterization. initialization,and model algorithms: as well as the error and variability in model predictions(Brown and Kulasiri 1996). The IPCC (2007) report contains protocol andclassification that they propose that all modelers use to describe the uncertainty assessment for modelers. Results must also be synthesized into variahland formats that are commonly employed by fire management.
4.4 A fire management simulation example
An example of how diverse simulation modeling approaches can be integratedtogether to create a viable management tool is presented to illustrate the useof landscape modeling in fire management.
pi
4.4 A fire management sirnulat.ion example 83
4.4.1 The LANDFIRE mapping project
The US Healthy Forest Restoration Act and the '1ational F,· I·e PI 'C h .. . an s 0 eSlveStrategy establIshed a national comntitment to reduce fire hazard and restorefire to those ecosystems where it had been excluded for decades (L' t d
r li' . 2 ).. .. . aver yanII It ams 000. rlus conumtment reqmred detailed multi-scale spat· I d t. ... I la a af~r pnontlzlllg. p anuing. and designing fuel reduction and ecosystem restora-tIOn treatments across the entire nation (GAO 2007). These spatial data layersmu~t ~rovIde essent18l fuel. fire regime. and vegetation information critical fordCSlgomg treatments and acthities at spatial scales compatible with effectiveland management (Hann and Bunnell 2001). The Fire Regime Condition Class(FRCC, an ordinal index with three categories that describe how rar the current landscape has departed fmm historical conditions) has been identifiedas. ~n~ of the primary metrics to be used for distributing resources and pri-ontlzlllg treatment areas to protect homes save I,·ves and rest d I· .
1 : ore ec Illmgfire-adapted ecosystems (Hann 2004). The LAXDFIRE project was initiated111 2005 to create a sCIentifically credible and ecologically meaningful nationalmap of FReC. along With developing a number of supporting maps of vegetatIon, fuels. and biophYSical seLtmgs, at 30 m pixel resolution that could be~sed a~ro~s mUltipl~ organizational scales. This project integrated mechanistIC ~tatlstlcal modellllg with landscape fire succession simulation to create thedesired products to serve as an example of how fire management might solvethe current challenges mentioned above.. .All LANDFIRE methods and protocols were based on a data-driven, em
pIrIcal approach where the majority of mapped and simulated entities werecreated from complex spatially explicit mechanistically based statistical modelmg (Keane et al. 2007) because managers required the LANDFIRE product~ to .b~ SCIentifically credible, repeatable I and accurate with a millimum ofSUbJectIVIty. To meet this challenge. the LAXDFIRE reference database wascreated. by. collecting georeferenced data from thousands of plots obtainedfrom a vanety of sources. most importantly. the USDA Forest Service Forest Inventory and Analysis program (Caratti 2006) (Table 4.1. Table 4.2).Th~ d~~a were used. ~or (i) ~eveloping training sites for imagery classificatIon, (n) parametenzmg. vahdating. and testing simulation models (iii)developll t t· I ill· . '
.. I g vege a Ion c ass catIons. (IV) creating statistical models, (v) de-termJ~mg data layer attributes, (Vi) describing mapped categories} and (vii);:jmg the accuracy of maps. models, and classifications (Rollins and Frame
6: The concept of lustoncal range and variability (HRV) was used as theilr~mlse of all FRCC calculations (Landres et al. 1999: Swetnam et al. 1999).,_ ds
Vwas defined as the quantification of temporal and spatial fluctuations of
",n cape comp·r ( . fw t OSI Ion portion a area by each vegetation map unit) prior toInn~ European-American settlemeut (Hann and Bunnell 2001). HistoricalFRec (Plies were then compared to current landscape composition to compute
ann 2004).
... i
1'] LA!'\DFlRE protot)'pe project developed the methods and protocols1e . . J . . tegratlOn
sed to map FRee acrosS the United States llsmg a camp ex In . al~r several ~CologicaJ models (Table 4.2) (Rollins et al. 2006). The hiS::~~n.spatial time series that. represented HRV were created from ~ands~la~~ :he US;HOIl modelina since historical maps and data are absent fot m~c d utputthe LA:\DSU:"I model was used to simulate landscape dynamIcs an 0
Chapler <1 Landscape Simulation in Fire )'lanagcmclIl
The linked models used in the LA;\"DFIRE project.
4.4 A fire managcment simulation cxamplc 85
\vXFIR8, BGG, See5
IVXFtRE, BGC, See5
IVXFIRE, BGC, See5
FIRE~ION
~Iodels
PLEAT
FIREHARM
HRVSTr\1'
LAl\DSU~1
HRVSTAT
Compute fire hazard and risk
(Kcane et al. 2010. keane andkarau 2010)Compute potential fire sever
ity (1<arau and I<eane 2010Iinprep))
Flow of LANDF1RE tasks
Compute departure (Pratt et
al. 2006)
Task
Compute FRCC (Steele et al.
2006)Use of LANDFIRE data
Compile LA:\'DFIRE
database (Caratti 2006)Build PVT map (Holsinger et
al. 2006)Create current cover typc and
structural stage maps (Zhu etal. 2006)Develop ancillary fucls data
layers (Keane et. al. 2006a)
Simulate HRV historical t,ime
series (Prat.t. et a!. 2006)
Data
FIA data. research data.
legacy dataLANDF'JRE database. Simu
lated outputs,LANDFIRE database, simll~
late<! outputs, satellite im
ageryLA DFlRE dat.abase, litcra
ture, PVT map. covcr typemap, structural stage map
LANDFIRE database. litera·ture, PVT map, cover type
map, structural stage map,N1F~tID database
LAKOSUM output. PVTmap, coyer type map, structural stage map
Departure estimates
LANDFIRE fuels and vegeta.tion maps
LA tDFIRE fuels maps
landscape composition over 5.000-year HRV simulations (Keane et al. 2006b:Pratt et al. 2006). Historical fire regime and vegetation succession field datacollected from numerous studies were used to parameterize LAi\DSU:\l (Longet al. 2006). LA"DSU~I stratifies these parameters by three \'egetation-basedclassifications: (i) Potential Vegetation Type (PVT) defined by biophysicalsetlings. (ii) coyer types described by dominant vegetation, and (iii) structuralstage described by \·ertical stand structure. The PVT approximates biophysical seLting by assuming that the unique nclima..( vegetation community thatwould eventually develop in the absence of disturbance can be used to identify uniqne environmental conditions (Daubenmire 1966). The LANDFfREPVT classification is a biophysically based site classification that uses plantspecies names as indicators of unique environmental conditions (Holsinger etal. 2006). Cover types were named for the species with plurality of canopycover or basal area. while structural stage was based on canopy cover andheight (Zhu et al. 2006).
Table 4.2 Flow of LAXDFIRE tasks to create the fire regime condit.ion class(FRee) and various fire management projects that use LAXDFIRE data (:\Iodelsare defined in Table 4.1).
Thornton(1998). Thorn
ton ct 01.(2002)
Steele et al.
(2006)
Thornt.on et. al.
(1997)
Lutcs et. al.
(2006)
Quinlan (2000)
Keane and
Holsinger
(2006)
(2002). 1<eane
et al. (2006b)
ources
Keane el al.Data
LA:"DSU'1.LANDFlREmapS
Legacy datafrom univer~
siLy, government, and
private agen-
cies\\'eather st.a~
t.ion data
collectedthroughout the
CSSimulated datafrom all models
WXFIRE.CLI~IET grid
ded database
CLI,IET grid
ded database,Digiwl eleva
tion models
\'egctationstudies. fire
history studies
Create theDAY~tET
database foruse in simula~
tion modelingCreate vcgeta·
tion and fuels
maps
gTiddedweather to
30 meters,compute climatc variables
for vegetation
mapping
Computeecosystem pro
cess variables
fOT vegetation
mapping
Computing
FRee from
HRV-current
comparison
Create theLANDFIRE
database
Purpose
time series
and map fire
regimesExtrapolate
coarse scale
Generate HR'·
Statisticalalgorithms
for regression
tree empirical
modeling
A model to
create griddeddaily weather
across the S
A statisti
cal analysis
program
A database and
analysis systemfor data man
agement
Description
A biogeochem
ical ccos~·stem
model
A biophysi
cal weather
extrapolation
model
model
A landscape
succession
r-.lodel
DAY~IET
FIRE~ION
II RVSTAT
BGC
\\,XFlHE
LA:"DSU~I
Table 4.1
84
""'"
."@'
'".."•~o3
"
t"'•5.~•'Co!!13
"~o'"
~
og-
~
'":-oc..-..-,~:-
"J
B
wv
t
'"-"---~--.• .. .. I....
A
'NY
t
'0
----~--.. .. .. ,. ,..
------"- - ....__...<lII_-------~-~--------------_.-------..----..._---------------,,,,_.--------------...."._--------......- .._--_....-::-
Fig.4.1 Critical maps creat.ed by the LANDPIR8 project to compute FRee fol' the northern H.ockies mapping zonentiaI vegetation type map; 13, cover type map; C, structural stage map; D, FRee map; E, fire behavior fuel model map.
-=-=-=-..- - - '. ...
~
:..:>
'";;3•~~o3"..g'~•3
'C..19. A, Pote·
'"-=-::-=_..• .. .. uo I.
wv
_.--.------"-------
D
t
wv
'"
},"".
.,. "',~ ':.), ..
'I-\.,!(\.~ . '.-,_.-_._..-...._-'-
""~
88 Chapter 4. Landscape Simulation in Fire ~Ianagclllelit
\"egetation maps portraying current conditions were thcu needed to compare with the simulated HRV time series to compute FRCC (Table 4.2). Thethree classifications used in the LA:'\DSU~jmodeling (PVT. cover lype. structural stage) were mapped to describe current conditions so that simulatedHllV data matched existing conditions, The PVT map (Fig, <I.IA) was crcated using a gradient modeling approach where plot-based assessmeuls ofPVT were modeled from a plethora of biophysical variables t hal werc summarized from field data using regression tree (CART) statistical techuiques(See5 model (QUinlan 2000)), The biophysical variables ,,'ere computed fromthe biophysical \vXFlRE model (Keane and Holsinger 2006). simulated using the BGC model (Holsinger et a1. 2006), or taken from the DAY:IJETgridded wealher database (Thornton et al. 1997) (Table 4,2), Resnltant empirical CART models \\·ere then used to map PVT across lhe region llsingindependent variables created from the same models (Holsinger et a1. 2006),The biophysical PVT map was subsequently used with other biophysical "ariabies and Landsat 7 Thematic :\lapper satellite imagery to create the covertype and structmc stage maps using a gradient modeling approach (Zhu eta1. 2006) (Fig. 4,lB,C),
Two methods were used to calculate the rlcpartul'e statistic that quantitatively compares the existing condition to the many historical landscape compositions, Steele et al. (2006) developed the I-1RVSTAT model that computeddeparture and statistical significance using advanced regression techniquesand Pratt et al. (2006) used a variation of the Sorenson's index based onmanagement-oriented FRCC methods (Hann 200<1, Barrett et al. 2006) (Fig,4,ID). Both departure indexes ranged from zero to 100 with 100 being themost departed, FRCC was finally created by making classes of the departurestatistic (Pratt et al. 2006),
Some products created fTom the LAl\DFIRE process (e,g.. biophysicalvariables) were then used to create additional fuels and fire regime layers thatare critical in the eventual planning and implementation of fuel and restoration treatments at local scales (Fig, 4.1E), Fire regimes were taken frOIll theLAl\DSU~1 model (Pratt et al, 2006) and described fire return interval andprobability of tlu'ee fire severity types, Canopy fuels maps were created using the gradient modeling approach where the same independent variablused for vegetation mapping were correlated to plot level fuels characteristics (Keane et al. 2006a). Surface fuel models were assigned to combinationsof categories in the three vegetation classifications (Fig, 4,IE), LANDFIREproducts have been extensively used by fire management for a variety of applications. Fire hazard and risk maps have been created for large regioWiusing the FIREHAR~j program (Hessburg et al. 2007: Keane et a1. 2008),The FLEAT program is being used to estimate fire severity and quantify ec0
logical benefits from wildfire using HRV simulations using LAKDFfRE inputdata (Karau and Keane 2010 [in press)), The LAl\DFIRE fuels data layer,are being used as inputs to FARSITE to simulate fire behavior on wildfire:;
...4,5 Research and llwllagemcllt needs and solutions 89
tlu'oughout the US,
4.5 Research and management needs and solutions
IrrespecLive of design, the performance of any simulation llIodel . db ' tl e q It' f I " IS governe
j 1 ua I} 0 tIe underlYlJlu SCIence the technl'cal '.' , .f I d I . .. o· ane CICatlve ablhtyo tIe .InO CeI to quantltauvcly implement this understandin did011 wInch the model' b d TI ,'," g, illl t ,e ata
, IS ase. lelclO1e. slIl1uluuon research nceds C) I . IqualIty, rele"ant science on which to base future models (1'1') I' I I I ug ,I
'I t b 'Id ' lIg I Y trame,mot e ers 0 til ,apply, test. and teach these models and (ii') ·t ' dt d I t t ' , , I' ' I ex enslve atao eve op. es. IIlltla lze~ and parameterize mOdels
The quality of the underlying science is deten;,ined b tl I'd' t f I ' Y lC Cumu atl,'ea \ancemen 0 p lyslcal. ecological, and climatic kno,,'ledge ' d b' . b ' pIne yc~
~f1ment.atlOn, 0. sen'atlOn. and publication in peer-reviewcd journ' I TJ".IlIformatlOn prOVides a mechanistic understanding fk. a s. lIS
d ' . 0 ey Plocesses that governecosystem ynamlCs as accepted br a broader s' t'fi 'I f · , '. . < Clen I c community. Because
tIe uture IS so uucertalJ1 l It IS vitally importal t tl ttl-' . '. 1 . '. 1 la lele al e comprehenSiverese81C 1 proglamS.8Ullmg at understanding novel CCos\'stelns hlel dOt'
d 'I ' I . ,J • con I lOllSan SOCIa Issues t I3t will evolve as climate are III ,'fi d I ' ''. oe Ie, JUmao populatIOnsIIlcrcase, and attitudes chanue. Since mechanist'lc' . I. • , _ 0 uPpl oac les are suggestedIt IS IlDportant that field researcll proJ'ects sho Id I, "I I' . u ene eavOl to quantify thecausa J:e ~tlOnslll~S that govern fire and ecosystem dynamics so that thesel~lecha..n~t~c eqtlatIO~ls can.~e developed for implementation in future simulation ~no e s. There IS a cnLJcal need for fundamental research into tb b 'phYSical processes that control fire behavior and subseqtle"t efl'ect F. e aSllcIi b 1 , '. S. '1I1C sea e
re .e la\ lOr o.utputs should leed detailed ecosystclllll1odels to mechanisticallypred~ct ecological response. This includes a new theory of wildland comb t'~~I~ICS 8
dnd ha more physiological approach to simulating vegetation rcs~~~:l~~
e an t e subsequent development.
Iin~Ood mOdelers. possessing comprehensive knowledge across diverse disci~irst 't~:ebr~e, .However. even good modelers are limited by many reasons,~O· ~c glotilld. knowledge. and experience of a modeler can limit the~o \~~iC~ua~lty•.and com~lexity of model design, Second, there is an extentexist" e'~"lIlg ecologIcal understanding can be feasibly incorporated into[ionslllogf IIl
I°t elstr~lctures based on the modeler's skills. There are also liluita-
'a a parsllnony d 'I b'I' "binder tl . . an aval a 1 Ity, and engmcenng restrictions. whichIe lIlcorporatlOn of ne ' k led' "('\"en the m . \\ now ge IIlto eXlstmg model structures by
peral and ost ,accomphshed mOdeler. ?\lore theoretical issues such as tem-ei spatIal scale reconciliation and representation (Urban 2005) a d
o-econollllC factors J dd' , ntnunit), TI ~ I a so pose a ItlOnal challenges to the modeling com-
" lerelore ed t'ltlUSt empl' J. u~a Ion programs. particularly at the graduate leveldl ~landinl~~~~ a dlver~lty of modeling ,approaches and multi-disciplinary un~
g Iture fil e models are gomg to possess the attributes needed in
-90 Chapter 4 Landscape Simulation in Fire ~Ianagcment
the future. In addition. modeling projects must involve collaboration acrossmultiple disciplines to ensure that current science is appropriately integrated
into model algorithms.At the center of future simulation research is a need for comprehensive
data to run and validate future models. The balance of data needs versusmodel advancement reRects a critical imperati\"e for cross~fertilization betweenfield ecologists. who provide data and equations to modelers. and modelers}who must then integrate that knowledge to provide descriptions of phenomena at different spatial and temporal scales. It is critical that extensive fieldprograms should be intimately integrated with simulation efforts to ensurethat sufficient parameter and validation data arc measured for model applications. Temporally deep. spatially explicit databases created from extensivefield measurements are needed to quantify input parameters. describe initialconditions, and provide a reference for model testing and validation. especially as landscape fire models are ported acroSS large geographic areas andto new ecosystems (Jenkins and Birdsey 1998), For example, Hessl et al.(2004) compiled a number of ecophysiological parameters for use in mechanistiC ecosystem models, which has increased parameter standardization anddecreased the time modelers spend on parameterization. :\ew sampling methods and techniques for collecting the data are needed to ensure that the rightvariables are being compared at the right scales. Field data useful in simulation modeling should be stored in standardized databases, such as F1RE\IOt\(Lutes et a!. 2006), and stored on web sites so that thcy are easily accessiblefor complex modeling tasks. Last. new instruments are needed to quantifyimportant simulation variables such as canopy bulk density: to initialize and
parametcrize firc behavior models (Keane et al. 2005),Model validation research is also critically needed to ensure that future
models axe behaving realistically and accurately (Rykiel 1996; Gardner andUrban 2003), There are many ways to validate models, The most preferable one is direct comparison of model results with field measurements in theproper spatial and temporal context. :\ext. intermediate results from modelalgoritlmls or modules can be compared against appropriate field data (Oderwald and Hans 1993), Results from complex sensitivity analyses can also beused to evaluate model behavior and to compare behavior against measureddata, expert opinion, or modeler experience (Cariboni et al. 2007), Comparative modeling exercises or ensemble modeling is also another potential loolfor validation where several models are applied to the same area (e.g. I standor landscape) under the same initial conditions with comparable parameterizations (Cary et a!. 2006), Results [TOm ensemble modeling can be used toevaluate the sensitivity. accuracy, and validity of model results and to explorenew ecosystem rcsponses (Cary et nl. 2009), ilfodel outputs can also be evaluated by a panel of experts to estimate the degree of accuracy and realism
(Keane et a1. 1996),New quantitative methods are also needed to evaluate the uncertainty
pi
'1.5 Research and management needs and solutions 91
around model predictiol1& (Gardner and rban 20 . .analysis methods are needed to sup t I'd' 03), StatIstIcal tests andpOl' va I allon compansons a d 't'ity tests that account for Sl)atial d t ' I n sensl IV-, 'ft an empma autocorrelatio d t 'slgm cance (~Iayer and Butler 1993) C "I _ n an est ,or_ ntlca to testing and \'a1idat' d Iis an assess.ment of whether the internal com lexit' of th· 1l1~ m~ e 5
tually malllfested in results (Cary et a1. 200~ anito bot:model deSIgn IS acestnnate of uncertainty (Klcijnen et al, 1992) illodeler d compleXIty m anthat compare the variance and trend of' I't' ,5 need statistIcal testsslmu a IOn resu ts to the x dcomes from model algorithms (O':\''-Il 1973) TI . e pecte outalgorithms and software to test the I~od I .. ' 1C) a.lso need both statisticale ovC! Its entIre range f I' I 'I'and creale response surfaces for variolls 1" 't', I I' . 0 app Ica )1 ityc 11 Icl cone ItiOns paramet ' t'and scenarios. There are also needs fa' d I' .' enza IOns:I a rno e lIlg SCIence resea) Iwhere new modeling approaches met! od d rc 1 agenc a
I. 1 s. an protocols are devcl I t
ensure t lat models are used correctlv b fi. opec 0
I, I d " y Ie management Optim .
atlon an scape size and shapes are needed to define t . . lim SllllU
future simulation projects (Karau and l{ 2007) he spattal context foreane and prope Tb'periods must be determined to ensure that ' r eqlll I ratIonresults (Pratt et al 2006) Th ' managers lllCorporate meaningfulfor stochastic mod~ls mus; be ~:~;:~~;,ate num,belr of simulation replicates. r ong Wit 1 t 1e appropriate sin ul t'
time spans lor creating fire regime ma (I{ 1 a Ion'd' I ps eane et al ?OO?) ~Iethod dgul es ,or se eding the most appropriate m d I ~ ,- -" s anare also needed. 0 e or a management application
There is a need for future modelinu d .code that is efficient. fast. and useful t 0 en eavOl s to create programmingmany purposes and appli~ations Ther 0 ~anagemellt and o~her modelers for
provide challenges for oPtil11all~odel Je:~;I~~~l:); ~:~I~;:;llllll1g concerns that
• ~~O~~~;'~~~~':I:t~~;i~:,~t~:~ity to compilc the modei on many machines
• ~~::ula~ tieSig~, 110del functions should be built in modular form with. co e so t lat modelers can modify coded algorithm r. .' .111 another model. s 01 II1tegratlon
• Graphicat User Inter' ' tlde t d Jace tnpU output. Easy way to enter input and unrs an output. -
• Open source anti integrated d S d 'and is posted or bl' I d rco e. ource co e IS written in modular style• At I . pu IS 1e lor others to use.
l! It-threaded executions, Abilit ' tcomputers. .) 0 nm on many processors across many
• Extensive document t" All'clearly defined ' ab~on, written code is fully documented includingand Lheir . t varia es and associated Ul1lts, descriptions of modulesUser man:~Pu and output structure, and descriptions of all functionslished and \m~del descriptions, and design descriptions should be pu~
• Sim1l.lat,' mh~ a- ata reco~'ded for all input/output parameters.on lStOry retentl AbTinform future s' I' on. I Ity to remember past simulations to
II11U atIons.
.----~-------- • 1
92 Chapter 4 Landscape Simulation ill Fire !\'lanagclllent
. must be developed to ensure efficient model-New software technologies d' ge modular sharing. and. r data accesS an stOl a .
iug building. rapid exccu lon, t t 1-11101ofn.r might be needed to. bT . 5 New compu er ec bJ
di,"erse debuggmg a 1tlle . . fi. 111anagement to evolve a new. f· 'C This may reqmre [c
support thIS new so twal . .~. I fi e models of the future. sug-. . I t run and mterpret t 1C r .' d
capaclt)' to Imp emen .' . f odelinu experts wlthm anucstinu specialized training and fOrilling tcams 0 III 0c cacrosS agencies.
4.6 Summary
References 93
fire models. Software and hardware technologies need to be developed thatfacilitate efficient and rapid simnlation. New test and validation statistical designs will be needed to evaluate the reliability and nncertainty iu simulationresnlts. And, modeling science research will need to develop suitable gnidelines for using and interpreting models. To effectively use these advances inmodeling technology, management will need to train modeling specialists toeffectively ntilize these models and interpret their resnlts. Simnlation holdsan important role in the fntnre of fire management bnt it is up to research todevelop comprehensive models that predict and explain important ecologicalphenomena, and it is up to fire management to understand these complexmodels so they can be used effectively in common analysis tasks.
References
Agee JK and Skinner CN (2005) Basic principles of forest fuel reduction treatments.Forest Ecology and Management 211: 83-96.
Albini FA (1999) Crown fire spread rate modeling. Progress Report RJVA RMRS99525, Fire Sciences Laboratory P.O. Box 8089} Missoula, MT SA.
Allen CD (2007) Interactions across spatial scales among forest dieback, fire, anderosion in northern New Mexico landscapes. Ecosystems 10: 797-808.
Allen TF and Starr TB (1982) Hierarchy: Perspectives for Ecological Complexity.The University of Chicago Press, Chicago.
Andrews PL and Bevins CD (1999) BEHAVE Fire Modeling System: Redesign andexpansion. Fire Management Notes 59: 16-19.
Araujo MB, Whittaker RJ, Ladle RJ and Erhard M (2005) Reducing uncertaintyin projections of extinction risk from climate change. Global Ecology and Biogeography Letters 14: 529-538.
Barrett S\OV, DeMeo T, Jones JL, Zeiler JD} Hutter \lVC (2006) Assessing ecological departure from reference conditions with the Fire Regime ConditionClass (FRCC) mapping tooL In: PL Andrews and BW Bntler (eds) Fuels Management - How to Measure Success. USDA Forest Service Rocky MountainResearch Station, Portland, OR, USA. 575-585.
Bart J (1995) Acceptance criteria for using individual-based models to make management decisions. Ecological Applications 5: 411-420.
Beukema SJ, Greenough JE, Robinson DC, Kurtz WA, Reinhardt ED, CrookstonNL, Brown JK, Hardy CC and Stage AR (1997) An introduction to the Fireand Fuels Extension to FVS. In: MMaJ A R Teck (ed) Proceedings: Forest Vegetation Simulator Conference. United States Department of Agriculture, ForestService, Intermountain Forest and Range Experiment Station, Ft. Collins, COUSA. 191-195.
Brown T t, Kulasiri D (1996) Validating models of complex, stochastic, biologicalsystems. Ecological Modelling 86: 129-134.
Bunnell FL (1989) Alchemy and nncertainty: What good are models? USDA ForestService General Technical Report P W-GTR-232.
C'ariboni J, Gatelli 0, Liska R, Saltelli A (2007) The role of sensitivity analysis inC ecological modelling. Ecological Modelling 203: 167-182.
ar)' G, Flannigan MO, Keane RE, Bradstock R, Davies ID, Lenihan JL, Li C,et aI. (2009) Relative importance of fuel management, ignition likelihood, andweather to area burned: Evidence from five landscape fire succession models.
~tajor user
1lana.g:emenl
Research and
management
Research
Research
managementHcsearch and
rc-
possible directions
);ew programming
software. open sourcecode development.
modular code designEnsemble modeling. mela-modeling.
novel field sampling
techniquesExpert cadres. centersof excellcnce, extensivetraining courses and
workshops
Standardizeddatabases, map creation procedures.database managelllcnt
technologiesEcophysiologicalscarch,
llesearch need
~lodel analyses andvalidation procedures
and technology
Ficld research in basic
physical process, inventofY systems thatquantify mechanisticparamet.erslnnovat.ive software design, better computer
resources
Need sampling meth
ods and protocols
managers
)"lodel use by
Challellgc
Accurate and
realistic models
)..lore efficient
models
Develop models thatare parsimonious but.explanatory. develop
effective training and
_______~al!?)p~l~ic~a~[~io~nc:'~·e:!:h~ic:!.les~-----------------
~lechanist.ic
design, balancing complexit.y
with utility
Data coHee-t,iolls
'11 t' l11e to delJCnd on computer simu-d research Wl con IIFire management ~l~ Heations. I-lOWC\"cr. man~r aspects nUist t:elation for many ploJecls and app fl' tl f ltll'C (Table ~.3). :'dodels Will
.' d Is to be use U in le uincorporated m new mo e ., d' I t s'IITIulate physical processes
I· . challlstIc eSH!TIS t 18 ~ .need spatially exp ICIL me . I I C 'I arrement-oriented Illodels mnst
-' t' saver lUUltIp e sea es. 1\ an (:) .and theu mterac lOn I I . ated for rcsearch e"llloratIons.
. d f tl TIj,lex fire moc e s CI ebe syntheslze rom lC COl· t . llect data that is useful to pa-
. I nt ficld efforts mus co t .,
Research anc manageme .) ... r. -' (inventory). and vahdatlOn oframeterization (research studIes 1 mltla IzatiOn
Id f simulation in fire management and
Table 4.3 Challenges and researc 1 nee s orresearch applications.
-94 Chapter 4 Landscape Simulation ill Fire ~Ianagement
International Journal of \oVildland Fire 18: 147-156.Cary GJ (2002) Importance of a changing climate for fire regimes in Australia.
In: Bradstock RA, AI\II Gill and JE Vv'illiams (cds) Flammable Australia: TheFire Regimes and Biodiversity of a Continent. Cambridge University Press,Cambridge, UK. 26-46.
Cary GJ, Keane RE, Gardner RH, et al. (2006) Comparison of the sensitivity oflandscape-fire-sliccession models to variation in terrain, fuel pattern and climate. Landscape Ecology 21: 121-137.
Daubcnmire R (1966) Vegetation: identification of typal communities. Science 151:291-298.
Edmonds RL (1991) Organic matter decomposition in western United States forests.In: Management and Productivity of \"'estern-montane Forest Soils. SDA Forest Service Intermountain Research Station General Technical Report, Boise,10. 118-125.
Ferry G\OV, RG Clark, RE Montgomery, R\V Mutch, \tVP Leenhouts, and 01' Zimmerman (1995) Altered fire regimes within fire-adapted ecosystems. U.S Department of the Interior - .ational Biological Service, '<\'ashington, DC.
Flannigan MD, BJ Stocks, ~IR Thretsky, and BM WoLton (2008) Impact of climatechange on fire activity and fire management in the circumboreal forest. GlobalChange Biology 15: 549-560.
GAO (2007) \oVildland fire management: Better information and a systematic pre?cess could improve agencies approach to allocating fuel reduction funds and selecting projects. GAO-07-1168, United States General Accounting Office, V\'ashington DC.
Gardner RH, MG Turner, RV O'Neill, and S Lavorel (1991) Simulation of the scale.dependent effects of landscape boundaries on species persistence and dispersal.In: MM Holland, PG Risser, and RJ Naiman (OOs) Ecotones: The Role ofLandscape Boundaries in the ~Ianagement and Restoration of Changing Envi·ronments. Chapman and Hall, New York. 76-89.
Gardner RH, DL Urban (2003) i\llodel validation and testing: Past lessons, presentconcerns, future prospects. In: CD Canham, JC Cole, and \VI< Lauenroth (OOs)Models in Ecosystem Science. Princeton Univ. Press, Princeton, NJ USA.
Greene DF and EA Johnson (1995) Long-distance wind dispersal of tree seeds. Can.J. Bot. 73: 1036-1045.
Hann \<\'J (2004) Mapping fire regime condition class: A method for watershedand project scale analysis. In: RT Engstrom, I<EM Galley, and WJ De Groot(eds) 22nd Tall Timbers Fire Ecology Conference: Fire in Temprate, Boreal,andMontane Ecosystems. Tall Timbers Research Station. 22-24.
Hann \tVJ and DL Bunnell (2001) Fire and land management planning and implementation across multiple scales. International Journal of Wildland Fire 10:389-403.
He H, RE Keane, and L Iverson (2008) Forest landscape models, a tool for under·standing the effect of the large-scale and long-term landscape processes. ForestEcology and Management 254: 371-374.
Hessburg PF, KM Reynolds, RE Keane, KM James, and RB Salter (2007) Evaluating wildland fire danger and prioritizing vegetation and fuels treatments. ForestEcology and Management 247: 1-17.
Hessl AE, C Milesi, MA Wbite, DL Peterson, and RE Keane (2004) Ecophysiologicalparameters for Pacific Northwest trees. General Tecllllical Report PNW-G'TR·618, USDA Forest Service Pacific Northwest Research Station, Portland, OR.USA.
Holsinger L, RE Keane, R Parsons, and E Karau (2006) Development of bioph)ical gradient layers. General Technical Report RMRS-GTR-175, USDA For~1
References 95
JPC~e(~~7)R~~kY MOlclllltain Research Station, Fort Collins. CO USA- Imate lange 2007 - The Phy'c IS· B. _.
versity Press, few York ew Yo k USA I a clence aSlS. Cambridge Uni-Jenkins JC a,~d R Birdsey (J998) Vali;a~ion d~ta '.
ground biomass and net primary productiVi~as~~~slln~latl~1lmo~lels: Ab~veFIA data. General Technical Report NC-212 YU~DA ~estlJnatlol~ usmg easwldetl"al Research Station, Boise, 10. USA.' orest SerVice, North Cen-
Karau E and RE Keane (2009) (in prep) B . .modeling and satellite imagery Inter t~lrn f~verlty mapPing using simulation
Karan EC and RE Keane (2007) Dete ~a. Ionia doufIlal of "Vildland Fire.d· b rmlnlllg an scape ext.ent ~IStUI" ance simluation modell·ng La d E or Succession andCE ( 00 ) . n scape . cology 22· 993-1006
Kay 2 7 Are lightning fires unnatural? A c .' .'ning ignition rates in the United Stat j. R~nTaflson of abonginal and lightProceedings of the 23rd Tall Timber:sp. n. E I' lasters and KEM Galley (eds)land and Shrubland Ecosystems 'I1 Ii ~~e b co ogy Conference: Fire in GrassFL. 16-28. . a lin ers Research Station, Tallahassee.
Keane HE, G Cary, 10 Davies. ~fD Flanni an RH Ghan, C Li, and TS Rupp (20011) A Ig.il . ardner, S Lavorel, JM Lennimodels: Spatially explicit models otfi
asSIc~tlOn of I?ndscape fire succession
Modelling 256: 3-27. re an vegetatIOn dynamic. Ecological
Keane RE, GJ Cary, and R Parsons (2003) U . . .An evaluation of approaches strategies s~nf sl.mu~atJon to map fire regimes:of \Vildland Pire 12: 309-322. ,an ImitatIOns. International Journal
Keane RE, SA Drury, E [(arau PF Hessbur dfor mapping fire hazard and risk acrossg, a',l
t. IJ<fv( Reynolds .(2008) A method
fire management. Ecological Modelling 2~~:u2~f8e scales and its application inKeane RE and l\fA Finney (2003) The simulation ~ .
ecosystem dynamics. In: TT Veblen WL Bak Jlandscape fire, chmate, andnam (eels) Fire and Global Change'. ~ er, Montenegro, and T\V SwetAmericas. Springer-Verlag I'ew Yor~nN emeerakteuES'cosystems of the \Vestern
Keane HE TL Fr· ' , ew lor. A. 32-68, escmo, MC Reeves, and J Lon (2006) I' .. •
across large regions for the LANDFIRE g a . Mapping Wildland fuelsC Fhme (eds) The LANDFIRE Proto p:o~~pe proJect. In: MG Rollins andLoc~lIy Relevant GeospatiaJ Data for \'~~Iand ~~~t: Nationally Co?sistent andService Rocky tl.Iountain Research Station. 367-3~~ Management. USDA Forest
Keane HE and L Holsinger (2006) S· I·. .modeling and ecncnrst . IInu ~tll1g bIOphysical environment for gradient
-~J em mappmg usmg the WXFlREO1en~ation and application. Research Pa er Rlvl S program: Model docu-
K Ser~oe Rocky Mountain Research Stati~n Fort~;I~TR~68~~AUSDA Forestea~le " L Hols ll1ger, and SPratt (2006b) S' I' ~ns, . 0, .
ICS using the landscape fire . . Inm atmg hlstoncal landscape dynam-Tecl . succession model LANDSUM . 40
llllcal Report RMRS-GTR-I71CD USD : - verSion .. GeneralK Research Station, Fort Collins, CO USA A Forest Service Rocky Mountain
.ane HE and EC Karau (2010) E I ..integrating fire and ec t va uatmg the.ecological benefits of wildfire by
K.ane HE, C Miller E S osy~ em models. Ecologtcal Modelling 221: 1162-1172~ressJ Represent'ing c~~tatIC~! ~ ~cI(enZie, D F~lk, .and L Kellogg (2011) 'finSimulation models G ej , IT'lslu~ ance, vegetation mteractions in landscapep . . enera ~ec 1111Cal Repo t P \V XXX
K 8Clfic orthwest Research Station P tl dr 0 I - . US Forest Service
eane HE PM. ' or an, R USA.ical' organ, and SW Running (1996) FIRE-BGC h. .
process model for simulatin fi . - a mee amstlc ecoIog-of the northern Rocky 1\1 t. g ~ SUcceSSion Oil coniferous forest landscapesDepartment of Agricultur~u~alliS. S ~ch Paper INT-RP-484, United States
, orest ervlce Intermountain Forest and Range Ex-
perimcnt Station, Ogden, UT USA. (2002) Estimating historical range andE R P and P Hessburg .' IKeane R I arsons, d' . Limitations of the sllTIulatlon appraae 1.
variation of landscape patch ynarTIlcs.
Ecological Modelling 151: 29-49. Gra '. and JJ Reardon (2005) EstimatingKeane RE. ED Reillhardt, ~ Scot.t. K. i;direct methods. Canadian Journal of
forest canopy bulk density usmg SiX
Forest Research 35: 724-739'h
(2007) Using simulated historical time series toKeane RE, ~lG Rollins, and Z Z u 55 the United States: The LAND-
prioritize fuel treatn~ents on lan~scapesda~~o 204: 485-502.FIRE prototype proJect. Ecologtcal Mo d~ ng't and stability' A reconciliation
King A\V, and SL Pimm (1~~3) Complexit)~ ~~~:~i~~n Naturalist '122: 229~239.of theoretical and empmcal results.)T~~. es for sensitivity analysis of sim
Kleijnen JP, Ham G, Rotmans J (19~2 h ec~1q~reenhouse effect. Simulation 58:ulation models: A case study 0 t e 2 0
410-4l7. . 9) Overview and use of natural variabilityLandres PB, ~lorgan ~,SwansOl~FJ (199 ms. Ecological Applications 9: 117~1l~.
concepts 111 managmg ecologlc~l ~yst~p et al. (1998) Simulation ffi.od.ellO.g III
Lauenroth \VI<, Canham CD, K111Zlgd
·ptvi G ffman (eds) Successes, LnmtatlOns,ecosystem science. In: ML Pace ~n 'S:o ~Verlag New York, New York land Frontiers in Ecosystem SCience. pnnger ,
SA. 404-415. . Ie and sustaining resources in fire.Laverty L, \~/illiams J (2000) Prot~ctmg Pt
eopForest Service Response to GAO
d t s - A cohesive stra egy. Cadapte ecosys err: SOA Forest Service, \~lashington 0 . .Report GAO/RCED 99-65 ~ D (2006) Vegetation succession modehng for the
Long DG, Loscnsky J, Beduna 1. Technical Report RMRS-GTR-175,LANDIFRE Prototype ProJect. Gene; earch Station Fort Collins, CO USA.USDA Forest Service Roc~y MountRl(20~~) FIREtV10N:\ Fire effects monitoring
Lutes DC, Keane RE, Carattl JF, e~:~i1lliCal Report'RMRS-GTR-164-CD, USDAand lllventory system. Gener~l . h Station Fort Collins, CO USA.Forest Service Rock~ Mou~ta111 ~ese~I~ f an~ual area burned in Canadian
Magnussen S (2008) JOll1t regional ~lmu ~ Ion 0 1 l' 37-53forest fires. The Open Forest S.cl~nc~v~r~~t~on' Ecolo~ical Modeling 68: 21-32.
Mayer DG, Butler DG (1993) Statlstlcaeucal pt for~sts. LeaOet Number 107, Com-McArthur AG (1967) Flr~ behavlOI JO nd Ti~ber Bureau. Sidney, Australia. .
monwealth of Austraha Forestry ~ (1996) Extrapolation problems in modehngMcKenzie 0\ Peterson DL, ~lvarado. A . International Journal of 'Wildland
fire effects at large spatial scales. reView.
Fire 6: 165-176. . wetnam TW, Peterson DL, Littell JS (2009)McKenzie 0, Sd~mol~tD, I<eane RE, ~ t bance interactions: Building ~ research
{in press} Chmatlc ch~nge and d1~~V_XXX US Forest Service Pacific Nortb-agenda. General Techlllcal Report 1 ,
west Research Station, Portland, ~R 'irSAt f fire management alternatives onMiller C, Urban DL (2000) M~de:ing ~:e~:;i~al Applications 10: 85-94. .
Sierra Nevada mIXed-com er ores . T han RJ Midgeley GF, Fragoso J~, L~1 eilson RP, Pitelka LF, Solomon AM, Na~ . 'nal to global plant migration Ul
hk H Thompson K (2005) Forecastmg reglOc e,. B" "5' 749-760. .response to clImate change.. IOSClenc~ ~. els Itr OJ Nelson (ed) Radlonu-
O'Neili RV (1973) Error analysIs of ecologIcal m~d .' Springfield Virgioia, USAelides in Ecosystems. Technical Information erVlce, '
898-908. RP (1993) Corroborating models with model properties. ForestOderwald RG, Hans a!
Ecology and Management, 62: 271-2~D (1993) CO SUME users guide. Gener
Oltmar RD, Burns 'I, Hall IN, Hanson
Technical Report PNW-GTR-304, USDA Forest Service.pacala S\-V, Tilman 0 (1994) Limiting similarity in mechanistic and spatial models
of plant competition in heterogeneous environments. The American Naturalist143: 222-257.
Peng C (2000) From static biogeographical model to dynamic global vegetationmodel: A global perspective on modelling vegetation dynamics. Ecological .\[odelling 135: 33-54.
Peters D. Herrick JE, Urban DL, Gardner RH, Breshears DD (2004) Strategies forecological extrapolation. Oikos 106: 627-663.
Pratt SO, Holsinger L, Keane RE (2006) ~lodcling historical reference conditions forvegetation and fire regimes using simulation modeling. General Technical ReportRMRS-GTR-175, USDA Forest Service Rocky Mountain Research Station, FortCollins, CO USA.
Quinlan JR (2000) Data mining tools See5 and C5.0. http://www.rulequest.com/see5-info.html.
Radeloff VC, Hammer BC, Stewart SI, Fried JS, Holcomb SS, ~lcKeefry JE (2005)The wildland-urban interface in the United States. Ecological Applications 15:799-805.
Rastetter EB, Aber JB, Peters D, Ojima D. Burke IC (2003) Using mechanisticmodels to scale ecological processes across space and time. Bioscience 53: 6877-
Rastetter EB, Ryan MG\ Shaver GR, et al. (1991) A general biogeochemical modeldescribing the responses of the C and N cycles in terrestrial ecosystems tochanges in C02, climate, and N deposition. Tree Physiology 9: 101-126.
Reinhardt ED, Keane RE, Brown JK (1997) First Order Fire Effects Model: FOFE~1
4.0 User's Guide. General Technical Report INT-GTR-344, USDA Forest Service.
Reinhardt ED I Keane RE, Brown JK (2001) Modeling fire effects. InternationalJournal of Wildland Fire 10: 373-380.
Rollins MG, Keane RE, Zhu Z (2006) An overview of the LANDFJRE Prototype Project. General Technical Report RMRS-GTR-175\ USDA Forest ServiceRocky Mountain Research Station, Fort Collins, CO USA.
Rothermel RC (1972) A mathematical model for predicting fire spread in wildlandfuels. Research Paper INT-1l5, United States Department of Agriculture\ ForestService, Intermountain Forest and Range Experiment Station, Ogden, Utah.
Running SW (2006) Is global warming causing more, larger wildfires? Science 313:927-928.
R)'kiel EJ (1996) Testing ecological models: The meaning of validation. EcologicalModeling 90: 229-244.
Ryu SR, Chen J, Zheng D, Bresee MK, Crow TR (2006) Simulating the effects ofprescribed burning on fuel loading and timber production (EcoFL) in managednorthern Wisconsin forests. Ecological Modelling 196: 395-406.
Sampson RN, Clark LR (1995) Wildfire and carbon emissions: A policy modelingapproach. Tn: American Forests, The Forest Policy Center, ,"Vashington DC1-23.
teele BM, Reddy SK, Keane RE (2006) A methodology for assessing departureof current plant communities from historical conditions over large landscapes.Ecological Modelling 199: 53-63.
..et?am TW, CD Allen, aod JL Betancourt (1999) Applied historical ecology:Th Usmg the past to maoage for the future. Ecological Applications 9: 1189-1206.
ornton P, Law B, Gholz HL, Clark KL, et al. (2002) Modeling and measuring(he effects of disturbance history and climate on carbon and water budgets inevergreen needleleaf forests. AgricuJtural and Forest ~leteorology 113: 185-222.
96Chapter 4 Landscape Simulation in Fire ~I{anagement
References 97
------=-------98 Chapt.er 4 Landscape Simulat.ion in Fire ;"lanagement.
Thornton PE, Running S\V·I'VhilC ~lA (1~9~mGp~~e~:~~:ftls~r~~~~SalO~fd~~~rl~11~~orological variables over arge regions 0 C .
190, 214-251. . I· C b·· 19 surface- audTh t PE (1998) Regional ecosystem SlnlU allon: om lJm . . d
orn on I· ka b t 'ee terrestnal energy ansatellite-based observations to sl.U?~ In. ges e \\ 11. ~IT USA.
. b d t PhD Dissertation. Umverslty of ~lontana. :-'1Jssoula, 'mass u ge s. 1 P I A Vos E Peterson D. I< nops J
Tilman 0, Reich PB, Phillips B, ~lenton ~, ale, 'colo 81: 2680-2685.(2000) Fires suppression and ecosystem carbon storage. E gy 10 '86- 1996-
Urban DL (2005) Modeling ecological processes across scales. Eco g) .
2006R· I" R . SW (1998) Forest Ecosystem" Analysis at ~lultiple Scales
\Vanng "I, unnmg . CA USA(2nd). Academic Press, Inc, San Dleg~" t 'm T\t\f (2006) \.\farming and earlier
\,Vesterling AL, Hidalgo HG, Cayan DR, ~~~fin.a ctivity Science 313: 940-943.spring increase in western US forest WI He a .. I henolo model for
\Vhit.e MA, Thornton P, Running S\V.(1997) A ~OI;~llle~~ta, PiabilitY~YGlobal Biomonitoring vegetation responses t.o lllterannua c una IC \81'
. geochemical ~Yc1e~11'L217.26j· Whitlock C (2000) Simulating historical vari-\Vlmberly ~IC, pies 'fongll r' t· tl e Oregon Coast Range. ConservatIOn
ability in the amount 0 °( ores 10 1
ZhuB~~I~~e\~";'~7J~8ghlen D, Kost J, Chen S, Tolk B, RollinS ~~~,~~~~k.~I:ping existing vegetation composItIon and structure. ~eneral h Station ~ortRMRSGTR-175, USDA Forest Service Rocky 110untam Researc ,
Collins, CO USA.
p:
Chapter 5 Using Landscape Disturbance andSuccession Models to Support Forest
Management
Eric J. Gustafson', Brian R. Sturtevant, Anatoly Z. Shvidenko
and Robert M. Scheller
AbstractManagers of forested landscapes must account for multiple, interacting ecological processes operating at broad spatial and temporal scales.These interactions can be of such complexity that predictions of futureforest ecosystem states are beyond the analytical capability of the humanmind. Landscape disturbance and succession models (LDSM) are predictive and analytical tools that can integrate these processes and providecritical decision support information. We briefly review the state of theart of LDSMs and provide two case stuclies to illustrate the applicationand utility of one LDSM, LANDIS. We conclude that LDSMs are ableto provide useful information to support management decisions for anumber of reasons: (i) they operate at scale that is relevant to manyforest management problems, (ii) they account for interactions anlOngecological and anthropogenic processes, (iii) they can produce objectiveand comparable projections of alternative management options or various global change scenarios, (iv) LDSMs are based on current ecologicalknowledge and theory, (v) LDSMs provide a vehicle for collaborationamong decision-makers, resource experts and scientists, (vi) LDSr-.1s arethe only feasible research tool that can be used to investigate long-term,large area dynamics.
• Eric J. Gustafson: Institute for Applied Ecosystem Studies, USDA Forest Service, NorthTh Research Station, 5985 Highway K, Rhinelander, WI USA. E-mail: [email protected] ~.S. ~overnment's right to retain a non-exclusive, royalty-free licence in and to any
Olpynght. IS acknowledged.
Chao LiRaffaele LafortezzaJiquan Chen
Landscape Ecology inForest Management andConservation
Challenges and Solutions for Global Change
With 73 figures
~ -;t 'fIt i ~ Pi. jJ.• HIGHER EDUCATION PRESS ~ Springer
EditorsDr. Chao LiCanadian Wood Fibre CentreCanadian Forest ServiceNatural Resource~ Canada5320-122 Street EdmonlonAlberta Canada T6H 355E-mail: Chao.Li@NRCan-R.!\iCan.gc.ca
Prof. Jiquan ChenLandscape Ecology & Ecos)stem Science (LEES)Department of Environmental Sciences (DES)Bowman-Odd)' Labor:llorie..;. ~lail Stop 6O-lUni\'crllit)' of Toledo. ToledoOH 43606-3390_ USAE-mail: [email protected]
Dr. Raffaele LafortezzagreenLab Dept. Scienze delleProduzioni VcgetaliUniversita degli Studi di BariVia Amendola I65/A 70126 Bari, hal)'E·mail: [email protected]
Foreword
ISBN 978-7-04-029136-0Higher Education Press, Beijing
ISBN 978-3-642-12753-3 c-ISBN 978-3-642-12754-0Springer Heidelberg Dordrecht London New York
Library of Congress Conlrol Number: 2010925386
©.Highcr ~duca~ion Press. Beijing and Springer-Verlag Berlin Hcidelberg 2011Tlll~ work IS SU?Jcct to cop~'right. All rights arc resen'cd. whether the whole or part of the material isconeeme~. speclfi~all)' the ng.hls of translation. reprinting. reuse of illustrations. recitation, broadcasting.reproduction on ~llcrofil.m or lfl any other way. and storage in data banks. Duplication of this publicationor pa~ ~hereof IS pem~ll1ed onl), un~e~ the provisions of the Gemlan Copyright Law of September 9.1.965, lfl ItS curre~t vcrslon. and permission for use must always be obtained from Springer. Violations areliable to prosecution under the German Copyright Law.
The u.se ofgeneral descriptive namcs. registered names. trademarks. ele. in this publication does nOf impl)'even III the .abscnce of a specific statement, that such names are exempt from the relevant protective law~and rcgulatlons and therefore free for general usc.
Corer desigll: Frido Sreinen-Broo. ESlUdio Calamar. Spain
Printed on acid-free paper
Springer is pan of Springer Science + Business Media (www.springer.com)
Like many others. my first exposure to the science of landscape ecologfrom the book entitled Landscape Ecology published by Richard Forma~lichel Godron in 1986. For me. this was a new and exciting way for lookthe world in which we live. It was obvious to me after reading this boothe science of landscape ecology had much to offer natural resources IllaoBut it is also important to recognize that a '"landscape perspective" ha:around for a long time in a variety of sources and in a variety of placf'~
example is a book published in 1962 by Paul B. Sears. an early ecololthe United States, entitled The Living Landscape. In this book writta general audience: Sears described with great elegance why a .Ilamperspective" is relevant (page 162):
"Compared to the noblest work of human genius: the landscape aboffers endless variety of interest and challenge. It is more than somet hlook at: it is something to comprehend and interpret. \Ve are inseparpart of it: and it is equally a part of us. Our destinies are linked) andNature will assuredly have the final judgment, modern man has the podetermine whether it will be thumbs up or down.: l
Aside from the gender bias that was common to that period. modemanity indeed will be making important choices that will profoundlyour children and many subsequent generations. Those choices should becated on the best available scientific knowledge. The current book editedLafortezza, and Chen is another valuable contribution to comprehendilinterpreting forested landscapes. It represents the latest work resultin.the bi-annual meetings sponsored by the IUFRO Landscape Ecology \VParty (08.01.02). The strength of this book is in the fact tbat it reflecexperience and knowledge gained by scientists in 15 different countries.provides a rich source of international literature.
It would be naive. howc\·er. to think that all we need to cure our citing environmental and human problems is to do good science. Humallto recognize what Sears stated so well in his book - "\\"e are inseparablyof it, and it is equally a part of us," ntil this linkage is clearly estill