Wildland-urban interface (W-UI) fires nce largely considered a Cali- When an apparently all-encompass- are a significant concern for federal, fornia problem, residential fire ing, seemingly unstoppable W-UI fire losses associated with wildland occurs, the rapid involvement of many state, and local land management and Ofires gained national attention homes over a wide area produces a sur- fire agencies. Research using in 1985 when 1,400 homes real impression; some homes survive modeling, experiments, and W-UI were destroyed nationwide (Laughlin amid the complete destruction of sur- case studies indicates that home and Page 1987). The wildland fire rounding residences. After the 1993 ignitability during wildland fires threat to homes is increasing and is Laguna Hills fire, some termed this depends on the characteristics of the commonly referred to as the wildland— seemingly inexplicable juxtaposition a home and its immediate surroundings. urban interface (W-UI) fire problem. “miracle.’ Miracles aside, the charac- These findings have implications for Since 1990, W-UI fires have threatened teristics of the surviving home and its hazard assessment and risk mapping, and destroyed homes in Alaska, immediate surroundings greatly influ- Arizona, , Colorado, Florida, enced its survival. effective mitigations, and iden- Michigan, New Mexico, New York, Wildland fire and home ignition re- tification of appropriate responsibility and Washington. Extensive or severe search indicates that a home’s exterior for reducing the potential far home fires in Yellowstone in 1988, Oakland and site characteristics significantly in- lass caused by W-UI fires, in 1991, and Florida in 1998 attracted fluence its ignitability and thus its much media coverage and focused chances for survival. Considering home national attention on wildland fire and site characteristics when designing, By Jack D. Cohen threats to people and property building, siting, and maintaining a Federal, state, and local land man- home can reduce W-UI fire losses. agement and fire agencies must directly and indirectly protect homes from W-UI Fire Loss Characteristics within and adjacent to W-UI residential fire losses differ wildlands. Davis (1990) indicated that from typical residential fire losses. since the mid-1940s, a major population Whereas residential fires usually increase has occurred in or adjacent to involve one structure with a partial loss, forests and woodland areas. Increasing W-UI fires can result in hundreds of residential presence near fire-prone totally destroyed homes. Particularly wildlands has prompted agencies to during severe W-UI fires, numerous take actions to reduce W-UI fire losses.

Journal of Forestry 15 homes can ignite in a very short time. The usual result is that a home either survives or is totally destroyed; only a few structures incur partial damage (Foote 1994). The W-UI Fire commonly originates in wildland fuels. During dry, windy conditions in areas with continuous fine fuels, a wildland fire can spread rapidly, outpacing the initial attack of firefighters. If residences arc nearby, a wildland fire can expose numerous homes to flames and lofted burning embers, or firebrands. A rapidly spreading wildland fire coupled with highly ignitable homes can cause many homes to burn simul- taneously. This multistructure involvement can overwhelm fire protection Capabilities and, in effect, result in unprotected residences. Severe W-UI fires can destroy whole neighborhoods in a few hours—much Figure 1. The structure survival process faster than the response time and suppression capabilities of even the best—equipped and staffed firefighting agencies. For example, 479 homes were destroyed during the 1990 Painted Cave fire in Santa Barbara, most of them within two hours of the initial fire report. The 1993 Laguna Hills fire in southern California ignited and burned nearly all of the 366 homes destroyed in less than five hours. Whether a home survives depends initially on whether it ignites; if ignitions with continued burning occur, survival then depends on effective fire suppression. Figure 1 shows that home survival begins with attention to the factors that influence ignition. These factors determine home ignitability and include the structure’s exterior materials and design combined with its exposure to flames and firebrands. The lower the home ignitability the lower the chance of incurring an effective ignition.

Ignition: A local Process Ignition and spread of fire, whether on structures or in wildland vegetation, is a combustion process. Fire spreads as a continuing ignition process whether from the propagation of flames or from the spot ignitions of firebrands. Unlike a flash Figure 2.The incident radiant heat flux is shown as a function of a wall’s distance flood or an avalanche, in which a mass from a flame 20 meters high by 50 meters wide, uniform, constant, 1,200 K, black- engulfs objects in its path, fire spreads body. The minimum time required for a piloted wood ignition is shown given the because the requirements for corresponding heat flux at that distance.

Journal of Forestry 16 combustion are satisfied at locations context of home ignitability involves receive from flames depending on its along the path. The basic requirements mitigating the fuel and heat compo- distance from the fire. The incident for combustion—the fire triangle—are nents sufficiently to prevent ignitions. radiant heat flux, defined as the rate of fuel, heat, and oxygen. An However, the question of sufficiency radiant energy per unit area received at insufficiency of any one of the three (or efficiency) remains: How much, or an exposed surface, decreases as the components, which can occur over a perhaps more appropriately, how little distance increases. relatively short distance, will prevent a fuel and heat reduction must be done Figure 2 also shows that the time specific location from burning. “Green to effectively reduce home ignitions? required for ignition depends on the islands” that remain after the passage To answer this question, we must first distance to a flame of a given size. At of a severe, stand-replacement fire quantify the heat source in terms of 40 meters the radiant heat transfer is demonstrate this phenomenon. the fuel’s ignition requirements; less than 20 kilowatts per square meter Commonly one can find a green, specifically, how close can flames be living tree canopy very close to a to a home’s wood exterior before an completely consumed canopy. ignition occurs? The requirements for combustion Research Insights equally apply to the W-UI fire situa- Diverse research approaches are tion. In the wildland fire context, fire providing clues for assessing the fuel managers commonly refer to vegeta- and heat requirements for residential tion as fuel. However, for the specific ignitions. Structure ignition modeling, context of W-UI residential fire losses, fire experiments, and W-UI fire case 2 a house becomes the fuel. Heat is sup- studies indicate that the fuel and heat (kW/m ), which translates to a mini- plied by the flames of adjacent required for home ignitions only mum piloted ignition time of more burning materials that could include involve the structure and its immediate than 10 minutes. firewood piles, dead and live surroundings—the home ignitability Ten minutes, however, is signifi- vegetation, and neighboring structures. context. cantly longer than the burning time of Firebrands from upwind fires also Modeling. The Structure Ignition wildland flame fronts at a location. supply heat when they collect on a Assessment Model (SIAM) (Cohen Large flames of wildland fires house and adjacent flammable 1995) is currently being developed to typically depend on fine dead and live materials. The atmosphere amply asses the potential for structure vegetation, which limits the intense supplies the third necessary ignitions from flame exposure and burning duration at a specific location component, oxygen. firebrands during W-UI fires. One to less than a few minutes. Recent A wildland fire cannot spread to function of SIAM is to calculate the experiments have homes unless the homes and their ad- total heat transferred, both radiation demonstrated a location-specific jacent surroundings meet those com- and convection, to a structure for burning duration of 50 to 70 seconds. bustion requirements. The home ig- varying flame sizes and from varying Experiments. Field studies con- nitability determines whether these re- distances. From the calculated heat ducted during the International Crown quirements are met, regardless of how transfer, SIAM calculates the amount Fire Modelling Experiment intensely or fast—spreading distant of heat over time that common (Alexander et al. 1998) provide data fires are burning. To use an extreme Piloted ignition When wood is for comparisons with SIAM model example, a concrete bunker would not sufficiently heated, it decomposes estimates. Total heat transfer ignite during any wildland fire to release combustible volatiles. At (radiation and convection) and ignition situation. At the other extreme, some a sufficient volatile—air mixture, a data were obtained from heat flux highly ignitable homes have ignited small flame or hot spark can ignite sensors placed in wooden wall without flames having spread to them. it to produce flaming; thus, a sections. These homes directly ignited from piloted ignition. The instrumented walls were lo- firebrands. exterior wood products can sustain be- cated on flat, cleared terrain at 10, 20, Firebrands are a significant ignition fore the occurrence of a piloted and 30 meters downwind from the source during W-UI fires, particularly ignition (Tran et al. 1992). edge of the forested plots. The wall when flammable roofs are involved. Based on severe-case assumptions section at 10 meters was 2.44 meters Foote (1994) found a significant of flame radiation and exposure time, wide and 2.44 meters high with a 1.22- difference in home survival solely SIAM calculations indicate that wild- meter eave and roof section (fig. 3a). based on roof f1ammability. Homes land flame fronts comparable to Exterior plywood (T-1-11) covered the with nonflammable roofs had a 70 crowning and torching trees (flames wall with oriented-strand board percent survival rate compared with 20 meters high and 50 meters wide) covering the roof section and the eave 19 percent for homes with flammable will not ignite wood surfaces at soffit. Trim boards were solid wood roofs. Davis (1990) reported similar distances greater than 40 meters with wood fiber composition board on results related to roof flammability. (Cohen and Butler, in press). Figure 2 the cave fascia. None of the materials Reducing W-UI fire losses in the shows the radiant heat a wall would were treated with fire retardant.

Journal of Forestry 17

The forest was variably composed of an sensors placed 1 meter apart in the wall. ficient for ignition and corresponding overstory of jack pine (Pinus banksiana) The amount of heat received by the wall to no actual occurrence of a wall about 14 meters high with an understory of increased as the flame front approached ignition. Therefore, a home at some black spruce (Picea mariana). The and decreased as the fine vegetation was distance from a large flame front, such spreading crown fire produced flames consumed. The initial heat flux “spike” as a crown fire, may not receive approximately 20 meters high. Figures 3b was caused by a nonuniform crowning sufficient energy to meet the minimum and 3c show examples of the experimental flame front. for ignition over any time period. In crown fire. The flux-time integral shown in addition, a home closer to a large Five burns were conducted where wall figure 4 indicates whether sufficient flame front can receive a high heat sections were exposed to a spreading heating has occurred to pilot-ignite wood flux (for example, 46 kW/m2 as shown crown fire. As the crown fires reached the (Tran et al. 992). SIAM uses the flux- in figure 4), but without the necessary downwind edge of the plot, turbulent time integral for calculating ignition duration to meet the threshold for flames extended into the clearing beyond potential, a correlation of the incident ignition. the forest edge. In two of the five burns, heat flux and the time required for pi- The flux-time integral plot flames extended beyond 10 meters to loted wood ignition. indicates the duration of the heat make contact with the 10-meter wall The flux-time correlation identifies transfer relevant to ignition. The heat section. When flame contact occurred, the two principal ignition criteria: (1) A transfer duration relevant to ignition 10-meter walls ignited; however, without minimum heat flux of 13 kW/m2 must combines the heat transfer from the flame contact, only scorch occurred, as occur before a piloted ignition can occur approaching crown fire plus the shown in figure 3d. The wooden panels at for any exposure time, and (2) piloted burning time of the fire after it has 20 meters experienced light scorch when ignition depends on attaining a critical reached the end of the plot. The flames extended beyond 10 meters from heating dosage level (heat transfer and observed time required for the flux- the experimental plot, and no scorch from its duration). These criteria are graphed time integral to increase from zero to the other burns. The 30-meter wall section in figure 4. The flux-time integral only its maximum value corresponds to the had no scorch from any of the crown fires. increases for incident heat fluxes greater heat transfer duration significant for Figure 4 displays the average total in- than the minimum of 13 kW/m2, and the ignition. Figure 4 indicates a duration cident heat flux (radiation and convection flux-time integral threshold value of of 65 seconds (flux-time plot from 75 combined) corresponding to the wall at 10 11,500 is shown as the ignition thresh- seconds to 140 seconds). meters (fig. 3d) and the crown fire shown old. As seen in the figure, the flux-time Case studies. Case studies of actual in figures 3b and 3c. The average total integral does not reach the ignition W-UI fires provide an independent incident heat flux is calculated from two threshold, indicating an exposure insuf- comparison with SIAM and the crown

Journal of Forestry 18 fire experiments. The actual fires incorporate a wide range of fire exposures. The case studies chosen examine significant factors related to home survival for two fires that destroyed hundreds of structures. The resulted in 484 homes destroyed (Howard et al. 1973) and the Painted Cave fire destroyed 479 homes (Foote 1994). Analyses of both fires indicate that home ignitions depend on the characteristics of a structure and its immediate surroundings. Howard et al. (1973) observed 86 percent survival for homes with nonflammable roofs and a clearance of 10 meters or more.

Dicussion A comparison of the SIAM model calculations in figure 2 with the observed heat flux from the experimental crown fire in figure 4 indicated that the model overestimates the heat flux. The model calculation at 10 meters reveals a radiant heat flux of 70 kW/m2, which exceeds the highest total heat flux of 46 kW/m2 observed At the 10-meter wall section in figure 4. SIAM calculations Figure 4. Actual average total incident heat flux and flux-time integral for the crown fire and 10-meter wall section shown in figure 3.

Journal of Forestry 19 overestimate the heat transfer because rate for homes with nonflammable roofs largely independent of wildland fuel the severe-case assumptions designate a and 10- to 20-meter vegetation management issues. This conclusion has homogeneous, black-body radiating clearances included fire-brands as an significant implications for the actions flame front. Real flame fronts do not ignition factor, thus indicating that and responsibilities of homeowners and meet these assumptions and produce a firebrand ignitions also depend on the fire agencies, such as defining and significantly smaller radiant heat flux ignition characteristics of the home and locating potential W-UI fire problems by comparison. For a given flame front, the adjacent flammable materials. (for example, hazard assessment and the SIAM calculations represent an mapping), identifying appropriate extreme-case estimate of radiant heat Conclusions mitigating actions, and determining who transfer, and thus an extreme-case The key to reducing W-UI home fire must take responsibility for home estimate of ignition potential. losses is to reduce home ignitability. ignitability Given the duration of the experi- SIAM modeling, crown fire experi- W-UI fire loss potential. Because mental heat flux (65 seconds), we can ments, and case studies indicate that a home ignitions depend on home ig- calculate the heat flux and correspond- home’s structural characteristics and its nitability, the behavior of wildland fires ing distance required for ignition. At 65 immediate surroundings determine a beyond the home or community site seconds, the ignition time graph (fig. 2) home’s ignition potential in a W-UI does not necessarily correspond to indicates ignition at a flame distance of fire. Using the model results as guid- W-UI home fire loss potential. Homes less than 30 meters. If the heat flux ance with the concurrence of experi- with low ignitability can survive high- duration is extended by a factor of five ments and case studies, we can con- intensity wildland fires, whereas highly to 325 seconds, the flame distance for clude that home ignitions are nor likely ignitable homes can be destroyed during ignition is less than 40 meters. By unless flames and firebrand ignitions lower-intensity fires. comparison, the 10-meter wall sections occur within 40 meters of the structure. This conclusion has implications for in the crown fire experiment did not This finding indicates that the spatial identifying and mapping W-UI fire ignite without flame contact and all scale determining home ignitions problem areas. Applying the term burns produced little or no scorch to corresponds more to specific home and wildland-urban interface to fire losses wall sections at 20 and 30 meters. The community sites than to the landscape might suggest that residential fire threat W-UI fire case studies indicated ap- scales of wildland fire management. occurs according to a geographic proximately 90 percent survival with a Thus, the W-UI fire loss problem location. In fact, the wildland fire threat vegetation clearance on the order of 10 primarily depends on the home and its to homes is not a function of where it to 20 meters for homes with nonflam- immediate site. happens related to wildlands, but rather mable roofs. Thus, the case studies sup- Consequently if the community or to how it happens in terms of home port the general flame-to-structure dis- borne site is not considered in reducing ignitability. Therefore, to reliably map tance range of 10 to 40 meters as found W-UI fire losses, extensive wildland the potential for home losses during through modeling and experiments. fuel reduction will be required. For wildland fires, home ignitability must However, firebrands can also cause highly ignitable homes, effective wild- be the principal mapping characteristic. homes to ignite during wildland fires. land fire actions must riot only prevent The home threat information must Although firebrands capable of ignition fires from burning to home sites, but correspond to the home ignitability can originate from a fire several kilo- also eliminate firebrands that would ig- spatial scale, that is, those character- meters away, homes can only be threat- nite the home and adjacent flammable istics of a home and its adjacent she ened if the firebrands ignite the home materials. To eliminate firebrands, within 40 meters. directly or ignite adjacent flammable wildland fuel reductions would have to Home fire loss mitigation. W-UI materials that then ignite the home. prevent firebrand production from home losses can be reduced by focusing Analyses of potential home ignitions wildland fires for a distance of several efforts on homes and their immediate using modeling, experiments, and case kilometers away from homes. surroundings. At higher densities where studies did not explicitly address neighboring homes may occupy the firebrand ignitions. However, firebrand Management Implications immediate surroundings, loss reductions ignitions were implicitly considered Because home ignitability is may necessarily involve a community. If because of the firebrand exposures that limited to a home and its immediate homes have a sufficiently low home occurred during the crown fire surroundings, fire managers can ignitability, a community exposed to a experiments and the case studies. The separate the W-UI structure fire loss severe wildfire can survive without major experimental crown fires provided a problem from other landscape-scale fire destruction. Thus, there is a need to firebrand exposure that resulted in spot fire management issues. The home and examine the reduction of wildland fuel ignitions in the dead wood and duff its surrounding 40 meters determine hazard for the specific objective of home around the wall sections hut not directly home ignitability, home ignitions protection. There are various land on the walls. In the case studies, depend on home ignitability, and fire management reasons for conducting firebrand ignitions occurred throughout losses depend on home ignitions. Thus, wildland vegetation management. the areas affected by the Bel Air and the W-UI fire loss problem can be However, when considering the use of Painted Cave fires. The high survival defined as a home ignitability issue wildland fuel

Journal of Forestry 20 hazard reduction specifically for pro- The fire services should become a COHEN, J., and J. SAVELAND. 1997. tecting homes, an analysis specific to community partner providing Structure ignition assessment can help reduce fire damages in the W-UI. Fire home ignitability should determine the homeowners with technical assistance Management Notes 57(4):19-23. treatment effectiveness. as well as fire response in a strategy of DAVIS, J.B. 1990. The wildland-urban Responsibility for home ignitability. assisted and managed community self- interface: Paradise or battleground? Journal If no or prescribed fires oc- sufficiency (Cohen and Saveland 1997). of Forestry 88(1):26-31. curred, the wildland fire threat to resi- For this approach to succeed, it must be DRYSDALE, D. 1985. An introduction to fire dynamics. New York: John Wiley & Sons. dential development would not exist. shared and implemented equally by FOOTE, E.I.D. 1994. Structure survival on the However, our understanding of the fire homeowners and the fire services. 1990 Santa Barbara “Paint” fire: A ecology for most of North America in- retrospective study of urban-wildland dicates that fire exclusion is neither Literature Cited interface fire hazard mitigation factors. MS possible nor desirable. Therefore, ALLXANDER, M.E., B.J. STOCKS, B.M. thesis, University of California at Berkeley. HOWARD, R.A., D.W. NORTH, F.L. homeowners who live in and adjacent to WOTTON, M.D. FLANNIGAN, J.B. TODD, B.W. BUTLER, and R.A. OFFENSEND, and C.N. SMART. 1973. the wildland fire environment most take LANOV1LLE. 1998. The International Decision analysis of fire protection strategy primary responsibility for ensuring that Crown Fire Modelling Experiment: An for the Santa Monica mountains: An initial their homes have sufficiently low home overview and progress report. In assessment (on file). Menlo Park, CA: ignitability. Homes should not be Proceedings of the Second Symposium on Stanford Research Institute. Fire and Forest Meteorology, 20-23. LAUGHLIN, J., and C. PAGE, eds. 1987. considered simply as potential victims Boston: American Meteorological Society. Wildfire strikes home! Report of the of wildland fire, but also as potential COHEN, J.D. 1995. Structure ignition national wildland/urban fire protection participants in the continuation of the assessment model (SIAM). In Proceedings conference. NFPA SPP-86. Quincy, MA: fire at their location. of Biswell Symposium: Fire Issues and National Fire Protection Association. A change needs to take place in the Solutions in Urban Interface and Wildland TRAN, H.C., J.D. COHEN, and R.A. CHASE. Ecosystems, 85-92. General Technical 1992. Modeling ignition of structures in relationship between homeowners and Report PSW-158. Albany, CA: USDA wildland/urban interface fires. In the fire services. Instead of home-re- Forest Service, Pacific Proceedings of the First International Fire lated presuppression and fire protection Southwest Research Station. and Material Conference, 253-62. London: responsibilities residing solely with fire COHEN, J.D., and B.W. BUTLER. In press. Inter Science Communications Limited. agencies, homeowners must take the Modeling potential ignitions from flame radiation exposure with implications for _ principal responsibility for ensuring wildland/urban interface fire management. Jack D. Cohen (e-mail: jcohen/rmrs_missoula adequately low home ignitability. In Proceedings of the 13th Conference on @fs.fed.us) is research physical scientist, Fire and Forest Meteorology. Fairfield, USDA Forest Service, Rocky Mountain WA: International Association of Wildland Research Station, Fire Sciences Laborator, PO Fire. Box 8089, Missoula, MT 59808.

Journal of Forestry 21