CHARACTERIZING FIREBRAND EXPOSURE DURING WILDLAND-URBAN INTERFACE FIRES. | NIST

This study examines the size distribution and other characteristics of firebrand exposure during the 2007 Angora Fire, a severe Wildland-Urban Interface (WUI) fire in California. Of the 401 houses that received direct interface fire exposure 61% were destroyed and 30% did not burn at all. The ignition of buildings by wind-blown embers, known as firebrands, and the starting of “spot fires” in unburned vegetation ahead of wildfires has been observed for centuries and studied extensively for decades. However, it is not yet possible to empirically quantify the exposure severity or describe how many firebrands of what size and over what duration and distance are causing ignition problems of concern. A seemingly rare opportunity to gather empirical firebrand data from an actual interface fire evolved in the days immediately following the Angora Fire. Firebrand size distributions are reported and compared to firebrand size distributions from experimental firebrand generation in both recent laboratory building ignition studies conducted by National Institute of Standards and Technology (NIST) and from historical firebrand field studies. Such data is needed to form the basis of effective and appropriate interface fire hazard mitigation measures as well as modeling fire spread. Comparisons are made to current wildfire protection building construction regulations and test standards. The most salient results of this study are the documentation of the consistently small size of firebrands and the close correlation of these results with the sizes of experimentally generated firebrands. INTRODUCTION AND BACKGROUND The quantitative results of this study are brief and narrowly focused on the sizes of firebrands (embers) lofted from burning vegetation and/or burning buildings, transported ahead of the propagating fire, and deposited on unburned fuel with some potential for igniting a new fire. This firebrand and spot fire phenomenon has been studied extensively and identified as an important fire spread mechanism for a number of “fire problems.” 1,2 One reason for conducting this investigation is similarly brief, it was identified as a specific work item in a Governor’s Blue Ribbon Fire Commission Task Force Report on disastrous interface fires in California. 3 There is, however, a broader context to the quantitative results that warrants a broader discussion. While the topic of firebrand ignition mechanisms in general, and the phenomenology of interface fire spread in particular, has been extensively studied, the problem of disastrous interface fire losses is widely perceived to be getting increasingly worse. To the extent that reduced interface fire loss is a desirable end product of fire investigations, studies should be clearly linked to the context and solution of a well identified fire problem. Firebrands have been observed in a wide range of fire problems. Firebrand ignition of “fire proof” buildings has been reported after the 1906 San Francisco earthquake 4 and following the nuclear attack on Hiroshima. 5 The problem of wildfire spreading to readily ignitable buildings has been specifically described for over a hundred years with thousands of buildings destroyed. Building loss on wildfires is a fire problem that has historically received a plethora of labels leading to confusion about the nature of the problem. 6 This problem has more recently been consistently and widely referred to as the “Wildland-Urban Interface.” However, the term “Wildland-Urban Interface” has a wide range of definitions including geographical descriptions of areas having a close proximity of vegetation to buildings with no history of large interface fires (nor any clearly identified risk of conflagrations). 7 The Wildland-Urban Interface can be defined simply in terms of fire spread mechanisms that provide clear paths to hazard mitigation solutions. Butler (1974) coined the “fire interface” terminology in publishing the first comprehensive description of wildfire related disastrous building loss as specific fire problem. 8 From a fire-loss reduction context, the interface fire problem can be most effectively defined as a sequence of factors; 1) exterior vegetation fire exposure during extreme weather conditions with, 2) rapid fire spread to many readily ignitable buildings, 3) overwhelming fire protection resource, 4) resulting in disastrous losses of buildings. For the characterization of firebrand exposure to meaningfully contribute to reduced interface fire losses, it must be directly related to the context and dominant ignition mechanisms of interface fire spread, as Cohen (2000) did for crown fire exposure. 9 For example, studies of very large firebrands or firebrand ignition of lumber at standard fuel moisture content (e.g. 12%) may contribute very little to fire loss reductions if the majority of buildings are ignited by very small firebrands under extreme weather conditions with very low fuel moisture. Conversely, identifying and expending limited hazard mitigation resources on specific building ignition mechanisms, even those involving very small firebrands under extreme conditions, may not be effective in reducing disastrous losses if the specified ignition mechanism is very rare or is overshadowed by other factors. There is substantial fire research to support the widely held observation that firebrand exposure has a significant role in the spread of disastrous interface fires 10 . Every multivariate retrospective study of interface fire building survival indicates firebrand exposure as a problem and those involving wood roofing have found a statistical correlation between wood roofing and increased building loss. 11,12,13,14 The most recent of these studies (1990 Santa Barbara Paint fire) found that for buildings without evidence of fire suppression, there was an 82% increase in the proportion of buildings surviving the fire where houses had: 1) a non-flammable roof and; 2) at least 10 m (30 ft) of brush clearance. The statistical analysis revealed that these two factors were mutually and independently associated with building survival and accounted for 59% of the variability in building survival on that fire. 15 It has been widely reported that firebrand exposure is a dominate fire spread mechanism in wood roof conflagrations which implies empirical support that firebrand exposure exists to some extent during interface fires with or without the wood roof factor being present. Recent building ignition modeling, full-scale crown fire exposure experiments, and case studies indicate that radiant heat transfer from forest fires is significantly less important as an interface fire buildingignition mechanism than previously assumed. There is limited but growing experimental evidence for firebrand exposure as an ignition mechanism of buildings not involving wood roofing. 16 Observational studies have long identified a building ignition mechanism where very small firebrands penetrate under non-combustible tile roof covering to ignite the building. 17,18 Recent full-scale firebrand exposure investigations of building ignition provided experimental confirmation of this ignition mechanism. 19 This work utilized an experimental apparatus recently developed by Manzello et al. known as the NIST Firebrand Generator (NIST Dragon) used to investigate ignition vulnerabilities of structures to firebrand exposure. The NIST Dragon is able to generate a controlled and repeatable size and mass distribution of glowing firebrands. The experimental results generated from the marriage of the NIST Dragon to the Building Research Institute’s (BRI) Fire Research Wind Tunnel Facility (FRWTF) have uncovered the vulnerabilities that structures possess to firebrand showers for the first time. 20 ,21 ,22 These detailed experimental findings are being considered as a basis for performance-based building standards with the intent of making structures more resistant to firebrand attack. The substantial public interest in this issue was illustrated during the recent triennial amendments to the International Building Code by the California Building Standards Commission. The issue of firebrand size and related building ignition mechanisms on interface fires received more public comments and Office of the State Fire Marshal agency response than any other issue during final consideration of the fire and life safety provisions for adoption of the 2010 California Building Code. 23 The firebrand size distributions developed from the Angora Fire are believed to be the first such data from an interface fire. These results together with the experimental work from NIST provide additional steps toward characterizing interface firebrand exposure. Angora Fire Study Area, Weather, Fuels and Interface Fire Exposure The Angora Fire started in a stand of dense unmanaged conifer forest located at the south end of Lake Tahoe approximately 240 km (150 mi) east-northeast of San Francisco California near the Nevada state boundary. The fire burned 1,243 ha (3,072 ac) and approximately 353 buildings of all types. Within the fire perimeter there were 401 houses with direct fire exposure, and of the 282 houses that burned, 87% were completely destroyed. The area around the unburned houses and the 13% of damaged houses were the source of the firebrand data collection portion of this study. The probability that firebrands will potentially ignite buildings or “spot fires” in vegetation requires that firebrands will be produced, lofted, transported, deposited on a receptive target fuel bed. This firebrand propagation phenomenon is dominated by weather conditions, the configuration and condition of both the source and target fuel beds, and the fire intensity. These factors were documented to the extent possible on the Angora Fire during the course of routine fire incident management and post-fire damage assessment. 24 The general weather over the Angora Fire area was dominated by a cold front moving th