Defining the Wildland–Urban Interface

ﬁre

Deﬁning the Wildland–Urban Interface

Susan I. Stewart, Volker C. Radeloff, Roger B. Hammer, and Todd J. Hawbaker

Federal wildland fire policy in the United States has been substantially revised over the past 10 years (HFRA), which became law in 2003—reit- and new emphasis has been given to the wildland–urban interface (WUI), which creates a need for erated the need for resource managers to information about the WUI’s location and extent. We operationalized a policy definition published in work with communities and homeowners in the Federal Register (US Department of the Interior [USDI] and US Department of Agriculture [USDA]), the WUI to reduce the risks associated with 2001, Urban wildland interface communities within vicinity of federal lands that are at high risk from wildfire. wildfire. Fed. Regist. 66(3):751–777) to create national maps and statistics of the WUI to guide The increasing national fire policy fo- strategic planning. Using geographic information system analysis, we evaluate the national WUI by cus on the WUI came in response to recent altering the definition’s parameters to assess the influence of individual parameters (i.e., housing housing trends in the United States. Home- density, vegetation type and density, and interface buffer distance) and stability of outcomes. The most owners want to be near open space and in sensitive parameter was the housing density threshold. Changes in outputs (WUI homes and area) were close contact with nature. From 1940 to

ABSTRACT much smaller than parameter variations suggesting the WUI definition generates stable results on most 2000, significant housing growth occurred landscapes. Overall, modifying the WUI definition resulted in a similar amount of WUI area and number in suburban and rural areas, especially in and of homes and affected the precise location of the WUI. near forests (Radeloff et al. 2005a). Housing growth was strong nationwide, including ar- Keywords: wildland–urban interface; wildland fire; housing growth; GIS sensitivity analysis; eas with short fire return intervals and high wildland fire policy departure from historic conditions such as the Rocky Mountains and the Sierra Nevada (Hammer et al. in press). The effects of re- he wildland–urban interface it clarified the role of the federal agencies in source management practices, climate (WUI) has become the central fo- fighting fires in the WUI (USDI and USDA changes, and insect and disease infestations, cus of wildland fire policy in the 1995). A 10-year overhaul of US wildland together with continued housing growth in T high fire-risk areas thus create an urgent United States. The tragic 1994 Storm King fire policy followed, spurred on by the ex- incident, in which 14 firefighters were killed, treme fire season of 2000. Each of the re- need to understand and manage fire risk in initiated intense scrutiny of wildfire policy ports and initiatives issued in successive the WUI (Pyne 2001). and management (US Department of the years—the Report to the President in Sep- Governments at all levels share respon- Interior [USDI] and US Department of Ag- tember of 2000, the 10-year Comprehensive sibility for wildland fire management in the riculture [USDA] 2006). When the Federal Strategy of 2001, its implementation plan WUI. The federal government’s role is to Wildland Fire Management Policy and Pro- and the Healthy Forest Initiative in 2002, provide leadership, coordination, and re- gram Review was issued the following year, and the Healthy Forest Restoration Act search across the country, a role that benefits

Received June 20, 2006; accepted April 17, 2007. Susan I. Stewart ([email protected]) is research social scientist, Northern Research Station, USDA Forest Service, 1033 University Avenue, Suite 360, Evanston, IL 60201. Volker C. Radeloff ([email protected]) is associate professor, Department of Forest Ecology and Management, University of Wisconsin, Madison, Wisconsin. Roger B. Hammer ([email protected]) is assistant professor, Department of Sociology, Oregon State University, Corvallis, Oregon. Todd J. Hawbaker ([email protected]) is Ph.D. student, Department of Forest Ecology and Management, University of Wisconsin, Madison, Wisconsin. The authors are grateful for the comments of Pam Jakes and Don Field on an earlier version of this article. Support for this research was provided by the USDA Forest Service Northern Research Station under the National Fire Plan, and by the University of Wisconsin–Madison. Copyright © 2007 by the Society of American Foresters.

Journal of Forestry • June 2007 201 from the ability to locate and compare the the Western Governor’s Association (Teie ulation data and determined vegetation WUI in different states and regions. To sup- and Weatherford 2000), with only minor proximity via detailed fuels mapping, al- port this aspect of WUI management and changes. The more recent HFRA moves though the Census data’s resolution, the strategic planning, we created a map of the away from this standardized approach to de- rules for determining vegetation and popu- WUI across the lower 48 states (Radeloff et fining the WUI by allowing communities to lation proximity, and the types of vegetation al. 2005b). Unlike a community-level WUI establish a buffer zone around the town, the included were not specified, and the distinc- map that can be enriched by using detailed civic infrastructure, and evacuation routes, tion between interface and intermix WUI local data, this “big picture” national map including these areas in the WUI as well. was made based on population density required nationally consistent data and a sin- This more flexible definition is consistent alone. A coarse-scale map of the wildland gle standardized WUI definition. Here, we with the HFRA’s emphasis on empowering fire risk to structures was developed as a tool sought to understand how the national WUI local communities to develop Community for strategic planning (Schmidt et al. 2002). map is influenced by the WUI definition Wildfire Protection Plans. Ambient population data (USDE 2005) was used, the relative effect of each part of the Throughout its evolution, the WUI used to derive housing densities and com- definition, and the overall usefulness of the definition always includes three compo- bined with extreme fire potential (derived map in identifying homes likely to be af- nents: human presence, wildland vegeta- from climate data) and potential fire expo- fected by wildfire. We used geographic in- tion, and a distance that represents the po- sure data (derived from vegetation data) to formation system (GIS) analysis to address tential for effects (e.g., wildland fire and categorize threat levels at a 1-km resolution. these questions. human activity) to extend beyond bound- These examples illustrate that despite aries and impact neighboring lands. Beyond all that has been written about the WUI and Literature Review these three components, most WUI discus- its significance in fire policy, there is no sin- A good WUI map provides a graphic sions are imprecise regarding what is or is gle operational definition. A review by the representation that matches the conceptual not included. For example, human presence Government Accountability Office (GAO) understanding of what and where the WUI has been defined by housing density, popu- of NFP implementation criticized the fed- is. This conceptual understanding of the lation density, number of houses, or eral agencies responsible for wildland fire WUI has evolved over time; the area where configuration of housing developments and management on this point (Hill 2001). In a houses and forests meet has attracted atten- neighborhood characteristics. Wildland formal response included in the GAO re- tion for many years. Bradley’s book (Bradley vegetation is always mentioned, but the den- port, Forest Service Chief Bosworth reaf- 1984) on the resource management issues in sity, extent, and type of vegetation that firmed his agency’s commitment to main- the interface is a major early contribution. makes some vegetation “wildland vegeta- taining enough flexibility in defining the Even earlier, Henry Vaux characterized the tion” is not well defined. The distance that WUI to accommodate the many different WUI as the “hotseat of forestry” (Vaux the WUI extends into wildlands or into a kinds of landscapes it manages. 1982) and cautioned foresters not to under- housing development has been described in Flexibility is valuable for both forest- estimate its political and policy significance. many different ways, including the distance and community-level mapping and manage- Both authors discussed a wide range of WUI a golf ball will fly off the porch or the dis- ment but is counterproductive when issues and neither equated the WUI with tance from which flames or firebrands can comparing WUI across places or over time wildland fire. However, the focus of WUI reach a structure (Summerfelt 2003), but periods. Our research was intended to assess discussions had narrowed by the end of the specific distances are rarely given. the WUI across regions and to track its 1980s, when Davis (1990) characterized the Conceptually, these many definitions growth and change over time, two purposes WUI as a setting where wildland fire is a all refer to the same basic idea, that the WUI that required standardizing the WUI defini- problem and where conflicts arise over re- is where houses and wildlands meet or over- tion. With no previous standard definition sponsibility for protecting homes from wild- lap. However, operationally, they differ and of the WUI to draw from, we tested the def- fire. A 1997 issue of the Journal of Forestry in previous WUI maps, definitions vary de- inition we developed to ensure its consis- featured a cover photo of a wildland fire en- pending on data available at the time and tency with our research aims and with stra- croaching on a subdivision and included two across the extent of the map. For example, tegic wildland fire planning. articles about the WUI that emphasized fire Greenburg and Bradley (1997) used remote The 2001 Federal Register (USDA and planning and management (Greenberg and sensed data to assess vegetative characteris- USDI 2001)WUI definition describes the Bradley 1997, Plevel 1997). The term “wild- tics related to human presence in two cities. characteristics of houses and vegetation and land–urban interface” is now used almost In their Eugene, Oregon, WUI map, popu- the relationship between them that must oc- exclusively in the context of wildland fire. lation and road densities captured human cur for an area to be classified as WUI. In Extensive references to the WUI in US presence; for the Seattle map, distance from this regard, the definition is a model or ab- fire policy reflect the fire management com- the city center was used as a proxy for human stract representation of reality. Sensitivity munity’s long familiarity with the concept. presence. Lein introduced fuzzy methods of analysis is a set of methods used in GIS anal- Published in the Federal Register in conjunc- determining the WUI’s location, relying on ysis to assess various characteristics of a tion with the National Fire Plan (NFP), the remote sensed data to detect the presence of model (Crosetto et al. 2000). The housing WUI definition states that “The WUI com- both structures and vegetation (Lein 2006). and vegetation characteristics and the buffer munity exists where humans and their devel- Two earlier maps attempted to represent risk distances are the parameters of the WUI def- opment meet or intermix with wildland fu- from fire as well as vegetation and housing inition, and the definition can be assessed by els” (USDA and USDI 2001, p. 752–753). characteristics. Space Imaging’s (2002) varying the parameters one at a time to iso- This definition was adapted from a report to Floridawide fire risk tool used Census pop- late the effect of each on the model output,

202 Journal of Forestry • June 2007 which is the WUI map (Hamby 1994). Sen- sitivity analysis also indicates how sensitive or stable a model is overall (Store and Kan- gas 2001). To fulfill its role as a policy tool, our WUI map needs to be relatively stable as individual parameters are changed. Stability would indicate that the specific levels we chose for housing density, vegetation characteristics, and so on, do not outweigh the importance of the components in combina- tion. Conceptually, the WUI is a conjunc- tion of housing and vegetation characteristics; all are important, so no single parameter should dramatically change the WUI map. However, housing growth is the most vola- tile factor influencing WUI growth. Hence, sensitivity to the housing density parameter is essential to the model’s ability to reflect the real world where the WUI is sensitive to housing growth (Rykiel 1996). This charac- Figure 1. The WUI definition. teristic of the model will ensure its suitability for sensing WUI change over time. Although explicit consideration of risk ence in this context because of the impor- Geological Service, a classification of Land- from fire is beyond the scope of this project, tance of structure protection in wildland sat Thematic Mapper imagery, at 30-m res- the policy intent of the WUI definition and firefighting. Because housing density has olution, to determine vegetative cover type map is to identify those homes most likely to grown faster than population density in re- and extent (Vogelmann et al. 2001). We de- be affected by wildland fire. For the states of cent decades, it is the better measure for eval- fined “wildland vegetation” as all types of Arizona, California, Colorado, Idaho, Mon- uating WUI change and projecting future vegetative cover except those that are clearly tana, New Mexico, Nevada, Oregon, Utah, WUI location and extent (Liu et al. 2003). not wild, such as urban grass, orchards, and Washington, and Wyoming, where fire pe- Furthermore, nonresidential structures are agricultural vegetation. After reviewing the rimeter data are relatively good, we assess the likely located close to houses; e.g., barns and WUI mapping efforts of the California Fire extent to which houses close to recent fires garages, bars, and restaurants tend to be clus- Alliance (2001) and analyzing vegetation were in areas classified as intermix or inter- tered near homes and are absent from areas density in several test areas, we chose a veg- face WUI and compare them with houses without homes. We used 2000 Census etation density threshold of 50% for inter- statewide. housing data to calculate housing density by mix. Thus, a census block was retained as census blocks. The size of census blocks var- potential intermix if at least 50% of its area Methods ies widely with population density but they consists of forest, shrubland, native grass- The policy definition on which we base are the smallest geographic unit for which land, transitional, or wetland vegetative our map was published in the Federal Regis- US Census data are reported (mean ϭ 25 ac; cover. Census blocks where less than 50% of ter and notes that “Generally, Federal agen- SD ϭ 2,674 ac), and are always subdivisions pixels were classified as wildland vegetation cies will focus on [interface and intermix] of counties. Housing density calculations in- were not included as potential intermix communities . . . Interface communities ex- clude all housing units as defined by the WUI, although portions of these blocks ist where structures directly abut wildland Census Bureau, such as apartments and could be classified as interface WUI when fuels....Intermix communities exist where houses, vacant and occupied, including sea- they were close enough to wildland vegeta- structures are scattered throughout a wild- sonal homes (US Census Bureau 2002). The tion to fall within the interface buffer (see land area” (USDA and USDI 2001, p. 753). minimum housing density for WUI speci- the following section). In the extensive text that follows, just fied in the Federal Register (USDA and Interface Vegetation and Buffer Dis- one of the three major components of the USDI 2001) definition is 1 housing unit per tance. The interface is the area where hous- WUI, human presence, is defined in detail 40 ac. ing is in close proximity to wildland vegeta- sufficient for analysis. We calibrated param- Wildland Vegetation Characteris- tion. Locating the interface required first eters for the other two components (wild- tics. Using land cover data to identify wild- identifying areas with wildland vegetation land vegetation characteristics and buffer land vegetation requires specific direction and then including areas within “close prox- size) to develop an operational definition. for including or excluding each type of veg- imity,” represented by a buffer some dis- The calibration of each parameter of the def- etative cover and for determining how much tance from the vegetation. Interface census inition is discussed next, and then sensitivity vegetation must be present for an area to be blocks must meet the same housing density analysis procedures are explained. considered “vegetated.” We assessed vegeta- minimum (at least one structure per 40 ac) Human Presence. Housing density is tive cover using the 1992/1993 National to be included in the WUI, but do not have the most appropriate metric for human pres- Land Cover Data (NLCD) from the US to have more than 50% wildland vegetation.

Journal of Forestry • June 2007 203 Figure 2. Distribution of WUI area and housing units by US county in 2000.

We chose a 1.5-mi buffer distance based on prevent areas surrounding small urban parks (housing density, vegetation type and den- the precedent of the California Fire Alliance from being classified as interface WUI. sity, and buffer distance) on WUI housing (2001). This distance represents an estimate WUI Definition. In summary, the units and area, with particular attention of how far, on average, a firebrand can fly WUI definition as we operationalized it is given to the influence of the housing density ahead of a fire front, a rationale clearly re- the area where houses exist at more than 1 parameter. In addition to the individual pa- lated to the fire policy application for which housing unit per 40 ac and (1) wildland veg- rameter changes, minimum and maximum the WUI map was developed. Although etation covers more than 50% of the land WUI scenarios were tested in which all pa- winds and topography affect the distance a area (intermix WUI) or (2) wildland vegeta- rameters were changed simultaneously to as- firebrand will actually travel, the information covers less than 50% of the land area, sess the extent of overlap across individual tion needed to capture these myriad fine- but a large area (over 1,235 ac) covered with parameter responses. scale variations in fire brand lofting is not more than 75% wildland vegetation is The size of the national WUI data set available. A comprehensive search of pub- within 1.5 mi (interface WUI). This defini- (8.9 million census blocks) precluded using lished WUI mapping work uncovered no al- tion is diagrammed in Figure 1, and the pa- the whole data set for sensitivity analyses. ternative rationale for the buffer distance rameters tested in the sensitivity analysis are Instead, the following states were chosen and no evidence against the validity of the shown in bold. representing different geographic regions of 1.5-mi distance we chose. The vegetated GIS Sensitivity Analysis. To test the the country, with different ecotypes and block from which the buffer extends was re- stability of our operationalization of the housing patterns: California, Colorado, quired to meet a higher vegetation density WUI definition, we specified that a stable Florida, Michigan, North Carolina, New criterion of 75% wildland vegetation. Cen- definition is one where the percent change in Hampshire, and Washington. For each pa- sus blocks that were only partially within the output (housing units and area classified as rameter modification, data were reprocessed 1.5-mi buffer distance were split and only WUI) is always smaller than the percent for all census blocks in each sample state (a the portions within 1.5 mi of the vegetated change made to any single parameter. One- total of 1.7 million census blocks, 20% of block were defined as interface. The inter- at-a-time variations in the definition’s pa- the national data set), and results were re- face buffer was extended only from vege- rameters were made to determine the influ- ported for WUI area and number of houses. tated areas of at least 1,235 ac (5 km2)to ence of each parameter individually Because one aim of the WUI mapping

204 Journal of Forestry • June 2007 Table 1. Percent change in WUI area and WUI housing units in response to parameter changes, by state.

California Colorado Florida Michigan North Carolina New Hampshire Washington Average

Original WUI Area (1,000,000 ac) 7.23 1.98 6.97 5.84 13.66 2.38 3.75 — Housing (100,000 HUs) 50.9 8.4 25.9 9.7 23.2 4.5 12.0 — Percent change in response to parameter changes (percent change in area, percent change in housing units) Housing density (original, Ͼ1 HU/40 ac) Ͼ1 HU/ 80 ac 40.6 71.7 39.8 62.3 34.0 40.0 36.9 46.5 1.0 3.0 1.9 6.7 3.7 3.9 2.0 3.2 Ͼ1 HU/ 20 ac Ϫ29.2 Ϫ39.6 Ϫ33.7 Ϫ45.6 Ϫ39.4 Ϫ39.0 Ϫ29.8 Ϫ36.5 Ϫ1.5 Ϫ3.3 Ϫ3.2 Ϫ9.7 Ϫ8.3 Ϫ7.4 Ϫ3.3 Ϫ5.2 Intermix vegetation density (origina,l Ͼ50% of pixel) Ͼ25% of pixel 4.9 5.4 11.1 47.4 13.7 0.6 6.4 12.8 14.0 5.8 17.8 57.1 20.3 2.6 22.5 20.0 Ͼ5% of pixel Ϫ2.3 Ϫ2.3 Ϫ8.2 Ϫ18.5 Ϫ14.1 Ϫ0.8 Ϫ3.8 Ϫ7.1 Ϫ3.7 Ϫ1.9 Ϫ6.2 Ϫ20.3 Ϫ11.5 Ϫ1.4 Ϫ8.2 Ϫ7.6 Interface buffer size (original, 1.5 mi) 3.0 mi 12.6 10.0 13.7 12.1 7.4 1.2 9.4 9.5 46.7 33.3 49.0 19.9 21.7 15.7 29.3 30.8 0.75 mi Ϫ9.9 Ϫ8.5 Ϫ9.6 Ϫ6.8 Ϫ5.3 Ϫ1.8 Ϫ7.5 Ϫ7.1 Ϫ3.7 Ϫ24.1 Ϫ27.9 Ϫ13.5 Ϫ14.7 Ϫ13.5 Ϫ20.2 Ϫ20.7 Vegetation (original all wildland vegetation) Upland only Ϫ0.6 0.2 Ϫ12.7 Ϫ30.8 Ϫ17.9 Ϫ1.8 Ϫ0.6 Ϫ13.9 Ϫ0.8 0.2 Ϫ71.5 Ϫ3.6 Ϫ18.8 Ϫ9.9 Ϫ1.9 Ϫ20.0 Forest only Ϫ65.9 Ϫ57.6 Ϫ20.7 Ϫ37.9 Ϫ18.8 Ϫ1.9 Ϫ19.5 Ϫ39.3 Ϫ87.8 Ϫ76.2 Ϫ86.8 Ϫ4.3 Ϫ20.3 Ϫ11.0 Ϫ31.0 Ϫ51.0 WUI scenariosa (simultaneous changes) Maximum 58.9 91.2 64.3 134.2 50.4 41.3 52.8 70 Extent 55.5 39.3 61.4 79.0 20.9 19.8 45.4 46 Minimum Ϫ84.3 Ϫ80.7 Ϫ94.4 Ϫ83.1 Ϫ50.9 Ϫ48.0 Ϫ52.9 Ϫ71 Extent Ϫ92.5 Ϫ83.4 Ϫ95.5 Ϫ75.7 Ϫ22.9 Ϫ37.9 Ϫ55.3 Ϫ66 aMaximum extent: housing density Ͼ 1 HU/80 ac, intermix vegetation density Ͼ25%, interface buffer 3 mi, vegetation all wildland; minimum extent: housing density Ͼ 1 HU/20 ac, intermix vegetation density Ͼ 75%, interface buffer 0.75 mi, vegetation forest only. Note: Figures in italics give percent change in housing units. project was to support fire policy, we made a The WUI and Recent Western Fires. Southwest. In western states, relatively small series of eight individual parameter changes Fire perimeters for 2006 available in the amounts of the land area, but a high percent- that represent plausible policy alternatives. GeoMac database (USDI and USDA 2007) age (well over 50% in many states) of the Plausible alternatives are those that do not representing 169 fires that covered 1.27 mil- housing units fell into the WUI. Intermix violate the underlying WUI concept in wild- lion ac were used to assess the amount of WUI covered more land area (81% of all land fire management. For example, if we set WUI near recent western fires. The number WUI area) than interface WUI; but because the housing density threshold too low (e.g., and classification of housing units within housing density is typically higher in inter- 1 house per 150 ac), the resulting map 1-mi-wide buffers at distances of 1–10 mi face areas, intermix WUI contained just over would no longer match what most land from the fire perimeters were analyzed. one-half (53%) of all WUI homes nation- managers envision as “WUI” because vast Within 10 mi from the fire, potential effects ally. areas would have no structures. The same will vary; closer homes would be more likely Varying Individual Parameters. logic guided all our test parameter choices; to experience evacuation orders, and homes Housing density threshold changes (100% the alternative parameters varied from start- that are more distant may be affected only by increase and 50% decrease) affected WUI ing values as far as was plausible, without smoke; but homes within this distance sel- area more than any other single parameter losing sight of the WUI concept and the dom would be considered too distant for change across all test states (Table 1). In map’s policy purpose. concern. both California and Florida, lowering the Two test values were chosen for each housing density threshold increased the parameter, typically a maximum and mini- Results WUI extent by over 11 million ac, and rais- mum. The housing density threshold and WUI in the Lower 48 States. Using ing it reduced the size of the WUI by 8.5 and interface buffer distance were both doubled the initial values for housing density, vege- 9.5 million ac, respectively. Effects on WUI and halved; intermix vegetation density was tation type, vegetation density, and interface houses are small, averaging a 3.2% increase increased and decreased by 50%. Vegetative buffer size, the NLCD and Census housing with a lower threshold and a 5.2% decrease cover types were eliminated to create two data were processed in a GIS and maps were with a higher threshold. more limited conceptions of “wildland,” one prepared (Figure 2). WUI occurred in all the Intermix vegetation density changes with only uplands (eliminating all wetland lower 48 states, with concentrations along (50% increase and 50% decrease) had little vegetation classes) and the other, just forests, the Eastern Seaboard, in amenity-rich re- impact in most states, averaging a 12.8% in- excluding wetlands, grasslands, and shrub- gions of the northern Great Lakes and the crease with a lower vegetation requirement lands. Responses were expressed as percent Missouri Ozarks, and around the metropol- and a 7.1% decrease in area when a higher change in WUI homes and area. itan areas of the Rocky Mountains and the density of vegetation is required. A lower

Journal of Forestry • June 2007 205 vegetation density threshold does, however, increase the number of WUI homes on average by 20%, with a wide variation in settings evidenced in the range of responses, from a 57.1% increase in Michigan to a 2.6% increase in New Hampshire. Taken together, the two modifications in intermix vegetation density tested affected the classification of 750,000 Michigan homes. Interface buffer distance changes (100% increase and 50% decrease) had small effects in terms of area and mixed effects on housing. The impact of using a larger buffer was particularly high in Califor- nia (46.7%) and Florida (49%) because ex- Figure 3. WUI classification of homes near 2006 western wildland fires. tensive areas of high density housing occurred within 3 mi of wildlands. However, in percentage terms across the test states, the never exceeded the magnitude of the param- the present time. Tensions over where best changes in interface buffer distance gener- eter change. The WUI scenarios with their to concentrate decisionmaking authority ated only a 26% change in the number of suite of simultaneous changes confirm that and funding responsibility are natural out- WUI homes and an 8% change in WUI many responses overlap, so that changing all comes of shared responsibility. In this con- area. the parameters simultaneously generates a text, defining and classifying the WUI takes Vegetation type definition changes smaller response than the sum of responses on greater significance. Our purpose here yielded mixed results because the distribu- to individual parameter changes. was to provide those who use the WUI data tion of nonforest vegetation varies across The WUI and Recent Western Fires. with more information about how a differ- states, and the changes to the vegetation type GIS analysis of the WUI classification of ent WUI definition would have affected the definition that we tested focused on nonfor- homes with a 1- to 10-mi distance from national WUI map and statistics. est vegetation. In areas where the wildland 2006 western fires revealed that nearly all Although a single national definition vegetation is mostly shrublands or grassland, housing units within 10 mi of the fires were such as the one we mapped may be unwork- such as California, the “forest only” defini- classified as WUI (73%), but the proportion able for day-to-day management of individ- tion eliminated large areas from the WUI. declined with increasing distance from the ual national forests in the United States, it For those where wetlands are extensive such fire (Figure 3). Within 1 mi of the fire pe- serves a valuable purpose for strategic plan- as Florida, the “uplands only” definition had rimeter, 92% of homes were classified as ning by providing consistency and credibil- similar effects. New Hampshire was least af- WUI. Homes 9–10 mi from the perimeter ity to estimates of the scope of the WUI na- fected by the changes because its land cover still include a majority of WUI homes tionwide. Trends suggest the WUI will is largely forest. (65%), which far exceeded the overall pro- continue to grow, which makes a national Overall Map Stability. When changes portion of WUI homes in these western perspective on the scope of the WUI fire were made in individual parameters, the states. The limited number of fire perimeters problems even more critical. Furthermore, magnitude of responses to each change as available made this analysis less comprehen- the USDA’s Office of the Inspector General measured by percent change in the number sive than our wall-to-wall WUI map, but issued a report in November of 2006 of homes or WUI area was less than the mag- results suggested that the WUI definition (USDA 2006) suggesting WUI growth is nitude of the parameter change (Table 1), mapped here tends to capture homes close to pushing firefighting costs higher, adding ur- satisfying our criterion for stability. The the areas where wildland fire occurs, sup- gency to our need to understand the WUI. highest average response was a 51% drop in porting the validity of the national map. Beyond its strategic purposes, the na- WUI housing units associated with use of tional WUI map also provides information the “forest only” vegetation parameter Implications and Conclusions for every state and community regardless of (dropping wetlands, grasslands, and shrub- Results of the sensitivity analysis sug- the other resources available to local plan- lands from the vegetative classes defined as gested that the operational definition devel- ners. Certainly, communities can go beyond wildlands) and a 46.5% increase in WUI oped for this mapping effort is stable. the WUI data in detail (e.g., locating indi- area in response to a 100% increase in the Changes made across a plausible range of vidual structures) and specificity to wildland housing density threshold. The states we alternatives for the WUI definition’s param- fire issues (e.g., distinguishing high- from sampled differed in their sensitivity to eters showed that at the national scale for low-hazard WUI areas based on locally changes. Michigan, Florida, and Colorado which the map was developed, modifying available fuels data), but for communities were more sensitive to changes in the WUI the settings to use other conceptually similar without the resources and technical staff to definition. In these three states, there are values did not change the overall pattern or create a customized local WUI map, this many housing units, a diverse array of vege- prevalence of WUI areas. classification is a suitable guide to the loca- tative characteristics, and large land areas. The involvement of governments at lo- tion and extent of WUI. Even with the greater sensitivity in these cal, state, and federal levels testifies to the Among the parameters we tested, the states to some individual changes, responses significance of wildland fire in US society at change that generated the largest response

206 Journal of Forestry • June 2007 was housing density, and housing density Literature Cited SPACE IMAGING. 2002. Florida fire risk assessment was the only parameter defined directly in BRADLEY, G.A. 1984. Land use and forest resources: final project report. Available online at www. the Federal Register. The rest of the WUI The urban/forest interface. University of Wash- flame.fl-dof.com/risk/final.pdf; last accessed definition was less specific. Testing these ington Press, Seattle, WA. 222 p. June 2006. STORE, R., AND J. KANGAS. 2001. Integrating spa- definition changes showed how the defini- CALIFORNIA FIRE ALLIANCE. 2001. Characterizing the fire threat to wildland-urban interface. Cal- tial multi-criteria evaluation and expert knowl- tion interacts with local conditions. We ifornia Fire Alliance, Sacramento, CA. 75 p. edge for GIS-based habitat suitability model- tested just those changes that seemed most CROSETTO, M., S. TARANTOLA, AND A. SALTELLI. ling. Landsc. Urban Plan. 55:79–93. plausible in light of the focus on forest policy 2000. Sensitivity and uncertainty analysis in SUMMERFELT, P. 2003. The wildland urban inter- and the availability of data, but even so, spatial modelling based on GIS. Ag. Ecosys. En- face: What’s really at risk? Fire Manage. Today most changes we tested would make the viron. 81:71–79. 63(1):4–7. DAVIS, J. 1990. The wildland-urban interface: WUI definition less representative of the un- TEIE, W.C., AND B.F. WEATHERFORD. 2000. Fire Paradise or battleground? J. For. 88:26–31. in the west: The wildland/urban interface fire derlying WUI concept. GREENBERG, J.D., AND G.A. BRADLEY. 1997. An- problem. Rep. to the Council of Western State The sensitivity of the WUI definition to alyzing the urban-wildland interface with GIS: Foresters, Deer Valley Press, Rescue, CA. 15 p. housing density was reassuring because it in- Two case studies. J. For. 95(10):18–22. US CENSUS BUREAU. 2002a. Census Bureau 2000 dicated that as housing growth occurs, the HAMBY, D.M. 1994. A review of techniques for technical documentation summary file 3A. US WUI will grow accordingly. Analysis of parameter sensitivity analysis of environmental Census Bureau, Washington, DC. 1,269 p. models. Environ. Monitor. Assess. 32:135–154. change over time interests those concerned US CENSUS BUREAU. 2002b. 2000 Census of pop- HAMMER, R.B., V.C. RADELOFF, J.S. FRIED, AND ulation and housing, summary file 1 [machine- with resource impacts, particularly because S.I. STEWART. Wildland-Urban Interface readable data files]/prepared by the US Census there is reason to expect that the full effect of growth during the 1990s in California, Ore- Bureau, Washington, DC. housing development on wildland resources gon and Washington. Int. J. Wildl. Fire (in US DEPARTMENT OF AGRICULTURE (USDA). press). will take time to become apparent. The abil- 2006. Forest Service large fire suppression costs. ity of the WUI definition to represent effects HILL, B.A. 2001. The national fire plan: Federal agencies are not organized to effectively and effi- Audit Rep. 08601-44-SF, USDA Office of In- of change over a long time period enhances ciently implement the plan. Testimony before spector General, Washington, DC. 48 p. its usefulness for resource management. the Subcommittee on Forests and Forest US DEPARTMENT OF AGRICULTURE (USDA) AND When the WUI data are combined with Health, Committee on Resources, House of US DEPARTMENT OF THE INTERIOR (USDI). data regarding fire risk, the exclusion of low- Representatives, US General Accounting Of- 1995. Federal wildland fire management policy risk WUI areas will alter the distribution of fice, GAO-01-1022T. 14 p. and program review. USDI and USDA, Wash- ington, DC. 40 p. WUI across the country. But the temporal LEIN, J.K. 2006. Toward a rapid characterization of the built environment within the wildland- US DEPARTMENT OF ENERGY (USDE). 2005. and spatial variability in fire hazard and the urban interface: A soft classification strategy. Landscan. Oak Ridge National Laboratory, many diverse factors (e.g., weather, fuels, GIScience Rem. Sens. 43(1):44–61. Oak Ridge, TN. Available online at www.ornl. and topography) that determine it will al- LIU, J., G.C. DAILY, P.R. EHRLICH, AND G.W. gov/sci/landscan/index.html. Last accessed ways limit our ability to make fire risk pre- LUCK. 2003. Effects of household dynamics on May 2007. dictions on the same scale and with the same resource consumption and biodiversity. Na- US DEPARTMENT OF THE INTERIOR (USDI) AND ture 421:530–533. confidence as we can map the WUI. Wild- US DEPARTMENT OF AGRICULTURE (USDA). PLEVEL, S.R. 1997. Fire policy at the wildland- 2001. Urban wildland interface communi- land vegetation and housing units change urban interface: A local responsibility. J. For. ties within vicinity of federal lands that are at only slowly over time relative to the com- 95(10):12–17. high risk from wildfire. Fed. Regist. 66(3): plex, dynamic factors determining fire risk. PYNE, S.J. 2001. The fires this time, and next. 751–777. Last but not least, although the current Science 294(2):1005–1006. US DEPARTMENT OF THE INTERIOR (USDI) AND RADELOFF, V.C., R.B. HAMMER, AND S.I. STEW- perception of the WUI centers on wildland US DEPARTMENT OF AGRICULTURE (USDA). ART. 2005a. Rural and suburban sprawl in the fire, the concept is applicable much more 2006. Protecting people and natural resources: A U.S. Midwest from 1940 to 2000 and its rela- cohesive fuels treatment strategy. USDI and broadly. The WUI is an area where the in- tion to forest fragmentation. Conserv. Biol. USDA, Washington, DC. 26 p. fluence of human development is mani- 19(3):793–805. US DEPARTMENT OF THE INTERIOR (USDI) AND fested in many different ways, among them, RADELOFF, V.C., R.B. HAMMER, S.I. STEWART, US DEPARTMENT OF AGRICULTURE (USDA) J.S. FRIED, S.S. HOLCOMB, AND J.F. MCKEE- through impacts on birds, mammals, plants 2007. Geomac wildland fire support. Available and other ecosystem structures, functions, FRY. 2005b. The wildland urban interface in the United States. Ecol. Appl. 15(3):799–805. online at http://geomac.usgs.gov/. Last ac- and services. The WUI also is an area where RYKIEL, E.J. JR. 1996. Testing ecological models: cessed May 2007. people come into contact with wildland and the meaning of validation. Ecol. Mod. 90:229– VAUX, H.J. 1982. Forestry’s hotseat: The urban/ its management, and the full range of public 244. forest interface. Am. For. 88(5):37, 44–46. responses to management play out. Thus, SCHMIDT, K.M., J.P. MENAKIS, C.C. HARDY, VOGELMANN, J.E., S.M. HOWARD,L.YANG, C.R. tracking the growth and change of the WUI W.J. HANN, AND D.L. BUNNELL. 2002. Devel- LARSON, B.K. WYLIE, AND N. VAN DRIEL. 2001. Completion of the 1990s national land is essential to dealing with the consequences opment of coarse-scale spatial data for wildland fire and fuel management. USDA For. Serv. cover data set for the conterminous United of housing growth and, ultimately, to sus- Gen. Tech. Rep. RMRS-GTR-87, Rocky States from Landsat thematic mapper data and taining the wildlands in the face of mount- Mountain Research Station, Fort Collins, CO. ancillary data sources. Photogramm. Eng. Rem. ing development pressures. 41 p. Sens. 67:650–662.

Journal of Forestry • June 2007 207