Effect of Thinning and Prescribed Burning on Crown Fire Severity in Ponderosa Pine Forests

Jolie Pollet¹ and Philip N. Omi²

¹Lakeview District Bureau of Land Management and Fremont National Forest, HC 10 Box 337, Lakeview, Oregon 97630 U.S.A.

Tel. 541/947-6175, FAX 541/947-6399 email jolie_pollet@or.blm.gov

²Department of Forest Sciences, Colorado State University, Fort Collins, Colorado 80523 U.S.A.

Tel. 970/491-5819, FAX 970/491-6754 email phil@cnr.colostate.edu

Abstract

Fire exclusion policies have affected stand structure and wildfire hazard in ponderosa pine forests. Wildfires are becoming more severe in stands where trees are densely stocked with shade-tolerant understory tree encroachment. Although forest managers have been employing fuel treatment techniques to reduce wildfire hazard for decades, little scientific evidence supports such fuel treatments. This research quantitatively examined fire effects in treated and untreated stands in western United States National Forests. Four ponderosa pine sites in Montana, Washington, California and Arizona were selected for study. Fuel treatments studied include: prescribed fire only, whole-tree thinning, and thinning followed by prescribed fire. On-the-ground fire effects were measured in adjacent treated and untreated forests. We developed post facto fire severity and stand structure measurement techniques to complete field data collection. We found that crown fire severity was mitigated in stands that had some type of fuel treatment compared to stands without any treatment. At all four of the sites, the fire severity and crown scorch was significantly lower at the treated sites (α=0.03). Results from this research indicate that fuel treatments removing small diameter trees may be beneficial for reducing crown fire hazard in ponderosa pine sites.

Keywords

Pinus ponderosa, Montana, Washington, California, Arizona, fuel treatment

Introduction

Ponderosa forests are the most widely distributed forest type in the western United States covering millions of hectares (Van Hooser and Keegan (1988). Fires in ponderosa pine (Pinus ponderosa) are becoming a growing concern (Arno and Brown 1991, Covington and Moore 1994). Colorado=s Buffalo Creek Fire in 1996 is one example of the recent and very large wildfires in ponderosa pine that have focused attention on controlling fire costs and damages (Gale 1977, Gonzalez-Caban 1995). Over 10,000 acres burned in about 6 hours, mostly as a crown fire (Orozco 1998). Millions of dollars were spent controlling the Buffalo Creek Fire, in replacing burned structures, for post-fire rehabilitation efforts and for consequent flood damage.

Fires in ponderosa pine forests often differ dramatically from those observed by early settlers. Many of today=s fires are stand-destroying crown fires as opposed to much lower intensity surface fires (Arno and Brown 1991, Agee 1993, Covington and Moore 1994, Mutch 1994). In addition to changes in fire behavior, stand structure in ponderosa pine forests also has been altered in the last century. Historical accounts describe large, park-like and open stands (Weaver 1943, Mutch et al. 1993, Covington and Moore 1994) that can be compared to the densely packed areas currently undergoingstand conversion as shade-tolerant trees out-compete ponderosa pine regeneration. These changes may be attributed to effective fire exclusion efforts over the past 100 years.

Forest managers have long contended that stand structural changes can be linked to more extreme wildfire behavior (Weaver 1943, Biswell 1960, Cooper 1960, Dodge 1972, VanWagner 1977, Rothermel 1991, McLean 1993, Fiedler et al. 1995, Williams 1998). For example, shade-tolerant species and dense regeneration may serve as ladder fuels to move fire into the tree crowns (Weaver 1943, Dickman 1978, Laudenslayer et al. 1989, MacCleery 1995). (Ladder fuels provide vertical continuity between the surface fuels and crown fuels, increasing the likelihood of of torching and crowning.) Fuel treatments such as prescribed fire and mechanical thinning are offered as ways to reduce or retard wildfire spread and intensity in ponderosa pine forests (Weaver 1961, Biswell et al. 1968, Babbitt 1995).

Many scientists and land managers assume that fuel treatments reduce wildfire hazard, but few studies have analyzed on-the-ground fire effects in treated versus untreated stands. Much of the evidence supporting the effectiveness of fuel treatments in mitigating wildfire damages has been inferred from informal observation, nonsystematic inquiry or computer modeling (Omi and Kalabokidis 1991, Edminster and Olsen 1995, Fiddler et al. 1995, Fiedler 1996, Kalabokidis and Omi 1998, Scott 1998a, 1998b, Stephens 1998). Only two studies have examined field wildfire effects in stands with fuel manipulations. First, Vihanek and Ottmar (1993) measured more severe post-wildfire effects in areas where slash was left compared to less severe effects in slash-treated areas. Another study attempted to quantify fire damage to ponderosa pine tree crowns by examining post-fire aerial photos and available databases (Weatherspoon and Skinner 1995). Weatherspoon and Skinner found that sites with harvest treatments that included complete slash removal had lower fire severity, but they did not complete field verification of the results. In contrast to these previously mentioned studies, our study systematically and quantitatively examines field observations following wildfire in treated versus untreated ponderosa pine stands.

Our hypothesis is that fuel treatments reduce fire severity and crown scorch. Fire severity, for the purpose of this study, refers to fire=s effect on the ecosystem and is directly related to post-fire vegetation survival (Ryan and Noste 1985). Study objectives are to compare crown scorch and crown consumption in untreated versus treated stands; and develop a methodology for making post facto comparisons of fire severity in untreated versus treated ponderosa pine stands.

Methods

Methods for study site selection and field data collection are described below. Both site selection and data collection were tailored to assure study integrity, i.e., eliminate intentional or unintentional bias.

Site Selection

During the initial stages of study development, wildfires occurring less than fifteen years prior that had fuel treatment activities within the wildfire perimeter were all potential candidates. Selection was eventually narrowed to those sites with mechanical fuel treatments, sites that could be sampled before deterioration of wildfire effects, sites with ponderosa pine as the dominant tree species and sites where wildfire behavior was not affected by suppression activities. Wildfires that had accurate pre-fire fuel treatment maps and records were also favored.

We began field searching for suitable wildfires in 1995 and ended our search in 1998. During that time, twelve sites were considered for inclusion in this study. Of those, only four wildfires met our selection criteria: Webb Fire in Montana; Tyee fire in Washington; Cottonwood Fire in California; and Hochderffer Fire in Arizona. Table 1 summarizes the 12 fires that were considered for this study, but not selected. Table 1 also provides anecdotal observations on the effectiveness of fuel treatments.

[Insert] Table 1. Candidate fires that were considered but not selected for this study (Omi 1997). This table provides anecdotal evidence supporting the benefits of fuel treatments in mitigating wildfire spread and related damages.

Sites were selected in ponderosa pine forests that had areas of adjacent untreated and treated stands and that were burned in wildfires. The following criteria were used to select sites for the study:

$ stands where ponderosa pine is the major species;

$ adjacent treated and untreated stands exposed to the same recent wildfire;

$ stands that had accurate treatment records (i.e., maps, timber sale inventories); and

$ stands that were treated within 15 years prior to wildfire. In ponderosa pine forests, stands that were treated greater than 15 years prior to wildfire may have out-grown the effects of the fuel treatment.

Stands from each category were adjacent to each other to facilitate comparisons. We avoided selecting sites with confounding influences such as roads, wide streams or constructed firelines that may have a significant effect on fire behavior. Since slash resulting from logging operations increases fire hazard, at least in the short run (Fahnestock 1968, Vihanek and Ottmar 1993), only thinned stands where slash residues were effectively removed prior to wildfire incidence were considered.

Field Data Collection

Selected ponderosa pine stands were categorized as either Atreated@ or Auntreated@ depending on the presence of a fuel treatment. We consulted with agency officials and reviewed forest records to determine the fitness of sites. Adjacent untreated and treated stands were assumed to be equivalent prior to the treatment. By selecting stands that were adjacent to each other and on similar topography, we minimized the differences in weather and topography between the untreated and treated areas.

The first site, Webb, had a prescribed fire only fuel treatment. (Jolie: do you want to comment later on the difficulty of sampling prescribed fire only sites—i.e., since treatment effects such as stumps won’t be apparent?) After sampling on that site, we limited fuel treatments to some type of mechanical tree removal, with or without subsequent prescribed fire. We focused the three later sites on mechanical fuel treatments since prescribed fire was already known to mitigate fire effects (Wagle and Eakle 1979) and we wanted to narrow the focus of this study.

Plots located along transects captured the variability in the untreated and treated areas. We sampled an equal number of plots in the untreated and treated areas. Transect locations were located based on terrain and topography, and on the treatment and wildfire boundaries. Depending on the site, three or four transects that spanned the treated and untreated areas were situated parallel 150 meters apart. Six to eight plots per transect were located 150 meters apart. By selecting plot transect locations prior to any field visits, we avoided locating plots in areas that would possibly introduce bias. Prior to starting field sampling, we mapped transects and plot locations on a 72 minute topographic map that delineated the treated and untreated stands.

We studied modifications of stand structure and canopy characteristics that are known to mitigate fire hazard. To determine the fuel treatment=s effect on stand characteristics, three variables describing stand structure were measured: stand density (trees/hectare), basal area (meters²/hectare) and average diameter (cm) of trees on the plot. Sample trees were selected using variable plot sampling using a Acruiser=s crutch@ angle gauge.

Crown characteristics, especially crown bulk density and height to the live crown, are known to affect crown fire initiation and propagation (VanWagner 1977, Rothermel 1991). Since crown bulk density estimates cannot be determined accurately from simple field measurements, crown weight was used as a substitute for crown bulk density (Brown 1978). Formulas to determine the crown weight (kg) for the Webb, Tyee and Cottonwood sites incorporated the ratio of crown to tree height, diameter at breast height (DBH) and crown position (Brown 1978). DBH was measured with a metric diameter tape, and crown length and tree height were calculated from clinometer measurements. Crown position, whether dominant, co-dominant or intermediate, was recorded for each tree in the plot. We accounted for the height to live crown factor by incorporating it into the crown weight formula computed for each tree. At the Hochderffer site, time constraints precluded crown weight measurements. By eliminating crown weight data, we could sample more plots over a shorter period of time. In addition, we collected ample data from the three previous wildfire sites to test any relationships between crown weights and severity measurements.

Fire severity was classified by observing foliage scorch and crown needle consumption (Wagener 1961, Wyant et al. 1986). Crown scorch percent and crown position were estimated ocularly (Peterson 1985, Wyant et al. 1986) and, crown scorch height and percent crown scorch measurements were adapted from Ryan and Noste (1985).

One estimate of fire severity rating per plot was ocularly determined based mostly on the condition of the aerial fuels. We did not complete a fire severity estimate from the soil/forest floor organic layer perspective because the elapsed time since the fire to sampling resulted in deterioration of much of that evidence. The following severity rating criteria were adapted from Omi and Kalabokidis (1991):

$ Unburned, fire did not enter the stand (rating=1);

$ Light, surface burn without crown scorch (rating=2);

$ Spotty, irregular crown scorch (rating=3);

$ Moderate, intense burn with complete crown scorch (rating=4);

$ Severe, high intensity burn with crowns totally consumed (rating=5).

We used multi-variate response permutation procedures (MRPP) for statistically testing differences between the untreated and treated groups in this study (Mielke 1986, Good 1994). Non-parametric tests, such as MRPP, have several advantages compared to using more well-known parametric procedures. While t-tests are frequently used for two sample comparisons, the validity of the assumptions of the t-test are questionable in this study. The various data sets in this study were relatively small and contained several outliers. MRPP techniques may be superior to t-tests when the sample size is small, if the assumption of normally distributed populations is not reasonable (i.e., samples contain extreme values or outliers), and if multivariate comparisons are desired. For other examples of MRPP used in forestry studies, see Huckaby and Moir (1995) and Reich (1991).

Selected Study Site Descriptions

The four sites we sampled all met the selection criteria, but each site was unique in terms of stand characteristics, treatment type, and wildfire behavior. Table 2 summarizes general descriptions for the four wildfires and treatment types.

[Insert] Table 2. Description of sampling sites at the Webb, Tyee, Cottonwood and Hochderffer wildfires

Figures 1, 2,3 and 4 show the adjacent treated and untreated stands at the four sampling sites.

[Insert] Figure 1. Photos of untreated and treated stands at the Webb Fire site at adjacent locations.

[Insert] Figure 2. Photos of untreated and treated stands at the Tyee Fire site at adjacent locations.

[Insert] Figure 3. Photos of untreated and treated stands at the Cottonwood Fire site at adjacent locations.

[Insert] Figure 4. Photos of untreated and treated stands at the Hochderffer Fire site at adjacent locations. Multiple stems with full crowns in the foreground of the treated photo mask the larger diameter trees in this plot.

Results

Tables 3, 4, 5, 6, and 7 summarize the results. Table 3 shows that post-fire basal area is higher in the untreated plots for all sites except Cottonwood (see below for further explanation). Slightly higher basal areas in the treated stands may be explained by understanding that a stand with many small trees may have similar basal area to a stand with few large trees.

The number of trees per hectare is much higher in the untreated stands at all four sites; the untreated Tyee site was especially dense with 1,244 trees per hectare. The average diameter of trees on the plots is higher for the treated stands which shows that the fuel treatment removed smaller diameter trees. The crown scorch percent and fire severity rating are higher for untreated stands at all four sites. The treated stands had higher crown weights. The formulas for estimating crown weights (Brown 1978) are most influenced by diameter. Thus, larger diameter trees, such as those found in the treated stands, will produce greater crown weights.

Some differences in topography are evident between the untreated and treated sites. Due to generally more active fire behavior in west-facing sites compared to northwest aspects, one may expect more severe fire effects on western aspects. However, at the Tyee site, higher fire severity was found in plots with a northwest aspect. Further, inspection of the slope data showed that for two sites the treated areas had steeper slopes and for the other two sites the untreated area had steeper slopes.

[Insert] Table 3. Key site characteristics for the four wildfires. (Standard deviations are in parentheses.) Identical superscripts indicate that the untreated and treated sites are not significantly different using univariate MRPP, a=.05 (Good 1994). Wildfires with different superscripts indicate that the sites are significantly different. Items without superscripts were not tested.

The basal area differences for the Cottonwood site yield some peculiar results. The two means are almost identical (30.0 versus 30.3 m²/ha) but are significantly different. Examining the Cottonwood site=s density and tree diameters, the slightly higher basal area in that treated stand may be attributed to that stand having fewer but larger trees. There are many outliers at the Cottonwood site and the data ranges are very different between the untreated and treated plots. Basal areas in the treated areas ranged from 21 to 39 m²/ha compared to the untreated range of 4 to 60 m²/ha. Only 16% (2 observations) of the observations for the untreated area fell within the range of the treated area. The likelihood that such extreme values (i.e., basal area <20 m²/ha or basal area>40 m²/ha) would be observed in the treated plots is very small. Therefore, the two plots have significantly different basal areas even though their means are almost identical.

Statistical analysis provided additional insights into structural differences between treated versus untreated stands. Univariate and multivariate stand structure comparisons between untreated and treated plots are analyzed (Tables 4 and 5). Differences in fire severity rating and percent crown scorch in untreated versus treated plots is presented statistically using MRPP (Table 6). Lastly, a correlation matrix (Table 7) shows associations between independent variables (density, basal area, diameter of trees on the plot, crown weight and slope) and the dependent variables (fire severity rating and crown scorch).

Results indicate that the untreated and treated stands are significantly different for the Webb, Tyee and Cottonwood sites. The lack of significant differences among the univariate and multivariate stand characteristic comparisons for the Hochderffer site is particularly noteworthy (Tables 4 and 5). Notice, however, the significant differences at that site for fire severity rating and crown scorch (Table 6). Something other than stand structure factors likely contributed to the differences in fire severity. Surface fuel loading or differences in fuel moistures rather than stand structure may have been the fire severity driver at this site.

[Insert]Table 4. P-values for univariate comparisons using MRPP comparing basal area (m²/ha), density (#stems/ha) and diameter (cm) between treated and untreated plots for the four sites (Good 1994).

[Insert] Table 5. Multivariate MRPP comparisons for basal area (m²/ha), density (#stems/ha), and diameter (cm) on the four sites (Good 1994). These data were standardized [(x-median)/range] to eliminate differences in units.

[Insert] Table 6. P-values for univariate comparisons using MRPP comparing fire severity rating and percent crown scorch between untreated and treated plots for the four sites (Good 1994).

Table 7 illustrates the correlation coefficients showing trends and relationships among the independent and dependent variables. Relationships among the independent and dependent variables are the most interesting and meaningful to this study. Density, basal area, diameter and crown weight all are significantly correlated with plot severity rating and percent crown scorch. The highest correlation coefficient (r=0.57, r²=0.32) between the independent and dependent variables is among density and fire severity rating. Thus 32% of the variation in fire severity rating can be explained by the variation in density. Slope does not appear to be related to fire severity or percent crown scorch.

(Jolie: This deletion could be moved to the Discussion, if you wish)

[Insert] Table 7. Summary of correlation coefficients (r) for Webb, Tyee, Cottonwood and Hochderffer sites for fire damage/severity variables (fire severity rating and percent crown scorch), stand structure variables (density, basal area, average diameter of trees on the plot and crown weight) and slope. P-values are in parentheses.

Discussion

The treated plots in this study have lower fire severity ratings and less crown scorch than the untreated plots. The null hypothesis (H_o), that both fire severity and crown scorch each do not differ significantly among untreated and treated plots, is rejected in favor of the research hypothesis (H_a), that fire severity and crown scorch are higher in untreated plots.

From these results we infer that the types of fuel treatments studied reduce fire severity rating and crown scorch. Based on the statistical results and field reconnaissance, sites with mechanical fuel treatment appear to have more dramatically reduced fire severity compared to sites with prescribed fire only. Although fire severity ratings and percent crown scorch are significantly different for untreated versus treated plots at all sites (Tables 3 and 6), the Webb site=s differences were the least extreme. Apparently, mechanical fuel treatments at the Tyee, Cottonwood and Hochderffer sites allow for more precise and controlled results compared to prescribed fire. For example, mechanical fuel treatment programs may specify the exact number of post-treatment residual trees per hectare and the treatment can be applied uniformly across the stand. By contrast, prescribed fire fuel treatment often varies across a stand and results in less precise stand structure changes.

For the Webb, Tyee and Cottonwood sites, the stand characteristics contributed to the differences in fire severity. The fuel treatments at these three sites resulted in forests with much lower density and larger trees. Stands with fewer trees have less continuous crown and ladder fuels. Larger trees generally have crowns higher off the ground and have thicker bark which makes them more fire resistant. This twofold benefit of treated stands results in lower potential for crown fire initiation and propagation and for less severe fire effects.

Stand structure for the Hochderffer site is not significantly different among the treated and untreated stands; other factors contributed to less severe fire effects in the treated stands since fire severity and percent crown scorch differences cannot be explained by stand structure manipulations. We did not study two factors that are known to drive fire behavior: surface dead downed fuel loading and fuel moistures. Due to the post facto nature of this study, we could not adequately quantify pre-fire fuel loadings or fuel moistures. Although fuel loading was not quantified in this study, we assume that the studied fuel treatments reduce surface fuel loading. For example, a recent prescribed burn will reduce surface fuel loading in the short-run. The Hochderffer site had a recent prescribed burn after the mechanical thinning treatment. Therefore, it is likely that surface fuel loading contributed to less severe fire effects in the treated stands at the Hochderffer site.

Fuel moistures may be affected by microclimate and probably do vary between the untreated and treated stands. A more open stand allows more wind and solar radiation resulting in a drier microclimate compared to a closed stand. A drier microclimate generally contributes to more severe fire behavior. However, our study does not support the assertion that more open stands experience higher fire severity. More open stands had significantly less fire severity compared to the more densely stocked untreated stands in this study. The degree of openness in the studied treated stands may not have been sufficient to increase fire activity.

Density and average tree diameter are closely related to fire severity (Table 7). Based on this study=s results, removing small diameter trees from a ponderosa pine stand reduces subsequent wildfire severity. At the four sites, the fuel reduction overcomes the microclimate=s effect on fire behavior. These findings agree with years of forest managers= field observations.

Many fire scientists have put much effort toward crown fire modeling and refinements are continually emerging (VanWagner 1977, Rothermel 1991, Agee 1996, Scott 1998c). Because crown fire modeling is such a complicated and heavily studied topic, we felt this paper would not provide unique insight into modeling procedures. Correlations between the independent variables (the stand characteristic variables) and the dependent variables (crown scorch and fire severity rating) indicate relationships and trends that reinforce understanding of crown fire and fire effects.

Critics of this research may consider alternative explanations for our study=s differences in fire severity: that the variability in fire effects is related to the random nature of wildfire behavior and not due to the stand=s fuel manipulation. In other words, the results were obtained by chance alone. All the interacting factors that govern the spread and intensity of wildfires are not clearly understood. The ways fire burn are influenced by multitudes of factors, some of which have been studied and measured and other factors have not. This study aimed to reduce influences other than fuel treatments in order to test the effectiveness of those treatments for reducing wildfire severity. Although we were unable to capture all of the subtleties that govern fire behavior and severity, the evidence is convincing that the treatment mitigates wildfire severity by whatever means. For three of the four study areas, stand structure differences between treated and untreated stands account for differences in fire severity (Table 5). And there is clear evidence that fuel treatments in this study reduce wildfire severity (Table 6) at all four sites. For the first time the assertion that fuel treatments reduce wildfire severity has been tested and analyzed.

Site selection was a significant and time-consuming aspect of this study. Most of the wildfires (8 of 12) that we seriously considered did not meet the specified criteria. Several of the proposed sites also had a surprising lack of accurate forest records that precluded sampling. For example, forest managers knew of fuel treatments in areas burned by subsequent wildfire, but had little data to confirm the treatment locations or the exact nature of the fuel treatment.

Sites that had fuel treatments could not be selected in advance since a wildfire had to take place before a treatment could be studied. Therefore, all measurements were taken after wildfire occurrence. Developing and synthesizing methods for assessing post facto fire effects that may be useful for similar studies elsewhere was an important aspect of this research. Problems involved with a post facto analysis and ways the problems were addressed include:

$ Wildfire events cannot be predicted in advance and quantitative pre-wildfire site information is often unavailable. Data may be lacking to quantify the fuel treatment to the degree desirable for a scientific analysis We developed sampling methods that allowed for quantification of the pre-wildfire condition in terms of: density, basal area, size of the trees, and crown weights. Additionally, we only chose sites that had sufficient fuel treatment data prior to the wildfire.

$ Confounding influences may cloud observed wildfire effects. These include: suppression activities disturbing the landscape and retarding wildfire spread; roads affecting fire spread; and salvage activities obliterating fire severity evidence.

$ Difficulty in minimizing the variability across and among the sites. Variability may mask treatment effects. We selected stands according to pre-specified conditions, including: ponderosa pine dominance; similar aspect for treated and untreated stands; treated and untreated stands that are adjacent; and treated stands that were treated 10 years or less before the wildfire. We selected sites that were adjacent with similar aspect and topography to minimize other fire behavior influences. Adjacent sites are also likely to have similar soils, similar vegetation types and similar weather patterns. These similarities reduce the possibility that factors other than the fuel treatment are the major determinants of fire severity.

$ Time since wildfire must be relatively short; otherwise evidence may be lost or erased (i.e., ground char, crown scorch/consumption). We kept the time since wildfire at three years or less. The Webb, Tyee and Hochderffer sites were sampled only one year post-wildfire and had more preserved evidence compared to the Cottonwood site that was sampled two years post-wildfire.

The best methods for assessing fire severity accurately require observing an active fire=s behavior or by immediate post-fire observation. That way, real-time fire behavior may be measured instead of estimated and there is no removal of the remaining evidence. Due to the unpredictable nature of wildfire timing and locations it was impossible for this study to take real-time fire behavior measurements or make immediate post-fire observations. We encountered delays of one year or more between the wildfire=s occurrence and field sampling. It took several months after a wildfire occurred for land managers to learn of this study, arrange a site visit and determine if a site met the selection criteria. Time was also lost gathering records and waiting for suitable weather for sampling.

Although several problems associated with a post facto study have been highlighted, few feasible study design alternatives exist. Researchers could potentially concentrate on quantifying fuels and fuel treatments over an area large enough to contain ample fuel treatment and control plots. A prescribed burn could then be conducted to examine the effects the treatment had on fire severity. However, it is difficult to replicate the scale and weather of a crowning wildfire. Another alternative is to pre-select an area and complete a careful fuel treatment program where all conceivable variables affecting fire behavior are measured pre-fire. Monitoring equipment could be set up for measuring wildfire observations. Then, researchers could wait for a wildfire to burn the entire area. However, these alternative methods are too impractical to successfully implement. Other studies have carefully quantified fuel treatments and used standard computer fire behavior modeling to determine the fuel treatment effects on fire behavior (Kalabokidis and Omi 1998, Scott 1998a, Stephens 1998). Computer simulations avoid problems associated with a post facto field study, but computer simulation is not always a good substitute for describing actual fire behavior.

Conclusions and Management Implications

Our findings indicate that fuel treatments do mitigate fire severity. Treatments provide a window of opportunity for effective fire suppression and protecting high-value areas. Although topography and weather may play a more important role than fuels in governing fire behavior (Bessie and Johnson 1995), topography and weather cannot be realistically manipulated to reduce fire severity. Fuels are the leg of the fire environment triangle (Countryman 1972) that land managers can change to achieve desired post-fire condition. However, in extreme weather conditions, such as drought and high winds, fuel treatments may do little to mitigate fire spread or severity.

This study shows that fuel treatments are effective in reducing severity in short fire-return interval ecosystems. However, fuel treatments in long fire-return interval ecosystems may be less effective. Most ponderosa pine forests have adapted to recurring, low severity fire. Wildfires in lodgepole pine forests, by contrast, are comparatively infrequent and have high severity. Due to the historical differences in wildfire frequency and severity in these types of forests, the effectiveness of fuel treatments in long fire return-interval ecosystems remains unclear.

Traditional fuel treatment programs were completed on distinct and often disjointed units, instead of on a landscape-level scale. In order to lessen the probability of a high-severity wildfire over a landscape, an entire landscape should be analyzed for determining the most appropriate scales and locations for fuel treatments. Intensively treating most of a landscape may not be necessary, but treating strategically located stands for fuel treatment or treating strips of fuels may be beneficial for reducing severe wildfire potential across a large area. The scale of fuel treatments likely affects their efficacy. For example, Tyee fire observers reported that in at least one instance the fire took over 100 meters to drop out of the crowns when it encountered the fuel treatment area. Many sites had similar fire severity in treated and untreated plots, especially near the treatment boundary. If only a small area was treated, a high-intensity crown fire may have enough momentum to burn right through a treated area. Mutch and Cook (p. 9, 1996) emphasize Aprescribed fire has not been used on a scale adequate for sustaining the productivity of fire-dependent ecosystems.@

There are at least three ways to reduce tree densities and accomplish fuel treatments: wildfire, prescribed fire and mechanical thinning. The first, natural fires, are often impractical. Letting natural fires play their historical role may have unwanted effects in forests that have undergone major stand structural changes over the past years of fire exclusion. Any fire started may result in historically uncharacteristic high severity. In many ponderosa pine forests choked with dense, small-diameter trees, or encroached by shade-tolerant trees, natural fires may no longer play a strategic role.

The second strategy for restoring these forests is large-scale prescribed burning. This is likely to be effective in stands that have moderate or low tree densities, little encroachment of ladder fuels, moderate to steep slopes which preclude mechanical treatment, and expertise in personnel to plan and implement such large prescribed burns. Large-scale implementation of this strategy will require funding for the planning and implementation over current expenditures and may require modifications to current air quality legislation. Future results of such expenditures may be seen down the road in lessened wildfire suppression costs, reduced fire severity, and reduced air quality impacts.

Mechanical tree removal, the third strategy, works best on forests that are too densely packed to burn, that have nearby markets for small-diameter trees, and areas where expertise and personnel are not available for prescribed burning programs. Mechanical tree removal may be accomplished by many different types of harvest, including precommercial thinning, selection or shelterwood harvest coupled with small-diameter tree removal, and thinning from below (Fiedler 1996). The goal is to manage forests for much lower tree densities leaving larger residual trees. Harvests to reduce wildfire hazard will remove small-diameter trees in contrast to traditional timber harvests. Mechanical fuel treatments can be very labor intensive, especially on steep slopes and in remote areas, and may not be commercially attractive due to the small diameter trees that need removal. To make fuel treatments more cost-effective for small-diameter trees, consistent markets are necessary (Nakamura 1996). Fiedler et al. (1997) assert that mechanized tree harvest on moderately-steep terrain coupled with removal of large amounts of biomass can generate considerable revenue. Periodic underburns and programs for restoring natural fire are critical to maintain these post-harvest stands.

Fuel treatment programs may be costly and time-consuming. But wildfire problems aren=t going away soon. We suggest focusing programs, funding and management attention where the risk resulting from severe wildfire is greatest: urban-interface, tree plantations, critical watersheds and habitat for threatened and endangered species. Treating high-volume areas using mechanized equipment could offset costs associated with fuel removal on steep slopes with little timber. Costs associated with wildfire suppression, in terms of funding suppression efforts and personal safety, far outweigh the costs of fuel treatment on similar landscapes.

Acknowledgements

The authors would like to thank the following people for their help in facilitating this research: Steve Arno retired from the Rocky Mountain Research Station, Fire Sciences Laboratory; Ron Hvizdak, Kootenai National Forest; Michelle Ellis, Wenatchee National Forest; Scott Abrams, Tahoe National Forest; Allen Farnsworth, Coconino National Forest; Lyn Morelan, Boise National Forest; and Robin Reich, Colorado State University.

References

Agee, J.K. 1993. Fire Ecology of Pacific Northwest Forests. Island Press, Washington, D.C. 493p.

Agee, J.K. 1996. The influence of forest structure on fire behavior. In: 17th Forest Vegetation Management Conference, Redding, CA. Feb. 16-18, 1996. pp.52-68.

Arno, S.F. and J.K. Brown. 1991. Overcoming the paradox in managing wildland fire. Western Wildlands. 17(1):40-46.

Babbitt, B. 1995. Return fire to its place in the West. Fire Management Notes. 55(4):6-8.

Bessie, W.D. and E.A. Johnson. 1995. The relative importance of fuels and weather on fire behavior in subalpine forests. Ecology. 76(3):747-762.

Biswell, H.H. 1960. Danger of wildfire reduced by prescribed burning in ponderosa pine. California Agriculture. 14(10):5-6.

Biswell, H.H., R.P. Gibbens and H. Buchanan. 1968. Fuel conditions and fire hazard reduction costs in a giant sequoia forest. National Parks Magazine. August:17-19.

Brown, J.K. 1978. Weight and density of crowns of Rocky Mountain conifers. USDA Forest Service Research Paper INT-197. 56p.

Cooper, C.F. 1960. Changes in vegetation, structure and growth of southwestern pine forests since white settlement. Ecological Monographs. 30:129-164.

Countryman, C. 1972. Fuel moisture and fire danger - some elementary concepts. USDA Forest Service Region 5, Aviation and Fire Management. 4p.

Covington, W.W. and M.M. Moore. 1994. Southwestern ponderosa pine forest structure: changes since Euro-American settlement. Journal of Forestry. 92(1):39-47.

Dickman, A. 1978. Reduced fire frequency changes composition of a ponderosa pine stand. Journal of Forestry. 76(1):24-25.

Dodge, M. 1972. Forest fuel accumulations - a growing problem. Science. 177:139-142.

Edminster, C.B. and W.K. Olsen. 1995. Thinning as a tool in restoring and maintaining diverse structure in stands of southwestern ponderosa pine. USDA Forest Service General Technical Report RM-278. pp.62-68.

Fahnestock, G.R. 1968. Fire hazard from precommercial thinning of ponderosa pine. USDA Forest Service Research Note PNW-57. 16p.

Fiddler, G.O., D.R. Hart, P.M. McDonald and S.J. Frankel. 1995. Silvicultural practices (commercial thinning) are influencing the health of natural pine stands in eastern California. USDA Forest Service General Technical Report RM-267. pp.208-212.

Fiedler, C.E., S.F. Arno and M.G. Harrington. 1995. Flexible silvicultural and prescribed burning approaches for improving health of ponderosa pine forests. USDA Forest Service General Technical Report RM-278. pp.69-74.

Fiedler, C.E. 1996. Silvicultural applications: restoring ecological and structural process in ponderosa pine forests. USDA General Technical Report INT-341. pp.39-40.

Fiedler, C.E., C.E. Keegan and S.F. Arno. 1997. Utilization as a component of restoring ecological processes in ponderosa pine forests. USDA General Technical Report FPL-100. pp.24-28.

Gale, R.D. 1977. Evaluation of fire management activities on the National Forests. USDA Forest Service Policy Analysis Staff Report. 127p.

Gonzalez-Caban, A., J. Loomis, R. Gregory. 1995. Protecting Oregon old-growth forests from fires: how much is it worth? USDA Forest Service General Technical Report PSW-156. p.139-143.

Good, P. 1994. Permutation Tests: A Practical Guide to Resampling Methods for Testing Hypotheses. Springer-Verlag, New York. 228p.

Huckaby, L.S. and W.H. Moir. 1995. Fire history of subalpine forests at Fraser Experimental Forest, Colorado. USDA Forest Service General Technical Report INT-310. pp.205-210.

Kalabokidis, K.D. and P.N. Omi. 1998. Reduction of fire hazard through thinning/residue disposal in the urban interface. International Journal of Wildland Fire. 8(1):29-35.

Laudenslayer, W.F., H.H. Darr and S. Smith. 1989. Historical effects of forest management practices on eastside pine communities in northeastern California. USDA Forest Service General Technical Report RM-185. pp.26-34.

MacCleery, D.W. 1995. The way to a healthy future for National Forest ecosystems in the west: what role can silviculture and prescribed fire play? USDA Forest Service General Technical Report RM-267. pp.37-45.

McLean, H.E. 1993. The Boise quickstep. American Forests. 99(182):11-14.

Mielke, P.W. 1986. Non-metric statistical analysis: some metric alternatives. Journal of Statistical Planning and Inference. 13:377-387.

Mutch, R.W., S.F. Arno, J.K. Brown, C.E. Carlson, R.D. Ottmar and J.L. Peterson. 1993. Forest health in the Blue Mountains: a management strategy for fire-adapted ecosystems. USDA Forest Service General Technical Report PNW-310. 14p.

Mutch, R.W. 1994. Fighting fire with prescribed fire: a return to ecosystem health. Journal of Forestry. 92(11):31-33.

Mutch, R.W. and W.A. Cook. 1996. Restoring fire to ecosystems: methods vary with land management goals. USDA Forest Service General Technical Report INT-341. pp.9-11.

Nakamura, G. 1996. Harvesting forest biomass reduces wildfire fuel. California Agriculture. 50(2):13-16.

Omi, P.N. and K.D. Kalabokidis. 1991. Fire damage on extensively versus intensively managed forest stands within the North Fork Fire, 1988. Northwest Science. 65:149-157.

Omi, P.N. 1997. Final Report: Fuels Modification to Reduce Large Fire Probability. Submitted to US Department of Interior, Fire Research Committee. 66p.

Orozco, P. 1998. Personal communication. USDA, Pike and San Isabel National Forest and Cimarron and Comanche National Grasslands, Fire Management Officer.

Peterson, D.L. 1985. Crown scorch volume and scorch height: estimates of post-fire tree condition. Canadian Journal of Forest Research. 15:596-598.

Reich, R.M. 1991. Spatial analysis of ponderosa pine trees infected with dwarf mistletoe. Canadian Journal of Forest Research. 21:1808-1815.

Rothermel, R.C. 1991. Predicting behavior and size of crown fires in the northern Rocky Mountains. USDA Forest Service General Technical Report INT-438. 46p.

Ryan, K.C. and N.V. Noste. 1985. Evaluating prescribed fires. USDA General Technical Report INT-182. pp.230-238.

Saab, V. and J.G. Dudley. 1998. Response of cavity nesting birds to stand replacement fire and salve logging in ponderosa pine/Douglas-fir forests of southwest Idaho. USDA Forest Service General Technical Report RMRS-11. 17p.

Scott, J. 1998a. Fuel reduction in residential and scenic forests: a comparison of three treatments in a western Montana ponderosa pine stand. USDA Forest Service Research Paper RMRS-5. 19p.

Scott, J. 1998b. Reduce fire hazards in ponderosa pine by thinning. Fire Management Notes. 58(1):20-25.

Scott, J. 1998c. Sensitivity analysis of a method for assessing crown fire hazard in the Northern Rocky Mountains, USA. In: 3rd International Conference on Forest Fire Research. November 16-20, 1998, Luso-Coimbra, Portugal.

Stephens, S.L. 1998. Evaluation of the effects of silvicultural and fuels treatment on potential fire behavior in Sierra Nevada mixed-conifer forests. Forest Ecology and Management. 105:21-35.

VanWagner, C.E. 1977. Conditions for the start and spread of crown fire. Canadian Journal of Forest Research. 7:23-34.

Vihanek, R.E. and R.D. Ottmar. 1993. When logged units burn in a wildfire, does slash treatment mitigate effects? In: 12th Conference on Fire and Forest Meteorology, October 26-28, 1993, Jekyll Island, Georgia. pp.709-714.

Wagener, W.W. 1961. Guidelines for estimating the survival of fire-damaged trees in California. USDA Forest Service Miscellaneous Paper PSW-60. 11p.

Wagle, R.F. and T.W. Eakle. 1979. A controlled burn reduces impact of a subsequent wildfire in a ponderosa pine vegetation type. Forest Science. 25:123-129.

Weatherspoon, C.P. and C.N. Skinner. 1995. An assessment of factors associated with damage to tree crowns from the 1987 wildfires in northern California. Forest Science 41(3):430-451.

Weaver, H. 1943. Fire as an ecological and silvicultural factor in the ponderosa pine region of the pacific slope. Journal of Forestry. 41:7-14.

Weaver, H. 1961. Implications of the Klamath fires of September 1959. Journal of Forestry. 59(8):569-572.

Williams, J. 1998. Ecosystem management brings concepts into practice. Fire Management Notes. 58(2):14-16.

Wyant, J.G., P.N. Omi, and R.D. Laven. 1986. Fire induced tree mortality in a Colorado ponderosa pine/Douglas-fir stand. Forest Science. 32(1):49-59.

Table 1. Candidate fires that were considered but not selected for this study during 1992-95 (Omi 1997).

Fire Name	Year	Size	Location	Description
Cleveland	1992	n/a	Eldorado National Forest, CA	Ponderosa pine plantations previously underburned survived a wildfire when suppression crews were able to backfire from the treated areas. This site was not selected due to suppression activities near the treatment boundary.
Paddock	1992	12 ha	Lakeview, OR	Fire spread into ponderosa pine that was underburned 3 years prior to wildfire. The fire had potential to reach 2,000 ha and spread on to private land. Land managers felt that the treatment limited the wildfire size, resource damage and suppression costs. This site was not selected because it did not have mechanical fuel treatment. We later limited studied fuel treatments to some type of mechanical treatment.
Star Gulch	1992	12,000 ha	Boise National Forest, ID	Ponderosa pine plantations that were thinned and underburned survived a wildfire. Untreated plantations experienced high mortality. The time lapse between the wildfire and study notification was too long for this study to be included since there was much deterioration of fire effects evidence.
Aspen	1994	664 ha	Salyer National Wildlife Refuge, SD	An escaped prescribed fire in aspen-grassland-shrubland became controllable in an aspen clear-cut. Ponderosa pine was not a dominant species.
Henry Peaks	1994	3,240 ha	Flathead Reservation, MT	An area thinned 20 years prior to the wildfire by uneven-aged logging (whole tree skidding with pile burning) experienced significantly lower fire severity and mortality compared to adjacent forest. The length of time since treatment precluded this site=s selection.
LeClair	1994	13,355 ha	Warm Springs Reservation, OR	Dozer line and prescribed burning one year prior to wildfire in sagebrush-grass held the fire at a subdivision boundary. Clear-cutting as a fuel treatment did not meet the study=s objectives.
Robinson	1994	3,400 ha	Yellowstone National Park, WY	Beetle-killed lodgepole pine (self-thinned to lower density) experienced significantly lower fire severity compared to adjacent burned areas. The dominant vegetation was not ponderosa pine and it was a naturally treated stand, not a mechanical fuel treatment.
Wind	1995	40 ha	Deschutes National Forest, OR	Fire behavior became more controllable in a grass and rabbitbrush area treated by prescribed fire in 1987. This enabled a dozer line to contain the wildfire. The dominant vegetation was not ponderosa pine.

Table 2. Description of sampling sites at the Webb, Tyee, Cottonwood and Hochderffer wildfires.

	Fire
	Webb	Tyee	Cottonwood	Hochderffer
Treatment Type	broadcast burn in 1989	precommercial thinning in 1970s with underburn for slash removal in 1983	whole tree thinning in 1989, 1990	undetermined tree harvest in 1970s with broadcast burn in 1995
Date of Fire	September, 1994	August, 1994	August, 1994	June, 1996
Date Sampled	July, 1995	October, 1995	September, 1996	October, 1997
Size of Fire	1,415 ha	56,780 ha	18,620 ha	6,640 ha
Elevation	1,067 m	762 m	2,012 m	2,408 m
Aspect	south	west	west	north
National Forest Location	Kootenai NF, Montana	Wenatchee NF, Washington	Tahoe NF, California	Coconino NF, Arizona

Table 3. Key site characteristics for the 4 wildfires. (Standard deviations are in parentheses.) Identical superscripts indicate that the untreated and treated sites are not significantly different using univariate MRPP, a=.05 (Good 1994). Wildfires with different superscripts indicate that the sites are significantly different. Items without superscripts were not tested.

		Sam- ple Size	Aspect	Slope (%)	Basal Area (m^2/ha)	Density (stems/ha)	Avg. Diameter (cm)	Fire Severity Rating	Crown Scorch (%)	Crown Weight (kg)
Webb	Untreated	9	S	29^a (13)	23.0^a (13.9)	637^a (498)	24.1^a (10.5)	3.2^a (0.8)	67^a (33)	165^a (179)
	Treated	9	S	39^a (9)	14.7^a (9.4)	73^b (42)	43.1^b (20.0)	2.6^b (0.5)	26^b (34)	334^a (197)
Tyee	Untreated	9	NW	38^a (17)	24.6^a (12.8)	1244^a (1417)	20.7^a (13.6)	4.4^a (0.5)	100^a (0)	63^a (71)
	Treated	9	W	22^a (11)	15.8^a (5.6)	218^b (98)	30.7^b (0.7)	3.0^b (0.0)	74^b (17)	142^b (75)
Cotton- wood	Untreated	12	W	21^a (14)	30.0^a (19.7)	578^a (571)	31.3^a (11.7)	4.0^a (0.8)	78^a (32)	194^a (146)
	Treated	12	S	11^b (5)	30.3^b (5.5)	262^b (184)	39.8^b (8.4)	2.7^b (0.6)	26^b (27)	225^a (210)
Hoch- derffer	Untreated	12	N	11^a (6)	25.0^a (8.1)	765^a (441)	21.9^a (7.0)	4.4^a (0.6)	99^a (1)	NOT MEAS-URED
	Treated	12	N	13^b (6)	23.3^a (11.0)	556^a (410)	24.2^a (6.5)	2.1^b (1.0)	29^b (38)

Table 4. P-values for univariate comparisons using MRPP comparing basal area (m²/ha), density (#stems/ha) and diameter (cm) between treated and untreated plots for the four sites (Good 1994).

	Basal Area	Density	Diameter
Webb	0.37	0.01*	0.01*
Tyee	0.14	0.01*	0.02*
Cottonwood	0.04*	0.05*	0.03*
Hochderffer	0.49	0.32	0.45

* Indicates that the treated and untreated plots are significantly different, a=.05.

Table 5. Multivariate MRPP comparisons for basal area (m²/ha), density (#stems/ha), and diameter (cm) on the four sites (Good 1994). These data were standardized [(x-median)/range] to eliminate differences in units.

	p-value
Webb	0.01*
Tyee	0.02*
Cottonwood	0.02*
Hochderffer	0.96

* Indicates that the treated and untreated plots are significantly different, a=.05.

Table 6. P-values for univariate comparisons using MRPP comparing fire severity rating and percent crown scorch between untreated and treated plots for the four sites (Good 1994).

	Fire Severity Rating	Percent Crown Scorch
Webb	0.01*	0.03*
Tyee	0.01*	0.01*
Cottonwood	0.01*	0.01*
Hochderffer	0.01*	0.01*

* Indicates that the treated and untreated plots are significantly different, a=.05.

Table 7. Summary of correlation coefficients (r) for Webb, Tyee, Cottonwood and Hochderffer sites for fire damage/severity variables (fire severity rating and percent crown scorch), stand structure variables (density, basal area, average diameter of trees on the plot and crown weight) and slope. P-values are in parentheses.

	DEPENDENT VARIABLES		INDEPENDENT VARIABLES
	Fire Severity Rating	Percent Crown Scorch	Density	Basal Area	Diameter	Crown Weight^*	Slope
Fire Severity Rating	1.00	0.79 (<0.01^**)	0.57 (<0.01^**)	0.40 (<0.01^**)	-0.48 (<0.01^**)	-0.40 (<0.01^**)	0.09 (0.43)
Percent Crown Scorch		1.00	0.45 (<0.01^**)	0.26 (0.03^**)	-0.45 (<0.01^**)	-0.43 (<0.01^**)	0.12 (0.28)
Density			1.00	0.56 (<0.01^**)	-0.63 (<0.01^**)	-0.46 (<0.01^**)	0.02 (0.88)
Basal Area				1.00	-0.11 (0.34)	-0.16 (0.22)	-0.05 (0.67)
Diameter					1.00	0.86 (<0.01^**)	0.05 (0.67)
Crown Weight^*						1.00	-0.03 (0.82)

*Crown Weight was computed for Webb, Tyee and Cottonwood sites only.

** Indicates that the correlation coefficients are significantly different from 0, a=.05.