Abstract: In this research, we simulated wildfires that originated from hypothetical human-caused ignition points to determine whether a broad-scale schedule of fuel reduction treatments would be effective in reducing wildfire size or severity. The study area was a large watershed in Northeastern Oregon (USA). The fuel reduction treatments included commercial thinning and thinning followed by a prescribed fire treatment. These fuel reduction treatments were distributed across the landscape in such a way as to simultaneously maximize both an even-flow of timber harvest volume and a spatial pattern of activity (dispersed, clumped, random and regular). We found that the clumped and regular patterns of management activity seemed to reduce simulated wildfire severity most effectively in two out of three cases. A dispersed pattern of management activity required scheduling more area for treatment since treatments spaced as far apart as possible produced lower scheduled timber volumes, thus had no recognizable effect. A random pattern of fuel reduction activities also seemed to have no effect on characteristics of simulated human-caused wildfires.
INTRODUCTION
Over the last few decades, managers and researchers have been investigating methods for reducing the risk of catastrophic wildfires in western forests of North America. Fuel reduction treatments have received considerable interest as a primary wildfire mitigation strategy and have been extensively applied to this region. The beneficial effects of fuel reduction treatments have been noted in many studies, but most studies were conducted at a local scale (Helms, 1979; Martin et al., 1989; Agee, 1998; Agee and Skinner, 2005; Stephens and Moghaddas, 2005). It has been suggested that fuel reduction activities be spatially arranged on a landscape to effectively and efficiently disrupt the progress of wildfires (Shang et al., 2004; Agee and Skinner, 2005).
The concept of spatially distributing fuel reduction treatments across a landscape has been previously reported and we have found that fuel reduction treatments, spatially allocated across a large landscape, moderately reduced simulated wildfire severity compared to control solutions (i.e., those with no fuel reduction treatments). One of the basic premises of this study is that budgetary resources of fuel reduction treatments are limited, thus to best utilize limited resources, treatments should be spread across the landscape in a logical manner. As a result, we tested the efficacy of a clumped, dispersed, random and regular arrangement of activities for reducing certain characteristics of wildfires. We found that the efficiency of reducing wildfire severity was closely associated with overall intensity of treatment, as measured by the amount of harvested timber. Our earlier study assumed that simulated wildfires were randomly distributed across the landscape, however the ignition of naturally-caused fires is more likely influenced by weather conditions (Rorig and Ferguson, 1999; Podur et al., 2003; Wotton and Martell, 2005) or fuel conditions (Diaz-Avalos et al., 2001; Wotton and Martell, 2005) than by geological or topographical considerations. Recent studies have reported that ignition locations of naturally-caused fires were not necessarily related to elevation or frequency of lightning (Diaz-Avalos et al., 2001) and field observations suggest that ignition locations of human-caused fires are not random either. Therefore, we are concerned with understanding whether spatial patterns of fuel reduction activities can effectively disrupt wildfires started by humans and in doing so, we assume that the ignition points of these fires are not random and that they are located within short distances of major roads.
By modeling ignition locations according to ignition type, we expect to obtain practical results that will benefit fuel management or management planning efforts. There is, however, limited information about potential ignition locations of human-caused wildfire (Wotton et al., 2003), but through anecdotal information one could draw the conclusion that human-caused fires are often ignited near developed areas such as highways, recreation sites, or the wildland-urban interface. In this research, we modeled several sets of wildfire ignition locations that were assumed to be human-caused and subsequently examined the effects of a spatial pattern of fuel reduction treatments on these simulated wildfires. Also, we adopted several management scenarios that were optimized for arranging fuel reduction treatments in spatial patterns across a large landscape in a related research effort. The management scenarios provide a diversity of treatment types and treatment intensities. Therefore, in this research, three hypotheses are examined for comparison of management scenarios: 1) treatment effects on human-caused wildfires vary according to treatment size, 2) the type of treatment activities influences the effects on simulated human-caused wildfires and 3) the pattern of treatment activity across the landscape will have an affect on simulated human-caused wildfires.
MATERIALS AND METHODS
Here, we describe the study site in which this research was centered, the modeling process used in developing alternative scenarios, the methods for simulating wildfires and the processes for analyzing the simulation results.
Study Site and Management Scenarios
The study site for this research is the Upper Grand Ronde River basin,
an area of approximately 178,000 hectares (Fig. 1) located
in northeastern Oregon (USA). Most of this area is managed by the US Forest
Service (Wallowa-Whitman National Forest), but small parcels of private
forestlands are also included in the basin. Geographic and forest structure
databases were provided by the Interior Northwest Landscape Analysis System
Project (La Grande Forestry and Range Science Lab, 2003) and included
geographic information systems databases describing the vegetation (management
units), roads, streams and topography of the area, as well as tree lists
pertaining to each management unit that were used as input into a growth
and yield simulator that allowed us project forest conditions into the
future (Bettinger et al., 2005).
| Fig. 1: | Study site: Upper Grand Ronde River basin in northeastern Oregon (USA) |
| Fig. 2: | Optimized spatial patterns of treatment units |
A forest scheduling model based on a heuristic algorithm was developed to optimize an even-flow of timber harvesting volume (10,000 MBF) and four spatial patterns of fuel reduction treatments (dispersed, clumped, random and regular pattern) for this same study site. Using the scheduling model, we were able to generate management scenarios that achieved an even-flow harvest volume target of 10,000 MBF per time period, that arranged treatments in four spatial patterns with various treatment sizes (Fig. 2). The five management scenarios (four scenarios for the spatial patterns and one control) used in the previous research were adopted for this research. Two types of fuel reduction treatment activities were used in the management scenarios: (1) commercial thinning of ladder fuels and (2) commercial thinning followed by prescribed burning. The heuristic scheduling model included rules for choosing amongst the management activities and was guided by the need to generate timber volume and to create a spatial pattern of activities across the landscape.
Wildfire Growth Simulation
For wildfire simulations, we used a wildfire growth model, FARSITE
(Finney, 1998), which has been utilized in several research projects (van
Wagtendonk, 1996; Stephens, 1998; Finney, 2001; Finney, 2003; Stratton,
2004). For each wildfire simulation, spatial information on topography
and fuel conditions was required. The databases (elevation, slope and
aspect) were prepared using geographic information systems (GIS) software.
The required forest structure GIS databases that contained fuel conditions
(fuel type, canopy cover, stand height and crown base height) were prepared
for each management scenario, since fuel conditions could be influenced
by management activities. Along with spatial information of topography
and fuels, the wildfire simulation also required a set of assumptions
regarding the weather conditions. A sample set of weather conditions (temperature,
humidity, wind and moisture) for a hypothetical severe wildfire season
in eastern Oregon was utilized for these wildfire simulations.
FARSITE provides several types of outputs describing a simulated wildfire and its simulated behavior. However, we primarily focused on fireline intensity and flame length characteristics for comparing the fuel reduction treatment effects. To compare the fuel reduction treatment effects, three sets of five ignition points were used to simulate wildfires that began along the main roads in the Upper Grande Ronde River basin. These ignition points were located by the authors and represent hypothetical human-caused wildfires from sources that original along the main roads. Three different sets were developed to determine how much the results vary based on spatial variability of both the scheduled fuel reduction treatments and the ignition points themselves. Average flame length and average fireline intensity were then used in this analysis to assess the effectiveness of the fuel reduction treatments on controlling human-caused wildfires.
To specify ignition locations of these hypothetical human-caused fires, a buffer zone was generated 10 m around highways passing through the study site and then five ignition points were selected randomly within the buffer zone. These five ignition points were considered as one set of hypothetical ignition locations of human-caused wildfires and applied consistently to all management scenarios (i.e., the same five ignition points were used in the wildfire simulation for the four patterns and the control solution). Two other sets of five hypothetical ignition locations were also developed, thus wildfire simulations (5 management scenarios x 3 sets of ignitions) were developed for this analysis.
Analysis of Simulated Outputs
Output files of wildfire characteristics from FARSITE are composed
of 30x30 m grid cells and each grid cell contains a value related to the
simulated wildfires (flame length or fireline intensity). From the grid
cells burned by simulated wildfires, the average, minimum and maximum
values of flame length or fireline intensity were calculated and compared
to determine the treatment effects related to the management scenarios.
In addition, wildfire grid cells in each FARSITE output file were categorized
into wildfire behavior classes, introduced by Rothermel and Rinehart (1983).
Rothermel and Rinehart (1983) classified wildfire behavior, such as flame
length and fireline intensity, into four severity classes and provided
interpretation of wildfire behavior for each class (Table
1). The wildfire severity classes were developed for the purpose of
wildfire suppression, but they also provide a reasonable interpretation
of wildfire behavior. Among the four severity classes, classes 3 and 4
are the ones in which wildfires were considered not controllable by suppression
efforts. In this research, fuel treatment activities were assumed to be
able to reduce the areas where wildfire behavior might be classified as
severe as proposed by Rothermel and Rinehart (1983). Thus, the number
of wildfire grid cells were counted and summarized by the wildfire severity
classes for the comparison of treatment effects.
| Table 1: | Interpretation of fire behavior (adapted from Rothermel and Rinehart, 1983) |
To enrich the analysis of the wildfire simulation outputs, the GIS data from FARSITE outputs again were categorized into three groups: grid cells in treatment units, grid cells adjacent to treatment units and grid cells outside treatment units. This categorization was intended to help us understand whether overall treatment effects would vary by treatment size (testing hypothesis 1). To investigate the effect of fuel reduction treatment type, the number of wildfire grid cells affected by fuel treatment activities was summed. These results were utilized to test the hypothesis 2, that the type of fuel treatment activities influences the effects on human-caused wildfires.
In this research, treatment intensity was explained by the amount of merchantable timber harvest volume scheduled from the resulting forest plans. Thus, we calculated how much volume was scheduled for harvest from the grid cells that were assumed burned in each wildfire simulation, to examine whether treatment intensity actually had an effect on disrupting the hypothetical human-caused wildfires (hypothesis 3).
RESULTS
When simulated wildfires were applied to the five forest plans that were developed (a control and four spatial patterns of fuel reduction activities), the effects of the fuel reduction treatments on specific wildfire events are not easy to ascertain. Figure 3, for example, illustrates the flame length reported from the wildfire simulations for the simulated human-caused fires. Given the scattered nature of the fuel reduction treatments across the landscape, specific impacts are difficult to visually determine. The same could be said about the second and third sets of human-caused wildfire ignition points (Fig. 4, 5), although these are but a portion of the output and only presented to illustrate the potential fire behavior. When viewed in tabular form, the summary of the wildfire characteristics provides more insight into the differences among the fuel reduction treatments. Table 2 shows that for ignition set 1, the actual size of the wildfires that were simulated was 20,483 ha in the control and ranged from 20,471 ha when using the dispersed pattern of fuel reduction treatments to 20,815 ha when using the regular pattern of fuel reduction treatments. While the dispersed pattern of treatments resulted in a fire area slightly less than the control, the simulated wildfire when using the regular pattern of fuel reduction treatments was about 332 ha larger than the control. The flame lengths for the ignition 1 simulations were very similar among the control and four treatments, thus the treatment effects were negligible in this regard. The fireline intensity was also very similar among the control and four treatments.
When ignition set 2 was applied to the control and four spatial patterns of fuel reduction treatments, we found significant treatment effects. For example, each of the four spatial patterns of fuel reduction treatments resulted in smaller fire areas (Table 2), ranging from 136 ha smaller when using the random pattern, to 426 ha smaller when using the regular pattern. There was very little difference in average flame lengths when the control was compared to the four spatial patterns of activities. However, fireline intensity was reduced when using the regular pattern of treatments.
| Fig. 3: | Results of fire simulations: Flame lengths associated with ignition set 1 |
| Fig. 4: | Results of fire simulations: Flame lengths associated with ignition set 2 |
| Fig. 5: | Results of fire simulations: Flame lengths associated with ignition set 3 |
| Table 2: | Fire simulation results |
| Table 3: | Fire simulation results by the fire behavior class: Area (ha) in severe fire behavior classes |
When ignition set 3 was applied to the control and four spatial patterns of fuel reduction treatments, we found mixed results. Two of the four spatial patterns of fuel reduction treatments resulted in smaller fire areas (Table 2), ranging from 166 ha smaller when using the regular pattern, to 255 ha smaller when using the clumped pattern. The other two spatial patterns of activities resulted in slightly larger simulated wildfires. As with the previous two ignition sets, there was very little difference in average flame length when the control was compared to the four spatial patterns of activities. However, fireline intensity was once again reduced when using the regular pattern of treatments and was also reduced when using the dispersed pattern of treatments.
Severe wildfire behavior results also varied when each set of ignition points was considered. For example, when using ignition set 1, the area in severe wildfire classes for all spatial patterns of activity actually increased slightly over the control (Table 3), while when using ignition set 2, we found that the regular pattern of activity led to significant reductions in severe wildfire area. Ignition set 3 allowed all of the spatial patterns of activities to show reductions in severe wildfire area, although the regular pattern of activity provided the greatest reduction. These results beg the question of whether the simulated fires came into contact with the scheduled fuel reduction treatments. In fact, on average, we found that with ignition set 1, about 1 to 4% of the area burned by wildfire was either within or touching fuel reduction treatment units (Table 4). Interestingly, more area was within or adjacent to dispersed treatments than the others, yet the dispersed treatments generally did not provide the greatest effect on wildfire behavior. Areas that were prescribed burned included only 0.4 to 1.9% of the wildfire area. Here again, the area of prescribed burns was greatest in the dispersed patterns of treatments, yet the dispersed treatments generally did not provide the greatest effect on wildfire behavior.
| Table 4: | Number of fire grid cells affected by fuel treatments: Ignition set 1 |
| aRatio of grid cells affected by treatments (grid cell in treatment units or adjacent to treatment units) over total fire grid cells. bRatio of grid cells affected by prescribed burning over total fire grid cells | |
| Table 5: | Number of fire grid cells affected by fuel treatments: Ignition set 2 |
| aRatio of grid cells affected by treatments (grid cell in treatment units or adjacent to treatment units) over total fire grid cells. bRatio of grid cells affected by prescribed burning over total fire grid cells | |
| Table 6: | Number of fire grid cells affected by fuel treatments: Ignition set 3 |
| aRatio of grid cells affected by treatments (grid cell in treatment units or adjacent to treatment units) over total fire grid cells. bRatio of grid cells affected by prescribed burning over total fire grid cells | |
With ignition set 2, again about 1 to 4% of the area burned by wildfire was either within or touching fuel reduction treatment units (Table 5). As with ignition set 1, more area was within or adjacent to dispersed treatments than the others, yet the dispersed treatments generally did not provide the greatest effect on wildfire behavior. Areas that were prescribed burned included only 0.4 to 1.8% of the wildfire area. Here again, the area of prescribed burns was greatest in the dispersed patterns of treatments. With ignition set 3, about 1 to 3% of the area burned by wildfire was either within or touching fuel reduction treatment units (Table 6). As with the other sets of ignition points, more area was within or adjacent to dispersed treatments than the others. Areas that were prescribed burned included only 0.4 to 1.6% of the wildfire area. Again, the area of prescribed burns was greatest in the dispersed patterns of treatments. With these results we can say that treatment effects do not vary according to treatment size (hypothesis 1). In addition, we can safely say that the type of fuel treatment activities does not seem to influence the broad-scale effects on human-caused wildfires during a severe wildfire season (hypothesis 2), given the level of fuel reduction activity we have modeled. However, we are only suggesting this because the dispersed pattern, which had a high ratio of prescribed fire to wildfire area, was not as effective at reducing broad-scale wildfire behavior in our study. As others have shown (Helms, 1979; Martin et al., 1989; Agee, 1998; Agee and Skinner, 2005; Stephens and Moghaddas, 2005) individual fuel reduction treatments may affect the behavior of small wildfires. Thus the local-scale effectiveness of fuel reduction treatments on the behavior of wildfire is evident (e.g., US Department of Agriculture, Forest Service, 2003) and not in dispute, however the broad-scale effectiveness may be masked during severe wildfire seasons.
| Table 7: | Scheduled harvest volumes (board feet) from fire grid cells in treatment units |
| aTotal harvest volume from grid cells which is in treatment units. bAverage harvest volume from each grid cell which is in treatment units | |
The scheduled timber harvest volumes shed more light on the varied results that we experienced (Table 7). As we noted in the methods section of this paper, we attempted to schedule an even level of timber harvest volume from the broader landscape being managed. Within the areas actually burned, we found that the total harvest volume prior to simulating the wildfires was relatively high for the dispersed pattern of activity, however the average volume removed per grid cell was relatively low compared to the other treatments. In effect, the scheduling model forced treated areas to be as far apart as possible and in doing so it resorted to applying sub-optimal management treatments to stands that resulted in low volumes removed. Thus for the dispersed pattern of activity, the scheduling of activities may not have resulted in the best residual stand density from a wildfire management perspective. On the other hand, the clumped pattern of activities resulted in scheduled timber harvest volumes that were relatively high, on average, per grid cell. Here, fewer harvests were required to schedule the volume necessary to meet the volume goal. The scheduling model, in effect, was able to group higher volume stands together for harvest, to both reduce the distance between harvests and to more efficiently meet the timber harvest volume target. Therefore, the residual stand density in these areas may have been more appropriate for reducing the effects of wildfires, yet fewer areas were treated across the landscape. As a result, we suggest that treatment intensity actually had no effect on disrupting the hypothetical human-caused wildfires (hypothesis 3).
DISCUSSION
One of our main assumptions of this study is that resources are limited and preclude the use of fuel reduction treatments in a widespread manner across the landscape. In a recent report (US Government Accountability Office, 2003), it was suggested that mitigating the risk of wildfires with fuel reduction treatments will require a long-term, sustained effort. In practice, local land management offices prioritize land for treatment because operational budgets are limited. Further, budgeted resources can easily be diverted from fuel reduction treatments to activities such as fire suppression. In addition, regulatory requirements and public resistance can limit the scope of fuel reduction treatment efforts. Consequently, local field offices generally identify the highest priority locations for fuel reduction treatments and a patterns is generally not considered. However, since it is doubtful that an extensive fuel reduction treatment program can be implemented, we assume that land managers may obtain the most benefit from limited budgets by spreading treatments across the broader landscape. Unfortunately, through our study we found mixed results when simulating human-caused wildfires in areas that were scheduled for patterns of fuel reduction treatments.
The clumped and regular pattern of fuel reduction treatments seemed to reduce simulated wildfire severity most effectively in two out of three cases. However, the regular pattern suggests that activities will be laid out in a grid across the landscape. This pattern requires a measure of regularity (i.e., how far apart should the treatments be spaced?) and may be more difficult to implement than the clumped pattern of treatments, given economies of scale associated with logging equipment. Thus from an economic efficiency point of view, the clumped treatments seem more appropriate. However, since only three ignition sets were analyzed in this study, some uncertainty still lingers over the fact that the results may simply be a matter of chance. One disadvantage of our approach to analyzing this issue is that the simulations provided here represent a small sample of the potential effects of wildfires and they were very time-intensive to process and analyze. Although broader, more extensive analysis may be required to thoroughly convince policy makers of the appropriate design of fuel reduction treatment strategies, policy makers need to understand that this will involve a significant commitment of resources.
While reductions in fire area and fireline intensity were seen in several cases, only when examining the third ignition set did we find that the severe wildfire conditions were reduced. The reductions were most pronounced when using the clumped and regular patterns of fuel reduction treatments, reinforcing our belief that when resources are limited, these types of spatial patterns of activities may be more appropriate for influencing human-caused wildfire behavior. The dispersed pattern of fuel reduction activity required scheduling more area for treatment, yet since the treatments were spaced as far apart as possible, it resulted in sub-optimal choices for many stands. Therefore, in two of three cases (ignition sets 1 and 3), we found that the dispersed fuel reduction treatments had no recognizable effect on human-caused simulated wildfires. The random pattern of fuel reduction treatments produced lukewarm results in each of the three ignition set cases.
Our analysis involved assessing broad-scale simulated wildfire behavior during a severe weather season. How these results translate to mild or moderate wildfire seasons may be of interest as well, although from a broader public perception point of view, wildfires may only be a concern during severe wildfire seasons. We postulate that modeling wildfires during severe wildfire seasons may mask the effects of fuel reduction treatments. Stated another way, it may be difficult to determine the broad-scale effects of the fuel reduction treatments on wildfires initiated during severe wildfire seasons, since weather conditions are extreme, thus fuels treatments may have limited effectiveness on controlling the spread of fires. Therefore, if one were to model milder wildfire weather conditions, the effectiveness of the treatments may be more pronounced. We leave this area of study, however, for others to pursue.
CONCLUSIONS
In this research, we examined whether a spatial arrangement of fuel reduction treatments would affect hypothetical human-caused wildfires. Flame length, fireline intensity and wildfire size were measured on simulated fires to assess the severity of wildfires ignited in hypothetical human-caused locations. The resulted outputs of wildfire simulation supported the fact that human-caused wildfire could be disrupted by fuel reduction treatments during severe wildfire seasons, but the effectiveness of treatments could be influenced by treatment type and intensity applied. While we found that the clumped and regular pattern of management activity seemed to reduce wildfire severity the most in two out of three cases, although more case studies would enhance the notion that these patterns are more suitable for large-scale treatment distribution than other patterns of activities. Further, human-caused fires were assumed to occur along major roads. This hypothesis could be expanded to heavily-traveled trails or heavily-used recreational areas. While the main issue in fuel reduction treatments may be to cover the most land area as possible, or the most highly fire-susceptible area as possible, organizational budgets may preclude this from happening. Therefore a pattern of activity across the landscape has been proposed.
This research serves as a starting point for considering the use of broad-scale fuels management treatments to affect the behavior of human-caused fires. As land managers, we can not predict the exact location of where wildfires will occur. In addition, we do not have the ability (manpower or budget) to make our entire forest fire-safe. Although stand-centric analyses have shown the effectiveness of fuels reduction treatments, a broader landscape-level analysis is also necessary. Here, we have illustrated the modeling of fire behavior in a spatially-explicit manner across a large area, using realistic information regarding both the landscape and the potential management actions. The contribution of this research to science is the conclusion that either a clumped or regular pattern of activity may be the most effective in affecting human-caused wildfire behavior during a severe wildfire season. These patterns of activity can be implemented in ways that both produce timber products and control fuel levels and they may be more effective at reducing wildfire size and severity than other patterns of activity. However, from an operational efficiency standpoint, a clumped pattern of activity may be the most effective. Further, we are among the first to illustrate that the wide-spread application of fuels management treatments across a broad landscape may have limited impact on human-caused wildfire behavior during severe fire seasons, since fuels management treatments are placed irrespective of fire ignition points. This is important even though human-caused wildfire ignition points may occur in a more geographically-limited area than those caused by lightning.