Megram Blowdown and Wildfire

INTRODUCTION

This case study uses a recent real world example to illustrate how the CRAFT process can help improve decision making. The management issues that we highlight are characteristic of fire and fuel issues that occur across the west. In this example, we demonstrate how fire risks and vegetation and habitat change can be characterized across broad spatial and long temporal scales.

The Megram example includes interactions between wide-ranging resource objectives, two interactive natural disturbances, and decision alternatives that need to weigh the effects of both active and passive management. Specifically, managers desire to implement a shaded fuel break included not just an expectation of reducing fire risk to the Wildland Urban Interface, but the risk of negatively affecting a Late Successional Reserve set aside for the northern spotted owl and associated wildlife and increased sediment delivery to streams that might affect salmonid habitat. The land allocation, wildlife habitat and landscape scale fire spread issues that surround this example are similar to those that pose problems across the West.

The purpose of this example is not to revisit the decisions that were made by the forest during the mid and late 1990s. Rather, we simply draw from this historical management issue and new data and models to demonstrate how CRAFT can support effective decision making related to landscape scale questions.

Back to Top

BACKGROUND

In December of 1995, severe windstorms caused an extensive blowdown in the mixed conifer forests in the Klamath Mountains of northern California. The increase in fuel loading from 5-50 tons per acre to 100-300 tons per acre heightened concerns of severe stand-replacing fire among managers of the Six Rivers National Forest. Such high severity fires were thought to be inconsistent with historical fire regimes and an undesirable outcome for nearby communities and high-value salmonid and spotted owl habitat. The proposed salvage was prevented following litigation because a court ruled that Forest managers had failed to adequately search for the Del Norte salamander, as was required by the Northwest Forest Plan. Although wildlife concerns had been incorporated into fire and fuel management planning, the legal requirements surrounding this species prevented active fuel and habitat management.

In August of 1999, lightning ignited multiple fires in the Trinity Alps Wilderness Area in the Klamath Mountains of northern California. Within a month, the fire had burned into the blowdown area. By the time of the wildfire, some salvage logging of the blowdown had reduced fuel near roads and as a result, fire severity was lower at those sites. When the fire was finally extinguished by November rains, the Megram fire and adjacent Onion fire (the Big Bar Complex) covered 560 km² on the Six Rivers and Shasta-Trinity National Forests. At one point, smoke had reduced air quality to where the governor declared a state of emergency for a three county area, and many residents of the nearby Hoopa Valley Indian Reservation and nearby communities were forced to leave.

The effects of the fire on vegetation and wildlife were complex. Nearly a third of the “old growth” forests had over 80% tree mortality. Following the wildfire, the Six Rivers National Forest sought to further manage fuels and reduce future wildfire risk to communities and ecosystems, but was prevented by additional litigation. In 2002, a District Court found that the Forest’s Environmental Impact Statement for the post-fire salvage plan had failed to address the cumulative effects of disturbance on wildlife; in particular, the Forest Service had not assessed the risk posed by increased soil erosion from salvage treatments and how that would affect the recovery of salmonid habitat. Once again, proposed fire, fuel and habitat management in the area was thwarted because of the complex ecological and social issues involved.

Back to Top

SPECIFYING OBJECTIVES

Public interest in the Megram area includes a wide range of objectives. These are often indicated by public comments and concerns expressed in NEPA documents, including the land management plan and others conducted more recently. These comments can be categorized in terms of the values that they reflect and combined with the specific objectives that the agency is required to value due to legislation and policy. Based on NEPA documents that relate to the Megram area, the following are all potential project objectives or they form the "raw material" for objectives. Take note that these vary widely in their level of specificity. This varying level of generality will help in the construction of an objectives hierarchy for the project area.

Selected Six Rivers National Forest Land Management Plan Selected Goals

Special Habitat Goal: Provide mature and old-growth habitat for plants and animals associated with mature and old-growth forests.

General Forest Goal: Provide multiple use development opportunities and a sustained yield of timber in a manner which preserves ecosystem function, biodiversity and landscape integrity.

Recent Fuel Treatment Project—Objectives from the Record of Decision

Protect remaining mature and old-growth stands from catastrophic loss, accelerate development of late-successional habitat, reduce fuel levels in strategic locations, and create stand conditions that would lower the potential for future catastrophic fire.

Reduce long-term impacts to soil productivity brought about by frequent high severity wildland fire.

Modify areas with high concentrations of standing dead and down trees or brush to a more open condition within strategically located fuel breaks and fuel treatment areas.

Provide for community protection from future wildfires and extended exposure to smoke.

Examples of public comments from a recent fuel treatment project Comments such as these may suggest lower level objectives for an objectives hierarchy.

Make firewood available to the public
Emphasize protection of houses and communities with firebreaks
Treat areas immediately next to communities, not father away
Salvage log before insect damage occurs
Maintain recreational access
Limit activity disturbance to northern spotted owls
Plan for future Del Norte salamander habitat in high severity areas
Protect goshawk sites near project areas
Limit planted tree density to minimize the need for future thinning treatments
Prevent damage to power lines from prescribed fire
Treat the forests to limit future risk to communities
Meet water quality standards and laws
Limit additional sediment in streams and restore the watershed
Minimize road construction
Do not salvage log or construct fuel breaks
Better describe the basis for locating fuel breaks in landscape
Consider jobs resulting from different treatment alternatives
Perform site-specific analyses for wildlife, vegetation and soil effects of treatments
Plan how to maintain fuel breaks over time
Prioritize understory treatments; retain overstory
Design fuel breaks for fire reintroduction, not just suppression
Maximize revenue from salvage
Maintain wildlife corridors in areas treated
Maintain sufficient number of snags and CWD cover for wildlife
Limit the effects of treatment on short term reburn potential
Protect soils from burning effects & compaction with post-fire treatment strategies
Limit tree removal in riparian reserves
Avoid any new road construction whatsoever
Limit sediment input from reopened roads
Maintain water quality for Mill Creek water users
Analyze successional changes in fuel breaks with and without maintenance

Due to the diverse range NEPA public comments, it is likely that they will occasionally contradict not only each other and even be inconsistent with the objectives of the forest plan. However, these may help provide "fodder" for building an objectives hierarchy, as they include both general and specific objectives. Note that some of these comments address are less clearly “objectives” as they are comments on proposed alternatives; that is, their objective is obscure. For example, “do not salvage log or construct fuel breaks” and “Avoid any new road construction whatsoever” support one of the proposed alternatives (i.e., no action), and the person’s values or objectives are unclear.

The general and specific objectives above were useful for creation of an objectives hierarchy. Others objectives were added that were not explicitly stated in these sources, but they become apparent as general and specific objectives are organized. The following is an example of an objectives hierarchy for the project area. Note that these are not "project objectives," but "place objectives." This distinction is critical the project should be designed, based on the objectives for the location.

An Objectives Hierarchy for sustaining forests and communities in the Megram area:

Back to Top

DESIGNING ALTERNATIVES

The proposed project is in a forest/habitat reserve that lies between an extensive wilderness area and the the wildland urban interface. This particular project was proposed within the context of a much larger forest management strategy. Following the 1999 wildfire, a strategic leadership team identified priorities for long-term fire and fuel management. The following is an abbreviated version of those priorities:

(1) Construct primary fuelbreaks by reducing fuel levels in high and moderate severity burned areas as well as unburned areas.

(2) Develop memorandum of understanding with the nearby Tribe for reducing fuel levels in unburned fuel treatment areas through large area burning.

(3) Reduce fuel levels in moderate and high severity burned fuel treatment areas.

(4) Continue with a maintenance-burning program over time.

The alternatives to be considered involve different types of treatment within the fuel treatment zone. These are typical of many project in the west and consist of the No Action alternative, a Burn Only treatment, a Thin and Pile-Burn treatment, and a more aggressive (and costly) Thin and Burn treatment.

In the actual post-fire EIS for the Megram fire, differences among alternatives consisted of tractor vs. helicopter logging and different intensities of thinning and salvage logging in different units within the forest reserve. In our example, we used alternatives that show greater differences in outcomes for illustrative purposes.

Back to Top

MODELING EFFECTS

The Megram example includes a wide range of objectives, and these are not easily included within a single belief network model. In particular, the aquatic and terrestrial issues may be more easily separated into separate models, even though fire behavior integrations can be used for both. The following influence diagram includes the terrestrial objectives in green. Future vegetation and habitat characteristics in the forest reserve are affected by both the proposed treatments and wildfire.

In that the question at hand is largely one of fire spread across a broad landscape, the model used here is based on that of the Spatial Movement Template. Multiple runs of fire spread software, such as FARSITE that incorporate a range of fire weather conditions will inform the effects of proposed treatments on the objectives.

Conversion from this conceptual model to a full-fledged belief network requires restructuring temporal components and decisions regarding which variables to use in the belief network. An illustration of how the diagram above can be developed into a belief network is shown below.

See this network in action: Megram.dne (Requires Netica software)

The following section describes this belief network in detail.

Unit of analysis

A single fire ignition event that may or may not be able to burn a forest reserve and/or a community.

Model logic

Proposed treatment alternatives within the forest reserve (shown in the blue decision node) may or may not alter wildland fire enough to (1) alter fire effects within the reserve and (2) affect fire spread into the WUI. In the utility node, four issues in the the costs of treatment and fire suppression are compared with fire effects in the reserve and the risk of altered WUI fire.

Supportive models and tools

Fire spread software (e.g., FARSITE); FireFamilyPlus for organization of fire weather data; GIS data for organizing vegetation and fuel data; Other data on the expected distribution of variables (e.g., % of homes with fire-safe landscaping, treatment costs across vegetation types, suppression costs, etc.)

Model discussion

The fire weather used in this belief network includes the conditions that follow historical ignitions during the fire season. Weather data is derived from RAWS station data after being manipulated in FireFamilyPlus and a spreadsheet. Streaming weather data is required for fire spread modeling (e.g., FARSITE) to determine the conditions under which wildfire arrives in the reserve and/or WUI. In this example, weather scenarios are sorted according by fire season severity to better understand the year-to-year risks associated with wildland fire use policies.

Ignitions are randomly selected from an ignition probability map derived from historical data. Although not shown in this example, lightning and human ignitions could be modeled separately for comparative purposes. Including both is important because human ignitions tend to be along roads while lightning ignitions are more often on less accessible ridgetops. The location of these ignitions can affect fire spread. In addition, the fire weather that follows lightning ignitions is unique because they start from frontal activity. Human ignitions are not similarly constrained.

Reserve flamelengths vary according to fuel conditions that are altered by treatment alternatives. This example uses custom fuel models that better distinguish fuel conditions in different vegetation types than do the basic fuel models. Fuel models are varied in GIS to be used as input layers for FARSITE runs. Grouping of fire weather scenarios this way allows managers to address season-related contingencies of fire effects and suppression costs.

Whether or not wildland fires are actually suppressed is difficult to model, as it depends on a host of stochastic variables including the number of concurrent fires, the available personnel, the equipment available and used, access limitations related to topography and land allocation, safety concerns, fuel conditions, and variability in fire weather. This probability could easily be modeled in a belief network of its own. To calculate suppression costs, the costs of fighting fire in the reserves and WUI are given comparable weight for calculation of conditional probability table values. WUI suppression is always classified as "high," but suppression costs in the Reserve are proportional to flamelength.

Given the 15 year time frame, no successional change due to in-growth is modeled in this example. This could be added to to address longer changes in vegetation due to growth by adding transitional probabilities. The vegetation classes used here were not defined with seral vegetation in mind, though. Instead, vegetation change is affected only by active treatment and wildfire.

Utility function

The utility function in this belief network allows comparison of suppression and treatment costs with houses burned and future vegetation conditions. Categories within each of these four nodes were ranked 1 to 5; then the three social values were grouped and evenly weighted with future forest conditions.

Adding components

Other endpoints needed for the effects analysis can be added to this belief network structure For example, the forest types shown provide different habitat quality for wildlife. Using wildlife habitat relations models, the relative value of different vegetation classes for a species can be modeled as shown below. In a modification of the prior example, utility values are only affected by future habitat and treatment costs. Habitat is arbitrarily weighted twice that of treatment costs in the utility node shown here.

See this network in action: MegramHabitat.dne (Requires Netica software)

Back to Top

SYNTHESIS

Using the values in the utility function, summary tables can be prepared to compare the sensitivity of outcomes on nodes of interest. The following summary table is one of many different summary tables that are possible given the complexity of the Megram.dne belief network. Values in bold have relatively high utility values.

Assumptions	No Action	Burn	Thin	Thin and Burn
WHEN ALL RESERVE VEGETATION IS CONSIDERED
Wilderness fires suppressed
Yes	.78	.66	.69	.78
No	.71	.62	.64	.75
Fire safe landscaping in WUI
Yes	.75	.65	.67	.78
No	.74	.64	.66	.76
Fire season
Above average	.69	.61	.63	.73
Below average	.79	.68	.70	.80
WHEN JUST PLANTATIONS ARE CONSIDERED
Wilderness fires suppressed
Yes	.68	.61	.74	.79
No	.64	.57	.68	.75
Fire safe landscaping in WUI
Yes	.66	.60	.72	.78
No	.66	.59	.71	.77
Fire season
Above average	.63	.56	.67	.73
Below average	.69	.63	.76	.81
WHEN JUST RESTORED FOREST IS CONSIDERED
Wilderness fires suppressed
Yes	.97	.90	.84	.88
No	.94	.87	.82	.86
Fire safe landscaping in WUI
Yes	.97	.89	.84	.88
No	.95	.88	.82	.87
Fire season
Above average	.92	.84	.79	.83
Below average	.99	.92	.86	.91

For the overall vegetation in the reserve, the Thin and Burn option is desirable under most circumstances. However, there are situations when No Action has a similar utility value. For example, if Wilderness fires were entirely suppressed, the No Action alternative is comparable to Thin and Burn. Similarly, No Action could be justified if future fire seasons were below average. In other words, No Action may be warranted when wildfires are less of a concern in the landscape.

A key concern in the objectives hierarchy is to maintain the viability of plantations. Based on the table shown above, the values for plantations indicate that plantations are vulnerable to fire regardless of the fire season or Wilderness fire management. With every scenario, thinning and burning provides the highest utility values.

Finally, the utility values for the areas that are already restored indicate that it is not worthwhile to perform any active management at this time. Remember, though, that this belief network does not have a successional component that would be affected by maintenance treatments. With successional change over time, the value of thinning, burning or thinning and burning would increase.

The following table provides a simple summary of how utility values change as one important independent driver node is varied. In this case, we consider the value of different treatment alternatives based on where wildfire ignitions occur on the landscape. With modifications, results such as these, could be used to describe the effectiveness of different proposed treatments on protecting the WUI from fire.

Ignition Distance to Reserve	No Action	Burn	Thin	Thin and Burn
Less than 1 mile	.69	.62	.63	.74
From 1 to 2 miles	.70	.62	.63	.75
From 2 to 4 miles	.74	.64	.66	.77
Greater than 4 miles	.76	.66	.68	.78
Overall ignitions	.74	.64	.66	.76

These results show that while the thin and burn alternative is highest regardless of where wildfires start, its relative benefits decrease with increasing ignition distance from the Reserve. Based simply on this single factor (ignition distance) and the overall pattern of likely ignitions, it is difficult to justify the thin and burn treatment over the no action alternative.

Similar summary tables can be constructed for any node of interest. These can be used to summarize results for both the decision maker and stakeholders.

Back to Top