2004 Starker Lecture Transcripts

The Role of Fire in Forest Restoration

James Agee
Virginia and Prentice Bloedel Professor of Forest Ecology,
University of Washington College of Forest Resources

Forests across the West are in trouble.  Wildfire area appears to be increasing year by year, and the severity of these fires appears to be outside of the historical range of variability.  Yet the gamut of solutions ranges from “heavy harvest” to “do nothing”.  Only in limited instances, and perhaps the Applegate Partnership in southern Oregon being one of these, do we seem to be able to develop consensus approaches for action.  Today, I’d like to provide a broader view of the problem and potential solutions (slide 2), recognizing the importance of priorities and place when applying these principles.  I’ll be referring to Powerpoint slides in the order of their appearance in the talk, and make references to appropriate literature along the way.

How Did We Get Here?

The fire problems we now face have a complex history.  They start with an attempt to forge a national forest management policy in the early 1900’s, and the critical role that the large fires in Idaho and Montana in 1910 played in creating that policy (Pyne 2001).  Foresters believed that European-style forest management could never be applied in America unless fire was controlled, and the fires of 1910 were the catalyst for a new fire exclusion policy.  Up into the 1920’s there were voices of dissent concerning this policy, which  advocated for the use of prescribed fire.  The case for “light burning” (slide 3) as it was then called was made primarily by industrial foresters, who were concerned that their old growth would be burned by intense fires due to fuel buildup before they could cut it.  Their pleas for short-term conservation of fire in dry forests were rejected (Agee 1993) and a century of fire exclusion resulted.  In dry forests, this turned out to be the disaster predicted by Harold Weaver in the 1940’s.  The high-canopied, low-fuel forests of the turn of the century morphed into fuel-choked forests (slides 4, 5, 6, 7) that now burn with high severity – but not everywhere did this happen, and not everywhere was high-severity fire out of character – it was an ecology of place.

In places where severe fires have replaced those more benign, Smokey Bear (slides 8, 9, 10) has been blamed for being too effective.  But fire prevention, as exemplified by Smokey, is still an important part of fire management, and is not the sole source of problems where they occur.  Pick-and-pluck selective logging (slide 11) removed the most fire-tolerant trees from the forest, and where the forests were predominately large trees, the forest was functionally clearcut.  Even low-intensity fires will have more severe effects when smaller trees, and trees of less fire-resistant species, replace large old ponderosa pines.  Grazing (slide 12) removed many of the fine herbaceous fuels that carried pre-European fires.  Over regional scale landscapes, the proportion of low-severity declined and the proportion of high-severity fire increased (slides 13, 14).

An Ecology of Place

High severity fires were always part of western landscapes, typically occurring in wetter coastal areas or forests at high elevation.  But they were uncommon in the drier forest types.  We can provide a context for this variability using the concept of the historical fire regime (Agee 1993, slide 15 15).  High severity fire regimes historically had fire return intervals exceeding 100 years, and when fires occurred they tended to be mostly stand-replacement in character.  Pacific Northwest forest types included here would be subalpine fir, mountain hemlock, Pacific silver fir, and western hemlock (both the Douglas-fir and spruce types).

The fire ecology of western hemlock/Douglas-fir forests (slide 16) is described elsewhere, but fires were typically separated by centuries.  After a fire, the growing space opened up for Douglas-fir allowed it to become a stand dominant in the next generation, and because it is long-lived, the stand dominant for hundreds of years.  Without fire, the stand would eventually become dominated by western hemlock and western redcedar, but few stands across the Pacific Northwest ever reached this stage before another stand-replacement fire occurred.  These fires were often weather-driven events such that fuels were a secondary consideration in the definition of either fire size or severity (Agee 1997).

In the Pacific Northwest, the forests with historical high-severity fire regimes are a low priority for active management to reduce fire hazard (slide 17).  Fire risk is low – many of these stands have persisted for centuries with very high fuel loads, so that short-term mitigation of hazard is not justified except in limited circumstances: other catastrophic events that excessively increase dead fuels, or adjacent to urban interface areas.

Historical mixed-severity fire regimes (slides 18, 19) present a more intermediate situation.  Drier Douglas-fir forests (westside southern Oregon), red fir, and grand fir/western larch forests fit into this category.  Fire return intervals might range from 30-100 years, with intermediate-sized patches of varying severity (slides 20, 21, 22, 23).  At a landscape scale, this provided diversity in both species composition and structure of these historical forests (slides 24, 25). This variability had significant effects on the ability of subsequent fires to spread, and helped to maintain this patchy character on the landscape.  Fuels, topography, and weather interacted to affect both fire spread and severity. Fires in these forests might have started in July and burned into October, under a wide variety of weather patterns, in a wide variety of forest patches with different structures and fuels, and across topography where it burned upslope, downslope and at night and during the day.

The case for active management in the mixed-severity fire regimes is easier to make than in the high-severity fire regimes (slide 26).  Fire risk is higher, and portions of these landscapes historically experienced low-severity fires.  The case is weaker than in the low-severity fire regimes where fire has been removed for many more “cycles”.

Low-severity fire regimes historically occurred in the warmer, drier forests where a substantial snow-free dry season existed (slide 27). These forests, usually with some ponderosa pine or pine mixed with Douglas-fir, white fir, or grand fir (slide 28), are found broadly across the western United States.  Although some of the Colorado Front Range and South Dakota pine forests appear to fit into mixed-severity fire regimes, the Southwest, California, and Pacific Northwest pine forests appear to fit the classic low-severity fire regime pattern of frequent, low-intensity surface fires (Allen et al. 2002).  It is those forests where the most dramatic shifts in fire severity have occurred (see slides 4, 5, 6, 7) and that the remainder of this lecture addresses.

Restoration of Firesafe Conditions

The principles of firesafe forests are clear (slide 29; Agee et al. 2000, Agee 2002a, Brown et al. 2004): reduce surface fuels, reduce ladder fuels (those fuels that bridge the gap between surface fuels and overstory canopy fuels), keep the large trees, and reduce crown density.  Also implied here is an order.  At the end of treatment, the most important actions are also in the same order.  Lowering surface fuels (slide 30) reduces the flame length of a potential wildfire.  Removing ladder fuels reduces the probability that a surface fire will transition to a crown fire.  Retaining large trees keeps the most fire-tolerant trees in the stand.  Reducing crown density lowers the probability that an independent crown fire will occur.

We know that prescribed fire does a pretty good job of reducing surface fuels – those are the fuels that carry the fire (slide 31), so by definition they have to decline after a burn.  But like most resource management actions, prescribed fire can be applied in many forms, under different weather conditions, as a heading, flanking, or backing fire.  One thing that is often overlooked is that the first prescribed also creates fuels by killing live vegetation that is not consumed in the first fires.  It can replace much of the original fuel load in 5 years (slide 32), although usually resulting in a much higher height to live crown.

Increasing the height to live crown reduces ladder fuel contributions that might help a surface fire transition to a crown fire (slide 33). The torching phenomenon is basically an interaction between the potential surface fire flame length, the moisture content of the understory foliage, and the height that the foliage occurs above the ground (slide 34). Two of these three variables are under managerial control.  By reducing surface fuels, potential surface fire flame length is reduced, and by increasing height to live crown by thinning understory trees, the required surface fire flame length to initiate crowning is increased.  Prescribed fire can be effective in doing both if properly scheduled (slide 35).

We’ll have an example of the value of keeping the large trees, and conversely the risk of leaving only the small ones.  The efficacy of reducing the crown density (slide 36) depends largely on a tree removal process that does both: reducing crown density while keeping the large trees.  It’s also important to remember that as thinning intensity increases, there are tradeoffs with surface fire intensity caused by drier surface fuels and increased midflame windspeeds in the thinned stands.  Often in the debates about active management we hear “Oh, we must thin the stand to save it!” but thinning comes in many forms, and only some forms will result in a firesafe forest condition.  Consider three types of classic thinning (slides 37, 38, 39, 40). A low thinning removes trees from below: the smallest ones.  A crown thinning takes a wider group of trees, and a selection thin is a thin from above: the largest ones first.  These classic graphs suffer from the exclusion of a structural component found in most current mixed-conifer stands: an unmerchantable tree layer (slide 41).

A simulation of the effect of various thinning and fuel treatment options was done using a stand much like the one shown in slide 30. It does have large trees (up to 100 cm [40 inches] in diameter), but there are also a lot of small ones (slide 42). It was assumed for this exercise that a commercial diameter limit was 15 cm (6 inches).  The thinning prescription was to reduce basal area to a threshold of 15 m² per hectare (60 ft² per acre) but in different ways.  The thinning treatments included no thin, low thin (start with smallest tree, increase tree size removed until basal area threshold is met), low thin – commercial limit (start with 15 cm tree, then increase as before until threshold is met), and selection thin (start with largest tree and move down in size until threshold is reached) (slide 43). The fuel treatment options included no treatment, and a prescribed fire with a 0.6 m (2 ft) flame length to reduce post-treatment fuels.  Then a worst-weather wildfire was simulated to burn across each stand, and survival was estimated using FOFEM (First order Fire Effects Model – Reinhardt et al. 2002).   Details can be found in Agee and Skinner (2005).  Obviously many other combinations could have been applied, and many other beginning stand structures could have been used.  But some basic principles emerge from this analysis (slide 44).

The treatments are arrayed from lowest to highest survival.  The unmanaged stand (left, slide 44) suffers a stand replacement event.  The surface fire flame length enables substantial torching in this stand.  Equally as bad was the selection thin with no fuel treatment, as the fire burned across increased surface fuels with increased fireline intensity, and only small trees were present.  The first treatment with any residual survival were the low thin-commercial limit with no fuel treatment and the selection thin that had fuel treatment.  In the former case, the unmerchantable understory combined with additional surface fuels from the thinning resulted in some torching and an intense surface fire, while in the latter case survival was minimal because all the trees in the residual stand were small.

The four options to the right had better survival.  They all had either a low thinning (one with a commercial limit) or fuel treatment by prescribed fire.  In this stand the best result was obtained with the stand was not thinned at all, but where a prescribed fire had been applied that reduced surface fuels and raised the height to live crown.  Of course, in the real world, such prescribed fires also create dead fuels, and those are not included in the simulation.

The basic principles emerging from this analysis (slide 45) are that “no action” is a disaster, thinning from above is also a disaster as it removes the most fire-tolerant trees, and low thinning is the best thinning method (from the standpoint of creating firesafe forest structures).  Prescribed fire shows up as being valuable, but in this simulation the dead fuels it creates are not included.  Treatments that reduced surface fuels, treated ladder fuels, and kept the large trees fared best.

Empirical Evidence for Firesafe Forests

The theory of firesafe forests derives primarily from research and empirical constants obtained from boreal (high latitude) forests (slide 46). Experimental crown fires there have been studied for decades.  Quantitative estimates of the relations between flame length and height to live crown, for example, or thresholds of mass flow rate (the quantity of crown fuel below which crown fire cannot operate) are largely derived from black spruce and jack pine forests.  Over the last decade evidence from the lower 48 states suggests that these principles also apply to western forests.  Five examples illustrate how these firesafe principles have been successfully applied and mitigated wildfire damage.

1.  1987 Hayfork fires (slides 47, 48, 49). These fires occurred during a massive outbreak of fires in northern California and southern Oregon.  Weatherspoon and Skinner (1995) evaluated fire severity as evidenced by crown scorch visible on post-fire aerial photography.  The forests burned in this study, mostly mixed-evergreen forests, were not specifically treated with firesafe principles in mind, but treated forests were classified as either cut-treated or cut-untreated.  Cutting was largely selective overstory removal, so cut units were implied to have average tree size smaller than uncut units.  Fuel treatment was either lop and scatter or patchy prescribed fire.  Forests experiencing the least damage were uncut, untreated forests that had the largest trees.  However, cut-treated forests did not significantly differ from uncut forests.  Fire severity in cut-untreated forests was significantly higher.

2.  Megram fire, 2002 (slides 50, 51, 52). This fire in northwestern California burned largely in fuels created after a large windsnap event in the winter of 1995-1996.  The Forest Service created limited fuelbreaks in this Douglas-fir/white fir forest.  In some fuelbreaks surface fuels, ladder fuels, and crown density were reduced, while in others only the surface and ladder fuels were treated.  One of these latter treatments is shown in slides 40 and 41. From the air and the ground, the fuelbreak edge is obvious, and even though substantial crown density was left, the fuelbreak forest, although it burned, suffered only a low severity fire compared to the untreated area.

3.  Tyee fire, 1994 (slides 53, 54, 55, 56). A large Washington wildfire burned across ponderosa pine/Douglas-forest, and created huge patches of stand replacement fire.  Areas where thinning and prescribed burning had been done fared much better than untreated areas, although scale of treatment was important.  Slide 53 shows an area where trees less than 15 cm (6 inches) had been removed within 3 years of the fire, residual trees pruned, and surface fuels piled and prescribed burned.  The crown fire approached this area, dropped to the ground with one tree length, and burned through as a surface fire, scorching about 50% of the crown volume and allowing a nearby residence to be saved.  An older, nearby narrow fuelbreak also showed better survival than untreated areas outside (slides 54-56).  The fuelbreak was created in the 1970s, and the trees in this thinned area had grown such that their average diameter was about 50% greater than in adjacent unthinned areas. Again a crown fire quickly transitioned to a surface fire upon encountering the fuelbreak, and then retransitioned to a crown fire on the far side of the fuelbreak.

4.  Cone fire, 2002 (slides 57, 58, 59, 60). This fire entered Black’s Mountain Experimental Forest in northwestern California where thinning and burning experiments had been underway for several years.  All treatments had been completed within 5 years of the wildfire.  In areas thinned and burned, the wildfire would not even spread.  A rapid transition in mortality occurred as one crossed into the boundary of treated units.

5.  B&B fire, Oregon 2002 (slides 61, 62, 63, 64). This 4-shot sequence shows an area affected by an eastern Oregon cascade wildfire.  The first shows untreated forest, the second thinned, the third prescribed fire after thinning, and the fourth (the image that appears on the slide) the post-wildfire condition, again showing a low-severity effect in the treated forest.

As a caution (slide 65) the Hayman fire of Colorado (2002) is also included.  Here, fuel treatment appeared to be effective under “normal” wildfire conditions, but treatments were not effective during exceptionally severe fire weather when the fire ran 18 miles in one day.

We appear to have good guidelines for stand-level treatment, and if the proper steps are taken, high-severity fire can be altered to low-severity fire under almost all conditions.  Fuel does make a difference in low-severity fire regimes.  Weather historically was responsible for larger spread of these fires, but severity was fuel-related.  This is still true.  Several issues remain outstanding, though.

How much of a landscape need be treated (slide 66)? This depends on assumptions about what will be done when a wildfire occurs (Finney 2001).  With aggressive fire suppression, probably 20-35% of a landscape will fragment fuels such that suppression can be effective.  If an aggressive fire suppression response is unlikely, then untreated areas of the landscape will burn severely, and treated area will burn less severely.  How much the landscape do we want to place at risk?

How long are these treatments effective (slide 67)? The answer to this depends on what is meant by “effective”.  Historical research (e.g., Heyerdahl et al. 2001, Wright and Agee 2004) shows that historic fires often stopped at the boundaries of area burned in the previous two years.  After that, spread was likely to pass over into previously burned areas.  So the effectiveness from a spread perspective is probably 5 years or less.  From a perspective of severity, the effectiveness depends on how long ladder fuels and surface fuels remain low, and how they interact with the residual tree fire tolerance in the face of wildfire.  In most cases where the first fuel treatment was effective, the answer might be 10-20 years.

Constraints on active management are many (slide 68). Some segments of society so fear active management that they apply the precautionary principle to an extreme, ignoring the fact that in dry forests the “no action” option is itself a large risk.  Species impacts, from large species such as hawks or owls to small organisms like mollusks and lichens, often constrain even “light on the land” management.  Soil impacts from any harvest, and effects of possible roadbuilding, limit the ability to treat large areas with thinning.  Some harvest techniques, like helicopter, have minimal soil impacts but require leaving much more surface fuel (tops, etc.) in the woods.  The biggest constraint for wildlands is the focus on the urban interface, as if there were no values at all to protect away from the interface.  That problem will only grow larger with time.  If we ever get serious about global climate change and carbon balances, more regulation of prescribed fire from strictly a carbon balance perspective is likely.  But how will that affect the tradeoff between wildfire carbon emitted and that emitted by practices like prescribed fire intended to reduce wildfire carbon emissions?  All of the issues involve risk management, and we do a lousy job of placing the choices for managers in a policy context.

Policy Issues

I’d like to close by moving a bit closer to two policy issues at our doorstep.  The first is the Healthy Forests Restoration Act (HFRA) of 2003 (slides 69, 70, 71, 72, 73). In my view, the Act provides the appropriate technical policy guidance for “doing the right thing” in our drier forests.  It is generally limited to drier forests (with some reasonable exceptions); it has an area limit, it directs a focus on small diameter trees, and allows both thinning and prescribed fire.  While the public involvement portions of the Act are troublesome in that they limit such involvement, my personal view as one close to public land management is that the appeal process was abused for years by some groups and needs an overhaul.  This might not be the appropriate overhaul, but I think it raises a flag that needed to be raised.  Some will cite various statistics that the appeal process has not be onerous, but those statistics do not count the plans that were scuttled or significantly reduced in scope specifically to avoid an appeal process.  That being said the effectiveness of HFRA remains to be demonstrated.  If the agencies and their administrators choose to follow the intent of the HFRA, I think it will improve forest health across the West.

The largest criticism I would make of HFRA is that it does not represent temporal scale well.  It focuses on short-term actions, and that is perhaps all we can ask of Congress these days.  I would be very surprised if Congress reopens this debate over the next four years.  But trees live along time, and our national policies should reflect that temporal scale.

Having placed myself on the frying pan with HFRA, let me throw myself into the fire by addressing timber salvage after wildfire.  Right now, we hear two voices about salvage: “we must recover maximum economic value by salvage”, and “we cannot salvage anything because of soil damage and the need for biological legacies (snags and logs) on the burned landscape”.  Both of these views are ignorant of place and incredibly myopic.  Let me take a more far-sighted view in the context of place – but I will do by ignoring the economic argument.  Is there an ecological rationale for salvage?  First, salvage is a terrible metaphor for forests (slide 74). It is reminiscent of the auto junkyard, where cars are stripped for anything of value and what is left is garbage.  Is that the  appropriate way to view forests?  Or should we think about post-fire timber removal as a means of long-term sustainability?  In this latter context, let me make a few brief statement about (aagh – I hate to even use the word) salvage.  Historical high-severity fire regimes are a low priority for salvage.  The boom and bust nature of coarse woody debris dynamics are natural in those systems (Agee 2002b) and not out of the historic range of variability.  Mixed severity fire regimes, I think, could be targeted, but primarily in those portions of that fire regime that were spatially coherent over time as low-severity fire regimes. 

Historically, coarse woody debris was limited in low-severity fire regimes.  They are warm and dry, and every summer log moisture declined to a level where if a log was encountered by fire, it was consumed (Agee 2002b).  We have emplaced a high-severity fire regime with its attendant characteristics, including spatial patch scale and high-severity coarse woody debris dynamics, into these historic low-severity fire regimes.  Without salvage, we carry an inordinate amount of coarse woody debris into the next forest generation, and significantly limit our ability to effectively manage the fuels in this next generation.  Prescribed burning in a 30-year stand with all that coarse woody debris smoldering will be difficult.  This fall I visited a site burned in 1970 and salvage logged.  My first visit there was in 1980, and it was not a very pretty site (slide 75). But on my revisit this year, the stand that emerged from that fire had been prescribed burned, and the fuel hazards were significantly lowered without significant damage to the young ponderosa pine there (slide 76). Most of the scorch seen in that slide will be gone by next year, and in its place will be a young stand resistant to wildfire.  In my view, the short-term ecological impacts of salvage have been outweighed by the long-term beneficial effects.  We still face a lot of questions: what proportion of the stand needs be left, how much of the landscape can be treated to balance the short term-long term effects, and many other issues.  My point is that salvage is not a black and white issue.  Like the ash after a wildfire, there is a lot of gray area.

In our dry forest types, we chose, for better or worse, to turn the friendly flame into a demon (slide 77). As a society, we have difficult choices about how to correct this policy nightmare that covers millions upon millions of acres across the West.  There are risks of action and no action, and people on both sides who want to do the right thing, and the wrong thing.  If the larger, broader society has a better understanding of what the “right thing” is, we will take back the forests from the advocates of the extreme, and allow fire and its surrogates to play ecologically appropriate roles in forest restoration.

References

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

Agee, J.K. 1997.  The severe weather wildfire: too hot to handle?  Northwest Science 71: 153-156.

Agee, J.K. 2002a.  The fallacy of passive management: managing for firesafe forest reserves.  Conservation Biology in Practice 3(1):18-25.

Agee, J.K. 2002b.  Fire as a coarse filter for snags and logs.  Pp. 339-368 In: Ecology and management of dead wood in western forests.  USDS Forest Service General Technical Report PSW-GTR-181.

Agee, J.K. and. Skinner, C.N. 2005.  Basic principles of forest fuel treatment.  Forest Ecology and Management (in press)

Agee, J.K., Bahro, B., Finney, M.A., Omi, P.N., Sapsis, D.B., Skinner, C.N.,  van Wagtendonk, J.W., and Weatherspoon, C.P. 2000.  The use of shaded fuelbreaks in landscape fire management.  Forest Ecology and Management 127: 55-66.

Allen, C.D., Savage, M., Falk, D.A., Suckling, K.F., Swetnam, T.W., Schulke, T., Stacey, P.B., Morgan, P., Hoffman, M., and Klingel, J.T. 2002.  Ecological restoration of southwestern ponderosa pine ecosystems: a broad perspective.  Ecological Applications 12: 1418-1433.

Brown, R.T., Agee, J.K., and Franklin, J.F. 2004.  Forest restoration and fire – principles in the context of place.  Conservation Biology 18: 903-912.

Finney, M.A.  2001.  Design of regular landscape fuel treatment patterns for modifying fire growth and behavior.  Forest Science 47: 219-228.

Heyerdahl, E.K., Brubaker, L.B., and Agee, J.K. 2001.  Spatial controls of historical fire regimes: a multiscale example from the interior West, USA.  Ecology 82: 660-678.

Pyne, S.J. 2001.  Year of the Fires: Story of the Great Fires of 1910.  Viking Press.  New York.

Reinhardt, E.D., Keane, R.E., and Brown, J.K. 2002.  First Order Fire Effects Model: FOFEM 4.0, user’s guide. USDA Forest Service Gen. Tech. Rep. INT-GTR-344. (updated to WindowsÔ version 5.0)

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

Wright, C.S., and Agee, J.K. 2004.  Fire and vegetation history in the eastern Cascade Mountains, Washington. Ecological Applications 14: 443-459.

back to the top