OVERRIDING THE COMPLEXITY OF NATURAL SYSTEMS: two examples of the revenge of nature

Control of forest fires

Forest fires are especially frequent in boreal forests and other seasonally dry ecosystems, such as prairies and savannas. In Canada, ~3 x 106 ha of forest burns each year, mainly due to natural ignition by lightning (Wein and MacLean 1983, Honer and Bickerstaff 1985). Natural forests are characterized by a mosaic of patchy, uneven-age stands and high heterogeneity in species abundance and distribution (Shugart 1984). Consequently, not all patches are equally susceptible to fires. Fires are usually limited to restricted areas and confined to the ground or understory. They are relatively modest in intensity; the high frequency of fire episodes prevents accumulation of fuel and reduces the probability of major catastrophic events. Fire suppression policies have been implemented in an attempt to reduce the frequency of fires in several U.S. National Parks. Remarkable side effects have followed the reduction in fire frequency, such as major changes in habitat structure and species composition, abundance, and spatial distribution. Some cases have been extensively documented: early successional forests in Oregon formerly had ~74 trees/ha and an average ponderosa pine (Pinus ponderosa) diameter >43 cm; after several decades of successful fire suppression, tree density has experienced a 10-fold increase (Daniel 1990), whereas average tree diameter has dropped to ~25 cm. The consequences of such changes in forest structure may be dramatic, because very high tree density implies increased vulnerability to insect pests and diseases and decreased resistance to drought (Habeck 1990). Furthermore, the ensuing accumulation of fuel in the forest greatly increases the system's susceptibility to catastrophic fire in drought years. Fire control in a mature forest eventually becomes too sensitive to monitoring errors: a small fire that is not localized and suppressed at once can rapidly spread over huge areas. Indeed, fires have occurred to an extent never before experienced: the well-publicized 1988 Yellowstone fires burned 570,000 ha, including ~50% of Yellowstone National Park.

The spruce budworm

This example relates to management of North American forest subject to periodic outbreaks of the spruce budworm (Chorisoneuma fumiferana). This important lepidopteran defoliator of conifers is responsible for tremendous damage to North American forests. Mature forest stands dominated by balsam fir (Abies balsamea) are believed to be particularly vulnerable to budworm outbreaks, but stands of white spruce (Picea glauca) and red spruce (P. rubens) may also suffer substantial damage. An outbreak may kill up to 75-90% of trees in a fir stand, whereas impacts on spruce are less catastrophic. Only small trees usually survive budworm infestation. When the epidemic episode dies out, young understory trees enter a series of successional stages, at the end of which a new mature community of fir and spruce is reestablished. This stand is again susceptible to anew outbreak that can eventually wipe out part of the forest and start a new trend of successional phases. The dynamic of the budworm-forest system thus can be viewed as a cyclic succession with long-term dynamic stability (Freedman 1995). Outbreaks of budworm have probably recurred on the landscape for thousands of years (Baskerville 1975, MacLean and Erdle 1984, Blais 1985, Freedman 1995). After the second World War, however, as a way to cope with this resource crisis, considerable effort was devoted to controlling budworm outbreaks by intensive spreading of insecticide. Initially, this policy led to higher biomass production by constraining the budworm population, limiting defoliation, and substantially reducing tree mortality. Eventually, it promoted an increasingly large biomass of susceptible tree species, mainly dominated by mature stands of balsam fir and white spruce. Since 1974, insecticide spraying has not been effective in controlling budworms; in subsequent years, an outbreak has covered an area of an extent and intensity never experienced before. On the basis of this experience, Clark et al. (1979) have devised an instructive lesson for ecologically sound policy design that enables resource managers to account explicitly for natural variability, spatial heterogeneity, and nonlinear causation due to the combination of the multiscale, dynamical mechanisms of the exploited ecosystem.