Forest response to catastrophic wind disturbance

Forests are dynamic entities on a variety of time scales. Many studies have examined the processes of colonization and regrowth following small-gap formation, which is common in many (perhaps most) forests worldwide. Some forests experience extreme, stand-replacing fires, which have also been the subject of much research. Wind disturbances and forest recovery after wind disturbances has received much less attention, yet wind is a very common source of disturbance. Moreover, the intermediate severity of wind damage may influence forest dynamics in ways distinct from low-severity gaps or insect defoliations, or high-severity fires, landslides, or volcanic eruptions. I have studied the damage and vegetation regeneration of several forests following tornado blowdown, in an effort to understand the consequences of disturbances that are intermediate in size and severity. Tornadoes are a major agent of disturbance to forests in the continental interior of North America (and are documented to a lesser extent on all continents). An average of nearly 1000 tornadoes per year occur in the U.S., and although nearly half are very small and weak, a small number do considerable damage to forests. Tornado intensity is rated on a six-level scale, from F0 to F5. F0 tornadoes are the weakest and most common; F5 tornadoes are the strongest and quite uncommon. There is a rough increase in tornado size and path length with increasing intensity, so the most powerful tornadoes tend to also be the largest and have the longest damage paths.

Details about current and former study sites:

• Tionesta Scenic and Research Natural Areas, Pennsylvania.
• Boundary Waters Canoe Area Wilderness, Minnesota.
• Blooming Grove Hunting and Fishing Club, Pike County, Pennsylvania.
• Cathedral Pines forest preserve, Connecticut.
• Mingo National Wildlife Refuge.
• Reviews and synthesis.

Tionesta Scenic and Research Natural Areas

The Tionesta forest is a 4100-acre stand of primary hemlock-northern hardwoods in northwestern Pennsylvania. It is centrally located within the Allegheny National Forest. It was struck by a tornado in May 1985, which destroyed the canopy across nearly 400 hectares of the old-growth preserve; this and another tornado damaged nearly 5000 acres in Allegheny National Forest overall. I studied damage patterns and reported them in Peterson & Pickett 1991 (abstract can be found here). Another, smaller tornado struck the Tionesta forest in 1994, about 2 km south of the path of the 1985 event. I have examined the damage and recovery of the 1994 disturbance to explicitly compare it to what I found after the 1985 event. The results of that research can be found in Forest Ecology & Management in September 2000; the abstract can be viewed here.

This image shows the view at Tionesta during the summer of 1986, one year after the tornado struck. Note the nearly complete canopy destruction, and lack of sprouting or regrowth of the damaged trees. This very low level of survival is in sharp contrast to findings from studies of hurricane damage.

I have used seedling tagging and monitoring, seed bank germination tests, and seed and seedling introduction and fencing to investigate patterns and causes of regeneration. One pattern that was obvious from the first few seasons of growth was the differences in revegetation among soil microsites created by tree uprooting, which is described in Peterson et al. 1990 (abstract here) and Peterson & Pickett 1990 (abstract here). The differences among microsites in terms of plant community composition are also notable in intact old-growth forest, as demonstrated in Peterson & Campbell 1993 (abstract here). However, several of the patterns of variation among microsites in intact forest are the opposite to that observed in the blowdown, suggesting that as microsites age, their relative suitability for plant establishment and growth shift.

Interestingly, regeneration at Tionesta differs from that observed in clearcuts of second-growth forests, being dominated by mid-successional Yellow Birch (Betula alleghaniensis) and late-successional Beech (Fagus grandifolia) and Striped Maple (Acer pensylvanicum). This contrasts with predictions of the existing conceptual models of forest regeneration. Overall dynamics of forest regeneration are presented in our 1995 Ecology paper (abstract here). Regeneration is influenced by the types and abundances of different patches (e.g. patches of surviving small beech seedlings; patches of thick leaf litter; patches of unvegetated disturbed soil; etc.) that were created at the time of the disturbance. I have found that seed germination, seedling growth, and intensity of mammalian herbivory all differ significantly among patch types. An earlier greenhouse study examined germination of birch and sumac seeds under various litter treatments, and is reported in Peterson & Facelli 1992 (abstract here).

A very intriguing finding has led me to start to think more about the interaction of disturbance with herbivore activity, in determining patterns of forest regeneration. Colleagues Zac Long, Walt Carson and I discovered that the presence of treefall pits and mounds from uprooted trees creates refugia from deer browsing for small hemlock (Tsuga canadensis) seedlings! It appears that although growing on top of treefall mounds is not favorable for hemlock seedlings in terms of growth rate, presumably because of moisture stress, it is a location where deer cannot reach the seedlings. Thus by establishing in a suboptimal location, these seedlings actually have much greater chance of survival than their compatriots that establish on level ground. While not the only time such a phenomenon has been documented, this finding suggests numerous fascinating possibilities for how disturbance and herbivory may interact in a more general sense. You can read the article in Journal of the Torrey Botanical Society, 1998, vol 125, pp. 165-168. The abstract is here. More recently, my graduate student Lisa Krueger has asked whether the effect documented in 1996 and 1997 was still having an effect in 2003. She re-sampled at Tionesta, and found that indeed, the "mound refuge" effect was still operating. Her findings are published in the October 2006 issue of American Midland Naturalist.

This figure shows how survival of transplanted red maple seedlings varied as a function of type of patch and exposure to mammalian herbivores. Note that survival of fenced seedlings was always higher than that of those open to herbivory. Seedling survival varied dramatically among patch types, with unprotected seedlings in Open and Fagus patch types suffering the greatest mortality. The effects of patch identity on germination, growth, and intensity of herbivory are presented in the journal Oikos; the abstract can be seen here.

Cathedral Pines preserve, western Connecticut.

Cathedral Pines is a small (25 hectare) stand of hemlock and white pines surrounded by agricultural land in northwestern Connecticut, and owned by the Connecticut chapter of The Nature Conservancy. Approximately 20 ha of the stand was destroyed by a tornado in June 1989. In 1990 and 1991, I investigated the extent and nature of damage, soil seed bank, and initial regeneration, and then resampled woody vegetation in late 1995 and in August 1999. I found that proportions of trees toppled by uprooting and trunk breakage were very similar (74% and 23%, respectively) to those at Tionesta. The soil seed bank at Cathedral Pines was substantially richer and more abundant than at Tionesta, perhaps due to the small size of the stand and resultant short dispersal distances from sources in the predominantly agricultural landscape. Tree seedling densities and composition were different from those observed at Tionesta, with fewer seedlings at Cathedral Pines, and a lesser component of beech. Several aspects of the damage and recovery at Cathedral Pines differ from patterns found in other windthrow studies, but are very similar to that observed in the old-growth Pisgah forest in southwestern New Hampshire after the 1938 hurricane. I conclude that much of the atypical damage and recovery at Cathedral Pines can be attributed to a) dominance by hemlock and white pine, which do not sprout when damaged; b) the large average tree size, thin soils, and exposed location, which probably contributed to the extreme severity of damage; and c) the landscape setting in which nearby sources can maintain a supply of propagules of pioneer and ruderal species.

Mingo National Wildlife Refuge, southeastern Missouri.

Mingo National Wildlife Refuge in southeastern Missouri was hit by a small tornado in June 1993. In August 1994, I studied the damage and initial vegetative recovery in collaboration with Dr. Alan Rebertus of Northern Michigan University. In agreement with most other wind damage studies, the largest trees were most likely to be damaged, although structural damage and mortality were not identical among sites, species, or sizes.

Here the predicted probability of damage is shown as a function of tree size (diameter at 1.4 m), based on a logistic regression of actual observations from the Mingo blowdown study. Clearly, the likelihood of damage increases greatly with slight size increases for small-to-medium sized trees. This type of relationship between tree size and probability of damage is likely to exist for most catastrophic wind disturbances, but the shape and position of the regression line will change for storms of differing intensity: slower winds will probably shift the curve to the right and lower the maximal risk of damage; faster winds will probably shift the curve to the left.

We studied three adjacent forests that differed in drainage but not topography. The differences in damage and recovery appear to be adequately explained by variation in tree sizes and forest species composition. This shows that independent of the strong topographic influences observed in other studies, species composition and tree size may be enough to cause differences in damage among sites. In this secondary forest with a long history of human effects, the response of the herbaceous species was much greater than at Tionesta, with composition being dominated by ruderal, shade-intolerant forbs. These findings were published in 1997 in the Journal of Vegetation Science.

The patterns I observed at these and other blowdown sites prompted me to begin work on a conceptual model of regeneration. This model, developed in collaboration with Dr. Walt Carson of the University of Pittsburgh, predicts the abundance of seeds of pioneer species (e.g. Pin cherry, Prunus pensylvanica) in the soil seed bank as a result of forest age and size at the time of disturbance.

This figure shows the predicted abundances of seed-bank seeds and living stems of a pioneer seed-bank species, such as pin cherry (Prunus pensylvanica). Note that stem density reaches high density immediately after a disturbance, and then declines rapidly when the pioneer species reaches its maximum age. Seeds decrease in abundance initially, as they are depleted by germination, but steadily increase in abundance until death of living stems eliminates the local input. The model assumes an approximate maximum duration of dormant seeds of roughly 75 years, after which they steadily die. The result is an important drop in seed bank seed density roughly 80 to 120 years after a disturbance, which in turn would greatly affect the regeneration that would occur if another disturbance occurs after this period.

Further developments will incorporate the effects of disturbance return time and size of area disturbed. This model and its implications are discussed in Peterson & Carson 1996 (abstract here). In relation to this model and the empirical studies described above, I am also interested in the significance of differential seed dispersal and germination to the organization of plant communities.

I am particularly pleased to have had the opportunity to share in the synthesis and exchange of ideas with a number of other workers who study disturbances, during a workshop in 1996 at the National Center for Ecological Analysis and Synthesis in Santa Barbara, CA. One of the results of that workshop was a collection of papers that was published in the journal, Ecosystems. The participants divided into several small groups by topic area, and created synthesis papers that attempt to both give an overview of the various topics related to disturbance, and to outline especially promising areas of research. I was pleased to work with Monica Turner, Bill Baker, and Robert Peet; our paper can be found in Ecosystems, volume 1, pages 511- 523. The abstract can be seen here. An outgrowth of these interactions was the development of a white paper through the National Assessment of Climate Change, which examined how forest disturbances in North America are likely to change in coming decades and centuries, as a result of climate change. This white paper has been distilled into a manuscript for the journal BioScience. Dr. Virginia Dale was the project leader for this effort; the introductory paragraphs can be seen here.

Wind-disturbance manuscripts currently underway

Several more wind-disturbance manuscripts are in the pipeline.

Another windthrow manuscript developed as a follow-up to my participation in an IUFRO- sponsored workshop on wind disturbance in forests, held in Joensuu, Finland in 1998. In this manuscript, I compare my older findings from the large 1985 Tionesta windthrow, to parallel findings from a smaller windthrow caused in 1994 by a less powerful tornado that struck the Tionesta forest preserve about 2 km south of the path of the 1985 storm. This was a very unusual opportunity to examine the effects of two infrequent events in the same forest, thus effectively eliminating confounding influences of different climate, soils, vegetation type, or land use history. The damage from the 1994 storm was less severe, and probably as a consequence, the regeneration was dominated by released understory Fagus saplings, rather than the many new Betula seedlings seen in the older & larger blowdown. This manuscript was mentioned above, and appeared in 1999 in Forest Ecology and Management; the abstract is here.

This figure shows differences in the risk of treefall as a function of size for five different blowdowns that I have studied. For this analysis, all species were pooled, so site differences are confounded with species differences. Note that different intensities of storms lead to great differences in risk of treefall: the Mingo and Tionesta 1994 events were rated F2, while Cathedral Pines was F3, and Tionesta 1985 and Gibbs Farm were F4 events. The 1985 and 1994 Tionesta events occurred on very similar sites and in similar forest types, thus illustrating differences almost entirely attributable to characteristics of the storm events themselves.

Reviews and Synthesis