Chapter 1FOREST DAMAGE AND RECOVERY FROM CATASTROPHIC WIND
"The oldest voice in the world is the wind. When you see it fitfully turning the blades of a mill lazily to draw water, you think of it as an unreliable servant of man. But in truth it is one of our masters, obedient only to the lord sun and the whirling of the great globe itself." Donald Culross PeattieINTRODUCTION
Catastrophic winds, including cyclonic storms, tornadoes, gales, and severe windstorms, impact forests around the globe in both temperate and tropical regions (Figure 10). Tree species differences, tree age or size, silvicultural treatment, topography, edaphic characteristics, and previous disturbance may all influence the spatial patterns and severity of damage and recovery from catastrophic wind. Although the influence of these factors is increasingly understood for chronic wind effects, particularly in forest plantations, there is little evidence that these relationships hold for less frequent catastrophic winds. Catastrophic winds have been observed and quantified for centuries, but few clear generalizations have been drawn about their impacts. Improved understanding of these impacts is required, as the need for conservation and ecosystem management increases, while at the same time both the frequency and intensity of these types of storm may increase, according to some meteorologists and global climate change models (Dunn & Miller 1960, Wendland 1977, Emanuel 1987, Gray 1990). In this review I attempt to answer seven questions: 1) How should catastrophic wind intensity be quantified? 2) How should catastrophic wind effects be quantified? 3) Is catastrophic wind damage homogeneous, or are there spatial patterns of damage, and can these patterns be predicted? 4) How do biotic factors (e.g. tree size, species, and stand differences) influence damage? 5) How do abiotic factors (e.g. topography, soil, and previous disturbance) influence damage? 6) Can any generalizations be drawn regarding the dynamics of recovery from catastrophic winds? 7) Are there consistent differences between tropical and temperate forests in terms of wind disturbance and response? For the purposes of this review, "catastrophic" refers to a disturbance that is intense and large scale. Not, as Harper (1977) used the term; a rare event unlikely to be repeated as a selective force. Also not necessarily having a detrimental impact, as many of the papers reviewed indicate. Rather, as Meffe and Carroll (1994) state, catastrophes are short in duration but wide spread in impact. Figure 10 - Major tracks of cyclonic storms Nine previous papers compared impacts of different storms on different forests (Smith 1946, Andersen 1954, Lugo et al. 1983, Liegel 1984, Glitzenstein & Harcombe 1988, Brokaw & Walker 1991, Peterson & Pickett 1991, Tanner et al. 1991, Everham 1995). I shall briefly summarize their findings before presenting a more extensive review. In addition, Biotropica and Journal of Coastal Research both had special issues in 1991 dealing with hurricane impacts, BioScience had a special section on Hurricane Andrew (Pimm et al. 1994), and Grace and Coutts (1995) published a collection of papers on the impacts of wind on forests. Liegel (1984) reported assessments of hurricane impacts on 11 conifer plantations. He found: 1) more damage at higher elevations, 2) an increase in some types of wind damage with increasing annual rainfall, 3) more damage in thinned than unthinned stands, 4) more uprooted than snapped stems, and 5) no clear differences between damage to conifers and broadleaf trees. He stressed the need to assess hurricane damage in order to develop models for predicting damage risk. Glitzenstein and Harcombe (1988) compared the impacts of a tornado, a large blowdown associated with a severe thunderstorm (Dunn et al. 1983), and a hurricane (Lugo et al. 1983) on six temperate and tropical broadleaf forests. They found that: 1) damage was positively correlated with average stem size in a stand, 2) the size of the area affected and the severity of disturbance were comparable for all three catastrophic wind events, 3) mortality tends to be low, 4) trees severely damaged often recover, and 5) the effects of catastrophic wind disturbance probably never approach the severity of catastrophic fire. Peterson and Pickett (1991) compared 16 studies of wind disturbance and identified a trend toward more uprooting under wetter conditions. Smith (1946) described the different types of wind disturbance that affect New England. He found that each type has characteristic wind directions that should result in predictable patterns of protected and damaged patches on the landscape. He also reported that: 1) local topography influences wind flow, 2) edaphic characteristics control root growth and therefore susceptibility to wind damage, and 3) silvicultural treatment (mixed-age management and thinning) affects wind resistance. Andersen (1954) reviewed the history of gale damage to Scotland (16 storms) and the literature on chronic and acute wind stress. He concluded: 1) topography influences wind and damage, 2) damage increases with tree height, 3) species differ in response, 4) rooting depth influences resistance, and 5) the juxtaposition of different silvicultural treatments can control damage. Brokaw and Walker (1991) presented a summary of damage reported for hurricanes impacting 14 tropical and temperate conifer and broadleaf forests. They found: 1) amount of stem breakage as opposed to uprooting varied among forests, sometimes correlated with stem size; 2) in some cases there is a bimodal distribution of damage versus stem diameter, large trees being more likely to be directly damaged by the wind, and small trees more likely to be indirectly damaged; 3) previous mechanical damage to trees may increase subsequent wind damage; 4) no clear relationship exists between damage of conifers versus broadleaf plants, or monocots versus dicots; 5) species differ in impacts, but no clear generalizations exist such as correlations with successional class; 6) there is no clear correlation between topographic exposure and severity of damage, and 7) although structural damage is high, mortality is generally low. Lugo et al. (1983) compared hurricanes to other natural disturbances. They proposed that hurricanes have a greater impact than either earthquakes or landslides, both because of higher frequencies and larger areas affected. Yet, they also predicted faster recovery from hurricane effects, because of adaptations to these frequent disturbances. Tanner et al. (1991), in a summary of impacts of hurricanes on forest ecosystems in the Caribbean, stated that the effects on ecosystems depend on: 1) hurricane intensity (including rainfall), 2) storm size and movement, 3) topography, and 4) susceptibility of the system to damage. An overall conclusion from reviewing these papers is that generalizations are lacking. A barrier to comparing the impacts of wind events is the lack of standard measures for storm intensity (Lugo et al. 1983). Another impediment to generalizations concerning impacts of catastrophic winds is a failure to standardize measures of impacts. Ackerman et al. (1991) reported the results of a conference on ecological effects of hurricanes and mention measures of damage gradients as one missing research tool. Thus I review both results and methodological issues, and I recommend methods that I feel will provide comparative results. In particular, I further develop the suggestion made in Everham (1995) that quantifying wind disturbance and expressing damage as both percent mortality and percent basal area damaged will facilitate comparisons among systems.LITERATURE REVIEW
I reviewed 119 papers reporting the impacts of 100 storms in 20 countries and in 21 states in the USA. Appendix F - I is a summary of the storms, locations, and references; with indications of each study's attempts to quantify the storm intensity or damage. The table is organized chronologically by storm event. In several cases, more than one paper reports on the same storm in the same forest. All were included whenever there are differences in plot location, sampling methodology, or analysis. For this review I have converted all measurements to metric equivalents. Storm Intensity As mentioned above, my first question in assembling this review was how should catastrophic wind intensity be quantified? Among the papers I reviewed, 101 included some description of storm intensity, usually either maximum wind gusts or average wind speed. Figure 11 represents the ranges of wind speed for each of four types of catastrophic wind events: windstorms, gales, hurricanes, and tornadoes (see literature cited for Appendix F-I). Several authors, however, mention that the critical factor in wind disturbance is gustiness; pulses of wind can cause more damage than a steady wind (Moss 1940, Conrad 1945, Smith 1946, Steven 1953b, Wyman 1954b, Cremer et al. 1977, 1982; Webb 1986). Therefore, the actual stress perceived at the point of assessment of damage is difficult to quantify. Few papers indicate the distance from the study site to the site where wind was measured and/or distance from the eye of the storm (Oliver & Mayhead 1974). Francis and Gillespie (1993) propose that storms that increase to maximum wind speed quickly may result in more damage, as the trees do not defoliate first. Figure 11 - Intensity of catastrophic wind events. Values from citations in Appendix F - I. Lugo et al. (1983) suggest that duration of hurricane winds and wind speed would be a minimum standard for comparison and that this might help separate differences in damage due to ecosystem response from those due to variations in the intensity of the "stressor". However, wind intensity of particular storms may vary significantly over scales of kilometers (Canam 1978, Boose et al. 1994) and the lack of detailed local meteorological data has made comparisons difficult (Tanner et al. 1991). Scatena and Larsen (1991) developed three indices of storm intensity, using: 1) maximum sustained wind and storm duration, 2) maximum sustained wind and proximity to the storm, and 3) rainfall totals (as percentage of annual average); each representing different views of hurricane impacts: total storm energy, impact at a specific point, and relative rainfall. Quantifying Damage Our second question was: how should catastrophic wind damage be quantified? There is little consistency in the methodologies used to quantify damage and assess recovery. Sampling plots vary from 2.5 m2 (litterfall studies) to 16 ha and include circular plots, rectilinear plots, gridded plots, and transects. Minimum stem diameters range from 2 to 20 cm dbh. The delay in sampling after the storm varies from ten days to thirteen years (not including analysis of historical storm data). The greatest variation of methodology is in the parameters used to quantify damage. For example, of the 48 papers reviewed in Everham (1995) that included quantification of damage, 28 different schemes or methodologies were used, including such difficult to convert measures as "trees/chain of perimeter" (Alexander 1964, 1967). Clearly, there are no standards for assessing impacts of catastrophic winds. One obvious explanation is that few researchers design experiments that anticipate naturally occurring catastrophic events. Instead, they capitalize on studies established before the disturbance or on surveys afterward (Lugo et al. 1983, Tanner et al. 1991). Efforts to quantify damage can be classified into six types: 1) stem damage, 2) branch damage, 3) canopy damage, 4) mortality, 5) volume or mass changes, 6) classification categories (using one or more of the above to set arbitrary damage categories of high, low, etc.). These types are not mutually exclusive. Many authors used more than one to describe the damage of catastrophic wind, but none reported damage in all six categories. Appendix F-II includes a summary of the different damage quantification methods reviewed. The variation in quantification techniques increases the difficulty of developing generalizations. STEM DAMAGE - Stem damage is most commonly reported as percent of downed trees, referred to as: windthrow, blowdown, windfall, or windblow. These terms are ambiguous and do not distinguish between stem breakage and uprooting (Schaetzl et al. 1989b). Secondarily, these types of damage are often differentiated as: uprooting, breakage, and bending or leaning. Most authors use actual counts from established plots, but aerial surveys and area estimates are reported also (Wilson 1976). These numbers may be given as raw counts (Versfeld 1980) without pre-disturbance stem numbers or densities, or as percentages. Researchers vary in their inclusion of leaning or bent stems and in the angle of leaning stems to include. Wilkinson et al. (1978) used twelve categories of stem lean, four categories of stem damage, and four categories of root exposure, in addition to categories relating to crown damage. Gresham et al. (1991) set eight categories from undamaged stems to downed stems. Scatena and Lugo (in press) used three main categories: little damage, standing damage, and down damage; but the standing category had nine subcategories and the prone category had five. Brewer and Linnartz (1973) suggest that a 45o lean from vertical is a critical point for survival [Pinus taeda (L.) stems with less lean had greater than 50% survival]. BRANCH DAMAGE - Branch damage is quantified either by counting branches damaged or by assignment to arbitrarily defined categories and varies based on the size of the branches assessed. Many researchers set no clear size categories and grade branch damage subjectively. Moreover, when these clearly defined categories are applied to standing damaged trees the assessment of branch diameters from the ground is subjective. Different minimum branch sizes, unquantified branch sizes, and categorization of damage all make it difficult to compare branch damage values. CANOPY DAMAGE - Canopy damage may be quantified directly from estimates of structure loss or defoliation, or determined indirectly using techniques to measure structure or changes in light levels. Both Whitmore (1989b) and Glitzenstein and Harcombe (1988) attempted to determine directly the percent canopy change for an entire study area. Whitmore (1989b) scored each plot as gap, building, or mature, and was then able to determine the percent of the area in gaps. Glitzenstein and Harcombe (1988) used aerial photographs to determine the percent cover after a tornado. Frangi and Lugo (1991) scored each stem for percent defoliated. More typically, researchers establish several categories for scoring canopy damage on each stem. This categorization of canopy damage makes it difficult to determine an overall average for the disturbed site. Canopy change has been assessed indirectly by measuring changes in light levels. Turton (1992) used hemispheric photography to determine the percent change in available light at the forest floor. Veblen et al. (1989), Fernandez and Fetcher (1991) and Walker et al. (1992) all measured light levels after wind disturbance; only the latter used a site established before the storm and were able to quantify changes. Whitmore (1989b) used photographs to assess canopy damage and monitor recovery of specific trees. No effort was made to determine an average for the area of study. Applegate and Bragg (1992) also used photographs to monitor recovery of the canopy, but no attempt was made to quantify these changes. Changes in canopy structure have been measured more directly by estimating vegetation cover at different heights above ground to produce vegetation profiles (Brokaw & Grear 1991, Wunderle et al. 1992). Both measures of light levels and of vegetation profiles promise to give clearer averages of canopy changes, if measurements are made before and after the disturbance. MORTALITY - Mortality measurements pose a number of problems. Mortality is often assumed to be the same as blowdown, which may be reasonable for some coniferous forests, but is not for many broadleaf forests - particularly in the tropics where percent resprouted stems may be as high as 64.8% of dicot stems (Zimmerman et al. 1994), 87% of snapped stems (Boucher 1990), or 56% of all trees (Bellingham et al. 1992). No clear relationship exists between damage and mortality. When and how mortality is assessed influences the values found. For example post-disturbance mortality appears to be elevated for months to years after catastrophic wind (Craighead & Gilbert 1962, Sauer 1962, Wiley 1965, Dittus 1985, Putz & Chan 1986, Veblen et al. 1989, Bellingham 1991, Smith et al. 1994), invalidating immediate post-disturbance estimates. Shaw (1983) reports that "overmature" trees may survive a storm, but rapidly decline and die from even minor damage. Webster (1963) found increased windfall and mortality in subsequent months, for trees that were damaged by a tornado. Walker (1995) states "mortality estimates made before trees begin to sprout after a hurricane may be too high ... estimates made during the intermediate period of vigorous sprouting may be too low if damaged trees continue to die". Often assessment of storm-associated mortality can be obscured by standing dead trees that died before the storm event. Most researchers appear to determine subjectively whether a stem was previously dead or was killed by the storm. Thirty-three of the papers reviewed by Everham (1995) that quantified damage did not include measures of mortality. This leads to a potential misunderstanding of the impacts of catastrophic wind damage; for example, O'Brien et al. (1992) ran computer simulations of hurricane damage with 100% mortality, although hurricane mortality never reaches that level. Mortality should be tracked for several years after catastrophic wind events to determine the extent of elevated mortality. VOLUME OR MASS CHANGES - Volume losses are typically reported by foresters concerned with timber yield. Often, estimates of volume lost are reported, without corresponding estimates of undamaged timber or conversion to percent loss. Biomass, or related measures such as basal area, are often measured or estimated by ecologists interested in energy or nutrient flows. Again, undamaged biomass, estimates of pre- disturbance levels, and totals from undisturbed areas are generally not reported. Sheffield and Thompson (1992) included a detailed assessment of the impacts of Hurricane Hugo on forests in South Carolina. Their assessment is based on volume losses, but includes pre-hurricane levels, pre-hurricane mortality, hurricane mortality, non-lethal hurricane damage, and an elaborate post-hurricane risk assessment that is forest- and age-class specific and includes consideration of crown loss, root damage, stem damage, and salt burn. CLASSIFICATION CATEGORIES - Five of the papers reviewed by Everham (1995), and five others (Wiley 1965, Touliatos & Roth 1971, Shaw 1983, Hook et al. 1991, Smith et al. 1994), included an overall damage categorization that could be used for comparing spatial patterns of damage. As might be expected, no two categorizations were the same. These schemes use three to five categories of percentages of types of stem damage. Although this type of categorization may facilitate analysis of spatial patterns, unless overall average values of damage are given, comparisons among sites and events are difficult. SUMMARY OF DAMAGE QUANTIFICATION - Apparent severity of damage from catastrophic wind can vary depending on how damage is measured, which, in turn, reflects the different goals of researchers. In addition, different methodologies may influence results even when the same measure is used. Measurements of damage severity are sensitive to species distributions and to stem size. Species differences are most pronounced when differences in types of stem damage or in stem and branch damage are considered. These inconsistencies make it difficult to develop generalizations about catastrophic wind disturbance; any standard for quantifying damage should minimize these influences. Therefore a summary damage parameter that incorporates all types of stem damage would tend to minimize the species effects. One approach to eliminating the problem of using different minimum stem sizes would be reporting data at a variety of stem sizes from the smallest stems up. Comparisons could then be made using the larger stem classes, common to more studies. Another approach would be to use a damage parameter that is less sensitive to change in minimum stem size. For example, basal area calculations minimize the effect of small stems and therefore vary little with changes in the minimum stem size. Pickett and White (1985) suggest either biomass destroyed or individuals killed as possible measures of disturbance effects. A single general measure of structural damage, such as basal area lost, may not be adequate to describe changes to a forest and, in particular, predict the path of recovery. Boucher et al. (1990) reported an inconsistency between damage and recovery from Hurricane Joan in Nicaragua: "damage to pine forests was considerably less than to rain forests", but "long- term prospects for [the recovery of] rain forest areas...[are]... brighter". In the rain forest, 75% of the stems were broken or uprooted compared to 44% of the stems in the pine forest, but mortality in the pine forest was 58% compared to only 13% in the rain forest. The rain forest was more affected only if damage is quantified by percent stems lost. But in terms of mortality, the rain forest was less affected and its more rapid recovery is predictable. The effects of a given catastrophic wind event might best be expressed as a combination of a general measure of structural damage (percent basal area or percent stems lost) and a measure of potential compositional change (percent mortality). Figure 12 diagrams this concept with a comparison among 20 disturbed forests. The points on this graph are located using percent mortality and percent stem damage or basal area lost. Several trends are evident: 1) mortality tends to be concentrated at the lower end of the gradient, 2) structural damage is distributed throughout the gradient, and 3) the region of gradient space defined by high mortality and low structural damage rarely exists. The maximum reported hurricane mortality is for a pine forest in Nicaragua (Boucher et al. 1990; 11 in Figure 12). The highest structural damage attributed to a hurricane is 94% for the 1938 storm in New England (Foster 1988b; based on estimated basal areas before and after the hurricane; 5 in Figure 12). The highest reported structural and compositional damage both occur in a tornado- impacted temperate forest in Pennsylvania (Peterson & Pickett 1991; 20 in Figure 12). Figure 12 - Secondary gradient space for severity of catastrophic wind events 1 & 2 Webb 1988, 3 Greene et al. 1992, 4 Dittus 1985, 5 Foster 1988b, 6 Whigham et al. 1991, 7 & 8 Lugo et al. 1983, 9 Liegel 1984, 10 & 11 Boucher 1989 and 1990, 12 & 13 Bellingham et al. 1992, 14 Frangi and Lugo 1991, 15 Walker 1991, 16 Dallmeier et al. 1991, 17 Zimmerman et al. 1994, 18 & 19 Glitzenstein and Harcombe 1988, 20 Peterson and Pickett 1991. To examine adaptations to wind, we must be able to standardize how we examine damage. Three recommendations can be made to facilitate comparisons. First, data reported with separate categories for size classes, vegetative types (conifer, broadleaf; early or late successional), and damage classifications would facilitate comparisons of the impacts of wind. Second, effort should be made to quantify undamaged trees, or to provide estimates of pre-disturbance forest composition. Neustein (1971) states "all surveys which did not include undamaged crops are basically weak". Finally, a standard for all researchers reporting average values for their study sites of both structural damage and mortality should help us to compare events and develop general hypotheses about catastrophic wind disturbance. Pattern of Damage Our third question was: Is catastrophic wind damage homogeneous, or are there patterns of damage, and can these patterns be predicted? Walker et al. (1992), Foster and Boose (1992), and Boose et al. (1994) all mention variation in damage over the landscape due either to proximity to the eye of cyclonic storms, or shifts in wind direction as the storm passes. At smaller scales, researchers in a variety of forests have noted variation in damage: Putz and Sharitz (1991) for hurricane effects in South Carolina, Bellingham et al. (1992) for hurricane damage in Jamaica, Foster (1988b) and Conrad (1945) for hurricane damage in Massachusetts; Webb (1958) for cyclone damage in Queensland, Shaw (1983) for cyclone damage in New Zealand, Wiley (1965) for hurricane damage in Washington, Browne (1949) for cyclone damage in Malaysia, Steven (1953b) for gale damage in Scotland, Stoeckeler and Arbogast (1955) for windstorm damage in Wisconsin, Furley and Newey (1979) for hurricane damage in Belize, Glitzenstein and Harcombe (1988) for tornado damage in Texas, and Weaver (1989) for hurricane- damaged recovering forests in Puerto Rico. Canham (1978) stated that successively more intense winds would tend to topple more successively resistant stems until the whole canopy is opened. However, there is no evidence of increasing homogeneity of damage with increasing intensity of wind. Minimum gap size caused by wind disturbance varies, but this variation is partly an artifact of methodology. Lorimer (1977) used surveyors' records to quantify historical blowdown patches as small as 25 ha, but proposed that damage may have been much more spotty, but not recorded by surveyors. Canham (1978) also used surveyors' data to document windstorm gaps from 0.65 to 3785 ha with a median of 32.4 ha. Ruth and Yoder (1953) reported windstorm damage in gaps down to 0.5 ha. Foster and Boose (1992), using aerial photos, found patches of hurricane damage of 0.04 (their arbitrarily set minimum size) to 35 ha, with most less than two ha. Foster (1988b) reported hurricane blowdown as "highly irregular" in size and shape, with gaps as small as 30 m2. Webb (1986) found gaps of 20 to 269 m2 in a pine-fir stand, and smaller gaps in a pine-maple stand, following a severe windstorm. Baldwin (1940) found patches "devoid of trees" up to 0.2 ha following the 1938 hurricane. Thus catastrophic windstorms can produce a great range in size of damaged patches. As Foster (1988b) observed, although there is generally uniform action of wind, local gusts and turbulence produced some otherwise inexplicable patterns of damage and survival. Biotic and Abiotic Influences on Damage Foster and Boose (1992) state that a "relatively small number of variables can be used to explain the damage" in their uncomplicated terrain. These factors can be divided into two groups: biotic and abiotic. Because catastrophic wind research principally involves post-disturbance surveys not coupled with pre-disturbance data, it is difficult to separate biotic and abiotic factors and determine their relative roles in influencing damage, the subject of my fourth question. BIOTIC FACTORS - Biotic factors that may influence the severity of wind on forests have been identified for decades; they include: a) stem size, b) stand conditions, c) species differences, and d) tree pathogens; e.g., rot, fungus, and insect infestation (Hubert 1918, Bromley 1939, Smith 1946, Day 1950, Stoeckeler & Arbogast 1955, Busby 1965). Stem Size - Most researchers have found a positive correlation between stem size and catastrophic wind damage. An increased risk for wind damage to larger trees has been observed by: Smith and Weitknecht (1915) for windstorm damage in Oregon; Behre (1921) for windstorm damage in New York; Brewer and Merritt (1978) for windstorm and possibly tornado damage in Michigan; Clapp (1938), Rowlands (1941), Curtis (1943), and Foster and Boose (1992) for hurricane damage in New England; King (1945) for cyclone damage in Mauritius; Lines (1953) and Andersen (1954) for gale damage in Scotland; Wadsworth and Englerth (1959) for hurricane damage in Puerto Rico; Neustein (1968) for severe windstorm damage in Scotland; Spinage and Guinness (1971) for windstorm damage in Rwanda; Whitmore (1974) for some species damaged by cyclones in the Solomon Islands; Dunn et al. (1983) for a windstorm in Wisconsin; Harcombe and Marks (1983) for windstorm damage in Texas; Dittus (1985) for cyclone damage in Sri Lanka; and Glitzenstein and Harcombe (1988) for tornado damage in Texas. Some studies report increasing damage with size up to some maximum, where damage stabilizes regardless of further increases in size. Damage ceased to increase above 30 m in height for pine plantations in Australia (Cremer et al. 1977, 1982). Scots pine damage levelled off in size classes above 18 m tall, although damage continued to increase for larger sizes classes of broadleaf species in Scotland (Andersen 1954). Damage ceased to increase in stems above 35 cm of diameter for softwoods in South Carolina (Sheffield & Thompson 1992). Three studies suggest the possibility of a negative correlation between size and damage. Boucher (1990) reported highest survival in the largest trees, after a hurricane in a rain forest in Nicaragua, yet in Boucher et al. (1990) both damage and mortality showed a unimodal response, with the greatest damage in the medium tree class. Alexander (1964) found more damage in understory trees when examining the wind damage to clear cuts in Colorado, but this may have been a unique response to the stand conditions and thinning history. Cremer et al. (1977), examining the impacts of a windstorm on pine plantations in Australia, reported 15% windthrow in stands 30-35 m, 13% windthrow in stands 35-40 m, and only 9% windthrow in the tallest stands of 40-45 m. However, these tallest stands had the longest history of thinning and may have been more preconditioned to wind. It is possible that the tallest trees were actually more resistant in many of the studies that indicated a positive correlation between tree size and damage, but sample sizes of the largest categories tend to be small and are often pooled over a large range. The actual relationship between tree size and damage may be more complex than a simple linear correlation. Most studies indicate a unimodal distribution of damage relative to stem size, with the smallest trees sheltered from the wind and the largest trees pre-conditioned to 'weather the storm'. Bromley (1939) found that the youngest and the oldest trees survived the 1938 New England hurricane best, while trees 30-100 years old were damaged the most. Peterson and Pickett (1991) found the most uprooting at intermediate (30-40 cm dbh) stem sizes. Both Webb (1958) and Applegate and Bragg (1992) found cyclone damage to a rain forest in Queensland to be concentrated in the intermediate-sized stems (10-20 cm dbh, in the former; and to the an intermediate canopy layer, with an emergent layer surviving, in the latter). Boe (1965), studying the impact of a hurricane on redwood stands in Northern California, found higher percent damage in intermediate size classes (30-60 cm dbh). Neustein (1971) found gale damage in Scotland increased in trees with stems up to 12 m tall, but then dropped off. Dittus (1985) reported the highest percentage felled to be in the canopy layer (0-90%) with lower percentages in both the emergent (10-70%) and the subcanopy layers (0-75%). In Nicaragua, both Boucher (1989) and Yih et al. (1991) reported damage and mortality concentrated in stems of 15 to 45 cm dbh. Putz and Sharitz (1991) described Hurricane Hugo damage in South Carolina as most severe in the non-dominant trees, with the largest trees receiving crown damage, but not snapping or uprooting. Their data for bottomland forest stand show a peak of damage in the middle (40-80 cm) diameter size class if the two types of stem damage are combined. This relationship also is indicated by the results of a wider survey of hurricane impacts in South Carolina (Sheffield & Thompson 1992). Dallmeier et al. (1991), studying the impact of Hurricane Hugo in Puerto Rico, found no relationship between stem damage and stem size, but higher crown damage to intermediate stem sizes of 20-30 cm dbh. Frangi and Lugo's (1991) assessment of hurricane damage in Puerto Rico clearly shows a unimodal response, with the greatest damage in the class 12-14 m tall, while the largest size class (24-26 m) suffered mostly defoliation and branch loss. Sauer (1962) found that young (<3 years) and old (>25 years) Casuarina equisetifolia (L.) were relatively less damaged. Glitzenstein and Harcombe (1988) compare five studies based on the average stem diameter and the percent basal area lost (their Figure 10). The trend appears to be a positive linear relationship, but if the point for the downburst in Wisconsin is considered an outlier, given its extreme intensity (253 km/hr), the graph could easily be viewed as unimodal. It appears there is ample evidence to support the hypothesis that the largest trees are more able to withstand some catastrophic winds, resulting in a unimodal response curve of stem size versus damage (Figure 13). But, this pre-conditioning of the largest stems may be obscured by small samples sizes in some studies, and may be unimportant under the most intense wind disturbance. Especially strong winds may obscure a difference in resistance of different-sized stems (Andersen 1954, Mergen 1954, Wilkinson et al. 1978, Hook et al. 1991). The impact of Hurricane Allen on forests in Jamaica is a good example. Where less than 50% of the stems were damaged, the remaining stems were larger than average, possibly indicating the survival of the largest stems. If the damage was greater than 50%, the surviving stems were smaller than average, possibly reflecting the removal of the largest stems and a downward shift in average stem size (Thompson 1983). Depending on sample size and the size categories used in the analysis, very different conclusions might be drawn. Four studies in three areas indicate a bimodal relationship between size and damage. Baker (1915) found that large trees (203-254 cm dbh) tended to be damaged more directly by wind during storms in Washington, while small stems were more affected by falling trees and branches. Thus the intermediate stem sizes were less damaged, although Baker did note that the very largest stems (>254 cm dbh) were relatively windfirm. Webb (1989) found size related to wind damage, but not mortality in her Minnesota study. The largest stems were damaged directly by the wind and the smaller stems indirectly. The intermediate-sized stems were presumably more sheltered from the wind but large enough to avoid the larger falling stems. However, only the counts of damaged trees per size category are reported, not percentages, so it is difficult to determine the actual distribution of damage. Both Scatena and Lugo (in press) and Basnet et al. (1992) found that an intermediate size class (20-25 m in height, 50-60 cm in diameter, respectively) was the least damaged in a Puerto Rican forest following Hurricane Hugo. Scatena and Lugo used six size categories of 5 m each (from 0-5 to >25) and found percent damaged stems (from smallest to largest) of 32, 21, 30, 43, 19, and 37 (differences not significant). There was not a clear bimodal distribution, and if the six categories are collapsed into three of 10 m each, the percent damaged are 24, 33, and 26, with the greatest damage in the intermediate size class of 10-20 m height. Basnet's data included less than five stems greater than 60 cm; making conclusions about the distribution of damage suspect. A number of researchers have found no significant relationship between stem size and wind damage. Lugo et al. (1983) found a size relationship with types of hurricane damage, but no clear trend for overall stem damage or mortality, but in their study no sample size in categories above 50 cm dbh exceeded ten stems. Wilkinson et al. (1978) found no relationship of hurricane damage to height or diameter, but this was in a 25 year old slash pine plantation with presumably little variation in either size parameter. Qinghong and Hytteborn (1991) found no size difference in windstorm damage, but observed that the trees may not have been large enough to exhibit differences. Putz and Sharitz (1991) reported "no clear relationships between types of damage and tree size" for hurricane effects in South Carolina, but used a minimum stem size of 20 cm and only three categories: 20-40, 40-80 and >80 cm. As mentioned above, their data actually suggest highest percent damage among intermediate sized stems. Alexander (1967) found no size difference in windstorm damage, but sampled only the perimeters of clear cuts where damage would presumably not vary closely with size. Zimmerman et al. (1994) found no overall trend toward the relationship between damage and size, but some species showed relationships between size and specific types of damage, most often an increase in damage to larger branches. A lack of clear trends in size versus damage, within and among studies, may be explained in several ways: 1) the difficulty of isolating differences in height that covary with other species-specific morphological differences, 2) variation in the relationship between height (more readily related to exposure) and diameter (more commonly measured), particularly in the largest diameter classes, 3) confounding of direct with indirect (stems hit by others) wind impacts, and 4) complications caused by different measures of damage, and 5) variations in minimum, maximum, and categories of size. Both Curtis (1943) and Pickering (1986) point out that a problem in correlating damage with stem diameter is that the form of the crown can greatly influence exposure to the wind. Deep crowns (distributed along the stem) may resist damage better than shallow crowns (concentrated at the top) (Boe 1965, Pickering 1986). Curtis (1943) refers to shallow-crowned trees as having a "high point" and also mentions crown density as positively correlated to wind damage, and the suppleness of the stem as negatively correlated to damage. Both crown form and crown density, at least in part, are independent of diameter. Curtis provides several figures to illustrate these morphological differences and provides data suggesting that species differences in damage actually relate to differences in crown development. King (1986) discusses increased exposure with increased height, but points out that the largest diameter tree may add additional diameter without growing taller. Cremer et al. (1982) suggests this change in height to diameter ratio in the largest trees increases their resistance to wind. Stem taper, related to stem flexibility and subsequently to damage (Cremer et al. 1982, Petty & Swain 1985), is also not distinguished by a simple diameter measure. Table 1 - Comparison of quantifications of stem size
Minimum Category Maximum Damaged Citation Stem Size Interval Stem Size Related to Size
Baker 1915 25 cm 25 cm 305 cm Yes Hansen 1937 2.5 cm 2.5 cm 25 cm Yes Boe 1965 25 cm 25 cm 229 cm Yes Reiners & Reiners 1965 15 cm 18 cm > 86 cm Yes Brewer & Merritt 1978 25 cm 25 cm > 76 cm Yes Harcombe & Marks 1983 4.5 cm 20 cm > 40 cm Yes Glitzenstein & Harcombe 1988 4.5 cm 5 cm 60 cm Yes Peterson & Pickett 1991 10 cm 10 cm 90 cm Yes Putz & Sharitz 1991 20 cm 40 cm > 80 cm Yes Yih et al. 1991 5 cm 15 cm > 46 cm Yes Reilly 1991 5 cm 2 cm 100 cm Yes Basnet et al. 1992 2.5 cm 10 cm 70 cm Yes Dallmeier et al. 1991 10 cm 5 cm > 55 cm Yes Qinghong & Hytteborn 1991 10 cm 10 cm 70 cm Yes Sheffield & Thompson 1992 15 cm 5 cm > 51 in Yes Zimmerman et al. 1994 10 cm 5,10 cm >100 cm No
Webb (1986) estimated that indirect causes account for 71% and 86% of the damage in two forest stands struck by a severe windstorm. With as many as two thirds of the stem damage caused by other falling trees, the size of neighboring trees potentially will be better correlated to damage than size of a damaged tree itself (Smith & Weitknecht 1915, and see Baker 1915, Reiners & Reiners 1965, Falinski 1978, Harcombe & Marks 1983, Dittus 1985, Basnet 1990, Frangi & Lugo 1991, Ogden et al. 1991, Reilly 1991, Applegate & Bragg 1992, Greene et al. 1992). Size may also influence the type of damage. Larger trees can be more subject to windthrow and smaller trees to stem snapping (Lugo et al. 1983, Thompson 1983, Healey 1990). This could be due to the difference between direct wind impacts on larger stems and indirect impacts on smaller ones (Figure 13). Bellingham et al. (1992) and Walker et al. (1992), however, found no differences in type of damage based on stem size. Putz et al. (1983) found higher numbers (percent of stems not reported) uprooted in middle-sized stems. Peterson and Pickett (1991) found the largest stems in a tornado disturbance were more likely to be snapped than uprooted. These complications have resulted in different interpretations of the relationship of stem size to damage: 1) no relationship, 2) a positive linear relationship, or 3) a more complex relationship. Determining relationships between diameter and damage has been complicated by methodological differences. All 16 studies in Table 1 used different minimum or maximum stem size, or intervals of stem sizes to examine size and damage relationships. Stand Conditions - There are several aspects of stand condition or history that may influence the severity of catastrophic wind damage: 1) even-aged versus mixed-age, 2) single-species versus mixed-species, 3) maturity, 4) density relative to thinning history, and 5) juxtaposition with other disturbances or contrasting silvicultural treatments. Many authors have noted increased damage in even-aged stands, however the impact due to stand structure is difficult to separate from impacts due to stand age, thinning history, and species composition in these usually single-species stands. This distinction is important, since it is hypothesized that the lower and more uniform canopy height (lacking emergents) of some wind impacted forests, makes them less susceptible to catastrophic wind (Webb 1958, Odum 1970b, Brown et al. 1983). Clapp (1938) found that uneven-aged stands lost all big trees, while even-aged stands were blown down entirely. Others have found mixed, irregular-shaped canopies to be more stable (Jensen 1941, Ruth & Yoder 1953, Andersen 1954, Boe 1965). An uneven-aged canopy exposes the largest trees to chronic winds and may precondition them to acute wind stress (Smith 1946). Cline (1939) found young even-aged conifer stands susceptible to the wind. However, Clapp (1938), Boe (1965), and Peart et al. (1992) found the youngest stands to be wind resistant. The difference may be in definitions of "young". Curtis (1943) found increased damage in older, dense, even-aged stands, but concluded "uneven forests in themselves are not more windfirm than even aged [stands]". Thompson (1983) also found greater damage in even-aged plantations, but this damage varied with the species. Plantations of Pinus caribea (Morelet) and Eucalyptus spp. were seriously windthrown, while plantations of Hibiscus elatus (Sw.) were only debranched. Natural forests, which were both uneven-aged and multi-species, were less damaged and suffered less mortality (58% versus 13%) than any plantations. Wadsworth and Englerth (1959) found spotty damage to individual trees in mixed species, uneven-aged stands, high or no damage in mixed species, even-aged stands, and high damage in even-aged plantations. There appears to be less preconditioning to wind in dense even-aged stands, making them more susceptible once the canopy is opened, but this effect may be either amplified or mitigated by species composition. Differences in susceptibility of multiple- compared to single-species stands are confounded by the fact that often single-species stands are also even-aged plantations, and there is little consistency in reports. Bellingham (1991) found higher damage in pine plantations (80% blowdown) than in mixed species stands (20% blowdown). Trousdell et al. (1965) also reported more damage to natural stands than to single-species plantations (age of plantations not given). Lugo et al. (1983) found native species-rich climax associations to be less damaged than less diverse palm brakes, however these forest types differ in stand structure, and the palm brake plots were closer to the path of the hurricane and presumably exposed to more intense winds. Clapp (1938) found no difference in mixed stands relative to pure stands. Cheston (1940) found more damage in mixed pine-hardwood stands than in pure hardwood stands, since the pines were selectively damaged. Andersen (1954) recommends mixing in windfirm species to decrease wind damage in plantations. In any case, higher species diversity increases the likelihood that some of the species will be relatively more wind-resistant. This could minimize damage where the alternative is unchecked damage in pure stands of susceptible species. Many researchers report a positive correlation between stand age and damage, and many of the biotic factors, e.g. stem size, species differences, and presence of pathogens, are associated with changes in the age of the stand. King (1986) found higher damage in mature stands than in smaller, denser stands of second growth. Steven (1953b) found damage concentrated in middle-aged and older forests. As reported above, although Cline (1939) found young, even-aged conifer stands susceptible to the wind, others report less damage in younger stands, or in young, dense, second growth vegetation (Clapp 1938, Trousdell 1955, Wadsworth & Englerth 1959, Cruickshank et al. 1962, Boe 1965, Touliatos & Roth 1971, Foster 1988b). Canham (1978), found old-growth blown down but not younger stands. Sheffield and Thompson (1992) recorded higher damage for older stands, and Curtis (1943) reported the same for old, dense, even-aged stands in which canopies tended to be concentrated high on the stem and root development was poor. Both Andersen (1954) and Foster (1988a) found ages at which the damage leveled off, not increasing with age. Andersen (1954) found that stand-level damage leveled off at ages with average stem height of 60 feet, though species within the stand were affected differently. He also reported on several studies in Europe where no relationship was found between age and windfirmness. Foster (1988a) found increasing damage with age up to 30 years. Others report decreases in damage with the oldest stems, and therefore the highest damage in stands of an intermediate age (Bromley 1939, King 1945, Wiley 1965). This may be due to the response noted above for stem size: the oldest stands have more stems that have survived previous storms and are now pre-conditioned to resist damage. Bromley (1939) found the most damage in stands 30-100 years old. Younger stands were not exposed to the wind, and older stands had developed deeper, stronger root systems. In the case of King (1945), the young plantations had 30-60% damage, the medium age stands 100%, and the oldest 90%, but the stands also differed in species composition. A relationship between stand age and damage may not be due to age per se but to age- related stand characteristics such as number of large trees, structural complexity, density, and species composition, all of which are influenced by stand management. More wind damage is reported in thinned stands (Bromley 1939, Ruth & Yoder 1953, Nelson & Stanley 1959, Wadsworth & Englerth 1959, Wilson 1976, Liegel 1984). Thinning increases wind flow in the canopy and removes the support of surrounding trees. Yet others report that the most heavily damaged forests were the least logged (Trousdell 1955, Merrens & Peart 1992, Peart et al. 1992), perhaps because they contained more old, susceptible trees. Curtis (1943) likewise stated that old, dense stands were more susceptible to wind damage. These conflicting reports might be resolved if the timing of thinning was included in the analysis. What appears critical about stand density, is the temporal proximity of thinning to subsequent wind exposure. In many cases stands are highly susceptible to wind damage immediately after thinning, but with further development can become more resistant. Alexander (1964) reports that stands growing in less dense conditions, as opposed to thinned, are more stable (see also Cremer et al. 1982). Versfeld (1980) suspected that delayed thinning had increased damage in some stands. Andersen (1954) reviewed five studies on the effect of thinning on subsequent damage but found increased damage only when the stand was "recently" thinned (and see Touliatos & Roth 1971, Foster 1988a). Wadsworth and Englerth (1959) reported more damage to thinned stands, in their case thinned less than two years before a hurricane. Cremer et al. (1977), studying the impact of catastrophic wind on plantations in Australia, had records of thinning history of each stand. Stands that were thinned less than five years before the disturbance had 22% windthrow. Stands thinned less recently had only 0.2% windthrow. In addition, unthinned stands downwind of clearcuts had 38% windthrow, but if these downwind stands were thinned less than six months before, the damage was 88%. Over time a thinned canopy fills in and the remaining stems thicken, reducing wind damage. Cremer et al. (1977, 1982) suggest that five years is required to recover from thinning, which is supported by Weidman's (1920a) results that two thirds or more of the damage occurred in the first five to six years after thinning. In one plot, 94% of wind damage in the fifteen years after thinning occurred in the first five years (Weidman 1920a). Mergen (1954) states that stem growth, and pre-conditioning to wind, is more rapid in young stems, so that stand recovery time may vary with stand age. Thinning increases stand wind resistance when weaker stems are culled. Boe (1965) found less damage in shelterwoods where 75% of the co-dominants were removed, presumably because the strongest stems were left, while many of the weaker stems had been removed. However, Nelson and Stanley (1959) found greater damage in stands that had been selectively cut to remove the diseased and poor-risk trees, but this thinning occurred less than a year before a hurricane. In another case of thinning less than a year before wind disturbance, Hansen (1937) found decreased damage and proposed that the most susceptible stems had been removed. Of course winds may remove weaker stems selectively, resulting in decreased damage over time (Ruth & Yoder 1953, Alexander 1967). Natural treefall gaps and clear-cuts may increase stand susceptibility when greater canopy roughness leads to further blowdown. Openings may result from harvesting or from tree death due to insects, pathogens, or wind (Andersen 1954). Greene et al. (1992) proposed a positive feedback from disturbance to blowdown to account for a continued decline in biomass after disturbance. Young and Hubbell (1991) showed that trees along edges of existing gaps in a Panamanian forest were more likely to fall than other trees. Cremer et al. (1977) found that damage increased below clearcuts (38%) as compared to non-isolated stands (0.2%). Others reported that subsequent storms will extend prior openings in canopies (Mergen 1954, Neustein 1968). Cruickshank et al. (1962) explain this effect as increased turbulence caused by the previous gaps or clear cuts. They found increasing damage over the course of three storms and proposed that each storm made the stand more vulnerable (though the intensity of the storms also increased). Cremer et al. (1977) state they found no evidence of patches of downed trees progressing from a weak point in the canopy, but Baker (1915) found clusters of damaged trees that he explained as one tree falling, opening the canopy, and weakening the roots of neighboring trees; later referred to as the "domino effect" by Canham (1978). Shaw (1983) reported that canopy gaps increased the likelihood of damage. Foster (1988b) reported that damage was increased by past logging and agricultural activities that create abrupt changes in stand height and density (Somerville 1980, DeWalle 1983). In one case where higher damage was detected in relatively unthinned stands, they were on the edges of clearcuts (Boe 1965). To summarize my review of the relationships between stand condition and wind damage: 1) even-aged stands may be relatively more resistant at younger ages, when the dense, uniform canopy reduces resistance to wind, but these stands become more susceptible if any canopy opening occurs, because the trees are not adequately pre-conditioned to wind; 2) wind resistance of more diverse stands may simply reflect the presence of more species that are wind-firm, rather than being an emergent property of species richness; 3) the relationship between stand age and damage is complex, with the youngest, densest stands resistant to wind damage; increasing age brings a more uneven canopy and increased incidence of pathogens (see below) weakening some trees, while others strengthen their root systems and become more windfirm; 4) wind disturbance increases when the stand has been thinned recently; and 5) wind damage in a stand may be increased by adjacent gaps, whether natural or human made. Species - Many studies have reported differences among tree species in susceptibility to wind damage. Appendix F-III summarizes the reported resistance to catastrophic wind for 242 species in 61 families. Resistance was classified as: 1) low resistance to wind disturbance, 2) intermediate resistance, and 3) high resistance. Species were assigned to these three categories: 1) by the researcher, 2) by division into three equal-sized groups, in cases where species' resistances were ranked in a list, or 3) according to naturally occurring clusters of damage severity in cases where species-specific damage data were presented. I arbitrarily divided the intensity of disturbance into low (<160 km/hr) and high (>160 km/hr) categories. Of the 42 papers reporting species- specific damage resistance, 21 were from tropical sites and 21 from temperate, 33 followed high intensity disturbance, and nine followed lower intensity disturbance. No distinct trends are apparent for individual families, nor for increasing intensity of wind for a given species. Species may differ in type of damage suffered, mortality following damage, or resprouting after damage. A number of researchers have reported species differences in stem breakage versus uprooting (Stoeckeler & Arbogast 1955, Wadsworth & Englerth 1959, Van Hooser & Hedlund 1969, Nonnemacher 1970, Touliatos & Roth 1971, Liegel 1982, Putz et al. 1983, Veblen 1986, Webb 1988, Hook et al. 1991, Peterson & Pickett 1991, Greene et al. 1992, Zimmerman et al. 1994). Others have examined differences in mortality (Sauer 1962, Harcombe & Marks 1983, Putz et al. 1983, Webb 1986, 1989; Glitzenstein & Harcombe 1988). Species differences in post-disturbance sprouting are often reported (Pickering 1986, Glitzenstein & Harcombe 1988, Putz & Brokaw 1989, Peterson & Pickett 1991, Walker et al. 1992, Bellingham et al. 1994, Zimmerman et al. 1994). Differences among species in either severity or type of damage can often be related to morphological differences such as: 1) strength of wood, 2) tree geometry, including shape of the bole or shape and size of crown, and 3) extent and depth of roots (Touliatos & Roth 1971, Webb 1986). Mergen (1954) and Wilson and Archer (1979) review stress, damage, and tree morphology. Numerous authors have linked damaged levels to wood strength or density. Curtis (1943) found that the bending force required for hardwoods is greater than twice that required for white pine (Pinus strobus L.). Weaver (1989) found that higher specific gravity is associated with less breakage and uprooting. Hook et al. (1991) concluded that the strong wood of Quercus virginiana (Mill.) resulted in lost branches only and little windthrow. Walker et al. (1992) found less snapping with higher density wood, and Van Hooser and Hedlund (1969) reported more snapping in pines and more uprooting in the denser broadleaf species. But Peterson and Pickett (1991) found no differences in proportion of uprooted to snapped trees among species that differed greatly in modulus of rupture. However, the intensity of tornado winds in the latter case may have obscured differences related to morphology that are important in less intense disturbances. Others have found a negative correlation between wood strength mortality (Putz et al. 1983, Webb 1989, Zimmerman et al. 1994). Putz et al. (1983) found that species with shorter, thicker stems, and denser wood tend to uproot rather than snap, and that species with low density wood tend to have higher mortality. Some authors have linked damage levels to crown and root characteristics. Foster (1988a) reported that species with full crowns and shallow roots are more susceptible than those with a vertical distribution of canopy, flexible branches, and a tapering shape. Cremer et al. (1977) found that sway, and therefore windthrow, increases with slender stems and wind drag of the crown. However, Andersen (1954) reported that slender, whippy stems survived gale damage, while full-crowned trees were blown down. The tendency for dense-crowned trees to be damaged and open foliage crowns to be spared has also been noted (Smith & Weitknecht 1915, Touliatos & Roth 1971, Kerr 1973). As mentioned previously, Curtis (1943) diagrammed several crown types. He found that the pattern of damage in the conifers he studied was related to crown development, greater crown development leading to higher damage. Putz et al. (1983) suggest that the presence of buttresses should lead to more snapped than uprooted stems, but they found no effect on damage or the frequency of uprooting (see also Walker et al. 1992). Basnet et al. (1992) attribute the resistance of Dacryodes excelsa (Vahl) to the presence of root grafts in this species. Lateral root development, in particular on the leeward side of the stem, can preventing uprooting (Smith 1946, Mergen 1954, Day 1950), but Steinbrenner and Gessel (1956) reported that windward-side root development explained damage patterns. Two possible broad trends in species' susceptibility to catastrophic wind damage must be discussed: 1) greater susceptibility to damage of conifers than broadleaf species, (Cline 1939, Curtis 1943, Stoeckeler & Arbogast 1955, Foster 1988a), and 2) greater susceptibility of pioneer than late successional species (Putz et al. 1983, King 1986, Webb 1986, Zimmerman et al. 1994). The first trend may actually be a manifestation of the second, as conifers play an early role in the secondary succession of many temperate forests. A third possible trend, increased damage to exotic species, has been documented by King (1945) and Wadsworth and Englerth (1959). This relationship is difficult to isolate, as exotic species tend to be in plantations that may be more susceptible to wind for independent reasons. Cline (1939) reported pure conifer stands to be blown down like "fields of grain", when hardwoods withstood the wind. In mixed stands, conifers tend to be windthrown selectively (Meyer & Plusnin 1945). Spurr (1956) found 93% of softwood stands blown down, 57% of mixed stands, and only 38% of hardwood stands. Ottenheimer (1992) reported 88% damage in white pine stands, and 28% in mixed hardwood stands. King (1945) found increased resistance in endemic hardwoods relative to pines and exotic hardwoods. When hardwood trees drop their leaves in the temperate zone cold season wind drag is reduced and these species may be less susceptible than evergreen conifers to wind damage, and especially to uprooting where frozen soil provides strong root anchorage (Moore 1988). Neustein (1971) observed that hardwoods and the deciduous conifer, larch (Larix decidua Miller), were less affected by a gale because they were leafless at the time of the storm. Mortality tends to be higher in early successional species (Webb 1986, Zimmerman et al. 1994); thus differences in susceptibility among broadleaf species may be due to their successional class. The susceptibility of pioneer species may be due to their fast growth of weak wood and to exposed positions in the canopy (Whitmore 1974, Veblen et al. 1989, Webb 1989, Nelson et al. 1994). Many fast growing broadleaf pioneer species achieve dominant, and therefore exposed, positions (Whitmore 1974, Foster 1988a, Veblen et al. 1989, Ottenheimer 1992). As mentioned, fast growth and weak wood of pioneers is also associated with higher mortality (Putz et al. 1983, Webb 1989, Zimmerman et al. 1994). King (1986), comparing two broadleaf species, found that the later successional species (Acer saccharum Marsh.) was more resistant than the shade intolerant one (Populus tremuloides Michaux). Early successional species also exhibit less sprouting ability (Putz & Brokaw 1989, Walker et al. 1992, Zimmerman et al. 1994). Not all researchers report these trends. Lindo (1968) estimated that native pines were more resistant than native hardwoods to hurricane damage in Belize. Cruickshank et al. (1962) found broadleaf species more impacted than conifers when a hurricane struck Ireland. Andersen (1954) states that hardwoods and softwoods are equally at risk when in full leaf. Boucher et al. (1990) reported higher percentage withthrow in broadleaf rain forest than in pines stands in Nicaragua, though actual mortality was greater in pine stands. Walker (1991) did not find successional status a good predictor of damage over a wide range of species. Francis and Gillespie (1993) found pioneer species, Ochroma lagopus (Sw.) and Cecropia peltata (L.), to suffer only minor structural damage because they shed their few large leaves early in the hurricane. Several researchers have reported no differences in damage among species. In these studies there may in fact have been no differences attributable simply to species identity, but differences may have been hidden by the measure of damage used or confounded with co-varying differences in size or exposure. Dittus (1985) reported no differences in species rank abundances before and after cyclone disturbance, but did find five species "significantly" reduced. Dallmeier et al. (1991) found uniform thinning (mortality) without regard to species, but did detect species differences in the type of crown damage. Clapp (1938) reported that species differences were actually differences in aspect and exposure, that hardwoods and softwoods blew down equally when equally exposed. Baker (1915), Webb (1988), and Glitzenstein and Harcombe (1988) all state that the species differences they observed were actually differences in size of the different species. Webb (1988) found no differences among species for large stems, which tended to be snapped and killed, while some species with only small stems differed in damage and survival. Species differences can be obscured by other factors. Schaetzl et al. (1989a) state "species that appear to be windfirm in one site context may be highly prone to uprooting elsewhere", and Webb (1986) states "within a species, one population may be susceptible to damage while another is resistant". Foster (1988a) explains these population differences as related to variation in shoot to root form, soil characteristics, historical factors, and chance. Thus differences in populations of a given species can be the result of differences in exposure due to topography or differences in root growth reflecting soil conditions (Clapp 1938, Andersen 1954). From this wealth of studies on differences among species in susceptibility to wind damage some generalizations can be made: 1) species differences do exist and can be explained by canopy architecture, wood density, bole shape, rooting patterns, and susceptibility to infection; 2) softwoods (shade intolerant pioneer species and some conifers) tend to be affected more heavily than hardwoods; and 3) species differences can be obscured by differences in exposure, edaphic conditions, silvicultural treatment, or wind intensity experienced by different populations of a species. Pathogens - A final biotic factor that influences damage is the presence of insect infestation, or fungus or other pathogens. The presence of these wood-weakening agents varies among species and with age, and therefore may affect correlations between these latter factors and damage (Wilson 1976). Bromley (1939) found that the presence of fungus and insect infestation increased hurricane damage to trees in New England. Hubert (1918), reporting on a survey of 90 windfall areas, found 18 percent of the sites to have fungi as the principal secondary cause of damage. Windstorm damage in Wisconsin was associated with the presence of rots and cankers; 75% of the broken stems had internal rot (Stoeckeler & Arbogast 1955). Trousdell (1955) also found hurricane damage associated with heart rot in trees in Virginia. Qinghong and Hytteborn (1991) found increased breakage associated with fungal infections and proposed that windstorms tended to not kill trees unless they were already weakened by fungus or insects. Ruth and Yoder (1953) found that the presence of root rot increased tree susceptibility to wind damage in Oregon. Alexander (1964) also found the presence of root and butt rots increased wind damage in conifers in Colorado. Thompson (1983) observed that 50% of the stumps examined after hurricane disturbance in Jamaica were infected and possibly weakened by fungi. In Washington state, Wiley (1965) found 40% of the stems damaged by a windstorm had either stem or root rot. Webb (1988) reported that wood-rotting fungi promoted breakage of some stems, and root rotting fungi may have stimulated uprooting of others hit by a windstorm in Minnesota. Putz and Sharitz (1991) found hurricane damage in South Carolina to be correlated with previous mechanical damage that may have led to fungal infection (see also Hubert 1918). Interestingly, the majority of these studies involve wind impacts in temperate regions. It is unclear whether pathogens are less of a contributory influence in the tropics, or whether tropical forest researchers have failed to investigate this possibility. Summary of Biotic Effects - I draw several conclusions about the influence of biotic factors on wind damage: 1) there is generally a unimodal relationship between stem size and susceptibility to direct wind damage (this is obscured when smaller stems, damaged by falling trees, are included in the analysis); 2) damage is related to canopy evenness, age of the stand (particularly resistance in the youngest even-aged stands), recent thinning, and proximity of canopy openings, but diversity of species does not increase overall resistance of a stand; 3) there are species differences in susceptibility to different types of damage, subsequent mortality, and ability to sprout following damage; some conifers and early successional species are particularly susceptible to severe stem damage and mortality, but these susceptible species are not necessarily more likely to sprout; and 4) weakening from pathogens has increased damage in temperate forests struck by catastrophic winds. However, any of these trends may be obscured by the highest intensity of wind disturbances, or by conflicting and confounding interactions among biotic and abiotic factors. ABIOTIC FACTORS - A variety of abiotic factors has been examined for their influence on severity of wind damage to forests, these include: 1) storm intensity, timing, and associated precipitation, 2) disturbance history, 3) topographic features, and 4) edaphic conditions. Storm Intensity - Storm intensity is the most obvious abiotic factor that influences severity of damage. The most commonly reported measures of wind intensity are either maximum sustained wind or maximum gusts. As mentioned earlier, Lugo et al. (1983) suggested duration and maximum wind speed as measures of intensity of disturbance. Scatena and Larsen (1991) suggest the additional measures of proximity to the storm and total rainfall. The perceived intensity of wind also may be a function of direction. There is a distinction between destructive winds that come from the same direction as prevailing winds, and those from different directions, for which the biota may not be pre-conditioned (Alexander & Buell 1955, Boe 1965, O'Cinneide 1975, Qinghong & Hytteborn 1991). Also, wind direction changes during a cyclonic storm, while the storm may be gaining or losing intensity (Boose et al. 1994). The impact of a catastrophic wind disturbance may partly depend on timing. During the cold season in temperate climates, deciduous trees will be less susceptible to uprooting, due both to reduce wind drag after dropping leaves and, in the more polar latitudes, to stronger anchorage in frozen soils (Mayer 1989). Precipitation is another aspect of storm intensity that influences damage (Day 1950). The 1938 hurricane in New England caused an abnormally high percent of damage as uprooting (87-100%) as opposed to stem snapping (Bromley 1939, Brake & Post 1941, Curtis 1943, Foster 1988a), perhaps due to the quantity of rain (15-35 cm) that fell with the storm (Brooks 1939c). Trousdell et al. (1965) thought the same mechanism accounted for their values of over 99% uprooting in Virginia and North Carolina from a hurricane (see also Moore 1988). Versfeld (1980), studying the impact of two windstorms at the same site in South Africa, found both a lower proportion of uprooting and a lower overall damage level for the storm preceded by drier conditions. However, Versfeld (1980) ignored the possibility that the decreased overall damage could be due to the removal of the most windthrow-prone individuals; the wetter windstorm occurred only three months before the second, drier windstorm. Cremer et al. (1977) also supported their assessment that damage was amplified by saturated soils during a windstorm in Australia by comparing a subsequent storm at the same site and of similar severity that occurred two years later under drier conditions. Again, damage was less in the second, drier, storm. Topography - Many researchers have examined the role of topographic features in increasing or decreasing exposure to wind, or affecting wind direction, and thus influencing the severity of wind disturbance. Changes in wind direction over a landscape can be examined by mapping directions of treefalls (Boe 1965, Cremer et al. 1977, Falinski 1978, Glitzenstein & Harcombe 1988, Dallmeier et al. 1991, Boose et al. 1994). One would naturally expect damage to be highest on ridges and windward slopes and least in valleys and leeward slopes. In a study of 159 windthrow areas, Neustein (1971) found 45% on windward slopes, 32% on ridges, 6% on lee slopes, and 7% on flats. But the ensuing discussion will show that the relationship between topography and wind damage can be complex, sometimes leading, in fact, to greater damage in valleys and on lee slopes. Exposure is a complex relationship of aspect, slope, topographic position, and landscape position (Foster & Boose 1992), while damage itself can reflect topographic correlates of species characteristics. Table 2 summarizes the reported interactions between topographic features and damage. Many observers have reported increased damage on windward slopes and ridges (Curtis 1943, Smith 1946, Ruth & Yoder 1953, Stoeckeler & Arbogast 1955, Webb 1958, Alexander 1964, Neustein 1968, 1971, Liegel 1984). And numerous others have observed the expected decreased damage in sheltered valleys or on lee slopes (Andersen 1954, Webb 1958, Liegel 1984, Wunderle et al. 1992). But Bellingham et al. (1992) found decreased damage on ridge tops affected by a hurricane in Jamaica. They suggested that the ridge top canopy was more streamlined and that chronic winds in these exposed positions selected for the more wind- resistant species. Scatena and Lugo (in press), and Basnet et al. (1992), all investigating the same watershed in Puerto Rico, report less damage and mortality on ridge tops and upland valleys. They proposed that this difference was due to: 1) the predominance of wind resistant species on ridge tops - at least one of which (Dacryodes excelsa Vahl) exhibits root grafting, 2) drier soil conditions on ridge tops that lead to more developed root systems, and 3) increased damage in exposure valleys due to trees dropping down the slope. The results of two studies indicated that slope alone does not explain damage and recorded increased damage where slope gradient changes rapidly (Ruth & Yoder 1953, Alexander 1964). Valleys may provide protection from catastrophic wind (Table 2), but also may amplify disturbance. Liegel (1984) found trees leaning or blown down like "spokes in a wheel", in protected valley sites. Many others report increased damage in valleys, and saddles between ridges, where the wind is constricted and accelerated (Weidman 1920a, Brooks 1938, Curtis 1943, Smith 1946, Ruth & Yoder 1953, Andersen 1954, Alexander 1967, Chandler 1968, Gloyne 1968, Irvine 1970, Cremer et al. 1977, Lugo et al. 1983, Brennan 1991, Scatena & Lugo in press). Reporting on earlier hurricane damage in Puerto Rico, Weaver (1986, 1989) noted increased damage in some valleys, where saturated soils seemed to increase windthrow. An interesting debate in the literature regards damage to lee slopes. As indicated in Table 2, Smith and Weitknecht (1915) for windstorm damage in Oregon, Boe (1965) for windstorm damage in California, Alexander (1967) for windstorm damage in Colorado, Irvine (1970) for windstorm damage in New Zealand, O'Cinneide (1975) for gale damage in Ireland, and Brooks (1939a) for the 1938 hurricane in New England, all reported increased damage on lee slopes. However, Clapp (1938) and Smith (1946) reported no lee slope damage for the 1938 hurricane in New England. Foster (1988b) reported little lee slope damage from the same storm, possibly because the 1938 hurricane came from the south and east and the prevailing winds are from the north and west. In this situation the lee slopes would be more pre-conditioned to chronic wind and possibly less susceptible to the hurricane winds. This does not explain the lack of lee side damage in Scotland for the gale of 1968, which came from the west as do most gales (Neustein 1971). Brokaw and Grear (1991) and Boose et al. (1994) found that hurricane damage to high elevation (cloud forest) in Puerto Rico did not vary from windward to leeward slopes and suggested that, while the windward slopes were subjected to more intense winds, they were also preconditioned by chronic winds from the same direction. Table 2a - Interaction between topographic features and wind damage: temperate forests + increased damage, - decreased damage, 0 refers to no significant effect
Exposed Lee Percent Gradient Aspect Ridges Valley Slope Slope Slope Change Citation
+ - Conrad 1945 + - + + Andersen 1954 + Baker 1915 0 + + 0 Smith & Weitnecht 1915 + Canham 1978 + + + + + Alexander 1964 + Stoeckeler & Arbogast 1955 + + Neustein 1968 + + + + Day 1950 + + Alexander 1967 + Brennan 1991 + Chandler 1968 + Weidman 1920b + + - Clapp 1938 + + - Smith 1946 + + + Ruth & Yoder 1953 + + + - Foster 1988b + + Foster & Boose 1992 + + Smith & Weitknecht 1915 + O'Cinneide 1975 + Boe 1965 + + + +,- +,- Neustein 1971 + + Smith 1946 + + + - Curtis 1943 + + Irvine 1970 + Brooks 1939a, 1939b + Brooks 1938 + + + Shaw 1983
Table 2b - Interaction between topographic features and wind damage - tropical forests + increased damage, - decreased damage, 0 refers to no significant affect
Exposed Lee Percent Gradient Aspect Ridges Valley Slope Slope Slope Change Citation
- - Wunderle et al. 1992 + + - Liegel 1984 + + + Wadsworth & Englerth 1959 0 + - + Webb 1958 - + Basnet et al. 1992 - + Scatena & Lugo In press - + + Lugo & Scatena 1994 + - Frangi & Lugo 1991 + + - Boose et al. 1994 + + + Weaver 1986 + Weaver 1989 - + Bellingham et al. 1992 + + Reilly 1991 + Brokaw & Grear 1991 + Frangi & Lugo 1991
O'Cinneide (1975) proposed that greater lee side damage was the result of: 1) turbulent flow over the ridge, 2) soil differences on windward and lee slopes, 3) pre-conditioning on the windward side, and/or 4) winds from unusual directions (affecting trees not pre-conditioned to wind). Others propose that lee slope damage is influenced by the steepness of the slope. With steep enough slopes, the wind flows over the ridge and past the vegetation on the lee side. Boose et al. (1994) use an inflection angle of 6o (13%) for winds passing over a ridge, in modeling the exposure to hurricane winds. Ruth and Yoder (1953) found decreased damage with lee slopes of greater than 70%, Andersen (1954) found that lee slopes steeper than 8o (18%) were protected, and both Gloyne (1968) and Foster and Boose (1992) proposed that lee slopes of greater than 10% are protected. Boe (1965) noted that the presence of uncut timber on windward slopes tended to protect lee slopes. It is possible that these differences in protected lee slopes reflect differences in the intensity of the wind disturbance, but, of the above, only Andersen (1954) reports wind speed (144-160 km/hr). Some studies reported no effect of topography on damage. Alexander and Buell (1955) found no difference in direction of treefalls related to different aspects and topographic positions, but they did not report whether total damage varied with different exposures. Canham (1978) reported little influence of topography on blowdowns, but his study involved reconstructing historical records of blowdowns. Variation in directions of individual storms, and the downward component of winds in thunderstorms (considered the primary source of severe blowdown), will obscure variation in exposure to wind. Canham (1978) did observe that "local topography figures in the pattern from any given storm", and local variations in exposure can account for small patches of blowdown. Smith and Weitknecht's (1915) detailed survey of over 640 ha in Oregon struck by two windstorms lead them to conclude that local topography does not influence windthrow. However, no description of the topographic variation of the site is given, they only report summary data and conclusions, and no description of the directions of the two storms are given. Baker (1915) investigated 1400 ha in Washington and concluded that only very steep slopes had an impact on windthrow. Again, no detailed analysis of the data is presented, so it is difficult to determine what comparisons were made. Cremer et al. (1977) used aerial photographs to assess damage in over 70 km2 in hilly terrain of Australia struck by a single windstorm and concluded "topographic features did not appear to be important". No analysis was presented, but Cremer et al. (1977) seemed to expect a simple association with ridge tops and windward exposure. They report "stands were thrown on both lee and windward slopes. In several places more throw occurred in the valley ...". A last point regarding the influence of topography on wind and wind damage is that the relationships seem to become more complex, or less predictable, at finer spatial scales. The large scale influence of topography is evident where the windward sides of mountains were more damaged than protected leeward sides (Browne 1949, Lugo et al. 1983, Boose et al. 1994), but there is much less consistency in damage patterns at smaller scales of topography (Bellingham 1991, Brokaw & Grear 1991, Boose et al. 1994). Gloyne (1968) stated that wind is canalized by valleys at scales of tens of miles, but that the ability of the wind to follow smaller scale contours is a function of the degree of slope. Lawrence et al. (1991) proposed that large scale variation in damage is controlled by landscape features and fine scale variation is a function of individual response of stems. At fine spatial scales, the interactions between landscape features, vegetation, and the moving air masses make the relationship between topography and wind damage more than simply a matter of exposure differences due to aspect. Soil Characteristics - Some apparent topographic associations with damage may actually reflect edaphic characteristics that co-vary with topography (O'Cinneide 1975, Furley & Newey 1979, Basnet 1990), especially soil depth. Associations between edaphic conditions and damage severity seem to be indirectly biotic, the critical factor being the effects on root development and anchorage. Shallow soils, high water table, a shallow impermeable soil layer, or soil textures may all restrict root growth and increase susceptibility to damage. Neustein (1968) suggested that sites with high wind and turbulence, such as ridges, upper slopes and shoulders, also have shallower soils. Valleys may have deeper soils but still be more affected by the wind, presumably because the soils are more saturated, limiting root development (Basnet et al. 1992, Scatena & Lugo in press). Weaver (1989) reported more windthrow in valleys due to saturated soils; and Wunderle et al. (1992) found the most severe damage on a flood plain (and see Lugo et al. 1983, Reilly 1991). A number of researchers have reported associations between wind damage and soil characteristics, independent of topography (Table 3). Damage may be higher for species with shallower root systems (Touliatos & Roth 1971) due to shallow soil (Neustein 1968, 1971) or high soil moisture (Sauer 1962). Smith (1946) stated that damage is increased on moist, hard, or thin soils; any condition that prevents the deep penetration of tree roots. Smith and Weitknecht (1915) found decreased damage in medium and deep soils. Alexander (1964) found more damage on shallow soils with less drainage. Trousdell et al. (1965) found 51% damage on plots with a layer that restricts root growth and only 7% damage on plots without this restrictive layer. Liegel (1984) found increasing damage with higher average rainfall. Ruth and Yoder (1953) found damage associated with thin soils or high water tables. Both Bromley (1939) and Foster (1988a) found increasing damage with water- logged soil, associated with an impermeable soil layer, which resulted in shallow-rooted trees. However, both Wadsworth and Englerth (1959) and Versfeld (1980) found little relation between soil depth and damage. In the former study, they found several cases contrasting with those reported above: heavy uprooting in deep soils in bottom lands (where soil depth was related to high soil moisture). They found breakage only in shallow soils on hills where the root systems were well anchored in rock. Putz and Sharitz (1991) observed an exception to the common relationship between soil moisture and damage: uprooting was uncommon in South Carolina sloughs during a hurricane. They proposed that resistance of these trees was due to a unique, well-developed root system adapted to wet soils. Stoeckeler and Arbogast (1955) found less damage on peat, and suggested that wet peat enabled trees to tilt downward in the wind and spring back afterward. Mayer (1989) found more uprooting when storms hit during warm periods and soil was wet, and more stem breakage when wind hit trees in frozen soils and anchorage was presumably firmer. Andersen (1954) observed looser soil, and more uprooting during warm, wet winters. Table 3 - Soil characteristics and catastrophic wind damage. 0 no association, * association exists, - negative association, + positive association
Soil Soil Soil Hard Root Forest Depth Water Texture Pan Growth Type Citation
- Temperate Felt 1939 * + Temperate Trousell et al. 1965 - - Temperate Steven 1953b * Temperate Brewer & Merritt 1978 - * - Temperate Pickering 1986 * - Temperate Mergen 1954 - + * - Temperate Boe 1965 - + * + - Temperate Neustein 1968, 1971 - Temperate Touliatos & Roth 1971 - + + - Temperate Smith 1946 - Temperate Smith & Weitknecht 1915 - + Temperate Alexander 1964 + Temperate Liegel 1984 - + Temperate Ruth & Yoder 1953 + + Temperate Foster 1988a + + - Temperate Bromley 1939 - - Temperate Putz & Sharritz 1991 + * + Temperate Andersen 1954 - + + - Temperate Cremer et al. 1977 0 Temperate Versfeld 1980 * - Temperate Touliatos & Roth 1971 + Temperate Hook et al. 1991 - - Temperate Stoeckeler & Arbogast 1955 - + + - Temperate Day 1950 + Temperate Brewer & Merritt 1978 - + Tropical Walker et al. 1992 + + Tropical Wadsworth & Englerth 1959 + Tropical Sauer 1962 + - Tropical Basnet et al. 1992 + - Tropical Scatena & Lugo In press + Tropical Weaver 1989 * Tropical Webb 1958
The development of tree root systems is also influenced by soil texture (Andersen 1954, Webb 1958, Touliatos & Roth 1971). Brewer and Merritt (1978) and Pickering (1986) report damage on clay soils; the latter observed that these soils are poorly drained and may cause higher root mortality. Trousdell et al. (1965) reported 30% damage on coarse-textured soils and only 5% on medium to fine soils. Mergen (1954) found that soil texture affects the distribution of roots and soil consistency, influencing its anchoring capacity. Disturbance History - Previous disturbance may influence the severity of damage for a given intensity of disturbance. As mentioned above, previous disturbances - natural or anthropogenic - that open the canopy can increase turbulent air flow, resulting in a positive feedback to wind disturbance (Mergen 1954, Cruickshank et al. 1962, Boe 1965, Neustein 1968, Cremer et al. 1977, Somerville 1980, DeWalle 1983, Shaw 1983, Veblen et al. 1989, Allen 1992, Greene et al. 1992). Non- lethal damage associated with one windstorm may predispose trees to subsequent windthrow (Hubert 1918, Webster 1963, Putz & Sharitz 1991), particularly when the later storm winds come from another direction (Trousdell 1955). Over longer time spans, pioneer species that invade and grow to maturity following a disturbance may create a community that is relatively more susceptible to subsequent wind disturbance (Whitmore 1974, Veblen et al. 1989, Webb 1989, Nelson et al. 1994). However, Smith et al. (1994) hypothesize that gaps that existed before hurricane disturbance, possibly created by small lightning-strike fires, had lower-statured vegetation that survived the hurricane disturbance. Also in the case of successive disturbances by wind the feedback may not always be positive. Both Ruth and Yoder (1953) and Alexander (1967) found less damage over time, with successive wind disturbances. Allen (1992), in studying the effects of two windstorms in southeast Britain found the second storm, though of equal or greater intensity, to result in one quarter the damage. Foster (1988b) found that recent, severe, disturbance (logging, agriculture, fires) may leave stands of young, dense second growth that are wind resistant. Miscellaneous Abiotic Effects - There are other forms of damage associated with windstorms in addition to the mortality or structural damage caused by wind: salt stress, from spray or high tides, causing loss of foliage (Moss 1940, Wyman 1954b, Morgan et al. 1959, Sauer 1962, Gunter & Eleuterius 1971, Chabreck & Palisano 1973, Gardner et al. 1991, Hook et al. 1991, Applegate & Bragg 1992, Sheffield & Thompson 1992), structural damage from waves (Morgan et al. 1959, Sauer 1962, Gunter & Eleuterius 1971, Chabreck & Palmisano 1973, Stoneburner 1978, Graham 1990), sand, silt, and mud deposition (Craighead & Gilbert 1962, Chabreck & Palmisano 1973, Bush 1991, Smith et al. 1994), and damage due to microclimate changes, such as more light and wind, higher temperature, and less humidity (Fernandez & Fetcher 1991, Applegate & Bragg 1992, Turton 1992, Loope et al. 1994). Summary of Abiotic Effects - Wind damage is related to intensity of wind and topographic exposure, but the variation in exposure over the landscape is complex, and may result in valleys or lee slopes being subjected to greater intensity of wind. Exceptions to the relationship between exposure and damage occur when those positions most exposed to catastrophic wind are also those most pre-conditioned by chronic wind. The relationship between soil and wind damage relates to root growth, which may be restricted by shallow soils, high water table, a shallow impermeable soil layer, or soil texture. Previous disturbances may influence subsequent wind damage by: a) increasing turbulence by opening the canopy, b) selectively removing susceptible trees, and c) shifting the vegetation composition toward more wind-resistant species. Both biotic and abiotic factors influence severity of wind disturbance. Variations in spatial patterns of damage may be difficult to interpret when: 1) species respond differently to damage, 2) species distribution over the landscape also varies, 3) species distributions are influenced by abiotic factors that also influence the severity of wind disturbance. Are the spatial patterns of damage driven by patterns of species distribution, or are differences in damage to species the result of their distribution in areas not equally exposed to wind? Since most studies of catastrophic wind are reactive rather than proactive, and - to date - no study sites have been designed to anticipate wind disturbance and separate the influence of various factors, it is difficult to answer this question. Some recent efforts by Borchard et al. (1992) to separate the species versus environment factors may help unravel this complex problem of the relative role of different factors determining the impact of catastrophic wind disturbance. Dynamics of Recovery So far, this paper has focused on the factors influencing the pattern of severity of catastrophic wind disturbance, but the more crucial issue is the response of the forest ecosystem to this disturbance. I examine the dynamics of recovery in terms of: 1) comparisons with other disturbances, 2) paths of recovery, and mechanisms that affect these paths, and 3) interactions with subsequent disturbance events. COMPARISON TO OTHER DISTURBANCES - Catastrophic wind disturbance has been compared to a variety of other disturbances in terms of impact and implications for recovery. As mentioned above, Lugo et al. (1983) compared the impacts of a hurricane to an earthquake and a landslide and concluded that hurricanes are more damaging because they affect larger areas more often. However, their estimate of recovery time is much faster for hurricanes, indicating a fundamental difference in the nature of the disturbance or the path to recovery. Boucher (1990) suggested that a hurricane, a single discrete disturbance event, differs from anthropogenic disturbances, such as logging, agriculture, and conversion to pasture, because these latter are of longer duration and involve repeated removals of vegetation. Veblen et al. (1989) proposed that wind disturbance is more similar to insect outbreaks than to fire disturbance, because fire affects the understory and bares mineral soil. Ackerman et al. (1991) suggest that less severe hurricanes result only in defoliation and de-branching, moderate intensity hurricanes cause damage comparable to background treefall gap dynamics, and more severe hurricanes may produce gaps of a much larger size. Each could result in a different array of microclimates and therefore a different dynamic of recovery. Dallmeier et al. (1991) state that post-hurricane forests "cannot be compared to a forest gap", rather they are more comparable to forests that have been uniformly thinned. Spurr (1956) also suggests that hurricane disturbance is similar to logging, then describes some interactions between the two that control the recovery process, involving substrate disturbance and damage to advanced regeneration that survives the wind disturbance, indicating some differences in their impacts. Merrens and Peart (1992) report recovery from hurricane disturbance to follow the same succession as following clear cutting. Treefall associated with catastrophic wind potentially differs from background tree mortality and gap creation in three ways: the gap size distribution, the nature of the damage, and the ability of the system to respond. It is inappropriate to make a clear distinction between background treefalls and those associated with catastrophic wind. It is more appropriate to visualize a gradient of disturbance from trees that die standing and create only small openings in the canopy, through larger gaps created by single trees falling and perhaps knocking down other trees, to groups of treefalls struck by intense winds and sometimes creating gaps with indistinct edges when surrounding vegetation is defoliated. Catastrophic disturbance can create the largest gaps as well as the relatively small, single treefall size (Foster 1988a, Allen 1992, Foster & Boose 1992, Greene et al. 1992). The studies presented in Table 4 support the assumption that catastrophic wind disturbance creates a range of gap sizes, with three possible exceptions. In a study of tornado disturbance in Texas, Glitzenstein and Harcombe (1988) found no small gaps, but their goal was not to identify all gaps associated with the storm, but to focus on two areas of discrete, severe blowdown. Nor did Lorimer (1977) find small gaps, but he was restricted to a 25 ha minimum gap size in his study of surveyor's records, because smaller gaps had not been recorded. Nelson et al. (1994) shows a minimum gap size of 30 ha because of the limited resolution of the remote sensing technique used. Canham (1978) shows gaps down to 0.65 ha but also was limited by the resolution of his data; he estimates that the gaps actually ranged down to 0.004 ha. Foster and Boose's (1992) study of hurricane damage in Puerto Rico is probably the most comprehensive analysis of gap size distribution following catastrophic wind. Although they found gaps up to 37 ha, the distribution was strongly concentrated in the smaller size class, indicating that hurricane disturbance creates gaps of all sizes. Table 4 - Gap size associated with catastrophic wind damage (area measurements in hectares, intensity is wind speed km/hr)
Median or Minimum Maximum Average Intensity Citation
- 37 ha 0.04 200 Foster & Boose 1992 0.025 5 - - Allen 1992 - 0.0315 0.0013 - Ogden et al. 1991 0.002 0.009 0.025 147 Webb 1989 0.002 0.027 0.035 147 Webb 1989 0.65 3786 32 - Canham 1978 10.4 31.1 - 180-250 Glitzenstein & Harcombe 1988 25 - - - Lorimer 1977 30 3370 30-100 - Nelson et al. 1994
Along the gradient of gap sizes we should see shifts not only in scale but in type of damage, which differentially influence resulting microclimate and path of recovery. Trees that die in place and gradually disintegrate rarely involve uprooting (Foster 1988a), do not form large root mounds, and should make relatively small gaps. However, Putz et al. (1983) found no difference in gap size caused by snapped trees as versus uprooted trees, although they were not examining catastrophic wind impacts. Individual treefalls not caused by catastrophic wind produce mostly coarse woody litter and are not surrounded by defoliated vegetation. Catastrophic wind disturbance knocks down trees and debranches and defoliates surrounding vegetation, increasing light levels dramatically, (Veblen et al. 1989, Fernandez & Fetcher 1991, Turton 1992, Walker et al. 1992) and sometimes producing litterfall exceeding average yearly totals (Lodge et al. 1991, Whigham et al. 1991). SPATIAL INFLUENCES ON RECOVERY - Differing dynamics of recovery based on gap size are well established (Hartshorn 1978, Runkle 1981, 1984, 1985; Brokaw 1985a, 1985b), regardless of the gap creation mechanism. After catastrophic wind disturbance, Webb (1986) found that trees grow into small gaps from the sides, but in large gaps shrubs and previously established seedlings reform the canopy. Dunn et al. (1983) reported that encroachment of surrounding vegetation and advanced growth closed small gaps, while pioneer species filled in only the larger gaps. Putz and Brokaw (1989) found this same pattern; in large gaps pioneer species have increased importance. Sprouts grow faster initially, but cannot compete with pioneer species over time. Ogden et al. (1991) state that it is the ratio between gap size (horizontal area) and canopy height that determines whether growth from below (of released understory) or lateral growth from surrounding vegetation will fill the gap. High values of this ratio indicate large gaps where recovery from below also will include the establishment of new individuals and the gap is unlikely to fill in from the side. Even though vertical growth is twice that of lateral growth, the majority of the gaps in their study were closed by lateral growth (their largest gap, 315 m2, was relatively small). When catastrophic winds strike an area, the resources that promote recovery in background-level treefall gaps may be eliminated, and certain forms of subsequent disturbance are encouraged. If defoliation is extensive, fruits and seeds will also be blown down, regardless of their developmental maturity. Even though many species will flower in response to defoliation (see below), a temporal gap of seed availability is created. Without a supply of mature fruit, the usual seed dispersers, if they survived the disturbance event, will themselves disperse to other less disturbed areas (Boucher 1990, Yih et al. 1991). The entire system is less able to provide and disperse new plant propagules, and recovery must therefore depend on soil seed banks (if exposed from under the thick litter layer), or vegetative recovery. Large quantities of woody litter over the landscape increases both the possibility of catastrophic fire and the likelihood of insect population explosions. TEMPORAL INFLUENCES ON RECOVERY - Estimates of recovery time after catastrophic wind vary tremendously, as the result of: differences in quantification of recovery, frequencies of disturbance, and, possibly, variations in storm intensity. Unfortunately, recovery times have not been documented frequently enough to evaluate these interacting factors (Table 5). Most authors suggest that the recovery process is "rapid and widespread" (Applegate & Bragg 1992) since they observe new leaves within a few weeks (Webb 1958, Craighead & Gilbert 1962, Lugo et al. 1983, Hopkins & Graham 1987, Fernandez & Fetcher 1991, Frangi & Lugo 1991, Reilly 1991, Whigham et al. 1991, You & Petty 1991, Walker et al. 1992, Wunderle et al. 1992, Pimm et al. 1994), or a few months (Bates 1930, Spurr 1956, Bellingham 1991, Dallmeier et al. 1991, Walker 1991, Yih et al. 1991, Applegate & Bragg 1992, Bellingham et al. 1992). Table 5 - Catastrophic wind recovery time (years)
Time Location Disturbance Citation
.92 Australia cyclone Turton 1992 50a Dominica hurricane Lugo et al. 1983 30 Dominica hurricane Beard 1948 2-6a USA hurricane Van Hooser & Hedlund 1969 45 Puerto Rico hurricane Weaver 1986 10s-100sa Jamaica hurricane Bellingham et al. 1992 30-100a New England hurricane Clapp 1938 40 New England hurricane Peart et al. 1992 50 New England hurricane Merrens & Peart 1992
a - predicted recovery time Foster (1988b) suggests a 5-10 year reorganization period before the recovery mechanisms are fully operational. For example, Bates (1930) noted a two month delay in refoliation following the 1928 hurricane in Puerto Rico. Silver (1992) found that fine root biomass continued to decline nine months after Hurricane Hugo in Puerto Rico. Moss (1940) found significant levels of salt still in the soil six months after the 1938 hurricane in New England. The reorganization period may include delayed mortality, recorded by several researchers after catastrophic wind. It would be interesting, if this reorganization period does occur, to compare the time required to reorganize with the frequency of catastrophic wind events in a given area. Weaver (1989) proposes three stages in recovery after a hurricane : 1) colonization and growth - establishment and growth of pioneer species, 2) building - a shift from pioneer species to shade intolerant species, and 3) maturity - shade tolerant species reaching a self-sustaining steady state of standing biomass. Weaver's (1989) data indicates a shift from the first to the second stage starting at 20 years after the disturbance. Unfortunately, his data do not include the dynamics of the first decade following the hurricane, so a possible reorganization period was not evaluated. Differences in measured recovery time partly are caused by differences in definitions and quantification of recovery. Turton (1992) states that light levels in the understory return to pre-cyclone levels after 11 months, but not that recovery is complete. Van Hooser and Hedlund (1969) estimate recovery time for the volume lost in a storm "if growth were to continue at the same rate as before the storm", a condition they acknowledge as "highly unlikely". Beard (1948) seemed to rely on the visual appearance of forest structure to evaluate recovery. Peart et al. (1992) also described post-storm recovery based on the structure of the canopy, specifically gap structure. Merrens and Peart (1992) report slightly longer recovery time for the same storm when comparing standing biomass. Weaver (1986) quantified recovery based on stands reaching an apparent steady state in biomass. Bellingham et al. (1992) predicted recovery times from Hurricane Gilbert in Jamaica based on the path of recovery: centuries if pioneer species establish, and decades if the recovery is through sprouting of surviving stems. To date, too few studies have actually documented the entire recovery process to draw any conclusions about factors that influence the time to recovery (Crow 1980). Frequent catastrophic wind disturbance has lead a number of authors to rethink the concept of recovery. Webb (1958) states that the return time of cyclones to north Queensland is much shorter than the life span of trees and proposed that these forest never reach a climax state but are in a constant state of recovery. Others emphasize the role of allogenic influences, as opposed to autogenic compositional changes, in creating new age classes (Henry & Swan 1974, Oliver & Stevens 1977, Foster & Boose 1992). Canham and Loucks (1984) propose that forests in Wisconsin may reach old-growth, but not for long before they are again disturbed. Several workers propose that ecosystems in areas of high hurricane frequency will not recover to a steady state, but are always in a state of disturbance and recovery (Lugo et al. 1983, Putz & Chan 1986, Basnet 1990, Boucher 1990, Clark 1990, Basnet et al. 1992, Lugo & Scatena 1994). In this situation, the concept of 'recovery time' is inappropriate. As mentioned above, the majority of studies of catastrophic wind have quantified only the impacts and the immediate response. But several studies have examined forests in the latter stages of recovery, following, for example: the New England hurricane of 1938 (Foster 1988a, 1988b, Foster & Boose 1992, Merrens & Peart 1992, Peart et al. 1992), a severe wind in Malaysia (Browne 1949, Wyatt-Smith 1954), and hurricanes in Puerto Rico (Crow 1980, Weaver 1986, 1987, 1989). However, these studies all fail to document the critical first stage of the dynamics of recovery. Probably only Whitmore's (1974, 1989b) plots in the Solomon Islands were established before a catastrophic wind disturbance and then followed for at least a decade thereafter. Several sites are now poised to follow recovery through several disturbance cycles on established plots, in: Jamaica (Bellingham 1991), Nicaragua (Boucher et al. 1990), Mexico (Whigham et al. 1991), and Puerto Rico (Dallmeier et al. 1991, Zimmerman et al. 1994). PATHS TO RECOVERY - Recovery from catastrophic wind disturbance might be expected to follow one or more of four paths: regrowth, release, recruitment, or repression (Figure 13). The critical question is "will the forests of 10-20 years time be dominated by regrowth of damaged trees or will saplings and seedlings 'released' by disturbance play a significant role?" (Sugden 1992). In other words, will there be a shift in the community structure during recovery? Both Oliver (1981) and Halpern (1988) stress the importance of disturbance severity in influencing response. As suggested in Figure 12, control of the path of recovery may be a function of both the type and severity of damage. Other factors may influence recovery, including: edaphic conditions, topography, species difference in response, and previous disturbance (Spurr 1956, Furley & Newey 1979, Glitzenstein & Harcombe 1983, Webb 1986, Brokaw & Walker 1991, Peterson & Pickett 1991, Tanner et al. 1991). Figure 13 - Paths to recovery from catastrophic wind disturbance. See text for description of recovery paths Regrowth - Regrowth refers to vegetative recovery of surviving stems. The prevalence of this type of recovery depends on the nature and extent of structural damage, e.g., uprooted versus snapped stems, and on mortality. The importance of regrowth should decline as mortality increases. As most catastrophic wind disturbances cause low to moderate mortality, this recovery path may well be the most common. Of the 47 instances reported in Table 6, 66% include regrowth as one observed path to recovery; a higher percentage (85.2% n=27) is reported in tropical as compared to temperate forests (40% n=20). Boucher (1989) introduced the phrase "direct regeneration", to describe recovery through sprouting. This recovery path has been documented and referred to in a variety of ways: Sauer (1962) described it as "vegetative advance of adjacent survivors", Dunn et al. (1983) referred to "encroachment of canopy trees", Pickering (1986) used the phrase "sprout regeneration", Walker (1991) called it "direct species recovery", and Marks (1974) presented sprouting as one form of "reorganization response". Regrowth can shift tree community structure, through a differential ability of species to sprout (Walker et al. 1992, Bellingham et al. 1994, Zimmerman et al. 1994). Shade tolerant species seem both to survive and sprout at higher frequencies than pioneer species. Rather than result in a shift to shade intolerant species, a catastrophic wind event may increase the dominance of established primary forest trees (Wyatt-Smith 1954, Dallmeier et al. 1991, Putz & Sharitz 1991, Scatena & Lugo in press). Table 6a - Dynamics of recovery from wind disturbance in temperate forests. Reg: regeneration, Rec: recruitment, Rel: release, Rep: repression
Intensity Mortality Damagec Sprouting Reg Rec Rel Rep Citation km/hr
15.3 7.4 + + Greene et al. 1992 91 + + + + Pickering 1986 90 7.1 + + Cremer et al. 1977 120 60 + Peart et al. 1992 120 + Smith 1946 120 + Henry & Swan 1974 120 + + + Merrens & Peart 1992 126 0.3 0.6 + - + Webb 1986, 1988, 1989 126 5.6 6.4 + Webb 1986, 1988, 1989 142 66 + Veblen et al. 1989 155 19 + - Putz & Sharitz 1991 155 48.7 + - Putz & Sharitz 1991 140-180 20 + + Craighead & Gilbert 1962 200 + + Spurr 1956 200 30.7 94 + + Foster 1988a 180-250 34 83 22 + Glitzenstein & Harcombe 1988 180-250 14 37 75 + Glitzenstein & Harcombe 1988 242 31 + + Loope et al. 1994 253 94 + + + + Dunn et al. 1983 430 93 99 24.9 - + Peterson & Pickett 1991
Table 6b - Dynamics of recovery from wind disturbance in tropical forests. Reg: regeneration, Rec: recruitment, Rel: release, Rep: repression
Intensity Mortality Damagec Sprouting Reg Rec Rel Rep Citation km/hr
+ Lindo 1968 + + Furley & Newey 1979 + + Weaver 1989 + Wyatt-Smith 1954 + + + Browne 1949 120 30 + Wadsworth & Englerth 1959 147 + + + Webb 1958 150 + + Wood 1970 166 1 10 27.5 + Frangi & Lugo 1991 176 32.8 + + Applegate & Bragg 1992 184 36.9 + + + + Whitmore 1974 185 40 32 + Dittus 1985 166-194 50 + Scatena & Lugo In press 166-194 9 16.7 64.8 + + Zimmerman et al. 1994 166-194 1 4.8 37.5 + + + Frangi & Lugo 1991 166-194 7 20 + + Walker 1991 166-194 + + Fernandez & Fetcher 1991 166-194 4 100 + + You & Petty 1991a 166-194 25 38.2 + Dallmeier et al. 1991 240 25 + + + Sauer 1962 240 + + Bates 1930 241 2 57.5 + + + Lugo et al. 1983 250 58 75 83.6 + + Boucher 1990 250 13 50 5 + Boucher 1990 255 7 55.3 56 + + Bellingham et al. 1992 255 8.1 14.3 59.3 + Bellingham 1991 300 9.7b 16.9 + Whigham et al. 1991
a single species studied, b mortality before fire, c damage calculated as percent loss of biomass, volume, or stems Simple leaf loss may not be especially damaging. Craighead and Gilbert (1962) found that deciduous trees refoliated following hurricane disturbance late in the year and retained their leaves well past the normal time for leaf loss. Putz and Sharitz (1991) reported that defoliation in South Carolina by Hurricane Hugo was not particularly damaging to the deciduous trees since it occurred close to the usual time of leaf loss. Recruitment - Recruitment refers to recovery of the forest through the establishment of seedlings of early successional species. This does not necessarily refer to classical models of secondary succession. Merrens and Peart (1992) suggest a "shortened" succession following the 1938 hurricane in New England, with short-lived pioneer species established briefly, immediately followed by the release of primary forest species also established at the time of the hurricane. Foster (1988b) suggests that shade intolerant species are established through canopy disturbance, then suppressed until released by the next disturbance. Recruitment may occur in localized parts of a wind affected forest, but appears to be relatively rare as the principal path of recovery after catastrophic wind, possibly because of the generally low levels of mortality in this type of disturbance. Yih et al. (1991) noted the paucity of pioneer species seedlings and suggested several possible causes: the lack of seed dispersers, thick litter, and inhibition by the root systems of surviving stems. However, the seedlings of shade tolerant species were found, so the explanation may simply have been the rarity of mature pioneer trees in the area before the hurricane (Yih et al. 1991). The dual damage parameter presented in Figure 12 may help predict the severity of damage required for recruitment. Recruitment would be expected after at least low to moderate mortality and/or moderate to high structural damage. The thick litter layer associated with high levels of structural damage might restrict recruitment (Yih et al. 1991). Seedlings numbers increased when litter was removed from experimental plots in a forest after hurricane disturbance (Guzman-Grajales & Walker 1991). However, Pickering (1986) still found recruitment as the primary path to recovery (followed by release and regrowth) in a tornado impacted forest with high structural damage (91%). It is possible that the two-dimensional damage gradient space (Figure 12) misses a critical factor that influences recruitment; disturbance of the substrate. Both Sauer (1962) and Stoneburner (1978) mention ocean waves exposing bare soil as a requirement for recruitment. Spurr (1956) found an association between post-hurricane salvage logging and invasion of pioneer species, which may be tied to soil disturbance. Following wind disturbance fire bares the substrate and can stimulate establishment of pioneer species (Smith 1946, Canham 1978, Furley & Newey 1979, Foster 1988b, Veblen et al. 1989, Putz & Sharitz 1991). Foster (1988b) suggests that the difference between the 1635 hurricane, when pines were established, and the 1938 hurricane when they were not, was due to the fires that followed the former and not the latter. Snook (1993) hypothesized that much of the mahogany (Swietenia macrophylla King) in the Yucatan becomes established when fires follow hurricane damage. However, Whigham et al. (1991) reported on a fire that followed Hurricane Gilbert in the Yucatan and did not find significant recruitment of pioneer species. Release - Release refers to the rapid growth of suppressed subcanopy trees or saplings. These may either be immature individuals of the dominant canopy species, or they may the vanguard of a significant shift in species dominance. This type of recovery is triggered by both structural damage and mortality to the canopy trees. Spurr (1956) observed stands 10 years after hurricane disturbance and found the canopy to be dominated by species prevalent in the understory at the time of the hurricane. Only in places where the seedling or sapling density was low when the hurricane hit did recruitment of pioneer species predominate. Thus, the tendency for recovery to include the release of the understory, and a possible shift in community composition, is partly tied to stand history. Glitzenstein and Harcombe (1988) found a shift to dominance of shade tolerant understory stems when the canopy had been dominated by large early successional species that were removed by a tornado. Veblen et al. (1989) found that a windstorm in Colorado removed the pioneer species (Pinus contorta Dougl.) established after a fire and released the shade tolerant fir and spruce from the understory. Repression - Repression refers to secondary succession that is suppressed by the establishment or growth of plants, often vines (Webb 1958, Wadsworth & Englerth 1959, Lindo 1968, Wood 1970, Whitmore 1974, Applegate & Bragg 1992), that restrict the regrowth or recruitment of trees, or succession that is suppressed by heavy litter. Former canopy vines may be blown to the ground but not killed and then are poised to dominate recovery. The lack of availability of tree seeds following extensive defoliation may allow dominance of herbs or shrubs that were not defoliated by the wind due to their protected position under the canopy. Some edaphic conditions, particularly drought that has followed many catastrophic wind disturbances (Holtam 1971, Touliatos & Roth 1971, Hopkins & Graham 1987, Applegate & Bragg 1992), may adversely affect survival of tree seedlings relative to herbaceous species. No investigation of this possibility has been reported. Heavy structural damage is associated with both high light levels and heavy litter and therefore may encourage or, conversely, repress establishment of early successional species (Whitmore 1974, Putz & Sharitz 1991). Whitmore (1974) suggested that indirect damage to seedlings and saplings that are crushed by falling debris may restrict recovery through release of previously suppressed plants. Yih et al. (1991) suggested that the thick layer of material knocked down by Hurricane Joan was one reason that establishment of pioneer species was so rare in Nicaragua. Canham (1978) suggested that pioneers seldom invade following wind disturbance unless fire, that removes litter, also occurs. Others have discussed the possible effects of wind disturbance and subsequent fire (Smith 1946, Furley & Newey 1979, Foster 1988b, Veblen et al. 1989, Snook 1993, Loope et al. 1994). The occurrence of these four paths of recovery is presented in Table 6, but too few studies have followed the recovery process long enough for us to develop strong generalizations. For example, Boucher (1990) emphasizes the predominance of regrowth (sprouting) in a rain forest in Nicaragua after a severe hurricane, but notes that mortality of sprouts, and additional delayed mortality of surviving trees could still change the mode of recovery. Dallmeier et al. (1991) state: "most previous studies of regeneration after disturbance do not help us to predict the path of succession" (for a hurricane-disturbed site in Puerto Rico). Peterson and Pickett (1991) state the need for "information from other wind disturbed-forests before we can generalize on the relative contribution of seed banks, surviving seedlings and saplings, colonizing seeds and sprouts to post-disturbance forest structure and before we can relate these sources of regeneration to underlying environmental gradients and disturbance intensities". One point can be made: because of low mortality, the classical secondary succession stage of complete establishment of pioneer species may never follow wind disturbance. MECHANISMS OF RECOVERY - A variety of species adaptations facilitate recovery from wind disturbance, including: 1) resistance to wind damage, 2) ability to sprout, and 3) ability to flower and fruit quickly. The first, resistance to wind, has been covered in some detail above. Species differ in their: susceptibility to damage; tendency toward types of damage such as debranching, uprooting, or snapping; and resulting mortalities. Dunn et al. (1983) state that few species have adaptations specifically for catastrophic disturbance, but, "some specializations that allow species to coexist in non-equilibrium communities dominated by small scale disturbance also pre-adapt them to large-scale disturbance". Sprouting - A number of researchers have reported differences among species in sprouting ability (Glitzenstein & Harcombe 1988, Putz & Brokaw 1989, Peterson & Pickett 1991, Walker et al. 1992, Bellingham et al. 1994, Zimmerman et al. 1994). Groups that seem to be poor at sprouting are pines (Boucher et al. 1990), palms (Frangi & Lugo 1991), and pioneers (Webb 1986, Putz & Brokaw 1989, Zimmerman et al. 1994). Several workers have noted connections between sprouting, damage level, and tree size: a higher frequency of sprouting with broken stems than with uprooted stems (Putz et al. 1983), increased incidence of sprouting with more branch damage and with broken stems (Zimmerman et al. 1994), more sprouting with broken stems or complete defoliation (Bellingham et al. 1994). The latter also found that larger undamaged stems and smaller damaged stems sprouted more. However, Harcombe and Marks (1983) determined that sprouting was not associated with higher damage; trees were either highly damaged and died, or had low to moderate damage and sprouted. A generalization emerging (Table 7 and see Glitzenstein & Harcombe 1988) is that overall sprouting following catastrophic wind disturbance is positively correlated to available water. Survival of sprouts varies across sites: 40.6% in gaps in Panama, (Putz & Brokaw 1989), 68% for a tornado impacted forest in Pennsylvania (Peterson & Pickett 1991), and 74% for gaps in Texas (Harcombe & Marks 1983). This small sample of studies does not allow us to generalize about the environmental factors that influence sprout survival. But stems with sprouts survived better than stems without them in Bellingham et al.'s (1994) study of hurricane damage. Among sprouting stems in treefall gaps, Putz and Brokaw (1989) found lower survival in larger than smaller stems. Despite sometimes high mortality of sprouts and sprouting stems, sprouting may still play an important role in recovery. After 12 years of recovery in a Panamanian forest sprouting was present in 38 of 46 gaps studied (Putz & Brokaw 1989), while pioneer trees were restricted to larger treefall gaps. In the large gaps sprouts grow faster initially, but cannot out-compete pioneer species over time. Table 7 - Sprouting response to catastrophic wind (intensity is measured as wind speed in km/hr)
Sprouting Moisture Intensity Region Citation percent
24 mesic - temperate Harcombe & Marks 1983 22.4 mesic 180-250 temperate Glitzenstien & Harcombe 1988 72.3 hydric 180-250 temperate Glitzenstein & Harcombe 1988 24.9 mesic 430 temperate Peterson & Pickett 1991 53 hydric - tropical Putz & Brokaw 1989 59.3 hydric 160-185 tropical Bellingham 1991 56 hydric 160-185 tropical Bellingham et al. 1992 61.4 hydric 160-185 tropical Bellingham et al. 1994 64.8 hydric 166-194 tropical Zimmerman et al. 1994 27.5 hydric 166-194 tropical Frangi & Lugo 1991 83.6 hydric 250 tropical Boucher et al. 1990 5 hydric 250 tropical Boucher et al. 1990
Flowering and Fruiting - Numerous researchers have noted unseasonal or abnormally intense flowering following catastrophic wind disturbance: a hurricane in Puerto Rico (Bates 1930), a hurricane in Florida (Craighead & Gilbert 1962), a hurricane in Dominica (Lugo et al. 1983), cyclones in Mauritius (King 1945), and cyclones in Australia (Webb 1958, Hopkins & Graham 1987, Applegate & Bragg 1992). All reported cases involved relatively intense disturbances. Either less intense wind disturbance does not elicit this response, or researchers in those cases did not look for it. Defoliation is thought to stimulate the flowering response (Vaughan & Wiehe 1937, King 1945, Webb 1958). Increased flowering would produce more fruiting. In Puerto Rico Walker and Neris (1993) found post-hurricane seedfall in gaps to be double that of pre-hurricane levels, and intact forest seedfall to be triple that of pre-hurricane levels. During their 20 month post-hurricane study they did not find an increase over time, and they reported that the spatial heterogeneity was greater than the temporal. Of course, immediately after a catastrophic storm that strips trees of flowers and fruits one would expect reduced fruiting. Thus several workers have noted decreased fruit abundance immediately after catastrophic wind disturbance (Synder & Synder 1979, Zamore 1981, Lynch 1991, Wiley & Wunderle 1994). Immediately after a cyclone hit a Sri Lankan forest, fruiting and flowering was significantly less than before, but then fruiting seemingly increased (not significantly) in later years to a higher level than before the disturbance (Dittus 1985). Damage Type - The form of stem damage, specifically uprooting versus stem breakage, can influence the pathway of recovery. Schaetzl et al. (1989a) use the term "floralturbation" for the mixing of soil associated with uprooting, which creates microsites for germination of pioneer species (Foster 1988a). Consequently, pioneer trees were more common in gaps caused by uprooting than in those caused by snapped stems in a Panamanian forest (Putz 1983). Frequency of sprouting may also vary with disturbance type. Putz et al. (1983) found higher sprouting in snapped stems (51%) than in uprooted stems (6%). Walker (1991) found the immediate and delayed mortality of snapped stems to be lower (15%) than uprooted stems (23%). Since the number of stems uprooting affects sprouting, survival of damaged stems, and establishment of pioneer species, it has been suggested as a measure of the severity of disturbance (Peterson & Pickett 1991). Table 8 summarizes the conditions associated with uprooting of stems reported in 39 papers (19 tropical, 20 temperate sites). The reported percentages are the proportion of total stems damaged that were uprooted rather than snapped. Cruickshank et al. (1962) suggested that stem breakage increases with intensity of wind disturbance, but this trend is not evident in either the tropical or temperate sites where low proportions of uprooting (and therefore high stem breakage) occurs with both chronic winds (23%, Putz et al. 1983; 0%, Harcombe & Marks 1983) and intense (>300 km/hr) winds (10%, Gunter & Eleuterius 1971; 25.8%, Whigham et al. 1991). Allen (1992) reported more uprooting in moist forests and swamps and in connection with storms accompanied by high rainfall. This trend seems evident for tropical sites, where uprooting does not exceed 26% for either xeric or mesic sites. No clear trend relative to site or storm moisture occurs in temperate sites. Table 8a - Percent uprooting in response to wind in temperate forests.
Percent Wind Forest Moisture Citation Uproot Intensity Type (km/hr)
0 chronic mixed mesic Harcombe & Marks 1983 25 chronic conifer mesic Gordon 1973 18 chronic conifer mesic Veblen 1986 30 chronic conifer mesic Veblen 1986 22 chronic broadleaf mesic Runkle 1982 87 chronic mixed mesic Falinski 1978 81.2 60- 75 conifer mesic Versfeld 1980 72.4 broadleaf mesic Reiners & Reiners 1965 84 broadleaf hydric Reiners & Reiners 1965 58 80-120 conifer hydric Stoeckeler & Arbogast 1955 7 80-120 broadleaf mesic Stoeckeler & Arbogast 1955a 38 80-120 conifer mesic Stoeckeler & Arbogast 1955a 63 80-120 conifer mesic Stoeckeler & Arbogast 1955a 20 96-120 conifer mesic Webb 1988 63 96-120 conifer mesic Webb 1988 99 144 conifer mesic Trousell et al. 1965 37 155 mixed hydric Putz & Sharitz 1991 45 155 mixed mesic Putz & Sharitz 1991 90 160 mixed mesic Cruickshank et al. 1962 30 160 mixed mesic Cruickshank et al. 1962 93 200c conifer mesic Foster 1988a 11 200c conifer mesic Foster 1988a 89 200c mixed mesic Foster 1988a 88 200c mixed mesic Foster 1988a 95 200c mixed mesic Foster 1988a 82 221 conifer mesic Boe 1965 86 221 broadleaf mesic Boe 1965 24 180-250 mixed mesic Glitzenstein & Harcombe 1988 48 180-250 broadleaf hydric Glitzenstein & Harcombe 1988 48 194-235 conifer mesic Hook et al. 1991 33 194-235 conifer mesic Hook et al. 1991 25-33 242 conifer mesic Loope et al. 1994 37.4 320 conifer mesic Van Hooser & Hedlund 1969b 56.9 320 broadleaf mesic Van Hooser & Hedlund 1969b 10 349 mixed mesic Gunter & Eleuterius 1971 66.5 430 conifer mesic Peterson & Pickett 1991
Table 8b - Percent uprooting in response to wind in tropical forests
Percent Wind Forest Moisture Citation Uproot Intensity Type (km/hr)
23 chronic broadleaf mesic Putz et al. 1983 80 chronic broadleaf hydric Uhl et al. 1988 90 chronic broadleaf hydric Hartshorn 1980 40 chronic broadleaf hydric Putz & Appanah 1987 51 chronic broadleaf hydric Lawton & Putz 1988 14.6 150 broadleaf mesic Wood 1970 51 166 broadleaf hydric Walker et al. 1992 43 166 broadleaf hydric Walker et al. 1992 55.6 166 palm hydric Frangi & Lugo 1991 45 166 broadleaf hydric Walker 1991 47.9 166-194 broadleaf hydric Zimmerman et al. 1994 17.8 220 mixed hydric Thompson 1983 78.5 212 broadleaf hydric Dallmeier et al. 1991 87.8 230 broadleaf hydric Basnet et al. 1992 58.3 241 broadleaf hydric Lugo et al. 1983 20 241 broadleaf xeric Lugo et al. 1983 29 250 broadleaf hydric Yih et al. 1991 62 250 conifer hydric Boucher et al. 1990 42 250 broadleaf hydric Boucher et al. 1990 60.6 205-255 broadleaf hydric Bellingham 1991 40.6 255 broadleaf hydric Bellingham 1991 26 255 broadleaf hydric Bellingham et al. 1992 28 255 broadleaf hydric Bellingham et al. 1992 41 255 broadleaf hydric Bellingham et al. 1992 66 255 broadleaf hydric Bellingham et al. 1992 53 255 broadleaf hydric Bellingham et al. 1992 48 255 broadleaf hydric Bellingham et al. 1992 25.8 300 broadleaf mesic Whigham et al. 1991
a single species, b average of multiple plots, c heavy rains accompanied storm Multiple Disturbance Interactions - Wind disturbance may render a forest more susceptible to other disturbances (Glitzenstein & Harcombe 1988, Ogden et al. 1991) that alter recovery. Wind disturbance may interact with fire, drought, insect invasion, anthropogenic activities, and subsequent wind disturbance. I have already shown that wind disturbance may increase or decrease susceptibility of a stand to subsequent wind events (see Biotic Factors: Stand Conditions; and Abiotic Factors: Disturbance History). Because catastrophic wind disturbance creates large fuel loads, fire is the most commonly predicted subsequent disturbance (Clapp 1938, Lindo 1968, Glitzenstein & Harcombe 1988, Gardner et al. 1991, Hook et al. 1991, Putz & Sharitz 1991). Glitzenstein and Harcombe (1988) and Hook et al. (1991) calculated the increased fuel load resulting from catastrophic wind events; the former found that a tornado increased fuel load 400% in a Texas forest, and the latter found eight times the normal fuel load after Hurricane Hugo struck forests in South Carolina. However, just a few cases of fire after catastrophic wind have been reported. Craighead and Gilbert (1962) reported that extensive fires followed the 1935 hurricane in Florida. Whigham et al. (1991) estimated that post-hurricane fires in the Yucatan had more impact on mortality of trees than did Hurricane Gilbert (and see Furley & Newey 1979). Fire after wind disturbance has a greater likelihood of significantly changing community composition than does wind alone. Windthrow and fire together increase the: severity of disturbance, establishment of early successional species, likely development of large, even-aged stands, and time for recovery (Wyatt-Smith 1954, Webb 1958, Canham 1978, Canham & Loucks 1984, Glitzenstein & Harcombe 1988, Veblen et al. 1989). Fires do not always follow wind disturbance. After gale damage in Britain there were no fires even though summers were unusually dry (Holtam 1971); fire risk, however, was reduced by timber salvage. Furley and Newey (1979) noted fires in Belize following a hurricane disturbance in 1961, but an earlier study (Wolffsohn 1967) reported that the fires were caused by an extremely dry year. In comparing hurricane disturbed areas to those protected from the wind, Wolffsohn (1967) found fire no more common or likely to spread in the former. However, timber downed by the hurricane did make access and fire control more difficult. Periods of low rainfall following catastrophic wind may increase the possibility of fire, and may limit the establishment and survival of recruiting or regrowing vegetation. A drought after Hurricane Camille struck Mississippi increased subsequent mortality (Touliatos & Roth 1971), and moisture stress delayed refoliation when drought followed a cyclone in Australia (Applegate & Bragg 1992). The woody debris brought down by wind disturbance also creates conditions for explosive growth of insect populations, such as wood-boring beetles, engraving beetles, weevils, and termites. Numerous researchers have noted the possibility of increasing insect populations following catastrophic wind disturbance (Steven 1953a, Nelson & Stanley 1959, Decker et al. 1962, Gardner et al. 1991), and actual increases have often been reported (Bates 1930, Gunter & Eleuterius 1971, Touliatos & Roth 1971, Luitjes 1977, Wilkinson et al. 1978). Fewer studies have documented commensurate damage by insects to vegetation (Fuhrer & Kerck 1978, Hook et al. 1991). Wilkinson et al. (1978) stated there was no evidence of insects spreading to adjacent vegetation after catastrophic wind in the southeast United States. Thompson (1983) reported that insect populations did not increase as expected after a hurricane in Jamaica and offered three explanations: 1) an unusually wet season, 2) too rapid decomposition, or 3) plentiful food on the forest floor that diverted invasion of living trees. Downed timber is sometimes salvaged after catastrophic wind disturbance. This human disturbance may reduce the risk of fire, as was the case following gale damage in Scotland (Holtam 1971). Soil disturbance by salvage logging, may, like fire, increase available sites for pioneer establishment. Pickering (1986) found twice the number of seedlings in tornado-disturbed sites that were salvage logged than in disturbed sites that were not logged. Spurr (1956) described the interaction between these two disturbances in three scenarios after a hurricane: 1) if advance regeneration exists in the understory and no salvage logging is done, the recovery will be an "accelerated succession" (release), 2) if advanced regeneration exists and salvage logging occurs, advanced regeneration will be mixed with pioneer species (recruitment), and 3) if no advance regeneration exists, pioneers will establish regardless of human activity. Temperate and Tropical Differences The frequency of disturbance from catastrophic wind relates to position on the globe and this difference in frequency may result in differences in the effects of and recovery from wind disturbance. Temperate and tropical systems may vary in: 1) disturbance frequency or 'return time', 2) type and severity of damage, 3) recovery time, and 4) path to recovery. Calculations of return time (Table 9) are hard to compare. They may indicate the average time between disturbances at a given point or for a region, or they may be 'turnover time', calculated by dividing the total area of interest by the amount of area affected per unit time by disturbances. Taking another approach, Barker and Miller (1990) looked at the number of hurricanes that struck Jamaica each year for the last century and calculated the probabilities for a given number to occur in a year (0.45 probability of more than one hurricane per year). It is also difficult to compare return times for storms of contrasting intensity. With these provisos in mind, it appears that return times for the tropics are generally shorter than for temperate regions, and tropical storms tend to be more intense events, cyclones and hurricanes. Where return times in the temperate zone are short these often involve less intense windstorms. Table 9 - Frequency of catastrophic wind disturbance
Return Time Based on Location Citation (years)
TEMPERATE
2 windstorm OR, USA Weidman 1920b 5.8 hurricane SC, USA Gresham et al. 1991 10-12 windstorm Australia Cremer et al. 1977 20-25 80 km/hr WI, USA Stoeckler & Arbogast 1955 34 gale Ireland Gallagher 1974 50 25 m/s Britain Allen 1992 75 gale Scotland Fraser 1971 70-100 hurricane MA, USA Foster 1988b 10 windstorm New Zealand Shaw 1983 100-150 windstorm Britian Moore 1988 100-280 soil turnover New Zealand Wardle 1970 220-400 turnover New Zealand Ogden et al. 1991 220-2439 soil turnover MI, USA Whitney 1986 1150 > 25 ha gaps ME, USA Lorimer 1977 1210 windstorm WI, USA Canham & Loucks 1984 1000-3000 windstorm NY, USA Seiscab & Orwig 1991 1316-2632 soil turnover MI, USA Whitney 1986 1587-3174 soil turnover MI, USA Whitney 1986 1730 Soil turnover IL, USA Schaetzl et al. 1989a 12500-25000 soil turnover MI, USA Whitney 1986
TROPICAL
1 cyclone Mauritius Vaughan & Wiehe 1937 11 cyclone Mauritius Baur 1964 1.2 hurricane Jamaica Bellingham 1991 1.25 cyclone Solomons Whitmore 1974 7.2 cyclone Queensland Baur 1964 3-40 cyclone Queensland Webb 1958 10 hurricane Puerto Rico Baur 1964 13 hurricane Belize Lindo 1968 6-20 cyclone Fiji Gane 1970 18 windstorm Sri Lanka Dittus 1985 20 hurricnae St. Vincent Beard 1945 24 cyclone Sri Lanka Dittus 1985 225-300 turnover Sri Lanka Dittus 1985
Species characteristics differ among these regions of the globe. Comparing all tropical species to all temperate species in Appendix F-III, the temperate sites have a greater proportion of susceptible species (53.7% n=95) than the tropical sites (40.2% n=164). The reverse is of course true for the resistant species. Temperate sites tend to have a greater proportion of uprooting (51.8% n=36, compared to 47.3% n=28; Table 8), and a corresponding lower average percentage of sprouting (35.9% n=4, compared to 51.5% n=8; Table 7). The increased sprouting at tropical sites may also relate to average differences in water availability. Putz et al. (1983) related the increased incidence of snapping in the tropics to a greater incidence of less dense wood. There appear to be latitudinal differences in the factors that influence severity of damage, but these may reflect the focus of investigators. In Table 3, temperate forest researchers have identified soil depth, the presence of an impermeable soil layer, and restricted root growth as important factors much more often than researchers at tropical sites. Soil water has been identified as an important factor more often in the tropics. Differences also exist in the perceived role of topography. Only tropical researchers have noted decreased damage on exposed ridges, and only temperate researchers have reported on the role of lee slopes in either increasing or decreasing damage. Measures of rate of recovery are too rare in the literature to identify clear differences among regions on the globe (see Table 5). Peterson and Pickett (1991) found only 44% plant cover on tip-up mounds following a tornado in Pennsylvania and compared this to recovery in Panama, where Putz (1983) found 100% cover after one year and proposed only a three to five month window of opportunity for establishment after gap creation. Peterson and Pickett (1991) propose several explanations for this contrast: mound size, rate of pit filling, seed availability, and growing season. Comparing differences in recovery rates is further hampered by a lack of a standard for comparing severity of damage. An examination of Table 6 shows differences in the reported paths to recovery for temperate and tropical forests. As reported above, regrowth (sprouting) has been reported twice as often in tropical sites (85.2%) as in temperate ones (40%). Both tropical sites and tropical forest species have been observed to have higher proportions of sprouting. Suppression also occurs slightly more often in the tropics (29.6% compared to 20%). Temperate forest researchers have reported release much more often that tropical forest researchers (65% n=20, compared to 22.2 n=27).SUMMARY
This review has identified some clear trends in forest damage and recovery from catastrophic wind, and it has illuminated questions for focusing future research. I presented seven critical questions at the start of this chapter and will now review the significant findings relative to each and suggest hypotheses derived from these findings. Intensity of wind disturbance should be quantified based on wind speed, both sustained and gusts. These values vary over the landscape, so distance from the wind measurement site must be included. In addition, the duration of the winds may influence the severity of impacts. Finally, hurricanes and cyclones often involve significant rainfall which may influence the type and severity of damage. Building on the work of Scatena and Larsen (1991), I suggest that all studies of catastrophic wind impacts should include these five measurements: 1) maximum sustained wind, 2) maximum gusts, 3) storm duration, 4) rainfall total and percent of average annual rainfall, and 5) distance between the study site and the site of measurement of the above parameters. How should catastrophic wind impacts be quantified? Different measures of hurricane damage result in different conclusions about severity of disturbance. To facilitate comparisons among sites I suggest a dual measure of severity: 1) structural loss - percent basal area lost, and 2) compositional loss - percent mortality. Regardless of individual research questions, catastrophic wind researchers should report totals for their sites that include structural loss and compositional loss. Two specific damage quantification issues require additional research. First, what are the conditions that control stem snapping versus stem uprooting. All researchers should distinguish between these two types of 'windthrow' and, where possible, report variations in their proportion relative to environmental conditions. Second, the issue of delayed mortality following catastrophic wind disturbance needs further examination. Evidence supports the hypothesis that mortality rate returns to background levels within five years following disturbance. This should be tested by following plots of damaged and undamaged forests through the first decade of recovery. Does catastrophic wind result in homogeneous damage across the landscape? Catastrophic wind damage does not uniformly affect the landscape. Increasing intensity of wind results in a gap size distribution increasingly skewed to larger gaps, but the full range of gap sizes from small to large is still created. With increasing damage, the landscape changes from a forest with gaps to one with isolated stands of intact forest. Future research should include analysis of gap size distribution and measures of landscape fragmentation to allow comparisons across a wind intensity gradient, and among regions of the globe. How do biotic factors influence severity of damage? Stem size, species, stand conditions, and pathogens may all influence the severity of damage during a windstorm. I hypothesize that the relationship between stem size and damage is unimodal and that studies indicating other patterns are limited in sample size or by the categories of stem size used in the analysis. The differential effects of wind on different species are so thoroughly documented in the literature, that further studies focusing exclusively on this issue are not recommended, as opposed to focusing on more general trends such as damage related to successional class. Pioneer species appear to be more susceptible to damage and more likely to die. The relationship between stand conditions and damage has been examined for managed stands, but not as thoroughly for natural forests. Another focus for future study should be the role of pathogens in influencing damage, particularly in the tropics. How do abiotic factors influence the severity of damage? Topography, disturbance history, and soil conditions all influence severity of damage. The role of topography in channeling wind is complex. Valleys are not always protected sites, even when their aspects do not align with the storm winds. Of particular interest is the influence of lee slopes, which may provide protection, and may be sites of turbulent airflow and therefore increased damage. Analysis of landscape patterns of damage should help us develop models to assess exposure accurately. Disturbances that partly open the canopy promote further disturbance from wind. More homogeneous disturbance can create even canopies that are resistant to wind. Multiple wind disturbance events often reduce further damage as each removes susceptible individuals and leaves more windfirm ones. Soil conditions that restrict root systems promote disturbance. Although root growth has been extensively studied in plantation forests, again, this would be a valuable focus for future research, particularly in the tropics. The relative roles of different biotic and abiotic factors in controlling catastrophic wind damage are difficult to identify when each varies over the landscape, usually not independently. Plots should be established, either before disturbance in areas of high storm frequency, or following a catastrophic wind event, specifically to isolate these variables and determine their relative importance. What are the dynamics of recovery from catastrophic wind impacts? Much more work is required in this area over longer time frames. Four distinct paths to recovery are identifiable: regrowth, recruitment, release, and repression. But the influence of disturbance severity and environmental gradients on these paths is little understood. It is critical to determine the factors that lead to these different recovery paths. I propose that the dual damage parameter of structural and compositional damage holds promise toward predicting the principal path to recovery. More data is needed to map regions of recovery paths within this gradient space. Several other issues of recovery deserve further analysis. 1) What is the role of soil moisture in influencing the rate and path to recovery? Here it would help to manipulate water in areas of different damage severity. 2) What cues stimulate flowering and fruiting after wind disturbance? Simulating disturbance of varying severity through defoliation, debranching, and stem snapping could help to identify the critical signal for this response, and how this response varies in different species. This could be an example of an adaptation specifically for catastrophic wind disturbance, since defoliation does not occurs with background tree mortality. Additionally, the question of whether this response relates in any way to successional class of the tree has not been examined. 3) Is there a reorganization period following catastrophic wind disturbance, and how does its length vary over regions of the globe? The difficulty here is in identifying a parameter to use in quantifying recovery. A measure of net system energy transfer, e.g., photosynthesis to respiration ratio, might be appropriate. 4) The issue of interacting disturbances is a complex one. But since most forests are subjected to a variety of types of disturbances, understanding the interactions between them and the resulting dynamics of recovery is critical. What are the differences between tropical forests and temperate forests in terms of disturbance and recovery? There appears to be a latitudinal gradient of increasing frequency and intensity of catastrophic wind toward the tropics and corresponding differences in the rate and path to recovery. Numerous authors have developed maps of storm events for different regions of the globe. Combining these efforts into one map that scores regions of the globe based on their catastrophic wind disturbance regime would be a valuable addition to the literature. Catastrophic wind disturbance may be an important factor in determining patterns of vegetation community structure. Much of the research examining these events has been merely descriptive, and only focused on the severity of disturbance and the initial response. To develop robust theories incorporating catastrophic wind disturbance, we must focus future research toward determining the relative roles of different biotic and abiotic factors in controlling the patterns of severity of damage, and we must relate environmental gradients and severity of damage to long-term dynamics of recovery.
Return to the Table of Contents