Project

BIOTROPICA 37(1): 16–24 2005 Factors Affecting Mortality and Resistance to Damage Following Hurricanes in a Rehabilitated Subtropical Moist Forest1 Rebecca Ostertag2,3,4 , Whendee L. Silver3,4 and Ariel E. Lugo4 3 Department 4 International of Environmental Science, Policy, and Management, 151 Hilgard Hall, University of California, Berkeley, CA 94720-3110, U.S.A. Institute of Tropical Forestry, Ceiba 1201, Jard´n Botanico Sur, R´o Piedras, PR 00926-1119, U.S.A. ı ı ´ ABSTRACT The ability to resist hurricane damage is a property of both individuals and communities, and can have strong effects on the structure and function of many tropical forests. We examined the relative importance of tree size, species, biogeographic origin, local topography, and damage from previous storms in long-term permanent plots in a rehabilitated subtropical moist forest in Puerto Rico following Hurricane Georges in order to better predict patterns of resistance. Severe damage included uprooted trees, snapped stems, or crowns with greater than 50 percent branch loss. Hurricane induced mortality after 21 mo was 5.2 percent/yr, more than seven times higher than background mortality levels during the nonhurricane periods. Species differed greatly in their mortality and damage patterns, but there was no relationship between damage and wood density or biogeographic origin. Rather, damage for a given species was correlated with mean annual increment, with faster growing species experiencing greater damage, suggesting that growth rate may reflect a variety of life history tradeoffs. Size was also predictive of damage, with larger trees suffering more damage. Trees on ridges and in valleys received greater damage than trees on slopes. A strong relationship was noted between previous hurricane damage and present structural damage, which could not solely be explained by the patterns with size and species. We suggest that resistance of trees to hurricane damage is therefore not only correlated with individual and species characteristics but also with past disturbance history, which suggests that in interpreting the effects of hurricanes on forest structure, individual storms cannot be treated as discrete, independent events. RESUMEN La habilidad de resistir da˜ os causados por un hurac´n es una propiedad com´ n de individuos y comunidades, y puede tener efectos marcados en la estructura y funci´ n de n a u o muchos bosques tropicales. Con el fin de predecir mejor los patrones de resistencia, nosotros examinamos la importancia relativa del tama˜ o de los arboles, las especies, el origen n ´ biogeogr´fico, la topografia local y el da˜ o por tormentas pasadas en parcelas permanentes de largo plazo en un bosque rehabilitado subtropical h´ medo en Puerto Rico, luego a n u del paso del hurac´n Georges. Las da˜ os severos incluyeron arboles con ra´ces socavadas, tallos rotos y/o doseles con p´rdida de ramas mayor del 50 por ciento. La mortalidad a n ı e ´ inducida por el hurac´n despu´s de 21 meses fue de 5.2 por ciento por a˜ o, siete veces superior que los niveles de mortalidad ocurridos durante per´odos sin huracanes. Las especies a e n ı difirieron mayormente en su mortalidad y patrones de da˜ o, pero no hubo relaci´ n entre el da˜ o causado y la densidad de la madera u origen biogeogr´fico. M´s bien, el da˜ o n o n a a n causado para una cierta especie estaba correlacionado con el incremento del promedio anual, con las especies de r´pido crecimiento experimentando mayor da˜ o, lo cual sugiere a n que la tasa de crecimiento puede reflejar variedad en las concesiones de historia de vida de las especies. El tama˜ o de los arboles tambi´n fue indicativo del da˜ o, con los arboles n e n ´ ´ m´s grandes sufriendo mayor da˜ o. Los arboles en las cimas y en los valles recibieron mayor da˜ o que los arboles en las pendientes. Se observ´ una fuerte relaci´ n entre el da˜ o a n n o o n ´ ´ causado por huracanes pasados y el actual da˜ o estructural, el cual no pudo ser explicado solamente por los patrones de tama˜ o y especies. Nosotros sugerimos que la resistencia n n de los arboles al da˜ o causado por el hurac´n se correlaciona no s´ lo con las caracter´sticas individuales y de las especies, sino tambi´n con el historial de da˜ os anteriores, lo cual n a o ı e n ´ sugiere que en la interpretaci´ n de los efectos de huracanes en la estructura del bosque, una tormenta individual no se puede tratar como un evento independiente y discreto. o Key words: disturbance; mean annual increment; mortality; resilience; resistance. THE ABILITY TO AVOID DAMAGE DUE TO NATURAL DISTURBANCES, defined as resistance, and the ability to recover after disturbances, defined as resilience, are important properties of individual species as well as communities (Holling 1973, Walker et al. 1999, Gunderson 2000). Differences in resistance and resilience among individual trees will alter competitive hierarchies and will therefore have strong effects on future forest structure. Hurricanes are an important natural disturbance in many tropical forests that act to define forest structure and function, particularly through variation in their intensity and frequency. Hurricane winds can lead to defoliation and large nutrient inputs to the forest floor, structural damage, decreased biomass, and mortality (e.g., Lugo et al. 1983; Whigham et al. 1991; Walker 1991, 1995; Weaver 1994; Lugo & Scatena 1996; Ostertag et al. 2003). These disturbances can also alter resource availability and heterogeneity, providing opportunities for 1 2 Received 2 March 2004; revision accepted 22 August 2004. Current address: Department of Biology, University of Hawaii at Hilo, 200 W. Kawili St. Hilo, HI 96720; e-mail: ostertag@hawaii.edu 16 regeneration, species invasion, and alteration of successional pathways (Guzm´n-Grajales & Walker 1991, Everham et al. 1996, Harrington a et al. 1997). Posthurricane surveys have shown that trees differ in their resistance to damage and in their mortality rates (Walker 1991, Zimmerman et al. 1994), but predicting damage and mortality has proven difficult because multiple factors may determine how a tree responds to hurricane winds and these factors may operate at differing spatial and temporal scales. The level of damage a tree experiences may relate to its size (Lugo et al. 1983, Walker 1991, Herbert et al. 1999), position on the landscape (Bellingham 1991, Boose et al. 1994, Weaver 1999), or to species-specific characteristics such as architecture, wood density (Zimmerman et al. 1994), or biogeographical origin (MacDonald et al. 1991). For example, after Hurricane Hugo in Puerto Rico, pioneer species, characterized by low wood densities, suffered greater damage than many nonpioneer trees (Zimmerman et al. 1994). Nonnative species displayed more damage than native species after a hurricane in the Mascarene Islands, and it was hypothesized that this result was because those species were not adapted Predicting Resistance to Hurricane Damage 17 to the disturbance regime (MacDonald et al. 1991). This suggestion that biogeographic origin affects damage resistance was also made for provenances of Pinus caribaea and P. oocarpa, after noting in plantations in Puerto Rico that provenances from locations with higher hurricane frequencies had lower mortality rates and less mechanical injury (Liegel 1984). Complicating these factors, however, is the previous damage to an individual tree and the length of time between storms. Other research (Sauer 1962, Putz & Sharitz 1991, Elmqvist et al. 1994, Peterson 2000) provide some circumstantial evidence that previously damaged trees are more susceptible to future damage, but this question has not been addressed on an individual tree level on tagged trees followed through multiple storms. We had a rare opportunity to examine all of these factors in parallel, after the passage of Hurricane Georges in 1998, which had particularly strong effects in a subtropical moist forest in the northeastern portion of Puerto Rico. This forest experienced a hurricane 9 yr previously, in an area in which trees were tagged, measured, and hurricane damage was quantified. Our objective was to examine the influence of hurricanes on forest structure and function by quantifying multiple factors including the relative importance of species, biogeographic origin, local topography, size, and previous damage as factors affecting resistance to hurricane damage and life history tradeoffs. We hypothesized that tree damage is strongly related to tree size, topography, and species characteristics but does not differ between native and nonnative species. We also hypothesized that previous damage from earlier hurricanes, although partially related to species and size characteristics, will in itself also be a useful predictor of future damage. Saffir–Simpson scale with sustained winds of 166 km/h and gusts of 194 km/h (Scatena & Larsen 1991). Hurricane Georges made landfall on the southeastern coast of Puerto Rico on 21 September 1998 and traversed the island from east to west. The island-wide overall statistics for hurricane were sustained winds of 184 km/h, gusts of 241 km/h, and a Category 3 storm on the Saffir–Simpson Scale (Ostertag et al. 2003). Although Hugo was an overall more powerful storm than Georges, all indications are that the Cubuy forest area was damaged more severely by Georges. Evidence of this comes from Zimmerman et al. (1995), who surveyed a nearby site in the Cubuy region in which they noted minimal damage and estimated peak sustained winds of 115 km/h. In addition, the Cubuy area was farther from the eye path in Hurricane Hugo (see Zimmerman et al. 1995), and during Hurricane Georges a different wind direction placed this forest much closer to the eye of the second hurricane (for a map of the eye of the storm, see USGS Fact Sheet 040-99, http://water.usgs.gov/pubs/FS/FS040-99/pdf/fs-040-99.pdf ). TREE DAMAGE AND GROWTH MEASUREMENTS.—During the 1992 forest survey described above, measurements on the effects of Hurricane Hugo were also taken in all 116 plots. Unfortunately, the data were collected 3 yr after the hurricane, and it is likely that these methods underestimated defoliation damage because this would be less evident after 3 yr, while evidence of snaps, uproots, or major branch damage remain. We used the same methodology as in the 1992 survey and took tree damage measurements in 1999 and 2000 (at 6 and 21 mo after Hurricane Georges). To compare the effects of the two hurricanes, we focused on a subset (N = 21) of the 116 plots. Fifteen of these plots were being used for a previous study (Silver et al. 2004) and the additional plots were chosen for the accessibility and to increase the range of tree species sampled. The plots were spread out across the original 9 ha area and contained a variety of key species. A total of 976 tagged trees were located and measured in 1999 and in 2000, these trees were checked for mortality only. In both the 1992 and 1999 sampling, tree status was classified as live, recently dead (due to the hurricane), or old dead (dead before most recent hurricane). Each tree was also classified by crown class (dominant, codominant, intermediate, or suppressed); when the top of the tree had snapped off, crown class was estimated based on the debris on the forest floor. Every live tagged tree was measured for DBH (DBH at 1.37 m) allowing for mean annual increment of diameter growth to be calculated based on the 7-yr growth period from 1992 to 1999. Mortality was calculated as (no. of trees dead) × 100/total no. of trees/7 yr. Wood density values were taken from a pantropical survey (Reyes et al. 1992). Damage to trees was quantified by noting damage type (snap, uproot, defoliation and branch damage, or no visible damage). Snap height (done only in 1999) and angle of uproot were also noted. Whenever possible, branch damage was assessed. A category of no branch damage was used for trees that had no visible defoliation or loss of branches. However, trees in all other categories experienced defoliation but differed in the degree of branch damage. Light damage was defoliation and branch loss of small terminal branches that affected 50 percent of crown and was loss of very large branches. This same classification METHODS SITE AND STORM DESCRIPTIONS.—The Cubuy Annex of the Luquillo Experimental Forest (LEF), Puerto Rico (18◦ 17 N, 65◦ 53 W) was once pasture but was reforested in the mid-to-late 1930s with single and mixed species plantings and natural regeneration (Marrero 1947). Today it is a diverse closed canopy secondary forest containing 75 tree species (Silver et al. 2004). The canopy is uneven with many of the larger diameter trees ranging in height from 15 to 30 m. The life zone is classified as subtropical moist forest (Ewel & Whitmore 1973), with elevations ranging from 300 to 550 m. Soils are generally highly weathered Ultisols derived from volcanoclastic material. Monitoring of this forest began in 1959 when the U.S. Forest Service established 116 permanent circular plots, which were of 0.04 ha in size and spaced 100 m apart (from center to center) in a 9 ha area of contiguous forest. At that time, only merchantable trees that were greater than 9.1 cm diameter at breast height (DBH) (DBH at 1.37 m above the ground) and of appropriate form were tagged and identified. These surveys were expanded in 1992 when all trees that were greater than 9.1 cm DBH were tagged and identified. Large trees >24.1 cm DBH outside of the 0.04 ha plots were also tagged, up to 16.05 m from the center of the plot. In this study, we searched for all trees within the 0.04 ha plots as well as these larger tagged trees so that we would have a representative range of sizes and species. This forest was hit by two hurricanes within a 9-yr period. On 18 September 1989, Hurricane Hugo hit northeastern Puerto Rico, including the LEF. This hurricane was a Category 4 storm on the 18 Ostertag, Silver, and Lugo TABLE 1. Mortality of trees at each sampling period for the most abundant tree species (>20 individuals) and total number of trees of all species in the Cubuy annex of LEF. Mortality from Georges was measured at 6 and 21 mo and calculated as the number of dead trees divided by total sample size at that time period. Sample sizes changed slightly between the two measuring periods because some trees could not be found at both periods. The background mortality rate was calculated as the species that were alive in the 1992 forest survey but were categorized as old dead (i.e., dead before the hurricane) in the 1999 survey. Species are listed from high to low total mortality with nonnative species in bold. Trees at 6 mo after Georges Mortality (percent/yr) 14.7 10.3 10.1 9.5 0.0 0.7 0.0 0.0 7.5 Trees at 21 mo after Georges Mortality (percent/yr) 14.7 9.3 3.3 2.7 2.4 1.3 1.0 0.0 5.2 No. of live in 1992 136 39 139 42 24 258 58 28 930 Trees before Georges No. of dead before hurricane 13 0 8 0 0 10 0 0 46 Background mortality (percent/yr) 1.2 0.0 0.8 0.0 0.0 0.5 0.0 0.0 0.7 N Calophyllum antillanum Roystonea boriquena Tectona grandis Syzygium jambos Hymenaea courbaril Tabebuia heterophylla Andira inermis Mangifera indica All species 136 39 139 42 24 268 58 28 930 N 136 37 139 42 24 268 57 27 928 N 149 39 147 42 24 268 58 28 976 was also used for the royal palm, Roystonea boriquena, a conspicuous component in this forest. Because the palm does not have branches it was evaluated as to whether it was missing 50 percent of its leaves. STATISTICAL ANALYSES.—Relationships for independence between categorical variables were evaluated using G-tests, which are log-likelihood ratio tests, where G = 2 ln (likelihood ratio), and where the P-value is based on a χ 2 distribution (Sokal & Rohlf 1995). Data were analyzed using JMP (SAS Institute 1995). G-tests were done to examine how both tree mortality and tree damage were affected by the following categorical variables: DBH class in 10-cm increments, crown class, local topography, and damage in the previous hurricane (Hugo) based on the 1992 measurements. In the 1992 sampling only one tree was listed as having light defoliation, and so this tree was combined with the no damage category. In addition, because there were few snaps and uproots in the 1992 data, we also combined these categories. In translating DBH into categories, we grouped all trees ≥40 cm due to the paucity of very large trees in this developing forest. To facilitate comparisons across groups with differing sample sizes, we created a “damage index,” which is percentage of trees with heavy branch damage, snaps, or uproots in a given class. This damage index thus concentrates on trees with severe crown loss or stem and root breakage and is comparable to one of the components calculated by Bellingham et al. (1995) that was used to group species into different degrees of resistance and recovery. The damage index was used to compare across species, size classes, crown classes, topography, and previous damage. When used to compare species, the damage index was only calculated for species in which there were 20 or more individuals; when used to compare the other variables, there were hundreds of individuals making up the percentage. The damage index represents the severity of damage relative to the total number of individuals within that category (e.g., a damage index of 33 for a species would mean that one-third of the individuals had severe damage in Hurricane Georges). RESULTS MORTALITY FROM HURRICANE GEORGES.—Mortality 6 mo after Hurricane Georges was only 7.5 percent/yr, but decreased to 5.2 percent/yr at 21 mo. This storm-related mortality was over seven times greater than the background mortality of 0.7 percent/yr for trees from the 1992 to 1999 period (Table 1). Mortality varied greatly by species: the nonnative species Mangifera indica had the lowest rate and the native species Calophyllum antillanum had the highest rate (Table 1). Trees that uprooted or snapped were much more likely to die after 21 mo, while those trees with none, light, or medium amounts of defoliation were more likely to survive (G = 183.2, df = 5, P < 0.0001, Table 2). Overall 3.2 percent of defoliated trees, 37.7 percent of snaps, and 41.9 percent of uprooted trees died within 21 mo of the storm. Although the type of damage affected the probability of death, mortality was independent of other characteristics such as crown class, DBH class, topography, or damage in the previous hurricane. RESISTANCE TO DAMAGE FROM HURRICANE GEORGES.—Of the trees that were not killed in the storm, only about 9 percent of trees had no damage (Table 2). Species varied greatly in their resistance to damage TABLE 2. Observed number of live and dead trees (and expected values in parentheses) suffering various types of hurricane damage. Trees were classified by whether they showed evidence of branch damage (light, medium, or heavy canopy damage), snap of trunk, or complete or partial uprooting. Snapped or uprooted trees died more often than would be expected. Branch damage None Dead Live 1 (6.2) Light 1 (23.8) Medium 4 (14.0) Heavy 16 (18.7) Snap 46 (10.1) 76 (111.9) Uproot 14 (2.6) 18 (29.4) 73 (67.9) 287 (264.2) 165 (155.0) 210 (207.3) Predicting Resistance to Hurricane Damage 19 TABLE 3. Percentage of the most abundant tree species (>20 individuals) suffering various types of hurricane damage. Trees were classified by whether they showed evidence of branch damage (light, medium, or heavy canopy damage), snap of trunk, or complete or partial uprooting. Species are listed according to their damage index, which is the sum of heavy canopy damage, snaps, and uprooted trees. Mean annual increment was also calculated for each species, except for the palm which does not have secondary growth. Nonnative species are in bold. Branch damage None Hymenaea courbaril Calophyllum antillanum Syzygium jambos Roystonea boriquena Mangifera indica Andira inermis Tectona grandis Tabebuia heterophylla All species 0.0 2.2 7.1 5.3 17.9 17.9 11.6 9.4 8.1 Light 16.7 10.3 26.2 21.1 10.7 25.0 40.6 48.9 31.5 Medium 4.2 19.1 9.5 18.4 17.9 16.1 22.5 24.4 18.5 Heavy 50.0 36.8 28.6 36.8 42.9 25.0 10.9 13.9 24.7 Snap 20.8 22.8 28.6 18.4 10.7 16.1 10.9 2.3 13.7 Uproot 8.3 8.8 0.0 0.0 0.0 0.0 3.6 1.1 3.5 Damage index (percent) 79.2 68.4 57.1 55.3 53.6 41.1 25.4 17.3 41.9 Mean annual increment (cm/yr) 0.381 0.232 0.220 0.328 0.103 0.220 0.101 0.176 (Table 3). Overall, the damage index across species was 41.9 percent, but was as low as 17.3 percent for Tabebuia heterophylla and as high as 79.2 percent for Hymenaea courbaril. Tree damage was not correlated with wood density (P > 0.4) and did not appear related to biogeographic origin of species, as both the native and nonnative species spanned the range in terms of damage index. Rather, the damage index of a species was positively correlated with mean annual diameter growth increment (Spearman’s rho = 0.81; P < 0.03), with species with faster diameter growth rates experiencing greater amounts of damage. This result suggests a life history tradeoff between growth rate and damage (Table 3), although it is influenced by tree size because growth rate is not a constant variable. When correlating basal area increment with damage index, the relationship was not significant. However, when using all variables in a stepwise multiple regression, damage index was related most strongly to basal area increment and average DBH of the species (y = 92,628 BAI (m2 ) − 6.7 AVERAGE DBH (cm) + 116.5; R 2 = 91.9 percent). Thus, in the multiple regression, the damage index of a species was greater for species that were of smaller size in the forest and were fast growing. Across all trees, most trees fell in a NW direction (275–30◦ ) or a SW direction (170–240◦ ), with a few trees falling E (85–102◦ ). Since the major path of the storm was from east to west, these treefall likely reflect in part average wind orientation, as demonstrated by Boose et al. (1994). Average snap height was 5.1 m, with 26 trees snapping between 0 and 2 m, 26 trees between 2.1 and 4 m, 32 trees between 4.1 and 6 m, 29 trees between 6.1 and 8 m, 9 trees between 8.1 and 10 m, and 6 trees snapping at heights > 10 m. Damage was not independent of DBH class (G = 65.1, df = 15, P < 0.0001). Heavy branch damage was less common than expected for smaller trees (