The Effects of Hurricane Hugo on Bats of the Luquillo Experimental Forest of Puerto

Natural disturbances can have large effects on ecosystem structure and function depending on their scale, intensity, and frequency. On 18 September 1989 Hurricane Hugo struck Puerto Rico, with the eye of the hurricane passing within 10 km of the Luquillo Experimental Forest. This provided a rare opportunity to evaluate the effects of an infrequent but large scale and high intensity disturbance on tropical bat species. Data on demographic parameters of three common phyllostomid bats (Artibeusjamaicensis, Stenoderma rufum, and Monophyllus redmani) were examined for three years prior and three years after the hurricane. Population levels as estimated by captures per net hour of all three species were affected by Hurricane Hugo. Populations of A. jamaicensis and M. redmani returned to predisturbance levels within two years. In contrast, population levels of S. rufum declined to about 30 percent of prehurricane levels and have not recovered after three years. Moreover, telemetry data indicate that foraging and home range size expanded to encompass an area approximately five times larger than its prehurricane size. The cost of foraging, in terms of time and energy, may be considerably elevated over prehurricane scenarios. In fact, a significant change in the age structure of the population (juvenile individuals have been absent from the population since Hurricane Hugo) as well as significant decline in the percent of reproductively active females indicate a failure to reproduce in the posthurricane environment. RESUMEN

bances that regularly occur throughout the Caribbean (Weaver 1989); their impact on the structure and function of tropical ecosystems is often great (Brokaw & Walker 1992, Walker et al. 1992).
Although the importance of disturbance regimes is a contemporary concept in ecological theory (Pickett & White 1985), few empirical studies document their effects on animal populations, with the notable exceptions of Covich et al. 1992, Reagan 1992, Waide 1992, Willig and Camilo 1992, and Woolbright 1992.Although bats are keystone agents of pollination and seed dispersal in the tropics, little is known about their response to and recovery from hurricanes.
Effects of a disturbance can be immediate or long-term.Immediate effects are directly attribut-I Received 22 March 1993, revision accepted 24 November 1993.able to the hurricane (high winds, heavy rains) and affect individual mortality and spatial distributions.
Indirect effects result in differential survivorship and reproductive success, with populations responding either positively or negatively to alterations that have occurred in the composition or structure of relevant habitats.
Frugivorous bats make a critical contribution to tropical forest succession by widely dispersing the seeds of early successional plants (e.g., de Foresta et a/. 1984, Charles-Dominique 1986).Indeed, the seeds of small, fleshy-fruited plants are often dispersed by bats and dominate seed banks of tropical soils.Moreover, such seeds are a source of colonists in recently disturbed areas (Uhl et al. 1981, Swaine & Hall 1983, de Foresta et al. 1984).Bats also have been implicated in having an important role in flower pollination in tropical systems (Howell 1974(Howell , 1978;;Stuart & Marshall 1976;Gould 1978;Ng 1978;Sazima & Sazima 1978).As such, frugivorous and nectarivorous bats can have a large impact on the distribution and genetic structure of plant species.
Compared to mainland areas of similar size and habitat diversity, Puerto Rico harbors few mammal species, and population numbers are generally low.Bats compose the major portion of the Puerto Rican mammal fauna in terms of species richness and density (Willig & Bauman 1984, Willig & Gannon 1994).Within the Luquillo Experimental Forest (LEF), four species of bats dominate, although some are common in certain habitats and rare or absent in others.Three of these species, Stenoderma rufum, Artibeus jamaicensis, and Brachyphylla cavernarum are principally frugivorous.The fourth, Monophyllus redmani, feeds on flower nectar.
Stenoderma rufum, the red fig-eating bat, has been found only at two localities on Puerto Rico, and on the nearby islands of St. John and St. Thomas (Genoways & Baker 1972).Until recent times, it was thought to be extinct and was known only from fossil records.As a result, it is a poorly known species and is in rare scientific collections.Only the population in the LEF has been sampled and studied to any extent during the past 25 years.It is currently designated as a "sensitive" species by the U.S. Forest Service.The status of other populations on Puerto Rico or the Virgin Islands is unknown.Studies conducted prior to Hurricane Hugo have examined various aspects of its population biology and ecology, including foraging and home range (Gannon 1991, Willig & Gannon 1994), reproduction (Gannon & Willig 1992), and diet (Willig & Bauman 1984, Willig & Gannon 1994).These works indicate that S. rufum consumes a variety of fruits, exhibits asynchronous bimodal polyestry, roosts in canopy foliage, and has a relatively small home range (mean = 2.1 ha) to which it exhibits high site fidelity for at least several months during the rainy season.
Artibeusjamaicensis, the Jamaican fruit bat, has a wide geographic distribution in tropical and subtropical America.It is a much-studied species and is known to consume a variety of fruits, as well as some flowers and insects.Although extensive work has examined aspects of the ecology of several mainland populations (Handley et al. 1991;Morrison 1975Morrison , 1978aMorrison , b, 1979)), little has been done with island populations in general, or with Puerto Rican populations specifically (see Kunz et al. 1983).Previous research from the Luquillo Experimental Forest indicates it comprises at least 60 percent of the bat fauna (Willig & Bauman, 1984) and exhibits aseasonal polyestry (Willig & Bauman 1984;Willig & Gannon 1994).

Monophyllus redmani, the Greater Antillean
Long-tongued bat, feeds primarily on flower nectar.

It has a distributional range restricted to the Greater
Antilles and several islands in the Bahamas (Homan & Jones 1975).Little is known of its ecology other than anecdotal observations.It is common on Puerto Rico, where it comprises a substantial portion of the LEF bat fauna (Willig & Bauman 1984, Willig & Gannon 1994).

STUDY SITE
Puerto Rico, the smallest and easternmost of the Greater Antilles, is subject to storms of great magnitude at intervals of approximately 60 years (Doyle 1981(Doyle , 1982)).The Luquillo Experimental Forest (18010'N, 65030'W), also known as the Caribbean National Forest, is located in the northeast corner of the island, within the Luquillo Mountains.Increasing elevation in the Luquillo Mountains is accompanied by changes in climate, soil, and vegetation structure and composition.As a result, three distinct life zones occur within the bounds of the LEF (Ewel & Whitmore 1973, Brown et at. 1983).The tabonuco rain forest, the largest life zone, is located on lower mountain slopes of the LEF below 650 meters.Rainfall is substantial and varies between 2000 and 4000 mm annually.The palo colorado forest is found in valleys and on mountain slopes above cloud condensation level at 600 meters elevation; average rainfall is 4700 mm.The dwarf forest occupies the highest mountain summits and ridge lines above 850 meters.It comprises dense stands of short trees and shrubs.This area is continuously exposed to winds and clouds and receives rain nearly 350 days per year.
The LEF is a disturbance mediated forest (Crow 1980, Doyle 1981).Natural disturbances vary in frequency and intensity and include tree fall gaps, landslides, and hurricanes.The greatest disturbance in recent years has been Hurricane Hugo, which had a dramatic effect on the forest (Brokaw & Walker 1992, Scatena & Larsen 1992, Walker et al. 1992)

METHODS
Bats were captured with mist nets at sites established within the tabonuco, palo colorado, and dwarf forest.Sites were chosen based on accessibility and our previous netting success.Age, sex, weight, and reproductive condition were determined for each captured animal.We also marked each bat by attaching small ball chain necklaces, each carrying a uniquely numbered aluminum band (Gey Band and Tag Co., Norristown, Pennsylvania, U.S.A.).
Individuals within bat populations are not equally catchable because of their ability to echolocate, and avoid nets after encounters with such devices.This is true of the bats in the LEF, where recaptures of bats are extremely low.Assumptions of most common population estimation techniques are violated severely when estimating bat density based upon netting.In an attempt to evaluate population trends of bats in the LEF, we utilized several techniques.We determined the number of bats captured per net hour during each sampling period, and compared those numbers over time.In addition, relative importance for each bat species (compared to other bats in the community) was assessed from netting records in two ways.Numerical dominance (ND) for each species was measured as n,/N, where ni is the number of captured individuals of species i and s N= n,Y, is the total number of captured animals regardless of taxonomic identity (Willig & Gannon 1994).
Biomass dominance (BD) for each species is gi- where Y, is the mean biomass of species i and s is the number of species in the community (Willig & Gannon 1994).
To evaluate foraging behavior and home range, S. rufum were selected and fitted with radiotransmitters using a previously described protocol (Gannon 1991).Radio-tracking was conducted in the same areas of the tabonuco rain forest during the rainy season Uune, July, and August) for two years prior and two years after the occurrence of Hurricane Hugo.Transmitters weighing 1.2 g (Holohil Systems, Ltd., Ontario, Canada, model BD-2) were attached to the dorsal pelage of bats using "Skin  (Gannon 1991).Telemetry accuracy using this protocol in the LEF was extremely high (Gannon 1991, Lindsey & Arendt 1991).
Home range is commonly considered to be the area traversed by an animal while it conducts essential activities (Burt 1943) based on telemetrically obtained location data using two methods: minimum convex polygon (Odum & Kuenzler 1955) and minimum area probability (Anderson 1982).
The minimum convex polygon (MCP) method is chosen frequently because of its simplicity.It assesses home range by drawing the smallest possible convex polygon around the outermost capture points.
Although this method may be sensitive to sample size, and over-or underestimate home range size   (Anderson 1982, Wilkinson & Bradbury 1988).These MAP values are suggested analytical standards (Anderson 1982, Wilkinson & Bradbury 1988).MAP 95 is used because it closely corresponds to the standard definition of home range suggested by Burt (1943).However, when using large MAP estimates, occasional outlier points lead to large errors (Anderson 1982).Because smaller percent MAPs are less sensitive to outlier generated error, we also conducted analyses using MAP 50 home range estimates.MAP 50 values are as useful as MAP 95 values for comparative purposes, but have the advantage of representing smaller core areas  1987 1988 1989 1990 1991 1992 Species  that reduce the effects of occasional outlier points on the resulting estimate of area.
The effects of hurricane (pre-versus post-Hugo) and sex (male versus female) on home range and foraging range size were evaluated via two-way analysis of variance (ANOVA), followed by orthogonal a priori contrasts (Sokal & Rohlf 1981).Separate analyses were conducted on home range and foraging range based upon MCP and MAP estimates.

RESULTS
POPULATION LEVELS.-Ourprimary netting site, near El Verde Field Station, in the tabonuco rain forest has been sampled for bats for the longest period, and thus should be the most useful in examining population trends within this life zone.This site was netted during dry and rainy seasons each year for six years beginning in 1987 (only the dry season was sampled in 1987).Population trends over this period were examined for each of the three dominant bat species at this locality, using numerical dominance, biomass dominance, and number of individuals captured per net hour (Fig. 1, Table 1).Both A. jamaicensis and S. rufum exhibited relatively high numbers and dominance prior to Hurricane Hugo.Numbers of A. jamaicensis dedined to near zero immediately after the hurricane, remained low for almost two years, and then recovered to prehurricane levels in year three.Numbers of S. rufum declined more gradually after the impact of Hurricane Hugo, reaching the lowest level in 1991.It has remained at densities far below those maintained prior to the disturbance.The dominance values of M. redmani fluctuated both before and after Hurricane Hugo; however, actual numbers of individuals have remained relatively stable, with only a slight increase soon after Hurricane Hugo.HOME RANGE AND FORAGING.-Statisticalcompar isons of MCP home ranges were made by two-wa ANOVA to assess effects of hurricane (pre-Hug versus 1 year post-Hugo versus 2 years post-Hugo sex (adult male versus adult female), and their in teraction (Table 2, Fig. 2a).Both home range an foraging range exhibited a consistent significant dif ference associated with the hurricane treatment, bu no difference related to sex of individuals.As consequence, the sexes were combined in subsequen a posteriori comparisons to evaluate mean differenc between all possible pairs of levels of hurrican treatments.Pre-Hugo (mean = 2.1 ha) and 2 year post-Hugo (mean = 3.6 ha) home range sizes wer statistically indistinguishable based upon Welsch step-up procedure; however, each was statistically distinguishable from the 1 year post-Hugo hom range size (mean = 13.4 ha).The same pattern o significance was obtained for MCP foraging range (means were 1.2 ha, 9.3 ha, and 2.3 ha for pre-Hugo, 1 year post-Hugo, and 2 years post-Hugo respectively).Thus, size of home range and foraging range were significandy affected by Hurricane Hugo but the effect was no longer detectable after 2 year (Fig. 3).
For analyses based on MAP, statistical comparisons were made with two-way ANOVA for effects due to hurricane (pre-Hugo versus 1 yea post-Hugo versus 2 years post-Hugo), sex (adult male versus adult female), and their interaction.For MAP 95 home range and foraging range, we de tected significant differences due only to Hurricane Hugo, and these differences were experienced b the sexes in the same manner (i.e., no significan interaction effect, Table 1, Fig. 2b).As a consequence, males and females were combined in sub sequent a posteriori comparisons to evaluate mea differences between all possible pairs of levels o Results for the more stable MAP 50 home range and foraging range showed significant differences due only to Hurricane Hugo, and these differences were experienced by males and females in the same manner (i.e., no significant interaction effect, Table 2, Fig. 2c).As a consequence, males and females were combined in subsequent a posteriori comparisons to evaluate mean differences between all possible pairs of levels of hurricane treatment.Pre-Hugo (mean = 0. 7 ha) and 2 year post-Hugo (mean = 1.1 ha) home range sizes were statistically indistinguishable based upon Welsch's step-up procedure; however, each was statistically distinguishable from the 1 year post-Hugo home range size (mean = 3.0 ha).Although the same trend toward recovery from hurricane effects is seen when attention is restricted to foraging range using MAP 50 techniques (means are 0.5 ha, 2.9 ha, and 1.2 ha for pre-Hugo, 1 year post-Hugo, and 2 years post-Hugo, respectively), each hurricane treatment level was different from the others (Welsch's step-up procedure).
Foraging ranges (MAP 50) were smaller 2 years after the hurricane than they were 1 year after the hurricane; nonetheless, they have not returned to the prehurricane sizes.percent results are illustrated (Fig. 4).These figures consistently suggest that bats forage and roost within larger areas 1 year after the hurricane, and that the utilization of space is more diffuse or homogeneous viduals regardless of species) at all sites wi particular forest type represents the com composition of that forest type (four yea 19911 in the tabonuco rain forest [Fig.51 as for tabonuco, palo colorado, and dwarf during 1991 [Fig.61).The proportional co sition of the bat fauna in the tabonuco rain forest was altered because of the effects of Hurricane Hugo (Fig. 5).In particular, the originally dominant A. jamaicensis became a less important component of the fauna within a year of Hurricane Hugo)s impact; however, it presently occupies a position of dominance equal to that during prehurricane periods.
Although the abundance of S. rufum declined after the hurricane, it declined less than the other species, consequently assuming a position of higher dominance within a year of Hurricane Hugo's impact.
Since the hurricane, we have noted a decline in the proportion of females that are breeding, with relatively few individuals producing young.This decline has persisted into 1992, with less than 5 percent of the females exhibiting reproductive activity.

DISCUSSION
Hurricane Hugo had a variety of effects on the bats of the LEF.These effects were best documented in the tabonuco rain forest, and to some extent, they were species-specific.
Artibeus jamaicensis was negatively affected by Hurricane Hugo, with a severe decrease in numbers, as well as in relative importance.Its levels remained low for two years posthurricane.Because prehurricane telemetry data indicated that this bat is a strong flier that moves large distances (M.Gannon, pers. obs.;Handley et al. 1991) The most striking results are for S. rufum; it was negatively affected by the hurricane, although apparently less so initially than other bat species.However, it decreased in numbers in a steady fashion after Hurricane Hugo and experienced its lowest level during the dry season of 1991.Inability to disperse out of the tabonuco rain forest, as suggested by its limited foraging and home ranges, combined with increased exposure to high temperature, precipitation, and wind at roost sites (tree canopy), as well as decreased availability of fruit, may account for its gradual decline after Hurricane Hugo.
The impact of Hurricane Hugo on potential reproductive status of S. rufum occurred in two stages.The first stage reflects the greater susceptibility of juveniles to hurricane-induced alteration of resource and refuge characteristics (on average, proportions of juveniles decreased from 40 percent before to 17 percent immediately after Hurricane Hugo; none have been found since the dry season of 1991).
The second stage reflects reduced fertility of adult females as well as reduced survivorship of pregnant females and their offspring; few of the posthurricane females have been reproductively active (on average, proportions decreased from 93% before to 29% after the hurricane).These trends appear to be continuing, and the future will probably see a further reduction in density of this sensitive species.Our surveys, along with historical records, indicate that S. rufum is restricted to this forest life zone.Its absence from both palo colorado and dwarf forest suggests that immigration from surrounding areas (the rescue effect: Brown & Kodric-Brown 1977, Willig & Moulton 1989)  We recognize that the alteration of the forest as a result of the hurricane could potentially alter our net-capture success.However, because the disturbance resulted in an almost complete removal of the forest canopy, the only remaining food source for frugivorous bats was early successional shrubs on the forest floor.In this case, with our nets located in close proximity to fruit sources, the netting bias would likely be to enhance capture success.Since our data reveal trends in the opposite direction, we feel we have taken a conservative approach in our interpretation.
Bats have been implicated as major dispersers of important species of tree in the tabonuco rain forest (Devoe 1990), and for some species, such as Manilkara bidentata, they appear to be the only dispersal agent (You 1991).Within the tabonuco Bond" surgical glue (Pfizer Hospital ProductsGroup, Inc., Largo, Florida).Subsequent to release, tagged individuals were tracked from the ground by two observers using hand-held telemetry receivers and antennas (Wildlife Materials, Inc., Carbondale, Illinois, model TRX-1000S).Eighteen adult S. rufum (8 d! and 10 99) were tagged and tracked during 1988 and 1989 (prehurricane); whereas, eight S. rufum (4 d! and 4 99) were tagged and tracked during 1990 (1 year posthurricane).Similarly, eight (4 d! and 4 99) were tagged and tracked during 1991 (2 years posthurricane).All radiotagged individuals were adults and nonreproductive at the time of tagging.Each bat's location was triangulated once during the day, and between 1 and 3 times during the night.Tracking continued for a minimum of 15 days.Day location points, indicating day roosts, and night location points, indicating night roosts, were recorded as coordinates on a gridded map.Previous work with bats in the LEF indicated this protocol resulted in stabilized areas and should therefore reflect an accurate estimate of space utilization FIGURE 1.Long-term population trends of three common phyllostomid bats, from a single netting locality within the Luquillo Experimental Forest, (a) based on number of specimens captured per net hour of sampling effort, (b) numerical dominance, and (c) biomass dominance.An arrow indicates the occurrence of Hurricane Hugo.The open circles represent Artibeus jamaicensis, the closed circles represent Stenoderma rufum, and the triangles represent Monophyllus redmani.
FIGURE 2. A comparison of mean home range (left) and foraging range (right) of Stenoderma rufum before (solid shading) and after Hurricane Hugo (1 year post-Hugo, unshaded; 2 years post-Hugo, stippled) based on (a) Minimum Convex Polygon, (b) 95 percent Minimum Area Probability, and (c) 50 percent Minimum Area Probability estimation methods.
FIGURE 3. Schematic representation on the Minimum Convex Polygon met different individuals were used to construct Minimum Convex Polygons; comparisons are illustrative and do not represent a change of space use by a single individual.
This content downloaded from 137.99.143.45 on Tue, 25 Apr 2023 19:52:33 UTC All use subject to https://about.jstor.org/termshurricane treatment.Pre-Hugo (mean = 2.1 ha) and 2 year post-Hugo (mean = 3.3 ha) home range sizes (MAP 95) were statistically indistinguishable based upon Welsch's step-up procedure; however, each was statistically distinguishable from the 1 year post-Hugo home range size (mean = 9.7 ha).Although the same trend toward recovery from hurricane effects was reflected in the analyses of foraging range using MAP 95 techniques (means are 1.5 ha, 8.5 ha, and 3.4 ha for pre-Hugo, 1 year post-Hugo, and 2 years post-Hugo, respectively), each hurricane treatment level was different from the others (Welsch's step-up procedure).Although MAP 95 foraging ranges are smaller 2 years after the hurricane than they were 1 year after the hurricane, they have not yet returned to the prehurricane sizes.
figures were obtained for ranges based on 95 percent and 50 percent approaches; consequently, only 95

(FIGURE 5 .
FIGURE 5. Proportional representation of spe the bar fauna of the tabonuco rain forest at El Ve 1988 to 1991 (Various netting sites are pooled rainy seasons are pooled).

FIGURE 6 .
FIGURE 6. Proportional representation of species in the bat fauna within each of three life zones (tabonuco rain forest, palo colorado forest, and dwarf forest) of the Luquillo Experimental Forest.
Rainy Dry Rainy Dry Rainy Dry Rainy Dry Rainy Dry Rainy Dry Rainy Dry Rainy

FIGURE 7 .
FIGURE 7. Histograms representing (a) the demographic structure, and (b) the proportion of breeding females, of the population of Stenoderma rufum in the tabonuco rain forest at El Verde.An arrow represents the occurrence of Hurricane Hugo.
, reduced numbers may reflect movement of individuals to less affected areas of the island.Data collected since the rainy season of 1991 indicate that this species has returned to the level of dominance prior to Hurricane Hugo, in terms of both proportional representation of the fauna and in terms of captures per net hour.Monophyllus redmani appeared positively affected by Hurricane Hugo.Both its biomass dominance and numerical dominance in the tabonuco rain forest increased compared to prehurricane levels.However, only a slight increase occurred in actual numbers of captured individuals.This relative increase is mostly due to the decrease of the other two previously dominant species.The small increase in actual numbers may also be attributed, in part, to the rapid and sizable increase in the presence of flowering plants in the open forest understory after Hurricane Hugo.
cannot countermand its decline.The localized occurrence of S. rufum in the tabonuco rain forest, its decline in density after the hurricane, and the dramatic reduction in juvenile representation in the population suggest that this species may be in danger of extirpation.This content downloaded from 137.99.143.45 on Tue, 25 Apr 2023 19:52:33 UTC All use subject to https://about.jstor.org/terms rain forest, S. rufum is the only frugivorous bat known to consume Manilkara in appreciable amounts, where it comprises approximately 23 percent of the diet of S. rufum(Willig & Gannon 1994).The extirpation of S. rufum could have longterm and far reaching adverse effects on the life history of this tree species in particular.Moreover, the absence of A. jamaicensis from the tabonuco rain forest immediately after the hurricane suggests that S. rufum initially may play a critical role in recovery as the only bat dispersal agent of some early successional plants.It is presently unclear how many other plants could be adversely affected, or how the absence of S. rufum could affect the structure of the forest as it recovers from the hurricane.This content downloaded from 137.99.143.45 on Tue, 25 Apr 2023 19:52:33 UTC All use subject to https://about.jstor.org/terms . On 18 September 1989, the eye of Hur-

TABLE 1 .
Sample size of three dominant bat species captured during the dry (D) and rainy (R) season from 1987 to 1991 at a single netting site in the Luquillo Experimental Forest near El Verde Field Station.

TABLE 2 .
Effects of sex and hurricane on home range and foraging range of S. rufum based upon a pure model I twoway analysis of variance.MCP = Minimum Convex Polygon, MAP (0.50) = 50% Minimum Area Probability, MAP (0.95) = 95% Minimum Area Probability, S x H = sex by hurricane interaction; home range is based upon both day roost and night roost captures, whereas foraging range is defined by night roost captures only.