Air temperature, vapor pressure deficit, and light intensity microclimatic gradients were examined along four forest edge and four paired forest interior transects in the East and West Usambara Mountains, Tanzania. Between 14 August 1995 and 11 August 1998, 287, 282, and 196 air temperature, vapor pressure deficit, and light intensity gradients, respectively, were measured along the four forest edge and four interior transects. The relationship between microclimate and distance from the forest edge was examined using piecewise linear regression. All microclimatic gradients were classified into one of nine shapes based on the sign and the size of the two estimated slopes. The relative frequency in the shapes of 65 percent of air temperature gradients, 52 percent of vapor pressure deficit gradients, and 62 percent of light intensity gradients along forest edge transects exceeded the relative frequency of these same shapes along forest interior transects, indicating that a majority of the forest edge microclimatic gradients measured were influenced by edge effects. Yet this result also indicated that approximately one-third of all air temperature and light intensity gradients and nearly one-half of all vapor pressure deficit gradients recorded during this study were affected by factors independent of edge effects per se, and that forest edge microclimatic gradients were temporally nonconstant. For air temperature and vapor pressure deficit gradients, low spatial but high temporal variation existed in estimated edge width and the relative change in microclimate between the forest edge and interior. For light intensity gradients, both high spatial and temporal variability characterized estimated edge width and relative change in microclimate between the forest edge and interior. The pooled mean edge width and relative change in microclimate between the forest edge and interior across the four forest edge transects for air temperature, vapor pressure deficit, and light intensity gradients were 94.1 m and 2.00°C, 82.6 m and 0.29 kPa, and 60.5 m and 10.6 joules/sec/m², respectively. These results suggest that forest edge microclimatic gradients in general may be inherently dynamic and nonconstant.

Temporal and spatial changes in floristic composition over a 15-year period in an Atlantic tropical forest in Brazil were analyzed. The data were recorded in five 0.5-ha (50 × 100 m) plots within a 22,000-ha forest reserve. In each plot, all stems ≥10 cm diameter at breast height (DBH) were repeatedly inventoried. The overall number of species among all plots was steady over time but changed within a given plot. In all plots, the proportion for the disappearance of species was 7 percent and that for immigration of new species into the plots was 9.5 percent. Approximately 65 percent of the disappearance occurred between 1986 and 1989, and 74 percent of the immigration happened after 1986. The observed changes in composition evidently were due to natural disturbance and climatic events, such as an unusual four-month drought. Disappearance of species from the plots did not denote regional extinction of those species, because they persisted elsewhere in the forest reserve. The changes in species composition within the plots and persistence outside them, however, indicated that large areas are required for in situ conservation of tropical tree species. These areas must be large enough for population changes to occur without negative effects on genetic and demographic structures of tree species.

A tree species replacement sequence for dry broadleaved forests (tropical hardwood hammocks) in the upper Florida Keys was inferred from species abundances in stands abandoned from agriculture or other anthropogenic acitivities at different times in the past. Stands were sampled soon after Hurricane Andrew, with live and hurricane-killed trees recorded separately; thus it was also possible to assess the immediate effect of Hurricane Andrew on stand successional status. We used weighted averaging regression to calculate successional age optima and tolerances for all species, based on the species composition of the pre-hurricane stands. Then we used weighted averaging calibration to calculate and compare inferred successional ages for stands based on (1) the species composition of the pre-hurricane stands and (2) the hurricane-killed species assemblages. Species characteristic of the earliest stages of post-agricultural stand development remains a significant component of the forest for many years, but are gradually replaced by taxa not present, even as seedlings, during the first few decades. This compositional sequence of a century or more is characterized by the replacement of deciduous by evergreen species, which is hypothesized to be driven by increasing moisture storage capacity in the young organic soils. Mortality from Hurricane Andrew was concentrated among early-successional species, thus tending to amplify the long-term trend in species composition.

The distribution of lianas (woody climbing plants) on trees in a lowland “liana forest” of northeastern Bolivia was clumped and varied with characteristics of individual trees and tree neighbors. In twenty-four 900-m² square plots established to estimate tree (≥10 cm DBH [diameter at breast height]) and liana (≥2 cm DBH) densities and to count the number of lianas a tree carried, we estimated a mean of 65 tree species and 51 liana species per hectare. Mean tree density at the study site (564 trees/ha, SE = 23.7) was similar to other tropical sites, but mean liana density was much higher (2471 lianas/ha, SE = 104.3). Basal area of trees ≥10 cm DBH was low in Oquiriquia (19.2 m²/ha) in comparison to other tropical forests. Liana diversity, as expressed by the ratio of liana/tree species, was higher in this forest than in any other so far reported. Of trees ≥10 cm DBH, 86 percent carried lianas. Four tree species (Astrocaryum aculeatum, Euterpe precatoria, Xylopia sericea, and Astronium fraxinifolium) had a lower proportion of liana-infested individuals than expected based on the mean percent of liana infestation in this forest. Forest plots with similar tree species composition did not have similar liana composition or liana loads per tree, which suggests that lianas and trees have no specific associations with each other. Lianas showed an aggregated distribution on trees, suggesting a facilitation process in which new lianas use already established ones to climb trees. Lianas of four different climbing mechanisms climbed a similar number of trees. Plots in the forest with high palm density also had high liana density, suggesting that palms and lianas respond positively to common forest conditions in the study site (perhaps related to successional forest status). Larger-diameter trees carried more lianas than slender trees, but this relationship was affected by the density of trees 10–30 cm DBH surrounding each tree, which suggests again that the successional stage of the forest in which a tree grows affects the number of lianas a tree carries. We found little evidence to support the idea that lianas were more likely to climb some tree species than others. Instead, larger trees and trees growing in the vicinity of trees 10–30 cm DBH, tended to have more lianas, perhaps as result of longer exposure to liana infestation.

Clonal reproduction is a common feature in tropical herbaceous plants and has been shown to be important for recruitment and population maintenance. Although numerous phenological responses (i.e., leaf flushing, flowering, and fruit maturation) have been shown to be associated with seasonality in the tropics, little is known about the relationship between seasonality and levels of clonal reproduction. I examined the temporal and spatial variation in clonal reproduction and its potential trade-off with growth and survival for a tropical understory herb, Aechmea magdalenae (André) André ex Baker, Handb. Bromel (Bromeliaceae). I addressed the following questions: (1) how do vegetative growth, survival, and clonal reproduction (i.e., rhizome and ramet production) vary between seasons and between years?; (2) how do these factors differ among populations?; and (3) do simulated dry and wet season light levels cause similar variation in rosette survival and vegetative and clonal reproduction, as found in natural populations? Rhizome and ramet production in A. magdalenae was found to be highly influenced by seasonality. During the transition from dry to wet season, there was a significant decrease in survival and a significant increase in clonal reproduction, and rosettes tended to grow at a slower rate. Rosettes in the location with the longest dry season had significantly lower survival, higher clonal reproduction, and a slower growth rate. Clonal reproduction also started at smaller rosette sizes in the most seasonal forest. When light levels were manipulated, rhizome and ramet production were significantly greater under the high light conditions, and rosette growth (change in longest leaf length) was significantly lower. Results from this study suggest that A. magdalenae responds to higher light levels by shifting investment from vegetative growth to clonal reproduction. Possible implications of this response to light for population demography are discussed.

Clidemia hirta is a highly invasive shrub in tropical forests throughout the world, but has had little success invading mainland sites and undisturbed forests. In the early 1990s, this plant was found to have invaded an unexpected site, an undisturbed continental tropical forest at Pasoh, Peninsular Malaysia. In 1997, a study was conducted of the C. hirta population at the Pasoh Forest Reserve. A demographic survey of the 50-ha long-term research plot at Pasoh located 1002 C. hirta individuals, 69 of which were reproductive at the time of the study. All but 8 individuals were located in high light gaps or gap edges. Relative growth rates were significantly higher in gaps and gap edges than in the understory, and no reproductive individuals were found in the understory. Mean plant size and dry biomass density increased steadily over the course of the study, while the observed mortality rate was 0 percent over two months. The biomass density of Clidemia at Pasoh was <500 g/ha at the conclusion of this study, but because the population is confined almost exclusively to high light environments, its density in these sites is much higher. The location of C. hirta plants in gaps was significantly correlated with past disturbance by wild pigs, suggesting that soil disturbance and light availability are essential for their establishment. The implication of this study is that by competing with native species in gaps, C. hirta invasion has the potential to alter forest regeneration at Pasoh. Changing land use practices near the reserve have increased the number of wild pigs, and thus the level of disturbance, which may explain the recent success of C. hirta at Pasoh.

We investigated the relationships among interstitial salinity, leaf sclerophylly, plant vigor, and population density for the leaf galling insect Cecidomyia avicenniae (Diptera: Cecidomyiidae) on its host plant Avicennia germinans (Avicenniaceae). Sampling was done in six mangrove stands and in one varzea forest of Maranhão, northeast Brazil. At each site, ten shoots were randomly taken on five A. germinans trees. From each shoot we counted the total number of galls and recorded the shoot length (cm). We also recorded the average length, width, total area, and biomass of leaves per shoot. Leaf sclerophylly was quantified by leaf biomass per unit area (g/cm²). Samples of interstitial water were taken by a 1.3-cm PVC tube with 80 cm of depth, and salinity (ppt) was measured with a refractometer. Leaf sclerophylly showed a positive relationship with interstitial salinity (R² = 0.77, P < 0.05). We also observed positive relationships between gall density per unit of leaf area (cm²) and salinity (r = 0.36, P < 0.05), and between gall density and leaf sclerophylly (r = 0.40, P < 0.05). The salinity and the leaf sclerophylly together explained 22 percent of the variation in gall density of C. avicenniae. We found a negative relationship between the number of galls per centimeter and shoot length (R² = 0.50, P < 0.05). Thus, longer shoots of A. germinans showed lower gall density. Our results suggest that the gall density of C. avicenniae on A. germinans is affected by the salinity of host plant habitat and by leaf sclerophylly along an interstitial salinity gradient.

We provide the first data on nocturnal pollination by bees. Heterophragma quadriloculare is a self-incompatible hermaphroditic tree solely pollinated at night by the carpenter bee Xylocopa (Mesotrichia) tenuiscapa Westwood, whose pollinating flights were lunar insensitive and unaffected by low nighttime temperatures (2–14°C). The number of carpenter bees visiting a tree per minute and the number of flowers visited per foraging bout were positively related to the size of the floral display. We found that floral display size was negatively related to fruit set but positively related to absolute fruit numbers. In pollen carryover experiments, cross-pollen did not travel far in sequences of non-emasculated flowers (up to the 4th flower), but traveled farther in emasculated flower sequences (up to the 15th flower) as evidenced by fruit set in these sequences. The relatively poor fruit set on trees with many open flowers may have resulted from deposition of self-pollen on touch-sensitive stigmas, which close on first contact. Because large-girth trees tend to have large floral displays, trees may increasingly realize a higher percentage of fitness through male rather than female function as they age. The steady-state flowering phenology of H. quadriloculare may limit reduction in the female function compared to the cornucopia strategy of many other members of the Bignoniaceae and may also provide a reliable resource for X. tenuiscapa.

In many areas of the humid tropics, relatively small patches of forest is the habitat most commonly available for forest-dwelling organisms. To assess resident hummingbird use of forest fragments near the Las Cruces Biological Station in southwestern Costa Rica, we mist-netted birds in five fragments (0.3–20 ha; ca 1300 m elev.) and the station's 226-ha forest preserve (ca 900–1280 m elev.). From January 1994 through mid-March 1999, we recorded 1069 captures of 21 hummingbird species during 49,900 net hours. Species richness, diversity indices, and capture rates increased asymptotically with patch size, as expected. We captured 16 hummingbird species in both the largest fragment (20 ha) and in the nearby forest preserve. Non-forest hummingbirds did not occur more frequently in the fragments than in the preserve, and all fragments supported a mixture of forest-interior and canopy-dwelling hummingbird species, along with a diverse group of hummingbird-pollinated plants. Phaethornis guy was common at all sites and visited >13 plant species in the fragments during the 1998 dry season (based on analyses of pollen collected from the birds' heads). In contrast, Eutoxeres aquila was found at only the largest three sites (10, 20, and 226 ha) and carried pollen from only one plant species during the 1998 dry season. Lampornis castaneoventris, a high-elevation hummingbird species, was far more common in the fragments than in the preserve, while three species typically found at lower elevations were rare in the fragments and common in the preserve. These distribution patterns could be influenced by local climatic differences because the fragments are higher and receive more cloud mist and annual precipitation than the preserve. Our data show that pollen loads on hummingbird species vary, and this has implications for gene flow among hummingbird-pollinated plants. Although our results are consistent with previous reports that hummingbird species may be less affected by deforestation than insectivorous avian species such as army ant followers, it is important to note that from the perspective of the genetics and demographics of hummingbird-pollinated plants, the important issue is whether appropriate pollinator species are able to move plant propagules among subpopulations of plants in fragmented landscapes. We conclude that the biodiversity of tropical forest fragments as small as 10 to 20 ha is high enough to warrant long-term preservation of these areas, especially if they are located near larger tracts of forest.

The fauna of water-filled tree holes in Neotropical forests is not well documented. Cumulatively, 54 macroinvertebrate and 5 vertebrate taxa were found in artificial and natural tree holes censused over four wet seasons on Barro Colorado Island, Panama. Most of the species were in the insect order Diptera, occurred as aquatic larvae in tree holes, and were detritivore/omnivores. Half (49%) of the collected species are considered specialists in this and similar container habitats, and three invertebrate taxa were previously unknown from tree holes. Successional patterns were weak in tree holes, but some taxa predictably colonized holes shortly after they were filled. The mosquito Culex urichii was more common and abundant in artificial than in natural tree holes; occurrence frequencies and densities of most other taxa were similar between hole types.

Baits and soil cores were used to quantify temporal variation in the use of cattle dung by two species of subterranean termites, Amitermes beaumonti Banks and Hoplotermes sp. nov. (hereafter referred to by genus), during the 1991 dry season in Guanacaste, Costa Rica. The densities of termite foragers of both species were evaluated in cattle dung patches (baits), in the soil beneath each patch (litter, 0–3.5 cm, and 3.5–7 cm), and in the same layers from adjacent dung-free control sites. Dung patch and control plots were sampled 3, 10, 60, or 110 days after 1450-ml dung patches were deposited on the pasture. Amitermes foragers were abundant in 3-, 10-, 60-, and 110-day-old dung patch plots with a mean (median) ± SD of 100 (59) ± 120 individuals/bait plot, and only 8 (3) ± 11 individuals/control plot. Hoplotermes foragers were less abundant, with a mean of 22 (0) ± 87 individuals in dung patch plots and 2 (0) ± 9 in control plots. Hoplotermes was most abundant in the surface litter under 10-day-old dung patches, where an average of 39 (0) ± 105 individuals were found. In association with dung patches, Amitermes density was highest inside or directly beneath the dung patch and lowest in the 3.5- to 7.0-cm microsite. In control areas, however, Amitermes was most abundant in the 3.5- to 7.0-cm microsite. Amitermes and Hoplotermes were found in 35 and 9 of the 54 control plots, respectively. We believe that the high termite densities associated with cattle dung patches can be explained by the energetic and microclimatic requirements of the termites and the physical structure of the dung. Although neither species of termite is a mound builder, these termites use specialized behaviors to exploit and modify dung patches such that foragers have access to concentrated food resources and partial protection from the dry climate and from their invertebrate predators.

The high diversity of amphibians in the central Amazon Rain Forest allowed us to examine the influence of tadpole antipredator defenses on assemblage structure and composition within bodies of water (α diversity) and among aquatic sites (β diversity) at a local scale. During a three-year study of tadpole assemblage composition, we found that the anuran community used a variety of bodies of water for reproduction; these ranged from streams and streamside ponds to isolated forest ponds. The distribution of several tadpole species was negatively related to fish density, while other species coexisted with high densities of fish. Tadpole size did not ensure survival against fish, and few tadpoles avoided fish by hiding in the leaf litter. Controlled predation experiments using a single tadpole species in a no-choice situation were conducted over 24- to 48-hour periods. Nearly all species of tadpoles that occurred in habitats with high fish densities were unpalatable to fish (except Centrolenella oyampiensis), indicating that unpalatability is a major adaptation allowing tadpoles and fish to coexist in this system. Unpalatability (to fish), however, was not an effective antipredator defense against odonate larvae, the other major tadpole predator in this system. The combination of predation pressure and the antipredator traits exhibited by individual species largely determined the composition of tadpole assemblages in individual bodies of water (α diversity). The heterogeneous distribution of predators among bodies water and the diversity of antipredator defenses exhibited by larval amphibians facilitated high diversity in this community (β diversity).

Two tropical upland forests, Mount Kupe in Cameroon and Bosque Protector Palo Seco in Panama, were compared in terms of herpetofaunal species richness and density of individuals. Based on rarefaction, whereby samples are standardized for abundance, Palo Seco had significantly more species of frogs and lizards. Extrapolations to total local species richness, by fitting the Michaelis–Menten equation to the species accumulation curves and by using Chao's estimator, yielded divergent results: more lizard species on Mount Kupe, and an equal number of frogs at both sites. These disparities can be accounted for by differences in evenness, which was higher in Palo Seco. Frog density was significantly higher on Mount Kupe, snake density significantly higher in Palo Seco, and lizards exhibited no density difference. Overall, the results revealed a less consistent pattern and more moderate differences than what is known from southeast Asian–Central American comparisons. This outcome is discussed in the light of available knowledge, but quantitative data from African forests are too sparse to allow general conclusions.

Increasing human activity in the Yucatan peninsula has led to declines in older stages of successional forest, threatening regional habitat diversity. To determine potential effects of this habitat loss on the region's avifauna, we examined the relative use of different forest stages by resident and migrant birds during the nonbreeding season. We used the fixed-width transect method to compare the distribution, abundance, and diversity of forest birds in early (five to ten years old), mid (15–25 years), and late (>50 years) successional forests in the state of Campeche, Mexico, in the south–central part of the peninsula. All stages of successional forest had highly similar bird assemblages and did not differ in bird abundance or diversity. Both migrant and resident birds occurred across the successional gradient. The majority of habitat specialists, however, were resident birds restricted to late-successional forest, indicating that early secondary growth may not be suitable for all species. Furthermore, resident birds that typically participate in mixed-species flocks attained their greatest densities in the oldest forest habitat. Rapid recovery of pre-disturbance physiognomic features, in addition to high levels of habitat connectivity in the region, may contribute to similar bird communities across a range of successional stages. The high degree of edge characterizing much of the forest mosaic also may allow birds access to different seral stages. Loss of late-successional forest, however, is likely to adversely affect the subset of resident avifauna that depends on unique features of mature habitat such as snags, large trees, and climatic buffering. Conservation efforts in Campeche should focus on the specialized requirements of the most habitat-restricted species while preserving the current landscape mosaic characteristic of the small-scale shifting cultivation system.

The Amazon Forest and the Cerrado of central Brazil share an extensive zone of integration. Each of these major ecosystems remains poorly studied; however, the composition of communities at the contact zone is even less well described. Small mammals were sampled during a ten-week period in a zone of contact between Amazonian forest and Cerrado. The vegetation in the area was a complex mixture of tropical and semi-deciduous forest, and grassland and savanna woodland formations. Mammals were marked and recaptured on 0.25- to 0.5-km transects and vouchers were collected. A total of 19 species was caught (13 rodents and 6 marsupials) in 4621 trap nights. Transitions from one vegetation type to another were abrupt with ecotones frequently < 5 m in width. The highest α-diversity was 7 species in cerrado. Ten of the 19 species were restricted to one habitat and 6 were restricted to two habitats. Three generalist grassland species occurred in more than two habitats, all variations of savanna. These data support other research indicating high habitat specificity of vertebrates and plants in the Brazilian savannas. The diversity of habitats within reserves in tropical grasslands must be given consideration in the design of protected areas in order to maximize the conservation of mammalian biodiversity.

Although there is growing recognition of the importance of small mammals in savanna ecosystems of east Africa, the ecology and behavior of these organisms are not well known. We conducted studies on the dietary preferences of one rodent, the pouched mouse (Saccostomus mearnsi), which is the dominant small mammal in a widespread habitat in central Kenya. To establish the food preferences of this species, we: (1) conducted cafeteria trials in the laboratory whereby individual mice were offered seeds and a selection of herbaceous vegetation from the habitat; (2) surveyed the habitat for piles of vegetation clipped by these mice and identified these clippings to species; and (3) estimated the abundance of plant species in the habitat to determine if the mice were clipping species in proportion to their availability. Our results indicated that S. mearnsi prefers green vegetation to seeds, at least during wet seasons, and that it has a strong preference for forbs. In cafeteria trials, S. mearnsi chose 93 percent green vegetation and only 7 percent seeds. The forb Commelina africana was highly preferred over other food items offered. In the habitat, two forbs, C. africana and Monsonia angustifolia, constituted > 40 percent of the piles of clippings, although these two species together made up < 4 percent of the available vegetation. Grasses also were clipped frequently, although in lower proportions than their availability in the habitat. The herbivorous habits of S. mearnsi establish the potential of this species to compete for resources with ungulates, which are abundant in this habitat.