Canopies house a major portion, perhaps the majority, of forest biodiversity (Erwin 1982; Lowman & Wittman 1996), but this assertion has been met with criticism (Hammond 1990; Hammond et al. 1997; Walter et al. 1998; Stork & Grimbacher 2006.). Advances in canopy access enabled the discovery of vertical stratification in forests, i.e., clear delineations in communities across the ground-overstory vertical height (Basset et al. 2003a). In general, tropical forests are thought to increase in species diversity with vertical height (Basset et al. 2003a), but this pattern has been disputed recently (Ulyshen 2011). In particular, there is debate regarding how temperate and tropical forests might vary in patterns of diversity with vertical height (Parker & Brown 2000; Basset 2001; Basset et al. 2003a).

Diversity generally was thought to increase with vertical height in tropical forests, but this hypothesis is still under debate (Basset et al. 2003a; Ulyshen 2011). For example, fruit-eating nymphalid butterfly communities in Ecuador, and beetle assemblages in Australia, showed that canopy and ground species diversity was about equal (DeVries et al. 1997), and small mammal diversity decreased with height in a Brazilian Atlantic rain forest (Vieira & Monteiro-Filho 2003). Collembola communities were stratified in tropical rainforests, but diversity did not vary between the ground and canopy (Rogers & Kitching 1998). Flying insects in rainforests in Panama, Papua New Guinea, and Brunei stratified vertically and, in some cases, were more abundant in the canopy (Sutton et al. 1983). Still, a family of moths in a Costa Rican rain forest showed an increase in species diversity with vertical height, but the pattern reversed in a different family of moths (Brehm 2006). Researchers point to tree architecture or resource variation to explain deviations from the general patterns, especially among smaller taxonomic groups (Basset 2001; Basset et al. 2003a, b).

In general, it is expected that tropical, but not temperate, forests show increasing diversity with vertical height (Basset 2001; Basset et al. 2003a). However, Florida biodiversity is largely driven by the confluence of temperate and tropical species commingling on the peninsula (Webb 1990; Kautz & Cox 2001). In fact, Florida forests are classified as temperate or tropical only by tree species composition (Greller 1980). Thus, it remains unclear what pattern of stratification, if any, might be structured in Florida forests (Su & Woods 2000). To further this endeavor, we examined how richness and diversity of arthropods in an upland hardwood forest in south-central Florida changed with vertical structure.

Materials and Methods

St. Sebastian River Preserve State Park encompasses about 8,800 ha in Brevard and Indian River counties, Florida (Florida Division of Recreation and Parks 2005). We delimited a 120 × 90 m (1.08 ha) plot in a 6.5 ha patch of mesic upland hardwood forest at the confluence of the North and South Prongs of the St. Sebastian River within St. Sebastian River Preserve State Park in Indian River County, Florida (generally located at 27.831400°N, 80.509100°W). The specific study site is characterized by sandy clay soils and upland hardwood forest, a rare habitat in the park (less than 0.04% of land area) (Florida Division of Recreation and Parks 2005). Upland hardwood forest is a type of mesic upland habitat that includes a high diversity of plants and animals, including the overstory trees Florida maple (Acer saccharum floridanum (Chapm.) Small & A.Heller) (Sapindaceae), pignut hickory (Carya glabra (Mill.) Sweet) (Juglandaceae), southern hackberry (Celtis occidentalis L.) (Cannabaceae), white ash (Fraxinus americana L.) (Oleaceae), sweetgum (Liquidambar styraciflua L.) (Altingaiceae), southern magnolia (Magnolia grandiflora L.) (Magnoliaceae), loblolly pine (Pinus taeda Blanco) (Pinaceae), white oak (Quercus alba L.) (Fagaceae), live oak (Q. virginiana Mill.) (Fagaceae), and laurel oak (Q. hemisphaerica W. Bartram ex Willd.) (Fagaceae) forming a closed canopy (Florida Natural Areas Inventory 2010).

All hardwood trees > 150 cm diam at breast height were included as candidates for sampling. Trees judged unsafe to climb (e.g., major signs of damage, lack of suitable anchor sites) were excluded, and 7 trees were randomly selected for sampling from the 22 that met the above criteria. Bole height, crown height, and diam at breast height were recorded. Additionally, drip line diameter at each cardinal direction and overstory density via convex spherical densitometer were recorded (Jennings et al 1999; Fiala et al. 2006).

Fig. 1.

Arboreal pitfall tap design. Two sheets of metal flashing are cut to length and punched with holes. Next, a cylinder and funnel are formed and secured with pop rivets or similar fastener. A plastic bag is inserted into the cylinder, and the cylinder and funnel are attached with tie straps. The bag is then filled with about 100 mL of 98% propylene glycol and secured to a tree trunk of branch such that arthropods slide down the funnel and into the collection fluid.

Each tree was fitted with 5 traps: 2 at the soil line and 3 arboreally. Two pitfall traps were installed by burying commercially available polystyrene drinking cups with a 9 cm diam flush with the soil surface. Both pitfall traps were positioned in random directions relative to the tree (determined by randomly selecting a compass azimuth with 2° resolution); one at 1 m from the tree base, and the other equidistant to the tree base and the drip line. The remaining 3 traps, of a new design inspired by Pinzón & Spence (2008), were placed aboveground in the tree (Fig. 1). These arboreal pitfall traps were constructed from two 40 cm lengths of 25 cm wide aluminum flashing measuring less than 0.01 cm thick. One length was used to form a funnel and the other a collector. The traps were tightened against trunks or branches of trees with string. Trap locations were in part determined by individual tree structure, such that the flexible funnel could bend to the contours of the tree. This formed a smooth transition zone across which arthropods may fall into a 6 mL plastic bag placed within the collector and secured with plastic tie straps (Fig. 1). About 100 mL of 98% propylene glycol was used as a preservative in all traps instead of the more typical 50% solution because of the probable dilution by rainwater (Hall 2006). We opted against using ethylene glycol to avoid poisoning non-target animals (Hall 2006; Jud & Schmidt-Entling 2008).

Arboreal traps were placed using a canopy access method, termed single rope technique (SRT) (Moffett et al. 1995; Lowman 2009). A line was placed over a crotch in the tree's upper crown. One end of the line was anchored to the base of another tree, and ascended the other end by use of a harness and typical rock-climbing and caving equipment. This allows for maximum maneuverability within the tree, and was far less disruptive than spike climbing or the use of heavy machinery (e.g., cherry picker, crane).

Thirty-four traps were installed in the 7 trees from 1 to 12 m aboveground (1 tree had suitable sites for only 2 arboreal traps). Traps caught arthropods for 10 to 22 d. We removed all traps at the project's end, filtered trap contents through filter paper, and preserved filtered contents in 70% isopropyl. All arthropods collected in the traps were identified to morphospecies using a Leica M80 stereomicroscope (Leica Microsystems, Buffalo Grove, Illinois, USA).

The number of morphospecies in each order were tallied in 2 m height increments. Alpha diversity using Shannon-Weaver and Simpson diversity indices and pairwise â diversity with the Sørensen-Dice index were calculated (Dice 1945; Sørensen 1948; Hill 1973). The Chao1 index and the abundance-based coverage estimator (ACE) were used to extrapolate species richness from our data (Chao 1984, 1987; Palmer 1990; Colwell & Coddington 1994; O'Hara 2005; Chiu et al. 2014). Rarefied species richness was calculated at 2 m height increments to predict total species richness in the habitat (Hurlbert 1971; Heck et al. 1975). Ground traps near or far from the tree base were tested to determine if they captured distinct arthropod communities using an analysis of similarity (ANOSIM) with the Bray-Curtis index and 10,000 permutations. Polynomial regression was used to test if arboreal trap height predicted morphospecies richness. Lastly, rarefaction curves were used to understand the level of sampling in the environment.

Results

We collected a total of 830 arthropods from the 34 traps, amounting to 103 morphospecies across 15 orders. Coleoptera was the most morphospecious order (31 morphospecies), followed by Diptera (21 morphospecies), Araneae (13 morphospecies), and Hymenoptera (10 morphospecies). Several other orders contained less than 10 morphospecies. We identified only a single morphospecies for 3 orders: Decapoda, Diplopoda, and Phthiraptera. Morphospecies richness negatively correlated with trap height (b = −2.21, 0.11; R ² = 0.93; F _{2, 9} = 76.17; P < 0.0001) (Fig. 2). Araneae, Collembola, and Lepidoptera all showed higher morphospecies richness on the ground than associated with trees (Fig. 3). Indeed, overall pairwise β diversity was highest between the ground and vertical classes (Table 1). Among ground-level pitfall traps, arthropod communities captured 1 m from the tree base had different compositions than those positioned equidistant to the tree and drip line (ANOSIM P < 0.001; 10,000 permutations). The tree base has higher species richness (64 vs 49 species) and individual abundance (448 vs 214), but equivalent diversity (Shannon: 2.96 vs 3.01; Simpson: 0.903 vs 0.906) (Table 2). The number of traps deployed was not predictive of trap morphospecies richness, but rarefaction curves by height class indicated that our survey is a conservative estimate of total species diversity in the habitat (Fig. 4).

Discussion

Tropical forests are expected to have increased diversity with vertical height, whereas temperate forests are expected to have equivalent diversity levels within the ground and canopy layers (Bassett 2001; Basset et al. 2003a, b). However, our study in Florida forests found that species diversity and richness decrease with height (Fig. 2). Alpha diversity was higher on the ground than the canopy. Arboreal traps decreased in diversity with height, and ground traps decreased in diversity with distance from the tree (Table 2). Species richness and individual abundance also followed these patterns (Table 2). Beta diversity was highest between the ground and arboreal traps, and more similar height classes had more similar beta diversities (Table 1). Although this pattern was unexpected, it is not unique among studies of diversity with vertical height. For example, neotropical litter-dwelling ants were more common on the ground than the canopy (Longino & Nadkarni 1990). Lepidoptera in Borneo showed a decrease in abundance in the forest canopy relative to lower levels, which might be due to variation in resources (Schulze et al. 2001).

Fig. 2.

Polynomial (b = −2.21, 0.11; R ² = 0.93; F _{2, 9} = 76.17; P < 0.0001) regressions of trap height and mean species richness for arthropods collected in an upland hardwood forest in Florida, USA. Species richness declines with vertical height.

Fig. 3.

Mean morphospecies richness categorized by order across 2 m height classes in an upland hardwood hammock in Florida, USA.

Most orders in our study were widely dispersed across the ground-canopy gradient (Fig. 3). For example, Acari were found in equivalent species richness at all levels (Fig. 3). Araneae were found on all but the highest areas of tree canopies, but the majority of Araneae morphospecies were collected on the ground. This likely reflects the location of prey abundance. Orthopterans seemed to be limited to the lower trunk and canopy but were never found on the ground, likely because they could evade terrestrial traps, though it also may reflect the location of food sources for Orthopterans. Collembola richness was highest on the ground, but at least 1 morphospecies was detected at nearly every height class (Fig. 3). Because Collembola commonly live in soil layers, surveys of canopy soils likely would reveal more species. Several orders normally associated with the terrestrial level, including Blattodea and Coleoptera, were found only higher up on trees. The reverse pattern was found for Lepidoptera, which were found only on the ground (Fig. 3). This is likely an effect of seasonality and trapping methodology, both of which resulted in more larval Lepidoptera than adults. These differences in stratification across orders might be indicative of the different microhabitats that primary and secondary consumers occupy or use differently. However, some of the differences we detected might change seasonally and interannually, as has been suggested elsewhere (Longino & Nadkarni 1990; Schulze et al. 2001).

In this sense, our study is spatially and temporally limited; multiyr research along a latitudinal gradient would reveal how stratified diversity varies with seasons, yr, and as the tree community shifts from tropical to temperate. Further, rarefaction curves showed that many more morphospecies would be identified with increased sampling across all height classes. Further study could better resolve how taxa are stratified. Such studies also could reveal the extent to which invasive species may impact canopy diversity (Kaspari 2000). Ours and future investigations will contribute toward resolving the debate regarding the pattern of vertical stratification in temperate vs tropical forests. Our study revealed an unexpected pattern of vertical stratification in diversity at the confluence of temperate and tropical assemblages.

Table 1.

Pairwise β diversity using the Sørensen-Dice index (x̄ = 3.5) for collections of arthropods at different height classes (m above soil) in an upland hardwood forest in Florida, USA.

Table 2.

Diversity and richness statistics for arthropods in an upland hardwood hammock in Florida, USA. Arthropods were sampled on the ground at 2 distances from the tree trunk, or at 5 heights (m) above the soil on tree trunks. H is Shannon diversity index; Simp is Simpson index; Rich is the mean species richness per trap; Ind is the mean number of individuals per trap; Chao1 and ACE (abundance-based coverage estimators) are species richness estimates.

Fig. 4.

Rarefaction curves for size height classes of arboreal and ground pitfall traps collecting arthropods in an upland hardwood forest in Florida, USA. Steep curves indicate that further sampling would reveal additional morphospecies.