Comparative gliding performance of Anolis lizards among three Caribbean ecomorphs

Abstract

Anole lizards (genus Anolis) have repeatedly and independently occupied arboreal niches, and in parallel have evolved associated morphologies and locomotor behaviors. In arboreal vertebrates more generally, parachuting and gliding have frequently evolved in response to selective pressure for targeting and controlled landings during either falls or intentional jumps. Here we describe morphological features and aerial performance among five anole species representing the trunk-ground, trunk-crown, and crown giant ecomorphs. Aerial behavior within a vertical wind tunnel was used to evaluate gliding performance, including body orientation, velocity, and glide angle. All study species exhibited controlled non-equilibrium aerial descent. When accounting for phylogenetic relatedness among species, heavier species exhibited greater downward velocities and glide angles. Interspecifically, glide angles were also inversely correlated with wing loading. By contrast, gliding performance did not differ significantly among the three Anolis ecomorphs. Overall, that all study species engaged in controlled aerial descent despite exhibiting substantial morphological variation and being phylogenetically distant within the genus. These results suggest that controlled aerial behaviors may be widespread within Anolis, and that morphological differentiation among species may derive in part from selection for locomotor performance in the air.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12862-026-02586-w.

Introduction

Animal locomotor capacity is tightly linked to habitat because environmental geometry can influence complexity of those behaviors either necessary or advantageous for successful transit [1, 21, 36]. As a consequence, natural selection acting on locomotor performance often has resulted in morphological adaptations that enhance behavioral efficacy, ranging from subtle modifications of existing structure to the evolution of novel morphologies (e.g., [62]). Furthermore, novel behavioral patterns can phylogenetically precede specific morphological adaptation [45, 62], and there is thus a broad diversity in locomotor ability and morphological traits relative to specific habitat demands. Selection on arboreal taxa for the ability to glide has been no exception.

Gliding (i.e., descending aerial behavior powered by gravity, and with a horizontal component of motion deriving from aerodynamic forces on axial and/or appendicular structures; [10]) has evolved in concert with arboreality in numerous taxa [5, 10, 40, 45, 68]. This inclusive and conventionally used definition for gliding captures the continuum of descent behaviors observed in nature that convert gravitational potential energy into useful aerodynamic forces [10, 11, 59], including those classically defined as “gliding” [45] and as directed aerial descent [68]. All organisms that live in trees are likely to either fall, jump, or be displaced by wind from the canopy (e.g., [55]). There is thus increased selective pressure to slow descent, and in many cases to target and control landings via aerial maneuvering so to avoid hazards below (e.g., [42]), to minimize risk of injury upon impact (e.g., [13]), and to avoid in-flight obstacles (see [7, 10]). Indeed, animals may become airborne either intentionally or unintentionally, and gliding provides advantages in such diverse contexts as predator evasion [8], intraspecific interactions [40], and routine transit [10, 43, 59]. Morphological and behavioral adaptations that enhance the generation and control of aerodynamic forces may thus be targets of selection.

In general, aerodynamic forces are directly proportional to the surface area of such relevant structures as wings and flaps, and morphological adaptations for gliding typically involve expansion of both axial and appendicular structures, including expanded skin folds [19, 53, 70], interdigital webbing [13, 53], and novel integumentary membranes (i.e., patagia) that extend laterally from the body [9, 24, 40]. Furthermore, arboreal habitats are often spatially heterogeneous, which may influence the extent to which such specialized structures evolve [14, 19]. However, controlled aerial behaviors may precede dedicated gliding morphologies [45, 62]; many taxa are certainly adept at gliding yet lack obvious airfoils (see [5, 58, 59, 60, 68]). Instead, these animals typically possess more subtle morphological adaptations such as elongated limbs and a high projected surface area relative to body mass (i.e., low wing loading; [53, 62]). Limb elongation is directly related to aerodynamic stability, as well as to control of roll, pitch, and yaw, because longer limbs increase the moment arm about which useful aerodynamic torques are created about body axes (e.g., [37]). Also, a low wing loading associates with slower gliding that may reduce the risk of injury upon collision [12, 62].

Although gliding has been observed in many invertebrate and vertebrate taxa [59], varying degrees of gliding ability, subtle intra- and interspecific morphological variation that facilitates aerodynamic force production and control, and those evolutionary pathways for species with no obvious aerodynamic features have not been fully characterized. Active modulation and control of aerodynamic forces are clearly of biomechanical advantage [2, 3, 7, 10, 28, 54, 57], yet little attention has been given to the evolution of aerial performance for those taxa with no clear morphological adaptations to gliding (but see [5, 6]). Here, we describe relationships among gliding performance, various features of aerial kinematics, and morphological traits for a clade of lizards that both lack dedicated morphological structures to this end, and that vary in vertical use of microhabitats.

Specifically, lizards in the genus Anolis have repeatedly and independently filled different arboreal niches in the Caribbean islands, and have evolved correlated morphologies including body elongation and the shortening of hindlimbs, forelimbs, and tails [32]. Jumping to the ground is in the escape behavior repertoire of many anole species [34], and it has been suggested that aerial descent is a feature of anole ecology [4, 45] enabled by control of body posture and orientation when in the air (see [27, 44, 45]). Furthermore, “a slight wiggling of the tail” has been observed in gliding A. carolinensis, which also can engage in large-amplitude and high-frequency tail oscillations under microgravity [65]. However, associations between morphological features, body kinematics, and gliding performance have not been systematically studied in anoles.

Accordingly, we measured kinematic, biomechanical, and morphological features relevant to gliding in five species of anole representing three distinct ecomorphs (i.e., trunk-ground, trunk-crown, and the crown giant). Although aerial abilities of the trunk-ground ecomorph have been called into question [4], we hypothesized that all such arboreal and semi-arboreal taxa possess some capacity for directed aerial descent (sensu [10]), and that gliding would vary among ecomorphs such that species found at lower heights in trees (i.e., the trunk-ground ecomorph) would have poorer aerial performance (i.e., a higher vertical velocity, lower horizontal velocity, and a steeper glide angle) than species in the trunk-crown ecomorph that occupy higher arboreal niches. Species in the crown giant ecomorph were predicted to have very poor aerial performance relative to the trunk-crown ecomorph due to their extremely large body size [34] and a high effective wing loading. Given documented advantages of long limbs in gliding frogs, [37], we also expected that relatively longer fore- and hindlimbs among ecomorphs would correlate with better aerial performance.

Methods

Study taxa and husbandry

A total of twenty-three individuals from among five species of anole lizard was used in this study (Table 1; Supplemental Figure 1; Supplemental Table 1). Species were selected to include the traditionally characterized anole ecomorphs (see [30, 34, 36]): trunk-ground (A. cristatellus and A. sagrei), trunk-crown (A. carolinensis and A. coelestinus), and the crown giant (A. equestris). We note that the nomenclature for A. coelestinus has changed such that the species is now known as A. chlorocyanus (see [20, 29]), but we use A. coelestinus to remain consistent with the chosen phylogeny [47]. We included five A. carolinensis (mass: 4.53 ± 1.54 g; SVL: 61.60 mm), five A. coelestinus (mass: 10.30 ± 4.04 g; SVL: 70.52 mm), four A cristatellus (mass: 8.19 ± 2.82 g; SVL: 64.6 mm), five A. equestris (mass: 13.08 ± 1.75 g; SVL: 91.92 mm) and four A. sagrei (mass: 5.43 ± 2.38 g; SVL: 55.03 mm; Table 1). Study individuals of A. equestris were juveniles because adults are too large to glide within the wind tunnel, but their body dynamics and kinematics are likely representative of adults given isometric scaling in the species [61].

Table 1.

Species means (± standard deviation) for all morphology, performance, and kinematics

Statistical values (DF, F-statistic, and p-values) listed are from one-way ANOVAs; values that indicate significant differences are in boldface

All lizards were acquired from commercial vendors approved by the Office of Laboratory Animal Care (OLAC) at the University of California, Berkeley, and underwent a 2–6-week quarantine. Anolis carolinensis, A. sagrei, and A. cristatellus were housed with conspecifics in large (width: 50.8 cm, length: 50.8 cm, height: 99 cm) mesh-walled tanks, whereas A. coelestinus and A. equestris were housed either individually or paired (see Supplemental Material) in medium-sized (width: 40.64 cm, length: 40.64 cm, height: 76.2 cm), mesh-walled tanks. All lizards were provided water and live crickets ad libitum and were kept within temperature-controlled rooms (24 °C) with 12:12 h light: dark cycle. At the end of experiment, all A. carolinensis and A. sagrei individuals were anaesthetized by isoflurane inhalant and sacrificed by Pentobarbital (100 mg/kg, intracoelomic injection) overdose followed by double-pithing administered by the OLAC veterinary staff. All A. coelestinus, A. cristatellus, and A. equestris individuals were adopted or transferred to collaborators at the conclusion of the experiment in consultation with OLAC.

Flight trials

Gliding was studied within the flight arena of a vertically oriented wind tunnel (Wind Generator 01–10, Aerolab, Laurel, MD, USA; see [6] for photograph and schematic of experimental setup), which consisted of a rigid Fiberglas cylinder (diameter of ~ 71 cm), at the base of which was positioned a rotary fan that generated upward airflow. Two layers of mesh and a 10 cm layer of metal honeycomb (with ~ 5 mm wide holes) were positioned above the fan to reduce turbulence. A square-shaped contracting funnel was fastened atop the cylinder to increase wind speed and to make it more uniform in cross-section. A flight arena of square cross section with transparent 5 mm thick acrylic walls (width: 30.5 cm, length: 30.5 cm) was positioned above the contracting funnel. Two taller walls (each of 50.8 cm in height) were used to lengthen vertically the spatial volume within which lizards could be filmed, whereas the other two walls were shorter (with a height of 25.4 cm) to allow investigator access to the inside of the arena. The base of the arena consisted of a cloth mesh (elliptical holes with major axis length of 4 mm and minor axis length of 2.5 mm) positioned beneath a layer of wire mesh (with 2.54 cm squares). Mosquito netting was fastened to the top edge of the walls and was extended upward to surround a camera positioned above the floor of the arena. The net was sealed above and laterally to prevent lizard escape, but a small gap (length: 15 cm) was left unsealed above the two shorter walls to enable experimenter access to the flight arena. Room temperature during all flight trials was approximately 25 °C.

Flight trials were filmed at 400 frames/s using three synchronized cameras (HiSpec, Fastec Imaging, San Diego, CA, USA); cameras were positioned at one dorsal (i.e., top perspective) and two lateral perspectives (front and side of the flight arena), each of which was tilted ~ 20 ± above horizontal. The MATLAB [35] function “detectCheckerboardPoints” [17] was used to calibrate the cameras (in pairs) by recording and then digitizing points on a checkerboard via stereo image pairs (assuming a value for sigma of 3, whereby sigma is the bandwidth of the Gaussian filter used for corner detection; see [17]). A 10 square x 7 square checkerboard (square length of 16.33 mm) was first held so as to be visible in all three camera views and was then slowly shifted in orientation and position within the arena (typically occupying approximately 25% of the field of view of any given camera). Recordings of a Styrofoam ball (diameter of 4 cm; mass of 0.034 g) falling vertically through the test arena was used to generate a gravity vector that was subsequently used to align axes of reconstructed landmark data relative to global coordinates. Extrinsic and intrinsic reconstruction parameters along with distortion coefficients for each camera (see [17]) were then estimated using the calibration image stacks. Extrinsic parameters included rotation and translation of the three-dimensional global coordinate system to match the local system for each camera, whereas intrinsic parameters included focal length, optical center, and a skew coefficient that related each camera’s three-dimensional coordinate system to its recorded two-dimensional images.

All cameras were focused ~ 10 centimeters above the flight arena’s floor at an approximate target height best suited for filming aerial behavior. Given individual variation in lizard mass, the vertical windspeed was set at different values nominally specific to each individual so as to approximately match its equilibrium velocity (as observed during preliminary trials and trials which concluded with the lizard falling to the bottom of the test arena). Mean wind speed at the center of the flight arena, and at the nominal target height for filming, were measured using a hand-held anemometer; VelociCalc model #8346; TSI, Inc.). Mean wind speed was estimated by averaging three instantaneous wind speeds, as measured by holding the probe in the airstream for thirty seconds and recording the speed over durations of 10, 20, and 30 s. Spatial variation in wind speed across the working section was assessed by partitioning the bottom of the flight arena into 144 squares (each 2.54 cm in length), and then measuring wind speed at a distance of 10 cm (i.e., at the target gliding height) above each square. These measurements were made at a nominal wind speed in the working section of ~ 9.6 m/s, a value about 17% slower than the mean speed (11.6 m/s) used in flight trials. Wind speed at each square (except for three corner squares) was within 25% of the overall mean among all squares (9.31 ± 0.726 m/s, with a turbulence intensity of 0.078 as indicated by the ratio of wind speed standard deviation to wind speed mean), and was thus less than the typical spatial variation (turbulence intensity ~ 1) found in tropical rainforests [38]. Furthermore, the mean windspeed used for biomechanical analyses was approximately 3% higher than the average speed among all 144 squares, as it was measured at the center of the flight arena where speeds tended to be highest (Supplemental Figure 2). The magnitude of vertical wind speed variation in the wind tunnel is comparable to that of naturally occurring spatial variation along vertical gradients within forest canopies (see [6]).

Fifteen gliding trials were conducted for each individual lizard; the duration of each airborne trial ranged from 1 to 10 s. Each trial per lizard was repeated following a minimum rest period of 24 h. Immediately prior to each trial, the lizard’s mass and SVL were recorded, and landmarks were then applied to the lizard’s skin using white non-toxic paint at four points along the lizard’s midline, on the head between the eyes, midway between the pectoral and pelvic girdle (or midbody), midway between the pelvic girdle and the tail tip, and at the tip of the tail (Supplemental Figure 3). To initiate a flight trial, a lizard was released by hand into the tunnel’s airflow at a height approximately 25 cm above the arena’s floor, such that the cranial direction of the lizard was directed towards the front camera. After release of the lizard and its subsequent translation in the airstream, the lizard either adhered to one of the acrylic walls of the arena, gripped the mesh floor of the arena upon gliding and landing, or reached the end of the filming duration. Given the cameras’ frame rates and the pre-specified spatial region of interest, only ~ 5 s of footage could be recorded per trial, and cameras were thus turned off immediately after a gliding event.

Kinematic analysis

Using footage from the top camera, we identified those videos in which the lizard unambiguously performed a gliding event, i.e., when traversing horizontally across any portion of the test section. A gliding event began when the lizard assumed a characteristic skydiving posture (see [45]), and when all regions of its body (excluding the tail) had lost physical contact with either the arena walls or with the investigator’s hand. As the tails of Anolis are relatively long, tails of gliding lizards often intermittently contacted the walls during filming. However, such contact was typically short in duration with no obvious adhesion, and we thus assumed any associated physical effects likely to be very small relative to other forces acting on the body (i.e., gravity, lift, and drag). Among trials in which a gliding event occurred, footage from two of the three cameras was used to select videos for detailed analysis, whereby all anatomical landmarks must have been visible in both two camera views at the start and conclusion of a gliding event. For each trial that met these criteria, we then identified those frames in which the gliding event began and ended.

For each individual lizard, we identified five exemplary glides from all video sequences, (i.e., those in which the animal either traversed a major portion of the test section or in which the lizard was distant from the walls of the arena). For each of these glides, aforementioned anatomical landmarks were tracked using DLTdv point-tracking software in MATLAB [18]. Only two-camera pairs (i.e., front and side, front and top, and side and top) were used for analysis because there was substantial variation in recorded body regions of interest among the cameras; the function “detectCheckerboardPoints” requires all cameras to record the same region of interest when calibrating more than two cameras. Furthermore, given the complexity of the gliding behavior and the fixed camera positions, some landmarks were occasionally occluded during glides. For all trials, we calculated the proportion of frames for which each landmark was visible from each camera. If a landmark was occluded in at least 40% of frames for one particular camera view, the camera pair excluding that camera was used for analysis. If a landmark was occluded in at least 40% of frames for more than one camera view, the trial was eliminated from subsequent analysis. If more than one camera pair had similarly low levels of landmark occlusion, analysis was initiated with both sets of views, but the pair with the lowest digital reconstruction error was then used for kinematic reconstruction.

A total of 114 trials (including five trials from each of 22 individuals, and four trials from one individual) met these criteria and were used to reconstruct body motions. Using custom MATLAB code, marker positions in three dimensions were reconstructed using calibration information for the corresponding camera pair. A quintic spline algorithm was used to smooth positional data, to interpolate positional data for frames in which landmarks were occluded, and to estimate instantaneous velocity and acceleration from the first- and second derivatives of the positional data function, respectively [64]. All splines were generated using the splinetool function in MATLAB and uniquely fit for each trial. We first fit a spline with a tolerance value of 0.01 to the x-, y-, or z- component of the positional data, then we increased or decreased the tolerance value until the smoothing spline represented the positional data without large accelerations due to noise (assessed visually via the positional fit and second derivative). We repeated this for all components of all trials.

These smoothed positional data were then used to derive (based on custom MATLAB code) various body and kinematic parameters (Table 1) for all 114 trials. For velocity estimates, the x-, y-, and z-components of the midbody velocity (calculated from the first derivative of the spline function used to smooth positional data in each trial). Despite likely variation in center of mass location among species, sexes, and sizes, this landmark was assumed to be nearest to the center of body mass given previously published calculations for A. carolinesis and A. sagrei body segments (see [31]). Vertical velocity was taken as the sum of the z-component of the body’s velocity vector and the wind tunnel speed, whereas horizontal velocity was estimated as the square root of the sum of the squares of the x- and y-components; we used the minimum and maximum values of these measures, respectively, for analysis. We also used velocities to estimate minimum glide angle (i.e., an instantaneous measure of the angle at which the arctangent of the quotient of vertical velocity and horizontal velocity is smallest). We used lizard’s vertical velocity and horizontal position data to model its glide path as if it had been falling (Fig. 1) and calculated maximum glide ratio as the maximum distance travelled horizontally relative to the simulated distance travelled vertically at that timepoint for an indication of overall glide performance. Similarly, we used the x-, y-, and z-components of the midbody’s acceleration, as calculated from the second derivative of the spline function used to smooth positional data, to obtain values for vertical and horizontal accelerations.

Fig. 1.

Smoothed positional data with superimposed mean stream flow for all trials to illustrate trajectories derived from wind tunnel trials, as viewed from (a), oblique; (b), top; and (c), lateral perspectives. All trajectories are translated to start from the 3D origin, and are rotated about the z-axis to align the vector from the origin to the point with greatest horizontal displacement relative to the origin with the x-axis

Lizard morphometrics

Body mass for each lizard was measured before each glide trial; the average body mass among all trials was taken as the representative individual mass. Snout-vent length (SVL) was obtained using digital calipers. The ventral surface area of each was obtained from digital photographs of lizards dorsoventrally on a transparent acrylic sheet alongside a ruler for scale, with the camera being located below the acrylic. ImageJ software (v. 1.53a; [56]) was then used to create a digital polygon around the lizard’s perimeter, from which projected ventral surface area was obtained (Supplemental Figure 3). Wing loading was calculated as the ratio of body mass to this surface area.

As juveniles, individuals of A. equestris individuals grew substantially during the six weeks of experimental trials (i.e., body mass increases of 36–113%). Ventral surface area for this species was however measured only at the conclusion of all trials, and estimated wing loading may correspondingly have been underestimated for this species. Differences in wing loading among species were thus compared using values for A. equestris derived only from the last recorded mass for each individual (i.e., the mass measurement directly prior to taking of its digital photograph for surface area measurement). Projected surface area was unavailable for A. cristatellus #2, so we estimated it using a linear regression (R² = 0.90) relating the logarithm of mass and the logarithm of wing loading from the other three individuals of the species, followed by interpolation of a value for wing loading based on this individual’s trial-averaged body mass (Supplemental Fig. 4). For all lizards, the extended forelimb length (from base of the humeral joint with the body to the end of the middle toe) and the extended hindlimb length (from base of the femur to the end of the longest toe) were measured using digital calipers.

Statistical analyses

All statistical analyses were performed in RStudio (R v. 4.4.1; [48, 50]). Difference among species were first analyzed using one-way Analysis of Variance (ANOVA) tests for each of the measured morphology and kinematic variables (i.e., mass, surface area, wing loading, maximum horizontal velocity, minimum vertical velocity, minimum glide angle, maximum glide ratio, mean horizontal acceleration, and mean vertical acceleration). As species that share common ancestry are not statistically independent [15], we ideally would have accounted for phylogenetic relationships in these tests. However, these tests were not phylogenetically corrected because the number of groups compared (i.e., species) would be equivalent to the number of observations (one per species), and would therefore be an inappropriate test. In addition, tests for Tukey honest significant difference (Tukey HSD) were used to identify significant pairwise differences between species.

When examining relationships among morphology and kinematics, we incorporated phylogenetic relationships among study taxa using a comprehensive Anolis phylogeny [47], as pruned to include only the five focal taxa of this study (Supplemental Figure 1). Univariate phylogenetic generalized least squares (PGLS, lambda = 1, delta = 1, and kappa = 1 for all) regressions [16] were used to correlate each of the species-averaged morphological variables (i.e., mass, surface area, and wing loading) with species-averaged kinematic variables, using the “pgls” function in the caper package [46, 52].

Among species, forelimb and hindlimb lengths were highly correlated with SVL (linear regression between the logarithm of forelimb length and the logarithm of SVL: p < 0.001, R² = 0.89; linear regression between the logarithm of hindlimb length and the logarithm of SVL: p < 0.001, R² = 0.84). To identify relationships between relative limb length and gliding performance, we corrected limb lengths for body mass by using phylogenetically-controlled residuals from a least-squares regression between mass and the forelimb length, and between mass and hindlimb length, using the “phyl.resid” function [51]. These residuals, i.e., relative limb lengths for both the fore- and hindlimbs among species, were then correlated using PGLS tests with variables for gliding performance.

Caribbean members of the genus Anolis exhibit well-documented relationships among morphological parameters, locomotor capacity, and microhabitat in accordance with correspondingly described ecomorphs [32, 49]. Although statistical non-independence cannot be assumed given shared evolutionary history among species, we used one-way ANOVA without phylogenetic control to descriptively compare morphology and kinematic variables among ecomorphs, as data were available for only a single representative for the crown ecomorph and were therefore inappropriate for a phylogenetic ANOVA. We used Tukey HSD tests to identify pairwise differences.

Results

Overall, the five Anolis species varied significantly in body mass (ANOVA, p < 0.001; DF = 4; F-statistic = 8.380) and in ventral surface area (ANOVA, p < 0.001; DF = 4; F-statistic = 28.100; Fig. 2; see Table 1 for all ANOVA results). However, wing loading did not differ significantly among species (ANOVA, p = 0.34; DF = 4; F-statistic = 1.200) in spite of substantial variation (e.g., mean wing loading for A. coelestinus was ~ 50% greater than that of A carolinensis; see Table 1). All tested species exhibited controlled gliding when released into the vertical airstream (Fig. 1). Several aerial performance variables differed significantly among species, namely minimum vertical velocity (ANOVA, p < 0.001; DF = 4; F-statistic = 8.300), minimum glide angle (ANOVA, p = 0.002; DF = 4; F-statistic = 6.790), and maximum glide ratio (ANOVA, p < 0.001; DF = 4; F-statistic = 8.360; see Fig. 2; Table 1). Anolis carolinensis descended at a lower velocity (-9.50 ± 0.38 m/s) than any other species except A. sagrei (-10.45 ± 0.97 m/s), which only descended slower than A. equestris (-12.12 ± 0.51 m/s). Anolis coelestinus (-11.70 ± 0.83 m/s), A. cristatellus (-11.18 ± 1.21 m/s), and A. equestris descended at similar vertical velocities. Anolis carolinensis had the highest maximum horizontal velocity (0.82 ± 0.11 m/s), which was ~ 17% faster than the next (i.e., A. equestris at 0.70 ± 0.13 m/s), but there was no significant difference among species (ANOVA, p = 0.119; DF = 4; F-statistic = 2.130). Minimum glide angle and maximum glide ratio showed approximately the same relationships, whereby A. carolinensis glided at a significantly lower angle and higher glide ratio than all other species (Fig. 1). The minimum glide angle (i.e., the maximum gliding performance) for A. carolinensis was ~ 85º, whereas the other species were achieved steeper angles of ~ 86.5º. Finally, there were no significant differences among species in either of the acceleration measurements (ANOVA, p > 0.05, DF = 4, and F-statistic < 1.760 in all cases; Table 1).

Fig. 2.

Variation among Anolis species in (a), body mass; (b), ventral surface area; (c), wing loading; (d), minimum glide angle; (e), maximum glide ratio; and (f) minimum vertical velocity. Column values represent species means (± standard deviation). Letters indicate statistically different groups such that columns that do not share a letter are significantly different according to the Tukey posthoc analyses (p < 0.05)

Among all of the 18 morphology-performance model configurations tested, only the PGLS models between body mass and minimum vertical velocity and between wing loading and maximum glide ratio (Fig. 3; Table 2) yielded significant correlations (DF = 3, p > 0.077, and F-statistic < 7.004 for all; see Table 2). Body mass was negatively correlated with vertical velocity (PGLS, p = 0.015, R² = 0.893; DF = 3; F-statistic = 25.060; see Fig. 3). Glide ratio was negatively correlated with wing loading (PGLS, p = 0.033; DF = 3; F-statistic = 14.109; see Fig. 3), such that animals with less mass relative to ventral surface area tended to perform shallower glides. There were no significant relationships between any kinematic variable and either relative forelimb length or relative hindlimb length (p > 0.05 in all cases; see Supplemental Table 2).

Fig. 3.

Negative interspecific correlations between (a), body mass and minimum vertical velocity (PGLS, p = 0.015) and (b), wing loading and maximum glide ratio (PGLS, 0.033). Data points represent the mean value per species

Table 2.

Results from PGLS analyses

Values that indicate significant differences are in boldface

Ecomorphs of Anolis differed only in mass (ANOVA, p = 0.008; DF = 2; F-statistic = 6.160) and surface area (ANOVA, p < 0.001; DF = 2; F-statistic = 19.5; Fig. 4; Table 3). Wing loading and all the kinematic variables were not statistically different among ecomorphs.

Fig. 4.

Variation among ecomorphs in (a), body mass and (b), surface area (one-way ANOVA, p < 0.05). Column values represent ecomorph means (± standard deviation). Columns that do not share a letter indicate significantly different groups according to the Tukey posthoc analyses

Table 3.

Results from the one-way ANOVAs comparing ecomorphs

Values that indicate significant differences are in boldface

Discussion

All Anolis individuals dropped into a vertical jet of air assumed a stereotypical skydiving posture, such that the fore- and hind limbs were positioned laterally while the body was maintained dorsoventrally upright (see [65]). This posture maximizes projected surface area while falling and enhances aerodynamic force production on the body [10, 45]. Additionally, all individuals of the five study species exhibited non-vertical glides with horizontal displacement, albeit at very steep angles best considered to be directed aerial descent [10, 68]; we too describe this behavior as “gliding” because its inclusive and conventionally used definition (see Introduction) captures this outcome within a continuum of controlled aerial behaviors (see [10, 59, 68]). Gliding performance was largely comparable among species, with substantial overlap in average values for vertical velocity and acceleration, horizontal velocity and acceleration, glide angle, and glide ratio (Table 1).

The glide ratios and angles measured here, although very steep, were meaningfully different from vertical (Table 1; Figs. 1 and 2). The minimum glide angle and maximum glide ratio per species (as averaged across individuals for all trials) reflects a time-averaged range of both vertical and horizontal accelerations (Table 1) that encompasses much behavioral variation, as lizards were dropped directly into a vertical airstream such that aerodynamic forces on the body would not instantaneously equilibrate to yield steady-state gliding. In nature, directed aerial descent begins with an initial jump from an elevated substrate, and is associated with an initial increase in horizontal velocity while simultaneously accelerating downward (see [10]). Some anoles reach initial horizontal velocities of at least 1 m/s after a jump [23, 61]; this value is substantially greater than even the maximum horizontal velocities recorded in this study (Table 1), but acceleration from an initial horizontal velocity of zero to a value of ~ 0.6 m/s measured indicates lateral force generation and impressive aerial control during a transient and atypical flight behavior. Generating horizontal motion with aerodynamic force may enhance control of trajectory control in non-vertical descent (even at steep glide angles), enabling better access to favorable landing sites [6] and avoidance of ground-based hazards [69, 70].

Descent velocities varied considerably among species (Table 1; Fig. 3). In steep glides, even small differences in vertical velocity meaningfully alter an animal’s kinetic energy and injury risk upon ground impact [63]. Therefore, even small variation in descent speeds can have survivorship consequences [12, 62]. The 2.5 m/s difference in vertical velocity (i.e., ~ 20%) observed between the fastest (A. equestris) and the slowest (A. carolinensis) species is sufficient to substantially change kinetic energy upon ground impact, although dedicated landing maneuvers are also relevant for diverse animal gliders [7, 28, 57].

The anole species studied here were chosen as representative examples of the well-characterized anole ecomorphs of the Caribbean [30, 32–36, 67], with substantial differences in both body mass and projected surface area. Body masses for these species were similar to those reported elsewhere (e.g., Elstrott and Irschick, [11, 36], such that individuals in the crown ecomorph are the most massive; the trunk-ground ecomorph is however comparable in body mass to the trunk-crown ecomorph (see [32]). The three-fold variation in body mass among study species was not, however, matched by systematic differences in wing loading. Instead, similar patterns of variation were present in ventral surface area such that wing loading did not statistically differ either among species or among ecomorphs. Anolis carolinensis, unsurprisingly, tended to have the lowest wing loading, but it was unexpected that A. coelestinus would have the highest value, which may have contributed towards the disparity in performance observed between these species. Interestingly, wing loading values for A. equestris (as measured at the end of the study) were very similar to those of A. cristatellus and A. sagrei. Given that these values were likely an underestimate for many of the individual A. equestris used in aerial trials, its lower assumed wing loading relative to the other species may explain its relatively similar gliding performance despite its greater mass.

Subtle differences in behavior and morphology can result in meaningful differences in locomotor performance. Previous studies of gliding animals have suggested that behavior changes may precede morphological adaptation [45, 53] and results are consistent with this prediction. None of the species in our study possess wings, webbing, or patagia that are typically associated with dedicated gliding [53, 62], yet all species demonstrated aerial control via the stereotypical skydiving posture that reduces descent speed and that facilitates stable gliding [45, 65]. In addition, reduced wing loading (via decreased mass or increased surface area) can enhance aerial performance in lieu of specialized structures as equilibrium glide velocity tends to be directly proportional to wing loading [62]. Wing loading was statistically invariant among the anole species studied here (Fig. 2), but a negative relationship was found among species between maximum glide ratio and wing loading (Fig. 3). This trend is consistent with observations that both total glide angle (including the initial descent and non-equilibrium stages of the glide) as well as net height lost can be strongly influenced by wing loading (see [39, 62]). The combination of dedicated behavioral responses and subtle morphological variation thus generates differences in aerial performance that can be of broad benefit to arboreal taxa [10].

Anoles residing higher in a forest canopy may experience selection for better aerial control and descent [12]. A variety of arboreal taxa tend to exhibit slenderized bodies [41], which have the effect of reduced wing loading and lower descent speeds [62] with a presumably increased likelihood of survival after a fall (see [10]). Here, vertical velocity among species was negatively correlated with body mass (Fig. 3), and the much higher body mass of the crown ecomorph (A. equestris) was associated with a higher descent speed (Table 1). Species from the trunk-crown ecomorph (i.e., A. carolinensis and A. coelestinus) were significantly dissimilar from one another in aerial performance, whereas trunk-ground ecomorphs showed no such differences among taxa. Species in the trunk-ground ecomorph tend to have stocky limbs and to remain relatively close to the ground [32, 34]. Whereas individuals in this ecomorph typically leap from tree trunks to the ground [22, 34], the behavioral cost of falling is low (see [12]), and their small size renders them unlikely to be injured on impact [63]. Their performance in aerial descent is therefore likely to be under weaker selective pressure relative to that of taxa that occupy niches higher in the trees [4]. Ontogenetic differences in behavior have also been observed in trunk-ground anoles, whereby juveniles tend to evade predators by leaping from vegetation more often than adults [22], thereby making directed aerial descent more ecologically relevant (see [10]). Nevertheless, the presence of a parachuting posture [45] as well as of controlled aerial descent in this ecomorph supports the hypothesis that controlled gliding may be widespread among arboreal lizards (see [40]), and in other vertebrate and invertebrate taxa that reside within even low-lying vegetation (see [5, 10]).

Whereas all study species assumed a stereotypical skydiving posture in the air [45, 65], we qualitatively observed some variation in this behavior among ecomorphs. Lizards in the trunk-crown ecomorph had similar body kinematics to each other. Both species frequently performed S-shaped lateral body oscillations (at frequencies similar to those previously reported; [65]) as they translated cranially after retracting their shoulders (Supplemental Movies 1 and 2). We expect that these oscillations must relate to thrust production during the glide, but associated kinematics remain to be quantified in detail. Yet only A. carolinensis exhibited relatively good gliding performance in the vertical wind tunnel (Fig. 2). The greater mass and (non-significantly) greater wing loading of A. coelestinus may thus impair its gliding performance relative to that of A. carolinensis, at least during the short glide durations simulated here. Additionally, both trunk-crown species perform large and rapid displacements of their long tails during glides that appear to facilitate yaw-based rotations, perhaps via inertial mechanisms that can have substantial consequences for maneuvering (see [26, 27]). By contrast, body movements for lizards in the trunk-ground (Supplemental Movies 3 and 4) and crown (Supplemental Movie 5) ecomorphs were qualitatively similar. These lizards maintained the limbs in the lateral positions, rarely performed S-shaped oscillations, and exhibited tail deviation from the midline to only a relatively small extent, albeit at frequencies comparable to those of the trunk-crown ecomorph. However, lizards in both ecomorphs were observed raising the tail above the plane of the body and engaging in either clockwise or anti-clockwise motions that likely served in yaw control [26, 27]. The posture and orientation of these species suggest stability, but the muted body dynamics perhaps indicate less active aerial control.

Although no statistically significant differences in body kinematics were detected among anole ecomorphs, there was a marginally nonsignificant trend (ANOVA, p = 0.059; DF = 2; F-statistic = 3.280) whereby the crown ecomorph descended faster (by ~ 2 m/s) relative to the other ecomorphs. Intriguingly, crown giants, along with trunk-crown anoles, occupy niches near the treetops [33, 67] where the costs of falling are likely higher, and especially for heavier animals [12, 63]. With their greater masses and even slightly faster descent speeds, crown giants would thus impact the ground with much greater kinetic energy. However, crown giants do spend more time near the trunk of trees [25, 34, 66] as compared to trunk-crown species, which often perch on branches and twigs distant from the trunk [25, 34, 41, 49]. To reach favorable landing sites such as those simply further down on a tree trunk, crown giants would thus need to translate less distance horizontally while falling (see [6, 10, 68] Most likely, an evolutionary tradeoff characterizes crown giant anoles, whereby the advantages of being a large predator in a forest canopy [34] simply outweigh the costs of reduced gliding performance.

Future studies quantifying relationships between limb and body kinematics relative to aerial maneuvers will be valuable in understanding those mechanisms by which such wingless animals can control glides. We also analyzed glides starting from a variety of different initial conditions, with the lizards losing contact with a wall and then falling sideways into the airflow, but subsequently capturing a glide posture. Further studies documenting takeoff behaviors for jumping arboreal taxa will provide knowledge about such behavioral diversity used to initiate aerial locomotion. Finally, increased taxonomic coverage would enhance comparisons among ecomorphs. Including additional crown giant species, and especially adults, would likely provide further evidence for their aerial abilities relative to other ecomorphs. It would be particularly interesting to include species from the “twig” ecomorph that have been documented to engage in gliding (e.g., Anolis pentaprion; [44]).

Consistent with findings for other aerially capable lizards [39, 62], gliding performance in the Anolis species studied here was inversely correlated with wing loading, suggesting that subtle morphological differences (i.e., relatively more surface area in less massive species) may precede the origins of specialized structures (e.g., patagia and skin flaps) that would enhance aerodynamic force production (e.g., [53]). Our data suggest that gliding behavior may vary among Anolis ecomorphs (possibly reflecting their variable habitat use within and below the canopy), although this pattern was not statistically significant given low sample sizes, and merits further investigation. Importantly, all anole species studied here exhibited a sky-diving posture during free fall, along with aerial control during steep (albeit non-equilibrium) glides. Such aerial behaviors may be thus widespread in this genus, and indeed among all arboreal lizards.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

ANOVA

Analysis of Variance

DF

Degrees of Freedom

OLAC

Office of Laboratory Animal Care

PGLS

Phylogenetic Generalized Least Squares

SVL

Snout–vent length

Tukey HSD

Tukey honest significant difference

Author contributions

E.A.S. and R.D. developed the concept and methodology of the manuscript. E.A.S. performed the methods described for the investigation. E.A.S. and R.D. both contributed to visualization. E.A.S. and R.D. acquired funding for the research. E.A.S. and R.D. both administered the project. R.D. provided supervision during this study. E.A.S. wrote the original draft of the manuscript and prepared all figures and tables. Both E.A.S. and R.D. reviewed and edited the manuscript.

Funding

This research was funded by the University of California, Berkeley, Class of 1933 Chair in the Biological Sciences, Berkeley Chapter of Sigma Xi’s Grant-in-aid-of-research, and The Herpetologists’ League’s EE Williams Research Grant.

Data availability

Data is provided within the manuscript or supplementary information files.

Declarations

Ethics approval and consent to participate

All animal acquisitions and methods using animals were approved by the University of California, Berkeley’s Animal Care and Use Committee (ACUC) under Animal Use Protocol # AUP-2017–03-9711. The methods used were in compliance with and followed guidelines of the United States Animal Welfare Act, the National Research Council Guide to the Care and Use of Laboratory Animals, and the Public Health Service Policy on the Humane Care and Use of Laboratory Animals. All animals were purchased from vendors approved by the University of California, Berkeley’s Office of Laboratory Animal Care in accordance with Animal Use Protocol # AUP-2017–03-9711 and institutional guidelines.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.