Transitions from animal to wind pollination have occurred repeatedly in flowering plants, driven by structural and biomechanical modifications to flowers. But the initial changes promoting wind pollination are poorly understood, especially those required to release pollen into airflows – the critical first stage of wind pollination.
Using a wind tunnel, we performed a comparative study of pollen release biomechanics in 36 species of animal- and wind-pollinated Thalictrum. We quantified pollination syndromes and stamen natural frequency (f_n), a key vibration parameter, to determine if floral traits reliably predicted pollen release probability. We then investigated if pollen release was caused by wind-induced resonance vibration of stamens.
We detected wind-induced stamen resonance in 91% of species and a strong effect of stamen acceleration on pollen release, inversely driven by f_n. However, unlike f_n, pollination syndromes did not reliably predict the probability of pollen release among species.
Our results directly link f_n to the capacity of stamens to release pollen by wind and suggest that structural mechanisms reducing f_n are likely to be important for initiating transitions from animal to wind pollination. Our inability to predict the probability of pollen release based on pollination syndromes suggests diverse phenotypic trajectories from animal to wind pollination.

Introduction

Wind pollination is characteristic of ancestral seed plants, but derived from animal pollination in angiosperms (Culley et al., 2002; Friedman & Barrett, 2009). Many of the evolutionary changes leading to the angiosperm flower involved increasing the probability of animal pollination while simultaneously limiting pollen dispersal by wind (Whitehead, 1969). The classical reason given for the predominance of animal pollination in angiosperms is that it promotes directional conspecific pollen transfer, often considered less wasteful of pollen (Regal, 1977). Wind pollination may be more random and imprecise, but has nevertheless evolved independently at least 65 times from animal-pollinated ancestors in many diverse lineages (Linder, 1998; Ackerman, 2000). This repeated evolution implies that wind pollination is a more effective means of pollination under some environmental conditions. A surprising feature of contemporary pollination biology is that the ecological drivers and phenotypic traits initiating the transition process from animal to wind pollination are not well understood despite the prevalence of wind pollination in many ecosystems.

The evolution of wind pollination (anemophily) is characterized by a series of structural changes to flowers (Linder, 1998; Weller et al., 1998; Friedman & Barrett, 2008; Welsford et al., 2016). Animal-pollinated flowers feature visual and olfactory attractants, rewards and specialized structures for interacting with pollinators (Faegri & van der Pijl, 1966; Fenster et al., 2004). By contrast, wind-pollinated flowers are inconspicuous, having lost such contrivances, and possess aerodynamic features for releasing, dispersing and capturing pollen in airflows (e.g. feathery stigmas, flexible stamens, small or no corollas, lighter pollen grains; Ackerman, 2000; Friedman & Barrett, 2009). Biomechanical models have linked some anemophilous traits to wind pollination mechanisms (Niklas, 1985, 1992), but few floral traits have been experimentally investigated to characterize the specific aerodynamic mechanisms governing pollen release and capture (but see Niklas, 1987; Sabban et al., 2012; Cresswell et al., 2010; Timerman et al., 2014; McCombe & Ackerman, 2018). An unresolved question is which traits in animal-pollinated ancestors are initially involved in the sequential process of floral modification that ultimately results in the distinctive wind pollination syndrome.

A key step in evolving wind pollination is acquiring a mechanism of effective pollen release into airflows (Cox, 1991). Pollen release is significant because it is the critical first step in successful pollen dispersal by wind, but is probably costly for animal-pollinated species owing to male gamete wastage (Timerman & Barrett, 2018). For both reasons, animal- and wind-pollinated species are likely to differ with respect to their capacities for releasing pollen. In general, pollen release by wind requires an external force to mobilize stationary pollen grains attached to anthers (Niklas, 1985). But models predict that aerodynamic forces on pollen grains are too weak to overcome forces of adhesion and gravity due to boundary layer effects – a layer of slow-moving air in proximity to solid surfaces such as anthers (Vogel, 1994; Jackson & Lyford, 1999; Urzay et al., 2009). Wind-pollinated plants may overcome the limitation of fluid-structure interactions through biomechanical optimization of stamens for vibrational pollen release (Timerman et al., 2014; Timerman & Barrett, 2018). This mechanism involves turbulence-induced stamen resonance causing rapid accelerations that impel pollen from anthers into airflows.

Stamen resonance occurs when swirling masses of air (eddies) impact stamens periodically at the stamen natural frequency, f_n, resulting in large amplitude oscillations owing to storage of vibrational energy (Timerman et al., 2014). f_n is the rate of exchange between potential and kinetic energy in a vibrating system, and is a composite biomechanical trait of stamens which can be approximated from anther mass and stamen length using a cantilever beam as a model for stamens (Spatz & Zebrowski, 2001; Brüchert et al., 2003; de Langre, 2019). For stamen resonance to occur, f_n must be sufficiently low because the kinetic energy of eddies declines with eddy turnover frequency (de Langre, 2008). Resonance additionally requires that dissipation of energy from the vibrating system due to friction be less than the energy stored (Denny et al., 1998). This decay in vibration can be quantified by the damping ratio, ξ, which for a resonating system is 0 < ξ < 1. Lower values within this range indicate a stronger response to excitation at f_n. We therefore expect that pollen release will depend on stamen acceleration mediated through f_n and ξ, and for wind-pollinated species to have higher probabilities of pollen release, and thus lower values than animal-pollinated species for both biomechanical parameters.

Here, we use Thalictrum (Ranunculaceae) as a study system for investigating vibrational pollen release and its relevance for the evolution of wind pollination. This genus comprises c. 190 herbaceous species distributed throughout the Northern Hemisphere, Africa and South America, with extensive variation in floral traits (Fig. 1) and pollination modes, including species pollinated by wind, insects or both pollen vectors, a condition referred to as ambophily (Kaplan & Mulcahy, 1971; Melampy & Hayworth, 1980; Steven & Waller, 2004; Timerman & Barrett, 2018). Several sexual systems occur in the genus including hermaphroditism, monoecy and dioecy (Kaplan & Mulcahy, 1971; Soza et al., 2012), and also several instances of cryptic dioecy in which females produce functioning stamens with non-germinating pollen grains (Davis, 1997; Humphrey, 2018). Phylogenetic analysis of pollination syndromes indicates that insect pollination is the ancestral state in the genus and wind pollination has evolved independently several times (Soza et al., 2012, 2013; Wang et al., 2019), possibly via ambophily (Timerman & Barrett, 2018). All Thalictrum species produce nectarless, apetalous flowers, but species vary extensively in inflorescence architecture (Keener, 1976), floral fragrances (Wang et al., 2019) and pollen morphology (Humphrey, 2016). Some species have large pigmented (yellow, purple) sepals that presumably function in pollinator attraction, but most have smaller, inconspicuous, ephemeral white or green sepals. Numerous stamens are often the most conspicuous feature of flowers, but also vary greatly among species in number, size, shape and pigmentation.

Details are in the caption following the image — **Figure 1**
Open in figure viewer PowerPoint

Examples of floral diversity in putatively animal- (AP) and wind-pollinated (WP) *Thalictrum* species. (a) *T. thalictroides* AP; (b) *T. pubescens* AP, WP; (c) *T. dioicum* WP; (d) *T. squamiferum* WP; (e) *T. pubescens* AP, WP; (f) *T. aquilegifolium* AP; (g) *T. petaloideum* AP; (h) *T. rochebrunianum* AP. Images courtesy of (d) Wei Zhou; (e) Tobias Mankis.

Here, we evaluate the hypothesis that selection for stamen resonance is a key factor initiating transitions from animal to wind pollination. To accomplish this, we performed a comparative study of floral trait variation and pollen release biomechanics in 36 species of Thalictrum. We measured the biomechanical properties of stamens, f_n and ξ, and investigated their correlates and effects on vibration-mediated pollen release. Because we investigated animal- and wind-pollinated species, we also investigated whether pollination syndromes predicted the probability of releasing pollen in airflows. Our study addressed three specific questions. (1) Do the pollination syndromes of species predict pollen release? We addressed this question by measuring selected floral traits for putatively animal- and wind-pollinated species, and determined pollen release probability in a wind tunnel assay. (2) Does wind cause stamens to resonate? We quantified wind-induced vibration of stamens using high speed videos recorded during the wind tunnel assay, and then used spectral analysis to determine if they resonated and at what frequency. (3) What is the role of stamen vibrations in pollen release? Controlling for phylogenetic relationships, we first determined the effects of f_n and wind speed on stamen acceleration, and then quantified the effects of stamen acceleration on pollen release in airflows.

Materials and Methods

Taxa and floral traits

We investigated interspecific variation in floral traits and pollen release in a diverse sample of Thalictrum species. Taxon sampling reflected our ability to procure either seeds or mature plants of species in the phylogeny of Soza et al. (2013). We propagated plants from seeds obtained through institutions or commercial suppliers in North America, Europe and China, purchased mature plants from nurseries and collected specimens from the field in Ontario, Canada (Table 1: species list and sources; Supporting Information Methods S1: germination and cultivation of plants).

Table 1. Thalictrum species included in our study with putative pollination system, sexual system and pollination index (the source of plants used in our study is also provided)

Vectora	Species	Sexual system	Natural frequency (Hz)	Pollen release probability	Pollination index	Source
Animal	T. aquilegiifolium	H	12.1	9.41 × 10⁻⁵	2.71£	KIB
	T. baicalense	H	10.8	1.15 × 10⁻¹	2.43§	LH
	T. chelidonii	H	27.3	6.23 × 10⁻⁶	2.29§	KIB
	T. clavatum	H	19.2	8.32 × 10⁻⁵	2.57§	LH
	T. finetii	H	7.7		2.40£	KIB
	T. flavum	H	28.9	2.60 × 10⁻³	2.29§	KIB
	T. glaucum	H	22.1	3.25 × 10⁻³	2.29§	HU
	T. ichangense	H	17.0	1.26 × 10⁻²	2.67¥	LH
	T. isopyroides *	H	13.1		2.00§	PWS
	T. kiusianum	H	56.0	4.40 × 10⁻⁴	2.14¥	PWNR
	T. lecoyeri	H	8.9	1.79 × 10⁻³	2.43§	KIB
	T. lucidum	H	19.9	1.30 × 10⁻²	2.43§	USDA
	T. petaloideum	H	24.3	9.44 × 10⁻⁵	2.57§	KIB
	T. reniforme *	H	32.1		2.14§	N/A
	T. rochebruneanum	H	20.0	6.36 × 10⁻³	2.67§	LH
	T. rutifolium	H	8.5	8.75 × 10⁻³	2.50¥	KIB
	T. saniculiforme	H	12.6	1.77 × 10⁻⁴	2.57§	KIB
	T. smithii	H	10.2	8.13 × 10⁻³	2.17§	KIB
	T. squamiferum	H	14.6	1.67 × 10⁻³	2.17¥	KIB
	T. thalictroides	H	28.8	3.84 × 10⁻⁵	3.00§	LH
	T. uchiyamae	H	43.1	1.81 × 10⁻³	2.17¥	LH
Wind	T. alpinum	H	12.1	9.41 × 10⁻⁵	2.00§	CT
	T. cultratum	H	8.0	2.33 × 10⁻²	1.57§	KIB
	T. dioicum	D	12.6	2.33 × 10⁻²	1.13§	F
	T. fendleri	D	5.7	3.20 × 10⁻²	1.00§	USDA
	T. foetidum	H	9.3	2.73 × 10⁻²	1.86§	KIB
	T. foliolosum	H	13.0	3.45 × 10⁻³	2.00§	KIB
	T. leuconotum	H	10.0	6.37 × 10⁻²		KIB
	T. polycarpum	D	9.1	4.08 × 10⁻³	1.14§	HU
	T. pubescens	D	16.8	1.79 × 10⁻²	1.14§	F
	T. revolutum	D	9.2	2.18 × 10⁻²	1.26§	WAF
	T. simplex	H	10.2	1.61 × 10⁻³	1.71§	KIB
	T. speciossimum	H	16.7	5.69 × 10⁻⁴		CT
	T. sphaerostachyum *	H	20.9			PWS
	T. squarrosum	H	12.2	8.04 × 10⁻³	1.71§	KIB
	T. venulosum	D	9.0	3.70 × 10⁻²	1.00§	WAF

Pollination index after Kaplan & Mulcahy (1971; see text); sexual system, D: dioecy, H: hermaphroditism; ¹putative pollination vectors inferred using the criteria of Soza et al. (2012; see text); *species not included in the multivariate analysis and wind tunnel study; ^§Kaplan & Mulcahy (1971), ^¥Soza et al. (2012), ^£Wang et al. (2019); CT, Chiltern Seeds (Wallingford, UK); F, collected in the field by DT (Ontario, CA); HU, National Botanical Garden (Vácrátót, HU); KIB, Kunming Institute of Botany (Kunming, CN); LH, Lost Horizons Nursery (Acton, CA); PWS, Plant World Seeds (Newton Abbot, UK); USDA, United States Department of Agriculture (Fort Collins, CO, USA); WAF, Wild About Flowers (Okotoks, CA).

On flowering, we measured stamen and carpel number, pollen production (pollen : anther; see Methods S2), pollen : ovule ratio, pollen diameter, stamen length, anther mass before dehiscence, anther volume, sepal length, orientation of flowers (erect or pendulous) and colour of sepals, anthers and filaments (green, white, purple or yellow). In total, we characterized floral traits for 67 plants, representing 1.88 ± 1.49 (± SD) individuals/species. In each plant, we measured or scored traits in representative samples of three flowers per plant. For metric traits, we sampled three stamens, three sepals and 10 pollen grains from each flower, and measured their properties from digital images.

Biomechanical properties of stamens

We measured the stamen natural frequency, f_n, and damping ratio, ξ, of species using experimental modal analysis (see Timerman et al., 2014). We used a vertically oriented K2004E01 electrodynamic shaker (The Modal Shop, Cincinnati, OH, USA) to vibrate the fixed ends of stamens through a range of sinusoidal frequencies (5–60 Hz). We evaluated three stamens per plant from different flowers in separate trials. Open flowers with fully extended stamens were randomly chosen before dehiscence, and mounted horizontally at their base to the shaker. We then excised all stamens except for one, which served as a test subject. We restricted our analysis to stamens commencing dehiscence to limit confounding effects of pollen and moisture content of anthers. In each trial, we recorded the vibrational response of the anther for two full cycles of excitation at 120 frames per second (fps) with a video filmed orthogonally to the direction of stamen motion (Elixim FH25; Casio Canada, Markham, ON, Canada). Afterwards, we preserved stamens in 60% ethanol for measurement of stamen length. We determined f_n by quantifying anther displacement in Tracker video analysis software (Open Source Physics, http://www.compadre.org/osp/) and using Fast Fourier Transform (FFT) to identify peaks in their frequency spectra. We calculated ξ using the half-power bandwidth equation: $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0001$ where f₊ and f₋ correspond to points on either side of f_n and in which peak amplitude decreased by a factor $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0002$ .

We also investigated the relation between f_n and ξ and the morphological correlates of f_n. Cantilever beam theory predicts that f_n for a slender structure anchored at one end (i.e. filament) with concentrated mass at the free end (i.e. anther) is directly proportional to both $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0003$ and $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0004$ , where H and M represent stamen length and anther mass, respectively (Niklas, 1992). Our analyses used linear regressions corrected for phylogenetic non-independence of species using independent contrasts (PIC; Felsenstein, 1985). Details of the phylogeny of Thalictrum used are provided below.

Experimental assay of pollen release

Wind tunnel experiment

We characterized stamen vibrations at different wind speeds and determined their effect on pollen release using a benchtop open-circuit wind tunnel. The test section of the wind tunnel had a length of 1.20 m and a rectangular cross-section measuring 0.25 × 0.18 m, which is equivalent to a hydraulic diameter of L = 0.21 m. Wind speed was measured at a rate of 5 Hz using an 8455 air velocity transducer (TSI, Minnesota, MN, USA) centred 0.55 m downstream in the test section. Reynolds number, Re = uL/v, of the test section ranged between c. 7000–70 000 for wind speeds μ = 1.0–5.0 m s⁻¹ and kinematic viscosity (21°C) $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0005$ = 15.2 × 10⁻⁶m² s⁻¹, indicating turbulent flow. In the wind tunnel, we exposed one flower at a time to a sequence of three incremental wind speeds (u_low = 0.66 ± 0.24 m s⁻¹, u_mid = 1.26 ± 0.20 m s⁻¹ and u_high = 2.11 ± 0.14 m s⁻¹; ± SD) for 5 min each. Transitions between wind speeds occurred in gradual steps over 30 s to limit flow disturbances. Higher wind speeds were not used because they caused flowers to move out of frame in the videos.

Before trials, we randomly sampled a single dehiscent flower from a focal plant (three flowers per plant) and placed it by the pedicel into a 0.1 ml water-filled PCR tube to prevent wilting. We sealed the tube using putty to prevent movement and spillage, and mounted it upside-down to a 12.0-cm-long, thin rod centred 0.6 m downstream in the test section. During trials, we measured pollen release at 1 Hz from flowers using a Solair Boulder airborne particle counter (Lighthouse Worldwide Solutions, Fremont, CA, USA), with its sampling tube positioned 0.05 m downstream of the flower. This device samples airborne particles using a constant rate of suction and counts particles 5–100 μm in diameter, an appropriate range for detecting pollen of Thalictrum (Humphrey, 2016). We also recorded stamen vibrations in 120 fps video using an a6300 digital camera with SEL90M28G 90 mm macro lens (Sony of Canada, Toronto, ON, Canada). The camera was positioned orthogonal to the direction of flow and centred on the flower at a fixed distance from the wind tunnel. After trials, we preserved flowers in 60% ethanol micro-tubes, and later determined the amount of pollen remaining (i.e. not released) in all dehisced anthers using a procedure similar to that described above. We tested a total of 137 flowers representing 33 species. Species had a mean sample size of 4.03 ± 2.34 flowers (mode = 4).

Vibrational response of stamens

To characterize the vibrational response of stamens to different airflows, we determined the acceleration of stamens in a random sample of pollen release events. For each level of wind speed, we randomly sampled five occasions (timestamps) when the particle counter detected > 0 pollen grains. We then quantified the motion of five stamens in the video recordings at times corresponding to the pollen release events. Each segment of video was 1 s in duration (120 frames), centred on the time of release. We measured stamen acceleration using Tracker (see ‘Biomechanical properties of stamens’ in the Materials and Methods section), and calculated the average root mean square (rms) acceleration for each video segment to quantify the magnitude of the vibration (see Denny & Gaines, 2000). We also investigated periodicity in the acceleration time series for each stamen by estimating the spectral density acceleration using the multitaper package in R software (function: ‘spec.mtm’, parameters: k = 15, nw = 30; Rahim et al., 2014).

Pollen release probability

We developed the metric pollen release probability (P_r) to quantify the propensity of species to release pollen in airflows. P_r is independent of interspecific differences in pollen production, which can influence the overall magnitude of pollen release for a species. Pollen release (by any mechanism) is a stochastic process dependent on random airflows, and therefore, P_r was also developed to account for variation in the frequency and magnitude of pollen release occurring within and among species.

We define the quantity P_r, as the conditional probability for pollen release of individual pollen grains given N ≥ 1 pollen release events. In our analysis, we considered these events to be instances where the particle counter recorded X_i ≥ 1 pollen grains (subscript i denotes the ordered sample number). To find P_r, we first determined the probability, P_e, of pollen release events. Because the particle counter sampled at a rate of 1 Hz, there were n = 320 measurements of pollen count for each 5-min interval. For simplicity, we define P_e as the proportion of those measurements for which pollen was detected such that: P_e = N/n. We then determined the probability, P_g, for a random pollen grain to be released during any of the N pollen release events. Because a pollen grain can only be released once, our model is based on sampling without replacement:

$urn:x-wiley:0028646X:media:nph15978:nph15978-math-0006$

Here, X_i and X_j represent the amount of pollen released in the i^th and j^th event, respectively, and Q represents the initial amount of pollen available. The first term in the equation gives the release probability for the first event, whereas the second, in series notation, gives the total probability for the remaining events, accounting for pollen released before each event. Finally, we calculated P_r by multiplying P_e and P_g. We calculated P_r for all flowers in the experiment, and then for species by averaging across flowers.

Statistical analysis

Our analyses address several specific questions enumerated below. All statistical analyses were performed using R software v.3.5.2 (R Core Team, 2018). To account for the evolutionary relationships of species, we used phylogenetically informed statistical methods. The phylogeny used in our analyses is pruned from the 80 species tree of Wang et al. (2019; http://purl.org/phylo/treebase/phylows/study/TB2:S32211). This Bayesian 50% majority rule consensus tree was constructed using data from several cpDNA regions and nuclear external and internal transcribed spacer regions. We obtained an ultrametric tree using penalized maximum likelihood with three calibration points (see Soza et al., 2013) implemented in the ape package (function: ‘chronos’; Paradis & Schliep, 2018).

Do pollination syndromes predict pollen release?

We investigated whether pollination syndromes and f_n reliably predicted pollen release probabilities. We considered the effects of pollination syndromes and F_n on P_r using phylogenetic least squares analysis. We assumed Brownian motion evolution and performed the analysis using the gls function in the package nlme (Pinheiro et al., 2018). Since P_r is bounded between 0 and 1, we used the logit transformation as our dependent variable. We used two metrics for quantifying pollination syndromes as independent variables in our models. The first metric used principal coordinates analysis (PCoA) in five axes to reduce the dimensionality of the floral trait data. We included all traits in the analysis except for anther mass, for which we were missing data from several species.

Our analysis used the function cmdscale and a Gower dissimilarity matrix calculated from the function daisy in the cluster package (Maechler et al., 2018). We then examined the contribution of each explanatory variable to the ordination using the envfit function in the vegan package (Oksanen et al., 2018). This function uses linear regression to fit explanatory variables to ordination scores. The estimated coefficients represent vector coordinates in ordination space, indicating the direction of maximum variation for a given trait. The length of each vector is proportional to the r² for that trait. Significance is then tested using a permutation test, which in our analysis involved 999 iterations. The second metric used published values for ‘pollination index’ (PI), an index developed for classifying pollination modes in Thalictrum (Kaplan & Mulcahy, 1971; Soza et al., 2012, 2013; Humphrey & Ossip-Drahos, 2018; Wang et al., 2019). PI is calculated by averaging scores assigned to floral traits based on their similarity to the average condition expected for wind- and animal-pollinated species (Kaplan & Mulcahy, 1971). Following Soza et al. (2012), species with PI < 2 were considered wind-pollinated, whereas species with PI > 2 were considered insect-pollinated. We used this classification scheme to test for morphological differences between wind and insect pollination using permutational multivariate analysis of variance of the dissimilarity matrix in the vegan package.

Does wind cause stamen resonance?

In the wind tunnel, we investigated stamen resonance and pollen release for a total of 150 flowers, representing 33 species. We tested for stamen resonance by identifying significant peaks in the spectral densities of stamens using an implementation of the Thomson F-test in multitaper. This test compares variance in power for a given frequency with the average background variance around that frequency. Significant peaks (F > F_2,2k-2) indicate that non-random periodic components contributed to the overall motion of stamens. This phenomenon was attributed to stamen resonance if peaks in spectral density occurred at frequencies corresponding to f_n. Because f_n, by definition, is the lowest frequency in a periodic waveform, we predicted a positive relation between the lowest significant frequency, f, and f_n. We tested this hypothesis for species by linear regression of f on f_n.

What is the role of stamen vibrations in pollen release?

We investigated the effects of stamen vibrations on pollen release using phylogenetic mixed effects models in the package MCMCglmm (Hadfield, 2010). For comparison, we repeated the analysis without correcting for phylogenetic relations using the package lme4 (Bates et al., 2015). We first modelled the number of pollen grains released as a function of rms stamen acceleration (fixed effect). We used a Poisson error distribution, and included random effects for phylogenetic relationships (in MCMCglmm), and intercepts for flowers nested in species and trials nested in flowers. The grouping factor trials accounted for correlations within flowers crossed with wind speeds. In lme4, we included an individual flower random slope for the effect of rms stamen acceleration, and used a likelihood ratio test to compare its effect to an intercept-only model. An individual-level factor was also included in the lme model to account for overdispersion.

Our second model quantified the relation between rms stamen acceleration and f (i.e. f_n measured in the wind tunnel using spectral analysis). The model was structured similarly to the first, but with wind speed (continuous) as an added fixed effect and a Gaussian error distribution with log link function in lme4. We included the interaction effect of f_n and wind speed in a separate lme4 model and tested its effect using a likelihood ratio test. Our Bayesian model used default priors for the fixed effects, and V = 1 and nu = 0.02 for residual variance and the variance components of each random effect. We ran 5 × 10⁶ MCMC iterations with a burn-in of 10 000 and a thinning interval of 500. We inspected trace and density plots of all estimated parameters and assessed mixing and convergence with Gelman-Rubin statistics (Gelman & Rubin, 1992).

Results

Pollination syndromes in Thalictrum

Multivariate analysis of floral traits revealed moderate clustering of species in the first two dimensions of the ordination (Fig. 2). Both axes explained a total of 38.1% of the variance and all but two traits were significantly correlated with the PCoA configuration (sepal length and pollen diameter; Table S1). In ordination space, species were grouped into two main clusters along the y-axis based on floral orientation. To the right of the y-axis were species with pendulous flowers, larger anthers, longer stamens and higher pollen–ovule ratios (Fig. 1c,d). Flowers were also drab with green or white sepals, yellow anthers, and green, white and occasionally purple filaments or sepals. To the left of the y-axis were species with erect flowers with many more stamens and pistils per flower, but with fewer pollen grains per anther and lower pollen–ovule ratios (Fig. 1a,b,e–h). Flowers had white or purple sepals and filaments, and yellow, white or purple anthers. Based on pollination index, species in the right-hand group were classified as predominantly wind-pollinated, with several exceptions, whereas all but one species in the left-hand group were insect-pollinated. Pollination index was not available for two species (triangles in Fig. 2), so they were classified according to their positions in the ordination. Despite the exceptions in this classification scheme, wind- and insect-pollinated species generally occupied significantly different regions of ordination space (F = 5.76, df = 1, P = 0.001; Table S2).

Biomechanical properties of stamens

Among species, stamens on the electrodynamic shaker had a mean f_n of 17.3 ± 10.6 (SD) Hz (n = 36) with a range of 5.7–56.0 Hz, and a mean ξ of 0.07 ± 0.04 (n = 26) with a range of 0.02–0.21. We also found a significant positive linear relation between the PICs of these variables (Fig. 3a; n = 23, F_1,21 = 4.38, P = 0.048). In accord with cantilever beam theory, significant positive linear relations occurred between PIC-scaled f_n, and both PIC-scaled H^-3/2 (Fig. 3b; n = 27, F_1,26 = 6.81, P = 0.015) and M^-1/2 (Fig. 3c; n = 17, F_1,16 = 25.54, P < 0.001).

Do pollination syndromes predict pollen release?

We found no significant relations between P_r and any of the five principal coordinates of floral trait variation; stepwise model selection indicated that the best fitting model included none of the variables (P > 0.05; Table S3). This was also the case for the relation between P_r and PI (P = 0.59; Table S3). These results indicate that our metrics for quantifying differences in pollination syndromes provided little information on pollen release biomechanics. By contrast, there was a significant relation between P_r and f_n (P < 0.019); as predicted, P_r was larger in species with smaller f_n (β = −0.17, 95% CI = −0.31 to −0.03; Table S3).

Does wind cause stamen resonance?

Stamens in the wind tunnel clearly exhibited non-random, periodic motion indicative of stamen resonance (see Fig. S1). We identified significant spectral peaks at low wind speed for 30 species (Fig. 4a) and the vast majority of stamens resonated. For each flower, we analysed 24.7 ± 1.1 (SD) video segments, of which 20.7 ± 3.6 (83.7% overall) exhibited resonance according to our F-test criteria (Fig. S1b). Visual inspection of spectral densities confirmed several of these peaks (Fig. S1c), but more importantly, there was a strong correspondence between f_n measured by the electrodynamic shaker and f estimated by spectral analysis (averaged across stamens and wind speeds; β = 0.24, SE = 0.07, F_1,29 = 10.9, P = 0.002), thus validating our methodology. Stamens continued resonating with increasing wind speeds, but with larger f at high wind speed (Fig. 4b,c).

What is the role of stamen vibrations in pollen release?

In the wind tunnel, pollen release occurred in episodic bursts of varying durations and times between occurrences. We found significant effects of rms stamen acceleration on pollen release, and f_n on rms stamen acceleration, both with (Table S4) and without phylogenetic corrections (Table S5). The first model indicated that increasing rms stamen accelerations resulted in larger pollen release events (P_MCMC < 0.001; $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0007$ = 56.6, P < 0.001; Fig. 5a). This effect was detected within samples (i.e. the five video segments per flower by wind speed) and among species, indicated by the significant variation in the slopes of flowers ( $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0008$ = 6.33, P = 0.042) and in the intercepts of species. Crucially, pollen release was never limited by pollen availability as only 6.44 ± 11.5% of available pollen was released from flowers during the experiment. Phylogenetic relatedness and species identity had an important effect on pollen release, accounting for 31% of the variation (Table S6). Differences among flowers within species had a much smaller effect, accounting for only 13% of the variation.

As expected, the second model revealed an inverse relation between rms stamen acceleration and f_n (P_MCMC < 0.003; lme4: $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0009$ = 21.9, P < 0.001; Fig. 5b). Unsurprisingly, the effect of wind speed was significant in both models (P_MCMC < 0.003; lme4: $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0010$ = 884.9 P < 0.001), as was its interaction with f_n in lme4 ( $urn:x-wiley:0028646X:media:nph15978:nph15978-math-0011$ = 8.0, P = 0.004).

Discussion

Our comparative study of Thalictrum species revealed four key insights into the process of pollen release relevant for understanding transitions from animal to wind pollination: (1) stamens respond to turbulence by resonating; (2) stamen acceleration is inversely related to stamen f_n; (3) pollen release depends on stamen acceleration, and thus is inversely related to f_n; (4) unlike f_n, pollination syndromes could not reliably predict the probability of pollen release. Our results suggest that f_n is critical in promoting or hindering pollen release in airflows, but that there are likely to be various phenotypic trajectories involved with transitions to wind pollination owing to the diverse floral syndromes of animal-pollinated ancestors. We now discuss the significance of these results and evaluate the utility of the dichotomy of animal and wind pollination commonly used to classify the pollination systems of angiosperm species.

Mechanism of vibrational pollen release

Our study demonstrated that wind-induced stamen vibration is an important cause of pollen release in Thalictrum. This conclusion is strongly supported by the positive association between pollen release and root mean square (rms) stamen acceleration (Fig. 5a) because rms is a measure of vibrational energy (Denny & Gaines, 2000). During trials, we typically observed little stamen motion between pollen release events but sudden frenetic motion at times when airborne pollen was detected, especially at low wind speed. Given how little pollen was released without vibration, it is unlikely that other mechanisms were important for releasing pollen in our experiment (but see Urzay et al., 2009).

Stamen f_n was an important determinant of pollen release due to its negative effect on rms stamen acceleration (Fig. 5b). This result may seem non-intuitive because the expected amplitude of a resonating harmonic oscillator is proportional to the natural frequency of vibration (Denny, 2016). However, this fundamental property of harmonic oscillators assumes that: (1) the excitation has, on average, constant amplitude at all frequencies, as in the case of white noise; and (2) there is no variation among oscillators in the levels of damping (Lalanne, 2014). Neither of these conditions occurred in our study or are likely to prevail during pollination in nature. Previous research on the dynamics of stamens and herbaceous stems indicates that the kinetic energy of turbulent eddies transferred into vibrations decreases with the passage frequency of eddies (Flesch & Grant, 1992; Sterling et al., 2003; Urzay et al., 2009; Timerman et al., 2014). Kinetic energy in turbulence cascades from larger to smaller eddies, with less energy available at each successive scale, until it is dissipated by viscosity (Wyngaard, 2010). Plant organs that respond to wind by resonating (0 < ξ < 1) are excited by a narrow band of frequencies around f_n (de Langre, 2019). Because the turnover frequency of eddies is inversely proportional to their length scale, the kinetic energy of vibration must also decline in relation to f_n, thus violating assumption 1. Damping may also have contributed to our results, given the positive relation between f_n and ξ, and that energy dissipation from damping reduces the amplitude of resonance vibrations (Denny, 2016; de Langre, 2019). Damping in stamens is caused by viscous drag due to air friction (Urzay et al., 2009), but it is unclear why ξ was correlated with f_n. Further experimentation is required to determine the mechanistic basis for this association, as well as the specific balance of turbulence to frictional forces acting on stamens with varying f_n.

The importance of f_n for pollen release was evident at each level of wind speed. Stamen resonance occurred at all wind speeds (Fig. 4), and the inverse effect of f_n on rms stamen acceleration also persisted (Fig. 5b). A lack of interaction between f_n and wind speed was not evident in our previous investigation of ambophilous T. pubescens, in which pollen release was inversely related to f_n for an average wind speed of 1.5 m s⁻¹ but not at a wind speed of 5.0 m s⁻¹ (Timerman & Barrett, 2018). However, the maximum wind speed used in the present study was 2.5 m s⁻¹, and the more limited range of speeds might have been too small to cause an interaction effect. An interaction is expected, as there are probably thresholds for wind speed above and below which stamens vibrate at their maximum amplitude or do not vibrate at all, irrespective of f_n. This may help to explain the flattening of lines for low and high wind speed evident in Fig. 5(b). The potential for this interaction suggests that the evolutionary dynamics of pollen release may depend upon prevailing wind conditions. Variation in f_n could have greater significance for pollen release in calmer than gustier environments, potentially resulting in interspecific differentiation and contrasting phenotypic trajectories during transitions to wind pollination.

Stamen resonance is a widespread phenomenon in wind-pollinated angiosperms and has probably evolved independently in numerous lineages as a simple solution to boundary layer constraints on pollen release (Niklas, 1992; Urzay et al., 2009; Timerman et al., 2014). This is undoubtedly the case in Thalictrum, as revealed by our study (and see Timerman & Barrett, 2018). Stamen resonance is the primary mechanism of pollen release in Plantago lanceolata (Plantaginaceae) and is characteristic of Plantago, the only wind-pollinated genus in ancestrally insect-pollinated Plantaginaceae (Reardon et al., 2009; Timerman et al., 2014). No other angiosperm taxa to our knowledge have been investigated experimentally, but we have observed wind-pollinated species with conspicuously vibrating stamens in Aceraceae, Amaranthaceae, Gunneraceae, Juncaceae, Poaceae, Polygonaceae, Salicaceae and Ulmaceae (D. Timerman, pers. obs.). Other forms of stamen motion have also been linked to pollen release. A mechanism similar to the one investigated here occurs in Halophytum ameghinoi (Halophytaceae) but involves movements of the anther instead of the filament (Pozner & Cocucci, 2006). An alternative mechanism in Cornus canadensis (Cornaceae; Whitaker et al., 2007), Morus alba (Moraceae; Taylor et al., 2006) and Boehmeria caudata (Urticaceae; Montoya-Pfeiffer et al., 2016) involves stamens developing under increasing elastic tension that when released catapults pollen from anthers.

Vibration is also an important feature in the reproductive ecology of other diverse taxa. For example, an analogous mechanism of resonance vibration occurs in some bryophytes, horsetails and fungi and functions in spore dispersal (Grace, 1977; Johansson et al., 2014; Zajączkowska et al., 2017). Pollen release in the c. 20 000 species of buzz-pollinated angiosperms occurs in response to resonance vibrations deliberately applied to flowers by pollinating bees (Harder & Barclay, 1994; King & Buchmann, 1995; Vallejo-Marín, 2019). Resonance vibration of stems may also promote pollen capture (Niklas, 1985; Krick & Ackerman, 2015) and is an important mechanism of wind pollination in grasses and Ephedra (Friedman & Harder, 2004; Niklas, 2015; McCombe & Ackerman, 2018). Despite the prevalence of wind-induced vibration mechanisms in the reproductive ecology of plants, their evolutionary histories, biomechanical mechanisms and effects on fitness have barely been considered. Our study demonstrates that comparative analysis of the biomechanical process of wind-induced vibration mechanisms can reveal new dimensions to the evolution of phenotypic diversity.

Floral diversity and evolution of pollination mechanisms in Thalictrum

Floral traits of Thalictrum species have been conventionally dichotomized as adaptations for either animal or wind pollination (Kaplan & Mulcahy, 1971). However, our study revealed a more complicated relation between floral form and function. Although species we investigated clustered by floral traits according to their presumptive pollination vectors (Fig. 2), the association between pollination syndrome and pollen release probability was not evident in many cases. We did not anticipate this result because species exhibited systematic floral trait differences that conventional wisdom would have attributed to differences in pollination mechanisms associated with the most effective pollen vectors for individual taxa (Faegri & van der Pijl, 1966; Fenster et al., 2004). The lack of association between pollination syndromes and pollen release probability is unlikely to have resulted from experimental error because we found a significant relation between pollen release probability and f_n, as predicted. Rather, our inability to predict the probability of vibrational pollen release from pollination syndromes stemmed from variation in stamen traits within clusters of putatively animal- and wind-pollinated species, through their joint effects on f_n (see Fig. 3b,c).

Several of the species in our comparative analysis exhibited floral traits and patterns of vibrational pollen release consistent with the earlier classification of pollination syndromes by Kaplan & Mulcahy (1971), and subsequently employed in phylogenetic studies of pollination systems in the genus (Soza et al., 2012, 2013; Humphrey & Ossip-Drahos, 2018; Wang et al., 2019). For example, T. cultratum, T. dioicum (Fig. 1c), T. foetidum, T. fendleri, T. leuconotum, T. revolutum and T. venulosum all possess pendulous stamens, a relatively high pollen : ovule ratio (34944 ± 17382), green to white sepals and no perceptible scent. In our wind tunnel analysis, these species ranked among the top 10 for pollen release probability (3.26 × 10⁻⁴ ± 1.48 × 10⁻⁴), consistent with wind pollination. By contrast, T. kiuasianum, T. petaloideum (Fig. 1g) and T. thalictroides (Fig. 1a) all possess erect stamens, a lower pollen–ovule ratio (8900 ± 8220), white to purple flowers and strong fragrances, and ranked among the 10 lowest species for pollen release probability (1.91 × 10⁻⁶ ± 2.17 × 10⁻⁶). Many of the remaining species did not fit well into the dichotomy of animal and wind pollination. For example, T. rutifolium and T. smithii were classified as insect-pollinated (Kaplan & Mulcahy, 1971; Soza et al., 2012), but more closely resembled wind-pollinated species (both species are represented by open circles in the bottom right quadrant of Fig. 2a) and had relatively high pollen release probabilities (8.44 × 10⁻⁴ ± 4.38 × 10⁻⁵). By contrast, T. alpinum was classified as wind-pollinated (filled circle in the top right quadrant of Fig. 2a), but had a very low probability of pollen release (9.41 × 10⁻⁷). Several of the putatively animal-pollinated species also had unexpectedly high pollen release probabilities, including T. baicalense (open circle in bottom left quadrant) with the highest value in our study (1.15 × 10⁻³). Significantly, the species whose typologies (see Kaplan & Mulcahy, 1971) did not conform to the expectations arising from our measurements of P_r did not exhibit a common syndrome of floral traits that could explain their ‘mis-categorization’.

Variation in floral traits of Thalictrum and their functional link to pollen release poses a challenge to traditional concepts of wind and animal pollination as binary states associated with specific adaptive syndromes (Faegri & van der Pijl, 1966; Whitehead, 1969; Friedman & Barrett, 2009). Our study revealed that for Thalictrum species, it may be more accurate to describe wind pollination as a continuously varying condition, at least in terms of pollen dispersal, whose relevance for a species must be evaluated in terms of performance-related traits such as f_n and ξ for pollen release, or capture efficiency η for pollen receipt (Krick & Ackerman, 2015; McCombe & Ackerman, 2018). An explicit prediction from this perspective is that wind pollination mechanisms were assembled from animal-pollinated ancestors, piecemeal through a sequence of various intermediate stages rather than by a single deterministic and predictable pathway of trait change. The existence of intermediate stages is not especially controversial as increasing numbers of species are reported as ambophilous, with a mixture of wind and animal pollination (Stelleman, 1984; Berry & Calvo, 1989; Vroege & Stelleman, 1990; Gomez & Zamora, 1996; Goodwillie, 1999; Totland & Sottocornola, 2001; Culley et al., 2002; Lázaro & Traveset, 2005; Gulías & Traveset, 2012; Yamasaki & Sakai, 2013; Ríos et al., 2014; Wang et al., 2017; Rosado et al., 2018; Saunders, 2018), including several species of Thalictrum (Kaplan & Mulcahy, 1971; Davis, 1997; Steven & Waller, 2004; Timerman & Barrett, 2018).

The evolutionary status of ambophilous species is often unclear (Friedman, 2011), but the condition is not limited to species with obvious animal pollination syndromes and can also occur in those exhibiting stereotypical characteristics of wind pollination (Saunders, 2018). Species with mixed animal and wind pollination may involve reversions to animal pollination, stages in the transition to wind pollination or an evolutionary stable state (Culley et al., 2002; Wragg & Johnson, 2011). Significantly, one of the species included in our comparative analysis (T. pubescens ambophilous; Fig. 1b,e; closed circle in bottom left quadrant of Fig. 2a) was recently the subject of more detailed laboratory and field studies of intraspecific variation in pollen release mechanisms (Timerman & Barrett, 2018). We found that f_n varied significantly among populations and that this variation was heritable. Moreover, the degree of pollinator visitation to experimental arrays resulted in divergent selection of f_n, either promoting or hindering pollen release. These results imply that at least some ambophilous species may maintain sufficient variation in pollen release mechanisms to allow populations to adaptively respond to local pollination environments, enabling transitions in either direction along the continuum from animal to wind pollination. Unfortunately, because of the paucity of field studies of pollination in Thalictrum species, we are uncertain to what extent ambophily may have contributed to variation in the magnitude of wind-induced vibrational pollen release revealed by our study. Future research linking the biomechanical analysis of pollen dispersal traits involved in pollen release, transport and capture to the pollination of natural populations of taxa comprised of both animal- and wind-pollinated species is required for understanding the mechanisms causing evolutionary transitions in pollination systems.

Acknowledgements

We thank Andrijana Stanic, Brina Alsmeyer, Baily McCullough, Cole Brookson, Dianna McAllister, Fernanda Ferreira Pazin, Rachel Woo, Rui Yuan, Shuyue Qiao and Tobias Mankis for technical assistance; Andrew Petrie and Bruce Hall for glasshouse support; Jannice Friedman and Wei Zhou for providing Thalictrum species; and Stuart A. Campbell for discussion. Research funded by a Discovery Grant from the Natural Sciences and Engineering Research Council of Canada (NSERC) to SCHB. DT supported by a CGS-D scholarship from NSERC.

Author contributions

DT and SCHB conceived the study and designed the research, DT performed the laboratory and wind tunnel work and carried out the data analysis, DT and SCHB wrote the manuscript.

Supporting Information

Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

Filename

Description

nph15978-sup-0001-SupInfo.pdfPDF document, 991.2 KB

Fig. S1 Representative example of procedure for determining f_n (arrows) of Thalictrum in the wind tunnel experiment.

Methods S1 Procedure for germinating seeds and cultivating species of Thalictrum.

Methods S2 Procedure for quantifying pollen production per anther in Thalictrum.

Table S1 Correlations between Thalictrum floral trait vectors and factors and ordination of species by principal coordinates analysis.

Table S2 Summary of permutational multivariate analysis of variance for investigating morphological differences between wind- and insect-pollinated species of Thalictrum. Bold: P < 0.05.

Table S3 Summary of phylogenetic least squares analyses for investigating morphological predictors of pollen release probability among species of Thalictrum.

Table S4 Summary of phylogenetic mixed effects models in MCMCglmm for investigating the effects of stamen acceleration on pollen count and f_n on stamen acceleration among species of Thalictrum.

Table S5 Summary of mixed effects models in lme4 for investigating the effects of root mean square (rms) stamen acceleration on pollen count and f_n on stamen acceleration among species of Thalictrum.

Table S6 Proportion of variance explained by random effects in model of pollen release vs stamen acceleration among species of Thalictrum using MCMCglmm.

Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.

References

Ackerman JD. 2000. Abiotic pollen and pollination: ecological, functional, and evolutionary perspectives. Plant Systematics and Evolution 222: 167–185.
10.1007/BF00984101
Web of Science®Google Scholar
Bates D, Mächler M, Bolker BM, Walker SC. 2015. Fitting linear mixed-effects models using lme4. Journal of Statistical Software 67: 1–51.
10.18637/jss.v067.i01
Web of Science®Google Scholar
Berry PE, Calvo RN. 1989. Wind pollination, self-incompatibility, and altitudinal shifts in pollination systems in the high Andean genus Espeletia (Asteraceae). American Journal of Botany 76: 1602–1614.
10.1002/j.1537-2197.1989.tb15145.x
Web of Science®Google Scholar
Brüchert F, Speck O, Spatz HC. 2003. Oscillations of plants’ stems and their damping: theory and experimentation. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences 358: 1487–1492.
10.1098/rstb.2003.1348
PubMedWeb of Science®Google Scholar
Cox PA. 1991. Abiotic pollination: an evolutionary escape for animal-pollinated angiosperms. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences 333: 217–224.
10.1098/rstb.1991.0070
Web of Science®Google Scholar
Cresswell JE, Krick J, Patrick MA, Lahoubi M. 2010. The aerodynamics and efficiency of wind pollination in grasses. Functional Ecology 24: 706–713.
10.1111/j.1365-2435.2010.01704.x
Web of Science®Google Scholar
Culley TM, Weller SG, Sakai AK. 2002. The evolution of wind pollination in angiosperms. Trends in Ecology & Evolution 17: 361–369.
10.1016/S0169-5347(02)02540-5
Web of Science®Google Scholar
Davis SL. 1997. Stamens are not essential as an attractant for pollinators in females of cryptically dioecious Thalictrum pubescens Pursch. (Ranunculaceae). Sexual Plant Reproduction 10: 293–299.
10.1007/s004970050101
Web of Science®Google Scholar
Denny M. 2016. Ecological mechanics: principles of life's physical interactions. Princeton, NJ, USA: Princeton University Press.
10.1515/9781400873951
Google Scholar
Denny M, Gaines S. 2000. Chance in biology: using probability to explore nature. Princeton, NJ, USA: Princeton University Press.
10.1515/9781400841400
Google Scholar
Denny M, Gaylord B, Helmuth B, Daniel T. 1998. The menace of momentum: dynamic forces on flexible organisms. Limnology and Oceanography 43: 955–968.
10.4319/lo.1998.43.5.0955
Web of Science®Google Scholar
Faegri K, van der Pijl L. 1966. The principles of pollination ecology. New York, NY, USA: Pergamon Press.

Google Scholar
Felsenstein J. 1985. Phylogenies and the comparative method. The American Naturalist 125: 1–15.
10.1086/284325
Web of Science®Google Scholar
Fenster CB, Armbruster SW, Wilson P, Dudash MR, Thomson JD. 2004. Pollination syndromes and floral specialization. Annual Review of Ecology, Evolution, and Systematics 35: 375–403.
10.1146/annurev.ecolsys.34.011802.132347
Web of Science®Google Scholar
Flesch TK, Grant RH. 1992. Corn motion in the wind during senescence: II. Effect of dynamic plant characteristics. Agronomy Journal 84: 748–751.
10.2134/agronj1992.00021962008400040038x
Web of Science®Google Scholar
Friedman J. 2011. Gone with the wind: understanding evolutionary transitions between wind and animal pollination in the angiosperms. New Phytologist 191: 911–913.
10.1111/j.1469-8137.2011.03845.x
CASPubMedWeb of Science®Google Scholar
Friedman J, Barrett SCH. 2008. A phylogenetic analysis of the evolution of wind pollination in the angiosperms. International Journal of Plant Sciences 169: 49–58.
10.1086/523365
Web of Science®Google Scholar
Friedman J, Barrett SCH. 2009. Wind of change: new insights on the ecology and evolution of pollination and mating in wind-pollinated plants. Annals of Botany 103: 1515–1527.
10.1093/aob/mcp035
PubMedWeb of Science®Google Scholar
Friedman J, Harder LD. 2004. Inflorescence architecture and wind pollination in six grass species. Functional Ecology 18: 851–860.
10.1111/j.0269-8463.2004.00921.x
Web of Science®Google Scholar
Gelman A, Rubin DB. 1992. Inference from iterative simulation using multiple sequences. Statistical Science 7: 457–472.
10.1214/ss/1177011136
Google Scholar
Gomez JM, Zamora R. 1996. Wind pollination in high-mountain populations of Hormathophylla spinosa (Cruciferae). American Journal of Botany 83: 580–585.
10.1002/j.1537-2197.1996.tb12742.x
Web of Science®Google Scholar
Goodwillie C. 1999. Wind pollination and reproductive assurance in Linanthus parviflorus (Polemoniaceae), a self-incompatible annual. American Journal of Botany 86: 948–954.
10.2307/2656611
CASPubMedWeb of Science®Google Scholar
Grace J. 1977. Plant response to wind. London, UK: Academic Press.

Google Scholar
Gulías J, Traveset A. 2012. Altitudinal variation in the reproductive performance of the Mediterranean shrub Rhamnus lycioides L. Journal of Plant Ecology 5: 330–336.
10.1093/jpe/rtr047
Web of Science®Google Scholar
Hadfield JD. 2010. MCMC methods for multi-response generalized linear mixed models: the MCMCglmm R package. Journal of Statistical Software 33: 1–22.
10.18637/jss.v033.i02
PubMedWeb of Science®Google Scholar
Harder LD, Barclay RMR. 1994. The functional significance of poricidal anthers and buzz pollination: controlled pollen removal from Dodecatheon. Functional Ecology 8: 509–517.
10.2307/2390076
Web of Science®Google Scholar
Humphrey RP. 2016. Pollen heteromorphism is pervasive in Thalictrum (Ranunculaceae). Plant Systematics and Evolution 302: 1171–1177.
10.1007/s00606-016-1312-8
Web of Science®Google Scholar
Humphrey RP. 2018. Sterile stamens do not enhance seed set in females of the cryptically dioecious Thalictrum macrostylum (Ranunculaceae). Journal of the Torrey Botanical Society 145: 82–90.
10.3159/TORREY-D-16-00040.1
Web of Science®Google Scholar
Humphrey RP, Ossip-Drahos AG. 2018. Selection imposed by pollination mode minimally influences evolution of pollen morphology in Thalictrum (Ranunculaceae). International Journal of Plant Sciences 179: 688–696.
10.1086/700175
Web of Science®Google Scholar
Jackson ST, Lyford ME. 1999. Pollen dispersal models in Quaternary plant ecology: assumptions, parameters, and prescriptions. Botanical Review 65: 39–75.
10.1007/BF02856557
Web of Science®Google Scholar
Johansson V, Lönnell N, Sundberg S, Hylander K. 2014. Release thresholds for moss spores: the importance of turbulence and sporophyte length. Journal of Ecology 102: 721–729.
10.1111/1365-2745.12245
Web of Science®Google Scholar
Kaplan SM, Mulcahy DL. 1971. Mode of pollination and floral sexuality in Thalictrum. Evolution 25: 659–668.
10.1111/j.1558-5646.1971.tb01923.x
PubMedWeb of Science®Google Scholar
Keener CS. 1976. Studies in the Ranunculaceae of the southeastern United States II. Thalictrum L. Rhodora 78: 457–472.

Google Scholar
King MJ, Buchmann SL. 1995. Bumble bee-initiated vibration release mechanism of Rhododendron pollen. American Journal of Botany 82: 1407–1411.
10.1002/j.1537-2197.1995.tb12677.x
Web of Science®Google Scholar
Krick J, Ackerman JD. 2015. Adding ecology to particle capture models: numerical simulations of capture on a moving cylinder in crossflow. Journal of Theoretical Biology 368: 13–26.
10.1016/j.jtbi.2014.12.003
PubMedWeb of Science®Google Scholar
Lalanne C. 2014. Random vibration. Hoboken, NJ, USA: John Wiley & Sons Ltd.
10.1002/9781118931158
Google Scholar
de Langre E. 2008. Effects of wind on plants. Annual Review of Fluid Mechanics 40: 141–168.
10.1146/annurev.fluid.40.111406.102135
Web of Science®Google Scholar
de Langre E. 2019. Plant vibrations at all scales: a review. Journal of Experimental Botany 70: 3521–3531.
10.1093/jxb/erz209
PubMedWeb of Science®Google Scholar
Lázaro A, Traveset A. 2005. Spatio-temporal variation in the pollination mode of Buxus balearica (Buxaceae), an ambophilous and selfing species: mainland–island comparison. Ecography 28: 640–652.
10.1111/j.2005.0906-7590.04228.x
Web of Science®Google Scholar
Linder HP. 1998. Morphology and the evolution of wind pollination. In: SJ Owens, PJ Rudall, eds. Reproductive biology in systematics, conservation and economic botany. London, UK: Kew Publishing, 123–135.

Google Scholar
Maechler M, Rousseeuw P, Struyf A, Hubert M, Hornik K. 2018. cluster: cluster analysis basics and extensions, v.2.0.7.1. [WWW document] URL https://CRAN.R-project.org/package=cluster [accessed 1 February 2019].

Google Scholar
McCombe D, Ackerman JD. 2018. Collector motion affects particle capture in physical models and in wind pollination. The American Naturalist 192: 81–93.
10.1086/697551
PubMedWeb of Science®Google Scholar
Melampy MN, Hayworth AM. 1980. Seed production and pollen vectors in several nectarless plants. Evolution 34: 1144–1154.
10.1111/j.1558-5646.1980.tb04057.x
PubMedWeb of Science®Google Scholar
Montoya-Pfeiffer PM, Kevan PG, González-Chaves A, Queiroz EP, Dec E. 2016. Explosive pollen release, stigma receptivity, and pollen dispersal pattern of Boehmeria caudata Sw. (Urticaceae) in a Brazilian rain forest. Botany 94: 607–614.
10.1139/cjb-2016-0031
CASWeb of Science®Google Scholar
Niklas KJ. 1985. The aerodynamics of wind pollination. Botanical Review 51: 328–386.
10.1007/BF02861079
Web of Science®Google Scholar
Niklas KJ. 1987. Pollen capture and wind-induced movement of compact and diffuse grass panicles: implications for pollination efficiency. American Journal of Botany 74: 74–89.
10.1002/j.1537-2197.1987.tb08581.x
Web of Science®Google Scholar
Niklas KJ. 1992. Plant biomechanics: an engineering approach to plant form and function. Chicago, IL, USA: University of Chicago Press.
10.1111/j.1365-3040.2006.01578.x
CASPubMedGoogle Scholar
Niklas KJ. 2015. A biophysical perspective on the pollination biology of Ephedra nevadensis and E. trifurca. Botanical Review 81: 28–41.
10.1007/s12229-014-9148-9
Web of Science®Google Scholar
Oksanen J, Blanchet FG, Friendly M, Kindt R, Legendre P, McGlinn D, Minchin PR, O'Hara RB, Simpson GL, Solymos P et al. 2018. vegan: community ecology package, v.2.5.3. [WWW document] URL https://CRAN.R-project.org/package=vegan [accessed 1 February 2019].

Google Scholar
Paradis E, Schliep K. 2018. ape 5.0: an environment for modern phylogenetics and evolutionary analyses in R. Bioinformatics 35: 526–528.
10.1093/bioinformatics/bty633
Web of Science®Google Scholar
Pinheiro J, Bates D, DebRoy S, Sarkar D, R Core Team. 2018. nlme: linear and nonlinear mixed effects models, v.3.1.137. [WWW document] URL https://CRAN.R-project.org/package=nlme [accessed 1 February 2019].

Google Scholar
Pozner R, Cocucci A. 2006. Floral structure, anther development, and pollen dispersal of Halophytum ameghinoi (Halophytaceae). International Journal of Plant Sciences 167: 1091–1098.
10.1086/508064
Web of Science®Google Scholar
R Core Team. 2018. R: a language and environment for statistical computing, v.3.5.2. Vienna, Austria: R Foundation for Statistical Computing. [WWW document] URL https://www.R-project.org [accessed 1 February 2019].

Google Scholar
Rahim KJ, Burr WS, Thomson DJ. 2014. Appendix A: multitaper R Package. Applications of multitaper spectral analysis to nonstationary data. Kingston, ON, Canada: Queen's University, 149–183.

Google Scholar
Reardon W, Fitzpatrick DA, Fares MA, Nugent JM. 2009. Evolution of flower shape in Plantago lanceolata. Plant Molecular Biology 71: 241–250.
10.1007/s11103-009-9520-z
CASPubMedWeb of Science®Google Scholar
Regal PJ. 1977. Ecology and evolution of flowering plant dominance. Science 196: 622–629.
10.1126/science.196.4290.622
CASPubMedWeb of Science®Google Scholar
Ríos LD, Fuchs EJ, Hodel DR, Cascante-Marín A. 2014. Neither insects nor wind: ambophily in dioecious Chamaedorea palms (Arecaceae). Plant Biology 16: 702–710.
10.1111/plb.12119
CASPubMedWeb of Science®Google Scholar
Rosado A, Vera-Vélez R, Cota-Sánchez JH. 2018. Floral morphology and reproductive biology in selected maple (Acer L.) species (Sapindaceae). Brazilian Journal of Botany 41: 361–374.
10.1007/s40415-018-0452-1
Web of Science®Google Scholar
Sabban L, Jacobson N-L, van Hout R. 2012. Measurement of pollen clump release and breakup in the vicinity of ragweed (A. confertiflora) staminate flowers. Ecosphere 3: 1–24.
10.1890/ES12-00054.1
Web of Science®Google Scholar
Saunders ME. 2018. Insect pollinators collect pollen from wind-pollinated plants: implications for pollination ecology and sustainable agriculture. Insect Conservation and Diversity 11: 13–31.
10.1111/icad.12243
Web of Science®Google Scholar
Soza VL, Brunet J, Liston A, Salles Smith P, Di Stilio VS. 2012. Phylogenetic insights into the correlates of dioecy in meadow-rues (Thalictrum, Ranunculaceae). Molecular Phylogenetics and Evolution 63: 180–192.
10.1016/j.ympev.2012.01.009
PubMedWeb of Science®Google Scholar
Soza VL, Haworth KL, Di Stilio VS. 2013. Timing and consequences of recurrent polyploidy in meadow-rues (Thalictrum, Ranunculaceae). Molecular Biology and Evolution 30: 1940–1954.
10.1093/molbev/mst101
CASPubMedWeb of Science®Google Scholar
Spatz H-C, Zebrowski J. 2001. Oscillation frequencies of plant stems with apical loads. Planta 214: 215–219.
10.1007/s004250100627
CASPubMedWeb of Science®Google Scholar
Stelleman P. 1984. Reflections on the transition from wind pollination to ambophily. Acta Botanica Neerlandica 33: 497–508.
10.1111/j.1438-8677.1984.tb01841.x
Web of Science®Google Scholar
Sterling M, Baker CJ, Berry PM, Wade A. 2003. An experimental investigation of the lodging of wheat. Agricultural and Forest Meteorology 119: 149–165.
10.1016/S0168-1923(03)00140-0
Web of Science®Google Scholar
Steven JC, Waller DM. 2004. Reproductive alternatives to insect pollination in four species of Thalictrum (Ranunculaceae). Plant Species Biology 19: 73–80.
10.1111/j.1442-1984.2004.00103.x
Google Scholar
Taylor PE, Card G, House J, Dickinson MH, Flagan RC. 2006. High-speed pollen release in the white mulberry tree, Morus alba L. Sexual Plant Reproduction 19: 19–24.
10.1007/s00497-005-0018-9
Web of Science®Google Scholar
Timerman D, Barrett SCH. 2018. Divergent selection on the biomechanical properties of stamens under wind and insect pollination. Proceedings of the Royal Society of London. Series B, Biological Sciences 285: 20182251.
10.1098/rspb.2018.2251
PubMedWeb of Science®Google Scholar
Timerman D, Greene DF, Urzay J, Ackerman JD. 2014. Turbulence-induced resonance vibrations cause pollen release in wind-pollinated Plantago lanceolata L. (Plantaginaceae). Journal of the Royal Society Interface 11: 20140866.
10.1098/rsif.2014.0866
PubMedWeb of Science®Google Scholar
Totland Ø, Sottocornola M. 2001. Pollen limitation of reproductive success in two sympatric alpine willows (Salicaceae) with contrasting pollination strategies. American Journal of Botany 88: 1011–1015.
10.2307/2657082
CASPubMedWeb of Science®Google Scholar
Urzay J, Llewellyn Smith SG, Thompson E, Glover BJ. 2009. Wind gusts and plant aeroelasticity effects on the aerodynamics of pollen shedding: a hypothetical turbulence-initiated wind-pollination mechanism. Journal of Theoretical Biology 259: 785–792.
10.1016/j.jtbi.2009.04.027
PubMedWeb of Science®Google Scholar
Vallejo-Marín M. 2019. Buzz pollination: studying bee vibrations on flowers. New Phytologist 224: 1068–1074.
10.1111/nph.15666
PubMedWeb of Science®Google Scholar
Vogel S. 1994. Life in moving fluids: the physical biology of flow. Princeton, NJ, USA: Princeton University Press.

Google Scholar
Vroege PW, Stelleman P. 1990. Insect and wind pollination in Salix repens L. and Salix caprea L. Israel Journal of Botany 39: 125–132.

Web of Science®Google Scholar
Wang L-L, Zhang C, Yang M-L, Zhang G-P, Zhang Z-Q, Yang Y-P, Duan Y-W. 2017. Intensified wind pollination mediated by pollen dimorphism after range expansion in an ambophilous biennial Aconitum gymnandrum. Ecology and Evolution 7: 541–549.
10.1002/ece3.2636
CASPubMedWeb of Science®Google Scholar
Wang TN, Clifford MR, Martínez-Gómez J, Johnson JC, Riffell JA, Di Stilio VS. 2019. Scent matters: differential contribution of scent to insect response in flowers with insect vs. wind pollination traits. Annals of Botany 123: 289–301.
10.1093/aob/mcy131
CASPubMedWeb of Science®Google Scholar
Weller SG, Sakai AK, Rankin AE, Golonka A, Kutcher B, Ashby KE. 1998. Dioecy and the evolution of pollination systems in Schiedea and Alsinidendron (Caryophyllaceae: Alsinoideae) in the Hawaiian Islands. American Journal of Botany 85: 1377–1388.
10.2307/2446396
CASPubMedWeb of Science®Google Scholar
Welsford MR, Hobbhahn N, Midgley JJ, Johnson SD. 2016. Floral trait evolution associated with shifts between insect and wind pollination in the dioecious genus Leucadendron (Proteaceae). Evolution 70: 126–139.
10.1111/evo.12821
PubMedWeb of Science®Google Scholar
Whitaker DL, Webster LA, Edwards J. 2007. The biomechanics of Cornus canadensis stamens are ideal for catapulting pollen vertically. Functional Ecology 21: 219–225.
10.1111/j.1365-2435.2007.01249.x
Web of Science®Google Scholar
Whitehead DR. 1969. Wind pollination in the angiosperms: evolutionary and environmental considerations. Evolution 23: 28–35.
10.2307/2406479
PubMedWeb of Science®Google Scholar
Wragg PD, Johnson SD. 2011. Transition from wind pollination to insect pollination in sedges: experimental evidence and functional traits. New Phytologist 191: 1128–1140.
10.1111/j.1469-8137.2011.03762.x
PubMedWeb of Science®Google Scholar
Wyngaard JC. 2010. Turbulence in the atmosphere. Cambridge, UK: Cambridge University Press.
10.1017/CBO9780511840524
Google Scholar
Yamasaki E, Sakai S. 2013. Wind and insect pollination (ambophily) of Mallotus spp. (Euphorbiaceae) in tropical and temperate forests. Australian Journal of Botany 61: 60–66.
10.1071/BT12202
Web of Science®Google Scholar
Zajączkowska U, Kucharski S, Nowak Z, Grabowska K. 2017. Morphometric and mechanical characteristics of Equisetum hyemale stem enhance its vibration. Planta 245: 835–848.
10.1007/s00425-017-2648-1
CASPubMedWeb of Science®Google Scholar

Citing Literature

Volume224, Issue3

Special Issue:The ecology, evolution, and genetics of plant reproductive systems

November 2019

Pages 1121-1132

Comparative analysis of pollen release biomechanics in Thalictrum: implications for evolutionary transitions between animal and wind pollination

Summary

Introduction