Flowering phenology, growth forms, and pollination syndromes in tropical dry forest species: Influence of phylogeny and abiotic factors

Abstract

PREMISE OF THE STUDY:

Analyses of the influence of temporal variation in abiotic factors on flowering phenology of tropical dry forest species have not considered the possible response of species with different growth forms and pollination syndromes, while controlling for phylogenetic relationships among species. Here, we investigated the relationship between flowering phenology, abiotic factors, and plant functional attributes, while controlling for phylogenetic relationship among species, in a dry forest community in Mexico.

METHODS:

We characterized flowering phenology (time and duration) and pollination syndromes of 55 tree species, 49 herbs, 24 shrubs, 15 lianas, and 11 vines. We tested the influence of pollination syndrome, growth form, and abiotic factors on flowering phenology using phylogenetic generalized least squares.

KEY RESULTS:

We found a relationship between flowering duration and time. Growth form was related to flowering time, and the pollination syndrome had a more significant relationship with flowering duration. Flowering time variation in the community was explained mainly by abiotic variables, without an important phylogenetic effect. Flowering time in lianas and trees was negatively and positively correlated with daylength, respectively.

CONCLUSIONS:

Functional attributes, environmental cues, and phylogeny interact with each other to shape the diversity of flowering patterns. Phenological differentiation among species groups revealed multiples strategies associated with growth form and pollination syndromes that can be important for understanding species coexistence in this highly diverse plant community.

The temporal distribution of flowering has an important influence on plant reproductive success (Munguía-Rosas et al., 2011). Therefore, strong selection favoring synchronization of flowering phenology with abiotic and biotic conditions that promote pollination and favor seed development can be expected (Elzinga et al., 2007; Kudo et al., 2008). Flowering phenology at the community level is a spectrum of different plant responses arising from both exogenous (biotic and abiotic) and intrinsic (i.e., genetic and phylogenetic constraints) factors (Rathcke and Lacey, 1985; van Schaik et al., 1993). For example, flowering response to abiotic factors such as water availability, temperature, and daylength (Borchert et al., 2004; Imaizumi and Kay, 2006; Springate and Kover, 2014) is determined by physiological processes associated with hormone balance, water content in plant tissues, or the amount of nonstructural carbohydrates (Wilczek et al., 2010).

In seasonally dry tropical forests, water is a temporally limited resource, and this seasonal variation is one of the most important abiotic factors influencing plant phenology (van Schaik et al., 1993; Bullock, 1995). Despite the high cost of producing flowers during the dry season, various plant groups flower outside the rainy season and instead, respond to other factors such as daylength (Borchert et al., 2004) and solar radiation (Zimmerman et al., 2007). Consequently, in these ecosystems it is possible to observe different flowering phenology patterns, which can be related to plant strategies to store water in stems and/or roots or to access water sources deep in the soil (Bullock, 1995; Borchert et al., 2004). One way to document the different phenological behaviors in a community is to compare species with contrasting strategies of resource acquisition, as is the case for species that differ in growth form (Sarmiento and Monasterio, 1983; Frankie et al., 2004; Marques et al., 2004). However, so far, most of the existing information on flowering phenology in seasonally dry tropical forests is based on woody plants, mainly trees (Bullock and Solis-Magallanes, 1990; Singh and Kushwaha, 2006).

Pollination vector is another factor influenced by flowering time, constituting one of the most important interactions for plant reproductive success (van Schaik et al., 1993; Bolmgren et al., 2003; Elzinga et al., 2007). This mutualism may influence the evolution of flowering time through competition for pollinators, which can cause selection for asynchronous flowering (Mosquin, 1971; Bell et al., 2005), or facilitation where pollinator service is improved by synchronous flowering (Rathcke and Lacey, 1985; van Schaik et al., 1993). Pollination systems have an important link with flowering period, because they determine the time available for pollination, which is important for both plant and pollinator. However, flowering duration has been examined less than flowering time, and few studies have considered time and duration as additional traits for recognizing pollination syndromes. For example, in seasonally dry tropical forests, species pollinated by bees produce flowers during the dry season, while plants pollinated by moths flower during the rainy season (Janzen, 2004; Frankie et al., 2004), and the duration of flowering will be longer in species flowering during the rainy season.

To understand the response of species to local abiotic conditions in seasonally dry forests, it is important to consider both growth form and pollination syndrome. However, another important factor influencing flowering phenology is the evolutionary history of the species (Davis et al., 2010). Previous studies have shown a phylogenetic signal in flowering patterns (Bolmgren and Cowan, 2008; Staggemeier et al., 2010; Davies et al., 2013; Lessard-Therrien et al., 2013; Du et al., 2015), however, in seasonally dry tropical forests, this subject has been scarcely explored, but see Silva et al. (2011).

This study was undertaken in a seasonally dry tropical forest in the floristic province of the Balsas River Basin, one of the most important in Mexico for its high diversity and proportion of endemism, making it a priority conservation region at a global scale (Olson and Dinerstein, 2002). One notable aspect of this tropical dry forest is the low annual precipitation (around 600 mm) and the long dry season (8-9 months). Although water availability is a limited resource in this type of forest at the worldwide scale, it is interesting to note that only a small fraction (10–37%) of tree species blooms during the rainy season (Kushwaha et al., 2011). This pattern suggests a diversity of strategies. Therefore, documenting the different flowering responses in relation to plant traits and environmental factors (e.g., precipitation or daylength) will be important for understanding the strategies of different species to survive and coexist in these resource-limited environments. Here, we considered two traits that are potentially related with flowering phenology—growth form and pollination syndrome—while simultaneously accounting for the influence of phylogenetic relationships among species. First, we described the flowering phenology of species in a tropical dry forest according to their growth form and pollination syndrome, and second we addressed the following questions: (1) Is there a relationship of growth form and pollination syndrome with flowering time and duration? (2) Do precipitation, daylength, and temperature influence the flowering time and duration?

MATERIALS AND METHODS

Study site

The study site is located in Churumuco, Michoacán, Mexico (18°38′–18°44′ N and 101°38′–101°41′W), which is part of the floristic province of the Balsas Depression. This site has a complex topography with altitudes varying between 490 and 1200 m a.s.l. The climate is dry and warm with a summer rain regime, and the temperatures oscillate between 23 and 36°C, with a mean annual temperature of 29.4°C. Total annual precipitation is 564 mm, with most rainfall concentrated between July and September (Appendix 1). In this forest, trees are deciduous during the dry season and are, on average, 6.3 m in height. The most frequent tree species are Acacia picachensis, Amphipterygium adstringens, Apoplanesia paniculata, Backebergia militaris, Bursera sarukhanii, Bursera infernidialis, Caesalpinia eriostachys, Gossypium lobatum, Handroanthus impetiginosus, Heteroflorum sclerocarpum, Lonchocarpus huetamoensis, Opuntia bensonii, Randia thurberi, and Stenocereus quevedonis.

Flowering phenology records

From January 2013 to December 2014, a monthly record was made of the presence of flower buds and flowers in anthesis (i.e., those in which reproductive structures such as stamens and/or pistils were visible) in 154 plant species—55 trees, 49 herbs, 24 shrubs, 15 lianas, and 11 vines. In the case of woody plants (trees, shrubs, and lianas) we followed 5–20 permanently labeled individuals per species along a 6 km long and 15 m wide trail, found between 500 and 650 m a.s.l. Because of the clonal growth of several of the herbaceous species (herbs and vines), their flowering phenology was recorded in 10 different points along the trail.

Pollination syndromes

From field observations, flower morphology, and specialized literature, we distinguished 10 pollination syndromes: bats, beetles, birds, diurnal butterflies, flies, large bees, moths, small bees, small insects, and wind (Endress, 1994). Bee pollination syndrome was divided into two categories (large and small), according to Frankie et al. (2004). Pollination by bees >1.2 cm in length (e.g., Apis mellifera, euglossines, and xylocopas) was associated with large flowers with bright colors, and frequently having bilateral symmetry, while pollination by bees <1.2 cm in length (anthophorids, halicitids, meliponines, and megachilids) was associated with small flowers, creamy colors, radial symmetry, and grouped in inflorescences. The small-insect syndrome included species with flowers <1 cm in diameter with green or pale yellow colors; these species were potentially pollinated by different insect groups such as bees, wasps, flies, and/or beetles (Bawa et al., 1985).

Phenological data analyses

The temporal distribution of flower buds and flowers in anthesis for species grouped according to growth form and pollination syndrome was examined using circular statistics. For this, we calculated the mean angle (ā), indicating the date when the flowering of a species occurs. To calculate ā, months were converted into angles (e.g., January corresponds to 0°, February to 30°, and December to 330°; Morellato et al., 2010), and the statistical significance of ā was determined using Rayleigh's test (z). A nonsignificant value of ā indicates that flowers are found throughout the year. In addition, we determined the vector r, which indicates the concentration of data around the mean angle as a proxy for synchrony. This vector is an inverse measure of circular variance and varies between 0, when the event is uniformly distributed throughout the year, to 1, when the event is concentrated in a specific time period. Flowering duration indicates the number of months one species was observed in anthesis, which was determined by averaging the number of months of flowering for each individual in every species.

For subsequent analysis, flowering dates (anthesis) were transformed from a circular to a linear scale, following the method of Staggemeier et al. (2010). From flowering ā, we constructed a matrix of phenological similitude using Euclidean distances. Through an eigenfunction analysis we obtained the ordination vectors derived from this matrix using the eigenfunction in R (R Core Team, 2015). The eigenvalues of the first eigenvector was associated with 74% of phenological variation; therefore, it was used as the phenological vector. To evaluate the equivalence between the phenological vector and the original data from ā, a second similitude matrix was constructed from the phenological vector. The correlation between the two matrices was determined with a Mantel test using the ‘ade4’ (Dray and Dufour, 2007) package in R. Because the association between the two matrices was 73% (P < 0.001), the phenological vector was considered an accurate representation of flowering time.

Phylogenetic reconstruction

The frequency of pollination syndromes and their association with growth forms were compared using a log-linear model (Crawley, 2012). To consider the effect of phylogenetic relationships on the phenological patterns of the species, we constructed a phylogenetic tree for the 154 species, using the Phylomatic software (Webb et al., 2008), based on an Angiosperm supertree (APG III, 2009). Using the ‘bladj’ algorithm in Phylocom, node age, and branch length in millions of years were assigned to the phylogeny. Divergence times were based on estimations for the major clades of angiosperms from Bell et al. (2010). The ‘bladj’ algorithm fixes the root node and other nodes based on the ages provided and, evenly places nodes that lack information in between those with known ages to produce a pseudo-chronogram (Webb et al., 2008). The resulting phylogenetic tree is shown in Appendix 2. The reconstruction method often leaves many polytomies at the genus level unresolved, but phylogenetic information is still scarce for species at the study region. However, this method has been used in different studies (e.g., Silva et al., 2011; Davies et al., 2013; Du et al., 2015) and is an important tool in the absence of more detailed phylogenetic information.

Influence of growth form and pollination syndrome on flowering time and duration

To account for the effect of shared evolutionary history, we used a phylogenetic generalized least squares model (PGLS) with the λ parameter estimated using maximum likelihood as a measure of phylogenetic signal. This analysis was performed with the ‘caper’ package (Orme et al., 2012) in R (Appendix S1, see the online Supplementary Data tab with this article). Using the PGLS model, we determined the correlation between flowering time and duration;, and we compared flowering duration in the four seasons of the year to visualize differences in flowering duration associated with flowering time. Finally, we evaluated the relationship between flowering time, duration (response variables), and growth form and pollination syndrome (predictor variables). Because of the low number of vine species (11) and their phenological similarity with the herbaceous species, we grouped both growth forms for this and subsequent analyses.

Influence of environmental factors on flowering time and duration

Monthly precipitation and temperature data were obtained from the nearest meteorological station to the study site (Appendix 1). Day-length data were obtained from the Applications Department of the U.S. Naval Observatory (http://www.usno.navy.mil). The effect of the abiotic environment on the phenological response of the species was evaluated in two ways. In the first analysis, following Staggemeier et al. (2010), we represented the environmental gradient occupied by the species during their flowering stage. Environmental eigenvectors were calculated by constructing two matrices. The first matrix was constructed with the species placed in columns and the 24 mo of observation in rows. Each cell was filled with the number of flowering individuals per species. The second matrix was constructed with environmental variables in columns and months in rows, and cells were filled with total monthly data of precipitation and mean monthly daylength and temperature. From these two matrices, we ran a canonical correspondence analysis with the ‘vegan’ package in R (Appendix S2). The first two ordination axes were significant (cca1 F = 4.40, P = 0.001; cca2 F = 3.64, P = 0.001) and were therefore used as predictor environmental variables. Then, a PGLS was used to evaluate the influence of environmental vectors on flowering time (phenological vector) and duration, while taking the phylogeny into account.

For the second analysis, we explored the influence of daylength, rainfall, and temperature on the temporal distribution of flowering species at the whole-community level and for groups of species according to their growth form and pollination syndrome without considering phylogenetic effects. In this analysis, a generalized linear mixed model was used, with the flowering time values predicted by rainfall and daylength (fixed factors) and month (random factor). Because temperature was highly correlated with daylength (coefficient of correlation, r = 0.69), for the analysis only the last variable was used because it explained a higher amount of variation. To evaluate the models, both the marginal and the conditional R² were calculated. The first one describes the proportion of variance explained by the fixed factors alone, while the second describes the proportion of variance explained by both the fixed and random factors. The choice of fixed and random effects was done considering the Akaike information criterion (AIC) value of alternative models. Final models were analyzed using the restricted maximum likelihood estimator (REML), with the ‘lme 4’ package in R.

RESULTS

Flowering time and pollination syndromes

All analyzed species in this tropical dry forest showed seasonal production of flower buds and flowers. At the community level, a significant concentration of flower buds and flowers in anthesis was observed during the rainy season in September (Fig. 1, Table 1). When species were grouped according to growth form, the ā of flowering was significant, indicating seasonal behaviors. This phenophase occurred during March (dry season) in tree species, in comparison to herbs (September), vines (September), and lianas (October). Shrubs showed a uniform pattern throughout the year (Table 1).

Table 1. Results of the circular statistics analyses to characterize the phenology of flower buds and flowers in anthesis, in species with different growth forms in a seasonally dry tropical forest.

Level	Flowering	Mean Angle	Mean Vector (r)	Rayleigh Test (z)
Community (154)	Flower bud	Sep 4 (239.85°)	0.2	12.54**
		Sep 27 (262.59°)	0.25	19.59**
	Flower	Sep 27 (262.81°)	0.22	16.81**
		Sep 28 (263.07°)	0.27	27.49**
Trees (55)	Flower bud	Mar 21 (76.64°)	0.33	13.60**
		Mar 10 (65.81°)	0.24	7.66**
	Flower	Mar 26 (81.04°)	0.28	11.31**
		Mar 28 (83.91°)	0.24	8.375**
Shrubs (24)	Flower bud	169.10°	0.04	0.083
		209.08°	0.18	1.579
	Flower	241.23°	0.13	0.81
		235.68°	0.16	1.73
Lianas (15)	Flower bud	Sep 9 (244.98°)	0.41	4.72**
		Sep 25 (260.1°)	0.26	2.87**
	Flower	Oct 7 (272.13°)	0.45	9**
		Oct 12 (277.72°)	0.45	8.70**
Vines (11)	Flower bud	Sep 15 (250.42°)	0.93	17.3**
		Sep 16 (251.77°)	0.95	16.35**
	Flower	Oct 2 (267.08°)	0.93	18.45**
		Oct 13 (278.69°)	0.87	43.95**
Herbs (49)	Flower bud	Sep 22 (257.03°)	0.89	48.70**
		Sep 22 (257°)	0.90	44.51**
	Flower	Oct 1 (265.5°)	0.87	54.38**
		Sep 26 (261.32°)	0.94	18.83**

• Note: Data from the first and second year are shown in the first and second line, respectively. The significance level of z corresponds to: *P≤ 0.05; **P≤ 0.001. The number of species analyzed for each category is shown in parenthesis in the first column.

Figure 1

Flowering phenology at the community level and for species groups with different growth forms and pollination syndromes during two years (2013–2014), in a seasonally dry tropical forest.

According to the log-linear model, at the community level there was a significant difference in the frequency of the different pollination syndromes (χ²₆ = 98.65, P < 0.001). The species with flower attributes related to pollination by large bees represented 46% of the total, and adding the species pollinated by small bees, this percentage increased to 57% (Fig. 2). Other frequent syndromes were pollination by small insects (14%) and moths (9%). There was also a significant association between pollination syndrome and growth form (χ²₁₇ = 67.4, P < 0.001), however, in all growth forms other than lianas, pollination by bees was the most frequent syndrome. On the other hand, two pollination syndromes were found in a single growth form (Fig. 2); pollination by flies (in lianas) and bats (in trees).

Figure 2

Frequency of species for each pollination syndrome at the community level and for the different growth forms in a seasonally dry tropical forest.

Because of the high phenological similarity between the small-bee and the large-bee syndrome, in Fig. 1 we pooled both as “pollination by bees”. The flowering occurred uniformly throughout the year in species pollinated by birds and small insects (Fig. 1, Table 2). For most pollination syndromes, flowering was significantly concentrated in the rainy season (August and September), with the exception of species pollinated by bats, which had flowers during the dry season (February and March, Table 2).

Table 2. Results of the circular statistics analyses to characterize the flowering phenology in species with different pollination syndromes in a seasonally dry tropical forest.

Pollination syndrome	Mean Angle	Mean Vector (r)	Rayleigh Test (z)
Bat (6)	Mar 13 (68.41°)	0.36	4.08**
	Feb 26 (51.2°)	0.34	3.25**
Bird (9)	218.26°	0.07	0.12
	318.06°	0.21	1.01
Butterfly (4)	Sep 10 (245.1°)	0.80	4.51**
	Sep 20 (255°)	0.75	3.97**
Large bee (71)	Sep 23 (258.79°)	0.27	13.22**
	Sep 28 (263.14°)	0.38	28.74**
Moth (14)	Sep 27 (262.81°)	0.51	9.66**
	Sep 10 (245.03°)	0.53	10.52**
Small bee (14)	Aug 14 (219.97°)	0.62	5.93**
	Sep 11(246.1°)	0.69	11.51**
Small insects (29)	263.08°	0.09	0.61
	253.11°	0.08	0.47
Wind (7)	Sep 10 (255°)	0.59	4.59**
	Sep 6 (241.54°)	0.35	1.76

• Note: Data from the first and second year are shown in the first and second line, respectively. The significance level of z corresponds to: *P≤ 0.05; **P≤ 0.001. The number of species in each category is shown in parenthesis.

Influence of growth form and pollination syndrome on flowering time and duration

Flowering duration of all species in the community was on average 1.74 mo, ranging from 1–5 mo (Table 3). A significant relationship between flowering duration and time was observed (R² = 0.14, P < 0.001, λ = 0.20). Flowering duration differed among the four seasons of the year (F_{3, 147} = 8.94, P < 0.001. λ = 0.30). Species that produced flowers during the winter had longer flowering periods than species that flowered in the other seasons (P < 0.001, in all cases).

Table 3. Duration of flowering for all species (whole community level) and groups of species, categorized by growth form and pollination syndrome. The different superscripted letters indicate significant differences of Tukey posthoc tests among groups. Number of species in each category is shown in parenthesis.

Level	Duration (months)	Coefficient of variation
Community (154)	1.74 ± 0.90	0.42
Growth form
Trees (55)	1.87 ± 0.77^b	0.39
Shrubs (24)	1.67 ± 0.74^ab	0.48
Lianas (15)	2.05 ± 0.90^ab	0.45
Vines (11)	1.57 ± 0.53^a	0.27
Herbs (49)	1.58 ± 0.44^a	0.16
Pollination syndrome
Bat (6)	3.25 ± 0.51^b	0.15
Bird (9)	2.06 ± 1.08^a	0.5
Large bee (71)	1.70 ± 0.96^a	0.44
Moth (14)	1.65 ± 1.24^a	0.52
Small bee (14)	1.49 ± 0.59^a	0.32
Small insects (29)	1.61 ± 0.65^a	0.31
Wind (7)	1.85 ± 0.54^a	0.38

Growth form was significantly related to flowering time (F_{3, 147} = 10.52, P < 0.001, λ = 0.30). Specifically, flowering time in herbs occurred later than trees, shrubs (P < 0.01, in all cases), and without differences in woody growth forms. We did not find a relationship between pollination syndrome and flowering time (F_{6, 144} = 1.44, P = 0.19, λ = 0.16).

Flowering duration differs according to the growth form (F_{3, 147} = 4.11, P < 0.01, λ = 0.53). Only tree species were significant related to more extended flowering periods (P = 0.001). Pollination syndrome was also significantly related to flowering duration (F_{6, 144} = 6.55, P < 0.001, λ = 0.00). Flowering was longer in bat-pollinated species, in comparison to other pollination syndromes (P < 0.01). Besides, we did not find differences in the flowering duration among other syndromes (P > 0.05, in all cases). In general, flowering duration increased from butterfly-pollinated species to small insect-pollinated, bee-pollinated, and bat-pollinated species (Table 3).

Influence of environmental factors on flowering time and duration

At the community level, the effects of environmental vectors on the flowering time were significant (R² = 0.79, λ = 0. 01, P < 0.001). In contrast, flowering duration was less related to environmental variation and showed a higher value of the phylogenetic signal (R² = 0.15, λ = 0. 34, P < 0.001).

The role of environmental factors differed according to growth form (Table 4). Flowering time in tree species was negatively related to rainfall and positively related to daylength. In shrubs, no significant relationships were detected between the phenology of reproductive structures and environmental factors. In lianas, flowering was negatively correlated with daylength and varied throughout the months. In herbaceous species, monthly variation had a highly significant effect on flowering phenology (Table 4). In contrast, when flowering of species were grouped by pollination syndrome, no significant effect of abiotic factors was identified. However, in all cases, the models including both fixed and random effects explained the highest proportion of variation in flowering phenology associated with pollination syndrome (Appendix 3).

Table 4. Results of the generalized linear mixed effects models to evaluate the relationship between flowering time and abiotic factors in species with different growth form of a seasonal tropical dry forest in Mexico. Abbreviations: AIC (Akaike information criterion), R²m (marginal R²), R²c (conditional), SE (standard error) and σ² (residual variance of random effect). Coefficients significance: *P ≤ 0.05, P ≤ **0.001.

Variables	Coefficients	SE	z	AIC	R^2m	R^2c	σ^2
Community
Fixed effects				175.9	0.05	0.83
Rainfall	0.04	0.18	−0.81
Daylength	−0.14	0.001	0.67
Random effects
Month							0.18
Trees
Fixed effects				135	0.62	0.68
Rainfall	−0.42	0.12	−3.3**
Daylength	0.27	0.1	2.6**
Random effects
Month							0.13
Shrubs
Fixed effects				105.9	0.01	0.22
Rainfall	0.057	0.11	0.48
Daylength	0.032	0.17	0.19
Random effects
Month							0.05
Lianas
Fixed effects				100.5	0.31	0.64
Rainfall	0.24	0.15	1.56
Daylength	−0.73	0.3	−2.38**
Random effects
Month							0.30
Herbs
Fixed effects				94.1	0.009	0.87
Rainfall	0.063	0.11	0.55
Daylength	0.44	1.58	0.27
Random effects
Month							4.44

DISCUSSION

Flowering phenology and pollination syndromes

At the community level, flowering was concentrated at the end of the rainy season, which is different from what has been documented in other seasonally dry tropical forests, where the flowering peak occurs at the beginning of the rainy season (Bullock and Solis-Magallanes, 1990; Silva et al., 2011; Carvalho and Sartori, 2015). The flowering pattern observed in this community is congruent with the view that most plant species in seasonally dry tropical forests use water and available resources to maximize their vegetative growth during the dry season. In woody species, this behavior has also been interpreted as a competitive strategy, given the light limitation for the expansion of leaves and shoots during the rainy season (Janzen, 1967).

Entomophilous pollination, particularly by bees, is the most frequent syndrome observed in seasonally dry tropical forests, both in woody and herbaceous species, and there is a low frequency of anemophilous species (Bullock, 1995; Frankie et al., 2004; Quirino and Machado, 2014; Hilje et al., 2015). We also found this pattern in our results; despite the association between pollination syndrome and growth form, 86% of the species showed entomophilous pollination, and only 4% of the species were wind pollinated. Also, pollination by bats and moths was more frequent in woody species (trees, shrubs, and lianas), as has also been reported in previous studies (Frankie et al., 2004; Quirino and Machado, 2014).

Influence of growth form and pollination syndrome on flowering time and duration

Even though the values of the phylogenetic signal were not high for most of the correlations, the evolutionary history of the species had an influence at different levels when we associated flowering phenology with other biological attributes. As in other studies, we found a relationship between flowering time and duration (Bawa et al., 2003; CaraDonna and Inouye, 2015). However, in this community, flowering had a longer duration in species that produce reproductive structures during the dry season (particularly during the winter) and not in species that flower during the rainy season (Bullock, 1995). Janzen (1967) suggested that species that maximize their vegetative growth during the rainy season can also store more resources and have more extended flowering periods.

We found an important relationship between growth form and flowering time. This observation is consistent with previous findings in other seasonal communities (Batalha and Martins, 2004; Stevenson et al., 2008). In our study, the phenology of herbaceous species (vines and herbs) was tightly linked with the rainy season. This pattern may be related to the shallow root systems of many herbaceous species, which allow them to respond rapidly to changes in water availability (Sarmiento and Monasterio, 1983).

In this forest, it is possible to observe woody species (lianas, shrubs, and trees) bearing flowers throughout the entire year. However, it is worth noting that the flowering peaks occurred at different times in these growth forms. As in other seasonally dry tropical forests, lianas showed a flowering peak at the beginning of the dry season (Morellato and Leitao-Filho, 1996). The phenological behavior of trees was consistent with the hypothesis suggested by Janzen (1967) that tree species in dry tropical forests should flower in the dry season because the rainy season is the major period for vegetative growth for these species. Flowering phenology in shrubs was not seasonal and contrasted with shrubby species in other seasonally dry tropical forests, in which flowering is related to the rainy season, indicating that many of those species respond to the first rains displaying their flowers (Opler et al., 1980; Frankie et al., 2004; de Vasconcelos et al., 2010). Woody species are capable of producing flowers during the driest time of the year, because of morphological adaptations such as water storing stems (e.g., Bursera spp., Dieterlea fusiformis, Jatropha spp.) or deep roots to access aquifers (e.g., Caesalpinia spp., Combretum farinosum, Randia thurberi).

Contrary to what has been reported for other tropical dry forest sites (Bullock, 1995; van Schaik et al., 1993; Frankie et al., 2004; Elzinga et al., 2007), flowering time was not related to pollination syndrome at the community level. This result arises from the inclusion of different growth forms into the same pollination syndrome. Therefore, when controlling for the effect of growth form we found different phenological patterns associated with pollination syndromes (Appendix 4). Flowering during the dry season in tree species pollinated by bees, and a tendency to flower in the rainy season for moth-pollinated liana species, were observations that were similar to previous records in other tropical dry forests (Janzen, 2004; Frankie et al., 2004). The staggered pattern of flowering in distinct pollination syndromes within woody species can be interpreted as a form of diffuse facilitation, because constant flower production can favor the maintenance of a community of pollinators throughout the year (Feinsinger, 1987).

Flowering duration differed among growth forms, and phylogeny had an important effect on this correlation. Our results contrast with the tendency of a more extended flowering in understory species (Frankie et al., 2004; Rathcke and Lacey, 1985). We found an increase in flowering duration from herbs to trees and to lianas. This pattern results from the short flowering period of many herbaceous species, which depend on the immediate availability of water, while canopy species can maintain longer flowering periods because of their capacity to store water or to access underground sources of water (Seghieri et al., 1995).

Pollination syndrome had an effect on flowering duration. Particularly, we found that bat-pollinated species had extended flowering compared to species with bee and insect syndromes. This pattern is similar to what has been found in woody species in other tropical dry forests (Heithaus et al., 1975; Bullock, 1995; Frankie et al., 2004). An extended duration of flowering may be advantageous for disseminating the risk of unpredictable peaks of visits from pollinators (Primack, 1985). Gentry (1974) stated that species with extended flowering durations commonly produce a few flowers a day over long periods (steady-state flowering). This pattern is consistent with our observations for bat-pollinated species and has been documented in other tropical dry forests (Lobo et al., 2003; Quesada et al., 2003). In relation to the short flowering duration, we also documented highly synchronous flowering in herbs and vines. Simultaneous flowering in different species has been interpreted as a strategy to attract pollinators (van Schaik et al., 1993). In this case, it is possible that synchronous flowering facilitates the detection of flowers by pollinators. During the rainy season, the high canopy cover and elevated foliage density can limit the visibility of flowers in the understory strata, and simultaneous flowering can attract pollinators such as bees and birds, which are known to respond to showy flower displays (Gentry, 1974).

Influence of environmental factors on flowering time and duration

The effect of phylogeny was not important in the correlation between flowering time and environmental factors. As in other communities where strong environmental seasonality occurs (e.g., alpine, cerrado, meadow communities), flowering time is explained mainly by environmental variation (Silva et al., 2011; Lessard-Therrien et al., 2013). Specifically, it has been proposed that flowering time is a plastic trait that responds to various environmental cues (Davis et al., 2010). Flowering duration was related to environmental variation to a lesser extent and, in this correlation, we found a stronger effect of phylogeny. This result might be explained by considering that flowering duration is affected by a suite of abiotic and biotic factors such as life history traits, pollination type, and seed development strategy (Primack, 1985; Cara-Donna and Inouye, 2015).

Among tree species, 63% bloomed in the months previous to the rainy season, coinciding with an increase in daylength (Table 4). Increased daylength can break the dormancy of flower buds, and thus induce flowering episodes at the height of the dry season (Borchert and Rivera, 2001). This abiotic factor is an important trigger of flowering time in species that have the capacity to store water, especially in their trunks (e.g., Burseraceae, Cactaceae, Euphorbiaceae). The shrubs did not show a significant relation between flowering and rainfall. This result can be related to the staggered flowering behavior documented in the shrub species studied. Like other woody growth forms, shrubs can access subterranean water sources because of the length of their root systems (Sarmiento and Monasterio, 1983), and this characteristic confers some independence between water availability and reproductive activity (Bullock, 1995; Borchert et al., 2004; Seghieri et al., 2012). The flowering of lianas responded to declining daylength where flowering occurs after the cessation of shoot growth when daylength starts to decline (autumn flowering) (Borchert et al., 2004). This result is in contrast with previous findings in other tropical dry forests, in which flowering of lianas reaches its maximum at the end of the dry season (Putz and Windsor, 1987; Morellato and Leitao-Filho, 1996).

In other seasonal communities, the phenology of herbaceous species is tightly linked with temporal variation in precipitation (Bullock, 1995; Batalha and Martins, 2004; Joshi and Janarthanam, 2004). In our study site, three flowering behaviors in herbaceous species occurred during the rainy season. A group of perennial, fast-growing species (5), with water storing roots, presented flowers at the beginning of the rainy season (e.g., Amaryllidaceae, Nyctaginaceae). A second group of perennial herbs (25 spp.) that grew during the first months of rain produced flowers in the middle and toward the end of this season (e.g., Acanthaceae, Fabaceae, Nyctaginaceae). Finally, a third group composed of annual and perennial species (27) that germinate after the first rains, grow and produce flowers at the end of the rainy season and at the beginning of the dry season (e.g., Asteraceae, Convolvulaceae, Cucurbitaceae, Euphorbiaceae, Poaceae).

Our results reveal a complex relationship between strategies of resource use and growth forms, and the environmental signals related to a particular set of floral traits (e.g., tree species and bat syndrome). For instance, species with floral traits associated with moths are mainly lianas and respond to an increase in precipitation and a decrease in daylength, but their flowering also coincides with an increase in moth abundance in the tropical dry forest (Frankie et al., 2004). Species with bat syndrome are arboreal species (mainly Bombacaceae and Cactaceae) that respond to day-length increase during the dry season, but also this phenological behavior matches the presence of their pollinators. The coincident occurrence of flowering in bird, small bee, and wind syndromes during the rainy season is highly related to the phenology of herbaceous species. This pattern reveals a strong pressure linked with availability of water. Thus, seasonal variation in rainfall apparently may be not only the proximate, but also the ultimate cause of flowering in these species groups.

The phylogenetic influence was more relevant in the correlations of biotic attributes with flowering duration, and the abiotic factors were more important in the flowering time variation. Taking into account the phylogenetic non-independence of species, our results demonstrated that flowering time is linked with strategies of resource use associated with growth forms. Likewise, an important relationship was found between pollination syndromes and flowering duration. Therefore, functional attributes, environment variables, and phylogenetic relationships interact with each other to shape the diversity of flowering patterns. However, our study represents a short period of time, and it is uncertain whether our results would differ in a longer time series. Consequently, long-term studies are necessary to include other phases of the reproductive cycle such as fruiting and seed germination, analyzed from the functional and phylogenetic perspectives.

ACKNOWLEDGEMENTS

This study—done through the Posgrado en Ciencias Biológicas de la Universidad Nacional Autónoma de México (UNAM)—is part of the doctoral research of the first author, JC-F, and is a requisite for JC-F to obtain his Ph.D. JC-F thanks both UNAM for the education he received there, and the Consejo Nacional de Ciencia y Tecnología (CONACyT) for the scholarship he received for graduate studies. This research was supported by the Programa de Apoyo a Proyectos de Investigación e Innovación Tecnológica (PAPIIT) of UNAM, as a part of Project IN207512 (Fenología de especies arbóreas del bosque tropical caducifolio en la Depresión del Balsas, Michoacán). The authors acknowledge the valuable comments of the two reviewers, which helped to improve the manuscript. Finally, the authors thank Ma. Guadalupe Cornejo-Tenorio for her help in identifying the plant specimens, and Atzimba López Maldonado, Alberto Valencia García, and Heberto Ferreira Medina for their technical assistance.

APPENDIX 1

(A) Climatic diagrams constructed with monthly precipitation and temperature data of the last 20 years and (B) with data from the two years of the study period. Data were obtained from the meteorological station at Infiernillo, Michoacán, Mexico.

APPENDIX 2

Phylogenetic tree constructed with Phylocom software for species of a tropical dry forest of Mexico.

APPENDIX 3

Results of the generalized linear mixed effects models to evaluate the relationship between flowering phenology and abiotic factors in species with different pollination syndrome of a seasonal tropical dry forest in Mexico. Abbreviations: AIC (Akaike information criterion), R²m (marginal R²), R²c (conditional), SE (standard error), and σ² (residual variance of random effect). Coefficients significance: *P≤ 0.05, P≤ **0.001.

Growth form/ Pollination syndrome	Mean Angle	Mean Vector (r)	Rayleigh Test (z)
Trees
Bat	Mar 13 (68.41°)	0.36	3.08**
	51.20°	0.34	2.25
Bee	Mar 01 (56.20°)	0.30	11.67**
	Feb 23 (48.95°)	0.23	7.13**
Bird	May 20 (135°)	0.83	3.69**
	May 05 (120°)	0.91	3.48**
Small insects	139.62°	0.18	1.02
	Apr 29 (114.67°)	0.33	3.63**
Moth	251.09°	0.38	1.44
	233.79°	0.42	1.94
Wind	30°	0.25	0.25
	69.89°	0.58	1.69
Shrubs
Bee	175.20°	0.18	0.80
	196.28°	0.30	2.54
Small insects	352.58°	0.48	2.73
	Dec 05 (335.66°)	0.47	3.12*
Moth	255°	0.41	1.4
	Aug 27 (232.91°)	0.57	2.93*
Woody vines
Bee	Oct 05 (270°)	0.82	6.19**
	Sep 22 (257°)	0.75	5.14**
Bird	30°	0.52	2.48
	Jan 22 (17.37°)	0.56	3.55**
Small insects	246.20°	0.51	2.37
	262.33°	0.48	1.90
Moth	Nov 03 (299.28°)	0.54	3.23**
	Oct 17 (282.62°)	0.62	3.90**
Wind	255°	0.96	1.86
	270°	1	1
Herbaceous vines
Bee	Oct 02 (268.39°)	0.93	16.67**
	Sep 27 (262.03°)	0.94	16.99**
Bird	Sep 27 (262.63°)	0.97	3.79**
	Sep 29 (264.89°)	0.93	5.26**
Herbs
Bee	Sep 22 (257.08°)	0.88	42.66**
	Oct 01 (265.84°)	0.87	52.58**
Bird	265°	0.94	1.44
	285°	0.96	1.86
Butterfly	Sep 11 (246.20°)	0.77	3.56**
	273°	0.66	2.23
Small insects	Sep 13 (248.10°)	0.91	7.19**
	Sep 16 (251.09°)	0.87	8.99**
Moth	Sep 08 (243.76°)	0.95	7.24**
	Sep 12 (247.91°)	0.89	5.65**
Wind	Sep 20 (255°)	0.96	5.59**
	Sep 13 (248.79°)	0.93	6.11**

• Note: Data from the first and second year are shown in the first and second line, respectively. The significance level of z corresponds to: *P ≤ 0.05; **P ≤ 0.001.

APPENDIX 4

Results of the circular statistics analyses to characterize flowering phenology in species having different pollination syndromes within growth forms in a seasonally dry tropical forest.

Variables	Coefficients	SE	z	AIC	R^2m	R^2c	σ^2
Bat
Fixed effects				82.5	0.23	0.32
Rainfall	−0.007	0.004	−1.41
Daylength	−0.04	0.23	0.19
Random effects
Month							0.05
Bird				84.6	0.04	0.31
Fixed effects
Rainfall	0.12	0.17	0.73
Daylength	−0.21	0.29	−0.73
Random effects
Month							0.18
Moth				102	0.15	0.48
Fixed effects
Rainfall	0.28	0.16	1.75
Daylength	−0.23	0.19	−1.15
Random effects
Month							0.22
Large bee				150.4	0.09	0.75
Fixed effects
Rainfall	0.01	0.10	0.17
Daylength	−0.24	0.21	−1.13
Random effects
Month							0.23
Small bee
Fixed effects				81.3	0.06	0.73
Rainfall	0.25	0.206	1.21
Daylength	0.4	0.79	0.5
Random effects
Month							2.44
Small insects				120.8	0.01	0.49
Fixed effects
Rainfall	0.06	0.12	0.50
Daylength	−0.07	0.20	0.34
Random effects
Month							0.16
Wind				70.2	0.09	0.32
Fixed effects
Rainfall	0.33	0.18	1.76
Daylength	−0.19	0.37	−0.51
Random effects
Month							0.24