Phenology, abundance and efficiency of pollinators drive the reproductive success of Sarcoglottis acaulis (Orchidaceae) at the Atlantic Forest

,


Introduction
Biotic and abiotic processes are central players in plant phenology (Lieth 1974;Morellato & Leitão Filho 1990;Wolf et al. 2017).In seasonal Rain Forests, phenology is limited by physiological stresses caused by variation in temperature and humidity.However, where climatic fluctuations are less pronounced, selective biotic pressures such as herbivory, predation, competition, pollination and dispersal, may influence phenological responses in plants (Aide 1988).It is known that global climate change interferes with the flowering patterns of plant species (Lesica & Kittelson 2010), and consequently the relationship with several pollinator groups might also be affected (Kharouba & Vellend 2015;Hutchings et al. 2018).
The Atlantic Forest biome is one of the thirty-six world hotspots due to its high biodiversity, high endemism rate, and advanced degree of deforestation, with current vegetation cover of 28 % (Rezende et al. 2018).In Atlantic Rain Forests of southeastern Brazil, most Orchidaceae species are known to flower during rainy months (Pansarin & Pansarin 2008).However, there is also evidence that flowering of many orchids is related to seasonality of pollinators such as euglossine bees (Ackerman 1983).Therefore, biotic and abiotic factors may contribute to the population and community dynamics of orchids.
With approximately 26,460 described species (Christenhusz et al. 2017), the family Orchidaceae is one of the most species-diverse among angiosperms, with a high rate of endemism (Govaerts 2003).In Brazil, 2,760 species from 250 genera are currently known (Flora do Brasil 2020Brasil 2021)), mostly from ombrophilous rainforests, including those from the Atlantic Forest biome, which harbors about half of all Brazilian orchid diversity (Barros et al. 2009).Orchidaceae presents a vast array of pollination strategy and may attract pollinators by offering different floral resources, such as nectar, oil, resin and fragrances, or by deceit (van der Pijl & Dodson 1969;Ackerman 1986;Nilsson 1992;Jersákova et al. 2006).
The present study aimed to record the reproductive and vegetative phenology of Sarcoglottis acaulis and to monitor the seasonality of its pollinators during the flowering period in three fragments of the Atlantic Forest in the Northeast Brazil, two of them inserted in an urban and very anthropized area.Specifically, we investigated if there is difference in the pollinator frequency, in the natural fruit set and in the pollinia removal among fragments.
The first two fragments are located within the urban perimeter of João Pessoa, and have typical seasonal semideciduous forest vegetation.These fragments are 1.5 km distant from each other and located southwest of João Pessoa's urban area (7º 6' Lat.Sul/ 34º 52' Long.Oeste), Paraíba state coast, at an average altitude of 45 m a.s.l (Barbosa 1996).Climate is typically hot and wet with mean annual temperature near 25 ºC and precipitation rates varying between 1500 to 1700 mm.Peak rainfall occurs between March and August and the dry season between October and December (Lima & Heckendorff 1985).The JBBM fragment has about 417 hectares.The DSE fragment is an Atlantic Forest remnant with 7 hectares, located at the Centro de Ciências Exatas e da Natureza of the UFPB, Campus I, João Pessoa, Paraíba (7º 8' S/ 34º 50' O).
The third fragment, Rebio Guaribas, is located at Mamanguape city, northern coastline of Paraíba state, 70 km from the state capital João Pessoa.The reserve has about 3,378 hectares, with vegetation consisting of a mosaic of Cerrado and Atlantic Forest remnants.Mean annual temperature varies between 24 and 26 ºC and total annual precipitation rates vary between 1,750 and 2,000 mm, with the rainiest months between April and June, and the driest months between October and December (Brasil 2003).

Phenology
The phenology of S. acaulis was monitored at the DSE and JBBM using two different approaches, one for observation of phenology phases (flowering longevity) of the individuals, and the other for recording the phenology in population level.In the DSE fragment, we marked 49 individuals, and monitored them in intervals of 2 -4 days during one flowering season in order to record the stages ranging from the emission of the scape, during the first flower bud stages, until seed release by ripe fruits (Fig. 1).
In the JBBM fragment, we delimited a 0.8 hectare area and subdivided it into smaller 10 x 20 m sections in order to monitor phenophases in population level and to evaluate  and  2012).Phenology monitoring included the assessment of the following reproductive phenophases: floral bud, opened flower, unripe and ripe fruits.The number of individuals in each phenophase was counted.Onset of leaf dormancy was determined by the disappearance of the vegetative structure, whereas the end was determined by its regrowth.
In some cases, the presence of a pseudobulb was detected on plants superficially connected to the substrate.Therefore, vegetative dormancy and presence/absence of pseudobulbs were also evaluated.Individuals that did not grow for four months or more were categorized as dead.Phenology data were correlated to precipitation and the mean temperature values recorded for the study period, and to historical data available for the study areas.Data follow non-normal distribution according to the Shapiro-Wilk test (Zar 1999), and the nonparametric Spearman's rank correlations were performed using R software version 3.3.1 (R Development Core Team 2013).Circular statistics were used to estimate the intensity, frequency and seasonality of each phenophase by converting months into angles (i.e.January: 0 o and December: 330º, with 30º intervals between months).This analysis generates the mean angles and standard deviations for the frequency of individuals on different phenophases, the mean months, the r vector value, and the significance level using Rayleigh's test.Tests were performed with the R software version 3.3.1.Precipitation data were provided by the Agência Executiva de Gestão das Águas do Estado da Paraíba (Paraíba 2020).

Comparison of the natural fruit set and pollinia removal among fragments
At the three study sites, flowers (N=333, 111 in Rebio, 69 in JBBM and 153 in DSE) of S. acaulis were marked and exposed to pollinators to assess natural fruit set from the different populations.Pollen removal was also quantified.To quantify pollinia removal in different periods of the day, 24 individuals per area were monitored for three days.On each day, observations were conducted early in the morning (05:30 am) and late in the afternoon (5 pm).Following the exposure period, total numbers of removed and nonremoved pollinia were compared among areas.
To test whether 1) natural fruit-set and 2) pollinia removal differed among the fragments, nonparametric Kruskal-Wallis tests were performed, since data were not normally distributed.The Anderson-Darling normality test was applied for residuals and a post hoc test of Dunn for comparison by groups.The tests were performed using the R software version 3.6.1.

Flower availability and pollinator frequency
To evaluate an eventual relationship between flowering intensity of S. acaulis and pollinator abundance, the frequency of bees and available plants were monitored at the three fragments.The Euglossine species, Eulaema atleticana and E. niveofasciata, were monitored.These bees had been chosen because they were previously registered as pollinators of S. acaulis (see Albuquerque et al. 2021).In the DSE and JBBM, phenology of plants and bees were observed throughout the year (from July 2016 to June 2017).In the Rebio Guaribas, observations were restricted to the plant reproductive period (from July to October 2016).During the initial and final flowering months, June and October, respectively, frequency of bees was assessed monthly, and during peak flowering months, August and September, biweekly evaluations were made.
To monitor phenology, 50 plants per fragment were marked and the number of open flowers was recorded on the same days in which bees were collected.To attract bees, we used four synthetic compounds, i.e., geraniol (97 %, Sigma-Aldrich), eucalyptol (99 %, Sigma-Aldrich), benzyl acetate (99 %, Sigma-Aldrich) and methyl salicylate (99 %, Sigma-Aldrich), which are known as potent attractants of euglossine bees (Bezerra & Martins 2001;Martins & Souza 2005).Filter papers were impregnated with the synthetic compounds individually and ranged at 1.5 m above the ground on different trees that were 3 m apart from each other.Bees were collected using an entomological net, cooled within an ice box, marked with non-toxic pens of colors corresponding to site of origin (Edding) and subsequently released.The bees with pollinia attached to the body were captured.Bees were collected on three consecutive days at the three areas, between 07:00 am and 2:00 pm.
Pollinator frequency and flower abundance were compared between sites using the Kruskal-Wallis test, as the data did not show normal distribution.The Anderson-Darling normality test was used for residuals and a post hoc test of Dunn for comparisons by groups.These tests were performed using the R software version 3.6.1 (http://www.rproject.org/).To test the correlation between the frequency of bees and the flowering frequency throughout the year, a Spearman Correlation was performed (no normal data) in the software R version 3.5.1 (http://www.r-project.org/).set per inflorescence was 3.78 ± 1.3 and 3.18 ± 1.39, respectively.Flowers lasted on average 5.95 ± 2.34 days (N=189) (Fig. 2).The interval ranging from the beginning of fruit development to fruit maturation and seed dispersion lasted 11.76 ± 3.36 days (Fig. 2).Sarcoglottis acaulis flowered between August and October, and flowering peaks differed between the two study years at JBBM.In 2011, flowering in the populations peaked in September (31 individuals; 23.48 %), whereas in 2012 it peaked in July and August (23 individuals; 17.4 %).Although a welldefined reproductive period was observed, phenophases were not concentrated on a particular period of the year at the population level (The Rayleigh test was not significant; Tab. 1).Of the 132 individuals monitored at the JBBM, 70 (53 %) flowered in 2011 and 27 (20.4%) in 2012.There was variation in precipitation observed during the two years, with higher rates in 2011 (about 2,355 mm) than in 2012 (1,651 mm).Reproductive period of S. acaulis correlated positively with local precipitation, and specific responses to water availability were observed for each reproductive phenophase: floral bud emission was positively correlated with precipitation with a lag time of one and two months (r = 0.61, p = 0.001 and r = 0.59, p = 0.001, respectively).Flowering, on the other hand, was correlated with precipitation with a lag time of two and three months (r = 0.55, p = 0.003 and r = 0.55, p = 0.003, respectively).During the first year of monitoring, both flowering and fruiting peaked in September.

Phenology
Fruiting frequency of S. acaulis varied slightly between years.In 2011, 14 (10.6 %) individuals set fruits, whereas 16 (12.1 %) individuals set fruits in the second year (2012).However, the period of fruit development varied between years.In 2011, fruiting began and ended in September, whereas in 2012, fruiting began in September but extended until October.Ripe fruits were correlated with precipitation of three months (r = 0.41, p = 0.03), whereas ripe fruit production correlated with precipitation of two and three months (r = 0.44, p = 0.02 and r = 0.40, p = 0.04, respectively).
Vegetative dormancy (leaf fall) was observed for 31 (23.5 %) of the individuals of S. acaulis monitored in 2011, peaking in December when 29 leafless individuals (22 %) were recorded.In 2012, the frequency was 69.7 %, and, again, December showed the highest proportion of leafless individuals (64.4 %).January 2012 and 2013 showed the highest rates of vegetative budding.Dormancy negatively Re-budding negatively correlated with precipitation and a lag time of two and three months was observed (r = -0.47,p = 0.001 and r = -0.55,p = 0.003, respectively).

Comparison of the natural fruit set and pollinia removal among fragments
Natural fruit-set varied significantly among the three fragments (H = 10; df = 2; p = 0.006).Post hoc tests showed that the difference was due to higher fruit set observed at the DSE fragment in relation to the Guaribas Rebio (H = 9.27; df = 1; p = 0.002).Fruit set at JBBM was similar to the two other study sites (Tab.2).Pollinia were always removed during the day.Removal rate differed significantly among the study sites (H = 6.41; df = 2; p = 0.04).Pollinia removal at the Rebio Guaribas was lower than those of both the JBBM (H= 2.021; p = 0.02) and the DSE fragments (H= 2.255; p = 0.01).Removal in the last two fragments was similar to each other (Tab.3).
In the scent baits, we collected a total of 115 individuals of two species of Eulaema, of which 14 were sampled at Guaribas (5 E. atleticana and 9 E. niveofasciata), 55 at DSE (37 E. atleticana and 18 E. niveofasciata), and 46 in JBBM (34 E. atleticana and 12 E. niveofasciata).One bee captured and marked at the DSE was later re-captured at the JBBM (Fig. 4).
At the three fragments, E. atleticana ( 14) and E. niveofasciata (2) were observed carrying pollinia of S. acaulis, which were attached to the ventral portion of their labrum.At the Rebio Guaribas, an E. niveofasciata individual carrying pollinia of an unidentified orchid species was also observed at the same flowering time as Sarcoglottis acaulis (Fig. 4).

Discussion
The development from buds to fruit dispersion of Sarcoglottis acaulis is completed in about one month.The short interval observed is very similar to Cyclopogon diversifolius, another species belonging to the Pelexia alliance (Singer & Cocucci 1999).Our data show a steady state flowering pattern to S. acaulis, according to the classification of Gentry (1974), as the plants produce a few flowers a day over an extended period of time.This pattern is common in flowers pollinated by trap-liners euglossine bees (Janzen 1971).According to the classification of Newstrom et al. (1994), the flowering pattern is annual, similar to those of several orchids occurring at the Atlantic Forest, which also exhibits one major flowering cycle per year (Pansarin & Pansarin 2008).The peak flowering month in the present study differed from other populations of S. acaulis recorded in other geographic regions, such as in southeastern Brazil, where flowering peaked in June (Cunha & Forzza 2007), corresponding to the dry season in the southeastern region, while in the northeastern occurred in September.This divergence may be related to the difference in the dry month periods in the regions of Brazil.This suggests that the precipitation of the previous months triggers the flowering of S. acaulis, since they bloom in the first dry month of each region.
High rain volumes during the flowering period result in high flower production, but not necessarily in high pollination rates, since euglossine bees are more abundant during the driest periods of the year (Farias et al. 2008).This further explains why similar fruiting rates were observed between years despite different precipitation regimes.Also, fruiting during the rainy months, due to S. acaulis's fast fruit maturation feature, seems to be disadvantageous, since seeds are dispersed by the wind.On the other hand, prolonged dry periods induce plants to undergo dormancy.Several terrestrial orchids have been previously observed to flower during the rainy period (Sahagun-Godinez 1996;Pansarin & Pansarin 2008), probably to increase energetic efficiency and to avoid the harsher and limited conditions of the dry period (Sahagun-Godinez 1996).Sarcoglottis acaulis, therefore, needs water to flower, to fruit and to maintain its leaves, but flowering needs to be extended to the less rainy months when pollinator frequency is higher to guarantee outbreeding.
The fragmentation of the Atlantic Forest reduces area and quality of habitat for euglossine bees (Nemésio 2013), which, although able to fly long distances in continuous forest areas (Wikelski et al. 2010), are negatively affected by habitat fragmentation (Nemésio & Silveira 2006).How vulnerable an organism is to fragmentation depends on its life history and ecological properties (Davies et al. 2004).For example, some euglossine species, such as Eulaema niveofasciata and E. atleticana, remain confined to forest fragments and, therefore, their foraging areas are limited to the frontiers of fragments (Milet-Pinheiro & Schlindwein 2005).This further increases reproductive isolation between plants located on different fragments due to a high dependency on bees for pollination.Alternatively, species such as E. nigrita Lepeletier, 1841 are more abundant in fragmented areas and, in fact, are frequently used as indicators of disturbance (Peruquetti et al. 1999;Nemésio & Silveira 2006).Therefore, fragment size and minimum distance between adjacent fragments need to be evaluated in order to guarantee the survival of bees and the plants that depend on them for pollination (Powell & Powell 1987).
Concerning fragmentation and pollinator dynamics, a lowest frequency of the main pollinator, E. atleticana, was observed at the largest fragment, i.e. the Rebio Guaribas.This might explain the lower rates of both pollinia removal and fruit set that were observed at this area.At the Rebio Guaribas, richness of orchids and other nectar-producing plants is higher than those of the other two areas (Araújo et al. 2009;Barbosa et al. 2011), suggesting the existence of a higher competition for pollinators.This idea may be supported by the fact that bees carrying pollinia from other species were also collected in this fragment in the same time the flowering of S. acaulis.
In the DSE fragment, although it is the smallest one, we observed at the same period, a high frequency of visits of Eulaema atleticana to S. acaulis flowers (Albuquerque et al. 2021), high fruit set and high pollinia removal rate.Bee abundance at this fragment may be a consequence of emigration from the Jardim Botânico, the Mata do DSE and other forest fragments within the Campus, which disperse to fulfill biological needs such as nidification sites and additional food sources on larger areas.Euglossine bee species fly over long distances (Wikelski et al. 2010) and similar species composition between forest fragments that are close (distance of 2 km) have been previously recorded (Ramalho et al. 2009).
In summary, our study highlights the importance of precipitation to flowering seasonality of S. acaulis and the dependency of this orchid on euglossine bees for pollination, considering that areas with higher abundance of bees showed higher pollinia removal and fruiting rates.The low rate of self-pollination of S. acaulis, as shown by our own data from the reproductive system of the same studied population (Albuquerque et al. 2021), reinforces the importance of pollinators for its reproduction.Furthermore, our data suggest that proximity of fragments is an important factor for conservation since even at small fragments, pollinators were abundant perhaps due to emigration from the larger, adjacent fragments.The lower frequency of bees from the scent baits in the largest fragment, that is, the Rebio Guaribas, which may have resulted in less pollinia removal and fruiting rates, is noteworthy and seems somewhat conflicting.Whether low abundance of bees was specific for the monitored period or, in fact, describe a particular feature of the conservation unit, is unknown.Future investigations should evaluate the causes of the low frequency of bees in this area.

Figure 1 .
Figure 1.Flower and fruit longevity in Sarcoglottis acaulis. A. Open flowers, B. Buds and flowers, C. Unripe fruits and D. Ripe fruits.
During population monitoring (N=49 individuals) at the DSE fragment in 2011 we recorded a total of 49 inflorescences and 189 flowers in an area of 11,000 m 2 .Mean number of open flowers per day was 3.05 ± 1.17 per inflorescence.The mean number of flowers and fruit Phenology, abundance and efficiency of pollinators drive the reproductive success of Sarcoglottis acaulis (Orchidaceae) at the Atlantic Forest Acta Botanica Brasilica, 2022, 36: e2021abb0121

Figure 2 .
Figure 2. Number of buds, flowers, unripe and ripe fruits produced during the longevity of 49 inflorescences in the fragment of Departamento de Sistemática e Ecologia, Atlantic Forest, Paraíba, NE-Brazil.

Figure 4 .
Figure 4. Floral visitors: A. Eulaema atleticana carrying a pollinia of Sarcoglottis acaulis; B. Eulaema niveofasciata carrying a pollinia of an unidentified species and C. Eulaema atleticana captured and marked at the Departamento de Sistemática e Ecologia and recaptured at the Jardim Botânico fragment, Paraiba, NE -Brazil.

Table 1 .
Circular statistics results for reproductive and vegetative phenophases and mean month for Sarcoglottis acaulis (Orchidaceae) in an Atlantic Forest area from northeastern Brazil.n.s.= not significant (p > 0.05).

Table 3 .
Rates of pollinia removal in three populations of