Introduction

Excess flower production and low fruit-to-flower ratios are a common phenomenon in hermaphrodite plants (Sutherland 1986). Holtsford (1985) categorized the ultimate factors of this phenomenon according to five hypotheses:

  1. 1.

    The male function hypothesis: excess flowers have a function as pollen donors.

  2. 2.

    The pollinator attractiveness hypothesis: excess flowers are useful in attracting pollinators.

  3. 3.

    The reproductive assurance hypothesis: seed production is assured when a part of the flower is damaged.

  4. 4.

    The resource boom hypothesis: excess flowers can fruit only in a year when conditions are good.

  5. 5.

    The selective fruiting hypothesis: many flowers enable selective bearing of good fruits.

The wider choice hypothesis and the bet hedging hypothesis have also been proposed (Burd 1998; Parra-Tabla and Bullock 1998; Melser and Klinkhamer 2001; Burd 2004). However, these hypotheses can be explained by any one of the hypotheses of Holtsford or by various combinations of them. Therefore no new additional hypothesis has been accepted since Holtsford (1985) published his paper.

In some tropical pioneer plant species the abortion rate is high not only in young ovaries just after flowering but also in flower buds and developed ovaries (our personal observations and the results of this study). These high abortion rates cannot be explained by the hypotheses on male function, pollinator attractiveness or selective fruiting.

The assumption underlying the resource boom hypothesis is that resources available for fruit production cannot be predicted at the flower stage, and that fruit set is high only when the resources for fruit production are sufficient. The resource boom hypothesis has been tested by controlled watering (Coulter 1979; Lee and Bazzaz 1982; Holtsford 1985; Shi et al. 2005), fertilization (Willson and Price 1980; Eriksson 1987; Gorchov 1988; Vaughton 1991; Nishikawa 1998; Parra-Tabla and Bullock 1998; Shi et al. 2005), thinning of reproductive organs (Lee and Bazzaz 1982), leaf cutting (Willson and Price 1980; Gorchov 1988), shading (Willson and Price 1980) and bark removal (Gorchov 1988). These manipulated experiments explain the proximate factor for abortion of reproductive organs, but they cannot explain the ultimate factor for excess production of reproductive organs. It is not surprising that the unnatural stresses placed on plants in the previous studies might have caused a high abortion rate. The ultimate factors that provide against the risk of environmental fluctuations such as in the resource boom hypothesis and the reproductive assurance hypothesis cannot be tested unless actual environmental fluctuations are measured.

To compensate for the shortcomings of the previous studies, we took a different approach. We measured fluctuations in the demand and supply of resources under natural conditions, and we investigated how the abortion of reproductive organs changed in response to the amount of resources.

Previous studies on the resource boom hypothesis dealt with the unpredictability of supra-annual environmental fluctuations (Coulter 1979; Willson and Price 1980; Lee and Bazzaz 1982; Holtsford 1985; Eriksson 1987; Gorchov 1988; Vaughton 1991; Nishikawa 1998; Parra-Tabla and Bullock 1998). However, in tropical regions some pioneer plants continuously reproduce (Momose et al. 1998a; Sakai et al. 1999), and they are considered to be affected by environmental fluctuations over much shorter time scales, namely, daily rather than supra-annual fluctuations.

We aim to test a hypothesis that the abortion of reproductive organs is an adaptation to environmental fluctuations over shorter time scales than previously reported in other plants.

Materials and methods

Study site and plant materials

The study was conducted in Lambir Hills National Park (4°12′N, 114°02′E, ca. 100 m altitude), Sarawak, Malaysia. The mean annual temperature is around 27°C, and the average annual rainfall is 2,740 mm (Kumagai et al. 2005). The climate is almost aseasonal (Kato et al. 1995).

The studied plants, Melastoma malabathricum L., are pioneer shrubs that are abundant at the wayside (Corner 1988). These plants reproduce almost continuously throughout the year, and reproductive organs at various stages, including young flower buds and mature fruits, are almost always found on the same individuals. All studied individuals were exposed to sunlight. Plant height ranged from approximately 1.2 to 2.3 m.

Inflorescences are terminal cymes of three to seven reproductive organs (Corner 1988). Of these, one flower (or in rare cases a few flowers) blooms at a time. Flowers are hermaphroditic, bisymmetrical and usually fivemerous. Petals are 25–35 mm long. Ovaries are inferior. The flowers open in the daytime for 1 day, and are pollinated by carpenter bees (Momose et al. 1998b). Fruits are capsules and urceolate-globular, 5–12 mm long. When fruits mature, they dehisce horizontally and purple pulp with hundreds of tiny seeds is exposed.

Basic data of resource economics for reproduction

For the purpose of testing the resource boom hypothesis and the male function hypothesis, we calculated the resource economics for reproduction using the following procedures (Fig. 1).

Fig. 1
figure 1

A flowchart showing the analysis of resource economics for reproduction. Eight measurements (rounded boxes) were conducted, and on the basis of the resulting data, ten edited sets of data (thick-line boxes) were obtained for analysis of the resource economics for reproduction. Two edited data items (Growing curve and resource allocation in each sub-organ and Efficiency of investment), were used for testing the male function hypothesis. The other edited data were used for testing the resource boom hypothesis, and Spearman’s rank correlation test was conducted for pairs of boxes connected by double-headed arrows. PPFD Photosynthetic photon flux density

Full size image

Size and dry weight

To continuously estimate the dry weight of the reproductive organs attached to plants, we determined the relationships between the size and weight. On 24–25 July 2004, ninety-four flower buds (nine individuals), 21 flowers (three individuals) and 71 fruits (seven individuals) were harvested, and their weight and dry weight were measured, as well as the length and width of the calyxes covering the ovaries.

Using harvesting samples, we made a multiple regression analysis of the relationships between log-transformed dry weight (the explained variable) and log-transformed length and width (independent variables). We also made a multiple regression analysis of the relationship between the water weight, the amount of water contained in the reproductive organs and size (length and width).

Growth of reproductive organs

From 5 June until 24 July 2004, we measured the lengths and widths of marked reproductive organs every day without damaging the organs. We used all measurable samples within our site. In total, the growth of 90 samples in six individual plants was monitored. The flower stage was included in only 41 samples, because the other samples aborted before the flower stage, or their monitoring started after the flower stage.

The growth in size of reproductive organs was transformed to a dry weight basis using the result of multiple regression between size and dry weight. Using all 90 samples, we obtained the abortion rate of reproductive organs on the next day. Samples including the flower stage were arranged according to the day of this stage, and dry weights [m(t)], were estimated for each growing day (t), i.e., each day from the earliest stage of flower bud (18 days before the flower stage). Similarly, water weight [w(t)] was calculated.

Monitoring the number of reproductive organs of plant individuals

We used all measurable plant individuals, that is, nine individuals, within our site, and monitored fluctuations of reproductive organs in each individual on each day from 5 June until 18 July 2004. The number of reproductive organs was counted in each reproductive stage every day. Seven stages of development were defined for the reproductive organs according to their size and color: young bud (length < 8 mm), intermediate bud (length ≥ 8 mm), large bud (1 day before the flower stage), flower (anthesis), young fruit (tops of ovaries are green), intermediate fruit (tops of ovaries are reddish, the ratio of width to length < 1) and large fruit (the ratio of width to length ≥ 1). The size of these individual plants was defined by trunk diameter at ground level.

Respiration and transpiration of reproductive organs

In the following measurements, the reproductive stages were distinguished on the basis of their dry weight and appearance: young bud (≤40 mg), intermediate bud (≥50 mg, except large buds), large bud (buds 1 day before the flower stage: petals have grown nearly as much as at anthesis), flower (anthesis), young fruit (≤50 mg), intermediate fruit (60–90 mg) and large fruit (≥100 mg). These definitions mostly agreed with the previous definitions according to sizes and colors.

The daily changes in respiration and transpiration were measured every hour in the daytime and every 3 h at night using a portable photosynthesis meter LI-6400 (LI-COR, Lincoln, Neb.) with a 6400–05 conifer chamber. Temperature was recorded at the same time. CO2 concentration was fixed at 360 μmol CO2mol−1. We used all measurable samples within our site, which were exposed to sunlight. Plant height ranged from approximately 1.2 to 2.3 m. Under natural light, each sample was measured for only 1 day on 2, 5, 7, 8, and10 May 2004, and then their dry weights were determined. Measurements were not made during rainfall. In total, 56 samples were measured from nine individual plants.

Photosynthesis of leaves

The daily changes in photosynthesis of leaves were measured on 13, 14, 16 and 19 May 2004 using the LI-6400 meter under natural light conditions at a fixed CO2 content (360 μmol CO2mol−1) (Kenzo et al. 2003). In our measurable range, 20 leaves were randomly selected from eight individual plants, and some selected leaves were not the same in which the respiration and transpiration of reproductive organs were measured. After the measurement, the area of these leaves was measured, and the area per single leaf was calculated.

Meteorological data

Solar radiation was measured every 10 min using a pyranometer (MS-801; EKO, Tokyo) at a fixed open site. The plants under study were considered to receive approximately the same amount of solar radiation as the sensor because they were located at the roadside. Rainfall was measured every hour using a tipping-bucket 0.5-mm tip rain gauge (no.34-T; OTA Keiki, Tokyo).

Dry mass in sub-organs

The dry mass of 13 buds (eight individuals), ten flowers (four individuals) and 13 fruits (nine individuals) was measured. To evaluate the male function hypothesis, buds and flowers were separated into petals, stamens, pistils and calyxes. Fruits were separated into ovaries and calyxes. The dry mass of their sub-organs was measured.

Analysis for the resource boom hypothesis

Based on the assumption that carbohydrate and water were demanded by reproductive organs and fluctuated with supply in the short-term, we analyzed the responses of plants to such fluctuations.

Respiration and transpiration

Because the respiration rate is generally dependent on temperature (Amthor 1989; Tjoelker et al. 2001), we conducted a simple regression analysis within each stage of development of the reproductive organs between log-transformed respiration rate per unit dry weight (r) and temperature (T) to obtain the constants r 0 and k in the following equation (Sutcliffe 1977; Meir and Grace 2002):

$$ \ln r = \ln r_{0} + kT . $$
(1)

If the coefficient k was not significantly larger than zero, we considered that the respiration rate was unaffected by temperature, and the average of measurements within the same reproductive stage was obtained. If the log-transformed respiration rate was not significantly different among two or more stages when temperature was the covariant in an analysis of covariance, the two or more stages were combined. Daily changes in temperature were calculated every 1–3 h from temperature measured at the same time as respiration and transpiration, and each mean temperature was obtained. Substituting the mean temperatures into Eq. 1 then multiplying by hour(s), and summing the estimated amount of respiration for a whole day, the total respiration per unit dry weight per day (R) for each reproductive stage was estimated. Because the transpiration rate of reproductive organs was related to the time of day but not to temperature, transpiration rate measurements were averaged every 2 or 3 h for each stage and multiplied by hours, to obtain a sum of the estimated amount of transpiration for a whole day. The total transpiration per unit dry weight per day (T r) for each reproductive stage was then calculated.

Daily demand for resources

Carbohydrate is translocated to reproductive organs, and it is used to produce essential energy for their growth and maintenance (Mooney 1972). Demand for carbohydrate [c(t)] was expressed as:

$$ c(t) = m(t) - m(t - 1) + uRm (t), $$
(2)

where m(t) − m(t − 1) is the increase in dry weight (g), Rm(t) is the respiration per reproductive organ per day (mol CO2day−1), and u is the constant used to convert CO2 (mol) into consumed carbohydrate, C6H12O6 (g), u = 30 g mol−1. The increase in dry weight representing a demand for carbohydrate was assumed to be all carbohydrate. Although for the day of the flower stage (t = 19), m(t) was measured in whole flowers, for 1 day after the flower stage (t = 20), m(t − 1) was measured from calyxes and pistils only (petals and stamens were not included), and the increase in dry weight was calculated.

Similarly, demand for water [v(t)] was expressed as:

$$ v(t) = w(t) - w(t - 1) + T_{r}m(t), $$
(3)

where w(t) − w(t − 1) is the increase in water weight (g), and T r m(t) is the transpiration per reproductive organ per day (g H2O day−1).

The periods of each stage were determined according to their reproductive stage defined by their dry weight and appearance. The average demands for carbohydrate and water in each stage were obtained. The number of reproductive organs counted in each reproductive stage on every day was multiplied by the demand for carbohydrate and water in each stage, and then the demand for carbohydrate and water in nine individual plants on each day was obtained.

Resource supply

The resource supply was determined from photosynthetic measurements and meteorological data.

The indexes, a, b, and R p, representing the relations between photosynthetic photon flux densities (PPFD) and photosynthetic rate as given by the following equation (Boote and Loomis 1991) were estimated:

$$ P_{{\text{g}}} = P + R_{{\text{p}}} = a{\text{ }}{\left[ {1 - \exp ( - ba^{{ - 1}} I)} \right]}, $$
(4)

where P g is the gross photosynthetic rate, P is the net photosynthetic rate, R p is the respiration rate (μmol m−2 s−1) and I is PPFD. The value of R p was obtained by averaging the measurements made at night. The value of b (initial slope) was given by the incline of the regression between P g and I when I ≤ 50 μmol m−2 s−1. After 1000 hours, stomatal conductance decreased and the photosynthetic rate under full light (I ≥ 600 μmol m−2 s−1) often reduced. Thus, the values of maximum gross photosynthetic rate (ba −1) were calculated separately before and after 1000 hours by averaging the measurements when I ≥ 600 μmol m−2s−1.

The relationship between total radiation and PPFD at our study site was provided by M. Yamashita (personal communication) using the following equation:

$${\text{PPFD }}(\mu {\text{mol}}\,{\text{m}}^{{ - 2}} {\text{s}}^{{ - 1}} ) = 2.2 \times {\text{radiation }}({\text{Wm}}^{{-2}}).$$
(5)

The radiation measured every 10 min on each day from 31 May to 17 July 2004 was transformed into PPFD. The PPFD value was substituted into Eq. 4, using different values of maximum photosynthetic rate according to the time of day, and the daily total photosynthetic productivity (g cm−2 day−1) was estimated for each day.

From the radiation measured every 10 min and the rainfall measured every hour, the daily total radiation and the daily total rainfall were calculated on each day.

Responses of plants

The amount of newly expanding leaves and the amount of fallen leaves per day are balanced according to the monitoring of marked twigs of M. malabathricum trees reaching the reproductive size (T. Hiromi, unpublished data). This means that the total amount of leaves on trees is regarded as constant. Thus, the daily fluctuation of carbohydrate supply is affected only by the daily fluctuation of productivity per leaf area.

Daily fluctuation patterns were compared between demand for a resource and supply of a resource, and between demand per supply of resource and abortion rate. Each day’s demand for carbohydrate (each individual and all individuals combined) was compared with the total photosynthetic productivity during the previous 1, 3 or 5 days by Spearman’s rank correlation test. Similarly, demand for water (each individual and all individuals combined) on the present day was compared with the total rainfall for the previous 5 days and the accumulated radiation for the previous 5 days. The daily abortion rate of the population on the next day was compared with the fraction of the demand for carbohydrate on the present day per the total photosynthetic productivity during the previous 1, 3 or 5 days, and with the fraction of the demand for water on the present day per the total rainfall during the previous 5 days, by Spearman’s rank correlation test.

Analysis for the male function hypothesis

In samples for which the flower stage was not monitored, the day of growth (t) was estimated from the dry weight [m(t)]. Using all 90 samples, we obtained the abortion rate from the previous day [d(t)] on each day of growth. The survival curve [L(t)] of reproductive organs starting from 18 days before the flower stage was obtained as follows:

$$ L{\text{(}}t{\text{)}} = {\prod\limits_{t = 1}^{45} {{d\text{(}}t - 1{\text{)}}} }, $$
(6)

where 45 is the number of days until the fruits mature (dehiscence: seeds coated with flesh are exposed).

Reproductive success and resource allocation

Expectation of the total amount of carbohydrate (C) consumed from bud formation until fruit maturation is expressed as:

$$ C = {\sum\limits_{t = 1}^{45} {c{\text{(}}t{\text{)}}L{\text{(}}t{\text{)}}} }, $$
(7)

where c(t) is the demand for carbohydrate given by Eq. 2, and L(t) is the survival curve given by Eq. 6.

Investments in each growing day were divided into successful and unsuccessful investments, and reproductive success was divided into male function and female function.

  1. 1.

    Successful investments in both male and female functions [c 1(t)] are expressed as:

    $$ c_1(t) = c(t)L(t) \frac{{L(45)}} {{L(t)}}\quad {\left( {1 \le t \le 19} \right)}, $$
    (8)

    where 19 is the day of the flower stage, 45 is the day of mature fruits, c(t) L (t) is the demand for carbohydrate including the survival curve for each growth day, and L(45) L(t)−1 is the proportion of reproductive organs that remained at fruit maturation. Thus, c 1(t) is carbohydrate, which the reproductive organs reaching fruit maturation consume during bud and flower stages.

  2. 2.

    Successful investments only as a male function [c 2(t)] are expressed as:

    $$ c_2(t) = c(t)L(t)\frac{{L(19)}} {{L(t)}}\quad {\left( {1 \le t \le 19} \right)}, $$
    (9)

    where L(19)L(t)−1 is the proportion of reproductive organs that remained until the flower stage. Thus, c 2(t) is carbohydrate, which the reproductive organs reaching the flower stage but not reaching fruit maturation consume during bud and flower stages.

  3. 3.

    Successful investments only as a female function, c 3(t), are expressed as:

    $$ c_3(t) = c(t)L(t) \frac{{L(45)}} {{L(t)}}\quad {\left( {19 < t \le 45} \right)}. $$
    (10)

Thus, c 3(t) is carbohydrate that the reproductive organs reaching fruit maturation consume during the fruit stages.

  1. 4.

    Investments in which both male and female functions were expected but were unsuccessful [c 4(t)] are expressed as:

    $$ c_4(t) = c(t)L(t) - c_1(t) - c_2(t)\quad {\left( {1 \le t \le 19} \right)}. $$
    (11)
  2. 5.

    Investments in which only the female function was expected but was unsuccessful [c 5(t)] are expressed as:

    $$ c_5(t) = c(t)L(t) - c_3(t)\quad {\left( {19 < t \le 45} \right)}. $$
    (12)

From the definitions, we find that:

$$ {\sum\limits_{t = 1}^{45} {(c_1(t) + c_2(t) + c_3(t) + c_4(t) + c_5(t)) = C} }. $$
(13)

Separation into sub-organs

The increase in dry mass in each sub-organ (M s) during the given period (t 1–t 2) was obtained using the following formula:

$$ M_{{\text{s}}} = {\sum\limits_{t = t_{1} }^{t_{2} } {\{ m_{{\text{s}}} (t) - m_{{\text{s}}} (t - 1)} }\} L(t), $$
(14)

where m s(t) was the estimated dry mass of the sub-organ, which is expressed as:

$$ m_{{\text{s}}} (t) = {\text{ }}gm(t)^{h} , $$
(15)

where m(t) was dry mass at the day of growth (t), and g and h were obtained from allometric relations between the dry mass of entire reproductive organs and sub-organs separately from the bud and flower stages or the fruit stages.

Damage to reproductive organs

To investigate the proportion of damaged reproductive organs among aborted ones under natural conditions, dropped reproductive organs were captured using nets with 2-mm mesh, and checked for damage.

Hand pollination

To test the selective abortion hypothesis, we conducted hand- pollination experiments on September 1995 and then checked fruit set under various conditions (listed below) 2 weeks later. The sample flowers were selected from ten individuals. Flower buds, flowers and fruits were not removed from those individuals.

Bagging

One day before flowering the flower buds were bagged, and the bags were removed 1 day after flowering.

Self-pollination

One day before flowering the flower buds were bagged, and when the flowers opened the anthers were collected. The stigmas of the same flowers were pollinated by pollen grains collected from the anthers, then the flowers were bagged again, and the bags were removed 1 day after flowering. To collect pollen, the anther was squeezed with tweezers, and the tweezers were moved toward the anther tip, where an anther pore was present, until pollen grains were pushed out. Pollen from the anther tip were directly attached to the stigma. The tweezers were wiped so as to avoid pollen contamination, although they seldom touched the pollen.

Cross-pollination

The procedure was the same as the self-pollination procedure, but we used anthers collected from more than three different individuals.

Open pollination

Flowers were marked without bagging and without hand pollination.

Pollinator visit

In order to test the pollinator attractiveness hypothesis, pollinator visits were observed on 22 September 1995, 23 September 1995, 29 April 2004, 3 September 2004 and 4 September 2004. The pollinators were three species of Xylocopa (species unidentified, probably not yet described) and Amegilla insularis. When these four species of bee visit flowers, their bodies (the under surfaces of the thorax or abdomen) touch the stigmas, and they then buzz the anthers to collect pollen, and pollen grains are deposited on their bodies. Smaller bees, Trigona spp., also visit flowers, and they collect the pollen remaining at the anther tips after the buzzing of the pollinator bees, but they seldom touch the stigmas. Thus, visits of Trigona spp. were not included in the pollinator visits.

All individuals and flowers found in the population were coded, and the pollinator visits were observed continuously from the opening of the flower until the flower wilted. However, on 29 April 2004, the individual plants were coded, and display size (the numbers of flowers per plant individual per day) were counted, but flowers appearing on the same individual were not separately coded. The number of pollinator visits per flower as a function of display size was obtained. Pollinator visits from different plant individuals and from neighboring flowers within the same individual were recorded, as were pollinator visits from a specific individual to different individuals and to neighboring flowers within the same individual.

Results

Resource economics for reproduction under fluctuating environments

Because log-transformed respiration rates were not significantly different between young and intermediate fruits (F = 1.0) when temperature was the covariance, those stages were combined and called the young–intermediate fruit stage.

Growth

Both dry weight and water weight were significantly regressed with length and width (Table 1). The duration of each stage was determined on the basis of dry weight (Table 2). Exponential growth was observed during the bud and large fruit stages, but growth decelerated during the young–intermediate fruit stages, and hyperbolic curves were fitted (Table 2; Fig. 2a), because among two common decelerating theoretical curves hyperbolic curves fit better than logistic curves.

Table 1 Regression between size (mm) width (x), length (y) and weight (mg). NS  Not significant (P > 0.05)
Full size table
Table 2 Basic data of resource economics for the reproductive organs in each stage. T Temperature (°C), t number days from earliest stage of flower bud
Full size table
Fig. 2
figure 2

a Dry weight of reproductive organs estimated using the result of multiple regression between size and dry weight. On the 19th day (flower stage), both the dry weight of whole reproductive organs and the dry weight after removing petals and stamens were plotted. Exponential weight growth was observed during the bud and large fruit stages, but growth slowed during the young–intermediate fruit stages, and a hyperbolic curve was fitted. b The survival rates were low in young fruits 3–5 days after flowering (ovaries had clearly grown from the flower stage) and in young buds. The survival rate from the youngest buds (1st day) until flowers (19th day) was 40.5%, and the survival rate from the youngest buds (1st day) until mature fruits (45th day) was 8.1%. The flower stage (19th day) is indicated by closed symbols

Full size image

Respiration and transpiration

Log-transformed respiration rates increased significantly with temperature except in large fruits (Fig. 3; Table 2). Transpiration rates per unit dry weight were similar among the stages, but exceptionally high transpiration rates were observed in flowers. The increase in water in the reproductive organs was much smaller than the amount of water consumed in transpiration (Table 2).

Fig. 3
figure 3

The relationship between respiration rates (log scale) and temperature at various stages of reproductive organs: a bud stage, b flower stage and c fruit stage. Because log-transformed respiration rates were not significantly different between young and intermediate fruits (F = 1.0) when temperature was the covariant, these stages were combined and expressed by one formula. Log-transformed respiration rates increased significantly with temperature except in large fruits. The respiration rate was regarded as constant in large fruits, and the mean respiration rate was calculated

Full size image

Resource supply

Both photosynthetic rates and PPFD vary in the daytime. Because stomatal conductance decreased after 1000 hours, the relationship between PPFD (I, μmol m−2 s−1) and photosynthetic rates (P, μmol CO2 m−2 s−1) was obtained separately (Fig. 4). The initial slope (b) was 0.038. Respiration rate (R p) was 0.80 μmol CO2 m−2 s−1. Maximum gross photosynthetic rates (ba −1) before and after 1000 hours were 11 and 7.3 μmol CO2 m−2 s−1, respectively.

Fig. 4
figure 4

The relationship between net photosynthetic rate and PPFD under natural light conditions. Because stomatal conductance decreased after 1000 hours, the photosynthetic curves were obtained separately before (solid line) and after (broken line) this time: initial slope is 0.038, respiration rate is 0.80 μmol CO2 m−2 s−1 and maximum gross photosynthetic rates before and after 1000 hours are 11 and 7.3 μmol CO2 m−2 s−1, respectively

Full size image

The transpiration rate from leaves was 376 mg cm−2 day−1, and the mean leaf area of a single leaf was 19.7 cm2. The amount of carbohydrate demanded by one flower (23 mg; Table 2) was supplied by 0.18–0.31 leaves, because daily photosynthetic production per leaf area was between 74.9 and 124.4 mg day−1. The amount of water demanded by one flower (2.0 g, Table 2) was the same amount as that transpired from 0.3 leaves, because a single leaf consumed 7.4 g H2O day−1 by transpiration.

Responses of plants

The carbohydrate demands of reproductive organs of nine individual plants for each day were compared with the total photosynthetic productivity during the previous day by Spearman’s rank correlation test. The correlation was positive in four out of nine individuals, insignificant in four individuals and negative in one (Table 3). Positive correlations were found to be predominant when they were compared with the total photosynthetic production over the previous 3 or 5 days. For the analyses at the population level (the total carbohydrate demands of reproductive organs of all individual combined were compared with the photosynthetic productivity), the relation was always significant (Table 3). The water demands of the reproductive organs on each day for each of the nine individuals and the total demand of all individuals combined did not correlate with either total rainfall for the previous 5 days or the radiation accumulated over the previous 5 days (Table 4).

Table 3 Each day’s carbohydrate demand of the reproductive organs of nine individual plants and all individuals combined were compared with the total photosynthetic productivity per leaf area during the previous 1, 3 or 5 days by Spearman’s rank correlation test. Also, the carbohydrate demands of all individuals combined per the total photosynthetic productivity per reproductive organs were compared by Spearman’s rank correlation test
Full size table
Table 4 Each day’s water demand of reproductive organs of nine individual plants and all individuals combined were compared with the total rainfall for the previous 5 days or the accumulated radiation for the previous 5 days by Spearman’s rank correlation test. Also, the water demands of all individuals combined per the total rainfall for the previous 5 days were compared with abortion rates of reproductive organs by Spearman’s rank correlation test
Full size table

The daily abortion rates of the population on the next day were compared with the following quantities: demand for carbohydrate on the present day per total photosynthetic production per leaf area during the previous 1, 3 or 5 days; and demand for water on the present day per total rainfall during the previous 5 days by Spearman’s rank correlation test. Significant correlations were found only with the former (Tables 3, 4). That is, the daily abortion rates were correlated with carbohydrate resource but not water resource.

Reproductive succession and resource allocation

Survival curve

The survival rates were low in young fruits 3–5 days after flowering (ovaries had clearly grown from the flower stage) and in young buds (Fig. 2b). The survival curve from the youngest buds for which monitoring started (t = 1) until flowers (t = 19) was 40.5%. The survival curve from flowers (t = 19) until mature fruits (t = 45) was 19.9%, i.e. L(19) = 40.5%, and L(45) = 8.1% (Fig. 2b).

Efficiency of investment

The expectation of the total amount of carbohydrate consumed by a reproductive organ (C) was 102.5 mg, and the proportion consumed in respiration was 35% of the total C; this proportion differed among the stages (24–49%; Table 2).

Of the total C, 69.8% was invested until the flower stage. The fraction of C that was abandoned, i.e., unsuccessful investments in either a male or female function, was 30.8% (Fig. 5).

Fig. 5
figure 5

Investments of the total carbohydrate demand (102.5 mg) were divided into five cases: successful investments in both male and female functions (investment of 13.5%) (c 1 ), successful investments only in a male function (45.1%) (c 2 ), successful investments only in a female function (13.2%) (c 3 ), investments in which both male and female functions were expected but were unsuccessful (13.5%) (c 4 ), and investments in which only the female function was expected but was unsuccessful (17.3%) (c 5 ). Of the total demands, 69.8% was invested until the flower stage, and unsuccessful investments in either a male or female function were 30.8%

Full size image

Separation into sub-organs

The growing patterns of the different sub-organs were different from one another (Fig. 6). Because calyxes grow quickly from the early stages of development, the dry mass allocated to calyxes until the flower stage (including those that were aborted on the way) was the largest among the sub-organs (37.4%, Table 5). Because the ovaries were completely covered with calyxes but the stamens and petals were exerted, the main function of the calyxes was regarded as protection of the ovaries. The dry mass allocated to calyxes and pistils until the flower stage was 53%, which was not necessary for male functions.

Fig. 6
figure 6

Reproductive organs were divided into four sub-organs: petals (circles), stamens (triangles), calyx (squares) and pistil or ovary (diamonds). The values of each sub-organ at the flower stage (19th day) are represented by closed symbols. The growing patterns of the different sub-organs were different from one another. Calyxes grew quickly from the early stages of development. By contrast, the growth of pistils was the slowest among sub-organs during the bud stage, but the ovary grew quickly during the fruit stage

Full size image
Table 5 Dry mass (mg) allocated to sub-organs of reproductive organs, and percentages in parentheses
Full size table

Damage to reproductive organs

Among the 211 dropped reproductive organs collected in nets, only four (1.9%) were eaten and fatally damaged. Some reproductive organs were found to have been sucked by Hemiptera, but the sucking was not fatal, and sucked organs often remained on the trees.

Hand pollination

The bagged flowers (n = 30) did not set fruit, but fruit did set under the other three treatments, that is, self-pollination (n = 13), cross-pollination (n = 30) and open pollination (n = 30), was 23, 23 and 30%, respectively. Fruit set under the bagging condition was significantly smaller than that under the other three treatments (P < 0.05, Fisher’s exact test), and no pairs of fruit set under self-pollination, cross-pollination or open pollination were significantly different (χ2-test or Fisher’s exact test when there were expectations smaller than 5). These results indicate that autogamy does not occur, that this species is self-compatible and that pollination processes do not limit the fruit set.

Pollinator visit

The flowers opened between 0750 hours and 0830 hours and closed between 1520 hours and 1700 hours. Pollinators visited before 1100 hours only. Out of 296 visits, 57.8% pollinators visited from the different individual plants, and 58.8% pollinators visited different individual plants. The average display size, the number of flowers per individual plant, was eight, and the mode was one. The number of pollinator visits per flower per day was not related to display size (F = 2.6, not significant; Fig. 7), that is, plants with more flowers are not visited more frequently by pollinators than plants with fewer flowers.

Fig. 7
figure 7

The relationship between the number of pollinator visits per flower per day and display size. The number of pollinator visits per flower per day was not related to display size (F = 2.6, not significant)

Full size image

Discussion

We tested five hypotheses by Holtsford (1985) with regard to the ultimate factor of excess production. Through tests on the hypotheses, we demonstrated that the abortion of reproductive organs of a tropical pioneer, M. malabathricum, that reproduces continuously was an adaptation to environmental fluctuations over shorter time scales than has been previously reported in other plants.

Selective abortion

Pollination processes do not limit fruit set in M. malabathricum, because fruit set under cross-pollination and natural conditions were similar.

Stigmatic receptivity and growth of pollen tubes have been studied with regard to paternal selection and pre-fertilization processes (Delph et al. 1998; Cowan et al. 2000; Bernasconi 2003); however, the mechanisms of post-fertilization paternal selection, and selective embryo abortion, are complicated and are not fully understood (Montalvo 1992; Melser et al. 1997; Korbecka et al. 2002; Raspé et al. 2004). Regardless of the mechanisms, pollen quality and the number of fertilized seeds have been reported to be factors in selective abortion (Stephenson and Winsor 1986; Obeso 1993), and no other factors have been indicated.

Burd (1998) showed theoretically that selective abortion due to these factors increased plant fitness. It has been reported that the fitness of selfed seeds is lower than that of crossed seeds in many plants (Huth and Pellmyr 2000; Shi et al. 2005). In plants without self-incompatibility of pre-fertilization such as M. malabathricum, it appears that selfing versus outcrossing is the most important factor of selective abortion based on the quality of pollen. If self-pollinated ovaries were selectively aborted, the fruit set under self-pollination would have to be low. However, the fruit set was not significantly different among self-pollinated, cross-pollinated and naturally pollinated ovaries.

In our cross-pollination experiments, a sufficient number of pollen grains collected from multiple donors were transferred to stigmas, but this did not result in a higher fruit set.

As for the factors of selective abortion previously reported, the selective fruiting hypothesis was not supported in our plant material. In addition, high abortion rates in the flower bud stages are not explained by the selective abortion hypothesis.

Pollinator attractiveness

Because the main pollinators, carpenter bees, show trap-lining foraging behavior (Momose et al. 1998b), it is predicted that the ideal free distribution of visits is realized (Fretwell and Lucas 1969). In such cases, plants should not have many flowers at the same time, but they should continue to produce one or a few flowers over a long period (Momose et al. 1998b). M. malabathricum is a typical example: it flowers throughout the year, producing one or a limited number of flowers per day, as reported in gingers pollinated by trap-lining bees (Sakai 2000). Our observation of pollinator visits also agreed with the prediction. Pollinator visits per flower were not significantly related with the display size.

The pollinator attractiveness hypothesis was not supported in our plant material.

Male function

It is difficult to assume the case that completely contradicts the male function hypothesis. That is, the function as pollen donor in hermaphrodite flowers is almost always at least partly present. In fact, because flowers were visited by pollinators 2.7 times on average, excess flowers had male functions. It should be discussed here whether the evolution of excess flowers can be explained from the male function hypothesis (Burd and Callahan 2000). As has been mentioned, not only young ovaries just after flowering, but also flower buds and developed ovaries, are frequently aborted. The low survival rates at such various stages are not explained by the male function hypothesis.

As seen in Fig. 5, 45.1% of total reproductive investment was distributed to the flower incidentally functioning only as a male. In addition, until the flower stage, 53% of dry mass was allocated to sub-organs not necessary for male functions. Such large dry mass allocation to calyxes and pistils is not consistent with the male function hypothesis.

Reproductive assurance

Among the reproductive organs aborted under the natural conditions, the rate of fatally damaged reproductive organs was low. One might suspect that this plant might sometimes suffer heavy seed damage. Even so, it is not adaptive to produce excess flowers to assure seed production under such rare conditions, because the plant will reproduce many times, not only once, in its lifetime. Therefore, the reproductive assurance hypothesis was not supported for the plant we studied.

Resource boom

The carbohydrate demands of the reproductive organs for each day were correlated with the total photosynthetic productivity per leaf area during the previous 1, 3 and 5 days (Table 3). Because the growing schedule of the reproductive organs was mostly fixed, this result suggests that the number of reproductive organs was regulated according to photosynthetic productivity. An increase in the number of reproductive organs can be realized through the creation of new buds, and a decrease can be realized only through the abortion of reproductive organs.

In support of the above statement, the daily abortion rates on the next day were correlated with the carbohydrate demand on that day divided by the total photosynthetic production per leaf area during the previous 1, 3 or 5 days (Table 3). Thus, the resource boom hypothesis was supported. The abortion of reproductive organs in M. malabathricum is an adaptation to environmental fluctuations over shorter time scales than previously reported in other plants.

Reproduction of tropical pioneers

Reproduction of tropical pioneer plant species is characterized by continuous reproduction (Momose et al. 1998a; Sakai et al. 1999), which is possible because plant and pollinator activity is not limited by seasonality. All previous studies on excess production of reproductive organs were conducted in locations with clear seasonality (Coulter 1979; Willson and Price 1980; Lee and Bazzaz 1982; Holtsford 1985; Eriksson 1987; Gorchov 1988; Vaughton 1991; Parra-Tabla and Bullock 1998; Nishikawa 1998), while our study was conducted in an aseasonal tropical wayside. Excess production and the abortion of reproductive organs of the plant studied, M. malabathricum, are adaptations to a short timescale without precedent.

Even without seasonal limitations, tropical climax species reproduce at intervals of 1 year or longer (Momose et al. 1998a; Sakai et al. 1999). They reserve resources: the higher their photosynthetic productivity, the more their resource reserves increase (Yoneda et al. 2002). In periods of reproduction, they produce as many reproductive organs as the size of reserves allows (Meletiou-Christou et al. 1998; Körner 2003; Ichie et al. 2005a, b). Such a strategy has the advantage that the loss of resources (30.8%; Fig. 5) in order to adjust demand to supply is avoided, but it also has some disadvantages. First, during the reproductive phase, long-lived plants utilize the resources that they have stored (Chapin et al. 1990); however, the resources are all lost when the individual dies. Second, when the resource is translocated, not only its own cost (Gamalei 2002; Lalonde et al. 2003), but also risk of attacks by shoot-sucking insects cannot be ignored. Third, there are costs for producing the storage organs and for maintaining the physical or chemical protective systems for the reserves (Chapin et al. 1990; Gatehouse 2002).

The most important difference related to the above disadvantages between climax and pioneer trees is in mortality. Because climax trees live longer, i.e., their mortality rate is relatively low (Clark and Clark 1992; Shimano 2000), the probability that their reserved resources will be lost as a result of death of an individual is relatively small. Thus, the reserve strategy is used only in climax species. Additionally, shoot-sucking insects are more abundant in open habitats or in gaps than in closed forest stands (our observations, with which most researchers working in the same field agree). Translocation of photosynthetic production over long distances from the leaves to the reserves and from the reserves to the reproductive organs are especially disadvantageous for plants growing in habitats with abundant shoot-sucking insects.

The tropical pioneer tree M. malabathricum reproduces continuously. When photosynthetic productivity is high, these trees let their reproductive organs grow and produce new flower buds. On the other hand, when photosynthetic productivity is low, reproductive organs are aborted. Such a reproductive strategy, which minimizes the reserved resources for reproduction, is explained by the typically high mortality of tropical pioneers and the abundance of shoot-sucking insects in their habitats.