POPULATION ECOLOGY OF PAEPALANTHUS POLYANTHUS: PREDISPERSAL HAZARDS AND SEED PRODUCTION1

(Population ecology of Paepalanthus polyanthus: predispersal hazards and seed production). This study aimed at evaluating seed production and predispersal hazards in a sand dune population of P. polyanthus (Eriocaulaceae) in Southern Brazil. Bad development of flowering capitula was caused by the wind and by interference among umbels. A positive correlation between the proportion of atrophied capitula and the number of capitula/ umbels also suggested resource limitation. A caterpillar of a Recurvaria Haworth (Lepidoptera: Gelechiidae) species that eats flowers and a boring caterpillar (Lepidoptera not identified) were the main herbivores. Plants reproducing during the flowering peak had a lower probability of being damaged by Recurvaria sp., suggesting an escape from herbivores by flowering synchronism. The proportion of capitula damaged by herbivores was low, causing a reduction of about 5% in the seed set/plant. The boring caterpillar may or may not cause umbel abortion. When the abortion occurs, seed production is reduced by 12%. Many capitula with no damage and individual flowers (up to 50%) did not produce seeds. Pollination failure could be related to this low rate of seed set. We discuss the fact that 318 Castellani & d’Eça-Neves: Population ecology of Paepalanthus polyanthus (Eriocaulaceae) the low recruitment of seedlings reported for populations of P. polyanthus did not seem to be limited by seed rain, which was estimated at an average of 8000 seeds/plant even when herbivore damage occurred.


Introduction
Insufficient pollination, resource deficiency, predation, and development failure due to deleterious genetic combinations are the main causes of mortality of ovules and seeds during the predispersal phase (Stephenson 1981;Lee & Bazzaz 1982;Bawa & Webb 1984;Fenner 1985).Flower and seed predation before dispersal reduces seedling recruitment in coastal environments and in prairies (Louda 1982;Louda et al. 1990).Herbivores on flowering stems decreased the number of capitula and the density in a sand dune population of Cirsium vulgare (Savi) Ten.(Compositae) (Klinkhamer et al. 1988).In Vulpia fasciculata (Forskal) Samp.(Gramineae), grazing on developing inflorescence may cause plant death and reduce the seed rain (Watkinson & Harper 1978).Flower, fruit and seed predation in the predispersal phase have been described for other sand dune plants, and the importance of these factors for seedling recruitment has often been discussed in the literature (Wilson 1977;Sterk et al. 1982;Zedler et al. 1983;van Groenendael 1986;Costa et al. 1991).
Paepalanthus polyanthus (Bong.)Kunth (Eriocaulaceae) is a monocarpic species, found in sand dunes in Southern Brazil.Populations of this plant undergo periods of marked local reduction due to the low emergence of seedlings without reposition of dead reproductive plants.Expressive seedling recruitment occurs in years of high rainfall, suggesting some dependence of germination and seedling survival on the water available (Castellani et al. 1995a, b).
In this study, our objective was to estimate the seed production of P. polyanthus and to evaluate the factors that reduce seed production during the predispersal phase.Since in other plant species abortion and the flower-and-seed predation rate may vary with the size of reproductive structures (Fenner 1985, Santos et al. 1992), and since the predation rate may be influenced by synchronisation of flowering and fruiting within the population (Janzen 1971, Silvertown 1980, Augspurger 1981, Santos et al. 1992), we also evaluated the importance of these factors in relation to P. polyanthus.

Materials and methods
Studied species -In Santa Catarina State, Brazil, Paepalanthus polyanthus (Fig. 1A) occurs in upland fields and coastal sand dunes.This is a monoecious species, presenting umbels with hundreds of capitula of male and female flowers (Moldenke & Smith 1976).In sand dunes, P. polyanthus can reproduce in the second year of life, or later, and is described as a short-lived monocarpic perennial plant (Castellani et al. 1995b).This species flowers from July to January, with a flowering peak in November.The fruiting period begins in September, with a peak in December (d'Eça Neves & Castellani 1994).The average number of umbels per plant was described as around 14 in 1988, 10 in 1989 and 8 in 1990, a positive correlation occurring between the number of umbels produced and the rosette diameter (Castellani 1990;d'Eça Neves & Castellani 1994).The flowering of umbels is synchronised within the plant, with the anthesis of male flowers occurring first, followed by female flowers and male flowers again (d'Eça Neves & Castellani 1994).
Study area -This study was carried out in a 450m 2 plot, located 500m from the sea in a sand dune slack at Joaquina Beach (27 o 36'40"S and 48 o 27'10"W), Santa Catarina State, Brazil.The local climate is characterised as Cfa mesothermic, according to Köppen (Souza et al. 1992).Rainfall occurs all year round, with higher indices in January and February, and lower precipitation from June to August (Castellani et al. 1999).During the study year (1990), the lowest average temperature occurred in June (14.5o C), and the highest in February (24.8 o C).Rainfall distribution followed the above pattern, the total annual precipitation being 1941mm (DEPV-FL, Florianópolis, SC).The predominant winds at Joaquina Beach are those from both Northern and Southern directions (Santos et al. 1995).
Agents inhibiting reproduction -A total of 123 plants began reproduction in the plot.These were marked and periodically checked to determine their reproductive phenological phases defined as: bud (umbels with closed and developing capitula), flowering (capitula with female or male flowers opening), and fruiting (capitula with developing seeds until dispersal began).For each sample, the umbels of each plant were counted and checked for the presence of herbivores and other damaging agents.Damaged umbels were marked and observed until dispersal.
Pupas of the two main herbivores associated with P. polyanthus inflorescences were collected in capitula, or entire umbels, and kept in containers under laboratory conditions until adult emergence.The adults obtained were identified by Dr. Vitor O. Becker (EMBRAPA/DF, Brazil).
Capitula production and damage -During the reproductive season, 152 fruiting umbels were collected before dispersal.In the laboratory, capitula were classified as: 1) developed; 2) developed with damage due to herbivores; and 3) atrophied.The atrophied capitula were classified as enclosed or not enclosed by the capitulum stem sheath (Fig. 1B).The number of capitula in each class was recorded.
Flower and seed production -In the fruiting plants, no flowers or seeds fell from the capitula.Utilising a stereoscopic microscope (Wild Leitz M3Z), 1% of the developed capitula damaged and 1% not damaged by herbivores were analysed to determine the number of male and female flowers/capitulum and the number of seeds/flower.Female flowers were quantified according to the occurrence of anthesis, the number that initiated seed formation, and those that produced seeds.

A B
herbivore damage in plants with different numbers of umbels and the occurrence of plants damaged during the flowering season were analysed by G-test (Sokal & Rohlf 1969).The Spearman rank correlation (Zar 1974) was applied to test relationships between the proportions of damaged plants, damaged capitula or atrophied capitula, and the number of reproductive structures.
Comparisons of flower and seed production in capitula with and without herbivore damage were done by the use of Mann-Whitney U tests.In calculations of significance, normal approximation was used, obtaining the Z-value, following the procedures suggested for tied rank data (Zar 1974).
The proportions of flowers that maturated seeds in developed capitula and in capitula damaged by herbivores were estimated.A comparison of these proportions was carried out using a 2x2 contingency table, considering proportions and calculating the Z-value with correction for continuity (Zar 1974).

Herbivores and percentage of plants damaged throughout the reproductive season -
The caterpillars of two Lepidoptera species were the main herbivores damaging the reproductive structures of P. polyanthus.Caterpillars of a Recurvaria Haworth (Lepidoptera: Gelechiidae) species were observed eating flowers.These caterpillars, about 5mm in length and creamcoloured, also use individual flowers adhered to the body for camouflage.Damage by this species was observed in 59 (47.97%) of the 123 monitored plants, and occurred in 197 (23.71%) of the 831 umbels produced.The Recurvaria sp.caterpillars were observed on flowering plants, but they were registered only once on plants in the budding phase (Fig. 2 A, B).In fruiting plants, Recurvaria sp. were registered in pupa or in the last larval instar.
The frequency of flowering plants damaged by Recurvaria sp. was significantly different during the flowering season (G-test 15.12, gl=7; P<0.05).Plants flowering during peak blooming (November) showed a lower probability of being damaged by these caterpillar species (Fig. 2B).Therefore, a negative correlation between the percentage of plants damaged and the number of plants in flower was found (Fig. 3).The occurrence of damage caused by Recurvaria sp. was independent of the number of umbels/ plant (G-test=5.69,gl=5, P>0.05) (Fig. 4).
The other Lepidoptera observed was a boring caterpillar.Since no adults emerged from the pupas kept in the laboratory, this species could not be identified.The boring caterpillar damaged the base and stem of the umbels, causing umbel abortion or loss of the individual capitula.Damage by this Lepidoptera species was observed in 28 plants (22.76%), with umbel abortion in 10 (8.13%).The boring caterpillar was observed from the budding phase onwards, and was mostly recorded in flowering plants (Figs.2A, B).The frequency of plants damaged by this species did   2B).This caterpillar pupates among capitula in flowering and fruiting plants.
Production of flowering capitula and loss by bad development and herbivores -Paepalanthus polyanthus umbels produced an average of 290 flowering capitula of which 62.42% were fully developed, having suffered no damage by herbivores (Tab.1), as described below.
About 29% of the capitula did not develop (Tab.1).A positive correlation between the proportion of atrophied capitula/umbel and the total number of capitula/umbel (r s = 0.25; P<0.002; n=152) was observed.Wind action caused desiccation, abrasion and atrophy of capitula in 36 plants (29.27%) affecting 234 umbels (28.16%).Part of the capitula of these umbels was atrophied, especially those exposed to the south wind (Fig. 5).Overlapping umbels also presented undeveloped capitula.Many capitula which were in mutual contact did not open.This fact was observed in 8 plants (6.5%) and 22 umbels (2.65%).
Among the 152 collected umbels, 134 were damaged by the caterpillars.Herbivores affected 8.36% of the total production of capitula/umbel (Tab.1).The proportion of capitula damaged by herbivores did not show a significant correlation with the number of capitula developed/umbel (r S = 0.0016; P>0.50; n=152).Herbivores on flowering capitula significantly reduced the number of male and female flowers per capitulum (Tab.3; Fig. 6).The frequency of capitula with no female flower increased from 0.75% to 11.0% (n=134) when herbivores occurred; capitula with no male flower increased only from 0.75% to 2.24% (n=134).The reduction in flower numbers (about 23-29%) contributed to a significant reduction in seed production (about 46% less) (Tab.3; Fig. 7).The occurrence of capitula with no seed set increased from 23.13% to 36.56% (n=134).It was observed that damaged capitula presented a higher proportion of flowers that reached anthesis but did not initiate seed development, thus significantly reducing the number of female flowers that produced mature seeds (Tab.4) (Z=3.32;P<0.001).Male flowers in damaged capitula were also affected, exhibiting 6.28% of chewed flowers and 4.02% of loose flowers (n=1639).
Despite herbivore action it was observed that a high number of developed capitula in P. polyanthus did not produce seeds (Fig. 7; Tab. 2, 4).
Production of seeds per plant -Based on   1994), a plant with no herbivore damage could produce about 9248 seeds.Assuming an average of 24.24 capitula damaged by herbivores and a production of 3.02 seeds/capitulum damaged, an umbel would produce 1089 seeds.A plant with this level of damage in all umbels would produce 8744 seeds, nearly 5.45% less than the produc-tion expected without herbivores.Seeing that the boring caterpillar can cause umbel abortion, a greater loss in seed production could occur.In 1990, we did not observe more than one aborted umbel/plant; therefore, we were able to estimate that in an average plant (8 umbels) this damage would generate a reduction to 8097 seeds, approximately 12.45% less than that estimated for a plant without herbivores.

Discussion
Paepalanthus polyanthus can produce hundreds of capitula/umbel, but about 40% are affected by faulty development or herbivore attack.
Three factors may be involved as causes of atrophied capitula in P. polyanthus: wind, resource deficiency and mechanical interference among umbels.Wind causes desiccation and atrophy of capitula in P. polyanthus affecting flower production in about one-third of the population.Although the influence of the wind on the vegetative development of plants in coastal zones is usually mentioned (Daubenmire 1974;Etherington 1982), the studies on seed production and predispersal hazards in sand dunes (Watkinson & Harper 1978;Sterk et al. 1982;Zedler et al. 1983;Westelaken & Maun 1985;van Groenendael 1986;Klinkhamer et al. 1988;Costa et al. 1991) do not refer to wind effect on the development of reproductive structures.Nevertheless, wind action is reported by Lee & Bazzaz (1982) as an abiotic factor which may cause the loss of young fruits or seeds.
Indirect evidence suggests that some limitation of resources influences the full development of capitula in P. polyanthus.A positive relationship between abortion rate and the size of inflorescence is indirect evidence for resource limitation (Stephenson 1980apud Fenner 1985;Santos et al. 1992) and in P. polyanthus such a correlation was significant.The high number of developed capitula with no seeds does not seem to be explained by failure in seed maturation; many flowers reached anthesis but did not initiate seed development.This fact may suggest a pollination failure.D'Eça Neves & Castellani (1994) suggest a dependence on pollination agents in P. polyanthus reproduction, and the visiting rate in this plant being low (Vera L. V. De Arruda, pers.comm.).In other sand dune species, low rates of fruit and seed production are related to the inefficiency of pollination and/or limitation of resources (Sterk et al. 1982;Zedler et al. 1983;Westelaken & Maun 1985;Cordazzo & Davy 1994).
Some hypotheses related to damage may also be discussed.Excessive reproductive structures could substitute those that are damaged or lost through predation (Lee & Bazzaz 1982) and may also help to satiate herbivores (Zedler et al. 1983).For P. polyanthus, a significant relationship between the rate of damaged capitula by herbivores and the number of capitula developed/umbel was not found.Nevertheless, the proportion of plants damaged by Recurvaria sp.tended to be lower during the flowering peak.This suggests that a synchronism within flowering plants could provide  1981).In Ipomoea cf.grandifolia (Dammer) O'Donell (Convolvulaceae), a reduction in the rate of fruit predation occurs during fruiting peaks (Santos et al. 1992).Reduction in predation rates, with synchronism of the population, may reflect satiation effects at moments of higher production (Janzen 1971;Silvertown 1980).
The action of herbivores on reproductive structures of P. polyanthus did not cause a representative loss in total seed production.Although the production of seeds/capitulum damaged is reduced by 50%, the average proportion of damaged capitula is low in the population.The frequency of umbel abortion is also low.
In relation to other sand dune plants, Cirsium rhothophilum Blake (Compositae) lost 25% of achene production due to insect herbivores (Zedler et al. 1983).In Cirsium vulgare (Savi) Ten.(Compositae), the loss due to damage in flowering capitula was of 3% and 17%.However, predation on stem flowering reduced the production of seeds by 80% and 15% thus diminishing seedling recruitment and population density (Klinkhamer et al. 1988).In Vulpia fasciculata (Forskal) Samp.(Gramineae), the reduction in seed production due to herbivores was from 4 to 31%, thus decreasing the seed rain (Watkinson & Harper 1978).Nevertheless, Keddy et al. (1983) considered that predation of flowers and seeds of Cypripedium passerinum Rich.(Orchidaceae) did not affect population recruitment; the emergence of seedlings was rare and dependent on humidity.Loss in seed production of a P. polyanthus plant damaged by predation was estimated at 5%, and was 12% in the case of umbel abortion by boring caterpillars.These low values observed in 1990 suggest that herbivore action in the predispersal phase is not a limiting factor to seedling recruitment in this population.Castellani et al. (1995a, b) refer to soil moisture as an important factor limiting regeneration, showing differences in the emergence of seedlings between years with different precipitation, and between areas with different distances from the water table.

Figure 2 .
Figure 2. Number of Paepalanthus polyanthus plants present in 450m 2 in different phenological phases (h) and percentage of plants damaged by Recurvaria sp.caterpillar ( >) and damaged by borer caterpillar (• ): A) plants in bud, (B) plants in flower.

Figure 3 .Figure 4 .
Figure 3. Percentage of plants damaged by Recurvaria sp.caterpillar in relation to the number of flowering plants in 450m 2 .r S : Spearman's rank correlation coefficient.

Figure 6 .
Figure 6.Number of female (A) and male flowers (B) of Paepalanthus polyanthus presented in developed capitula damaged (grey) and not damaged (white) by herbivores.

Table 2 .
Success of flowering and rate of seed maturation in female flowers of Paepalanthus polyanthus in developed capitula.The percentages (%) were calculated in relation to the 1498 flowers surveyed in 274 capitula from 152 umbels.

Table 3 .
Comparison of the male and female flower production and seed set per developed capitulum damaged and undamaged by herbivores.Values of Z and P in the Mann-Whitney U-test using normal approximation.Rates of reduction (%) were calculated by: (number presented on damaged -undamaged capitula / number in undamaged capitula) x 100.

Table 4 .
Success of flowering and rate of seed maturation in female flowers of Paepalanthus polyanthus in developed capitula damaged and undamaged by herbivores.The percentages were calculated in relation to the total number of flowers present in the capitula analysed (n=134 damaged and 134 undamaged).