SEED PRODUCTION AND SEED QUALITY OF THE DUNE BUILDING GRASS PANICUM RACEMOSUM SPRENG

(Seed production and seed quality of the elune-building grass Panicum racem.oslIm. Spreng.). Seeel production, poIlination requirement, seed characteristics relateel to quality and the re lationship between number anel mass of seeds were examineel for Palliculll racelllOSlIlll in three successional populations in southern Brazilian coastal dunes. The seed prod uction lVas generaIly low and eleclined further betlVeen the frontal dunes anel the backelunes, elropping fro m 4.05 seeds per panicle in the former to 1.8 seeds in the latter. HOlVever fettility (% fettile florets) eliel not differ among the three habitats. Plants cross-poIlinateel in a glasshouse showeel an increase in seeel proeluction to 4 1.4 seeels compared to no seed production in self-poIlinated plants. Caryopses varied in mass from 3.2 to 12 .2 mg lV ith a mean of 7.98 mg. A strong negative correlation was found between mean individual seed mass and the total number of seeds per panide in a natural population. However, this relationship elid not persist in seeds proeluceel by cultivated plants in the glassho use. The causes of 101V seed production appear to be mainly poIlen se lf-incompatib ility and additionally competition for nutrients between sexual reproduction and aIlocation to donal growth. Under conditions of nutrien t shOltage, Panicllll1 raCell10SlIII1 probably aIlocates resources more to cIonal growth and to fewer, but weIl-endowed seeds. This would pennit emergence from deeper burial sand , faster growth and greater survival of seeellings. RESUMO (Produção e qualidade da semente da gramínea Panicwn racel/wsU/11 Spreng. em um gradiente de dunas costeiras). A produção de sementes, po linização, características das sementes e a relação entre o nÚJllerO e peso de senlentes foranl exuJnin adas em três populações de Panicul11 racelllOSlIlIl ao longo de um gradiente nas dunas costeiras no sul do Brasil. A produção de sementes foi pequena e decresceu ao longo do gradiente, diminuindo de 4,05 sementes por esp iga nas dunas frontais para 1,8 sementes por espiga nas dunas mais estáveis. Entretanto não ocorreram diferenças na feitilidade entre as popul ações. Plantas culti vadas em casas de vegetação e que foram polinizadas cruzadamente mostraram um aumento na produção de sementes (4 1,4 sementes) em comparação com as plantas auto-polini zadas (nenhuma semente). As sementes variaram em relação à massa de 3,2 a 12,2 mg com uma média de 7,98 mg. Uma forte corre lação negativa foi encontrada entre a massa indi vidual das sementes e o número total de sementes por espiga nas sementes coletadas nas dunas frontai s. As causas da baixa proelução de sementes parecem ser principalmente devido à I Depattamento de Oceanografia, Un ive rsidade do Rio Grande, Caixa Postal 474, CEP 96500-900, Rio Grande-RS, Brasil. , School of Biological Sciences, Un iversity of East Angl ia, NorlVich, NR4 7TJ, UK.


SEED PRODUCTION AND SEED QUALITY OF THE DUNE
BUILDING GRASS PANICUM RACEMOSUM SPRENG.

Introduction
Panicul1l racemosumSpreng. is a warm-season, C4, perennial grass with a wide distribution from Argentina (35°S) to northeastern Brazil (8 0 S) (Smith et al.,1982;  Costa et aI.1991).On southern Brazilian foredunes , Panicul1l racemosum is the dominant dune-building grass on the first dune ridge (Cordazzo & Seeliger, 1993).It has considerable economic potential because of its dune-building and sand-binding abilities (Pfadenhauer 1980, Costa et aI. 1991, Cordazzo & Seeliger 1993) which are similar to those of Uniola paniculata and Anunophila species.P. racemosum appears to be very well adapted to the conditions prevailing in this environment (Pfadenhauer 1980) and strongly influences the plant cbmmunity and dune formation (Cordazzo & Seeliger 1993).
Despite its notable importance in the coastal environment, little is known specifically about the reproducti ve biology and ecology of Panicum racemosum.Most studies have been concerned with taxonomic and distributional aspects.Only a few studies have considered its growth, leaf demography and decline in vigour along the environmental gradient of a dune succession (Costa et aI. 1984(Costa et aI. , 1991)).Costa et ai.(1984, 1991) reported that Panicum racemosum can reproduce sexually and spread by c10naI growth.However, populations of Panicum racemosum in southern Brazil are maintained principally by c10nal growth (Pfadenhauer 1978), similarly to other dune-building species such as Uniola paniculata (Wagner 1964), Ammophila arenaria (Huiskes 1979) and A. breviligulata (Maun & Baye 1989).Harris & Davy (1986) found that a very low percentage of tillers of Elytrigia juncea (El)'mus farctus) produced inflorescences but, after a catastrophic tidal surge, seeds and rhizome fragments were of similar importance in giving rise to new colonies.According to Laing (1958) only 0.02% of new shoots in a population of Ammophila breviligulata had originated from seedlings.The survival of plants originating from rhizome fragments is usually high, probably because of stored reserves and the presence of root primordia on rhizome fragments (Maun & Baye 1989).Thus , dune-.building plants tend to allocate a small proportion of resources to tlowering culms (Maun 1985).In addition, many authors have supposed that the number of seeds is limited by competition among developing ovules for pollen or nutrients, and between fruits and other organs for limiting resources (Lloyd 1980).Another important element in the reproductive strategy of a plant concerns the partitioning of its seed output either into many small seeds or a few large ones .Within the constraints of a given reproductive allocation there is c1early antagonism between seed size and number (Fenner 1985, Silvertown & Lovett Doust 1993).
The objectives of this study were (a) to determine seed production rates and the mechanisms that limit them in Panicul1l racemosul1l populations across an environmental gradient in southern Brazil, (b) to examine seed mass and dimensions , and (c) to investigate whether the number of seeds produced is related to seed quality .

Material and methods
Studies were conducted on Panicum racemosum plants growing in coastal foredunes ofthe state ofRio Grande do Sul (southern Brazil).For details oflocations and site descriptions see Costa et aI. (1991) and Seeliger (1992).
Total culm density and flowering culm density were recorded in December 1991 in 100 randomly selected quadrats (0.5 x 0.5 m.) in each of the three P. racemosum habitats (frontal, top and backdunes).Seed production and floral characteristics were determined by collecting 25 flowering culms from each habitat.Florets were removed manually from each panicle and separated into filled (those containing a caryopsis) and unfilled (those without a caryopsis).The number of seeds (caryopses) per panicle was recorded as well as the number of florets per panicle and the panicle length.
Fecundity and percentage fertility were determined using the methodology described by K.rajnyk & Maun (1982) for Ammophila breviligulata populations.Thus, fecundity was definied as potential reproductive performance, determined by counting the number of florets on a panicle.It was assumed that each floret had the potential to produce a normal well-developed caryopsis.Fertility was calculated as the number of well-developed caryopses per panicle di vided by the number of florets.
In January 1991 seedlings of Panicul1l racemosum (5 days old) were individually planted in 26-cm diameter plastic pots filled with 4: 1 sand and potting compostmixture (John Innes no .2). Ali plants were maintained in an unheated glasshouse at the University of East Anglia, Norwich until they produced inflorescences.Nutrient so lution (NPK 7:7:7, 19/1) was supplied every two months.Fifty flowering culms of Panicum racemosum from the glasshouse were separated in two groups early the in flowering season.Twenty-five flowering culms were enclosed in glazed paper bags, so that each panicle was completely isolated from externai pollen.In the seco nd group, the panicles were not bagged and were manually cross-pollinated.Mature flowerin g culms were analysed as described previously.
Mass and length of 450 caryopses selected at random from field collections were recorded using an electronic balance and calipers, respectively .Cumulative means of seed mass were calculated with increasing sample size.When sample size exceeded 300 seeds, mean seed mass and variance remained fairly constant, and it was thus assumed that the 450-seed sample reflectecl true variation in the seed population .Mean mass of seeds and number of seeds per panicle were recorded .
Differences in seed production and floral characteristics were tested for statisti-cal significanee with a one-way analysis of varianee (ANOVA).Tllkey's mllltiple range test was llsed when the ANOV A showed a signifieant differenee (Sokal & Rohlf 1981 ).
Five replicates of 500 g of sand were colleeted monthly from eaeh habitat at 5 em depth between Oetober and Deeember 1991.These samples were used to determine phosphate, nitrate, ealcium, potassi.umand pH.Potassium and ealcium were extraeted with IM ammonium acetate at pH 7 (Allen et aI.1976) and determined by atomie absorption speetrometry.Nitrate was extraeted from sand with NaCI (6%) solution, and phosphate was extraeted with 0.5 M sodium biearbonate at pH 8 (Allen et aI.1976).Both, phosphate and nitrate were measured using eolorimetrie methocls, similar to the methodology deseribed by Costa et ai. (1991) for the same sites.The pH was cletermined using a Cole Palmer pH meter, on aI: 1 mixture by volume of sand and clistilled water.

Results
Panicles of Panicum racemosum are spike-like, ereet, eontraeted, more or less interrupted below, eontinuous and moclerately clense above and 10.5 to 27 em long (Table I).The densities of eulms and flowering eulms (those with panicles) are heterogeneous in eaeh habitat (Table I) .The ranges of density of eulms were 20-84, 16-80 and 8-76 per square meter for frontal, top and baekdune habitats, respeetively (Table 1).Mean number of eulms per square meter did not differ signifieantly between frontal and top dunes, but it was signifieantly lower in the baekdunes (Table 1).
Feellndity (total nllmber of f10retsper paniele) varied among the different habitats ofthe dune system, ranging from 140 to 68.3 (Table 1).In ali tlu'ee habitats a large proportion offlorets on eaeh panicle (about 91 %) eontained shrivelled earyopses which had either not developed beyond the stage of floral m' aturation or had developed for a short period after fertilization ancl then abortecl (the periearp and testa had developed but were empty).A small number offlorets (5 .08%)was entirely empty and only 3.38% hacl well-developec l earyopses (seeds).
Plants of Panicum racemosum started to produee infloreseenees only in the third spring after sowing (2 years old).The results of glasshouse experiments (Table 2) showecl that the mean number of seeds produeed in eross-pollinate d plants was tenfold higher than in the field, with a mean of 41.4 ± 2.9 seecls.The ratio of seedslflorets from eross-pollinate d plants was 0.3, fifteenfold higher than in the field.The bagged flowering eulms (self-pollinated ) did not produee any seed.Length of panicles and number of florets produeed were not statistieally different among fielcl, glasshouse eross-pollinatec l ancl glasshouse self-pollinated plants.* Means with lhe same letter in the row are not stati stically different according TlIkey 's multi pie range test.
Panicum racemosum is not a heavy seed producer.Mean number of seeds per panicle declined along the habitat gradient from 4.1 ± 0.7 to 2.0 ± 0.3 to 1.8 ± 0.5 in the frontal , top and backdunes, respectively (Table I).However, fertility (ratio of number of seeds per panicle to ali florets in the panicle) did not differ significantly along the habitat gradient (Table 1).
Seed length varied from 4.2 to 5.3 mm .Mean seed length was 4.75 ± 0.02 (mean ± standard error), which was lower than the mode (5.02 mm).Seed mass of Panicul1l racemosul1l varied from 3.2 to 12.2 mg per seed.Mean individual seed mass ofthis whole population was 7.92 ± 0.08, which was higher than the mode (7.8 mg) .The freqllency distribution of seed mass was skewed to the right (gl = -0.3273)and platykurtic (g2 = -0.3155)(Figure 1).
Correlation between seed mass and length did not show a significant relationship (F = 0.5227 ; P < 0.5), with a correlation coefficient of 0.073.Thus, heavier seeds were not necessarily longer.
A multiplicative model regression analysis of mean individual seed mass per panicle on the total number of seeds per panicle (Fig. 2) showed a significant relationship (F = 10.67 ; P<O.OI), with a correlation coefficient of -0.7783.Thus, panicles of Panicul1l racemosum that produced a smaller number of seeds allocated greater mass per seed.However, the relationship between seed mass and seed number per panicJe from cultived plants in glasshouse, was not significant (F = 0.517 ; P < 0.5), with a conelation coeficient of -0.155 .PWÚClIIll raC e 1l10JU/11 panicJes frolll a southern Brazilian fronla l dune population.
Soils of ali three habitats had very low concentrations of phosphate, nitrate and potassium, and a clear gradient of decreasing nutrient avail abi lity occurred from frontal to backdunes (Table 3).

Discussion
A marked variation in Panicum racemosum culm density and flowering culm density along the environmental gradient in southern Brazili an coastal foredunes was detected, confirming the observations made in the same area by Costa et aI. (1991).However, culm density and flowe ring culm density recorded in this study were substantially different from those in previous studies .Costa et aI. (1991) fo und culm densities of 75; 18.75 and 15.62 cul ms per square meter, and 28.82; 6.16 and zero flowering culms per square meter in frontal, top and backdunes , respectively.The inconsistencies with this study are probably because the earlier studies employed only a single permanent quadrat (2 x 2 m) in each habitat, and Panicum racemosum has a very heterogenous distribution in each habitat (Cordazzo & Seeliger 1993).Hence, the previous studies probably over-estimated the density in frontal dunes and underestimated the densities in top and backdunes.There may also be variation between years.
Like other dune-bui lding grasses such as Uniola paniculata (Wagner 1964), Ammophila arena ria (Huiskes 1979) and A. breviligulata (Maun & Baye 1989), Panicum racemosum allocated a small proportion of its resources to flowering culms, and plant vigour and density were COITelated with moderate sand accretion (Pfadenhauer 1980, Costa et aI. 1991).Flowering culms showed higher densities in habitats with active or recent sand accretion.Following dune stabili zation , the density and proportion of flowerin g culms diminished markedly , probably because of the decline of vigour of the population (Costa et aI. 1991).Only a small proportion of florets on a panicle contains normal well-developed caryopses.The reasons for this are obscure and need further investigation.Krajnyk & Maun (1982) suggested that the causes of low fertility in Ammophila breviligulata popul ations could be non-viable pollen , insect damage, sensitivity to environmental factors , shortage of pollen and meiotic abnormalities.Studies on seed production in another dune grass, Uniola paniculata also demonstrated low seed production (Hester & Mendelssohn, 1987), with an average of 2.55 ± 0.64 seeds per culm , which is even lower than that found in the Panicum racemosum population studied.Like most grasses, Panicum racemosum is wind-pollinated and each floret contains a single ovule and so can produce only one seed (caryopsis).One possible explanation for thi s extremely low seed production is that cross-pollination may be required for Panicul1l racem.osum to produce an appreciable number of seeds; this is not an unusual requirement among perennial grasses (Hubbard 1984, Schemske & Lande 1985).For cross-pollination, the species would require pollen from parents with a different genetic constitution, which may not be readily available because the population may have low genetic variability as a result of very extensive clonal growth (vegetative reproduction by rhizomes) (Pfadenhauer 1980) and rare establishment of new genetic individuaIs from caryopses (Cordazzo & Seeliger 1988a, Costa et aI. 1991).This hypothesis was supported by the results obtained in glasshouse experiments; panicles that received pollen from another genetic individu al showed a significant increase in seed production over plants in the field , whereas panicles that were bagged did not produce any seed at ali.This result also precludes apomixis, which occurs widely in the family Poaceae (Bashaw & Hanna 1990).A further factor that may affect successful pollination in the field is rai nfall, because a moderate shower virtually eliminates pollen grains from the atmosphere, and the flowerin g period in Panicum racemosum is very short (Cordazzo & Seeliger 1988 b) .
A large number of caryopses was wrinkled ; pericarp and testa remained empty despite an apparent initial growth.Emptiness of pericarp and testa may be caused by: (a) the abortion of Iess vigorous zygotes due to competition within the genetically eliverse zygote populations, similar to that which occurs in Ammophila breviligularo (Krajnyk & Maun 1982); (b) shrivelling of caryopses, a common occurrence in the family Poaceae, which coulel result from waterdeficits at the flowering stage (Shaw & Laing 1966, Ekanayake et aI. 1989), extremes of temperature at anthesis (Clary 1965) or meiotic abnormalities (Laing 1958); (c) selective elimination; by eliminating some seeds, the plant allocates more resources to vegetative growth, which is active in Panicum racemosum eluring spring-summer (Costaet ai.1984 );(d) an in-built strategy which leads to the proeluction of only a small number of relatively large seeds.
The relation between seeel mass anel number is subject to marked variation in some species (Salisbury 1942, Wlllff 1986, Michaels et ai. 1988), with a possible trade-off between seed size and seed number (Silvertown 1989).Sometimes there is a negative correlation between seed mass and number, sometimes a positive one (Sil vertown 1989), or none at alI.A plant that has ample nlltrients may invest more in seed resources than a plant experiencing a shortage of nlltrients bllt plants presllmab1y optimize their seed mass according to the amollnt of resources available to the maternal plant (Harper 1977).Although, coastal sand dunes are exposed to a constant nutrient input from salt spray (van eler Valk 1974), the capacity of the sandy soils to retain these nutrients is low, in the face of high leaching activity.The relatively eleficient nutrient status of southern Brazilian dunes may restrict seed number and seed size.Thus, uneler less favourable conditions of nutrient availability, especially during the later phases of reproeluction, the competition between the developing seeds for resources may lead to a negative correlation between their number and individual mass (Werner & Platt 1976, Wulff 1986).However, when nutrients are not limiting, as in the glasshouse experiments, the competition for resources was not detectable.The abortion of seeds may be of adaptive significance because a fixeel quantity of assimilates would be available for fewer ovules thus increasing or maintaining the size of the seed produced (Hawke & Maun 1989).So, the low percentage seed set and the negative correlation between seeel mass and seed number in Panicum racemosum ll1ay be of adaptive significance for two reasons.First, low seed set implies greater amounts of assimilates for maturing ovules, resulting in greater ll1ass per seed (Krajnyk & Maun 1982).Second, larger seeds may represent more advanced embryo development, with a subsequent reduction in the time taken to germinate (Marks & Truscott 1985), a greater possibility of ell1ergence from greater depth ofburial by sand (Maun & Lapierre 1986), anel faster growth anel greater survival probability of the seedlings (Zhang & Maun 1991).
Damage by a beetle ofthe family Anthicidae also accounts for a small reduction in the number of viable caryopses in the Panicum racemosum populations studied.
However, insect infestation of the panicles was not aelelressed in this study and ll1ay require further investigation .
In conclusion, the extremely low seed production in natural populations of Panicum racemosum appears to be due mainly to self-incoll1patibility; it requires cross-pollination to proeluce viable seeels.In southern Brazilian coastal dunes, clones may be very extensive giving little opportunity for out-breeding.Additionally, under conditions oflow nutrient availability, common in coastal sand dunes, the competition between sexual reproduction and allocation to clonal growth may lead to a negative correlation between the seed number and individual mass.Further studies are necessary to investigate the effect of pollination and nutrient supply on seed production in natural environments and examine variations in seed production among different populations of Panicum racemosum throughout its geographic distribution.

Figure I .
Figure I. Frequency di slribuli oll of seed mass in Pall; c lIIlI mL'eIllIlSlIlIl (n= 450) from a frontal natural population in coastal salld dune s in so uthern Brazil.

Table l .
Reproducti ve variation of Pal/icum racenzosul1l poplllations 01' Southern Brazil.Means ± SE are based on 100 samples in each area.

Table 2 .
Comparison of reproductive characteristics of Panicum racemosum from frontal natural population in southern Brazil with cross-and self-pollination plants cu lti vated in a glasshouse.Means ± SE are based on 100 samples from the natural habitat and 50 samples from glass house.

Table 3 .
Nutrient concentrations (means ± SE) of sands (5 cm depth) in three different habitats occupied by Panicum racemosum populations in Southern Brazilian coastal dunes