Ocotea

Abstract: Seed fall in a Brazilian subtropical lower montane, moist forest was measured over a year using 120 traps that had a total collecting surface of 174 M2. Twenty-two thousand seeds from 227 plant species were collected. Wind-dispersed species released seeds mainly during the late dry season and early wet season. Among nonwind-dispersed species, large-seeded species matured seeds primarily during the wet season. Small-seeded species were much less seasonal. This difference in seasonality may stem from a correlation between seed size and mode of animal dispersal. Arrival probability per collection period, approximated as the percentage of traps hit by free seeds of each species, was regressed against several independent variables. For nonwind-dispersed species, the logarithm of seed area best predicted arrival probability of free seeds, being inversely related to it. Existence of this inverse relationship has often been supposed but only now demonstrated. THE REPRODUCTIVE BIOLOGY of any seed-plant population is believed to reflect evolutionary compromise among various conflicting aspects of reproduction (Harper et al. 1970, Stebbins 1971). Of these, seed size is fundamental because it apparently influences many other aspects of reproduction. This paper examines relationships in a tropical forest between seed size and two other reproductive parameters: the temporal and spatial extent of seed fall. Numerous workers have assumed that small seeds are more widely dispersed than large ones (e.g., Harper et al. 1970), yet substantial data to test this idea have been lacking. Because of the great temporal and spatial spans over which seed dispersal occurs, the dispersal process at the community level is likely to remain intractable to direct study. Instead of attempting to measure dispersal, I measured arrival probabilities of free seeds at fixed locations. Arrival probability is clearly a crucial component of dispersibility. Seed size has also been hypothesized to influence the temporal pattern of seed fall in tropical forests. Smythe (1970) reported fruiting of largeseeded species in a Panamanian forest to be synchronous, whereas small-seeded species were less seasonal in fruiting time. I sought to determine whether such temporal patterns occur in another neotropical forest. STUDY SITE AND METHODS The study was conducted in a virgin evergreen forest at 800-900 m elevation in Reserva Biologica Nova Lombardia, Municipio de Santa Tereza, Espirito Santo, Brazil, at latitude 19? 53' S. The site is surrounded by several square kilometers of old-growth forest, although some agricultural clearings exist less than 2 km away. Jackson (1978) provides climatic data for the site. Reserva Nova Lombardia experiences a moderate temperature differential between winter (June-August) and summer (December-February) and a long but not severe dry season (April-September). In the Holdridge Life Zone System (Holdridge 1967) the vegetation is classified as Subtropical Lower Montane Moist Forest. As measured by frequency along line transects, the commonest trees are Manilkara elata (Sapotaceae), Lucumza laurifolia (Sapotaceae), Campomazesia warmingiana (Myrtaceae), Guarea petiolulata (Meliaceae), Emimotuni zitens (Icacinaceae), Didymnopanax longipetiolatuxm (Araliaceae), Nectandra puberula (Lauraceae), Psidiuni macrospermnuni (Myrtaceae), Ocotea declinata (Lauraceae), Andira stipulacea (Leguminosae), Clusia grandiflora (Guttiferae), Terminalia ijanuariensis (Combretaceae), Leandra alpestris (Melastomaceae), Sideroxylon gardnerianum (Sapotaceae), Cybianzthus cunieifolius (Myrsinaceae), Ocotea nmacropoda (Lauraceae), Copaifera langsdorffi (Leguminosae), and Ficus doliaria (Guttiferae) (A. Ruschi, pers. comm.). Falling seeds and fruits were sampled with 120 horizontally placed rectangles of green, plastic window screen. Each corner of a screen was secured with plastic cord to a vertical bamboo stake at one meter above the ground. Screens measured 1.4x1.2 m when flat (area 1.68 Mi2), but trap sag reduced the average collecting surface per trap to 1.44 m2?. The mesh in the screen was 1.2 mm square. The seed traps were set in two straight lines approximately 200 m apart. Sixty traps comprised each transect, with 6 m between adjacent traps. One transect began at midslope on a hill (836 m elevaBIOTROPICA 13(2): 121-13