FOREST FRAGMENTATION, POLLINATION, AND PLANT REPRODUCTION IN A CHACO DRY FOREST, ARGENTINA

In a fragmented, dry subtropical forest in northwestern Argentina, we compared pollination levels, fruit set, and seed set among small (<1 h a) forest fragments, large (>2 ha) fragments, and continuous forest in 16 plant species representing a wide range of pollination systems, br eeding systems, and growth forms. For three species, Prosopis nigra (M imosoideae), Cercidium australe (Caesalpinoideae), and Atamisquea emar ginata (Capparaceae), the three treatments were replicated across four sites; we achieved less replication for other species. Because compar isons between forest and fragment populations for different species to ok place in different sites, however, by treating all 16 species as a unit we lessened the potential bias of confounding site effects and co uld evaluate the overall impact of fragmentation. Significant or margi nal (P < .10) fragmentation-related declines in number of pollen tubes per style, fruit set, and seed set occurred in 9 of 16, 5 of 15, and 3 of 14 species, respectively. Overall, significant or nonsignificant declines occurred in 81%(pollen tubes), 73% (fruit set), and 79% (seed set) of the species. In all cases these proportions were greater (P l ess than or equal to.06) than the null binomial expectation of a 1:1 r atio of increases to decreases. Breeding system did not explain sensit ivity to fragmentation: the magnitudes of declines in pollen tubes, fr uit set, and seed set were virtually indistinguishable between self-co mpatible and self-incompatible species. At least 4 of the 10 self-inco mpatible species, however, were heavily visited in small fragments by Africanized honey bees, which may have compensated for a decline in vi sits by native pollinators. The exact nature of responses varied among plant species. In some, the absolute quantity of pollen grains transf erred to stigmas decreased with fragmentation, and sometimes this was reflected in reduced fruit or seed set. In Cercidium, Prosopis, and At amisquea, the quality of the grains transferred apparently changed: nu mber of pollen tubes produced per pollen grain on the stigma declined with increasing fragmentation, and at least in the latter two species seed production declined as well. Overall, levels of pollination and s eed production undoubtedly integrated many idiosyncratic effects of fr agmentation on particular plant and animal populations, and indicated that ''community health'' of fragments suffered in comparison with tha t of continuous forest. Median decreases in pollination levels and see d output from forest to fragments approached 20%. The impact of these declines on plant recruitment is less clear, however, because cattle g razing and trampling of seedlings and saplings in fragments may consti tute a much more serious short-term conservation problem.