Structure of a Tropical Stream Fish Community: A Role for Biotic Interactions

Abstract

We surveyed the fish community of a tropical watershed in the Northern Range Mountains of Trinidad and conducted experiments to test the hypothesis that species interactions structure the fish community. We censused the fish community at 86 sites, measuring each fish collected and recording physical variables for each site (pool size, substrate composition, cover, etc.) One member of the fauna, a killfish, Rivulus hartii, was widely distributed in the drainage, occurring alone in each headwater but also encountering other fish species below barrier waterfalls. This pattern allowed us to use Rivulus to test four predictions derived from the species interaction hypothesis: (1) Rivulus density (in grams per square metre) would be depressed at sites with other fish species present, when compared to the density expected from physical characteristics of each site as if Rivulus were alone; (2) Rivulus density would decline at points of contact with the piscivorous fish Hoplias malabaricus; (3) local populations of Rivulus found in pools occupied by the predator would show a smaller maximal size than in pools without the predator; and (4) the growth rate of Rivulus would be lowered in the event of a local invasion by the guppy, Poecilia reticulata, another abundant and sometimes co—occurring species in the watershed. The first three predictions were tested with data from the survey, while the fourth prediction was tested by experiments in a nearby stream. We used a factor analysis to reduce the dimensionality of the physical variables to four factors. A regression analysis of Rivulus density on the four factors, using 43 Rivulus—only pools, was used to predict Rivulus density in nine pools, each in a different tributary, in which Rivulus first encountered a second species after being the only species present upstream of this pool. The expected density was based solely on the physical characteristics of the pool and represented the expected Rivulus density purely from physical attributes, as if Rivulus were the only species present. Comparison of these expected densities with the actual Rivulus densities revealed a consistent negative impact of the presence of other species on Rivulus density, in support of Predition 1. Overall, actual Rivulus densities at points of contact with other species were about one—third of the density expected if the other species were absent. The distribution of Rivulus was essentially complementary to that of the piscivorous fish, although only one site provided unambiguous test of Prediction 2, since Hoplias did not ascend the tributaries in the watershed. The size distribution of Rivulus in populations occurring in pools with predators was dominated by small individuals, with none larger than the median size of Rivulus in predator—free populations, supporting Prediction 3. Finally, the competition experiments revealed that adult Poecilia depressed the growth rate of young Rivulus. We conclude that interspecific interactions are important in structuring the geographical distribution and abundance of Rivulus throughout this tropical watershed, and suggest that interspecific interactions can fragment Rivulus into local populations.