Niels Bouton

The destruction of an endemic species flock: quantitative data on the decline of the haplochromine cichlids of Lake Victoria

SynopsisThe Lake Victoria fish fauna included an endemic cichlid flock of more than 300 species. To boost fisheries, Nile perch (Lates sp.) was introduced into the lake in the 1950s. In the early 1980s an explosive increase of this predator was observed. Simultaneously, catches of haplochromines decreased. This paper describes the species composition of haplochromines in a research area in the Mwanza Gulf of Lake Victoria prior to the Nile perch upsurge. The decline of the haplochromines as a group and the decline of the number of species in various habitats in the Mwanza Gulf was monitored between 1979 and 1990. Of the 123+ species originally caught at a series of sampling stations ca. 80 had disappeared from the catches after 1986. In deepwater regions and in sub-littoral regions haplochromine catches decreased to virtually zero after the Nile perch boom. Haplochromines were still caught in the littoral regions where Nile perch densities were lower. However, a considerable decrease of species occurred in these regions too. It is expected that a remnant of the original haplochromine fauna will survive in the littoral region of the lake. Extrapolation of the data of the Mwanza Gulf to the entire lake would imply that approximately 200 of the 300+ endemic haplochromine species have already disappeared, or are threatened with extinction. Although fishing had an impact on the haplochromine stocks, the main cause of their decline was predation by Nile perch. The speed of decline differed between species and appeared to depend on their abundance and size, and on the degree of habitat overlap with Nile perch. Since the Nile perch upsurge, the food web of Lake Victoria has changed considerably and the total yield of the fishery has increased three to four times. Dramatic declines of native species have also been observed in other lakes as a result of the introduction of alien predators. However, such data concern less speciose communities and, in most cases, the actual process of extinction has not been monitored.

A procedure to extract phylogenetic information from morphometric data

To reconstruct a phylogeny, data with sufficient resolving power are necessary. When morphological or molecular data fail this criterion, other data types are needed. Morphometric characters can sometimes offer a solution. However, environmental effects due to differences in location have to be eliminated or at least reduced before these characters can be applied in phylogeny reconstruction. Here we present a new procedure to extract phylogenetic information from morphometric characters. The procedure is based on assigning weights to these characters. The procedure starts with a principal component analysis of the morphometric measurements. Then an analysis of variance (ANOVA) is done on principal components scores in order to determine species effects, environmental effects due to differences in location, and interaction effects. Subsequently, components are chosen for weighting characters on the grounds of two criteria. First, there must be only a significant species effect on their scores. Second, they have to account for a large proportion of total variance. The measurements are discretized by the homogeneous subset coding procedure. The discretized characters are weighted by means of the proportions of variance that are contributed to the chosen PCs. We used head measurements of haplochromines to test the procedure. Evaluations indicated that our procedure is capable of extracting phylogenetic information from morphometric data.

THE STATISTICAL SIGNIFICANCE OF DIETS AND OTHER RESOURCE UTILIZATION PATTERNS

The assessment of the statistical significance of differences in diets is surprisingly rare. In this note we discuss a powerful randomization test, that can be applied to diets as well as to other patterns of resource utilization and, for example, to patterns of species composition. We argue that ecological questions often require that both between-group overlaps and within-group overlaps are compared with their expected distribution under the null-hypothesis of no differences between groups.