Testing alternative hypotheses for the decline of cichlid fish in Lake Victoria using fish fossils time series from sediment cores

Lake Victoria is well known for its high diversity of endemic fish species that provide livelihoods for millions of people. The lake garnered widespread attention during the twentieth century as major environmental and ecological changes modified the fish community with the extinction of ~40% of endemic cichlid species by the 1980s. Suggested causal factors include anthropogenic eutrophication, fishing, and introduced non-native species but their relative importance remains unresolved because monitoring data started in the 1970s when changes were already underway. Here, for the first time, we reconstruct two time series, covering the last ~200 years, of fish assemblage using fish teeth preserved in lake sediments. Two sediment cores Lake Victoria (Mwanza Gulf), were subsampled continuously at intra-decadal resolution, and teeth were identified to major taxa: Cyprinoidea, Haplochromini, Mochokidae, and Oreochromini. None of the fossils could be confidently assigned to non-native Nile Perch. Our data show significant decreases in haplochromine and oreochromine cichlid fish abundances began long before Nile Perch's arrival, while cyprinoids have generally been increasing. Our study is the first to reconstruct a time series of fish assemblage in Lake Victoria extending deeper back in time than the past 50 years, helping shed light on processes underlying Lake Victoria's biodiversity loss.

Methods
Two sediment cores (SC9 and SC14) were collected in 2018 from the Mwanza Gulf of Lake Victoria, Tanzania. The cores were taken from 12.5 m and 14.5 m water depths and were subsampled contiguously at 1-cm intervals and macrofossil analysis. Once the fossils were recovered, we calculated the fossil concentration (number of fossils divided by sediment volume). To identify temporal change points in the estimates of fossil teeth concentration (teeth cm3), we used the segmented package v1.6-4 in R (Regression Models with Break-Points and Change-Points Estimation) (Muggeo, 2008) and Spearman’s correlation to analyze the trends.