A quantitative inference model for conductivity using non-marine ostracode assemblages on San Salvador Island, Bahamas: Paleosalinity records from two lakes

Abstract

Quantitative records of past environments are needed to understand natural variability in ecosystems and their responses to climate change. Changes in ostracode assemblages through time can provide such records as ostracode species are sensitive to changes in their local environments. Before they can be used to indicate past environments, however, is it necessary to understand how distributions of assemblages change across environmental gradients. To that end, thirty-two lakes on San Salvador Island, Bahamas were sampled for both ostracodes and nineteen limnological variables. Multivariate fuzzy set ordination indicates that change in ostracode assemblages is significantly and independently correlated with three environmental variables: electrical conductivity (salinity), dissolved oxygen and alkalinity. A transfer function was created to reconstruct past conductivity since changing conductivity of lakes on San Salvador is influenced by changes in climate and sea-level. A 2-component weighted-averaging partial least squares model performed best as a transfer function for conductivity with an apparent r2 of 0.76 and an r2 of 0.69 between observed and predicted conductivity, as assessed by leave-one-out cross validation. The resulting transfer function was then applied to two mid- to late-Holocene sediment cores from which ostracode assemblages were sampled. The late Holocene conductivity records show that changes in conductivity of lakes on San Salvador are broadly synchronous with times of enhanced ENSO activity corresponding to elevated conductivity in response to lower Atlantic hurricane occurrence. These results demonstrate that changing ostracode assemblages through time provide a reliable means to reconstruct past salinity and demonstrates the strong effect of ENSO on Bahamian aridity.