Diatom–environment relationships and a transfer function for conductivity in lakes of the Badain Jaran Desert, Inner Mongolia, China

Abstract

We describe a dataset of 26 modern diatom samples and associated environmental variables from the Badain Jaran Desert, northwest China. The influence of electrical conductivity (EC) and other variables on diatom distribution was explored using multivariate analyses and generalized additive modeling of species response curves. A transfer function was derived for EC, the variable with the largest unique effect on diatom variance, as shown by partial canonical correspondence analysis. Weighted-averaging partial least squares regression and calibration provided the best model, with a high coefficient of determination ( $$ {\text{r}}_{\text{boot}}^{2} $$ = 0.91) and low prediction error (RMSEPboot = 0.136 log10 μS cm−1). To assess its potential for palaeosalinity and palaeoclimate reconstructions, the EC transfer function was applied to fossil diatom assemblages from 210Pb-dated short sediment cores collected from two subsaline lakes of the Badain Jaran Desert. The diatom-inferred (DI) EC reconstructions were compared with meteorological data for the past 50 years and with remote sensing data for the period AD 1990–2012. Changes in DI–EC were small and their relationship with climate was weak. Moreover, remote sensing data indicate that the surface areas and water depths of these lakes did not change, which suggests that water loss by evaporation is compensated by groundwater inflow. These results suggest that the response of these lakes to climate change is mediated by non-climatic factors such as the hydrogeological setting, which control recharge from groundwater, and may be non-linear and non-stationary.