Assessment of a diatom-based transfer function for the reconstruction of lake-level changes in Boqueirão Lake, Brazilian Nordeste

Abstract

Transfer functions have been developed to quantify the relations between organism assemblages and environmental variables and to apply these models to reconstruct past conditions. We present an assessment of a transfer function developed to reconstruct the depth of Boqueirão Lake in the Brazilian Northeast region. This region is affected by the Intertropical Convergence Zone and El Niño that create rainy and drought periods and that affect lake water level and can cause algal community compositional shifts. To reconstruct an algal depth profile of Boqueirão Lake, we used diatom calibration data-sets from 72 lake superficial sediment samples. After excluding rare species (<1%), we developed a transfer function model using sixty species, almost all exclusively benthic. Five environmental variables (silt, clay, pH, dissolved oxygen and depth) were also selected. Partial Canonical Correspondence Analysis indicated that silt and depth have the highest shared effect in terms of partially explained variance and λ1/λ2. The reduced set of explanatory variables captured 14.3% of the total variance in diatom distributions. We assess six models to reconstruct past lake levels. No one model generated the best results in all aspects of transfer functions generated to map the water depth, but all reconstructions fall within the range of variation of the system. Although their results did not converge to a specific depth, they converged for depicting periods of lowered water levels or periods of floods. This convergence suggests that the diatom-based transfer function for the reconstruction of the paleodepth of Boqueirão Lake is reliable.