Application of artificial neural networks (ANN) to high-latitude dinocyst assemblages for the reconstruction of past sea-surface conditions in Arcticand sub-Arctic seas

The artificial neural network (ANN) method was applied to dinoflagellate cy st (dinocyst) assemblages to estimate palaeoceanographical conditions. The ANN method was adapted to three distinct data bases covering the northern N orth Atlantic (N = 371), plus the Arctic seas (N = 540) and the Bering Sea (N = 646). The relative abundance of 23 dinocyst taxa was calibrated agains t hydrographic variables (sea-surface temperature, salinity and density in February and August, and seasonal extent of sea-ice cover) using ANNs. The estimation of hydrographical parameters based on an ANN yields high coeffic ients of correlation between observations and reconstructions for each vari able selected. The validation tests performed on the different data bases s uggest more accurate calibration at the scale of the North Atlantic and Arc tic (N = 540) than on a multibasin scale, i.e. when including the subpolar North Pacific (N = 646). The ANN calibrations and the modern analogue techn ique (MAT) have been applied to two sequences from the northwest North Atla ntic spanning the past 25 000 yr for the purpose of comparison. Both approa ches yielded similar results, generally within the range of their respectiv e uncertainties, demonstrating their suitability. The main discrepancies ge nerally correspond to assemblages with poor modern analogues for which we h ave to admit a higher degree of uncertainties in the reconstruction, whatev er the approach used. Copyright (C) 2001 John Wiley & Sons, Ltd.