Relating West Antarctic Ice Cores to Climate with Artificial Neural Networks

This NSF funded proposal focused on three areas: (1) filling gaps in and extending the length of automatic weather station (AWS) records using an objective technique based on feed-forward artificial neural networks (FF-ANNs) and atmospheric reanalysis data from ECMWF (ERA-15); (2) extracting generalized circulation patterns from ERA-15 annual fields using self-organizing maps; (3) relating ice core data to the patterns found in (2) using FF-ANNs so as to reconstruct past circulation from ice core data.

Part 1 resulted in complete 15-year records (merged observations and FF-ANN-based predictions) at 6-hourly resolution for these six AWS for 1979-1993: Siple Station, Byrd, Lettau, Elaine, Marilyn and Ferrell. As these datasets are a synthesis of previously existing observations (available from the University of Wisconsin, uwamrc.ssec.wisc.edu/) and model-based predictions they are not posted here (but they are available from the author on request).