Enhancing the West Antarctic Meteorological Record With Artificial Neural Networks

Improved interpretation of the ever growing body of ice-core-based paleoclimate records from Antarctica requires a deeper understanding of Antarctic meteorology. New field campaigns and improved numerical forecasting models will ultimately provide long-term benefits but neither addresses the existing observational archive. In contrast, our work with automatic weather station (AWS) data addresses this issue directly. AWS currently provide the only year-round, continuous direct measurements of weather on the ice sheet. As the spatial coverage of the network has expanded year to year (thanks to C. Stearns and his University of Wisconsin AWS group), so has our meteorological database. Unfortunately, many of the records are relatively short (less than 10 years) and/or incomplete (to varying degrees) due to the vagaries of the harsh environment. Climate downscaling results in temperate latitudes suggest it is possible to use GCM-scale meteorological data sets (e.g., ECMWF reanalysis products) to both fill gaps in the AWS records and extend them back in time to create a uniform and complete database of West Antarctic surface meteorology (at AWS sites). Such records are highly relevant to the improved interpretation of the expanding library of snow-pit and ice-core data sets. Our solution uses artificial neural network (ANN) techniques to predict the near-surface meteorology recorded by AWS instruments (e.g., temperature, pressure) using large-scale features of the atmosphere (e.g., 500 mb geopotential height) from a region around the AWS. ANNs are trained to predict observed AWS data from the corresponding GCM-scale data. Intrayear prediction (of observations in the training year) has been very successful (e.g., RMS errors < 2 mbar for pressure). Interyear prediction (of observations not in the training year) remains a work-in-progress (e.g., RMS errors are 4-5 mbar). Three ANN architectures yield similar results suggesting our approach is valid but our training methodology needs refinement. These results support high confidence in the ANN-based predictions from the GCM-scale data for periods where AWS data are unavailable, e.g., before installation. ANNs thus provide a means to expand our surface meteorological records significantly in West Antarctica.