Recent Climate in Greenland Through Ice Cores and Self-organizing Maps

Recent years have seen a large increase in the availability of high-quality ice cores from the Greenland Ice Sheet (GIS) and the subsequent development of many new high-resolution proxy climate records. Here we apply a relatively new, nonlinear approach using self-organizing maps (SOMs) to study the spatial and temporal variability seen in accumulation records from 50-plus GIS sites covering part or all of the period 1957-2002. SOMs provide an unsupervised classification of complex geophysical data sets, e.g., time series of the atmospheric circulation or sea-ice extent, into a fixed number of distinct generalized patterns, or modes, that represent the probability density function (PDF) of the input data. These patterns collectively provide a nonlinear classification of the continuum of the PDF into a two-dimensional, spatially organized grid form. In contrast to principal component analysis, SOMs do not force orthogonality or require subjective rotations to produce interpretable patterns. Results from analyses of annual accumulation show that the SOM readily captures the high spatial diversity seen in climate records from the GIS, including nonlinear gradients in latitude and elevation. For example, sites in the northern and central regions (e.g. Humboldt, NASA-U) tend to be unrelated to sites in the south and east (e.g., Das1, STUNUA). Sites within the south/southeast also show richer patterns of variability than simply above or below average accumulation. Understanding these relationships, and the spatial complexity of Greenland's climate, is key to improving our ice core-based reconstructions of past climate and projecting possible future changes in the GIS. The SOM algorithm is widely held to be robust in the presence of incomplete input data, a characteristic typical of multi-site/project ice core analyses. Here we examine the sensitivity of the SOM-derived accumulation patterns and reconstructed site records to changes in the number of sites and record lengths. This explores the information available from this climate dataset and gives insight to which sites "matter" the most in understanding ice sheet history (as well as testing the analysis technique itself). When combined with ongoing SOM-based analyses of the atmospheric circulation, we anticipate new insights into the complex climate of this region, including relationships with phenomena such as the North Atlantic Oscillation/Arctic Oscillation.