Construction of consistent ice core accumulation time series from large-scale meteorological data: Development and description of a regression model for one North Greenland ice core

Abstract

An empirical downscalling relationship is constructed which consistently estimates annual ice accumulation from large-scale meteorological data. The method by which the relationship was developed and the application of the method to one ice core are described. The statistical technique is based on a stepwise multiple linear regression using yearly accumulation as the predictand and the principle components (PCs) of seasonally averaged large-scale atmospheric fields as predictors. In order to separate the accumulation variations related to circulation and thermodynamics, the first step involves the stream-function field as the predictor. In the second step, temperature PCs are related to the residuals between real ice accumulation and accumulation described by stream-function PCs. One model is fitted for a North Greenland ice core. The atmospheric data are monthly NCEP Reanalysis data from 1948 to 1992. A statistical relationship is found which reproduces about 71.5% of local ice accumulation variability. The relationship involves 3 physically plausible stream-function patterns, representing a seasonal mean (May to August) over Greenland, the North Atlantic and North Europe, which describes 64% of variance. In the second step, an additional contribution of the temperature field to the explained variance of 7.5% is achieved. The temperature PCs represent the annual mean 700 hPa pattern covering the area from east Canada to east Greenland.