Abstract. Ice-core-based records of isotopic composition are a proxy for past
temperatures and can thus provide information on polar climate variability
over a large range of timescales. However, individual isotope records are
affected by a multitude of processes that may mask the true temperature
variability. The relative magnitude of climate and non-climate contributions
is expected to vary as a function of timescale, and thus it is crucial to
determine those temporal scales on which the actual signal dominates the
noise. At present, there are no reliable estimates of this timescale
dependence of the signal-to-noise ratio (SNR). Here, we present a simple
method that applies spectral analyses to stable-isotope data from multiple
cores to estimate the SNR, and the signal and noise variability, as a
function of timescale. The method builds on separating the contributions from
a common signal and from local variations and includes a correction for the
effects of diffusion and time uncertainty. We apply our approach to firn-core
arrays from Dronning Maud Land (DML) in East Antarctica and from the West
Antarctic Ice Sheet (WAIS). For DML and decadal to multi-centennial
timescales, we find an increase in the SNR by nearly 1 order of magnitude
(∼0.2 at decadal and ∼1.0 at multi-centennial scales). The estimated
spectrum of climate variability also shows increasing variability towards
longer timescales, contrary to what is traditionally inferred from single
records in this region. In contrast, the inferred variability spectrum for
WAIS stays close to constant over decadal to centennial timescales, and the
results even suggest a decrease in SNR over this range of timescales. We
speculate that these differences between DML and WAIS are related to
differences in the spatial and temporal scales of the isotope signal,
highlighting the potentially more homogeneous atmospheric conditions on the
Antarctic Plateau in contrast to the marine-influenced conditions on WAIS. In
general, our approach provides a methodological basis for separating local
proxy variability from coherent climate variations, which is applicable to a
large set of palaeoclimate records.