Principal component analysis of the power spectra of high-resolution ice core records

High-resolution ice core records allow for analysis of variability on short timescales (annual, sub-annual, decadal) in addition to longer timescales (centennial, millennial), as well as how variability on different timescales changes across time.  Spectral analysis of these time series data is used to evaluate the amplitudes of signals in the records at particular frequencies. Principal component analysis (PCA) is a technique for dimensionality reduction while retaining the maximum possible variance from the original data. PCA can be used with Empirical Orthogonal Functions (EOFs) to identify the dominant spatial patterns (EOFs) with their corresponding time variations (PCs). Applying this technique to the spectra of ice core records will help to explore the spatial patterns of variability in the frequency domain by identifying the dominant modes of spectral variability (PCs) and the spatial pattern of this variability across an ice sheet (EOFs).
The power spectra will be produced for a suite of high-resolution ice core water isotope records, with methodological choices for resampling resolution being informed by results from tests with synthetic data. PCA will then be applied to the dataset of power spectra, where each spectrum is an observation and the different frequency bins are the variables. This analysis will be applied to the Holocene section of the ice cores in particular, to create a more comprehensive picture of the high frequency variability on short timescales (annual, sub-annual, decadal) and regional climate dynamics. The Holocene-only focus has the ability to resolve high-frequency signals that are often lost in older ice due to ice thinning and diffusion. Further, the relatively stable Holocene climate will allow for a more focused study of the regional mechanisms in the Arctic that operate on shorter timescales.