Optimising the recovery of water isotope signals in deep ice cores

Understanding the Earth's climate is a complex and challenging task, especially considering that direct observations of climate parameters have only been documented relatively recently. Information on longer timescales has to be gleaned from proxies, preserved physical characteristics which serve as natural recorders of the climate. One such proxy is the isotopic composition of water molecules stored as snow and ice in ice sheets, as they relate to past atmospheric temperatures. The retrieval of ice cores from Earth's largest ice sheets, Greenland and Antarctica, provides continuous isotopic records dating back up to hundreds of thousands of years. However, this information is not perfectly preserved within the ice sheet, due to the constant random motion of the water molecules. This movement, called diffusion, homogenises the initial climate signal, generating a smoothing effect that dampens high-frequency variability. Additional isotopic perturbations are introduced during the sampling and measurement processes, further obscuring the desired climate information.

Consequently, for accurate interpretation of the proxy, these effects must first be considered. Diffusion can be corrected for, but the efficacy of such a correction depends on two key factors. Firstly, the magnitude of diffusion (or "diffusion length") should be accurately known, to reliably amplify the attenuated frequencies without overcompensating or underestimating the effect. The diffusion length can be determined through models constrained using knowledge of the diffusive process and the physical parameters, or it can be statistically estimated from the isotopic data itself through inspection of the variability on different timescales. Secondly, it is important to obtain precise water isotope measurements, as the correction process cannot separate the climate signal from the measurement noise. High frequency information which has undergone significant diffusion can be dominated by this subsequently added measurement noise, rendering the signal irretrievable and limiting the maximum obtainable resolution. This dissertation investigates both of these factors, and explores the additional effects of discrete sampling, with a focus on deep ice cores.

In this thesis, an adapted method for statistically estimating the diffusion length in deep ice is described, which assumes more realistic climate variability structure than conventional approaches. The method is applied to the deepest, oldest isotopic data from the EPICA Dome C ice core and finds a notable reduction in diffusion length compared to previous estimates, in closer agreement with the physical models. Analysis of the resolution limitation imposed by measurement noise is performed, resulting in a method to calculate the potentially recoverable timescales of an ice core for different measurement precisions. Application of the method to the Beyond EPICA - Oldest Ice Core suggests 10,000 year cycles are realistically retrievable from 1.5 million year old ice if care is taken to optimise precision. Expressions outlining the quantitative effect of discrete sampling are also derived, and simulations reveal how sample size impacts the representativeness of the isotopic record and the accuracy of diffusion length estimations. The results suggest a sample size equal to half the diffusion length is sufficient for both purposes. The thesis concludes with a discussion of all the results and their collective implications, detailing the necessary considerations for optimal water isotope signal recovery from ice cores.