Inverse stochastic–dynamic models for high-resolution Greenland ice core records

Niklas Boers,Anders Svensson,Mickael D Chekroun,Michael Ghil,Dmitri Kondrashov,Denis-Didier Rousseau,Honghu Liu,Matthias Bigler

doi:10.5194/esd-8-1171-2017

Abstract

Abstract. Proxy records from Greenland ice cores have been studied for several decades, yet many open questions remain regarding the climate variability encoded therein. Here, we use a Bayesian framework for inferring inverse, stochastic–dynamic models from δ18O and dust records of unprecedented, subdecadal temporal resolution. The records stem from the North Greenland Ice Core Project (NGRIP), and we focus on the time interval 59–22 ka b2k. Our model reproduces the dynamical characteristics of both the δ18O and dust proxy records, including the millennial-scale Dansgaard–Oeschger variability, as well as statistical properties such as probability density functions, waiting times and power spectra, with no need for any external forcing. The crucial ingredients for capturing these properties are (i) high-resolution training data, (ii) cubic drift terms, (iii) nonlinear coupling terms between the δ18O and dust time series, and (iv) non-Markovian contributions that represent short-term memory effects.

Highlights

Data-driven stochastic difference equation models have recently been successfully applied to a wide range of climatic phenomena (Kondrashov et al, 2005, 2006; Kravtsov et al, 2005, 2009)
As a training set for the parameter optimization of our stochastic–dynamic model in Eq (5), we choose the time interval 59–22 ka b2k; this interval roughly coincides with Marine Isotope Stage 3
We have shown that a coupled, two-dimensional stochastic– dynamic model with cubic drift term and linear delay terms is capable of reproducing the statistical properties of δ18O and dust time series derived from the high-resolution North Greenland Ice Core Project (NGRIP) record for the interval 59–22 ka b2k, which roughly corresponds to Marine Isotope Stage 3

Summary

Introduction

Data-driven stochastic difference equation models have recently been successfully applied to a wide range of climatic phenomena (Kondrashov et al, 2005, 2006; Kravtsov et al, 2005, 2009). The specific functional form of F may use some a priori knowledge of the system under study, while the parameters of the proposed model are always inferred by training it on a given set of time series produced by the system. In this sense, the approach is semiempirical, rather than being entirely hypothesis-free. Most existing methodologies for empirical-model derivation are based on least-squares fitting to determine optimal parameters for F

Results

Discussion

Conclusion