Skilful decadal predictions of subpolar North Atlantic SSTs using CMIP model-analogues

Matthew B Menary,Jon Robson,Juliette Mignot

doi:10.1088/1748-9326/ac06fb

Abstract

Predicting regional climate variability is a key goal of initialised decadal predictions and the North Atlantic has been a major focus due to its high level of predictability and potential impact on European climate. These predictions often focus on decadal variability in sea surface temperatures (SSTs) in the North Atlantic subpolar gyre (NA SPG). In order to understand the value of initialisation, and justify the high costs of such systems, predictions are routinely measured against technologically simpler benchmarks. Here, we present a new model-analogue benchmark that aims to leverage the latent information in uninitialised climate model simulations to make decadal predictions of NA SPG SSTs. This system searches through more than one hundred thousand simulated years in Coupled Model Intercomparison Project archives and yields skilful predictions in its target region comparable to initialised systems. Analysis of the underlying behaviour of the system suggests the origins of this skill are physically plausible. Such a system can provide a useful benchmark for initialised systems within the NA SPG and also suggests that the limits in initialised decadal prediction skill in this region have not yet been reached.

Highlights

As the global climate continues to change in response to anthropogenic influences (Bindoff et al 2013), estimates of how climate might evolve in the near term and on a regional level are becoming increasingly important (Kushnir et al 2019)
We focus on decadal timescales as the ability of simple models to largely capture the dynamics of the North Atlantic subpolar gyre (NA subpolar gyre (SPG)) on these timescales (Born et al 2015), should reduce the degrees of freedom and make the search for good analogues more attainable (Dool 1994)
In addition to the SPG, the skill of the modelanalogue system remains high globally (figure 4(a)). This is despite the fact that the system is (a) only choosing simulation segments based on their spatial similarity to observations within the North Atlantic, (b) does not include any information about the timing of external forcings, and (c) that this analogue design was chosen for its potential skill in the NA SPG only

Summary

Introduction

As the global climate continues to change in response to anthropogenic influences (Bindoff et al 2013), estimates of how climate might evolve in the near term and on a regional level are becoming increasingly important (Kushnir et al 2019) These estimates can take the form of projections of the century using different shared socioeconomic pathways from the 6th Coupled Model Intercomparison Project (CMIP6) (Eyring et al 2016). Previous work has shown that sea surface temperatures (SSTs) in the North Atlantic (NA) including the subpolar gyre (SPG) can impact climate both locally and remotely (Sutton and Hodson 2005, Monerie et al 2018) They are potentially predictable on decadal timescales (Collins et al 2006) and successfully initialising them can provide skill elsewhere (Dunstone et al 2011). Technologically simpler benchmarks are a useful tool to help quantify the uncertainty in climate model projections/predictions (Brunner et al 2020)

Objectives

Methods

Results

Conclusion