Bringing Together Simulated ~20 Year Variability in Coupled Climate Models

Thursday, 18 December 2014
Matthew Menary1,2, Dan Hodson2, Jon Robson2, Rowan Sutton3 and Richard A. Wood1, (1)Met Office Hadley center for Climate Change, Exeter, United Kingdom, (2)University of Reading, Reading, United Kingdom, (3)University of Reading, Reading, RG6, United Kingdom
Many climate models simulate significant spectral power in large scale, North Atlantic subpolar gyre indices at timescales of around 20 years. Despite similar periodicities, the underlying mechanisms reported in individual models can vary greatly. For example, the timescale can be set by any combination of geostrophic self advection, Rossby wave propagation, or advection by the mean circulation. The role of the overturning circulation can either be active or passive, and the ultimate driver of density changes in the deep water formation regions is split roughly evenly in the literature between salinity and temperature – with implications for any feedback mechanisms. These simulations typically span many centuries with constant external forcings to capture internal climate variability. The extent to which either this periodicity carries over to the real world under increasingly strong external forcing, or which, if any, of the modelled feedbacks are applicable, is unclear.

We present new results from a state-of-the-art high resolution coupled climate model (HadGEM3) in which the mechanism of internal decadal variability in the North Atlantic is diagnosed and discuss the causes of differences from previous work. Due to the non-linear equation of state, biases in the simulated mean state can explain some of the inter-model differences via the relative importance of temperature or salinity in density changes. These biases can then propagate throughout the mechanistic chain resulting in fundamentally different simulated mechanisms. For example, whether temperature or salinity control densities in the Labrador Sea influences whether a strengthening overturning circulation acts as a negative or positive feedback.

Although analysis of the model proceeds via lagged regressions, this is generally not possible with observational data. We use a combination of palaeo reconstructions and targeted process-based analysis to investigate whether there is any signal of bidecadal periodicity within the real world subpolar gyre and to what extent our simulated mechanisms exist. While uncertainties remain large, we conclude that there is some hope for determining those models more likely to simulate real world decadal variability. This knowledge will hopefully help understand and add-value to decadal forecasts.