Understanding the predictability of the winter North Atlantic Oscillation using dynamical seasonal forecast models and machine learning techniques – Laura Baker – University of Reading
The winter North Atlantic Oscillation (NAO) is the leading mode of winter atmospheric variability in the North Atlantic region, and greatly influences the winter weather experienced in the UK and North-West Europe. The current generation of operational seasonal forecast systems are able to capture the NAO variability with some significant, but modest, levels of skill. However, there is some uncertainty in the skill of these forecasts. This is seen both by intermittency of skill in individual years within hindcast periods, and also in the decadal variability of skill over long hindcasts. In this work, we aim to understand more about the key processes leading to NAO variability, and how well they are represented by the seasonal forecast systems.
We do this using two approaches. First, we study a multi-model ensemble of seasonal forecast systems (C3S) to investigate common drivers of predictability of the NAO. We find that strong tropical forcing, mostly from the El Niño Southern Oscillation (ENSO), is an important factor in the more predictable NAO winters. Second, we use a machine learning technique (NARMAX) to learn more about the key drivers of predictability of the NAO and the temporal intermittency of this predictability. The Nonlinear AutoRegressive Moving Average with eXogenous inputs (NARMAX) systems identification approach is an interpretable machine learning method which can be used to identify and model linear and non-linear dynamic relationships between meteorological and related variables. We apply NARMAX to both reanalysis data and dynamical seasonal hindcasts of the NAO. We compare the predictors identified by NARMAX in each case. These results will help to understand how NAO predictability differs between the “real world” and the “model world”, and identify potential deficiencies in the dynamical seasonal forecast models.