| description abstract | A conceptual framework is developed for quantifying the relationship between low-frequency variability and extreme events. In this framework, variability is decomposed into low-frequency and synoptic components using complementary 10-day low-pass and high-pass filters, and a distinction is made between two ways that low-frequency variability influences extremes: the additive effect, which neglects the dependence of synoptic variability on the low-frequency state, and the multiplicative effect, which is due to the dependence of synoptic variability on the low-frequency state. The influence of various factors on the relationship between low-frequency variability and extreme events is decomposed and quantified by generating a series of simple synthetic datasets based on different assumptions about low-frequency and synoptic variability and their relationship. These techniques are used to study the relationship between low-frequency variability and extreme westerly wind events in three datasets, an 1158-yr GCM simulation and two reanalysis datasets, with similar results for all three. Geographical variations in the low-frequency?extreme relationship are only partially explained by geographical variations in the low-frequency?synoptic variance ratio; the non-Gaussianity of low-frequency and synoptic variability and the relationship between synoptic variance and the low-frequency state are also found to be important. The simple synthetic datasets that include these factors provide good estimates of the magnitude and probability of extremes. Implications for predictability and applications to more complex low-frequency?extreme relationships are discussed. | |
