Abstract

Climate variability on timescales longer than a year is often characterized by temporal scaling (“power law”) behaviour for which spectral power builds up at low frequencies in contrast to red-noise behaviour for which spectral power saturates at low frequencies. We estimate temporal power-law exponents (“Hurst exponents”) for the global atmospheric circulation of pre-industrial control and 20th century simulations for the troposphere and stratosphere. We consider 17 most established climate models from the CMIP3 archive. We show that current-generation climate models generally simulate the spatial distribution of the Hurst exponents well. We also use simulations of an atmospheric GCM with different climate forcings to explain the Hurst exponent distribution and to account for discrepancies in scaling behaviour between different observational products. Our analysis demonstrates that at the surface regions of large power law exponents coincide with the regions of strong decadal variability, namely northern North Atlantic, northern and tropical Pacific, and the Southern Ocean. In the free atmosphere these regions are confined to the tropical and subtropical troposphere and stratosphere. The spectral steepness in the former is explained by its strong coupling to the surface and in the later by its sensitivity to the volcanic aerosols. We conclude that characterization of temporal power-law behaviour provides a valuable tool for cross-validating low-frequency variability in various datasets, for elucidating the physical mechanisms underlying this variability, and for statistical testing of trends and periodicities in climate time series.