Attribution of Atlantic Multidecadal Variability to External Forcing, Internal Variability, and Weather Noise
Date
2014-10-14
Authors
Colfescu, Ioana
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Detection of externally forced climate change and attribution of the causes of the externally forced and internally generated climate variability during the last century are the central scientific issues of current climate science and the subject of important controversies. This thesis systematically addresses fundamental problems in detection and attribution. A novel three-tier model ensemble strategy is developed and applied in the model world to address these issues. At the top tier, an ensemble of CGCMs with the same external forcing applied to each member is used to separate the results from each ensemble member into the externally forced and internally generated components. At the second tier, an ensemble of atmospheric GCMs (AGCM) with each member forced by the same SST, taken from a member of the CGCM ensemble, is used to separate the atmospheric variability in that CGCM member into SST-and-externally-forced and weather noise components. The third tier, uses an interactive ensemble version of the CGCM, in which the AGCM is replaced by an AGCM ensemble, so that atmospheric weather noise in the CGCM is suppressed. Controlled experiments forcing the interactive ensemble with the atmospheric noise diagnosed in the AGCM ensemble tier isolate the role of the weather noise in generating the internal SST variability found in the CGCM Ensemble tier. The strategy is employed to examine three important detection and attribution issues. The first is why the AGCM ensemble forced by observed SST does not simulate the observed 20th century sea level trends in the Indian Ocean. It has been suggested that this is because of an intrinsic failure of the AGCM Ensemble to correctly represent the SST forced response of the coupled system. The results show that the AGCM and CGCM ensembles are consistent with each other, and suggest that the failure to simulate the observed trends is due to model bias rather than coupling. Next, the spatial and temporal properties of the weather noise obtained from the CGCM and AGCM ensembles are examined in a preliminary fashion. This is the first attempt to document these properties. The temporal and spatial structures of the weather noise in the CGCM and AGCM simulations are very similar. The temporal structures of the noise spectra are white at timescales larger than approximately 5 months, although the noise is temporally non-Gaussian, while the spatial structures resemble those of major modes of observed climate variability. No change is detected between the statistical properties of the noise in the early and late 20th century. The Atlantic Multidecadal Variability (AMV) sea surface temperature is decomposed into externally forced and internally generated components using the CGCM ensemble, and the weather noise contribution to the internal component is isolated using the AGCM ensemble and interactive ensemble. The AMV has a strong contribution from the external 20th century forcing. The internal AMV variability is primary forced by the weather noise, but other sources of internal variability are also important. An important contribution to the internal AMV is associated with the internal variability of the oceanic Atlantic Meridional Overturning Circulation, and this contribution is distinct from the weather noise forced component.
Description
Keywords
Climate change, Climate Models, Climate variability in 20th century, Detection and attribution of climate change, North Atlantic Variability, Weather Noise