Statistical attribution of temporal variability in global gridded temperature data

Abstract:

Spatiotemporal variability within the climate system results from a complex interaction of various exogenous and endogenous factors, yet the understanding of the specific role of individual climate-forming agents is still incomplete. In this contribution, near-surface monthly temperature anomalies from several gridded datasets (GISTEMP, Berkeley Earth, MLOST, HadCRUT4, 20^th Century Reanalysis) are investigated for presence of components attributable to external forcings (anthropogenic, solar and volcanic) as well as to internal forcings related to major climate variability modes (El Niño / Southern Oscillation, North Atlantic Oscillation, Atlantic Multidecadal Oscillation and Pacific Decadal Oscillation). Statistical methodology based on multiple linear regression is employed, and applied to monthly temperature data for the 1901-2010 period. The results presented illustrate the spatial fingerprints of individual forcing factors and their robustness (or lack thereof) among individual temperature datasets. Particular attention is devoted to the specific features of the 20^th Century Reanalysis: It is demonstrated that while most of the response patterns are represented similarly in the reanalysis data and in their analysis-type counterparts, some distinctions appear, especially for the components associated with anthropogenic forcing and volcanic activity.