Detection and Attribution of Global Surface Air Temperature Change in the Instrumental Record

Thesis (Ph.D.)--University of Washington, 2013This study investigates the spatial and temporal variability of surface air temperature (SAT) in the instrumental record of climate, emphasizing the detection of SAT change in the presence of natural variability as well as the attribution of SAT variability and trends to atmospheric circulation changes, as reflected in fluctuations in sea-level pressure (SLP) teleconnection patterns. In order to better understand the magnitude and significance of SAT trends observed over land during the 20th Century, we identify and characterize the background variability on time scales of days to decades. The magnitude of day-to-day and month-to-month SAT variability exhibits a pattern of polar amplification, particularly during the boreal cold season. The ratio of extratropical to tropical variability is approximately 4:1 on daily to monthly time scales, decreasing to about 2:1 on the time scale of several decades. Similarly, it is suggested that variability on the order of the length of the instrumental record exhibits an extratropical to tropical ratio of approximately 2:1. It follows that despite polar amplifications, global warming is more clearly distinguishable from the background variability in the tropics than at high latitudes. The ecological impacts of global warming are also likely to be larger in the tropics. We employ a variety of statistical methods to estimate the fraction of Northern Hemisphere cold season SAT variability that is attributable to fluctuations in atmospheric circulation patterns. These methods each yield a dynamical adjustment to SAT fields to account for the influence of patterns in dynamical fields on local SAT variability. After comparing a set of seven statistical approaches, we argue that when applied in a pointwise manner, partial least squares regression provides the most parsimonious and statistically robust dynamical adjustment. The dynamical adjustment accounts for approximately 50% of the variance of NH cold season SAT. Changes in the NH cold season circulation patterns encompassed by the dynamical adjustment account for approximately 0.7°C of the 1.7°C of warming observed poleward of 40°N during the historical reference interval 1965-2000, and approximately 0.5°C of the 1.6°C warming observed over the same area during the period 1920-2010, reconciling much of the spatial and seasonal differences in SAT trends observed regionally and globally. Our analysis yields a dynamically adjusted SAT field; this provides a dataset capable of being analyzed for residual variability and trends associated with thermodynamic and radiative phenomena. The circulation patterns that influence local SAT variability are closely associated with the dominant patterns of NH wintertime SLP variability. Whereas previous descriptions of the dominant patterns of NH wintertime SLP variability have been limited to just a few patterns, we provide evidence that there exists a set of 6-9 reproducible NH wintertime teleconnection patterns that are robust with respect to the choice of analysis method and to the background climatology on which they are imposed. The leading three patterns are hemispheric to continental in scale. The higher order patterns are synoptic to continental in scale, limited to specific regions. Although the hemispheric patterns have a broader influence, the regional patterns contribute equally to local climate impacts