Abstract
The low-frequency signals (LFS) of climate variables such as temperature and pressure often contain variability as a result of the nonlinear and non-stationary nature of Earth's climate system. Occasionally, as in the case of the North Pacific climate regime shift of the mid-1970s, this variability appears in the form of an abrupt shift in climate states. Because such variability can have large impacts on agriculture, wildfire frequency/intensity, and ecological systems, it is important to pursue a more complete understanding of low frequency climate interactions. Previously, techniques such as fourier, windowed fourier, and wavelet analyses were used to extract the LFS. However, these techniques rely on an assumption of linearity, and thus when applied to nonlinear climate data, can potentially cloud the physical meaning of the extracted LFS.
In this study a recently developed adaptive and temporally local analysis method—ensemble empirical mode decomposition (EEMD)—is applied to extract the LFS from observed daily minimum temperature data. The analysis uses data from 115 weather stations scattered throughout North Carolina, South Carolina, Georgia, Alabama, and Florida for the period from 1955 through 2007. An EOF analysis of the minimum temperature LFS reveals a large drop in the first PC time series in the mid-1970s. Further EOF-based analysis of the low-frequency variability leads to different interpretations of the characteristics of surface temperature variability. Most notably, the widely recognized shift of low-frequency variability around the mid-1970s can be alternatively interpreted in the Southeastern United States as phase coincidence between individual quasi-oscillatory components of interannual to decadal timescales.
|
