Intelligent Engineering Systems through Artificial Neural Networks Volume 18
27 Identifying Climate Teleconnection Signatures by Data Feature Extraction through Wavelet Spectral Decomposition
Download citation file:
- Ris (Zotero)
- Reference Manager
In this paper we present an analysis of various climate data sets by utilizing a discrete wavelet transform spectral decomposition. The spectral decomposition allows for feature extraction from the data which identifies teleconnection signals between the systems at various spectral frequencies. Feature extraction through spectral decomposition provides a means for understanding the likely time course response of individual systems to climate changes. Additionally, the spectral decomposition can yield insight into the time course response of climate changes propagating through various systems studied. By studying the spectral relationships, teleconnections between systems can be identified. The discrete wavelet transform allows for more information to be extracted from the data than a Fourier transform, by capturing both a map of the frequency content of the data, through examining it at different scales, as well as the temporal content of the data, by yielding the times at which these frequencies occur. Our overall research is focused on quantifying the uncertainty in climate model predictions of the effects of global warming on various systems. Understanding the frequency content and time course response of systems provides valuable information in identifying the evolution of signals through the data.