The Bird's Ear View: Audification for the Spectral Analysis of Heliospheric Time Series Data.

Abstract

The sciences are inundated with a tremendous volume of data, and the analysis of rapidly expanding data archives presents a persistent challenge. Previous research in the field of data sonification suggests that auditory display may serve a valuable function in the analysis of complex data sets. This dissertation uses the heliospheric sciences as a case study to empirically evaluate the use of audification (a specific form of sonification) for the spectral analysis of large time series. Three primary research questions guide this investigation, the first of which addresses the comparative capabilities of auditory and visual analysis methods in applied analysis tasks. A number of controlled within-subject studies revealed a strong correlation between auditory and visual observations, and demonstrated that auditory analysis provided a heightened sensitivity and accuracy in the detection of spectral features. The second research question addresses the capability of audification methods to reveal features that may be overlooked through visual analysis of spectrograms. A number of open-ended analysis tasks quantitatively demonstrated that participants using audification regularly discovered a greater percentage of embedded phenomena such as low-frequency wave storms. In addition, four case studies document collaborative research initiatives in which audification contributed to the acquisition of new domain-specific knowledge. The final question explores the potential benefits of audification when introduced into the workflow of a research scientist. A case study is presented in which a heliophysicist incorporated audification into their working practice, and the “Think-Aloud” protocol is applied to gain a sense for how audification augmented the researcher’s analytical abilities. Auditory observations are demonstrated to make significant contributions to ongoing research, including the detection of previously unidentified equipment-induced artifacts. This dissertation provides three primary contributions to the field: 1) an increased understanding of the comparative capabilities of auditory and visual analysis methods, 2) a methodological framework for conducting audification that may be transferred across scientific domains, and 3) a set of well-documented cases in which audification was applied to extract new knowledge from existing data archives. Collectively, this work presents a “bird’s ear view” afforded by audification methods—a macro understanding of time series data that preserves micro-level detail.