Publication Abstract
Guided Analysis of Hurricane Trends Using Statistical Processes Integrated with Interactive Parallel Coordinates.
Steed, C., Swan II, J. E., Jankun-Kelly, T. J., & Fitzpatrick, P. J. (2009). Guided Analysis of Hurricane Trends Using Statistical Processes Integrated with Interactive Parallel Coordinates. IEEE Symposium on Visual Analytics Science and Technology 2009. Atlantic City, NJ.
Abstract
This paper demonstrates the promise of augmenting interactive
multivariate representations with information from statistical processes
in the domain of weather data analysis. Statistical regression,
correlation analysis, and descriptive statistical calculations are
integrated via graphical indicators into an enhanced parallel coordinates
system, called the Multidimensional Data eXplorer (MDX).
These statistical indicators, which highlight significant associations
in the data, are complemented with interactive visual analysis capabilities.
The resulting system allows a smooth, interactive, and
highly visual workflow.
The system’s utility is demonstrated with an extensive hurricane
climate study that was conducted by a hurricane expert. In the
study, the expert used a new data set of environmental weather data,
composed of 28 independent variables, to predict annual hurricane
activity. MDX shows the Atlantic Meridional Mode increases the
explained variance of hurricane seasonal activity by 7-15% and removes
less significant variables used in earlier studies. The findings
and feedback from the expert (1) validate the utility of the data
set for hurricane prediction, and (2) indicate that the integration of
statistical processes with interactive parallel coordinates, as implemented
in MDX, addresses both deficiencies in traditional weather
data analysis and exhibits some of the expected benefits of visual
data analysis.