Interactive Visual Analysis of Multi-run Climate Data

Johannes Kehrer

MISC, December, 2009

Abstract

The increasing complexity of data stemming from climate models and observations creates new challenges for data analysis. Traditional approaches are often based on computing statistical data properties. Interactive visual analysis, on the other hand, allows the stepwise exploration of the data in a guided human-computer dialog. It uses graphical representations of the data to interactively explore the data in multiple linked views. This allows the analyst to rapidly generate and analyze hypotheses, to identify data deficiencies, and to explore data trends and outliers. In an ongoing cooperation between the University of Bergen, Norway, the Potsdam Institute for Climate Impact Research (PIK), and the SimVis GmbH, Vienna, we used and extended our visual analysis framework to also work with multi-run climate data. In the framework, we relate the original multi-run data and derived statistical properties to each other. This allows the analyst to work in parallel with both, the aggregated data representation and the original multi-run data. We demonstrate this in a visual sensitivity analysis of the multi-run data.

Published

Invited talk at Potsdam Institute for Climate Impact Research (PIK)

• Location: Potsdam, Germany
• Date: December 2009

Media

BibTeX

@misc{kehrer09potsdam,
  author = {Johannes Kehrer},
  title  ={Interactive Visual Analysis of Multi-run Climate Data},
  year = {2009},
  month = {December},
  howpublished = {Invited talk at Potsdam Institute for Climate Impact Research (PIK)},
  location = {Potsdam, Germany},
  abstract = {The increasing complexity of data stemming from climate models and 
observations creates new challenges for data analysis.  Traditional approaches 
are often based on computing statistical data properties.  Interactive visual 
analysis, on the other hand, allows the stepwise exploration of the data in a 
guided human-computer dialog.  It uses graphical representations of the data to 
interactively explore the data in multiple linked views.  This allows the analyst 
to rapidly generate and analyze hypotheses, to identify data deficiencies, and 
to explore data trends and outliers.

In an ongoing cooperation between the University of Bergen, Norway, the Potsdam 
Institute for Climate Impact Research (PIK), and the SimVis GmbH, Vienna, we used 
and extended our visual analysis framework to also work with multi-run climate 
data.  In the framework, we relate the original multi-run data and derived 
statistical properties to each other.  This allows the analyst to work in 
parallel with both, the aggregated data representation and the original 
multi-run data.  We demonstrate this in a visual sensitivity analysis of the 
multi-run data.},


}