In the last few years, network construction from climate data has developed to a promising
analysis method. Here, the similarities of time series from grid or station based climate data
are transfered into a network structure of nodes and edges. Then, this structure is being
analyzed using network measures (e.g. betweenness, centrality) and visualization
techniques. Major challenges for the visualization of such networks are their size, their
geo-reference and the multi-variate information coming with the nodes and edges:
The size results from the typical grid sizes (e.g. 720x360 in longitude, latitude
and often as well with z-levels) or the number of measurement stations and the
time dependency of the underlying climate data sets.
The geo-reference information is of high importance interpreting the underlying
physical processes of network structures, thus, network layout techniques are
often inadequate, and edge clutter can not be easily avoided. Considering
network comparison, which is of increasing importance in climate network
analyses, this clutter problem becomes even worse.
The multi-variate information in climate networks results from both underlying
data and node and edge measures, and can not be easily presented in the
visualization.
Facing these challenges, the talk will introduce strategies for the interactive visual
analysis of large, climate networks. This includes the discussion of different network
visualization techniques and state-of-the art visual analytics tools, including first solutions for
3D networks and for a visual comparison of climate networks. |