Dynamic network visualization accounting for geographic information

March 2016
Dynamic network visualization accounting for geographic information

Networks are complex objects whose representation generates a large number of constraints and questions. If we add a temporal component it becomes even more difficult to produce a satisfactory overview for reporting the dynamic aspect of this type of data. So, why not try to represent this data dynamically?

If the graph representation can be readable and understandable for a few tens or even hundreds of vertices (in the case where the density of links remains low), it becomes illusory to think of producing a readable image once the network contains few thousand vertices or more. If maps or graphs of this type are regularly produced, the issues are mainly technical and / or aesthetic because the representation will finally have little informative value.


brittanyNetworkByCommunes_2005_2007 brittanyNetworkByCommunes_2008_2010 brittanyNetworkByCommunes_2011_2013

Network representation of cattle trade data in Brittany (Northwestern France) between dairy farms for different periods of time, (a) 2005-2007, (b) 2008-2010 and (c) 2011-2013. Diagram shows animal movement data aggregated spatially by municipality and temporally over the different periods. Size of filled circles corresponds to the number of animals (yearly average herd size) present in each municipality (represented on the map at its geographical location), and their colour to the polarity (blue when rather seller and red when rather buyer). Lines represent animal movements between municipalities (direction is neglected), and their thickness is proportional to the number of traded animals. Movements from and to outside the metapopulation are not shown.


Above is three snapshots of a network. While it is quite easy to notice that there is an evolution of these network between the three periods of time, it is not possible to account for the existing real dynamic which is here rather strong using this type of visualization.

Below is an animation designed using Processing, with the aim to present the data used during my PhD thesis. Here the message remains clear, this animation show that we have an oriented, weighted and dynamic network.


Animation showing cattle trade network in Brittany (Northwestern France) between dairy farms over the year 2013. The visualization shows animal movement data aggregated spatially by municipality. Size of filled circles corresponds to the number of animals (yearly average herd size) present in each municipality (represented on the map at its geographical location), and their colour to the polarity (blue when rather seller and red when rather buyer over the whole period). Lines represent animal movements between municipalities, and the size of the black filled circles is proportional to the number of traded animals. Movements from and to outside the metapopulation are not shown.


Code used to create this animation will be soon available (when I will find time to annotate it a bit more).