Good Visualization Example

This image represents a simulation of a fictional category 3 hurricane, as it surges from the Gulf of Mexico and into Lake Pontchartrain, flooding New Orleans . A category 3 hurricane is relatively slow moving, but the models showed that it is enough to break the levees that keep the lake in check. Hurricane Katrina grew to a category 5 hurricane at one point. In fact, everything that happened in New Orleans was predicted 100%, three years ago during a multidisciplinary study, according to Ivor van Herdeen.


I chose this as an example of a good visualization because with only a little rudimentary knowledge of maps, one is able to infer and correctly recognize most of the key points of the simulation.

The vector field intuitively communicates the flow of the hurricane, and the sequential images imply time flow as the hurricane moves across the land. Colour coding the wind velocity in this case is a better idea than varying the shade of one colour. The colour gradient chosen is well ingrained into our society as a threat indicator, with most of us accustomed to red meaning danger, and green/blue meaning safety. The high velocity wind is a direct threat to the safety of people living in that area, and thus the gradient was a good choice to communicate the urgency.

One drawback however is that with the shades of colour representing wind velocity, it is hard to differentiate between land mass and water.





Source: John Travis, "Hurricane Katrina: Scientists' Fears Come True as Hurricane Floods New Orleans", Science Vol 309, Issue 5741, 1656-1659 , 9 September 2005





  Fig. 4. Lightning and masses for the multicell: (a) Total, CG (negative and positive), and positive CG (CG+) lightning frequency averaged over 5 min, time of radar volume scans is marked on the upper time-axis, the horizontal bar indicates when the storm is observed by the ITF system. (b) Total mass derived for the complete storm versus CG lightning frequency, the arrows indicate the time sequence of the radar scans. (ch) Scatter plots of the CG lightning frequency and the mass fraction of rain (R), graupel (G), snow (S), and hail (H) in the (c) base layer (BL), (e) mid-level (ML), and (g) top-level (TL) of the storm. (d), (f), (h) same as (c), (e), (g) but for the mass of the hydrometeor classes and the total mass of all hydrometeors (T).



Bad Visualization Example

These graphs (as far as I understand) map the frequency of lightning strikes to different types of storm mass measurements (rain, snow, hail), gathered from a radar station. The first row displays totals, and the subsequent rows represent data from the BaseLevel (BL) of the storm, MidLevel (ML) and TopLevel (TL).


With more knowledge in the field, these graphs would probably make more sense. Regardless, they are trying to cram too much information in a small space, and taking a lot of shortcuts with their data plotting.

  • Instead of overlapping letters for the different data points, different shaped and coloured bullet points could have been used.
  • The graphs contain a mismatch of uppercase and lowercase letters denoting different things, which can lead to confusion.
  • The legend for the different graph lines is found only on the first graph, which has different scales and seems to measure something else alltogether, leading to ambiguous lines on the other graphs.
  • There is no legend for the data points apart from the wordy description in the caption. The graphs should be self-explanatory as much as possible, in my opinion.
  • The relation between base, mid and top level of the storm could have been better emphasized and represented visually, if it were organized to imply the actual physical distinction.


Source: Thorsten Fehr, Nikolai Dotzek, Hartmut Holler, "Comparison of lightning activity and radar-retrieved microphysical properties in EULINOX storms", Atmospheric Research Vol 76, Issues 1 - 4, July-August 2005, Pages 167-189