It can be difficult to visualize categorical data that fluctuate over space and time. Listing the frequency of different crimes, for example, can be accomplished with a table showing the crime type on one axis, and the neighborhood or police beat on the other axis. While this is a fine method to report crimes statistics, it may not reveal spatial patterns that are easily discerned through a map-based approach. The method employed for displaying these type of data in Salt Lake City is considered one brand of “Data Rose” and could be used for many different topics, anything from crime statistics to bird migration patterns.
 
To begin, a look at the legend quickly orients you to proper interpretation of this icon. There are 12 rings that represent the months of the year, with January being the internal ring and December on the outside. These rings are split into 10 sectors that represent different crime types. Color coding the rings according to the monthly total allows you to see 120 data points in a single region, and it’s easy to read!

A sample section of the map. Click to view full poster (9 MB PDF).

 
It is important to note that the green to red color ramp is consistent within a given crime type across the city, but not between types. There are five frequency classes per type, so while green always indicates at least one occurrence, the significance of red can vary widely. For example a few acts of arson may show as red because arson is rare, while car prowls require many more occurrences to indicate an elevated level and therefore be shown in red. Also, these data are not normalized by population.
 
While all of the information addresses an overall topic, minor categories can be subtly delineated as well.  Crimes that are directed at people are on the left side of the rose, and crimes that are focused on property are on the right. A quick glance at the map and you can easily see the different crime types that occur across the city. One area has an elevated rate of vandalism during the summer months, while another has car prowls that are at a consistent level throughout the year.
 
A comparison of the data rose to other graphing methods reveals clear aesthetic advantages. A stacked bar chart in each police beat, for example, may use each bar to represent the individual months with the portions of the bar indicating the amount of crime. While this accomplishes the goal of placing many variables in the graph, these variables, in the case of our crimes types, are not aligned making a temporal comparison within a given type more difficult.
 
This method to display data over space and time was inspired by the work of Guilan Huang of the Division of Integrated Biodefense at Georgetown University, which was published in the Winter 2008 copy of ArcUser Magazine. Huang's approach applied rings around an entire county, with each sector showing a single variable through time for a given ZIP Code. I modified this approach to shrink the rings within the area of interest and convert the sectors to show 10 variables rather than one.