Ask A Cartographer

showing classes in legends for continuous rasters

October 16 2008 | 0 comments
Categories: Map Elements

What would be the proper convention for displaying raster classes in a legend for continuous rasters? Would floating point values and integer rasters be treated differently? In my case, I have both types. I am inclined to display the classes in the legend for a floating point, say an interpolated precipitation surface for example, like the following:

Precipitation (over the previous 18 hours)
<= 0.5mm
0.5 - 1.5mm
1.5 - 5.0mm
5.0 - 15.0mm
> 15mm

There are a few issues here. One is, whether to use the mathematical operators to describe "less than or equal to" and "greater than".

Another is whether to put the units (mm) the way that I had it, or in the layer title, or after ever number (i.e., "1.5mm - 5.0mm"). My example of Dew Point further down was done differently just to show one of these other ways of doing it.

But likely my biggest issue is whether to include the dividing values ("Break Values") of the classes in the legend's classes on both sides of the value (i.e., "0.5" in both the 1st class description and the 2nd one, and the same with "1.5", "5.0", and "15.0"). I have seen someone else in my office use several decimal places to avoid repeating a value in the legend (eg. "0.50000001 - 1.5mm"), but I find that clunky and dependent on the bit depth of the pixel. On a floating point raster, will the audience be bothered by the question of which class a value of 0.500000000 belongs to? ArcGIS seems to always assign the break value to the upper limit of the lower range, which is different than what has historically been done in my office, but I am inclined to change our ways to match this ArcGIS default way.
If the interpolated continuous surface raster was composed of integer values, would it be correct to label classes like the following (dew point as example)?

Dew Point (degrees Celcius)
<= -5
-4 - 0
1 - 5
6 - 10
11 - 15
16+

With integer values, I am less inclined to repeat the break values in the legend. Does this sound logical?

Your advise would be greatly appreciated. Thanks.

### Mapping Center Answer:

For continuous data I would do a custom graphic.  The default legends for rasters are pretty weak in terms of their potential for rapid communication.  See the attachment I added.

The main things I did were:

1. Converted to graphics and ungrouped the result until I could edit all the pieces
2. Replaced the graphic tiles which are not actually fill symbols with graphic rectangles that I made the same size as the originals using the graphics toolbar.  I did this to remove the outlines--your pixels in the map have no outlines and neither should your legend.
3. Rounded the values to make them easy to read
4. Instead of labeling the class with values that are ambiguous in terms of where they stop and start, label the breaks between the class with a single break value.  This works very well for floating point datasets--see below for integer data.

One thing about your precipitation legend that is just wrong is the use of < and > for the smallest and largest classes.  There is a concept called "truth in mapping" and the idea is to present all the information.  Legends that use the convention you described do not communicate the minimum and maximum values--which are quite possibly the most interesting values to readers.

For integer rasters, the ranges can come back, like you depict, but use your true minimum and maximum values to bound the first and last class's ranges in the legend.  Based on "truth in mapping" your label of 16+ is of no use.  I don't know whether the maximum value is 17 or 27,000.

If you would like to post a comment, please login.